filename stringlengths 13 34 | markdown stringlengths 0 359k |
|---|---|
10.1371_journal.pone.0217459 | RESEARCH ARTICLE
Dietary multi-enzyme complex improves In
Vitro nutrient digestibility and hind gut
microbial fermentation of pigs
Neeraja Recharla1, Duwan Kim2, Sivasubramanian Ramani1, Minho Song3,
Juncheol Park2, Balamuralikrishnan BalasubramanianID
Sungkwon ParkID
1, Pradeep Puligundla4,
1*
1 Department of Food Science and Biotechnology, Sejong University, Seoul, Korea, 2 National Institute of
Animal Science, Swine division, RDA, Sunghwan, Korea, 3 Division of Animal and Dairy Science, Chungnam
National University, Daejeon, Korea, 4 Department of Food Science & Biotechnology, Gachon University,
Seongnam, Korea
* sungkwonpark@sejong.ac.kr
Abstract
This study was conducted in two stages to investigate the potential of multi-enzyme supple-
mentation on the nutrient digestibility, growth performance, and gut microbial composition of
pigs. In stage 1, effects of multi-enzyme complex (xylanase, α-amylase, β-glucanase, and
protease) supplementation on the ileal and total tract dry matter (DM) digestibility of feed-
stuffs were investigated with in vitro two-stage and three-stage enzyme incubation methods.
A wide range of feed ingredients, namely, corn meal, wheat meal, soybean meal, fish meal,
Oriental herbal extract, Italian rye-grass (IRG) and peanut hull were used as substrates.
Supplementation of the multi-enzyme complex increased (P < 0.05) the digestibility of the
Oriental herbal extract and corn meal. In stage 2, in vivo animal studies were performed to
further investigate the effects of the dietary multi-enzyme complex on the nutrient utilization,
growth performance, and fecal microbial composition of pigs. A total of 36 weaned pigs
were fed corn- and soybean meal-based diets without (control) and with the multi-enzyme
complex (treatment) for 6 weeks. Fecal samples were collected from 12 pigs to analyze the
microbial communities by using DNA sequencing and bioinformatics tools. Multi-enzyme
supplementation had no effect on apparent digestibility of nutrients and growth performance
of pigs compared to control. Taxonomic analysis of the fecal samples indicated that the
bacteria in both control and treatment samples predominantly belonged to Firmicutes and
Bacteroidetes. In addition, the proportion of the phylum Firmicutes was slightly higher in
the treatment group. At the genus level, the abundance of Treponema and Barnesiella
increased in the treatment group; whereas the numbers ofthe genera including Prevotella,
Butyricicoccus, Ruminococcus and Succinivibrio decreased in the treatment group. These
results suggest that multi-enzyme supplementation with basal diets have the potential to
improve nutrient digestibility and modify microbial communities in the hind-gut of pigs.
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: Recharla N, Kim D, Ramani S, Song M,
Park J, Balasubramanian B, et al. (2019) Dietary
multi-enzyme complex improves In Vitro nutrient
digestibility and hind gut microbial fermentation of
pigs. PLoS ONE 14(5): e0217459. https://doi.org/
10.1371/journal.pone.0217459
Editor: Juan J Loor, University of Illinois, UNITED
STATES
Received: July 26, 2018
Accepted: May 13, 2019
Published: May 28, 2019
Copyright: © 2019 Recharla et al. This is an open
access article distributed under the terms of the
Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: All relevant data are
within the paper and its Supporting Information
files.
Funding: This work was carried out with the
support of ‘‘Cooperative Research Program for
Agriculture science & Technology Development
(Project No. PJ012615)” Rural Development
Administration and Basic Science Research
Program through the National Research
Foundation of Korea (NRF) funded by the Ministry
of Science, ICT & Future planning
PLOS ONE | https://doi.org/10.1371/journal.pone.0217459 May 28, 2019
1 / 19
(2015R1C1A1A01051767), South Korea). The
funders had no role in study design, data collection
and analysis, decision to publish, or preparation of
the manuscript.
Competing interests: The authors have declared
that no competing interests exist.
The effect of enzyme complex supplement on pig microbiota and performance
Introduction
Many plant-based feed ingredients used in swine diets, especially cereal grains, contain large
quantities of non-starch polysaccharides (NSPs) with anti-nutritional factors (ANFs). NSPs
are poorly used by pigs because they lack specific endogenous enzymes for their breakdown,
and, consequently, NSPs are fermented and used by pig intestinal microbes [1]. The major
NSPs of the plant cell wall are composed of cellulose (linear β-glucan chains), hemicellulose
or non-cellulosic polymers (arabinoxylans, mixed-linked β-glucans, mannans, galactans, and
xyloglucan) and pectic polysaccharides (polygalacturonic acids) [2]. Elevated levels of NSPs in
swine diets have adverse effects on nutrient digestibility and absorption rate. Especially, soluble
NSPs increase the viscosity of digesta and alter the intestinal transit time. These effects could,
in turn, lead to changes in the physiology and ecosystem of the gut [2,3]. The addition of exog-
enous NSP- degrading enzymes can improve nutrient availability in swine diets by breaking
down the nutrient encapsulating cell wall as well as ameliorating viscosity problems associated
with certain NSPs, particularly arabinoxylans and β-glucans [4,5].
Exogenous enzymes are being successfully used in poultry diets to reduce the negative
effects of NSPs in cereals such as barley, wheat, and rye [6]. Recently, there has been consider-
able interest in the use of exogenous enzymes in the swine industry to address the anti-nutri-
tive effects of NSPs on animal performance. A previous study has shown that supplemental
exogenous digestive enzymes may improve average daily gain (ADG), feed conversion ratio
(FCR), and digestibility of dry matter (DM) in pigs [7]. Thus, supplementation of swine diets
with exogenous enzymes has gained popularity, and this approach has the potential to improve
the nutrient utilization of high-fiber diets.
Different types of exogenous feed enzymes, such as proteases, carbohydrases, phytases, and
xylanases are commercially used in livestock feeds [8]. The inclusion of proteases as feed addi-
tives enhances the protein digestion and can increase the growth performance of adult pigs [9,
10]. Proteases also stimulate gut development, maturation, and health in weaning and weaned
piglets in the early stages by degrading protein-bound complexes to release other nutrients
along with protein [10, 11, 12]. Carbohydrases are enzymes that catalyze the breakdown of
complex carbohydrates into oligosaccharides, disaccharides and monosaccharides. Carbohy-
drases are classified into starch-degradable and NSP-degradable enzymes [13]. The starch-
degrading enzymes are not often used in animal feed (except in creep feed) because animals
are able to synthesize them endogenously. However, for efficient feed-nutrient utilization, cer-
tain enzymes may be exogenously supplemented as feed additives to target NSP hydrolysis in
the livestock production industry. Such accessory enzymes include xylanase, mannanase, and
glucanase [14, 15]. These enzymes hydrolyze plant cell-wall components such as xylan, man-
nan, and beta-glucan and assist in the release of nutritional constituents, such as protein,
starch, lipids and other minerals, that are trapped within the cell-wall matrix [15, 16, 17].
Upon hydrolysis of NSPs and availability of the entrapped nutrients, the resultant products are
readily accessible for the intestinal microflora, which can have multiple beneficial effects on
animal gut health and whole animal [18]. Such health-promoting microorganisms enhance
gut physiology, for example, reduction of relative weight of organs in the digestive system and
increased villus height [18, 19].
Many studies have reported that multi-enzyme supplementation had more positive effects
on feed utilization and animal performance when pigs were fed with mixed grain-based diets
because of the synergistic interaction between enzymes [20, 21]. The potential of a multi-
enzyme preparation for the hydrolysis of different feed-stuffs is principally determined by
the digestibility rate. This rate can be measured directly in in vivo animal models, but it is
practically difficult because of the high number of samples and ethical objection to animal
PLOS ONE | https://doi.org/10.1371/journal.pone.0217459 May 28, 2019
2 / 19
The effect of enzyme complex supplement on pig microbiota and performance
experimentation. In addition, in vivo methods are time consuming and expensive procedures.
In contrast, in vitro incubation techniques that mimic in vivo digestion can be considered as
rapid methods for the prediction of in vivo digestibility values [22, 23]. In vitro techniques are
relatively less expensive, simpler, and rapid when compared with animal experiments [24].
Therefore, prior to introducing novel exogenous enzymes in swine diets, in vitro digestibility
methods can be used to determine the efficacy of the exogenous enzymes.
The objective of the present study was to screen swine feed ingredients suitable for hydroly-
sis by multi-enzyme complex with in vitro digestibility methods and to evaluate the effects
of enzyme supplementation of corn-soybean meal-based diets on the nutrient digestibility,
growth performance, and gut microbial composition of pigs.
Materials and methods
2.1. In vitro study
2.1.1. Feed ingredients and enzymes. Seven samples of ground feed ingredients (soybean
meal, corn meal, wheat meal, fish meal, Oriental herbal extract, pea nut hull, and Italian rye
grass) were obtained and sieved with a 0.85 mm sieve and stored in air-tight containers until
digestibility analysis. A commercial exogenous enzyme complex was provided by Feed Best
Inc. (Seoul, Korea). The complex contained a mixture of β-pentosanase (xylanases; 6000 EPU/
g) and synergetic enzymes, namely, β-glucanase, α-amylase and protease (32,000, 17,600, and
142 EU/g, respectively). The test samples were divided into two groups, feed samples without
enzymes that served as the controls, and feed samples with 0.1% multi-enzyme (treatment
group).
2.1.2. Experiment procedures: In vitro ileal digestibility method. To predict the in vitro
ileal digestibility (IVID) of the feed-stuffs for pigs, we used the method described by Boisen
and Fernandez [23]; it consisted of two-step enzymatic incubations under different pH condi-
tions that simulated digestion in the stomach and small intestine. The experiments were per-
formed with three replicates.
According to the protocol described by Boisen and Fernandez [23], in step one, 1 g of the
ground feed sample to an accuracy of ± 0.1 mg (1 g ± 0.1 mg) was placed in a conical flask, 25
mL of 0.1 M phosphate buffer at pH 6 (sodium phosphate buffer prepared with monosodium
phosphate and its conjugate base, disodium phosphate) and 10 mL of 0.2 M HCl were added
in the flask, and the pH was adjusted to 2 with 1 M HCl or NaOH solution. Then 1 mL of
freshly prepared pepsin solution (10 mg/mL; � 250 U/mg solid, P7000, pepsin from porcine
gastric mucosa; Sigma-Aldrich, St. Louis, MO, USA) and 0.5 mL of chloramphenicol solution
(to prevent bacterial contamination) were added to the mixture. The flasks were incubated in a
shaking water bath (shaking speed, 80 rpm) at 39 ˚C for 6 h. In the second step (after 6 h of
incubation), 10 mL of phosphate buffer (0.2 M, pH 6.8) and 5 mL of 0.6 M NaOH solution
were added to the mixture in the flasks. Then, pH was adjusted to 6.8 with 1 M HCl or NaOH,
and 1 mL of freshly prepared pancreatin solution (50 mg/mL; 4 × USP, P1750, pancreatin
from porcine pancreas; Sigma-Aldrich) was added. After enzyme addition, the test flasks were
incubated in a shaking water bath at 39 ˚C for 18 h.
After the incubation, all the samples were placed in an ice bath to stop enzyme action until
sample filtration. The undigested residues were filtered using dried and pre-weighed glass filter
crucibles containing 500 mg of Celite filter aid (Sigma-Aldrich) with vacuum pump support.
All undigested samples were transferred to the filtration crucible by rinsing the flasks with dis-
tilled water, and the residue was further washed with about 10 mL of 95% ethanol and 99.5%
acetone to remove the lipid content. After filtration, the undigested residue in the crucibles
was dried at 101 ˚C overnight. Then, the residue was weighed and cooled for 1 h in a
PLOS ONE | https://doi.org/10.1371/journal.pone.0217459 May 28, 2019
3 / 19
The effect of enzyme complex supplement on pig microbiota and performance
desiccator. The in vitro digestibility of DM was calculated from the difference between DM in
the sample and undigested residue after correction for DM in the blank.
2.1.2a: In vitro total tract digestibility method. In vitro total tract digestibility (IVTTD)
was measured using a three-step multi-enzymatic incubation method, which consists of gastric
(pepsin added), small intestine (pancreatin added), and large intestine (viscozyme added)-sim-
ulated digestive phases according to Boisen and Fernandez [23]. The experiments were per-
formed with three replicates.
The ground feed sample (0.5 g, particle size �0.85 mm, accuracy of 0.001 g) was placed in a
100 mL conical flask, and 25 mL of 0.1 M phosphate buffer (pH 6) was added to the flask, pH
of the solution was reduced with 10 mL of 0.2 M HCl and the pH adjusted to 2.0 ± 0.1 with 1
M HCl or 1 M NaOH. Then, 1 mL of pepsin solution, which was prepared by adding 25 mg of
pepsin per mL of distilled water (pepsin from porcine gastric mucosa, P 7000, � 250 U/mg
solid; Sigma-Aldrich) was added. To prevent microbial contamination, 0.5 mL chlorampheni-
col solution (0.5 g of chloramphenicol in 100 mL of ethanol) was added. The flasks were closed
with silicon stoppers and placed in a water bath with mild agitation (50 rpm) at 39 ˚C for 2
h ± 5 min (first step). After the first incubation, 10 mL of 0.2 M phosphate buffer (pH 6.8) and
5 mL of 0.6 M NaOH were added to the flasks, and the pH was adjusted to 6.8 ± 0.1 with 1 M
HCl or 1 M NaOH. Then, 1 mL of pancreatin solution (Sigma-Aldrich) containing 100 mg of
pancreatin per 1 mL of distilled water was added to the mixture. The flasks were closed with
silicon stoppers and incubated in a shaking water bath with mild agitation (50 rpm) at 39 ˚C
for 4 h ± 5 min (second step). After the incubation, 10 mL of 0.2 M EDTA solution was added
to the mixture, the pH was adjusted to 4.8 ± 0.1 with 30% acetic acid solution, and 0.5 mL of
Viscozyme L (V-2010; Sigma-Aldrich) was added to each flask; the flasks were incubated in a
shaking water bath for 18 h with agitation at 39 ˚C (third step). Then, the flasks were placed in
ice water to stop enzymatic reaction. The schematic representation of the IVTTD method is
shown in Fig 1.
IVID and IVTTD of DM (%) were calculated using the following equation;
IVID or IVTTD of DM % ¼ ½Sample DM (cid:0) ðResidue DMI (cid:0) Blank DMÞ�=Sample DM � 100
DM = Dry matter
2.2. In vivo study
2.2.1. Pigs and housing. A total of 36 weaned pigs [Duroc × (Landrace × Yorkshire) with
an initial average body weight (BW) of 6.02 ± 0.32 kg were used for this study. The experimen-
tal protocol for this study was reviewed and approved by the Animal Care and Use Committee
of Chungnam National University (Approval# CNU-00611). The pigs were randomly assigned
to two different dietary treatments, either control or treatment, with three pigs per pen. Each
pen was equipped with a feeder and waterer in an environmentally controlled room. The pigs
had free access to water and feed.
2.2.2. Experimental diets. The diets were based on corn and soybean meal and formu-
lated to meet required amounts of vitamins and minerals for pigs [25]. The composition of
the experimental diets is listed in Table 1. The diets were fed without (control) and with the
multi-enzyme supplement (treatment). Treatment diet was the control diet supplemented
with 0.1% (w/w) multi-enzyme complex including xylanase, α-amylase, β-glucanase, and
protease (6000 EPU/g, 17,600, 32,000, and 142 EU/g, respectively). The pigs were fed for 6
weeks according to a two-phase feeding program: weeks 1 to 3 (phase 1, 21 days) and weeks
4 to 6 (phase 2, 21 days). During the experimental period, two dietary compositions (for
phases 1 & 2) were formulated to meet the nutrient requirements based on age. The diets did
PLOS ONE | https://doi.org/10.1371/journal.pone.0217459 May 28, 2019
4 / 19
The effect of enzyme complex supplement on pig microbiota and performance
Fig 1. Schematic representation of IVTTD.
https://doi.org/10.1371/journal.pone.0217459.g001
not include spray-dried plasma, antibiotics, or zinc oxide to avoid their antibacterial or phys-
iological effects.
2.2.3. Performance monitoring and sample collection for digestibility analysis. The
pigs were individually weighed on day 1, 21, and 42 of the experiment and the amount of feed
supplied per pen and remaining feed were recorded at the beginning and end of the experi-
ment for each phase. These values were used to measure ADG, average daily feed intake
(ADFI), and feed efficiency (gain to feed ratio, G:F).
Diet samples were also collected from each batch of the manufactured feed and stored at
-20 ˚C until analyses. To determine the apparent digestibility of nutrients, 0.25% chromic
oxide (as the marker) was added to the diets during days 36 to 42 of the experiment [26]. Fecal
samples were collected for 3 days after the 4-day adjustment period for apparent total tract
digestibility. The collected fecal samples were pooled and stored at -20 ˚C until analyses. Two
pigs per pen were euthanized using CO2 after injection of zoletil and exsanguinated at the end
of the experiment. Digests samples from the ileum were collected and stored at -20 ˚C until
analyses.
2.2.4. Analytical methods. The diet, feces, and digesta samples were dried in a force-
air drying oven at 60 ˚C and ground through a cyclone mill (Foss Tecator Sycltec 1093,
HillerØd, Denmark) before analysis. The diet, feces, and digesta samples were analyzed
for DM (method 930.15) and nitrogen (method 988.05) according to the AOAC methods
[27]. Gross energy was measured using a bomb calorimeter (Parr 1281 Bomb Calorimeter:
Parr Instrument Co., Moline, IL, USA), and chromium content was determined using an
PLOS ONE | https://doi.org/10.1371/journal.pone.0217459 May 28, 2019
5 / 19
The effect of enzyme complex supplement on pig microbiota and performance
Table 1. Composition of experimental diets fed for weaned pigs.
Ingredient (%)
Control
Phase 1x
Treatment
Control
Phase 2y
Treatment
Corn
Soybean meal, 44%
Soy protein concentrate
Dried whey
Lactose
Soybean oil
Limestone
Multi-enzyme supplementa
Monocalcium phosphate
Vitamin pre-mixb
Mineral pre-mix
L-lysine-HCl
DL-methionine
Total
Calculated energy and nutrient content
ME, Mcal/kg
CP, %
Calcium, %
Phosphorus, %
Lysine, %
31.57
18.00
16.96
24.00
4.00
3.00
1.00
-
0.90
0.20
0.20
0.08
0.09
100
3.53
24.49
0.81
0.69
1.54
31.57
18.00
16.96
24.00
4.00
3.00
1.00
+
0.90
0.20
0.20
0.08
0.09
100
3.53
24.49
0.81
0.69
1.54
51.56
26.56
8.00
10.00
-
1.35
1.00
-
0.90
0.20
0.20
0.17
0.07
100
3.42
22.51
0.73
0.63
1.41
51.56
26.56
8.00
10.00
-
1.35
1.00
+
0.90
0.20
0.20
0.17
0.07
100
3.42
22.51
0.73
0.63
1.41
Phase 1x = week 1 to 3 (21 days), phase 2y = week 4 to 6 (21 days).
Multi-enzyme supplementa = 1 kg multi-enzyme mixture was mixed per 1 ton of control diets. The multi-enzyme contained xylanase, α-amylase, β-glucanase, and
protease.
Vitamin pre-mixb = Provided per kilogram of diet: vitamin A, 12,000 IU; vitamin D3, 2,500 IU; vitamin E, 30 IU; vitamin K3, 3 mg; D-pantothenic acid, 15 mg; nicotinic
acid, 40 mg; choline, 400 mg; and vitamin B12, 12 μg.
ME = Metabolizable energy; Mcal/kg = megacalories per kilogram
CP = Crude protein
https://doi.org/10.1371/journal.pone.0217459.t001
absorption spectrophotometer (Hitachi Z-5000 Absorption Spectrophotometer; Hitachi
High-Technologies Co., Tokyo, Japan), according to Williams et al. [26]. Apparent ileal
and total tract digestibility of DM, CP, and energy were calculated using the index method
[28].
2.2.5. Microbial community analysis. To analyze the bacterial diversity in the pig feces,
fecal samples were collected from the two groups of pigs (6 pigs per group). The collected fecal
samples were immediately stored at -80 ˚C for further analyses.
2.2.5a: DNA isolation and PCR amplification. The DNA was extracted from the fecal
samples by using the PowerSoil DNA Isolation Kit (Cat. No. 12888, MO BIO) according to the
manufacturer’s protocol. Each sequenced sample was prepared according to the Illumina 16S
Metagenomic Sequencing Library protocols. The DNA quantity and quality were measured
using PicoGreen (Invitrogen, Eugene, OR, USA) and NanoDrop (Thermo Scientific, Wal-
tham, MA, USA). The 16S rRNA genes were amplified using 16S V3-V4 primers: 16S ampli-
con PCR forward primer
5' TCGTCGGCAGCGTCAGATGTGTATAAGAGACAGCCTACGGGNGGCWGCAG
16S amplicon PCR reverse primer
5' GTCTCGTGGGCTCGGAGATGTGTATAAGAGACAGGACTACHVGGGTATCTAATCC
PLOS ONE | https://doi.org/10.1371/journal.pone.0217459 May 28, 2019
6 / 19
The effect of enzyme complex supplement on pig microbiota and performance
Input gDNA (12.5 ng) was amplified with the primers, and a subsequent limited-cycle
amplification step was performed to add multiplexing indices and Illumina sequencing adapt-
ers. The final products were normalized and pooled using PicoGreen, and the library size were
verified using the LabChip GX HT DNA High Sensitivity Kit (PerkinElmer, Massachusetts,
USA). The sequencing was performed using the MiSeq platform (Illumina, San Diego, CA,
USA).
2.2.5b: Next-generation sequencing and data analysis. Next-generation sequencing
(NGS) analysis was performed using the fecal samples by Macrogen Inc. (Seoul, Korea). The
amplicons were sequenced using the 454 FLX titanium system. The paired-end reads obtained
using NGS were assembled with FLASH software, and the sequencing reads were filtered and
trimmed using the CD-HIT-OUT software and rDNA Tools. For the taxonomic analysis,
operative taxonomic units (OTUs) were selected on the basis of 97% threshold of sequence
similarity with the QIIME-UCLUST program. The filtered reads were clustered, and OTUs
were generated using CD-HIT-DUP. The filtered sequences were analyzed using the QIIME
pipeline, which includes features to calculate diversity indices and phylogenetic diversity (PD)
rarefaction curves. The diversity indices (OTUs, Chao1, Shannon, and Simpson index) were
measured. The Ribosomal Database Project classifier was used for taxonomic classification of
the fecal microbiome of the control and treatment groups.
2.2.6. Statistical analysis. The data were analyzed using the GLM procedure of SAS (SAS
Institute Inc., Cary, NC, USA) in a randomized complete block design. The statistical model
for growth performance and digestibility included effects of dietary treatment as fixed effects
and initial BW as a covariate. Statistical significance and tendency were considered at P < 0.05
and 0.05 � P < 0.10, respectively. The in vitro digestibility results were analyzed and compared
using the t-test and SPSS software. A P value < 0.05 was considered statistically significant.
Results
3.1. In vitro digestibility study
3.1.1. IVID. The IVID of DM without multi-enzyme supplementation for corn meal, Ori-
ental herbal extract, wheat meal, soybean meal, fish meal, IRG, and peanut hull was 95.24%,
85.91%, 83.02%, 76.12%, 67.19%, 31.6%, and 8.85%, respectively; IVID of DM with multi-
enzyme supplementation was 97.66%, 87.62%, 85.76%, 76.29%, 66.56%, 31.44%, and 8.61%,
respectively (Table 2). Among all the tested feed stuffs, corn meal showed a significant
(P = 0.01) increase in DM digestibility with multi-enzyme supplementation. DM digestibility
of wheat meal, soybean meal and Oriental herbal extract (P > 0.05) quantitatively increased
with multi-enzyme supplementation (S1 Fig). However, IVID of DM for fish meal slightly
Table 2. IVID of feed ingredients (values are mean ± standard deviation, n = 3).
Feed Ingredient
Wheat meal
Soybean meal
Fish meal
Oriental herbal extract
Peanut hull
Corn meal
IRG
Control mean ± SD
83.02 ± 5.17
76.12 ± 0.51
67.19 ± 1.09
85.91 ± 0.54
8.85 ± 0.31
95.23 ± 0.17
31.6 ± 0.78
Treatment mea ± SD
85.76 ± 2.41
76.29 ± 0.57
66.56 ± 1.36
87.62 ± 1.29
8.61 ± 0.51
97.67 ± 0.27
31.44 ± 0.45
P-value
0.452
0.719
0.564
0.102
0.529
0.001
0.769
SD, standard deviation; Control, without enzyme supplementation; Treatment, with enzyme supplementation IRG, Italian ryegrass
https://doi.org/10.1371/journal.pone.0217459.t002
PLOS ONE | https://doi.org/10.1371/journal.pone.0217459 May 28, 2019
7 / 19
The effect of enzyme complex supplement on pig microbiota and performance
Table 3. IVTTD of feed ingredients (values are mean ± standard deviation, n = 3).
Feed Ingredient
Wheat meal
soybean meal
fish meal
Oriental herbal extract
peanut hull
Corn meal
IRG
Control mean ± SD
84.33 ± 3.05
90.64 ± 0.37
85.67 ± 1.10
95.54 ± 0.35
11.8 ± 0.9
88.92 ± 0.39
33.6 ± 0.43
Treatment mean ± SD
87.47 ± 1.08
91.32 ± 0.23
84.93 ± 0.85
96.84 ± 0.12
9.6 ± 0.76
97.6 ± 0.18
33.14 ± 0.72
P-value
0.169
0.053
0.411
0.004
0.032
0.003
0.395
SD, standard deviation; Control, without enzyme supplementation; Treatment, with enzyme supplementation IRG, Italian ryegrass
https://doi.org/10.1371/journal.pone.0217459.t003
(0.63%) decreased (P > 0.05). The in vitro ileal digestibility of peanut hull and IRG did not
change with multi-enzyme supplementation.
3.1.2. IVTTD. The IVTTD of DM without multi-enzyme supplementation for Oriental
herbal extract, soybean meal, corn meal, fish meal, wheat meal, IRG, and peanut hull was
95.54%, 90.64%, 88.91%, 85.67%, 84.33%, 33.6% and 11.8%, respectively; the IVTTD of DM
with multi-enzyme supplementation was 96.84%, 91.32%, 97.61%, 84.93%, 87.47%, 33.14%
and 9.6%, respectively (Table 3). On the basis of in vitro analysis, IVTTD of DM for corn
meal (P = 0.0039) and Oriental herbal extract (P = 0.004) increased significantly (S2 Fig).
However, the digestibility of peanut hull decreased (P = 0.032) with multi-enzyme supple-
mentation. Among all the feed-stuffs, corn meal had the highest IVTTD, followed by Orien-
tal herbal extract, and peanut hull which exhibited the lowest digestibility with multi-enzyme
supplementation.
3.2. In vivo study
3.2.1. Nutrient digestibility and apparent ileal and total tract digestibility. The appar-
ent ileal and total tract digestibility of nutrients from control and treatment are provided in
Table 4. Digestibility of DM, energy, and CP in pigs fed with control diet were not statistically
different (P > 0.05) from that of pigs fed with treatment diet. Apparent ileal digestibility of
DM in the control and treatment groups was similar (78.34% and 78.44%, respectively). The
values of apparent ileal digestibility of energy in the control and treatment groups (71.86% and
Table 4. Apparent digestibility of nutrients in pigs fed diets with multi-enzyme supplementation.
Item
Controlx
Treatmenty
SEMz
P-Value
78.44
72.56
72.14
80.78
81.44
75.46
0.4
0.35
0.28
0.52
1.45
1.32
0.913
0.245
0.546
0.908
0.678
0.528
Apparent ileal tract digestibility, %
DMa
CPb
GEc
Apparent total tract digestibility, %
71.86
78.34
71.75
DM
CP
GE
80.45
80.76
74.76
Controlx; without multi-enzyme supplement
Treatmenty; with multi-enzyme supplement
SEMz; standard error of mean
DMa; dry matter CPb; crude protein GEc; gross energy
https://doi.org/10.1371/journal.pone.0217459.t004
PLOS ONE | https://doi.org/10.1371/journal.pone.0217459 May 28, 2019
8 / 19
Table 5. Growth performance of weaned pigs fed dietary treatment1.
The effect of enzyme complex supplement on pig microbiota and performance
Items
Phase 1 (d 1–21)
Initial BW, kg
Final BW, kg
Feed intake, kg
ADG, g/d
ADFI, g/d
G: F, g/g
Phase 2 (d 21–42)
Initial BW, kg
Final BW, kg
Feed intake, kg
ADG, g/d
ADFI, g/d
G: F, g/g
Overall (d 1–42)
Initial BW, kg
Final BW, kg
Feed intake, kg
ADG, g/d
ADFI, g/d
G: F, g/g
Controlx
Treatmenty
SEMz
P-value
6.02
13.41
35.15
352
558
0.631
13.41
25
53.93
552
856
0.645
6.02
25
89.08
452
707
0.639
6.03
14.09
37.49
384
595
0.645
14.09
26.36
55.19
584
876
0.667
6.03
26.36
92.67
484
735.5
0.658
0.42
0.54
1.78
28.12
22.29
0.034
0.54
1.54
3.81
27.65
34.75
0.032
0.42
1.54
4.45
28.91
32.68
0.031
0.975
0.714
0.485
0.371
0.456
0.514
0.714
0.465
0.712
0.765
0.647
0.698
0.975
0.465
0.624
0.601
0.374
0.862
1Values are presented as the least squares mean of 6 replicates (3 pigs/replicate).
Controlx = diet based on corn and soybean meal.
Treatmenty = control with 0.1% multi-enzyme.
SEMz = standard error of mean.
https://doi.org/10.1371/journal.pone.0217459.t005
72.14%, respectively) were similar to apparent ileal digestibility of CP (71.75% and 72.56%,
respectively). The apparent total tract digestibility of DM, CP, and GE for control group was
80.45%, 80.76% and 74.76%, respectively, and 80.78%, 81.44%, and 75.46% for the treatment,
respectively, and these values were not significantly (P > 0.05) different between the groups.
3.2.2. Growth performance. The average initial BW of piglets was 6.02 ± 0.32 kg and final
BW of pigs in control and treatment group was 25 and 26.36 kg, respectively. Multi-enzyme
supplementation had no significant effect on feed intake, ADFI, feed conversion efficiency, or
growth performance of pigs (P > 0.05; Table 5).
3.2.3. Microbial community analysis. 3.2.3a. DNA sequence data and bacterial diver-
sity: The pyrosequencing analyses generated a total of 523,130 valid sequences for the control
and treatment groups by using the fecal samples of 12 pigs. The average OTUs at 97% confi-
dence intervals were 514 OTUs for the control group and 479 OTUs for the treatment group
(Fig 2a). Microbial diversity was calculated using the Shannon-Weaver and Simpson diversity
indices (Fig 2b and 2c). The average Shannon-Weaver index of the control and treatment
groups was 6.22 (SD = 0.16) and 6.07 (SD = 0.17), respectively. The Simpson index values of
the control and treatment groups were 0.97 (SD = 0.005) and 0.96 (SD = 0.009), respectively.
The alpha diversity measurement, chao1 estimator of total species richness value was 585 in
the control and 556 in the treatment group, but not statistically significant (P = 0.208) differ-
ences were observed (Fig 2d). These results indicate that the treatment diet did not affect the
PLOS ONE | https://doi.org/10.1371/journal.pone.0217459 May 28, 2019
9 / 19
The effect of enzyme complex supplement on pig microbiota and performance
Fig 2. Variations in alpha diversity of pigs.
https://doi.org/10.1371/journal.pone.0217459.g002
diversity of microbiota in the gut (S1 Table). The total number of observed species in the con-
trol and treatment groups was presented using rarefaction curves (Fig 3).
3.2.3b. Taxonomic analysis: The results of the taxonomic analysis at the phylum level are
shown in Fig 4. The bacteria in both control and treatment samples were predominantly
belonged to Firmicutes and Bacterioidetes, which account for more than 65% of the total
sequences. At the phylum level, the bacteria in the control group belonged primarily to the
phyla Bacteroidetes (34.99%), Firmicutes (34.73%), Spirochaetes (13.50%), and Proteobacteria
(3.72%); the other phyla and non-bacteria comprised 2.3% and 10.69%, respectively, of the
total sequences analyzed. The bacteria in the treatment group primarily belonged to the phyla
Firmicutes (38.52%), Bacteroidetes (32.82%), Spirochaetes (14.00%), Proteobacteria (1.55%)
and Euryarchaeota (1.30%); the non-bacteria comprised 10.97% of the total analyzed
sequences. Both groups shared similar phyla, with a trend toward higher abundance of
PLOS ONE | https://doi.org/10.1371/journal.pone.0217459 May 28, 2019
10 / 19
The effect of enzyme complex supplement on pig microbiota and performance
Fig 3. Rarefaction curves of observed species in groups. Control = without multi-enzyme; Treatment = with multi-
enzyme.
https://doi.org/10.1371/journal.pone.0217459.g003
Firmicutes and a corresponding decrease in the abundance of Bacteroidetes and Proteobac-
teria in the treatment group.
At the class level (Fig 5), the proportion of Clostridia increased and the proportion of Bac-
teroidia and Gammaproteobacteria decreased in the treatment group. At the genus level (Fig
6), a total of 135 genera were identified. In the control group, more than 70% of the total
sequences belonged to 13 genera: Meniscus, Barnesiella, Porphyromonas, Prevotella, Lactoba-
cillus, Christensenella, Clostridium, Roseburia, Oscillibacter, Sporobacter, Mitsuokella, Succi-
nivibrio, and Treponema. Prevotella (13.71%) and Treponema (13.13%) were the most
abundant genera. A total of 124 genera were identified in the treatment group, and more
than 65% of the total sequences belonged to 13 genera: Methanobrevibacter, Meniscus, Barne-
siella, Porphyromonas, Prevotella, Alistipes, Lactobacillus, Christensenella, Roseburia, Oscilli-
bacter, Sporobacter, Phascolarctobacterium, and Treponema. Barnesiella, Prevotella, and
Treponema were the most abundant genera. However, the proportion of Prevotella decreased
from 13.71% (control) to 8.06% in the treatment group, and the proportion of Barnesiella
and Treponema slightly increased in the treatment group. The proportion of other genera,
including Meniscus, Butyricicoccus, Ruminococcus, and Succinivibrio, decreased in the treat-
ment group (S2 Table).
Discussion
4.1. In vitro DM digestibility of feed-stuffs
In vitro ileal and total tract digestibility methods are effective techniques to evaluate the feed
efficiency before using an animal diet. When compared with animal experiments, these tech-
niques are cost–effective, rapid, and repeatable [24, 29]. In this study, we examined the DM
digestibility of feed-stuffs that can be used as a basal diet (corn, fish, wheat, and soybean meal)
PLOS ONE | https://doi.org/10.1371/journal.pone.0217459 May 28, 2019
11 / 19
The effect of enzyme complex supplement on pig microbiota and performance
Fig 4. Bacterial taxonomic composition of phylum level.
https://doi.org/10.1371/journal.pone.0217459.g004
and feed additives (Oriental herbal extract, IRG, and peanut hull). Numerous studies have
reported the effects of exogenous supplementation of dietary enzymes (single enzyme/ multi-
enzymes) on pig nutrition [7, 30]. In the present study, we used in vitro digestibility studies to
evaluate feed stuffs with or without multi-enzyme supplementation. The multi-enzyme prepa-
ration mainly contained NSP-degrading enzymes (xylanase and glucanase). The purpose of
this study was primarily to screen the digestibility of these feed ingredients for further animal
experiments.
The multi-enzyme supplementation did not influence the IVID of DM digestibility for all
the feed–stuffs, except corn meal. The corn meal ileal DM digestibility increased (P = 0.001)
by 2% with enzyme addition. Park et al. [31] have reported similar findings. The results of
IVTTD of DM digestibility for soybean, corn, and wheat meal were similar to those of previous
PLOS ONE | https://doi.org/10.1371/journal.pone.0217459 May 28, 2019
12 / 19
The effect of enzyme complex supplement on pig microbiota and performance
Fig 5. Bacterial taxonomic compositions of class level.
https://doi.org/10.1371/journal.pone.0217459.g005
studies [23, 31, 32]. The average IVTTD of DM digestibility for IRG (33.6%) was similar to
that reported by Anderson and Ralston [33]. The DM digestibility of peanut hull decreased
with multi-enzyme addition in this study. Similar results have been reported previously [34].
Because of a high proportion of crude fiber (60–67%) and only 6–7% crude protein, the digest-
ibility of peanut hull is less than 20% in ruminants [35]. Lindermann et al. [36] studied the
effects of peanut hull substitution (7.5, 15, and 30%) in swine diets. They reported no differ-
ences in daily gain among the treatments.
PLOS ONE | https://doi.org/10.1371/journal.pone.0217459 May 28, 2019
13 / 19
The effect of enzyme complex supplement on pig microbiota and performance
Fig 6. Bacterial taxonomic compositions of genus level.
https://doi.org/10.1371/journal.pone.0217459.g006
4.2. Effects of enzyme addition on the digestibility and growth
performance of pigs
Corn meal and soybean meal are the main energy and protein sources in typical swine diets.
All plant-based feed-stuffs have some ANF such as cellulose, gums, and hemicelluloses (ara-
binoxylans, mannan, and glucomannan). Corn contains 9.7% NSP, mainly 4.3% arabinoxy-
lans and 0.21% phytate P [37, 38, 39]. Soybean meal has 21.7% NSP, mainly α-galactosides
and β-galactomannan [40] and 0.38% phytate P [25]. However, NSP also has beneficial
effects due to microbial fermentation of NSP in the hindgut [41]. The effects of NSP-degrad-
ing enzymes as feed additives for pigs have been reported in high fiber diets with wheat,
rye, barley, and rice [30, 42]. However, only a few studies have investigated the effects of
enzyme supplementation on corn-soybean meal diets [39, 43]. Therefore, the objective of
this study was to investigate the effects of multi-enzyme supplementation of corn-soybean
meal-based diets for pigs. Quantitatively, multi-enzyme supplementation increased
(P > 0.05) the apparent total tract DM digestibility, energy, and CP digestibility. Jo et al. [39]
have reported similar results. In contrast, Omogbenigun et al. [44] reported that the pigs fed
enzyme-supplemented diets had higher (P = 0.001 to 0.014) total-tract digestibility of DM,
CP, and GE than those fed the control diet. They used many feed stuffs as a basal diet, includ-
ing corn and soybean meal.
The results of the growth performance experiment are consistent with those of previous
studies [39, 45]. In the present study, the results of growth performance did not show signifi-
cant (P > 0.05) differences among two groups. Similarly, Willami et al. [46] reported that
enzyme supplementation to corn- soybean-based diets exerted no effect on the nutrient digest-
ibility and growth performance of growing pigs. However, enzyme supplementation of high
NSP containing wheat-rye barley-based diets showed positive effects on growth performance.
4.3. Effect of enzyme addition on the pig gut microbial diversity and
community changes
The mammalian gut microbiota is a complex and diverse ecosystem composed of different
microbial communities and there is evidence that the gut microbiota plays an important role
in host health [47, 48]. The symbiotic relationship between the host and gut microbiota is well
PLOS ONE | https://doi.org/10.1371/journal.pone.0217459 May 28, 2019
14 / 19
The effect of enzyme complex supplement on pig microbiota and performance
established [49, 50]. The composition of the microbiota can be affected by several factors such
as age, diet, environment and host genetics. Few studies have examined the changes in micro-
bial communities in response to various diets in pigs [51, 52]. In this study, we examined
microbial community changes in young pigs fed a basal diet supplemented with multi-enzyme
complex by using NGS. As shown in previous studies, gram-positive Firmicutes followed by
gram-negative Bacterioidetes were the most dominant phyla in both groups [53]. The abun-
dance of Firmicutes increased in the treatment group. The ratio of Firmicutes to Bacteroidetes
shifts according to BW, and this proportion increases with weight gain [54, 55, 56]. The abun-
dance of Proteobacteria, which includes various pathogen species, such as Campylobacter and
Succinivibrio, declined with the enzyme supplementation. At the genus level, Prevotella and
Treponema were dominant in both groups. Prevotella is mainly composed of gram-negative
bacteria in the gastro-intestinal tract of pigs. The results were confirmed using previous culture
[57] and DNA-based culture-independent studies [58, 59, 60, 61].
Dietary enzyme supplementation resulted in the higher populations of two genera, Trepo-
nema (Spirochaetes) with abundant Treponema porcinum, and Barnesiella (Bacteroidetes)
with Barnesiella intestinihominis. In contrast, Zhang et al. [52] demonstrated that the abun-
dance of Treponema decreased with enzyme supplementation in pigs fed wheat bran-based
(WB) and soybean hull-based (SH) diets, and they observed that the proportion of Treponema
was higher in the SH pigs than in the WB diet fed pigs. Furthermore, the treatment group
showed decreased numbers of Prevotella (Prevotella copri), Clostridium, Butyricicoccus, Menis-
cus, and Succinivibrio, which can act as pathogens and negatively affect the gut immunity. For
instance, Prevotella species such as Prevotella intermedia and Prevotella copri can act as poten-
tial opportunistic pathogens [62, 63]. The Shannon index, which represents the species rich-
ness and evenness, and Simpson’s index, which accounts for proportional abundance or
probability, were similar in both groups. Similarly, Zhang et al. [52] did not find significant
differences in microbial diversity between two different fibrous diets with or without enzyme
supplementation.
Conclusions
Enzyme addition increased the in vitro total tract digestibility of Oriental herbal extract and
corn meal. In the treatment group, multi-enzyme supplementation of the basal diet did not
influence the nutrient digestibility and growth performance of the pigs. However, the micro-
bial composition of the microbiota in the hindgut was modified by enzyme supplementation
of the corn-soybean meal diet.
Supporting information
S1 Fig. In vitro ileal dry matter digestibility rates of feed stuffs without multi-enzyme (C)
and with multi -enzyme (T). The results shown as mean ± standard deviation (n = 3). Signifi-
cance measured at P < 0.05 and shown as asterisk (�).
(TIF)
S2 Fig. In vitro total tract digestibility rates of feed stuffs without multi-enzyme (C) and
with multi- enzyme (T). The results shown as mean ± standard deviation (n = 3). Significance
measured at P < 0.05 and shown as asterisk (�).
(TIF)
S1 Table. Community richness and diversity between control and treatment groups.
(DOCX)
PLOS ONE | https://doi.org/10.1371/journal.pone.0217459 May 28, 2019
15 / 19
The effect of enzyme complex supplement on pig microbiota and performance
S2 Table. Relative abundance of taxa in control and treatment groups representing > 0.1%
of total sequences.
(DOCX)
Author Contributions
Conceptualization: Sungkwon Park.
Formal analysis: Sivasubramanian Ramani.
Investigation: Minho Song.
Methodology: Duwan Kim.
Project administration: Juncheol Park.
Supervision: Sungkwon Park.
Validation: Sungkwon Park.
Writing – original draft: Neeraja Recharla.
Writing – review & editing: Balamuralikrishnan Balasubramanian, Pradeep Puligundla,
Sungkwon Park.
References
1.
Jha R, Berrocoso JD. Review: Dietary fiber utilization and its effects on physiological functions and gut
health of swine. Animal. 2015; 9: 1441–1452. https://doi.org/10.1017/S1751731115000919 PMID:
25997437
2. Choct M. Feed non-starch polysaccharides: chemical structures and nutritional significance. Feed Mill-
ing International.1997; 191: 13–26.
3. Choct M, Dersjant-Li Y, McLeish J, Peisker M. Soy oligosaccharides and soluble non-starch polysac-
charides: a review of digestion, nutritive and anti-nutritive effects in pigs and poultry. Asian Austal J
Anim. 2010; 23: 1386–1398.
4. O’Neill HVM, Smith JA, Bedford MR. Multicarbohydrase enzymes for non-ruminants. Asian Austral J
Anim. 2014; 27: 290–301.
5. Yuan L, Wang M, Zhang X, Wang Z. Effects of protease and non-starch polysaccharide enzyme on per-
formance, digestive function, activity and gene expression of endogenous enzyme of broilers. Plos
One. 2017; 12.
6. Slominski BA. Recent advances in research on enzymes for poultry diets. Poult Sci. 2011; 90: 2013–23.
https://doi.org/10.3382/ps.2011-01372 PMID: 21844268
7. Patience JF, Bedford MR, Classen HL, Inborr J. The effect of dietary enzyme supplementation of rye-
and barley-based diets on digestion and subsequent performance in weanling pigs. Can J Anim Sci.
1992; 72: 97–105.
8. Ravindran V. Feed enzymes: The science, practice, and metabolic realities. J Appl Poult Res. 2013; 22:
628–636.
9. Upadhaya SD, Yun HM, Kim IH. Influence of low or high-density corn and soybean meal-based diets
and protease supplementation on growth performance, apparent digestibility, blood characteristics and
noxious gas emission of finishing pigs. Anim Feed Sci Tech. 2016; 216: 281–7.
10.
Tactacan GB, Cho SY, Cho JH, Kim IH. Performance responses, nutrient digestibility, blood character-
istics, and measures of gastrointestinal health in weanling pigs fed protease enzyme. Asian Austral J
Anim. 2016; 29: 998.
11. Wang H, Guo Y, Shih JC. Effects of dietary supplementation of keratinase on growth performance,
nitrogen retention and intestinal morphology of broiler chickens fed diets with soybean and cottonseed
meals. Anim Feed Sci Tech. 2008; 140: 376–84.
12. Prykhodko O, Pierzynowski SG, Nikpey E, Sureda EA, Fedkiv O, Westro¨ m BR. Pancreatic and pancre-
atic-like microbial proteases accelerate gut maturation in neonatal rats. Plos One 2015; 10: 2
PLOS ONE | https://doi.org/10.1371/journal.pone.0217459 May 28, 2019
16 / 19
The effect of enzyme complex supplement on pig microbiota and performance
13. Adeola O, Bedford MR. Exogenous dietary xylanase ameliorates viscosity-induced anti-nutritional
effects in wheat-based diets for White Pekin ducks (Anas platyrinchos domesticus). Brit J Nutr. 2004;
92: 87–94. https://doi.org/10.1079/BJN20041180 PMID: 15230991
14. Vahjen W, Osswald T, Scha¨ fer K, Simon O. Comparison of a xylanase and a complex of non-starch
polysaccharide-degrading enzymes with regard to performance and bacterial metabolism in weaned
piglets. Arch Anim Nutr. 2007; 61: 90–102. https://doi.org/10.1080/17450390701203881 PMID:
17451108
15.
Li S, Sauer WC, Huang SX, Gabert VM. Effect of β-glucanase supplementation to hulless barley- or
wheat-soybean meal diets on the digestibilities of energy, protein, β-glucans, and amino acids in young
pigs. J Anim Sci. 1996; 74: 1649–1656. https://doi.org/10.2527/1996.7471649x PMID: 8818811
16. Meng X, Slominski BA. Nutritive values of corn, soybean meal, canola meal, and peas for broiler chick-
ens as affected by a multicarbohydrase preparation of cell wall degrading enzymes. Poultry Sci. 2005;
84: 1242–1251.
17. Nortey TN, Patience JF, Simmins PH, Trottier NL, Zijlstra RT. Effects of individual or combined xyla-
nase and phytase supplementation on energy, amino acid, and phosphorus digestibility and growth per-
formance of grower swine fed wheat-based diets containing wheat millrun. J Anim Sci. 2007; 85: 1432–
1443. https://doi.org/10.2527/jas.2006-613 PMID: 17325125
18. Yin F, Zhang Z, Huang J, Yin Y. Digestion rate of dietary starch affects systemic circulation of amino
acids in weaned pigs. Br J Nutr. 2010; 103: 1404–1412. https://doi.org/10.1017/S0007114509993321
PMID: 20102672
19. Sieo CC, Abdullah N, Tan WS, Ho YW. Influence of β-glucanase-producing Lactobacillus strains on
intestinal characteristics and feed passage rate of broiler chickens. Poult Sci. 2005; 84: 734–41. https://
doi.org/10.1093/ps/84.5.734 PMID: 15913185
20. Owusu-Asiedu A, Simmins PH, Brufau J, Lizardo R, Peron A. Effect of xylanase and β-glucanase on
growth performance and nutrient digestibility in piglets fed wheat–barley-based diets. Livest Sci. 2010;
134: 76–78.
21. De Lange C. F. M, Pluske J, Gong J, Nyachoti C. M. Strategic use of feed ingredients and feed additives
to stimulate gut health and development in young pigs. Livest Sci. 2010; 134: 124–134.
22. Boisen S, Fernandez JA. Prediction of the apparent ileal digestibility of protein and amino acids in feed-
stuffs and feed mixtures for pigs by in vitro analyses. Anim Feed Sci Tech. 1995; 51: 29–43.
23. Boisen S, Ferna´ndez JA. Prediction of the total tract digestibility of energy in feedstuffs and pig diets by
in vitro analyses. Anim Feed Sci Tech. 1997; 68: 277–286.
24. Regmi PR, Ferguson NS, Zijlstra RT. In vitro digestibility techniques to predict apparent total tract
energy digestibility of wheat in grower pigs. J Anim Sci. 2009; 87: 3620–3629. https://doi.org/10.2527/
jas.2008-1739 PMID: 19648497
25. NRC. Nutrient requirements of swine. 11th revised ed. Washington, DC: National Academy press;
2012.
26. Williams CH, David DJ, Iismaa O. The determination of chromic oxide in faeces samples by atomic
absorption spectrophotometry. J Agri Sci. 1962; 59: 381–5.
27. AOAC. Official methods of analysis, 15th ed. Arlington, VA; 2000
28. Adeola O. Digestion and balance techniques in pigs. In: Lewis AJ, Southern LL, editors. Swine nutrition.
Boca Raton, FL: CRC Press; 2001. pp. 903–916.
29.
Jha R, Tiwari UP. Rapid techniques for feed evaluation: scope and limitations. In: Proceedings of the
NZ Poultry Industry Conference. 2016;13: 84–102.
30. Ho¨gberg A, Lindberg JE. Influence of cereal non-starch polysaccharides and enzyme supplementation
on digestion site and gut environment in weaned piglets. Anim Feed Sci Tech. 2004; 116: 113–128.
31. Park KR, Park CS, Kim BG. An enzyme complex increases in vitro dry matter digestibility of corn
and wheat in pigs. Springerplus. 2016; 5: 598. https://doi.org/10.1186/s40064-016-2194-5 PMID:
27247894
32. Kong C, Park CS, Kim BG. Effects of an enzyme complex on in vitro dry matter digestibility of feed ingre-
dients for pigs. Springer plus. 2015; 4: 261. https://doi.org/10.1186/s40064-015-1060-1 PMID:
26090308
33. Anderson DC, Ralston AT. Chemical treatment of ryegrass straw: in vitro dry matter digestibility and
compositional changes. J Anim Sci.1973; 37: 148–152.
34. Kornegay ET, Rhein-Welker D, Lindemann MD, Wood CM. Performance and nutrient digestibility in
weanling pigs as influenced by yeast culture additions to starter diets containing dried whey or one of
two fiber sources. J Anim Sci. 1995; 73: 1381–1389. https://doi.org/10.2527/1995.7351381x PMID:
7665367
PLOS ONE | https://doi.org/10.1371/journal.pone.0217459 May 28, 2019
17 / 19
The effect of enzyme complex supplement on pig microbiota and performance
35. Kerr TJ, Windham WR, Woodward JH, Benner R. Chemical composition and in-vitro digestibility of ther-
mochemically treated peanut hulls. J Sci Food Agric. 1986; 37: 632–6.
36.
Lindemann MD, Kornegay ET, Moore RJ. Digestibility and feeding value of peanut hulls for swine. J
Anim Sci. 1986; 62: 412–21.
37. Knudsen KEB. Carbohydrate and lignin contents of plant materials used in animal feeding. Anim Feed
Sci Tech. 1997; 67: 319–338.
38.
dos Passos AA, Kim SW. Use of enzymes in pig diets. In: VI Congresso Latino Americano de Nutricao
Animal-SALA SUINOS; 2014.
39. Ward NE, Slominski BA, Fernandez SR. Nonstarch polysaccharide content of corn grain. Poult Sci.
2008; 87; 91.
40.
Jo JK, Ingale SL, Kim JS, Kim YW, Kim KH, Lohakare JD, et al. Effects of exogenous enzyme supple-
mentation to corn-and soybean meal-based or complex diets on growth performance, nutrient digestibil-
ity, and blood metabolites in growing pigs. J Anim Sci. 2012; 90: 3041–3048. https://doi.org/10.2527/
jas.2010-3430 PMID: 22585797
41. Bakker GCM, Dekker RA, Jongbloed R, Jongbloed AW. Non-starch polysaccharides in pig feeding. Vet
Quart. 1998; 20: 59–64.
42. Yin YL, Baidoo SK, Schulze H, Simmins PH. Effects of supplementing diets containing hulless barley
varieties having different levels of non-starch polysaccharides with β-glucanase and xylanase on the
physiological status of the gastrointestinal tract and nutrient digestibility of weaned pigs. Livest Prod
Sci. 2001; 71: 97–107.
43. Shim YH, Chae BJ, Lee JH. Effects of phytase and carbohydrases supplementation to diet with a partial
replacement of soybean meal with rapeseed meal and cottonseed meal on growth performance and
nutrient digestibility of growing pigs. Asian Austral J Anim. 2003; 16: 1339–47.
44. Omogbenigun FO, Nyachoti CM, Slominski BA. Dietary supplementation with multienzyme prepara-
tions improves nutrient utilization and growth performance in weaned pigs. J Anim Sci. 2004; 82: 1053–
1061. https://doi.org/10.2527/2004.8241053x PMID: 15080327
45.
l’Anson KA, Choct M, Brooks PH. Effect of feed processing and enzyme supplementation on diet digest-
ibility and performance of male weaner pigs fed wheat-based diets in dry or liquid form. Anim Prod Sci.
2013; 53: 531–539.
46. Willamil J, Badiola I, Devillard E, Geraert PA, Torrallardona D. Wheat-barley-rye-or corn-fed growing
pigs respond differently to dietary supplementation with a carbohydrase complex. J Anim Sci. 2012; 90:
824–32. https://doi.org/10.2527/jas.2010-3766 PMID: 22345107
47. Nicholson JK, Holmes E, Wilson ID. Gut microorganisms, mammalian metabolism and personalized
health care. Nat Rev Microbiol. 2005; 3: 431–438. https://doi.org/10.1038/nrmicro1152 PMID:
15821725
48.
Lozupone CA, Stombaugh JI, Gordon JI, Jansson JK, Knight R. Diversity, stability and resilience of the
human gut microbiota. Nature 2012; 489: 220–230. https://doi.org/10.1038/nature11550 PMID:
22972295
49. Hooper LV, Gordon JI. Commensal host-bacterial relationships in the gut. Science. 2001; 292: 1115–
1118. PMID: 11352068
50. Hooper LV, Littman DR, Macpherson AJ. Interactions between the microbiota and the immune system.
Science. 2012; 336: 1268–1273. https://doi.org/10.1126/science.1223490 PMID: 22674334
51. Yang Y, Kiarie E, Slominski BA, Bruˆle´ -Babel A, Nyachoti CM. Amino acid and fiber digestibility, intesti-
nal bacterial profile, and enzyme activity in growing pigs fed dried distillers grains with solubles-based
diets. J Anim Sci. 2010; 88: 3304–12. https://doi.org/10.2527/jas.2009-2318 PMID: 20525935
52.
Zhang YJ, Liu Q, Zhang WM, Zhang ZJ, Wang WL, Zhuang S. Gastrointestinal microbial diversity and
short-chain fatty acid production in pigs fed different fibrous diets with or without cell wall-degrading
enzyme supplementation. Livest Sci. 2018; 207: 105–16.
53. Kim HB, Isaacson RE. The pig gut microbial diversity: understanding the pig gut microbial ecology
through the next generation high throughput sequencing. Vet Microbiol. 2015; 177: 242–251. https://
doi.org/10.1016/j.vetmic.2015.03.014 PMID: 25843944
54.
Ley RE, Turnbaugh PJ, Klein S, Gordon JI. Microbial ecology: human gut microbes associated with
obesity. Nature. 2006; 444: 1022–1023. https://doi.org/10.1038/4441022a PMID: 17183309
55. De Filippo C, Cavalieri D, Di Paola M, Ramazzotti M, Poullet JB, Massart S, et al. Impact of diet in shap-
ing gut microbiota revealed by a comparative study in children from Europe and rural Africa. P Natl
Acad Sci USA. 2010; 107: 14691–14696.
56. Koliada A, Syzenko G, Moseiko V, Budovska L, Puchkov K, Perederiy V, et al. Association between
body mass index and Firmicutes/Bacteroidetes ratio in an adult Ukrainian population. BMC Microbiol.
2017; 17: 120. https://doi.org/10.1186/s12866-017-1027-1 PMID: 28532414
PLOS ONE | https://doi.org/10.1371/journal.pone.0217459 May 28, 2019
18 / 19
The effect of enzyme complex supplement on pig microbiota and performance
57. Robinson IM, Whipp SC, Bucklin JA, Allison MJ. Characterization of predominant bacteria from the
colons of normal and dysenteric pigs. Appl Environ Microb. 1984; 48: 964–969.
58. Pryde SE, Richardson AJ, Stewart CS, Flint HJ. Molecular analysis of the microbial diversity present in
the colonic wall, colonic lumen, and cecal lumen of a pig. Appl Environ Microb. 1999; 65: 5372–5377.
59.
60.
Leser TD, Amenuvor JZ, Jensen TK, Lindecrona RH, Boye M, Moller K. Culture-independent analysis
of gut bacteria: the pig gastrointestinal tract microbiota revisited. Appl Environ Microb. 2002; 68: 673–
690.
Liu H, Ivarsson E, Dicksved J, Lundh T, Lindberg JE. Inclusion of chicory (Cichorium intybus L.) in pigs’
diets affects the intestinal microenvironment and the gut microbiota. Appl Environ Microb. 2012; 78:
4102–4109.
61. Ramayo-Caldas Y, Mach N, Lepage P, Levenez F, Denis C, Lemonnier G, et al. Phylogenetic network
analysis applied to pig gut microbiota identifies an ecosystem structure linked with growth traits. ISME
J. 2016; 10: 2973–2977. https://doi.org/10.1038/ismej.2016.77 PMID: 27177190
62. Ruan Y, Shen L, Zou Y, Qi Z, Yin J, Jiang J, et al. Comparative genome analysis of Prevotella interme-
dia strain isolated from infected root canal reveals features related to pathogenicity and adaptation.
BMC Genomics. 2015; 16(1):122.
63.
Toscano M, De Grandi R, Miniello VL, Mattina R, Drago L. Ability of Lactobacillus kefiri LKF01
(DSM32079) to colonize the intestinal environment and modify the gut microbiota composition of
healthy individuals. Digest Liver Dis. 2017; 49: 261–7.
PLOS ONE | https://doi.org/10.1371/journal.pone.0217459 May 28, 2019
19 / 19
|
10.1371_journal.pone.0229102 | RESEARCH ARTICLE
Venous hematology, biochemistry, and blood
gas analysis of free-ranging Eastern
Copperheads (Agkistrodon contortrix) and
Eastern Ratsnakes (Pantherophis
alleghaniensis)
Anthony J. CerretaID
1☯¤, Sarah A. Cannizzo2☯, Dustin C. Smith3☯, Larry J. MinterID
1,3☯*
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
1 Department of Clinical Sciences, College of Veterinary Medicine, North Carolina State University, Raleigh,
North Carolina, United States of America, 2 Fort Worth Zoo, Fort Worth, Texas, United States of America,
3 North Carolina Zoo, Asheboro, North Carolina, United States of America
☯ These authors contributed equally to this work.
¤ Current address: Department of Small Animal Clinical Sciences, College of Veterinary Medicine, University
of Tennessee, Knoxville, Tennessee, United States of America
* Jb.Minter@nczoo.org
Abstract
Hematology, plasma biochemistry, and blood gas analysis were performed on venous sam-
ples obtained from free-ranging Eastern Copperheads (Agkistrodon contortrix) and Eastern
Ratsnakes (Pantherophis alleghaniensis) in central North Carolina during a mark-recapture
study conducted from April to October 2015 at the North Carolina Zoo. Blood samples were
collected from 31 (15 male and 16 female) free-ranging copperheads and 34 (20 male and
14 female) free-ranging ratsnakes at the beginning and end of restraint. Restraint was per-
formed for morphometric measurements, sex determination, and identification via place-
ment of intracelomic passive integrated transponder (PIT) tags and marking of ventral
scutes with a handheld electrocautery unit. Blood gas analytes were measured at the begin-
ning of restraint and compared to analytes measured at the end to evaluate for changes sec-
ondary to handling. Total restraint time prior to the first blood sampling was 1.4 ± 0.4 mins
(mean ± SD) and 1.0 ± 0.2 mins (mean ± SD) and restraint time prior to second blood sam-
pling was 12.5 ± 2.4 mins (mean ± SD) and 13.5 ± 3.4 mins (mean ± SD) for copperheads
and ratsnakes, respectively. Blood lactate concentrations at the beginning of restraint were
similar for both species. Lactate concentrations increased significantly and pH decreased
significantly for both species at the end of restraint when compared to the beginning of
restraint. Furthermore, lactate concentrations at the end of restraint were significantly ele-
vated in ratsnakes compared to copperheads. This study provides guidelines for interpreta-
tion of venous hematology, plasma biochemistry, and blood gas values for free-ranging
copperheads and ratsnakes in central North Carolina and demonstrates the physiological
response to venous blood gas analytes secondary to capture and restraint.
OPEN ACCESS
Citation: Cerreta AJ, Cannizzo SA, Smith DC,
Minter LJ (2020) Venous hematology,
biochemistry, and blood gas analysis of free-
ranging Eastern Copperheads (Agkistrodon
contortrix) and Eastern Ratsnakes (Pantherophis
alleghaniensis). PLoS ONE 15(2): e0229102.
https://doi.org/10.1371/journal.pone.0229102
Editor: Pal Bela Szecsi, Copenhagen University
Hospital Holbæk, DENMARK
Received: October 21, 2019
Accepted: January 29, 2020
Published: February 14, 2020
Copyright: © 2020 Cerreta et al. This is an open
access article distributed under the terms of the
Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: All relevant data are
within the paper and its Supporting Information
files.
Funding: The authors have no support or funding
to report.
Competing interests: The authors have declared
that no competing interests exist.
PLOS ONE | https://doi.org/10.1371/journal.pone.0229102 February 14, 2020
1 / 15
Venous Analytes of Copperheads and Ratsnakes
Introduction
Eastern Copperheads (Agkistrodon contortrix), a venomous snake of the subfamily Crotalinae
in the family Viperidae and the nonvenomous colubrid Pantherophis alleghaniensis, com-
monly called the Eastern Ratsnake, are endemic to North America and commonly exhibited in
natural history museums, zoological facilities, and aquariums throughout the United States.
Despite their popularity in managed collections, published hematologic, biochemical, and
blood gas reference values for these species are limited. Baseline physiological data serve as an
important reference for the health assessment of both free-ranging and managed populations
[1–4]. The increased availability of portable point-of-care analyzers in veterinary medicine has
facilitated evaluation of clinical pathology analytes, thus permitting plasma biochemistry and
venous blood gas analysis to be conducted in field settings [5,6].
In addition to assessing health status in the field, data from point-of-care analyzers can be
used to evaluate the physiological effects of different capture methods on free-ranging animals.
Increases in plasma lactate levels have been attributed to stressful events in both humans and
avian species [7,8]. This increase in lactate concentration has also been reported in sea turtles
following two separate capture techniques and during the manual restraint of free-ranging
birds for banding and morphometric data collection [9–11]. Rising lactate concentrations
often indicate anaerobic metabolism, thus lactate measurements by portable point-of-care ana-
lyzers may provide valuable information regarding decreased tissue perfusion and the physio-
logical effects of exertion or restraint [12].
In this study, hematological, biochemical, and blood gas parameters from 31 copperheads
and 34 ratsnakes were evaluated to develop guidelines for the interpretation of these parame-
ters in these species. Venous blood gas analytes were assessed at the beginning and end of
restraint to evaluate the impact of capture and restraint on these values. It was hypothesized
that lactate concentration, a marker associated with exertion and stress, would significantly
increase between the beginning and end of handling due to the stress of restraint in both spe-
cies and that there would be a concurrent decrease in pH. It was hypothesized that there would
be no difference in lactate concentration or pH between the two species.
Materials and methods
Ethics statement
This research was conducted within guidelines and approval of the North Carolina Zoo
Research Committee. All handling and sampling procedures were consistent with standard
vertebrate protocols and veterinary practices, and all efforts were made to minimize pain.
Animals, sample collection, and sample handling
This study was performed in conjunction with a mark-recapture study conducted on the
grounds of the North Carolina Zoo (Asheboro, NC) between 3 April 2015 and 12 October
2015. Peripheral venous blood samples were collected from 31 (15 male and 16 female) free-
ranging copperheads and 34 (20 male and 14 female) free-ranging ratsnakes at the beginning
and end of restraint. Restraint was performed for morphometric measurements, sex determi-
nation, and identification via placement of intracelomic passive integrated transponder (PIT)
tags and marking of ventral scutes with a handheld electrocautery unit.
Zoo employees or visitors reported snake sightings, and personnel trained to handle ven-
omous snakes responded to each sighting. All snakes were manually restrained. Non-venom-
ous species were manually restrained by grasping snakes at the base of the head and
supporting the body of the snake, as previously described [13]. Venomous species were
PLOS ONE | https://doi.org/10.1371/journal.pone.0229102 February 14, 2020
2 / 15
Venous Analytes of Copperheads and Ratsnakes
manually restrained with the head and cranial half of the body in a clear, open-ended, acrylic
tube (McMaster-Carr, Douglasville, GA) according to previously described methods [14].
Blood was collected from the ventral coccygeal vein with a 3 mL non-heparinized syringe
using a 22-gauge or 25-gauge needle. The first blood sample was collected at the beginning of
restraint. Next, cloacal probing was performed to determine the sex of each snake [15]. Mor-
phometric measurements for each snake were then obtained by weighing and measuring total
length (TL) and snout-vent length (SVL) using a flexible tape measure. Each snake was given
two unique identifiers. Passive integrated transponders (PIT) (Biomark, Boise, ID) were placed
in the caudal coelomic cavity and the ventral scutes were marked with a handheld electrocau-
tery unit corresponding to the snake’s identification number [16,17]. After the morphometric
measurements, sex determination, and identification procedures were complete, a second
blood sample was obtained from the ventral coccygeal vein as previously described. The total
blood sample for both collections was less than 1% of the body mass. The total time (mins)
from first contact to collection of the first blood sample, and from first blood collection to col-
lection of the second blood sample, were recorded. Based on examinations performed in the
field, no abnormalities were detected on physical examination. Snakes were returned to their
collection site immediately after the second blood sample was acquired. Individual snakes
were only included in this study once.
Hematology, plasma biochemistry, and venous blood gas analysis were performed on the
first blood sample (beginning-restraint). The second sample (end-restraint) was used for
venous blood gas analysis. Blood smears were made in the field immediately after blood collec-
tion from the non-heparinized syringe. The remaining blood was placed into a lithium hepa-
rin-coated microtainer tube (BD Microtainer, Becton Dickinson, Franklin Lake, NJ). Any
samples that had visible evidence of a blood clot or lymph contamination during venipuncture
were excluded.
Hematology. Non-heparinized microhematocrit tubes were filled from the lithium hepa-
rin-coated microtainer tubes and centrifuged to determine the packed cell volume (PCV).
Plasma total solids (TS) were measured via a refractometer (Schuco clinical, Allied Healthcare,
St. Louis, MO) and used as a correlate for plasma protein. Blood smears were air dried, heat-
fixed, and stained with a modified Giemsa stain (Protocol Hema 3, Fisher Scientific, Kalama-
zoo, MI). Each slide was first assessed for quality to ensure cells were evenly distributed and
that there was no evidence of clumping prior to interpretation. Estimated white blood cell
(WBC) counts were measured by taking the average WBC count of 10 fields in the monolayer
region viewed at 40X and multiplying by 1700 [18,19]. A minimum of 200 WBCs were
counted for each sample to determine the differential leukocyte count. Leukocytes were classi-
fied as heterophils, lymphocytes, monocytes, azurophils, eosinophils, or basophils based on
cell morphology [18]. The presence of hemogregarine-like parasites, appearing as oblong baso-
philic intraerythrocytic inclusions in erythrocytes, was classified based on morphology and
noted as a comment [20].
Plasma biochemistry. Plasma biochemical assays were performed using a tabletop ana-
lyzer (VetScan VS2, Avian/Reptilian Profile Plus, Abaxis, Union City, CA). Plasma samples
were analyzed for aspartate aminotransferase (AST), creatine kinase (CK), uric acid (UA), glu-
cose (Glucosepc), total calcium (Ca), phosphorus (P), total protein (TP), albumin (Alb), globu-
lin (Glo), potassium (Kpc), and sodium (Napc). Assay methods are displayed in Table 1.
Globulin is a calculated value (total protein–albumin). No biochemical analyses were per-
formed on whole blood.
Repeated analysis on the same blood specimen to assess analyzer between run variability
was not performed due to the low specimen volume and cost. Therefore, coefficients of varia-
tion (CV) for each analyte were obtained from the manufacturer-reported quality control
PLOS ONE | https://doi.org/10.1371/journal.pone.0229102 February 14, 2020
3 / 15
Venous Analytes of Copperheads and Ratsnakes
Table 1. VetScan assay methods, precision and calibration.
Analytes
Assay
Aspartate aminotransferase
(U/L)
L-aspartate +αketoglutarate
Creatine kinase (U/L)
Creatine phosphate+ADP
Uric acid (mg/dL)
Uricase+peroxidase
Glucose (mg/dL)
Calcium (mg/dL)
Hexokinase
Arsenazo III method
Phosphorus (mg/dL)
Sucrose phosphorylase
+phosphoglucomutase
Total Protein (g/L)
Biuret reaction
Albumin (g/L)
Bromcresol green dye-binding assay
Potassium (mmol/L)
Pyruvate kinase, lactate dehydrogenase
Sodium (mmol/L)
B-galactosidase
Coefficient of Variation
(%)
Calibration Standard
Calibration Method
2.0
6.0
4.8
1.6
3.4
4.9
1.9
4.3
6.3
1.8
IFCC
IFCC
Correlation to Beckman LX-20/
DX-20
NIST SRM #909
Correlation to Beckman LX-20/
DX-20
Colorimetric
Enzymatic
Enzymatic
Enzymatic
Arsenazo III Dye
NIST SRM #909B
Enzymatic
NIST SRM #909
Biuret (Copper II)
Correlation to Beckman LX-20/
DX-20
Dye Binding Bromocresol
Purple
NIST SRM #909
Correlation to Beckman LX-20/
DX-20
Enzymatic
Enzymatic
Assay methods, coefficients of variation, and calibration standards and methods for biochemistry analytes using the VetScan. When several CVs were reported by the
manufacturer, the largest was selected for this study.
IFCC indicates International Federation of Clinical Chemistry; NIST, National Institute of Standards and Technology; SRM, Standard Reference Manual
https://doi.org/10.1371/journal.pone.0229102.t001
available on the product inserts (Table 1). The CVs were determined using avian plasma sam-
ples for the VetScan Avian/Reptilian Profile Plus rotors. The VetScan uses internal calibrators
(in-house calibrators) or reference materials (National Institute of Standards and Technology)
to calibrate each parameter. The in-house calibrators are assigned by the previous in-house cal-
ibrators, which are considered the gold standard, and verified using a comparative and/or ref-
erence method. The methods and materials used in control value assignment procedures are
traceable to the standards listed in Table 1.
Blood gas analysis. Venous blood gas analyses was performed in the field immediately
after collection. A portable point-of-care analyzer (Elemental POCTM Rapid Blood Analyzer,
Heska, Loveland, CO) using Element POCTM Test Cards and a subsample of whole blood (0.1
mL) from the lithium heparin tube was used for blood gas analyses. The point-of-care analyzer
measured the following analytes: pH, partial pressure of carbon dioxide (pCO2), partial pres-
sure of oxygen (pO2), sodium (Na), chloride (Cl), potassium (K), ionized calcium (iCa), creati-
-),
nine, glucose, lactate, hematocrit (HCT), total carbon dioxide (TCO2), bicarbonate (HCO3
base excess of extracellular fluid (BE(ecf)), base excess of blood (BE(b)), anion gap (AG), oxy-
gen saturation (sO2), and hemoglobin (Hgb). The point-of-care analyzer calculated the follow-
ing analytes: TCO2, HCO3
was measured at actual pH. The partial pressure of oxygen was not reported because strict
anaerobic conditions were not maintained. The following analytes were also not reported: cre-
atinine, sO2, BE, Hgb, and HCT. Creatinine is not considered diagnostic in reptiles and oxygen
saturation, BE, Hgb, and HCT are measured or calculated using assumptions based on human
hemoglobin properties which are not accurate in reptiles [21–23]. Values for pH, pO2 and
pCO2 are presented as uncorrected and corrected based on ambient temperature (TA). We
performed temperature corrections using the following formulae [24–26].
pH (TA) = pHI—0.0147 x (TA—37) + 0.0065 x (7.4—pHI) x (TA—37)
pO2 (TA) = pO2I x 10−0.0058�(T
-37)
pCO2 (TA) = pCO2I x 100.0019�(T
-, BE(ecf), BE(b), AG, sO2, and Hgb. The ionized calcium reported
-37)
A
A
PLOS ONE | https://doi.org/10.1371/journal.pone.0229102 February 14, 2020
4 / 15
Venous Analytes of Copperheads and Ratsnakes
HCO3-(TA) = αCO2 x pCO2(TA) x 10(pH(T
A
TCO2(TA) = HCO3 - (TA) + (αCO2 x pCO2(TA))
αCO2 = 9.174 x 10−2–3.269 x 10−3 x TA + 6.364 x 10−5 x TA
TA
pKa = 6.398–1.341 x 10−2 x TA + 2.282 x 10−4 x TA
(1.011 + 10pH(T
-10.241 + 10pH(T
A
-8.889)
A
)+0.011xT
)+0.001xT
)-pKa)
3
A
A
2–5.378 x 10−7 x
2–1.516 x 10−6 x TA
3—log
Statistical analysis. Descriptive statistics including mean, median, standard deviation
(SD), and range (i.e. minimum and maximum values) were calculated for all variables. We
assessed the data for normality using the Anderson–Darling test (Reference Value Advisor)
[27,28]. We then used the Dixon test and the Tukey test to identify outliers and suspect outliers
(Reference Value Advisor) [27]. Following the recommendations found in the Guidelines of
the American Society for Veterinary Clinical Pathology [29], we calculated reference intervals
(90% of reference values) by the robust method after Box–Cox transformation of the data. The
number of animals for some analytes assessed in this study do not fully comply with the strict
ASVCP guidelines for establishing a reference interval; therefore, these results should only be
considered as an estimate rather than true reference ranges [29].
Statistical analysis for morphology (weight, snout-vent length, and total length), hematol-
ogy, and plasma biochemistry data were analyzed with JMP Pro, Version 12.0.1 (SAS, Cary,
NC). Differences by sex for each species were assessed with the Wilcoxon ranked sum. Venous
blood gas analytes were compared between the species with the Wilcoxon ranked sum. Blood
gas analytes collected at the beginning and end of restraint for each species were compared
with the Wilcoxon signed rank test. Paired blood gas analyses from samples obtained at the
beginning and end of restraint were available for eight copperheads and 13 ratsnakes. Blood
gas results were not analyzed by sex due to small sample size. A chi-squared test was used to
compare electrolyte values from the Abaxis and venous blood gas analyzers. The relationships
between restraint time and end-restraint lactate concentrations for each species were assessed
with Kendall’s τ rank correlation. Statistical significance for all analyses was set at P < 0.05.
Results
Morphometric data for copperheads and ratsnakes are reported in “Table 2”. Male copper-
heads were significantly heavier and longer than their female counterparts (P = 0.02, P = 0.03,
and P = 0.03; weight, snout-to vent length, and total length, respectively), but there was no sig-
nificant difference in size between male and female ratsnakes (P = 0.20, P = 0.14, and P = 0.18;
weight, snout-to vent length, and total length, respectively) (Table 2). Ambient temperature
Table 2. Copperhead and ratsnake morphometric parameters.
A
Parameters
Weight (g)
Snout-to-vent length (cm)
Total length (cm)
B
Parameters
Weight (g)
Snout-to-vent length (cm)
Total length (cm)
Mean
240�
67.3�
77.5�
Mean
529
111
133
Male (n = 14)
Median
255
66.5
77.8
Male (n = 20)
Median
388
109
131
SD
96.9
12.5
13.4
SD
383
28.9
34.0
Range
100–410
49.0–87.0
57.5–99.0
Range
52.0–1500
61.0–155
74.0–187
Female (n = 16)
Mean
134�
52.0�
59.8�
Female (n = 14)
Mean
334
98.0
117
SD
44.1
5.50
6.10
SD
199
23.4
27.7
Median
130
52.7
60.5
Median
322
99.0
119
Range
40.0–200
36.0–59.0
42.4–67.2
Range
58.0–700
64.0–132
77.0–156
Body weight, snout-to-vent length, and total length for male and female free-ranging E. Copperheads (Agkistrodon contortrix) (A) and free-ranging E. Ratsnakes
(Pantherophis alleghaniensis) (B) from central North Carolina. Significant differences between the sexes (P < 0.05) are denoted with an asterisk (�).
https://doi.org/10.1371/journal.pone.0229102.t002
PLOS ONE | https://doi.org/10.1371/journal.pone.0229102 February 14, 2020
5 / 15
Venous Analytes of Copperheads and Ratsnakes
Table 3. Copperhead hematology values.
Analytes
PCV (%)a
Total solids (g/L)a
Est. WBC (103/μL)
Heterophils (103/μL)
Bands (103/μL)
Lymphocytes (103/μL)
Monocytes (103/μL)
Eosinophils (103/μL)
Basophils (103/μL)
Azurophils (103/μL)
Mean
28
55
13
0.76
0
8.8
0.5
0.01
0.73
1.4
SD
5.7
10
6.8
0.49
0
5.7
0.9
0.01
0.35
1.4
Median
Min–Max
LRL 90% CI
URL 90% CI
29
54
11
0.67
0
6.6
0.23
0
0.71
1.3
13–38
40–74
3.6–27
RIA
12–38
33–78
3.8–32
11–20
31–41
2.8–5.1
0.07–2.1
0.08–2.0
0.04–0.18
0
1.6–20
0–0.33
0–0.07
0.08–1.3
0–5.7
0
1.5–26
0–0.4
0–0.09
0.04–1.5
0–5.4
0
1.1–2.1
0–0.05
0–0.03
0–0.2
0–0.9
36–45
69–85
23–39
1.6–2.6
0
18–33
0.28–0.49
0.04–0.14
1.3–1.7
3.2–5.6
PCV, total solids, and estimated WBC and differential results for free-ranging E. Copperheads (Agkistrodon contortrix) from central North Carolina (n = 29 unless
otherwise noted).
an = 17
PCV indicates packed cell volume; WBC, white blood cell; RIA, reference interval approximation; LRL, lower reference limit; URL, upper reference limit
https://doi.org/10.1371/journal.pone.0229102.t003
was 26.4˚C ± 3.4˚C (mean ± SD) and ranged from 20.4˚C to 33.9˚C throughout the duration
of this study.
PCV, TS, estimated WBC counts and differential results are reported in “Table 3” (copper-
heads) and “Table 4” (ratsnakes). There were no statistically significant sex-specific differences
in hematological values for either copperheads (Table 3) or ratsnakes (Table 4). Hemogregar-
ine-like parasites were observed in 20% (12/60) of all snakes captured in this study. The hemo-
protozoa genera could not be identified based on morphology [4,20,30,31]. Ten percent of
copperheads (3/29) were infected with hemogregarine-like parasites, while 29% (9/31) of rats-
nakes were infected.
Plasma biochemistry results obtained from the Abaxis are reported in “Table 5” (copper-
heads) and “Table 6” (ratsnakes). Male copperheads had significantly higher albumin than
female copperheads (P = 0.04). They also had a significantly higher CK than female
Table 4. Ratsnake hematology values.
Analytes
PCV (%)a
Total solids (g/L)a
Est. WBC (103/μL)
Heterophils (103/μL)
Bands (103/μL)
Lymphocytes (103/μL)
Monocytes (103/μL)
Eosinophils (103/μL)
Basophils (103/μL)
Azurophils (103/μL)
Mean
27
62
8.9
1.1
0
6.1
0.6
0.03
0.19
0.01
SD
5.4
13
3.5
0.7
0
3.3
0.7
0.17
0.19
0.01
Median
Min–Max
28
61
8.7
1.1
0
5.6
0.4
0
0.11
0
14–38
40–92
2.9–16
0.1–3.0
0
1.1–14
0–2.0
0–0.9
0–0.64
0–0.4
RIA
14–37
36–91
2.4–17
0–2.9
0
1.2–15
0–2.2
0–0.7
0–0.8
0–0.5
LRL 90% CI
URL 90% CI
7.0–19
30–43
1.4–3.6
0–0.2
0
0.8–1.8
0–0.8
0–0.01
0–0.1
0–0.01
35–39
81–99
14–19
2.4–3.5
0
12–17
1.6–2.5
0.02–0.8
0.5–1.0
0.2–0.6
PCV, total solids, and estimated WBC and differential results for free-ranging E. Ratsnakes (Pantherophis alleghaniensis) from central North Carolina (n = 31 unless
otherwise noted).
an = 22
PCV indicates packed cell volume; WBC, white blood cell; RIA, reference interval approximation; LRL, lower reference limit; URL, upper reference limit
https://doi.org/10.1371/journal.pone.0229102.t004
PLOS ONE | https://doi.org/10.1371/journal.pone.0229102 February 14, 2020
6 / 15
Venous Analytes of Copperheads and Ratsnakes
Table 5. Copperhead biochemistry parameters.
Analytes
Aspartate aminotransferase (U/L)
Creatine kinase (U/L)
Uric acid (mg/dL)
Glucose (mg/dL)
Calcium (mg/dL)
Phosphorus (mg/dL)
Total Protein (g/L)
Albumin (g/L)
Globulin (g/L) a
Potassium (mmol/L)
Sodium (mmol/L)
Mean
48
837
5.0
51
15.1
4.1
52
11
42
6.4
158
SD
85
1170
4.0
29
2.6
1.1
5.8
1.8
4.7
0.5
8
Median
25
507
2.8
47
14.0
4.1
52
11
41
5.9
157
Range
11–342
205–4800
1.5–14.6
19–122
12.0–20.1
2.1–6.7
43–66
7–14
34–53
4.8–9.8
143–172
Plasma biochemistry values for free-ranging E. Copperheads (Agkistrodon contortrix) from central North Carolina (n = 14 unless otherwise noted).
an = 12
https://doi.org/10.1371/journal.pone.0229102.t005
copperheads (P = 0.04). There were no additional statistically significant sex-specific differ-
ences in biochemistry values for either copperheads (Table 5) or ratsnakes (Table 6).
Venous blood gas analysis results from the beginning and end of restraint are reported in
“Table 7” (not temperature corrected) and “Table 8” (temperature corrected). There were no
significant differences in the analytes (sodium, potassium, and glucose) obtained from the
Abaxis compared with the venous blood gas analysis (P = 0.37, P = 0.25 and P = 0.41, respec-
tively). The total time (mins) from first contact to collection of the first blood sample, was
1.4 ± 0.4 mins (mean ± SD) and 1.0 ± 0.2 mins (mean ± SD) for copperheads and ratsnakes,
respectively. The total time (mins) from first blood collection to collection of the second blood
sample, was 12.5 ± 2.4 mins (mean ± SD) and 13.5 ± 3.4 mins (mean ± SD) for copperheads
and ratsnakes, respectively. There were no significant differences in restraint times between
the two species (P = 0.43 and P = 0.52, respectively). Lactate concentrations increased signifi-
cantly between beginning-restraint sample and end-restraint sample in both species (P = 0.03
and P = 0.01 for copperheads and ratsnakes, respectively). Venous blood pH significantly
decreased between beginning-restraint sample and end-restraint sample in both species
Table 6. Ratsnake biochemistry parameters.
Analytes
Aspartate aminotransferase (U/L)
Creatine kinase (U/L)
Uric acid (mg/dL)
Glucose (mg/dL)
Calcium (mg/dL)
Phosphorus (mg/dL)
Total Protein (g/L)
Albumin (g/L)
Globulin (g/L)
Potassium (mmol/L)
Sodium (mmol/L)
Mean
22
616
5
64
15.4
5.1
58
21
37
5.7
164
SD
16
615
6.3
14
1.1
1.9
8.3
3
6.2
1.3
7
Median
18
392
3.0
64
15.3
4.4
59
21
38
5.7
162
Range
9–72
88–2701
1.0–25.1
42–95
13.8–18
2.5–9.2
43–74
14–26
29–49
2.2–7.6
156–180
Plasma biochemistry values for free-ranging E. Ratsnakes (Pantherophis alleghaniensis) from central North Carolina (n = 18).
https://doi.org/10.1371/journal.pone.0229102.t006
PLOS ONE | https://doi.org/10.1371/journal.pone.0229102 February 14, 2020
7 / 15
Venous Analytes of Copperheads and Ratsnakes
Table 7. Venous blood gas results (not temperature corrected).
Beginning-restraint (n = 14)
E. Copperheads
Analytes
pH
pCO2 (mmHg)
Sodium (mmol/L)
Chloride (mmol/L)
Potassium (mmol/L)
Ionized Calcium (mmol/L)
Glucose (mmol/L)
Glucose (mg/dL)
Lactate (mmol/L)
TCO2 (mmol/L)
- (mmol/L)
HCO3
Anion Gap (mmol/L)
Analytes
pH
pCO2 (mmHg)
Sodium (mmol/L)
Chloride (mmol/L)a
Potassium (mmol/L)b
Ionized Calcium (mmol/L)c
Glucose (mmol/L)
Glucose (mg/dL)
Lactate (mmol/L)
TCO2 (mmol/L)
- (mmol/L)
HCO3
Anion Gap (mmol/L)d
Mean
7.19�
25.3
158
127
5.2
1.71
2.38
43
10.9�
10.4
9.6
27
Mean
7.30�
36.1
162
124
5.2
1.51
3.05
55
10.3�
17.8
16.7
25
SD
0.21
10.6
6
7
1.1
0.23
1.55
28
6.75
3.9
3.8
6
Median
7.16
21.8
157
129
5.0
1.71
2.11
38
9.61
10.0
9.1
25
Range
6.91–7.62
11.8–41.9
149–176
109–138
3.9–8.0
1.08–2.12
1.11–6.94
20–125
2.13–20.1
5.0–17.3
4.4–16.8
20–42
Beginning-restraint (n = 19)
E. Ratsnakes
SD
0.19
19.2
5
10
1.0
0.19
0.72
13
5.37
8.2
7.8
5
Median
7.33
31.9
162
126
5.1
1.56
3.22
58
10.1
16.3
15.0
25
Range
6.97–7.74
11.6–78.5
153–180
93–140
3.8–8.0
1.17–1.76
1.67–4.22
30–76
2.66–20.1
8.9–47.5
8.6–45.3
16–34
Mean
7.12�
27.6
158
128
5.0
1.71
2.22
40
12.7�†
9.9
9.0
27†
Mean
7.22�
26.0
165
127
5.5
1.53
3.72
67
17.4�†
11.2
10.5
33
End-restraint (n = 12)
SD
Median
0.22
13.6
4
7
0.6
0.21
0.78
14
5.58
5.1
4.9
4
7.06
25.7
159
130
4.9
1.75
2.33
42
11.6
9.2
8.1
27†
End-restraint (n = 18)
SD
Median
0.23
10.2
7
9
0.6
0.22
0.61
11
2.93
3.8
3.8
6
7.18
25.3
164
127
5.4
1.59
3.66
66
17.86
11.6
10.6
32
Range
6.82–7.45
7.8–51.6
150–163
114–135
4.1–5.9
1.16–1.92
1.11–3.28
20–59
5.48–20.1
3.3–18.5
3.1–17.2
20–32
Range
6.86–7.81
9.8–45.1
151–180
99–140
4.6–7.0
0.96–1.87
2.39–4.88
43–88
10.98–20.1
6.8–18.8
6.0–18.5
24–48
First blood sample (beginning-restraint) and second blood sample (end-restraint) blood gas values for free-ranging E. Copperheads (Agkistrodon contortrix) and E.
Ratsnakes (Pantherophis alleghaniensis) from central North Carolina. Significant differences between the sampling periods (P < 0.05) for each species are denoted with
an asterisk (�); only the paired samples (n = 8 for E. Copperheads and n = 13 for E. Ratsnakes) were compared. Significant differences between the species (P < 0.05) are
denoted with a cross (†).
aBeginning-restraint n = 18
bEnd-restraint n = 17
cEnd-restraint n = 16
dBeginning-restraint n = 17, End-restraint n = 17
https://doi.org/10.1371/journal.pone.0229102.t007
(P = 0.04 and P = 0.04 for copperheads and ratsnakes, respectively). End-restraint lactate con-
centrations were significantly greater in ratsnakes compared to the copperheads (P = 0.01).
Discussion
This study aimed to establish guidelines for the interpretation of venous hematology, plasma
biochemistry, and blood gas analytes in free-ranging copperheads and ratsnakes while also
assessing for changes in venous blood gas analytes secondary to restraint. This study docu-
mented that blood lactate concentrations at the beginning of restraint were similar for both
PLOS ONE | https://doi.org/10.1371/journal.pone.0229102 February 14, 2020
8 / 15
Venous Analytes of Copperheads and Ratsnakes
Table 8. Venous blood gas results (temperature corrected).
Analytes
pH
pCO2 (mmHg)
TCO2 (mmol/L)
- (mmol/L)
HCO3
Analytes
pH
pCO2 (mmHg)
TCO2 (mmol/L)
- (mmol/L)
HCO3
E. Copperheads
Beginning-restraint (n = 14)
SD
0.24
10.3
8.4
8.5
Median
7.34
20.6
6.4
5.4
Range
6.99–7.78
11.2–40.4
2.2–30.4
1.6–29.8
E. Ratsnakes
Beginning-restraint (n = 19)
SD
0.21
18.4
17.2
17
Median
7.46
30.0
12.8
12.2
Range
7.07–7.96
10.9–75.9
7.5–76.4
5.0–73.4
Mean
7.24�
26.3
7.4
6.3
Mean
7.36�
24.9
12.5
11.6
End-restraint (n = 12)
SD
Median
0.25
13.0
6.5
6.6
7.18
24.5
4.4
3.2
End-restraint (n = 18)
SD
Median
0.26
9.8
19.6
19.7
7.32
24.6
5.9
5.2
Mean
7.33�
24.3
9.0
8.1
Mean
7.45�
34.5
18.2
16.9
Range
6.91–7.64
7.4–49.5
1.5–20.3
0.88–19.5
Range
6.93–8.07
9.4–43.2
2.4–86.8
1.2–86.4
First blood sample (beginning-restraint) and second blood sample (end-restraint) blood gas values corrected for ambient temperature for free-ranging E. Copperheads
(Agkistrodon contortrix) and E. Ratsnakes (Pantherophis alleghaniensis) from central North Carolina. Significant differences between the sampling periods (P < 0.05) for
each species are denoted with an asterisk (�); only the paired samples (n = 8 for E. Copperheads and n = 13 for E. Ratsnakes) were compared. Significant differences
between the species (P < 0.05) are denoted with a cross (†).
https://doi.org/10.1371/journal.pone.0229102.t008
species. Additionally, lactate concentrations increased significantly in both species at the end
of restraint when compared to the beginning of restraint. The significant decrease in venous
blood pH between the beginning-restraint sample and end-restraint sample in both species
likely occurred secondary to the aforementioned changes in lactate.
Although previous studies have reported baseline biochemical, blood gas, and hematology
values for other viperid and colubrid snakes, this is the first study to establish guidelines for
the interpretation of these analytes in copperheads and ratsnakes. The hematological and
plasma biochemical results for copperheads and ratsnakes reported in this study were consis-
tent with previously published values for both viperid and colubrid snakes, with the exception
of higher leukocyte counts in the copperheads [4,32–39]. This difference may be attributed to
many variables such as species, age, size, time of year, methodology, and instrumentation
[4,34,40]. Due to the nucleated nature of reptile erythrocytes, automated methods of obtaining
white blood cell (WBC) counts and differentials are not accurate; manual methods are utilized
in these species, with the accuracy of results dependent on the cytologist’s experience [18]. In
this study, estimated leukocyte counts and differentials were performed with the standard pro-
tocol used at the North Carolina Zoo, where the monolayer regions of the blood smear are
scanned for estimating the total number of leukocytes.
Point-of-care analyzers are valuable diagnostic tools, especially in field settings or medical
emergencies where rapid results are needed. Commonly used point-of-care analyzers perform
blood gas analysis that assess respiratory function, metabolic status, and tissue perfusion, and
provide near-instantaneous results [41–43]. However, many of the blood gas components are
influenced by temperature, so when applied to research involving reptiles and other poikilo-
therms, clinicians and researchers must decide whether to correct the results for ambient tem-
perature. Temperature corrections may not always be warranted and instituting them can
preclude comparison of data sets if the methods are not stated, vary among sources, or if the
temperatures are different [44–46]. There are multiple formulae for manually performing tem-
perature corrections, but none have been validated for snake blood gas analysis performed
with portable point-of-care analyzers, such as the iSTAT (Abbott Point of Care, Princeton, NJ)
PLOS ONE | https://doi.org/10.1371/journal.pone.0229102 February 14, 2020
9 / 15
Venous Analytes of Copperheads and Ratsnakes
or the EPOC [9,22,47,48]. Venous blood gas results from copperheads and ratsnakes are pre-
sented both without temperature correction “Table 7” and with temperature correction to
ambient temperature “Table 8” to facilitate comparison with future studies and clinical appli-
cations. Temperature corrections did not affect statistical comparisons between copperheads
and ratsnakes.
Despite all of the snakes in this study appearing outwardly healthy on physical examination,
lactate concentrations both at the beginning and end of restraint were higher than the reported
lactate concentrations from other reptiles, except cold-stunned Kemp’s ridley sea turtles on
their first day of rehabilitation and manually restrained Nile crocodiles [5,43,47–53]. The
higher lactate concentrations in copperhead and ratsnakes may reflect species-specific differ-
ences in anaerobic metabolism or differences in instrumentation, however, additional consid-
erations for elevated lactate in reptiles include hepatocellular damage, renal disease,
hypoventilation, poor perfusion, and cardiovascular disease [12,54]. Based on physical exami-
nations, there was no evidence for these conditions in the animals used in this study. These
comparisons illustrate the need for species-specific reference intervals and underscore the
importance of understanding clinicopathologic responses to restraint. Improved understand-
ing of the effects of capture and restraint in these species will enable clinicians to interpret lac-
tate results more appropriately and improve clinical reasoning.
The lactate concentrations from the samples obtained at the beginning and end of restraint
for copperheads were comparable to lactate concentrations observed in male copperheads 60
minutes after losing intra-sexual conspecific fights [55]. Both the winners of these ritualized
competitions and control animals, which did not fight, were found to have lower lactate con-
centrations than the copperheads in this study; however, these animals were anesthetized prior
to blood collection which may have reduced their level of stress and lactate production [55].
Due to the similarity of lactate concentrations from our study to those in studies on the meta-
bolic costs of reproductive behaviors [55], the restraint of animals may be as physiologically
demanding as the physical interactions that occur during the breeding season.
Blood gas analyses, and more specifically lactate measurements, can serve as a means of
evaluating the physiological response to capture, restraint, and human interactions in snakes;
however, the physiological response to restraint techniques and handling can vary by species
[10,56]. In this study, lactate concentrations increased and pH decreased in both species when
the first blood sample (beginning-restraint) and second blood sample (end-restraint) were
compared, but lactate was significantly increased in ratsnakes at the end of restraint. The sig-
nificant increase in ratsnakes may be associated with a stronger sympathetic nervous system
response to the more hands-on restraint that these non-venomous snakes experienced. This
increase in sympathetic nervous system response may have resulted in increased physical resis-
tance to restraint and subsequently, higher end-restraint lactate concentrations. In contrast,
copperheads were restrained in open-ended, acrylic tubes, which reduced the amount of
hands-on contact. Our results differ from Kreger and Mench [56] who found higher cortico-
sterone concentrations in Ball Pythons (Python regius) restrained in polyvinyl chloride (PVC)
restraint tubes compared to non-handled controls while animals that underwent handheld
restraint did not exhibit significantly different corticosterone concentrations compared to con-
trols. It is possible that lactate and corticosterone are not interchangeable proxies for measur-
ing response to restraint. Alternatively, differences in resting metabolic rate between pit vipers
and colubrids may explain the differences in lactate following restraint. These colubrids are
active foragers and generally have higher resting metabolic rates than many pit vipers which
are ambush foragers [57,58]. Thus, the higher resting oxygen requirements of ratsnakes could
translate into an earlier switch to anaerobic metabolism during restraint, which would be
reflected in higher post-restraint lactate.
PLOS ONE | https://doi.org/10.1371/journal.pone.0229102 February 14, 2020
10 / 15
Venous Analytes of Copperheads and Ratsnakes
A shift from aerobic to anaerobic metabolism often results from failure of the oxygen sup-
ply system in meeting the energy demands of tissues. In reptiles, this shift occurs to varying
degrees in response to high energy utilization during intense muscular activity, or anoxia in
aquatic species [59,60]. Increased lactate concentrations can lead to acidemia, which potenti-
ates the anaerobic cycle and may lead to electrolyte imbalances [61]. Point-of-care analyzers
enable clinicians to perform analyses to elucidate the metabolic status of a patient in the field,
without the delay imposed by measuring these parameters in a laboratory. The physiologic
effect of handling on snakes may be apparent when the pH and lactate results of this study are
compared to results from South American Rattlesnakes (Crotalus durissus terrificus), which
had arterial catheters with long extensions to eliminate the effects on handling on lactate
results [52]. The lower pH and higher lactate results in copperheads and ratsnakes compared
to the non-handled rattlesnakes suggests that even brief handling has an immediate effect on
metabolism and subsequently, blood pH and lactate. The role of species differences in response
to restraint and metabolism as well as differences in the instrumentation used to assess these
parameters may also affect these results; thus further research is necessary to establish refer-
ence intervals to facilitate interspecies comparisons.
The pCO2 and HCO3
- results in this study were lower than those reported in other reptiles
including Marine Iguanas (Amblyrhynchus cristatus), anesthetized Green Iguanas (Iguana
iguana), Inland Bearded Dragons (Pogona vitticeps), and sea turtles [5,43,47,48,50,53,62–64].
In conjunction with the metabolic lactic acidosis detected in the snakes of this study, the
observed decrease in pCO2 and HCO3
hyperventilation (compensatory respiratory alkalosis) and the titration of HCO3
ing lactic acid. The lower pCO2 and HCO3
represent taxonomic differences. Lower HCO3
sample processing, but Kirshbaum [65] found minimal changes in HCO3
serum samples exposed to air for up to two hours.
- may also be explained by air exposure during
- concentrations in
- concentrations were interpreted as a correction by
- concentrations in both snake species could also
- by circulat-
This study provides baseline venous hematology, plasma biochemistry, and blood gas data
for free-ranging copperheads and ratsnakes. These results add to a growing database of knowl-
edge about health management in wild and managed reptiles. Although the sample size for
some analytes in this study do not meet the specifications for formal determination of refer-
ence intervals [28], these results provide researchers and veterinarians with species-specific
guidelines, which will be valuable for field health assessments and captive management of
these species. Future research is necessary to establish formal reference intervals for these spe-
cies to facilitate comparisons of blood values across age groups, geographical localities, and
subspecies.
This study also demonstrates free-ranging copperheads and ratsnakes’ physiological
response to capture and restraint. These differences should be considered when interpreting
the results of blood chemistry analysis and when planning medical procedures and population
studies in the field. Further research comparing the two methods of restraint used in this study
may contribute to our understanding of this physiological response and improve animal wel-
fare in scenarios necessitating restraint. An improved understanding of the effects of restraint
in reptiles will enable researchers and clinicians to interpret venous blood gas results more
appropriately.
Conclusions
• This study provides guidelines for the interpretation of venous hematology, plasma bio-
chemistry, and blood gas values for free-ranging copperheads and ratsnakes in central North
Carolina.
PLOS ONE | https://doi.org/10.1371/journal.pone.0229102 February 14, 2020
11 / 15
Venous Analytes of Copperheads and Ratsnakes
• Capture and restraint resulted in a significant increase in lactate and a significant decrease in
pH for both species, underscoring the effects of stress and handling on clinicopathologic
data.
Supporting information
S1 File.
(XLSX)
Acknowledgments
The authors thank Janice Coakley, Ezra Ellis, Jenessa Gjeltema, Katherine Hagen, David Hill,
Mark Lewis, Michael Overly, Andrea Persson, Brianne Phillips, Cheryl Purnell, Heather
Shaub, Hendrik Smock, and David Strickler from the North Carolina Zoo for their contribu-
tions in sample collection and technical assistance for this project. The authors also thank Drs.
Matt Allender and Maria Correa for their statistical guidance.
Author Contributions
Conceptualization: Dustin C. Smith, Larry J. Minter.
Data curation: Anthony J. Cerreta, Sarah A. Cannizzo, Dustin C. Smith, Larry J. Minter.
Formal analysis: Anthony J. Cerreta, Larry J. Minter.
Investigation: Sarah A. Cannizzo, Dustin C. Smith, Larry J. Minter.
Methodology: Sarah A. Cannizzo, Dustin C. Smith, Larry J. Minter.
Project administration: Sarah A. Cannizzo, Dustin C. Smith, Larry J. Minter.
Resources: Larry J. Minter.
Supervision: Larry J. Minter.
Writing – original draft: Anthony J. Cerreta, Sarah A. Cannizzo.
Writing – review & editing: Anthony J. Cerreta, Sarah A. Cannizzo, Dustin C. Smith, Larry J.
Minter.
References
1. Dujowich M, Mazet JK, Zuba JR. Hematologic and biochemical reference ranges for captive California
condors (Gymnogyps californianus). J Zoo Wildl Med. 2005; 36: 590–597. https://doi.org/10.1638/04-
111.1 PMID: 17312714
2. Dallwig RK, Paul-Murphy J, Thomas C, Medlin S, Vaughan C, Sullivan L, et al. Hematology and clinical
chemistry values of free-ranging basilisk lizards (Basiliscus plumifrons) in Costa Rica. J Zoo Wildl Med.
2011; 42: 205–213. https://doi.org/10.1638/2008-0202.1 PMID: 22946396
3.
Travis EK, Watson P, Dabek L. Health assessment of free-ranging and captive Matschie’s tree kanga-
roos (Dendrolagus matschiei) in Papua New Guinea. J Zoo Wildl Med. 2012; 43: 1–9. https://doi.org/10.
1638/2010-0082.1 PMID: 22448504
4. Minter LJ, Dombrowski DS, Stoskopf MK, Purnell CA, Loomis MR, DeVoe RS. Hematology and plasma
biochemical values for free ranging cottonmouths (Agkistrodon piscivorus) in central North Carolina,
USA. Herpetol Conserv Biol. 2013; 8: 321–334.
5.
Lewbart GA, Hirschfeld M, Denkinger J, Vasco K, Guevara N, Garcı´a J, et al. Blood gases, biochemis-
try, and hematology of Gala´pagos green turtles (Chelonia mydas). PLoS One. 2014; 9:e96487. https://
doi.org/10.1371/journal.pone.0096487 PMID: 24824065
PLOS ONE | https://doi.org/10.1371/journal.pone.0229102 February 14, 2020
12 / 15
Venous Analytes of Copperheads and Ratsnakes
6. Stoot LJ, Cairns NA, Cull F, Taylor JJ, Jeffrey JD, Morin F, et al. Use of portable blood physiology point-
of-care devices for basic and applied research on vertebrates: a review. Conserv Physiol. 2014; 2:
cou011. https://doi.org/10.1093/conphys/cou011 PMID: 27293632
7. Sive WJ, Hattingh J. The measurement of psychophysiological reactions of pilots to a stressor in flight
simulator. Aviat Space Environ Med. 1991; 62: 831–836. PMID: 1930070
8.
Le Maho Y, Karmann H, Briot D, Handrich Y, Robin JP, Mioskowski E, et al. Stress in birds due to rou-
tine handling and a technique to avoid it. Am J Physiol. 1992; 263: 775–781.
9. Harms CA, Mallo KM, Ross PM, Segars A. Venous blood gases and lactates of wild loggerhead sea tur-
tles (Caretta caretta) following two capture techniques. J Wildl Dis. 2003; 39: 366–374. https://doi.org/
10.7589/0090-3558-39.2.366 PMID: 12918445
10. Harms CA, Harms RV. Venous blood gas and lactate values of mourning doves (Zenaida macroura),
boat-tailed grackles (Quiscalus major), and house sparrows (Passer domesticus) after capture by mist
net, banding, and venipuncture. J Zoo Wildl Med. 2012; 43: 77–84. https://doi.org/10.1638/2011-0114.1
PMID: 22448512
11. Harms CA, Jinks MR, Harms RV. Blood gas, lactate, and hematology effects of venipuncture timing and
location after mist-net capture of mourning doves (Zenaida macroura), boat-tailed grackles (Quiscalus
major), and house sparrows (Passer domesticus). J Wildl Dis. 2016; 52: 54–64.
12. Allen SE, Holm JL. Lactate: physiology and clinical utility. J Vet Emerg Crit Care. 2008; 18: 123–132.
Fowler ME. Reptiles. In: Restraint and Handling of Wild and Domestic Animals. 3rd ed. Ames, Iowa,
USA: Wiley-Blackwell; 2008. pp. 423–425.
13.
14.
15.
16.
Lock B. Venomous snake restraint and handling. J Exot Pet Med. 2008; 17: 273–284.
Funk RS. Snakes. In: Mader DR, Divers SJ, editors. Reptile Medicine and Surgery. 2nd ed. St. Louis,
MO, USA: Saunders Elsevier; 2006. pp. 49–50.
Jemison SC, Bishop LA, May PG, Farrell TM. The impact of PIT-tags on growth and movement of the
rattlesnake, Sistrurus miliarius. J Herpetol. 1995; 29: 129–132.
17. Winne CT, Willson JD, Andrews KM, Reed RN. Efficacy of marking snakes with disposable medical
cautery units. Herpetol Rev. 2006; 37: 52–54.
18. Sykes JM, Klaphake E. Reptile hematology. Clin Lab Med. 2015; 35:661–680. https://doi.org/10.1016/j.
cll.2015.05.014 PMID: 26297412
19. Keller JM, Kucklick JR, Stamper MA, Harms CA, McClellan-Green PD. Associations between organo-
chlorine contaminant concentrations and clinical health parameters in loggerhead sea turtles from
North Carolina, USA. Environ Health Perspect. 2004; 112: 1074–1079. https://doi.org/10.1289/ehp.
6923 PMID: 15238280
20. Campbell TW. Hemoparasites. In: Mader DR, Divers SJ, editors. Reptile Medicine and Surgery. 2nd ed.
St. Louis, MO, USA: Saunders Elsevier; 2006. pp. 801–805.
21. Campbell TW. Clinical pathology of reptiles. In: Mader DR, Divers SJ, editors. Reptile Medicine and Sur-
gery. 2nd ed. St. Louis, MO, USA: Saunders Elsevier; 2006. pp. 453–470.
22.
Innis CJ. Master Class: Sea Turtles Part C. Interpretation of plasma biochemical, hematological, acid-
base, and blood gas data of sea turtles. Proceedings Association of Reptilian and Amphibian Veterinari-
ans. 2009; 51–56.
23. Campbell TW. Clinical pathology. In: Mader DR, Divers SJ, editors. Current Therapy in Reptile Medicine
and Surgery. St. Louis, MO, USA: Elsevier Saunders; 2014. pp. 70–92.
24. Stabenau EK, Heming TA. Determination of the constants of the Henderson–Hasselbalch equation,
(alpha)CO2 and pKa, in sea turtle plasma. J Exp Biol. 1993; 180: 311–314.
25. CLSI. Blood Gas and pH Analysis and Related Measurements; Approved Guideline. 2001. CLSI docu-
ment C46-A [ISBN 1-56238-444–9]. Clinical and Laboratory Standards Institute, 940 West Valley Road,
Suite 1400, Wayne, Pennsylvania 19087–1898, USA.
26. Anderson ET, Harms CA, Stringer EM, Cluse WM. Evaluation of hematology and serum biochemistry
of cold-stunned green sea turtles (Chelonia mydas) in North Carolina, USA. J Zoo Wildl Med. 2011; 42:
247–255. https://doi.org/10.1638/2010-0217.1 PMID: 22946402
27. Geffre´ A, Concordet D, Braun JP, Trumel C. Reference Value Advisor: a new freeware set of macroin-
structions to calculate reference intervals with Microsoft Excel. Vet Clin Pathol. 2011; 40: 107–112.
https://doi.org/10.1111/j.1939-165X.2011.00287.x PMID: 21366659
28. Geffre´ A, Friedrichs K, Harr K, Concordet D, Trumel C, Braun JP. Reference values: a review. Vet Clin
Pathol. 2009; 38: 288–298. https://doi.org/10.1111/j.1939-165X.2009.00179.x PMID: 19737162
29.
Friedrichs KR, Harr KE, Freeman KP, Szladovits B, Walton RM, Barnhart KF, et al. ASVCP reference
interval guidelines: determination of de novo reference intervals in veterinary species and other related
topics. Vet Clin Pathol. 2012; 41: 441–453. https://doi.org/10.1111/vcp.12006 PMID: 23240820
PLOS ONE | https://doi.org/10.1371/journal.pone.0229102 February 14, 2020
13 / 15
Venous Analytes of Copperheads and Ratsnakes
30.
Telford SR Jr. The hemogregarines. In: Hemoparasites of the Reptilia: Color Atlas and Text. Boca
Raton, FL, USA: CRC Press; 2009. pp. 199–260.
31. Stacy NI, Alleman AR, Sayler KA. Diagnostic hematology of reptiles. Clin Lab Med. 2011; 31: 87–108.
https://doi.org/10.1016/j.cll.2010.10.006 PMID: 21295724
32. Alleman AR, Jacobson ER, Raskin RE. Morphologic, cytochemical staining, and ultrastructural charac-
teristics of blood cells from eastern diamondback rattlesnakes (Crotalus adamanteus). Am J Vet Res.
1999; 60: 507–514. PMID: 10211697
33. Dutton CJ, Taylor P. A comparison between pre- and posthibernation morphometry, hematology, and
blood chemistry in viperid snakes. J Zoo Wildl Med. 2003; 34: 53–58. https://doi.org/10.1638/1042-
7260(2003)34[0053:ACBPAP]2.0.CO;2 PMID: 12723800
34. Allender MC, Mitchell MA, Phillips CA, Gruszynski K, Beasley VR. Hematology, plasma biochemistry,
and antibodies to select viruses in wild-caught eastern massasauga rattlesnakes (Sistrurus catenatus
catenatus) from Illinois. J Wildl Dis. 2006; 42: 107–114. https://doi.org/10.7589/0090-3558-42.1.107
PMID: 16699153
35. Guzman DSM, Mitchell MA, Acierno M. Determination of plasma osmolality and agreement between
measured and calculated values in captive male corn snakes (Pantherophis [Elaphe] guttatus guttatus).
J Herpetol Med Surg. 2011; 21: 16–19.
36. Wack RF, Hansen E, Small M, Poppenga R, Bunn D, Johnson CK. Hematology and plasma biochemis-
try values for the giant garter snake (Thamnophis gigas) and valley garter snake (Thamnophis sirtalis
fitchi) in the Central Valley of California. J Wildl Dis. 2012; 48: 307–313. https://doi.org/10.7589/0090-
3558-48.2.307 PMID: 22493106
37. Species 360 [Internet]. Zoological Information Management System (ZIMS) for Medical: Expected test
results for Agkistrodon contortrix. 2019 September 05. Available from http://zims.Species360.org/
38. Species 360 [Internet]. Zoological Information Management System (ZIMS) for Medical: Expected test
results for Pantherophis obsoletus. 2019 September 05. Available from http://zims.Species360.org/
39.
LaGrange SM, Kimble SJ, MacGowan BJ, Williams RN. Seasonal variation in hematology and blood
plasma chemistry values of the timber rattlesnake (Crotalus horridus). J Wildl Dis. 2014; 50: 990–993.
https://doi.org/10.7589/2013-10-267 PMID: 25098306
40. Russo EA, McEntee L, Applegate L, Baker JS. Comparison of two methods for determination of white
blood cell counts in macaws. J Am Vet Med Assoc. 1986; 189: 1013–1016. PMID: 3505917
41. Bernal W, Donaldson N, Wyncoll D, Wendon J. Blood lactate as an early predictor of outcome in para-
cetamol-induced acute liver failure: a cohort study. Lancet. 2002; 359: 558–563. https://doi.org/10.
1016/S0140-6736(02)07743-7 PMID: 11867109
42.
Zacher LA, Berg J, Shaw SP, Kudej RK. Association between outcome and changes in plasma lactate
concentration during presurgical treatment in dogs with gastric dilatation-volvulus: 64 cases (2002–
2008). J Am Vet Med Assoc. 2010; 236: 892–897. https://doi.org/10.2460/javma.236.8.892 PMID:
20392189
43. Keller KA, Innis CJ, Tlusty MF, Kennedy AE, Bean SB, Cavin JM, et al. Metabolic and respiratory
derangements associated with death in cold-stunned Kemp’s ridley turtles (Lepidochelys kempii): 32
cases (2005–2009). J Am Vet Med Assoc. 2012; 240: 317–323. https://doi.org/10.2460/javma.240.3.
317 PMID: 22256849
44. Shapiro BA, Peruzzi WT, Kozlowski-Templin R. Temperature correction of blood gas values. In: Sha-
piro B, Peruzzi WT, Kozlowski-Templin R. Clinical Application of Blood Gases. 5th ed. St. Louis, MO,
USA: Mosby-Year Book Inc; 1994. pp. 227–233.
45. Hess JC, Benson GJ, Grimm KA, Sarr R, Constable PD, Tranquilli WJ. Minimum alveolar concentration
of isoflurane and arterial blood gas values in anesthetized green iguanas, Iguana iguana. J Herpetol
Med Surg. 2007; 17: 118–124.
46. Stockham SL, Scott MA. Blood gases, blood pH, and strong ion difference. In: Stockham SL, Scott MA.
Fundamentals of Veterinary Clinical Pathology. 2nd ed. Ames, IA, USA: Wiley-Blackwell;
2008. pp.559–592.
47. Chittick EJ, Stamper MA, Beasley JF, Lewbart GA, Horne WA. Medetomidine, ketamine, and sevoflur-
ane for anesthesia of injured loggerhead sea turtles: 13 cases (1996–2000). J Am Vet Med Assoc.
2002; 221: 1019–1025. https://doi.org/10.2460/javma.2002.221.1019 PMID: 12369681
48.
Innis CJ, Tlusty M, Merigo C, Weber ES. Metabolic and respiratory status of cold-stunned Kemp’s ridley
sea turtles (Lepidochelys kempii). J Comp Physiol B. 2007; 177: 623–630. https://doi.org/10.1007/
s00360-007-0160-9 PMID: 17431640
49. Al-Badry KS, El-Deib S, Nuzhy S. Changes in serum chemistry parameters of the Egyptian cobra (Naja
haje haje) in correlation with the hibernation cycle. J Therm Biol. 1992; 17: 111–114.
PLOS ONE | https://doi.org/10.1371/journal.pone.0229102 February 14, 2020
14 / 15
Venous Analytes of Copperheads and Ratsnakes
50.
Innis CJ, Merigo C, Cavin JM, Hunt K, Dodge KL, Lutcavage M. Serial assessment of the physiological
status of leatherback turtles (Dermochelys coriacea) during direct capture events in the northwestern
Atlantic Ocean: comparison of post-capture and pre-release data. Conserv Physiol. 2014; 2: cou048.
https://doi.org/10.1093/conphys/cou048 PMID: 27293669
51. Pfitzer S, Ganswindt A, Fosgate GT, Botha PJ, Myburgh JG. Capture of farmed Nile crocodiles (Croco-
dylus niloticus): comparison of physiological parameters after manual capture and after capture with
electrical stunning. Vet Rec. 2014; 175: 304–308. https://doi.org/10.1136/vr.102438 PMID: 25096588
52. Bertelsen MF, Buchanan R, Jensen HM, Leite CAC, Abe AS, Nielsen SS, et al. Assessing the influence
of mechanical ventilation on blood gases and blood pressure in rattlesnakes. Vet Anaesth Analg. 2015;
42: 386–393. https://doi.org/10.1111/vaa.12221 PMID: 25174935
53.
Lewbart GA, Hirschfeld M, Brothers JR, Muñoz-Pe´rez JP, Denkinger J, Vinueza L, et al. Blood gases,
biochemistry and haematology of Gala´pagos marine iguanas (Amblyrhynchus cristatus). Conserv Phy-
siol. 2015; 3: cov034. https://doi.org/10.1093/conphys/cov034 PMID: 27293719
54. Mader DR, Rudloff E. Emergency and Critical Care. In: Mader DR, Divers SJ, editors. Reptile Medicine
and Surgery. 2nd ed. St. Louis, MO, USA: Saunders Elsevier; 2006. pp. 533–548.
55. Schuett GW, Grober MS. Post-fight levels of plasma lactate and corticosterone in male copperheads,
Agkistrodon contortrix (Serpentes, Viperidae): differences between winners and losers. Physiol Behav.
2000; 71: 335–341. https://doi.org/10.1016/s0031-9384(00)00348-6 PMID: 11150566
56. Kreger MD, Mench JA. Physiological and behavioral effects of handling and restraint in the ball python
(Python regius) and the blue-tongued skink (Tiliqua scincoides). Appl Anim Behav Sci. 1993; 38: 323–
336.
57. Secor SM, Nagy KA. Bioenergetic correlates of foraging mode for the snakes Crotalus cerastes and
Masticophis flagellum. Ecology, 1994; 75, 1600–1614.
58. Beaupre SJ, Montgomery CE. Chapter 11: The meaning and consequences of foraging mode in snakes
In: Reilly SM, McBrayer LB, Miles DB, editors. Lizard Ecology: The Evolutionary Consequences of For-
aging Mode. Cambride: Cambridge University Press; 2007; 334–368.
59. Hochachka PW, Storey KB. Metabolic consequences of diving in animals and man. Science. 1975;
187: 613–621. https://doi.org/10.1126/science.163485 PMID: 163485
60. Bennett AF, Dawson WR. Metabolism. In: Gans C, Dawson WR, editors. Biology of the Reptilia, Vol 5
(Physiology A). New York and San Francisco, USA: Academic Press; 1976. pp. 127–223.
61. Dantzler WH. Renal Function (with Special Emphasis on Nitrogen Excretion). In: Gans C, Dawson WR,
editors. Biology of the Reptilia Vol 5 (Physiology A). New York and San Francisco, USA: Academic
Press; 1976. pp. 447–498.
62. Mosley CAE, Dyson D, Smith DA. Minimum alveolar concentration of isoflurane in green iguanas and
the effect of butorphanol on minimum alveolar concentration. J Am Vet Med Assoc. 2003; 222: 1559–
1564. https://doi.org/10.2460/javma.2003.222.1559 PMID: 12784962
63. Mosley CAE, Dyson D, Smith DA. The cardiovascular dose–response effects of isoflurane alone and
combined with butorphanol in the green iguana (Iguana iguana). Vet Anaesth Analg. 2004; 31: 64–72.
https://doi.org/10.1111/j.1467-2995.2004.00135.x PMID: 14756755
64. Christiansen EF, Stoskopf MK, Harms CA. Pre- and post-surgical evaluation of bearded dragons under-
going sterilization. J Herpetol Med Surg. 2013; 23: 83–90.
65. Kirschbaum B. Loss of carbon dioxide from serum samples exposed to air. Effect on blood gas parame-
ters and strong ions. Clinica Chimica Acta. 2003; 334: 241–244.
PLOS ONE | https://doi.org/10.1371/journal.pone.0229102 February 14, 2020
15 / 15
|
10.1371_journal.pone.0229895 | RESEARCH ARTICLE
A simulation based difficult conversations
intervention for neonatal intensive care unit
nurse practitioners: A randomized controlled
trial
Roberta Bowen1, Kate M. Lally2,3, Francine R. Pingitore3,4, Richard Tucker1, Elisabeth
C. McGowan1,3, Beatrice E. LechnerID
1,3*
1 Department of Neonatology, Women & Infants Hospital, Providence, RI, United States of America,
2 Program in Palliative Care, Care New England Health System, Providence, RI, United States of America,
3 Warren Alpert Medical School of Brown University, Providence, RI, United States of America,
4 Department of Pediatrics, Hasbro Children’s Hospital, Providence, RI, United States of America
* blechner@wihri.org
Abstract
Background
Neonatal nurse practitioners are often the front line providers in discussing unexpected
news with parents. This study seeks to evaluate whether a simulation based Difficult Con-
versations Workshop for neonatal nurse practitioners leads to improved skills in conducting
difficult conversations.
Methods
We performed a randomized controlled study of a simulation based Difficult Conversa-
tions Workshop for neonatal nurse practitioners (n = 13) in a regional level IV neonatal
intensive care unit to test the hypothesis that this intervention would improve communica-
tion skills. A simulated test conversation was performed after the workshop by the inter-
vention group and before the workshop by the control group. Two independent blinded
content experts scored each conversation using a quantitative communication skills per-
formance checklist and by assigning an empathy score. Standard statistical analysis was
performed.
Results
Randomization occurred as follows: n = 5 to the intervention group, n = 7 to the control
group. All participants were analyzed in each group. Participation in the simulation based
Difficult Conversations Workshop increases participants’ empathy score (p = 0.015) and the
use of communication skills (p = 0.013) in a simulated clinical encounter.
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: Bowen R, Lally KM, Pingitore FR, Tucker
R, McGowan EC, Lechner BE (2020) A simulation
based difficult conversations intervention for
neonatal intensive care unit nurse practitioners: A
randomized controlled trial. PLoS ONE 15(3):
e0229895. https://doi.org/10.1371/journal.
pone.0229895
Editor: Karen-Leigh Edward, Swinburne University
of Technology, AUSTRALIA
Received: December 6, 2019
Accepted: February 16, 2020
Published: March 9, 2020
Peer Review History: PLOS recognizes the
benefits of transparency in the peer review
process; therefore, we enable the publication of
all of the content of peer review and author
responses alongside final, published articles. The
editorial history of this article is available here:
https://doi.org/10.1371/journal.pone.0229895
Copyright: © 2020 Bowen et al. This is an open
access article distributed under the terms of the
Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: All relevant data are
within the manuscript.
PLOS ONE | https://doi.org/10.1371/journal.pone.0229895 March 9, 2020
1 / 12
PLOS ONE Funding: The authors received no specific funding
for this work.
Competing interests: The authors have declared
that no competing interests exist.
Conclusions
Our study demonstrates that a lecture and simulation based Difficult Conversations Work-
shop for neonatal nurse practitioners improves objective communication skills and empathy
in conducting difficult conversations.
Difficult conversations simulation in the NICU
Introduction
The ability to communicate effectively with patients’ families is an essential skill for those caring
for infants in the neonatal intensive care unit (NICU). Delivering bad news is a skill set not typi-
cally taught in the formal education of advanced practice registered nurses. In the United States,
these advanced practice registered nurses, or nurse practitioners (NPs), provide care alongside
physicians, often in a role similar to the physician’s role and sometimes in lieu of the physician.
In the NICU, neonatal NPs diagnose and treat infants, perform procedures and interact with
and provide support to parents. Thus, acquisition of skills for leading difficult conversations is
essential for nurse practitioners to be successful in their full scope of practice. Conducting
research on communication skills training in the clinical setting is challenging and the current
status of the field does not allow for the identification of one gold standard [1, 2]. Even fewer
studies exist in the context of neonatology. Neonatal NPs feel that their education is lacking in
this key component of practice [3], and studies of NICU communication skills did not include
NPs in the assessment [4] or only measured NPs’ self-reported and thus subjective outcomes
[5]. The complicated communication task of delivering bad news to the parents of infants is
fraught with discomfort and uncertainty for the practitioner delivering the news [6], especially
given that bad news around the birth of an infant is not in line with parental expectations.
Most clinicians rely on skills demonstrated by mentors or those learned by trial and error,
despite the fact that taking part in a formal program to enhance communication skills leads to
an improvement in communication skills [7, 8], while studies have demonstrated that patients
desire good communication [9] and that communication skills can be taught and retained
[10]. Parents of infants in the NICU are at very high risk for adverse mental health outcomes
[11]. Thus, communication approaches used by the medical team, including NPs, gain utmost
importance. When working in level 1 and 2 community hospital nurseries, neonatal nurse
practitioners are often the front line providers in discussing unexpected news with parents.
Thus, we sought to evaluate the hypothesis that a lecture and simulation based Difficult Con-
versations Workshop for the neonatal nurse practitioners will increase skill in conducting dif-
ficult conversations with patients’ families.
Materials and methods
We performed a randomized controlled prospective study of a simulation based Difficult Con-
versations Workshop for NICU nurse practitioner staff at a large regional level IV NICU in the
Northeast of the United States. The research related to human use has been approved by the
Women & Infants Hospital Institutional Review Board. Written informed consent was
obtained. In this 80 bed level 3 NICU, a simulation based Difficult Conversations Workshop is
part of the training program for the neonatal-perinatal medicine fellows.
Participants
The clinical NICU nurse practitioner group consists of 31 nurse practitioners, who work in a
level IV regional NICU as well as multiple level II community hospital NICUs. All NPs were
PLOS ONE | https://doi.org/10.1371/journal.pone.0229895 March 9, 2020
2 / 12
PLOS ONE Difficult conversations simulation in the NICU
invited to participate in the study. The NPs were recruited to participate in the study using
email as well as a presentation of the study by one of the study authors at a monthly NP staff
meeting. Recruitment and workshop were performed from May 2016 to July 2016. Each three
hour session of the simulation based Difficult Conversations Workshop consisted of 4–6 par-
ticipants. Participants in each session were randomized using the web-based randomization
tool Randomizer.org to either the intervention or control group. Simple randomization was
performed with a randomization allocation of 1:1. Randomization was performed at the begin-
ning of the workshop. Both groups participated in a three hour workshop.
Study structure (Fig 1)
Prior to randomization, all study participants (intervention group and control group) filled
out an anonymous pre-workshop survey. Then, after randomization, the control group per-
formed the Test Scenario, which was a standardized clinically relevant simulation scenario
using trained improvisational actors as parents. They then took part in the simulation based
Difficult Conversations Workshop so as to allow them the opportunity to benefit from the
learning opportunity. The intervention group, on the other hand, took part in the simulation
based Difficult Conversations Workshop prior to performing the Test Scenario. At the end of
the Workshop and Test Scenarios, all participants filled out a post-workshop survey. Data col-
lection on the pre- and post-workshop surveys ascertained demographics, past experiences
with communication skills training, past experiences leading difficult conversations in the clin-
ical setting, as well as feedback on the workshop. The workshop took place in the Care New
England Simulation Center at Women & Infants Hospital.
Simulation based difficult conversations Workshop
The simulation based Difficult Conversations Workshop was a 4.5 hour workshop that con-
sisted of three components (Fig 1). First, the participants were presented with a lecture on dif-
ficult conversation communication skills. This lecture was 30 minutes long and highlighted
the basic tenets of communication skills in healthcare. Next, each participant took part in a
simulation Teaching Scenario, a clinically relevant practice difficult conversation with a
trained improvisational actor that was about ten minutes long, while remaining participants
observed the scenario via live video. Finally, at the end of the Teaching Scenarios, a facilitated
debriefing session was held for all participants. This debriefing session was usually an hour to
two hours in length. The workshop was led by a neonatologist who is the director of and
trainer in the Difficult Conversations for Neonatal Fellows Training program. Each simulated
Teaching Scenario reflected a situation typical of the NICU NP’s work environment. The
trained actors functioned in the role of a parent during the simulated difficult conversations.
Performance assessment
The Test Scenario was a 10 minute conversation with a trained improvisational actor in a sim-
ulated standardized clinical scenario. The encounter took place in the Women & Infants Hos-
pital Simulation Center and was videotaped, but not shown via live video to any intervention
NPs, control NPs or trainers (in contrast to the Teaching Scenarios). This was done to main-
tain the integrity of the standardized Test Scenario for all intervention and control NPs. The
Test Scenario simulation was scored at a later date independently by two blinded content
expert observers. One observer was a board certified palliative care physician; the other
observer was a doctorally prepared pediatric psychiatric clinical nurse specialist with expertise
in interpersonal communication and relationships. These observers did not work with or
know any of the participants and were blinded to participant group. In order to assess the
PLOS ONE | https://doi.org/10.1371/journal.pone.0229895 March 9, 2020
3 / 12
PLOS ONE Difficult conversations simulation in the NICU
Fig 1. Study flow diagram. NNP = neonatal nurse practitioner.
https://doi.org/10.1371/journal.pone.0229895.g001
performance of each participant, the observers completed a quantitative communication skills
performance checklist as well as assigning an empathy score to rate the participant’s level of
empathy on a scale of 1 (no empathy) to 10 (extremely empathetic) (Fig 2). The quantitative
communication skills performance checklist was developed using a two-step approach. A
review of the literature was performed for communication skill checklists, then the final check-
list was curated by the authors via expert consensus. The empathy score was developed via
expert consensus.
PLOS ONE | https://doi.org/10.1371/journal.pone.0229895 March 9, 2020
4 / 12
PLOS ONE Difficult conversations simulation in the NICU
Fig 2. Evaluation tool utilized by blinded independent content experts to evaluate recorded Difficult Conversations Test Scenarios
performed by participants. NNP = neonatal nurse practitioner.
https://doi.org/10.1371/journal.pone.0229895.g002
PLOS ONE | https://doi.org/10.1371/journal.pone.0229895 March 9, 2020
5 / 12
PLOS ONE Difficult conversations simulation in the NICU
Teaching & Test Scenarios
In Teaching Scenario #1, a mother was informed that child protective services had informed the
hospital that they would investigate the mother after her twins were born. In Teaching Scenario
#2, a mother was informed that her infant had failed a congenital heart disease screening and
needed to be transferred to a regional NICU to rule out congenital heart disease. In the Test Sce-
nario, a mother was told that there was clinical suspicion of Down Syndrome in her newborn.
Data analysis
Statistical analysis was performed as follows. Differences in exhibition of communication skills
between the groups was tested using Fisher’s exact test, and numbers of skills demonstrated
and empathy scores were compared via the Student’s t-test.
Inter-rater reliability on the scoring of the Test Scenario was measured for the communica-
tion skills items using a pooled kappa statistic. Rater agreement on empathy scores was calcu-
lated using the two one-sided t-tests (TOST) method, with agreement limits of ±3 points.
Results
13 out of 31 participated; n = 5 in the intervention group, n = 7 in the control group. One
video could not be assessed due to technical difficulties with sound recording. Demographics
of the group and experience with difficult conversations as a trainee and in the clinical setting
are presented in Table 1.
Table 1. Participant demographics and experience with difficult conversations.
Survey questions
Number of years as an NP taking care of infants
0–1
2–5
6–10
> 10
Average number of weekly hours worked
12–24
25–32
33–40
41–55
> 55
NICU level most often worked in
3 or 4
2
1
Any work in level 1/2 community hospital nursery
yes
Take transport call
yes
Received education during training/career on communicating bad news to the family of an infant
yes
Number of times present in the past year when bad news was given to the family of an infant
0
n = 13 (%)
1 (8)
5 (38)
2 (15)
4 (31)
0 (0)
2 (17)
3 (25)
3 (25)
4 (33)
12 (92)
1 (8)
0 (0)
9 (69)
7 (54)
2 (15)
0 (0)
(Continued )
PLOS ONE | https://doi.org/10.1371/journal.pone.0229895 March 9, 2020
6 / 12
PLOS ONE Difficult conversations simulation in the NICU
Table 1. (Continued)
Survey questions
1–2
3 or more
Number of times in the past year you gave bad news to the family of an infant
0
1–2
3 or more
Extent to which you feel competent to deliver bad news to a family of an infant in your care
not competent
somewhat competent
moderately competent
competent
don’t know
https://doi.org/10.1371/journal.pone.0229895.t001
n = 13 (%)
3 (23)
10 (77)
3 (23)
5 (38)
5 (38)
2 (15)
7 (54)
3 (23)
1 (8)
0 (0)
Participation in the simulation based Difficult Conversations Workshop increases the use of
communication skills in a simulated clinical encounter and increases participants’ empathy
score.
In the intervention group, the mean number of predefined communication skill behaviors
exhibited by each participant was higher than in the control group (12 skills compared to 8
skills per scenario; p = 0.013). Among the individual communication skill behaviors compared
individually between the groups, only “asks parents open-ended questions” was significantly
higher in the post intervention group (p = 0.047). In the intervention group, the mean empa-
thy score was higher compared to the control group (8.4 compared to 6.2; p = 0.015).
Independent of participation in the simulation based Difficult Conversations Workshop, some
communication skills are used more often than others.
The frequency with which individual communication skills were applied in simulated clini-
cal encounters was similar among the two groups. Some skills, such as “Introduces/Re-intro-
duces self”, were almost always displayed, while others, such as “Asks parents to repeat back”
were never displayed (Table 2).
Interobserver agreement between the two independent blinded reviewers in communica-
tion skill scores was 74% agreement overall (range for individual participants between 59 and
94% and for individual skills between 42% and 100%) with an interrater reliability pooled
kappa of 0.77. In the empathy score, the top three scores and the bottom two were identical
between the two reviewers, while there were some differences in the middle of the field. The
two one sided t-tests demonstrated equivalence of empathy score differences, with differences
within three points considered equivalent (p = 0.0030).
On the post-intervention survey, participants rated the workshop between 5.8 and 6.0 on a
variety of measures on a 6.0 scale (Table 3).
Discussion
Our study demonstrates that an intervention consisting of a structured lecture and simulation
based communication skills workshop for neonatal NPs leads to an increase in the use of spe-
cific communication skills as well as improvement in a perceived empathy score in a simulated
difficult conversation setting. This is the first study assessing these objective outcomes in neo-
natal nurse practitioners, while a previous study demonstrated improved self-reported confi-
dence in difficult conversations in neonatal fellows and nurse practitioners [5]. As nurse
PLOS ONE | https://doi.org/10.1371/journal.pone.0229895 March 9, 2020
7 / 12
PLOS ONE Difficult conversations simulation in the NICU
Table 2. Utilized communications skills.
Communications skill
Introduces/Re-introduces self
Body Position (Seated/Positioned at eye level to parent; not hovering
over parent; lean forward toward parent)
Makes statements that furnish hope (“I hope I am wrong about this”)
Summarizes and makes a follow up plan. Assures parents they will be
available
Avoids medical jargon (“atypical features” instead of dysmorphic
features/Down Syndrome/Trisomy 21)
Uses expressions that communicate empathy (“I wish I had better
news”)
Uses the baby’s name during the conversation
Suggests additional supportive resources for the parents (chaplain,
social worker, etc)
Asks what the parent(s) know/suspect
Speaks slowly in short simple sentences
Acknowledges the parents’ emotions (“I can see how worried you
are,” “I know this must be shocking,” “It’s OK to cry,” “I can see that
you don’t know what to say”)
Asks parents open-ended questions
Asks parent(s) if there is anyone else they would like to be present for
the meeting
Foreshadows the bad news (“I’m sorry but I have bad news”)
Pauses consciously and allows for silence after delivering bad news
If visitors present, gives family a choice on who should be present for
the meeting
Asks parents to repeat back what they have been told
https://doi.org/10.1371/journal.pone.0229895.t002
Intervention group
(n = 5) (%)
Control group
(n = 7) (%)
100
100
100
100
100
80
80
90
50
80
40
80
70
70
60
00
00
86
86
86
79
64
64
50
50
43
43
43
36
29
29
21
00
00
practitioners are important members of the multidisciplinary teams providing care for neo-
nates in many NICUs across the United States, it is important to train neonatal NPs in difficult
conversations and breaking bad news, particularly in the current changing climate in health-
care, where NPs are providing more and more care in academic medical centers as well as
community hospitals.
Objectively assessable specific communication skills are important components of difficult
conversations in the clinical setting. Our results are supported by other studies, which have
shown that simulation is an effective tool for realistic training in difficult conversations in the
context of neonatal care, for example in decision-making at the limits of viability [12], as well
as in pediatrics [8], and leads to improved actual communication skills [7]. Studies also
Table 3. Participant workshop evaluations.
Mean score (1–6 [extremely ineffective/unsatisfactory—extremely effective/outstanding])
The lecture on communication skills was helpful/informative
The simulation was helpful/informative
The facilitated debriefing was helpful/informative
The environment felt safe to ask questions/share thoughts
After attending the workshop, I feel more competent to lead a difficult conversation
Overall satisfaction with the session
The workshop should be part of neonatal NP orientation/training
https://doi.org/10.1371/journal.pone.0229895.t003
n = 13
5.9
5.9
6.0
6.0
5.8
6.0
6.0
PLOS ONE | https://doi.org/10.1371/journal.pone.0229895 March 9, 2020
8 / 12
PLOS ONE Difficult conversations simulation in the NICU
demonstrate that taking part in a formal program aimed at improving communication skills in
difficult conversations influences pediatric provider confidence in managing difficult clinical
scenarios [13], as well as leading to better humanistic skills and better delivery of bad news
[14] and to improved knowledge and comfort levels in communication [15].
Communication skills training has also been used successfully to improve residents’ skills
in code status discussions [16], for genetic counseling training [17], communication for anes-
thesia residents[10] and the disclosure of medical errors [18]. Furthermore, simulation has
been shown to improve long term retention of skills and self-reported changes in behavior
[19] [15] [20] as well as the long term retention of confidence in one’s communication skills in
breaking bad news [21].
While it is important to assess objective communication skills in difficult conversations,
not every aspect of the level of skill that care providers demonstrate can be assessed using a
specific skill checklist. In addition to the objective communication skills that are necessary for
breaking bad news, empathy plays a significant role in patient-provider communication.
Parents prioritize communication[9] and want caring providers, for example when receiving
prenatal consults by neonatologists for congenital anomalies [22]. Additionally, studies have
shown that physician empathy is associated with increased adherence to therapy and improved
clinical outcomes [23–25]. While possible differences in communication style between physi-
cians and nurse practitioners have not been studied, nurses’ and physicians’ patterns of com-
munication differ in enacted NICU conversations; physicians provide more biomedical
information while nurses provide more psychosocial information [4].
Furthermore, empathy is an important component of the patient-practitioner interaction
from the practitioner perspective as well. For example, empathy in medical students is associ-
ated with a decrease in burnout [26, 27]. Thus, the empathy score was utilized as an additional
marker of the interaction between the provider and patient.
Our study demonstrates that lecture plus simulation based training improves empathy per-
ceived by an expert observer in addition to improving objective communication skills. Empa-
thy does not lend itself to one simple definition. One approach to categorizing empathy is into
cognitive empathy vs. affective empathy, reviewed in [28], in which cognitive empathy is asso-
ciated with external traits that can be learned, while affective empathy is not. Thus, for the pur-
poses of this study, we defined empathy as a cognitive and thus behavioral trait, which
consequently is a characteristic that lends itself to modification by training. Despite the abun-
dance of alternative definitions for empathy [28], the two independent content expert observ-
ers were able to assess empathy in the Test Scenarios with high interrater reliability. They
ranked the level of empathy that participants displayed in scenarios similarly: both their high-
est and lowest ranking participants were identical, irrespective of the actual number of the
empathy score on the scale. Such concordance was achieved despite the fact that the content
expert observers were not trained to look for specific signs, but received the sole instructions
to score scenarios on a scale of 1 (no empathy) to 10 (extremely empathic).
One limitation of this study is that individual participants were not tested using both a pre-
and a post-intervention scenario, given that the increased time commitment necessary for that
experimental model was not possible due to participants’ clinical staffing requirements. The
disadvantage of not having the same participants in both the pre- and post-intervention group
is a decrease in the signal to noise ratio. However, given that we nonetheless saw significant
improvement in both specific skills and overall empathy scores, we hypothesize that the results
would have been even stronger if interpersonal differences had been accounted for using the
same participants for both arms of the study.
Another limitation of the study is that the communication skill result and the empathy
score result may not be independent variables, as it is possible that the observing content
PLOS ONE | https://doi.org/10.1371/journal.pone.0229895 March 9, 2020
9 / 12
PLOS ONE Difficult conversations simulation in the NICU
expert evaluators were subconsciously influenced in their assessment of the empathy score
based on the number of communication skills demonstrated. If this were the case, this would
not detract from the validity of the results. In fact, this mechanism, if it were at play in these
assessments, would support the hypothesis that empathy can be learned as a specific skill set,
aligning with the cognitive/behavioral definition of empathy, thus suggesting that specific
learned communicative behavioral skills may impact the patient’s perception of empathy.
Furthermore, we recognize that only 13/31 NPs took part in this workshop. Upon further
investigation, the most common reason for non-participation included the complex schedul-
ing of clinical load. To see if these results are generalizable, a larger cohort may be needed.
Nonetheless, this small study demonstrated a difference between the intervention and control
groups.
Since others have shown that trainees and their program directors are more lenient in their
assessment of communication simulation performance compared to patients and communica-
tion experts [29], an advantage of this study is that we utilized independent blinded content
experts to perform the video assessments for both the skills assessment and the empathy score.
An additional approach that may improve difficult conversation skill and empathy scores
may be to incorporate erroneous examples into the lecture component of the workshop, as
these have been shown to improve breaking bad news simulation performance in nursing stu-
dents [30].
In summary, our study demonstrates that a lecture and simulation based Difficult Conver-
sations Simulation workshop improves objective communication skills and empathy in neona-
tal nurse practitioners in conducting difficult conversations with patients’ families as perceived
by an expert observer. Future studies will need to address the long term retention of learned
communication skills as well as the transfer of communication skills from simulation settings
to actual clinical practice.
Author Contributions
Conceptualization: Roberta Bowen, Elisabeth C. McGowan, Beatrice E. Lechner.
Data curation: Roberta Bowen, Kate M. Lally, Francine R. Pingitore, Elisabeth C. McGowan,
Beatrice E. Lechner.
Formal analysis: Richard Tucker, Elisabeth C. McGowan, Beatrice E. Lechner.
Investigation: Roberta Bowen, Kate M. Lally, Francine R. Pingitore, Elisabeth C. McGowan,
Beatrice E. Lechner.
Methodology: Kate M. Lally, Richard Tucker, Elisabeth C. McGowan, Beatrice E. Lechner.
Project administration: Roberta Bowen, Elisabeth C. McGowan, Beatrice E. Lechner.
Resources: Beatrice E. Lechner.
Supervision: Beatrice E. Lechner.
Writing – original draft: Roberta Bowen, Beatrice E. Lechner.
Writing – review & editing: Roberta Bowen, Kate M. Lally, Francine R. Pingitore, Richard
Tucker, Elisabeth C. McGowan, Beatrice E. Lechner.
References
1. Ditton-Phare P, Loughland C, Duvivier R, Kelly B. Communication skills in the training of psychiatrists:
A systematic review of current approaches. Aust N Z J Psychiatry. 2017; 51(7):675–92. https://doi.org/
10.1177/0004867417707820 PMID: 28462636
PLOS ONE | https://doi.org/10.1371/journal.pone.0229895 March 9, 2020
10 / 12
PLOS ONE Difficult conversations simulation in the NICU
2. Deveugele M. Communication training: Skills and beyond. Patient Educ Couns. 2015; 98(10):1287–91.
https://doi.org/10.1016/j.pec.2015.08.011 PMID: 26298220
3. Botwinski C. NNP education in neonatal end-of-life care: a needs assessment. MCN Am J Matern Child
Nurs. 2010; 35(5):286–92. https://doi.org/10.1097/NMC.0b013e3181e62440 PMID: 20706099
4.
Lamiani G, Meyer EC, Browning DM, Brodsky D, Todres ID. Analysis of enacted difficult conversations
in neonatal intensive care. J Perinatol. 2009; 29(4):310–6. https://doi.org/10.1038/jp.2008.228 PMID:
19148109
5. Boss RD, Urban A, Barnett MD, Arnold RM. Neonatal Critical Care Communication (NC3): training
NICU physicians and nurse practitioners. J Perinatol. 2013; 33(8):642–6. https://doi.org/10.1038/jp.
2013.22 PMID: 23448940
6. Dosanjh S, Barnes J, Bhandari M. Barriers to breaking bad news among medical and surgical residents.
Med Educ. 2001; 35(3):197–205. https://doi.org/10.1046/j.1365-2923.2001.00766.x PMID: 11260440
7. Meyer EC, Brodsky D, Hansen AR, Lamiani G, Sellers DE, Browning DM. An interdisciplinary, family-
focused approach to relational learning in neonatal intensive care. J Perinatol. 2011; 31(3):212–9.
https://doi.org/10.1038/jp.2010.109 PMID: 20706191
8.
Tobler K, Grant E, Marczinski C. Evaluation of the impact of a simulation-enhanced breaking bad news
workshop in pediatrics. Simul Healthc. 2014; 9(4):213–9. https://doi.org/10.1097/SIH.
0000000000000031 PMID: 24787559
9. Meyer EC, Ritholz MD, Burns JP, Truog RD. Improving the quality of end-of-life care in the pediatric
intensive care unit: parents’ priorities and recommendations. Pediatrics. 2006; 117(3):649–57. https://
doi.org/10.1542/peds.2005-0144 PMID: 16510643
10. Karam VY, Barakat H, Aouad M, Harris I, Park YS, Youssef N, et al. Effect of a simulation-based work-
shop on breaking bad news for anesthesiology residents: an intervention study. BMC Anesthesiol.
2017; 17(1):77. https://doi.org/10.1186/s12871-017-0374-7 PMID: 28615002
11. Roque ATF, Lasiuk GC, Radunz V, Hegadoren K. Scoping Review of the Mental Health of Parents of
Infants in the NICU. J Obstet Gynecol Neonatal Nurs. 2017; 46(4):576–87. https://doi.org/10.1016/j.
jogn.2017.02.005 PMID: 28506679
12. Boss RD, Donohue PK, Roter DL, Larson SM, Arnold RM. "This is a decision you have to make": using
simulation to study prenatal counseling. Simul Healthc. 2012; 7(4):207–12. https://doi.org/10.1097/SIH.
0b013e318256666a PMID: 22569285
13. Dickens DS. Building competence in pediatric end-of-life care. J Palliat Med. 2009; 12(7):617–22.
https://doi.org/10.1089/jpm.2009.0032 PMID: 19594346
14. Vetto JT, Elder NC, Toffler WL, Fields SA. Teaching medical students to give bad news: does formal
instruction help? J Cancer Educ. 1999; 14(1):13–7. https://doi.org/10.1080/08858199909528567
PMID: 10328318
15. Vadnais MA, Dodge LE, Awtrey CS, Ricciotti HA, Golen TH, Hacker MR. Assessment of long-term
knowledge retention following single-day simulation training for uncommon but critical obstetrical
events. J Matern Fetal Neonatal Med. 2012; 25(9):1640–5. https://doi.org/10.3109/14767058.2011.
648971 PMID: 22191668
16. Szmuilowicz E, Neely KJ, Sharma RK, Cohen ER, McGaghie WC, Wayne DB. Improving residents’
code status discussion skills: a randomized trial. J Palliat Med. 2012; 15(7):768–74. https://doi.org/10.
1089/jpm.2011.0446 PMID: 22690890
17. Holt RL, Tofil NM, Hurst C, Youngblood AQ, Peterson DT, Zinkan JL, et al. Utilizing high-fidelity crucial
conversation simulation in genetic counseling training. Am J Med Genet A. 2013; 161A(6):1273–7.
https://doi.org/10.1002/ajmg.a.35952 PMID: 23633180
18. Matos FM, Raemer DB. Mixed-realism simulation of adverse event disclosure: an educational method-
ology and assessment instrument. Simul Healthc. 2013; 8(2):84–90. https://doi.org/10.1097/SIH.
0b013e31827cbb27 PMID: 23334365
19. Hughes S, Cusack J, Fawke J. PC.68 Learning point recall and self-perceived behavioural change fol-
lowing multi-disciplinary high fidelity point of care simulation training. Arch Dis Child Fetal Neonatal Ed.
2014; 99 Suppl 1:A59.
20. Bender J, Kennally K, Shields R, Overly F. Does simulation booster impact retention of resuscitation
procedural skills and teamwork? J Perinatol. 2014; 34(9):664–8. https://doi.org/10.1038/jp.2014.72
PMID: 24762413
21. Karkowsky CE, Landsberger EJ, Bernstein PS, Dayal A, Goffman D, Madden RC, et al. Breaking Bad
News in obstetrics: a randomized trial of simulation followed by debriefing or lecture. J Matern Fetal
Neonatal Med. 2016; 29(22):3717–23. https://doi.org/10.3109/14767058.2016.1141888 PMID:
26786087
PLOS ONE | https://doi.org/10.1371/journal.pone.0229895 March 9, 2020
11 / 12
PLOS ONE Difficult conversations simulation in the NICU
22. Miquel-Verges F, Woods SL, Aucott SW, Boss RD, Sulpar LJ, Donohue PK. Prenatal consultation with
a neonatologist for congenital anomalies: parental perceptions. Pediatrics. 2009; 124(4):e573–9.
https://doi.org/10.1542/peds.2008-2865 PMID: 19736266
23. Hojat M, Louis DZ, Markham FW, Wender R, Rabinowitz C, Gonnella JS. Physicians’ empathy and clin-
ical outcomes for diabetic patients. Acad Med. 2011; 86(3):359–64. https://doi.org/10.1097/ACM.
0b013e3182086fe1 PMID: 21248604
24. Rakel D, Barrett B, Zhang Z, Hoeft T, Chewning B, Marchand L, et al. Perception of empathy in the ther-
apeutic encounter: effects on the common cold. Patient Educ Couns. 2011; 85(3):390–7. https://doi.
org/10.1016/j.pec.2011.01.009 PMID: 21300514
25. Del Canale S, Louis DZ, Maio V, Wang X, Rossi G, Hojat M, et al. The relationship between physician
empathy and disease complications: an empirical study of primary care physicians and their diabetic
patients in Parma, Italy. Acad Med. 2012; 87(9):1243–9. https://doi.org/10.1097/ACM.
0b013e3182628fbf PMID: 22836852
26. Brazeau CM, Schroeder R, Rovi S, Boyd L. Relationships between medical student burnout, empathy,
and professionalism climate. Acad Med. 2010; 85(10 Suppl):S33–6. https://doi.org/10.1097/ACM.
0b013e3181ed4c47 PMID: 20881699
27. Gleichgerrcht E, Decety J. Empathy in clinical practice: how individual dispositions, gender, and experi-
ence moderate empathic concern, burnout, and emotional distress in physicians. PLoS One. 2013; 8
(4):e61526. https://doi.org/10.1371/journal.pone.0061526 PMID: 23620760
28. Preusche I, Lamm C. Reflections on empathy in medical education: What can we learn from social neu-
rosciences? Adv Health Sci Educ Theory Pract. 2016; 21(1):235–49. https://doi.org/10.1007/s10459-
015-9581-5 PMID: 25597025
29. Wayne DB, Cohen E, Makoul G, McGaghie WC. The impact of judge selection on standard setting for a
patient survey of physician communication skills. Acad Med. 2008; 83(10 Suppl):S17–20. https://doi.
org/10.1097/ACM.0b013e318183e7bd PMID: 18820492
30. Schmitz FM, Schnabel KP, Stricker D, Fischer MR, Guttormsen S. Learning communication from erro-
neous video-based examples: A double-blind randomised controlled trial. Patient Educ Couns. 2017;
100(6):1203–12. https://doi.org/10.1016/j.pec.2017.01.016 PMID: 28179074
PLOS ONE | https://doi.org/10.1371/journal.pone.0229895 March 9, 2020
12 / 12
PLOS ONE |
10.1371_journal.pone.0226644 | RESEARCH ARTICLE
Epidemiology of cardiovascular diseases
related admissions in a referral hospital in the
South West region of Cameroon: A cross-
sectional study in sub-Saharan Africa
Clovis NkokeID
Cyrille Nkouonlack1, Anastase Dzudie2,3,5
1,2*, Ahmadou Musa Jingi2,3, Christelle Makoge3, Denis Teuwafeu1,4,
1 Buea Regional Hospital, Buea, Cameroon, 2 Clinical Research Education, Networking and Consultancy,
Douala, Cameroon, 3 Faculty of Medicine and Biomedical Sciences, University of Yaounde, Yaounde,
Cameroon, 4 Faculty of Health Sciences, University of Buea, Buea, Cameroon, 5 Douala General Hospital,
Douala, Cameroon
* cnkoke@yahoo.com
Abstract
Background
Sub-Saharan Africa (SSA) is experiencing an epidemic of cardiovascular diseases (CVD)
as a result of a rapid epidemiological transition. Little is known about the admission for CVD
and outcome in rural and semi-urban settings in Cameroon in this era of epidemiological
transition. The aim of this study was to determine the frequency and the pattern of CVD
admissions in the South West region of Cameroon.
Methods
This retrospective descriptive study included all adult patients admitted for CVD in the medi-
cal unit of the Buea Regional Hospital between Jan 2016 and December 2017.
Results
Out of the 3140 patients admitted, 499(15.9%) had CVD. There were 304(60.9%) females.
The mean age was 58.7±16.2 years. There was no age difference between men and
women (59.7 years vs 58.1years, p = 0.29). The most commonly affected age group was
those aged 50–59 years (22%). Heart failure (38.5%), stroke (33.3%) and uncontrolled
hypertension (22.4%) were the most prevalent CVDs. The length of hospital stay ranged
from 1 to 37 days with a median length of hospital stay of 7 days. In-hospital case fatality
was 78(15.8%). Mortality was higher in women compared to men (9% vs 7%, p = 0.43). The
case fatality for stroke was higher compared to case fatality for heart failure (21.7% vs
16.7%, p = 0.23).
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: Nkoke C, Jingi AM, Makoge C, Teuwafeu
D, Nkouonlack C, Dzudie A (2019) Epidemiology of
cardiovascular diseases related admissions in a
referral hospital in the South West region of
Cameroon: A cross-sectional study in sub-Saharan
Africa. PLoS ONE 14(12): e0226644. https://doi.
org/10.1371/journal.pone.0226644
Editor: Chiara Lazzeri, Azienda Ospedaliero
Universitaria Careggi, ITALY
Received: July 23, 2019
Accepted: December 1, 2019
Published: December 19, 2019
Copyright: © 2019 Nkoke et al. This is an open
access article distributed under the terms of the
Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: All relevant data are
within the paper and its Supporting Information
files.
Funding: The author(s) received no specific
funding for this work
Competing interests: The authors have declared
that no competing interests exist.
PLOS ONE | https://doi.org/10.1371/journal.pone.0226644 December 19, 2019
1 / 8
Epidemiology of cardiovascular disease in Cameroon
Conclusion
CVDs are a common cause of hospital admission in this semi-urban setting, dominated by
heart failure. Women were disproportionately affected and it was associated with high mortality.
Prevention, early detection and management of risk factors for cardiovascular disease are
imperative given the growing burden of CVD in SSA to reduce CVD morbidity and mortality.
Introduction
Sub-Saharan Africa (SSA) is facing a double burden of infectious and non-communicable dis-
eases including cardiovascular disease (CVD). Cardiovascular diseases (CVD) are a major
public health problem worldwide. It is the leading cause of death worldwide with about 80% of
the deaths occurring in low and middle-income countries including SSA [1].
Sub-Saharan Africa (SSA) is experiencing an epidemic of CVD as a result of the rapid epi-
demiological transition with urbanization and adoption of western lifestyles [2, 3]. As a result
of this epidemiological transition, the incidence of CVD in SSA has been on a rise in the last
few years [4]. This places an additional burden on the health care systems which are already
overwhelmed by infectious diseases and limited resources. In 2015, about 1.2 million people
died from CVD in Africa [5] and the CVD burden of SSA is projected to double by 2030, pre-
dominantly driven by increased rates of hypertension, smoking, and obesity [6]. It is postu-
lated that CVD is set to overcome infectious diseases in developing countries to become the
most common causes of death [7, 8]. A study in Ghana, on the trend of CVD admission,
showed an increase in the percentage of CVD admissions from 4.6% to 8.2%, representing a
78% increase over a decade [9].
There is a paucity of data regarding the burden of CVD admissions in Cameroon in this era
of epidemiological transition which limits the formulation of data-driven national policies.
Also there are no morbidity and mortality registries for CVD in Cameroon. The goal of this
study was to determine the frequency and pattern of CVD admission in the Buea Regional
Hospital, South West region of Cameroon. The study could support policy makers to develop
strategies to fight CVD.
Methodology
Study setting and design
This was a retrospective study of patients with a diagnosis of CVD admitted to the medical
wards of the Buea Regional Hospital between January 2016 and December 2017. This is a sec-
ondary level Hospital and serves as one of the two main referral centers in the region, with a
bed capacity of about 111 beds, and a catchment population of about 200,000 inhabitants. The
Hospital which is a public hospital also serves as one of the teaching hospitals of the University
of Buea. Buea is a semi-urban setting, the headquarters of the South West region of Cameroon.
The main economic activity in the region is agriculture. The hospital receives referrals from
other health facilities in the region.
Data collection
The data was collected from the ward admission and discharge register. Details obtained
included age, sex, a final diagnosis of CVD, duration of hospitalization and outcome. All cases
were diagnosed by a physician using clinical features and investigations. Patients with an
unclear diagnosis of cardiovascular disease or incomplete data were excluded.
PLOS ONE | https://doi.org/10.1371/journal.pone.0226644 December 19, 2019
2 / 8
Epidemiology of cardiovascular disease in Cameroon
Ethical statement
The study was approved by the administrative authorities of the hospital acting as the local eth-
ics committee. We carried out this work in accordance with the declarations of Helsinki. We
report this work following the preferred standard in reporting observational studies in epide-
miology (STROBE). No informed consent was required since it was a retrospective review the
using admission and discharge register.
Sample size and statistical analysis
A convenient sample of all eligible patients was considered for this study. The data collected
were analyzed using SPSS version 20. Data were expressed as frequency and percentages for
discrete variables, and as means (with standard deviation and confidence interval of the
means) for continuous variables. We have presented the distribution of CVDs according to sex
and age group. We compared the proportions using the Chi-squared test or Fischer exact test
where appropriate. We have compared mean values using the Student t-test. We considered a
p-value <0.05 to be statistically significant for the observed differences or associations.
Results
During the two year study period, a total of 3140 patients were admitted and 499(15.9%) were
admitted for CVD. There were 304(60.9%) females. The mean age was 58.7±16.2 years. There
was no age difference between men and women (59.7 years vs 58.1 years, p = 0.29). The most
commonly affected age group (Fig 1) was those aged 50–59 years (22%). Heart failure (38.5%),
stroke (33.3%), and uncontrolled hypertension (22.4%) were the most common CVDs. The
distribution of CVDs according to sex is shown in Table 1. The proportion of patients admit-
ted for stroke was significantly higher in men, while uncontrolled hypertension was signifi-
cantly more common in females. The distribution of CVDs according to the age group is
shown in Table 2. Stroke and heart failure admissions increased with increasing age. Uncon-
trolled hypertension was more common between 40 and 60 years. Poor outcome (death) and
mean duration of hospitalization are shown in Table 3. The highest burden of death was seen
Fig 1. Age distribution of patients.
https://doi.org/10.1371/journal.pone.0226644.g001
PLOS ONE | https://doi.org/10.1371/journal.pone.0226644 December 19, 2019
3 / 8
Epidemiology of cardiovascular disease in Cameroon
Table 1. Distribution of CVD by sex.
Cardiovascular disease
Total(N = 499)
Sex
Male(n = 195)
Female(n = 304)
Stroke
Heart failure
Uncontrolled hypertension
Acute myocardial infarction
Venous thromboembolism
Pericardial disease
Arrhythmias
https://doi.org/10.1371/journal.pone.0226644.t001
166(33.3)
192(38.5)
112(22.4)
4(0.8)
23(4.5)
3(0.6)
5(1)
78(40)
76(39)
32(16.4)
3(1.5)
8(4.1)
1(0.5)
2(1)
88(29)
116(38.2)
80(26.3)
1(0.3)
15(4.9)
2(0.7)
3(1)
p value
0.011
0.858
0.0098
0.135
0.677
0.782
1
in those with heart failure and stroke. The length of hospital stay ranged from 1 to 37 days with
a median length of hospital stay of 7 days. In-hospital case fatality was 78(15.8%). Mortality
was higher in women compared to men (9% vs 7%, p = 0.43). The case-fatality for stroke was
higher compared to case-fatality for heart failure (21.7% vs 16.7%, p = 0.23).
Discussion
The aim of this study was to report on the epidemiology of CVD admissions in a semi-urban
setting in Cameroon. We found that CVD accounted for 15.9% of admissions in the medical
ward. Women were significantly more affected than men; the most common CVDs were heart
failure, stroke and uncontrolled hypertension. In-hospital case-fatality was 15.8%.
The epidemic of CVD associated with lifestyle and epidemiological transitions occurring all
over the world is increasingly being recognized. Cardiovascular disease accounts for one-third
of global deaths and it is a leading contributor to the global burden of disease. This is the first
study describing the epidemiology of CVD admissions in the South West region of Cameroon,
a semi- urban setting in this era of epidemiological transition. The incidence of CVD admis-
sion in this study which was about 16% was similar to the admission for CVD reported by
Osuji et al in Nigeria [10]. This was however higher than the admission for CVD reported in
the Northern part of Cameroon (9.9%) [11]. It was also higher than that reported by previous
studies that showed that CVD accounted for approximately 7%-9% of admissions among
adults in medical wards in Africa [12]. But more recent reports have shown prevalence of up
to 20.1% in Nigeria [13]. This increasing prevalence of CVD admission in SSA can be attrib-
uted to the deteriorating cardiovascular risk factors profile of populations in SSA as a result of
epidemiological transition with a rising prevalence of hypertension, diabetes, obesity and over-
weight [4,6]. A study in Ghana demonstrated an increasing percentage of CVD admissions
from 4.6% to 8.2%, representing a 78% increase over a decade [9]. In Cameroon, stroke
Table 2. Distribution of CVD by age distribution.
Cardiovascular Disease
Age range (Years)
Stroke
Heart failure
Uncontrolled hypertension
Acute myocardial infarction
Venous thromboembolism
Pericardial disease
Arrhythmias
<30
1(0.6)
5(2.6)
11(9.8)
0(0)
1(4.3)
1(33.3)
0(0)
30–39
11(6.6)
13(6.8)
20(17.9)
0(0)
5(21.7)
0(0)
0(0)
40–49
27(16.3)
15(7.8)
26(23.2)
0(0)
5(21.7)
0(0)
0(0)
50–59
36(21.7)
44(22.9)
26(23.2)
1(25)
6(26.1)
0(0)
0(0)
60–69
34(20.5)
38(19.8)
15(13.4)
2(50)
4(17.4)
2(66.7)
2(40)
� 70
57(34.3)
77(40.1)
14(12.5)
1(25)
2(8.7)
0(0)
3(60)
Total
166(100)
192(100)
112(100)
4(100)
23(100)
3(100)
5(100)
https://doi.org/10.1371/journal.pone.0226644.t002
PLOS ONE | https://doi.org/10.1371/journal.pone.0226644 December 19, 2019
4 / 8
Epidemiology of cardiovascular disease in Cameroon
Table 3. Poor outcome and mean hospital stay.
Died in Hospital
Duration in Hospital
Cardiovascular disease
Frequency
Stroke
Heart failure
Uncontrolled hypertension
Acute myocardial infarction
Venous thromboembolism
Pericardial disease
Arrhythmias
https://doi.org/10.1371/journal.pone.0226644.t003
36/166
32/192
8/112
0/4
2/23
2/3
1/5
%
21.7
16.7
7.1
0
8.2
66.7
20
Mean (SD)
95% CI of Mean
8 (5.1)
7.9 (5.3)
5.2 (3.7)
5 (2)
10.9 (6.2)
13.3 (9.3)
5.6 (7.8)
7.3–8.8
7.2–8.7
4.5–5.9
1.8–8.2
8.2–13.5
-9.6–36.4
-4.1–15.3
admission in a major referral hospital increased from 2.5% in 1999–2000 to 13.1% in 2011–
2012 [14].
The incidence of CVD increases with increasing age. The mean age in our study which was
58.7 years was close to that reported by Osuji et al in Nigeria [15]. It was however higher than
that reported by other authors in other African countries [16, 17]. The majority of patients
admitted for CVD were between 50–69 years, an observation that was similar to that reported
in Nigeria where the majority of patients admitted for CVD were aged 51–74 years [13]. The
greater incidence of CVD in patients over 50 years of age is consistent with the overall trend of
CVD which increases with increasing age [18]. But studies have shown that CVD occurs at a
younger age in low and middle-income countries including SSA, affecting the economically
active proportion of the population [19]. This will affect the economic productivity of these
countries. The distribution of cardiovascular disease admissions by sex in Africa has been
inconsistent. Overall, more women (60%) were admitted for CVD compared to men in our
study. This is in contrast with reports from Nigeria where the proportion of men (60%) admit-
ted for CVD was higher [15]. In Ghana, there was a nearly equal distribution of CVD admission
by sex [9]. The risk of CVD in women increases after menopause. This may explain the higher
prevalence of CVD admission in women in this study [20–22]. This higher prevalence of CVD
admission in women in this study may also be attributed to differences in access to healthcare.
In our study, heart failure constituted the first cause of CVD admission with 38.5%. This
was a similar finding in Ghana where it represented 80% of CVD admissions [9]. Stroke was
the second cause of CVD admission in our study. On the contrary, stroke was the first CVD
necessitating admission in medical wards in some studies from Nigeria [15, 23]. Heart failure
has emerged as the most common primary diagnosis for patients admitted to the hospital with
suspected cardiac disease in SSA [24].
Uncontrolled hypertension was the third cause of CVD admission in our study. Hyperten-
sion is by far the largest contributor to the CVD burden worldwide. A recent nationwide sur-
vey in the Cameroonian adult population reported a prevalence of hypertension of 29.2% [25].
Hypertension was the single most common etiological risk factor reported in studies on heart
failure and stroke in Cameroon [14, 26]. This high prevalence of hypertension is associated
with low awareness and poor control [25, 27]. Awareness of CVD and its risk factors by the
general population is low. In a population-based survey in Cameroon, Aminde et al reported
that more than half of the participants had a poor knowledge on CVD and its risk factors [28].
Also, only one quarter of the participants could correctly identify different types of CVD. In
sub-Saharan Africa, the prevention, detection, management and control of hypertension
should now be regarded as a high priority.
The in-hospital case fatality for CVD in our study was 15.8%. This was lower than that
reported in Nigeria where the authors reported a mortality of 32%. But similar to study,
PLOS ONE | https://doi.org/10.1371/journal.pone.0226644 December 19, 2019
5 / 8
Epidemiology of cardiovascular disease in Cameroon
mortality was highest in patients admitted for stroke [15]. According to the WHO, over three
quarters of CVD death occur in low and middle income countries [1].
Limitations
This study may underestimate the occurrence of cardiovascular diseases in the population.
Patients with severe disease might have died at home or before reaching the hospital. These
patients would not be reported in the hospitalization registries, thus would not be captured in
this study. However, this study provides useful information on the epidemiology of cardiovas-
cular diseases in this region, in an era of epidemiological transition.
Conclusion
Cardiovascular disease accounts for a significant proportion of medical admission. There was
a female predisposition to stroke and uncontrolled hypertension related admissions. Preven-
tion, early detection and management of risk conditions should now be regarded as high a pri-
ority given the growing burden.
Supporting information
S1 Database.
(XLSX)
Acknowledgments
We thank all the staff of the medical unit of the Buea Regional Hospital, Buea
Author Contributions
Conceptualization: Ahmadou Musa Jingi, Christelle Makoge, Denis Teuwafeu, Cyrille
Nkouonlack, Anastase Dzudie.
Data curation: Clovis Nkoke, Ahmadou Musa Jingi, Christelle Makoge.
Formal analysis: Clovis Nkoke, Ahmadou Musa Jingi.
Methodology: Anastase Dzudie.
Supervision: Anastase Dzudie.
Writing – original draft: Clovis Nkoke, Ahmadou Musa Jingi, Christelle Makoge, Denis Teu-
wafeu, Cyrille Nkouonlack, Anastase Dzudie.
Writing – review & editing: Clovis Nkoke, Ahmadou Musa Jingi, Christelle Makoge, Denis
Teuwafeu, Cyrille Nkouonlack, Anastase Dzudie.
References
1. World Health Organization Noncommunicable Diseases. [(accessed on 1 June 2018)]; Available online:
http://www.who.int/news-room/fact-sheets/detail/noncommunicable-diseases
2.
Fezeu L., Minkoulou E., Balkau B., Kengne A.-P., Awah P., Unwin N., et al. Association between socio-
economic status and adiposity in urban Cameroon. Int. J. Epidemiol. 2005; 35:105–111. https://doi.org/
10.1093/ije/dyi214 PMID: 16258058
3. Mayosi B.M., Flisher A.J., Lalloo U.G., Sitas F., Tollman S.M., Bradshaw D. The burden of non-commu-
nicable diseases in South Africa. Lancet. 2009; 374:934–947. https://doi.org/10.1016/S0140-6736(09)
61087-4 PMID: 19709736
PLOS ONE | https://doi.org/10.1371/journal.pone.0226644 December 19, 2019
6 / 8
Epidemiology of cardiovascular disease in Cameroon
4. Hamid Suzan, Groot Wim & Pavlova Milena (2019) Trends in cardiovascular diseases and associated
risks in sub-Saharan Africa: a review of the evidence for Ghana, Nigeria, South Africa, Sudan and Tan-
zania, The Aging Male, https://doi.org/10.1080/13685538.2019.1582621 PMID: 30879380
5. WHO. Global Health Estimates 2015: Death by Cause, Age, Sex, by Country and by Region, 2000–
2015. Geneva, World Health Organization. 2016
6. Kariuki JK, Stuart-Shor EM, Leveille SG, et al. Methodological challenges in estimating trends and bur-
den of cardiovascular disease in sub-Saharan Africa. Cardiol Res Pract. 2015; 2015:921021. https://
doi.org/10.1155/2015/921021 PMID: 26697260
7. Yach D, Hawkes C, Gould CL, Hofman KJ. The global burden of chronic diseases: overcoming impedi-
ments to prevention and control. JAMA. 2004 Jun 2; 291(21):2616–22 https://doi.org/10.1001/jama.
291.21.2616 PMID: 15173153
8. Mathers CD, Loncar D. Projections of global mortality and burden of disease from 2002 to 2030.PLoS
Med. 2006 Nov; 3(11):e442 https://doi.org/10.1371/journal.pmed.0030442 PMID: 17132052
9. Appiah LT, Sarfo FS, Agyemang C, et al. Current trends in admissions and outcomes of cardiac dis-
eases in Ghana. Clin Cardiol. 2017; 40:783–788. https://doi.org/10.1002/clc.22753 PMID: 28692760
10. Osuji CU, Onwubuya EI, Ahaneku GI, Omejua EG. Pattern of cardiovascular admissions at Nnamdi Azi-
kiwe University Teaching Hospital Nnewi, South East Nigeria. Pan Afr Med J. 2014; 17:116. https://doi.
org/10.11604/pamj.2014.17.116.1837 PMID: 25120860
11. Pancha Mbouemboue Olivier Ngoufack Jacques Olivier, Koona Koona Adonis Maha Falmata, Kingue
Samuel. Epidemiological aspects of cardiovascular diseases in Ngaounde´ re´ Regional Hospital. Health
Sci. Dis: Vol 16 (1) January–February—March 2015
12. Mocumbi AO. Lack of focus on cardiovascular disease in sub-Saharan Africa. Cardiovasc Diagn Ther.
2012; 2:74–77. https://doi.org/10.3978/j.issn.2223-3652.2012.01.03 PMID: 24282699
13. Ukpabi OJ, Uwanurochi K. Comparing indications for cardiovascular admissions into a Nigerian and an
Israeli Hospital. Ann Afr Med. 2017; 16(2):70–73. https://doi.org/10.4103/aam.aam_30_16 PMID:
28469120
14.
Lekoubou A, Nkoke C, Dzudie A, Kengne AP. Stroke admission and case-fatality in an urban medical
unit in sub-Saharan Africa: a fourteen year trend study from 1999 to 2012. J Neurol Sci. 2015 Mar 15;
350(1–2):24–32. https://doi.org/10.1016/j.jns.2015.02.002 Epub 2015 Feb 7 PMID: 25684340
15. Charles Ukachukwu Osuji Emmanuel Ikechukwu Onwubuya, Gladys Ifesinachi Ahaneku Emeka God-
win Omejua. Pattern of cardiovascular admissions at Nnamdi Azikiwe University Teaching Hospital
Nnewi, South East Nigeria.The Pan African Medical Journal. 2014; 17:116. https://doi.org/10.11604/
pamj.2014.17.116.183
16. Ansa VO, Ekott JU, Bassey EO. Profile and outcome of cardiovascular admissions at the University of
Uyo Teaching Hospital, Uyo: a five year review. Niger J Clin Pract. 2008; 11(1):22–24. PMID:
18689134
17. Amendezo E, Twagirumukiza M, Sebantunzi O, Kagame A. Inhospital Cardiovascular morbidity and
mortality in the department of internal Medicine at CHU Kigali (Rwanda). Ann Trop Med Public Health.
2008; 1(1): 9–14.
18.
Fleg JL, Aronow WS, Frishman WH. Cardiovascular drug therapy in the elderly: benefits and chal-
lenges. Nat Rev Cardiol. 2011 Jan; 8(1):13–28. https://doi.org/10.1038/nrcardio.2010.162 PMID:
20978470
19. Damasceno A, Mayosi BM, Sani M, Ogah OS, Mondo C, Ojji D. et al. The causes, treatment, and out-
come of acute heart failure in 1006 africans from 9 countries. Arch Intern Med. 2012; 172:1386–1394.
https://doi.org/10.1001/archinternmed.2012.3310 PMID: 22945249
20. Mosca L, Benjamin EJ, Berra K, et al. Effectiveness-based guidelines for the prevention of cardiovascu-
lar disease in women– 2011 update: a guideline from the American Heart Association. Circulation 2011;
123:1243–62. https://doi.org/10.1161/CIR.0b013e31820faaf8 PMID: 21325087
21. Kip KE, Marroquin OC, Kelley DE, et al. Clinical importance of obesity versus the metabolic syndrome
in cardiovascular risk in women: a report from the Women’s Ischemia Syndrome Evaluation (WISE)
study. Circulation 2004; 109:706–13 https://doi.org/10.1161/01.CIR.0000115514.44135.A8 PMID:
14970104
22. Huxley R, Barzi F, Woodward M. Excess risk of fatal coronary heart disease associated with diabetes in
men and women: meta-analysis of 37 prospective cohort studies. BMJ 2006; 332:73–8. https://doi.org/
10.1136/bmj.38678.389583.7C PMID: 16371403
23. Okunola OO, Akintunde AA, Akinwusi PO. Some emerging issues in medical admission pattern in the
tropics. J Dent Med Med Sci. 2011; 1(11): 005–008.
PLOS ONE | https://doi.org/10.1371/journal.pone.0226644 December 19, 2019
7 / 8
Epidemiology of cardiovascular disease in Cameroon
24. Damasceno A, Cotter G, Dzudie A, Sliwa K, Mayosi BM. Heart failure in sub-saharan Africa: time for
action. J Am Coll Cardiol 2007: 50: 1688–1693. https://doi.org/10.1016/j.jacc.2007.07.030 PMID:
17950152
25. Dzudie A, Djomou A, Ba H, Njume E, Ndom MS, Mfekeu LK, et al.; Cameroon Cardiac Society and
MMM Cameroon Investigator Group. MMM17-Cameroon, analysis and opportunities-Sub-Saharan
Africa. Eur Heart J Suppl. 2019 Apr; 21(Suppl D):D31–D33. https://doi.org/10.1093/eurheartj/suz081
Epub 2019 Apr 24 PMID: 31043871
26. Nkoke C, Jingi AM, Aminde LN, Teuwafeu D, Nkouonlack C, Noubiap JJ, et al. Heart failure in a semi-
urban setting in Cameroon: clinical characteristics, etiologies, treatment and outcome. J Xiangya Med
2019; 4:11
27. Menanga A, Edie S, Nkoke C, et al. Factors associated with blood pressure control amongst adults with
hypertension in Yaounde, Cameroon: a cross-sectional study. Cardiovasc Diagn Ther. 2016; 6(5):439–
445. https://doi.org/10.21037/cdt.2016.04.03 PMID: 27747167
28. Aminde LN, Takah N, Ngwasiri C, et al. Population awareness of cardiovascular disease and its risk fac-
tors in Buea, Cameroon. BMC Public Health. 2017; 17(1)
PLOS ONE | https://doi.org/10.1371/journal.pone.0226644 December 19, 2019
8 / 8
|
10.1371_journal.pone.0234905 | RESEARCH ARTICLE
Top predator reveals the stability of prey
community in the western subarctic Pacific
Dongming LinID
1,2,3,4, Xinjun Chen1,2,3,4,5*
1 College of Marine Sciences, Shanghai Ocean University, Shanghai, China, 2 Key Laboratory of
Sustainable Exploitation of Oceanic Fishery Resources, Ministry of Education, Shanghai, China, 3 National
Distant-water Fisheries Engineering Research Center, Shanghai, China, 4 Key Laboratory of Oceanic
Fisheries Exploration, Ministry of Agriculture and Rural Affairs, Shanghai, China, 5 Laboratory for Marine
Fisheries Science and Food Production Processes, Qingdao National Laboratory for Marine Science and
Technology, Qingdao, China
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
* xjchen@shou.edu.cn
Abstract
OPEN ACCESS
Citation: Lin D, Chen X (2020) Top predator
reveals the stability of prey community in the
western subarctic Pacific. PLoS ONE 15(6):
e0234905. https://doi.org/10.1371/journal.
pone.0234905
Editor: Andrea Belgrano, Swedish University of
Agricultural Sciences and Swedish Institute for the
Marine Environment, University of Gothenburg,
SWEDEN
Received: October 24, 2019
Accepted: June 4, 2020
Published: June 19, 2020
Peer Review History: PLOS recognizes the
benefits of transparency in the peer review
process; therefore, we enable the publication of
all of the content of peer review and author
responses alongside final, published articles. The
editorial history of this article is available here:
https://doi.org/10.1371/journal.pone.0234905
Copyright: © 2020 Lin, Chen. This is an open
access article distributed under the terms of the
Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: All relevant data are
within the paper and its Supporting Information
files.
The stability of the ecosystems depends on the dynamics of the prey community, but
changes in the composition and abundance of prey species are poorly understood, espe-
cially in open ocean ecosystems. We used neon flying squid Ommastrephes bartramii, an
active top predator, as a biological sampler to investigate the dynamics of the prey commu-
nity in the southwestern part of the Western Subarctic Gyre in the northwestern Pacific
Ocean. Squid were collected monthly from July to November 2016. There were no signifi-
cant differences among months in stable isotopes (δ13C and δ15N) in the digestive gland, a
fast turnover organ reflecting recent dietary information. Similar findings were obtained from
analyses of isotopic niche width and fatty acid profiles. The potential influence of the envi-
ronment (monthly mean sea surface temperature, SST, and chlorophyll-a, Chl-a) on the
prey community was examined with SST and Chl-a both varying significantly among sam-
pling months. We found little evidence for significant effects of SST and Chl-a on the isotopic
values, nor on the fatty acid profiles except for 20:4n6 and 24:1n9. These lines of evidence
indicate that the prey community in the southwestern part of the gyre remains stable, with lit-
tle evidence for systematic changes at the community level. This study provides a novel
understanding of the dynamics of the prey community and highlights the use of top preda-
tors to study the trophic dynamics of an oceanic system where a long-term scientific survey
is unavailable.
Introduction
Concern about ecosystem functioning [1–3] highlights the need for a better understanding of
how the composition and abundance of species in natural communities respond to environ-
mental change. For instance, predatory animals are susceptible to reduction or extirpation of
available prey due to environmental processes, which undermines the stability of ecosystems
and the services they provide [3, 4]. Stability is central to ecosystem functioning, which
includes the ecological processes controlling the fluxes of energy, nutrients and organic matter
PLOS ONE | https://doi.org/10.1371/journal.pone.0234905 June 19, 2020
1 / 17
PLOS ONE Funding: This work was supported by National Key
Research and Development Project
(2019YFD0901404), National Natural Science
Foundation of China (41876144, 41876141),
Natural Science Foundation of Shanghai
(16ZR1415400) and Shanghai Science and
Technology Innovation Program (19DZ1207502).
The funders had no role in study design, data
collection and analysis, decision to publish, or
preparation of the manuscript.
Competing interests: The authors have declared
that no competing interests exist.
Top predator reveals the stability of prey community
[5–7]. Therefore, insight into the prey community and its stability is important for a compre-
hensive understanding of how an ecosystem responds to ongoing environmental change [8].
In the northwest Pacific Ocean, large areas are highly productive and support large popula-
tions of pelagic predators including squids [9–11]. Previous studies in this region have identi-
fied that oceanographic productivity is significantly driven by spatial-temporal variation of the
anticyclonic and cyclonic gyres [12], which greatly influences the abundance of higher trophic
level predators [9, 13]. However, the status of the prey community that supports higher trophic
level predators is poorly studied, which limits our understanding of the functioning of the
overall northwest Pacific ecosystem. The Western Subarctic Gyre, a cyclonic gyre in the north-
west Pacific Ocean, is one region with limited scientific monitoring and hence understanding
of the dynamics of prey communities, even though this information is needed for assessing
ecosystem functioning.
Squids grow rapidly, have short lifespans, and semelparous reproduction [14, 15]. They
impose considerable predation pressure on low- and mid-trophic level species [16] due to
their voracious and active feeding [17–19], and simultaneously support the productivity of
other predators [20]. They consequently play a key role in ecosystem functioning [20, 21].
Squid are highly adapted to the environment to exploit a diverse range of prey and habitat
resources [22, 23]. They occupy medium to top trophic positions in many marine food webs
and their trophic niche width differs among species and ecosystems [24, 25]. These character-
istics reflect not only their flexible feeding strategy [23, 26], but also provide information on
the trophic structure of the system in which they are found [27, 28]. Increasingly, squid have
been highlighted as indicators to examine major changes in trophic structure and ecosystem
functioning [20, 26, 29].
Many naturally occurring biochemical tracers such as stable isotopes and fatty acids have
increased the ability to quantify and characterize complex food webs and community dynam-
ics [24, 30, 31]. These techniques can assess a predator’s dietary history over a range of tempo-
ral scales, reflecting “you are what you eat” [32, 33]. Biochemical tracers are considered to be a
complementary or even alternative and cost-effective tool to stomach content analysis for
examining major changes in trophic structure and ecosystem productivity [26, 34]. For exam-
ple, Pethybridge et al. [26] reported that the comparison of fatty acid profiles of Todarodes
filippovae with those of its potential prey taxa revealed temporal dietary shifts related to site-
specific oceanography and ecosystem structure in continental slope waters in the Southern
Ocean. Stable isotope ratios of nitrogen (δ15N) and carbon (δ13C) for higher trophic organisms
match of those of their prey [32, 34]: δ15N values are enriched by about 3‰ per trophic level,
while δ13C values change little among trophic levels in marine food webs. It is possible to esti-
mate the trophic width of species, populations and ecosystems by analyzing δ15N and δ13C
data together [34–36]. In relation to fatty acids, marine heterotrophs are subject to biochemical
limitations in biosynthesis and modification of carboxylic acids, and generally assimilate die-
tary fatty acids with little or no modification [37]. Many individual fatty acid tracers (e.g.,
20:4n6, 20:5n3, 22:6n3) have been used to study trophic ecology and have revealed the overlap-
ping influences of temperature, habitat, trophic guild and phylogeny (see Meyer et al. [38]).
Thus, by selecting an appropriate predator, stable isotopes and fatty acids could allow the esti-
mation of trophic structure and its dynamics at multiple time scales.
We use neon flying squid Ommastrephes bartramii as a biological sampler to investigate the
trophic dynamics of the prey community of the southwestern part of the Western Subarctic
Gyre in the northwest Pacific Ocean. This region is characterized by high productivity that
supports a large population of higher trophic level species including O. bartramii [9, 39]. O.
bartramii is an extremely widely distributed ommastrephid with a worldwide oceanic bi-sub-
tropical distribution, and inhabits the entire water column through the epipelagic, mesopelagic
PLOS ONE | https://doi.org/10.1371/journal.pone.0234905 June 19, 2020
2 / 17
PLOS ONE Top predator reveals the stability of prey community
and upper bathypelagic zones [15]. More importantly, O. bartramii is a high trophic level spe-
cies, with an average δ15N value up to 13.6‰ [17, 19, 40, 41], which occupies a similar trophic
position as other top predators such as albatrosses (mean δ15N, 12.0‰ for Diomedea immut-
abilis; 14.4‰ for Diomedea nigripes) [40], and sharks (Prionace glauca, mean δ15N 12.1‰)
[42]. O. bartramii is an opportunistic generalist that preys on a wide variety of species, includ-
ing crustaceans, fishes and cephalopods [15, 17, 18, 43]. The diet of O. bartramii varies spatial-
temporally given the associated prey community, e.g., it may feed on transitional-water species
during its northward feeding migration [18], migratory mesopelagic species in the epipelagic
zone at night [44], and non-migratory species during the day in the mesopelagic zone [18]. O.
bartramii therefore has the potential to be an ideal trophic indicator of ecosystem functioning
[14], and represents a way of integrating ecological dynamics over a large area and across sev-
eral ecosystems that are difficult to study directly [4]. We analyzed carbon (δ13C) and nitrogen
(δ15N) stable isotope ratios and fatty acids from the digestive gland of O. bartramii—the diges-
tive gland having been shown to provide information on recent diet (10–14 days) of cephalo-
pods [45–48].
We aim to (a) determine the isotopic trophic niche and variation of the prey community of
O. bartramii; and (b) assess the dynamics of the prey community over a relatively long period.
These results will increase our understanding of the systematic changes in the ecological com-
munity in the region, and provide a basis for quantifying community dynamics in response to
environmental change.
Materials and methods
Ethics statement
Specimens were collected as dead squids from the small-scale trawl fishery landings, from July
to November 2016. The specimens were analyzed in the laboratory using methods consistent
with current Chinese national standards, namely Laboratory Animals—General Requirements
for Animal Experiment (GB/T 35823–2018). There was no requirement for ethics approval of
sampling protocols because all the material analyzed in this paper were obtained from com-
mercial fishermen and were already dead.
Study area
The Western Subarctic Gyre is the western cyclonic subgyre in the North Pacific Ocean and is
found in the northern Kuroshio-Oyashio transition zone [49]. It is nutrient rich owing to
upwelling, presumably due to the Oyashio Current in the southwest and Subarctic Current in
the south [49, 50]. It has shallow mixed depth and photic zone [51, 52]. The phytoplankton
biomass is maximal during spring and does not differ significantly during summer, autumn
and winter [53]. The zooplankton community is relatively simple [54], and the biomass assem-
blage is dominated by large interzonal copepods [55]. It is supposed that microzooplankton
and other mesozooplankton taxa replace phytoplankton as the primary food source for domi-
nant mesozooplankton species, which are then preyed on by micronekton and larger zoo-
plankton [54].
Biological data collection
Ommastrephes bartramii were collected monthly from July to November 2016 from commer-
cial fishing operations in the Western Subarctic Gyre (see the sample stations in Fig 1). This
period is considered to be one of active feeding and growth for the winter-spring cohort in the
northwest Pacific Ocean [15]. The specimens were frozen immediately onboard under -30˚C,
PLOS ONE | https://doi.org/10.1371/journal.pone.0234905 June 19, 2020
3 / 17
PLOS ONE Top predator reveals the stability of prey community
Fig 1. Study area showing sample locations and selected bathymetric contour lines in the western subarctic Pacific
with a schematic illustration of western subarctic gyre and the near-surface current. The schematic diagram of
Western Subarctic Gyre and its currents is redrawn from Qiu [49].
https://doi.org/10.1371/journal.pone.0234905.g001
and shipped to the laboratory for further analyses. After defrosting at room temperature, 129
specimens that covered all the sampling months were randomly selected (Table 1). Dorsal
mantle length (ML, 1 mm), body weight (BW, 1 g), and sexual maturation were recorded for
each specimen. Macro-scale maturity stages were assigned following [56], and all specimens
were found to be subadults with developing and maturing gonads.
The whole digestive gland of each selected specimen was dehydrated in a freeze-drying
chamber (Crist Alpha 1-4/LDplus; (Martin Christ Gefriertrocknungsanlagen GmbH, Ger-
many). After dehydration, each digestive gland was ground up in a mortar and pestle, and a
subsample (~1 g) was taken for fatty acid and stable isotope analysis.
Fatty acid analysis
The subsample of each digestive gland was extracted using a 2:1 (v/v) chloroform:methanol
solution [57]. The lipids were used for fatty acid analysis, while the lipid-extracted samples
were lyophilized again for at least 24 hours for stable isotope analysis.
Table 1. Mantle length and body weight of O. bartramii from which samples of digestive gland were taken.
Sampled month
July
August
September
October
November
pooled
N
33
16
20
26
34
129
https://doi.org/10.1371/journal.pone.0234905.t001
Mantle length (ML, mm)
Body weight (BW, g)
Range
194; 283
209; 275
236; 288
219; 345
218; 370
194; 370
mean±SD
225.39±20.82
251.56±20.3
261.4±13.39
278.69±44.47
277.41±33.19
258.67±36.35
Range
163; 795
246; 628
408; 780
312; 1274
302; 1491
163; 1491
mean±SD
333.73±136.58
436.19±122.91
513.75±83.49
667.23±305.07
622.88±232.46
517.78±237.94
PLOS ONE | https://doi.org/10.1371/journal.pone.0234905 June 19, 2020
4 / 17
PLOS ONE Top predator reveals the stability of prey community
The extracted lipids were used for fatty acid determination following the “Determination of
total fat, saturated fat, and unsaturated fat in foods—Hydrolytic extraction-gas chromatogra-
phy” [58] protocol. Fatty acid methyl esters (FAME) were analyzed separately for each sample
using an Agilent 7890B Gas Chromatograph (GC) coupled to a 5977A series Mass Spectrome-
ter Detector (MSD, Agilent Technologies, Inc. USA). The fatty acid 19:0 was used as an inter-
nal standard. The separation was carried out with helium as the carrier gas, and a thermal
gradient programmed from 125˚C to 250˚C, with an auxiliary heater at 280˚C. The total fatty
acids were determined as dry tissue weight (mg/g dry weight), and each fatty acid was
expressed as a percentage of total fatty acids in the sample [37].
Stable isotope analysis
Due to contaminants when lyophilized again, 56 lipid-extracted subsamples of the digestive
gland were not used for stable isotope analysis. Consequently, a total of 73 lipid-extracted sub-
samples were used and ground separately to a homogeneous fine powder, and a ~1.0 mg sub-
sample for each subsample was used for stable isotope analysis. Stable isotope ratios (δ13C and
δ15N) were measured separately for each sample using an IsoPrime 100 isotope ratio mass
spectrometer (IsoPrime) and vario ISOTOPE cube elemental analyzer (Elementar Analysen-
systeme). The standards for carbon and nitrogen followed Gong et al. [59]: using international
reference materials (USGS 24 [δ13C = −16.049‰], USGS 26 [δ15N = 53.7‰]) and the labora-
tory running standard (protein [δ13C = −26.98‰ and δ15N = 5.96‰]). The measurement
errors were approximately 0.05‰ and 0.06‰ for δ13C and δ15N, respectively.
Statistical analysis
Isotopic values and fatty acids were tested for significant differences between sampling
months. All data were first checked for normality using the one-sample Kolmogorov-Smirnoff
test and for homogeneity of variances using Levene’s test [60]. One-way ANOVA was then
applied to test for differences, and a Tukey’s post hoc test [60] performed to determine where
the difference occurred when significant differences were found. Data were analyzed using a
Kruskall-Wallis nonparametric one-way ANOVA test and a Games-Howell post hoc test [60]
when normality and/or homoscedasticity were rejected.
Stable Isotope Bayesian Ellipses (SIBER) [61] implemented in R [62] were used to analyze the
stable isotope data in the context of isotopic niche between sampling months. We calculated the
prey community isotopic niche widths for each sampling month, including the standard ellipse
area (SEAb), the corrected standard ellipse area (SEAc, an ellipse containing 40% of the data
regardless of sample size) and the overlap as the proportion of the sum of the non-overlapping
ellipse areas (non-overlap SEAc proportion) based on 1,000 replications [61]. The non-overlap
SEAc proportion ranges from 0 (completely distinct ellipses, indicating zero overlap in the isoto-
pic niche widths between groups), to 1 (completely coincidental ellipses, indicating a complete
overlap in the isotopic niche widths between groups) [61]. SEAb was used to test for differences
in the isotopic niche area of the prey community between months, while SEAc and the non-
overlap SEAc proportion were used to compare the niche width of the prey community over
months. These analyses allowed the trophic dynamics of the prey community to be identified.
Non-metric multidimensional scaling (nMDS) and analysis of similarities (ANOSIM) were
applied to assess the similarities of fatty acid profiles between months. These analyses could
allow for the identification of potential differences in the trophic structure of the prey commu-
nity among months, similar to the analyses of dietary data for a specific species [48]. Each fatty
acid was expressed as a percentage of total fatty acids, and a square-root transformation was
used to avoid over-emphasis of extreme values [37]. A Bray–Curtis dissimilarity measure was
PLOS ONE | https://doi.org/10.1371/journal.pone.0234905 June 19, 2020
5 / 17
PLOS ONE Top predator reveals the stability of prey community
employed in the nMDS and ANOSIM [63, 64]. The analyses were performed in the package
‘vegan’ [65] in R.
Generalized additive mixed models (GAMMs) [66] with sampling month as the random
effect were used to access the potential effects of the ambient environment on the dynamics of
the prey community. This involved testing for potential relationships between isotopic values,
fatty acids, and the environmental variables. The dependent variables were δ13C, δ15N, and the
individual fatty acid that was found to differ significantly between sampling months. Key pre-
dictors were monthly mean sea surface temperature (SST, ˚C) and chlorophyll-a concentration
(Chl-a, mg m-3). SST and Chl-a were downloaded from the National Oceanic and Atmo-
spheric Administration (NOAA) ERDDAP (Version 1.82) (https://oceanwatch.pifsc.noaa.gov/
erddap/index.html), at a resolution of 0.05˚× 0.05˚. Prelimilary analysis indicated that both
SST and Chl-a were not correlated with each other (variance inflation factor = 1.54). The effect
of sampling month was taken to be random to account for temporal effects in the data and
unexplained differences among the prey community. We used the function ‘gamm’ with a
Gaussian error distribution in the package ‘gamm4’ [67] in R.
Results
The sampled O. bartramii ranged from 194 to 370 mm ML and from 163 g to 1491 g. Body
size increased significantly with sampling month (ML, F = 16.99, P<0.05; BW, F = 13.35,
P<0.05) (Table 1).
Stable isotopic and niche analyses
δ13C ranged between -22.18‰ and -19.13‰, with an average of -20.49 ± 0.70‰, and δ15N ran-
ged between 5.18‰ and 9.88‰, with an average of 8.42 ± 0.96‰ (Table 2). The highest values
of δ13C and δ15N occurred during October, but no significant differences in δ13C and δ15N
were detected among months (Kruskal-Wallis test, δ13C, χ2 = 0.84, P = 0.93; δ15N, χ2 = 4.99,
P = 0.29). The variation of δ13C between the minimum and the maximum values was similar
among months (range 2.60‰ to 3.02‰). Similar findings were obtained for δ15N, where the
variation ranged from 3.27‰ to 4.49‰ (Table 2).
The Bayesian isotopic niche analyses did not find significant differences in the standard
ellipse area (SEAb) among months (Kruskal-Wallis, χ2 = 6.26, P = 0.18) (Fig 2A). The cor-
rected standard ellipse area (SEAc) ranged from 1.36 to 1.50, and indicated considerable
Table 2. Stable isotopic values and isotopic niche width metrics. Isotopic values were determined from the digestive gland of O. bartramii. The variation between the
minimum and the maximum isotopic values is given in parenthesis under the ranges. SEAc, corrected standard ellipse area; non-overlap SEAc proportion, proportion of
the sum of the non-overlapping ellipse areas.
Sampling month
N
July
August
September
October
November
Pooled
12
15
16
18
12
73
δ13C (‰)
Range
-22.04; -19.41 (2.63)
mean ± SD
-20.57 ± 0.77
δ15N (‰)
range
6.59; 9.88 (3.29)
Isotopic niche width
mean ± SD
8.46 ± 0.88
SEAc
1.46
non-overlap SEAc proportion
-22.09; -19.49 (2.60)
-20.54 ± 0.67
5.18; 9.67 (4.49)
8.38 ± 1.04
1.36
-22.15; -19.37 (2.78)
-20.46 ± 0.74
6.12; 9.37 (3.27)
8.47 ± 0.91
1.43
-22.15; -19.13 (3.02)
-20.37 ± 0.68
5.72; 9.79 (4.07)
8.65 ± 1.05
1.41
-22.18; -19.57 (2.61)
-22.18; -19.13 (3.05)
-20.53 ± 0.71
-20.49 ± 0.70
5.60; 9.37 (3.77)
5.18; 9.88 (4.70)
8.15 ± 0.99
8.42 ± 0.96
1.50
-
0.70
0.76
0.77
0.72
-
https://doi.org/10.1371/journal.pone.0234905.t002
PLOS ONE | https://doi.org/10.1371/journal.pone.0234905 June 19, 2020
6 / 17
PLOS ONE Top predator reveals the stability of prey community
Fig 2. Estimates of isotopic niche area for each sampling month based on δ13C and δ15N. Isotopic values were determined
from the digestive gland of O. bartramii. (A) Bayesian standard ellipse area (SEAb) estimates for each sampling month. The boxes
cover the central 50% of the distributions and bars the 90% intervals, with grey solid circles and horizontal lines indicating the
means and medians; black points indicate whiskers. (B) 40% Bayesian credible intervals for the standard ellipse for each sampling
month.
https://doi.org/10.1371/journal.pone.0234905.g002
overlap in the isotopic data between each consecutive sampling month, confirmed by the high
non-overlap SEAc proportions (Table 2) and the considerable overlap of the ellipses for the
different months (Fig 2B).
Fatty acids and dissimilarity analyses
No significant difference in the total fatty acids was found among months (ANOVA, F = 1.56,
P = 0.19) (Table 3). Similarly, no significant differences in the proportions of the main fatty
acid classes were detected among months, except for the saturated fatty acids (SFA) where July
was significantly higher than the remaining months (Tukey HSD, P<0.05) (Table 3). In terms
of individual fatty acid profiles, 16 fatty acids varied significantly between months (11:0, 13:0,
17:0, 18:0, and 16:1n7 higher in July; 20:1, 22:1n9 and 20:5n3 higher in September, 24:1n9 and
20:2 higher in October, and 14:1n5, 18:2n6t, 18:3n6, 20:3n6, 20:4n6 and 22:2n6 higher in
November; Table 3).
In contrast, nMDS revealed a considerable overlap in the overall fatty acid profiles (Fig 3).
These findings were confirmed by ANOSIM, in which the dissimilarity value (ANOSIM statis-
tical R value) between each two consecutive months ranged from 0.08 to 0.18, with a global
value of 0.15 for all months pooled (Table 4).
Potential relations to sea surface environments
Monthly mean sea surface temperature (SST) at the sampling stations varied significantly
among months (F = 38.19, P<0.01), with the highest temperature in August (mean±SD, 19.83
PLOS ONE | https://doi.org/10.1371/journal.pone.0234905 June 19, 2020
7 / 17
PLOS ONE Table 3. Fatty acids in the digestive gland of O. bartramii sampled in the western subarctic gyre of northwest pacific ocean, from July to November 2016. SFA, satu-
rated fatty acids; MUFA, monounsaturated fatty acids; PUFA, polyunsaturated fatty acids; TFA, total fatty acids. Values are mean ±SD; different superscript letters within
rows represent significant differences (P<0.05) detected using the post hoc test.
Terms
July
August
September
October
November
Top predator reveals the stability of prey community
Fatty acid (%∑TFA)
10:0
11:0
12:0
13:0
14:0
15:0
16:0
17:0
18:0
20:0
21:0
22:0
23:0
24:0
14:1n5
16:1n7
18:1n9t
18:1n9c
20:1
22:1n9
24:1n9
18:2n6t
18:2n6c
18:3n6
18:3n3
20:2
20:3n6
20:4n6
22:2n6
20:5n3
22:6n3
0.06±0.01
0.25±0.09b
0.14±0.03
0.45±0.15b
3.58±0.57
1.16±0.17
16.07±2.31
1.48±0.53b
9.58±1.39b
1.06±0.32
0.58±0.19
1.07±0.19
0.63±0.17
1.13±0.35
0.68±0.24ab
2.93±1.07b
1.00±0.29
15.28±2.92
4.93±1.37a
0.90±0.16a
1.56±0.40a
1.16±0.44ab
1.27±0.14
0.59±0.20ab
0.85±0.17
1.09±0.16a
0.50±0.13a
1.79±0.50a
0.62±0.21ab
6.32±0.93ab
20.95±3.90
Main FA Classes (%∑TFA)
∑SFA
∑MUFA
∑PUFA
37.24±2.81b
27.29±4.13
35.47±5.03
Total fatty acids (mg/g dry weight)
∑TFA
141.56±21.8
https://doi.org/10.1371/journal.pone.0234905.t003
0.06±0.04
0.16±0.05a
0.11±0.04
0.37±0.14ab
3.14±0.53
1.04±0.27
15.51±3.28
1.36±0.44ab
6.07±2.33a
1.02±0.45
0.55±0.21
1.01±0.26
0.59±0.25
1.06±0.53
0.62±0.25ab
2.54±1.38ab
0.92±0.31
16.24±3.82
5.98±2.25ab
1.81±1.34ab
1.77±0.30ab
1.42±0.74ab
1.46±0.35
0.89±0.50ab
1.22±0.34
1.28±0.28ab
0.86±0.43ab
2.17±1.49ab
1.03±0.61abc
5.55±1.35a
21.33±2.19
32.05±4.20a
30.61±5.01
37.33±4.57
0.05±0.01
0.13±0.08a
0.12±0.03
0.25±0.16a
3.29±0.65
1.09±0.16
15.73±3.63
1.08±0.37a
8.31±2.68b
1.04±0.31
0.57±0.31
1.05±0.42
0.62±0.61
1.11±0.34
0.41±0.27a
2.62±1.06ab
0.67±0.35
12.91±6.75
7.41±2.66b
3.47±4.70b
1.98±0.78ab
0.76±0.64a
1.31±0.35
0.46±0.43a
0.98±0.34
1.24±0.28ab
0.52±0.36a
1.71±1.39a
0.52±0.51a
7.04±1.55b
20.64±3.92
34.44±3.50ab
29.76±4.28
35.80±4.50
0.05±0.02
0.19±0.08ab
0.12±0.06
0.43±0.22ab
3.18±0.69
1.04±0.36
15.98±4.57
1.46±0.45ab
5.91±2.14a
1.06±0.53
0.57±0.31
1.05±0.59
0.58±0.34
1.09±0.49
0.72±0.40ab
1.97±0.55a
1.06±0.51
12.97±4.75
5.61±1.71ab
1.86±1.22ab
2.11±0.57b
1.66±1.12b
1.43±0.5
1.06±0.74b
1.29±0.68
1.64±0.44c
1.03±0.64b
2.82±1.81ab
1.25±0.87bc
6.24±1.82ab
21.1±4.68
32.73±4.16a
27.53±4.20
39.74±4.85
0.05±0.04
0.18±0.06a
0.12±0.03
0.43±0.15b
3.15±0.83
1.05±0.25
16.03±4.14
1.47±0.29ab
5.02±1.29a
1.06±0.26
0.58±0.17
1.04±0.33
0.60±0.22
1.08±0.38
0.73±0.26b
2.13±0.40ab
1.06±0.35
15.08±4.19
5.58±1.45ab
1.65±0.34ab
2.02±0.33ab
1.74±0.72b
1.57±0.61
1.12±0.47b
1.33±0.44
1.54±0.36bc
1.08±0.42b
3.08±1.03b
1.31±0.55c
5.26±0.92a
20.58±3.06
31.86±4.48a
29.33±4.09
38.81±5.35
150.48±18.31
147.08±13.66
145.68±17.44
154.5±15.10
±0.22˚C) and the lowest in November (13.44±0.74˚C) (Fig 4A). Monthly sea surface chloro-
phyll-a (Chl-a) also varied significantly among months (F = 10.92, P<0.01), being the lowest
in August (mean±SD, 0.25±0.02 mg m-3) and the highest in October (mean±SD, 0.61±0.07 mg
m-3) (Fig 4B).
There were no significant effects of sea surface temperature on either δ15N or δ13C
(GAMM, δ15N, F = 0.00, P = 0.92; δ13C, F = 0.00, P = 0.67), nor were there significant effects of
PLOS ONE | https://doi.org/10.1371/journal.pone.0234905 June 19, 2020
8 / 17
PLOS ONE Top predator reveals the stability of prey community
Fig 3. Non-metric multidimensional scaling (nMDS) ordination based on the Bray–Curtis dissimilarity metric
showing overlap in the fatty acid composition from different sampling months.
https://doi.org/10.1371/journal.pone.0234905.g003
chlorophyll-a on these isotopic ratios (GAMM, δ15N, F = 0.06, P = 0.22; δ13C, F = 0.00,
P = 0.55) (S1 Table). For those individual fatty acids that differed significantly among months
(Table 2), only 24:1n9 and 20:4n6 were significantly related to the Chl-a (S2 Table), with their
amounts increasing with increasing Chl-a (Fig 5).
Discussion
Our work indicates that neon flying squid, O. bartramii, can provide information about the
prey community in the southwestern part of the Western Subarctic Gyre in the northwest
Pacific Ocean. Ommatrephes bartramii is an appropriate biological sampler for this region
because it feeds throughout the water column [15], the digestive gland provides information
Table 4. Results of analysis of similarities (ANOSIM) for the change in fatty acid compositions between months.
The ANOSIM R value ranges from -1 to 1, where a 1 indicates complete difference between groups, and 0 indicates
high similarity.
Terms
July vs. August
August vs. September
September vs. October
October vs. November
pooled
https://doi.org/10.1371/journal.pone.0234905.t004
R-value
P-value
0.18
0.08
0.10
0.10
0.15
PLOS ONE | https://doi.org/10.1371/journal.pone.0234905 June 19, 2020
0.024
0.057
0.040
0.003
0.001
9 / 17
PLOS ONE Top predator reveals the stability of prey community
Fig 4. Monthly mean sea surface temperature at the sampling stations in western subarctic gyre of northwest pacific ocean from July to November 2016. (A) Sea
surface temperature (SST) and (B) Sea surface chlorophyll-a concentration (Chl-a).
https://doi.org/10.1371/journal.pone.0234905.g004
about recent feeding [45–48], and lipids are stored with little or no modification [45, 68, 69].
To our knowledge, this is the first study that uses squid as a biological sampler, combined with
the use of stable isotopes and fatty acids to explore trophic dynamics for an oceanic ecosystem.
Fig 5. Smooth plots from generalized additive mixed models (GAMMs) showing the significant influence of sea surface
chlorophyll-a concentration (Chl-a) on fatty acids 24:1n9 (A) and 20:4n6 (B). Solid line is the estimate of the smooth, grey
shade represents 95% confidence intervals, and blue circles represent the raw data.
https://doi.org/10.1371/journal.pone.0234905.g005
PLOS ONE | https://doi.org/10.1371/journal.pone.0234905 June 19, 2020
10 / 17
PLOS ONE Top predator reveals the stability of prey community
The similar pattern of isotopic values and the obvious overlap of the fatty acids reveal a stable
trophic structure of the prey species community during the second half of the year in this
region. Specifically, the δ13C and δ15N ratios did not change significantly during this period.
The δ13C is effective for determining foraging habitat [34, 36], and the non-significant differ-
ences could reflect that the prey species share similar habitats in the region, regardless of tempo-
ral fluctuations. On the other hand, the similar pattern of δ15N could indicate that a stable prey
community was available to O. bartramii, given that a stable nitrogen isotope ratio can be effec-
tive at identifying the trophic structure for marine organisms [34, 36]. The Western Subarctic
Gyre is characterized by highly diverse species and abundant biomass, particularly in the mar-
gin areas of the gyre [36, 52, 54], and it is one of the most important feeding grounds for many
higher trophic level species, including O. bartramii [19, 42, 70]. A stable prey community
appears to be vital to support the large populations of these species in this region [11, 42, 71].
Prey availability, to a lesser extent, is responsible for the trophic characteristics of organisms
[2]. As such, the stability of δ15N over months, along with similar niche widths would be
expected if the trophic dynamics of the prey community were stable throughout the five sam-
pling months. The variance in the isotope space is an integrated measure of niche width and
reflects the variation in the diets of consumers [34, 61, 72]. The dynamics of prey species will
tend to result in the highest variance of isotopic niche space in a given ecosystem [73], and ulti-
mately determine the isotopic niche width of the consumers [34, 61]. It is documented that the
stomach contents of O. bartramii effectively reflect the prey availability locally, and for exam-
ple, indicate the endemic species Maurolicus imperatorius in the transitional zone of the Cen-
tral North Pacific in July [18] and the migratory myctophids such as Engraulis japonicas and
Watasenia scintillans in the Kuroshio–Oyashio transition of the western North Pacific during
the Autumn and early Spring [71]. The δ15N values in the mantle muscle of O. bartramii from
the northern part of the Central North Pacific showed moderate variation, due to the prey
items prevalently composed by myctophid Symbolophorus evermanni and squid families Ony-
choteuthidae and Enoploteuthidae [17, 41, 43]. Therefore, it would be not unexpected that the
prey community occupies a similar niche space and does not temporally change over the sam-
pling months in the southwestern part of the gyre.
The obvious overlap of fatty acid compositions further supports the inference of the stability
of the prey community. The multivariate analyses showed that the fatty acid compositions in
different sampling months are very similar, evidenced by the clear overlap of the nMDS scat-
terplots (Fig 3) and low ANOSIM statistic R values for each two consecutive sampling month
period (Table 4). These observations suggest that the prey community is composed of either
single species or many species that consistently occur in the gyre region throughout July to
November, as the fatty acids in higher-order consumers match their diets [47, 48, 74, 75].
There is no reasonable evidence that O. bartramii would prey on a single species, because the
variation of δ15N in the digestive gland is larger than the typical enrichment of the nitrogen
isotope per trophic level (about 3‰ per trophic level [32]). Indeed, O. bartarmii is a well-
known voracious generalist that preys on many food items (e.g., Watanabe et al.[18]), and
exhibits more variation of nitrogen isotopes than a typical trophic level [19, 76].
Nearly half of the individual fatty acids varied significantly between months (Table 3). In
marine environments, many fatty acids have been identified as good tracers of distinct taxa.
For example, 16:1n7 and 20:5n3 are indicators of first-order carnivores, 16:0, 18:0 and 22:6n3
of second-order carnivores, and 20:4n6 and 22:4n6 of top predators [38]. 18:0 is also an impor-
tant tracer of herbivores, and 22:4n6 of planktivores [38]. Accordingly, the lowest values of
16:1n7 in October and 20:5n3 in November may imply that the first-order carnivores were at
lower abundance during these months. By contrast, the top predators in the prey community
could be much more abundant in November, as suggested by the high value for 20:4n6.
PLOS ONE | https://doi.org/10.1371/journal.pone.0234905 June 19, 2020
11 / 17
PLOS ONE Top predator reveals the stability of prey community
Regarding the second-order carnivores, they should be relatively stable from July to November
because no significant differences among months were found for 16:0 and 22:6n3. Coupling
with the obvious overlap and similarity of the fatty acid compositions (Fig 3; Table 4), such
findings highlight that the prey community in the Western Subarctic Gyre is likely to be in
dynamic equilibrium. This is because variation among species is essential for ecosystem stabil-
ity [77]. Large populations of predators including suspension feeders to carnivores seasonally
migrate to the subarctic northwest Pacific [13], so a dynamic equilibrium of the prey commu-
nity would be expected in the southwestern part of the gyre region, and this stability may be
maintained by the high productivity of prey species at the same trophic level, along with sea-
sonal fluctuations.
Monthly sea surface temperature (SST) and chlorophyll-a (Chl-a) differed significantly
among months at the sampling stations. However, our findings indicate that the general pattern
of isotopic values was not correlated with the environmental variables, suggesting that the tro-
phic structure of the prey community is stable regardless of changes of the ambient environ-
ment. Similar results were obtained for the fatty acids that varied significantly from July to
November. There was no evidence that the variation of the individual fatty acids was a function
of SST or Chl-a, with the exception of 24:1n9 and 20:4n6 (Fig 5). These findings seem to contra-
dict the general arguments about marine species responding to oceanic environments. For
example, populations or species may differ in their life-history traits (e.g. growth rate) and sub-
sequent biomass due to changes in water temperature and/or primary productivity (indication
through Chl-a) [78]. However, it is noteworthy that life-history traits and subsequent ecology
for individual species may depend on community composition and demography [6, 79]. First,
conditions become more favorable for some species and less favorable for others [1, 80], thereby
influencing a species’ ecological relevance and ultimately altering the prey species available for
top predators [27, 28]. Second, life-history traits and the diversity expressed within species are
evolutionarily flexible, such that shifts in life-history strategies, such as staggered age structure,
may reduce the risk that an entire cohort will encounter unfavorable environmental conditions
[81]. Such flexibility may enable the community to be more resilient to environmental variation.
Finally, the physiological tolerance of a species to ambient environmental conditions might
increase with its ontogeny [2, 14, 23]. This ultimately contributes to the dynamic equilibrium of
a trophic community. Cumulatively, the stable trophic community in the southwestern part of
the gyre may have evolved to have high resilience to the regional environment, possibly through
high productivity or shifts in life-history strategies among species.
The zooplankton-based food web in the western subarctic Pacific [54, 55] may be another
reason for the stability of the trophic community. Large interzonal copepods predominate the
zooplankton assemblage [55], and many mesozooplankton including copepods, euphausiids
and salps are spatially patchy, creating local zones of high prey availability for predators [54].
However, more detailed information on population connectivity, interaction webs, and struc-
ture-forming species is necessary to specifically examine the stability of the community. Addi-
tionally, although our survey covered a relatively long time period from July to November,
information about the prey community in other seasons is lacking. Further work is needed to
address the status of the prey community in other seasons, even though the life history patterns
of some important copepods in the gyre may be independent of variable environments [55].
Conclusions
We demonstrate that stable isotopes and fatty acid composition data from the digestive gland
of Ommatrephes bartramii, an opportunistic top predator, varies little from July to November
in the southwestern part of the Western Subarctic Gyre of the northwest Pacific Ocean. These
PLOS ONE | https://doi.org/10.1371/journal.pone.0234905 June 19, 2020
12 / 17
PLOS ONE Top predator reveals the stability of prey community
findings imply a stable prey community in the gyre region. The prey community may be resil-
ient to fluctuations in the environment due to high productivity within trophic levels and shifts
in life-history strategy with ontogeny. Although trophic analyses at the taxonomic level are still
necessary to evaluate the dynamics of prey communities, our work enhances understanding of
trophic dynamics in this region, and highlights the use of top predators as biological samplers
to better understand trophic dynamics. Voracious and active top predators, combined with
stable isotopes and fatty acid techniques can provide trophic information at multiple time
scales, allowing an assessment of trophic dynamics. This methodology should be generally
applicable to an oceanic system that is poorly sampled.
Supporting information
S1 Table. GAMM results for stable isotopes (δ15N and δ13C) for Ommastrephes bartramii
modeled in relation to monthly mean sea surface temperature (SST) and chlorophyll a
concentration (Chl-a) in the western subarctic gyre of northwest pacific ocean. AIC, Akaike
information criterion; BIC, Bayesian information criterion; DF, degree of freedom; logLik.,
maximum log-likelihood ratio; edf, estimated degrees of freedom; Ref.df, reference degree of
freedom (prior to deductions); R-sq.(adj), adjusted R-squared; Std.Dev., standard deviation;
Std.Error, standard error.
(DOCX)
S2 Table. GAMM results for fatty acids of Ommastrephes bartramii modeled in relation to
monthly mean sea surface temperature (SST) and chlorophyll a concentration (Chl-a) in
the western subarctic gyre of northwest pacific ocean. The fatty acids used for GAMM are
those differed significantly between sampling months (details see Table 1). Akaike information
criterion; BIC, Bayesian information criterion; DF, degree of freedom; logLik., maximum log-
likelihood ratio; edf, estimated degrees of freedom; Ref.df, reference degree of freedom (prior
to deductions); R-sq.(adj), adjusted R-squared; Std.Dev., standard deviation; Std.Error, stan-
dard error.
(DOCX)
Acknowledgments
This is a contribution of the Distant Squid Fisheries Sci-Tech Group (SHOU). We thank the
staff members of the Key Laboratory of Sustainable Exploitation of Oceanic Fisheries
Resources, Ministry of Education, Shanghai Ocean University for assistance in the laboratory.
We are grateful to technicians Shaoqin Wang and Chunxia Gao for the determination of the
fatty acids. We thank Kai Zhu, Fei Han, Sipeng Xuan, Yanran Wei, Zimo Chen for the biologi-
cal data collected, and Dr. Stuart Corney for his insightful comments. We also appreciate Dr.
Andre Punt for his cordial help with English edits and insightful comments.
Author Contributions
Conceptualization: Dongming Lin, Xinjun Chen.
Data curation: Dongming Lin.
Formal analysis: Dongming Lin.
Funding acquisition: Dongming Lin, Xinjun Chen.
Investigation: Dongming Lin.
Methodology: Dongming Lin, Xinjun Chen.
PLOS ONE | https://doi.org/10.1371/journal.pone.0234905 June 19, 2020
13 / 17
PLOS ONE Top predator reveals the stability of prey community
Project administration: Xinjun Chen.
Resources: Dongming Lin.
Software: Dongming Lin.
Supervision: Dongming Lin, Xinjun Chen.
Visualization: Dongming Lin.
Writing – original draft: Dongming Lin, Xinjun Chen.
Writing – review & editing: Dongming Lin, Xinjun Chen.
References
1. Griffith GP, Strutton PG, Semmens JM. Climate change alters stability and species potential interac-
tions in a large marine ecosystem. Glob Chang Biol. 2018; 24:e90–e100. https://doi.org/10.1111/gcb.
13891 PMID: 28869695
2. Sa´ nchez-Herna´ ndez J, Amundsen P-A. Ecosystem type shapes trophic position and omnivory in fishes.
Fish Fish. 2018; 19:1003–1015. https://doi.org/10.1111/faf.12308
3. Dee LE, Cowles J, Isbell F, Pau S, Gaines SD, Reich PB. When Do Ecosystem Services Depend on
Rare Species? Trends Ecol Evol. 2019; 34(8):746–758. https://doi.org/10.1016/j.tree.2019.03.010
PMID: 31104954
4. Bell G, Fortier-Dubois E´ . Trophic dynamics of a simple model ecosystem. Proc Royal Soc B. 2017; 284
(1862):20171463. https://doi.org/10.1098/rspb.2017.1463 PMID: 28904142
5. Cardinale BJ, Duffy JE, Gonzalez A, Hooper DU, Perrings C, Venail P, et al. Biodiversity loss and its
impact on humanity. Nature. 2012; 486:59–67. https://doi.org/10.1038/nature11148 PMID: 22678280
6. Magurran AE, Deacon AE, Moyes F, Shimadzu H, Dornelas M, Phillip DAT, et al. Divergent biodiversity
change within ecosystems. Proc Natl Acad Sci. 2018; 115(8):1843–1847. https://doi.org/10.1073/pnas.
1712594115 PMID: 29440416
7. Ke´ fi S, Domı´nguez-Garcı´a V, Donohue I, Fontaine C, The´bault E, Dakos V. Advancing our understand-
ing of ecological stability. Ecol Lett. 2019; 22(9):1349–1356. https://doi.org/10.1111/ele.13340 PMID:
31286641
8.
Ives AR, Carpenter SR. Stability and Diversity of Ecosystems. Science. 2007; 317(5834):58–62. https://
doi.org/10.1126/science.1133258 PMID: 17615333
9. Okuda T, Kiyota M, Yonezaki S, Murakami C, Kato Y, Sakai M, et al. Delineating the boundary and
structure of higher trophic level assemblages in the western North Pacific Ocean. Deep-Sea Research
II. 2017; 140:45–54. https://doi.org/10.1016/j.dsr2.2016.11.013.
10. Ohshimo S, Hiraoka Y, Sato T, Nakatsuka S. Feeding habits of bigeye tuna (Thunnus obesus) in the
North Pacific from 2011 to 2013. Mar Freshwater Res. 2018; 69(4):585–606. https://doi.org/10.1071/
MF17058.
11. Ohshimo S, Madigan DJ, Kodama T, Tanaka H, Komoto K, Suyama S, et al. Isoscapes reveal patterns
of δ13C and δ15N of pelagic forage fish and squid in the Northwest Pacific Ocean. Prog Oceanogr.
2019; 175:124–138. https://doi.org/10.1016/j.pocean.2019.04.003.
12. Polovina JJ, Howell EA, Abecassis M. Ocean’s least productive waters are expanding. Geophys Res
Lett. 2008; 35(3):L03618. https://doi.org/10.1029/2007gl031745
13. Yamaguchi A, Matsuno K, Abe Y, Arima D, Imai I. Latitudinal variations in the abundance, biomass, tax-
onomic composition and estimated production of epipelagic mesozooplankton along the 155˚E longi-
tude in the western North Pacific during spring. Prog Oceanogr. 2017; 150:13–19. https://doi.org/10.
1016/j.pocean.2015.04.011.
14. Boyle P, Rodhouse P. Cephalopods: ecology and fisheries. Oxford, UK: Wiley-Blackwell; 2005.
15.
Jereb P, Roper CFE. Cephalopods of the world. An annotated and illustrated catalogue of cephalopod
species known to date. Volume 2. Myopsid and Oegopsid Squids. Rome: FAO; 2010.
16. Rodhouse PG, Nigmatullin CM. Role as Consumers. Philos Trans R Soc Lond B Biol Sci. 1996; 351
(1343):1003–1022. https://doi.org/10.1098/rstb.1996.0090
17. Parry MP. The trophic ecology of two ommastrephid squid species, Ommastrephes bartramii and Sthe-
noteuthis oualaniensis, in the north Pacific sub-tropical gyre. Manoa: University of Hawaii; 2003.
PLOS ONE | https://doi.org/10.1371/journal.pone.0234905 June 19, 2020
14 / 17
PLOS ONE Top predator reveals the stability of prey community
18. Watanabe H, Kubodera T, Ichii T, Kawahara S. Feeding habits of neon flying squid Ommastrephes bar-
tramii in the transitional region of the central North Pacific. Mar Ecol Prog Ser. 2004; 266:173–184.
https://doi.org/10.3354/meps266173
19. Kato Y, Sakai M, Nishikawa H, Igarashi H, Ishikawa Y, Vijai D, et al. Stable isotope analysis of the gla-
dius to investigate migration and trophic patterns of the neon flying squid (Ommastrephes bartramii).
Fish Res. 2016; 173:169–174. https://doi.org/10.1016/j.fishres.2015.09.016.
20. Arkhipkin AI. Squid as nutrient vectors linking Southwest Atlantic marine ecosystems. Deep-Sea
Research II. 2013; 95:7–20. https://doi.org/10.1016/j.dsr2.2012.07.003.
21. Clarke MR. The role of cephalopods in the world’s oceans: an introduction. Philos Trans R Soc Lond B
Biol Sci. 1996; 351:979–983.
22. Pecl GT, Moltschaniwskyj NA, Tracey SR, Jordan AR. Inter-annual plasticity of squid life history and
population structure: ecological and management implications. Oecologia. 2004; 139(4):515–524.
https://doi.org/10.1007/s00442-004-1537-z PMID: 15054657
23. Coll M, Navarro J, Olson RJ, Christensen V. Assessing the trophic position and ecological role of squids
in marine ecosystems by means of food-web models. Deep-Sea Research II. 2013; 95:21–36. https://
doi.org/10.1016/j.dsr2.2012.08.020
24. Guerreiro M, Phillips RA, Cherel Y, Ceia FR, Alvito P, Rosa R, et al. Habitat and trophic ecology of
Southern Ocean cephalopods from stable isotope analyses. Mar Ecol Prog Ser. 2015; 530:119–134.
https://doi.org/10.3354/meps11266
25. Rosas-Luis R, Navarro J, Martı´nez-Baena F, Sa´nchez P. Differences in the trophic niche along the gla-
dius of the squids Illex argentinus and Doryteuthis gahi based on their isotopic values. Reg Stud Mar
Sci. 2017; 11:17–22. https://doi.org/10.1016/j.rsma.2017.02.003
26. Pethybridge HR, Nichols PD, Virtue P, Jackson GD. The foraging ecology of an oceanic squid, Todar-
odes filippovae: The use of signature lipid profiling to monitor ecosystem change. Deep-Sea Research
II. 2013; 95:119–128. https://doi.org/10.1016/j.dsr2.2012.07.025
27. Navarro J, Coll M, Somes CJ, Olson RJ. Trophic niche of squids: Insights from isotopic data in marine
systems worldwide. Deep-Sea Research II. 2013; 95:93–102. https://doi.org/10.1016/j.dsr2.2013.01.
031
28. Espinoza P, Lorrain A, Me´ nard F, Cherel Y, Tremblay-Boyer L, Argu¨elles J, et al. Trophic structure in
the northern Humboldt Current system: new perspectives from stable isotope analysis. Mar Biol. 2017;
164(4):86. https://doi.org/10.1007/s00227-017-3119-8
29.
de la Chesnais T, Fulton EA, Tracey SR, Pecl GT. The ecological role of cephalopods and their repre-
sentation in ecosystem models. Rev Fish Biol Fisheries. 2019; 29:313–334. https://doi.org/10.1007/
s11160-019-09554-2
30. Cherel Y, Hobson KA. Stable isotopes, beaks and predators: a new tool to study the trophic ecology of
31.
cephalopods, including giant and colossal squids. Proc Royal Soc B. 2005; 272(1572):1601–1607.
https://doi.org/10.1098/rspb.2005.3115 PMID: 16048776
Lourenc¸o S, Roura A´ , Ferna´ ndez-Reiriz M-J, Narciso L, Gonza´lez A´ F. Feeding Relationship between
Octopus vulgaris (Cuvier, 1797) Early Life-Cycle Stages and Their Prey in the Western Iberian Upwell-
ing System: Correlation of Reciprocal Lipid and Fatty Acid Contents. Front Physiol. 2017; 8(467):1–11.
https://doi.org/10.3389/fphys.2017.00467.
32. Post DM. Using stable isotopes to estimate trophic position: models, methods, and assumptions. Ecol-
ogy. 2002; 83(3):703–718. https://doi.org/10.1890/0012-9658(2002)083[0703:USITET]2.0.CO;2
33. Goetsch C, Conners MG, Budge SM, Mitani Y, Walker WA, Bromaghin JF, et al. Energy-Rich Mesope-
lagic Fishes Revealed as a Critical Prey Resource for a Deep-Diving Predator Using Quantitative Fatty
Acid Signature Analysis. Front Mar Sci. 2018; 5:430. https://doi.org/10.3389/fmars.2018.00430
34.
Layman CA, Araujo MS, Boucek R, Hammerschlag-Peyer CM, Harrison E, Jud ZR, et al. Applying sta-
ble isotopes to examine food-web structure: an overview of analytical tools. Biol Rev. 2012; 87(3):545–
562. https://doi.org/10.1111/j.1469-185X.2011.00208.x PMID: 22051097
35. Kuhnen VV, Romero GQ, Linhares AX, Vizentin-Bugoni J, Porto EAC, Setz EZF. Diet overlap and spa-
tial segregation between two neotropical marsupials revealed by multiple analytical approaches. PLOS
ONE. 2017; 12(7):e0181188. https://doi.org/10.1371/journal.pone.0181188 PMID: 28704561
36. Sogawa S, Sugisaki H, Tadokoro K, Ono T, Sato E, Shimode S, et al. Feeding habits of six species of
euphausiids (Decapoda: Euphausiacea) in the northwestern Pacific Ocean determined by carbon and
nitrogen stable isotope ratios. J Crustacean Biol. 2017; 37(1):29–36. https://doi.org/10.1093/jcbiol/
ruw014
37.
Iverson SJ. Tracing aquatic food webs using fatty acids: from qualitative indicators to quantitative deter-
mination. In: Kainz M, Brett MT, Arts MT, editors. Lipids in Aquatic Ecosystems. New York, NY:
Springer New York; 2009. p. 281–308.
PLOS ONE | https://doi.org/10.1371/journal.pone.0234905 June 19, 2020
15 / 17
PLOS ONE Top predator reveals the stability of prey community
38. Meyer L, Pethybridge H, Nichols PD, Beckmann C, Huveneers C. Abiotic and biotic drivers of fatty acid
tracers in ecology: A global analysis of chondrichthyan profiles. Funct Ecol. 2019; 33: 1243–1255.
https://doi.org/10.1111/1365-2435.13328
39. Chen X, Chen Y, Tian S, Liu B, Qian W. An assessment of the west winter–spring cohort of neon flying
squid (Ommastrephes bartramii) in the Northwest Pacific Ocean. Fish Res. 2008; 92(2):221–230.
https://doi.org/10.1016/j.fishres.2008.01.011
40. Gould P, Ostrom P, Walker W. Trophic relationships of albatrosses associated with squid and large-
mesh drift-net fisheries in the North Pacific Ocean. Can J Zool. 1997; 75(4):549–562. https://doi.org/10.
1139/z97-068
41. Parry M. Trophic variation with length in two ommastrephid squids, Ommastrephes bartramii and Sthe-
noteuthis oualaniensis. Mar Biol. 2008; 153(3):249–256. https://doi.org/10.1007/s00227-007-0800-3
42.
Fujinami Y, Nakatsuka S, Ohshimo S. Feeding Habits of the Blue Shark (Prionace glauca) in the North-
western Pacific Based on Stomach Contents and Stable Isotope Ratios. Pac Sci. 2018; 72(1):21–39.
https://doi.org/10.2984/72.1.2
43. Parry M. Feeding behavior of two ommastrephid squids Ommastrephes bartramii and Sthenoteuthis
oualaniensis off Hawaii. Mar Ecol Prog Ser. 2006; 318:229–235. https://doi.org/10.3354/meps318229
44.
Lipiński MR, Linkowski TB. Food of the squid Ommastrephes bartramii (Lesueur, 1821) from the South-
West Atlantic Ocean. S Afr J Mar Sci. 1988; 6(1):43–46. https://doi.org/10.2989/025776188784480573
45. Phillips KL, Jackson GD, Nichols PD. Predation on myctophids by the squid Moroteuthis ingens around
Macquarie and Heard Islands: stomach contents and fatty acid analyses. Mar Ecol Prog Ser. 2001;
215:179–189. https://doi.org/10.3354/meps215179
46. Phillips KL, Nichols PD, Jackson GD. Size-related dietary changes observed in the squid Moroteuthis
ingens at the Falkland Islands: stomach contents and fatty-acid analyses. Polar Biol. 2003; 26(7):474–
485. https://doi.org/10.1007/s00300-003-0509-9
47. Stowasser G, Pierce GJ, Moffat CF, Collins MA, Forsythe JW. Experimental study on the effect of diet
on fatty acid and stable isotope profiles of the squid Lolliguncula brevis. J Exp Mar Bio Ecol. 2006; 333
(1):97–114. https://doi.org/10.1016/j.jembe.2005.12.008
48.
Fluckiger M, Jackson G, Nichols P, Virtue P, Daw A, Wotherspoon S. An experimental study of the
effect of diet on the fatty acid profiles of the European Cuttlefish (Sepia officinalis). Mar Biol. 2008; 154
(2):363–372. https://doi.org/10.1007/s00227-008-0932-0
49. Qiu B. Kuroshio and Oyashio Currents. In: Steele JH, editor. Encyclopedia of Ocean Sciences. Oxford:
Academic Press; 2001. p. 1413–1425.
50. Yuan X, Talley LD. The subarctic frontal zone in the North Pacific: Characteristics of frontal structure
from climatological data and synoptic surveys. J Geophys Res: Oceans. 1996; 101(C7):16491–16508.
https://doi.org/10.1029/96jc01249
51. Harrison PJ, Boyd PW, Varela DE, Takeda S, Shiomoto A, Odate T. Comparison of factors controlling
phytoplankton productivity in the NE and NW subarctic Pacific gyres. Prog Oceanogr. 1999; 43(2):205–
234. https://doi.org/10.1016/S0079-6611(99)00015-4
52. Harrison PJ, Whitney FA, Tsuda A, Saito H, Tadokoro K. Nutrient and Plankton Dynamics in the NE
and NW Gyres of the Subarctic Pacific Ocean. J Oceanogr. 2004; 60(1):93–117. https://doi.org/10.
1023/B:JOCE.0000038321.57391.2a
53.
Imai K, Nojiri Y, Tsurushima N, Saino T. Time series of seasonal variation of primary productivity at sta-
tion KNOT (44˚N, 155˚E) in the sub-arctic western North Pacific. Deep-Sea Research II. 2002; 49
(24):5395–5408. https://doi.org/10.1016/S0967-0645(02)00198-4
54. Mackas DL, Tsuda A. Mesozooplankton in the eastern and western subarctic Pacific: community struc-
ture, seasonal life histories, and interannual variability. Prog Oceanogr. 1999; 43(2):335–363. https://
doi.org/10.1016/S0079-6611(99)00012-9
55.
56.
Tsuda A, Saito H, Kasai H. Life histories of Eucalanus bungii and Neocalanus cristatus (Copepoda:
Calanoida) in the western subarctic Pacific Ocean. Fish Oceanogr. 2004; 13(s1):10–20. https://doi.org/
10.1111/j.1365-2419.2004.00315.x
ICES. Report of the Workshop on Sexual Maturity Staging of Cephalopods, 8–11 November 2010,
Livorno, Italy. ICES CM 2010/ACOM:49. 2010.
57. Parrish CC, Abrajano TA, Budge SM, Helleur RJ, Hudson ED, Pulchan K, et al. Lipid and Phenolic Bio-
markers in Marine Ecosystems: Analysis and Applications. In: Wangersky PJ, editor. Marine Chemistry.
Berlin, Heidelberg: Springer Berlin Heidelberg; 2000. p. 193–223.
58. GAQSIQ. Determination of total fat, saturated fat, and unsaturated fat in foods: Hydrolytic extraction-
Gas chromatography. Beijing: Standards Press of China; 2008.
PLOS ONE | https://doi.org/10.1371/journal.pone.0234905 June 19, 2020
16 / 17
PLOS ONE Top predator reveals the stability of prey community
59. Gong Y, Ruiz-Cooley RI, Hunsicker ME, Li Y, Chen X. Sexual dimorphism in feeding apparatus and
niche partitioning in juvenile jumbo squid Dosidicus gigas. Mar Ecol Prog Ser. 2018; 607:99–112.
https://doi.org/10.3354/meps12768.
60.
61.
Zar JH. Biostatistical Analysis, fourth Edition. New Jersey: Prentice Hall; 1999.
Jackson AL, Inger R, Parnell AC, Bearhop S. Comparing isotopic niche widths among and within com-
munities: SIBER–Stable Isotope Bayesian Ellipses in R. J Anim Ecol. 2011; 80(3):595–602. https://doi.
org/10.1111/j.1365-2656.2011.01806.x PMID: 21401589
62. R Core Team. R: A language and environment for statistical computing. 3.5.0 ed. Vienna, Austria: R
Foundation for Statistical Computing; 2018.
63.
Zuur A, Ieno EN, Smith GM. Analyzing Ecological Data. 1 ed. New York: Springer-Verlag; 2007.
64. Anderson MJ. Permutational Multivariate Analysis of Variance (PERMANOVA). In: Balakrishnan N,
Colton T, Everitt B, Piegorsch W, Ruggeri F, Teugels JL, editors. Wiley StatsRef: Statistics Reference
Online: John Wiley & Sons, Ltd.; 2017.
65. Oksanen J, Blanchet FG, Friendly M, Kindt R, Legendre P, McGlinn D, et al. vegan: Community Ecol-
ogy Package. 2019. https://cran.r-project.org, https://github.com/vegandevs/vegan.
66.
Zuur A, Ieno EN, Walker N, Saveliev AA, Smith GM. Mixed Effects Models and Extensions in Ecology
with R. New York: Springer-Verlag; 2009.
67. Wood S, Scheipl F. gamm4: Generalized Additive Mixed Models using ’mgcv’ and ’lme4’. 2017. https://
cran.r-project.org/package=gamm4
68. Ballantyne JS, Hochachka PW, Mommsen TP. Studies on the metabolism of the migratory squid, Loligo
opalescens: enzymes of tissues and heart mitochondria. Mar Biol Lett. 1981; 2:75–85.
69. Mommsen TP, Hochachka PW. Respiratory and enzymatic properties of squid heart mitochondria. Eur J
Biochem. 1981; 120(2):345–350. https://doi.org/10.1111/j.1432-1033.1981.tb05710.x PMID: 7318831
70. Murase H, Kawabata A, Kubota H, Nakagami M, Amakasu K, Abe K, et al. Basin-scale distribution pat-
tern and biomass estimation of Japanese anchovy Engraulis japonicus in the western North Pacific.
Fish Sci. 2012; 78(4):761–773. https://doi.org/10.1007/s12562-012-0508-2
71. Watanabe H, Kubodera T, Ichii T, Sakai M, Moku M, Seitou M. Diet and sexual maturation of the neon
flying squid Ommastrephes bartramii during autumn and spring in the Kuroshio–Oyashio transition
region. J Mar Biol Assoc UK. 2008; 88(2):381–389. https://doi.org/10.1017/S0025315408000635
72. Bearhop S, Adams CE, Waldron S, Fuller RA, Macleod H. Determining trophic niche width: a novel
approach using stable isotope analysis. J Anim Ecol. 2004; 73(5):1007–1012. https://doi.org/10.1111/j.
0021-8790.2004.00861.x
73. Yeakel JD, Bhat U, Elliott Smith EA, Newsome SD. Exploring the Isotopic Niche: Isotopic Variance,
Physiological Incorporation, and the Temporal Dynamics of Foraging. Front Ecol Evol. 2016; 4(1).
https://doi.org/10.3389/fevo.2016.00001
74. Graeve M, Kattner G, Wiencke C, Karsten U. Fatty acid composition of Arctic and Antarctic macroalgae:
indicator of phylogenetic and trophic relationships. Mar Ecol Prog Ser. 2002; 231:67–74. https://doi.org/
10.3354/meps231067
75. Roo J, Estefanell J, Betancor MB, Izquierdo M, Ferna´ ndez-Palacios H, Socorro J. Effects of supple-
mentation of decapod zoea to Artemia basal diet on fatty acid composition and digestive gland histology
in common octopus (Octopus vulgaris) paralarvae. Aquac Res. 2017; 48(2):633–645. https://doi.org/
10.1111/are.12910
76.
Fang Z, Thompson K, Jin Y, Chen X, Chen Y. Preliminary analysis of beak stable isotopes (δ13C and
δ15N) stock variation of neon flying squid, Ommastrephes bartramii, in the North Pacific Ocean. Fish
Res. 2016; 177:153–163. https://doi.org/10.1016/j.fishres.2016.01.011.
77. McCann KS. The diversity–stability debate. Nature. 2000; 405(6783):228–233. https://doi.org/10.1038/
35012234 PMID: 10821283
78. Brown JH, Gillooly JF, Allen AP, Savage VM, West GB. Toward a metabolic theory of ecology. Ecology.
2004; 85(7):1771–1789. https://doi.org/10.1890/03-9000
79.
Zhao Q, Van den Brink PJ, Carpentier C, Wang YXG, Rodrı´guez-Sa´ nchez P, Xu C, et al. Horizontal
and vertical diversity jointly shape food web stability against small and large perturbations. Ecol Lett.
2019; 22:1152–1162. https://doi.org/10.1111/ele.13282 PMID: 31095883
80. Sunday JM, Pecl GT, Frusher S, Hobday AJ, Hill N, Holbrook NJ, et al. Species traits and climate veloc-
ity explain geographic range shifts in an ocean-warming hotspot. Ecol Lett. 2015; 18(9):944–953.
https://doi.org/10.1111/ele.12474 PMID: 26189556
81. Cline TJ, Ohlberger J, Schindler DE. Effects of warming climate and competition in the ocean for life-his-
tories of Pacific salmon. Nature Ecol Evol. 2019; 3(6):935–942. https://doi.org/10.1038/s41559-019-
0901-7 PMID: 31133724
PLOS ONE | https://doi.org/10.1371/journal.pone.0234905 June 19, 2020
17 / 17
PLOS ONE |
10.1371_journal.pone.0236717 | RESEARCH ARTICLE
Genetic diversity, relatedness and inbreeding
of ranched and fragmented Cape buffalo
populations in southern Africa
Deon de JagerID
1*, Cindy Kim Harper2, Paulette Bloomer1
1 Molecular Ecology and Evolution Programme, Department of Biochemistry, Genetics and Microbiology,
Faculty of Natural and Agricultural Sciences, University of Pretoria, Pretoria, Gauteng, South Africa,
2 Veterinary Genetics Laboratory, Faculty of Veterinary Science, University of Pretoria, Pretoria, Gauteng,
South Africa
* dejager4@gmail.com
Abstract
Wildlife ranching, although not considered a conventional conservation system, provides a
sustainable model for wildlife utilization and could be a source of valuable genetic material.
However, increased fragmentation and intensive management may threaten the evolution-
ary potential and conservation value of species. Disease-free Cape buffalo (Syncerus caffer
caffer) in southern Africa exist in populations with a variety of histories and management
practices. We compared the genetic diversity of buffalo in national parks to private ranches
and found that, except for Addo Elephant National Park, genetic diversity was high and sta-
tistically equivalent. We found that relatedness and inbreeding levels were not substantially
different between ranched populations and those in national parks, indicating that breeding
practices likely did not yet influence genetic diversity of buffalo on private ranches in this
study. High genetic differentiation between South African protected areas highlighted their
fragmented nature. Structure analysis revealed private ranches comprised three gene
pools, with origins from Addo Elephant National Park, Kruger National Park and a third,
unsampled gene pool. Based on these results, we recommend the Addo population be sup-
plemented with disease-free Graspan and Mokala buffalo (of Kruger origin). We highlight
the need for more research to characterize the genetic diversity and composition of ranched
wildlife species, in conjunction with wildlife ranchers and conservation authorities, in order to
evaluate the implications for management and conservation of these species across differ-
ent systems.
Introduction
Wildlife ranching, where wild animals are managed in fenced areas, is practiced in various
regions around the world, including North America, Europe and Africa [1–5]. Whether wild-
life ranching is beneficial for conservation is often debated [6–8]. While wildlife ranching pre-
serves habitat that would otherwise be converted for other land-use types [9], it also comes
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: de Jager D, Harper CK, Bloomer P (2020)
Genetic diversity, relatedness and inbreeding of
ranched and fragmented Cape buffalo populations
in southern Africa. PLoS ONE 15(8): e0236717.
https://doi.org/10.1371/journal.pone.0236717
Editor: Elissa Z. Cameron, University of Tasmania,
AUSTRALIA
Received: January 10, 2020
Accepted: July 13, 2020
Published: August 14, 2020
Peer Review History: PLOS recognizes the
benefits of transparency in the peer review
process; therefore, we enable the publication of
all of the content of peer review and author
responses alongside final, published articles. The
editorial history of this article is available here:
https://doi.org/10.1371/journal.pone.0236717
Copyright: © 2020 de Jager et al. This is an open
access article distributed under the terms of the
Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: The genotype file is
available from the Dryad Digital Repository (https://
doi.org/10.5061/dryad.1c2d3).
Funding: DdJ hereby acknowledges the financial
contribution of the National Research Foundation
PLOS ONE | https://doi.org/10.1371/journal.pone.0236717 August 14, 2020
1 / 20
PLOS ONE (https://www.nrf.ac.za/), under the grant number
SFH150630122321. Opinions expressed, and
conclusions arrived at are those of the authors and
are not necessarily to be attributed to the NRF. The
funders had no role in study design, data collection
and analysis, decision to publish, or preparation of
the manuscript.
Competing interests: The authors have declared
that no competing interests exist.
Population genetics of ranched and fragmented Cape buffalo
with the pitfalls of small population sizes and fragmentation of populations due to fencing that
could result in loss of genetic diversity of many species, not just those of economic interest on
the ranch [5, 10]. Furthermore, wildlife ranching often involves intensive breeding of wildlife,
or intentional genetic manipulation [5], for a single, or multiple, traits. This can exacerbate the
fixation or loss of alleles at a rate much higher than would occur through drift or natural selec-
tion [11]. Consequently, the frequency of alleles that are identical by descent increases in the
population, thus increasing the risk of inbreeding and inbreeding depression [12]. Deleterious
alleles that are in linkage disequilibrium with the genes underlying the desired traits will also
be inherited by offspring and increase in frequency in the population [11]. Such intentional
genetic manipulation of wildlife may decrease the evolutionary potential of species and reduce
the conservation value that may be provided by privately-owned wildlife populations [5, 13].
Few studies have been done comparing genetic diversity of species on intensively managed
wildlife ranches to more open or natural systems in state protected areas (reserves and trans-
frontier- or national parks) and extensive game ranches. A comprehensive study of popula-
tions of red deer (Cervus elaphus) on ranches and in a national park in Spain found no
significant differences in genetic diversity [2]. In South Africa, Grobler et al. [14] found signifi-
cantly lower allozyme diversity in impala (Aepyceros melampus) populations in smaller, man-
aged reserves compared to a large population, although these reserves were not intensively
breeding impala. In a study of blue and black wildebeest (Connochaetes taurinus and C. gnou),
Grobler et al. [15] found no significant difference in genetic diversity between ranch and refer-
ence populations of these species, although this was not the aim of the study. Finally, Grobler
et al. [16] used polymorphic allozyme loci and found substantially lower genetic diversity in a
ranch population of Cape buffalo (Syncerus caffer caffer) compared to reference samples from
the Kruger National Park (KNP). However, the ranch population was established with only
two bulls and six cows from Addo Elephant National Park (AENP), which is now known to
have relatively low genetic diversity [16, 17]. This indicates that the low diversity of the ranch
population was likely due to founder effect, given that 25 years (~3.5 buffalo generations) had
passed between establishment of the population and sampling.
At the time of the Grobler et al. [16] study, private ranches in South Africa could only be
stocked with disease-free buffalo from AENP and zoological gardens around the world. This
was due to the prevalence of bovid diseases in other potential source populations, such as KNP
or Hluhluwe-iMfolozi Park (HiP), that are controlled in South Africa through strict veterinary
regulations [18]. In the 1890s and early 1900s rinderpest and foot-and-mouth disease epidem-
ics resulted in the loss of an estimated 95% of the buffalo population in southern Africa [18].
During the 1980s and 1990s, outbreaks of bovine tuberculosis and Corridor disease in the buf-
falo populations of South Africa (excluding AENP) prompted South African National Parks
(SANParks) to start a disease-free breeding programme for KNP buffalo which was carried out
between 1999 and 2007 [19]. The goal was to maintain a disease-free population outside dis-
ease-affected areas that would represent the high genetic diversity of KNP. Two of these dis-
ease-free KNP-derived populations are now maintained in Mokala National Park (MNP) and
the nearby breeding centre Graspan (GNP). Consequently, the wildlife ranching industry had
a new source of disease-free buffalo with presumably high genetic diversity.
Concurrently, the wildlife ranching industry in South Africa experienced rapid growth, not
only due to the popularity of disease-free buffalo, but also due to the breeding, selling and
eventual hunting of rare phenotypes of various antelope species, such as exotic colour varia-
tions, unique horn morphology and increased horn length [5, 20, 21]. Buffalo is one of the
most expensive species and are generally bred to obtain exceptional horn length and “spread”,
with buffalo of East African origin being more expensive and sought after than southern Afri-
can individuals due to their apparent superior phenotype in this regard [5].
PLOS ONE | https://doi.org/10.1371/journal.pone.0236717 August 14, 2020
2 / 20
PLOS ONE Population genetics of ranched and fragmented Cape buffalo
In this study, we aimed to: (i) Determine how successful the SANParks disease-free breed-
ing programme was in maintaining the high genetic diversity of KNP buffalo; (ii) Investigate
whether the genetic diversity of buffalo under intensive management regimes on private
ranches was significantly different compared to their source populations in national parks; (iii)
Determine how the relatedness and inbreeding levels of buffalo on private ranches compared
to those in national parks and; (iv) Characterize the genetic composition of private ranch buf-
falo populations. We discuss how the results relate to the general breeding practices on private
ranches and what the implications are for the conservation and management of Cape buffalo
and other wildlife species that are popular in wildlife ranching. Finally, we make recommenda-
tions regarding the genetic management of Cape buffalo in national parks in southern Africa.
Materials and methods
Samples and ethics statement
This study was performed in collaboration with SANParks (Project code: HARC1227) and the
Ministry of Environment and Tourism, Namibia. Samples from MNP (Northern Cape, South
Africa, GPS: -29.162613, 24.321083), GNP (located near MNP in the Northern Cape) and
AENP (Eastern Cape, South Africa, GPS: -33.483474, 25.750269) were obtained from SAN-
Parks, as blood stored in ethylenediaminetetraacetic acid (EDTA). Samples from Waterberg
Plateau National Park (WPP, Namibia, GPS: -20.352369, 17.337493) and 12 private wildlife
ranches (P001 –P012) were previously obtained by the Veterinary Genetics Laboratory (VGL)
at the University of Pretoria. All samples were collected between 2008 and 2015. Thus, within
one buffalo generation (5–7.5 years), with 99% of the samples collected between 2011 and
2015 (Table 1). Private buffalo owners send samples to the VGL for individual genotyping and
parentage analysis. The number of samples from each locality is shown in Table 1. Permission
Table 1. Sample information and selected summary statistics of Cape buffalo from the 16 localities included in
this study.
Locality
AENP
GNP
MNP
WPP
P001
P002
P003
P004
P005
P006
P007
P008
P009
P010
P011
P012
Collection year
2008, 2013, 2014
2011, 2012, 2013
2011, 2012, 2013, 2014
2014
2012, 2013, 2014
2012, 2014
2014
2009, 2011, 2012, 2013, 2014
2013, 2014
2013
2014
2014
2011, 2013, 2014
2013
2015
2014
�
NC
800
80
400
600
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N/A
N
79
21
35
95
153
308
21
262
57
164
17
54
99
35
22
37
NA
37
68
80
62
89
93
59
99
86
96
64
78
85
70
69
77
PA
0
1
1
0
2
1
0
1
0
2
0
1
1
1
0
0
�Approximate figures. NC: Population census size. N: Sample size. NA: Total number of alleles across the 11
microsatellite markers used in this study. PA: Private (or unique) alleles present in that population. N/A: Not
available.
https://doi.org/10.1371/journal.pone.0236717.t001
PLOS ONE | https://doi.org/10.1371/journal.pone.0236717 August 14, 2020
3 / 20
PLOS ONE Population genetics of ranched and fragmented Cape buffalo
was obtained from the Namibian Department of Environmental Affairs and Tourism and each
private ranch owner to use the genotype data of their samples in this study. Data from the pri-
vate ranches, such as sample and ranch names, have been anonymized for this study. Ranch
locations have also been omitted as this may compromise the anonymity of the participating
ranches. The location of the ranches was not important for the analyses and conclusions of the
study, except to say that all ranches are located in South Africa. This project was approved by
the Animal Ethics Committee of the University of Pretoria (code: ec005-16).
DNA extraction and microsatellite genotyping
DNA was extracted from EDTA-blood samples from AENP, MNP and GNP using the
PrepFiler1 Forensic DNA Extraction Kit (Applied Biosystems, CA, USA) and genotyped at
18 variable microsatellite loci and a sex marker using the following PCR conditions: initial
denaturation at 95˚C for 3 min, 35 cycles of 95˚C for 15 s, 60˚C for 30 s and 72˚C for 30 s, fol-
lowed by a final extension step of 72˚C for 10 min. Fragments were separated on an ABI
3500xl Genetic Analyzer (Applied Biosystems) together with GeneScan™ LIZ™ 500 dye Size
Standard (Applied Biosystems). Allele binning and scoring was performed using STRand ver-
sion 2.4.110 [22] (http://www.vgl.ucdavis.edu/informatics/strand.php). Samples from WPP
and the 12 private ranches were previously genotyped at between 15 and 18 loci. Microsatellite
loci that were not common to all data sets were removed, thus the final data set consisted of
1,459 buffalo genotyped at 11 microsatellite loci and a sex marker (S1 Table). The microsatel-
lite loci used in this study have previously been deemed satisfactory in terms of conforming to
Hardy-Weinberg equilibrium in this exact data set [23].
Study system and population histories
All sampling localities in this study are geographically isolated. The AENP population is a rem-
nant population of buffalo that was fenced in during 1931 when the borders of AENP were set
up and remains free of bovid diseases [17]. The number of buffalo fenced in when the original
borders were set up is not known. However, 130 buffalo were removed in 1981, reducing the
population to 75 individuals. By 1983 the population had increased to approximately 220 indi-
viduals, but was reduced again to 52 buffalo in 1985, potentially due to drought during that
time (Pers. Comm. D. Zimmerman 2015). The census size reported in 1998 was 85 buffalo
[17]. The current census size is approximately 800 buffalo (Table 1). There are no records of
any human-mediated introduction of buffalo to AENP at any point in the history of this popu-
lation (Pers. Comm. D. Zimmerman 2015).
The GNP and MNP populations were established through the SANParks disease-free
breeding programme in 1999. Thereafter, each was supplemented with yearlings, after disease
testing, from the KNP breeding group until 2007 (Pers. Comm. D. Zimmerman 2015). The
breeding group consisted of buffalo (approximately 140 cows and 10 bulls) that originated
mainly from northern KNP and contained in a fenced-off camp in KNP (Pers. Comm. D. Zim-
merman 2015). No introductions have been made to GNP or MNP since 2007 (Pers. Comm.
D. Zimmerman 2015). The current census sizes of GNP and MNP are approximately 80 and
400 buffalo, respectively (Table 1).
A disease-free buffalo population unrelated to those managed by SANParks was established
in Namibia in 1981 in what is now known as Waterberg Plateau National Park (WPP). The
buffalo population in WPP was founded with seven buffalo from AENP in 1981, with addi-
tional introductions of 26 AENP buffalo between 1986 and 1991 (Pers. Comm. M. Lindeque
2017). Two introductions of five and six buffalo from the Willem Pretorius Nature Reserve
(WPNR) in the Free State Province of South Africa occurred in 1985 and 1986, respectively
PLOS ONE | https://doi.org/10.1371/journal.pone.0236717 August 14, 2020
4 / 20
PLOS ONE Population genetics of ranched and fragmented Cape buffalo
(Pers. Comm. M. Lindeque 2017). Although a detailed history of WPNR buffalo population
could not be established, it has been reported that this population was founded with two bulls
and six cows originating from AENP around 1986 [16]. Lastly, four buffalo were introduced in
1986 from a Namibian game dealer who had imported buffalo from a zoo in the Czech Repub-
lic. The origin of the buffalo from the zoo are suspected to be East African (most likely Tanza-
nia or Kenya), but this could not be confirmed. There have been no additional introductions
of buffalo to WPP since 1991 (Pers. Comm. M. Lindeque 2017). The current census size of
WPP is approximately 600 buffalo (Table 1).
The individual histories of the private ranch buffalo populations included in this study are
not known. However, many of the buffalo in the wildlife ranching industry were sourced from
the AENP population (disease-free), as well as from the disease-free KNP breeding programme
(GNP and MNP). The introduction of buffalo from other areas, such as East Africa and Zim-
babwe, onto private ranches is thought to occur. Therefore, the private ranch buffalo popula-
tions in this study represent various combinations of disease-free buffalo from AENP, GNP,
MNP and potentially or parts of southern Africa and/or East Africa.
Population statistics
The number of alleles (NA), allelic richness (AR) and inbreeding coefficient (FIS) of each sam-
pling location were calculated in R v3.1.3 [24], using the divBasic function in the diveRsity
package v1.9.90 [25]. The divBasic function calculates AR by normalising all populations to the
smallest sample size and subsampling 1,000 times, with replacement, and thus provides 95%
confidence intervals. This is considered one of the most precise ways to obtain unbiased esti-
mates of AR when sample sizes are unequal [26]. The number of private alleles (PA) in each
population was determined from the tables of allele frequencies calculated in Genetix v4.05.2
[27]. Observed (HO) and expected heterozygosity (HE) were calculated in Cervus v3.0.7 [28].
Effective population size (Ne) for each sampling locality was estimated using NeEstimator
v2.01 [29]. The linkage disequilibrium method was used at a critical value of 0+ and 95% confi-
dence intervals were calculated by jack-knifing over loci 1,000 times. Ne was also estimated for
GNP-MNP combined and all private ranches combined.
Relatedness and individual inbreeding
Pairwise relatedness was estimated in COANCESTRY v1.0.1.8 [30] using the likelihood esti-
mator, TrioML [31]. This estimator was chosen as it had the smallest variance of the seven esti-
mators available in COANCESTRY (S2 Table) and the estimates produced are between zero
and one, thereby facilitating interpretation. Relatedness was estimated independently for each
sampling locality, using the allele frequencies of that locality, and without accounting for
inbreeding, as most localities showed little or no evidence of inbreeding. For example, while
AENP had low genetic diversity, its population-level FIS was not significantly greater than
zero, and while P006 had an FIS significantly greater than zero, it had high genetic diversity.
Two statistics to describe distributions, the skewness (a measure of asymmetry) and kurtosis (a
measure of how strongly the distribution is tailed), were calculated for the distribution of pair-
wise relatedness values, to assist in more robust interpretation. These were calculated using the
package moments v0.14 (https://cran.r-project.org/web/packages/moments/index.html) in R
v3.5.0 [32]. In order to determine whether there was a disproportionate number of close rela-
tives (second- and first-order relatives) in private ranches compared to national parks, possibly
due to breeding practices, the proportion of pairwise relatedness values greater than or equal
(�) to 0.25 was calculated for each sampling locality. Mean relatedness within sexes per locality
was also calculated to determine whether there was any sex-bias in relatedness distributions.
PLOS ONE | https://doi.org/10.1371/journal.pone.0236717 August 14, 2020
5 / 20
PLOS ONE Population genetics of ranched and fragmented Cape buffalo
The genotype at the amelogenin sex marker was used to group individuals into sexes. Individ-
ual inbreeding coefficients, F, were estimated independently for each sampling locality in
COANCESTRY using the TrioML method, by selecting the “Account for Inbreeding” option.
Population differentiation
To investigate genetic differentiation between sampling localities, the population differentia-
tion estimator, Jost’s D (DJOST) [33], and Weir and Cockerham’s fixation index (FST) [34],
were calculated, for all pairs of localities, in R v3.1.3 using the diffCalc function in the diveRsity
package. The 95% confidence intervals for these statistics were calculated with the same func-
tion, using 999 bootstrap replicates, to determine significant deviation from zero. Correlation
between DJOST and FST was calculated in R v3.5.0 with the cor.test function using the non-
parametric Spearman’s rank correlation, rho, since the data were not normally distributed (as
shown by a Shapiro-Wilk normality test in R). Genetic structure was also investigated using
STRUCTURE v2.3.4 [35] and a discriminant analysis of principal components (DAPC) [36]
using the package adegenet v2.1.1 [37] in R v3.5.0. The STRUCTURE analysis was conducted
on the full data set, as well as a data set where relatives were removed. See S1 Appendix for
details of these analyses.
The R scripts used to create most of the figures in this manuscript and to perform the
DAPC analysis are available at the following URL: https://github.com/DeondeJager/Rscripts-
for-buffalo-microsat-paper. Rstudio v1.2.5033 [38] was used for all R-based analyses.
Results
Genetic diversity of national parks and private ranches
In this study, Graspan (GNP) and Mokala National Park (MNP) were used as “benchmark”
localities in terms of genetic diversity due to the Kruger National Park (KNP) origin of these
buffalo. Addo Elephant National Park (AENP) had the fewest number of alleles, despite having
the seventh largest sample size in the study. Nine of 11 private alleles were present in the pri-
vate ranch samples, with the remaining two private alleles present in GNP and MNP (Table 1).
The allelic richness (AR) analysis showed that AENP had significantly lower genetic diver-
sity than any other population in this study (Fig 1A, S3 Table). WPP and P003 had significantly
lower AR than the mean across all the populations, whereas P001, P002 and P006 had signifi-
cantly higher AR than the mean (Fig 1A). When the mean was calculated without AENP, only
P001 remained with significantly higher AR. Despite its low AR, AENP showed no significant
population-level inbreeding (FIS, Fig 1B). Conversely, P006 showed a low, but significant FIS
value, with an excess of homozygotes, despite having the highest AR. P007 showed an excess of
heterozygotes, with an FIS significantly less than zero. The observed and expected heterozygos-
ity (Fig 1C) values had large standard deviations, but mirrored the inbreeding coefficients,
where a lower observed than expected heterozygosity corresponded to a positive inbreeding
coefficient and vice versa.
Effective population size (Ne) estimates were similar for most localities (range: 3.3–48.8), as
well as for GNP-MNP and all private ranches combined (88.7 and 81.9, respectively) (S1 Fig,
S3 Table). AENP had an Ne estimate of 24.2 (S3 Table), although this was not significantly
lower than GNP-MNP, based on (slightly) overlapping 95% confidence intervals (S1 Fig).
Relatedness and individual inbreeding
The pairwise relatedness (r) values had an L-shaped distribution, with a slight increase in fre-
quency of values around r = 0.5 in all localities, albeit to varying degrees (Fig 2). Mean pairwise
PLOS ONE | https://doi.org/10.1371/journal.pone.0236717 August 14, 2020
6 / 20
PLOS ONE Population genetics of ranched and fragmented Cape buffalo
Fig 1. Genetic diversity indices of the 16 sampling localities analysed in the present study. A Allelic richness, where the mean is shown by the dashed
line and the mean without AENP is shown by the dotted line. B Inbreeding coefficient, where an asterisk (�) indicates significant deviation from zero
(dotted line). C Observed (HO) and expected (HE) heterozygosity. Vertical bars indicate 95% confidence intervals on plots A and B, and standard
deviation on plot C.
https://doi.org/10.1371/journal.pone.0236717.g001
relatedness was generally low, where P003 (0.110) and AENP (0.096) had the highest and GNP
(0.044) and MNP (0.049) had the lowest mean relatedness (Fig 2, S4 Table). The positive, and
relatively high, skewness values confirmed the distributions were right-tailed and highly asym-
metrical (range: 1.75–3.31, S4 Table). A normal distribution has a skewness of zero and kurto-
sis of three. The kurtosis values of GNP and MNP were the highest at 14.64 and 12.56 (S4
Table), meaning their relatedness distributions had the most extreme tails, i.e. their tails were
the “thinnest”. As stated by Bradley [39]: “The coefficients of skewness and kurtosis of an L-
shaped distribution increase rapidly as the long positive tail of the distribution becomes thin-
ner and thinner. . .”. In other words, GNP and MNP had the lowest proportion of close rela-
tives, while those localities with lower kurtosis values, such as P003 (3.49) and AENP (5.14),
had a larger proportion of close relatives.
PLOS ONE | https://doi.org/10.1371/journal.pone.0236717 August 14, 2020
7 / 20
PLOS ONE Population genetics of ranched and fragmented Cape buffalo
Fig 2. Pairwise relatedness distributions per locality. Density histograms were constructed with ggplot2 v3.3.0 [40] in R, with 30 bins for all histograms. The dashed
line on each plot indicates the mean relatedness, with the text inset showing the value of the mean. Note that the y-axes scales are independent between plots to aid in
visualization of the distribution for each locality, due to the large range of sample sizes in this data set.
https://doi.org/10.1371/journal.pone.0236717.g002
Indeed, this was illustrated by the strong negative correlation between kurtosis of the
relatedness distributions and the proportion of pairwise relatedness values � 0.25 (Fig 3). The
proportion of close relatives, as estimated by the proportion of r values � 0.25, was not signifi-
cantly higher in private ranches (mean = 10.0%) compared to national parks (mean = 8.1%)
(p-value = 0.2, one-tailed t-test). However, private ranches generally had a proportion of close
relatives approximately twice the value seen in GNP (4.8%) and MNP (4.2%), while this value
in AENP (14.5%) was approximately three times higher than the latter two (S4 Table). In gen-
eral, males were more related to each other than females, except for P003 and P011, although
this difference was not significant when the mean was calculated across all localities (p-
value = 0.077, one-tailed t-test) (S5 Table).
Individual inbreeding (F) estimates were substantially higher in AENP (mean F = 0.156)
compared to the other localities, except for P006 which had a mean F of 0.112 (Fig 4, S4
Table). The higher levels of F in these two localities was also evident in their distributions in
Fig 4, with larger interquartile ranges and higher 75% quantiles than all other localities. The
remainder of the localities had relatively low inbreeding estimates, with means ranging
between 0.023 and 0.097 (S4 Table).
PLOS ONE | https://doi.org/10.1371/journal.pone.0236717 August 14, 2020
8 / 20
PLOS ONE Population genetics of ranched and fragmented Cape buffalo
Fig 3. Scatterplot showing the correlation between kurtosis and proportion of relatedness values � 0.25. Pearson’s correlation coefficient (R) and the associated p-
value are shown.
https://doi.org/10.1371/journal.pone.0236717.g003
Population differentiation
The majority of the 120 population pairs were significantly differentiated based on the DJOST
and FST analyses (Table 2). These two metrics were significantly correlated, Spearman’s
rho = 0.90, p-value < 2.2e-16. Differentiation estimates were deemed significant if the 95% con-
fidence interval did not encompass zero (S6 Table). Only four pairs were not significantly dif-
ferentiated based on FST, namely GNP-MNP, P003-P006, P003-P008 and P003-P011, while
ten pairs were not significantly differentiated based on DJOST (Table 2). In terms of the
national parks, GNP and MNP were more similar to each other than to any other population
and their differentiation (DJOST) was not significant, while AENP and WPP were also more
similar to each other than to any other populations, but were still significantly differentiated.
Both results were unsurprising given the shared origin of GNP and MNP buffalo and that
AENP provided founders for WPP. The mean FST across all populations was 0.077 and the
mean DJOST was 0.116. Between the four national parks, the mean FST was 0.147 and mean
DJOST was 0.184. The corresponding values among the private buffalo populations only were
0.055 and 0.098. AENP was the most differentiated from the private ranches, with a DJOST of
PLOS ONE | https://doi.org/10.1371/journal.pone.0236717 August 14, 2020
9 / 20
PLOS ONE Population genetics of ranched and fragmented Cape buffalo
Fig 4. Box and whisker plots showing the distribution of individual inbreeding (F) estimates. The lower and upper hinges of the box indicate the 25% and 75%
quantiles, respectively, with the horizontal line in the box showing the median. The lower whisker shows the smallest observation greater than or equal to the lower hinge
minus 1.5 times the interquartile range and the upper whisker shows the largest observation less than or equal to the upper hinge plus 1.5 times the interquartile range.
Points indicate outliers beyond the whiskers. The turquoise squares show the mean.
https://doi.org/10.1371/journal.pone.0236717.g004
0.203 (mean calculated across P001 –P012), followed by WPP (0.161), MNP (0.089) and finally
GNP (0.080).
The STRUCTURE analysis of the full data set revealed K = 2 as the most likely (based on
DeltaK), while the analysis with the relatives removed data set revealed K = 3 as the most likely
number of genetic clusters (based on both the log likelihood of the data and DeltaK) (S2 Fig).
STRUCTURE assumes Hardy-Weinberg equilibrium (HWE), and linkage equilibrium
between loci, within populations [35]. In the full data set, five private ranches did not conform
to HWE, after Bonferroni corrections (S7 Table). However, in the relatives removed data set,
all localities conformed to HWE (S7 Table). In the full data set, some pairs of loci were not in
linkage equilibrium after Bonferroni corrections were performed, although there was no pair
consistently in linkage disequilibrium (S8 Table). However, in the relatives removed data set,
all pairs of loci were in linkage equilibrium in all localities (S8 Table). The relatives removed
data set therefore satisfied the assumptions of STRUCTURE and thus we concluded that K = 3
represents the most likely number of genetic clusters in this data set.
PLOS ONE | https://doi.org/10.1371/journal.pone.0236717 August 14, 2020
10 / 20
PLOS ONE Table 2. Pairwise DJOST (above diagonal) and FST (below diagonal) values for the 16 sampling localities in this study.
Population genetics of ranched and fragmented Cape buffalo
MNP WPP
-
AENP
AENP
-
GNP
MNP
WPP
P001
P002
P003
P004
P005
P006
P007
P008
P009
P010
P011
P012
0.255
0.257
0.053
0.144
0.135
0.129
0.112
0.248
0.106
0.229
0.099
0.247
0.272
0.198
0.265
GNP
0.289
-
0.303
0.003
0.005
-
0.156
0.060
0.072
0.069
0.070
0.038
0.041
0.076
0.059
0.025
0.026
0.029
0.037
0.155
0.074
0.082
0.075
0.078
0.038
0.049
0.080
0.062
0.021
0.023
0.037
0.018
0.038
0.227
0.246
-
0.079
0.074
0.069
0.075
0.159
0.055
0.131
0.052
0.169
0.165
0.111
0.173
P001
0.222
0.112
0.141
0.138
-
0.016
0.054
0.025
0.064
0.028
0.028
0.038
0.078
0.069
0.030
0.082
P002
0.166
0.136
0.173
0.110
0.029
-
0.059
0.029
0.067
0.034
0.041
0.035
0.083
0.077
0.043
0.090
P003
0.066
0.077
0.125
0.049
0.081
0.061
-
0.050
0.085
0.014
0.090
0.022
0.094
0.090
0.042
0.104
P004
0.131
0.143
0.147
0.124
0.054
0.057
0.076
-
0.064
0.025
0.047
0.025
0.076
0.075
0.042
0.071
P005
0.287
0.058
0.055
0.233
0.146
0.148
0.124
0.134
-
0.056
0.075
0.065
0.047
0.046
0.050
0.057
P006
0.137
0.064
0.081
0.093
0.068
0.057
0.012
0.043
0.124
-
0.045
0.014
0.061
0.061
0.024
0.066
P007
0.221
0.113
0.086
0.192
0.055
0.071
0.129
0.058
0.118
0.072
-
0.069
0.076
0.092
0.062
0.094
P008
0.095
0.090
0.111
0.081
0.086
0.066
0.009
0.040
0.120
0.023
0.111
-
0.073
0.068
0.041
0.077
P009
0.307
0.019
0.025
0.274
0.167
0.175
0.130
0.154
0.067
0.119
0.119
0.13
-
0.031
0.060
0.032
P010
0.310
0.045
0.028
0.244
0.138
0.154
0.111
0.173
0.074
0.138
0.184
0.12
0.059
-
0.04
0.03
P011
0.207
0.052
0.079
0.141
0.047
0.074
0.079
0.085
0.088
0.044
0.098
0.07
0.095
0.086
-
0.04
P012
0.286
0.048
0.020
0.255
0.164
0.178
0.167
0.139
0.104
0.137
0.137
0.16
0.032
0.034
0.07
-
Bold, underlined values indicate population pairs that were not significantly differentiated for the relevant statistic, based on the 95% confidence intervals (S6 Table). All
other population pairs were significantly differentiated.
https://doi.org/10.1371/journal.pone.0236717.t002
The individual assignment plots at both K = 2 and K = 3 showed that the private ranches
predominantly constituted buffalo originating from AENP and GNP-MNP, as expected given
the known history of the private ranch populations (S3 Fig). However, at K = 3, some localities
showed a signal from a third, unsampled gene pool. P001 and P002, in particular, showed a
substantial contribution from this third gene pool, while P004 and P011 also had a fairly strong
contribution. In the national parks, WPP showed a not insignificant signal from this third
gene pool, but clustered predominantly with AENP, as expected, while AENP was mostly
homogeneous. GNP and MNP clustered together and showed some variation contributed
from AENP and the third gene pool (S3 Fig).
The discriminant analysis of principal components (DAPC) at K = 3 (the optimum K as
determined above) showed distinct clustering of samples corresponding to those identified in
the STRUCTURE analysis, namely an AENP cluster, GNP-MNP cluster and a third, unidenti-
fied gene pool “Other” (S4 Fig).
Discussion
In this study, we evaluated and compared the genetic diversity, relatedness and inbreeding, as
well as characterized the structure of disease-free Cape buffalo (Syncerus caffer caffer) in pri-
vate ranches and national parks in southern Africa. Disease-free buffalo populations in south-
ern Africa could be important reservoirs of genetic diversity for the species if future disease
outbreaks occur on the scale of those of the 1890s/1900s [19]. The unique population histories
of each sampling locality offered an opportunity to investigate how recent (GNP and MNP)
and older (AENP) population contractions, as well as admixture of different gene pools (WPP
and private ranches) affected the genetic diversity of the populations. Furthermore, given the
more intensive management of buffalo populations on private ranches, in general, as com-
pared to national parks, it was important to determine whether these practices may have
PLOS ONE | https://doi.org/10.1371/journal.pone.0236717 August 14, 2020
11 / 20
PLOS ONE Population genetics of ranched and fragmented Cape buffalo
affected the genetic diversity, and how the relatedness and inbreeding distributions of these
populations compare against more natural or less managed populations.
The high allelic richness, effective population size (Ne), low inbreeding coefficient (FIS) and
low relatedness in GNP and MNP indicated that the disease-free breeding programme
appeared to have been designed in a robust enough manner to prevent a significant founder
effect. The observed heterozygosity of GNP and MNP (0.66 and 0.61, respectively) was slightly
lower than that previously estimated in the north of KNP (0.71–0.75), but overlapped with the
distribution of heterozygosity throughout the entire park (0.62–0.75) [17, 41–44]. Thus, short
of a direct comparative study between GNP-MNP and KNP using the same loci, we can con-
clude that the disease-free breeding programme has likely maintained the high genetic diver-
sity of KNP buffalo in the GNP and MNP populations.
The older population contraction (due to the disease outbreaks in the 1890s/1900s) in
AENP was reflected in the significantly lower allelic richness (AR), the low heterozygosity, low
Ne and high individual inbreeding (F) estimates observed in this population. The observed het-
erozygosity in AENP (0.40) was slightly lower than previously reported (0.48) by O’Ryan et al.
[17], who used seven loci compared to 11 used here. There was no overlap of loci between the
studies. While the observed reduction in heterozygosity of AENP may be due to the different
loci used, it is likely that the genetic diversity of this population decreased in the last two
decades, given the severity of the known population contraction that occurred and the fact
that there has been no gene flow into this population between O’Ryan et al. [17] and the cur-
rent study (Pers. Comm. D. Zimmerman 2015).
The admixed populations in this study (WPP and the private ranches) had genetic diversity
statistically equivalent to that of GNP and MNP, although WPP and P003 had significantly
lower AR than the mean across all populations. However, the significantly higher AR in WPP,
compared to AENP, the population from which the majority of its founders originated, was
most likely a consequence of unique variation contributed by the non-AENP/Czech Republic
zoo/East African buffalo. This highlighted the positive effect that even a few breeding migrants
(whether natural or human-mediated) can have on the genetic diversity of a population, how-
ever the risk of outbreeding depression should also be taken into account when mixing indi-
viduals from isolated populations [49].
Inbreeding coefficients were low for all populations, with only P006 and P007 significantly
deviating from zero, in opposite directions. P006 showed an excess of homozygosity (low, but
significantly positive FIS) and P007 showed an excess of heterozygosity (significantly negative
FIS). The latter result for P007 was most likely a result of the small sample size from this popu-
lation (N = 17) making statistical inferences less robust, as indicated by the large confidence
intervals around the FIS estimate. P006 was an anomaly, because it had high genetic diversity,
low mean relatedness, high kurtosis and a low proportion of relatedness values � 0.25, but still
had a significantly positive FIS value and a relatively high mean individual inbreeding value
(F = 0.112). One potential explanation for these observations follows. On private ranches with
intensive management, stud breeding is often practiced, where a single male is generally the
only breeding bull in a breeding group. This breeding strategy may have been implemented
for several consecutive generations in the P006 population without sufficient turnover of cows
in the breeding group, thus resulting in significant inbreeding. The high genetic diversity and
low relatedness could be a result of recent introductions of newly acquired buffalo that were
sampled and included in this study before they could contribute their genetic variation to the
population, and thus before they could have reduced the inbreeding levels in P006. The fact
that this FIS deviation and high individual F values were not seen in any of the other private
ranches could be due to a faster turnover of breeding bulls and/or cows, or that the samples
included in this study were more randomly sampled from those populations.
PLOS ONE | https://doi.org/10.1371/journal.pone.0236717 August 14, 2020
12 / 20
PLOS ONE Population genetics of ranched and fragmented Cape buffalo
Given the high genetic diversity of the private ranches, the relatively low estimates of Ne
were somewhat surprising. The private ranches had Ne estimates close to those estimated for
AENP and WPP. The Ne in AENP was only 3% of the population census size, while the Ne of
WPP was only 4.2% of the census size. Thus, while the Czech Republic zoo buffalo may have
increased the genetic diversity of WPP, the effect of the population decline that its main source
population (AENP) experienced had not been completely negated. For comparison, the Ne of
GNP-MNP was 18% of their combined population census size, which is comparable to the Ne/
Nc ratios reported for KNP by van Hooft et al. [44] and O’Ryan et al. [17] of 26% and 10–30%,
respectively. The low Ne estimated for the private ranches may reflect the management
approaches on these ranches. However, the Ne estimates in this study should be interpreted
with caution, particularly for the private ranches, since these populations are atypical and do
not necessarily satisfy the assumptions of typical population genetics models.
The L-shaped relatedness distributions indicated that most buffalo from each sampling
locality were unrelated. This was a somewhat surprising result given the observed social struc-
ture that exists in unmanaged buffalo herds, where females mate with one or a few (in large
herds) dominant bulls [45, 46], and the genetic evidence from herds in KNP that supports this
hypothesis [44]. Therefore, the expectation was that there would be a relatively high propor-
tion of close relatives in each locality, and particularly in private ranches, due to the breeding
strategies employed. However, the high kurtosis values and low proportions of close relatives
showed that this was not the case. There was a slight increase in the frequency of first-order
relatives (r = ~0.5) indicating some family structure was present, albeit to a lower extent than
expected.
These observations may be explained by several individual factors, or a combination
thereof. First, the expected ecological and genetic patterns of buffalo herds may not hold true
in the case of private ranches, and even in some smaller natural populations, due to the break-
down of natural demographic processes and social structure. Second, the samples included in
this study may have originated from multiple herds within each locality (unfortunately, sam-
pling information at this scale was not available). This would explain why most buffalo within
localities were unrelated, as well as the lower-than-expected proportion of close relatives in
each locality. The increase in frequency of relatedness values around r = 0.5 could thus be even
more pronounced in individual herds than was observed here. Lastly, the relatively frequent
introduction of buffalo on private ranches (immigrants) and removal of other buffalo (emi-
grants) would result in lower overall relatedness and lower proportions of close relatives than
expected in these populations. Only a small number of breeding migrants per generation are
necessary to affect the genetic diversity and relatedness of a population [17]. This may also
explain the unexpected relatedness results in GNP-MNP, which showed some evidence of
(indirect) gene flow from an unsampled population- likely via their source population, KNP
(see discussion of STRUCTURE results below).
Females were expected to be more related to each other than males, given their lower inter-
herd migration levels (5–20% per generation in females vs 100% in males) [44]. Again, the
results indicated that this was not the case for the localities in this study, where relatedness was
generally higher within males rather than females, although overall this difference was not sig-
nificant. Perhaps the non-conformity of private ranches to this sex-biased relatedness hypothe-
sis is not entirely surprising, given that the ranches do not represent a natural system in the
way that AENP does, for example. Thus, not seeing this sex-biased relatedness pattern in a
more natural population such as AENP was more surprising but could again be explained by
considering that samples may have originated from multiple herds.
Interestingly, samples from P003 may, in fact, represent a single herd, as the relatedness
parameters of this locality most closely matched the expectations discussed above. P003 had
PLOS ONE | https://doi.org/10.1371/journal.pone.0236717 August 14, 2020
13 / 20
PLOS ONE Population genetics of ranched and fragmented Cape buffalo
the highest mean relatedness (0.115), the lowest skewness and kurtosis (thus the most even
relatedness distribution), the highest proportion of close relatives (20.5%) and a substantially
higher mean relatedness between females (0.158) than between males (0.095), while still main-
taining low individual inbreeding levels (mean = 0.083).
While DJOST is generally a more accurate measure of genetic differentiation than FST [33],
we predominantly compare FST here, as most previous buffalo studies did not compute DJOST,
except Smitz et al. [42], and the two metrics were significantly and strongly correlated in this
data set. The genetic differentiation between AENP and GNP/MNP (DJOST = 0.289/0.303, FST
= 0.255/0.257) was, to the best of our knowledge, the highest recorded (in terms of FST)
between any two buffalo populations in southern Africa. It was substantially higher than the
differentiation found between AENP and KNP (FST = 0.131) by O’Ryan et al. [17]. It was also
higher than the differentiation found between Hluhluwe-iMfolozi Park (HiP) and KNP (DJOST
= 0.213, FST = 0.141) by Smitz et al. [42] and between AENP and HiP (FST = 0.227) estimated
by O’Ryan et al. [17].
These levels of genetic differentiation between buffalo populations in South African pro-
tected areas are two to four times higher than the differentiation observed between Cape buf-
falo populations across Africa. AENP and HiP are more differentiated from KNP, than KNP is
from populations in the East African countries of Kenya and Tanzania (FST = 0.078–0.100)
[42, 43]. Incredibly, southern African Cape buffalo (S. c. caffer) populations are more differen-
tiated from each other than KNP buffalo are from forest buffalo (S. c. nanus) (FST = 0.116–
0.155), a different subspecies [43]. This amount of intraspecific genetic differentiation within
the single subspecies of Cape buffalo populations in southern Africa (and particularly South
Africa) is indicative of the highly fragmented nature of South Africa’s protected areas and is
another example of the human-induced population fragmentation and consequent reproduc-
tive isolation experienced by more than 25% of species in the world [47]. Thus, if past subdivi-
sion of South African buffalo populations was minimal, which is likely, as suggested by O’Ryan
et al. [17], then action should be taken to re-establish gene flow [48], either through the estab-
lishment of natural corridors between populations or through translocations (e.g. from the dis-
ease-free GNP-MNP to AENP). Potential concerns about outbreeding depression and/or
genomic swamping of the AENP buffalo population following this proposed genetic supple-
mentation could be alleviated by considering the current support for an estimated short time
since isolation (~100 years) [17, 49] and by conducting regular genetic monitoring of the pop-
ulation. Furthermore, recent studies showed local genes, and alleles related to local adaptation,
were maintained in the recipient population after genetic supplementation, while the benefits
of genetic rescue were still observed [50–52].
The results obtained with the STRUCTURE analyses showed that the removal of close rela-
tives assisted in the more accurate identification of the most likely number of genetic clusters
present in the data set, by satisfying the assumptions of the models implemented. The results
suggested that these approaches were effective in mitigating the biases introduced into
STRUCTURE analyses by the sampling of close relatives [53].
The individual assignment plots at K = 3 illustrated the significant genetic differentiation
between AENP and GNP-MNP. Apparent contributions to each other between these two clus-
ters are most likely a result of shared ancestral polymorphism and not recent gene flow, given
the small proportion of the contributions. At K = 3, a contribution from a third gene pool
(dark green) was prevalent in many individual buffalo in P001, P002, P004 and P007, and in a
few individuals in the WPP population. The DAPC analysis also clustered a large proportion
of individuals into this third gene pool. This third gene pool may have originated from one of
two broadly defined regions: East Africa or the northern parts of southern Africa. We define
East Africa here as the area north-north-east of Tanzania, inclusive. Thus, Tanzania, Kenya,
PLOS ONE | https://doi.org/10.1371/journal.pone.0236717 August 14, 2020
14 / 20
PLOS ONE Population genetics of ranched and fragmented Cape buffalo
Burundi, Rwanda, Uganda and the southern parts of South Sudan, Ethiopia and Somalia. The
support for an East African origin is that WPP had four buffalo potentially of Tanzanian or
Kenyan origin introduced into its population via a Czech Republic zoo. Additionally, numer-
ous private ranches in South Africa breed with East African buffalo, although the definition of
East Africa may differ from ranch to ranch.
The northern parts of southern Africa are here defined as the “northern cluster” identified
by Smitz et al. [42] and consists of northern Botswana, northern Zimbabwe, Angola and central
and northern Mozambique (excluding Gorongosa National Park). Smitz et al. [42] identified
two additional clusters; central (comprising southern Zimbabwe, southern Mozambique and
northern South Africa, which included KNP- the source of GNP-MNP buffalo) and a southern
cluster consisting mainly of buffalo from the isolated Hluhluwe-iMfolozi Park in the east of
South Africa. The authors found evidence of the northern cluster gene pool in KNP, indicating
gene flow between the northern cluster and KNP. It is thus conceivable that the signal of the
third gene pool identified in the present study originated from northern Botswana/Zimbabwe
buffalo and was also present in some of the buffalo that formed part of the disease-free breeding
programme of KNP and therefore is present in both GNP-MNP and the private ranches (which
source their buffalo from GNP-MNP, AENP and elsewhere). Some private ranches may also
have independently introduced buffalo from this northern cluster to their properties, thus giv-
ing a more extensive signature of this third gene pool in the private ranch populations.
Both the above-mentioned scenarios are likely and may be occurring at the same time given
the extensive translocations of buffalo in southern Africa. A more comprehensive data set con-
sisting of Cape buffalo samples from across the subspecies’ range, including private ranches,
may reveal a more detailed genetic history regarding the origin of buffalo on private ranches
and the natural genetic structure that exists across the continent.
In conclusion, we echo O’Ryan et al. [17] in recommending that the AENP population be
augmented with buffalo of KNP origin to increase its genetic diversity- one breeding bull per
generation (every ~7.5 years), to prevent genomic swamping and maintain unique diversity and
local adaptation (if present) in AENP. This is more feasible now than in 1998, given that dis-
ease-free populations of KNP buffalo now exist in GNP and MNP and have high genetic diver-
sity. Likewise, the population in WPP, with moderate genetic diversity and most of its founders
from AENP, would benefit from the introduction of disease-free buffalo from GNP-MNP (at a
similar frequency and rate to AENP) to firstly increase and then maintain genetic diversity. Dis-
ease-free buffalo populations (including private ranches, AENP, GNP-MNP and WPP) repre-
sent an important insurance policy for the species in southern Africa. Confirming that GNP,
MNP and the private ranch populations maintained high genetic diversity was thus an impor-
tant result. However, it is equally important to continually monitor genetic diversity of these
populations, as the impact on genetic diversity of their recent establishment (i.e. founder effect)
and fragmentation has likely not yet manifested. Furthermore, we found that genetic diversity,
relatedness and inbreeding levels did not appear to be affected at the present time by the breed-
ing strategies employed for buffalo on private ranches in this study, except perhaps for P006. It
should be cautioned, however, that the genetic diversity on private ranches is highly dependent
on the management practices on each ranch and the exchange of buffalo between ranches. If
not managed actively and adequately, genetic diversity could be lost due to breeding practices,
small population sizes (genetic drift) and inbreeding. We also showed that private ranch buffalo
were predominantly of AENP and GNP-MNP origin, but that there was substantial contribu-
tion from a third, unsampled gene pool, most likely representing buffalo from East Africa, or
the northern parts of southern Africa.
Private ranches contain a significant proportion of individuals of different wildlife species,
not only in South Africa, but also in other countries such as Namibia, the USA and Spain [2,
PLOS ONE | https://doi.org/10.1371/journal.pone.0236717 August 14, 2020
15 / 20
PLOS ONE Population genetics of ranched and fragmented Cape buffalo
5]. The present study is the first in southern Africa to evaluate the genetic diversity, at a large
scale, of one of these species, the Cape buffalo, on private ranches compared to their source
populations in national parks. Given that genetic diversity is a key component of biodiversity
[54] and private ranches harbour a significant proportion of certain wildlife species, there is a
clear case for similar studies to be conducted for other popular wildlife ranching species. This
will require cooperation between private ranch owners, researchers and conservation authori-
ties. The goal should be to catalogue the genetic diversity contained in private wildlife popula-
tions, as compared to national parks, for the benefit of the species. Furthermore, there appears
to be no unified guidelines for private ranchers to assist in the genetic management of wildlife
on their ranches in southern Africa to ensure sustainable use and long-term survival of the spe-
cies [5]. Additional studies and appropriate guidelines are particularly pertinent given the
increased pressures on extensive areas for ranching and conservation and on wildlife popula-
tions due to human expansion, anthropogenic activities and climate change [55, 56].
Future research on African buffalo could include investigating at a finer scale how specific
ranching approaches (e.g. intensive vs. extensive, or particular breeding protocols) affect the
genetic diversity, relatedness and inbreeding of populations and individual herds. Addition-
ally, researchers could take advantage of the relative affordability of generating genome-wide
data to investigate which loci might be under selection in private ranches and natural popula-
tions, to explore the demographic history of the species to investigate how it may have
responded to historical climate change, to estimate divergence times of different populations,
and to inspect what signatures historical population declines may have left in the genomes of
individuals from different populations.
Supporting information
S1 Appendix. Supplementary methods.
(PDF)
S1 Fig. Estimated effective population size (Ne) of the buffalo population from each
locality. Vertical lines indicate 95% confidence intervals. Numbers inserted for GNP and
GNP-MNP indicate the value of the upper bound of the 95% CI. The dashed line indicates the
lower 95% CI of GNP. Values are also shown in S3 Table. PVT: Private ranches combined.
(PDF)
S2 Fig. Statistical support for K. The first column of graphs [L(K)] show the mean log likeli-
hood of each value of K with its associated standard deviation, while the second column (Del-
taK) shows the most likely value of K as determined by the Evanno method. Rows indicate the
full data set (FDS) and the relatives removed (RR) data set. The graphs were generated using
StructureHarvester and further organized in Inkscape v0.92 (https://inkscape.org/).
(PDF)
S3 Fig. Individual assignment plots of the STRUCTURE analyses at K = 2 and K = 3. A–full
data set, B–relatives removed. The plots were generated using the online version of Clumpak
and further organized in Inkscape v0.92 (https://inkscape.org/).
(PDF)
S4 Fig. Discriminant analysis of principal components (DAPC) of the full data set at K = 3.
AENP Cluster: Addo Elephant National Park cluster, GNP-MNP Cluster: Graspan and Mokala
National Park cluster, “Other” Cluster: Third, unknown origin cluster.
(PDF)
PLOS ONE | https://doi.org/10.1371/journal.pone.0236717 August 14, 2020
16 / 20
PLOS ONE Population genetics of ranched and fragmented Cape buffalo
S1 Table. Summary statistics of microsatellite loci used in this study. Calculated in Cervus
v3.0.7.
(DOCX)
S2 Table. Mean and variance of the relatedness estimators available in COANCESTRY.
TrioML (values in bold) had the lowest variance for each sampling locality and produces posi-
tive relatedness estimates between zero and one (as does DyadML).
(DOCX)
S3 Table. Population summary statistics for each sampling locality.
(DOCX)
S4 Table. Relatedness and individual inbreeding statistics.
(DOCX)
S5 Table. Mean relatedness within sexes.
(DOCX)
S6 Table. Pairwise DJOST and FST values with 95% confidence intervals.
(DOCX)
S7 Table. Hardy-Weinberg Equilibrium (HWE) probability tests of each sampling locality,
with the full data set and relatives removed. Data sets from sampling localities conformed to
HWE after relatives were removed.
(DOCX)
S8 Table. Original and Bonferroni-corrected linkage disequilibrium p-values of all pairs of
loci in all sampling localities. Both the full data set (FDS) and relatives removed (RR) data set
are shown.
(XLSX)
Acknowledgments
We acknowledge the contribution of Amy Clarke, previously of the Veterinary Genetics Labo-
ratory in genotyping many of the samples over numerous years. We would like to thank Dave
Zimmerman of SANParks for providing detailed population histories of the AENP, GNP and
MNP buffalo populations. We also thank Mark Jago (previously of the Namibia Ministry of
Environment and Tourism (MET)) and Malan Lindeque (Namibia MET), for providing a
detailed population history of the WPP buffalo population and the MET for releasing this
information for use in this research, as well as for providing samples. We thank the owners of
the private ranches included in this study for their permission to publish the data in anon-
ymized form.
Author Contributions
Conceptualization: Deon de Jager, Cindy Kim Harper, Paulette Bloomer.
Data curation: Deon de Jager, Cindy Kim Harper.
Formal analysis: Deon de Jager.
Funding acquisition: Deon de Jager, Cindy Kim Harper, Paulette Bloomer.
Investigation: Deon de Jager.
Methodology: Deon de Jager.
PLOS ONE | https://doi.org/10.1371/journal.pone.0236717 August 14, 2020
17 / 20
PLOS ONE Population genetics of ranched and fragmented Cape buffalo
Project administration: Deon de Jager, Cindy Kim Harper, Paulette Bloomer.
Resources: Cindy Kim Harper, Paulette Bloomer.
Supervision: Cindy Kim Harper, Paulette Bloomer.
Validation: Cindy Kim Harper, Paulette Bloomer.
Visualization: Deon de Jager.
Writing – original draft: Deon de Jager.
Writing – review & editing: Deon de Jager, Cindy Kim Harper, Paulette Bloomer.
References
1. Butler MJ, Teaschner AP, Ballard WB, McGee BK. Commentary: Wildlife ranching in North America—
Arguments, issues, and perspectives. Wildl Soc Bull. 2005; 33(1):381–9. https://doi.org/10.2193/0091-
7648(2005)33[381:Cwrina]2.0.Co;2
2. Galarza JA, Sa´ nchez-Ferna´ ndez B, Fandos P, Soriguer R. Intensive management and natural genetic
variation in red deer (Cervus elaphus). J Hered. 2017; 108(5):496–504. https://doi.org/10.1093/jhered/
esx052 PMID: 28863451
3.
4.
Lindsey PA, Barnes J, Nyirenda V, Pumfrett B, Tambling CJ, Taylor WA, et al. The Zambian wildlife
ranching industry: Scale, associated benefits, and limitations affecting its development. PLoS One.
2013; 8(12):e81761. https://doi.org/10.1371/journal.pone.0081761 PMID: 24367493
Lindsey PA, Havemann CP, Lines RM, Price AE, Retief TA, Rhebergen T, et al. Benefits of wildlife-
based land uses on private lands in Namibia and limitations affecting their development. Oryx. 2013; 47
(1):41–53. Epub 2013/01/08. https://doi.org/10.1017/S0030605311001049
5. Russo I-RM, Hoban S, Bloomer P, Kotze´ A, Segelbacher G, Rushworth I, et al. ‘Intentional Genetic
Manipulation’ as a conservation threat. Conserv Genet Resour. 2019; 11(2):237–47. https://doi.org/10.
1007/s12686-018-0983-6
6. Cousins JA, Sadler JP, Evans J. Exploring the role of private wildlife ranching as a conservation tool in
South Africa: stakeholder perspectives. Ecol Soc. 2008; 13(2):43.
7. Nogueira SSC, Nogueira-Filho SLG. Wildlife farming: An alternative to unsustainable hunting and defor-
estation in Neotropical forests? Biodivers Conserv. 2011; 20(7):1385–97. https://doi.org/10.1007/
s10531-011-0047-7
8.
Tensen L. Under what circumstances can wildlife farming benefit species conservation? Global Ecology
and Conservation. 2016; 6:286–98. http://dx.doi.org/10.1016/j.gecco.2016.03.007.
9. Carruthers J. "Wilding the farm or farming the wild"? The evolution of scientific game ranching in South
Africa from the 1960s to the present. Trans R Soc S Afr. 2008; 63(2):160–81.
10.
Lacy RC. Loss of genetic diversity from managed populations: Interacting effects of drift, mutation,
immigration, selection, and population subdivision. Conserv Biol. 1987; 1(2):143–58. https://doi.org/10.
1111/j.1523-1739.1987.tb00023.x
11. Druet T, Pe´ rez-Pardal L, Charlier C, Gautier M. Identification of large selective sweeps associated with
major genes in cattle. Anim Genet. 2013; 44(6):758–62. https://doi.org/10.1111/age.12073 PMID:
23859468
12.
Frankham R, Briscoe DA, Ballou JD. Introduction to conservation genetics. 2nd ed. Cambridge: Cam-
bridge University Press; 2010.
13. Cousins JA, Sadler JP, Evans J. The challenge of regulating private wildlife ranches for conservation in
South Africa. Ecol Soc. 2010; 15(2):28.
14. Grobler JP, Van der Bank FH. Allozyme variation in South African impala populations under different
management regimes. S Afr J Wildl Res. 1994; 24(4):89–94.
15. Grobler P, van Wyk AM, Dalton DL, van Vuuren BJ, Kotze A. Assessing introgressive hybridization
between blue wildebeest (Connochaetes taurinus) and black wildebeest (Connochaetes gnou) from
South Africa. Conserv Genet. 2018; 19(4):981–93. https://doi.org/10.1007/s10592-018-1071-x
WOS:000439164500018.
16. Grobler JP, Van Der Bank FH. Genetic diversity and isolation in African Buffalo (Syncerus caffer). Bio-
chem Syst Ecol. 1996; 24(7):757–61. https://doi.org/10.1016/S0305-1978(96)00078-6.
PLOS ONE | https://doi.org/10.1371/journal.pone.0236717 August 14, 2020
18 / 20
PLOS ONE Population genetics of ranched and fragmented Cape buffalo
17. O’Ryan C, Harley EH, Bruford MW, Beaumont M, Wayne RK, Cherry MI. Microsatellite analysis of
genetic diversity in fragmented South African buffalo populations. Anim Conserv. 1998; 1(2):85–94.
https://doi.org/10.1111/j.1469-1795.1998.tb00015.x
18. Winterbach HEK. Research review: The status and distribution of Cape buffalo Syncerus caffer caffer in
southern Africa. S Afr J Wildl Res. 1998; 28(3):82–8.
19.
20.
21.
22.
23.
Laubscher L, Hoffman L. An overview of disease-free buffalo breeding projects with reference to the dif-
ferent systems used in South Africa. Sustainability. 2012; 4(12):3124–40. https://doi.org/10.3390/
su4113124
Janovsky E. Wildlife industry expected to continue to grow. ABSA report. 2015:1–3.
Taylor WA, Lindsey PA, Davies-Mostert H. An assessment of the economic, social and conservation
value of the wildlife ranching industry and its potential to support the green economy in South Africa.
The Endangered Wildlife Trust. 2016:Johannesburg.
Toonen RJ, Hughes S. Increased throughput for fragment analysis on an ABI Prism® 377 automated
sequencer using a membrane comb and STRand software. Biotechniques. 2001; 31(6):1320–5. PMID:
11768661
de Jager D, Swarts P, Harper C, Bloomer P. Friends and family: A software program for identification of
unrelated individuals from molecular marker data. Mol Ecol Resour. 2017; 17(6):e225–e33. https://doi.
org/10.1111/1755-0998.12691 PMID: 28503747.
24. R Core Team. R: A language and environment for statistical computing Vienna, Austria.: R Foundation
for Statistical Computing; 2015. Available from: http://www.R-project.org/.
25. Keenan K, McGinnity P, Cross TF, Crozier WW, Prodo¨ hl PA. diveRsity: An R package for the estimation
and exploration of population genetics parameters and their associated errors. Methods Ecol Evol.
2013; 4(8):782–8. https://doi.org/10.1111/2041-210X.12067
26.
Leberg PL. Estimating allelic richness: Effects of sample size and bottlenecks. Mol Ecol. 2002; 11
(11):2445. https://doi.org/10.1046/j.1365-294x.2002.01612.x PMID: 12406254
27. Belkhir K, Borsa P, Chikhi L, Raufaste N, Bonhomme F. GENETIX 4.05: Population genetics software
for WindowsTM. Universite´ de Montpellier II Montpellier. 2004.
28. Kalinowski ST, Taper ML, Marshall TC. Revising how the computer program Cervus accommodates
genotyping error increases success in paternity assignment. Mol Ecol. 2007; 16(5):1099–106. https://
doi.org/10.1111/j.1365-294X.2007.03089.x PMID: 17305863
29. Do C, Waples RS, Peel D, Macbeth GM, Tillett BJ, Ovenden JR. NeEstimator v2: Re-Implementation of
software for the estimation of contemporary effective population size (Ne) from genetic data. Mol Ecol
Resour. 2014; 14(1):209–14. https://doi.org/10.1111/1755-0998.12157 PMID: 23992227
30. Wang J. Coancestry: A program for simulating, estimating and analysing relatedness and inbreeding
coefficients. Mol Ecol Resour. 2011; 11(1):141–5. https://doi.org/10.1111/j.1755-0998.2010.02885.x
PMID: 21429111
31. Wang J. Triadic IBD coefficients and applications to estimating pairwise relatedness. Genet Res
(Camb). 2007; 89(3):135–53. Epub 2007/09/27. https://doi.org/10.1017/S0016672307008798 PMID:
17894908.
32. R Core Team. R: A language and environment for statistical computing Vienna, Austria: R Foundation
for Statistical Computing; 2018. Available from: http://www.R-project.org/.
33.
Jost L. GST and its relatives do not measure differentiation. Mol Ecol. 2008; 17(18):4015–26. https://doi.
org/10.1111/j.1365-294x.2008.03887.x PMID: 19238703
34. Weir BS, Cockerham CC. Estimating F-statistics for the analysis of population structure. Evolution.
1984; 38(6):1358–70. https://doi.org/10.1111/j.1558-5646.1984.tb05657.x PMID: 28563791
35. Pritchard JK, Stephens M, Donnelly P. Inference of population structure using multilocus genotype
data. Genetics. 2000; 155(2):945–59. PMC1461096. PMID: 10835412
36.
37.
Jombart T, Devillard S, Balloux F. Discriminant analysis of principal components: A new method for the
analysis of genetically structured populations. BMC Genet. 2010; 11(1):94. https://doi.org/10.1186/
1471-2156-11-94 PMID: 20950446
Jombart T. Adegenet: A R package for the multivariate analysis of genetic markers. Bioinformatics.
2008; 24(11):1403–5. https://doi.org/10.1093/bioinformatics/btn129 PMID: 18397895
38. RStudio Team. RStudio: Integrated Development for R. Boston, MA: RStudio, Inc.; 2019.
39. Bradley JV. The insidious L-shaped distribution. Bulletin of the Psychonomic Society. 1982; 20(2):85–8.
https://doi.org/10.3758/BF03330089
40. Wickham H. ggplot2: elegant graphics for data analysis: Springer; 2016.
PLOS ONE | https://doi.org/10.1371/journal.pone.0236717 August 14, 2020
19 / 20
PLOS ONE Population genetics of ranched and fragmented Cape buffalo
41. Greyling BJ. Genetic variation, structure and dispersal among Cape buffalo populations from the Hluh-
luwe-Imfolozi and Kruger National Parks of South Africa [PhD thesis]. Pretoria, South Africa: University
of Pretoria; 2007.
42. Smitz N, Corne´ lis D, Chardonnet P, Caron A, Garine-Wichatitsky Md, Jori F, et al. Genetic structure of
fragmented southern populations of African Cape buffalo (Syncerus caffer caffer). BMC Evol Biol. 2014;
14:203–22. https://doi.org/10.1186/s12862-014-0203-2 PMID: 25367154
43.
44.
45.
van Hooft WF, Groen AF, Prins HHT. Microsatellite analysis of genetic diversity in African buffalo (Syn-
cerus caffer) populations throughout Africa. Mol Ecol. 2000; 9(12):2017–25. https://doi.org/10.1046/j.
1365-294x.2000.01101.x PMID: 11123614
van Hooft WF, Groen AF, Prins HHT. Genetic structure of African buffalo herds based on variation at
the mitochondrial D-loop and autosomal microsatellite loci: Evidence for male-biased gene flow. Con-
serv Genet. 2003; 4(4):467–77. https://doi.org/10.1023/A:1024719231545
du Toit JT. Order RUMINANTIA. In: Chimimba CT, Skinner JD, editors. The Mammals of the Southern
African Sub-region. 3rd ed. Cambridge: Cambridge University Press; 2005. p. 616–714.
46. Sinclair ARE. The African Buffalo: A Study of Resource Limitation of Populations. Chicago: The Uni-
versity of Chicago Press; 1977.
47.
Frankham R, Bradshaw CJA, Brook BW. Genetics in conservation management: Revised recommen-
dations for the 50/500 rules, Red List criteria and population viability analyses. Biol Conserv. 2014;
170:56–63. http://dx.doi.org/10.1016/j.biocon.2013.12.036.
48. Ralls K, Ballou JD, Dudash MR, Eldridge MDB, Fenster CB, Lacy RC, et al. Call for a paradigm shift in
the genetic management of fragmented populations. Conserv Lett. 2018; 11(2):e12412. https://doi.org/
10.1111/conl.12412
49.
50.
51.
Frankham R, Ballou JD, Eldridge MDB, Lacy RC, Ralls K, Dudash MR, et al. Predicting the probability
of outbreeding depression. Conserv Biol. 2011; 25(3):465–75. https://doi.org/10.1111/j.1523-1739.
2011.01662.x PMID: 21486369
Fitzpatrick SW, Bradburd GS, Kremer CT, Salerno PE, Angeloni LM, Funk WC. Genetic rescue without
genomic swamping in wild populations. bioRxiv. 2019:701706. https://doi.org/10.1101/701706
Fitzpatrick SW, Gerberich JC, Kronenberger JA, Angeloni LM, Funk WC. Locally adapted traits main-
tained in the face of high gene flow. Ecol Lett. 2015; 18(1):37–47. https://doi.org/10.1111/ele.12388
PMID: 25363522
52. Poirier M-A, Coltman DW, Pelletier F, Jorgenson J, Festa-Bianchet M. Genetic decline, restoration and
rescue of an isolated ungulate population. Evol Appl. 2019; 12(7):1318–28. https://doi.org/10.1111/eva.
12706 PMID: 31417617
53. Rodriguez-Ramilo ST, Wang J. The effect of close relatives on unsupervised Bayesian clustering algo-
rithms in population genetic structure analysis. Mol Ecol Resour. 2012; 12(5):873–84. https://doi.org/10.
1111/j.1755-0998.2012.03156.x PMID: 22639868
54. Noss RF. Indicators for monitoring biodiversity: A hierarchical approach. Conserv Biol. 1990; 4(4):355–
64. https://doi.org/10.2307/2385928
55. Di Marco M, Santini L. Human pressures predict species’ geographic range size better than biological
traits. Global Change Biol. 2015; 21(6):2169–78. https://doi.org/10.1111/gcb.12834 PMID: 25504910
56. Ceballos G, Ehrlich PR, Barnosky AD, Garcı´a A, Pringle RM, Palmer TM. Accelerated modern human–
induced species losses: Entering the sixth mass extinction. Science Advances. 2015; 1(5):e1400253.
https://doi.org/10.1126/sciadv.1400253 PMID: 26601195
PLOS ONE | https://doi.org/10.1371/journal.pone.0236717 August 14, 2020
20 / 20
PLOS ONE |
10.1371_journal.pone.0237277 | RESEARCH ARTICLE
Causal graph analysis of COVID-19
observational data in German districts reveals
effects of determining factors on reported
case numbers
Edgar Steiger*, Tobias MussgnugID, Lars Eric Kroll
Central Research Institute of Ambulatory Health Care in Germany (Zi), Berlin, Germany
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
* esteiger@zi.de
Abstract
OPEN ACCESS
Citation: Steiger E, Mussgnug T, Kroll LE (2021)
Causal graph analysis of COVID-19 observational
data in German districts reveals effects of
determining factors on reported case numbers.
PLoS ONE 16(5): e0237277. https://doi.org/
10.1371/journal.pone.0237277
Editor: Sungwoo Lim, New York City Department
of Health and Mental Hygiene, UNITED STATES
Received: July 29, 2020
Accepted: May 5, 2021
Published: May 27, 2021
Peer Review History: PLOS recognizes the
benefits of transparency in the peer review
process; therefore, we enable the publication of
all of the content of peer review and author
responses alongside final, published articles. The
editorial history of this article is available here:
https://doi.org/10.1371/journal.pone.0237277
Copyright: © 2021 Steiger et al. This is an open
access article distributed under the terms of the
Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: All relevant data are
available from public sources and are aggregated
in the github repository pertaining to the
Several determinants are suspected to be causal drivers for new cases of COVID-19 infec-
tion. Correcting for possible confounders, we estimated the effects of the most prominent
determining factors on reported case numbers. To this end, we used a directed acyclic
graph (DAG) as a graphical representation of the hypothesized causal effects of the deter-
minants on new reported cases of COVID-19. Based on this, we computed valid adjustment
sets of the possible confounding factors. We collected data for Germany from publicly avail-
able sources (e.g. Robert Koch Institute, Germany’s National Meteorological Service, Goo-
gle) for 401 German districts over the period of 15 February to 8 July 2020, and estimated
total causal effects based on our DAG analysis by negative binomial regression. Our analy-
sis revealed favorable effects of increasing temperature, increased public mobility for essen-
tial shopping (grocery and pharmacy) or within residential areas, and awareness measured
by COVID-19 burden, all of them reducing the outcome of newly reported COVID-19 cases.
Conversely, we saw adverse effects leading to an increase in new COVID-19 cases for pub-
lic mobility in retail and recreational areas or workplaces, awareness measured by searches
for “corona” in Google, higher rainfall, and some socio-demographic factors. Non-pharma-
ceutical interventions were found to be effective in reducing case numbers. This compre-
hensive causal graph analysis of a variety of determinants affecting COVID-19 progression
gives strong evidence for the driving forces of mobility, public awareness, and temperature,
whose implications need to be taken into account for future decisions regarding pandemic
management.
Introduction
As the COVID-19 pandemic progresses, research on mechanisms behind the transmission of
SARS-CoV-2 shows conflicting evidence [1–3]. While effects of mobility have been extensively
discussed, less is known on other factors such as changing awareness in the population [4–6]
or the effects of temperature [7–9]. A limiting factor in many studies is the lack of a causal
PLOS ONE | https://doi.org/10.1371/journal.pone.0237277 May 27, 2021
1 / 22
PLOS ONE manuscript: https://github.com/zidatalab/
causalcovid19.
Funding: The authors received no specific funding
for this work.
Competing interests: The authors have declared
that no competing interests exist.
Causal graph analysis for effects of determining factors on COVID-19 infections
approach to assess the causal contributions of various factors [10]. This can lead to distorted
estimates of the causal factors with observational data [10–12].
With COVID-19, we find ourselves in a situation in which information on the causal con-
tribution of various influencing factors in the population is urgently needed to inform politi-
cians and health authorities. On the other hand, trials cannot be carried out for obvious ethical
and legal reasons. Therefore, when assessing the effects of determinants of SARS-CoV-2
spread, special attention must be paid to strategies for the selection of confounding factors.
Another problem with assessing the effects of various determinants of SARS-CoV-2 spread
is the heterogeneity of the countries and regions examined for example in the Johns Hopkins
University (JHU) COVID-19 database [13]. The comparison of time series of case numbers
from different countries and observational periods can be strongly distorted by different fac-
tors like testing capacities and regional variations.
Our objective is to provide valid estimates of the effects of the main drivers of the pandemic
with a causal graph approach. We conducted a scoping review of the available studies regard-
ing signaling pathways and determinants of the spread of SARS-CoV-2 infections and the
reported new COVID-19 cases. Then we integrated the current findings into a directed acyclic
graph for the progress of the pandemic at the regional level. Using the resulting model and the
do-calculus we found identifiable effects without blocked causal paths whose effects can be
analyzed with observational data [14]. We used regional time series data of all German districts
(401) from various publicly available sources to analyze these questions on a regional level.
Germany is a good choice in this regard, because it has ample data on contributing factors
on the regional level and has had high testing and treatment capacities from early on in the
pandemic.
Causal model
We used a directed acyclic graph (DAG) [11, 12] as a tool to analyze the causal relationships
between several exposures and SARS-CoV-2 spread. To get an overview on published associa-
tions, a scoping review was conducted from 20th to 22nd of May 2020 within Pubmed and
Google scholar. Restrictions were applied to English and German language and the publication
date in the last one year. The following search terms were applied to abstracts and title in
Pubmed (“COVID-19” OR “COVID19” OR “Corona” OR “Coronavirus” OR “SARS-CoV-2”)
and connected separately in each case with the exposure variables (“mobility”, “public aware-
ness”, “awareness”, “google trends”,“ambient temperature”, “temperature”). For “mobility”,
we analyzed n = 8 studies, N = 103 were scanned in Pubmed, together with the first ten pages
(100 results) in Google scholar (“awareness”/“public awareness”/“google trends” n = 9,
N = 215; “temperature”/“ambient temperature” n = 16, N = 235). We integrated these findings
where possible into the construction of our DAG, which can be seen in Fig 1.
A number of studies report a strong association of mobility restrictions on the number of
new COVID-19 cases: Restrictive measures (e.g. “stay-at-home” orders, travel bans, or school
closures) are shown to possibly reduce the COVID-19 incidence [2, 15–21]. However, some
studies point out the combination of various non-pharmaceutical interventions (NPIs) is deci-
sive to prevent new infections [22, 23].
Google Trends [24] data can be used as a tool to get insights into public interest (aware-
ness) in the coronavirus disease. Several recent studies imply a connection of relative search
volumes (RSV) indices and reported new COVID-19 cases [4–6, 25–30]. Some search terms
e.g. “COVID-19” or “coronavirus” predated newly infected cases/total number of cases by
roughly 7 to 14 days for different countries [4–6, 26]. Additionally, we acknowledged that
PLOS ONE | https://doi.org/10.1371/journal.pone.0237277 May 27, 2021
2 / 22
PLOS ONE Causal graph analysis for effects of determining factors on COVID-19 infections
Fig 1. DAG of determinants of reported COVID-19 cases on the district level. Unobserved variables are light gray,
variables marked with an asterisk (�) are confounded by weekday/holiday.
https://doi.org/10.1371/journal.pone.0237277.g001
individual risk-aware behavior might be a reaction to the current COVID-19 burden (mea-
sured as reported cases at the day of exposure).
Mixed evidence is available regarding the effect of temperature: On the one hand several
papers report an association between increase in temperature and decrease in newly infected
COVID-19 cases [7–9, 31–36]. On the other hand, also the opposite has been found [37, 38].
Some studies found no association at all [22, 39–42]. It should be noted that few studies con-
sidered other confounding variables than meteorological ones (especially age and population
density among others [22, 36, 39]). In addition, the transferability of results between different
climate zones is questionable. To avoid possible bias caused by weather variables other than
temperature, we included rain, wind, and humidity in our model.
When investigating causal determinants of SARS-CoV-2 infections, a number of confound-
ers have to be considered. Well-known risk factors for SARS-CoV-2 as well as for other infec-
tions are demographic factors such as age, gender, socio-economic status (SES), population
density, and foreign citizenship/ethnicity [13, 43, 44]. In Germany along with other countries
(i.e. Brazil, USA, or the UK), populist parties or politicians and their electorate tend to be more
PLOS ONE | https://doi.org/10.1371/journal.pone.0237277 May 27, 2021
3 / 22
PLOS ONE Causal graph analysis for effects of determining factors on COVID-19 infections
sceptical about effects of containment measures than the other part of the electorate [45, 46].
Therefore we considered both “right-wing populist party votes” and “voter turnout” as possi-
ble confounders. Public health interventions were also taken into account (contact restrictions,
school closures etc.), as their implementation showed strong correlations with controlling the
spread of SARS-CoV-2 [22, 23, 47]. To avoid bias due to reporting delay of case numbers we
had to include weekday and German holidays. We included some unobserved variables in our
DAG (e.g. “Herd immunity”), too. Please note that “Exposure to SARS-CoV-2” is itself an
unobserved variable: German case numbers are reported with delay after date of exposure and
symptom onset. Exposure to the virus should not be confused with the formal exposure vari-
ables of the DAG.
Materials and methods
Data
We collected and aggregated data on reported COVID-19 cases, regional socio-demographic
factors, weather, and general mobility on district and state level in Germany for the period of
15 February 2020 to 8 July 2020. Our observation period for the outcome consisted of all dates
from 20 February 2020 to 8 July 2020 (T = 140), since we used a lag of 5 days for all confound-
ers. We did not exclude any states or districts (K = 401). We analyzed the daily reported num-
ber of new cases as outcome (K � T = 56 140 observations). The set of possible predictors was
derived from our causal DAG (see Table 1 and Fig 1). Due to modelling and data limitations,
some of the predictors were unobserved or were modelled as a construct consisting of several
variables. For our causal graph analysis, we computed adjustment sets separately for all
observed exposures within the DAG (if the respective exposure was identifiable within the
DAG causal analysis framework).
Variables. We downloaded German daily case numbers on district level reported by Rob-
ert Koch Institute (RKI, [48]) and aggregated them by date. The number of daily active cases
for day d was derived by subtracting the total number of reported cases on day d and day d
− 14 (14 days as a conservative estimate for the infectious period, which corresponds here to
the required quarantine time in Germany).
To assess the mobility of the German population, we used data publicly available on Ger-
man state level from Google [49]. Measurements are daily relative changes of mobility in per-
cent compared to the period of 3 January 2020 to 6 February 2020. Missing values (25 out of
13 488) were imputed with value 0 and the state level measurements were passed onto districts
within the corresponding state. Google mobility data was available for six different sectors of
daily life (“retail and recreation”, “grocery and pharmacy”, “parks”, “transit stations”, “work-
places”, “residential”) which means that “mobility” is a construct consisting of several vari-
ables. All variables but “residential” mobility are relative changes of daily visitor numbers to
the corresponding sectors compared to the reference period. “Residential” mobility is the rela-
tive change of daily time spent at residential areas. The six mobility variables showed high
correlations among each other and with other variables. To reduce multicollinearity, we trans-
formed them by principal component analysis (PCA) into six uncorrelated principal compo-
nents which were used in place of the original variables.
The notion of awareness in the population of COVID-19 describes the general state of alert-
ness about the new infectious disease. As such, it was hard to measure directly. As a proxy, we
used the relative interest in the topic term “corona” as indicated by Google searches. The daily
data was available on state level [24] and passed onto district level. As a second proxy for
awareness, we used the daily reported number of COVID-19 cases on the day of the exposure:
PLOS ONE | https://doi.org/10.1371/journal.pone.0237277 May 27, 2021
4 / 22
PLOS ONE Table 1. Observed model variables.
Variable
Weekday
Holiday (report)
Holiday (exposure)
Mobility
Retail and recreation
Grocery and pharmacy
Parks
Workplaces
Residential
Transit stations
Awareness
Searches corona
COVID-19 burden
Weather
Rainfall
Temperature
Humidity
Wind
Interventions
Ban of mass gatherings
School and kindergarten closures
Contact restrictions
Mandatory face masks
Socio-demographic
Age
Gender
Population density
Foreign citizens
Socio-economic status
Turnout
Right-wing populist party votes
Nursing homes
Case numbers
Causal graph analysis for effects of determining factors on COVID-19 infections
Dynamics
Level
Type
Unit/comment
daily
daily
daily
daily
daily
daily
daily
daily
daily
daily
daily
daily
daily
daily
daily
daily
daily
daily
daily
national
categorical
Sat through Thu as six binary variables, Fri as baseline
national
national
binary
binary
-
-
state
state
state
state
state
state
numeric
numeric
numeric
numeric
numeric
numeric
percent change compared to reference period
percent change compared to reference period
percent change compared to reference period
percent change compared to reference period
percent change compared to reference period
percent change compared to reference period
state
district
numeric
numeric
percent relative to other states and observation period
reported cases on day of exposure
district
district
district
district
numeric
numeric
numeric
numeric
mm (l/sqm)
˚C
relative humidity (%)
m/s
national
binary
-
state
numeric
0 for no closure, 1 for full closure, 0.5 for partial reopening
national
district
binary
binary
-
-
Source
-
-
-
Google [49]
Google [49]
Google [49]
Google [49]
Google [49]
Google [49]
Google [24]
RKI [48]
DWD [50]
DWD [50]
DWD [50]
DWD [50]
-
-
-
IZA [52]
INKAR [51]
INKAR [51]
INKAR [51]
INKAR [51]
INKAR [51]
INKAR [51]
INKAR [51]
RKI [48]
RKI [48]
constant
constant
constant
constant
constant
constant
constant
constant
district
district
district
district
district
district
district
district
numeric
numeric
numeric
numeric
numeric
numeric
numeric
numeric
2 variables: share of population > = 65 years & <18 years
share of female population
population per sqkm
2 variables: share of foreign citizens & of population seeking refuge
INKAR [51]
share of households with low income
voter turnout in last election
share of votes for AfD in last election
number of nursing (retirement) homes
Reported new cases of COVID-19
Active cases
daily
daily
district
district
numeric
numeric
-
active cases on day of report
https://doi.org/10.1371/journal.pone.0237277.t001
Since media reported case numbers prominently, we assumed that this could reflect individual
awareness, too.
We constructed daily weather from four variables (“temperature”, “rainfall”, “humidity”,
“wind”). Weather data was downloaded from Deutscher Wetterdienst (DWD, [50]) for all
weather stations in Germany below 1000 meters altitude with daily records for our observation
period. District level daily weather data was aggregated per district by averaging the data from
the three nearest weather stations (which includes weather stations inside the district). Missing
values were imputed with mean values (n = 59 for wind).
The reported number of COVID-19 cases varied strongly by day of the week. Thus, we
included “weekday” as a categorical variable. Similarly, the reported cases and the exposure
to the virus were affected by official holidays. Within the observation period, this included
PLOS ONE | https://doi.org/10.1371/journal.pone.0237277 May 27, 2021
5 / 22
PLOS ONE Causal graph analysis for effects of determining factors on COVID-19 infections
among others Good Friday, Easter Monday, and Labor Day. To correct for effects of these
days, we included two variables in the model, “Holiday (report)” (indicates if the day of the
report was a holiday, because governmental health departments were less likely to be on full
duty) and “Holiday (exposure)” (indicates if the day of exposure to the virus was a holiday,
because the population behaves differently on holidays).
For different official and political interventions on a daily basis and the district level we
used one-hot encoded daily variables, i.e. ban of mass gatherings, school and kindergarten clo-
sures and their gradual reopening, contact restrictions, and mandatory face masks for shop-
ping and public transport.
We included several social, economic, and demographic factors on the district level with
direct or indirect influence on the risk of exposure to SARS-CoV-2 in our analysis. All are
readily available from INKAR database [51]. We used the share of population that is 65 years
or older and the share of population that is younger than 18 years (Age), the share of females
in population (Gender), the population density, the share of foreign citizenships and the share
of the population seeking refuge (Foreign citizenship), the share of low-income households
(Socio-economic status), voter turnout, share of right-wing populist party votes, and the num-
ber of nursing (retirement) homes.
All continuous variables but the outcome “Reported new cases of COVID-19” and the offset
“Active cases” were centered and scaled by one standard deviation for numerical stability,
while we left binary variables as-is. After estimating the effects of variables, we re-scaled con-
tinuous variables’ effects to their original scale. Additionally for mobility variables, we re-trans-
formed the effects of the principal components to the original mobility variables. Furthermore,
we lagged the effect of all variables (but outcome, offset, and the non-dynamic socio-demo-
graphic variables) by 5 days (optimal lag found by cross-validation) which means that we
assumed that their effects on the outcome will be visible after 5 days.
Methods
Causal analysis with DAG and adjustment sets. We used a directed acyclic graph as a
graphical representation of the hypothesized causal reasoning that leads to exposure to the
SARS-CoV-2 virus, onset of COVID-19, and finally reports of COVID-19 cases. We use the
terms “causal effect” or “causal relationship” for effect estimates that are based on this causal
graph framework. Every node vi in the graph is the graphical representation of an observed or
unobserved variable xi, a directed edge eij is an arrow from node vi to vj that implies a direct
causal relationship from variable xi onto variable xj. The set of all nodes is denoted by V, the
set of all edges by E, as such, the complete DAG is the tuple G = (V, E). The seminal works of
Spirtes and Pearl [53, 54] introduce the theory of causal analysis, do-calculus, and how to ana-
lyze a DAG to estimate the total or direct causal effect from a variable xi onto a variable xj. The
direct effect is the effect associated with the edge eij only (if it exists), while the total effect takes
indirect effects via other paths from vi to vj into account, too. Here we estimated total effects
only, since most of our variables were not hypothesized to have a direct effect on the reported
number of new COVID-19 cases. In contrast to prediction tasks, where one would include all
variables available, it is actually ill-advised to use all available variables to estimate causal
effects, due to introducing bias by adjusting for unnecessary variables within the causal DAG.
This is why we need to identify a valid set of necessary variables (an adjustment set) to estimate
the proper causal effect [54]. The “minimal adjustment set” [55] is a valid adjustment set of
variables that does not contain another valid adjustment set as a subset. However, identifying a
minimal adjustment set might not be enough to reliably estimate the causal effect. Thus, we
PLOS ONE | https://doi.org/10.1371/journal.pone.0237277 May 27, 2021
6 / 22
PLOS ONE Causal graph analysis for effects of determining factors on COVID-19 infections
identified the “optimal adjustment set” [56] as the set of variables which is a valid adjustment
set while having the lowest Akaike information criterion (AIC).
We analyzed the DAG from Fig 1 with the R Software [57] and the R packages dagitty
(formal representation of the graph and minimal adjustment sets [12]) and pcalg (for find-
ing an optimal adjustment set [58]). For the defined exposures and the outcome “Reported
new cases of COVID-19”, we computed the minimal and optimal adjustment sets. Since it was
possible that these sets contained unobserved variables that needed to be left out of the regres-
sion model, we chose the valid set with the lowest AIC (see next section) to estimate the final
total causal effect from exposure to outcome.
Regression with negative binomial model. We can estimate the causal effect from expo-
sure to outcome by regression [54]. Since the outcome “Reported new cases of COVID-19” is a
count variable, one should not employ a linear regression model with Gaussian errors, but
instead we assumed a log-linear relationship between the expected value of the outcome Y
(new cases) and regressors x, as well as a Poisson or negative binomial distribution for Y:
logðE½Yjx�Þ ¼ a þ
X
bi � xi;
i2S
ð1Þ
where α is the regression intercept, S is the set of adjustment variables for the exposure i�
including the exposure variable itself, βi are the regression coefficients corresponding to the
variables xi. As such βi� is the total causal effect from exposure variable xi� on the outcome Y.
The Poisson regression assumes equality of mean and variance. If this is not the case one
observes so-called overdispersion (the variance is higher than the mean), this indicates one
should use regression with a negative binomial distribution instead to estimate the variance
parameter separately from the mean.
We needed to account for the fact that our outcome is not counted per time unit (one day)
only, but depends on the number of active COVID-19 cases: Holding all other variables fixed,
the number of new cases Y is a constant proportion of the number of active cases A. This was
modeled by including an offset log(A + 1) in the regression model Eq (1):
logðE½Yjx�Þ ¼ a þ logðA þ 1Þ þ
X
bi � xi
i2S
�
, log
�
E½Yjx�
A þ 1
¼ a þ
P
bi � xi
,
E½Yjx�
A þ 1
¼ expðaÞ �
Y
expðbiÞxi :
ð2Þ
ð3Þ
Here we added a pseudocount “+1” to ensure a finite logarithm and avoid division by 0.
One can interpret the model as approximating the log-ratio of new cases and active cases by
a linear combination of the regressor variables in Eq (2). If all variables xi are centered in Eq
(3), we have for the baseline 8i xi ¼ 0 ) E½Yjx ¼ 0� ¼ expðaÞ � ðA þ 1Þ. In other words, the
exponentiated intercept is the baseline daily infection rate (how many people does one infected
individual infect in one day). If we hold all variables xi fixed (e.g. at baseline 0) in Eq (3) but
now increase the exposure variable xi� = 0 by one unit to xi� + 1 = 0 + 1, we have
E½Yjx0� ¼ expðaÞ � ðA þ 1Þ � expðbxi� þ1
i�
Y
Þ
i6¼i�
expðbiÞ0
¼ expðaÞ � ðA þ 1Þ � expðbi� Þ;
PLOS ONE | https://doi.org/10.1371/journal.pone.0237277 May 27, 2021
7 / 22
PLOS ONE Causal graph analysis for effects of determining factors on COVID-19 infections
which means the exponentiated coefficient βi� describes the rate change of the outcome by one
unit increase of the exposure.
In practice, given observations of Y and x we estimate the regression coefficients α and βi by
maximum likelihood [59]. Our observational measurements are ykt and xikt, where k indicates
the corresponding district and t the date of measurement.
We conducted a log-linear regression (function glm with family = poisson() for
Poisson regression, and glm.nb from the MASS package for the negative binomial regression
[60]) for the full data set to assess general model adequacy and to estimate the θ parameter of
the negative binomial. The proper lag between exposures and outcome was found by 10-fold
cross-validation on different lags between 1 and 20 days. Model diagnostics on the final full
model did not show severe problems with model assumptions (linearity, distribution of residu-
als, independence of observations). Analysis of variance inflation factors revealed some prob-
lems with multicollinearity. To reduce the effects of multicollinearity, first we transformed the
highly correlated mobility variables by PCA as described above. Second, we used a ridge
regression approach [61], which is a regularization method that shrinks regression coefficients
and alleviates the effect of correlation between variables on their respective regression coeffi-
cients. Furthermore, regularized regression allows for better fits on unseen data, thus prevent-
ing overfitting the data, too. The hyper-parameter λ of the ridge regression was chosen by
10-fold cross-validation, where the folds were constructed from random subsets of the 401 dis-
tricts. We used this hyper-parameter with the cv.glmnet function from the R package
glmnet [62] with family = negative.binomial(theta) and chose the λ value
within one standard deviation from the minimal λ as regularization hyper-parameter. After-
wards, we calculated the effects of separate exposures on the outcome. For every exposure, we
analyzed the different valid adjustment sets given by analysis of the causal DAG (i.e. the mini-
mal and optimal adjustment sets). Then, we first checked if the respective set included unob-
served variables. If this was the case for the optimal adjustment set, we discarded the
unobserved variables from the set and checked if it was still a valid adjustment set (function
gac in package pcalg [63]). If a minimal adjustment set contained unobserved variables, we
discarded the whole set. If no valid adjustment set for a given exposure was available, we con-
cluded that the effect of this exposure was unidentifiable within our causal graph. We used the
function glmnet with the parameters θ and λ as above on every remaining valid adjustment
set as regressors (that is, we applied ridge regression) and calculated the Akaike information
criterion (AIC) for this model/set of regressors. Finally, for every exposure, we decided for the
model/adjustment set (if available) with the lowest AIC. We report the exponentiated esti-
mated coefficients for the separate exposures on their original scale.
Results
Descriptive statistics for the included variables are presented in Table 2.
In the observational period, the number of daily reported COVID-19 cases increased till the
end of March/beginning of April and continually decreased afterwards till the beginning of
June 2020 with a slight increase and decrease afterwards (Fig 2A). On the other hand, the (log-
)ratio of reported cases over active cases decreased steeply till the mid of April and increased
steadily afterwards with a slight decrease close to the end of the observation period (Fig 2B).
Both figures examplify a considerable variation among the districts (light blue points are indi-
vidual district’s data).
In Germany, we observed a rebound in mobility after the initial political measures, reduc-
tions in incident cases were associated with a diminishing public interest in COVID-19, and
temperatures were overall increasing (cf. Fig 3); with correlations between temporal
PLOS ONE | https://doi.org/10.1371/journal.pone.0237277 May 27, 2021
8 / 22
PLOS ONE Causal graph analysis for effects of determining factors on COVID-19 infections
Table 2. Descriptive statistics for observed variables.
Variable
n
Mobility
Retail and recreation
Grocery and pharmacy
Parks
Workplaces
Residential
Transit stations
Awareness
Searches corona
COVID-19 burden
Weather
Rainfall
Temperature
Humidity
Wind
Interventions
Ban of mass gatherings
School and kindergarten closures
Contact restrictions
Mandatory face masks
Socio-demographic
Age (pop. 65 and older)
Age (pop. younger 18)
Gender
Population density
Foreign citizens
Foreign citizens (refugees)
Socio-economic status
Turnout
Right-wing populist party votes
Nursing homes
Case numbers (Outcome and offset)
Reported new cases COVID-19
Active cases
https://doi.org/10.1371/journal.pone.0237277.t002
mean (SD)
56140
-26.62 (24.60)
-3.94 (22.77)
47.26 (58.20)
-22.96 (20.35)
8.13 (6.49)
-29.58 (21.11)
26.94 (18.23)
3.50 (10.28)
1.89 (4.01)
10.90 (5.33)
67.81 (13.03)
3.63 (1.66)
0.83 (0.38)
0.54 (0.36)
0.74 (0.44)
0.49 (0.50)
22.09 (2.74)
16.17 (1.25)
50.59 (0.64)
533.75 (701.84)
10.03 (5.14)
1.88 (1.14)
30.64 (6.02)
75.08 (3.79)
13.39 (5.32)
36.11 (30.69)
3.53 (10.29)
48.76 (120.86)
progression and mobility in retail and recreation rA,B = 0.02, awareness (“Searches corona”)
rA,C = -0.3, and temperature rA,D = 0.8.
Main results
We list the results of our causal analysis for the effects of different exposure variables in
Table 3. The estimates are multiplicative rates of increase/decrease for a one unit increase of
the respective variable: Values above 1 lead to an increase, below 1 to a decrease of the infec-
tion rate. To put these estimates into perspective, Fig 4 shows the relative causal effect of the
different exposure variables on the number of reported COVID-19 cases on a range of sensible
values of the exposure variables (95 percent quantiles of data points).
PLOS ONE | https://doi.org/10.1371/journal.pone.0237277 May 27, 2021
9 / 22
PLOS ONE Causal graph analysis for effects of determining factors on COVID-19 infections
Fig 2. Temporal and district level variation of outcome (log-scale).
https://doi.org/10.1371/journal.pone.0237277.g002
Within our framework, we saw very different effects for individual mobility variables. For
mobility in retail/recreation, an increase of 1 percent point mobility compared to the reference
period (03 January to 06 February 2020) leads to an increase of the daily reported case number
by about 0.11 percent. Similarly, mobility on workplaces showed an effect of 0.33 increase in
case numbers for every 1 percent point increase in mobility, while mobility on transit stations
showed an effect of 0.26 increase in case numbers for every 1 percent point increase.
Fig 3. Temporal variation of outcome and main determinants.
https://doi.org/10.1371/journal.pone.0237277.g003
PLOS ONE | https://doi.org/10.1371/journal.pone.0237277 May 27, 2021
10 / 22
PLOS ONE Causal graph analysis for effects of determining factors on COVID-19 infections
Table 3. Effect estimates from causal graph analysis.
Cause
Mobility
Retail and recreation
Grocery and pharmacy
Parks
Transit stations
Workplaces
Residential
Awareness
Searches corona
COVID-19 burden
Weather
Temperature
Rainfall
Humidity
Wind
Interventions
Interventions (ban of mass gatherings)
Interventions (school and kindergarten closures)
Interventions (contact restrictions)
Interventions (mandatory face masks)
Demographic
Age (pop. 65 and older)
Age (pop. younger 18)
Foreign citizens
Foreign citizens (refugees)
Gender
Nursing homes
Population density
Socio-economic status
https://doi.org/10.1371/journal.pone.0237277.t003
Effect estimate
1.0011
0.9977
0.9997
1.0026
1.0033
0.9903
1.0089
0.9980
0.9905
1.0121
1.0057
1.0329
0.9729
0.9277
0.8314
0.9064
0.9953
1.0120
1.0048
0.9985
0.9925
1.0011
1.0000
0.9982
Contrarily, the remaining three mobility variables showed negative effects on the number of
reported COVID-19 cases. An increase of 1 percent point mobility for the areas of grocery/
pharmacy leads to a decrease in the reported case number by approximately 0.23 percent,
while increased mobility of 1 percent point within parks leads to a decrease in the reported
case number by approximately 0.03 percent, and finally an increase of 1 percent point in resi-
dential mobility leads to a decrease by approximately 0.97 percent. Fig 4 shows the effects of
mobility on a range of possible values. Thus, we expect an increase of daily cases by approxi-
mately 7.8 percent if mobility in workplaces reaches baseline levels of 0 percent difference to
the reference period. On the other hand, an increase of mobility for residential areas by 10
percent points compared to the reference period leads to a reduction of the infection rate by
approximately 1.8 percent.
“Awareness” had two opposite effects on the outcome in our DAG. Awareness measured by
Google searches for corona had a positive effect on the number of reported cases. An one per-
cent point increase of the state’s Google searches (relative to other states and the observation
period) leads to an increase of approximately 0.89 percent. For example, if a district shows 10
percent points more relative searches for corona than another one, we expect approximately
PLOS ONE | https://doi.org/10.1371/journal.pone.0237277 May 27, 2021
11 / 22
PLOS ONE Causal graph analysis for effects of determining factors on COVID-19 infections
Fig 4. Relative causal effects of exposures.
https://doi.org/10.1371/journal.pone.0237277.g004
PLOS ONE | https://doi.org/10.1371/journal.pone.0237277 May 27, 2021
12 / 22
PLOS ONE Causal graph analysis for effects of determining factors on COVID-19 infections
9.3 percent more infections for this district after 5 days. COVID-19 burden (reported number
of cases on day of exposure) affected the outcome negatively, where every additional daily case
in the district leads to a 0.2 percent decrease in newly reported case numbers. The correspond-
ing plot in Fig 4 visualizes this relationship: For a local outbreak with 20 daily cases as COVID-
19 burden, we estimate as total causal effect a subsequent reduction of infection rate by 3.9
percent.
Within our model, we observed effects of temperature and all other weather variables.
Every increase of 1 degree Celsius in temperature leads to a reduction of the daily reported
case numbers by approximately 0.95 percent. On the other hand, we found an increasing effect
of rainfall: One millimeter (=1 liter per square meter) more rainfall leads to an increase of
reported case numbers by approximately 1.21 percent. We observe effects for humidity and
wind as well (higher humidity and stronger wind leading to more cases). In perspective (Fig
4), with temperature we expect an increase by approximately 21 percent at a daily average tem-
perature of 0˚C compared to a day with 20˚C. For rainfall, we expect on a rainy day with 10
mm rainfall a corresponding increase of the infection rate by approximately 12.8 percent com-
pared to a day with no precipitation.
The different intervention variables showed the strongest effects in our analysis, see
Table 3. While the first intervention (ban of mass gatherings) reduced subsequent daily case
numbers by 2.7 percent, the closure of schools/kindergartens reduced infections by an addi-
tional 7.2 percent and mandatory face masks reduced this by another 9.4 percent. The effect of
contact restrictions was the strongest in our observation period, with an reduction of the case
rates by 16.9.
The effects of the different socio-demographic factors are quite small in comparison to the
effects described above. We see an increasing effect on case numbers by additional nursing
homes between districts. Districts with a younger population, more foreign citizens, higher
population density and a lower average social-economic status showed higher case numbers,
too.
For all exposures, our analysis pipeline opted to use the (reduced) optimal adjustment set
over the minimal adjustment sets because of lower AICs, except for exposure variable “nursing
homes”, for which the minimal adjustment set had the lowest AIC. For an overview of all final
adjustments sets, see Table 4. We found that there were no valid adjustment sets for the non-
identifiable variables turnout and right-wing populist party votes.
We decided for a lag of 5 days based on cross-validation. Similarly, negative binomial
regression was chosen over Poisson regression, because the latter showed overdispersion and
an higher AIC value.
Discussion
Main findings
Our objective was to identify effects of determining factors for COVID-19 cases within a causal
framework. We found that weather affects the reported number of infections, especially tem-
perature (which has a reducing effect on case numbers) and rainfall (which increases case
numbers). We saw that reports of high case numbers in districts led to a reduction in new
infection numbers, which indicates risk-averse awareness in the population and/or effective
public health measures to suppress a local outbreak. Mobility showed distinct effects: Increas-
ing activity in retail and recreational areas, as well as transit stations and workplaces increased
reported case numbers, while increased movement for essential shopping (grocery and phar-
macy) and in parks or residential areas led to reduced case numbers. All interventions consid-
ered (ban of mass gatherings, school/kindergarten closures, contact restrictions, mandatory
PLOS ONE | https://doi.org/10.1371/journal.pone.0237277 May 27, 2021
13 / 22
PLOS ONE Causal graph analysis for effects of determining factors on COVID-19 infections
-
o
i
c
o
S
c
i
m
o
n
o
c
e
s
u
t
a
t
s
n
o
i
t
a
l
u
p
o
P
y
t
i
s
n
e
d
g
n
i
s
r
u
N
s
e
m
o
h
r
e
d
n
e
G
n
g
i
e
r
o
F
s
n
e
z
i
t
i
c
e
g
A
s
n
o
i
t
n
e
v
r
e
t
n
I
d
n
i
W
y
t
i
d
i
m
u
H
l
l
a
f
n
i
a
R
e
r
u
t
a
r
e
p
m
e
T
9
1
-
D
I
V
O
C
s
e
h
c
r
a
e
S
y
t
i
l
i
b
o
M
n
e
d
r
u
b
a
n
o
r
o
c
.
s
i
s
y
l
a
n
a
l
a
s
u
a
c
r
o
f
s
t
e
s
t
n
e
m
t
s
u
j
d
a
l
a
n
i
F
.
4
e
l
b
a
T
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
)
e
r
u
s
o
p
x
e
(
y
a
d
i
l
o
H
)
t
r
o
p
e
r
(
y
a
d
i
l
o
H
y
a
d
k
e
e
W
a
n
o
r
o
c
s
e
h
c
r
a
e
S
I
9
1
-
D
V
O
C
r
e
h
t
a
e
W
n
e
d
r
u
b
y
t
i
l
i
b
o
M
s
s
e
n
e
r
a
w
A
y
t
i
l
i
b
o
M
e
r
u
t
a
r
e
p
m
e
T
y
t
i
d
i
m
u
H
l
l
a
f
n
i
a
R
d
n
W
i
c
i
h
p
a
r
g
o
m
e
d
-
o
i
c
o
S
s
n
o
i
t
n
e
v
r
e
t
n
I
s
n
o
i
t
n
e
v
r
e
t
n
I
s
n
e
z
i
t
i
c
n
g
i
e
r
o
F
c
i
m
o
n
o
c
e
-
o
i
c
o
S
s
e
m
o
h
g
n
i
s
r
u
N
t
u
o
n
r
u
T
i
g
n
w
-
t
h
g
i
R
y
t
r
a
p
t
s
i
l
u
p
o
p
s
e
t
o
v
s
u
t
a
t
s
r
e
d
n
e
G
e
g
A
n
o
i
t
a
l
u
p
o
P
y
t
i
s
n
e
d
4
0
0
t
.
7
7
2
7
3
2
0
.
e
n
o
p
.
l
a
n
r
u
o
j
/
1
7
3
1
.
0
1
/
g
r
o
.
i
o
d
/
/
:
s
p
t
t
h
PLOS ONE | https://doi.org/10.1371/journal.pone.0237277 May 27, 2021
14 / 22
PLOS ONE Causal graph analysis for effects of determining factors on COVID-19 infections
face masks) reduced case numbers considerably. Socio-demographic variables had small
effects individually, but in conjunction they explained larger case numbers in (urban) areas
with younger population, lower socio-economic status, and higher population density.
Furthermore, we made a strong case for the use of causal DAGs in epidemiology and a pan-
demic like COVID-19: DAGs allow to choose confounders for the analysis in a principled and
statistically correct way while reducing possible causes for bias. Also, the DAG formalization
allows for discussion about the underlying causal assumptions.
Comparison with previous research
Most research on determinants affecting case numbers of COVID-19 is restricted to single
aspects [5, 16, 32, 35]. To reliably identify (causal) drivers, one must adjust for confounders.
To this end, we used an integrated model with variables from different aspects like mobility,
awareness, weather, or socio-demographics and identified confounders by causal analysis with
a directed acyclic graph. A causal approach is used in another current COVID-19 analysis
[64]. There, however, they identify the causal relationships (reconstruct a DAG), while we esti-
mated effects for a given hypothesized causal DAG.
Several studies assessing the impact of public health measures on mobility have each
observed a downward trend accompanied by a decrease in the number of newly reported cases
[15–17, 19, 23, 47].
Our findings regarding awareness/Google Trends analysis are in good agreement with the
correlations found by others [4, 6, 26], who conclude that alertness to COVID-19 rises several
days before the highest number of cases are reported. At this point it should be noted, that
awareness is substantially influenced by public media coverage, which should be considered, if
possible, in future studies [4]. As such, awareness is difficult to measure and here the number
of Google searches for “corona” could only be a proxy for this concept.
In addition, in alignment with other recent published studies, our results confirm evidence
which associated a negative effect of temperature on new COVID-19 cases [7–9, 31–36]. It is
however controversial to other scientific literature describing no effects [22, 39–42] or even
converse correlations [37, 38]. The conflicting results might be explained by different climates
and characteristics of the populations under study. While we are confident that our strict
causal analysis resulted in effect estimates as undistorted as possible, there might be unconsid-
ered bias in those other studies. Further research needs to be done to elucidate the biological
characteristics of the novel virus SARS-CoV-2 regarding its ambient temperature survival
and transmission. Finally, we found a positive effect of increased precipitation and a raise in
COVID-19 cases, which supports previous observations [33].
A recent review on COVID-19 based on evidence from the US and UK concludes that low
socio-economic status groups are being hit harder by the pandemic [65]. Albeit specific path-
ways remain unclear, many studies found associations with poverty or its correlates such as
poor and potentially overcrowded housing conditions. For Germany, a higher case fatality of
COVID-19 cases in districts with higher socio-economic deprivation has also been reported
just recently, which was especially pronounced in the second wave of the pandemic [66]. Simi-
larly, our analysis identified a decreasing effect on COVID-19 case numbers within districts
with a higher socio-economic status during the first wave.
Limitations and strengths
While use of a causal DAG is itself a strong tool to identify causal effects (and not just statistical
associations), it introduces two limitations: causal assumptions within the graph (depicted by
edges) need to be well justified, and the statistical regression model that calculates total causal
PLOS ONE | https://doi.org/10.1371/journal.pone.0237277 May 27, 2021
15 / 22
PLOS ONE Causal graph analysis for effects of determining factors on COVID-19 infections
effects needs to be appropriate for the task at hand. We endorse our graph as a basis for discus-
sion on residual confounding. We did not try to construct the DAG from the available data (cf.
[64]). As such, our proposed DAG is not entirely consistent with the data and there are condi-
tional dependencies between variables that cannot be dissolved by adding edges to the DAG
(e.g. between the interventions like contact restrictions and mandatory face masks). Another
way to identify potential problems in the proposed DAG is to perform a sensitivity analysis of
its structure by inspecting its maximal ancestral graph (MAG) or its Markov equivalence class
represented by a complete partially DAG (CPDAG) and the existence of valid adjustment sets
for these generalized graphs [67]. For the MAG derived from our DAG, only the effects for
exposures mobility and searches for corona can be estimated with valid adjustment sets, while
for the Markov equivalence class all exposures but COVID-19 burden lead to valid adjust-
ments sets. A further analysis of these implications is out of the scope of this paper.
We observed overdispersion and a substantial increase in model performance with a nega-
tive binomial regression compared to Poisson regression, which is in line with the results on
COVID-19 daily case counts of [17] and others [7, 9, 68]. We did not model case counts with a
differential equation model like the classic SIR-model [69] and its successors, since these are
more suited to prediction e.g. [70], while our choice of a negative binomial regression frame-
work allowed us to estimate the effects of confounders more reliably. There are more advanced
statistical methods for count data, e.g. zero-inflated models and mixed models. We tested both
approaches as extensions to the negative binomial regression and experienced numerical prob-
lems and increased computing time, along with an insubstantial increase in model perfor-
mance. Furthermore, our model assumed that all variables have effects proportional to the size
of their measurements. It is possible that some variables show saturation effects or opposite
effects for low, medium, or high values. This could be modeled with polynomial or other trans-
formations of the variables, which we did not employ due to limited temporal and spatial data
availability. Interaction effects of variables and confounding effects or mediating variables
are explicitly taken care of by deriving the valid adjustment sets for a given exposure based on
the causal DAG. Use of a fixed DAG with effect estimation via regression assumes that data
was generated by the same underlying process for the observation period. By inclusion of the
successive mitigation interventions as binary variables we were able to explain some of the var-
iance caused by the changing dynamics of case numbers (similar to [68]). While multicolli-
nearity of variables poses less of a problem for a proper causal graph analysis [71], we
addressed the problem of multicollinearity in our predictors by two approaches: principal
component analysis for the highly collinear mobility variables as well as a regularized regres-
sion approach (ridge regression). The latter (in conjunction with cross-validation) also
reduced the problem of overfitting.
We stress the point that our effects were deduced on an aggregate (district) level in the
absence of available data on an individual level. As such, conclusions about effects cannot be
transferred on individuals without the possibility for an ecological fallacy. Furthermore, as we
were using administrative data for our analysis, the results are susceptible to the Modifiable
Area Unit Problem (MAUP) [72]. The MAUP postulates that different regional aggregations
of the units of observation may lead to different results and conclusions. Due to limited avail-
able data for the different variables, there is currently no way to overcome these problems that
are inherent to all analyses on aggregated data level.
Our observation period was restricted to succession from late winter to spring and summer
(February to July). Nevertheless, this transition with increasing temperature was a natural
experiment that allowed clues on weather effects.
We could not include data on health care utilization during the pandemic into our models
due to the lack of available resources. This is planned for a later follow up to this paper since
PLOS ONE | https://doi.org/10.1371/journal.pone.0237277 May 27, 2021
16 / 22
PLOS ONE Causal graph analysis for effects of determining factors on COVID-19 infections
we rank health care utilization and mobility within health care facilities among the strong fac-
tors for COVID-19 progression: personnel in hospitals and private practices is particularly
exposed to infection, while the lack of adequate care for other diseases has severe effects on
general health of the population. At the same time, health care facilities are key for testing and
surveillance of COVID-19 patients.
Social determinants of health are important factors to consider in an epidemiological frame-
work of a pandemic disease like COVID-19. To account for this problem, we included several
socio-economic confounders that were available on a district level in Germany. While our anal-
ysis is not an exhaustive analysis of the effects of social determinants on COVID-19 infections,
we emphasize the necessity of their inclusion and our results add to the growing body of evi-
dence that these factors interact with each other and cluster especially among people or within
areas of underprivileged conditions, with detrimental effects on population health [73].
While our analysis focused on Germany and its districts, we assume that results may be
transferred to other countries by adjusting for their respective weather conditions, mobility
habits, socio-demographic characteristics, and other determining factors.
The code and resources for our analysis are available on Github, we invite other researchers
to replicate our analysis with different assumptions using the files provided in the repository of
the article (https://github.com/zidatalab/causalcovid19).
Discussion of causal effects
In our analysis, the adverse effects of mobility in retail/recreation and workplaces and the
favorable effect of mobility in grocery/pharmacy and residential areas indicate that interven-
tions like contact restrictions which limit the number of individual interactions can lead to
reduced infection numbers. This is due to retail/recreational and workplace areas encompass-
ing mostly places of (social) gatherings, while if people are doing more of their essential shop-
ping at supermarkets and stay at home with less contact to other people, they are less likely to
come in contact with infected individuals.
The effects of awareness measured via searches for “corona” and the COVID-19 burden are
harder to interpret. We assume that within our model, the searches for “corona” are an insuffi-
cient proxy for awareness, while the decreasing effect for future case numbers of high daily
COVID-19 burden indicates it affects individual risk-behavior and entails effective non-phar-
maceutical interventions.
Similarly, the effects of temperature and rainfall can be interpreted as causal effects for
indoor and outdoor activities, such that higher temperatures and low rainfall indicate more
people spending time outdoor while lower temperatures and high rainfall result in indoor
activities, which lead to more infections. Current research suggests this to be due to the preva-
lent airborne and respiratory droplets and aerosol transmission of the SARS-CoV-2 virus [74].
In this light, we advocate for precautious measures like increased hygiene, face masks, and air
ventilation for unavoidable indoor activities.
Furthermore, our analyses strongly support the effectiveness of non-pharmaceutical inter-
ventions. To a lesser extent, the adverse effects of some socio-demographic factors might help
to identify areas that are at higher risk of local COVID-19 outbreaks and more severe out-
comes of infection cases.
Conclusion
To the best of our knowledge, this is the most comprehensive analysis of causes for COVID-19
infections which integrates different data sources (all publicly available). Causal reasoning
with a DAG allows us to estimate the possible causal effects more reliably.
PLOS ONE | https://doi.org/10.1371/journal.pone.0237277 May 27, 2021
17 / 22
PLOS ONE Causal graph analysis for effects of determining factors on COVID-19 infections
Our findings suggest that the infection-driving effects of mobility, awareness, and weather
(and to some extent socio-demographic factors) need to be taken into account when deciding
for mitigation and suppression interventions, depending on the recent and future COVID-19
pandemic development.
Acknowledgments
We are thankful for feedback from Thomas Czihal, Johannes Textor, Ralph Brinks, and an
anonymous reviewer who gave helpful suggestions on earlier versions of the manuscript.
Author Contributions
Conceptualization: Edgar Steiger, Lars Eric Kroll.
Data curation: Edgar Steiger.
Formal analysis: Edgar Steiger.
Investigation: Edgar Steiger, Tobias Mussgnug.
Methodology: Edgar Steiger, Lars Eric Kroll.
Project administration: Lars Eric Kroll.
Software: Edgar Steiger.
Supervision: Lars Eric Kroll.
Visualization: Edgar Steiger.
Writing – original draft: Edgar Steiger, Tobias Mussgnug, Lars Eric Kroll.
Writing – review & editing: Edgar Steiger, Tobias Mussgnug, Lars Eric Kroll.
References
1. WHO Team. Report of the WHO-China Joint Mission on Coronavirus Disease 2019 (COVID-19); 2020,
accessed 2020-06-25. Available from: https://www.who.int/publications-detail/report-of-the-who-china-
joint-mission-on-coronavirus-disease-2019-(covid-19).
2. Chinazzi M, Davis JT, Ajelli M, Gioannini C, Litvinova M, Merler S, et al. The effect of travel restrictions
on the spread of the 2019 novel coronavirus (COVID-19) outbreak. Science. 2020; 368(6489):395–400.
https://doi.org/10.1126/science.aba9757 PMID: 32144116
3. Guan W, Ni Z, Hu Y, Liang W, Ou C, He J, et al. Clinical Characteristics of Coronavirus Disease 2019 in
China. New England Journal of Medicine. 2020; https://doi.org/10.1056/NEJMoa2002032
4. Higgins TS, Wu AW, Sharma D, Illing EA, Rubel K, Ting JY. Correlations of Online Search Engine
Trends With Coronavirus Disease (COVID-19) Incidence: Infodemiology Study. JMIR Public Health
and Surveillance. 2020; 6(2):e19702. https://doi.org/10.2196/19702 PMID: 32401211
5.
Li C, Chen LJ, Chen X, Zhang M, Pang CP, Chen H. Retrospective analysis of the possibility of predict-
ing the COVID-19 outbreak from Internet searches and social media data, China, 2020. Euro Surveil-
lance. 2020; 25(10). https://doi.org/10.2807/1560-7917.ES.2020.25.10.2000199
6. Yuan X, Xu J, Hussain S, Wang H, Gao N, Zhang L. Trends and Prediction in Daily New Cases and
Deaths of COVID-19 in the United States: An Internet Search-Interest Based Model. Exploratory
research and hypothesis in medicine. 2020; 5(2):1–6. https://doi.org/10.14218/ERHM.2020.00023
PMID: 32348380
7. Bannister-Tyrrell M, Meyer A, Faverjon C, Cameron A. Preliminary evidence that higher temperatures
are associated with lower incidence of COVID-19, for cases reported globally up to 29th February 2020.
medRxiv. 2020; https://doi.org/10.1101/2020.03.18.20036731
8. Demongeot J, Flet-Berliac Y, Seligmann H. Temperature Decreases Spread Parameters of the New
Covid-19 Case Dynamics. Biology. 2020; 9(5). https://doi.org/10.3390/biology9050094 PMID:
32375234
PLOS ONE | https://doi.org/10.1371/journal.pone.0237277 May 27, 2021
18 / 22
PLOS ONE Causal graph analysis for effects of determining factors on COVID-19 infections
9.
Liu J, Zhou J, Yao J, Zhang X, Li L, Xu X, et al. Impact of meteorological factors on the COVID-19 trans-
mission: A multi-city study in China. Science of the Total Environment. 2020; 726:138513. https://doi.
org/10.1016/j.scitotenv.2020.138513 PMID: 32304942
10. Greenland S, Robins JM, Pearl J. Confounding and Collapsibility in Causal Inference. Statistical Sci-
ence. 1999; 14(1):29–46. https://doi.org/10.1214/ss/1009211805
11. Schipf S, Knu¨ ppel S, Hardt J, Stang A. Directed Acyclic Graphs (DAGs)—Die Anwendung kausaler
Graphen in der Epidemiologie. Gesundheitswesen. 2011; 73(12):888–892. https://doi.org/10.1055/s-
0031-1291192 PMID: 22193898
12.
Textor J, van der Zander B, Gilthorpe MS, Liśkiewicz M, Ellison GT. Robust causal inference using
directed acyclic graphs: the R package ‘dagitty’. International Journal of Epidemiology. 2017; 45
(6):1887–1894. https://doi.org/10.1093/ije/dyw341
13. Center for Systems Science and Engineering (CSSE). COVID-19 Data Repository by the Center for
Systems Science and Engineering (CSSE) at Johns Hopkins University; 2020. Available from: https://
github.com/CSSEGISandData/COVID-19.
14. Pearl J, Bareinboim E. External Validity: From Do-Calculus to Transportability Across Populations. Sta-
tistical Science. 2014; 29(4):579–595. https://doi.org/10.1214/14-STS486
15. Chang MC, Kahn R, Li YA, Lee CS, Buckee CO, Chang HH. Variation in human mobility and its impact
on the risk of future COVID-19 outbreaks in Taiwan. BMC Public Health. 2021; 21(1):226. https://doi.
org/10.1186/s12889-021-10260-7 PMID: 33504339
16.
Fowler JH, Hill SJ, Obradovich N, Levin R. The Effect of Stay-at-Home Orders on COVID-19 Cases and
Fatalities in the United States. medRxiv. 2020; https://doi.org/10.1101/2020.04.13.20063628
17. Kraemer MUG, Yang CH, Gutierrez B, Wu CH, Klein B, Pigott DM, et al. The effect of human mobility
and control measures on the COVID-19 epidemic in China. Science (New York, NY). 2020; 368
(6490):493–497. https://doi.org/10.1126/science.abb4218
18.
19.
Lasry A, Kidder D, Hast M, Poovey J, Sunshine G, Winglee K, et al. Timing of Community Mitigation
and Changes in Reported COVID-19 and Community Mobility—Four U.S. Metropolitan Areas, February
26-April 1, 2020. MMWR Morbidity and mortality weekly report. 2020; 69(15):451—457. https://doi.org/
10.15585/mmwr.mm6915e2 PMID: 32298245
Linka K, Peirlinck M, Sahli Costabal F, Kuhl E. Outbreak dynamics of COVID-19 in Europe and the
effect of travel restrictions. Computer Methods in Biomechanics and Biomedical Engineering. 2020; p.
1–8. https://doi.org/10.1080/10255842.2020.1759560 PMID: 32367739
20. Mazzoli M, Mateo D, Hernando A, Meloni S, Ramasco JJ. Effects of mobility and multi-seeding on the
propagation of the COVID-19 in Spain. medRxiv. 2020; https://doi.org/10.1101/2020.05.09.20096339
21. Xiong C, Hu S, Yang M, Younes HN, Luo W, Ghader S, et al. Data-Driven Modeling Reveals the Impact
of Stay-at-Home Orders on Human Mobility during the COVID-19 Pandemic in the U.S. arXiv e-prints.
2020; p. arXiv:2005.00667.
22.
23.
Ju¨ ni P, Rothenbu¨ hler M, Bobos P, Thorpe KE, da Costa BR, Fisman DN, et al. Impact of climate and
public health interventions on the COVID-19 pandemic: A prospective cohort study. Canadian Medical
Association Journal. 2020; https://doi.org/10.1503/cmaj.200920 PMID: 32385067
Lai S, Ruktanonchai NW, Zhou L, Prosper O, Luo W, Floyd JR, et al. Effect of non-pharmaceutical inter-
ventions to contain COVID-19 in China. Nature. 2020; https://doi.org/10.1038/s41586-020-2293-x
PMID: 32365354
24. Google LLC. Google Trends, search term “corona”; 2020, accessed 2020-06-25. Available from:
https://www.google.com/trends.
25. Ayyoubzadeh SM, Ayyoubzadeh SM, Zahedi H, Ahmadi M, R Niakan Kalhori S. Predicting COVID-19
Incidence Through Analysis of Google Trends Data in Iran: Data Mining and Deep Learning Pilot Study.
JMIR Public Health and Surveillance. 2020; 6(2):e18828. https://doi.org/10.2196/18828 PMID:
32234709
26. Effenberger M, Kronbichler A, Shin JI, Mayer G, Tilg H, Perco P. Association of the COVID-19 pan-
demic with Internet Search Volumes: A Google Trends(TM) Analysis. International Journal of Infectious
Diseases. 2020; 95:192–197. https://doi.org/10.1016/j.ijid.2020.04.033 PMID: 32305520
27.
Lin YH, Liu CH, Chiu YC. Google searches for the keywords of “wash hands” predict the speed of
national spread of COVID-19 outbreak among 21 countries. Brain, Behavior, and Immunity. 2020;
https://doi.org/10.1016/j.bbi.2020.04.020 PMID: 32283286
28. Mavragani A. Tracking COVID-19 in Europe: Infodemiology Approach. JMIR Public Health and Surveil-
lance. 2020; 6(2):e18941. https://doi.org/10.2196/18941 PMID: 32250957
29. Walker A, Hopkins C, Surda P. Use of Google Trends to investigate loss-of-smell-related searches dur-
ing the COVID-19 outbreak. International Forum of Allergy & Rhinology. 2020; 10(7):839–847. https://
doi.org/10.1002/alr.22580 PMID: 32279437
PLOS ONE | https://doi.org/10.1371/journal.pone.0237277 May 27, 2021
19 / 22
PLOS ONE Causal graph analysis for effects of determining factors on COVID-19 infections
30.
Zhou WK, Wang AL, Xia F, Xiao YN, Tang SY. Effects of media reporting on mitigating spread of
COVID-19 in the early phase of the outbreak. Mathematical Biosciences and Engineering. 2020; 17
(3):2693–2707. https://doi.org/10.3934/mbe.2020147 PMID: 32233561
31. Qi H, Xiao S, Shi R, Ward MP, Chen Y, Tu W, et al. COVID-19 transmission in Mainland China is associ-
ated with temperature and humidity: A time-series analysis. Science of the Total Environment. 2020;
728:138778. https://doi.org/10.1016/j.scitotenv.2020.138778 PMID: 32335405
32. Shi P, Dong Y, Yan H, Zhao C, Li X, Liu W, et al. Impact of temperature on the dynamics of the COVID-
19 outbreak in China. Science of the Total Environment. 2020; 728:138890. https://doi.org/10.1016/j.
scitotenv.2020.138890 PMID: 32339844
33. Sobral MFF, Duarte GB, da Penha Sobral AIG, Marinho MLM, de Souza Melo A. Association between
climate variables and global transmission of SARS-CoV-2. Science of the Total Environment. 2020;
729:138997. https://doi.org/10.1016/j.scitotenv.2020.138997 PMID: 32353724
34.
Tosepu R, Gunawan J, Effendy DS, Ahmad LOAI, Lestari H, Bahar H, et al. Correlation between
weather and Covid-19 pandemic in Jakarta, Indonesia. Science of the Total Environment. 2020;
725:138436. https://doi.org/10.1016/j.scitotenv.2020.138436 PMID: 32298883
35. Wang M, Jiang A, Gong L, Luo L, Guo W, Li C, et al. Temperature significant change COVID-19 Trans-
mission in 429 cities. medRxiv. 2020; https://doi.org/10.1101/2020.02.22.20025791
36. Wu Y, Jing W, Liu J, Ma Q, Yuan J, Wang Y, et al. Effects of temperature and humidity on the daily new
cases and new deaths of COVID-19 in 166 countries. Science of the Total Environment. 2020;
729:139051. https://doi.org/10.1016/j.scitotenv.2020.139051 PMID: 32361460
37. Auler AC, Ca´ ssaro FAM, da Silva VO, Pires LF. Evidence that high temperatures and intermediate rela-
tive humidity might favor the spread of COVID-19 in tropical climate: A case study for the most affected
Brazilian cities. Science of the Total Environment. 2020; 729:139090. https://doi.org/10.1016/j.
scitotenv.2020.139090 PMID: 32388137
38. Xie J, Zhu Y. Association between ambient temperature and COVID-19 infection in 122 cities from
China. Science of the Total Environment. 2020; 724:138201. https://doi.org/10.1016/j.scitotenv.2020.
138201 PMID: 32408450
39. Briz-Redo´n A´ , Serrano-Aroca A´ . A spatio-temporal analysis for exploring the effect of temperature on
COVID-19 early evolution in Spain. Science of the Total Environment. 2020; 728:138811. https://doi.
org/10.1016/j.scitotenv.2020.138811 PMID: 32361118
40.
41.
Iqbal N, Fareed Z, Shahzad F, He X, Shahzad U, Lina M. The nexus between COVID-19, temperature
and exchange rate in Wuhan city: New findings from partial and multiple wavelet coherence. Science of
the Total Environment. 2020; 729:138916. https://doi.org/10.1016/j.scitotenv.2020.138916 PMID:
32388129
Jahangiri M, Jahangiri M, Najafgholipour M. The sensitivity and specificity analyses of ambient tempera-
ture and population size on the transmission rate of the novel coronavirus (COVID-19) in different prov-
inces of Iran. Science of the Total Environment. 2020; 728:138872. https://doi.org/10.1016/j.scitotenv.
2020.138872 PMID: 32335407
42. Yao Y, Pan J, Liu Z, Meng X, Wang W, Kan H, et al. No association of COVID-19 transmission with tem-
perature or UV radiation in Chinese cities. The European Respiratory Journal. 2020; 55(5). https://doi.
org/10.1183/13993003.00517-2020
43.
de Lusignan S, Dorward J, Correa A, Jones N, Akinyemi O, Amirthalingam G, et al. Risk factors for
SARS-CoV-2 among patients in the Oxford Royal College of General Practitioners Research and Sur-
veillance Centre primary care network: a cross-sectional study. The Lancet Infectious Diseases; https://
doi.org/10.1016/S1473-3099(20)30371-6
44. Wahrendorf M, Rupprecht CJ, Dortmann O, Scheider M, Dragano N. Erho¨ htes Risiko eines COVID-19-
bedingten Krankenhausaufenthaltes fu¨ r Arbeitslose: Eine Analyse von Krankenkassendaten von 1,28
Mio. Versicherten in Deutschland. Bundesgesundheitsblatt—Gesundheitsforschung—Gesund-
heitsschutz. 2021; 64(3):314–321. https://doi.org/10.1007/s00103-021-03280-6
45. Dohle S, Wingen T, Schreiber M. Acceptance and Adoption of Protective Measures During the COVID-
19 Pandemic: The Role of Trust in Politics and Trust in Science. Social Psychological Bulletin. 2020; 15
(4):1–23. https://doi.org/10.32872/spb.4315
46. Engle S, Stromme J, Zhou A. Staying at home: mobility effects of COVID-19. Available at SSRN. 2020;
https://doi.org/10.2139/ssrn.3565703
47. Cowling BJ, Ali ST, Ng TWY, Tsang TK, Li JCM, Fong MW, et al. Impact assessment of non-pharma-
ceutical interventions against coronavirus disease 2019 and influenza in Hong Kong: an observational
study. The Lancet Public Health. 2020; 5(5):e279–e288. https://doi.org/10.1016/S2468-2667(20)
30090-6 PMID: 32311320
48. Robert Koch-Institut (RKI). Fallzahlen in Deutschland (COVID-19); 2020, accessed 2020-07-12. Avail-
able from: https://www.rki.de/DE/Content/InfAZ/N/Neuartiges_Coronavirus/Fallzahlen.html.
PLOS ONE | https://doi.org/10.1371/journal.pone.0237277 May 27, 2021
20 / 22
PLOS ONE Causal graph analysis for effects of determining factors on COVID-19 infections
49. Google LLC. Google COVID-19 Community Mobility Reports; 2020, accessed 2020-06-25. Available
from: https://www.google.com/covid19/mobility/.
50. Deutscher Wetterdienst (DWD) Climate Data Center (CDC). Recent daily station observations (temper-
ature, pressure, precipitation, sunshine duration, etc.) for Germany, quality control not completed yet,
version recent; 2020, accessed 2020-07-12. Available from: https://opendata.dwd.de/climate_
environment/CDC/observations_germany/climate/daily/kl/recent/.
51. Bundesinstitut fu¨ r Bau-, Stadt- und Raumforschung (BBSR). INKAR—Indikatoren und Karten zur
Raum- und Stadtentwicklung; 2020, accessed 2020-06-25. Available from: https://www.inkar.de/.
52. Mitze T, Kosfeld R, Rode J, Wa¨ lde K. Face Masks Considerably Reduce COVID-19 Cases in Germany.
Proceedings of the National Academy of Sciences. 2020; 117(51):32293–32301. https://doi.org/10.
1073/pnas.2015954117 PMID: 33273115
53. Spirtes P, Glymour CN, Scheines R, Heckerman D. Causation, prediction, and search. MIT Press;
2000.
54. Pearl J. Causality. Cambridge: Cambridge University Press; 2009. Available from: https://www.
cambridge.org/core/books/causality/B0046844FAE10CBF274D4ACBDAEB5F5B.
55. Greenland S, Pearl J, Robins JM. Causal Diagrams for Epidemiologic Research. Epidemiology. 1999;
10(1):37–48. https://doi.org/10.1097/00001648-199901000-00008 PMID: 9888278
56. Henckel L, Perković E, Maathuis MH. Graphical Criteria for Efficient Total Effect Estimation via Adjust-
ment in Causal Linear Models. arXiv e-prints. 2020; p. arXiv:1907.02435.
57. R Core Team. R: A Language and Environment for Statistical Computing; 2019. Available from: https://
www.R-project.org/.
58. Kalisch M, Ma¨chler M, Colombo D, Maathuis MH, Bu¨hlmann P. Causal Inference Using Graphical Mod-
els with the R Package pcalg. Journal of Statistical Software. 2012; 47(11):1–26. https://doi.org/10.
18637/jss.v047.i11
59. Hilbe JM, Greene WH. 4—Count Response Regression Models. In: Rao CR, Miller JP, Rao DC, editors.
Essential Statistical Methods for Medical Statistics. Boston: North-Holland; 2011. p. 104–145. Avail-
able from: http://www.sciencedirect.com/science/article/pii/B9780444537379500074.
60. Venables WN, Ripley BD. Modern Applied Statistics with S. 4th ed. New York: Springer; 2002. Avail-
able from: http://www.stats.ox.ac.uk/pub/MASS4.
61. Hoerl AE, Kennard RW. Ridge Regression: Biased Estimation for Nonorthogonal Problems. Techno-
metrics. 1970; 12(1):55–67. https://doi.org/10.1080/00401706.1970.10488634
62.
Friedman J, Hastie T, Tibshirani R. Regularization Paths for Generalized Linear Models via Coordinate
Descent. Journal of Statistical Software. 2010; 33(1):1–22. https://doi.org/10.18637/jss.v033.i01 PMID:
20808728
63. Perković E, Textor J, Kalisch M, Maathuis MH. A Complete Generalized Adjustment Criterion. arXiv e-
prints. 2015; p. arXiv:1507.01524.
64. Gencoglu O, Gruber M. Causal Modeling of Twitter Activity During COVID-19. Computation. 2020; 8(4).
https://doi.org/10.3390/computation8040085
65. Wachtler B, Michalski N, Nowossadeck E, Diercke M, Wahrendorf M, Santos-Ho¨vener C, et al. Socio-
economic inequalities and COVID-19—A review of the current international literature. Journal of Health
Monitoring. 2020;(S7):3–17. https://doi.org/10.25646/7059
66. Hoebel J, Michalski N, Wachtler B, Diercke M, Neuhauser H, Wieler LH, et al. Sozioo¨ konomische
Unterschiede im Infektionsrisiko wa¨ hrend der zweiten SARS-CoV-2-Welle in Deutschland. Dtsch Arz-
tebl International. 2021; 118(15):269–270. https://doi.org/10.3238/arztebl.m2021.0188
67. Perković E, Textor J, Kalisch M, Maathuis MH. Complete graphical characterization and construction of
adjustment sets in Markov equivalence classes of ancestral graphs. The Journal of Machine Learning
Research. 2017; 18(1):8132–8193.
68.
Islam N, Sharp SJ, Chowell G, Shabnam S, Kawachi I, Lacey B, et al. Physical distancing interventions
and incidence of coronavirus disease 2019: natural experiment in 149 countries. BMJ. 2020; 370.
https://doi.org/10.1136/bmj.m2743 PMID: 32669358
69. Kermack WO, McKendrick AG. Contributions to the mathematical theory of epidemics–I. 1927. Bulletin
of mathematical biology. 1991; 53(1-2):33—55. https://doi.org/10.1007/bf02464423
70.
an der Heiden M, Buchholz U. Modellierung von Beispielszenarien der SARS-CoV-2-Epidemie 2020 in
Deutschland. 2020; https://doi.org/10.25646/6571.2
71. Schisterman EF, Perkins NJ, Mumford SL, Ahrens KA, Mitchell EM. Collinearity and Causal Diagrams:
A Lesson on the Importance of Model Specification. Epidemiology. 2017; 28(1). https://doi.org/10.1097/
EDE.0000000000000554 PMID: 27676260
PLOS ONE | https://doi.org/10.1371/journal.pone.0237277 May 27, 2021
21 / 22
PLOS ONE Causal graph analysis for effects of determining factors on COVID-19 infections
72. Openshaw S. Ecological Fallacies and the Analysis of Areal Census Data. Environment and Planning
A: Economy and Space. 1984; 16(1):17–31. https://doi.org/10.1068/a160017 PMID: 12265900
73. Solar O, Irwin A. A conceptual framework for action on the social determinants of health. WHO Docu-
ment Production Services; 2010. Available from: https://drum.lib.umd.edu/handle/1903/23135.
74. World Health Organization, et al. Transmission of SARS-CoV-2: implications for infection prevention
precautions: Scientific Brief, 09 July 2020. World Health Organization; 2020.
PLOS ONE | https://doi.org/10.1371/journal.pone.0237277 May 27, 2021
22 / 22
PLOS ONE |
10.1371_journal.pone.0228134 | RESEARCH ARTICLE
Proteomic profiling of proteins in the dorsal
horn of the spinal cord in dairy cows with
chronic lameness
Daniel HerzbergID
Marianne Werner2, Hedie Bustamante1*
1*, Pablo Strobel2, Heine Mu¨ ller1, Constanza Meneses3,
1 Veterinary Clinical Sciences Department, Faculty of Veterinary Science, Universidad Austral de Chile,
Valdivia, Chile, 2 Animal Science Department, Faculty of Veterinary Science, Universidad Austral de Chile,
Valdivia, Chile, 3 Comparative Biomedical Science Graduate Program, College of Veterinary Medicine, North
Caroline State University, Raleigh, North Carolina, United States of America
* hbustamante@uach.cl (HB); danielherzberg@gmail.com (DH)
Abstract
Chronic lameness affects bovine welfare and has a negative economic impact in dairy
industry. Moreover, due to the translational gap between traditional pain models and new
drugs development for treating chronic pain states, naturally occurring painful diseases
could be a potential translational tool for chronic pain research. We therefore employed liq-
uid chromatography tandem mass spectrometry (LC-MS/MS) to stablish the proteomic pro-
file of the spinal cord samples from lumbar segments (L2-L4) of chronic lame dairy cows.
Data were validated and quantified through software tool (Scaffold® v 4.0) using output data
from two search engines (SEQUEST® and X-Tandem®). Search Tool for the Retrieval of
Interacting Genes/Proteins (STRING) analysis was performed to detect proteins interac-
tions. LC-MS/MS identified a total amount of 177 proteins; of which 129 proteins were able
to be quantified. Lame cows showed a strong upregulation of interacting proteins with chap-
erone and stress functions such as Hsp70 (p < 0.006), Hsc70 (p < 0.0079), Hsp90 (p <
0.015), STIP (p > 0.0018) and Grp78 (p <0.0068), and interacting proteins associated to gly-
colytic pathway such as; γ-enolase (p < 0.0095), α-enolase (p < 0.013) and hexokinase-1 (p
< 0.028). It was not possible to establish a clear network of interaction in several upregulated
proteins in lame cows. Non-interacting proteins were mainly associated to redox process
and cytoskeletal organization. The most relevant down regulated protein in lame cows was
myelin basic protein (MBP) (p < 0.02). Chronic inflammatory lameness in cows is associated
to increased expression of stress proteins with chaperone, metabolism, redox and structural
functions. A state of endoplasmic reticulum stress and unfolded protein response (UPR)
might explain the changes in protein expression in lame cows; however, further studies
need to be performed in order to confirm these findings.
Introduction
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: Herzberg D, Strobel P, Mu¨ller H, Meneses
C, Werner M, Bustamante H (2020) Proteomic
profiling of proteins in the dorsal horn of the spinal
cord in dairy cows with chronic lameness. PLoS
ONE 15(1): e0228134. https://doi.org/10.1371/
journal.pone.0228134
Editor: Jon M. Jacobs, Pacific Northwest National
Laboratory, UNITED STATES
Received: May 2, 2019
Accepted: January 8, 2020
Published: January 28, 2020
Copyright: © 2020 Herzberg et al. This is an open
access article distributed under the terms of the
Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: All relevant data are
within the paper and its Supporting Information
files.
Funding: This work was supported by Conicyt/
Scholarship Nº21150613. The funders had no role
in study design, data collection and analysis,
decision to publish, or preparation of the
manuscript.
Competing interests: The authors have declared
that no competing interest exist.
Chronic pain represents a dysfunction of the nervous system [1, 2], and similar to humans,
this type of pain impact negatively the quality of life of affected animals [3]. Naturally
PLOS ONE | https://doi.org/10.1371/journal.pone.0228134 January 28, 2020
1 / 18
Spinal proteomic profile of chronic lame dairy cows
occurring painful diseases in animals may represent an alternative approach to investigate
nociceptive mechanisms involved in chronic pain [4, 5]. Painful lameness in dairy cows is
common, and one of the most important causes of economic losses for the dairy industry [6]
due to increasing culling [7]https://paperpile.com/c/QlNkzH/hxba, reduced milk production
and reproductive performance [8]. https://paperpile.com/c/QlNkzH/Hj7c+4D5eLameness is a
multifactorial condition and its prevalence has been associated with housing and nutritional
management [9]https://paperpile.com/c/QlNkzH/Cn8P+388y+UI6B+9Lth, environmental
factors [10]https://paperpile.com/c/QlNkzH/6tmv, metabolic status [11], inadequate claw
trimming [9], gestation and stage of lactation [11]. The characteristic features of chronic lame-
ness associated-pain in dairy cows make this condition a potential new translational model for
the study of chronic pain.
Within the damaged tissue, inflammatory mediators released from immune cells can
reduce the threshold of somatic and visceral primary afferent nociceptors, leading to periph-
eral sensitization [12]https://paperpile.com/c/QlNkzH/2xFo+o6FS. Prostanoids [13], kinins
[14], growth factors [15], chemokines [16], cytokines [17], protons and ATP [18] can promote
peripheral sensitization, activating multiple intracellular transduction signals that lead to an
increase in membrane receptor expression (i.e., TRPV1, TRPVA1, Nav1.7, Nav1.9, among oth-
ers). Moreover, persistent peripheral sensitization can potentially induce a state of central sen-
sitization [12]. Central sensitization is the facilitation of synaptic transmission in central
nervous system (CNS), which often turns into maladaptive and prolonged molecular changes
in the nociceptive pathway [1]. This increased sensitivity to noxious and harmful stimuli will
create pain behaviors known as hyperalgesia and allodynia, respectively [19]. It has been previ-
ously demonstrated that reduced nociceptive umbral thresholds can be detected in cattle with
mild to severe lameness associated to chronic inflammatory lesions of the hoof [20].
Two temporal phases can be distinguished during central sensitization: an early phase
dependent on kinase activation that results from rapid changes in the glutamate receptor and
ion channels function [21, 22], and a later and long-lasting phase in which transcriptional and
translational events drive the synthesis of proteins responsible for persistent pain [1, 23].
Proteins expression can be determined through proteomic analysis [24]. Proteomic tech-
niques have been used in order to investigate the dynamics of protein expression under patho-
logical pain states, with an increased potential for identification of pain biomarkers [25]. Most
of the proteomic studies focusing on pain have been performed either in the spinal cord or in
the dorsal root ganglion (DRG) of rodents using experimental pain models [26]. Recently,
increasing concern about the translational impact of basic science research in the development
of new drugs has been discussed [25]. According to Mao (2012) [27], the time frame of pain in
experimental models may not adequately reproduce the impact of prolonged nociception of
clinical pain. Moreover, few proteomic studies have focused on naturally occurring pathologi-
cal pain. Recently, the cerebrospinal fluid, serum and plasma proteome of human patients
with neuropathic pain, rheumatoid arthritis and widespread back pain has been described [28,
29, 30].
The aim of this study was to describe the proteomic profile in the dorsal horn spinal cord of
cows with chronic inflammatory lameness.
Materials and methods
Bioethics statement
The experimental protocol was approved by the Ethics Committee of Animal Research of the
Universidad Austral de Chile (resolution number 323/2018).
PLOS ONE | https://doi.org/10.1371/journal.pone.0228134 January 28, 2020
2 / 18
Spinal proteomic profile of chronic lame dairy cows
Animals
Twelve dairy cows were selected from a commercial dairy farm (Agricola Los Rı´os, single live-
stock role 10.5.07.0760). All selected animals were Kiwi cross and 2 years of age or older, origi-
nating in different herds within the same farm with similar breeding, feeding and other
routine practices. Spinal cord samples from 5 lame dairy cows (Lame group; n = 5) with a hind
limb lameness history of at least 5 months were obtained after euthanasia via intravenous gen-
eral anesthesia and intrathecal lidocaine injection at the atlanto-occipital foramen as previ-
ously reported. Similarly, spinal cord samples from 7 non-lame animals (Control Group;
n = 7) without apparent lameness and without history of previous lameness episodes were
selected from a commercial slaughterhouse (Frigorifico Balmaceda SPA, single livestock role
10.512.0882) after euthanasia by mechanical stunning and exsanguination according to
national regulations. Researchers were not involved in the decision for euthanasia or
slaughter.
Lameness assessment
Given that bovine lameness is a multifactorial condition, cows were selected considering the
most prevalent causes of lameness in Southern Chile, including white line disease, sole hemor-
rhage, sole ulcer and digital dermatitis [31]. Lameness was confirmed and classified according
to the mobility score previously described by Reader et al. (2011) [32]. Briefly, cows were clas-
sified into two groups (Lame or Control), and lameness was scored as follow: MS 0 not lame;
MS 1 imperfect mobility/uneven; MS 2 impaired mobility/mildly lame; and MS 3 severely
impaired mobility/very lame. Lame group consisted only of cows with a MS of 3. Exclusion cri-
teria for both groups included the presence of visible acute wounds, visible neurological gait
alteration (central or peripheral ataxia) and acute or chronic mastitis.
Spinal cord sampling and protein extraction
Lumbar spinal cord sections (L2-L4) were aseptically obtained post-mortem after removal of
the dorsal aspect of lumbar vertebrae. A 20 cm segment was obtained from each animal. Dura
mater and arachnoids meninges were gentle dissected and after carefully washing the tissue
with cold phosphate buffered saline (PBS) samples were snap frozen in liquid nitrogen and
transported to laboratory for further processing. Several segments of approximately 300 mg of
the ipsilateral dorsal horn were stored in a mixture of 1 mL of PBS and 10 uL of protease inhib-
itor. For protein extraction, samples were sonicated three times for 30 seconds each and then
centrifuged at 20,000 x g for 20 min at 4˚C. The supernatant was collected and stored at -80˚C
until further analysis. Total protein quantification was performed using the BCA Assay (Pier-
ceTM Thermo Scientific, Rochford USA) according to the manufacturer’s instructions. Protein
extraction was evaluated using a 5–12% sodium dodecyl sulfate polyacrylamide gel electropho-
resis (SDS-PAGE) stained with Coomassie blue.
Sample preparation for proteomics analysis
100 μg of protein were lyophilized for 120 minutes and then re-solubilized in 6M guanidine
hydrochloride and 25 mM NH4HCO3 at pH 7.5 during 60 minutes. Subsequently, proteins
were reduced at room temperature using 2 mM dithiothreitol and alkylated with 10 mM
iodoacetamide at room temperature for 60 minutes. These reactions were then diluted seven
times with 25 mM NH4HCO3 at pH 7.5. Modified trypsin 1:50 (Promega, Madison, WI, USA)
was added, and the reaction was incubated at 37˚C for 16 hours. Trypsin digestion was stopped
adding 5% formic acid, pH 3.0.
PLOS ONE | https://doi.org/10.1371/journal.pone.0228134 January 28, 2020
3 / 18
Spinal proteomic profile of chronic lame dairy cows
Protein identification by tandem mass spectrometry (LC-MS/MS)
Protein samples were concentrated using a CentriVap concentrator (Labconco, Kansas City,
MO, USA) to a final volume of 20 μL and loaded into a 350 μm internal diameter (ID) fused
silica peptide trap column with 3 cm of C18 reverse-phase desalting resin. Immediately after
washing with 0.1% formic acid for 30 min at 0.5 μL/min, the efflux of the peptide trap column
was directed to a 10 cm resolving reversed-phase column (100 Å, 5 μm Magic C18 particles,
Michrom Bioresources), which was mounted on the electrospray stage of a VELOSPRO mass
spectrometer (LTQ, Thermo Finnigan LLC) by a 0–90% acetonitrile gradient for 120 min at a
flow rate of 600 nL/min. An electrospray voltage of 2.2 kV with the ion transfer temperature
set to 270ºC was used. The mass spectrometer was controlled by the Xcalibur software, which
continuously performed mass-scan analysis of the six most intense ions during MS/MS scans
of the ion traps. For this, one repeat scan of the same ion was dynamically excluded, using a 30
sec repeat duration and 90 s exclusion duration. Normalized collision energy for the MS/MS
was set to 35%.
Database search algorithm and protein identification criteria. SEQUEST (Thermo
Fisher Scientific, San Jose, CA, USA) and X!Tandem software (www.thegpm.org) were used to
analyse all tandem mass spectra MS/MS samples. SEQUEST was set up to search Bostaur-
os_NCBI_250722016.fasta (60.090 entries) with the inclusion of trypsin. Similarly, X!Tandem
was set up to search a reverse concatenated subset of the uniprot-bos+taurus database (200
entries) also assuming trypsin digestion. Both databases were searched with a fragment ion
mass tolerance of 0.80 Da and a parent ion tolerance of 2.5 Da. Carbamidomethyl of cysteine
was specified in SEQUEST and X!Tandem as a fixed modification. Deamidated of asparagine
and glutamine and oxidation of methionine were specified in SEQUEST as variable modifica-
tion. Glu!pyro-Glu of the n-terminus, ammonia-loss of the n-terminus, gln!pyro-Glu of
the n-terminus, asparagine and glutamine deamidation, methionine oxidation were specified
in X!Tandem as variable modifications.
In order to validate MS/MS peptide and protein identification, data were loaded into Scaf-
fold v.4.8.6 (Proteome Software Inc., Portland, OR). Peptide and Protein probabilities were
assigned using the peptide/protein prophet algorithm [33]. Peptide and protein criteria of
identification included a 95% confidence in the peptide/protein prophet algorithm, with a
minimum of 4 identified peptideshttps://paperpile.com/c/QlNkzH/LDZX [34]https://
paperpile.com/c/QlNkzH/oLuB. Proteins that contained similar peptides and could not be dif-
ferentiated based on MS/MS analysis were grouped to satisfy the principles of parsimony. Pro-
teins sharing significant peptide evidence were grouped into clusters. Proteins were there
annotated with gene ontology (GO) terms from goa uniprot_all.gaf (downloaded Jan 12, 2017)
[35]. Unique peptide counts are included as supplemental information (Supp 1).
Additionally, a search tool for the retrieval of interacting genes/proteins (STRING) neigh-
borhood analysis was performed on all identified proteins by searching the STRING database
to detect possible interactions (https://string-db.org). Protein names were obtained by match-
ing the gen-info identifier number (GI) of the proteins to the UniProtKB database (https://
www.uniprot.org). STRING was set to identify the evidence of the type of interaction, selecting
database and experiments as active interaction sources and a minimum required interaction
score of 0.900 [36].s
Western blot analysis
Western blot analysis was performed in order to validate some proteins of interest detected in
proteomics. Protein extraction was performed using the same method described previously for
LC-MS/MS. Proteins were quantified by BSA Assay (PierceTM Thermo Scientific Rockford
PLOS ONE | https://doi.org/10.1371/journal.pone.0228134 January 28, 2020
4 / 18
Spinal proteomic profile of chronic lame dairy cows
USA) and twenty microgram of each sample was separated by SDS-PAGE and transferred to a
0.45 μm nitrocellulose membrane (Amersham™ Potan™ GE Healthcare Germany). Membranes
were blocking during 1 hour with TBS (10 mM Tris-HCl pH 7,6) and 5% BSA at room temper-
ature. Membranes were incubated for 16 h at 4˚C with the following antibodies: mouse mono-
clonal anti gamma-enolase-enolase (1:500 w/v, Santa Cruz Biotechnology), mouse monoclonal
anti-Hsp70 (1:500 w/v, Santa Cruz Biotechnology) and mouse monoclonal anti-β-actin (1:500
w/v, Santa Cruz Biotechnology) as loading control. Following primary antibody incubation,
membranes were washed three times (10 minutes each) with TBS and 0.1% Tween-20, and
incubated during 60 minutes at room temperature with anti-mouse IgG HRP-secondary anti-
body (1:2000 v/w, Santa Cruz Biotechnology). Following secondary antibody reaction mem-
branes were washed three times for 10 minutes each. The intensity of the bans were visualized
with enhanced chemiluminescence (Pierce ECL Thermo Scientific Rockford USA) and cap-
tured by Odyssey Imaging System (LI-COR Bioscience, Lincoln, NE). For quantification of
blot intensities Image Studio Lite ver 5.2.5 software (LI-COR Bioscience, Lincoln, NE) was
used.
Statistical analysis
Label free total spectra counting and total abundance of proteins for lame and control cows
were tested for normality using the Kolmogorov-Smirnov test and then compared between
groups using the t-test. Band intensities, expressed as mean ±SEM, were tested for normality
using the Kolmogorov-Smirnov test and then compared between groups using the t-test. Dif-
ferences were considered significant when p � 0.05. All statistical tests were performed on
Graphpad Prism 7.0.
Results
Tandem mass spectrometry analysis was able to identify a total amount of 177 proteins in the
dorsal horn of the spinal cord. Proteins with significant peptide evidence were grouped into 47
clusters. From the totality of identified proteins, Scaffold software was able to quantify 129 of
them. Quantitative analysis showed that 10 proteins were significantly downregulated, and 27
proteins were significantly upregulated in cows with chronic lameness compared to controls
(Tables 1 and 2). Moreover, 26 proteins were only detected in the dorsal horn of the spinal
cord of lame cows (Table 2).
Gene ontology (GO) annotations were performed in order to obtain an overall biological
and functional background of the quantified proteins. Sequence distribution by GO level is
shown in Fig 1. GO analysis indicate that proteins were distributed by activity as binding
(39.5%), catalytic (32.4%) followed by antioxidant (3.4%) structural (2.2%), carrier (2.2%) and
transport (1.1%). Regarding biological process distribution, cellular processes (45.2%), meta-
bolic processes (32.1%) and biological regulation (31.1%) were the most frequent annotated
distributions. Gene ontology for cellular localization showed that proteins were primary dis-
tributed in cell parts (42,2%) and organelles (31,2%), followed by protein-containing complex
(17.9%), membranes (11.6%), membrane enclosed lumen (10.5%), with few sequences distrib-
uted at the extracellular region (8.4%), cell junction (2.1%) and nucleoid (1.1%) levels (Fig 1).
The identified and quantified proteins were subjected to STRING interaction analysis. Out
of 129 proteins, 121 were successfully recognized in the STRING database and represented as a
network of proteins connected with evidence-based edges. The resulting network shows four
evident clusters in which one protein is connected to at least three other proteins with highest
confidence bond (Fig 2). One cluster consisted of 10 proteins identified as chaperones and co-
chaperones from which 7 were only detected in lame cows (Fig 2A). A second cluster consisted
PLOS ONE | https://doi.org/10.1371/journal.pone.0228134 January 28, 2020
5 / 18
Spinal proteomic profile of chronic lame dairy cows
Table 1. Downregulated proteins detected in the spinal cord of lame cows.
Proteins
Accesion (GI)
MW (kDa)
Gene Name
p-value
Fold Change
Albumin
ALBU_BOVIN
Dihydropyrimidinase-related protein 2
Glyceraldehyde 3-phosphate Dehydrogenase
Hemoglobin beta
Hemoglobin beta Bali
Hemoglobin chain C
Hemoglobin fetal subunit beta
Hemoglobin subunit alpha
Myelin basic protein
Ubiquitin carboxyl-terminal hydrolase
https://doi.org/10.1371/journal.pone.0228134.t001
528959240
694270100
294459577
223864
97724899
62460494
359061887
741972060
528952847
69
65
38
16
15
15
33
25
ALB
DPYSL2
GAPD
HBB
HBG
HBA
MBP
UCHL1
0,0033
0,0013
0,036
0,00016
0,00028
0,001
0,00013
0,0022
0,02
0,0018
0,3
0,6
0,7
0,2
0,3
0,5
0,3
0,3
0,7
0,4
of 16 proteins associated to glycolysis, gluconeogenesis and the pentose phosphate pathway
(Fig 2B), from which 5 of them were upregulated in lame cows (Table 3). Also, a third cluster
(Fig 2C) consisting of mitochondrial proteins and a fourth cluster (Fig 2D), consisting of tubu-
lin isoforms showed no differences between lame and non-lame cows. A marked interaction
between clusters A and D (Fig 2) could be observed. The Kyoto Encyclopedia of Genes and
Genome (KEGG) pathways most representative in the interacting network are depicted in
Table 3. The most notorious finding is that antigen processing and presentation pathway
(KEGG pathway ID 0416) was only integrated by proteins strongly upregulated in lame cows.
Validation of LC-MS/MS results was performed by western blot analysis of one protein
from the chaperone cluster (Hsp70) and one protein from the glycolytic enzymes cluster
(gamma-enolase) (Fig 3).
Discussion
In this study we describe the differential expression of proteins in the dorsal horn of the spinal
cord of cows with chronic painful inflammatory lameness using a proteomic analysis. Peptide
spectra were identified using SEQUEST search engine and validated through X!Tandem using
different modification parameter in order avoid identification redundancy. The totality of the
lame dairy cows included in our study had a history of lameness of more than 5 months with
notorious and evident hyperalgesia and allodynia of the affected limb during the mobility
score evaluation. Lame cows had a higher number of expressed proteins in the dorsal horn of
the spinal cord compared to sound cows. Different proteomic profile studies in dorsal root
ganglia and spinal cord in various animal models of pain have previously reported an
increased number of upregulated proteins compared to downregulated proteins [37].
We describe an increased amount of stress-associated proteins in the spinal cord of lame
cows. A 20-time fold change in the constitutive Hsc70 (HSPA8) protein was observed in lame
cows, with the inducible forms of Hsp70 (HSPA1L) and Hsp90 (HSPA90AA1) only detected
in lame cows. Similar to our findings, several proteomic profiles in chronic neuropathic pain
in rodents have reported increasing levels of heat shock proteins (HSP´s) after injury in both
the peripheral and central nervous system [37, 38, 39]. Additionally, upregulation of HSP’s has
been reported after experimental lower lumbar and sacral terminal nerve compression in dogs
[40]. HSP’s are a family of intracellular chaperones that bind to different proteins facilitating
their folding into their native and active conformation [41]. In response to cellular stress, upre-
gulation of HSP’s help to protect cells from abnormal protein aggregation, thus preventing cell
death [42, 43]. However, newer roles of HSP´s have been described recently, including antigen
presentation, immune cell activation and inflammation [44]. Furthermore, neurons, glial and
PLOS ONE | https://doi.org/10.1371/journal.pone.0228134 January 28, 2020
6 / 18
Table 2. Upregulated proteins detected in the spinal cord of lame cows.
Proteins
Accesion (GI)
MW kDa
Gene Name
p value
Spinal proteomic profile of chronic lame dairy cows
4-aminobutyrate aminotransferase, mitochondria
4-trimethylaminobutyraldehyde dehydrogenase
78 kDa glucose-regulated protein
Acyl-CoA-binding protein
Cdc42
Acetyl-CoA acetyltransferase, mitochondrial
Aconitate hydratase
AHNAK2
Aldehyde dehydrogenase, mitochondrial
Aldose reductase
Alpha-aminoadipic semialdehyde dehydrogenase
Arylsulfatase B precursor
Aspartate aminotransferase, cytoplasmic
Aspartate aminotransferase, mitochondrial
ATP-citrate synthase
Dihydrolipoyl dehydrogenase, mitochondria
Dihydropyrimidinase-related protein 1
Dihydropyrimidinase-related protein 3
Enolase 1
Enolase 2
Enolase 3
Fascin
Galectin-1
Glutathione S-transferase P
Heat shock 70 kDa protein 1
Heat shock 70 kDa protein 4
Heat shock 70 kDa protein 6
Heat shock cognate 71 kDa protein
Heat shock protein HSP 90-alpha
Heat shock-related 70 kDa protein 2
Hexokinase-1
Hyaluronan and proteoglycan link protein 2
L-isoaspartate(D-aspartate) O-methyltransferase
Microtubule-associated protein 1A
Microtubule-associated protein 1B
N(G),N(G)-dimethylarginine dimethylaminohydrolase
Peroxiredoxin-1
Peroxiredoxin-4
Phosphoglucomutase-1
Pyridoxal phosphate phosphatase
Rab GDP dissociation inhibitor alpha
Rab GDP dissociation inhibitor beta
retinal dehydrogenase 1
Stress-70 protein, mitochondrial
Stress-induced-phosphoprotein 1
Superoxide dismutase [Cu-Zn]
Tenascin
125991950
114051782
115495027
ACBP_BOVIN
7245833
114050959
27806769
983004191
115496214
113594
296485604
155372077
29135295
27807377
82697335
329663954
741930532
155371867
296479148
528950986
77736349
78045491
999589
29135329
529003643
166795319
297472417
296480084
60592792
296482938
60592784
528942294
296483921
741967576
329663571
156121049
296488840
27806085
116004023
78045487
27806617
76253900
27806321
77735995
296471478
SODC_BOVIN
56
54
78
10
25
45
85
ABAT
ALDH9A1
GRP78
DBI
CDC42
ACAT1
ACO2
186
AHNAK2
57
36
59
59
?
48
17
54
74
74
47
47
47
58
14
24
72
70
70
70
90
70
102
38
25
336
330
31
22
31
62
32
51
50
55
74
63
16
ALDH2
AKR1B1
ALDH7A1
ARSB
GOT2
GOT1
ACLY
DLD
CRMP1
DPYSL3
ENO1
ENO2
ENO3
FSCN1
LGALS1
GSTP1
HSPA1L
HSPA4
HSPA6
HSPA8
HSP90AA1
HSPA2
HK1
HAPLN2
PCMT1
MAP1A
MAP1B
DDAH1
PRDX1
PRDX4
PGM1
PDXP
GDI1
GDI2
ALDH1A1
HSPA9
STIP1
SOD1
TNC
528959916
260
0,0012
0,017
0,0068
0,048
0,0001
0,038
0,045
0,0025
0,012
0,029
0,014
0,00026
0,00063
0,0049
0,032
0,0011
0,0053
0,00082
0,013
0,0095
0,011
0,00045
0,024
0,017
0,006
0,006
0,0029
0,0079
0,015
0,0063
0,028
0,00047
0,036
0,0019
0,0002
0,0014
0,0039
0,0001
0,0002
0,016
0,031
0,022
0,014
0,008
0,0018
0,021
0,0001
Fold Change�
5,9
6,1
INF
3,1
INF
INF
4,3
8,3
11
INF
INF
INF
19
4,9
17
INF
INF
9,9
1,9
1,8
2
21
1,9
6,7
INF
INF
INF
24
INF
INF
INF
14
10
INF
INF
4,9
5,3
INF
6,4
INF
INF
22
INF
INF
4,3
INF
(Continued )
PLOS ONE | https://doi.org/10.1371/journal.pone.0228134 January 28, 2020
7 / 18
Spinal proteomic profile of chronic lame dairy cows
Table 2. (Continued)
Transferrin
Transketolase
Versican
Zeta-crystallin
Proteins
Accesion (GI)
MW kDa
Gene Name
p value
602117
148744821
296485061
4097831
24
68
370
35
TF
TKT
VCAN
CRYZ
0,034
0,033
0,0046
0,0039
Fold Change�
24
8,6
3,6
INF
� Fold change was calculated by category (Lame and Control) using Scaffold v.4.8.6 as the ratio between the average of quantitative values in the lame group (numerator)
and the control group (denominator). INF indicates zero quantitative values in the control group.
https://doi.org/10.1371/journal.pone.0228134.t002
immune cells can secrete stress proteins under specific conditions [45, 46]. In humans, plasma
concentration of Hsp90 has been associated with tumor malignancy [47], and similarly, a
Fig 1. GO annotation for A) molecular function and B) biological process of identified proteins in the dorsal horn of the spinal cord of chronically lame cows.
https://doi.org/10.1371/journal.pone.0228134.g001
PLOS ONE | https://doi.org/10.1371/journal.pone.0228134 January 28, 2020
8 / 18
Spinal proteomic profile of chronic lame dairy cows
Fig 2. STRING network graphic of proteins detected in the spinal cord. Edges represents evidence of the type of interaction, selecting database, experiments co-
expression and co-occurrence as active interaction sources with a high confident of interaction score (0.900). Circles distinguish four groups of interaction. A: consisted
in 10 proteins identified as chaperones and co-chaperones in which 7 were only detected in lame cows and 1 was strongly upregulated in lame cow. B: conformed by 16
proteins involved in glycolysis, gluconeogenesis and pentose phosphate pathway. C: composed mainly by mitochondrial proteins involved in the tricyclic acid pathway
with a relative even expression in both groups. D: interaction generated by tubulin isoforms with no difference in the level of expression among groups but interacting
with chaperone network. Light blue edges represent known interactions obtained from curated data base, violet edges represent known interactions experimentally
determined, and black edges represent protein association by co-expression.
https://doi.org/10.1371/journal.pone.0228134.g002
positive correlation between plasma Hsp70 and plasma cytokines has been described [48].
Also, glial cells can secrete Hsp70 and Hsp90 [49], which can be recognized by several surface
receptor including: TLR2/TLR4 [50], CD40 [51] and CCR5 [52]. Interestingly, CD40 and
CCR5 are involved in microglial activation and chronic pain development [53, 54]. Moreover,
PLOS ONE | https://doi.org/10.1371/journal.pone.0228134 January 28, 2020
9 / 18
Spinal proteomic profile of chronic lame dairy cows
Table 3. KEGG pathway ID of interconnected proteins obtained from STRING database.
KEGG Pathway
ID
Pathway Description
00010
Glycolysis
count in
genes
18
00020
Citrate cycle (TCA cycle)
04612
00030
04066
Antigen processing and
presentation
Pentose phosphate pathway
HIF-1 signalling pathway
6
7
5
5
p value
1.66e-
24
7.21e-
07
4.15e-
06
Proteins
ENO1 ("), ENO2 ("), ENO3 ("), GAPD (#), HK1 ("), GPI ($), ALDOA ($), PGK1 ($),
PGAM1 ($), LDH ($)
ACLY ("), ACO2 ("), DLD (""), IDH2 ($), MDH1 ($), MDH2 ($)
HSP90AA1 (""), HSP90AB1 ($), HSPA1 (""), HSPA1L (""), HSPA4 (""), HSPA6 (""),
HSPA8 (")
8.4e-06 ALDOA ($), ALDOC ($), GPI ($), PGM1 ($), TKT ("")
0.00276 ENO1 ("), ENO2 ("), ENO3 ("), HK1 ("), TF (")
"Increased expression in lame cows, ""Only detected in lame cows, #Reduced expression in lame cows, $no difference in expression
https://doi.org/10.1371/journal.pone.0228134.t003
during nerve damage, HSP´s are release from damaged neurons acting as damage-associated
molecular patterns (DAMP’s), which can trigger glial and immune cell activation through the
TLR2 and TLR4 pathways, contributing to neuropathic pain development [55]. Hsp90 can
induce microglial activation in the spinal cord, enhancing hyperalgesia through TLR4 receptor
activation [56]. Recently, Nascimento et al. (2018) [57] reported a marked upregulation of
Fig 3. A) Representative western blot of gamma-enolase and Hsp70 from the spinal cord obtained from lame and
control cows. B) Densitometric analysis of western blot showing gamma-enolase and Hsp70 upregulation in the spinal
cord of lame cows. L: lame (n = 5), C: control (n = 7).
https://doi.org/10.1371/journal.pone.0228134.g003
PLOS ONE | https://doi.org/10.1371/journal.pone.0228134 January 28, 2020
10 / 18
Spinal proteomic profile of chronic lame dairy cows
Hsp90 in DRG in a rat arthritic model, describing that the pharmacological inhibition of
Hsp90 induced analgesia and a reduction in astrocyte activation. Furthermore, increased levels
of Hsp90 have been described in different chronic inflammatory diseases [58] and Hsp90 has
been associated with the development of neurodegenerative diseases due to its role in stabiliz-
ing the transcriptional factor NF-κB [59].
The increased number of chaperone proteins here described could be associated with endo-
plasmic reticulum (ER) stress inducing an unfolding protein response (UPR) [43, 60]. ER
stress promotes cell protection against different insults by altering the transcriptome and pro-
teome. However, prolonged ER stress may disrupt the protective mechanisms of UPR, leading
to the activation of inflammatory and apoptotic processes [43]. In our study, the ER chaperone
Grp78 (also known as BiP or HSPA5) was strongly upregulated in lame cows. Grp78 has been
described as a marker of ER stress and UPR [61]. UPR is associated with an increased demand
of protein folding and secretion in the ER, which could explain the elevated number of chaper-
one proteins found in lame cows. UPR has been reported in several neuroinflammatory condi-
tions in which protein aggregates are found [43]. Recently, Zhou et al. (2017) [62] reported an
increased expression of Grp78 in the dorsal horn of the spinal cord in a rat model of inflamma-
tory pain. Increased levels of Grp78 have also been reported in an orofacial pain model in rats
[63] and in the spinal cord after L5-spare nerve ligation (SNL) in a rat model of neuropathic
pain [64]. Interestingly, increased expression of Grp78 has been shown to participate in the
activation of astrocytes and microglia [65, 66]. Moreover, chronic morphine administration
causes an upregulation of Grp78 and UPR leading to morphine tolerance in rats [67]. This
finding may explain in part the mechanisms involved in the lack of analgesic effect of opioids
during chronic pain states or might be involved in the downregulation of the endorphin anti-
nociceptive descending pathway described during chronic pain.
Another cluster of interacting proteins detected in the STRING analysis corresponded to
metabolic enzymes. Additionally, GO and KEGG pathway analysis showed a marked number
of proteins involved in the glycolytic pathway, tricarboxylic acid pathway and the pentose
phosphate pathway. Similar to our findings, changes in the expression of metabolism-associ-
ated proteins are frequently reported in pain studies using a proteomic approach [37, 38, 39].
These findings could be explained by the increasing metabolic demand of the neural tissue fol-
lowing nerve damage, since protein synthesis is an energy demanding process [68]. Enolase is
a glycolytic and multifunctional enzyme that is frequently detected in proteomic studies [69].
Here, enolase 1 (α-enolase; ENO1), enolase 2 (γ-enolase or neuronal specific enolase; ENO2)
and enolase 3 (β-enolase; ENO3) were strongly upregulated in lame cows. Non-glycolytic func-
tions of metabolic enzymes have gained attention, in particular enolase, which can act as plas-
minogen receptor during pathological states [70]. Inflammatory signals promote enolase
translocation to the cell membrane [71] enhancing plasminogen activation [72] and promot-
ing extracellular matrix degradation, metalloproteinase activation, macrophage migration and
cytokines synthesis [71, 73]. Cell surface enolase has been found in neurons, activated micro-
glia [74], astrocytes [73] and different immune cells [71]. The role of enolase in chronic pain
was confirmed by Polcyn et al. (2017) [75], after selectively blocking its non-glycolytic func-
tions, thus reducing spinal glial activation and cytokine synthesis in a rat model of spinal cord
injury.
Antioxidant proteins SOD1, GST, PRDX1 and PRDX4 were significantly upregulated in
lame cows. Different modifications in the expression of antioxidant proteins have been
reported in pain studies using proteomics analysis [38, 39]. Reactive oxygen species (ROS)
increase in the spinal cord during chronic pain [76] participating in the phosphorylation of the
NMDA receptor [77], in the activation of TRPA1 and TRPV1 channels [78] and decreasing
GABA release by inhibitory interneurons [79]. Furthermore, ROS signaling is an important
PLOS ONE | https://doi.org/10.1371/journal.pone.0228134 January 28, 2020
11 / 18
Spinal proteomic profile of chronic lame dairy cows
regulator of ER stress and UPR [80, 81] and acts as a potent stimulus for chaperone proteins
upregulation [82], which is consistent with the findings previously described.
Peroxiredoxin 1 (Prdx1) and Peroxiredoxin 4 (Prdx4) are active mediators of the ROS sig-
naling pathway [83]. Prdx1 has previously been reported to increase during chronic pain [39].
Additionally, Prdx1 can be secreted by non-classical pathway, activating the immune response
[84]. In contrast, Prdx4 was only detected in the spinal cord of lame cows suggesting a high
demand of antioxidant activity in the ER. Prdx4 is a peroxidase with chaperone functions pre-
dominantly localized in the ER that assist protein folding by reducing H2O2, thus preventing
oxidative stress in the ER [85]. Increased levels of Prdx have been associated with chronic
inflammatory conditions and neurodegeneration [86]. We believe that Prdx proteins may play
an important role in the development of chronic pain of lame cows.
Fascin is another protein that was strongly upregulated in the dorsal horn of lame cows.
Fascin is an actin-bundling protein that regulates cell motion [87]. Fascin upregulation in the
CNS promotes microglial activation; cell migration and proinflammatory cytokine release
[88]. Similarly, dihydropyrimidinase-like proteins (Dpysl) participate in synaptic formation
during CNS development [89] interacting with actin and microtubules [90]. Two proteins of
this family were differentially expressed in lame cows. Dpysl2 was downregulated and Dpysl3
was strongly upregulated. Dpysl2 are frequently downregulated in proteomics studies after
nerve damage in order to control axonal guidance [26]. In contrast, microglial activation leads
to an increase in the expression of Dpysl3 following stimulation by LPS [91]. Additionally,
microtubule-associated protein 1A (MAP1A) and 1B (MAP1B) were only present in lame
cows. Both proteins are expressed along the axon and dendritic processes of neurons, both in
the CNS and the peripheral nervous system (PNS), where they bind to microtubules and
microfilaments [92]. MAP1B is only expressed during embryogenesis and its role in nervous
system development has been described [93, 94]. A potential role of MAP1B in the mature
CNS has not been defined, but its upregulation could be associated with the presence of pro-
tein aggregates in neurodegenerative disease, proteosomal degradation and autophagy [94,
95].
A limitation of our study is that several upregulated proteins in lame cows were not inte-
grated in the enrichment pathway analysis. This could be explained by the incomplete Bos tau-
rus annotation database due to the reduced scientific information of the species and to the
highest confidence score selected in order to reduce false positive rate. This must be taken into
account when bioinformatics analyses are performed in less common species. Also, the small
sample size in the lame group must be taken into account before extrapolating these results.
Furthermore, the proteomic results here presented must be confirmed using additional func-
tional molecular analysis.
Conclusions
The results here presented demonstrate that persistent pain originated by chronic inflamma-
tory lameness in dairy cows is partly mediated by ER stress. Evident changes in chaperones,
metabolic and redox proteins that are frequently upregulated under cellular stress are
described. Moreover, the proteome of the dorsal horn from chronically lame cows showed
increased expression of several proteins with non-canonical functions. This non-canonical
function might be triggered by stress signals originated from the persistent painful stimulus.
Reactive oxygen species, ER stress and UPR are known to play an important role in the main-
tenance of chronic pain states. Further molecular analysis is necessary in order to confirm the
findings here described.
PLOS ONE | https://doi.org/10.1371/journal.pone.0228134 January 28, 2020
12 / 18
Spinal proteomic profile of chronic lame dairy cows
Supporting information
S1 Data.
(XLSX)
Acknowledgments
We want to thanks to Mauricio Herna´ndez for LC-MS/MS technical support, and Pablo Saldi-
via and Guillermo Nourdin for bioinformatics analysis support of the data.
Author Contributions
Conceptualization: Daniel Herzberg, Pablo Strobel, Constanza Meneses, Hedie Bustamante.
Formal analysis: Daniel Herzberg, Pablo Strobel, Heine Mu¨ller.
Funding acquisition: Marianne Werner, Hedie Bustamante.
Investigation: Daniel Herzberg.
Methodology: Daniel Herzberg, Pablo Strobel.
Project administration: Marianne Werner, Hedie Bustamante.
Resources: Hedie Bustamante.
Supervision: Marianne Werner, Hedie Bustamante.
Validation: Daniel Herzberg.
Visualization: Heine Mu¨ller, Constanza Meneses.
Writing – original draft: Daniel Herzberg.
Writing – review & editing: Pablo Strobel, Heine Mu¨ller, Constanza Meneses, Marianne Wer-
ner, Hedie Bustamante.
References
1.
Latremoliere A, Woolf CJ. Central sensitization: a generator of pain hypersensitivity by central neural
plasticity. J Pain. 2009; 10: 895–926. https://doi.org/10.1016/j.jpain.2009.06.012 PMID: 19712899
2. Kuner R. Spinal excitatory mechanisms of pathological pain. Pain. 2015; 156 Suppl 1: S11–7. https://
doi.org/10.1097/j.pain.0000000000000118 PMID: 25789427
3. Grandin T. Welfare Problems in Cattle, Pigs, and Sheep that Persist Even Though Scientific Research
Clearly Shows How to Prevent Them. Animals (Basel). 2018; 8. https://doi.org/10.3390/ani8070124
PMID: 30037055
4. Mogil JS, Davis KD, Derbyshire SW. The necessity of animal models in pain research. Pain. 2010; 151:
12–17. https://doi.org/10.1016/j.pain.2010.07.015 PMID: 20696526
5. Klinck MP, Mogil JS, Moreau M, Lascelles BDX, Flecknell PA, Poitte T, et al. Translational pain assess-
ment: could natural animal models be the missing link? Pain. 2017; 158: 1633–1646. https://doi.org/10.
1097/j.pain.0000000000000978 PMID: 28614187
6. Kossaibati MA, Esslemont RJ. The costs of production diseases in dairy herds in England. Vet J. 1997;
154: 41–51. Available: https://www.ncbi.nlm.nih.gov/pubmed/9265852 https://doi.org/10.1016/s1090-
0233(05)80007-3 PMID: 9265852
7. Booth CJ, Warnick LD, Gro¨hn YT, Maizon DO, Guard CL, Janssen D. Effect of lameness on culling in
dairy cows. J Dairy Sci. 2004; 87: 4115–4122. https://doi.org/10.3168/jds.S0022-0302(04)73554-7
PMID: 15545373
8. Garbarino EJ, Hernandez JA, Shearer JK, Risco CA, Thatcher WW. Effect of lameness on ovarian
activity in postpartum holstein cows. J Dairy Sci. 2004; 87: 4123–4131. https://doi.org/10.3168/jds.
S0022-0302(04)73555-9 PMID: 15545374
PLOS ONE | https://doi.org/10.1371/journal.pone.0228134 January 28, 2020
13 / 18
Spinal proteomic profile of chronic lame dairy cows
9. Griffiths BE, Grove White D, Oikonomou G. A Cross-Sectional Study Into the Prevalence of Dairy Cattle
Lameness and Associated Herd-Level Risk Factors in England and Wales. Front Vet Sci. 2018; 5: 65.
https://doi.org/10.3389/fvets.2018.00065 PMID: 29675419
10. Ranjbar S, Rabiee AR, Gunn A, House JK. Identifying risk factors associated with lameness in pasture-
based dairy herds. J Dairy Sci. 2016; 99: 7495–7505. https://doi.org/10.3168/jds.2016-11142 PMID:
27394954
11. Sepu´ lveda-Varas P, Lomb J, von Keyserlingk MAG, Held R, Bustamante H, Tadich N. Claw horn
lesions in mid-lactation primiparous dairy cows under pasture-based systems: Association with behav-
ioral and metabolic changes around calving. J Dairy Sci. 2018; 101: 9439–9450. https://doi.org/10.
3168/jds.2018-14674 PMID: 30100516
12. Woolf CJ, Ma Q. Nociceptors—noxious stimulus detectors. Neuron. 2007; 55: 353–364. https://doi.org/
10.1016/j.neuron.2007.07.016 PMID: 17678850
13. Chen L, Yang G, Grosser T. Prostanoids and inflammatory pain. Prostaglandins Other Lipid Mediat.
2013; 104–105: 58–66. https://doi.org/10.1016/j.prostaglandins.2012.08.006 PMID: 22981510
14. Couture R, Harrisson M, Vianna RM, Cloutier F. Kinin receptors in pain and inflammation. Eur J Phar-
macol. 2001; 429: 161–176. Available: https://www.ncbi.nlm.nih.gov/pubmed/11698039 https://doi.org/
10.1016/s0014-2999(01)01318-8 PMID: 11698039
15. Mizumura K, Murase S. Role of nerve growth factor in pain. Handb Exp Pharmacol. 2015; 227: 57–77.
https://doi.org/10.1007/978-3-662-46450-2_4 PMID: 25846614
16.
17.
Zhang N, Inan S, Cowan A, Sun R, Wang JM, Rogers TJ, et al. A proinflammatory chemokine, CCL3,
sensitizes the heat- and capsaicin-gated ion channel TRPV1. Proc Natl Acad Sci U S A. 2005; 102:
4536–4541. https://doi.org/10.1073/pnas.0406030102 PMID: 15764707
Jin X, Gereau RW 4th. Acute p38-mediated modulation of tetrodotoxin-resistant sodium channels in
mouse sensory neurons by tumor necrosis factor-alpha. J Neurosci. 2006; 26: 246–255. https://doi.org/
10.1523/JNEUROSCI.3858-05.2006 PMID: 16399694
18. Sun WH, Chen CC. Roles of Proton-Sensing Receptors in the Transition from Acute to Chronic Pain. J
Dent Res. 2016; 95: 135–142. https://doi.org/10.1177/0022034515618382 PMID: 26597969
19. Sandku¨ hler J. Models and mechanisms of hyperalgesia and allodynia. Physiol Rev. 2009; 89: 707–758.
https://doi.org/10.1152/physrev.00025.2008 PMID: 19342617
20.
Tadich N, Tejeda C, Bastias S, Rosenfeld C, Green LE. Nociceptive threshold, blood constituents and
physiological values in 213 cows with locomotion scores ranging from normal to severely lame. Vet J.
2013; 197: 401–405. https://doi.org/10.1016/j.tvjl.2013.01.029 PMID: 23499542
21. Santos SD, Carvalho AL, Caldeira MV, Duarte CB. Regulation of AMPA receptors and synaptic plastic-
ity. Neuroscience. 2009; 158: 105–125. https://doi.org/10.1016/j.neuroscience.2008.02.037 PMID:
18424006
22. Chen B-S, Roche KW. Regulation of NMDA receptors by phosphorylation. Neuropharmacology. 2007;
53: 362–368. https://doi.org/10.1016/j.neuropharm.2007.05.018 PMID: 17644144
23. Khoutorsky A, Price TJ. Translational Control Mechanisms in Persistent Pain. Trends Neurosci. 2018;
41: 100–114. https://doi.org/10.1016/j.tins.2017.11.006 PMID: 29249459
24.
Zhang Y, Fonslow BR, Shan B, Baek M-C, Yates JR 3rd. Protein analysis by shotgun/bottom-up proteo-
mics. Chem Rev. 2013; 113: 2343–2394. https://doi.org/10.1021/cr3003533 PMID: 23438204
25. Gomez-Varela D, Barry AM, Schmidt M. Proteome-based systems biology in chronic pain. J Proteo-
mics. 2019; 190: 1–11. https://doi.org/10.1016/j.jprot.2018.04.004 PMID: 29653266
26. Niederberger E, Ku¨ hlein H, Geisslinger G. Update on the pathobiology of neuropathic pain. Expert Rev
Proteomics. 2008; 5: 799–818. https://doi.org/10.1586/14789450.5.6.799 PMID: 19086860
27. Mao J. Current challenges in translational pain research. Trends Pharmacol Sci. 2012; 33: 568–573.
https://doi.org/10.1016/j.tips.2012.08.001 PMID: 22959652
28. Serada S, Fujimoto M, Ogata A, Terabe F, Hirano T, Iijima H, et al. iTRAQ-based proteomic identifica-
tion of leucine-rich alpha-2 glycoprotein as a novel inflammatory biomarker in autoimmune diseases.
Ann Rheum Dis. 2010; 69: 770–774. https://doi.org/10.1136/ard.2009.118919 PMID: 19854709
29. Ba¨ ckryd E, Ghafouri B, Carlsson AK, Olausson P, Gerdle B. Multivariate proteomic analysis of the cere-
brospinal fluid of patients with peripheral neuropathic pain and healthy controls—a hypothesis-generat-
ing pilot study. J Pain Res. 2015; 8: 321–333. https://doi.org/10.2147/JPR.S82970 PMID: 26170714
30. Wåhle´n K, Olausson P, Carlsson A, Ghafouri N, Gerdle B, Ghafouri B. Systemic alterations in plasma
proteins from women with chronic widespread pain compared to healthy controls: a proteomic study. J
Pain Res. 2017; 10: 797–809. https://doi.org/10.2147/JPR.S128597 PMID: 28435317
31.
Tadich N, Flor E, Green L. Associations between hoof lesions and locomotion score in 1098 unsound
dairy cows. Vet J. 2010; 184: 60–65. https://doi.org/10.1016/j.tvjl.2009.01.005 PMID: 19211281
PLOS ONE | https://doi.org/10.1371/journal.pone.0228134 January 28, 2020
14 / 18
Spinal proteomic profile of chronic lame dairy cows
32. Reader JD, Green MJ, Kaler J, Mason SA, Green LE. Effect of mobility score on milk yield and activity
in dairy cattle. J Dairy Sci. 2011; 94: 5045–5052. https://doi.org/10.3168/jds.2011-4415 PMID:
21943755
33. Nesvizhskii AI, Keller A, Kolker E, Aebersold R. A statistical model for identifying proteins by tandem
mass spectrometry. Anal Chem. 2003; 75: 4646–4658. Available: https://www.ncbi.nlm.nih.gov/
pubmed/14632076 https://doi.org/10.1021/ac0341261 PMID: 14632076
34. Keller A, Purvine S, Nesvizhskii AI, Stolyar S, Goodlett DR, Kolker E. Experimental protein mixture for
validating tandem mass spectral analysis. OMICS. 2002; 6: 207–212. https://doi.org/10.1089/
153623102760092805 PMID: 12143966
35. Ashburner M, Ball CA, Blake JA, Botstein D, Butler H, Cherry JM, et al. Gene ontology: tool for the unifi-
cation of biology. The Gene Ontology Consortium. Nat Genet. 2000; 25: 25–29. https://doi.org/10.1038/
75556 PMID: 10802651
36. Szklarczyk D, Franceschini A, Wyder S, Forslund K, Heller D, Huerta-Cepas J, et al. STRING v10: pro-
tein-protein interaction networks, integrated over the tree of life. Nucleic Acids Res. 2015; 43: D447–52.
https://doi.org/10.1093/nar/gku1003 PMID: 25352553
37. Kang SK, So HH, Moon YS, Kim CH. Proteomic analysis of injured spinal cord tissue proteins using 2-
DE and MALDI-TOF MS. Proteomics. 2006; 6: 2797–2812. https://doi.org/10.1002/pmic.200500621
PMID: 16586436
38. Komori N, Takemori N, Kim HK, Singh A, Hwang S-H, Foreman RD, et al. Proteomics study of neuro-
pathic and nonneuropathic dorsal root ganglia: altered protein regulation following segmental spinal
nerve ligation injury. Physiol Genomics. 2007; 29: 215–230. https://doi.org/10.1152/physiolgenomics.
00255.2006 PMID: 17213366
39. Huang H-L, Cendan C-M, Roza C, Okuse K, Cramer R, Timms JF, et al. Proteomic profiling of neuro-
mas reveals alterations in protein composition and local protein synthesis in hyper-excitable nerves.
Mol Pain. 2008; 4: 33. https://doi.org/10.1186/1744-8069-4-33 PMID: 18700027
40. Cı´zkova´ D, Luka´ cova´ N, Marsala M, Kafka J, Luka´ c I, Jergova´ S, et al. Experimental cauda equina com-
pression induces HSP70 synthesis in dog. Physiol Res. 2005; 54: 349–356. Available: https://www.
ncbi.nlm.nih.gov/pubmed/15974836 PMID: 15974836
41. Yon JM. Protein folding in the post-genomic era. J Cell Mol Med. 2002; 6: 307–327. https://doi.org/10.
1111/j.1582-4934.2002.tb00511.x PMID: 12417049
42.
van Noort JM. Stress proteins in CNS inflammation. J Pathol. 2008; 214: 267–275. https://doi.org/10.
1002/path.2273 PMID: 18161755
43. Chakrabarti A, Chen AW, Varner JD. A review of the mammalian unfolded protein response. Biotechnol
Bioeng. 2011; 108: 2777–2793. https://doi.org/10.1002/bit.23282 PMID: 21809331
44. Calderwood SK, Mambula SS, Gray PJ Jr, Theriault JR. Extracellular heat shock proteins in cell signal-
ing. FEBS Lett. 2007; 581: 3689–3694. https://doi.org/10.1016/j.febslet.2007.04.044 PMID: 17499247
45. Guzhova I, Kislyakova K, Moskaliova O, Fridlanskaya I, Tytell M, Cheetham M, et al. In vitro studies
show that Hsp70 can be released by glia and that exogenous Hsp70 can enhance neuronal stress toler-
ance. Brain Res. 2001; 914: 66–73. Available: https://www.ncbi.nlm.nih.gov/pubmed/11578598 https://
doi.org/10.1016/s0006-8993(01)02774-3 PMID: 11578598
46. Mambula SS, Stevenson MA, Ogawa K, Calderwood SK. Mechanisms for Hsp70 secretion: crossing
membranes without a leader. Methods. 2007; 43: 168–175. https://doi.org/10.1016/j.ymeth.2007.06.
009 PMID: 17920512
47. Wang X, Song X, Zhuo W, Fu Y, Shi H, Liang Y, et al. The regulatory mechanism of Hsp90alpha secre-
tion and its function in tumor malignancy. Proc Natl Acad Sci U S A. 2009; 106: 21288–21293. https://
doi.org/10.1073/pnas.0908151106 PMID: 19965370
48. Perac¸oli JC, Bannwart-Castro CF, Romao M, Weel IC, Ribeiro VR, Borges VTM, et al. High levels of
heat shock protein 70 are associated with pro-inflammatory cytokines and may differentiate early- from
late-onset preeclampsia. J Reprod Immunol. 2013; 100: 129–134. https://doi.org/10.1016/j.jri.2013.08.
003 PMID: 24051131
49.
Tytell M, Lasek RJ, Gainer H. Axonal maintenance, glia, exosomes, and heat shock proteins.
F1000Res. 2016; 5. https://doi.org/10.12688/f1000research.7247.1 PMID: 26962444
50. Asea A, Rehli M, Kabingu E, Boch JA, Bare O, Auron PE, et al. Novel signal transduction pathway uti-
lized by extracellular HSP70: role of toll-like receptor (TLR) 2 and TLR4. J Biol Chem. 2002; 277:
15028–15034. https://doi.org/10.1074/jbc.M200497200 PMID: 11836257
51. Wang Y, Kelly CG, Karttunen JT, Whittall T, Lehner PJ, Duncan L, et al. CD40 is a cellular receptor
mediating mycobacterial heat shock protein 70 stimulation of CC-chemokines. Immunity. 2001; 15:
971–983. Available: https://www.ncbi.nlm.nih.gov/pubmed/11754818. https://doi.org/10.1016/s1074-
7613(01)00242-4 PMID: 11754818
PLOS ONE | https://doi.org/10.1371/journal.pone.0228134 January 28, 2020
15 / 18
Spinal proteomic profile of chronic lame dairy cows
52.
Floto RA, MacAry PA, Boname JM, Mien TS, Kampmann B, Hair JR, et al. Dendritic cell stimulation by
mycobacterial Hsp70 is mediated through CCR5. Science. 2006; 314: 454–458. https://doi.org/10.
1126/science.1133515 PMID: 17053144
53. Cao L, Palmer CD, Malon JT, De Leo JA. Critical role of microglial CD40 in the maintenance of mechan-
ical hypersensitivity in a murine model of neuropathic pain. Eur J Immunol. 2009; 39: 3562–3569.
https://doi.org/10.1002/eji.200939657 PMID: 19750482
54.
Laudati E, Currò D, Navarra P, Lisi L. Blockade of CCR5 receptor prevents M2 microglia phenotype in a
microglia-glioma paradigm. Neurochem Int. 2017; 108: 100–108. https://doi.org/10.1016/j.neuint.2017.
03.002 PMID: 28279751
55. Milligan ED, Watkins LR. Pathological and protective roles of glia in chronic pain. Nat Rev Neurosci.
2009; 10: 23–36. https://doi.org/10.1038/nrn2533 PMID: 19096368
56. Hutchinson MR, Ramos KM, Loram LC, Wieseler J, Sholar PW, Kearney JJ, et al. Evidence for a role of
heat shock protein-90 in toll like receptor 4 mediated pain enhancement in rats. Neuroscience. 2009;
164: 1821–1832. https://doi.org/10.1016/j.neuroscience.2009.09.046 PMID: 19788917
57. Nascimento DSM, Potes CS, Soares ML, Ferreira AC, Malcangio M, Castro-Lopes JM, et al. Drug-
Induced HSP90 Inhibition Alleviates Pain in Monoarthritic Rats and Alters the Expression of New Puta-
tive Pain Players at the DRG. Mol Neurobiol. 2018; 55: 3959–3975. https://doi.org/10.1007/s12035-
017-0628-x PMID: 28550532
58.
Tukaj S, Węgrzyn G. Anti-Hsp90 therapy in autoimmune and inflammatory diseases: a review of pre-
clinical studies. Cell Stress Chaperones. 2016; 21: 213–218. https://doi.org/10.1007/s12192-016-0670-
z PMID: 26786410
59. Bohonowych JE, Hance MW, Nolan KD, Defee M, Parsons CH, Isaacs JS. Extracellular Hsp90 medi-
ates an NF-κB dependent inflammatory stromal program: implications for the prostate tumor microenvi-
ronment. Prostate. 2014; 74: 395–407. https://doi.org/10.1002/pros.22761 PMID: 24338924
60. Sprenkle NT, Sims SG, Sa´nchez CL, Meares GP. Endoplasmic reticulum stress and inflammation in
the central nervous system. Mol Neurodegener. 2017; 12: 42. https://doi.org/10.1186/s13024-017-
0183-y PMID: 28545479
61. Wang M, Wey S, Zhang Y, Ye R, Lee AS. Role of the unfolded protein response regulator GRP78/BiP
in development, cancer, and neurological disorders. Antioxid Redox Signal. 2009; 11: 2307–2316.
https://doi.org/10.1089/ARS.2009.2485 PMID: 19309259
62.
Zhou F, Zhang W, Zhou J, Li M, Zhong F, Zhang Y, et al. Involvement of endoplasmic reticulum stress
in formalin-induced pain is attenuated by 4-phenylbutyric acid. J Pain Res. 2017; 10: 653–662. https://
doi.org/10.2147/JPR.S125805 PMID: 28360534
63. Yang ES, Bae JY, Kim TH, Kim YS, Suk K, Bae YC. Involvement of endoplasmic reticulum stress
response in orofacial inflammatory pain. Exp Neurobiol. 2014; 23: 372–380. https://doi.org/10.5607/en.
2014.23.4.372 PMID: 25548537
64.
Zhang E, Yi M-H, Shin N, Baek H, Kim S, Kim E, et al. Endoplasmic reticulum stress impairment in the
spinal dorsal horn of a neuropathic pain model. Sci Rep. 2015; 5: 11555. https://doi.org/10.1038/
srep11555 PMID: 26109318
65. Kakimura J, Kitamura Y, Taniguchi T, Shimohama S, Gebicke-Haerter PJ. Bip/GRP78-induced produc-
tion of cytokines and uptake of amyloid-beta(1–42) peptide in microglia. Biochem Biophys Res Com-
mun. 2001; 281: 6–10. https://doi.org/10.1006/bbrc.2001.4299 PMID: 11178952
66. Mha´ille AN, McQuaid S, Windebank A, Cunnea P, McMahon J, Samali A, et al. Increased expression of
endoplasmic reticulum stress-related signaling pathway molecules in multiple sclerosis lesions. J Neu-
ropathol Exp Neurol. 2008; 67: 200–211. https://doi.org/10.1097/NEN.0b013e318165b239 PMID:
18344911
67.
68.
Liu D, Zhou Y, Peng Y, Su P, Li Z, Xu Q, et al. Endoplasmic Reticulum Stress in Spinal Cord Contributes
to the Development of Morphine Tolerance. Front Mol Neurosci. 2018; 11: 72. https://doi.org/10.3389/
fnmol.2018.00072 PMID: 29559889
Lindqvist LM, Tandoc K, Topisirovic I, Furic L. Cross-talk between protein synthesis, energy metabolism
and autophagy in cancer. Curr Opin Genet Dev. 2018; 48: 104–111. https://doi.org/10.1016/j.gde.2017.
11.003 PMID: 29179096
69. Petrak J, Ivanek R, Toman O, Cmejla R, Cmejlova J, Vyoral D, et al. De´ jà vu in proteomics. A hit parade
of repeatedly identified differentially expressed proteins. Proteomics. 2008; 8: 1744–1749. https://doi.
org/10.1002/pmic.200700919 PMID: 18442176
70. Dı´az-Ramos A, Roig-Borrellas A, Garcı´a-Melero A, Lo´ pez-Alemany R. α-Enolase, a multifunctional pro-
tein: its role on pathophysiological situations. J Biomed Biotechnol. 2012; 2012: 156795. https://doi.org/
10.1155/2012/156795 PMID: 23118496
PLOS ONE | https://doi.org/10.1371/journal.pone.0228134 January 28, 2020
16 / 18
Spinal proteomic profile of chronic lame dairy cows
71. Bae S, Kim H, Lee N, Won C, Kim H-R, Hwang Y-I, et al. α-Enolase expressed on the surfaces of mono-
cytes and macrophages induces robust synovial inflammation in rheumatoid arthritis. J Immunol. 2012;
189: 365–372. https://doi.org/10.4049/jimmunol.1102073 PMID: 22623332
72. Bock A, Tucker N, Kelher MR, Khan SY, Gonzalez E, Wohlauer M, et al. α-Enolase Causes Proinflam-
matory Activation of Pulmonary Microvascular Endothelial Cells and Primes Neutrophils Through Plas-
min Activation of Protease-Activated Receptor 2. Shock. 2015; 44: 137–142. https://doi.org/10.1097/
SHK.0000000000000394 PMID: 25944790
73. Haque A, Ray SK, Cox A, Banik NL. Neuron specific enolase: a promising therapeutic target in acute
spinal cord injury. Metab Brain Dis. 2016; 31: 487–495. https://doi.org/10.1007/s11011-016-9801-6
PMID: 26847611
74. Hafner A, Glavan G, Obermajer N, Zˇ ivin M, Schliebs R, Kos J. Neuroprotective role of γ-enolase in
microglia in a mouse model of Alzheimer’s disease is regulated by cathepsin X. Aging Cell. 2013; 12:
604–614. https://doi.org/10.1111/acel.12093 PMID: 23621429
75. Polcyn R, Capone M, Hossain A, Matzelle D, Banik NL, Haque A. Neuron specific enolase is a potential
target for regulating neuronal cell survival and death: implications in neurodegeneration and regenera-
tion. Neuroimmunol Neuroinflamm. 2017; 4: 254–257. https://doi.org/10.20517/2347-8659.2017.59
PMID: 29423430
76. Guedes RP, Bosco LD, Teixeira CM, Arau´jo ASR, Llesuy S, Bello´ -Klein A, et al. Neuropathic pain modi-
fies antioxidant activity in rat spinal cord. Neurochem Res. 2006; 31: 603–609. https://doi.org/10.1007/
s11064-006-9058-2 PMID: 16770731
77. Gao X, Kim HK, Chung JM, Chung K. Reactive oxygen species (ROS) are involved in enhancement of
NMDA-receptor phosphorylation in animal models of pain. Pain. 2007; 131: 262–271. https://doi.org/10.
1016/j.pain.2007.01.011 PMID: 17317010
78. Nishio N, Taniguchi W, Sugimura YK, Takiguchi N, Yamanaka M, Kiyoyuki Y, et al. Reactive oxygen
species enhance excitatory synaptic transmission in rat spinal dorsal horn neurons by activating TRPA1
and TRPV1 channels. Neuroscience. 2013; 247: 201–212. https://doi.org/10.1016/j.neuroscience.
2013.05.023 PMID: 23707800
79. Yowtak J, Lee KY, Kim HY, Wang J, Kim HK, Chung K, et al. Reactive oxygen species contribute to
neuropathic pain by reducing spinal GABA release. Pain. 2011; 152: 844–852. https://doi.org/10.1016/j.
pain.2010.12.034 PMID: 21296500
80. Sharma HS, Gordh T, Wiklund L, Mohanty S, Sjo¨quist PO. Spinal cord injury induced heat shock protein
expression is reduced by an antioxidant compound H-290/51. An experimental study using light and
electron microscopy in the rat. J Neural Transm. 2006; 113: 521–536. https://doi.org/10.1007/s00702-
005-0405-2 PMID: 16550329
81. Eletto D, Chevet E, Argon Y, Appenzeller-Herzog C. Redox controls UPR to control redox. J Cell Sci.
2014; 127: 3649–3658. https://doi.org/10.1242/jcs.153643 PMID: 25107370
82. Niforou K, Cheimonidou C, Trougakos IP. Molecular chaperones and proteostasis regulation during
redox imbalance. Redox Biol. 2014; 2: 323–332. https://doi.org/10.1016/j.redox.2014.01.017 PMID:
24563850
83. Netto LES, Antunes F. The Roles of Peroxiredoxin and Thioredoxin in Hydrogen Peroxide Sensing and
in Signal Transduction. Mol Cells. 2016; 39: 65–71. https://doi.org/10.14348/molcells.2016.2349 PMID:
26813662
84. Riddell JR, Wang X-Y, Minderman H, Gollnick SO. Peroxiredoxin 1 stimulates secretion of proinflamma-
tory cytokines by binding to TLR4. J Immunol. 2010; 184: 1022–1030. https://doi.org/10.4049/jimmunol.
0901945 PMID: 20018613
85.
Tavender TJ, Bulleid NJ. Peroxiredoxin IV protects cells from oxidative stress by removing H2O2 pro-
duced during disulphide formation. J Cell Sci. 2010; 123: 2672–2679. https://doi.org/10.1242/jcs.
067843 PMID: 20627953
86. Park MH, Jo M, Kim YR, Lee C-K, Hong JT. Roles of peroxiredoxins in cancer, neurodegenerative dis-
eases and inflammatory diseases. Pharmacol Ther. 2016; 163: 1–23. https://doi.org/10.1016/j.
pharmthera.2016.03.018 PMID: 27130805
87. Yamashiro S, Yamakita Y, Ono S, Matsumura F. Fascin, an actin-bundling protein, induces membrane
protrusions and increases cell motility of epithelial cells. Mol Biol Cell. 1998; 9: 993–1006. Available:
https://www.ncbi.nlm.nih.gov/pubmed/9571235 https://doi.org/10.1091/mbc.9.5.993 PMID: 9571235
88. Kim J-K, Lee S-M, Suk K, Lee W-H. A novel pathway responsible for lipopolysaccharide-induced trans-
lational regulation of TNF-α and IL-6 expression involves protein kinase C and fascin. J Immunol. 2011;
187: 6327–6334. https://doi.org/10.4049/jimmunol.1100612 PMID: 22102721
89. Yamashita N, Goshima Y. Collapsin response mediator proteins regulate neuronal development and
plasticity by switching their phosphorylation status. Mol Neurobiol. 2012; 45: 234–246. https://doi.org/
10.1007/s12035-012-8242-4 PMID: 22351471
PLOS ONE | https://doi.org/10.1371/journal.pone.0228134 January 28, 2020
17 / 18
Spinal proteomic profile of chronic lame dairy cows
90. Yu-Kemp H-C, Brieher WM. Collapsin Response Mediator Protein-1 Regulates Arp2/3-dependent Actin
Assembly. J Biol Chem. 2016; 291: 658–664. https://doi.org/10.1074/jbc.C115.689265 PMID:
26598519
91. Manivannan J, Tay SSW, Ling E-A, Dheen ST. Dihydropyrimidinase-like 3 regulates the inflammatory
response of activated microglia. Neuroscience. 2013; 253: 40–54. https://doi.org/10.1016/j.
neuroscience.2013.08.023 PMID: 23988434
92. Mohan R, John A. Microtubule-associated proteins as direct crosslinkers of actin filaments and microtu-
bules. IUBMB Life. 2015; 67: 395–403. https://doi.org/10.1002/iub.1384 PMID: 26104829
93.
Tortosa E, Montenegro-Venegas C, Benoist M, Ha¨ rtel S, Gonza´ lez-Billault C, Esteban JA, et al. Micro-
tubule-associated protein 1B (MAP1B) is required for dendritic spine development and synaptic matura-
tion. J Biol Chem. 2011; 286: 40638–40648. https://doi.org/10.1074/jbc.M111.271320 PMID: 21984824
94. Villarroel-Campos D, Gonzalez-Billault C. The MAP1B case: an old MAP that is new again. Dev Neuro-
biol. 2014; 74: 953–971. https://doi.org/10.1002/dneu.22178 PMID: 24700609
95. Sayas CL, A´ vila J. Crosstalk between axonal classical microtubule-associated proteins and end binding
proteins during axon extension: possible implications in neurodegeneration. J Alzheimers Dis. 2014; 40
Suppl 1: S17–22. https://doi.org/10.3233/JAD-132315 PMID: 24531158
PLOS ONE | https://doi.org/10.1371/journal.pone.0228134 January 28, 2020
18 / 18
|
10.1371_journal.pone.0216465 | RESEARCH ARTICLE
Nrf2 downregulates zymosan-induced
neutrophil activation and modulates
migration
Doumet Georges HelouID
Marie-He´ lène Damien1, Marc Pallardy1, Saadia Kerdine-Ro¨ mer1☯, Sylvie Chollet-
MartinID
1, Sarah Braham1, Luc De Chaisemartin1,2, Vanessa Granger1,2,
1,2☯*
1 Inflammation, Chimiokines et Immunopathologie, INSERM UMR996, Univ. Paris-Sud, Universite´ Paris-
Saclay,Chaˆtenay-Malabry, France, 2 Laboratoire d’immunologie, « Autoimmunite´ et Hypersensibilite´s »,
Hoˆpital Bichat-Claude Bernard, AP-HP, Paris, France
☯ These authors contributed equally to this work.
* sylvie.chollet-martin@u-psud.fr
Abstract
Polymorphonuclear neutrophils (PMNs) are the first line of defense against pathogens and
their activation needs to be tightly regulated in order to limit deleterious effects. Nrf2
(Nuclear factor (erythroïd-derived 2)-like 2) transcription factor regulates oxidative stress
and/or represses inflammation in various cells such as dendritic cells or macrophages. How-
ever, its involvement in PMN biology is still unclear. Using Nrf2 KO mice, we thus aimed to
investigate the protective role of Nrf2 in various PMN functions such as oxidative burst, neto-
sis, migration, cytokine production and phagocytosis, mainly in response to zymosan. We
found that zymosan induced Nrf2 accumulation in PMNs leading to the upregulation of
some target genes including Hmox-1, Nqo1 and Cat. Nrf2 was able to decrease zymosan-
induced PMN oxidative burst; sulforaphane-induced Nrf2 hyperexpression confirmed its
implication. Tnfα, Ccl3 and Cxcl2 gene transcription was decreased in zymosan-stimulated
Nrf2 KO PMNs, suggesting a role for Nrf2 in the regulation of proinflammatory cytokine pro-
duction. However, Nrf2 was not involved in phagocytosis. Finally, spontaneous migration of
Nrf2 KO PMNs was lower than that of WT PMNs. Moreover, in response to low concentra-
tions of CXCL2 or CXCL12, Nrf2 KO PMN migration was decreased despite similar CXCR2
and CXCR4 expression and ATP levels in PMNs from both genotypes. Nrf2 thus seems to
be required for an optimal migration. Altogether these results suggest that Nrf2 has a protec-
tive role in several PMN functions. In particular, it downregulates their activation in response
to zymosan and is required for an adequate migration.
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: Helou DG, Braham S, De Chaisemartin L,
Granger V, Damien M-H, Pallardy M, et al. (2019)
Nrf2 downregulates zymosan-induced neutrophil
activation and modulates migration. PLoS ONE 14
(8): e0216465. https://doi.org/10.1371/journal.
pone.0216465
Editor: Pavel Strnad, Medizinische Fakultat der
RWTH Aachen, GERMANY
Received: October 5, 2018
Accepted: July 9, 2019
Published: August 16, 2019
Copyright: © 2019 Helou et al. This is an open
access article distributed under the terms of the
Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: All relevant data are
within the manuscript and its Supporting
Information files.
Funding: The authors received no specific funding
for this work.
Competing interests: The authors have declared
that no competing interests exist.
Introduction
Abbreviations: BM, Bone marrow; DC, Dendritic
cell; GILZ, Glucocorticoïd-induced leucine zipper;
GPX, Glutathione peroxidase; HMOX-1, Heme
Polymorphonuclear neutrophils (PMNs) are the first cells to be mobilized against pathogens
present in both blood and tissues. They possess a variety of killing mechanisms such as reactive
oxygen species (ROS) release during oxidative burst, secretion of cytolytic enzymes and
PLOS ONE | https://doi.org/10.1371/journal.pone.0216465 August 16, 2019
1 / 21
oxygenase-1; IL-, Interleukin; KEAP 1, Kelch-like
ECH-associated protein 1; MPO, Myeloperoxidase;
NET, Neutrophil extracellular traps; NOX2, NADPH
oxidase 2; NQO1, NAD(P)H quinone
oxidoreductase 1; NRF2, Nuclear factor (erythroïd-
derived 2)-like 2; PMNs, Polymorphonuclear
neutrophils; ROS, Reactive oxygen species; SFN,
Sulforaphane; TLR, Toll-like receptor; TNFα, Tumor
necrosis factor α.
The role of Nrf2 transcription factor in neutrophil biology
cytokines, phagocytosis and formation of neutrophil extracellular traps (NETs). They are cen-
tral players in critical illness [1,2] and in chronic inflammation [3]; therefore, their activation
needs to be tightly regulated to avoid tissue damage. Over the past two decades, a large number
of studies have evidenced that PMNs can also behave like immune-regulatory cells [2,4].
Among the recently described mechanisms, one can emphasize the following: i) myeloperoxi-
dase (MPO), a key PMN enzyme, can decrease mortality in a sepsis model and regulate in-
flammation [5,6], ii) our group and others have described PMN-induced dendritic cell
modulation in particular via NETs [7,8], iii) NADPH oxidase 2 (NOX2) can limit inflamma-
tion in some situations [9] iiii) several inhibitory receptors and mediators have been described
such as immunoglobulin-like transcript 4 or Glucocorticoïd- Induced Leucine Zipper (GILZ)
[10,11].
Among the antioxidant and cytoprotective factors, nuclear factor (erythroïd-derived 2)-like
2 (Nrf2) is a transcription factor known as a master cell protector from ROS and electrophilic
insult [12]. At basal state, Nrf2 is repressed by its negative regulator kelch-like ECH-associated
protein 1 (Keap 1) [13]. Cellular exposure to oxidative stress or electrophiles can alter Keap1
conformation leading to nuclear translocation of Nrf2 [14]. Consequently, Nrf2 activates a bat-
tery of cytoprotective genes, such as Nqo1 [NAD(P)H quinone oxidoreductase 1], Hmox-1
(heme oxygenase-1) and Cat (catalase), all characterized by their antioxidant response element
(ARE) regulatory sequence [12,15]. Typical Nrf2 activators, such as sulforaphane (SFN) and
tert-butylhydroquinone (tBHQ), interact with certain cysteine residues of Keap1 implicating
an electrophilic modification [16]. In addition, Nrf2 can be activated by endogenous inflam-
matory products such as 15-deoxy-Δ12,14-prostaglandin J2 (15d-PGJ2) and NO-derived prod-
ucts [17].
The Nrf2-Keap1 pathway is involved in dendritic cell (DC) and macrophage functions [18].
In particular, we have shown that Nrf2 controls cell death induced by contact sensitizers in
DCs and upregulates antioxidant genes like Gsr, Cat, Gpx and Nos2 controlling ROS produc-
tion [19,20]. Recently, in a recent publication we described Nrf2 involvement in the control of
fibrosis and autoimmunity during sclerodermia [21]. The absence of Nrf2 in immature DCs
(iDCs) raises the intracellular levels of ROS and results in an enhanced co-stimulatory receptor
expression associated with an increased antigen-specific CD8 T cell stimulation capacity [22].
Furthermore, the absence of Nrf2 downregulates the phagocytic functions of DCs [22] and
macrophages [23,24], in particular via CD36 expression. In addition, Nrf2 downregulates the
transcription of pro-inflammatory cytokine genes in macrophages such as Il-6, Il-1β and Tnfα,
independently of redox control [19,25] and is involved in the expansion of suppressive mye-
loid-derived suppressor cells in steady state and during sepsis [26].
In contrast to DCs and macrophages, few studies have evaluated the role of Nrf2 in PMNs.
It has been suggested that ROS regulation, one of the major roles of Nrf2, could contribute to
the overall Nrf2 anti-inflammatory effect [24]. As Nrf2 activation is one of the ROS regulation
pathways (both mitochondrial [27] and NOX2 derived-ROS [9,28,29], it can be assumed that
Nrf2 could be a potential regulator of NOX2-dependent PMN functions, such as netosis. The
ex vivo ROS production capacity of PMNs from Nrf2 Knock-out mice (Nrf2 KO) has indeed
been shown to be increased during sepsis [24,30], in severe periodontitis [31] or in a model of
wound healing [32].
Our aim was to better understand the role of Nrf2 in the regulation of oxidative burst and
several other PMN functions. In order to do this, we chose to use zymosan as a stimulus. This
insoluble cell wall preparation from Saccharomyces cerevisiae is known to activate phagocytes
(TLR-2 and Dectin ligand) via the phosphorylation of p47phox subunit and rac2 activation
[33]. We thus compared in vitro, several functions of bone marrow (BM)-derived PMNs from
wild type (WT) mice and Nrf2 Knock-Out (KO) mice. We found that Nrf2 exhibited a
PLOS ONE | https://doi.org/10.1371/journal.pone.0216465 August 16, 2019
2 / 21
The role of Nrf2 transcription factor in neutrophil biology
protective role in zymosan-stimulated PMNs. Nrf2 was indeed shown to activate the transcrip-
tion of the cytoprotective genes Nqo-1, Hmox-1 and Cat, to downregulate the pro-inflamma-
tory genes Tnfα, Cxcl2 and Ccl3 and to reduce ROS production. Furthermore, optimal
migration was linked to Nrf2 expression. Interestingly, Nrf2 was not required for phagocytosis
in our model. These findings could help clarify the implication of Nrf2 in clinical situations
associated with PMN recruitment and/or activation. Moreover, these new findings could
improve the yet to be developed comprehensive evaluation for the Nrf2-targeted therapy, as
recently discussed by Cuadrado et al. [34].
Material and methods
Ethics statements
All animal studies were performed according to European Commission guidelines in compli-
ance with French Animal Welfare Law (law n˚2013–1118 from February 1st 2013, article
R214.89). Mice were killed for the sole purpose of collecting tibias and femurs to isolate bone-
marrow. According to the French law cited above, this is not considered as an experimental
procedure and no ethical approval is needed from the French Ministry of Research nor the
French Ministry of Agriculture.
Tibias and femurs were collected immediately after cervical dislocation. This euthanasia
procedure is in agreement with Directive 2010/63/UE of September 22 2010 annex IV and law
n˚ 2013–1118 of February 1st 2013.
Mice
Wild-Type (WT, Nfe2l2+/+) and Nrf2 Knock-Out mice (Nrf2 KO, Nfe2l2−/−) were generated
from inbred C57BL/6J background nrf2 heterozygous mice. Nrf2−/− mice [35] were provided
by the RIKEN BRC in accordance with a material transfer agreement (MTA) signed with Prof.
S. Kerdine-Ro¨mer. The donating investigator reported that these mice were backcrossed to
C57BL/6J for at least 10 generations. Mice were housed in a pathogen-free facility and handled
in accordance with the principles and procedures outlined in Council Directive 2010/63/EU.
Mice were bred side by side in ventilated racks within a specific pathogen-free facility. Age-
and sex-matched mice were used at 8–14 weeks of age. Genotyping was performed by PCR
using genomic DNA that was isolated from tail snips as described [35].
BM-derived PMN isolation
PMNs were isolated from BM using the mouse neutrophil isolation Kit (Miltenyi Biotec, Berg-
ish Glabash, Germany) according to the manufacturer’s instructions. In brief, cells were col-
lected from the femur and tibia and resuspended in phosphate buffered saline (PBS) solution
containing 0.5% bovine serum albumin (BSA) and 2 mM Ethylene diamine tetraacetic acid
(EDTA). Cell suspension was then filtrated through pre-separation filters 70 μM (Miltenyi Bio-
tec) to remove cell aggregates or large particles and ensure effective magnetic cell labeling.
Cells were incubated with biotin-antibody cocktail for 10 min, washed and incubated with
anti-biotin microbeads for 15 min. Finally, cell suspension was washed and applied onto a LS
column placed in a magnetic field. Flow-through containing unlabeled cells representing the
enriched PMN suspension, were collected and suspended in Hank’s balanced salt solution
(HBSS) supplemented with 0.5% heat-inactivated fetal calf serum (FCS). Cells were kept at
4–8˚C during staining and magnetic bead isolation. Cell purity and viability were assessed by
flow cytometry (FACS Calibur, BD Biosciences, San Jose, USA) using antibodies against
CD11b and Ly6G (BioLegend, London, UK) and was always �95% (S1 Fig).
PLOS ONE | https://doi.org/10.1371/journal.pone.0216465 August 16, 2019
3 / 21
The role of Nrf2 transcription factor in neutrophil biology
In vitro Nrf2 activation by sulforaphane
In some experiments, sulforaphane (SFN) (Sigma-Aldrich, St. Louis, MO, USA) was used in
order to strongly activate Nrf2 in WT PMNs. Cells (1 × 106/ml) were pre-incubated with SFN
1 μM for 4 h at 37 ˚C with 5% CO2. As SFN can be toxic in some conditions, we checked that
none of the tested SFN concentrations (1 and 5 μM) were either toxic or induced apoptosis
within 4 h, using the Annexin V/7-amino-actinomycin (AnnV/7AAD) counterstaining (Bio-
Legend) followed by flow cytometry (FACS Calibur). Cells positive for Annexin V and nega-
tive for 7-AAD were considered as apoptotic cells (consistently under 3%), whereas double
positive cells were considered as necrotic cells (consistently under 8%) (S2 Fig).
Nrf2 quantification using flow cytometry
PMNs (1 × 106/ml) from WT mice were stimulated with SFN (1 μM), or with 5 μg/ml of zymo-
san A (Sigma-Aldrich, suspended uniformly in HBSS) for 4 h at 37 ˚C with 5% CO2. After
fixation and nuclear permeabilization using a commercial kit (ThermoFisher Scientific, Cali-
fornia, USA), cells were incubated with a rabbit monoclonal anti-Nrf2 antibody detecting
nuclear and cytoplasmic Nrf2 (ab62352, Abcam, Cambridge, UK), and then with a goat anti-
rabbit Alexa Fluor 488 IgG (ThermoFisher Scientific). IgG antibody isotype control was used
as a negative control (BD Biosciences). Intracellular total Nrf2 expression was quantified in all
the samples using an Attune NxT flow cytometer (Thermofisher Scientific). In some experi-
ments, whole BM cells were stained directly without PMN isolation; in that case, the anti-
Ly6G antibody was used to identify PMNs.
Quantitative reverse transcription-polymerase chain reaction (RT-qPCR)
PMNs (1 × 106/ml) from WT and Nrf2 KO mice were stimulated with SFN (1 μM), or with
zymosan (5 μg/ml) for 4 h at 37 ˚C with 5% CO2. Total RNA was extracted after PMN lysis
with RNA-PLUS reagent (MP Biomedicals, Santa Ana, CA, USA). Total RNA pellets were
resuspended in RNAse-free water and quantified by spectrophotometry. First-strand cDNA
was synthesized from total RNA on a thermocycler (Biometra, Go¨ttingen, Germany). The
reaction used 1 μg of total RNA, a dNTP mixture (containing 25 mM dATP, dGTP, dCTP,
and dTTP) and 50 μM oligo (dT) primers (MWG Biotech, Ebersberg, Germany). Reverse tran-
scription was performed in 1× AMV reverse transcriptase reaction buffer (Promega, Charbon-
nières-les-Bains, France), with RNase inhibitor (RNasine; Promega) at 40 U/μl, AMV reverse
transcriptase (Promega) at 10 U/μl, and RNase-free water, to a final volume of 10 μl. A control
without reverse transcriptase was used to confirm the absence of DNA contamination. RT-
qPCR was performed with SYBR Green technology on a CFX96 system (Bio-Rad, Marnes-la-
Coquette, France). Each reaction mix consisted of 1:50 diluted cDNA in a 4 μl final volume of
nuclease-free water; 0.5 μM of each forward and reverse primer for Hmox-1, Nqo1, Cat, Cxcl1,
Cxcl2, Ccl3, Il-6, Tnfα, Il-1β, Gapdh, β-actin; and Sso Advanced Supermix (Bio-Rad) in a total
reaction volume of 10 μl. The following specific primers were used (forward and reverse,
respectively): Hmox-1: 5’-AGG GTC AGG TGT CCA GAG AA-3’ and 5’-CTT CCA
GGG CCG TGT AGA TA-3’; Nqo1: 5’-ACG GGG ACA TGA ACG TCA TTC T-3’ and
5’-AGT GTG GCC AAT GCT GTA AAC C-3’; Cat: 5’-GTG GTT TTC ACT GAC
GAG ATG GCA-3’ and 5’-TCG TGG GTG ACC TCA AAG TAT CC-3’; Cxcl1: 5’-
GGC CCC ACT GCA CCC AAA CC-3’ and 5’-CCG AGC GAG ACG AGA CCA GGA
GA-3’; Cxcl2: 5’-CTC TCA AGG GCG GTC AAA AAG TT-3’ and 5’-TCA GAC AGC
GAG GCA CAT CAG GTA-3’; Ccl3: 5’-ACC ACT GCC CTT GCT-3’ and 5’-TGG
AAT CTT CCG GCT-3’; Il-6: 5’-AGT TGC CTT CTT GGG ACT GA-3’ and 5’-CAG
AAT TGC CAT TGC ACA AC-3’; Tnfα: 5’-CAC CAC GCT CTT CTG TCT AC-3’; Il-
PLOS ONE | https://doi.org/10.1371/journal.pone.0216465 August 16, 2019
4 / 21
The role of Nrf2 transcription factor in neutrophil biology
1β: 5’-ACA GCA GCA CAT CAA-3’ and 5’-GCA GGT TAT CAT CAT-3’; Gapdh:
5’-TGC ACC ACC AAC TGC TTA G-3’ and 5’-GAT CCA GGG ATG ATG TTC-3’;
β-actin: 5’-CCT TCT TGG GTA TGG AAT C-3’ and 5’-AGG TCT TTA CGG ATG
TCA AC-3’.
After 30 s at 95˚C for Sso7dfusion polymerase activation, amplification was allowed to pro-
ceed for 44 cycles, each consisting of denaturation at 95˚C for 5 s and annealing/extension at
60˚C for 5 s. Eightfold serial dilutions of mixed cDNA (from different samples) were analyzed
for each target gene, enabling us to construct linear standard curves from which the efficiency
(E) of each PCR run was evaluated. SYBR green fluorescence was detected at the end of each
elongation cycle, after which a melting curve was constructed to confirm the specificity of the
PCR products. Quantification was performed with CFX Manager Software (Bio-Rad), and
data were analyzed by the ΔΔCt method. Ratios were calculated as the geometric mean of (1
+ E) − ΔΔCt, where E is the efficiency and ΔΔCt is the target gene expression of treated cells
compared with normal levels in untreated cells, with correction for the expression of the refer-
ence genes β-actin and Gapdh. Results are expressed as the fold factor increase (i.e., ratio of (1
+ E)− ΔΔCt of treated cells/(1 + E) − ΔΔCt of untreated WT cells).
Analysis of ROS production
The sum of intra- and extra-cellular ROS produced was quantified by luminol (5-amino-
2,3-dihydro-1,4-phthalazindione)-amplified chemiluminescence assay. Isolated WT and Nrf2
KO PMNs were seeded at 1 × 105/well in a white flat bottom 96-well plate (Costar, Kennebunk
ME, USA) and treated with 0.06 mM luminol (Sigma-Aldrich). Cells were then stimulated
with increasing concentrations of zymosan (1, 5 and 10 μg/ml). In some experiments, a prein-
cubation with SFN 1 μM for 4 h was carried out before cell stimulation, allowing for optimal
Nrf2 expression. PMNs without stimulation were used as controls. ROS-dependent chemilu-
minescence was analyzed immediately using a multimode microplate reader (TristarTM
LB941 Berthold, Bad Wildbad, Germany). ROS release was monitored for 60 min every 30 sec
at 37˚C. All samples were tested in triplicate. The area under the curve (AUC) of each sample
was calculated.
Induction and quantification of neutrophil extracellular traps (NETs)
NET analysis was performed as previously described [36]. Briefly, staining with the non-cell-
permeable DNA dye SYTOXgreen (Invitrogen, Carlsbad, USA) was used to evaluate the kinet-
ics of extracellular DNA release. PMNs (1 × 105) in HBSS medium were seeded to a Cellstar
black 96-well plate (Greiner Bio-One, Frickenhausen, Germany). SYTOXgreen (5 μM) was
added 20 min before PMN stimulation or not with PMA 100 nM or zymosan 50 μg/ml. The
fluorescence of NET-bound SYTOXgreen (excitation: 488 nm, emission: 510 nm) was ana-
lyzed for a period of 3 h every 15 min at 37˚C using LB 941 Multimode reader TriStar. NET
release was calculated as the difference between the mean relative fluorescence unit (RFU) at
time 15 min and the RFU at time 180 min.
Phagocytosis analysis
Phagocytosis analysis was performed as per supplier’s instructions using pHrodo Red zymosan
BioParticles (Life Technologies, Carlsbad, USA) conjugate for phagocytosis. Briefly, isolated
WT and Nrf2 KO PMNs (1 × 106/ml) were seeded in duplicate wells, pretreated or not with
cytochalasin D (20 μM) (Sigma-Aldrich) at 37 ˚C with 5% CO2 for 30 min and then incubated
for 90 min in the dark, alone or with increasing concentrations of zymosan bioparticles (5–
50 μg/ml). Using flow cytometry (FACS Calibur), phagocytosis was quantified by the increase
PLOS ONE | https://doi.org/10.1371/journal.pone.0216465 August 16, 2019
5 / 21
The role of Nrf2 transcription factor in neutrophil biology
in particle fluorescence in acidic compartments. Cells were subjected to a one-color analysis
(FL-3, PerCP) for the percent of zymosan positive cells.
Chemotaxis assay
Chemotaxis assay was performed on freshly isolated WT and Nrf2 KO PMNs using transwell
migration assay, as previously described [37]. Briefly, PMNs (1 × 106 cells) were added to the
upper chamber of Transwell filters (3 μm pore diameter, Costar). These chambers were placed
in 24-well cell culture plates containing 600 μL assay buffer without chemoattractant or with
N-Formylmethionyl-leucyl-phenylalanine (fMLP, 1 μM) (Sigma-Aldrich), increasing concen-
trations of CXCL2 (0.5–200 nM) and increasing concentrations of CXCL12 (50–400 nM) (both
from BioLegend). In some experiments, cells were preincubated with AMD3100 octahy-
drochloride 50 nM (Sigma-Aldrich) for 30 min before being placed in the upper Transwell
chamber to confirm the specificity of CXCR4-dependent migration. Chambers were then incu-
bated for 60 min at 37˚C with 5% CO2 and the cells that had migrated to the bottom chamber
were recovered and stained with antibodies against Ly6G (PercP. Cy5.5) and CD11b (FITC)
(both from BioLegend) for flow cytometry analysis (FACS Calibur). Chemotactic indexes were
then calculated by dividing the number of PMNs counted in chemokine-stimulated wells by the
number of PMNs counted in filter-free wells (input well without any chemokine).
Quantification of receptor expression
The expression of TLR2, CXCR2 and CXCR4 on freshly isolated WT and KO PMNs was eval-
uated using flow cytometry. First, PMNs were incubated with anti-FcR antibody (anti-CD16/
CD32, BD Biosciences) at 4˚C for 15 min. Then, cells were washed and incubated in the dark
with antibodies against Ly6G, TLR2, CXCR2 and CXCR4 or with corresponding isotypes (all
from BioLegend). Ly6G positive cells were subjected to a double-color analysis to measure the
mean fluorescence intensity (MFI) for receptors.
ATP measurement
The level of ATP was measured in freshly isolated PMNs using Luminescence ATP detection
Assay Kit (Abcam) following the manufacturer’s instructions. Isolated resting WT and Nrf2
KO PMNs were seeded at 1 × 105/well in a white flat bottom 96-well plate (Costar) without
any stimulation. Samples were tested in triplicate. Luminescence was quantified using the mul-
timode microplate reader (TristarTM LB941 Berthold) and then converted to ATP concentra-
tion (in μM) using standard curve.
Statistical analysis
Nonparametric analyses were performed using GraphPad Prism software: the Mann Whitney
test was used to compare two independent groups and the Kruskal-Wallis test for more than
two independent groups. Data are expressed as means ± SEM. P<0.05 was considered to
denote statistical signficance.
Results
Nrf2 is inducible in BM-derived WT PMNs and triggers the activation of its
target genes
As a first step, we assessed Nrf2 expression in WT PMNs since Nrf2 is an ubiquitous transcrip-
tion factor, constantly ubiquitinilated by its cytosolic repressor Keap1 [38]. In a first set of
experiments, BM cells from WT mice were stained directly after filtration in order to quantify
PLOS ONE | https://doi.org/10.1371/journal.pone.0216465 August 16, 2019
6 / 21
The role of Nrf2 transcription factor in neutrophil biology
Nrf2 endogenous expression in PMNs, in comparison to other BM cells. As expected, Nrf2
was expressed in almost all BM cells; interestingly, its expression in PMNs (identified as Ly6G+
cells) was at least 2-fold higher than in the other BM cells (Fig 1A).
As Nrf2 was highly expressed in WT PMNs, we evaluated its accumulation in response to
SFN (a well-known Nrf2 activator) used as a positive control stimulus, or to zymosan a strong
PMN stimulus. Freshly isolated BM-derived WT PMNs were thus incubated alone, with SFN
Fig 1. Nrf2 endogenous expression is high in PMNs and increases in response to SFN and zymosan. Freshly obtained BM cells were stained with
anti-Ly6G and rabbit anti-Nrf2 IgG or with corresponding isotype, then with goat anti-rabbit Alexa Fluor 488 Ab. Isolated PMNs were incubated for 4
h alone (NS), with 1 μM of SFN before staining or with zymosan 5 μg/ml. (A) Endogenous Nrf2 was quantified in BM cells using intracellular staining
followed by flow cytometry. PMNs were identified as Ly6G+, while Ly6G- cells represents the rest of BM cells. Nrf2 expression was quantified as the
mean fluorescence intensity (MFI) and compared between Ly6G+ and Ly6G- cells (#p<0.05 and ##p<0.01, Kruskal-Wallis test). (B) Nrf2 accumulation
in isolated PMNs after incubation with SFN 1 μM or zymosan 5 μg/ml, was measured using flow cytometry. The ratio of Nrf2 expression was calculated
by dividing the MFI of treated PMNs by the MFI of untreated PMNs (NS). Ratios greater than 1 indicate Nrf2 accumulation in stimulated PMNs
(Mann-Whitney test). Results are the mean ± SEM of samples from 4 independent experiments, n = 4.
https://doi.org/10.1371/journal.pone.0216465.g001
PLOS ONE | https://doi.org/10.1371/journal.pone.0216465 August 16, 2019
7 / 21
The role of Nrf2 transcription factor in neutrophil biology
or with zymosan, for 4 h. Using flow cytometry, we evidenced that zymosan as well as SFN
1 μM were able to significantly increase the intracellular staining of Nrf2 in WT PMNs, indi-
cating its accumulation (Fig 1B).
Finally, the transcription of three downstream target genes of Nrf2 (Nqo1, Hmox-1 and
Cat) was evaluated in WT and Nrf2 KO PMNs, in response to SFN and zymosan. In accor-
dance with flow cytometry results, SFN and zymosan induced the transcription of Nqo1,
Hmox-1 and Cat in WT PMNs (Fig 2).
Taken together, these results strongly suggest that like SFN, zymosan can induce Nrf2 accu-
mulation and activate its target genes in WT PMNs. As expected, no Nrf2 target gene tran-
scription was observed in Nrf2 KO PMNs.
Nrf2 activation in PMNs participates in the regulation of zymosan-induced
oxidative burst
As Nrf2 is inducible in PMNs, we aimed to study its implication in oxidative burst. First, a
luminol-amplified chemiluminescence assay was used to evaluate the role of Nrf2 on both
Fig 2. Nrf2 upregulates the transcription of 3 main target genes in response to SFN and zymosan. WT and Nrf2 KO PMNs were incubated or not (NS)
with SFN 1μM or zymosan 5 μg/ml for 4 h. mRNA expression of Nqo1, Hmox-1 and Cat was quantified using RT-qPCR. Results are expressed as fold
increase normalized to WT NS and corrected by the expression of the housekeeping genes β-actin and gapdh in all RT-qPCR experiments. Results are the
mean ± SEM of samples from 4 independent experiments, n = 4 (� indicates a significant difference between WT and Nrf2 KO PMNs throughout the study,
�p<0.05, Mann-Whitney test).
https://doi.org/10.1371/journal.pone.0216465.g002
PLOS ONE | https://doi.org/10.1371/journal.pone.0216465 August 16, 2019
8 / 21
The role of Nrf2 transcription factor in neutrophil biology
intra- and extracellular ROS production in response to zymosan. In comparison to WT, PMNs
from Nrf2 KO mice exhibit a shift in the kinetics of ROS production, characterized by an early
peak at 25 min following stimulation with zymosan (Fig 3A). In response to increasing con-
centrations of zymosan (1, 5 and 10 μg/ml), we quantified a significantly higher production of
ROS in Nrf2 KO PMNs than in WT PMNs suggesting an implication of Nrf2 in the regulation
of ROS production (Fig 3A and 3B). As zymosan-induced activation is dependent on TLR2
activation, we wondered if Nrf2 KO PMNs displayed a higher basal expression of TLR2. Flow
cytometric analysis showed that PMNs from both genotypes displayed a similar expression of
TLR2 (Fig 3C). Similar results were observed with dectin-1 expression, another receptor
involved in zymosan-induced PMN activation (data not shown).
Next, the effect of a significant Nrf2 accumulation on the regulation of ROS production was
assessed. WT and Nrf2 KO PMNs were thus pre-treated or not with 1 μM of SFN at 37˚C for 4
h, before stimulation with zymosan at 5 μg/ml. As shown in Fig 3D, SFN-preincubation more
significantly decreased ROS production in WT PMNs than in Nrf2 KO PMNs suggesting that
Nrf2 activation decreases ROS production in PMNs.
As netosis can be related to ROS production in most conditions, we wanted to examine the
possibility that Nrf2 might also modulate this function, using the SYTOXgreen assay. We
found that PMA-induced extracellular DNA release was similar in WT and KO PMNs, as
zymosan failed to induce netosis in PMNs from WT and Nrf2 KO mice (S3 Fig). We could
thus rule out a potential effect for Nrf2 in these conditions.
Altogether, these results provide evidence that Nrf2 activation participates in the regulation
of ROS production in BM-derived PMNs in response to zymosan independently of TLR2
expression level.
Nrf2 modulates the transcription of some pro-inflammatory genes
TNFα, IL-6 and IL-1β are key pro-inflammatory cytokines produced by PMNs. We sought to
demonstrate whether Nrf2 could control the transcription of genes encoding these cytokines
as well as the chemokines CCL3, CXCL1 and CXCL2. For that, their transcriptional level was
quantified in WT and Nrf2 KO PMNs stimulated or not with zymosan 5 μg/ml for 4 h. Inter-
estingly, the lack of Nrf2 allowed for considerable transcription of genes encoding CCL3,
CXCL2, TNFα and IL-1β in response to zymosan. As shown in Fig 4, Nrf2 KO PMNs dis-
played a significantly enhanced Tnfα, Cxcl2 and Ccl3 mRNA expression as compared to WT
PMNs, while Il-6, Cxcl1 and Il-1β mRNA expressions were not significantly increased in the
absence of Nrf2.
Nrf2 is not involved in the phagocytic capacity of PMNs
We then compared the phagocytic capacity between WT and Nrf2 KO PMNs. This function
was analyzed using pHrodo Red Zymosan Bioparticles Conjugates. We found that WT and
Nrf2 KO PMNs showed similar capacity to phagocyte zymosan bioparticles after 90 min, in a
concentration-dependent manner. Similarly, the inhibition effect of Cytochalasin D (inhibitor
of actin polymerization) was similar in both genotypes (Fig 5A and 5B). These data suggest
that Nrf2 does not seem to be involved in the phagocytosis of zymosan particles by PMNs.
Nrf2 is required for an optimal PMN migration independently of
chemokine receptor expression and ATP level
We finally studied the role of Nrf2 in the modulation of PMN migration. We focused on two
axes (CXCR2 and CXCR4) because in vivo PMN recruitment relies heavily on their ligands
CXCL2 and CXCL12 respectively; fMLP was used as a positive control. Using a transwell
PLOS ONE | https://doi.org/10.1371/journal.pone.0216465 August 16, 2019
9 / 21
The role of Nrf2 transcription factor in neutrophil biology
Fig 3. Nrf2 activation decreases PMN ROS production in response to zymosan. WT and Nrf2 KO PMNs were stimulated or not
with increasing concentrations of zymosan. The sum of intra- and extracellular ROS was monitored for 60 min by the luminol-
amplified chemiluminescence assay. (A) Graphs from one representative experiment show the kinetic of ROS production in
unstimulated PMNs (left panel) or in response to zymosan 10 μg/ml (right panel). (B) The area under curve (AUC) from the kinetic
curves of 6 independent experiments was used to calculate the ratio of ROS production (KO AUC/WT AUC). Ratios greater than 1
correspond to higher ROS production in KO PMNs (��p<0.01, Mann-Whitney test). (C) The expression of TLR2 was assessed on
freshly isolated WT and Nrf2 KO PMNs (grey filled curves) in comparison to corresponding isotypes (black unfilled curves). The
associated graph represents the mean ± SEM of TLR2 MFI, n = 4. (D) To ensure an important Nrf2 accumulation, PMNs were
pretreated or not (NT) with SFN 1 μM for 4 h then stimulated with 5 μg/ml of zymosan. ROS production was then measured for 60
PLOS ONE | https://doi.org/10.1371/journal.pone.0216465 August 16, 2019
10 / 21
The role of Nrf2 transcription factor in neutrophil biology
min by the luminol-amplified chemiluminescence assay. Graphs show the kinetics of ROS production from one representative
experiment (left panel) and the mean ± SEM of AUC from the kinetic curves of 6 independent experiments (right panel, n = 6;
�p<0.05, Mann-Whitney test).
https://doi.org/10.1371/journal.pone.0216465.g003
migration assay, PMNs were allowed to migrate for 1 h spontaneously, toward 1 μM of fMLP
or toward gradual concentrations of CXCL2 and CXCL12. In the absence of Nrf2, PMN spon-
taneous migration and migration in response to fMLP were significantly decreased as com-
pared with WT PMN migration. Concerning the response to increasing concentrations of
CXCL12, both WT and Nrf2 KO PMNs displayed a relatively weak capacity to migrate. In-
terestingly, Nrf2 KO PMNs displayed a significantly decreased capacity to migrate in response
to the lowest concentration of CXCL12 as compared to WT PMNs. This migration was
completely abolished in the presence of AMD, a competitive inhibitor of CXCL12. Concern-
ing the dose-response of CXCL2, two distinct bell curves could be observed, one shifted in
relation to the other. Indeed, Nrf2 KO PMNs displayed a significantly decreased capacity to
migrate in response to the two lower concentrations of CXCL2, while migration in response
to higher concentrations was similar in both WT and Nrf2 KO PMNs (Fig 6A). These results
suggest that Nrf2 could participate in PMN motility by increasing their sensitivity to
chemoattractants.
In order to assess whether these differences in migration in the absence of Nrf2 could be
linked to differences in the expression of chemokine receptors, we analyzed CXCR4 and
CXCR2 membrane expression on freshly isolated PMNs. As observed, WT and Nrf2 KO
PMNs displayed similar expression of both receptors (Fig 6B), suggesting that Nrf2 can partici-
pate in PMN migration independently of chemokine receptor expression.
In order to better understand the mechanism of Nrf2 KO PMN affected migration, we ana-
lyzed the PMN phenotype after 1 h of migration by quantifying the expression of the β2 integ-
rin CD11b/CD18 adhesion molecule. We found that the migration-induced upregulation of
CD11b was significantly increased in Nrf2 KO PMNs as compared with WT PMNs, in
response to low concentrations of CXCL2 (Fig 6C).
Finally, as ATP is mandatory for an optimized migration we hypothesized that ATP levels
could differ in Nrf2 KO PMNs. We found no significant difference between both genotypes at
resting state, even if ATP levels seemed to be slightly higher in WT PMNs (Fig 6D).
Discussion
PMNs play a key role in host defense against pathogens, but inadequate or excessive activation
can lead to deleterious effects, contributing to the pathophysiology of many acute and chronic
inflammatory diseases. In parallel, Nrf2 plays an active role in the control of inflammation, via
several mechanisms. In this study, using an in vitro model, we found that Nrf2 participates in
the regulation of murine BM-derived PMN response to the fungal stimulus zymosan, via the
activation of cytoprotective genes and the downregulation of pro-inflammatory genes. ROS
production depends on Nrf2 activation, while phagocytosis is similar in WT and Nrf2 KO
mice PMNs. Our results thus highlight the impact of Nrf2 on PMN activation, providing new
findings relevant to the regulation of neutrophil activation by zymosan.
PLOS ONE | https://doi.org/10.1371/journal.pone.0216465 August 16, 2019
11 / 21
The role of Nrf2 transcription factor in neutrophil biology
Fig 4. Nrf2 modulates Ccl3, Cxcl2 and Tnfα transcription. WT and Nrf2 KO PMNs were incubated or not with zymosan 5 μg/ml for 4
h. mRNA expression of Ccl3, Cxcl2, Cxcl1, Tnfα, Il-1β and Il-6, were measured using RT-qPCR. All results are expressed as fold increase
normalized to WT NS and corrected by the expression of the housekeeping genes β-actin and gapdh (4 independent experiments, n = 4;
�p<0.05 Mann-Whitney test).
https://doi.org/10.1371/journal.pone.0216465.g004
Several studies have led to the consensus that PMNs are able to finely regulate both innate
and adaptive immune responses, engaging with T lymphocytes or antigen-presenting cells,
releasing anti-inflammatory mediators or expressing regulatory receptors; this broader role in
immunity also led to the definition of several functional PMN subpopulations [4,39–41]. In
particular, our group evidenced that isolated NETs or neutrophil-derived ectosomes were able
to downregulate LPS-induced DC maturation and their capacity to induce T lymphocyte pro-
liferation [7,42]. We also evidenced that GILZ, a potent anti-inflammatory mediator
PLOS ONE | https://doi.org/10.1371/journal.pone.0216465 August 16, 2019
12 / 21
The role of Nrf2 transcription factor in neutrophil biology
Fig 5. The lack of Nrf2 does not affect PMN phagocytic capacity. Cells were incubated for 90 min alone or with the
determined concentrations of pHRodo Red Zymosan Bioparticles. (A) Flow cytometry-based analysis was used to
evaluate PMN zymosan uptake. PMNs were subjected to a one-color analysis for the percent of zymosan (25 μg/ml)
positive events. (B) Graph representing the percentage of PMNs that phagocytosed increasing concentrations of
zymosan (5, 25 and 50 μg/ml). PMNs were pretreated for 30 min with 20 μM of cytochalasin D for negative control. All
data are presented as the mean ± SEM, 4 independent experiments, n = 4.
https://doi.org/10.1371/journal.pone.0216465.g005
PLOS ONE | https://doi.org/10.1371/journal.pone.0216465 August 16, 2019
13 / 21
The role of Nrf2 transcription factor in neutrophil biology
Fig 6. Nrf2 is necessary for optimal PMN migration toward CXCL12 and CXCL2 without regulating receptors expression. (A) WT
and Nrf2 KO PMNs were allowed to migrate for 1 h across transwell filters spontaneously or toward fMLP 1 μM, CXCL12 or CXCL2. The
negative control in CXCL12-dependent migration was obtained by pretreating PMNs with 50 μM of AMD300. Chemotactic indexes were
then calculated by dividing the number of PMNs that were counted in the chemokine-stimulated well by the number of PMNs that were
counted in the input well (�p<0.05 ��p<0.01 Mann-Whitney test). (B) The expression of CXCR2 (B, left panel) and CXCR4 (B, right panel)
were assessed on freshly isolated WT PMNs (solid grey line) and Nrf2 KO PMNs (dashed black line) in comparison to corresponding
isotypes (unfilled curves). The associated graph represents the mean ± SEM of CXCR2 and CXCR4 MFI, n = 4. (C) The MFI fold variation
of CD11b was assessed on PMNs that have migrated through the transwell in comparison to PMNs in the input well. (D) ATP levels in
PLOS ONE | https://doi.org/10.1371/journal.pone.0216465 August 16, 2019
14 / 21
The role of Nrf2 transcription factor in neutrophil biology
freshly isolated PMNs were quantified using an ATP luminescent assay. The expression levels from 6 independent experiments are shown
as the mean of MFI ± SEM (n = 6).
https://doi.org/10.1371/journal.pone.0216465.g006
implicated in cell survival, was present in human neutrophils, promoted apoptosis via Mcl-1
downregulation, and was upregulated in patients with the acute respiratory distress syndrome
in relation to severity [11,43]. Our group is also very involved in studying the role of Nrf2 in
several pathophysiological conditions. For instance, we showed that Nrf2 can control inflam-
mation in a model of allergic contact dermatitis (both sensitization and effector phases) [44]
and during sclerodermia [21]. We also highlighted the role of Nrf2 in the control of DC cell
death induced by chemical sensitizers and we found that Nrf2 positively controls antioxidant
genes like gsr, catalase, gpx, nos2 and some immune genes in response to contact sensitizers
[19]. We thus addressed the question of the potential immunoregulatory role of Nrf2 in
numerous PMN functions.
Our first aim was to study Nrf2 activation in BM-derived PMNs. Joshi et al. recently dem-
onstrated that the transcription level of Nrf2 was strong in circulating PMNs and in wound tis-
sue PMNs [32]. Here, using flow cytometry, we confirm that Nrf2 is highly expressed in BM-
derived WT PMNs as compared to other BM-derived cells. Moreover, using SFN, a well-char-
acterized Nrf2 activator, we demonstrate that Nrf2 can be mobilized and activated in WT
PMNs as several target genes were induced; SFN was particularly potent for the activation of
Nqo1 and Cat. In addition, our study evidences for the first time that zymosan, a TLR2 agonist
is also able to mobilize Nrf2 in WT PMNs and to activate the same target genes. These results
showing the important activity of Nrf2 in PMNs are consistent with those of Joshi et al. in
mouse blood PMNs [32], and with those of R Thimmulappa et al. in human PMNs [30] and
VC Araujo et al. in human PMNs from fungal oral granuloma [45].
Since Keap1/Nrf2 signaling maintains redox homeostasis in the cell, we studied the role of
Nrf2 in PMN oxidative burst. We thus performed a set of experiments aiming to document
the contribution of Nrf2 in ROS production. C Sima et al. have recently reported that in
response to PMA, Nrf2 KO PMNs displayed a normal ROS production [31]. Conversely, an
enhanced ROS release in Nrf2 KO PMNs has been described in the literature in response to
LPS; however, experimental conditions were quite different as PMNs were collected from the
peritoneal fluid 4 h after thioglycolate injection [24]. In our study, we were able to detect a low
but significant Nrf2-mediated regulation of ROS production in response to zymosan, indepen-
dently of TLR2 and dectin-1 expression level. This suggests that Nrf2 could rather intervene in
the downstream signaling of these receptors. We chose non opsonized zymosan as a stimulus
as it has been demonstrated that it induces NADPH oxidase activation in human blood PMNs
leading to high ROS production [33]. Since zymosan induces Nrf2 accumulation, the antago-
nistic activities of Nrf2 and NADPH oxidase [27] could explain the exacerbated zymosan-
induced ROS production in the absence of Nrf2. Our second set of experiments was designed
to greatly upregulate Nrf2 before stimulating oxidative burst. We indeed found that SFN pre-
treatment reduced zymosan-induced ROS to a greater extent in WT PMNs than in Nrf2 KO
PMNs, suggesting that Nrf2 has a crucial role in ROS regulation in BM-derived PMNs. Of
note, we only used here low concentrations of SFN to avoid activation of other signaling path-
ways [46]. The latter results are consistent with those of Thimmulappa et al. using the triterpe-
noid CDDO-Im as Nrf2 enhancer or those of Dias et al. using SFN in human PMNs [47,48].
PLOS ONE | https://doi.org/10.1371/journal.pone.0216465 August 16, 2019
15 / 21
The role of Nrf2 transcription factor in neutrophil biology
Using live cell imaging of brain hippocampal glio-neuronal cultures, other authors also found
that a graded expression of Nrf2 paralleled a graded production of ROS [27].
Several studies have documented the Nrf2-induced decrease in transcriptional expression
of some pro-inflammatory cytokines in various cell types, such as human epithelial cells or
macrophages [49,50]. In PMNs, few results have been reported. Here we have clearly shown
that, in response to zymosan, Nrf2 KO PMNs displayed a significantly increased Ccl3, Cxcl2
and Tnfα mRNA expression. This suggests that Nrf2 can modulate inflammatory gene tran-
scription in BM-derived mice PMNs, as already suggested in peritoneal PMNs in a model of
LPS-induced inflammation [30] and in human blood PMNs [47]. Moreover, we would suggest
that Nrf2 activation in zymosan-induced peritonitis could limit neutrophil activation and
auto-recruitment mainly via the regulation of TNFα, CXCL2 and CCL3 production in the
peritoneal fluid [51]. Concerning IL-6, WT and Nrf2 KO PMNs exhibited similar transcript
levels, in accordance with N Joshi et al. recent study [32]. Although Nrf2 decreases the tran-
scriptional upregulation of Il-6 and Il-1β in LPS-stimulated macrophages through the inhibi-
tion of RNA Pol II recruitment, no direct interference could be observed in PMNs [25].
Interestingly, it was also reported that Nrf2 and its target genes were not needed for interferon
γ production by lung PMNs in a mouse model of pneumonia [52]. We can thus suggest that,
Nrf2 is needed for the transcriptional regulation of several but not all inflammatory cytokines
and chemokines in PMNs.
We also examined whether the lack of Nrf2 can affect PMN capacity to phagocyte zymosan
particles. We found that Nrf2 KO and WT PMNs exhibited similar capacity to phagocyte
zymosan bioparticles. While Nrf2 KO peritoneal macrophages have been described to exhibit
impaired phagocytosis during sepsis as compared with WT cells [24], our results suggest that
Nrf2 is not involved in PMN phagocytosis, unlike macrophages. This result highlights several
specific roles of Nrf2 depending on cell type.
The role of ROS in PMN migration has been widely studied over the past few years [53,54].
A potential link between Nrf2 and PMN migration can thus be suspected. Although some
studies have shown an enhanced PMN recruitment in Nrf2 KO mice in different inflammatory
settings [55,56], the use of a full Nrf2 KO model could not lead to clear conclusions concerning
Nrf2 role in PMN migration; in particular, Sima et al. suggested that Nrf2 could slow down
fMLP-induced PMN migration in vitro as early as 15 min of migration [57]. We thus further
investigate this mechanism, using an one hour-transwell migration assay with fMLP and
increasing concentrations of two other chemokines, CXCL2 and CXCL12 that antagonistically
modulate PMN chemotaxis [58,59]. In contrast with ROS shifted curves, the chemokine bell-
shaped migration curves suggested that Nrf2 improved PMN ability to migrate toward fMLP
or low concentrations of CXCL2 and CXCL12. Of note, WT and Nrf2 KO PMNs displayed a
similar expression of CXCR2 and CXCR4. Nrf2 has been previously reported to partially con-
trol hematopoiesis through the regulation of CXCR4 signaling [60]. Here, we saw that Nrf2
could also regulate PMN CXCR2- and CXCR4-dependent migration independently from the
expression level of both receptors. To better understand the mechanism of Nrf2-modulated
migration, we compared the modification of expression of β2 integrin CD11b/CD18 after 1 h
of migration and found that Nrf2 KO PMNs exhibit a significant upregulation of CD11b even
in response to low concentrations of CXCL2. CD11b upregulation on in vitro-migrating
PMNs has already been described, in particular in a recent model of airway epithelial cells
[61]. This suggests that Nrf2 ensures an optimal migration of PMNs via the control of migrat-
ing cell activation level. Some studies have evidenced that Nrf2 could potentially enhance the
synthesis of ATP in various cell models [62,63]. As ATP is crucial for PMN migration [64], we
thus hypothesized that ATP could be involved in the modified migration observed in Nrf2 KO
PMNs. However, we failed to evidence a significant difference between ATP levels in resting
PLOS ONE | https://doi.org/10.1371/journal.pone.0216465 August 16, 2019
16 / 21
The role of Nrf2 transcription factor in neutrophil biology
WT and Nrf2 KO PMNs suggesting that Nrf2 plays a weak role in this setting. Based on our
findings, we can suggest that the intense recruitment of PMNs observed in Nrf2 KO mice in
several inflammatory in vivo models [55,56] might not be related to an intrinsic control of
PMN migration by Nrf2. Nevertheless, Nrf2 could ensure an optimal PMN motility and che-
mokine sensing needed, in particular to drive PMN migration from the BM to the tissues at
the very early stages of inflammation.
This in vitro evaluation of Nrf2 implication in PMN functions need also to be analyzed
from a clinical point of view as Nrf2 diseasome and drugome have been recently defined [34].
Indeed, in the numerous clinical settings associated with significant PMN tissue infiltration
and/or PMN activation, it is important to better understand the implication and the efficiency
of regulatory mechanisms including Nrf2 pathway. For instance, in the lung, activating Nrf2
leads to protective effects during acute lung injury or asthma, but also enhances advanced
stages of carcinogenesis [65,66]. We can thus assume that activating or inhibiting Nrf2 in
PMN-associated diseases is an important issue that needs to be fully considered.
Taken together, our results contribute to a better understanding of the Nrf2 protective role
in PMNs, particularly in response to zymosan that is a physiological stimulus mimicking yeast
infection, and widely used in experimental models of arthritis or peritonitis. This study indi-
cates that Nrf2 cytoprotective target genes are inducible in PMNs and highlights its oxidative
burst regulation capacity. In addition, this study demonstrates that Nrf2 is needed for an opti-
mal PMN migration. Future research into understanding the role of Nrf2 in PMN recruitment
in a context of inflammation may provide insight and novel approaches in the field of
inflammation.
Supporting information
S1 Fig. Flow cytometry-based assessment of PMN purity. Following negative isolation,
PMN purity was assessed using antibodies against Ly6G and CD11b. Ly6G+, CD11b+ cells rep-
resent BM PMN. PMN purity was > 95% in all experiments.
(TIF)
S2 Fig. Assessment of SFN toxicity through AnnexinV-7AAD staining. PMN were incu-
bated alone or with the indicated concentrations of SFN for 4 h, and then stained with Annex-
inV and 7-AAD. AnnV+ cells represent apoptotic cells while double positive cells
(AnnV+,7-AAD+) represent necrotic cells. Data are shown as representative FACS analysis (A)
and as the mean ± SEM of 3 independent experiments (B).
(TIF)
S3 Fig. Zymosan fails to induce DNA release in netosis. PMN were incubated for 3 h, alone
or with PMA 100 nM and zymosan 50 μg/ml. Time dependent DNA release was monitored
using the fluorescent SYTOXgreen. (A) Data from one representative experiment shows the
kinetic of DNA release in response to PMA 100 nM. (B) Results from 4 independent experi-
ments are expressed as the difference between RFU at time 15 min and the RFU at time 180
min.
(TIF)
Acknowledgments
We thank Dr. Viviana Marin-Esteban and Dr. Karl Balabanian for providing expertise in neu-
trophil migration and Dr. Claudine Delome´nie for her expertise in transcriptional analysis.
We also thank Vale´rie Domergue, Ayma Galland and the Institut Fe´de´ratif de Recherche
IPSIT for excellent technical assistance for animal testing.
PLOS ONE | https://doi.org/10.1371/journal.pone.0216465 August 16, 2019
17 / 21
The role of Nrf2 transcription factor in neutrophil biology
Author Contributions
Conceptualization: Sarah Braham, Sylvie Chollet-Martin.
Formal analysis: Doumet Georges Helou.
Funding acquisition: Saadia Kerdine-Ro¨mer, Sylvie Chollet-Martin.
Methodology: Luc De Chaisemartin, Vanessa Granger, Marie-He´lène Damien, Saadia Ker-
dine-Ro¨mer.
Supervision: Saadia Kerdine-Ro¨mer, Sylvie Chollet-Martin.
Validation: Saadia Kerdine-Ro¨mer, Sylvie Chollet-Martin.
Writing – original draft: Doumet Georges Helou, Saadia Kerdine-Ro¨mer, Sylvie Chollet-
Martin.
Writing – review & editing: Doumet Georges Helou, Luc De Chaisemartin, Vanessa Granger,
Marc Pallardy, Saadia Kerdine-Ro¨mer, Sylvie Chollet-Martin.
References
1. McDonald B, Davis RP, Kim S-J, Tse M, Esmon CT, Kolaczkowska E, et al. Platelets and neutrophil
extracellular traps collaborate to promote intravascular coagulation during sepsis in mice. Blood. 2017;
blood-2016-09-741298. https://doi.org/10.1182/blood-2016-09-741298 PMID: 28073784
2.
Leliefeld PHC, Wessels CM, Leenen LPH, Koenderman L, Pillay J. The role of neutrophils in immune
dysfunction during severe inflammation. Crit Care. 2016; 20: 73. https://doi.org/10.1186/s13054-016-
1250-4 PMID: 27005275
3. Soehnlein O, Steffens S, Hidalgo A, Weber C. Neutrophils as protagonists and targets in chronic inflam-
mation. Nat Rev Immunol. 2017; 17: 248. https://doi.org/10.1038/nri.2017.10 PMID: 28287106
4. Scapini P, Marini O, Tecchio C, Cassatella MA. Human neutrophils in the saga of cellular heterogeneity:
insights and open questions. Immunol Rev. 2016; 273: 48–60. https://doi.org/10.1111/imr.12448 PMID:
27558327
5. Reber LL, Gillis CM, Starkl P, Jo¨ nsson F, Sibilano R, Marichal T, et al. Neutrophil myeloperoxidase
diminishes the toxic effects and mortality induced by lipopolysaccharide. J Exp Med. 2017; 214: 1249–
1258. https://doi.org/10.1084/jem.20161238 PMID: 28385925
6. Odobasic D, Kitching AR, Holdsworth SR. Neutrophil-Mediated Regulation of Innate and Adaptive
Immunity: The Role of Myeloperoxidase. J Immunol Res. 2016; 2016. https://doi.org/10.1155/2016/
2349817 PMID: 26904693
7. Barrientos L, Bignon A, Gueguen C, Chaisemartin L de, Gorges R, Sandre´ C, et al. Neutrophil Extracel-
lular Traps Downregulate Lipopolysaccharide-Induced Activation of Monocyte-Derived Dendritic Cells.
J Immunol. 2014; 193: 5689–5698. https://doi.org/10.4049/jimmunol.1400586 PMID: 25339673
8. Breedveld A, Groot Kormelink T, van Egmond M, de Jong EC. Granulocytes as modulators of dendritic
cell function. J Leukoc Biol. 2017; 102: 1003–1016. https://doi.org/10.1189/jlb.4MR0217-048RR PMID:
28642280
9. Singel KL, Segal BH. NOX2-dependent regulation of inflammation. Clin Sci Lond Engl 1979. 2016; 130:
479–490. https://doi.org/10.1042/CS20150660 PMID: 26888560
10. Baudhuin J, Migraine J, Faivre V, Loumagne L, Lukaszewicz A-C, Payen D, et al. Exocytosis acts as a
modulator of the ILT4-mediated inhibition of neutrophil functions. Proc Natl Acad Sci U S A. 2013; 110:
17957–17962. https://doi.org/10.1073/pnas.1221535110 PMID: 24133137
11. Espinasse M-A, Hajage D, Montravers P, Piednoir P, Dufour G, Tubach F, et al. Neutrophil expression
of glucocorticoid-induced leucine zipper (GILZ) anti-inflammatory protein is associated with acute respi-
ratory distress syndrome severity. Ann Intensive Care. 2016; 6. https://doi.org/10.1186/s13613-016-
0210-0 PMID: 27807817
12.
13.
Taguchi K, Motohashi H, Yamamoto M. Molecular mechanisms of the Keap1–Nrf2 pathway in stress
response and cancer evolution. Genes Cells. 2011; 16: 123–140. https://doi.org/10.1111/j.1365-2443.
2010.01473.x PMID: 21251164
Zhang DD. Mechanistic Studies of the Nrf2-Keap1 Signaling Pathway. Drug Metab Rev. 2006; 38: 769–
789. https://doi.org/10.1080/03602530600971974 PMID: 17145701
PLOS ONE | https://doi.org/10.1371/journal.pone.0216465 August 16, 2019
18 / 21
The role of Nrf2 transcription factor in neutrophil biology
14. Hayes JD, McMahon M, Chowdhry S, Dinkova-Kostova AT. Cancer Chemoprevention Mechanisms
Mediated Through the Keap1–Nrf2 Pathway. Antioxid Redox Signal. 2010; 13: 1713–1748. https://doi.
org/10.1089/ars.2010.3221 PMID: 20446772
15. Dinkova-Kostova AT, Holtzclaw WD, Kensler TW. The role of Keap1 in cellular protective responses.
Chem Res Toxicol. 2005; 18: 1779–1791. https://doi.org/10.1021/tx050217c PMID: 16359168
16. Kobayashi E, Suzuki T, Yamamoto M. Roles Nrf2 Plays in Myeloid Cells and Related Disorders. Oxid
Med Cell Longev. 2013;2013. https://doi.org/10.1155/2013/529219 PMID: 23819012
17.
Itoh K, Mochizuki M, Ishii Y, Ishii T, Shibata T, Kawamoto Y, et al. Transcription Factor Nrf2 Regulates
Inflammation by Mediating the Effect of 15-Deoxy-Δ12,14-Prostaglandin J2. Mol Cell Biol. 2004; 24:
36–45. https://doi.org/10.1128/MCB.24.1.36-45.2004 PMID: 14673141
18. Al-Huseini LMA, Yeang HXA, Sethu S, Alhumeed N, Hamdam JM, Tingle Y, et al. Nuclear Factor-ery-
throid 2 (NF-E2) p45-related Factor-2 (Nrf2) Modulates Dendritic Cell Immune Function through Regu-
lation of p38 MAPK-cAMP-responsive Element Binding Protein/Activating Transcription Factor 1
Signaling. J Biol Chem. 2013; 288: 22281–22288. https://doi.org/10.1074/jbc.M113.483420 PMID:
23775080
19. El Ali Z, Delome´ nie C, Botton J, Pallardy M, Kerdine-Ro¨mer S. Dendritic cells’ death induced by contact
sensitizers is controlled by Nrf2 and depends on glutathione levels. Toxicol Appl Pharmacol. 2017; 322:
41–50. https://doi.org/10.1016/j.taap.2017.02.014 PMID: 28219650
20. Mussotter F, Tomm JM, El Ali Z, Pallardy M, Kerdine-Ro¨mer S, Go¨ tz M, et al. Proteomics analysis of
dendritic cell activation by contact allergens reveals possible biomarkers regulated by Nrf2. Toxicol
Appl Pharmacol. 2016; 313: 170–179. https://doi.org/10.1016/j.taap.2016.11.001 PMID: 27816475
21. Kavian N, Mehlal S, Jeljeli M, Saidu NEB, Nicco C, Cerles O, et al. The Nrf2-Antioxidant Response Ele-
ment Signaling Pathway Controls Fibrosis and Autoimmunity in Scleroderma. Front Immunol. 2018; 9.
https://doi.org/10.3389/fimmu.2018.01896 PMID: 30177933
22. Yeang HXA, Hamdam JM, Al-Huseini LMA, Sethu S, Djouhri L, Walsh J, et al. Loss of Transcription
Factor Nuclear Factor-Erythroid 2 (NF-E2) p45-related Factor-2 (Nrf2) Leads to Dysregulation of
Immune Functions, Redox Homeostasis, and Intracellular Signaling in Dendritic Cells. J Biol Chem.
2012; 287: 10556–10564. https://doi.org/10.1074/jbc.M111.322420 PMID: 22311972
23. Olagnier D, Lavergne R-A, Meunier E, Lefèvre L, Dardenne C, Aubouy A, et al. Nrf2, a PPARγ Alterna-
tive Pathway to Promote CD36 Expression on Inflammatory Macrophages: Implication for Malaria.
PLoS Pathog. 2011; 7. https://doi.org/10.1371/journal.ppat.1002254 PMID: 21949655
24. Kong X, Thimmulappa R, Kombairaju P, Biswal S. NADPH Oxidase-Dependent Reactive Oxygen Spe-
cies Mediate Amplified TLR4 Signaling and Sepsis-Induced Mortality in Nrf2-deficient Mice. J Immunol
Baltim Md 1950. 2010; 185: 569–577. https://doi.org/10.4049/jimmunol.0902315 PMID: 20511556
25. Kobayashi EH, Suzuki T, Funayama R, Nagashima T, Hayashi M, Sekine H, et al. Nrf2 suppresses
macrophage inflammatory response by blocking proinflammatory cytokine transcription. Nat Commun.
2016; 7: ncomms11624. https://doi.org/10.1038/ncomms11624 PMID: 27211851
26. Ohl K, Fragoulis A, Klemm P, Baumeister J, Klock W, Verjans E, et al. Nrf2 Is a Central Regulator of
Metabolic Reprogramming of Myeloid-Derived Suppressor Cells in Steady State and Sepsis. Front
Immunol. 2018; 9. https://doi.org/10.3389/fimmu.2018.01552 PMID: 30034396
27. Kovac S, Angelova PR, Holmstro¨ m KM, Zhang Y, Dinkova-Kostova AT, Abramov AY. Nrf2 regulates
ROS production by mitochondria and NADPH oxidase. Biochim Biophys Acta. 2015; 1850: 794–801.
https://doi.org/10.1016/j.bbagen.2014.11.021 PMID: 25484314
28.
Fuchs TA, Abed U, Goosmann C, Hurwitz R, Schulze I, Wahn V, et al. Novel cell death program leads
to neutrophil extracellular traps. J Cell Biol. 2007; 176: 231–241. https://doi.org/10.1083/jcb.200606027
PMID: 17210947
29. Segal AW. The function of the NADPH oxidase of phagocytes and its relationship to other NOXs in
plants, invertebrates, and mammals. Int J Biochem Cell Biol. 2008; 40: 604–618. https://doi.org/10.
1016/j.biocel.2007.10.003 PMID: 18036868
30.
Thimmulappa RK, Scollick C, Traore K, Yates M, Trush MA, Liby KT, et al. Nrf2 dependent protection
from LPS induced inflammatory response and mortality by CDDO-Imidazolide. Biochem Biophys Res
Commun. 2006; 351: 883–889. https://doi.org/10.1016/j.bbrc.2006.10.102 PMID: 17097057
31. Sima C, Aboodi GM, Lakschevitz FS, Sun C, Goldberg MB, Glogauer M. Nuclear Factor Erythroid 2-
Related Factor 2 Down-Regulation in Oral Neutrophils Is Associated with Periodontal Oxidative Dam-
age and Severe Chronic Periodontitis. Am J Pathol. 2016; 186: 1417–1426. https://doi.org/10.1016/j.
ajpath.2016.01.013 PMID: 27070823
32.
Joshi N, Werner S. Nrf2 is highly expressed in neutrophils, but myeloid cell-derived Nrf2 is dispensable
for wound healing in mice. PLOS ONE. 2017; 12: e0187162. https://doi.org/10.1371/journal.pone.
0187162 PMID: 29073253
PLOS ONE | https://doi.org/10.1371/journal.pone.0216465 August 16, 2019
19 / 21
The role of Nrf2 transcription factor in neutrophil biology
33. Makni-Maalej K, Chiandotto M, Hurtado-Nedelec M, Bedouhene S, Gougerot-Pocidalo M-A, Dang PM-
C, et al. Zymosan induces NADPH oxidase activation in human neutrophils by inducing the phosphory-
lation of p47phox and the activation of Rac2: Involvement of protein tyrosine kinases, PI3Kinase, PKC,
ERK1/2 and p38MAPkinase. Biochem Pharmacol. 2013; 85: 92–100. https://doi.org/10.1016/j.bcp.
2012.10.010 PMID: 23085266
34. Cuadrado A, Manda G, Hassan A, Alcaraz MJ, Barbas C, Daiber A, et al. Transcription Factor NRF2 as
a Therapeutic Target for Chronic Diseases: A Systems Medicine Approach. Pharmacol Rev. 2018; 70:
348–383. https://doi.org/10.1124/pr.117.014753 PMID: 29507103
35.
Itoh K, Chiba T, Takahashi S, Ishii T, Igarashi K, Katoh Y, et al. An Nrf2/small Maf heterodimer mediates
the induction of phase II detoxifying enzyme genes through antioxidant response elements. Biochem
Biophys Res Commun. 1997; 236: 313–322. https://doi.org/10.1006/bbrc.1997.6943 PMID: 9240432
36. Granger V, Faille D, Marani V, Noe¨ l B, Gallais Y, Szely N, et al. Human blood monocytes are able to
form extracellular traps. J Leukoc Biol. 2017; 102: 775–781. https://doi.org/10.1189/jlb.3MA0916-411R
PMID: 28465447
37. Pionnier N, Brotin E, Karadjian G, Hemon P, Gaudin-Nome´ F, Vallarino-Lhermitte N, et al. Neutropenic
Mice Provide Insight into the Role of Skin-Infiltrating Neutrophils in the Host Protective Immunity against
Filarial Infective Larvae. PLoS Negl Trop Dis. 2016; 10. https://doi.org/10.1371/journal.pntd.0004605
PMID: 27111140
38.
39.
Li Y, Paonessa JD, Zhang Y. Mechanism of Chemical Activation of Nrf2. PLOS ONE. 2012; 7: e35122.
https://doi.org/10.1371/journal.pone.0035122 PMID: 22558124
Leliefeld PHC, Koenderman L, Pillay J. How Neutrophils Shape Adaptive Immune Responses. Front
Immunol. 2015; 6. https://doi.org/10.3389/fimmu.2015.00471 PMID: 26441976
40. Koenderman L, Buurman W, Daha MR. The innate immune response. Immunol Lett. 2014; 162: 95–
102. https://doi.org/10.1016/j.imlet.2014.10.010 PMID: 25455595
41. Hong C-W. Current Understanding in Neutrophil Differentiation and Heterogeneity. Immune Netw.
2017; 17: 298–306. https://doi.org/10.4110/in.2017.17.5.298 PMID: 29093651
42.
Turbica I, Gallais Y, Gueguen C, Tharinger H, Al Sabbagh C, Gorges R, et al. Ectosomes from neutro-
phil-like cells down-regulate nickel-induced dendritic cell maturation and promote Th2 polarization. J
Leukoc Biol. 2015; 97: 737–749. https://doi.org/10.1189/jlb.3A0314-132RR PMID: 25673294
43. Espinasse M-A, Pe´ pin A, Virault-Rocroy P, Szely N, Chollet-Martin S, Pallardy M, et al. Glucocorticoid-
Induced Leucine Zipper Is Expressed in Human Neutrophils and Promotes Apoptosis through Mcl-1
Down-Regulation. J Innate Immun. 2016; 8: 81–96. https://doi.org/10.1159/000439052 PMID:
26384220
44. El Ali Z, Gerbeix C, Hemon P, Esser PR, Martin SF, Pallardy M, et al. Allergic Skin Inflammation
Induced by Chemical Sensitizers Is Controlled by the Transcription Factor Nrf2. Toxicol Sci. 2013; 134:
39–48. https://doi.org/10.1093/toxsci/kft084 PMID: 23564646
45. Arau´jo VC, Demasi APD, Soares AB, Passador-Santos F, Napimoga MH, Martinez EF, et al. Neutro-
phils in Oral Paracoccidioidomycosis and the Involvement of Nrf2. PLOS ONE. 2013; 8: e76976. https://
doi.org/10.1371/journal.pone.0076976 PMID: 24204715
46. Garber K. Biochemistry: A radical treatment. Nature. 2012; 489: S4–S6. https://doi.org/10.1038/
489S4a PMID: 23013714
47.
Thimmulappa Rk, Fuchs Rj, Malhotra D, Scollick C, Traore K, Bream Jh, et al. Preclinical Evaluation of
Targeting the Nrf2 Pathway by Triterpenoids (CDDO-Im and CDDO-Me) for Protection from LPS-
Induced Inflammatory Response and Reactive Oxygen Species in Human Peripheral Blood Mononu-
clear Cells and Neutrophils. Antioxid Redox Signal. 2007; 9: 1963–1970. https://doi.org/10.1089/ars.
2007.1745 PMID: 17822364
48. Dias IHK, Chapple ILC, Milward M, Grant MM, Hill E, Brown J, et al. Sulforaphane Restores Cellular
Glutathione Levels and Reduces Chronic Periodontitis Neutrophil Hyperactivity In Vitro. PLoS ONE.
2013; 8. https://doi.org/10.1371/journal.pone.0066407 PMID: 23826097
49.
Liu X, Zhang X, Ding Y, Zhou W, Tao L, Lu P, et al. Nuclear Factor E2-Related Factor-2 Negatively Reg-
ulates NLRP3 Inflammasome Activity by Inhibiting Reactive Oxygen Species-Induced NLRP3 Priming.
Antioxid Redox Signal. 2017; 26: 28–43. https://doi.org/10.1089/ars.2015.6615 PMID: 27308893
50. Ahmed SM, Luo L, Namani A, Wang XJ, Tang X. Nrf2 signaling pathway: Pivotal roles in inflammation.
Biochim Biophys Acta. 2017; 1863: 585–597. https://doi.org/10.1016/j.bbadis.2016.11.005 PMID:
27825853
51. Watzlawick R, Kenngott EE, Liu FDM, Schwab JM, Hamann A. Anti-Inflammatory Effects of IL-27 in
Zymosan-Induced Peritonitis: Inhibition of Neutrophil Recruitment Partially Explained by Impaired Mobi-
lization from Bone Marrow and Reduced Chemokine Levels. PLOS ONE. 2015; 10: e0137651. https://
doi.org/10.1371/journal.pone.0137651 PMID: 26360023
PLOS ONE | https://doi.org/10.1371/journal.pone.0216465 August 16, 2019
20 / 21
The role of Nrf2 transcription factor in neutrophil biology
52. Gomez JC, Yamada M, Martin JR, Dang H, Brickey WJ, Bergmeier W, et al. Mechanisms of Interferon-
γ Production by Neutrophils and Its Function during Streptococcus pneumoniae Pneumonia. Am J
Respir Cell Mol Biol. 2015; 52: 349–364. https://doi.org/10.1165/rcmb.2013-0316OC PMID: 25100610
53. Hattori H, Subramanian KK, Sakai J, Jia Y, Li Y, Porter TF, et al. Small-molecule screen identifies reac-
tive oxygen species as key regulators of neutrophil chemotaxis. Proc Natl Acad Sci. 2010; 107: 3546–
3551. https://doi.org/10.1073/pnas.0914351107 PMID: 20142487
54. Wang G, Cao L, Liu X, Sieracki NA, Di A, Wen X, et al. Oxidant Sensing by TRPM2 Inhibits Neutrophil
Migration and Mitigates Inflammation. Dev Cell. 2016; 38: 453–462. https://doi.org/10.1016/j.devcel.
2016.07.014 PMID: 27569419
55.
Iizuka T, Ishii Y, Itoh K, Kiwamoto T, Kimura T, Matsuno Y, et al. Nrf2-deficient mice are highly suscepti-
ble to cigarette smoke-induced emphysema. Genes Cells. 2005; 10: 1113–1125. https://doi.org/10.
1111/j.1365-2443.2005.00905.x PMID: 16324149
56. Gomez JC, Dang H, Martin JR, Doerschuk CM. Nrf2 Modulates Host Defense during Streptococcus
pneumoniae Pneumonia in Mice. J Immunol. 2016; 197: 2864–2879. https://doi.org/10.4049/jimmunol.
1600043 PMID: 27566827
57. Sima C, Aboodi GM, Lakschevitz FS, Sun C, Goldberg MB, Glogauer M. Nuclear Factor Erythroid 2-
Related Factor 2 Down-Regulation in Oral Neutrophils Is Associated with Periodontal Oxidative Dam-
age and Severe Chronic Periodontitis. Am J Pathol. 2016; 186: 1417–1426. https://doi.org/10.1016/j.
ajpath.2016.01.013 PMID: 27070823
58. Suratt BT, Petty JM, Young SK, Malcolm KC, Lieber JG, Nick JA, et al. Role of the CXCR4/SDF-1 che-
mokine axis in circulating neutrophil homeostasis. Blood. 2004; 104: 565–571. https://doi.org/10.1182/
blood-2003-10-3638 PMID: 15054039
59. Sadik CD, Kim ND, Luster AD. Neutrophils cascading their way to inflammation. Trends Immunol. 2011;
32: 452–460. https://doi.org/10.1016/j.it.2011.06.008 PMID: 21839682
60.
Tsai JJ, Dudakov JA, Takahashi K, Shieh J-H, Velardi E, Holland AM, et al. Nrf2 regulates haemato-
poietic stem cell function. Nat Cell Biol. 2013; 15: 309–316. https://doi.org/10.1038/ncb2699 PMID:
23434824
61. Deng Y, Herbert JA, Smith CM, Smyth RL. An in vitro transepithelial migration assay to evaluate the
role of neutrophils in Respiratory Syncytial Virus (RSV) induced epithelial damage. Sci Rep. 2018; 8:
6777. https://doi.org/10.1038/s41598-018-25167-4 PMID: 29712964
62. Holmstro¨ m KM, Baird L, Zhang Y, Hargreaves I, Chalasani A, Land JM, et al. Nrf2 impacts cellular bio-
energetics by controlling substrate availability for mitochondrial respiration. Biol Open. 2013; 2: 761–
770. https://doi.org/10.1242/bio.20134853 PMID: 23951401
63. Dinkova-Kostova AT, Abramov AY. The emerging role of Nrf2 in mitochondrial function. Free Radic Biol
Med. 2015; 88: 179–188. https://doi.org/10.1016/j.freeradbiomed.2015.04.036 PMID: 25975984
64. Harada Y, Kato Y, Miyaji T, Omote H, Moriyama Y, Hiasa M. Vesicular nucleotide transporter mediates
ATP release and migration in neutrophils. J Biol Chem. 2018; 293: 3770–3779. https://doi.org/10.1074/
jbc.M117.810168 PMID: 29363573
65. Vega MR de la, Dodson M, Chapman E, Zhang DD. NRF2-targeted therapeutics: New targets and
modes of NRF2 regulation. Curr Opin Toxicol. 2016; 1: 62. https://doi.org/10.1016/j.cotox.2016.10.005
PMID: 29082352
66. Cho H-Y, Kleeberger SR. Noblesse Oblige: NRF2 Functions in the Airways. Am J Respir Cell Mol Biol.
2014; 50: 844–847. https://doi.org/10.1165/rcmb.2014-0116PS PMID: 24783956
PLOS ONE | https://doi.org/10.1371/journal.pone.0216465 August 16, 2019
21 / 21
|
10.1371_journal.pone.0232245 | RESEARCH ARTICLE
Cost-effectiveness of psychological
treatments for post-traumatic stress disorder
in adults
Ifigeneia MavranezouliID
Jonathan Leach7, Caitlin Daly8, Sofia DiasID
Sharif El-Leithy11, Neil GreenbergID
1,2*, Odette Megnin-Viggars1,2, Nick Grey3,4, Gita Bhutani5,6,
8¤, Nicky J. Welton8, Cornelius Katona9,10,
12, Sarah Stockton2, Stephen Pilling1,2,13
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: Mavranezouli I, Megnin-Viggars O, Grey
N, Bhutani G, Leach J, Daly C, et al. (2020) Cost-
effectiveness of psychological treatments for post-
traumatic stress disorder in adults. PLoS ONE 15
(4): e0232245. https://doi.org/10.1371/journal.
pone.0232245
Editor: Scott McDonald, Hunter Holmes McGuire
VA Medical Center, UNITED STATES
Received: August 12, 2019
Accepted: April 10, 2020
Published: April 30, 2020
Peer Review History: PLOS recognizes the
benefits of transparency in the peer review
process; therefore, we enable the publication of
all of the content of peer review and author
responses alongside final, published articles. The
editorial history of this article is available here:
https://doi.org/10.1371/journal.pone.0232245
Copyright: © 2020 Mavranezouli et al. This is an
open access article distributed under the terms of
the Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: Full details on the
methods and the clinical studies included in the
network meta-analysis that informed the economic
analysis are provided in Mavranezouli et al.,
1 Centre for Outcomes Research and Effectiveness, Research Department of Clinical, Educational & Health
Psychology, University College London, London, United Kingdom, 2 National Guideline Alliance, Royal
College of Obstetricians and Gynaecologists, London, United Kingdom, 3 Sussex Partnership NHS
Foundation Trust, Hove, United Kingdom, 4 School of Psychology, University of Sussex, Brighton, United
Kingdom, 5 Lancashire & South Cumbria NHS Foundation Trust, Preston, United Kingdom, 6 University of
Liverpool, Liverpool, United Kingdom, 7 Davenal House Surgery, Bromsgrove, United Kingdom,
8 Population Health Sciences, Bristol Medical School, University of Bristol, Bristol, United Kingdom, 9 Helen
Bamber Foundation, London, United Kingdom, 10 Division of Psychiatry, University College London, London,
United Kingdom, 11 Traumatic Stress Service, Springfield Hospital, London, United Kingdom, 12 King’s
Centre for Military Health Research, King’s College London, London, United Kingdom, 13 Camden and
Islington NHS Foundation Trust, St Pancras Hospital, London, United Kingdom
¤ Current address: Centre for Reviews and Dissemination, University of York, York, United Kingdom
* i.mavranezouli@ucl.ac.uk
Abstract
Background
Post-traumatic stress disorder (PTSD) is a severe and disabling condition that may lead to
functional impairment and reduced productivity. Psychological interventions have been
shown to be effective in its management. The objective of this study was to assess the cost-
effectiveness of a range of interventions for adults with PTSD.
Methods
A decision-analytic model was constructed to compare costs and quality-adjusted life-years
(QALYs) of 10 interventions and no treatment for adults with PTSD, from the perspective of
the National Health Service and personal social services in England. Effectiveness data
were derived from a systematic review and network meta-analysis. Other model input
parameters were based on published sources, supplemented by expert opinion.
Results
Eye movement desensitisation and reprocessing (EMDR) appeared to be the most cost-
effective intervention for adults with PTSD (with a probability of 0.34 amongst the 11 evalu-
ated options at a cost-effectiveness threshold of £20,000/QALY), followed by combined
somatic/cognitive therapies, self-help with support, psychoeducation, selective serotonin
reuptake inhibitors (SSRIs), trauma-focused cognitive behavioural therapy (TF-CBT), self-
PLOS ONE | https://doi.org/10.1371/journal.pone.0232245 April 30, 2020
1 / 22
PLOS ONE Psychol Med. 2020 Mar;50(4):542-555. doi: 10.
1017/S0033291720000070. All other relevant data
are within the paper and its Supporting Information
files.
Funding: The economic analysis presented in this
paper was initiated by the National Collaborating
Centre for Mental Health (NCCMH) and continued
by the National Guideline Alliance (NGA) at the
Royal College of Obstetricians and Gynaecologists
(RCOG) from 1 April 2016, with support from the
National Institute of Health and Care Excellence
(NICE) Guidelines Technical Support Unit (TSU),
University of Bristol, which is funded by the Centre
for Guidelines (NICE). NCCMH and NGA received
funding from NICE to develop clinical and social
care guidelines. IM, OMG, SS and SP received
support from the National Collaborating Centre for
Mental Health and the National Guideline Alliance,
which were in receipt of funding from the National
Institute for Health and Care Excellence (NICE), for
the submitted work. CD, SD and NJW received
support from the NICE Guidelines Technical
Support Unit, University of Bristol, with funding
from the Centre for Clinical Practice (NICE). The
views expressed in this publication are those of the
authors and not necessarily those of the RCOG,
NGA, NCCMH or NICE. The funder had no role in
study design, data collection and analysis, decision
to publish, or preparation of the manuscript. All
authors had full access to all the data in the study
and had final responsibility for the decision to
submit for publication. National Institute for Health
and Care Excellence (2018) Post-traumatic stress
disorder. Available from: https://www.nice.org.uk/
guidance/ng116
Competing interests: IM, OMV, SS and SP
received support from the National Collaborating
Centre for Mental Health and the National Guideline
Alliance, which were in receipt of funding from the
National Institute for Health and Care Excellence
(NICE), for the submitted work. CD, SD and NJW
received support from the NICE Guidelines
Technical Support Unit, University of Bristol, with
funding from the Centre for Clinical Practice
(NICE). SD and NJW were co-applicants on a grant
(unrelated to this work) from the MRC
Methodology Research Programme which
included an MRC Industry Collaboration
Agreement with Pfizer Ltd, who part-funded a
researcher to work on statistical methodology in a
project underlated to this work. This does not alter
our adherence to PLOS ONE policies on sharing
data and materials. GB is a coinvestigator on a
NIHR RfPB grant, Eye Movement Desensitization
and Reprocessing Therapy in Early Psychosis
(EYES): A feasibility randomised controlled trial.
Cost-effectiveness of psychological treatments for PTSD in adults
help without support, non-TF-CBT and combined TF-CBT/SSRIs. Counselling appeared to
be less cost-effective than no treatment. TF-CBT had the largest evidence base.
Conclusions
A number of interventions appear to be cost-effective for the management of PTSD in
adults. EMDR appears to be the most cost-effective amongst them. TF-CBT has the largest
evidence base. There remains a need for well-conducted studies that examine the long-
term clinical and cost-effectiveness of a range of treatments for adults with PTSD.
Introduction
A considerable proportion of people exposed to trauma, around 5.6%, will develop post-trau-
matic stress disorder (PTSD) [1]. PTSD is a severe and disabling condition that may lead to
functional impairment and reduced productivity [2]. A number of psychological interventions
have been shown to be effective in the treatment of PTSD in adults, predominantly eye move-
ment desensitisation and reprocessing (EMDR) and trauma-focused cognitive behavioural
therapy (TF-CBT) [3]. However, many people with PTSD delay seeking help or are not identi-
fied by health services [4]. Given the variety of available interventions and the need for efficient
use of healthcare resources, the objective of this study was to examine the cost-effectiveness of
a range of psychological interventions for the treatment of PTSD in adults from the perspective
of the National Health Service (NHS) and Personal Social Services (PSS) in England, using
decision-analytic economic modelling.
The analysis presented here is part of the work that informed the updating of national guid-
ance for the management of PTSD in England, published by the National Institute for Health
and Care Excellence (NICE) [5]. The guideline was developed by a guideline committee, an
independent multi-disciplinary group of clinical academics, health professionals and service
user and carer representatives with expertise and experience in the field of PTSD. The commit-
tee contributed to the development of the economic model by providing advice on issues relat-
ing to the natural history and treatment patterns of PTSD in the UK, and on model inputs in
areas where evidence was lacking.
Methods
Population
The study population comprised adults presenting in primary care with clinically important
post-traumatic stress symptoms, defined by a diagnosis of PTSD according to the Diagnostic
and Statistical Manual of Mental Disorders (DSM), the World Health Organization (WHO)
International Classification of Diseases (ICD) or similar criteria, or by clinically significant
PTSD symptoms, indicated by a PTSD symptom score above threshold on a validated scale,
that are present for more than 3 months after a traumatic event.
The starting age of the cohorts in the economic model was 39 years, to reflect the mean age
of adults with PTSD presenting to healthcare services in the UK [6]. The percentage of women
in each cohort at the start of the model was 51.6%, calculated using national statistics for the
general population [7], and data on the percentage of people screened positive for PTSD by
age and sex in England [4]. The starting age and gender mix of the cohorts was used to esti-
mate mortality risks and gender-specific quality-adjusted life-years (QALYs).
PLOS ONE | https://doi.org/10.1371/journal.pone.0232245 April 30, 2020
2 / 22
PLOS ONE NGreenberg is the Royal College of Psychiatrists
Lead for Military and Veterans’ Health and is a
trustee of two military charities. He is also a senior
researcher with King’s College London working on
a number of military mental health r studies. NGrey
is a member of the Wellcome Trust Anxiety
Disorders Group developing, testing and
disseminating Cognitive Therapy for PTSD (CT-
PTSD), a trauma-focused cognitive behavioural
therapy (TF-CBT). He has published papers and
book chapters on CT-PTSD, and facilitates teaching
workshops for which payment is received. As
editor, he receives royalties from sales of a trauma
book, A Casebook of Cognitive Therapy for
Traumatic Stress Reactions. CK is Medical Director
of the Helen Bamber Foundation (a human rights
charity) and refugee and asylum mental health lead
for the Royal College of Psychiatrists. He writes
expert psychiatric reports in the context of asylum
mental health. JL is NHS England Medical Director
for Military and Veterans Health. SP receives
funding from NICE for the development of clinical
guidelines and is also supported by the NIHR
UCLH Biomedical Research Centre. The authors
report no other relationships or activities that could
appear to have influenced the submitted work.
Cost-effectiveness of psychological treatments for PTSD in adults
Interventions
The interventions considered in the economic analysis were selected from those considered in
a network meta-analysis (NMA) of randomised controlled trials (RCTs) of psychological treat-
ments for adults with PTSD ([3]; see S2 Appendix for inclusion criteria for the NMA). We
included only interventions that had been tested on at least 100 individuals in the NMA of
changes in PTSD symptoms at treatment endpoint, as this was deemed the minimum adequate
evidence base that would enable robust conclusions to be drawn on clinical and cost-effective-
ness. Moreover, we included only interventions that had shown a higher mean effect in com-
parison with waitlist. Selective serotonin reuptake inhibitors (SSRIs) were included in the
analysis as they were relevant comparators to psychological interventions.
The economic analysis evaluated the following interventions:
• EMDR
• TF-CBT
• Non-TF-CBT
• Combined somatic/cognitive therapies
• SSRIs
• Combined TF-CBT/SSRIs
• Self-help with support
• Self-help without support
• Counselling
• Psychoeducation
• No treatment, reflected in waitlist RCT arms.
TF-CBT is a broad class of psychological interventions that predominantly use trauma-
focused cognitive, behavioural or cognitive-behavioural techniques and exposure approaches
to treatment. Although some interventions place their main emphasis on exposure and others
on cognitive techniques, most use a combination of both. TF-CBT includes therapies such as
cognitive therapy, cognitive processing therapy, exposure therapy/prolonged exposure, virtual
reality exposure therapy, mindfulness-based cognitive therapy and narrative exposure therapy.
Although the specific interventions that make up a class do not include exactly the same con-
tent or follow the same manual, they use the same broad approach and there is considerable
overlap in the proposed mechanisms. In the economic analysis that informed the NICE guide-
line [5] we divided the TF-CBT class by the number of sessions and format of delivery and cre-
ated separate categories of TF-CBT treatment according to their intensity, as these differences
in resource use comprised practical considerations that informed the guideline recommenda-
tions; in addition to different intervention costs, each TF-CBT category had its own clinical
effectiveness, estimated in the guideline NMAs. However, in the analyses we present here, we
considered TF-CBT as an umbrella term of interventions that share a similar approach to
treatment, in order to investigate the overall performance of the TF-CBT class relative to other
treatments, regardless of its mode of delivery. Resource use for TF-CBT in the economic analy-
sis we present here was determined by the average resource use reported in the TF-CBT trials
informing the analysis, considering also that their vast majority assessed individual forms of
TF-CBT.
PLOS ONE | https://doi.org/10.1371/journal.pone.0232245 April 30, 2020
3 / 22
PLOS ONE Cost-effectiveness of psychological treatments for PTSD in adults
Non-TF-CBT is a class of interventions that focus on current symptoms of PTSD without
re-visiting the trauma experience. Combined somatic/cognitive therapies, such as emotional
freedom techniques and thought field therapy, are exposure-based therapies with both cogni-
tive and somatic components that utilise the tapping of points on the body [8, 9]. EMDR is
based on a theoretical model which posits that the dysfunctional intrusions, emotions and
physical sensations experienced by trauma victims are due to the improper storage of the trau-
matic event in implicit memory. The EMDR procedures are based on stimulating the patient’s
own information processing in order to help integrate the targeted event as an adaptive contex-
tualised memory [10]. Counselling is a type of psychological treatment that builds on the con-
cept of client-centred therapy by Rogers [11]. It has been described in the relevant literature as
non-directive counselling, supportive counselling, supportive psychotherapy, or person-cen-
tred counselling. Individuals are helped to focus on their thoughts, feelings and behaviour; to
reach clearer self-understanding; and to find and use their strengths so that they cope more
effectively with their lives by making appropriate decisions, or by taking relevant action. Coun-
selling is primarily non-directive and non-advisory, but recognises that some situations require
positive guidance by means of information and advice. Psychoeducation involves the provision
of information about the nature and causes of PTSD, and strategies and treatments that can
help address PTSD symptoms. It can be delivered individually, but is commonly delivered to
groups. Psychoeducation is usually non-directive and takes an educational didactic format.
Finally, self-help therapies include interventions such as internet-based TF-CBT or other com-
puterised psychological therapies, expressive writing and cognitive bibliotherapy. Self-help
with support includes interventions in which therapist’s input is an integral part of the inter-
vention; in self-help without support the therapist’s input is minimal or absent.
The guideline analysis included interventions tested on at least 50 people in the NMA of
changes in PTSD symptoms at treatment endpoint, whereas here we used a threshold of at
least 100 people to improve the robustness of the results. The impact of increasing this thresh-
old was the omission of interpersonal psychotherapy and present-centered therapy from the
analysis presented here; both interventions were shown to occupy middle-to-lower cost-effec-
tiveness rankings in the guideline analysis, and therefore their omission had no impact on the
overall conclusions of our analysis.
Economic model structure
A hybrid decision-analytic model consisting of a decision-tree followed by a three-state Mar-
kov model was constructed using Microsoft Office Excel 2013 to estimate total costs and
QALYs associated with each treatment. The model structure was determined by the natural
history of PTSD, its treatment patterns in the UK, and the availability of relevant clinical and
epidemiological data (Fig 1).
The model followed hypothetical cohorts of adults with PTSD, initiated on each of the treat-
ments assessed. The treatment duration for each of the assessed options equalled 3 months (12
weeks), according to the average duration of interventions in trials and routine clinical practice
(range 4–20 weeks). Following a course of treatment, people in each cohort either remitted
(entering a state of ‘no-PTSD’) or failed to remit, remaining in a ‘PTSD’ state. Those initiated
on SSRIs alone or in combination were given 3 months of maintenance pharmacological ther-
apy if they had remitted. Death was not considered during provision of interventions, as no
relevant differential mortality data are available. In the next 3 months of follow-up, those who
had remitted could remain in remission, relapse to PTSD or die. Those who had not remitted
could remain in the ‘PTSD’ state, remit (and move to ‘no-PTSD’) or die. The length of the
PLOS ONE | https://doi.org/10.1371/journal.pone.0232245 April 30, 2020
4 / 22
PLOS ONE Cost-effectiveness of psychological treatments for PTSD in adults
Fig 1. Schematic diagram of the economic model.
https://doi.org/10.1371/journal.pone.0232245.g001
follow-up period immediately post-treatment was set at 3 months as this is the period for
which most follow-up data are reported in RCTs of psychological interventions for PTSD.
After that point, people in each cohort entered the Markov model, run in 3-month cycles,
for consistency with the duration of the two periods of the decision-tree. In each cycle, they
remained in the same health state or moved between the states of ‘PTSD’ and ‘no-PTSD’ or
moved to death (absorbing state). People in both the ‘PTSD’ and the ‘no-PTSD’ states received
primary, community and secondary healthcare and personal social services, as relevant to their
health state. A half-cycle correction was applied.
The time horizon of the analysis was 3 years (36 months), comprising 6 months in the deci-
sion tree and 2.5 years (10 x 3-month cycles) in the Markov component of the model. This
time frame was deemed adequate to capture longer-term costs and effects of treatment, with-
out making significant extrapolations and assumptions over the course of PTSD.
Effectiveness data
We obtained effectiveness data from a systematic review and NMA of psychological treatments
for adults with PTSD [3]. We utilised the results of 2 NMAs of changes in PTSD symptoms:
between baseline and treatment endpoint; and between baseline and 1-4-month follow-up.
Details on the NMAs, including the studies and data that informed them, the selection of the
effectiveness data and the transformations required for use in the economic model are pro-
vided in S1 File.
The outputs of the NMA of changes in PTSD symptoms between baseline and treatment
endpoint informed the intervention effects in the model period of 0–3 months. For the 3–6
month follow-up period, the base-case economic analysis conservatively assumed that the
active intervention effects were not retained and equalled the effect of no treatment; this was
decided because the results of the NMA of changes in PTSD symptoms between baseline and
1-4-month follow-up were based on limited evidence and showed considerable uncertainty.
Nevertheless, data from this NMA were used in a secondary analysis, to inform effects for each
active intervention during 3–6 months after treatment initiation.
PLOS ONE | https://doi.org/10.1371/journal.pone.0232245 April 30, 2020
5 / 22
PLOS ONE Cost-effectiveness of psychological treatments for PTSD in adults
Baseline probability of remission
The probability of remission for no treatment (baseline) and for all model arms beyond treat-
ment endpoint (i.e. for all treatment options during 3–6 months after treatment initiation in
the base-case analysis and for all treatment options during 6–36 months after treatment initia-
tion in both the base-case and secondary analysis) was estimated using data from an Australian
national mental health survey [12]. We considered the survey participants to be representative
of our study population, which was adults presenting in primary care with symptoms of
PTSD. Details on the methods used for the estimation of the baseline probability of remission
are provided in S2 File.
Risk of relapse
Due to lack of published evidence, an annual risk of relapse of 0.10 was assumed across all
treatment arms, based on the committee’s expert opinion; this was translated into a 3-month
probability of relapse of 0.026 assuming an exponential function, which was applied in the
3-month follow-up period of the decision-tree and over the whole duration of the Markov
model. This assumption was tested in a sensitivity analysis.
Risk of development of side effects from SSRI treatment
The probability of developing common side effects from SSRIs (headaches, nausea or vomit-
ing, agitation, sedation and sexual dysfunction) was estimated from a retrospective analysis of
a large US managed care claims database on 40,017 people with depression who were initiated
on antidepressant monotherapy, including SSRIs [13]. Serious side effects from SSRIs (such as
death, attempted suicide or self-harm, falls, fractures, stroke, epilepsy/seizures) were not con-
sidered; however, their omission is not expected to have had a significant impact on the eco-
nomic results, due to their low incidence in the study population [14, 15].
Mortality
PTSD is associated with an increased risk of mortality. The hazard ratio of death associated
with PTSD, adjusted for confounders such as age, gender, diabetes mellitus and hypertension
was obtained from a study on 637 veterans in the US [16]. This ratio was applied onto general
mortality statistics for England [17] to estimate the annual mortality risk in people with PTSD
over the time period they experienced PTSD symptoms. People without PTSD symptoms had
the mortality risk of the general population.
Utility data
Utility scores express preferences for the health-related quality of life (HRQoL) in distinct
health states; they are necessary for the estimation of QALYs. Following a systematic literature
search of utility data for PTSD, the base-case economic analysis used utility scores generated
from HRQoL ratings of Australian adult participants in a national mental health survey, some
of whom had a diagnosis of PTSD according to DSM-4 criteria [18]. HRQoL was assessed with
the Assessment of Quality of Life measure (http://www.aqol.com.au). The study provided gen-
der-specific data for people with PTSD and people who were PTSD-free following evidence-
based treatment, which corresponded directly to the model health states of interest.
In a sensitivity analysis, we used utility data derived from a sample of 808 US veterans
attending primary care clinics, 97 of whom had developed PTSD, adjusted for confounders
such as gender, employment status, presence of disability, and mental and physical health
comorbidities [19]. HRQoL was assessed using SF-36, which was converted to utility scores
PLOS ONE | https://doi.org/10.1371/journal.pone.0232245 April 30, 2020
6 / 22
PLOS ONE Cost-effectiveness of psychological treatments for PTSD in adults
using the UK algorithm [20]. Data from this study indicated a narrower utility benefit fol-
lowing remission from PTSD compared with the utility data used in the base-case analysis.
The mean utility decrement in people experiencing common side effects from SSRIs was
estimated using EQ-5D scores of participants in a US medical expenditure survey, some of
whom had depression and experienced side effects from antidepressant treatment [21]. This
utility decrement was applied only over the period that people received SSRI treatment.
Resource use and cost data
The analysis included intervention costs (healthcare professional time, drug acquisition and
equipment/infrastructure required for self-help interventions), costs of managing side effects
from medication, and costs relating to the ‘PTSD’ and ‘no-PTSD’ health states including costs
of primary, community and secondary healthcare and PSS costs.
Intervention costs (Table 1) were calculated by combining resource use reported in RCTs
included in the NMA that informed the economic analysis (i.e. number and duration of thera-
peutic sessions and mean daily dosage of sertraline, which was the most commonly used SSRI
in trials), modified to represent routine UK practice, with respective national unit costs. All
psychological interventions, with the exception of self-help and psychoeducation, were
assumed to be delivered in a primary care setting by Band 7 psychological therapists according
to the NHS Agenda for Change (AfC) pay-scales for community-based scientific and profes-
sional staff, to reflect routine practice in the UK. Psychoeducation was assumed to be delivered
by AfC band 5 Psychological Well-being Practitioners (PWPs) and self-help by AfC band 6
psychological therapists. People receiving SSRIs attended general practitioner (GP) monitor-
ing visits and undertook routine laboratory testing. Those experiencing side effects had one
extra GP contact every 3 months and received medication for their management. For self-help
therapies we included the cost of the provider of digital mental health programmes, and costs
of hardware and capital overheads.
Unit costs were estimated using a combination of data derived from national sources and
other published evidence [22–26]. Health professional unit costs included wages/salary, salary
on-costs, capital and other overheads, qualification costs where available, and supervision
costs. The ratio of direct (face-to-face) to indirect (preparation and administrative tasks) health
professionals’ time was taken into account. Details on the methods and sources used to esti-
mate therapists’ unit costs are reported in S3 File.
Annual costs associated with the PTSD and no-PTSD health states were estimated using
predominantly NHS and PSS usage data from a national psychiatric morbidity survey con-
ducted in England in 2014 [4], supplemented with resource use data from other published
sources [27–29] and expert opinion, which were subsequently combined with national unit
costs [22, 23, 30]. Costs for each state included inpatient hospital stays and outpatient visits,
contacts with GPs, psychiatrists, psychologists, social workers, community psychiatric and
learning disability nurses, other nursing services, self-help and support groups, home help or
home care, outreach or family support workers and community day-care centres. Details on
the methods and data used to estimate annual costs associated with the PTSD and no-PTSD
health states are provided in S4 File. These were then translated into 3-month cost figures that
informed the economic model. Because the estimated health state-related costs were based to a
large extent on expert opinion, a sensitivity analysis was conducted, in which PTSD costs were
varied by ±50%.
Costs were expressed in 2017 prices, uplifted, where necessary, using the Hospital and
Community Health Services Pay and Prices Index [22].
PLOS ONE | https://doi.org/10.1371/journal.pone.0232245 April 30, 2020
7 / 22
PLOS ONE Table 1. Intervention costs of treatments for adults with PTSD (2017 prices).
Cost-effectiveness of psychological treatments for PTSD in adults
Resource use details
Intervention cost
Intervention
EMDR
TF-CBT
non-TF-CBT
Combined somatic/cognitive
therapies
6 x 1.5 hr individual sessions (9 hours) delivered by a Band 7 psychological therapist
9 x 1.5 hr individual sessions (13.5 hours) delivered by a Band 7 psychological therapist
9 x 1 hr individual sessions (9 hours) delivered by a Band 7 psychological therapist
4 x 1 hr individual sessions (4 hours) delivered by a Band 7 psychological therapist
SSRIs (sertraline)
Mean daily dosage 150mg, 4 GP visits at 0–3 months + 1 visit at 3–6 months, monitoring lab testing
For people experiencing side effects: 1 extra GP visit over 3 months, medication for management.
Combined TF-CBT/SSRIs
Sum of the individual treatment components
Self-help with support
Self-help without support
Fixed cost of provider of digital programmes, hardware & capital overheads, 180 minutes of
support by a band 6 psychological therapist
Fixed cost of provider of digital programmes, hardware & capital overheads, 40 minutes of support
by a band 6 psychological therapist
Counselling
Psychoeducation
No treatment
10 x 1 hr individual sessions (10 hours) delivered by a Band 7 psychological therapist
3 x 1 hr individual sessions (3 hours) delivered by a band 5 PWP
No resource use
£912
£1,368
£912
£405
0–3 months: £155
3–6 months: £39
If side effects: £37
0–3 months: £1,523 3–6
months: £39
£266
£98
£1,014
£127
£0
EMDR: eye movement desensitisation reprocessing; IPT: interpersonal psychotherapy; PWP: psychological well-being practitioner; TF-CBT: trauma-focused cognitive
behavioural therapy
Unit cost of band 7 psychological therapists: £101 per hour of direct contact with the client–see S3 File for details on estimation of unit cost
Unit cost of band 5 PWPs: £42 per hour of direct contact with the client–see S3 File for details on estimation of unit cost
Unit cost of band 6 psychological therapist: £72 per hour of direct contact with the client (mean value of unit costs of band 7 psychological therapist and band 5 PWP)
Unit cost of GP: £37 per patient contact lasting 9.22 minutes [22]
Sertraline acquisition cost: 100mg, 28 tab, £0.99 [23]– 3-month cost £1.59
Cost of monitoring lab testing (SSRIs): £5 per person (expert advice)
Cost of medication for management of side effects (SSRIs): £3 per person over 3 months (expert advice)
Fixed cost of provider of digital programmes: £36.20 per person (expert advice)
Cost of hardware & capital overheads: £14 per person [24]
https://doi.org/10.1371/journal.pone.0232245.t001
Discounting
Costs and QALYs were discounted at 3.5% annually as recommended by NICE [31].
Analysis
To account for the uncertainty around input parameter point estimates, a probabilistic analysis
was undertaken, in which input parameters were assigned probabilistic distributions [32]. Sub-
sequently, 10,000 iterations were performed, each drawing random values out of the distribu-
tions fitted onto the model input parameters. Mean costs and QALYs for each treatment were
calculated by averaging across the 10,000 iterations. The Net Monetary Benefit (NMB) for
each intervention was estimated for each iteration and averaged across the 10,000 iterations,
determined by the formula
NMB ¼ E � l (cid:0) C
where E and C are the effects (QALYs) and costs of each intervention, respectively, and λ rep-
resents the willingness-to-pay per unit of effectiveness, set at the NICE lower cost-effectiveness
threshold of £20,000/QALY [31]. The intervention with the highest NMB is the most cost-
effective option [33].
PLOS ONE | https://doi.org/10.1371/journal.pone.0232245 April 30, 2020
8 / 22
PLOS ONE Cost-effectiveness of psychological treatments for PTSD in adults
The mean ranking by cost-effectiveness is reported for each intervention (out of 10,000 iter-
ations), where a rank of 1 suggests that an intervention is the most cost-effective amongst all
evaluated treatment options. The probability of the intervention with the highest NMB being
the most cost-effective option is also provided, calculated as the proportion of iterations in
which the intervention has had the highest NMB amongst all interventions considered in the
analysis. The probability of cost-effectiveness has been estimated in a step-wise approach,
according to which the most cost-effective intervention is omitted at each step, and the proba-
bility of cost-effectiveness of the next most cost-effective intervention amongst the remaining
treatment options is re-calculated. The probabilities estimated following this approach reflect
the uncertainty around the cost-effectiveness not only of the most cost-effective intervention,
but also of the second, third, fourth, etc. most cost-effective intervention in ranking, after
more cost-effective interventions have been omitted from analysis. Finally, the cost-effective-
ness acceptability frontier has been plotted, showing the treatment with the highest mean
NMB over different cost-effectiveness thresholds (λ), and the probability that this treatment is
the most cost-effective among those assessed [33].
Table 2 reports the values of all model input parameters. Deterministic values were used in
deterministic one-way sensitivity analyses. The probability distributions show the types and
range of distributions assigned to each parameter; estimation of distribution ranges was based
on data reported in the published sources of evidence.
Two probabilistic analyses were undertaken, each using different assumptions on the effec-
tiveness of interventions at the 3-month follow-up:
• Base-case analysis: treatment effect between 3–6 months equalled that of no treatment
• Secondary analysis: treatment effect between 3–6 months equalled that estimated from the
NMA of changes in PTSD symptoms between baseline and 1-4-month follow-up
One-way deterministic sensitivity analyses explored the following scenarios applied onto
the base-case analysis:
• change in the annual risk of relapse between 0.05 and 0.20
• change in the PTSD health state cost by ± 50%
• use of alternative utility scores for the PTSD and no-PTSD states [19].
Validation of the economic model
The economic model was developed in collaboration with members of the guideline commit-
tee. All inputs and model formulae were systematically checked. The model was tested for logi-
cal consistency by setting input parameters to null and extreme values and examining whether
results changed in the expected direction. Results were discussed with the committee to con-
firm their plausibility.
Results
Table 3 shows the results of the base-case economic analysis. Interventions have been ordered
from the most to the least cost-effective. The table provides the mean number of QALYs, inter-
vention costs and total costs per person, the mean NMB and ranking of each intervention, and
its probability of being cost-effective in a step-wise approach at a threshold of £20,000/QALY.
EMDR was found to be the most cost-effective intervention for adults with PTSD, with the
highest NMB at the cost-effectiveness threshold of £20,000/QALY. This was followed by
PLOS ONE | https://doi.org/10.1371/journal.pone.0232245 April 30, 2020
9 / 22
PLOS ONE Cost-effectiveness of psychological treatments for PTSD in adults
Table 2. Economic model input parameters.
Input parameter
Deterministic
value
Characteristics of study population
Starting age of cohort
(years)
Proportion of women
39
0.52
Probability distribution (type, range)
Sources–comments
No distribution
No distribution
[6]; mean age of adults referred for assessment for possible PTSD in a UK NHS
outpatient clinic
Calculated using the proportion of women in the general population aged 39
years [7], and data on the percentage of people screened positive for PTSD by
age and sex [4].
Odds ratios of remission versus no treatment at treatment endpoint
EMDR
TF-CBT
non-TF-CBT
Combined somatic/
cognitive therapies
SSRIs
Combined TF-CBT/SSRIs
Self-help with support
Self-help without support
Counselling
Psychoeducation
42.18
14.06
9.09
21.33
7.99
9.06
13.98
5.17
3.70
8.99
95% CrI: 13.59 to 132.42
95% CrI: 6.76 to 29.81
95% CrI: 2.50 to 33.62
95% CrI: 3.84 to 121.63
95% CrI: 1.50 to 44.61
95% CrI: 1.15 to 69.34
95% CrI: 2.74 to 70.74
95% CrI: 1.29 to 20.27
95% CrI: 1.12 to 12.38
95% CrI: 0.26 to 276.72
[3]; standardised mean differences converted to odds ratios according to [34];
distribution based on 300,000 samples from posterior distributions outputted
from NMAs, thinned by 30 to obtain 10,000 values
Odds ratios of remission versus no treatment at 3-month follow-up (secondary analysis only)
EMDR
TF-CBT
non-TF-CBT
Combined somatic/
cognitive therapies
SSRIs
Combined TF-CBT/SSRIs
Self-help with support
Self-help without support
Counselling
Psychoeducation
7.53
3.80
2.18
8.08
No data
No data
10.11
8.85
1.73
2.58
95% CrI: 1.55 to 35.77
95% CrI: 1.49 to 9.72
95% CrI: 0.37 to 12.40
95% CrI: 0.41 to 155.56
No data
No data
95% CrI: 2.03 to 48.96
95% CrI: 0.73 to 105.43
95% CrI: 0.37 to 8.15
95% CrI: 0.42 to 15.43
[3]; standardised mean differences converted to odds ratios according to [34];
distribution based on 300,000 samples from posterior distributions outputted
from NMAs, thinned by 30 to obtain 10,000 values
3-6-month probability of remission for SSRIs assumed to equal the probability
of remission of SSRIs during initial treatment (0–3 months); 3-6-month
probability of remission for combined TF-CBT/SSRIs borrowed from TF-CBT.
Probability of remission–no treatment (also applied to all interventions between 3–6 months in the base-case analysis & all interventions beyond 6 months in
the base-case and secondary analyses)
0–3 months from PTSD
onset
0–12 months from PTSD
onset
0–24 months from PTSD
onset
0–36 months from PTSD
onset
Risk of relapse–all treatments
0.03
0.15
0.27
0.32
Beta: α = 17.26; β = 646.74
[12]. See S2 File for details
Beta: α = 98.94; β = 565.06
Beta: α = 176.62; β = 487.38
Beta: α = 212.48; β = 451.52
3-month risk
0.026
Beta: α = 2.60; β = 97.40
Expert opinion
Risk of developing common side effects from SSRIs
3-month risk
Mortality
Hazard ratio–PTSD vs no
PTSD
Baseline mortality–
general population
0.029
Beta: α = 687; β = 22,933
[13]
1.77
Log-normal: 95% CI 1.02 to 3.14
[16]
Age/sex specific
No distribution
General mortality statistics for the UK population [17]
(Continued )
PLOS ONE | https://doi.org/10.1371/journal.pone.0232245 April 30, 2020
10 / 22
PLOS ONE Cost-effectiveness of psychological treatments for PTSD in adults
Table 2. (Continued)
Input parameter
Utility values
Base-case analysis
PTSD, men
PTSD, women
No PTSD, men
No PTSD, women
Sensitivity analysis
PTSD, all
No PTSD, all
Deterministic
value
Probability distribution (type, range)
Sources–comments
0.54
0.57
0.63
0.64
0.61
0.64
[18]; distribution estimated based on method of moments
Beta: α = 26.83; β = 22.86
Beta: α = 86.75; β = 65.44
Beta: α = 5.11; β = 3.00
Beta: α = 14.11; β = 7.93
No distribution
[19]
Disutility due to side effects from SSRIs
% reduction in health
state utility
Intervention costs–resource use
Number of sessions
EMDR
TF-CBT
non-TF-CBT
Combined somatic/
cognitive therapies
Counselling
Psychoeducation
Therapist time (minutes)
Self-help with support
Self-help without support
Number of GP contacts–SSRIs
0–3 months
3–6 months
Treatment of side effects
Intervention costs—unit
costs
SSRI– 3-month drug
acquisition
Laboratory testing–SSRIs
Medication for side
effects–SSRIs
Self-help infrastructure
GP unit cost
Band 7 clinical
psychologist unit cost
Band 5 PWP unit cost
Band 6 therapist unit cost
10.3
Beta: α = 89.64; β = 784.07
[21]; disutility applied as a percentage onto the health state (PTSD or no
PTSD) utility
6
9
9
4
10
3
180
40
4
1
1
£1.59
£5
£3
£50
£37
£101
£42
£72
0.70: 5–6, 0.16: 4, 0.14: 3
0.70: 7–9, 0.16: 5–6, 0.14: 3–4
0.70: 7–9, 0.16: 5–6, 0.14: 3–4
0.70: 4, 0.30: 2–3
0.70: 8–10, 0.16: 5–7, 0.14: 3–4
0.70: 3, 0.16: 2, 0.14: 1
Different probabilities assigned to different numbers of sessions for each
therapy, based on resource use reported in the RCTs included in the NMAs
that informed the economic analysis, supplemented by further assumptions.
Combined TF-CBT/SSRIs: resource use was the sum of the resource use of the
individual treatment components.
Details on intervention costs are provided in Table 1.
Normal: SD = 0.30�mean
Normal: SD = 0.30�mean
SD based on assumption
0.70: 4, 0.30: 2–3
0.70: 1, 0.30: 0
0.80: 1, 0.20: 2
Different probabilities assigned to different numbers of sessions; number of
visits based on expert opinion; distribution based on assumption.
No distribution
[23]
No distribution
No distribution
No distribution
Assumption
Assumption
Fixed digital therapy provider cost based on expert advice; capital cost based
on [24]
Normal, SE = 0.05 of the mean
[22]; distribution based on assumption
Normal: SE = 0.05 of the mean
See S3 File; distribution based on assumption
Normal, SE = 0.05 of the mean
See S3 File; distribution based on assumption
Determined by distribution of Band 7
and Band 5 therapist unit costs
Assumed to be the mean of Band 7 and Band 5 therapist unit costs
3-month NHS/PSS health state cost
PTSD
No-PTSD
£293
£27
Gamma: SE = 0.30 of the mean
Gamma: SE = 0.30 of the mean
Based on resource use data reported in national and other published sources
[4, 27–29], supplemented with expert opinion and combined with national
unit costs [22, 23, 30], expressed in 2017 prices; see S4 File for details.
(Continued )
PLOS ONE | https://doi.org/10.1371/journal.pone.0232245 April 30, 2020
11 / 22
PLOS ONE Cost-effectiveness of psychological treatments for PTSD in adults
Table 2. (Continued)
Input parameter
Deterministic
value
Probability distribution (type, range)
Sources–comments
Annual discount rate
0.035
No distribution
Applied to costs and QALYs [31]
CI: confidence intervals; CrI: credible intervals; EMDR: eye movement desensitisation reprocessing; GP: general practitioner; IPT: interpersonal psychotherapy; NHS:
National Health Service; PSS: personal social services; PTSD: post-traumatic stress disorder; PWP: psychological well-being practitioner; SD: standard deviation; SE:
standard error; SSRI: selective serotonin reuptake inhibitor; TF-CBT: trauma-focused cognitive behavioural therapy
https://doi.org/10.1371/journal.pone.0232245.t002
combined somatic/cognitive therapies, self-help with support, psychoeducation, SSRIs,
TF-CBT, self-help without support, non-TF-CBT, combined TF-CBT/SSRIs, no treatment and
counselling. The probability of EMDR being the most cost-effective treatment amongst the 11
options assessed was only 0.34. The probabilities of cost-effectiveness for next interventions in
ranking up to (and including) self-help without support did not exceed 0.42, although increas-
ingly fewer interventions were included in the analysis, indicating uncertainty in the results.
Notably, counselling was less cost-effective than no treatment; this finding was attributed to
the relatively low clinical effectiveness of counselling (the lowest amongst all active treatments
assessed in the economic analysis), which did not offset its relatively high intervention cost.
The cost-effectiveness plane (Fig 2) depicts the mean incremental costs and QALYs of all inter-
ventions versus no treatment (placed at the origin). According to the cost-effectiveness accept-
ability frontier (Fig 3), combined somatic/cognitive therapies appeared to be most cost-
effective amongst the 11 treatment options assessed for thresholds up to £2,500/QALY, with a
low probability that reached 0.25 at maximum, whereas EMDR became the most cost-effective
option at higher thresholds, with a probability that ranged from 0.19 to 0.41 amongst the 11
options assessed.
Results of the secondary analysis, which utilised 3-month NMA follow-up data, were not
very different. The top 3 interventions (EMDR, combined somatic/cognitive therapies, self-
Table 3. Base-case results of economic modelling.
Intervention
EMDR
Combined somatic/cognitive therapies
Self-help with support
Psychoeducation
SSRIs
TF-CBT
Self-help without support
non-TF-CBT
Combined TF-CBT/SSRIs
No treatment
Counselling
Mean per person
NMB (£/ person)
Mean rank
Prob�
QALY
Intervention cost (£)
Total cost (£)
(at a threshold of £20,000/
QALY)
1.80
1.77
1.75
1.74
1.72
1.74
1.71
1.73
1.73
1.67
1.69
746
360
266
108
146
1,058
98
705
1,204
0
785
2,047
1,963
2,047
1,982
2,143
2,854
2,253
2,670
3,140
2,488
3,043
33,928
33,364
32,880
32,754
32,316
32,042
31,865
31,860
31,451
30,915
30,854
2.31
3.28
4.01
4.90
5.15
6.38
6.19
6.79
8.19
9.14
9.66
0.34
0.35
0.32
0.42
0.37
0.26
0.41
0.50
0.48
0.64
1.00
EMDR: eye movement desensitisation reprocessing; NMB: net monetary benefit; Prob: probability of cost-effectiveness; SSRIs: selective serotonin reuptake inhibitors;
TF-CBT: trauma-focused cognitive behavioural therapy
�estimated in a step-wise approach, according to which the most cost-effective intervention is omitted at each step, and the probability of cost-effectiveness of the next
most cost-effective intervention amongst the remaining treatment options is re-calculated
https://doi.org/10.1371/journal.pone.0232245.t003
PLOS ONE | https://doi.org/10.1371/journal.pone.0232245 April 30, 2020
12 / 22
PLOS ONE Cost-effectiveness of psychological treatments for PTSD in adults
Fig 2. Cost-effectiveness plane: Base-case analysis results for 1,000 adults with PTSD.
https://doi.org/10.1371/journal.pone.0232245.g002
help with support) remained the same. The ranking of self-help without support improved
and counselling became better than no treatment. The probabilities of cost-effectiveness of the
top 5 interventions were low, ranging between 0.14 and 0.47, indicating uncertainty around
the results. Self-help without support appeared to be the most cost-effective option at a zero
cost-effectiveness threshold and combined somatic/cognitive therapies were most cost-effec-
tive at higher thresholds up to £18,000/QALY; EMDR was the most cost-effective option at
higher thresholds, with a 0.14 probability at the threshold of £20,000/QALY. Results of the
secondary analysis are provided in S5 File.
In deterministic sensitivity analyses, results were, overall, robust to changes in the risk of
relapse and in the PTSD health state cost and rankings were not affected. TF-CBT was the only
option that dropped (by one place) in ranking when the baseline risk of relapse was increased
by 50% or the PTSD health state cost was reduced by 50%. Use of alternative utility data that
assumed narrower HRQoL benefits associated with remission had a small impact on the
results, with the relative cost-effectiveness of TF-CBT alone or combined with SSRIs and non-
Fig 3. Cost-effectiveness acceptability frontier: Base-case analysis.
https://doi.org/10.1371/journal.pone.0232245.g003
PLOS ONE | https://doi.org/10.1371/journal.pone.0232245 April 30, 2020
13 / 22
PLOS ONE Cost-effectiveness of psychological treatments for PTSD in adults
TF-CBT being reduced. However, results for other interventions were not affected. Results of
deterministic sensitivity analyses are shown in S6 File. For information, results of the NICE
guideline economic analysis are shown in S7 File.
Discussion
Overview of findings
EMDR appears to be the most cost-effective intervention for adults with PTSD more than 3
months after trauma, followed by combined somatic/cognitive therapies, self-help with sup-
port, psychoeducation, SSRIs, TF-CBT, self-help without support, non-TF-CBT and combined
TF-CBT/SSRIs. Counselling appears to be less cost-effective than no treatment, due to its rela-
tively low clinical effectiveness (the lowest amongst all active treatments assessed in the eco-
nomic analysis), as shown in the NMAs that informed the economic analysis [3], which was
not enough to offset its relatively high intervention cost. The low effectiveness found for coun-
selling can be attributed to counselling’s non-directive person-centred approach, which is less
likely to help the person overcome avoidance (which is one of the criteria for PTSD), and thus
less likely to reduce PTSD symptoms. On the other hand, counselling’s effectiveness may have
been underestimated to some extent due to researcher allegiance, since in the vast majority of
the RCTs on counselling that informed the economic analysis, counselling served as a control
treatment to other active interventions, primarily TF-CBT [3].
Results were characterised by uncertainty, as reflected in relatively low probabilities of each
intervention being cost-effective amongst alternative treatment options; however they were
overall robust to assumptions tested through deterministic sensitivity analyses.
Strengths and limitations
Our analysis utilised effectiveness data derived from a systematic review and NMA of changes
in PTSD symptoms [3]. This methodology enabled us to consider information from direct and
indirect comparisons between interventions, and allowed simultaneous comparisons across all
options while preserving randomisation [35]. This approach for evidence synthesis is essential
for populating model-based economic studies assessing more than two competing interven-
tions. No inconsistency was detected between direct and indirect evidence. We used 10,000
iterations of the NMA models in the economic analysis, which represent the full uncertainty
and correlation in the relative effects.
The NMA that informed the base-case economic analysis (changes in PTSD symptoms
between baseline and treatment endpoint) used a large evidence base and produced robust
data. The NMA of 1-4-month follow-up changes in PTSD symptoms, which informed the
secondary analysis, showed considerable uncertainty for most interventions, due to the small
number and size of the included studies; TF-CBT and EMDR were the only treatments in this
NMA with data on at least 100 people at 1-4-month follow-up that showed evidence of sus-
tained effect. Thus, results of this secondary economic analysis should be interpreted with
caution. Both NMAs were characterised by high between-trial heterogeneity, which is likely to
have been caused by heterogeneity across populations included in the trials considered in the
NMAs, for example, in terms of the presence of a formal PTSD diagnosis, the severity, com-
plexity and chronicity of PTSD symptoms, the type, extent and multiplicity of trauma expo-
sure, the presence of comorbidity, and also the variability of interventions within each assessed
option and the differences across settings, e.g. inpatient versus outpatient delivery of interven-
tions [3]. Heterogeneity may also have been caused by the type, multiplicity and timing of pre-
vious treatments in trial participants, but relevant information was not available in the
PLOS ONE | https://doi.org/10.1371/journal.pone.0232245 April 30, 2020
14 / 22
PLOS ONE Cost-effectiveness of psychological treatments for PTSD in adults
majority of the RCTs included in the NMAs; therefore, the impact of previous treatments on
the effectiveness of the interventions cannot be assessed.
Regarding the potential heterogeneity across populations in the trials that informed the eco-
nomic model, it should be noted that, of the 82 studies that were included in the NMAs that
informed the economic analysis, 50 (61%) recruited people with a formal PTSD diagnosis
whereas the remaining 32 (39%) recruited people with clinically important PTSD symptoms, as
indicated by baseline scores above a predefined threshold on a validated PTSD symptom scale.
The percentage of trials recruiting people with a formal PTSD diagnosis was relatively high
(range 70–100%) among trials assessing TF-CBT alone or combined with SSRIs, EMDR and
counselling; moderate (range 50–66%) among trials assessing non-TF-CBT and self-help with
support; and rather low (range 25–33%) among trials assessing combined somatic/cognitive
therapies, self-help without support and psychoeducation ([3]; see S5 Appendix for population
characteristics of included studies in the NMA). It may be hypothesised that a formal diagnosis
of PTSD is associated with more severe symptoms at baseline, resulting in interventions tested
on such populations appearing to be less clinically and cost-effective compared with interven-
tions tested on people with clinically important PTSD symptoms, who may have had less severe
symptoms at baseline. Nevertheless, the NMA and economic modelling results did not confirm
such a hypothesis, since, for example, EMDR and, to a lesser degree, TF-CTB, showed high
clinical and cost-effectiveness, despite both having been tested in RCTs that, in their majority,
recruited people with a formal PTSD diagnosis. On the other hand, it is possible that partici-
pants in the trials that recruited people with clinically important PTSD symptoms might meet
criteria for (and might have received) a formal PTSD diagnosis, but they were not required to
do so in order to participate in the trial, and this could well have been for pragmatic reasons
related to trial management rather than an intention to recruit people with potentially lower
symptom severity. Furthermore, for PTSD symptom scales that are based closely on diagnostic
criteria, e.g. PTSD Checklist (PCL), scoring above the clinical threshold may be regarded as
comparable to receiving a formal diagnosis of PTSD. Related to this point, we should note that
all RCTs that evaluated combined somatic/cognitive therapies and recruited people with clini-
cally important PTSD symptoms used scales that are based on diagnostic criteria (e.g. PCL and
Modified PTSD Symptom Scale [MPSS-SR]); similarly, the majority of the studies that evalu-
ated psychoeducation and recruited people with clinically important PTSD symptoms used
scales based on diagnostic criteria (e.g. PCL and Davidson Trauma Scale). Consequently, the
symptom severity of trial populations receiving combined somatic/cognitive therapies and psy-
choeducation (who, in the majority of trials, were not required to have a formal diagnosis of
PTSD) was likely similar to the symptom severity of populations with a formal diagnosis of
PTSD receiving other interventions considered in our NMA and economic analysis.
The strengths and limitations of the NMAs that informed the economic analyses should be
considered when interpreting the cost-effectiveness results. Moreover, the quality and limita-
tions of RCTs considered in the NMAs have unavoidably impacted on the quality of the model
input parameters. Although all interventions included in the economic analysis had been
tested on at least 100 trial participants for treatment endpoint, the size of the evidence base
differed considerably across interventions. TF-CBT had by far the largest evidence base in
both NMAs (N = 903 across 29 RCTs at treatment endpoint and N = 753 across 13 RCTs at 1-
4-month follow-up), which gives us more confidence in the results on its clinical (and, conse-
quently, cost-) effectiveness. The evidence base was more limited for other interventions
included in the economic analysis (each tested on N<350 in each of the two NMAs). For
comparison, EMDR, which was shown to be the most cost-effective treatment option in our
economic analysis, was tested on N = 260 across 11 RCTs at treatment endpoint and N = 121
across 4 RCTs at 1-4-month follow-up in the NMAs that informed the economic analysis [3].
PLOS ONE | https://doi.org/10.1371/journal.pone.0232245 April 30, 2020
15 / 22
PLOS ONE Cost-effectiveness of psychological treatments for PTSD in adults
The economic model structure did not incorporate discontinuation due to limited data avail-
ability. However, the NMAs that informed the economic analysis utilised intention-to-treat data,
where available, so that discontinuation has been implicitly considered in the economic analysis.
The probabilistic analysis took into account, where possible, the completion rates of the interven-
tions in the RCTs that informed the economic analysis, so that the number of sessions reflected,
up to a degree, the attrition rates of each intervention. The time horizon of the analysis was 3
years, in order to capture longer-term effects and costs associated with a course of treatment for
PTSD without significant extrapolation over the natural course of PTSD.
The baseline risk of remission was estimated from a large longitudinal study on adults with
PTSD in the community [12], as the survey’s target population was deemed to be directly rele-
vant to our study population. The risk of relapse was not available in published literature, and
was therefore based on expert opinion. Utility data were derived from a systematic literature
review. Costs incurred by adults with PTSD and those remitting from PTSD were based on
published national survey data, supplemented with other published evidence and expert opin-
ion, due to lack of more accurate information. Sensitivity analysis showed that results were
robust to use of alternative values for the risk of relapse, utility and costs. The risk of side
effects from SSRIs was based on an uncontrolled study that did not examine the rate of side
effects that were attributable to SSRIs. Therefore, our economic analysis may have overesti-
mated the impact of common side effects from SSRIs relative to other treatments and thus
may have underestimated the relative cost effectiveness of SSRIs.
In conclusion, our study is characterised by different strengths and limitations, which we
have considered when constructing our model and interpreting the results of our analysis. We
carried out probabilistic analyses, which took into account the uncertainty around model
parameters and, where possible, we conducted secondary and deterministic sensitivity analyses
to address uncertainties and gaps in the evidence.
Comparison with existing economic evidence
Published economic evaluations of interventions for PTSD in adults have concluded that expo-
sure therapy (a form of TF-CBT) is more cost-effective than no treatment [36]. TF-CBT and
SSRIs are likely more cost-effective than usual care [18]; prolonged exposure (TF-CBT) has
been found to be more cost-effective than SSRIs [37]. Finally, self-management was shown to
be no more effective but overall less costly than psychoeducation [38]. These economic studies
evaluated a limited range of interventions for adults with PTSD and made very few compari-
sons between active interventions; notably, EMDR, which was shown to be the most cost-effec-
tive intervention in our analysis, has not been evaluated in previously published economic
literature on adults with PTSD.
Overall, our findings are in agreement with previously published evidence. Our economic
analysis estimated the cost-effectiveness of a wider range of interventions available for adults
with PTSD, such as EMDR, combined somatic/cognitive therapies, self-help, non-TF-CBT
and counselling and allowed, for the first time, simultaneous comparisons of cost-effectiveness
across interventions, and their ranking from the most to the least cost-effective.
On the other hand, an economic evaluation of psychological interventions for PTSD in chil-
dren and young people, which also used efficacy data derived from a NMA and adopted a sim-
ilar approach and methodology to the analysis described here, concluded that individual forms
of TF-CBT were most cost-effective in the treatment of children and young people with PTSD,
whereas EMDR occupied middle cost-effectiveness rankings amongst the treatment options
assessed [39]. This finding was attributed to the lower effectiveness of EMDR relative to other
treatments in children and young people compared with adult populations [40].
PLOS ONE | https://doi.org/10.1371/journal.pone.0232245 April 30, 2020
16 / 22
PLOS ONE Cost-effectiveness of psychological treatments for PTSD in adults
Generalisability of the results and implications of the study
Our analysis was conducted from the perspective of the NHS/PSS in England. Results may be
generalisable to other settings with similar funding and structure of healthcare and personal social
services and comparable care pathways for adults with PTSD. Conclusions on cost-effectiveness
ultimately rely on the cost-effectiveness threshold adopted, and this depends on the policy makers’
willingness-to-pay for treatment benefits, which may vary across countries and health systems.
Our analysis estimated the resource use relating to the delivery of each intervention based
on information reported in the RCTs that informed the economic analysis; for example, the
mode number of hours for a course of EMDR and TF-CBT was 9 and 13.5, respectively. If the
duration and therefore the cost of an intervention is considerably different from our estimates,
then its relative cost-effectiveness is expected to be affected. However, reducing the number of
sessions of an intervention will improve its cost-effectiveness only if its clinical effectiveness
remains unaffected. In practice, a reduction in the number of sessions below a point that is
critical for the optimal delivery of the intervention is expected to reduce its clinical effective-
ness, too; the impact on its cost-effectiveness will depend on the trade-off between a lower
intervention cost and a lower clinical effectiveness.
Based on the results of the NMAs and the economic analysis, the NICE guideline on PTSD
recommended EMDR and individual TF-CBT for the treatment of adults with PTSD presenting
more than 3 months after trauma [5]. Both interventions were shown to be effective in reducing
PTSD symptoms post-treatment and were the only ones with sufficient evidence to suggest sus-
tainment of effect beyond treatment. EMDR appeared to be the most cost-effective intervention
amongst those assessed. TF-CBT appeared to be less cost-effective than other interventions (i.e.
combined somatic/cognitive therapies, psychoeducation, self-help with support and SSRIs), but
had by far the largest evidence base and the guideline economic analysis showed that brief indi-
vidual TF-CBT (delivered in fewer than 8 sessions) had the highest clinical and cost effectiveness
amongst all options assessed; the finding that brief individual TF-CBT had the highest clinical
effectiveness was explained by inspection of the clinical data, which revealed that participants in
trials of brief individual TF-CBT had less severe PTSD symptoms at baseline, and therefore were
likely to have a better response to treatment, compared with participants in trials of more inten-
sive forms of individual TF-CBT. The recommendation for EMDR was restricted to people with
non-combat-related trauma, as evidence suggested a non-significant effect on people with com-
bat-related trauma.
The NICE guideline recommendations on TF-CBT and EMDR for adults with PTSD are
consistent with other published PTSD clinical practice guidelines (compared in [41]). Three
more guidelines make recommendations of equal strength for TF-CBT and EMDR [42–44],
whereas one guideline makes a strong recommendation for TF-CBT while EMDR has been
given a moderate rating [45].
Self-help with support was shown to be the third most cost-effective option amongst those
assessed, owing to a combination of its high effectiveness at treatment endpoint (informed by
N = 198 across 5 RCTs in the respective NMA [3]) and its low intervention cost. There was
also limited evidence (N = 85 across 3 RCTs in the respective NMA [3]) that it can sustain
effects beyond treatment endpoint. All 5 RCTs on self-help with support that informed the
economic analysis focused on computerised TF-CBT, which is consistent with TF-CBT deliv-
ered by a therapist, and this element may have been the driver of the intervention’s clinical
effectiveness. The NICE guideline committee considered the clinical and cost-effectiveness of
self-help with support and, also, that of SSRIs, but noted their narrower evidence base and
made weaker (‘consider’) recommendations for people who expressed a preference for these
interventions, and, in the case of self-help, did not have severe PTSD symptoms and were not
PLOS ONE | https://doi.org/10.1371/journal.pone.0232245 April 30, 2020
17 / 22
PLOS ONE Cost-effectiveness of psychological treatments for PTSD in adults
at risk of harm to themselves or others. Based on its middle rankings in the NMA and eco-
nomic analysis, a ‘consider’ recommendation was also made for non-TF-CBT targeted at
specific symptoms, for people who are unable or unwilling to engage in a trauma-focused
intervention or have residual symptoms after treatment. Psychoeducation was shown to be
cost-effective based on limited and inconclusive clinical evidence; therefore, it was not recom-
mended as a stand-alone intervention, but as part of individual TF-CBT. Finally, the commit-
tee noted the evidence of high clinical and cost-effectiveness for combined somatic/cognitive
therapies, but also considered their particularly limited evidence base beyond treatment end-
point and the lack of specific indications for these interventions, and decided not to recom-
mend them but instead to make a recommendation for further research [5].
Conclusion
EMDR appears to be the most cost-effective intervention for adults with PTSD more than 3
months after trauma, followed by combined somatic/cognitive therapies, self-help with sup-
port, psychoeducation, SSRIs, TF-CBT, self-help without support, non-TF-CBT and combined
TF-CBT/SSRIs. Counselling appears to be less cost-effective than no treatment. Results were
characterised by uncertainty, and relatively limited evidence base for interventions other than
TF-CBT. There is a need for well-conducted studies that examine the relative clinical and cost-
effectiveness of a range of psychological treatments for adults with PTSD, in particular assess-
ment of longer-term costs and effects, to reduce the uncertainty and limitations characterising
current evidence.
Supporting information
S1 File. Selection of effectiveness data and transformation for use in the economic analysis.
(DOCX)
S2 File. Estimation of the baseline probability of remission.
(DOCX)
S3 File. Estimation of the unit cost of therapists delivering psychological interventions for
PTSD in the British National Health Service (NHS).
(DOCX)
S4 File. Estimation of annual health and personal social service costs incurred by adults
with PTSD and adults without PTSD.
(DOCX)
S5 File. Results of secondary probabilistic economic analysis [beneficial effect up to
3-months post-treatment].
(DOCX)
S6 File. Results of deterministic sensitivity analyses.
(DOCX)
S7 File. Results of the NICE guideline economic analysis.
(DOCX)
S1 Appendix. Search strategy.
(DOCX)
PLOS ONE | https://doi.org/10.1371/journal.pone.0232245 April 30, 2020
18 / 22
PLOS ONE Cost-effectiveness of psychological treatments for PTSD in adults
S2 Appendix. Study protocol.
(DOCX)
S3 Appendix. Details of the statistical analysis and WinBUGS codes for data synthesis.
(DOCX)
S4 Appendix. Details of the inconsistency checks and WinBUGS codes for inconsistency
models.
(DOCX)
S5 Appendix. Characteristics of studies included in the network meta-analysis, and full ref-
erences.
(DOCX)
S6 Appendix. List of excluded studies with reasons for exclusion.
(DOCX)
S7 Appendix. NMA data files.
(DOCX)
S8 Appendix. Risk of bias of studies included in the NMA.
(DOCX)
S9 Appendix. Model fit statistics.
(DOCX)
S10 Appendix. Inconsistency checks.
(DOCX)
S11 Appendix. Relative effects between all pairs of interventions: Direct, indirect and com-
bined (NMA) results.
(DOCX)
S12 Appendix. Results of the NICE guideline NMA.
(DOCX)
S13 Appendix. Pairwise sub-analyses.
(DOCX)
S14 Appendix. References in the online supplementary material.
(DOCX)
Acknowledgments
We thank other members of the Guideline Committee for the NICE guideline on ‘Post-trau-
matic stress disorder’ for their contributions to this work. Members of the Guideline Commit-
tee were: Steve Hajioff, Philip Bell, Gita Bhutani, Sharif El-Leithy, Neil Greenberg, Nick Grey,
Cornelius Katona, Jonathan Leach, Richard Meiser-Stedman, Rebecca Regler, Vikki Touzel,
and David Trickey.
Author Contributions
Conceptualization: Ifigeneia Mavranezouli, Odette Megnin-Viggars, Nick Grey, Gita Bhutani,
Jonathan Leach, Caitlin Daly, Sofia Dias, Nicky J. Welton, Cornelius Katona, Sharif El-
Leithy, Neil Greenberg, Sarah Stockton, Stephen Pilling.
PLOS ONE | https://doi.org/10.1371/journal.pone.0232245 April 30, 2020
19 / 22
PLOS ONE Cost-effectiveness of psychological treatments for PTSD in adults
Data curation: Ifigeneia Mavranezouli, Odette Megnin-Viggars, Nick Grey, Gita Bhutani, Jon-
athan Leach, Caitlin Daly, Sofia Dias, Sarah Stockton.
Formal analysis: Ifigeneia Mavranezouli.
Funding acquisition: Stephen Pilling.
Methodology: Ifigeneia Mavranezouli, Odette Megnin-Viggars, Caitlin Daly, Sofia Dias,
Nicky J. Welton.
Project administration: Ifigeneia Mavranezouli.
Validation: Nick Grey, Gita Bhutani, Jonathan Leach, Cornelius Katona, Sharif El-Leithy, Neil
Greenberg, Stephen Pilling.
Visualization: Ifigeneia Mavranezouli.
Writing – original draft: Ifigeneia Mavranezouli.
Writing – review & editing: Odette Megnin-Viggars, Nick Grey, Gita Bhutani, Jonathan
Leach, Caitlin Daly, Sofia Dias, Nicky J. Welton, Cornelius Katona, Sharif El-Leithy, Neil
Greenberg, Sarah Stockton, Stephen Pilling.
References
1. Koenen KC, Ratanatharathorn A, Ng L, McLaughlin KA, Bromet EJ, Stein DJ, et al. Posttraumatic stress
disorder in the World Mental Health Surveys. Psychol Med. 2017; 47(13):2260–74. https://doi.org/10.
1017/S0033291717000708 PMID: 28385165
2. Alonso J, Angermeyer MC, Bernert S, Bruffaerts R, Brugha TS, Bryson H, et al. Disability and quality of
life impact of mental disorders in Europe: results from the European Study of the Epidemiology of Men-
tal Disorders (ESEMeD) project. Acta Psychiatr Scand Suppl. 2004; (420):38–46. https://doi.org/10.
1111/j.1600-0047.2004.00329.x PMID: 15128386
3. Mavranezouli I, Megnin-Viggars O, Daly C, Dias S, Welton N, Stockton S, et al. Psychological treat-
ments for post-traumatic stress disorder in adults: a network meta-analysis. Psychol Med. 2020; 50
(4):542–55. https://doi.org/10.1017/S0033291720000070 PMID: 32063234
4. McManus S, Bebbington P, Jenkins R, Brugha T (eds). Mental health and wellbeing in England: Adult
Psychiatric Morbidity Survey 2014. Leeds: NHS Digital; 2016.
5. National Institute for Health and Care Excellence. Post-traumatic stress disorder. NICE; 2018. https://
www.nice.org.uk/guidance/ng116
6. Ehlers A, Grey N, Wild J, Stott R, Liness S, Deale A, et al. Implementation of cognitive therapy for
PTSD in routine clinical care: effectiveness and moderators of outcome in a consecutive sample. Behav
Res Ther. 2013; 51(11):742–52. https://doi.org/10.1016/j.brat.2013.08.006 PMID: 24076408
7. Office for National Statistics. Population Estimates for UK, England and Wales, Scotland and Northern
Ireland: Mid-2016. ONS; 2017. http://www.ons.gov.uk/peoplepopulationandcommunity/
birthsdeathsandmarriages/lifeexpectancies/bulletins/nationallifetablesunitedkingdom/2015-09-23
8. Church D, Hawk C, Brooks AJ, Toukolehto O, Wren M, Dinter I, et al. Psychological trauma symptom
improvement in veterans using emotional freedom techniques: a randomized controlled trial. J Nerv
Ment Dis. 2013; 201(2):153–60. https://doi.org/10.1097/NMD.0b013e31827f6351 PMID: 23364126
9. Robson RH, Robson PM, Ludwig R, Mitabu C, Phillips C. Effectiveness of thought field therapy provided
by newly instructed community workers to a traumatized population in Uganda: A randomized trial. Curr
Res Psychol. 2016; 7(1):1–11.
10. Shapiro F. Eye Movement Desensitisation and Reprocessing: Basic Principles, Protocols, and Proce-
dures. 2nd ed. New York, NY: Guilford Press; 2001.
11. Rogers CR. Client-Centered Therapy: Its Current Practice, Implications and Theory. Boston: Houghton
Mifflin; 1951.
12. Chapman C, Mills K, Slade T, McFarlane AC, Bryant RA, Creamer M, et al. Remission from post-trau-
matic stress disorder in the general population. Psychol Med. 2012; 42(8):1695–703. https://doi.org/10.
1017/S0033291711002856 PMID: 22166813
PLOS ONE | https://doi.org/10.1371/journal.pone.0232245 April 30, 2020
20 / 22
PLOS ONE Cost-effectiveness of psychological treatments for PTSD in adults
13. Anderson HD, Pace WD, Libby AM, West DR, Valuck RJ. Rates of 5 common antidepressant side
effects among new adult and adolescent cases of depression: a retrospective US claims study. Clin
Ther. 2012; 34(1):113–23. https://doi.org/10.1016/j.clinthera.2011.11.024 PMID: 22177545
14. Coupland C, Dhiman P, Morriss R, Arthur A, Barton G, Hippisley-Cox J. Antidepressant use and risk of
adverse outcomes in older people: population based cohort study. BMJ. 2011; 343:d4551. https://doi.
org/10.1136/bmj.d4551 PMID: 21810886
15.
Jakobsen JC, Katakam KK, Schou A, Hellmuth SG, Stallknecht SE, Leth-Moller K, et al. Selective sero-
tonin reuptake inhibitors versus placebo in patients with major depressive disorder. A systematic review
with meta-analysis and Trial Sequential Analysis. BMC Psychiatry. 2017; 17(1):58. https://doi.org/10.
1186/s12888-016-1173-2 PMID: 28178949
16. Ahmadi N, Hajsadeghi F, Mirshkarlo HB, Budoff M, Yehuda R, Ebrahimi R. Post-traumatic stress disor-
der, coronary atherosclerosis, and mortality. Am J Cardiol. 2011; 108(1):29–33. https://doi.org/10.
1016/j.amjcard.2011.02.340 PMID: 21530936
17. Office for National Statistics. National Life Tables, England, 1980–82 to 2014–2016. ONS; 2017. http://
www.ons.gov.uk/peoplepopulationandcommunity/birthsdeathsandmarriages/lifeexpectancies/
bulletins/nationallifetablesunitedkingdom/2015-09-23
18. Mihalopoulos C, Magnus A, Lal A, Dell L, Forbes D, Phelps A. Is implementation of the 2013 Australian
treatment guidelines for posttraumatic stress disorder cost-effective compared to current practice? A
cost-utility analysis using QALYs and DALYs. Aust N Z J Psychiatry. 2015; 49(4):360–76. https://doi.
org/10.1177/0004867414553948 PMID: 25348698
19.
Freed MC, Yeager DE, Liu X, Gore KL, Engel CC, Magruder KM. Preference-weighted health status of
PTSD among veterans: an outcome for cost-effectiveness analysis using clinical data. Psychiatr Serv.
2009; 60(9):1230–8. https://doi.org/10.1176/ps.2009.60.9.1230 PMID: 19723738
20. Brazier J, Roberts J, Deverill M. The estimation of a preference-based measure of health from the SF-36.
J Health Econ. 2002; 21(2):271–92. https://doi.org/10.1016/s0167-6296(01)00130-8 PMID: 11939242
21. Sullivan PW, Valuck R, Saseen J, MacFall HM. A comparison of the direct costs and cost effectiveness
of serotonin reuptake inhibitors and associated adverse drug reactions. CNS Drugs. 2004; 18(13):911–
32. https://doi.org/10.2165/00023210-200418130-00006 PMID: 15521793
22. Curtis L, Burns A. Unit Costs of Health & Social Care 2017. Canterbury: PSSRU; 2017.
23. NHS Business Services Authority, NHS Prescription Services. NHS England and Wales. Electronic
Drug Tariff. Issue: March 2018. Compiled on the behalf of the Department of Health and Social Care.
2018. https://www.nhsbsa.nhs.uk/pharmacies-gp-practices-and-appliance-contractors/drug-tariff
24. Kaltenthaler E, Brazier J, De NE, Tumur I, Ferriter M, Beverley C, et al. Computerised cognitive behav-
iour therapy for depression and anxiety update: a systematic review and economic evaluation. Health
Technol Assess. 2006; 10(33) https://doi.org/10.3310/hta10330 PMID: 16959169
25. British Association for Behavioural & Cognitive Psychotherapies. Criteria and guidelines for re-accreditation
as a behavioural and/or cognitive psychotherapist. British Association for Behavioural & Cognitive Psycho-
therapies; 2016. http://www.babcp.com/files/Accreditation/CBP/Full/CBP-Full-Guidelines-V5-0614.pdf
26. National College for Teaching and Leadership. Review of clinical and educational psychology training
arrangements. NHS Health Education England; 2016.
27. Kontopantelis E, Olier I, Planner C, Reeves D, Ashcroft DM, Gask L, et al. Primary care consultation
rates among people with and without severe mental illness: a UK cohort study using the Clinical Practice
Research Datalink. BMJ Open. 2015; 5(12):e008650. https://doi.org/10.1136/bmjopen-2015-008650
PMID: 26674496
28. NHS Digital. Hospital Episode Statistics for England. Admitted Patient Care statistics, 2016–17. NHS
Digital; 2017.
29. NHS Digital, Community and Mental Health Team. Psychological Therapies: Annual report on the use
of IAPT services. England, 2015–16. NHS Digital; 2016.
30. NHS Improvement. National Schedule of Reference Costs, 2016–17—NHS trusts and NHS foundation
trusts. NHS Improvement; 2017. https://improvement.nhs.uk/resources/reference-costs/
31. National Institute for Health and Care Excellence. Developing NICE guidelines: the manual (last
updated October 2018). NICE; 2014. https://www.nice.org.uk/process/pmg20
32. Briggs A, Sculpher M, Claxton K. Decision Modelling for Health Economic Evaluation. New York, NY:
Oxford University Press; 2006.
33.
Fenwick E, Claxton K, Sculpher M. Representing uncertainty: the role of cost-effectiveness acceptability
curves. Health Econ. 2001; 10(8):779–87. https://doi.org/10.1002/hec.635 PMID: 11747057
34. Chinn S. A simple method for converting an odds ratio to effect size for use in meta-analysis. Stat Med.
2000; 19(22):3127–31. https://doi.org/10.1002/1097-0258(20001130)19:22<3127::aid-sim784>3.0.
co;2-m PMID: 11113947
PLOS ONE | https://doi.org/10.1371/journal.pone.0232245 April 30, 2020
21 / 22
PLOS ONE Cost-effectiveness of psychological treatments for PTSD in adults
35. Caldwell DM, Ades AE, Higgins JP. Simultaneous comparison of multiple treatments: combining direct
and indirect evidence. BMJ. 2005; 331(7521):897–900. https://doi.org/10.1136/bmj.331.7521.897
PMID: 16223826
36.
37.
Tuerk PW, Wangelin B, Rauch SAM, Dismuke CE, Yoder M, Myrick H, et al. Health service utilization
before and after evidence-based treatment for PTSD. Psychol Serv. 2013; 10(4):401–9. https://doi.org/
10.1037/a0030549 PMID: 23148769
Le QA, Doctor JN, Zoellner LA, Feeny NC. Cost-effectiveness of prolonged exposure therapy versus
pharmacotherapy and treatment choice in posttraumatic stress disorder (the optimizing ptsd treatment
trial): A doubly randomized preference trial. J Clin Psychiatry. 2014; 75(3):222–30. https://doi.org/10.
4088/JCP.13m08719 PMID: 24717377
38. Dunn NJ, Rehm LP, Schillaci J, Souchek J, Mehta P, Ashton CM, et al. A randomized trial of self-man-
agement and psychoeducational group therapies for comorbid chronic posttraumatic stress disorder
and depressive disorder. J Trauma Stress. 2007; 20(3):221–37. https://doi.org/10.1002/jts.20214
PMID: 17598141
39. Mavranezouli I, Megnin-Viggars O, Trickey D, Meiser-Stedman R, Daly C, Dias S, et al. Cost-effective-
ness of psychological interventions for children and young people with post-traumatic stress disorder. J
Child Psychol Psychiatry. 2019. Epub 2019/10/28. https://doi.org/10.1111/jcpp.13142 PMID: 31654414
40. Mavranezouli I, Megnin-Viggars O, Daly C, Dias S, Stockton S, Meiser-Stedman R, et al. Research
Review: Psychological and psychosocial treatments for children and young people with post-traumatic
stress disorder: a network meta-analysis. J Child Psychol Psychiatry. 2020; 61(1):18–29. https://doi.
org/10.1111/jcpp.13094 PMID: 31313834
41. Hamblen JL, Norman SB, Sonis JH, Phelps AJ, Bisson JI, Nunes VD, et al. A guide to guidelines for the
treatment of posttraumatic stress disorder in adults: An update. Psychotherapy (Chic). 2019; 56
(3):359–73.
42. Department of Veterans Affairs and Department of Defence (VA/DoD). VA/DoD clinical practice guide-
line for the management of posttraumatic stress disorder and acute stress disorder. Washington, DC:
VA/DoD; 2017. https://www.healthquality.va.gov/guidelines/MH/ptsd/
43. Phoenix Australia Centre for Posttraumatic Mental Health. Australian Guidelines for the treatment of
Acute Stress Disorder & Posttraumatic Stress Disorder. Phoenix Australia Centre for Posttraumatic
Mental Health; 2013. https://www.phoenixaustralia.org/resources/ptsd-guidelines/
44.
International Society for Traumatic Stress Studies. ISTSS PTSD Prevention and Treatment Guidelines
Methodology and Recommendations. ISTSS; 2019. www.istss.org/treating-trauma/new-istss-
prevention-and-treatment-guidelines
45. American Psychological Association. Clinical Practice Guideline for the Treatment of PTSD. Washing-
ton, DC: American Psychological Association; 2017. https://www.apa.org/ptsd-guideline/ptsd.pdf
PLOS ONE | https://doi.org/10.1371/journal.pone.0232245 April 30, 2020
22 / 22
PLOS ONE |
10.1371_journal.pone.0242533 | RESEARCH ARTICLE
Low prevalence of bloodstream infection and
high blood culture contamination rates in
patients with COVID-19
David Yu1,2¤, Karolina Ininbergs1,3, Karolina Hedman3, Christian G. Giske1,3,
Kristoffer Strålin4,5, Volkan O¨ zenciID
1,3¤*
1 Division of Clinical Microbiology, Department of Laboratory Medicine, Karolinska Institutet, Stockholm,
Sweden, 2 Functional Area of Emergency Medicine, Karolinska University Hospital, Stockholm, Sweden,
3 Department of Clinical Microbiology, Karolinska University Hospital, Stockholm, Sweden, 4 Department of
Medicine Huddinge, Karolinska Institutet, Stockholm, Sweden, 5 Department of Infectious Diseases,
Karolinska University Hospital, Stockholm, Sweden
¤ Current address: Division of Clinical Microbiology F 72, Karolinska Institutet, Karolinska University Hospital,
Stockholm, Sweden
* volkan.ozenci@ki.se
Abstract
Purpose
In the management of COVID-19, knowledge is lacking on the frequency of secondary bac-
terial infections and on how empirical antibiotic therapy should be used. In the present
study, we aimed to compare blood culture (BC) results of a COVID-19 patient cohort with
two cohorts of patients without detected COVID-19.
Methods
Using a retrospective cohort study design of patients subjected to BC in six tertiary care hos-
pitals, SARS-CoV-2 positive patients from March 1 to April 30 in 2020 (COVID-19 group)
were compared to patients without confirmed SARS-CoV-2 during the same period (control
group-2020) and with patients sampled March 1 to April 30 in 2019 (control group-2019).
The outcomes studied were proportion of BC positivity, clinically relevant growth, and con-
taminant growth.
Results
In total 15,103 patients and 17,865 BC episodes were studied. Clinically relevant growth
was detected in 197/3,027 (6.5%) BC episodes in the COVID-19 group compared to 717/
6,663 (10.8%) in control group-2020 (p<0.0001) and 850/8,175 (10.4%) in control group-
2019 (p<0.0001). Contamination was present in 255/3,027 (8.4%) BC episodes in the
COVID-19 group compared to 330/6,663 (5.0%) in control group-2020 (p<0.0001) and 354/
8,175 (4.3%) in control group-2019 (p<0.0001).
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: Yu D, Ininbergs K, Hedman K, Giske CG,
Strålin K, O¨ zenci V (2020) Low prevalence of
bloodstream infection and high blood culture
contamination rates in patients with COVID-19.
PLoS ONE 15(11): e0242533. https://doi.org/
10.1371/journal.pone.0242533
Editor: Surbhi Leekha, University of Maryland
School of Medicine, UNITED STATES
Received: July 2, 2020
Accepted: November 5, 2020
Published: November 23, 2020
Peer Review History: PLOS recognizes the
benefits of transparency in the peer review
process; therefore, we enable the publication of
all of the content of peer review and author
responses alongside final, published articles. The
editorial history of this article is available here:
https://doi.org/10.1371/journal.pone.0242533
Copyright: © 2020 Yu et al. This is an open access
article distributed under the terms of the Creative
Commons Attribution License, which permits
unrestricted use, distribution, and reproduction in
any medium, provided the original author and
source are credited.
Data Availability Statement: All relevant data are
within the manuscript and its Supporting
Information files.
PLOS ONE | https://doi.org/10.1371/journal.pone.0242533 November 23, 2020
1 / 10
PLOS ONE Funding: The authors received no specific funding
for this work.
Competing interests: The authors have declared
that no competing interests exist.
Conclusion
In COVID-19 patients, the prevalence of bloodstream bacterial infection is low and the con-
tamination rate of BC is high. This knowledge should influence guidelines regarding blood
culture sampling and empirical antibiotic therapy in COVID-19 patients.
Blood cultures in COVID-19 patients
Introduction
Secondary bacterial infections are a major clinical problem in patients with influenza virus
infections and have previously been reported to be associated with poor disease outcome [1, 2].
Bloodstream infections (BSIs) remain one of the most common and life-threatening complica-
tions in patients with severe viral infections. Epidemiological data of secondary BSIs might
therefore play a significant role in reducing mortality and morbidity rates due to COVID-19. A
retrospective study from USA reported that patients with COVID-19 have low bacteremia rates
than controls [3]. The blood culture routines and the characteristics of COVID-19 patients dif-
fer significantly between centers and geographic locations. Therefore, there is an imminent
need for studies on BSIs in COVID-19 to understand their importance for disease outcome.
The aim of the present study is to analyze blood culture data of a cohort of COVID-19
patients and compare it with two cohorts of patients without COVID-19.
Materials and methods
Setting
The study was performed between 1 March 2020 and 30 April 2020 at Karolinska University
Hospital, which serves a population of 2,436,767. The Karolinska University Laboratory
receives blood culture specimens from six tertiary care hospitals in the greater Stockholm area
and surrounding cities and suburbs. Historically, we have a yearly 10% increase in numbers of
our blood culture bottles at Karolinska University Laboratory without any change in contami-
nation rates.
Study design
Blood cultures collected from patients with COVID-19 and controls were analyzed retrospec-
tively. The blood culture data was retrieved from the laboratory information system (wwLab/
ADBakt, Autonik AB, Nykoping, Sweden) using QlikView (Qlik, King of Prussia, PA, USA).
Study population
Patients with COVID-19. Patients were considered to have COVID-19 if they were posi-
tive for SARS-CoV-2 RNA by reverse transcriptase PCR in respiratory secretions. Blood cul-
ture results from the COVID-19 patients registered between 1 March and 30 April in 2020 and
were included and referred to as “COVID-19 group” herein after.
Control groups. The study included two control groups, i.e. one historical control group
with blood culture results registered between 1 March and 30 April in 2019 (referred to as
“control group-2019”) and one contemporary control group with blood culture results regis-
tered between 1 March and 30 April in 2020, and with no confirmed PCR-positivity for
SARS-CoV-2 (referred to as “control group-2020”). In the beginning of the pandemic, testing
was only done in patients with symptoms consistent with COVID-19 and not all admitted
patients. In the present study, the control group-2020 therefore consisted of a mix of negative
PLOS ONE | https://doi.org/10.1371/journal.pone.0242533 November 23, 2020
2 / 10
PLOS ONE Blood cultures in COVID-19 patients
patients and patients not tested. However, all patients with COVID-19-like symptoms were
tested. Therefore, it is reasonable to assume that the patients not tested for SARS-COV-2 did
not have clinical findings of COVID-19.
Laboratory methods
Blood cultures. Three different blood culture bottles were used in the study; BacT/Alert
FA Plus aerobic, BacT/ALERT-PF Plus pediatric and BacT/Alert FN Plus anaerobic plus bot-
tles. Bottles were incubated in BacT/ALERT Virtuo (bioMe´rieux, Durham, NC, USA) blood
culture system until they signaled positive or for a maximum of five days.
The Gram stains were done directly from positive blood culture bottles. According to the
result from the staining, specimen from the positive bottles were subcultured onto relevant
agar plates. The microorganisms grown on the agar plates were identified by Bruker MALDI--
TOF MS. Antimicrobial susceptibility testing was performed by disc diffusion and the results
were interpreted following EUCAST recommendations (www.eucast.org).
SARS-CoV-2 RT-PCR. Testing for SARS-CoV-2 was performed using three different
RT-PCR assays: cobas SARS-CoV-2 (Roche Molecular Systems, Inc., Branchburg, NJ), Xpert
Xpress SARS-CoV-2 (XPRSARS-COV2-10) (Cepheid, Sunnyvale, CA) or an in-house devel-
oped assay based on Corman et al. [4] targeting the E- and RdRP-genes, with modifications of
primers according to Ede´n A et al. (Neurology, in revision); (Supplementary methods).
Data analysis
The data on blood cultures were presented as individual BSI episodes, from here on called
only “episode”. More than one episode from the same patient could be included, however, to
be defined as a new episode a minimum of 72 h had to pass between sampling of the same
patient. In case of more than 4 bottles taken in a single episode, only the first four bottles were
considered in the analysis. Following isolates were considered as contaminants if they grew in
less than 3 out of 4 blood culture bottles: Bacillus spp., Corynebacterium spp., Cutibacterium
spp., coagulase negative staphylococci (CoNS), Micrococcus spp., Cellulomonas spp., Lactoba-
cillus spp., Dermabacter spp., Facklamia spp., Rothia spp., Exiguobacterium spp., Brevibacter-
ium spp., and Trueperella spp.
Statistical analysis
The statistical analyses were performed with GraphPad Prism 5.0 (GraphPad Software, San
Diego, CA). The blood culture results in patients with COVID-19, control group-2020 and
control group-2019 were compared using the Pearson’s chi-square test. Values of P <0.05
were considered as statistically significant.
Results
In total, 58,704 blood culture bottles from 17,865 episodes in 15,103 patients were studied. The
study flow chart is depicted in Fig 1. The patients of the study groups had the following charac-
teristics; COVID-19 group, 790/2,240 (35.3%) female, mean (Standard Deviation [SD]) age 64
(18) years; control group-2020, 2,789/6,022 (46.3%) female, mean (SD) age 57 (26) years; and
control group-2019, 3,322/6,841 (48.6%) female, mean (SD) age 60 (26) years.
Overall blood culture positivity
The COVID-19 group consisted of 2240/8262 (27%) of all patients sampled for blood cultures
during the study period in 2020. The total number of blood cultured patients during the 2020
PLOS ONE | https://doi.org/10.1371/journal.pone.0242533 November 23, 2020
3 / 10
PLOS ONE Blood cultures in COVID-19 patients
Fig 1. Flow chart of the study population.
https://doi.org/10.1371/journal.pone.0242533.g001
study period, 8262, is an increase of 1421 (20%), compared to the same period last year (total
6841 patients). In 511/2,240 (22.8%) patients in the COVID-19 group, there were two or more
episodes during the study period. In the control groups, BCs were obtained from 6022 and
6841 patients in control group-2020 and control group-2019, respectively. In control group-
2020, 459 (7.6%) patients had two or more episodes. In control group-2019, 910 (13.3%)
patients had two or more episodes. In total 3,027 episodes in the COVID-19 group, 6,663 in
control group-2020 and 8,175 in control group-2019 were studied. Considering episodes,
growth of microorganisms in BC was detected in 433/3,027 (14.3%) episodes in the COVID-
19 group, compared with 1,015/6,663 (15.2%) in control group-2020 (non-significant) and
1,153/8,175 (14.1%) in control group-2019 (non-significant) (Table 1).
Table 1. Bloodstream infection episode data for patients with COVID-19 and both control groups.
Episode type
Included episodes, N
Episodes with growth, n (%)
Episodes with clinically relevant growth, n (%)
• Gram positive�
• Gram negative�
• Yeast�
• Polymicrobial episodes��
Episodes with contaminant growth, n (%)
• Only contaminant growth
• Both contaminant and clinically relevant growth
COVID-19
3027
433 (14.3)
197 (6.5)
116 (3.8)
64 (1.7)
2 (0.07)
27 (0.89)
255 (8.4)
236 (7.8)
19 (0.63)
Control group-2020
Control group-2019
6663
1015 (15.2)
717 (10.8)
344 (5.2)
306 (4.6)
11 (0.17)
56 (0.84)
330 (5.0)
298 (4.5)
32 (0.48)
8175
1153 (14.1)
851 (10.4)
420 (5.1)
351 (4.3)
8 (0.10)
72 (0.88)
354 (4.3)
302 (3.7)
52 (0.64)
�Monomicrobial episodes.
��Polymicrobial episode is defined as an episode with occurrence of more than one clinically relevant isolate.
https://doi.org/10.1371/journal.pone.0242533.t001
PLOS ONE | https://doi.org/10.1371/journal.pone.0242533 November 23, 2020
4 / 10
PLOS ONE Blood cultures in COVID-19 patients
Fig 2. Blood culture episodes with clinically relevant growth (Panel A) and with contaminant growth (Panel B). Total number of episodes included for analysis
were COVID-19 group: 3,027, Control group-2020: 6,663, Control group-2019: 8,175.
https://doi.org/10.1371/journal.pone.0242533.g002
Clinically relevant growth
Clinically relevant growth was detected in 197/3,027 (6.5%) of episodes in the COVID-19
group, compared with 717/6,663 (10.8%) in control group-2020 (p<0.0001) and 851/8,175
(10.4%) in control group-2019 (p<0.0001) (Table 1, Fig 2 [Panel A]).
When blood cultures with polymicrobial bacteremia were analyzed there was no difference
among the three groups studied (Table 1).
Contaminant growth
Contamination in blood cultures were detected in 255/3,027 (8.4%) episodes in patients with
COVID-19, compared with 330/6,663 (4.95%) episodes in control group-2020 (p<0.0001) and
354/8,175 (4.33%) episodes in control group-2019 (p<0.0001). The two control groups had sim-
ilar numbers of episodes with contaminant growth (non-significant) (Table 1, Fig 2 [Panel B]).
When relationship between the frequency of contaminants and the hospital localization
was analyzed, control group-2019 had higher contamination rates in ICUs than in emergency
departments and other clinics. In contrast, the contamination rates were similar in all hospital
locations for control group-2020. In the COVID-19 group, contamination rates were higher in
all hospital locations compared to the control groups, but more so in the emergency depart-
ments and ICUs (Table 2).
Table 2. Numbers and proportions of contamination in blood cultures from different hospital locations.
Hospital location
Emergency
department
Intensive care unit
Other hospital
locations
COVID-19
Control group 2020
Control group 2019
Contaminant (% of
total)
Total episodes
(n)
Contaminant (% of
total)
Total episodes
(n)
Contaminant (% of
total)
Total episodes
(n)
112 (9.2)
70 (10.6)
73 (6.4)
1221
659
1147
130 (4.8)
18 (5.6)
182 (5.0)
2705
319
3639
110 (4.0)
16 (7.9)
228 (4.4)
2769
202
5204
https://doi.org/10.1371/journal.pone.0242533.t002
PLOS ONE | https://doi.org/10.1371/journal.pone.0242533 November 23, 2020
5 / 10
PLOS ONE Table 3. Distribution of microorganisms isolated from blood cultures in patients with COVID-19 and both control groups.
COVID-19
Control group-2020
Control group-2019
Blood cultures in COVID-19 patients
All isolates, N
Gram-positive bacteria� n (%)
• Coagulase negative staphylococci��
• Staphylococcus aureus
• Enterococcus spp.
• Viridans group streptococci
• Beta-hemolytic streptococci
• Streptococcus pneumoniae
• Other Gram-positive bacteria†
Gram-negative bacteria� n (%)
• Escherichia coli
• Other Enterobacterales
• Pseudomonas aeruginosa
• Other Gram-negative bacteria†
Anaerobic bacteria n (%)†
Yeast n (%)†
�Not including anaerobic bacteria
��Including Staphylococcus epidermidis
†Other bacteria and yeast are shown in S1 Table.
https://doi.org/10.1371/journal.pone.0242533.t003
226
150 (66)
49 (22)
44 (19)
23 (10)
22 (9.7)
6 (2.7)
3 (1.3)
3 (1.3)
66 (29)
34 (15)
24 (11)
7 (3.1)
1 (0.4)
7 (3.1)
3 (1.3)
781
385 (49)
55 (7.1)
124 (16)
46 (5.9)
64 (8.2)
47 (6.0)
21 (1.4)
28 (3.6)
352 (45)
213 (27)
111 (14)
13 (1.7)
15 (1.9)
32 (4.1)
11 (1.4)
940
487 (52)
58 (6.2)
162 (17)
75 (8.0)
63 (6.7)
50 (5.3)
51 (1.6)
28 (3.0)
397 (42)
242 (26)
117 (13)
14 (1.5)
24 (2.6)
40 (4.3)
15 (1.6)
Similar results were observed when numbers of bottles with contaminants were analyzed.
In total 337/10,504 (3.2%) bottles in the COVID-19 group were contaminated, compared to
433/21,261 (2.0%) in control group-2020 (p<0.0001) and 470/26,939 (1.7%) in control group-
2019 (p<0.0001).
Diversity of microorganisms in blood cultures
There was a significant diversity in microorganisms detected from blood cultures among the
three groups studied. Gram-positive growth was significantly higher in patients with COVID-
19, 150/226 (66%) isolates, than in control group-2020 and -2019, 385/781 (49%) 487/940
(52%) isolates respectively (p<0.0001 for both comparisons) (Table 3). The two control groups
had similar levels of numbers of Gram-positive growth (non-significant). In contrast, Gram-
negative isolates was significantly fewer in patients with COVID-19, 66/226 (29%) isolates,
than in control group-2020 353/781 (45%) isolates and -2019 398/940 (42%) isolates
(p<0.0001 and p<0.001 respectively). The two control groups had similar levels of numbers of
Gram-negative isolates (non-significant).
The most common three microorganisms detected in blood cultures from COVID-19
patients were CoNS, Staphylococcus aureus and Escherichia coli. In contrast, for both control
groups the three most common isolates were E. coli, S. aureus and other Enterobacterales.
Patients with COVID-19 had higher numbers of episodes with clinically relevant CoNS
than in control group-2020 and -2019 (p<0.0001 for both comparisons). There was no differ-
ence in numbers of episodes with CoNS between the two control groups (non-significant). In
contrast, lower numbers of Escherichia coli were observed in patients with COVID-19 (15%)
compared to control group-2020 (27%) and -2019 (26%) (p = 0.0001 and p = 0.0005, respec-
tively). There was no significant difference for S. aureus, anaerobes and yeasts among the three
groups studied. A detailed list of all microorganisms can be found in supplements (S1 Table).
PLOS ONE | https://doi.org/10.1371/journal.pone.0242533 November 23, 2020
6 / 10
PLOS ONE Blood cultures in COVID-19 patients
Table 4. Time to detection of microorganisms in positive blood cultures in patients with COVID-19 and both control groups.
Time to detection
Mean (SD [h])
Median (IQR [h])
COVID-19
(n = 820)
22.8 (17.6)
18.5 (11.5)
Control group
2020 (n = 2185)
18.4 (15.3)
14.0 (10.7)
Control group
2019 (n = 2663)
18.2 (17.2)
13.2 (10.4)
SD: standard deviation. IQR: interquartile range.
https://doi.org/10.1371/journal.pone.0242533.t004
Time to detection in blood cultures
The time to detection (TTD) in positive blood culture bottles differed between the three
groups analyzed. The mean (SD) TTD was 22.8 (17.6) in COVID-19 group whereas control
group-2020 and control group-2019 had 18.4 (15.3) and 18.2 (17.2) h (Table 4).
When the incubation period for blood cultures were analyzed in a total of 5568 bottles, we
observed that the vast majority (93–96%) of the bottles signaled positive in 48 h. An additional
4–5% signaled positive in 4 days. The remaining BC that signaled positive in 5 days were 2% in
COVID-19 group and 1% in both control groups (Table 5).
Discussion
Blood culture is the gold standard for detection of microorganisms in patients with BSI. We
presented the blood culture findings in patients with COVID-19 and in other patients with
suspected BSIs from a contemporary and a historical control group.
In the present study the overall blood culture positivity rate was similar in all three groups
analyzed. However, the proportion of episodes with clinically relevant growth was significantly
lower in in patients with COVID-19 than both control groups. Although in our study the true
incidence of bacteremia was not known, the proportion of episodes with clinically relevant
growth correlates with previous data on clinical characteristics in COVID-19, where bacter-
emia was observed in 5.6% of cases [5] and septic shock in 4% of cases [6]. The reason for
lower bacteremia rates in patient with COVID-19 is largely unknown. Patients with severe
COVID-19 fulfill the sepsis-3 criteria for sepsis [7] and the term viral sepsis has been intro-
duced [8]. In patients who are hospitalized for COVID-19, it is thus difficult to use clinical and
laboratory parameters to differentiate between the viral component and a potential bacterial
component. The low rate of relevant bacteremia indicates that the viral component is predom-
inant in COVID-19. There was a higher proportion of patients with more than one suspected
BSI episode in the COVID-19 group compared to the control groups. The reason for this is
not known.
In a recent study, from New York, USA, it is reported that only 3.8% of COVID-19 patients
had positive blood cultures which was significantly lower than the controls in that study [3].
The present data differs from the NY study since the overall BC positivity rate in our COVID-
Table 5. Proportions of positive blood cultures signaling positive during given time intervals in the blood culture system.
Time to detection
Day 1–2
Day 3
Day 4
Day 5
https://doi.org/10.1371/journal.pone.0242533.t005
COVID-19
(n = 820)
93%
4%
1%
2%
Control group 2020
(n = 2185)
Control group 2019
(n = 2663)
96%
3%
1%
1%
95%
3%
1%
1%
PLOS ONE | https://doi.org/10.1371/journal.pone.0242533 November 23, 2020
7 / 10
PLOS ONE Blood cultures in COVID-19 patients
19 group was 14.3% and did not differ from controls. Moreover, clinically relevant growth in
the present COVID-19 cohort was 6.5% in comparison to 1.8% in COVID-19 patients
reported by Sepulveda et al. [3]. The underlying reason for these differences is unknown. It is
reasonable to assume that the blood culture routines and the characteristics of COVID-19
patients were different. However, both studies showed that the clinically relevant growth is
lower in patients with COVID-19 than in controls.
The present results show that bacterial and fungal BSIs are uncommon in patients with
COVID-19 and it is warranted to establish stringent clinical criteria for empiric antibiotic
treatment for BSI in these patients.
The species composition of microorganisms isolated from blood cultures from patients
with COVID-19 and controls differed. When the three most common microorganisms iso-
lated from the positive bottles were considered, patients with COVID-19 had significantly
higher rates of clinically relevant CoNS than both control groups (p<0.0001 for both compari-
sons). In contrast, both control groups had higher rates of E. coli compared to COVID-19
patients (p = 0.0001 and p = 0.0005, respectively). All three groups had similar levels of S.
aureus (non-significant). The underlying differences in occurrence of BSI with CoNS and E.
coli in patients with COVID-19 and controls might be important in empiric antibiotic treat-
ment of these patients. Overall, our results emphasize the importance of antimicrobial stew-
ardship in the treatment of COVID-19 patient to minimize the threat of superinfections [9].
Contamination is a major problem in blood cultures. It was recently reported that COVID-
19 patients had higher proportion of cultures that likely represented contamination with nor-
mal skin microbiota than controls [3]. However, the study did not analyze the growth of con-
taminants further. Growth of normal skin microbiota might be clinically relevant. The present
study analyzed the contaminants in detail both at episode and BC bottle level, by using an algo-
rithm to discriminate possible clinically relevant growth of normal skin flora bacteria [10, 11].
Patients with COVID-19 had 3.2% blood culture bottles (8.4% episodes) contaminated as com-
pared to 2.04% (5.0% episodes) and 1.74% (4.2% episodes) in the two control groups, respec-
tively. Under normal circumstances, blood cultures received in our center has low
contamination rates as shown in the two control groups studied. In contrast, the blood culture
bottles in the COVID-19 group exceeded the recommended rate of contamination of <3%
according to CLSI guidelines [12], which may lead to unnecessary antibiotic use and longer
hospital stay for these patients. The underlying reason for high contamination rates in patients
with COVID-19 is not known. In the present study, contamination rates were higher in emer-
gency departments and ICUs compared to other units in the COVID-19 group. The relation-
ship between high stress environments and BC contamination rates has been previously
reported [13, 14]. The patient characteristics and the work pace in emergency departments
and ICUs differs from the other units. It is reasonable to suggest that the stressful working
environment in these two units with well-known risk to be exposed to SARS-CoV-2 might
play an important role in higher contamination rates observed in COVID-19 group.
TTD of positive blood cultures might be a relevant parameter in comparing different
patient populations [15]. In the present study we showed that the mean TTD in COVID-19
was approximately 20% longer compared to controls. It is reasonable to suggest that the longer
TTD in COVID-19 group is based on higher rates of contaminants in this group.
The recommended incubation period for blood cultures is 5 days. In line with a recently
published study, we observed that 98–99% of the BC bottles in all three groups signaled posi-
tive in 4 days [3]. The present data support the assumption that BC can be incubated for a
maximum of 4 days when it is necessary.
To our knowledge, this is the first European study analyzing blood culture data from
patients with COVID-19 and has several strengths. The inclusion of over 15,000 patients from
PLOS ONE | https://doi.org/10.1371/journal.pone.0242533 November 23, 2020
8 / 10
PLOS ONE Blood cultures in COVID-19 patients
six tertiary care hospital is an important strength of the study design. The study analyzed the
clinically relevant growth and contaminants in detail and had relatively high positive blood
culture rates in all three groups studied.
The present study also has important limitations. First, we did not have access to baseline
clinical data such as comorbidity, disease duration, length of hospital stay and treatment of
patients in the three study groups. Therefore, assessment of the impact of differences in patient
characteristics on the BC results could not be analyzed. Second, we did not include a control
group with another viral respiratory infection, such as influenza, during the same season.
Third, as the present study focused on BSI in COVID-19, data regarding other culture results
were not analyzed, which precluded the analysis of other secondary infections such as
pneumonia.
Conclusions
The present study shows that patients with COVID-19 have low prevalence of BSI and a higher
rate of contamination in blood cultures. Further clinical studies are warranted in order to
improve blood culture-based diagnostics in patients with COVID-19.
Supporting information
S1 Table. Distribution of all microorganisms isolated from blood cultures.
(DOCX)
Author Contributions
Data curation: David Yu, Karolina Ininbergs, Karolina Hedman.
Formal analysis: David Yu, Karolina Ininbergs.
Methodology: David Yu, Christian G. Giske, Kristoffer Strålin, Volkan O¨ zenci.
Project administration: Volkan O¨ zenci.
Supervision: Volkan O¨ zenci.
Writing – original draft: David Yu, Christian G. Giske, Volkan O¨ zenci.
Writing – review & editing: Volkan O¨ zenci.
References
1.
Liderot K, Ahl M, Ozenci V (2013) Secondary bacterial infections in patients with seasonal influenza A
and pandemic H1N1. Biomed Res Int 2013:376219 https://doi.org/10.1155/2013/376219 PMID:
23865050
2. Morens DM, Taubenberger JK, Fauci AS (2008) Predominant role of bacterial pneumonia as a cause of
death in pandemic influenza: implications for pandemic influenza preparedness. J Infect Dis 198
(7):962–970 https://doi.org/10.1086/591708 PMID: 18710327
3. Sepulveda J, Westblade LF, Whittier S, Satlin MJ, Greendyke WG, Aaron JG, et al. (2020) Bacteremia
and Blood Culture Utilization During COVID-19 Surge in New York City. J Clin Microbiol https://doi.org/
10.1128/JCM.00875-20 PMID: 32404482
4. Corman VM, Landt O, Kaiser M, Molenkamp R, Meijer A, Chu DK, et al. (2020) Detection of 2019 novel
coronavirus (2019-nCoV) by real-time RT-PCR. Euro Surveill 25 (3)
5. Goyal P, Choi JJ, Pinheiro LC, Schenck EJ, Chen R, Jabri A, et al. (2020) Clinical Characteristics of
Covid-19 in New York City. New England Journal of Medicine https://doi.org/10.1056/NEJMc2010419
PMID: 32302078
PLOS ONE | https://doi.org/10.1371/journal.pone.0242533 November 23, 2020
9 / 10
PLOS ONE Blood cultures in COVID-19 patients
6. Chen N, Zhou M, Dong X, Qu J, Gong F, Han Y, et al. (2020) Epidemiological and clinical characteristics
of 99 cases of 2019 novel coronavirus pneumonia in Wuhan, China: a descriptive study. Lancet 395
(10223):507–513 https://doi.org/10.1016/S0140-6736(20)30211-7 PMID: 32007143
7. Singer M, Deutschman CS, Seymour CW, Shankar-Hari M, Annane D, Bauer M, et al. (2016) The Third
International Consensus Definitions for Sepsis and Septic Shock (Sepsis-3). JAMA 315 (8):801–810
https://doi.org/10.1001/jama.2016.0287 PMID: 26903338
8.
Li H, Liu L, Zhang D, Xu J, Dai H, Tang N, et al. (2020) SARS-CoV-2 and viral sepsis: observations and
hypotheses. Lancet 395 (10235):1517–1520 https://doi.org/10.1016/S0140-6736(20)30920-X PMID:
32311318
9. Clancy CJ, Nguyen MH (2020) COVID-19, superinfections and antimicrobial development: What can
we expect? Clin Infect Dis
10. Yu D, Larsson A, Parke A, Unge C, Henning C, Sunden-Cullberg J, et al. (2020) Single-Sampling Strat-
egy vs. Multi-Sampling Strategy for Blood Cultures in Sepsis: A Prospective Non-inferiority Study. Front
Microbiol 11:1639 https://doi.org/10.3389/fmicb.2020.01639 PMID: 32793149
11.
Leyssene D, Gardes S, Vilquin P, Flandrois JP, Carret G, Lamy B (2011) Species-driven interpretation
guidelines in case of a single-sampling strategy for blood culture. Eur J Clin Microbiol Infect Dis 30
(12):1537–1541 https://doi.org/10.1007/s10096-011-1257-3 PMID: 21499970
12. Dargere S, Cormier H, Verdon R (2018) Contaminants in blood cultures: importance, implications, inter-
pretation and prevention. Clin Microbiol Infect 24 (9):964–969 https://doi.org/10.1016/j.cmi.2018.03.
030 PMID: 29621616
13. Self WH, Mickanin J, Grijalva CG, Grant FH, Henderson MC, Corley G, et al. (2014) Reducing blood cul-
ture contamination in community hospital emergency departments: a multicenter evaluation of a quality
improvement intervention. Acad Emerg Med 21 (3):274–282 https://doi.org/10.1111/acem.12337
PMID: 24628752
14. Self WH, Speroff T, Grijalva CG, McNaughton CD, Ashburn J, Liu D, et al. (2013) Reducing blood cul-
ture contamination in the emergency department: an interrupted time series quality improvement study.
Acad Emerg Med 20 (1):89–97 https://doi.org/10.1111/acem.12057 PMID: 23570482
15.
Lamy B (2019) Reprint of: Blood culture time-to-positivity: making use of the hidden information. Clin
Microbiol Infect</References> https://doi.org/10.1016/j.cmi.2019.03.005 PMID: 30898587
PLOS ONE | https://doi.org/10.1371/journal.pone.0242533 November 23, 2020
10 / 10
PLOS ONE |
10.1371_journal.pone.0234902 | RESEARCH ARTICLE
A deep learning approach to predict visual
field using optical coherence tomography
Keunheung Park1,2, Jinmi Kim3, Jiwoong LeeID
1,2*
1 Department of Ophthalmology, Pusan National University College of Medicine, Busan, South Korea,
2 Biomedical Research Institute, Pusan National University Hospital, Busan, South Korea, 3 Department of
Biostatistics, Clinical Trial Center, Biomedical Research Institute, Pusan National University Hospital, Busan,
South Korea
* alertlee@naver.com
Abstract
We developed a deep learning architecture based on Inception V3 to predict visual field
using optical coherence tomography (OCT) imaging and evaluated its performance. Two
OCT images, macular ganglion cell-inner plexiform layer (mGCIPL) and peripapillary retinal
nerve fibre layer (pRNFL) thicknesses, were acquired and combined. A convolutional neural
network architecture was constructed to predict visual field using this combined OCT image.
The root mean square error (RMSE) between the actual and predicted visual fields was cal-
culated to evaluate the performance. Globally (the entire visual field area), the RMSE for all
patients was 4.79 ± 2.56 dB, with 3.27 dB and 5.27 dB for the normal and glaucoma groups,
respectively. The RMSE of the macular region (4.40 dB) was higher than that of the periph-
eral region (4.29 dB) for all subjects. In normal subjects, the RMSE of the macular region
(2.45 dB) was significantly lower than that of the peripheral region (3.11 dB), whereas in
glaucoma subjects, the RMSE was higher (5.62 dB versus 5.03 dB, respectively). The deep
learning method effectively predicted the visual field 24–2 using the combined OCT image.
This method may help clinicians determine visual fields, particularly for patients who are
unable to undergo a physical visual field exam.
Introduction
Glaucoma is one of the leading causes of blindness in the world [1,2]. It is a widespread
chronic, irreversible optic neuropathy characterised by the progressive and permanent loss of
retinal ganglion cells (RGCs) and their axons. It is associated with visual field abnormalities,
the loss of which can greatly impact the quality of life [3,4]. In practice, monitoring visual field
examination is an important process in preventing vision loss.
However, visual field exams are very subjective tests and depend largely on patient compli-
ance. They inherently involve several random errors and fluctuations, which can be affected by
various factors, that result in a low signal-to-noise ratio [5]. The fluctuations are more severe
in glaucomatous patients than in normal subjects [6,7]. It is often thought that a visual field
exam is difficult to perform as several factors can affect the quality of the exam, some of which
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: Park K, Kim J, Lee J (2020) A deep
learning approach to predict visual field using
optical coherence tomography. PLoS ONE 15(7):
e0234902. https://doi.org/10.1371/journal.
pone.0234902
Editor: Ireneusz Grulkowski, Nicolaus Copernicus
University, POLAND
Received: September 18, 2019
Accepted: June 4, 2020
Published: July 6, 2020
Peer Review History: PLOS recognizes the
benefits of transparency in the peer review
process; therefore, we enable the publication of
all of the content of peer review and author
responses alongside final, published articles. The
editorial history of this article is available here:
https://doi.org/10.1371/journal.pone.0234902
Copyright: © 2020 Park et al. This is an open
access article distributed under the terms of the
Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: Data is provided with
supporting information file. Please refer S1 File. All
python source code, trained model, and test
images are available on Github web site: https://
github.com/climyth/VFbySD-OCT (DOI: 10.5281/
PLOS ONE | https://doi.org/10.1371/journal.pone.0234902 July 6, 2020
1 / 19
PLOS ONE zenodo.3757702). The training images are
available from the first (climyth@naver.com) or
corresponding author (alertlee@naver.com) on
reasonable request. The provided data must be
used for only research purpose and should not be
shared with other unauthorized institutions. Data
will be stored in our research NAS (network
attached storage) and cloud storage for backup
purpose up to December 2022, after which it will
be discarded as required by local regulations.
Although the authors cannot make their study’s
data publicly available at the time of publication, all
authors commit to make the data underlying the
findings described in this study fully available
without restriction to those who request the data,
in compliance with the PLOS Data Availability
policy. For data sets involving personally
identifiable information or other sensitive data, data
sharing is contingent on the data being handled
appropriately by the data requester and in
accordance with all applicable local requirements.
Funding: KHP, This research was supported by the
Bio & Medical Technology Development Program
of the National Research Foundation (NRF) funded
by the South Korean government (MSIT) (No.
NRF-2018M3A9E8066254).
Competing interests: The authors have declared
that no competing interests exist.
Visual field prediction from OCT
include patient attention, fatigue, artefacts such as ptosis, lens rim defects, and incorrect refrac-
tive error correction. Despite attempts to minimise or control the influence of all these factors,
a patient’s learning curve may also affect the outcome of the visual field exam [8,9].
In contrast to the visual field exam, optical coherence tomography (OCT) is an objective
test and its reproducibility is known to be excellent [10–15]. Patients show evidence of struc-
tural changes, including optic nerve head (ONH) damage and retinal nerve fibre layer (RNFL)
thinning, before functional loss is detected by standard automated perimetry [16,17]. It is reli-
able when performed on both normal and glaucoma patients [18]. Structural changes mea-
sured by OCT are closely related to the functional changes in the visual field [19]. The pattern
of correlation and corresponding locations between structure and function have been investi-
gated by several previous studies [20–25]. Taken together, this suggests that it may be possible
to deduce a visual field test from OCT images, which would be very helpful in monitoring
patients who are unable to undergo visual field testing, including children, the elderly, and
those with dementia. There have been previous attempts to predict visual fields using OCT
images [26–28]; however, they used a pointwise numerical regression method.
More recently, computer technology has tremendously improved and with the aid of GPUs
(graphic processing units), parallel processing capability, which is important in neural network
computation, is also greatly advanced. Artificial intelligence algorithms have also improved
and recently, ‘deep learning algorithms’ have emerged, performing at levels almost comparable
to that of humans [29–31]. The biggest advantage of deep learning algorithm is that it is an
end-to-end learning algorithm, i.e. a precise mechanism does not need to be provided to
resolve complex problems; rather, such mechanisms are learnt during training. Structure-
function relationship is a complex and non-linear problem with many unpredictable errors
and large variations among patients. Neural network computation may be a good choice to
deal with these types of complex problems.
The purpose of this study was to construct a deep learning architecture to predict visual
fields using OCT images and evaluate its performance. We built a model using a state-of-the-
art convolutional neural network (CNN) architecture and tested its accuracy globally and
regionally. We also attempted to identify various factors which affected visual field prediction.
Materials and methods
This retrospective study was performed in accordance with the tenets of the Declaration of
Helsinki. The study was approved by the institutional review board (IRB) of Pusan National
University Hospital, South Korea. The requirement for patient consent was waived by the IRB
due to the retrospective nature of the study.
All training and test data were obtained from subjects who had visited the glaucoma clinic
at Pusan National University Hospital from 2013 to 2018. The demographic characteristics of
the training group are summarised in Table 1. The training dataset consisted of 2,811 eyes
from 1,529 subjects and was not labelled by diagnosis. Therefore, normal fundus images, as
well as data from subjects with glaucoma and other optic neuropathies, were included. How-
ever, eyes with retinal disease or severe media opacity (such as cataracts) were excluded. The
mean ± standard deviation (SD) age of the test group was 62.1 ± 16.6 years. A total of 2,811
records from the training dataset was randomly split into training and validation data at a
ratio of 9:1. Validation data were used to check the current fitness of the neural network during
training to prevent overfitting.
In addition to the training dataset, a separate test dataset was prepared with 290 eyes from
290 subjects, with no patient overlap between datasets. For all subjects in the test group, a ret-
rospective review of the detailed results of ophthalmic examinations was performed, including
PLOS ONE | https://doi.org/10.1371/journal.pone.0234902 July 6, 2020
2 / 19
PLOS ONE Visual field prediction from OCT
Table 1. Demographic characteristics of the training group.
Total number of eyes
Total number of patients
Age (years; mean ± SD)
Number of eyes binned by visual field mean deviation (MD)
MD > –3 dB
–3 dB � MD > –6 dB
–6 dB � MD > –9 dB
–9 dB � MD > –12 dB
–12 dB � MD
SD: standard deviation.
https://doi.org/10.1371/journal.pone.0234902.t001
Values
2811
1529
62.1 ± 16.6
1019 (36.2%)
656 (23.3%)
328 (11.7%)
180 (6.4%)
628 (22.3%)
best corrected visual acuity (BCVA), Goldmann applanation tonometry (GAT), slit-lamp
examination, funduscopy, biometry using the IOLMaster (Carl Zeiss Meditec, Dublin, CA,
USA), Humphrey visual field test (Carl Zeiss Meditec), central corneal thickness (CCT) using
ultrasonic pachymetry (Pachmate; DGH Technology, Exton, PA, USA), keratometry using the
Auto Kerato-Refractometer (ARK-510A; NIDEK, Hiroshi, Japan), and Cirrus high definition
optical coherence tomography (HD-OCT; Carl Zeiss Meditec). Glaucomatous optic neuropa-
thy was defined if one or more of the following criteria were met: focal or diffuse neuroretinal
rim thinning, localised notching, cup-to-disc ratio asymmetry �0.2, and the presence of reti-
nal nerve fibre layer (RNFL) defects congruent with visual field defects [32]. Normal subjects
were defined as those with no history of ocular disease, an intraocular pressure <21 mm Hg,
an absence of a glaucomatous optic disc appearance, and a normal visual field. To ensure
representation of the full range of disease, normal subjects also included those who were clini-
cally suspected of having glaucoma, based on optic disc or RNFL appearance, or elevated intra-
ocular pressure, but had normal visual field. Patients with corneal or ocular media opacity, a
refractive error �±6.0 dioptres, optic neuropathies other than glaucoma, or recent ocular sur-
gery or trauma were excluded.
Spectral Domain Optical Coherence Tomography (SD-OCT)
The Cirrus spectral domain (SD)-OCT instrument (Carl Zeiss Meditec) was used to acquire
macular ganglion cell-inner plexiform layer (mGCIPL) and peripapillary retinal nerve fibre
layer (pRNFL) thickness maps. Two consecutive OCT exams, 6 mm × 6 mm macular cube
scan 200 × 200 protocol and 6 mm x 6 mm optic disc cube 200 × 200 scan, were performed at
the same time to obtain both mGCIPL and pRNFL thickness maps. Following pupil dilation
using 0.5% tropicamide and 0.5% phenylephrine, the subject was seated and properly aligned.
The eye was then brought into view using the mouse-driven alignment system and the line
scanning image was focused by adjusting for refractive error. The macular centre or ONH was
shown at the centre of the live image and, further centring (Z-offset) and enhancement were
optimised. The laser scanned over a 6 mm x 6 mm square area, capturing a cube of data con-
sisting of 200 × 200 A-scans to make B-scans (40,000 points) in about 1.5 seconds (27,000 A-
scans/sec). The ganglion cell analysis algorithm automatically segmented the GCIPL and then
calculated the thickness of the mGCIPL within a 6 mm x 6 mm square area centred on the
fovea. The RNFL analysis algorithm automatically segmented the peripapillary RNFL layer
and calculated its thickness within a 6 mm x 6 mm square area centred at the ONH centre. For
quality control, only good quality scans defined as having a minimum signal strength of 6, no
PLOS ONE | https://doi.org/10.1371/journal.pone.0234902 July 6, 2020
3 / 19
PLOS ONE Visual field prediction from OCT
involuntary eye movements, blinking artefacts, or being without misalignment or segmenta-
tion failures were used for analysis.
Input image generation and visual field region definition
We developed a custom software to generate a combined image of mGCIPL and pRNFL thick-
ness maps. In Fig 1, an example of the combined OCT image is shown on top of the neural net-
work architecture. Our custom software utilised two report images exported from Cirrus
OCT: 1) Ganglion Cell OU Analysis: Macular Cube 200 × 200, and 2) ONH and RNFL OU
Analysis: Optic Disc Cube 200 × 200 protocol. The mGCIPL and pRNFL thickness maps from
both eyes were present in the report image. The custom software automatically detected the
location of these blue-toned thickness maps by searching for the rectangular boundary of the
blue image, starting from a predefined location, and cropped and combined them (mGCIPL
map on the left, ONH map on the right). All left eye images were flipped horizontally to match
the format of the right eye.
In Fig 2A, the fundus photo with Humphrey visual field 24–2 test points and two superim-
posed OCT thickness maps are shown. This image was drawn by our custom software which
precisely locates the visual field test points at their designated locations. The fundus photo was
taken with the Nidek AFC-330 camera (Hiroshi, Japan) with a 45˚ horizontal field of view.
The original fundus image had a resolution of 2438 x 2112 (width x height); however, the
actual fundus area without the margin had a resolution of 2290 x 2112 (width x height), gener-
ating a linear scaling of 51 pixels per degree. Based on this information, the custom software
located the visual field test points. A user manually overlapped the mGCIPL and pRNFL OCT
images on the fundus photos by exactly matching retinal vessels and the shape of the ONH. In
this overlapped image, we noted that the central 4 × 4 visual field test points were inside the
mGCIPL scan area. We defined this central visual field area as the ‘macular OCT scan area’
and the surrounding area as the ‘peripheral OCT scan area’ (Fig 2B). We defined another set
of regions using the Garway-Heath sectorisation [23]. It consisted of six sectoral areas on the
ONH corresponding to the visual field test points (Fig 2B).
Visual field examination
Within 6 months of the OCT exam, automated perimetry was performed on all training and
test subjects using a Humphrey Visual Field Analyzer 750i instrument (Carl Zeiss Meditec)
with the Swedish interactive threshold algorithm (SITA) 24–2 or 30–2. Of the 54 test points of
the 24–2 test pattern, 2 points of physiologic scotoma were excluded and the remaining 52 test
points of the total threshold value were used as the ground truth visual field of the training and
test sets. The 30–2 test pattern was converted to 24–2 by using overlapping test points. Reliable
visual field tests were defined as having a false-positive rate <33%, false-negative rate <33%,
and fixation loss <20%. Normal subjects were defined as those with a glaucoma hemifield test
(GHT) within the normal limits, and with a mean deviation (MD) and pattern standard devia-
tion (PSD) within 95% of the normal population. Glaucomatous visual fields were those that
met at least one of the following criteria: GHT outside the normal limits and/or PSD probabil-
ity outside of 95% of the normal population. Glaucoma severity was determined based on the
MD of the visual field test: early >–6 dB and moderate to severe �–6 dB.
Deep learning architectures and training
The open source deep learning platform, Keras library [33], running on the TensorFlowTM
backend (Google, Mountain View, CA, USA) python API r1.10, was used. Python language
version 3.5 was used with the CUDA toolkit 9.0 and cuDNN 7.0 library to utilise the GPU
PLOS ONE | https://doi.org/10.1371/journal.pone.0234902 July 6, 2020
4 / 19
PLOS ONE Visual field prediction from OCT
PLOS ONE | https://doi.org/10.1371/journal.pone.0234902 July 6, 2020
5 / 19
PLOS ONE Visual field prediction from OCT
Fig 1. Deep learning architecture. The shape of the tensor (input/output) is described on the right side of each layer
box. The global average pooling layer and four fully connected network (dense) layers were connected after the
Inception V3 backbone CNN (convolutional neural network) architecture. The four dense layers used ReLu (rectified
linear unit) as the activation function.
https://doi.org/10.1371/journal.pone.0234902.g001
computation power. The hardware environment used for training and test runs was Intel i5-
8400 CPU, 32 GB RAM, and a GeForce Titan Volta (NVIDIA, Santa Clara, CA, USA).
The final deep neural network architecture used in this study is shown in Fig 1. A state-of-
the-art CNN architecture, Inception V3 [34] developed by Google, was used as the backbone
structure to extract global features. A bottleneck layer of the Inception V3 was removed and
replaced with one global average pooling layer followed by four consecutive densely connected
Fig 2. Visual field test pattern and Garway-Heath map. (A) A colour fundus photo with Humphrey 24–2 visual field
test pattern and two optical coherence tomography (OCT) images were overlapped on the fundus photo. Garway-
Heath sectorisation (white radiating line) is drawn on the optic nerve head centre. (B) Regions of visual field test points
outlined by Garway-Heath sectorisation map. The central dashed square shows the boundary of the macular OCT scan
area and the surrounding area is defined as the peripheral OCT scan area. IN: inferonasal, IT: inferotemporal, N: nasal,
SN: superonasal, ST: superotemporal, T: temporal.
https://doi.org/10.1371/journal.pone.0234902.g002
PLOS ONE | https://doi.org/10.1371/journal.pone.0234902 July 6, 2020
6 / 19
PLOS ONE Visual field prediction from OCT
layers. All dense layers used ReLu (rectified linear unit) as the activation function. A combined
OCT image, which had a size of 322 × 161 (width × height), was used as input data. The Incep-
tion V3 used the input image to produce a 3 × 8 × 2048 (height × width × depth) image con-
taining global features. The global average pooling layer flattened the output of Inception V3
and made averaged 2048 features. Four dense layers condensed these features into 52 final out-
put neurons which corresponded to 52 visual field threshold values (two points of physiologic
scotoma were excluded from prediction).
Before training began, Inception V3 pretrained on the ImageNet dataset was downloaded
and applied. No layer was frozen during training and all layers were fine-tuned. A total of
2,811 records were randomly aplit into training and validation datasets in a 9:1 ratio and
batches of 64 were supplied to the neural network. The optimizer was ‘rmsprop’ and the loss
function was ‘mean squared error’. Training was monitored by reference to the loss trends of
both the training and validation sets. When no further performance gain was observed over
100 epochs, training finished. To prevent overfitting, the repeated random sub-sampling cross
validation technique [35] was used. The training data were again randomly split in a 9:1 ratio,
the last trained weight file was loaded, and training resumed until no further performance gain
was evident over 100 epochs. This process was repeated five times.
Statistical analyses
The Shapiro-Wilk test was performed to check the normality of the data distribution. To com-
pare parameters between normal subjects and glaucoma patients, we used Student’s t-test or
Mann-Whitney U test depending on the normality of the data. The chi-square test was used
for categorical variables. Visual field prediction error was calculated as the root mean square
error (RMSE) using the following formula:
RMSE ¼
s
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
X52
ðtrue THVn (cid:0) predicted THVn Þ2
52
n¼1
n ¼ nth test point of visual field exam; THV ¼ visual field threshold value
The above formula is an example of the global RMSE calculation (i.e. includes all 52 test
points). When we calculated regional prediction error, only a select number of visual field test
points inside the target region were used. Those regions are defined in Fig 2B.
We performed correlation analysis and simple linear regression analysis to identify factors
affecting visual field prediction. Depending on data normality, Pearson’s correlation coeffi-
cient or Spearman’s rank correlation coefficient were used. Multiple linear regression analyses
with the ENTER method were also used to identify the importance of possible factors affecting
visual field prediction. For conducting statistical analyses, SPSS (version 21.0 for Windows;
SPSS, Chicago, IL, USA) and MedCalc (version 12.5 for Windows; Ostend, Belgium) were
used, and P < 0.05 (single comparison) and P < 0.017 (multiple comparisons) were consid-
ered to indicate statistical significance.
Results
A total of 290 eyes from 290 subjects were recruited for the test group, including 112 normal
subjects (60 normal, 52 suspected glaucoma), and 178 glaucoma patients (115 early stage, 63
moderate to severe glaucoma). The demographic characteristics are summarised in Table 2.
Although the visual acuity (logarithm of the minimum angle of resolution; logMAR) and
spherical equivalence were significantly different (P = 0.027, P = 0.006 respectively) between
PLOS ONE | https://doi.org/10.1371/journal.pone.0234902 July 6, 2020
7 / 19
PLOS ONE Table 2. Demographic characteristics of the test group.
Normal (n = 112)
Glaucoma (n = 178)
Age (years)
Female / male (number)
Visual acuity (logMAR)
Spherical equivalence (dioptre)
Intraocular pressure (mm Hg)
Axial length (mm)
Central corneal thickness (μm)
Visual field test
• Mean deviation (dB)
• Pattern standard deviation (dB)
• Visual field index (%)
Optical coherence tomography
• Average mGCIPL thickness (μm)
• Average pRNFL thickness (μm)
52.5 ± 15.4
56 / 56
0.066 ± 0.112
–1.44 ± 2.84
15.6 ± 3.9
24.34 ± 1.62
548.8 ± 37.8
–1.62 ± 2.10
1.98 ± 1.11
97.8 ± 3.2
79.3 ± 6.9
91.0 ± 10.2
51.7 ± 14.0
85 / 93
0.093 ± 0.121
–2.36 ± 3.13
15.6 ± 4.1
24.70 ± 1.65
548.9 ± 32.6
–6.29 ± 6.18
6.12 ± 4.10
84.2 ± 18.8
69.2 ± 9.0
72.5 ± 12.4
Values are presented as mean ± standard deviation.
mGCIPL: macular ganglion cell-internal plexiform layer, pRNFL: peripapillary retinal nerve fibre layer.
a Mann-Whitney U test.
b χ2 test.
https://doi.org/10.1371/journal.pone.0234902.t002
Visual field prediction from OCT
P value
0.537a
0.780b
0.027a
0.006a
0.744a
0.058a
0.917a
< 0.001a
< 0.001a
< 0.001a
< 0.001a
< 0.001a
the normal subjects and glaucoma patients, age, gender, intraocular pressure, axial length, and
CCT were not. All visual field parameters were also significantly different between these two
groups, including average visual field MD which was -1.62 dB for normal subjects and -6.29
dB for glaucoma patients. The average OCT parameters were also significantly different with
average mGCIPL thickness measuring 79.3 μm and 69.2 μm, and average pRNFL thickness
measuring 91.0 μm and 72.5 μm, respectively.
Global and regional visual field prediction error and representative examples of prediction
are shown in Table 3 and Fig 3. Globally (the entire visual field area), the RMSE was
4.79 ± 2.56 dB including all patients and, 3.27 dB and 5.27 dB in normal subjects and glaucoma
patients, respectively. The prediction error in normal subjects was always significantly lower
than that in glaucoma patients (all P < 0.001), regardless of regions. By sector, the prediction
error of the superior visual field region was generally lower than that of the corresponding
inferior visual field region. The nasal region showed the lowest prediction error (3.50 dB) fol-
lowed by the superotemporal (3.93 dB), temporal (4.08 dB), superonasal (4.55 dB), inferotem-
poral (4.66 dB), and inferonasal (5.23 dB) regions. In comparing the OCT scan areas of all
subjects, the prediction error of the macular region (4.40 dB) was significantly higher than that
of the peripheral region (4.29 dB; P = 0.031, Mann-Whitney U test). However, in normal sub-
jects, the prediction error of the macular region (2.45 dB) was significantly lower (P < 0.001,
Mann-Whitney U test) than that of the peripheral region (3.11 dB), whereas in glaucomatous
patients, the prediction error of macular region (5.62 dB) was higher than that of the periph-
eral region (5.03 dB), though not significantly (P = 0.741, Mann-Whitney U test).
The representative example of the class activation map (CAM) is shown in Fig 4. In this
map, the red colour denoted the area where the CNN was highly activated and produced a
high sensitivity value for the visual field test point, whereas the blue colour denoted the oppo-
site. The actual visual field test result (Fig 4A) showed low sensitivity in the superonasal area
and the CAM images at the corresponding location in the collection (Fig 4C) showed low acti-
vation in the inferotemporal sectors of the ONH OCT scan image. In contrast, the inferonasal
PLOS ONE | https://doi.org/10.1371/journal.pone.0234902 July 6, 2020
8 / 19
PLOS ONE Table 3. Global and regional root mean square error of visual field prediction.
Visual field prediction from OCT
Subject group
Normal
3.27 ± 1.50
2.41 ± 1.63
2.36 ± 1.36
3.07 ± 1.53
3.30 ± 1.93
2.69 ± 1.72
4.41 ±2.39
2.45 ± 1.20
3.11 ± 1.47
Glaucoma
5.75 ± 2.63
4.88 ± 4.04
5.16 ± 3.74
5.66 ± 3.46
5.33 ± 3.16
4.01 ± 2.86
5.74 ± 3.10
5.62 ± 3.54
5.03 ± 2.27
P valuea
<0.001
<0.001
<0.001
<0.001
<0.001
<0.001
<0.001
<0.001
<0.001
Global
All subjects
4.79 ± 2.56
Region by Garway-Heath sectorisation
Superotemporal
Temporal
Inferotemporal
Superonasal
Nasal
Inferonasal
Region by OCT scan area
Macularb
Peripheralc
3.93 ± 3.53
4.08 ± 3.34
4.66 ± 3.14
4.55 ± 2.92
3.50 ± 2.56
5.23 ± 2.92
4.40 ± 3.26
4.29 ± 2.20
Values are presented as mean ± standard deviation.
OCT: optical coherence tomography.
a Mann-Whitney U test between normal and glaucoma group.
b Visual field test points inside the macular OCT scan area.
c Visual field test points outside the macular OCT scan area.
https://doi.org/10.1371/journal.pone.0234902.t003
area showed high values for visual field sensitivity and the CAM images at the corresponding
location showed high activation (seen as the red colour) in the superotemporal sectors of the
ONH OCT scan image.
In each CAM image, red indicates strongly activated points yielding high threshold values;
blue (or no color) indicates the opposite. In this example, the visual field damage is principally
superonasal; the inferonasal region is relatively intact (i.e., exhibits high threshold values). The
CAM images in (C) that are numbered 27~30, 35~38, 43~46, and 49~52 are intensely red
(high visual field threshold values). Note that these activated areas in the CAM images (also
intensely red) exactly match the orange and dark grey regions in (D). In contrast, the CAM
images numbered 5, 6, 11, 12, and 19~22 are not colored (and thus not activated) and generate
low threshold visual field values. These areas match the blue and light grey regions in (D).
OCT: optical coherence tomography, ONH: optic nerve head.
Correlation analysis (Spearman’s rho) was performed to determine factors affecting visual
field prediction (Table 4). Age, sex, spherical equivalence, CCT, axial length, and macular
OCT signal strength all showed no significant correlation with visual field prediction. Visual
acuity (logMAR) positively correlated (r = 0.157, P = 0.007), and visual field MD (r = –0.543,
P < 0.001), average mGCIPL thickness (r = –0.553, P < 0.001), ONH OCT signal strength (r =
–0.126, P = 0.032), and average pRNFL thickness (r = –0.597, P < 0.001) negatively correlated
with visual field prediction.
Multiple linear regression analysis was performed to investigate the relative influence of
possible factors affecting visual field prediction (Table 5). The model was constructed using
the ENTER method and the RMSE as the outcome variable. Age, visual acuity (logMAR),
spherical equivalence, CCT, axial length, visual field MD, macular OCT signal strength, aver-
age mGCIPL thickness, ONH OCT signal strength, and average pRNFL thickness were used as
the input variables. The final model had R2 = 0.463 and P < 0.001. No multicollinearity was
found between variables (all variance inflation factors, VIFs � 3.051). Three out of 10 input
variables were significantly correlated with prediction error. The visual field MD was the most
PLOS ONE | https://doi.org/10.1371/journal.pone.0234902 July 6, 2020
9 / 19
PLOS ONE Visual field prediction from OCT
PLOS ONE | https://doi.org/10.1371/journal.pone.0234902 July 6, 2020
10 / 19
PLOS ONE Visual field prediction from OCT
Fig 3. Representative cases of visual field prediction. (A) The combined OCT images, which were input into the
deep learning architecture, are shown on the left column. The actual threshold values of visual field exams are shown
in the (B) middle panel and the threshold values predicted by Inception V3 based deep learning architecture are shown
on the (C) right panel. The colour reference for the threshold values are shown at the bottom. Despite the artificial
intelligence having never seen the actual visual field, the predicted visual field looked very similar to the actual visual
field exam.
https://doi.org/10.1371/journal.pone.0234902.g003
influential variable (β = –0.433, P < 0.001) followed by average pRNFL thickness (β = –0.252,
P = 0.002) and average mGCIPL thickness (β = –0.170, P = 0.028).
Fig 5 shows the relationship between prediction error ratio and the visual field MD using a
scatter plot. The prediction error was defined as the ratio of average prediction error inside the
macular OCT scan area (Fig 2) divided by the prediction error in the peripheral scan area (i.e.
outside the macular OCT scan area). In linear regression analysis, the prediction error ratio
was negatively associated with the visual field MD and its slope was –0.020 (P < 0.001). In
other words, as the MD decreased, the macular prediction error became greater than the
peripheral prediction error.
Discussion
The main objective of this study was to develop a deep learning architecture to predict Hum-
phrey visual field 24–2 threshold values from macular and ONH OCT imaging. We evaluated
the performance of visual field prediction globally and regionally, and tried to identify the fac-
tors that affected prediction. Globally, the RMSE of the deep learning algorithm was
4.70 ± 2.56 dB for all test subjects. In glaucoma subjects, the prediction error was significantly
higher than that of normal subjects (5.75 dB versus 3.27 dB, respectively). Visual field MD was
the most influential factor for prediction, followed by average pRNFL and mGCIPL thickness.
In the ophthalmology field, this study is the first to use OCT and a deep learning algorithm to
predict Humphrey visual field 24–2.
Retinal ganglion cells and their nerve fibre layer are closely related to the glaucomatous
visual field defect. This structure-function relationship has been investigated extensively. Woll-
stein et al. [20] studied the relationship between OCT-measured macular retinal thickness,
pRNFL thickness, and visual field. In their report, macular retinal thickness was able to detect
glaucomatous visual field damage and also correlated with pRNFL thickness. Sato et al. [21]
reported that GCIPL thickness, measured by Cirrus HD-OCT, was significantly correlated
with the central visual field. Similarly, Raza et al. [22] showed that GCIPL thickness was well
correlated with visual field loss within 7.2˚ of the fovea. Kim et al. [36] noted that both macular
ganglion cell complex (GCC) thickness and pRNFL thickness showed similar diagnostic per-
formance in detecting glaucoma. In our previous study [25], we found that the macular OCT
scan area mostly overlapped with visual field 10–2 test points and closely correlated with each
other. Wu et al. [37] reported that localised pRNFL thinning measured by SD-OCT was well
correlated with localised glaucomatous visual field defects. Garway-Heath et al. [23] mapped
visual field locations to the ONH sectors derived by overlapping visual field test points on
RNFL photographs. Gardiner et al. [24] also showed a topographical map between visual field
locations and ONH sectors. In those previous reports, mGCIPL thickness was more related to
the central visual field whereas pRNFL thickness was related to gross glaucomatous change.
Therefore, we could theorise that these two OCT exams may have complementary roles in pre-
dicting visual field.
Although attempts to use machine learning algorithms for structure-function relationships
in glaucoma are not novel [38–40], there have been few studies that predict visual field using
OCT imaging. In a recent study similar to ours, Christopher et al. [41] used a deep learning
PLOS ONE | https://doi.org/10.1371/journal.pone.0234902 July 6, 2020
11 / 19
PLOS ONE Visual field prediction from OCT
Fig 4. Representative example of Class Activation Map (CAM). The figure shows (A) the actual threshold values and (B) the predicted threshold values of the
visual field examination. (C) Fifty-two CAMs were placed at individual visual field test points. Each CAM image is numbered at the top left. (D) Structure-
function mapping between the combined OCT image (left) and the visual field (right). The macular scan in the combined OCT image corresponds to the dashed
rectangle in the visual field. Color-coded Garway-Heath sectors are superimposed on the ONH scan of the combined OCT image and the corresponding visual
field regions are similarly colored. The numbers in the visual field image are the same as those in the CAM images of (C).
https://doi.org/10.1371/journal.pone.0234902.g004
PLOS ONE | https://doi.org/10.1371/journal.pone.0234902 July 6, 2020
12 / 19
PLOS ONE Visual field prediction from OCT
Table 4. Correlation coefficients and simple linear regression analyses between visual field prediction error and various factors.
Correlation coefficients
Spearman’s rho
Age
Sex
Visual acuity (logMAR)
Spherical equivalence
Central corneal thickness
Axial length
Visual field MD
Macular OCT signal strength
Average mGCIPL thickness
ONH OCT signal strength
Average pRNFL thickness
–0.003
–0.018
0.157
–0.018
–0.051
0.071
–0.543
–0.039
–0.553
–0.126
–0.597
P value
0.957
0.756
0.007
0.765
0.404
0.257
<0.001
0.510
<0.001
0.032
<0.001
Simple linear regression analysis
Slope
0.011
–0.108
2.803
0.015
–0.003
0.015
–0.296
–0.103
–0.145
–0.303
–0.102
Intercept
4.230
4.842
4.556
4.818
6.705
4.399
3.463
5.625
15.415
7.280
12.896
R2
0.004
<0.001
0.017
<0.001
0.002
<0.001
0.403
0.002
0.296
0.019
0.338
P value
0.301
0.722
0.028
0.766
0.452
0.878
<0.001
0.410
<0.001
0.020
<0.001
MD: mean deviation, mGCIPL: macular ganglion cell-internal plexiform layer, OCT: optical coherence tomography, ONH: optic nerve head, pRNFL: peripapillary
retinal nerve fibre layer.
https://doi.org/10.1371/journal.pone.0234902.t004
method to predict glaucomatous visual fields from OCT images. The deep learning architec-
ture employed to predict visual field global indices was ResNet and the inputs were Spectralis
SD-OCT ONH images. Various image types (RNFL thickness maps, RNFL en-face images,
and confocal scanning laser ophthalmoscopic images) were input and the predictions using
each type were compared. Unlike our deep learning method, which predicts the entire visual
field from both macular and ONH images, the method of Christopher et al. uses only ONH
image as input and predicts visual field global indices including the mean deviation (MD), the
pattern standard deviation (PSD), and the mean sectoral pattern deviation. The best mean
absolute errors, between the real and predicted values, were 2.5 dB (MD) and 1.5 dB (PSD).
Zhu et al. [42] introduced a method akin to the neural network with a radial basis function
customised under a Bayesian framework (BRBF), to predict visual field from pRNFL thickness.
In their report, the mean absolute error of the BRBF was 2.9 dB, which was better than the
Table 5. Multiple linear regression analyses between visual field prediction error and various factors.
Age
Visual acuity (logMAR)
Spherical equivalence
Central corneal thickness
Axial length
Visual field MD
Macular OCT signal strength
Average mGCIPL thickness
ONH OCT signal strength
Average pRNFL thickness
Adjusted β
–0.081
–0.059
0.130
–0.007
0.020
–0.433
0.097
–0.170
0.040
–0.252
P value
0.216
0.253
0.081
0.889
0.772
0.000
0.127
0.028
0.540
0.002
VIF
1.983
1.204
2.558
1.065
2.159
1.846
1.849
2.731
1.914
3.051
Final model: Outcome = root mean square error (RMSE) of prediction. Adjusted R2 = 0.463, P < 0.001, ENTER
method used. No multicollinearity was found between variables (all VIFs � 3.051).
MD: mean deviation, mGCIPL: macular ganglion cell-internal plexiform layer, OCT: optical coherence tomography,
ONH: optic nerve head, pRNFL: peripapillary retinal nerve fibre layer, VIF: variance inflation factors.
https://doi.org/10.1371/journal.pone.0234902.t005
PLOS ONE | https://doi.org/10.1371/journal.pone.0234902 July 6, 2020
13 / 19
PLOS ONE Visual field prediction from OCT
Fig 5. Scatter plot of the prediction error ratio (macular/peripheral) versus the visual field Mean Deviation (MD).
The slope was –0.020 (P < 0.001) which suggested that as the MD decreased, the macular prediction error became
higher than the peripheral prediction error. In other words, as glaucoma progressed, the peripheral prediction became
more accurate than the macular prediction.
https://doi.org/10.1371/journal.pone.0234902.g005
classical linear regression model (4.9 dB). Though their result showed better performance than
that noted in our study (4.70 dB), absolute comparison was not possible as the performance
metric used in their study was mean absolute error and their test dataset (Blue Mountains Eye
Study [BMES] data) largely consisted of healthy subjects (230 healthy subjects and 76 glau-
coma patients). Considering that the prediction error was worse in glaucoma patients than
normal subjects, the large proportion of healthy subjects in their study likely reduced their pre-
diction error compared to that determined in this study.
Another study using machine learning to predict visual field threshold values was con-
ducted by Guo et al. [26], who predicted visual field 24–2 from wide field composite OCT.
They used a 9-field per eye protocol which fixated the patient’s eye on a 3 × 3 grid spot pattern
to obtain a total of nine OCT images. These images were stitched together to generate a single
wide field composite OCT image. They constructed four predefined topological structure-
function maps and applied a support vector machine (SVM) algorithm to these maps to pre-
dict visual field. Among the four predefined maps, the best map showed an RMSE of 5.42 dB.
The performance of this method was excellent, but depended largely on how the structure-
function map was defined. In the different maps, the prediction error increased up to 7.24 dB.
However, our deep learning method creates a structure-function map by itself during the
training process. In Fig 4C, the CAM showed how Inception V3 based deep learning architec-
ture constructed this map and noted that it was similar to previous studies such as the Garway-
Heath map. Moreover, the deep learning algorithm not only considers a specific mapping spot
but also broad neighbouring areas as well. This likely made the prediction more accurate as a
predefined mapping spot could contain errors; however, by considering a wider area, this
error can be overcomed.
PLOS ONE | https://doi.org/10.1371/journal.pone.0234902 July 6, 2020
14 / 19
PLOS ONE Visual field prediction from OCT
Among the regions of the Garway-Heath sectorisation, we observed that the superior sec-
tors of the ONH had fewer prediction errors than the corresponding inferior sectors, suggest-
ing that the superior retina is better correlated with functional tests than the inferior retina.
This is corroborated by similar findings reported by Guo et al. [26], who found that the corre-
lation between structure and function was higher in the superior than the inferior retina. It
was suggested that this was due to the superior retina, which is responsible for the inferior
visual field, being more important for survival in nature and may be an evolutionary conse-
quence. However, we propose another possible reason for this observation. From previous
studies, glaucomatous damage is known to occur sequentially in sectors. It begins in the infer-
otemporal ONH region, and then progresses to the superotemporal sectors. [43] In our study,
as glaucoma progressed, the overall prediction error was increased. Since the inferotemporal
ONH region is the first to be damaged, its prediction error could be higher than that of the
superior region.
In regression analysis, we found that both mGCIPL and pRNFL thickness are significantly
correlated to the visual field prediction error with the pRNFL thickness being slightly more
influential than mGCIPL thickness (adjusted β = –0.170 and –0.252 for mGCIPL and pRNFL,
respectively). We suggest that this is because the mGCIPL only provides information regarding
the macular area whereas the pRNFL offers a more generalised view across all areas of the ret-
ina. However, the regional prediction errors, and macular and peripheral OCT scan areas were
different between normal and glaucoma subjects. In normal subjects, the macular OCT scan
area showed a lower prediction error than the peripheral OCT scan area, whereas in glaucoma
patients, the opposite was noted. This result is consistent with previous studies. Wollstein et al.
[20] reported that the pRNFL thickness was more sensitive to glaucomatous damage. Kim
et al. [36] reported that mean GCC thickness, instead of pRNFL thickness, was a better diag-
nostic indicator of early glaucoma cases as well as a non-glaucomatous condition. In normal
or early glaucoma patients, information provided by the mGCIPL is relatively more important
than that provided by the pRNFL which is likely to have made visual field prediction using the
macular OCT scan area more accurate. As glaucoma progressed, the information provided by
the pRNFL thickness became more influential and reversed the prediction accuracy between
the macular and peripheral OCT scan areas. This may provide an explanation for the negative
correlation of the prediction error ratio (macular versus peripheral) with visual field MD in
Fig 5.
Regression analysis also revealed that age, visual acuity, spherical equivalence, axial length,
and signal strength of OCT were not correlated with prediction error. In several previous
reports, the reproducibility of OCT imaging has been considered to be generally reliable [10–
15]. Axial length has been shown to affect the RNFL thickness measurement [44]; however,
another study has noted that both macular GCC and pRNFL thickness measurements show
good diagnostic performance in individuals with high myopia [45]. In our study, extreme
cases including spherical equivalence >6.0 dioptres and axial length >26.0 mm were excluded,
and only eyes with relatively good visual acuity, no media opacity, and no diseases other than
glaucoma were included. With these inclusion criteria, the reproducibility of OCT imaging in
predicting visual field appears to be strong and negligibly affected by the factors noted above.
Even though the quality of the OCT scan is inevitably associated with the accuracy of visual
field prediction, there was no significant correlation between OCT image quality and predic-
tion error.
This study had a few limitations. First, only those patients with glaucoma were included to
predict visual field. If patients with diseases other than glaucoma, where the visual field defect
was not altitudinal in nature, such as temporal hemianopsia, were included, the prediction
accuracy may be different than that reported in our study. Second, the training and test
PLOS ONE | https://doi.org/10.1371/journal.pone.0234902 July 6, 2020
15 / 19
PLOS ONE Visual field prediction from OCT
datasets were comprised of a primarily Korean population and the possibility of performance
being different in populations comprising of other ethnicities should be considered. Girkin
et al. [46] evaluated the ONH, RNFL, and macular parameters yielded by SD-OCT in terms of
age and race; all parameters varied by race. Thus, researchers must be aware that visual field
predictions derived from SD-OCT will also vary by race.
In conclusion, the deep learning method effectively predicted the visual field 24–2 using the
combined OCT image (mGCIPL and pRNFL thickness map) with a prediction error of 4.70
dB. The accuracy of the visual field prediction was not influenced by factors such as age, visual
acuity, spherical equivalence, axial length, and OCT signal strength. This may help clinicians
perform visual field testing, especially of patients who are unable to undergo real visual field
examinations (young children, dementia patients, and mentally retarded patients).
Supporting information
S1 File. All test dataset.
(XLSX)
Author Contributions
Conceptualization: Keunheung Park.
Data curation: Keunheung Park, Jinmi Kim.
Formal analysis: Keunheung Park.
Funding acquisition: Keunheung Park.
Investigation: Keunheung Park.
Methodology: Keunheung Park, Jinmi Kim.
Project administration: Jiwoong Lee.
Resources: Keunheung Park.
Software: Keunheung Park.
Supervision: Jiwoong Lee.
Validation: Keunheung Park.
Visualization: Keunheung Park.
Writing – original draft: Keunheung Park.
Writing – review & editing: Jiwoong Lee.
References
1. Resnikoff S, Pascolini D, Etya’ale D, Kocur I, Pararajasegaram R, Pokharel GP, et al. Global data on
visual impairment in the year 2002. Bull World Health Organ. 2004; 82: 844–851. PMID: 15640920
2.
Tham Y-C, Li X, Wong TY, Quigley HA, Aung T, Cheng C-Y. Global prevalence of glaucoma and projec-
tions of glaucoma burden through 2040: a systematic review and meta-analysis. Ophthalmology. 2014;
121: 2081–2090. https://doi.org/10.1016/j.ophtha.2014.05.013 PMID: 24974815
3. Gutierrez P, Wilson MR, Johnson C, Gordon M, Cioffi GA, Ritch R, et al. Influence of glaucomatous
visual field loss on health-related quality of life. Arch Ophthalmol. 1997; 115: 777–784. https://doi.org/
10.1001/archopht.1997.01100150779014 PMID: 9194730
4. Nelson P, Aspinall P, Papasouliotis O, Worton B, O’Brien C. Quality of life in glaucoma and its relation-
ship with visual function. J Glaucoma. 2003; 12: 139–150. https://doi.org/10.1097/00061198-
200304000-00009 PMID: 12671469
PLOS ONE | https://doi.org/10.1371/journal.pone.0234902 July 6, 2020
16 / 19
PLOS ONE Visual field prediction from OCT
5.
6.
Fogagnolo P, Sangermani C, Oddone F, Frezzotti P, Iester M, Figus M, et al. Long-term perimetric fluc-
tuation in patients with different stages of glaucoma. Br J Ophthalmol. 2010; bjo–2010.
Langerhorst CT, Van den Berg T, Van Spronsen R, Greve EL. Results of a fluctuation analysis and
defect volume program for automated static threshold perimetry with the scoperimeter. Sixth Interna-
tional Visual Field Symposium. Springer; 1985. pp. 1–6.
7. Brenton RS, Argus WA. Fluctuations on the Humphrey and Octopus perimeters. Invest Ophthalmol Vis
Sci. 1987; 28: 767–771. PMID: 3570687
8. Wild JM, Dengler-Harles M, Searle AET, O’Neill EC, Crews SJ. The influence of the learning effect on
automated perimetry in patients with suspected glaucoma. Acta Ophthalmol (Copenh). 1989; 67: 537–
545. https://doi.org/10.1111/j.1755-3768.1989.tb04105.x PMID: 2589053
9. Marra G, Flammer J. The learning and fatigue effect in automated perimetry. Graefes Arch Clin Exp
Ophthalmol. 1991; 229: 501–504. https://doi.org/10.1007/BF00203309 PMID: 1765286
10. Carpineto P, Nubile M, Agnifili L, Toto L, Aharrh-Gnama A, Mastropasqua R, et al. Reproducibility and
Repeatability of CirrusTM HD-OCT Peripapillary Retinal Nerve Fibre Layer Thickness Measurements in
Young Normal Subjects. Ophthalmologica. 2012; 227: 139–145. https://doi.org/10.1159/000334967
PMID: 22261709
11. Gu¨rses-O¨ zden R, Teng C, Vessani R, Zafar S, Liebmann JM, Ritch R. Macular and Retinal Nerve Fiber
Layer Thickness Measurement Reproducibility Using Optical Coherence Tomography (OCT-3). J Glau-
coma. 2004; 13: 238. https://doi.org/10.1097/00061198-200406000-00012 PMID: 15118470
12. Paunescu LA, Schuman JS, Price LL, Stark PC, Beaton S, Ishikawa H, et al. Reproducibility of Nerve
Fiber Thickness, Macular Thickness, and Optic Nerve Head Measurements Using StratusOCT. Invest
Ophthalmol Vis Sci. 2004; 45: 1716–1724. https://doi.org/10.1167/iovs.03-0514 PMID: 15161831
13. Hong S, Kim CY, Lee WS, Seong GJ. Reproducibility of peripapillary retinal nerve fiber layer thickness
with spectral domain cirrus high-definition optical coherence tomography in normal eyes. Jpn J Ophthal-
mol. 2010; 54: 43–47. https://doi.org/10.1007/s10384-009-0762-8 PMID: 20151275
14. Garcia-Martin E, Pinilla I, Idoipe M, Fuertes I, Pueyo V. Intra and interoperator reproducibility of retinal
nerve fibre and macular thickness measurements using Cirrus Fourier-domain OCT. Acta Ophthalmol
(Copenh). 2011; 89: e23–29. https://doi.org/10.1111/j.1755-3768.2010.02045.x PMID: 21106044
15. Budenz DL, Fredette M-J, Feuer WJ, Anderson DR. Reproducibility of peripapillary retinal nerve fiber
thickness measurements with stratus OCT in glaucomatous eyes. Ophthalmology. 2008; 115: 661–
666. https://doi.org/10.1016/j.ophtha.2007.05.035 PMID: 17706287
16. Wollstein G, Kagemann L, Bilonick RA, Ishikawa H, Folio LS, Gabriele ML, et al. Retinal nerve fibre
layer and visual function loss in glaucoma: the tipping point. Br J Ophthalmol. 2012; 96: 47–52. https://
doi.org/10.1136/bjo.2010.196907 PMID: 21478200
17. Alasil T, Wang K, Yu F, Field MG, Lee H, Baniasadi N, et al. Correlation of retinal nerve fiber layer thick-
ness and visual fields in glaucoma: a broken stick model. Am J Ophthalmol. 2014; 157: 953–959.
https://doi.org/10.1016/j.ajo.2014.01.014 PMID: 24487047
18. Carpineto P, Ciancaglini M, Zuppardi E, Falconio G, Doronzo E, Mastropasqua L. Reliability of nerve
fiber layer thickness measurements using optical coherence tomography in normal and glaucomatous
eyes. Ophthalmology. 2003; 110: 190–195. https://doi.org/10.1016/s0161-6420(02)01296-4 PMID:
12511365
19.
Lee J-W, Kim E-A, Otarola F, Morales E, Yu F, Afifi AA, et al. The Fast Component of Visual Field
Decay Rate Correlates With Disc Rim Area Change Throughout the Entire Range of Glaucomatous
Damage. Invest Ophthalmol Vis Sci. 2015; 56: 5997–6006. https://doi.org/10.1167/iovs.15-17006
PMID: 26393466
20. Wollstein G, Schuman JS, Price LL, Aydin A, Beaton SA, Stark PC, et al. Optical coherence tomogra-
phy (oct) macular and peripapillary retinal nerve fiber layer measurements and automated visual fields.
Am J Ophthalmol. 2004; 138: 218–225. https://doi.org/10.1016/j.ajo.2004.03.019 PMID: 15289130
21. Sato S, Hirooka K, Baba T, Tenkumo K, Nitta E, Shiraga F. Correlation Between the Ganglion Cell-
Inner Plexiform Layer Thickness Measured With Cirrus HD-OCT and Macular Visual Field Sensitivity
Measured With Microperimetry. Invest Ophthalmol Vis Sci. 2013; 54: 3046–3051. https://doi.org/10.
1167/iovs.12-11173 PMID: 23580483
22. Raza AS, Cho J, Moraes CGV de, Wang M, Zhang X, Kardon RH, et al. Retinal Ganglion Cell Layer
Thickness and Local Visual Field Sensitivity in Glaucoma. Arch Ophthalmol. 2011; 129: 1529–1536.
https://doi.org/10.1001/archophthalmol.2011.352 PMID: 22159673
23. Garway-Heath DF, Poinoosawmy D, Fitzke FW, Hitchings RA. Mapping the visual field to the optic disc
in normal tension glaucoma eyes. Ophthalmology. 2000; 107: 1809–1815. https://doi.org/10.1016/
s0161-6420(00)00284-0 PMID: 11013178
PLOS ONE | https://doi.org/10.1371/journal.pone.0234902 July 6, 2020
17 / 19
PLOS ONE Visual field prediction from OCT
24. Gardiner SK, Johnson CA, Cioffi GA. Evaluation of the structure-function relationship in glaucoma.
Invest Ophthalmol Vis Sci. 2005; 46: 3712–3717. https://doi.org/10.1167/iovs.05-0266 PMID:
16186353
25.
Lee J-W, Morales E, Sharifipour F, Amini N, Yu F, Afifi AA, et al. The relationship between central visual
field sensitivity and macular ganglion cell/inner plexiform layer thickness in glaucoma. Br J Ophthalmol.
2017; 101: 1052–1058. https://doi.org/10.1136/bjophthalmol-2016-309208 PMID: 28077369
26. Guo Z, Kwon YH, Lee K, Wang K, Wahle A, Alward WLM, et al. Optical Coherence Tomography Analy-
sis Based Prediction of Humphrey 24–2 Visual Field Thresholds in Patients With Glaucoma. Invest
Ophthalmol Vis Sci. 2017; 58: 3975–3985. https://doi.org/10.1167/iovs.17-21832 PMID: 28796875
27. Bogunovic H, Kwon YH, Rashid A, Lee K, Critser DB, Garvin MK, et al. Relationships of retinal structure
and humphrey 24–2 visual field thresholds in patients with glaucoma. Invest Ophthalmol Vis Sci. 2014;
56: 259–71. https://doi.org/10.1167/iovs.14-15885 PMID: 25491294
28.
Zhang X, Bregman CJ, Raza AS, De Moraes G, Hood DC. Deriving visual field loss based upon OCT of
inner retinal thicknesses of the macula. Biomed Opt Express. 2011; 2: 1734–42. https://doi.org/10.
1364/BOE.2.001734 PMID: 21698033
29. Peng Y, Dharssi S, Chen Q, Keenan TD, Agro´n E, Wong WT, et al. DeepSeeNet: A Deep Learning
Model for Automated Classification of Patient-based Age-related Macular Degeneration Severity from
Color Fundus Photographs. Ophthalmology. 2019; 126: 565–575. https://doi.org/10.1016/j.ophtha.
2018.11.015 PMID: 30471319
30. Burlina P, Joshi N, Pacheco KD, Freund DE, Kong J, Bressler NM. Utility of Deep Learning Methods for
Referability Classification of Age-Related Macular Degeneration. JAMA Ophthalmol. 2018; 136: 1305–
1307. https://doi.org/10.1001/jamaophthalmol.2018.3799 PMID: 30193354
31. Xu Y, Wang Y, Liu B, Tang L, Lv L, Ke X, et al. The diagnostic accuracy of an intelligent and automated
fundus disease image assessment system with lesion quantitative function (SmartEye) in diabetic
patients. BMC Ophthalmol. 2019; 19: 184. https://doi.org/10.1186/s12886-019-1196-9 PMID:
31412800
32.
Foster PJ, Buhrmann R, Quigley HA, Johnson GJ. The definition and classification of glaucoma in prev-
alence surveys. Br J Ophthalmol. 2002; 86: 238–242. https://doi.org/10.1136/bjo.86.2.238 PMID:
11815354
33. Chollet F. Keras. GitHub; 2015. Available: https://github.com/fchollet/keras
34. Szegedy C, Vanhoucke V, Ioffe S, Shlens J, Wojna Z. Rethinking the inception architecture for com-
puter vision. Proceedings of the IEEE conference on computer vision and pattern recognition. 2016. pp.
2818–2826.
35. Dubitzky W, Granzow M, Berrar DP. Fundamentals of data mining in genomics and proteomics.
Springer Science & Business Media; 2007.
36. Kim NR, Lee ES, Seong GJ, Kim JH, An HG, Kim CY. Structure–Function Relationship and Diagnostic
Value of Macular Ganglion Cell Complex Measurement Using Fourier-Domain OCT in Glaucoma.
Invest Ophthalmol Vis Sci. 2010; 51: 4646–4651. https://doi.org/10.1167/iovs.09-5053 PMID:
20435603
37. Wu H, de Boer JF, Chen L, Chen TC. Correlation of localized glaucomatous visual field defects and
spectral domain optical coherence tomography retinal nerve fiber layer thinning using a modified struc-
ture–function map for OCT. Eye. 2015; 29: 525–533. https://doi.org/10.1038/eye.2014.317 PMID:
25633881
38. Bowd C, Chan K, Zangwill LM, Goldbaum MH, Lee T-W, Sejnowski TJ, et al. Comparing neural net-
works and linear discriminant functions for glaucoma detection using confocal scanning laser ophthal-
moscopy of the optic disc. Invest Ophthalmol Vis Sci. 2002; 43: 3444–3454. PMID: 12407155
39. Brigatti L, Hoffman D, Caprioli J. Neural networks to identify glaucoma with structural and functional
measurements. Am J Ophthalmol. 1996; 121: 511–521. https://doi.org/10.1016/s0002-9394(14)75425-
x PMID: 8610794
40. Uchida H, Brigatti L, Caprioli J. Detection of structural damage from glaucoma with confocal laser image
analysis. Invest Ophthalmol Vis Sci. 1996; 37: 2393–2401. PMID: 8933756
41. Christopher M, Bowd C, Belghith A, Goldbaum MH, Weinreb RN, Fazio MA, et al. Deep Learning
Approaches Predict Glaucomatous Visual Field Damage from OCT Optic Nerve Head En Face Images
and Retinal Nerve Fiber Layer Thickness Maps. Ophthalmology. 2020; 127: 346–356. https://doi.org/
10.1016/j.ophtha.2019.09.036 PMID: 31718841
42.
Zhu H, Crabb DP, Schlottmann PG, Lemij HG, Reus NJ, Healey PR, et al. Predicting Visual Function
from the Measurements of Retinal Nerve Fiber Layer Structure. Invest Ophthalmol Vis Sci. 2010; 51:
5657–5666. https://doi.org/10.1167/iovs.10-5239 PMID: 20505207
PLOS ONE | https://doi.org/10.1371/journal.pone.0234902 July 6, 2020
18 / 19
PLOS ONE Visual field prediction from OCT
43.
Jonas JB, Ferna´ ndez MC, Stu¨ rmer J. Pattern of Glaucomatous Neuroretinal Rim Loss. Ophthalmology.
1993; 100: 63–68. https://doi.org/10.1016/s0161-6420(13)31694-7 PMID: 8433829
44. Savini G, Barboni P, Parisi V, Carbonelli M. The influence of axial length on retinal nerve fibre layer
thickness and optic-disc size measurements by spectral-domain OCT. Br J Ophthalmol. 2012; 96: 57–
61. https://doi.org/10.1136/bjo.2010.196782 PMID: 21349942
45. Kim NR, Lee ES, Seong GJ, Kang SY, Kim JH, Hong S, et al. Comparing the ganglion cell complex and
retinal nerve fibre layer measurements by Fourier domain OCT to detect glaucoma in high myopia. Br J
Ophthalmol Lond. 2011; 95: 1115. http://dx.doi.org.eproxy.pusan.ac.kr/10.1136/bjo.2010.182493
46. Girkin CA, McGwin G, Sinai MJ, Sekhar GC, Fingeret M, Wollstein G, et al. Variation in Optic Nerve and
Macular Structure with Age and Race with Spectral-Domain Optical Coherence Tomography. Ophthal-
mology. 2011; 118: 2403–2408. https://doi.org/10.1016/j.ophtha.2011.06.013 PMID: 21907415
PLOS ONE | https://doi.org/10.1371/journal.pone.0234902 July 6, 2020
19 / 19
PLOS ONE |
10.1371_journal.pone.0232176 | RESEARCH ARTICLE
Analysis of adult disease characteristics and
mortality on MIMIC-III
Zheng DaiID
1, Siru Liu2, Jinfa Wu1, Mengdie Li1, Jialin Liu3,4*, Ke Li1*
1 School of Life Science & Technology, University of Electronic Science & Technology of China, Chengdu,
China, 2 Department of Biomedical Informatics, University of Utah, Salt Lake City, Utah, United States of
America, 3 Department of Medical Informatics, West China Medical School, Sichuan University, Chengdu,
China, 4 Information Center, West China Hospital, Sichuan University, Chengdu, China
* colinlike@163.com (KL); DLJL8@163.com (JL)
Abstract
Purpose
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: Dai Z, Liu S, Wu J, Li M, Liu J, Li K (2020)
Analysis of adult disease characteristics and
mortality on MIMIC-III. PLoS ONE 15(4):
e0232176. https://doi.org/10.1371/journal.
pone.0232176
Editor: Omid Beiki, Karolinska Institutet, SWEDEN
To deeply analyze the basic information and disease information of adult patients in the
MIMIC-III (Medical Information Mart for Intensive Care III) database, and provide data refer-
ence for clinicians and researchers.
Materials and methods
Tableau2019.1.0 and Navicat12.0.29 were used for data analysis and extraction of disease
distribution of adult patients in the MIMIC-III database.
Received: September 16, 2019
Result
Accepted: April 8, 2020
Published: April 30, 2020
Copyright: © 2020 Dai et al. This is an open access
article distributed under the terms of the Creative
Commons Attribution License, which permits
unrestricted use, distribution, and reproduction in
any medium, provided the original author and
source are credited.
Data Availability Statement: The data underlying
the results presented in the study are available
from the Medical Information Mart for Intensive
Care (MIMIC-III), a large, single-center database
comprising information relating to patients
admitted to critical care units at a large tertiary care
hospital. Because the data contain sensitive patient
information, MIMIC-III has not made the data sets
publicly available. More information about MIMIC-
III can be found on their website (https://mimic.
mit.edu/about/mimic/). To access this data,
interested researchers must first complete the CITI
“Data or Specimens Only Research” course
(https://www.citiprogram.org/index.cfm?pageID=
A total of 38,163 adult patients were included in the MIMIC-III database. Only 38,156
patients with the first diagnosis were selected. Among them, 21,598 were males accounting
for 56.6% the median age was 66 years (Q1-Q3: 53–78), the median length of a hospital
stay was 7 days (Q1-Q3: 4–12), and the median length of an ICU stay was 2.1 days (Q1-Q3:
1.2–4.1). Septicemia was the disease with the highest mortality rate among patients and the
total mortality rate was 48.9%. The disease with the largest number of patients at the last
time was other forms of chronic ischemic heart disease.
Conclusion
By analyzing the patients’ basic information, the admission spectrum and the disease mor-
bidity and mortality can help more researchers understand the MIMIC-III database and facili-
tate further research.
1. Introduction
In order to understand the nature of the disease, many studies have been carried out to deter-
mine the pathogenesis, course, and complications of the disease [1–3]. MIMIC-III (Medical
PLOS ONE | https://doi.org/10.1371/journal.pone.0232176 April 30, 2020
1 / 12
PLOS ONE 154&icat=0&ac=0) and then submit an application
for credentialed access through PhysioNet (https://
physionet.org/content/mimiciii/). For more detailed
instructions on how to apply for data access,
please see the Supporting Information files of this
article. This study used the publicly available
Multiparameter Intelligent Monitoring in Intensive
Care (MIMIC) III database version 1.4.
Funding: This research did not receive any specific
grant from funding agencies in the public,
commercial, or not-for-profit sectors.
Competing interests: The authors declare that they
have no competing interests.
Analysis of adult disease characteristics and mortality on MIMIC-III
Information Mart for Intensive Care III) is a large, freely-available database comprising de-
identified health-related data associated with over forty thousand patients who stayed in criti-
cal care units of the Beth Israel Deaconess Medical Center between 2001 and 2012 [4]. The
data in the MIMIC-III database is more reliable and complete than the data in the MIMIC-II
database. Many researchers have conducted a lot of research on the MIMIC-III database. Kur-
niati A P [5] used process mining to study oncology. Mengling F’s [6] study was designed to
examine the association of transthoracic echocardiography with 28-day mortality specifically
in patients who suffered from sepsis. Parreco J P [7] used supervised machine learning to pre-
dict central line-associated bloodstream infection (CLABSI). Some scholars have used deep
learning [8,9] and visualization methods [10,11] to study the MIMIC-III database and get bet-
ter results. There are also many other researchers [12–15] using other methods to study the
MIMIC-III database.
Recent researches have leveraged machine learning techniques to predict the mortality rate
of various diseases. In order to solve clinical problems in a meaningful way, researchers need
to understand clinicians’ needs. Because clinicians lack knowledge of SQL programming and
EHR database programming [16–18], the clinical questions that they raised cannot be sup-
ported by data. Therefore, in order to promote the communication and cooperation between
clinicians and researchers [18], a detailed analysis of disease information based on the MIMI-
C-III database was conducted in this paper to provide data reference for researchers.
2. Materials and methods
2.1. Data source
This study used the publicly available Multiparameter Intelligent Monitoring in Intensive Care
(MIMIC) III database version 1.4 [4]. The MIMIC-III database is derived from the private
medical record of the ICU of the Boston Medical Center in the United States. All patients’ data
were anonymized prior to extraction and data analysis. The creation, maintenance, and use of
the MIMIC-III database were approved by the institutional review boards of the Massachusetts
Institute of Technology and Beth Israel Deaconess Medical Center [19]. The database contains
detailed information on patients’ vital signs, laboratory tests, and diagnosis codes for research
by global scholars. The MIMIC-III database contains 46,520 patients at the unique patient
level, of whom 38,163 are adult patients, 38,161 patients were last admitted to the ICU, and
38,156 patients have the first diagnosis.
2.2. Data selection
Because juvenile patients have different physical status from adult patients, this article only
selected adult patients to conduct data analysis. The research in this paper is based on unique
patient levels. All data were obtained at a unique patient level. Because a patient may have mul-
tiple admission records, we selected the last admission record. In the last admission record, we
selected the last ICU record for statistical analysis. 38,161 patients were last admitted to the
ICU. In this study, only the first diagnosis was selected as the object of data analysis, and the
disease classification was gradually refined according to the primary International Classifica-
tion of Diseases (ICD-9) codes [20]. Because the MIMIC-III database only includes the
patients’ admission time and discharge time, we used the discharge time minus the admission
time to obtain the patients’ hospitalization days. The date of death of each patient was recorded
in the MIMIC-III database. We used the date of death of each patient minus the date of dis-
charge to obtain how many days the patients died after being discharged. Then we selected
patients who died within 90 days for analysis.
PLOS ONE | https://doi.org/10.1371/journal.pone.0232176 April 30, 2020
2 / 12
PLOS ONE Analysis of adult disease characteristics and mortality on MIMIC-III
2.3. Research methods
The research in this article was divided into two parts the first part was the statistics and analy-
sis of basic information of all patients in MIMIC-III database, and the second part was the dis-
tribution of diseases. The first part describes the patients’ age, gender, ethnicity, admission
time and distribution of ICU types. We used SQL programming in Navicat12.0.29 to extract
the required indicator data from the database and then imported the extracted data into Tab-
leau2019.1.0. We used table query and calculated fields in Tableau2019.1.0 to filter unique data
and then performed the visual analysis to get the final result.
In the second part, we analyzed the basic information of ten diseases with the largest num-
ber of patients, and the mortality and the ICU distribution of the ten diseases. Firstly, accord-
ing to the ICD-9 coding standard, the number of patients and deaths of each disease were
counted, and the ten diseases with the largest number of patients and the ten diseases with the
highest mortality were screened. Then, we counted the basic information of patients for each
disease. According to the analysis methods mentioned in the first part of this section, the ICU
distribution of the ten diseases with the largest number of patients was made to reflect the
main diseases of each ICU.
3. Result
3.1. Basic information of all patients in MIMIC-III
We divided patients into three parts: death in hospital, 90-day death after discharge, and no
death within 90-day after discharge. We divided patients into three categories and gained their
information separately, so we can understand the disease characteristics of different popula-
tions more clearly.
We analyzed such as length of stay, age, gender. Age and length of stay are expressed in
median (inter-quartile range). Two of the 90-day-dead patients who were discharged from the
hospital had incorrect information that the time of death was greater than the time of admission.
Therefore, the basic information of the two patients was not counted in the statistics. It can be
seen from the Table 1 that the dead patients (include death in the hospital and 90-day death after
discharge) are older, 74 years (Q1-Q3: 60–83) and 76 years (Q1-Q3: 64–84), while the non-dead
patients are relatively young, 64 years (Q1-Q3: 51–76). The mortality of patients in the MIMI-
C-III database is 23.2%. Data analysis showed that older people were more likely to be admitted
to the ICU and had a higher risk of death. We should pay more attention to the physiological
indicators of patients above the median age of 74 years old to improve their survival rate.
We counted the distribution of patients in different ICUs (Fig 1). The ICU with the largest
number of patients is the Medical Intensive Care Unit (MICU), and the number of patients
who died in the hospital and the number of patients non-dead within 90-day after discharge in
Table 1. Patients’ basic information in MIMIC-III.
Type
Age, years, median
(Q1-Q3)
Male (%)
Days in hospital, median days
(Q1-Q3)
Days in ICU, median days
(Q1-Q3)
Number of patients
(%)
Death in the hospital
74 (60–83)
90-day of death after discharge
76 (64–84)
No death within 90-day after
discharge
Sum
64 (51–76)
66 (53–78)
https://doi.org/10.1371/journal.pone.0232176.t001
3046
(54.5%)
1690
(53.8%)
16862
(57.5%)
21598
(56.6%)
6 (2–13)
9 (5–16)
7 (4–11)
7 (4–12)
3.1 (1.3–7.5)
5697 (15.0%)
2.7 (1.5–5.3)
3140+2(8.2%)
2.0 (1.2–3.5)
29324(76.8%)
2.1 (1.2–4.1)
38163(100%)
PLOS ONE | https://doi.org/10.1371/journal.pone.0232176 April 30, 2020
3 / 12
PLOS ONE Analysis of adult disease characteristics and mortality on MIMIC-III
Fig 1. ICU distribution of all patients admitted to the ICU (Total number of patients is 38161).
https://doi.org/10.1371/journal.pone.0232176.g001
the MICU are also much higher than those in other ICUs. It shows that the internal disease is
the disease with the largest number of patients and the highest mortality rate. In subsequent
studies, we can study the physiological indicators of patients in MICU and use machine learn-
ing to predict the health of patients. Fig 1 also shows that the number of patients admitted to
Cardiac Surgery Intensive Care Unit (CSRU) was second only to MICU, but the number of
deaths in hospital and the number of 90-day deaths after discharge is the lowest compared to
other ICUs. In Fig 2, we counted the mortality of patients admitted to each type of ICU, and
we found that the highest mortality rate was 32.6% in MICU. Mortality exceeds 24% in both
CCU (24.3%) and SICU (24.4%).
We analyzed the ethnic mortality of patients (Fig 3). We find Asians have the highest mor-
tality rate, at 24.8%. Fig 4 depicts a pie chart of the ethnic distribution of all adult patients in
the MIMIC-III database. The ethnic distribution and proportion can be clearly seen. Because
the database does not completely record the race of patients, many patients only say their
nationality, so it is not possible to specify which race. Therefore, “UNKOWN” and “OTHER”
exist in the ethnic classification. In Fig 5 we have calculated the age distribution of patients in
MIMIC-III. On the whole, we can find that ICU patients are mainly elderly people. Specifi-
cally, we can see that the number of patients in the 71–80 age group is the largest, followed by
the 61–70 age group. Therefore, we think that the elderly at these ages have a high risk of dis-
eases, and it is particularly important to protect the elderly in their daily lives.
3.2. The distribution of diseases in MIMIC-III
According to the coding standard of ICD-9, this article counted the top 10 diseases with the
largest number of patients admitted to the hospital according to the first diagnostic criteria
(Table 2). A total of 38,163 patients were admitted, and 38,156 patients had the first diagnosis.
The total number of the top 10 diseases with the largest number of patients is 15020. Because
we only selected the patients’ first diagnosis as the data analysis, it is equivalent to each patient
PLOS ONE | https://doi.org/10.1371/journal.pone.0232176 April 30, 2020
4 / 12
PLOS ONE Analysis of adult disease characteristics and mortality on MIMIC-III
Fig 2. ICU mortality of all patients (in-hospital + 90-day after discharge).
https://doi.org/10.1371/journal.pone.0232176.g002
suffering from only one disease. No cross-replicated data for the diseases. The disease informa-
tion of patients was counted according to the first diagnosis. The most common disease of
patients admitted to hospital is Other forms of chronic ischemic heart disease, but the mortal-
ity rate of this category is the lowest in the ten diseases. The 90-day mortality rate and in-hospi-
tal mortality of Septicemia disease are the highest among the ten diseases. Heart failure has the
highest median age among the ten diseases it is 77 years (Q1-Q3:66–85).
PLOS ONE | https://doi.org/10.1371/journal.pone.0232176 April 30, 2020
5 / 12
PLOS ONE Analysis of adult disease characteristics and mortality on MIMIC-III
Fig 3. Mortality of different race (It is calculated as: The number of deaths per race/the total number of patients
per race).
https://doi.org/10.1371/journal.pone.0232176.g003
Fig 6 shows the mortality rate of the top 10 diseases with the largest number of patients.
The disease with the highest mortality rate is Septicemia, and the mortality rate is 48.9%. Fol-
lowed by Intracerebral hemorrhage and Other diseases of lung, the mortality rate of which also
exceeded 44%. Patients with Other diseases of lung have the longest ICU stay. The days
reached 4.68 (Q1-Q3: 2.1–9.8).
PLOS ONE | https://doi.org/10.1371/journal.pone.0232176 April 30, 2020
6 / 12
PLOS ONE Analysis of adult disease characteristics and mortality on MIMIC-III
Fig 4. Ethnic distribution of all patients (Total number of patients is 38161).
https://doi.org/10.1371/journal.pone.0232176.g004
Fig 5. Age distribution of all admitted patients (Total number of patients is 38163).
https://doi.org/10.1371/journal.pone.0232176.g005
PLOS ONE | https://doi.org/10.1371/journal.pone.0232176 April 30, 2020
7 / 12
PLOS ONE Table 2. Basic information of ten diseases with the largest number of patients.
Analysis of adult disease characteristics and mortality on MIMIC-III
No.
Disease/ICD-9
Age, years,
median
(Q1-Q3)
68(60–75)
1
2
3
4
5
6
7
8
9
Other forms of chronic ischemic
heart disease /414
Acute myocardial infarction/410
70(59–81)
Septicemia/038
72(59–83)
Other diseases of endocardium
/424
71(60–79)
Other diseases of lung/518
69(56–80)
Heart failure/428
77(66–85)
Intracerebral hemorrhage/431
72(60–82)
Subarachnoid, subdural, and
extradural hemorrhage, following
injury/852
76(58–85)
Cardiac dysrhythmias /427
70(59–81)
10 Occlusion of cerebral arteries/434
76(63–85)
https://doi.org/10.1371/journal.pone.0232176.t002
Male (%) Days in hospital,
median days
(Q1-Q3)
Days in ICU,
median days
(Q1-Q3)
Death in the
hospital (%)
90 days of death
after discharge (%)
Number of
patients
2371
(77.3%)
1723
(64.1%)
1358
(52.5%)
865
(59.4%)
544
(51.0%)
533
(56.2%)
491
(52.9%)
484
(54.8%)
463
(60.6%)
292
(46.5%)
6(5–9)
2.0(1.2–3.1)
34(1.1%)
47(1.5%)
6(3–10)
2.1(1.2–3.9)
305(11.3%)
173(6.4%)
7(4–13)
2.9(1.6–6.5)
952(36.8%)
314(12.1%)
7(5–10)
2.1(1.2–3.3)
24(1.6%)
28(1.9%)
8(4–15)
4.7(2.1–9.8)
331(31.0%)
144(13.5%)
8(5–13)
2.7(1.4–5.1)
182(19.2%)
153(16.1%)
6(3–12)
2.1(1.1–5.1)
334(36.0%)
83(8.9%)
5(3–10)
1.8(1.0–3.4)
164(18.6%)
69(7.8%)
5(2–8)
1.9(1.1–3.4)
125(16.4%)
56(7.3%)
6(4–11)
2.1(1.2–3.9)
158(25.2%)
75(11.9%)
3069
2689
2587
1456
1067
948
929
883
764
628
Fig 7 shows the distribution of ten diseases with the largest number of patients in ICU. The
disease with the greatest number of patients is Other forms of chronic ischemic heart disease
in the five ICUs. Followed by Other diseases of endocardium and Acute myocardial infarction.
In the CCU, the disease with the greatest number of patients is Acute myocardial infarction. In
the MICU, the disease with the greatest number of patients is Septicemia. In SICU and TSICU,
the number of patients is relatively small. The main diseases are Intracerebral hemorrhage and
Subarachnoid, subdural, and extradural hemorrhage, followed by injury and Occlusion of
cerebral arteries.
4. Discussion
The MIMIC-III database is the first large ICU database that is open to the public free of charge.
It has a really large dataset with rich types of medical data, providing high-quality data
resources for clinical research, mining and building a knowledge base. This study was based
on the latest version of the MIMIC-III database to analyze the information of diseases and the
basic information of adult patients who had entered the ICU for the last time.
The research in this paper was mainly divided into two parts. In the first part, we divided
the admission patients into three categories: death in the hospital, 90-day of death after dis-
charge, and no death within 90-day after discharge. Referring to some scholars’ research
[21,22], this paper divided patients into three categories, which can better understand the dis-
ease characteristics of patients in the MIMIC database. The second part was to introduce the
basic information about the top 10 diseases with the largest number of patients, including the
length of hospital stay, gender, age, in-hospital mortality, and 90-day mortality. Based on the
disease information of the admitted patients, the mortality of each disease and the distribution
of ICUs were calculated.
PLOS ONE | https://doi.org/10.1371/journal.pone.0232176 April 30, 2020
8 / 12
PLOS ONE Analysis of adult disease characteristics and mortality on MIMIC-III
Fig 6. Mortality of ten diseases with the largest number of patients (all dead patients: death in-hospital + 90-day
of death after discharge).
https://doi.org/10.1371/journal.pone.0232176.g006
Fan [18] also made a statistical analysis of the MIMIC-III database. The distribution of dis-
ease spectrum in patients of the same age is described in detail. However, the whole study basi-
cally analyzed the patients as a whole and does not classify the patients. It is not clear about the
changes in patients at different stages. Therefore, we divided the patients into three parts for
the analysis of related indicators and mortality. This can help us understand the patients’ situa-
tion more clearly. And they selected the patients who were admitted for the first time. In our
study, we selected the patients who were admitted to the hospital for the last time and they
were more representative. Their research is based on the human body systems. There is no
analysis of specific diseases. In our study, the basic information and mortality of the top 10 dis-
eases with the largest number of patients were analyzed. We also calculated the prevalence of
the top 10 diseases with the largest number of patients in different ICUs, so that we can see the
types of the main diseases in each ICU more clearly.
This study was subject to some limitations: First, we extracted data from the MIMIC-III
database, which is a retrospective study with selective bias. Second, this study mainly involved
Caucasian, African American and Asians are very few. Due to ethnic differences, the results of
this study need to be further verified in other ethnic groups. Third, we did not conduct influ-
encing factors analysis and predictive analysis for these diseases. Influencing factors analysis
and predictive analysis can better remind people how to prevent the diseases from the onset,
predict the progression of the diseases according to the patients’ condition and basic physio-
logical characteristics, and provide patients with better prognosis and rational allocation of
medical resources. This will be the main direction of our next research.
PLOS ONE | https://doi.org/10.1371/journal.pone.0232176 April 30, 2020
9 / 12
PLOS ONE Analysis of adult disease characteristics and mortality on MIMIC-III
Fig 7. The distribution of ten diseases with the largest number of patients in ICU.
https://doi.org/10.1371/journal.pone.0232176.g007
5. Conclusion
In summary, this study provides a detailed analysis and description of the case records and dis-
ease distribution of adult patients entering the ICU for the MIMIC-III database. It provides
the statistics of patients’ basic information and the top 10 diseases with the largest number of
patients and the distribution of the ten diseases. This paper implements a complete data analy-
sis process for the MIMIC-III database, and systematically paves the way for clinicians and
researchers to conduct data research in the medical field.
Supporting information
S1 Data.
(DOCX)
S1 Checklist. PLOS ONE clinical studies checklist.
(DOCX)
Author Contributions
Conceptualization: Mengdie Li.
Data curation: Jinfa Wu, Mengdie Li.
Formal analysis: Mengdie Li.
Resources: Zheng Dai, Ke Li.
Software: Zheng Dai.
Supervision: Ke Li.
PLOS ONE | https://doi.org/10.1371/journal.pone.0232176 April 30, 2020
10 / 12
PLOS ONE Analysis of adult disease characteristics and mortality on MIMIC-III
Validation: Ke Li.
Visualization: Zheng Dai.
Writing – original draft: Zheng Dai.
Writing – review & editing: Siru Liu, Jialin Liu.
References
1. Aboelsoud M, Javaid AI, Al-Qadi MO, et al. Hypoxic hepatitis—its biochemical profile, causes and risk
factors of mortality in critically-ill patients: A cohort study of 565 patients. Journal of Critical Care.2017;
41:9–15. https://doi.org/10.1016/j.jcrc.2017.04.040 PMID: 28460210
2. Shen Y, Zhang W, Cheng X, et al. Association between postoperative fluid balance and acute kidney
injury in patients after cardiac surgery: A retrospective cohort study. Journal of Critical Care.2017;
44:273. https://doi.org/10.1016/j.jcrc.2017.11.041 PMID: 29220757
3. Calvert J, Mao Q, Hoffman JL, et al. Using electronic health record collected clinical variables to predict
medical intensive care unit mortality. Annals of Medicine and Surgery.2016; 11:52–57. https://doi.org/
10.1016/j.amsu.2016.09.002 PMID: 27699003
4.
Johnson AEW, Ppllard TJ, Shen L, et al. MIMIC-III, a freely accessible critical care database. Sci Data,
2016, 3: 160035. https://doi.org/10.1038/sdata.2016.35 PMID: 27219127
5. Kurniati AP, Hall G, Hogg D, et al. Process mining in oncology using the MIMIC-III dataset. Journal of
Physics Conference Series. 2018, 971.
6.
Feng M, McSparron JI, Kien DT, et al. Transthoracic echocardiography and mortality in sepsis: analysis
of the MIMIC-III database. Intensive Care Med.2018; 44: 884–892. https://doi.org/10.1007/s00134-018-
5208-7 PMID: 29806057
7. Parreco JP, Hidalgo AE, Badilla AD, et al. Predicting central line-associated bloodstream infections and
mortality using supervised machine learning. Journal of Critical Care. 2018; 45:156–162. https://doi.org/
10.1016/j.jcrc.2018.02.010 PMID: 29486341
8. Sanjay P, Chuizheng M, Zhengping C, et al. Benchmarking Deep Learning Models on Large Healthcare
Datasets. Journal of Biomedical Informatics.2018; 83:112–134. https://doi.org/10.1016/j.jbi.2018.04.
007 PMID: 29879470
9. Beaulieu-Jones BK, Orzechowski P, Moore JH. Mapping Patient Trajectories using Longitudinal Extrac-
tion and Deep Learning in the MIMIC-III Critical Care Database. Pacific Symposium on Biocomputing
Pacific Symposium on Biocomputing.2018.
10. Van PS, Zhongheng Z, Martin S, et al. Scalable Predictive Analysis in Critically Ill Patients Using a
Visual Open Data Analysis Platform. PLOS ONE 2016, 11(1):e0145791. https://doi.org/10.1371/
journal.pone.0145791 PMID: 26731286
11.
Lee J, Ribey E, Wallace JR. A web-based data visualization tool for the MIMIC-II database. BMC Medi-
cal Informatics and Decision Making.2015; 16(1):15.
12. Desautels T, Calvert J, Hoffman J, et al. Prediction of Sepsis in the Intensive Care Unit With Minimal
Electronic Health Record Data: A Machine Learning Approach. Jmir Medical Informatics.2016, 4(3):
e28. https://doi.org/10.2196/medinform.5909 PMID: 27694098
13.
Zimmerman JE, Kramer AA, Knaus WA. Changes in hospital mortality for United States intensive care
unit admissions from 1988 to 2012. Critical Care.2013, 17:R81. https://doi.org/10.1186/cc12695 PMID:
23622086
14. Marshall DC, Salciccioli JD, Goodson RJ, et al. The association between sodium fluctuations and mor-
tality in surgical patients requiring intensive care. Journal of Critical Care.2017; 40:63–68. https://doi.
org/10.1016/j.jcrc.2017.02.012 PMID: 28347943
15.
16.
Johnson AE, Stone DJ, Celi A, et al. The MIMIC Code Repository: enabling reproducibility in critical
care research. Journal of the American Medical Informatics Association.2018; 25(1):32–39. https://doi.
org/10.1093/jamia/ocx084 PMID: 29036464
Lee J. Patient-Specific Predictive Modeling Using Random Forests: An Observational Study for the Crit-
ically Ill. Jmir Medical Informatics.2017; 5(1):e3. https://doi.org/10.2196/medinform.6690 PMID:
28096065
17. Raghupathi W, Raghupathi V. Big data analytics in healthcare: promise and potential[J]. Health Infor-
mation Science and Systems.2014; 2(1):3.
18.
Fan Y, Zhao Y, Li P, et al. Analysis of diseases distribution in Medical Information Mart for Intensive
Care III database. Zhonghua Wei Zhong Bing Ji Jiu Yi Xue. 2018; 30(6):531–537 https://doi.org/10.
3760/cma.j.issn.2095-4352.2018.06.006 PMID: 30009726
PLOS ONE | https://doi.org/10.1371/journal.pone.0232176 April 30, 2020
11 / 12
PLOS ONE Analysis of adult disease characteristics and mortality on MIMIC-III
19.
Lokhandwala Sharukh, Ned McCague Abdullah Chahin. One-year mortality after recovery from critical
illness: A retrospective cohort study.PLoS One 2018; 13(5).
20. Anon. Ninth revision of the International Classification of Diseases(ICD-9) [EB/OL]. [2017-12-21]. http://
www.cdc.gov/nchs/data/statab/gmwki_98.pdf
21.
Lacroix J, He´ bert Paul, Fergusson D, et al. The Age of Blood Evaluation (ABLE) Randomized Controlled
Trial: Study Design. Transfusion Medicine Reviews.2011; 25(3):197–205. https://doi.org/10.1016/j.
tmrv.2011.03.001 PMID: 21550205
22. Rood PJT, van de Schoor F, van Tertholen K, et al. Differences in 90-day mortality of delirium subtypes
in the intensive care unit: A retrospective cohort study. Journal of Critical Care.2019; 53:120–124.
https://doi.org/10.1016/j.jcrc.2019.06.007 PMID: 31228762
PLOS ONE | https://doi.org/10.1371/journal.pone.0232176 April 30, 2020
12 / 12
PLOS ONE |
10.1371_journal.pone.0215651 | RESEARCH ARTICLE
Simulating beach and dune evolution at
decadal to centennial scale under rising sea
levels
Caroline HallinID*, Magnus Larson, Hans Hanson
Department of Water Resources Engineering, Lund University, Lund, Sweden
* caroline.hallin@tvrl.lth.se
Abstract
A numerical model for simulating beach-dune evolution at decadal to centennial time scales
is developed. The work builds on an existing semi-empirical cross-shore model, the CS-
model, to which the effect of sea level rise is added and the routines for aeolian transport and
morphological dune evolution are improved. The model development is based on established
conceptual models from the literature, which are translated into mathematical formulations
and solved numerically. The capability of the proposed model is demonstrated through a
case study at A¨ ngelholm Beach, Sweden. The model is calibrated and validated against a
seven-year long data set on morphological evolution and sediment grain-size samples.
Beach and dune evolution is then simulated from 2018 to 2100 for a range of sea level rise
scenarios. The model results are promising, and suggest that the model has potential to be
used for long-term assessment of climate change impact on beaches and dunes.
Introduction
Coastal dunes play an important role to protect against flooding and beach erosion, while pro-
viding nature and recreation values [1,2]. Therefore, both natural and artificial dunes are
becoming increasingly popular flood protection methods in many coastal areas [3,4]. Mean-
while, rising sea levels increase the probability of flooding as well as the pressure on beaches
and dunes [5,6]. The capability to predict future beach and dune evolution is thus of major
importance for flood risk assessment and design of coastal protection. Several models have
been developed to separately estimate dune erosion during storms [7–9], dune build-up due to
aeolian transport [10–12], and long-term coastline evolution [13]. However, models for the
evolution of the beach and dune system at decadal time scales, which are of particular interest
in coastal management [14], require a coupling between constructive and destructive near-
shore, beach, and dune processes [15]. In recent years, there has been advances in the coupling
of cross-shore and longshore transport processes at decadal time scales, e.g., the CoSMoS--
COAST model [16] and the LX-model [17]; but, these models do not include dune processes.
To improve the understanding of how processes in the land-sea interface control the foredune
evolution, Cohn et al [18] combined three process-based models—X-BEACH [8], CDM [19],
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: Hallin C, Larson M, Hanson H (2019)
Simulating beach and dune evolution at decadal to
centennial scale under rising sea levels. PLoS ONE
14(4): e0215651. https://doi.org/10.1371/journal.
pone.0215651
Editor: João Miguel Dias, Universidade de Aveiro,
PORTUGAL
Received: October 10, 2018
Accepted: April 7, 2019
Published: April 19, 2019
Copyright: © 2019 Hallin et al. This is an open
access article distributed under the terms of the
Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: The wind and water
level data is open access provided by the Swedish
Meteorological and Hydrological Institute, SMHI,
through www.smhi.se. The grain size data is
available through an FTP server with address www.
tvrl.se/caf/ftp/Sieved_grain_size_samples.xlsx.
Model code, input and results files will not be
provided open access but can be made available
upon request. Contact Caroline Hallin at Lund
University, Sweden, at caroline.hallin@tvrl.lth.se.
Topographic and bathymetric data cannot be
shared publicly because it has not been collected
by the authors and is not open access. The
PLOS ONE | https://doi.org/10.1371/journal.pone.0215651 April 19, 2019
1 / 30
topographic data are available from A¨ngelholm
municipality. Requests to access the data should
be sent to info@engelholm.se. The bathymetric
data is owned by the Swedish Geological Survey
(SGU) and can only be accessed by researchers
who meet the criteria for access to this confidential
data. For more information on how to access the
bathymetric data, contact kundservice@sgu.se.
Since the authors are not in possession of the
topographic and bathymetric data, we cannot
guarantee that the organizations who own this data
will make it available to other researchers.
However, the observed volumetric changes that are
presented in the manuscript are sufficient for
calibration and validation of the model.
Funding: The authors received no specific funding
for this work.
Competing interests: The authors have declared
that no competing interests exist.
Simulating beach and dune evolution at decadal to centennial scale under rising sea levels
and Aeolis [20]—in the model framework Windsurf. Windsurf successfully simulated dune
evolution in one cross-shore transect during one year, however, longer time and spatial scales
require reduced complexity approaches. In this study, a model is developed that simulates sed-
iment transport and morphological evolution of the beach and dune on time scales of decades
to centuries. The aim is to create a robust, computationally efficient model that couples near-
shore, beach, and dune processes.
Simulations of long-term beach and dune evolution require integration of sediment trans-
port processes acting in both the cross-shore and longshore direction on multiple time scales
[14,21]. This work builds on the CS-model [15], a semi-empirical beach profile evolution
model that simulates sediment transport due to dune erosion and dune overwash caused by
wave action, dune build-up from aeolian transport, and beach-bar material exchange. Long-
shore sediment transport gradients are accounted for as constant sources or sinks, and nour-
ishment episodes can be added to the beach or dune [22]. Previously, sea level rise (SLR) has
not been included in the CS-model and the dune build-up from wind-blown sand did not take
into account sediment availability, which is an important factor for long-term dune evolution
[23,24]. In this study, the CS-model is developed through introducing SLR and mathematical
methods to compute aeolian transport and the associated dune growth.
The included transport processes are active over a range of spatial and temporal scales (Fig
1). Sediment transport and morphological evolution are simulated in cross-shore transects,
which may be aggregated to represent coastal evolution at spatial scales of kilometres. Model
input is SLR, wind, deep water waves, and simultaneous still water levels (SWL). From a
coastal management perspective model output of interest is e.g. long-term shoreline change,
the shoreline variation envelope, and dune height and dune volume in relation to storm ero-
sion rates [25]. The model is computationally efficient due to reduced complexity transport
equations and a simplified representation of the coastal morphology, which makes it suitable
for probabilistic approaches.
The main objective of this study is to develop a simple, robust, and computationally effi-
cient model for evolution of the beach-dune system under rising sea levels at decadal to cen-
tennial time scales. Established conceptual morphological models are translated into
mathematical formulations to extend the CS-model [15] to include the dynamic and intercon-
nected effects of SLR and aeolian transport.
The paper starts by presenting the theoretical background for the proposed model develop-
ment, followed by a model description. The proposed model is then applied to simulate the
long-term beach and dune evolution at A¨ ngelholm Beach, Sweden. First, the model is cali-
brated and validated against a seven-year data set and thereafter, the evolution from 2018–
2100 is simulated considering a range of SLR scenarios. The paper ends with a discussion on
model performance and suggestions for further development. Notations used throughout the
paper are summarized in a list of symbols at the end.
Theoretical background
Aeolian sediment transport
The mechanics of sediment transport by wind and the associated morphological evolution of
coastal dune systems have been studied extensively [26,27], which has provided physical
insights to aeolian transport processes. Still, existing models do not manage to accurately pre-
dict transport rates at time scales relevant for coastal management [23,25,28]. Sediment trans-
port formulas only accounting for wind speed and grain size [26,29–33] are commonly
overestimating transport rates when compared to field data from beach environments [23,34].
PLOS ONE | https://doi.org/10.1371/journal.pone.0215651 April 19, 2019
2 / 30
Simulating beach and dune evolution at decadal to centennial scale under rising sea levels
Fig 1. Conceptual model of decadal to centennial scale beach and dune evolution.
https://doi.org/10.1371/journal.pone.0215651.g001
When modelling yearly to decadal sediment transport rates, limits to transport are of inter-
est, rather than time-varying forcing conditions [23]. Such limiting factors are sediment avail-
ability [24], beach width and fetch length [35], beach slope [34], surface moisture [35], snow
and ice cover [36], vegetation, and crust development on the beach surface [20].
Dune build-up by wind requires sediment with a grain size fine enough to be mobilised by
the wind, and coarse enough to be deposited in the foredune, where the wind shear stress
PLOS ONE | https://doi.org/10.1371/journal.pone.0215651 April 19, 2019
3 / 30
Simulating beach and dune evolution at decadal to centennial scale under rising sea levels
decreases due to the effect of vegetation and topography [35]. Grain sizes in aeolian deposits
typically range between 0.15–0.30 mm [37], and are often smaller than those found on the
beach [38]. Eroding beaches tend to be drained of fine sediment and accreting beaches to be
supplied with fine sediment due to the selection of grain size in transport processes, where
smaller grains are more likely to be picked up [39]. In this way, the aeolian sediment transport
depends on the longshore and cross-shore transport processes, which control the supply of
sediment of appropriate grain size.
Dune morphology
Transport rates alone cannot be used to predict the morphological evolution of foredunes. It is
also important where the aeolian sediment is deposited; on the crest, the seaward or landward
slopes. The beach sediment budget, which could be positive, negative, or stable, affects both
the rate of transport and the morphological evolution [24,25]. According to Psuty’s conceptual
model [24], dunes at an accreting beach will grow rapidly and create a prograding dune ridge
topography, where a new foredune is formed in front of the existing ones, creating low dunes
with mild slopes [40,41]. The dunes will be low because there is not enough time for them to
grow in height before a new foredune is built in front of them [42]. At a stable beach, the dune
stays in place and grows higher due to scarping and recovery. Eroding beaches may develop in
two different ways; if they are slightly eroding, the dune will maintain or even increase its vol-
ume, grow higher, and be displaced inland through scarping in combination with aeolian
transport and/or overwash. If the beach is eroding rapidly and overwash processes are domi-
nant, the dune will be flattened and move landward.
The explanatory mechanisms behind these different morphological behaviours are sedi-
ment supply and the influence of vegetation. If vegetation is present at the dune foot, sediment
is trapped in an embryonal foredune and without vegetation sediment is deposited near the
foot of the foredune, forming a dune ramp (Fig 2), which facilitates the passage of wind-blown
sand over the dune crest [43,44].
The interaction between dune ramps and morphological evolution was studied at Skallin-
gen spit in Denmark, which is a retreating coastal feature, subject to overwash and landward
foredune migration [44]. Christiansen and Davidson-Arnott [44] found that aeolian transport
over the dune crest increased rapidly as the vertical distance from the top of the ramp to the
Fig 2. Vegetated embryonal foredune (left) at an accreting beach, and non-vegetated dune ramp (right) at a stable beach in A¨ ngelholm, Sweden. The sites are later
referred to as profile C and B, respectively.
https://doi.org/10.1371/journal.pone.0215651.g002
PLOS ONE | https://doi.org/10.1371/journal.pone.0215651 April 19, 2019
4 / 30
Simulating beach and dune evolution at decadal to centennial scale under rising sea levels
crest became equal to or less than 1 m, and that the wind during wet conditions eroded the
ramp to approximately this height.
The dune ramp concept was further developed in a study at Greenwich Dunes on Prince
Edward Island in Canada [36]. In 2002–2009 measurements were performed at two stretches of a
beach with different littoral sediment budgets; one negative and one positive or neutral, although
both showing a long-term transgressive trend. After scarping, if no ramp was present, very little
sediment reached the dune crest or landward slope. When a dune ramp was present, accumula-
tion was observed at the upper seaward slope, crest and landward slope. The beach with a tempo-
rarily positive sediment transport had vegetated embryonal dunes present for 2–3 years and
through that time most sediments were deposited at the seaward side of the dune. When the
embryonal dune was damaged and a vegetated ramp had not yet been formed, the observed
transport to the crest and landward side of the dune was significant. These studies support that
the sediment budget affects the morphological evolution, since embryonal foredunes are more
likely to occur on accreting beaches whereas dune ramps form on eroding or stable beaches.
For long-term modelling of foredune evolution, dune height is essential as it affects flood
safety. The conceptual models of Psuty [24] and Sherman and Bauer [25] suggest that dune
height is depending on the sediment budget. The ratio between height/width of coastal fore-
dunes are typically about 0.11 [12], but the maximum height varies between different environ-
ments. Dura´n and Moore [19] proposed a linear relationship between foredune height and
wave height, assuming that the dune height was depending on wave impact. Hesp [27] found a
strong relationship between increasing foredune height and increasing dissipativeness but
concluded that the trend might also be explained by higher exposure to storm winds at dissipa-
tive beaches. In summary, the dune height is site-specific and depends on the sediment budget,
wind and wave climate, and sediment and vegetation characteristics.
The Bruun rule
The Bruun rule provides a simple method to estimate the shoreline retreat, RBruun, under a
slowly rising sea level [45],
RBruun ¼ SSLR
BBruun
h
ð1Þ
where SSLR is the sea level rise, BBruun and h the width and the height of the active profile,
respectively. The width BBruun represents the distance from the shoreline to the depth closure,
and h is the sum of the depth of closure and the beach berm height.
The Bruun rule is based on the assumption that SLR creates accommodation space for sedi-
ment within the subaqueous part of the active profile. The Bruun Rule only considers cross-
shore transport, assuming that the profile consists entirely of sand, that there is a full spectrum
of waves, wind and water levels to force the profile to its new equilibrium, and that the SLR is
slow. All sediment transport is in the seaward direction to the subaqueous part of the profile.
However, infilling sediment can come from various sources, not only the beach. The sediment
may also originate from a negative longshore sediment transport gradient, artificial nourish-
ments, or from offshore supplies outside the depth of closure, also known as the Dean equilib-
rium concept [46–48].
The original version of the Bruun rule does not consider any vertical adjustment of the sub-
aerial part of the profile to SLR. If the mean sea level rises, also the beach berm height and
dune foot are expected to adjust to the new sea level [49]. For this purpose, extended versions
of the Bruun Rule that accounts for landward transport due to aeolian processes and overwash
of the beach berm and dune have been proposed [49–51].
PLOS ONE | https://doi.org/10.1371/journal.pone.0215651 April 19, 2019
5 / 30
Simulating beach and dune evolution at decadal to centennial scale under rising sea levels
It should be noted that the Bruun rule is a much discussed method that has been subject to
numerous critical evaluations [52]. The first main argument against the Bruun rule concerns
its two-dimensionality and the assumption that other transport processes such as gradients in
longshore transport and overwash are excluded. In the CS-model, however, the Bruun rule
will only be applied to compute the transport directly related to SLR. Other relevant transport
processes are dealt with in separate transport equations, and the morphological evolution of
the beach and dune is a result of all of these types of transports.
The second common critique against the Bruun rule is the validity of the concepts of equi-
librium profiles and depth of closure. In a recent extensive laboratory study [53], both the orig-
inal Bruun rule [45] and a modified version including landward transport [49] showed good
agreement with observations and confirmed the underlying assumption of a profile adjust-
ment to SLR.
Last, the Bruun rule has also been criticized for ignoring import of sediment from deeper
water outside depth of closure [47]. However, these processes are more common at geological
time scales than at engineering time scales. If known to be present they could be added as a
sediment source in the model.
Model for long-term foredune and beach evolution
The CS-model simulates sediment transport and the associated morphological evolution based
on a set of transport and continuity equations [15]. In this paper, the profile schematization is
developed and routines are added for aeolian transport, morphological dune evolution, and
the effect of SLR. For more detailed description of the other model components we refer to
Larson et al [15].
Profile schematization
The beach profile is schematized through distances, slopes, characteristic heights, and volumes
(Fig 3), which are updated every time step. Length coordinates, relative to a reference point
Fig 3. Definition of heights, length coordinates, volumes and transports (red arrows).
https://doi.org/10.1371/journal.pone.0215651.g003
PLOS ONE | https://doi.org/10.1371/journal.pone.0215651 April 19, 2019
6 / 30
Simulating beach and dune evolution at decadal to centennial scale under rising sea levels
behind the dune, are defined for landward and seaward dune foot, yL and yS, and dune crest,
y'L and y'S, and the intersection with mean sea level, yG. Landward and seaward dune slopes,
tan βL and tan βS, are assumed to be constant and a maximum foreshore slope, tan βF,max, is
defined. If tan βF,max is exceeded, βF >βF,max, the entire beach is considered a swash zone.
Under this condition, sediment that would have been eroded from the beach is instead eroded
from the dune.
Depth of closure, DC, and dune foot height, DF, are constants referenced to the mean sea
level (MSL). The dune height, S, is a variable, specified relative to DF. A maximum value of S,
Smax, is limiting dune height growth if the sediment budget is positive, whereas for negative
and stable budgets S has no restrictions [12]. Smax can be approximated as a representative
height of prograding dune ridges, present in or near the study area.
Volumes and transport rates are given per meter of beach width. The specified volumes are
dune volume, Vdune, where the dune ramp volume, Vramp, is an integral part, total beach vol-
ume, Vbeach, where the berm volume, Vberm, is an integral part, and bar volume, Vbar (Fig 3).
The bar volume represents nearshore subaqueous deposits involved in material exchange with
the beach and may consist of none, one, or multiple bars. The rest of the subaqueous part of
the profile, from MSL to DC, is assumed to be in equilibrium [54].
The beach volume Vbeach is defined as the volume of sediment from a reference point
behind the dune to the intersection with mean sea level, yG, which is vertically limited by the
depth of closure, DC, and the dune foot height, DF. The sub-volume Vberm is used to calculate
beach width; the quantity yG-yS is defined as a function of Vberm based on site specific data.
Thus, the shape of the beach is not specified; it is only described by its width and volume, indi-
cated with a dashed line in Fig 3.
The purpose of the ramp volume Vramp is to determine how the sediment distribution over
the dune occurs. If the dune is eroded, Vramp is eroded first and if the ramp disappears
completely, the main part of the dune (behind the seaward dune crest, y'S) will start to erode.
The changes in profile schematization compared to previous versions of the CS-model
require adjustment of the frictional losses over the beach. The runup height significant for
ffiffiffiffiffiffiffiffiffiffiffiffiffi
dune impact, R, is defined as R ¼ 0:158
, where Hrms and L0 are deep water root-
HrmsL0
mean-square wave height and mean wavelength, respectively [55]. R is corrected for friction
over the beach according to [56],
p
R0 ¼ Rexpð(cid:0) 2cf xÞ þ ðDF (cid:0) SWLÞð1 (cid:0) expð(cid:0) 2cf xÞÞ
ð2Þ
where R' is the adjusted runup height, cf is an empirical friction coefficient, and x is the hori-
zontal travel distance of the wave front. If SWL+ R' > DF dune erosion will occur. Since the
beach in the CS-model has no shape, the travel distance cannot be exactly known. Instead the
travel distance x is expressed as a function of the volume Vberm,
x ¼
2Vberm
DF
�
1 (cid:0)
�
SWL
DF
ð3Þ
Aeolian sediment transport
The proposed aeolian sediment transport scheme is based on the assumption that, on a decadal
time scale, the most important limiting factor for aeolian sediment transport is the supply of
material of appropriate grain size. Sediment available for aeolian transport is assumed to be
present as a location-specific fraction of the transported sediment volumes longshore and
cross-shore. Availability is computed through a balance including sediment transport and
PLOS ONE | https://doi.org/10.1371/journal.pone.0215651 April 19, 2019
7 / 30
Simulating beach and dune evolution at decadal to centennial scale under rising sea levels
nourishments, to and from Vbeach. The volume of sediment available for aeolian transport, VW,
in time step i is calculated as,
VW;i ¼ VW;i(cid:0) 1 þ ðqS;i(cid:0) 1 (cid:0) qW;i(cid:0) 1 (cid:0) A q ðqSLR;i(cid:0) 1 (cid:0) qLS;i(cid:0) 1ÞÞDt þ AbVnour;i(cid:0) 1
VW � 0
ð4Þ
where qS is the transport rate of eroded sediment from the dune to the beach, qW is the aeolian
transport rate from the beach to the dune, qSLR is the transport rate of sediment to compensate
for SLR, ‘Bruun Rule’ transport, qLS is the transport rate due to gradients in longshore trans-
port, Δt is length of time step, and Vnour volume of beach nourishments. The coefficients Aq
and Ab describe the fractions of the transport rate and nourished volume, respectively, that are
within the proper range of grain size and available for aeolian transport.
If VW,i > 0, aeolian transport will take place during time step t = i. If VW,i � 0, the aeolian
transport will be turned off and Vw,i is set to zero. The initial available volume, Vw,0, depends
on conditions prior to the simulation period, e.g. nourishments and large dune erosion events.
The potential aeolian sediment transport rate, mWE, in kg/s/m, can be estimated by [31],
mWE ¼ KW
s
ffiffiffiffiffiffiffiffi
D50
Dref
50
ra
u2
�
g
ðu� (cid:0) u�cÞ
ð5Þ
ref is the median reference grain size of 0.25 mm, ρa the
where D50 is the median grain size, D50
density of air, g the acceleration due to gravity, u� the shear velocity at the bed, u�c the critical
shear velocity at the bed, and Kw an empirical coefficient. If u� < u�c, then mWE = 0.
The shear velocity, u�, can be calculated using the law of the wall,
uz
u�
¼
1
k
� �
z
z0
ln
ð6Þ
where uz is the wind velocity at z meter above ground, z0 is the aerodynamic roughness height,
and κ is von Karman’s constant (� 0.41). For wind conditions below the critical shear velocity
for initiation of transport, u�<u�c, the roughness height, z0, can be parameterized as z0 =
0.081log10(d/0.18), where d is the grain size in mm [33]. The critical shear velocity can be esti-
mated through [26],
u�c ¼ AW
s
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ðrs (cid:0) raÞ
ra
gD50
ð7Þ
where ρs is the density of sand (typically 2650 kg/m3) and AW is a coefficient of about 0.1.
To determine the shear velocity during transport, two wind measurements at different
heights or one wind measurement and an estimate of the roughness height affected by trans-
port are required. For long-term simulations, wind data is typically not available from the
beach of interest but from a wind gauge that is not influenced by the sand transport. Consider-
ing the spatial and temporal resolution of available wind data, the relationship between uz and
2(uz-uz,c) [58]. The critical velocity for
u� can be approximated as linear [57] so that mWE/uz
transport at z m height, uz,c can be computed through combining Eqs 6 and 7. Eq 5 can then
be rewritten as,
s
mWE ¼ CW
where CW is an empirical coefficient.
PLOS ONE | https://doi.org/10.1371/journal.pone.0215651 April 19, 2019
ffiffiffiffiffiffiffiffi
D50
Dref
50
ra
u2
z
g
ðuz (cid:0) uz;cÞ
ð8Þ
8 / 30
Simulating beach and dune evolution at decadal to centennial scale under rising sea levels
The mass flux mWE is converted to a volumetric equilibrium transport rate (qWE) of sand to
the dunes by,
qWE ¼
mWE
rsð1 (cid:0) PÞ
ð9Þ
where P is the porosity (typically 40%).
To reach the transport rates described by the equilibrium equations, a critical fetch length is
required [35,59]. The fetch length, F, depends on the wind angle towards shore normal, θ, and
the dry beach width, Bdry. For 0˚ � θ � 80˚,
F ¼
Bdry
cosðyÞ
ð10Þ
The beach is assumed to be dry landward of the runup limit. A runup length coordinate, yR,
is defined as,
�
yR ¼ yS þ 1 (cid:0)
�
R þ SWL
DF
ðyG (cid:0) ySÞ
ð11Þ
The dry beach width, Bdry, is then calculated as the horizontal distance from the runup limit
to the seaward dune foot Bdry = yR-yS. If yR � yS, the whole beach is assumed to be wet and
there will be no aeolian transport.
A simplified equation for the potential transport rate, corrected for fetch-limited condi-
tions, qWF, was developed by Larson et al [15] based on the work by Sauermann et al [60],
qWF ¼ qWEð1 (cid:0) expð(cid:0) d FÞÞ
ð12Þ
where δ is an empirical coefficient in the order of 0.1–0.2 m-1.
Oblique wind angles have longer fetches and may therefore generate higher aeolian trans-
port rates. This effect is counteracted by the cosine effect, implying that only the onshore com-
ponent, qW, adds to the dune volume,
qW ¼ qWFcosðy Þ
ð13Þ
Aeolian transport due to offshore directed wind is neglected.
When neglecting the impact of grain size, the effect of slope correction on transport rates
has been found to be small at beaches with slopes below 15˚ [34]. Here, grain size is accounted
for in Eqs 5 and 6 and since most beach slopes are milder than 15˚ [61], the effect of beach
slope is assumed to be negligible.
Morphological dune evolution
Dunes evolve differently depending on where the aeolian transported sediment is deposited, in
front of the dune, on the seaward slope, on the crest, or on the landward slope. Here we intro-
duce a sediment deposition scheme based on Psuty’s sediment budget formulation [24] and
the dune ramp concept [36,44].
The sediment budget is calculated as the change of volume in the beach-dune system (ΔVT)
over a relevant time scale (Tbud) and can either be negative (ΔVT < 0), stable (ΔVT � 0), or pos-
itive (ΔVT > 0),
DVT ¼
1
n
Xt¼i
(cid:0)
t¼i(cid:0) n
ð(cid:0) qSLR;t þ qLS;tÞDt þ Vnour;t
�
PLOS ONE | https://doi.org/10.1371/journal.pone.0215651 April 19, 2019
ð14Þ
9 / 30
Simulating beach and dune evolution at decadal to centennial scale under rising sea levels
where n = Tbud/Δt. Here all nourishments are considered in Vnour, also dune nourishments.
The time scale, Tbud, is in the order of years and should be long enough to represent long-term
trends and not seasonal variations. The proposed distribution scheme generates different dune
evolution depending on the sediment budget (Fig 4).
In the case of a positive sediment budget, ΔVT > 0, the ramp is filled until the ramp height
is equal to the dune height, S. After the ramp has filled up, if S < Smax, for a trapezoidal shape,
the sediment will be deposited on the crest, and for a triangular shape, the dune grows while
maintaining a constant dune crest length coordinate, y'L = y'S = constant, until Smax is reached.
If S � Smax, all sediment is deposited on the seaward side of the dune and the dune grows sea-
ward maintaining a trapezoidal shape.
For a stable sediment budget, ΔVT � 0, the ramp is filled until the ramp height is equal to
the dune height, S. Thereafter, in the case of a trapezoidal dune shape, a fraction of the sedi-
ment, As, is deposited on the crest and the fraction 1-As is deposited on the seaward side of the
dune. If the dune has a triangular shape, the dune grows in height while maintaining a con-
stant dune crest length coordinate, y'L = y'S = constant.
When the sediment budget is negative, ΔVT < 0, the ramp is filled until a critical ramp
height is reached. Thereafter, in the case of a trapezoidal dune shape, a fraction of the sedi-
ment, Ae, is deposited on the crest and the fraction 1–Ae is deposited on the landward side
of the dune. If the dune has a triangular shape, y'L = y'S, the dune grows on its landward
side.
The critical ramp height is here defined as 1 m below the crest level [44]. If the model is
applied in an area where local observations suggest otherwise, a site-specific value of the criti-
cal ramp height should be adopted. If there are no observations of sediment distribution across
the dune, the distribution coefficients, As and Ae, can be set to 0.5 based on the observations by
Ollerhead et al [36].
Fig 4. Sediment distribution scheme on a dune due to wind-blown sand.
https://doi.org/10.1371/journal.pone.0215651.g004
PLOS ONE | https://doi.org/10.1371/journal.pone.0215651 April 19, 2019
10 / 30
Simulating beach and dune evolution at decadal to centennial scale under rising sea levels
Impact of sea level rise
The Bruun Rule is used to calculate the required sediment transport from the beach volume,
Vbeach, to the active part of the profile to compensate for SLR. In agreement with previous
Bruun rule extensions [49–51], also landward transport is accounted for to compensate for an
elevation adjustment of the dry beach (Fig 5). Analogous to the original Bruun Rule concept,
the dry beach is assumed to be subject to hydrodynamic forcing capable of adjusting it to a
slowly rising sea level. The explanatory mechanism is that the beach is built by waves; with a
higher MSL the waves will build a higher beach. Therefore, the width of the active profile, B, is
here extended to encompass the horizontal distance from the seaward dune foot, yS, to the
depth of closure, DC.
The dune foot height is kept constant relative to MSL. Thus, the dune foot is moving up
with the same rate as SLR. Since the dune crest elevation is not changing, the lowest part of the
dune is per definition converted from dune volume, Vdune, to beach volume, Vbeach. Also the
depth of closure, DC, is constant, so that the lower vertical limit of the beach volume is moving
up with the same rate as SLR.
Vbeach is eroding to supply the active profile with sediment equivalent to the demand from
the increased accommodation space due to SLR. The difference between the old and new
Vbeach is denoted VSLR. Thus, the volume VSLR eroded from the beach is derived from the
Bruun Rule with the new definition of the active profile width, B,
VSLR ¼ RBruunðDF þ DCÞ ¼ SSLRB
ð15Þ
The sediment transport required to maintain the equilibrium profile, qSLR, is taken as the
time derivative of VSLR,
qSLR ¼
dVSLR
dt
¼
dSSLR
dt
B
ð16Þ
Consequently, there are two direct effects of SLR in the CS-model: first, adjustment of the
equilibrium profile to the new sea level by an offshore ‘Bruun rule’ transport, qSLR, from the
Fig 5. Impact of sea level rise on beach volume and dune height. The beach and dune volumes are for simplicity of
the visualisation schematized as rectangular shapes.
https://doi.org/10.1371/journal.pone.0215651.g005
PLOS ONE | https://doi.org/10.1371/journal.pone.0215651 April 19, 2019
11 / 30
Simulating beach and dune evolution at decadal to centennial scale under rising sea levels
beach volume Vbeach; and second, decrease of dune volume, Vdune, and height, S, due to an
upward shift of the reference SWL (Fig 5). The aeolian transport and morphological dune evolu-
tion scheme are indirectly impacted through qSLR (Eqs 4 and 14). Furthermore, the ‘Bruun rule’
transport is a sink for the beach volume, which will affect wave energy dissipation over the beach
(Eqs 2 and 3). This leads to an increase in the probability of dune erosion and overwash trans-
port, which are processes already included in the CS-model [15]. In this way, the dune may
translate landwards and the new scheme for morphological dune evolution allows the dunes to
adjust in height to the new sea level by keeping Smax constant relative the dune foot elevation.
Case study A¨ ngelholm Beach, Sweden
To demonstrate the capability of the proposed model to simulate long-term beach and dune evo-
lution, it was applied to a seven-year data set from A¨ ngelholm Beach in south Sweden (Fig 6).
Forcing
Wind data at 10 m height was collected from stations 1, 3, 6, and 7 (Fig 6), operated by the
Swedish Meteorological and Hydrological Institute (SMHI). Predominant wind direction is
west to southwest and the shore normal is oriented 310˚N (Fig 7).
Deep-water wave conditions at the mouth of Ska¨lderviken Bay were hindcasted from the
wind data using a modified SMB-method [22,62] (Fig 8). The largest computed significant
wave height and peak period is 5.3 m and 9.3 s, respectively, which occurred on 06/12/2013.
Water level data was collected from SMHI station 5 (Fig 6). The water levels were corrected
for local wind setup, following the method described by Palalane et al [22]. The current decade
has seen several extreme events compared to the period 1990–2010 (Fig 9). The highest
observed water level, adjusted for wind setup, was 185 cm above MSL on 27/11/2011. There is
no significant astronomical tide.
Morphology
Beach morphology varies alongshore, with higher dunes and a narrower beach in the north,
and lower dunes and a wider beach in the south. The observed long-term coastline evolution is
erosion in the north and accretion in the south (Fig 6). Profiles A, B, and C have been selected
Fig 6. Maps over the study area and measurement stations.
https://doi.org/10.1371/journal.pone.0215651.g006
PLOS ONE | https://doi.org/10.1371/journal.pone.0215651 April 19, 2019
12 / 30
Simulating beach and dune evolution at decadal to centennial scale under rising sea levels
Fig 7. Wind rose compiled with data from 1976–2017.
https://doi.org/10.1371/journal.pone.0215651.g007
to represent stretches of the beach with different long-term evolution (Fig 10). Analysis of
aerial photos since the 1940’s suggests that the vegetation line has been retreating about 0.3 m/
year in profile A, stable in profile B, and accreting about 0.3 m/year in profile C [22].
Topographic observations are available from 2010–2017. In Table 1, the dune volume
changes over time are presented together with the data source.
Storms causing dune erosion occurred in November 2011, December 2013, January 2015,
December 2015, and December 2016, of which the storms in 2011 and 2013 were the most
severe.
Since 2010, the dune volume has decreased in all three profiles, including the long-term sta-
ble and accreting profiles, B and C. This is mainly due to the storm erosion in December 2013
from which the dunes have not recovered. In profile A, the observed accretion in November
2014 and September 2015 is due to dune nourishments when sediment was scraped from just
seaward of the water line and deposited in the dune front. These operations were carried out
in April 2012, April 2014, and March 2015 to replace the sediment lost from the dunes during
storms.
In a bathymetric survey from October 2012 (provided by the Swedish Geological Survey)
no bars are observed in the subaqueous part of the studied profiles. However, from aerial
PLOS ONE | https://doi.org/10.1371/journal.pone.0215651 April 19, 2019
13 / 30
Simulating beach and dune evolution at decadal to centennial scale under rising sea levels
Fig 8. Hindcasted significant wave height at the mouth of Ska¨lderviken Bay during 1976–2017.
https://doi.org/10.1371/journal.pone.0215651.g008
photos, parallel bars can be detected south of profile C. Along the entire beach temporary cres-
centic subaqueous deposit are visible within 70 m from the shoreline. These features indicate
that there is a cross-shore exchange of sediment stored in nearshore deposits, although, no
fully developed bars are observed within the measured profiles.
Sediments
Sediment samples have been collected at four occasions during 2015 and 2016 to determine
D50 across the beach and dune (Fig 11).
The median grain size at sample points 1–5 on the foreshore and berm is in general decreas-
ing from profile A to profile C, indicating a longshore sorting. As the sediment is fairly well
graded, the grain size found in the dune, about 0.2–0.3 mm, is assumed to be representative
for dune building sediments. The samples from the dune in profile A with lower D50 are mixed
with silty material from the artificial fill, and is therefore not considered representative for aeo-
lian transported sediment.
In profile C, dune-building sediment was available at all sampling occasions, except for the
first in October 2015 when it was finer, with D50 below 0.2 mm. In Profile A, sediment of
appropriate grain size was only available on the last sampling occasion in October 2016. Profile
PLOS ONE | https://doi.org/10.1371/journal.pone.0215651 April 19, 2019
14 / 30
Simulating beach and dune evolution at decadal to centennial scale under rising sea levels
Fig 9. Yearly maximum SWL in 1976–2016 based on measurements from Viken (dashed grey line) and the yearly maximum corrected for local wind
setup in Ska¨lderviken Bay (solid black line).
https://doi.org/10.1371/journal.pone.0215651.g009
Fig 10. Profiles from DEM dated 12/04/2010, dashed lines display model schematization of the dune.
https://doi.org/10.1371/journal.pone.0215651.g010
PLOS ONE | https://doi.org/10.1371/journal.pone.0215651 April 19, 2019
15 / 30
Simulating beach and dune evolution at decadal to centennial scale under rising sea levels
Table 1. Dune volume changes between consecutive measurements. A summation of the changes since the first topographic observation in 2010 is given within
brackets.
Survey date
12/04/2010
30/11/2012
26/11/2014
21/01/2015
30/09/2015
15/12/2015
22/04/2016
27/10/2016
05/01/2017(A,C)
17/01/2017(B)
Volumetric change since previous survey in m3/m (accumulated change since
2010)
Profile B
Profile A
Profile C
initial profile
initial profile
initial profile
-2 (-2)
13 (11)
-28 (-17)
13 (-4)
-14 (-18)
3 (-15)
0 (-15)
-10 (-25)
-1 (-1)
-10 (-11)
-1 (-12)
0 (-12)
-2 (-14)
-1 (-15)
0 (-15)
1 (-14)
1 (1)
-10 (-9)
-2 (-11)
2 (-9)
-2 (-11)
0 (-11)
0 (-11)
-1 (-12)
Data source
DEM LiDAR (resolution 1×1 m)
DEM LiDAR (resolution 1×1 m)
DEM Photogrammetry (resolution 1×1 m)
DEM Photogrammetry (resolution 1×1 m)
Profile RTK-GPS
DEM Photogrammetry (resolution 1×1 m)
Profile RTK-GPS
Profile RTK-GPS
Profile RTK-GPS
https://doi.org/10.1371/journal.pone.0215651.t001
B had only complete samples from three of the occasions and then showed availability in some
of the sampling points along the profile, but not all.
Fig 11. Median grain size D50 for sediment samples in Profile A, B, and C sampled from the shoreline (1) to the dune
crest (7).
https://doi.org/10.1371/journal.pone.0215651.g011
PLOS ONE | https://doi.org/10.1371/journal.pone.0215651 April 19, 2019
16 / 30
Simulating beach and dune evolution at decadal to centennial scale under rising sea levels
Model setup
The initial profile was schematized so that the total sand volume in the subaerial part of the
profile was represented (Fig 10). Profile C, which consists of three dune rows, was schematized
into one dune, while preserving dune volume and seaward dune crest position. This procedure
is necessary as the CS-model can only simulate one row of dunes. The seaward dune slope
angle, βS, was set equal to the angle of repose of sand (approximately 30–34 ˚), whereas the
landward dune slope angle, βL, and maximum foreshore slope angle, βF,max, was obtained from
the topographic observations.
The relationship between the beach width, yG-yS, and the beach volume, Vberm, is deter-
mined from the topographic observations covering the profile from the dune foot (yS) to the
intersection with MSL (yG) (Fig 12). A linear function was least-square fitted to the data yield-
ing yG-yS = 1.2Vberm-4.3, with a coefficient of determination, R2, of 0.76.
The D50 used to calculate the fall velocity in the beach-bar exchange formulas [15] and
employed in the aeolian transport equations was set to 0.2 mm for all three profiles. The dune
foot height, DF, was estimated to 2 m for all profiles, based on an average of the topographic
observations.
The depth of closure, DC, was calculated to be 5.7 m according to Hallermeier [63] and the
width of the subaqueous part of the active profile was estimated to be 800 m (B = 800+yG-yS).
The average sea temperature is assumed to be about 10˚C [64].
The gradients in the longshore sediment transport rate was estimated to be 4.1 m3/m/year
in profile A, 0 m3/m/year in profile B, and -3.0 m3/m/year in profile C, based on the vegetation
line analysis [22]; thus, qLS equals -4.1, 0, and 3 m3/m/year for profile A, B, and C respectively.
The simulation time step was set to one hour.
Calibration and validation
The model was calibrated for the period 12/04/2010–21/04/2016 and then validated for the
period 22/04/2016–17/01/2017. Literature values were used for all empirical coefficients [22]
except for CS and CW, which are the most important calibration parameters for the dune
evolution.
Dune erosion was calibrated by adjusting the impact coefficient, CS, to fit the observed
dune erosion. CS describes the resistance of the dunes to wave erosion; the eroded volume
from the dune during wave attack is proportional to CS [15]. For profiles B and C, which
Fig 12. Observed relation between beach volume (Vberm) and beach width (yG-yS).
https://doi.org/10.1371/journal.pone.0215651.g012
PLOS ONE | https://doi.org/10.1371/journal.pone.0215651 April 19, 2019
17 / 30
Simulating beach and dune evolution at decadal to centennial scale under rising sea levels
involve dunes built up by aeolian transport with similar vegetation, CS was calibrated to a com-
mon value of 7.5�10−5. In profile A, the dune had initially aeolian transported sand, but was
after April 2012 partly nourished with a silty material. The impact coefficient, CS, was here cali-
brated to a larger value of 5�10−4, meaning that the dune is less resistant to erosion.
The empirical coefficient in the aeolian transport equation (Eq 7) CW was calibrated to
1.0�10−5, a common value for all profiles. Initial available sediment volume for aeolian trans-
port, Vw, was set to 0 m3/m in the calibration phase.
For this study, there was no information on the morphological evolution of the subaqueous
part of the profile. The initial bar volume, which represents subaqueous deposits interacting
with the beach deposits, was set to 20 m3/m for profile A, 70 m3/m for profile B, and 50 m3/m
for profile C, to reproduce the evolution of the beach volume. For the validation, the bar vol-
ume at the end of the calibration period was used as the initial value.
During the calibration period minor dune erosion events occur almost yearly and there are
two major storm events with significant dune erosion in November 2011 and December 2013
(Fig 13). The sudden increases in dune volume in profile A are due to the dune nourishment
episodes. In agreement with field observations, the aeolian transported volumes throughout
the simulation period are small compared to the eroded volumes during these storm events.
The dunes have not fully recovered at the end of the calibration period.
The validation period contains one storm event in December 2016, with a clear signal of
erosion in profile A and C (Fig 14). In profile B no erosion is observed, which is probably due
to an error in the measured topographic data.
Fig 13. Dune and beach volume evolution for calibration period.
https://doi.org/10.1371/journal.pone.0215651.g013
PLOS ONE | https://doi.org/10.1371/journal.pone.0215651 April 19, 2019
18 / 30
Simulating beach and dune evolution at decadal to centennial scale under rising sea levels
Fig 14. Dune and beach volume evolution for validation period.
https://doi.org/10.1371/journal.pone.0215651.g014
The performance of the calibration and validation is satisfactory. For the calibration period,
the coefficient of determination, R2, equals 0.73 for the dune volume evolution and 0.63 for the
berm volume evolution. For the validation period, R2 equals 0.69 for the dune evolution, but
for the berm evolution, R2 equals 0 because of few observations with only small volume
changes. The model sensitivity to the calibrated parameters was investigated by comparing a
modified model to the calibrated model. (Table 2). The mean absolute deviation of dune and
berm volume was calculated for a simulation period encompassing both the calibration and
validation period. The sensitivity analysis showed that the model was most sensitive to changes
Table 2. Result from sensitivity analysis of calibrated parameters.
Parameter
Modification
Mean absolute deviation from calibrated
model (m3/m)
Initial VW
Initial bar volume, Vbar,1
CW
CS
https://doi.org/10.1371/journal.pone.0215651.t002
+10 m3
+25%
-25%
+25%
-25%
+25%
-25%
Vdune
0.26
0.37
0.39
0.27
0.27
4.58
4.28
Vberm
0.14
1.24
1.22
0.17
0.19
1.59
1.66
PLOS ONE | https://doi.org/10.1371/journal.pone.0215651 April 19, 2019
19 / 30
Simulating beach and dune evolution at decadal to centennial scale under rising sea levels
in the duneha erosion impact coefficient, CS. Changes to the other parameters had only a
minor impact on the dune and berm volume evolution.
However, there are uncertainties associated both with the topographic observations and the
forcing. The beach profiles are derived from DEMs with resolution 1×1 m based on both
LiDAR data and photogrammetry, and topographic surveys with GPS are carried out by both
the local municipality and by researchers within this study. Therefore, it is uncertain whether
some of the minor observed topographic developments are due to actual morphologic changes
or artefacts related to different data acquisition methods. The eroded volumes under storms
are on some occasions overestimated, and on other occasions underestimated. This can partly
be due to both uncertainties in the topographic data and in the forcing.
In the dune erosion equation, the water level and wave properties both have large impact
on the eroded volume. The lack of in situ water level gauge data introduces uncertainty when
the water levels within the bay are corrected for the local wind setup. It is also possible that
local variation in the wave climate occurs within the study area, which is not considered here;
this could partly explain the large variation in the value of the impact coefficient, CS, between
profile A and profiles B-C. Other possible explanations can be the mix of sediment in the dune
in profile A, where the main part of the dune consists of sand with D50 around 0.2–0.3 mm
and the front part consists of finer, silty material from the dune nourishment; and further, that
vegetation is less abundant in profile A compared to B and C. The dune erosion formula used
in the CS-model [55] was developed for sandy dunes with D50 ranging from 0.15 to 0.5 mm,
and its validity outside this range of grain sizes is uncertain. Future coastal management
approaches in the area are unknown; possibly the sediment characteristics in the dune may
change. In the long-term simulations from 2018–2100, dune evolution in profile A is simulated
with both the high and low calibrated value of CS.
The simulated yearly average aeolian transport amounts to 0.3 m3/m/year in profile A, 0.5
m3/m/year in profile B and 0.6 m3/m/year in profile C (Fig 15), which is significantly lower
than in previous CS-model applications where a constant rate of about 10–11 m3/m/year was
applied in all three profiles [22]. The new results are more realistic when compared to the
observed long-term evolution. In the accreting profile C, the vegetation line has advanced 0.3
m/year since the 1940s [22], which is approximately equal to an accumulation of 0.6 m3/year
considering the average dune height of 2 m.
The simulation results, with the highest aeolian transport in profile C and lower transport
in profile A, are in line with the topographic observations. The simulated differences in the
Fig 15. Available sediment for aeolian transport (VW) and accumulated aeolian transport (qW) during the calibration and validation
periods.
https://doi.org/10.1371/journal.pone.0215651.g015
PLOS ONE | https://doi.org/10.1371/journal.pone.0215651 April 19, 2019
20 / 30
Simulating beach and dune evolution at decadal to centennial scale under rising sea levels
average annual transport rates are due to variations in sediment supply between the different
profiles as well as the variation in beach width. The four sets of sediment samples taken from
October 2015 to October 2016 display variations both in space and time (Fig 11), where profile
C has the most available sediment, profile B less, and profile A only intermittent availability
during the period 2010–2016. This pattern is reflected in the simulated available sediment (Fig
15). However, the simulation results show that there should be sediment available for aeolian
transport within all three profiles during the actual sampling period, which was not supported
by the sediment grain size data. An explanation for this could be that the material eroded from
the dune in profile A before the sampling period was finer than the appropriate grain size for
dune build up, due to the silty nourishment, and thus should not have been accounted for in
the available sediment volume, VW.
Long-term simulation with sea level rise
Long-term beach and dune evolution from 2018–2100 was simulated with a range of SLR sce-
narios from IPCC [65]. The simulations were run with 10 sets of 83 years of input data,
obtained by randomly shuffling years (July–June) from 01/07/1976–30/06/2016 with simulta-
neous values of wind, waves, and water levels relative to MSL. The three different scenarios of
SLR until 2100 were defined as mean low 0.44 m, mean high 0.74 m, and likely upper range
high 0.98 m, relative to the average MSL during the years 1986–2005. The first scenario corre-
sponds to the RCP 2.6 scenario and the latter two to the RCP 8.5 scenario in the fifth assess-
ment report by IPCC [65]. The SLR rate was derived from fitting quadratic polynomials to the
data provided by IPCC (Table AII.7.7 in [66]). A post glacial uplift of 1.3 mm/year [67] was
subtracted from the computed SLR rates and the scenarios were compared to a baseline sce-
nario with no relative SLR, i.e. where SLR compensates exactly the post-glacial uplift.
The regional climate change prognosis does not predict any changes to the wind climate
[67]; therefore no changes in wind speed or direction were considered in this study. The same
holds for the wave climate, which is dominated by regionally wind-generated waves.
Profile A was simulated with two different values on the impact coefficient, CS; the calibrated
value of 5�10−4, which is assumed to partly depend on the dune nourishments during the cali-
bration period, and the lower value of 7.5�10−5, which was estimated for profiles B and C.
In all profiles, dune volumes decrease for all three scenarios, including profile C despite
input of sediment due to a negative gradient in longshore transport (Fig 16). In Profile A, the
dune erodes away completely before year 2100, both with the large and small dune erosion
coefficient, CS. In the simulations with large CS, dune erosion is more important than the effect
of SLR, also without SLR the dune is eroded away quickly. Taking into account all SLR scenar-
ios, the dunes are eroded away completely within the period 2050–2075 with the lower CS coef-
ficient, compared to 2027–2037 with the higher CS coefficient. This evolution is explained by
the long-term trend of beach erosion and the severe storms in the last decade that are present
in the data set.
For profile B, the dune only disappears completely in the upper high SLR scenario, but the
volume is markedly diminished also in the other two scenarios. In Profile C, the effect of SLR
is compensated by the negative gradient in the longshore transport, which is added as a con-
stant transport rate to the beach volume. Still, the long-term simulation without SLR does not
show a net dune growth, which could be expected on an accreting beach. This can be explained
by the severe storm events in 2011 and 2013 that are present in the data set.
The long-term observation of dune evolution based on aerial photos from the 1940s con-
tains a period during 1960s and 1970s that was unusually calm with respect to storminess [68].
If the wind climate from 1976–2016 is representative for future conditions, model results
PLOS ONE | https://doi.org/10.1371/journal.pone.0215651 April 19, 2019
21 / 30
Simulating beach and dune evolution at decadal to centennial scale under rising sea levels
Fig 16. Simulated dune volume evolution. Solid lines represent mean value, and the dashed lines represent min and max values from the
10 simulations for each scenario.
https://doi.org/10.1371/journal.pone.0215651.g016
indicate that the aeolian transport capacity will not be sufficient to repair storm erosion. In
profiles B and C, the available sediment for aeolian transport accumulates faster than the aeo-
lian transport rate, the gradient in longshore transport cause the beach volume to increase and
the shoreline to extend seawards (Fig 17). The wider beach increases frictional losses, so that
dune erosion decreases which explains the trend change in dune volume evolution in profile C
for the low and baseline scenarios (Fig 16). The dune erosion and recovery during the simula-
tion period are subtracted from, or added to, the foredune. In reality, new foredune ridges
could form during periods of recovery, however, this is not possible with the present simplified
dune schematization of the CS-model.
The dunes are also losing sediment due to the change of reference dune foot elevation when
the sea level is rising, causing the dune crest height to decrease (Fig 17). In profile B the hori-
zontal dune crest erodes away completely and the dune attains a triangular shape, so that the
dune crest is further lowered due to storm erosion. None of the simulations of profile B and C
results in landward dune translation. In profile B, with no gradient in the longshore transport,
SLR changes the sediment budget from stable to eroding for all scenarios. Profile C is still
accreting in the low scenario but changes into a negative sediment budget in the mean high
and upper high scenarios. In the scenarios with negative sediment budget, the aeolian trans-
port capacity is too low to restore the dune ramp, which is supposed to produce landward
translation between consecutive erosion events.
PLOS ONE | https://doi.org/10.1371/journal.pone.0215651 April 19, 2019
22 / 30
Simulating beach and dune evolution at decadal to centennial scale under rising sea levels
Fig 17. Simulated beach–dune evolution for profile B and C. The ‘+’ indicates intersection with MSL.
https://doi.org/10.1371/journal.pone.0215651.g017
Discussion and concluding remarks
In this study, the CS-model was expanded to simulate beach and dune evolution at decadal to
centennial time scales under SLR. Model development was based on morphological concepts
from the literature, which were translated into mathematical formulations that were numeri-
cally solved. The proposed model was applied to a data set from A¨ ngelholm Beach, Sweden.
First, the model was calibrated and validated for a seven-year period, and then the beach and
dune evolution was simulated from 2018–2100 for a range of SLR scenarios.
Topography and grain size data suggests that the morphological evolution at A¨ ngelholm
Beach conform to the concepts that were introduced in the CS-model. The different profile
characteristics fit the conceptual sediment budget model by Psuty [24] where the eroding and
stable dunes in profiles A and B are the highest and the accumulating profile C displays a
lower, prograding dune-ridge topography. The differences in distribution of D50 over the pro-
files are in line with the hypothesis that available sediment is a controlling factor for aeolian
transport and that preconditions for aeolian transport are variable both in space and time.
However, although there was sediment available, the aeolian transport that was observed in
the stable and accreting profiles B and C were not of sufficient rates to repair the storm erosion
during the study period.
Overall, the results of the calibration and validation against field data were satisfactory and
the long-term predictions of beach and dune evolution under different SLR scenarios showed
reasonable results. The CS-model illustrated the relative importance of longshore transport
gradients, storm erosion, aeolian transport, and SLR for the profile evolution. The simulations
demonstrated the importance of extreme storm events for decadal-scale dune evolution, and
confirmed the need of integration of physical processes acting on multiple time scales [21].
At A¨ ngelholm Beach the aeolian transport capacity, rather than sediment availability, lim-
ited the dune growth in the long-term simulations. In the case with a positive sediment budget
(the low SLR scenario in profile C), the available sediment was accumulating faster than the
aeolian transport capacity, so that the shoreline extended seawards, widening the beach. For
the profiles with negative sediment budget, the dune ramp did not become fully developed
until the sediment available for aeolian transport had eroded away, suggesting that aeolian
transport capacity in relation to dune erosion rates is too low to result in landward dune
migration. The model assumes that the dune ramp is built entirely by wind-blown sand. Field
studies suggest that the main part of the ramp is built from wind-blown sand [36], however, in
PLOS ONE | https://doi.org/10.1371/journal.pone.0215651 April 19, 2019
23 / 30
Simulating beach and dune evolution at decadal to centennial scale under rising sea levels
future studies it would be worthwhile exploring if an avalanche mechanism should be included
in the model to facilitate the development of dune ramps.
Precipitation is not included in the model, but may be an important limiting factor [35]. At
beaches with seasonal ice and snow cover the effect on aeolian transport may be significant,
especially if transport is inhibited during periods with strong winds [36]. In areas with signifi-
cant impact of rain, snow, and ice, the model could be improved through modifying sediment
availability during wet or frozen conditions to restrict aeolian transport. Blowouts may also be
an important factor for landward foredune migration [40] but are difficult to include in a 1-D
model.
The CS-model assumes an equilibrium profile and therefore the Bruun rule fits well with
the model concept. Introducing a more realistic profile representation to the CS-model would
introduce a higher degree of model complexity, which is not desired. It is however important
to keep the limitations of the equilibrium concept in mind when applying the CS-model and
the Bruun rule. Furthermore, the depth of closure is time scale dependent. Over longer peri-
ods, the probability of storm events mobilizing sediment at deeper water increases [69]. There-
fore, the depth of closure adopted in the CS-model should be chosen with consideration to the
length of the simulation period.
In conclusion, the modified CS-model, including SLR related processes and a more
advanced transport and dune evolution scheme, provides fast and robust simulations with rea-
sonable results. The simplified model approach enables robust, long-term simulations and
makes the model suitable for probabilistic simulations. The results presented in this study are
promising, but more studies with longer-term data sets are required to validate the proposed
numerical formulations more thoroughly.
List of symbols
Ab Coefficient describing fraction of nourishments available for aeolian transport [–]
Ae Coefficient describing fraction of sediment deposited on dune crest in a negative sediment
budget [–]
Aq Coefficient describing fraction of longshore and ‘Bruun Rule’ transport available for aeolian
transport [–]
As Coefficient describing fraction of sediment deposited on dune crest in a stable sediment
budget [–]
AW Coefficient in critical shear velocity equation [–]
B Width of active profile [m]
BBruun Width of active profile in the original Bruun rule [m]
Bdry Dry beach width [m]
cf Empirical friction coefficient for wave runup [–]
CS Dune erosion impact coefficient [–]
CW Empirical coefficient in aeolian equilibrium transport formula [–]
d Grain size [mm]
D50 Median grain size [m]
D50
ref Median reference grain size [m]
PLOS ONE | https://doi.org/10.1371/journal.pone.0215651 April 19, 2019
24 / 30
Simulating beach and dune evolution at decadal to centennial scale under rising sea levels
DC Depth of closure [m]
DF Dune foot height [m]
F Fetch (aeolian transport) [m]
g Standard acceleration due to gravity [m/s2]
h Height of active profile [m]
Hm0 Energy based significant wave height [m]
Hrms Deep water root-mean-square wave height [m]
i Index, time step [–]
KW Empirical coefficient in aeolian equilibrium transport formula [–]
L0 Mean deep water wave length [m]
MSL Mean sea level [m]
mWE Potential aeolian transport rate [kg/s/m]
n Number of timesteps [–]
P Porosity [–]
qD Transport rate of eroded sediment from the dune [m3/s/m]
qL Transport rate of eroded sediment from the dune front to the landward side of the dune
[m3/s/m]
qLS Transport rate due to gradients in longshore transport [m3/s/m]
qS Transport rate of eroded sediment from the dune to the beach [m3/s/m]
qSLR Transport rate of sediment to compensate for sea level rise, ‘Bruun Rule’ transport [m3/s/
m]
qW Onshore component of aeolian transport [m3/s/m]
qWE Volumetric aeolian transport rate [m3/s/m]
qWF Potential aeolian transport rate corrected for fetch length [m3/s/m]
R Runup [m]
R' Runup corrected for friction over beach [m]
R2 Coefficient of determination [–]
RBruun Shoreline retreat due to sea level rise [m]
S Dune height [m]
Smax Maximum dune height limiting vertical dune growth for positive sediment budget [m]
Snew Dune height after SLR [m]
Sold Dune height before SLR [m]
SSLR Height of sea level rise [m]
SLR Sea level rise [–]
PLOS ONE | https://doi.org/10.1371/journal.pone.0215651 April 19, 2019
25 / 30
Simulating beach and dune evolution at decadal to centennial scale under rising sea levels
SWL Still water level [m]
t Time step [–]
Tbud Significant time scale for sediment budget [s]
u� Shear velocity [m/s]
u�
c Critical shear velocity [m/s]
uz Wind speed at z m elevation [m/s]
uz,c Critical wind speed at z m elevation [m/s]
Vbar Volume of subaqueous deposits [m3/m]
Vberm Volume of beach above mean sea level, between shoreline and dune foot [m3/m]
Vbeach Total beach volume [m3/m]
Vbeach, new Total beach volume after SLR [m3/m]
Vbeach, old Total beach volume before SLR [m3/m]
Vdune Dune volume [m3/m]
Vnour Nourishment volume [m3/m]
Vramp Dune ramp volume [m3/m]
VSLR Eroded volume from Vbeach, due to SLR [m3/m]
Vw Volume of sediment available for aeolian transport [m3/m]
x Horizontal travel distance of wave front [m]
y'L Landward dune crest length coordinate [m]
y'S Seaward dune crest length coordinate [m]
yG Shoreline length coordinate [m]
yL Landward dune foot length coordinate [m]
yR Runup limit length coordinate [m]
yS Seaward dune foot length coordinate [m]
z Wind gauge elevation [m]
z0 Roughness height [m]
βF Average beach slope angle [˚]
βF,max Maximum foreshore slope angle [˚]
βL Landward dune slope angle [˚]
βS Seaward dune slope angle [˚]
δ Empirical coefficient to account for fetch effect [m-1]
Δt Length of time step [s]
ΔVT Change of volume in beach- dune system to determine sediment budget [m3/m]
PLOS ONE | https://doi.org/10.1371/journal.pone.0215651 April 19, 2019
26 / 30
Simulating beach and dune evolution at decadal to centennial scale under rising sea levels
κ von Karman’s constant [–]
Θ Wind angle against shore normal [˚]
ρa Air density [kg/m3]
ρs Sand density [kg/m3]
Author Contributions
Conceptualization: Caroline Hallin.
Data curation: Caroline Hallin.
Formal analysis: Caroline Hallin.
Investigation: Caroline Hallin.
Methodology: Caroline Hallin.
Supervision: Magnus Larson, Hans Hanson.
Writing – original draft: Caroline Hallin.
Writing – review & editing: Magnus Larson, Hans Hanson.
References
1.
Louisse CJ, Van Der Meulen F. Future coastal defence in the Netherlands: strategies for protection and
sustainable development. J Coast Res. 1991; 7: 1027–1041.
2. Doody P, Ferreira M, Lombardo S, Lucius I, Misdorp R, Niesing H, et al. Living with Coastal Erosion in
Europe: Sediment and Space for Sustainability: Results from the EUROSION Study. [Internet]. Euro-
pean Commission, Luxembourg. 2004. Available: http://repository.tudelft.nl/view/hydro/
uuid:483327a3-dcf7-4bd0-a986-21d9c8ec274e/VN—readcube.com
3. Hanson H, Brampton A, Capobianco M, Dette HH, Hamm L, Laustrup C, et al. Beach nourishment proj-
ects, practices, and objectives—A European overview. Coast Eng. 2002; 47: 81–111. https://doi.org/
10.1016/S0378-3839(02)00122-9
4. Borsje BW, van Wesenbeeck BK, Dekker F, Paalvast P, Bouma TJ, van Katwijk MM, et al. How ecologi-
cal engineering can serve in coastal protection. Ecol Eng. Elsevier B.V.; 2011; 37: 113–122. https://doi.
org/10.1016/j.ecoleng.2010.11.027
5.
Leatherman SP, Zhang K, Douglas BC. Sea level rise shown to drive coastal erosion. EOS Trans Am
Geophys Union. 2000; 81: 55–57.
6. Wong PP, Losada IJ, Gattuso JP, Hinkel J, Khattabi A, McInnes KL, et al. Coastal systems and low-
lying areas. Climate Change 2014 Impacts, Adaptation and Vulnerability: Part A: Global and Sectoral
Aspects. 2015. https://doi.org/10.1017/CBO9781107415379.010
7.
Larson M, Kraus NC. SBEACH: Numerical model for simulating storm-induced beach change, report 1:
empirical foundation and model development. Tech Rep CERC-89-9 US Army Eng Waterw Exp Station
Coast Eng Res Center, Vicksburg, MS. 1989;
8. Roelvink D, Reniers A, van Dongeren A, van Thiel de Vries J, McCall R, Lescinski J. Modelling storm
impacts on beaches, dunes and barrier islands. Coast Eng. Elsevier B.V.; 2009; 56: 1133–1152. https://
doi.org/10.1016/j.coastaleng.2009.08.006
9. Vellinga P. Beach and dune erosion during storm surges. Doctoral Thesis, Delft University of Technol-
ogy, Delft, The Netherlands Publication 372, Delft Hydraulics. 1986. https://doi.org/10.1016/0378-3839
(82)90007-2
10.
11.
Luna MCM de M, Parteli EJR, Dura´n O, Herrmann HJ. Model for the genesis of coastal dune fields with
vegetation. Geomorphology. Elsevier B.V.; 2011; 129: 215–224. https://doi.org/10.1016/j.geomorph.
2011.01.024
van Boxel JH, Arens SM, van Dijk PM. Aeolian processes across transverse dunes. I: Modelling the air
flow. Earth Surf Process Landforms. 1999; 24: 255–270. https://doi.org/10.1002/(SICI)1096-9837
(199903)24:3<255::AID-ESP962>3.0.CO;2–3
PLOS ONE | https://doi.org/10.1371/journal.pone.0215651 April 19, 2019
27 / 30
Simulating beach and dune evolution at decadal to centennial scale under rising sea levels
12.
van Dijk PM, Arens SM, van Boxel JH. Aeolian processes across transverse dunes. II: Modelling the
sediment transport and profile development. Earth Surf Process Landforms. 1999; 24: 319–333.
13. Hanson H, Kraus NC. GENESIS: Generalized model for simulating shoreline change. 1989.
14. Stive MJF, Aarninkhof SGJ, Hamm L, Hanson H, Larson M, Wijnberg KM, et al. Variability of shore and
shoreline evolution. Coast Eng. 2002; 47: 211–235. https://doi.org/10.1016/S0378-3839(02)00126-6
15.
Larson M, Palalane J, Fredriksson C, Hanson H. Simulating cross-shore material exchange at decadal
scale. Theory and model component validation. Coast Eng. Elsevier B.V.; 2016; 116: 57–66. https://doi.
org/10.1016/j.coastaleng.2016.05.009
16. Vitousek S, Barnard PL, Limber P, Erikson L, Cole B. A model integrating longshore and cross-shore
processes for predicting long-term shoreline response to climate change. J Geophys Res Earth Surf.
2017; 122: 782–806. https://doi.org/10.1002/2016JF004065
17. Robinet A, Idier D, Castelle B, Marieu V. A reduced-complexity shoreline change model combining long-
shore and cross-shore processes: the LX-Shore model. Environ Model Softw. Elsevier; 2018; 109: 1–
16. https://doi.org/10.1016/j.envsoft.2018.08.010
18. Cohn N, Hoonhout BM, Goldstein EB, Vries S De. Exploring Marine and Aeolian Controls on Coastal
Foredune Growth Using a Coupled Numerical Model. J Mar Sci Eng. 2019; 7: 1–25. https://doi.org/10.
3390/jmse7010013
19. Dura´ n O, Moore LJ. Vegetation controls on the maximum size of coastal dunes. Proc Natl Acad Sci U S
A. 2013; 110: 17217–22. https://doi.org/10.1073/pnas.1307580110 PMID: 24101481
20. Hoonhout BM, de Vries S. A Process-based Model for Aeolian Sediment Transport and Spatiotemporal
Varying Sediment Availability. J Geophys Res. 2016; https://doi.org/10.1002/2015JF003692
21.
French J, Payo A, Murray B, Orford J, Eliot M, Cowell P. Appropriate complexity for the prediction of
coastal and estuarine geomorphic behaviour at decadal to centennial scales. Geomorphology. 2016;
256: 3–16. https://doi.org/10.1016/j.geomorph.2015.10.005
22. Palalane J, Fredriksson C, Marinho B, Larson M, Hanson H, Coelho C. Simulating cross-shore material
exchange at decadal scale. Model application. Coast Eng. Elsevier B.V.; 2016; 116: 26–41. https://doi.
org/10.1016/j.coastaleng.2016.05.009
23.
de Vries S, Southgate HN, Kanning W, Ranasinghe R. Dune behavior and aeolian transport on decadal
timescales. Coast Eng. Elsevier B.V.; 2012; 67: 41–53. https://doi.org/10.1016/j.coastaleng.2012.04.
002
24. Psuty NP. Sediment budget and dune / beach interaction. J Coast Res. 1988; 3: 1–4.
25. Sherman DJ, Bauer O. Dynamics of beach-dune systems. Prog Phys Geogr. 1993; 17: 413–447.
26. Bagnold RA. The transport of sand by wind. Geogr J. 1937; 89: 409–438.
27. Hesp P. Surfzone, beach and foredune interactions on the Australian southest coast. J Coast Res.
1988; 3: 15–25.
28. Barchyn TE, Martin R, Kok J, Hugenholtz CH. Fundamental mismatches between measurements and
models in aeolian sediment transport prediction: The role of small-scale variability. Aeolian Res. Else-
vier B.V.; 2014; 15: 245–251. https://doi.org/10.1016/j.aeolia.2014.07.002
29. Hsu S-A. Wind stress criteria in eolian sand transport. J Geophys Res. 1971; 76: 8684–8686.
30. Kawamura R. Study on Sand Movement by Wind. Reports Phys Sci Res Inst Tokyo Univ. 1951; 5(3–4):
95–112.
31.
Lettau K, Lettau H. Experimental and micrometeorological field studies of dune migration. In: Lettau H,
Lettau K, editors. Exploring the World’s Driest Climate. University of Wisconsin-Madison; 1978.
32. Owen P. Saltation of uniform grains in air. J Fluid Mech. 1964; 20: 225–242.
33.
Zingg A. Wind tunnel studies of the movement of sedimentary material. Proc 5th Hydraul Conf Bull.
1953; 34: 111–135.
34. Sherman DJ, Jackson DWT, Namikas SL, Wang J. Wind-blown sand on beaches: an evaluation of
models. Geomorphology. 1998; 22: 113–133. https://doi.org/10.1016/S0169-555X(97)00062-7
35. Bauer BO, Davidson-Arnott RGD, Hesp P, Namikas SL, Ollerhead J, Walker IJ. Aeolian sediment trans-
port on a beach: Surface moisture, wind fetch, and mean transport. Geomorphology. Elsevier B.V.;
2009; 105: 106–116. https://doi.org/10.1016/j.geomorph.2008.02.016
36. Ollerhead J, Davidson-Arnott R, Walker IJ, Mathew S. Annual to decadal morphodynamics of the fore-
dune system at Greenwich Dunes, Prince Edward Island, Canada. Earth Surf Process Landforms.
2013; 38: 284–298. https://doi.org/10.1002/esp.3327
37. Bagnold RA. The physics of blown sand and desert dunes. London: Meuthen; 1941.
38. Narra P, Coelho C, Fonseca J. Sediment grain size variation along a cross-shore profile–representative
d50. J Coast Conserv. 2015; 19: 307–320. https://doi.org/10.1007/s11852-015-0392-x
PLOS ONE | https://doi.org/10.1371/journal.pone.0215651 April 19, 2019
28 / 30
Simulating beach and dune evolution at decadal to centennial scale under rising sea levels
39. McLaren P, Bowles D. The effects of sediment transport on grain-size distributions. J Sediment Petrol.
1985; 55: 457–470. https://doi.org/10.1306/212F86FC-2B24-11D7-8648000102C1865D
40. Hesp P. Foredunes and blowouts: initiation, geomorphology and dynamics. Geomorphology. 2002; 48:
245–268. https://doi.org/10.1016/S0169-555X(02)00184-8
41. Moore LJ, Dura´ n Vinent O, Ruggiero P. Vegetation control allows autocyclic formation of multiple dunes
on prograding coasts. Geology. 2016; 44: 559–562. https://doi.org/10.1130/G37778.1
42. Davidson-Arnott R, Hesp P, Ollerhead J, Walker I, Bauer B, Delgado-Fernandez I, et al. Sediment Bud-
get Controls on Foredune Height: Comparing Simulation Model Results with Field Data. Earth Surf Pro-
cess Landforms. 2018; https://doi.org/10.1002/esp.4354
43. Kuriyama Y, Mochizuki N, Nakashima T. Influence of vegetation on aeolian sand transport rate from a
backshore to a foredune at Hasaki, Japan. Sedimentology. 2005; 52: 1123–1132. https://doi.org/10.
1111/j.1365-3091.2005.00734.x
44. Christiansen MB, Davidson-Arnott R. Rates of Landward Sand Transport over the Foredune at Skallin-
gen, Denmark and the Role of Dune Ramps. Geogr Tidsskr J Geogr. 2004; 104: 31–43. https://doi.org/
10.1080/00167223.2004.10649502
45. Bruun P. Sea level rise as a cause of shore erosion. J Waterw Harb Div. 1962;ASCE 88: 117–130.
46. Dean RG. Additional sediment input to the nearshore region. Shore and Beach. 1987; 76–81.
47. Dean RG, Houston JR. Determining shoreline response to sea level rise. Coast Eng. Elsevier B.V.;
2016; 114: 1–8. https://doi.org/10.1016/j.coastaleng.2016.03.009
48. Houston JR. Shoreline Response to Sea-Level Rise on the Southwest Coast of Florida. J Coast Res.
2015; 314: 777–789. https://doi.org/10.2112/JCOASTRES-D-14-00161.1
49. Rosati JD, Dean RG, Walton TL. The modified Bruun Rule extended for landward transport. Mar Geol.
2013; 340: 71–81. https://doi.org/10.1016/j.margeo.2013.04.018
50. Wolinsky MA, Brad Murray A. A unifying framework for shoreline migration: 2. Application to wave-dom-
inated coasts. J Geophys Res Earth Surf. 2009; 114: 1–13. https://doi.org/10.1029/2007JF000856
51. Davidson-Arnott RGD. Conceptual Model of the Effects of Sea Level Rise on Sandy Coasts. J Coast
Res. 2005; 216: 1166–1172. https://doi.org/10.2112/03-0051.1
52. Cooper JAG, Pilkey OH. Sea-level rise and shoreline retreat: Time to abandon the Bruun Rule. Glob
Planet Change. 2004; 43: 157–171. https://doi.org/10.1016/j.gloplacha.2004.07.001
53. Atkinson AL, Baldock TE, Birrien F, Callaghan DP, Nielsen P, Beuzen T, et al. Laboratory investigation
of the Bruun Rule and beach response to sea level rise. Coast Eng. Elsevier Ltd; 2018; 136: 183–202.
https://doi.org/10.1016/j.coastaleng.2018.03.003
54. Dean RG. Equilibrium beach profiles: U.S. Atlantic and Gulf coasts. Dep Civ Eng Ocean Eng Rep No
12, Univ Delaware, Newark, Delaware. 1977;
55.
Larson M, Erikson L, Hanson H. An analytical model to predict dune erosion due to wave impact. Coast
Eng. 2004; 51: 675–696. https://doi.org/10.1016/j.coastaleng.2004.07.003
56. Hanson H, Larson M, Kraus NC. Calculation of beach change under interacting cross-shore and long-
shore processes. Coast Eng. Elsevier B.V.; 2010; 57: 610–619. https://doi.org/10.1016/j.coastaleng.
2010.02.002
57. Belly P-Y. Sand movement by wind; Technical Memorandum No. 1, US. Army Corps of Engineers.
1964.
58.
Fryberger SG, Dean G. Dune Forms and Wind Regimes. In: McKee ED, editor. A Study of Global Sand
Seas. Reston, VA, USA: U.S. Geological Survey; 1979. pp. 137–169.
59. Davidson-Arnott RGD, Law MN. Seasonal patterns and controls on sediment supply to coastal fore-
dunes, Long Point, Lake Erie. Coastal Dunes: Form and Process. 1990. pp. 177–200.
60. Sauermann G, Kroy K, Herrmann HJ. Continuum saltation model for sand dunes. Phys Rev E Stat Non-
lin Soft Matter Phys. 2001; 64: 031305. https://doi.org/10.1103/PhysRevE.64.031305 PMID: 11580330
61. Bascom WN. The relationship between sand size and beach-face slope. Trans Am Geophys Union.
1951; 32: 866–874.
62. USACE (US Army Corps of Engineers). Shore Protection Manual (SPM). Washington: U.S. Govern-
ment Printing Office; 1984.
63. Hallermeier RJ. A profile zonation for seasonal sand beaches from wave climate. Coast Eng. 1981; 4:
253–277. https://doi.org/10.1016/0378-3839(80)90022-8
64. MacKenzie BR, Schiedek D. Long-term sea surface temperature baselines-time series, spatial covaria-
tion and implications for biological processes. J Mar Syst. 2007; 68: 405–420. https://doi.org/10.1016/j.
jmarsys.2007.01.003
PLOS ONE | https://doi.org/10.1371/journal.pone.0215651 April 19, 2019
29 / 30
Simulating beach and dune evolution at decadal to centennial scale under rising sea levels
65. Church JA, Clark PU, Cazenave A, Gregory JM, Jevrejeva S, Levermann A, et al. Sea level change.
Clim Chang 2013 Phys Sci Basis Contrib Work Gr I to Fifth Assess Rep Intergov Panel Clim Chang.
2013; 1137–1216. https://doi.org/10.1017/CB09781107415315.026
66. Prather M, Flato G, Friedlingstein P, Jones C, Lamarque J-F, Liao H, et al. IPCC, 2013: Annex II: Cli-
mate System Scenario Tables. Clim Chang 2013 Phys Sci Basis Contrib Work Gr I to Fifth Assess Rep
Intergov Panel Clim Chang. 2013;
67. Persson G, Sjo¨ kvist E, Åstro¨m S, Eklund D, Andre´ asson J, Johnell A, et al. Rapport Nr 2011–52 Klima-
tanalys fo¨ r Skåne la¨ n [in Swedish]. 2012.
68. BACC II Author Team. Second assessment of climate change for the Baltic Sea Basin. 2015. https://
doi.org/10.1007/978-3-319-16006-1
69. Nicholls RJ, Birkemeier WA, hong Lee G. Evaluation of depth of closure using data from Duck, NC,
USA. Mar Geol. 1998; 148: 179–201. https://doi.org/10.1016/S0025-3227(98)00011-5
PLOS ONE | https://doi.org/10.1371/journal.pone.0215651 April 19, 2019
30 / 30
|
10.1371_journal.pone.0221824 | RESEARCH ARTICLE
The extraordinary osteology and functional
morphology of the limbs in Palorchestidae, a
family of strange extinct marsupial giants
Hazel L. RichardsID
W. Adams2,5
1,2*, Rod T. Wells3,4, Alistair R. Evans1,2, Erich M. G. Fitzgerald2, Justin
1 School of Biological Sciences, Monash University, Clayton, Victoria, Australia, 2 Geosciences, Museums
Victoria, Melbourne, Victoria, Australia, 3 Ecology and Evolution, College of Science and Engineering,
Flinders University, Adelaide, South Australia, Australia, 4 Palaeontology, South Australian Museum,
Adelaide, South Australia, Australia, 5 Department of Anatomy & Developmental Biology, School of
Biomedical Sciences, Faculty of Medicine, Nursing and Health Sciences, Monash University, Clayton,
Victoria, Australia
* hazel.l.richards@gmail.com
Abstract
The Palorchestidae are a family of marsupial megafauna occurring across the eastern Aus-
tralian continent from the late Oligocene through to their extinction in the Late Pleistocene.
The group is known for their odd ‘tapir-like’ crania and distinctive clawed forelimbs, but their
appendicular anatomy has never been formally described. We provide the first descriptions
of the appendicular skeleton and body mass estimates for three palorchestid species, pre-
senting newly-identified, and in some cases associated, material of mid-Miocene Propa-
lorchestes, Plio-Pleistocene Palorchestes parvus and Pleistocene Palorchestes azael
alongside detailed comparisons with extant and fossil vombatiform marsupials. We propose
postcranial diagnostic characters at the family, genus and species level. Specialisation in
the palorchestid appendicular skeleton evidently occurred much later than in the cranium
and instead correlates with increasing body size within the lineage. We conclude that
palorchestid forelimbs were highly specialised for the manipulation of their environment in
the acquisition of browse, and that they may have adopted bipedal postures to feed. Our
results indicate palorchestids were bigger than previously thought, with the largest species
likely weighing over 1000 kg. Additionally, we show that P. azael exhibits some of the most
unusual forelimb morphology of any mammal, with a uniquely fixed humeroulnar joint unlike
any of their marsupial kin, living or extinct.
Introduction
Australia’s marsupial megafauna became largely extinct during the Late Pleistocene, likely due
to a combination of climatic change and human activity [1, 2]. The Vombatomorphia were an
adaptively-diverse clade that once included carnivorous ambush predators (thylacoleonids),
giant arboreal specialists (Nimbadon spp.), and the largest marsupial to have ever lived,
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: Richards HL, Wells RT, Evans AR,
Fitzgerald EMG, Adams JW (2019) The
extraordinary osteology and functional morphology
of the limbs in Palorchestidae, a family of strange
extinct marsupial giants. PLoS ONE 14(9):
e0221824. https://doi.org/10.1371/journal.
pone.0221824
Editor: Julien Louys, Griffith University,
AUSTRALIA
Received: May 8, 2019
Accepted: August 15, 2019
Published: September 13, 2019
Copyright: © 2019 Richards et al. This is an open
access article distributed under the terms of the
Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: All relevant data are
within the paper and its Supporting Information
files.
Funding: This research was supported by an
Australian Government Research Training Program
Scholarship, Robert Blackwood Partnership PhD
Top-Up Scholarship (HLR), internal funds from the
Department of Anatomy and Developmental
Biology, Monash University (JWA), and the
Australian Research Council DP180101797 (ARE).
PLOS ONE | https://doi.org/10.1371/journal.pone.0221824 September 13, 2019
1 / 69
Osteology and functional morphology of the limbs in Palorchestidae
The funders had no role in study design, data
collection and analysis, decision to publish, or
preparation of the manuscript.
Competing interests: ARE is a Section Editor of
Paleontology and JWA is an Academic Editor with
PLOS ONE; however, they confirm that their
positions do not alter the authors’ adherence to the
PLOS ONE policies. They and the other authors
have declared that no other competing interests
exist. This does not alter our adherence to PLOS
ONE policies on sharing data and materials.
Abbreviations: AM F, Australian Museum fossil
collection; FU, Flinders University collection;
NHMUK, Natural History Museum, London; MZ,
Monash University zoology collection; NMV C,
Museums Victoria mammal collection; NMV P,
Museums Victoria Palaeontology Collection; NTM,
Museum and Art Gallery of the Northern Territory;
SAMA P, South Australian Museum palaeontology
collection.
Diprotodon optatum (see Table 1). Perhaps least understood among these vombatomorphs are
the Palorchestidae. Palorchestids were a sister family to the better-known Diprotodontidae,
whose closest living relatives are wombats (Vombatidae) and, more distantly, koalas (Phasco-
larctidae) [3–8]. These giant quadrupeds possessed a suite of morphological features totally
unlike any living mammal, including a skull with hyper-retracted nasals, accompanied by pow-
erful specialised forelimbs equipped with enormous scimitar-like claws. This led palaeontolo-
gists to speculate about ecological similarities between palorchestids and other large extinct
clawed mammals like ground sloths and chalicotheres [9, 10], an ecomorph notably absent
from modern faunas [11]. Despite an assortment of postcranial material identified in Austra-
lian museum collections since the 1970s [9, 10], historical emphasis towards craniodental
research has meant that even basic descriptions of the palorchestid appendicular skeleton have
yet to be published.
Here we present the first formal descriptions of palorchestid appendicular morphology,
highlighting morphological trends within the lineage over evolutionary time and proposing
diagnostic characters at the family, genus and species level where possible. We also provide the
first quantitative body mass estimates for the group. This work represents an important first
step toward further taxonomic, biomechanical and palaeoecological analyses of these unusual
marsupials.
Historic and current understanding of palorchestids
Palorchestids have been recorded at sites across eastern mainland Australia and Tasmania,
and as a group occurred from the late Oligocene though to the major megafaunal extinction in
the Late Pleistocene [7, 25, 30, 31]. Their narrow face, scooped incisors and bilophodont molar
morphology suggest a selective browser diet, supported by dental microwear analysis of a sin-
gle P. azael molar [32], while a specialised bark-feeding niche has also been proposed [10].
Palorchestid body mass has been roughly approximated as ‘lamb-sized’ in the earliest Propa-
lorchestes species [28], 100 kg in the Plio-Pleistocene P. parvus, and 500 kg in the largest, most
abundant and latest-surviving Palorchestes azael [29] (Table 1).
Despite being known principally from craniodental material, speculative reconstructions of
palorchestid body form abound in the popular literature (reviewed in [33]). The form of such
reconstructions has varied widely in the 145 years since the initial description of the type spe-
cies Palorchestes azael. Owen’s [34] original misinterpretation of the macropod-like midlinks
between molar lophs in the fragmentary holotype (NHMUK PV OR 46316) led to assumptions
of a kangaroo-like body [35]. Eventual taxonomic reassessment by Woods [3], noting the lack
of a diagnostic macropodine masseteric canal, placed them within the Diprotodontoidea,
which suggested a more vombatomorph body shape. More recent reconstructions often
resemble a ‘marsupial tapir’ largely due to the elongate palorchestid snout and autapomorphic
retracted nasal bones bearing a superficial resemblance to those of tapirs [25]. Recent discovery
of new fossils and subsequent re-examination of all known palorchestid cranial material have
challenged this longstanding tapir analogy, omitting the hypothesised proboscis in favour of a
sensitive prehensile lip accompanied by a long protrusible tongue [36]. Whatever its fleshy
appearance, this unique rostral morphology was already evident in the earliest-known palorch-
estid crania, indicating early cranial specialisation at some point after their divergence from
the Diprotodontidae [28, 37].
Palorchestid fossil finds are exceptionally rare, but have occurred over a large geographic
range of what would have been forested environments [38, 39]. As a result, some have specu-
lated they may have been solitary animals with slow reproductive rates and large home ranges
[10, 40]. This can be contrasted with the fossils of other large diprotodontoids, such as
PLOS ONE | https://doi.org/10.1371/journal.pone.0221824 September 13, 2019
2 / 69
Osteology and functional morphology of the limbs in Palorchestidae
Table 1. Relationships, occurrences, appendicular osteology sources and previously published mass estimates for comparative vombatiform taxa.
Taxonomy
Stratigraphic range
Appendicular osteology
Published mass estimate (kg)
Vombatiformes
Phascolarctomorphia
Phascolarctidae
Phascolarctos
cinereus
Quaternary
Lee and Carrick [12]
Vombatomorphia
Thylacoleonidae†
Vombatidae
Diprotodontoidea†
Diprotodontidae†
Zygomaturinae
Diprotodontinae
Palorchestidae†
Thylacoleo
Phascolonus†
Lasiorhinus
Vombatus
carnifex
gigas
latifrons
ursinus
Plio-Pleistocene
Finch and Freedman [14]
Pleistocene
Quaternary
Quaternary
Stirling [17]
Scott and Richardson [18]
Owen [19]
Nimbadon
Neohelos
Zygomaturus
Ngapakaldia
Diprotodon
Propalorchestes
Palorchestes
lavarackorum
stirtoni
trilobus
tedfordi
bonythoni
optatum
sp.
parvus
azael
Miocene
Miocene
Pleistocene
Oligo-Miocene
Oligo-Miocene
Pleistocene
Oligo-Miocene
Plio-Pleistocene
Pleistocene
Black et al. [20]
Murray et al. [21]
Scott [23]
Munson [24]
Munson [24]
Owen [26]
This study
This study
This study
12[13]
87[15, 16]
250[6, 16]
25[13]
26[13]
70[16, 20]
250[16, 22]
675[16, 22]
‘sheep sized’ [25]�
‘sheep sized’[25]�
2786[27]
‘lamb sized’[28]�
100[29]�
500[29]�
† indicates extinct taxa. Taxonomy and occurrences adapted from Murray [6], Black [7] and Black et al. [5] (see [8] for an alternative position for Thylacoleonidae).
Mass estimation method used by these authors was that of Anderson et al. [16] using stylopodial circumferences, and are means except where taxa are extant, or � where
no method was reported.
https://doi.org/10.1371/journal.pone.0221824.t001
Nimbadon [41] and Diprotodon [42], which are often recovered in groups of many individuals
of various ontogenetic stages. Several sympatric palorchestid species occur at multiple localities
and time periods across eastern Australian deposits [30, 39, 40, 43], and recently named addi-
tional species show that the evolution within the lineage was more complex than previously
thought [38, 44–46].
The only palorchestid postcrania so far described as such in the peer-reviewed literature are
a series of six caudal vertebrae assigned to P. azael [47], indicating the likelihood of a well-
developed tail in that species. From later excavations at Victoria Fossil Cave, Wells [9] reported
finding huge laterally-compressed ungual phalanges in association with some limb material
distinct from those of other megafauna. These were tentatively assigned to Palorchestes but
never formally described. In the popular book Kadimakara, Flannery and Archer [10] recount
their recognition of an unidentified partial skeleton in the collection at Museums Victoria as
P. azael based on diagnostic associated molars. Although this specimen (NMV P157144) was
in poor condition and lacked locality data, its distinctive limb elements enabled them to iden-
tify another set of postcrania from Buchan, Victoria (NMV P159792) due to key morphologi-
cal overlap between the two. This then led to recognition of a further palorchestid specimen
recovered from Wee Jasper, New South Wales (Australian Museum AM F58870), which was
subsequently assigned to P. parvus on the basis of its smaller size than the Victorian specimens,
and association with a diagnostic premaxilla fragment. Although Flannery and Archer’s chap-
ter provides a tantalising glimpse of the unique morphology of these limb bones, no photo-
graphs or formal descriptions have been presented.
Geological settings
Table 2 provides an overview of the localities and geological contexts in which the main associ-
ated material described here was found.
PLOS ONE | https://doi.org/10.1371/journal.pone.0221824 September 13, 2019
3 / 69
Osteology and functional morphology of the limbs in Palorchestidae
Table 2. Geological settings for the main associated palorchestid material described.
Locality
Locality description
Formation
Age
Buchan Caves,
Victoria
Buchan Caves Reserve, East
Gippsland, 370 km east of
Melbourne
Keilor,
Victoria
Confluence of Maribyrnong
River and Dry Creek,
approximately 16 km northwest
of Melbourne
Caves are formed within Early Devonian
Buchan Group limestone [48]. Pleistocene
deposits found both within soft terra rossa
sediments lining the cave floor and
embedded in calcite flowstone of the walls
and ceiling [49]
D Clay fossil bearing unit of the Arundel
Formation, composed of fine-grained
overbank silts [50, 51]
Uranium-thorium dating of calcite flowstone
surrounding a Palorchestes azael skull yielded a
date of 275 ka [49]
The 40–25 ka date range estimated by Joyce and
Anderson [52] for the D Clay was questioned by
Duncan [51] due to likely contamination of
samples. They are currently assumed to be older
than 31 ka
Taxa
Palorchestes
azael
P. parvus
P. azael
Wee Jasper,
New South
Wales
Either Punchbowl or Signature
cave in the Wee Jasper system,
approximately 40 km north west
of Canberra
Cave within Early Late Devonian Taemas
Limestone formation, consisting of
fossiliferous interbedded limestone and
calcareous mudstone [53]
Sediments determined through
magnetostratigraphy to date between 2.0 and 0.73
Ma [54]. One megafaunal fossil was dated to at
least 45–30 ka via U-series analysis [55]
P. parvus
Bullock Creek,
Northern
Territory
In a river valley situated on
Camfield Station, approximately
500 km south of Darwin
https://doi.org/10.1371/journal.pone.0221824.t002
Camfield Limestone Beds freshwater
carbonate unit
Biocorrelation and mammalian stage-of-evolution
comparisons constrain the site to Mid-Miocene,
between 16–13 Ma [22, 31, 55–57]
Propalorchestes
sp.
Methods
Descriptions and comparisons
Here we collate and describe the appendicular material for Palorchestes azael, P. parvus and
Propalorchestes accessible across five Australian museum collections (with one specimen from
Natural History Museum, London). We figure the best-preserved examples of the representa-
tive morphology throughout. Over the course of this research additional material has been rec-
ognised in collections and referred where possible to the species level.
Palorchestes azael, the type species and that with the most known material, is first described
in detail relative to other vombatiform marsupials. Palorchestes parvus and Propalorchestes are
then described in terms of key differences to the type species.
Where possible, apomorphic diagnostic features at the family, genus and species level are
proposed for palorchestid species, although full character state and phylogenetic analyses are
beyond the scope of the present study. These emended diagnoses are aggregated at the begin-
ning of the Systematic Palaeontology section to maximise accessibility of this information for
the reader.
Throughout the text, comparisons are made with other vombatiform marsupials referred to
in-text by generic names except for Ngapakaldia, where two species are referred to as Ng.
bonythoni and Ng. tedfordi. The specific comparative taxa used and their relationships to
palorchestids are detailed in Table 1 along with literature sources for their own appendicular
osteological descriptions. Museum registration details for comparative vombatiform speci-
mens are provided in S1 Table, and key limb elements from these comparative taxa are illus-
trated in S1–S3 Figs.
Select measurements are given in-text, with comprehensive measurements of all palorches-
tid specimens also available in S1 Table. Measurements < 150 mm were taken with digital cal-
lipers to the nearest tenth of a millimetre. Measurements > 150 mm were taken with vinyl
tape to the nearest millimetre.
Osteological, myological and directional terminology follow Illustrated Veterinary Anatom-
ical Nomenclature [58] except when inappropriate for marsupials, where nomenclature follows
recent descriptive work by Harvey and Warburton [59], Warburton et al. [60] and Warburton
and Marchal [61].
PLOS ONE | https://doi.org/10.1371/journal.pone.0221824 September 13, 2019
4 / 69
Osteology and functional morphology of the limbs in Palorchestidae
Body mass estimation
The body masses of many extinct vombatiform taxa have previously been estimated using an
equation derived by Anderson et al. [16], where minimum midshaft circumferences of the
humerus and femur were used to predict mammalian mass (Table 1). The relationship between
stylopodial circumference and body mass is thought to be highly conserved due to the funda-
mental weight-bearing role of these elements in quadrupedal terrestrial animals, regardless of
posture [62–64]. However, such circumference-based methods may yield inflated mass esti-
mates for taxa that have stylopodia of particularly irregular (non-circular) section profile (see
Megatherium femora in Casinos [65]). Campione and Evans [63], in their work revisiting the
Anderson et al.[16] method, concluded that their predictive equation based on combined
humeral and femoral circumferences was inappropriate for highly fossorial species such as
moles, due to their exceedingly apomorphic humeral morphology. We anticipate similar issues
with the Palorchestes azael humerus, as its minimum circumference is distorted by distally-posi-
tioned pectoral and proximally-extensive supracondylar crests projecting from the diaphysis.
To assess this, we generated predictive equations based on humeral, femoral and combined cir-
cumferences so mass estimates could be compared. As per the Campione and Evans [63] method,
we measured minimum humeral and femoral circumferences using thin measuring tapes. To
their tetrapod dataset we added koala (Phascolarctos cinereus) and common wombat (Vombatus
ursinus) data from Wroe et al. [27] to represent the closest living palorchestid relatives (total sam-
ple species n = 255, S2 Table). From this modified dataset we derived predictive equations using
Model I (OLS) bivariate regressions using femoral, humeral and combined circumferences as the
independent variables and body mass as the dependent variable (all log10 transformed). For each
equation we report percent prediction error (PPE) and standard error of the estimate (SEE) as cal-
culated through the MASSTIMATE package [66] in R [67]. Depending on availability of associ-
ated material, we then used one or more of these equations to estimate body mass for each
palorchestid specimen. Additionally, we applied this method to revise previously-reported mass
estimates for our extinct comparative vombatiform taxa, provided in S2 Table.
Systematic palaeontology.
Supercohort MARSUPIALIA Illiger, 1811; sensu Cuvier, 1817
Order DIPROTODONTIA Owen, 1866
Suborder VOMBATIFORMES Woodburne, 1984
Infraorder VOMBATOMORPHIA Aplin & Archer, 1987
Superfamily DIPROTODONTOIDEA Archer & Bartholomai 1978
Family PALORCHESTIDAE Tate, 1948; sensu Archer & Bartholomai 1978
Emended diagnosis
Members of Family Palorchestidae can be diagnosed based on craniodental features described
by Woods [3] and emended by Trusler [36]. Additionally, they can be recognised from the fol-
lowing shared appendicular diagnostic features:
Humerus. Palorchestid humeri distinguished from diprotodontids by: Medial epicondyle
strongly developed, causing substantial widening of distal humerus relative to length; Insertion
facet for mm. latissimus dorsi & teres major ovoid in medial view, located at least halfway down
diaphysis; Insertion facet for mm. latissimus dorsi & teres major has a posterolateral margin
recurved as a ridge overhanging the posterior diaphysis.
Unguals. Palorchestid ungual phalanges distinguished from diprotodontids by: Ungual
processes dorsoventrally deep and laterally compressed; Contour of ventral margin in lateral
view angled and not a continuous arc; Extensor process dorsally deflected.
Genus PALORCHESTES Owen 1873
PLOS ONE | https://doi.org/10.1371/journal.pone.0221824 September 13, 2019
5 / 69
Osteology and functional morphology of the limbs in Palorchestidae
Emended diagnosis
Members of the genus Palorchestes can be diagnosed based on craniodental features described
by Woods [3] and emended by Trusler [36]. Additionally, they can be recognised from the fol-
lowing shared diagnostic appendicular features:
Humerus. Palorchestes humeri distinguished from diprotodontids by: Pectoral crest
inferior margin falcate in medial and lateral views; Pectoral crest distal tip curls medially
over bicipital sulcus. Differs from Propalorchestes in these humeral characters, as well as:
Deltoid tuberosity well developed laterally; Capitulum and trochlea distal surfaces project
equally distally in anterior view; Medial epicondyle more expansive relative to humeral
length.
Species Palorchestes azael Owen, 1873
Emended diagnosis
Palorchestes azael can be diagnosed based on craniodental features described originally by
Woods [3] and emended in detail by Trusler [36]. Additionally, it may be distinguished from
other Palorchestes species by the following diagnostic appendicular features:
Humerus. Differs from P. parvus in: Pectoral crest entirely separate from deltoid tuberosity
on the lateral diaphysis; Olecranon fossa absent; Trochlear facet almost completely flat and
inferiorly facing (articular surface not visible in anterior view); Trochlea deeper anteroposter-
iorly than capitulum in distal view; Rugose muscle scar for m. epitrochleoanconeus present on
posterior surface of medial epicondyle.
Ulna. Differs from P. parvus in: Trochlear surface narrow, elongate and flattened.
Species Palorchestes parvus De Vis, 1895
Emended diagnosis
Palorchestes parvus can be diagnosed based on craniodental features described by Woods [3]
and supplemented by Trusler [36]. Additionally, they can be recognised from the following
diagnostic appendicular features:
Humerus. Differs from P. azael in: Distal tip of pectoral crest and deltoid tuberosity con-
nected by superolaterally-oriented oblique ridge as in extant wombats; Olecranon fossa pres-
ent; Trochlear surface slightly convex and just visible in anterior view; Rugose muscle scar for
m. epitrochleoanconeus absent.
Ulna. Differs from P. azael in: Trochlear surface wide and concave both proximodistally
and mediolaterally.
Genus PROPALORCHESTES Murray 1986
Species Propalorchestes novaculocephalus Murray 1986
Propalorchestes ponticulus Murray 1990
Emended diagnosis
Specimens belonging to the genus Propalorchestes can be diagnosed based on cranial fea-
tures described in detail by Murray [68] and emended to include dental features in Murray
[28]. Additionally, they can be recognised from the following diagnostic appendicular
features:
Humerus. Differs from Palorchestes by: Pectoral crest thin and gracile medially; Pectoral
crest terminates in distinct tubercle at distal end; Deltoid tuberosity on lateral diaphysis weak;
Capitulum more distally extensive than trochlea in anterior view; Radial fossa present and dis-
tinct; Olecranon fossa present and deep.
PLOS ONE | https://doi.org/10.1371/journal.pone.0221824 September 13, 2019
6 / 69
Osteology and functional morphology of the limbs in Palorchestidae
Comparative descriptions
Palorchestes azael Owen 1873
Referred material. Measurements for all referred material below are provided in
S1 Table.
NMV P157144. Associated partial skeleton including: mandible with diagnostic molars;
left scapula fragment (glenoid cavity and base of spine with partial supraspinous/infraspinous/
subscapular fossae only); right humerus fragment (distal two-thirds of shaft and epiphysis,
missing capitulum and lateral epicondyle); left ulna fragment (proximal two-fifths, olecranon
and anconeal surface damaged); left and right os coxae fragments (acetabular regions and par-
tial ischia only); and femur fragment (diaphysis only). Unknown collector from unknown
locality.
NHMUK PV OR 46914. Right humerus (diaphyseal cortex fractured and repaired). This
element was figured and attributed to Nototherium mitchelli by Owen ([69], plate CXXVII Figs
1–6). Collected by W. L. R. Gipps from Castlereagh River, Mendooran, NSW, in 1875.
NMV P159792. Associated partial skeleton including: left humerus (superficial damage to
upper third of pectoral crest, lateral supracondylar ridge); right humerus (middle third of shaft
with partial pectoral crest, both epiphyses missing); left ulna (proximal two-thirds only); right
ulna; right radius (open distal metaphysis with missing distal epiphysis); right os coxa; left os
coxa (fragment with acetabulum, superior iliopubic ramus and ischial tuberosity only) with
detached symphyseal epiphysis; left femur (shaft fragment only); right tibia (lateral proximal
epiphysis damaged, hole in cortex of superior posterior diaphysis); ungual phalanx (dorsal pro-
cess missing). Unfused metaphyses indicate bones were not yet fully mature. Humeral and
ulnar morphology consistent with NMV P157144 but is smaller (S1 Table), other elements are
here referred to the species by association with this humerus and ulna. Collected by F. Spry
from Buchan Caves (probably Foul Air Cave), VIC, in 1907. Identified by Flannery and Archer
in 1980, articulated humerus and ulna illustrated in Flannery and Archer [10].
NMV P26534. Right femur (superficial damage to rim of head, lateral margin of lateral
condyle). Associated with NMV P159792 but given own registration when figured by Scott
([23], Plate 21) as a ‘phascolomyform’ femur. Collected by F. Spry from Buchan Caves (proba-
bly Foul Air Cave), VIC, in 1907.
Fig 1. Left scapula fragment of Palorchestes azael NMV P157144. (A) lateral; (B) anterior; (C) medial; (D) posterior; (E) distal views. Scale bar 50 mm.
https://doi.org/10.1371/journal.pone.0221824.g001
PLOS ONE | https://doi.org/10.1371/journal.pone.0221824 September 13, 2019
7 / 69
Osteology and functional morphology of the limbs in Palorchestidae
Fig 2. Labelled illustrations of Palorchestes azael left scapula NMV P157144. (A) lateral; (B) posterior; (C) distal
views. Hatching indicates surface damage to cortical bone, dashed lines indicate inferred bone contours.
Abbreviations: cop, coracoid process; gc, glenoid cavity; igt, infraglenoid tubercle; isf, infraspinous fossa; sbf,
subscapular fossa; sn, scapular notch; ssf, supraspinous fossa; ssp, scapular spine. Scale bar 50 mm.
https://doi.org/10.1371/journal.pone.0221824.g002
AM F5458. Right ulna fragment (proximal third, olecranon damaged). Unknown collector
from Wellington Caves, NSW, pre-1898.
NHMUK PV OR 47828. Left ulna fragment (proximal fourth, olecranon damaged). Col-
lected by G. Bennett from unknown locality, circa 1877.
AM F58934. Right humerus fragment (middle third, both epiphyses missing). Collected by
R. Wright from Ginnagulla, NSW, in 1977.
NMV P252196. Ungual phalanx (some erosion to edges of ungual crests). This element was
tentatively attributed to Thylacoleo and photographed by Wakefield [70], pg. 79. Collected by
N. A. Wakefield from Mount Hamilton Caves, VIC, in 1963.
SAMA P28945, P28946, P28947. Associated podial elements including: Two diagnostic
ungual phalanges; associated intermediate phalanx. Collected by R. T. Wells from Victoria
Fossil Cave, South Australia, 1970s.
SAMA P55199. Left humerus (distal half of immature bone with open articular meta-
physes, and distal articular and medial traction epiphyses missing). Collected by J. S. Lockie
from Puralka, VIC.
PLOS ONE | https://doi.org/10.1371/journal.pone.0221824 September 13, 2019
8 / 69
Osteology and functional morphology of the limbs in Palorchestidae
Fig 3. Humeri of Palorchestes azael. Left side humerus NMV P159792 (A-D, G-H) and right side humerus NHMUK PV OR 46914 (E-F, I-J, mirrored for comparison).
(A) anterior; (B) lateral; (C) posterior; (D) medial; (G) proximal and (H) distal views. Right humerus NHMUK PV OR 46914 in (E) anterior; (F) medial; (I) proximal and
(J) distal views. Scale bar 50 mm.
https://doi.org/10.1371/journal.pone.0221824.g003
PLOS ONE | https://doi.org/10.1371/journal.pone.0221824 September 13, 2019
9 / 69
Osteology and functional morphology of the limbs in Palorchestidae
Fig 4. Labelled illustrations of the Palorchestes azael left humerus NMV P159792. (A) anterior; (B) lateral; (C) posterior; (D) medial views. Hatching indicates
surface damage to cortical bone, dashed lines indicate inferred bone contours. Abbreviations: bg, bicipital groove; brf, fossa for m. brachialis origin; ca, capitulum; del,
deltoid insertion; eta, origin for m. epitrochleoanconeus; gt, greater tubercle; hh, humeral head; inf, fossa for insertion of m. infraspinatus; ldtm, insertion for mm.
latissimus dorsi and teres major; le, lateral epicondyle; lsc, lateral supracondylar crest; lt, lesser tubercle; me, medial epicondyle; pec, pectoral crest; subf, fossa for
insertion of m. subscapularis; supf, fossa for insertion of m. supraspinatus; sf, supracondylar foramen; tr, trochlea; tri, origin for humeral heads of m. triceps brachii.
Scale bar 50 mm.
https://doi.org/10.1371/journal.pone.0221824.g004
All following specimens were collected as part of the 1970s Gallus excavations at Keilor,
VIC, found associated in the same ‘D Clay’ layer, and assigned to Zygomaturus trilobus by
Marshall [50]. While no palorchestid material is yet reported from Keilor, these elements
strongly differ from the morphology seen in Zygomaturus, and some were identified as
palorchestid by Szalay [71], so we refer them to Palorchestes azael on the basis of their very
large size and morphological similarity to elements of AM F58870.
NMV P29619. Metatarsal 5 (proximolateral tuberosity eroded).
NMV P29620. Metatarsal 4.
NMV P29621. Right cuboid.
NMV P29622. Right ectocuneiform.
NMV P29623. Right navicular.
NMV P254089. Left astragalus (posterolateral part missing). Figured in Szalay ([71], Fig
8.46 E-F).
NMV P30723. Right calcaneus (tip of calcaneal tuber missing). Figured in Szalay ([71], Fig
7.83 A-C).
Scapula (Figs 1 and 2). The only known scapula for P. azael is part of the associated skele-
ton NMV P157144 (Figs 1 and 2). It is very eroded and encrusted with matrix which could not
be safely removed. Preserving only the glenoid and distal portions, its proximodistal length,
PLOS ONE | https://doi.org/10.1371/journal.pone.0221824 September 13, 2019
10 / 69
Osteology and functional morphology of the limbs in Palorchestidae
Fig 5. Right ulna of Palorchestes azael NMV P159792. (A) anterior; (B) medial; (C) posterior; (D) lateral; (E) distal views. Scale
bar 50 mm.
https://doi.org/10.1371/journal.pone.0221824.g005
anteroposterior width and overall shape are not known. What little morphology is retained
appears wombat-like, with the notable exception of the enlarged infraglenoid tubercle.
Glenoid fossa. The glenoid fossa is shallow, anteroposteriorly elongate and mediolaterally
narrow as is typical of vombatiforms, although none of the outer circumference is preserved so
the details of its shape are uncertain. The coracoid process is damaged and not clearly discernible.
Supraspinous and infraspinous fossae. Little of the supra- and infraspinous fossae are pre-
served, although the scapular notch is intact and appears more deeply curved than in vomba-
tids and Diprotodon but less than in zygomaturines.
Scapular spine. The base of the distal scapular spine is preserved, showing it to curl anteri-
orly over the supraspinous fossa as in zygomaturines. The dorsal margin and acromion are
missing.
PLOS ONE | https://doi.org/10.1371/journal.pone.0221824 September 13, 2019
11 / 69
Osteology and functional morphology of the limbs in Palorchestidae
Fig 6. Labelled illustrations of the Palorchestes azael right ulna NMV P159792. (A) anterior; (B) medial; (C) lateral
views. Hatching indicates surface damage to cortical bone. Abbreviations: ap, anconeal process; apl, origin for m.
abductor pollicis longus; cf, capitular facet; cp, coronoid process; edp, origin for m. extensor digitorum profundus; fdp,
fossa for origin of m. flexor digitorum profundus; op, olecranon process; pq, origin for m. pronator quadratus; rn,
radial notch; sp, styloid process; tn, trochlear notch; ut, ulnar tuberosity. Scale bar 50 mm.
https://doi.org/10.1371/journal.pone.0221824.g006
Infraglenoid tubercle. The infraglenoid tubercle, though damaged, is very extensive med-
iolaterally and superoinferiorly, and deeply pitted. Roughly triangular in shape, it would have
provided large origin for the long head of m. triceps brachii. It does not appear to have pro-
jected away from the glenoid to alter the posterior scapular border as it does in derived dipro-
todontids like Diprotodon and Zygomaturus, instead resembling the vombatid condition,
though being thicker mediolaterally.
Humerus (Figs 3 and 4). The humerus of Palorchestes azael is remarkable in its blade-like
pectoral crest and its curiously flat trochlear surface (Figs 3 and 4). Of the two complete
humeri figured, the larger (NHMUK PV OR 46914) shows marked increase in relative size
and rugosity of muscle attachments, while the smaller humerus with its unfused metaphyses
PLOS ONE | https://doi.org/10.1371/journal.pone.0221824 September 13, 2019
12 / 69
Osteology and functional morphology of the limbs in Palorchestidae
likely represents an osteologically-immature animal. Descriptions below are based on the
former.
Head. The domed head extends its articular surface posteroinferiorly to provide increased
posterior surface area, similar to that of other large-bodied diprotodontoids, rather than the
superiorly-oriented head in vombatids. Viewed medially, this articular surface ‘beaks’ out
from the humeral neck at its posteroinferior tip. The head is flanked anteriorly by greater and
lesser tubercles separated by a broad and shallow bicipital sulcus.
Greater tubercle. The greater tubercle is rugose and projects above the humeral head, nar-
rowing to a proximally rounded tip resembling that of Lasiorhinus in shape and proportion.
Discrete proximal (for m. supraspinatus) and lateral (for m. infraspinatus) facets are present.
The tubercle continues inferomedially to give rise to the pectoral crest. Viewed laterally, the
greater tubercle is broad and occupies two thirds of the total anteroposterior depth of the prox-
imal humerus.
Lesser tubercle. The lesser tubercle in P. azael is comparatively poorly developed among
diprotodontoids. In medial view, it extends obliquely from its uppermost tip anteriorly to its
distal edge posteriorly, terminating proximal to the tip of the beaked humeral head. A medi-
ally-oriented facet for insertion of m. subscapularis is, like in other palorchestids, narrow and
steeply oblique, though less steep than in Nimbadon.
Deltoid tuberosity. Viewed anteriorly the deltoid tuberosity is extensive and approximately
triangular in shape, projecting strongly from the lateral diaphysis and contour of the greater
tubercle above. It tapers to a rugose, bulbous terminus at its apex which lies a third of the way
down the humeral shaft. The posterolateral lip of this deltoid tuberosity curls posteriorly to
form the lateral margin of the fossa for m. brachialis. The clear separation of this large tuberos-
ity from the pectoral insertion crest, contrasting with morphology of other palorchestids and
vombatids, suggests a strongly developed scapular deltoid in P. azael and a reduction in the
clavicular deltoid, with no distinct attachment site visible for this muscle. This arrangement
resembles that seen in Diprotodon, although in P. azael the deltoid tuberosity differs strongly
from that taxon in its triangular form and far more proximal position relative to the pectoral
crest.
Pectoral crest. The pectoral crest is a distinctive feature of the P. azael humerus, originating
from the greater tubercle just lateral to the humeral midline. It follows an oblique inferomedial
path along the anterior shaft, becoming a laminar crest with a rectangular sectional profile
which increases in height and thickness as it curls around medially. At its distal tip it swells to
a bulbous terminus in the largest specimens, which then re-joins the shaft via a recurved,
sickle-shaped ‘pulley’, presumably for the passage of m. biceps brachii descending along its
medial plane. At its zenith the anterior projection of this pectoral crest from the shaft is greater
than in any other vombatiform studied, in both relative and absolute extent.
Tuberosity for mm. teres major and latissimus dorsi. An elongate, ovoid muscle scar lays
halfway down the medial aspect of the humeral shaft for the insertion of mm. teres major and
latissimus dorsi. Its posterior edge overhangs and sharply demarcates the posterior from the
anterior shaft surface. This insertion is relatively larger, better developed and more distal than
in any other vombatiform studied.
Diaphysis. The humeral diaphysis is highly irregular in section and in profile due to the
many large muscle attachment crests along its relatively short length. It is stout and broad in
anterior view with an almost ‘hourglass’ silhouette due to the flared proximal and distal ends
and narrow constriction in line with the lateral supracondylar crest. In lateral view the poste-
rior contour of the shaft is near-flat (as is typical for diprotodontoids), while the anterior con-
tour is dominated by the hooked pectoral crest. The posterior diaphyseal surface features
obvious scars for the origins of the humeral heads of m. triceps brachii. At the base of the
PLOS ONE | https://doi.org/10.1371/journal.pone.0221824 September 13, 2019
13 / 69
Osteology and functional morphology of the limbs in Palorchestidae
humeral neck is a distinct fossa for the origin of m. brachialis, so expansive as to nearly reach
the shaft midline as in Phascolonus. This fossa curves laterally beneath the deltoid tuberosity
and anterior to the lateral supracondylar crest. There is no olecranon fossa on the posterior
aspect of the distal humeral shaft, which is flat and smooth but for a marked rugose patch
immediately proximal to the trochlea. This forms the origin for m. epitrochleoanconeus.
Lateral epicondyle. The lateral epicondyle projects only slightly beyond the lateral margin
of the capitulum. In lateral view, the lateral epicondyle is thick inferiorly before curving proxi-
mally up and tapering into the lateral supracondylar crest, a sheet of bone that presumably ter-
minated in a hooked process as in most other vombatiform marsupials. Almost the entire edge
of this crest is damaged in all specimens observed. However, the concave, arced profile that
remains at the top of the crest, curling superiorly from the lateral shaft, suggests such a hooked
process was once present. This crest would have provided origin for mm. brachioradialis and
extensor carpi radialis.
Trochlea. The trochlea is another notably unique feature of the P. azael humerus. Its articu-
lar area is completely flattened and would have allowed almost no parasagittal movement of
the ulna. Rather than a pulley-shaped or domed articulation as in other vombatiforms, it is a
simple, distolaterally-facing ovoid facet oriented ~115˚ to the neighbouring capitulum. In dis-
tal view the trochlea is offset anteriorly from the dorsal (coronal) plane of the humerus, pro-
jecting cranially much more than the smaller capitulum. In the larger P. azael specimens the
posterior trochlear surface curves superiorly to provide slightly more anconeal articulation
than in the smaller humeri.
Capitulum. In anterior view the capitulum is near hemi-spherical and lacks an obvious
radial fossa proximally. In distal view it is ~20% smaller than the trochlea. Posteriorly, the joint
surface for articulation with the anconeal process of the ulna extends slightly higher than on
the anterior side of the capitulum and is slightly elevated from the posterior shaft surface.
Supracondylar foramen. The bridge defining the ceiling of the supracondylar foramen is
broad proximally and tapers somewhat at its mediodistal end. This bridge lies almost horizon-
tally along the corresponding contour of the medial epicondyle, giving the underlying foramen
a vertical orientation unlike the more oblique foramina of other vombatiforms (where pres-
ent). The foramen is approximately four times longer than its breadth.
Medial epicondyle. The medial epicondyle in anterior view is rounded and projects
strongly medially, its axis perpendicular to the vertical axis of the humeral shaft in both dorsal
and transverse planes. The medial epicondyle constitutes approximately 32% of the total distal
humeral width, making it the largest, both relatively and absolutely, among the vombatiforms.
Ulna (Figs 5 and 6). Generally, the ulnar shape and proportions of P. azael resembles that
of extant wombats, Ngapakaldia and Thylacoleo much more than other large diprotodontoids,
albeit with some significant alterations (see S2 Fig). Overall the ulna is straight when viewed
anteriorly, while in lateral view it is slightly convex posteriorly, giving a bowed appearance
(Figs 5 and 6).
Olecranon process. The olecranon is very enlarged and elongate, representing nearly a
quarter of the total ulnar length, and in lateral view appears quadrangular with a slight bulbous
projection posteriorly. Unlike in the diprotodontid ulna, the olecranon lies along the axis of
the shaft with no posterior deflection, and medial deflection is only very slight–less than in
extant wombat or Ng. bonythoni ulnae and much less than the medially-curled olecranon in
Phascolonus. The olecranon is deep in the sagittal plane, slightly more so than at the midshaft
of the bone, and does not taper proximally as extant wombats do. The transverse width at the
dorsal edge of the olecranon is around half its dorso-ventral depth.
Trochlear notch. The trochlear notch and trochlear surface are flattened (in agreement
with the corresponding articular surface of the humerus), with only the slightest concavity as
PLOS ONE | https://doi.org/10.1371/journal.pone.0221824 September 13, 2019
14 / 69
Osteology and functional morphology of the limbs in Palorchestidae
the trochlear surface ramps distally to form the coronoid process. The orientation of the troch-
lear surface is slightly less steep than in vombatids, being approximately 45˚ to the axis of the
shaft. This contrasts with the near-horizontal equivalent in diprotodontids and is not nearly as
concave as the latter. In P. azael this flat trochlear surface lies on a medially-projecting plat-
form, creating a deep elongate fossa beneath it on the medial shaft surface as in vombatids and
Ng. bonythoni. The floor of this fossa represents the narrowest lateral dimension of the ulna
overall and provided origin for m. flexor digitorum profundus. The lateral semilunar facet for
the humeral capitulum is concave and less than half the proximodistal length of the trochlear
surface.
Coronoid process. In lateral view the coronoid process is robust with a triangular profile,
owing to the flattened trochlear surface on the proximal side and the gradual slope of diaphy-
seal bone on the distal side. It is less dorsoventrally extensive relative to the diaphyseal thick-
ness than in all other comparative taxa observed.
Anconeal process. The anconeal process is strongly laterally deflected, its axis in the trans-
verse plane offset by ~90˚ to that of the shaft so that in medial view it is not visible.
Radial notch. The radial notch projects laterally and is similar to the anconeal process in its
extent. The radial notch is more posteriorly offset than any other taxon studied, positioned
very near the dorsal border of the shaft. A deep, inset ligamentous pit lies directly mediad to it
for attachment of the annular ligament. Like the notch, this pit is more dorsally positioned
than in other vombatomorphs.
Ulnar tuberosity. The ulnar tuberosity for insertion of mm. brachialis and biceps brachii is
approximately 15 mm in diameter. It is raised, bulging and is situated well distal to the coro-
noid process and radial notch on the lateral aspect of the ventral shaft margin. In relative dis-
placement of the tuberosity from the coronoid, the P. azael ulna is surpassed only by that of
Phascolonus.
Diaphysis. The diaphysis varies in section profile along its length from ovo-quadrangular
proximally, lachrymiform at midshaft due to the thickened posterior border, to subcircular at
the distal epiphysis. This circular distal section contrasts with the mediolaterally-flattened
epiphyses of vombatid and Ng. tedfordi ulnae, being more similar to those of Diprotodon and
Zygomaturus, likely corresponding to changes in directional stresses associated with increased
body mass. The P. azael ulnar shaft also varies in anteroposterior depth along its length, being
deepest at the ulnar tuberosity and gently tapering distally. The lateral diaphyseal surface is
largely smooth and lacks the marked longitudinal scarring seen in diprotodontid ulnae. The
most obvious muscle attachments are two large ovoid scars–one arising laterally at the approx-
imate midshaft (mm. abductor pollicis longus and extensor digitorum profundus), the other
occupying the distal third of the lateral diaphysis (m. pronator quadratus). The latter scar
strongly resembles the condition in Ngapakaldia.
Styloid process. The styloid process is thick and distinctly anteromedially deflected like in
vombatids, tapering from a wide, dorsolaterally-elongate base to a small bulb featuring a dis-
tally-facing convex articular facet for the triquetrum. The styloid is ovoid in distal view and
much smaller than the surrounding epiphysis, similar to that of Ng. tedfordi in proportions
and shape and unlike the inflated, bulbous styloid in Ng. bonythoni, Neohelos, Diprotodon and
Zygomaturus.
Radius (Figs 7 and 8). The palorchestid radius is most like that of Vombatus in overall
proportions (Figs 7 and 8). The angle of the head and neck, and the degree to which the diaph-
ysis curves medially down toward the styloid process is very vombatid. Like wombats, the
radial midshaft is triangular in section. It becomes progressively more mediolaterally expanded
to appear flatter and bulkier in distal section than the wombat though still distinctly trapezoi-
dal, unlike the triangular distal profile of Diprotodon and Phascolonus radii.
PLOS ONE | https://doi.org/10.1371/journal.pone.0221824 September 13, 2019
15 / 69
Osteology and functional morphology of the limbs in Palorchestidae
Fig 7. Right radius of Palorchestes azael NMV P159792. (A) dorsal; (B) medial; (C) ventral; (D) lateral; (E) proximal;
(F) distal views. Scale bar 50 mm.
https://doi.org/10.1371/journal.pone.0221824.g007
Head. In proximal view, the perimeter of the capitular fossa is suboval, with the slight trans-
verse elongation seen in wombats and Ng. tedfordi, rather than the more truly circular fossae
of Diprotodon and Phascolonus radial heads. The hemispherical fossa is deeply cupped, closest
in relative depth to that of Ng. tedfordi, though lacking the distinct craniolateral tilt and sharp
rim edge of the latter. The articular surface for the ulna occupies two-fifths of the capitular cir-
cumference, roughly equivalent to the vombatid condition and less than in Ng. tedfordi.
Radial tuberosity. The radial tuberosity for insertion of m. biceps brachii is more distally
situated on the shaft than in Vombatus, closely resembling Lasiorhinus and Ngapakaldia spp.
in relative positioning. However, the radial tuberosity in P. azael is not prominent and does
not project from the shaft anything like the marked tuberosities seen in the radii of Phascolo-
nus, extant wombats or Ngapakaldia spp., suggesting reduced size or contractile force in m.
biceps brachii.
PLOS ONE | https://doi.org/10.1371/journal.pone.0221824 September 13, 2019
16 / 69
Osteology and functional morphology of the limbs in Palorchestidae
Fig 8. Labelled illustrations of the Palorchestes azael right radius NMV P159792. (A) dorsal; (B) medial; (C) ventral
views. Hatching indicates surface damage to cortical bone, dashed lines indicate inferred bone contours.
Abbreviations: cf, capitular fossa; ecr, notch for tendon of m. extensor carpi radialis; ioc, interosseous crest; rt, radial
tuberosity; uls, articular surface for ulna. Scale bar 50 mm.
https://doi.org/10.1371/journal.pone.0221824.g008
Diaphysis. The interosseous crest begins immediately distal to the radial tuberosity and is
less sharply defined than in extant wombats, and dramatically less so than the broad flange
present in Phascolonus. The crest continues along the lateral border of the diaphysis, forming
the ‘apex’ of the triangular midshaft section profile and terminating in a slightly rugose area ¾
of the way down the shaft. This rugosity differs strongly from the Vombatus and Phascolonus
forms in which it is deeply pitted. The shaft becomes extremely thick and robust distally, more
so than any other taxon studied. The medial diaphyseal border is dominated by the insertion
scar for m. pronator quadratus, which is very similar in positioning and extent to that seen in
extant wombats.
Distal end. The distal section profile of the P. azael radius is trapezoidal, with a flattened
dorsal surface parallel to the slightly concave ventral surface. This contrasts with the dorsoven-
trally broad, triangular distal profile of diprotodontid and Vombatus radii, the flattened
palorchestid state being most similar to that of Lasiorhinus and Ngapakaldia.
Ungual (Figs 9 and 10). Palorchestid unguals are morphologically distinct from all other
vombatiforms in their deep, laterally-compressed, lunate shape (Figs 9 and 10). Overall, they
most closely resemble those of Phascolarctos or Nimbadon, but in palorchestids they are dorso-
ventrally deeper and larger relative to the proximal and intermediate phalanges.
PLOS ONE | https://doi.org/10.1371/journal.pone.0221824 September 13, 2019
17 / 69
Osteology and functional morphology of the limbs in Palorchestidae
Fig 9. Unassociated ungual phalanges of Palorchestes azael. NMV P252196 (A-E) and NMV P159792 (F-J) in (A,F)
lateral; (B, G) medial; (C, H) proximal; (D, I) dorsal; (E, J) volar views. Scale bar 50 mm.
https://doi.org/10.1371/journal.pone.0221824.g009
Ungual crests. When preserved, these crests are very well-developed plates of bone, project-
ing distally to an even greater extent than in Nimbadon or extant Phascolarctos. The resulting
sulcus would have provided a deep and secure attachment for the proximal rim of the kerati-
nous claw sheath.
Proximal end. The articular facets for the intermediate phalangeal condyles are transversely
narrow, deep semilunate notches with a low median crest. In proximal view the facets are
slightly tapered dorsally and broader ventrally, giving an elongate trapezoidal shape overall in
this aspect. The dorsal extensor process is expanded and is more dorsally deflected than in any
other vombatiform, overhanging the joint while allowing a degree of hyperextension. Dilated
and robust flexor tubercles on the plantar surface are larger relative to the rest of the ungual
than other large diprotodontoids. These tubercles are as mediolaterally broad as the articular
facets, giving a more quadrate shape to this region of the ungual in proximal view unlike the
ventrally tapered appearance in Ngapakaldia and Neohelos.
Whether the assigned P. azael unguals are manual or pedal is not known.
Fig 10. Labelled illustration of ungual phalanx of Palorchestes azael NMV P252196. (A) Lateral; (B) proximal views.
Hatching indicates surface damage to cortical bone, dashed lines indicate inferred bone contours. Abbreviations: af,
articular facet; dp, dorsal process; ft, flexor tubercle; mk, median keel; uc, ungual crest (estimated); up, ungual process.
Scale bar 50 mm.
https://doi.org/10.1371/journal.pone.0221824.g010
PLOS ONE | https://doi.org/10.1371/journal.pone.0221824 September 13, 2019
18 / 69
Osteology and functional morphology of the limbs in Palorchestidae
Os coxa (Figs 11 and 12). The os coxa overall is gracile and slender, lacking broad muscle
attachment surfaces or pronounced tuberosities (Figs 11 and 12).
Iliac blades. In P. azael the iliac blades are sickle-shaped, with mediolaterally narrow
dimensions, quite unlike the laterally flared ilia of Zygomaturus and Diprotodon. In relative
length the ilia are intermediate between those taxa and the elongate ilia of extant wombats.
They resemble Neohelos ilia but have a more deeply concave lateral margin and are slightly
broader. The best-preserved specimen is missing the lateral tip of the blade and the epiphyseal
rim along its proximal margin–their contours are estimated in Fig 12. In lateral view the ilium
is oriented to create a 140˚ angle with the body of the ischium, a much more dorsally-rotated
position than in Zygomaturus or the almost aligned Diprotodon. The ilium is slightly twisted
anteriorly causing the posterior (gluteal) blade surface to be visible in lateral aspect.
Ischial body. The body of the ischium in P. azael is long, with a flattened ovoid profile. It is
relatively longer than in Diprotodon or Zygomaturus, with a similarly upturned distal end for
the ischial tuberosity. Unlike extant wombat ischia, there is no ischial spine present.
Ischial tuberosity. The ischial tuberosity is a gracile, elongate, weakly rugose structure in
posterior view and lacking the distinct triangular shape with posterolateral projections seen in
wombats. It appears narrower and shorter than in all other diprotodontoids studied.
Acetabulum. The shape of the acetabular margin in lateral view is arched, with a slightly
tapered point at its superior margin greatly overhanging the acetabular fossa. Inside the acetab-
ulum the notch is narrow, leaving a large articular surface area within the socket resembling
the condition in Zygomaturus in extent and shape, though less buttressed around the perime-
ter than the latter. There is a deeply excavated fossa for the ligamentum teres. The acetabulum
overall is shallower and less circular than in diprotodontids and vombatids, and more postero-
laterally oriented than in the latter.
Anterior inferior iliac spine. The anterior inferior iliac spine for the origin of m. rectus
femoris is located inferiorly and close to the acetabular margin as in Zygomaturus, though it is
a narrower muscle scar in P. azael.
Iliopectineal eminence. The iliopectineal eminence is smaller, sharper and more distinct in
P. azael than any other taxon studied.
Iliopubic ramus. The iliopubic ramus in P. azael is slender overall, with a large ovoid fossa
for articulation with the epipubic bone. This ramus is relatively longer and more slender than
in Ngapakaldia, Phascolonus or Diprotodon and more inferiorly directed than in Neohelos,
with more circular section profile.
Obturator foramen. The obturator foramen is ovotriangular and slightly elongate, with the
apex of its shape directly under the pectineal origin, similar to that seen in Ngapakaldia and
Neohelos, though relatively broader than both.
Inferior/ischiopubic ramus. The ischiopubic ramus is long relative to the ischial body,
indicating a deep pelvic cavity as in diprotodontids. It is dorsoventrally shallow at its pubic
end. It deepens dorsoventrally as it extends posteriorly, though not as much as in Phascolonus,
upturning but not increasing in transverse thickness as it gives rise to the ischial tuberosity.
Auricular surface. The auricular surface is rugose, with two distinct articular facets; a small
one facing medially and a longer, deeper one facing more anteriorly. The sacral surface is cra-
niocaudally short compared to that of Zygomaturus and much shorter than in Phascolonus,
being more similar to the Ngapakaldia form. Inferior to the auricular surface, a small greater
sciatic notch is visible in posterior view.
Detached symphysial epiphysis. Associated with the NMV P159792 os coxae of P. azael is
the detached symphyseal epiphysis from the ventral pubis. This provides the subpubic angle
and angle of the anterior pelvic brim, indicating P. azael had a more V-shaped girdle as in Nga-
pakaldia than the U-shape in vombatids or broad, open girdle of large diprotodontids. The
PLOS ONE | https://doi.org/10.1371/journal.pone.0221824 September 13, 2019
19 / 69
Osteology and functional morphology of the limbs in Palorchestidae
PLOS ONE | https://doi.org/10.1371/journal.pone.0221824 September 13, 2019
20 / 69
Fig 11. Associated pelvic elements of Palorchestes azael NMV P159792. Right side os coxa in (A) lateral view; (B) dorsal; (C) medial; (D) ventral views. Left side os coxa
fragment in (E) lateral view. Pubic symphyseal epiphysis in (F) anterior; (G) posterior view. Scale bar 50 mm.
https://doi.org/10.1371/journal.pone.0221824.g011
Osteology and functional morphology of the limbs in Palorchestidae
reconstruction in Fig 12D shows that the ilia are less laterally extensive than in vombatids or
large diprotodontids, possible evidence for a slenderer body form in P. azael.
Femur (Figs 13 and 14). The Palorchestes azael femur is elongate and gracile for its size,
with a slender sub-cylindrical shaft and narrow epiphyses. In proportions and overall shape it
strongly resembles the femur of extant Vombatus (see S3 Fig), but is of course significantly
larger, with a weaker lesser trochanter and no gluteal tuberosity (Figs 13 and 14).
Femoral head. The apex of the hemispherical head lies slightly proximal to the greater tro-
chanter when the femoral condyles are placed on a flat surface. The neck is stout and very
short relative to femora of similarly-sized diprotodontids, and projects anteromedially, posi-
tioning the head such that ~30% of its bulk overhangs the anterior shaft when viewed medially.
In proximal view the head is significantly deeper cranio-caudally than the diaphysis or neigh-
bouring greater trochanter, being very similar in this respect to the anatomy of Vombatus and
differing from other large diprotodontoids.
Greater trochanter. In anterior view the greater trochanter is tapered proximally as it proj-
ects strongly above the femoral neck unlike in diprotodontids. The anterolateral surface of the
greater trochanteric epiphysis projects anteriorly from the shaft and curls slightly medially. In
lateral view the trochanter appears expanded anteroposteriorly to give a rounded appearance,
and the anterior projection of its distal portion is clearly appreciated. Posteriorly, the trochan-
teric fossa is very deep and relatively longer superoinferiorly than any other taxon studied.
There is no discernible intertrochanteric crest or gluteal tuberosity.
Lesser trochanter. The lesser trochanter is an extensive plate-like flange on the proximo-
medial femoral shaft, arising at the femoral neck and extending distally to become a rugose
insertion scar that extends to a third of the way down the entire bone length. Some damage to
the proximal portion of this crest obscures its full extent, but overall the lesser trochanter
appears to be similar to those of Diprotodon and Zygomaturus in that it is a broad crest rather
than a distinct tubercle, especially appreciated in posterior view.
The proximal femur as a whole resembles that of Phascolonus more than any other vomba-
tiform, though with key differences in P. azael including; a more acute angle between the
higher greater trochanter and shorter femoral neck, more distally-extensive articular surface
around all edges of the head, and a mediolaterally narrower proximal femur overall.
Diaphysis. The femoral diaphysis in P. azael is smooth overall. Like in other diprotodon-
toids there is no linea aspera as such, however the femoral shaft shows elliptical pits, grooves
and muscle scarring in the area immediately distal to the lesser trochanter on the medial and
posteromedial midshaft. These mark insertion areas for adductor musculature and are much
more pronounced than in other vombatiforms studied. On the distolateral diaphysis a raised
area bordered by a crest marks the origin for lateral head of m. gastrocnemius, similarly situ-
ated to that in Zygomaturus but sharper and more distinct than in the latter.
Medial condyle. In distal view the anteroposterior extent of the medial condyle is ~30%
greater than that of the lateral (as is typical of large vombatiforms). It is mediolaterally nar-
rower and more oblique to the sagittal axis of the bone than the lateral condyle. The medial
surface of the condyle is deeply pitted for the attachment of the medial collateral ligament. In
posterior view, the medial condyle presents a narrower articular surface than the lateral, with a
hooked process curling in toward the bone midline as in vombatids and Thylacoleo, though
more pronounced than both. In anterior view the distal surfaces of the medial and lateral con-
dyles sit approximately level to a plane perpendicular to the femoral shaft, unlike the marked
distal offset of the lateral condyle in large diprotodontids and Phascolonus.
PLOS ONE | https://doi.org/10.1371/journal.pone.0221824 September 13, 2019
21 / 69
Osteology and functional morphology of the limbs in Palorchestidae
Fig 12. Labelled illustrations of the Palorchestes azael os coxae. (A) right side os coxa, lateral view; (B) right side os coxa, dorsal view; (C) right side os
coxa, medial view; (D) reconstructed pelvic girdle, superoventral view (this view not to scale). Hatching indicates surface damage to cortical bone, dashed
PLOS ONE | https://doi.org/10.1371/journal.pone.0221824 September 13, 2019
22 / 69
Osteology and functional morphology of the limbs in Palorchestidae
lines indicate inferred bone contours. Abbreviations: acf, acetabular fossa; acn, acetabular notch; aiis, anterior inferior iliac spine; aur, auricular surface;
epi, facet for articulation with epipubic bone; glf, gluteal fossa; gsn, greater sciatic notch; ilb, iliac blade; ilc, iliac crest; ilf, iliacus fossa; ilpr, iliopubic
ramus; ipe, iliopectineal eminence; isb, ischial body; ispr, ischiopubic ramus; ist, ischial tuberosity; obf, obturator foramen; psy, pubic symphysis. Scale
bar 50 mm.
https://doi.org/10.1371/journal.pone.0221824.g012
Fig 13. Right femur of Palorchestes azael NMV P26534. (A) anterior; (B) medial; (C) posterior; (D) lateral; (E) proximal; (F) distal views. Scale bar 50 mm.
https://doi.org/10.1371/journal.pone.0221824.g013
PLOS ONE | https://doi.org/10.1371/journal.pone.0221824 September 13, 2019
23 / 69
Osteology and functional morphology of the limbs in Palorchestidae
Fig 14. Labelled illustrations of the Palorchestes azael right femur NMV P26534. (A) anterior; (B) medial; (C) posterior; (D) lateral views. Hatching indicates surface
damage to cortical bone, dashed lines indicate inferred bone contours. Abbreviations: add, insertion scars for adductor muscles; fh, femoral head; gt, greater trochanter;
gas, origin for m. gastrocnemius lateral head; lc, lateral condyle; lt, lesser trochanter; mc, medial condyle; ps, patellar surface; tf, trochanteric fossa. Scale bar 50 mm.
https://doi.org/10.1371/journal.pone.0221824.g014
Lateral condyle. The lateral condyle in distal view is anteroposteriorly shorter, mediolater-
ally broader and oriented more closely along the sagittal axis than its medial counterpart. In
posterior view, the lateral condyle is very short proximodistally relative to the medial condyle.
This contrasts both with the anatomy of Diprotodon and Phascolonus in which the lateral is
PLOS ONE | https://doi.org/10.1371/journal.pone.0221824 September 13, 2019
24 / 69
Osteology and functional morphology of the limbs in Palorchestidae
taller, and with the subequal heights of the condyles in most other diprotodontids. The Ngapa-
kaldia femur presents a similar condition to palorchestids, however in the latter overall both
condyles appear short proximodistally relative to femoral length when compared to all other
vombatiforms.
Patellar surface. The patellar surface in P. azael is quadrangular and only shallowly
indented. In distal view the medial crest of the patellar surface only modestly protrudes from
the distal femur, far less than the protrusion seen in large diprotodontids or Phascolonus,
being more similar to extant wombats.
Tibia (Figs 15 and 16). Other than thicker transverse dimensions along the diaphysis, the
tibia of P. azael resembles that of Vombatus in general proportions in that it is longer relative
to the epiphyses than in diprotodontids or Phascolonus (Figs 15 and 16).
Proximal articulations. The medial condylar surface is suboval, much broader in all
dimensions than in P. parvus and closer in extent to the Phascolonus tiba. It is concave but not
as deep or regular as those seen in Diprotodon and Zygomaturus. The convex lateral condylar
surface is elevated on the tibial plateau relative to the medial condyle and is a near-level
Fig 15. Left tibia of Palorchestes azael NMV P159792. (A) anterior; (B) medial; (C) posterior; (D) lateral; (E) proximal; (F)
distal views. Scale bar 50 mm.
https://doi.org/10.1371/journal.pone.0221824.g015
PLOS ONE | https://doi.org/10.1371/journal.pone.0221824 September 13, 2019
25 / 69
Osteology and functional morphology of the limbs in Palorchestidae
Fig 16. Labelled illustrations of the Palorchestes azael left tibia NMV P159792. (A) anterior; (B) medial; (C) posterior; (D) lateral; (E) proximal views.
Hatching indicates surface damage to cortical bone, dashed lines indicate inferred bone contours. Abbreviations: af, astragalar facet; gr, insertion of m. gracilis;
ie, intercondylar eminence; iol, interosseus line; lc, lateral condylar surface; mc, medial condylar surface; mm, medial malleolus; pf, popliteal fossa; ptm,
posterior tubercle of medial malleolus; st, insertion of m. semitendinosus; tc, tibial crest; tt, tibial tuberosity. Scale bar 50 mm.
https://doi.org/10.1371/journal.pone.0221824.g016
surface, rather than the gently posteriorly-sloped surfaces in Phascolonus and Diprotodon or
the steeply sloped equivalent in Zygomaturus. The anterior plateau and tibial tuberosity com-
prise almost half of the total surface of the proximal tibia. This creates a proximal tibial surface
that is expanded anteroposteriorly, approaching the vombatid condition more than that of
diprotodontoids. The intercondylar eminence is a quadrangular protuberance in anterior
aspect, similar to that of vombatids but more robust in proximal view. The tibial tuberosity for
insertion of the patellar ligament is an inverted isosceles triangle, more elongate than in vom-
batids and unlike the trapezoidal diprotodontid condition. The lateralmost portion of the
proximal tibia is eroded so the shape and extent of the fibular articulation in P. azael is not
known. A concave popliteal fossa lies beneath the tibial plateau on the posterior side.
Diaphysis. The diaphysis in midsection is flattened medially and convex laterally, creating
distinct anterior and posterior borders between these surfaces. The principal difference with
the Vombatus tibia is that the tibial crest in P. azael is more proximally positioned and not as
convex in mediolateral view. The line for attachment of the tibiofibular interosseus ligament
lies obliquely along the lower anterolateral diaphysis, while medially the scars for m. gracilis
and m. semitendinosus are strongly expressed.
Distal articulations. The distal articular surface is similar to that of vombatids in basic
shape, contours and orientation. The astragalar facet is highly convex and continues further
anterolaterally than in vombatids to provide a large laterally-facing articular surface, larger
than and totally unlike the flat facet in diprotodontids. The posterior tubercle of the malleolus
is a small conical process, more distinct than in diprotodontids and vombatids while being
shorter and barely extending past the distal articular surface (it is absent in Phascolonus).
Behind this lies an incised groove for the ankle and digital flexor tendons.
PLOS ONE | https://doi.org/10.1371/journal.pone.0221824 September 13, 2019
26 / 69
Osteology and functional morphology of the limbs in Palorchestidae
Pes (Figs 17 and 18). The pes of P. azael is very unlike the highly modified pedes of other
giant vombatiforms (Figs 17 and 18).
Calcaneus. The sustentaculum is better developed and more distinct than in Zygomaturus
and Diprotodon, presenting a steeply sloped astragalar facet laterally. The sustentaculum has
a curved articular facet on its superior edge and is dorsoventrally thickened compared to the
calcaneus of Ng. tedfordi. The calcaneal tuber is comparatively straight. The posteriormost
part of the tuber is eroded so the total length and extent to which it curves medially is not
known. However, it is clear that the tuber is not the narrow sigmoid-shaped structure of Zygo-
maturus and instead appears to resemble the morphology seen in Nimbadon, Thylacoleo and
Ng. tedfordi. The distolateral process is very pointed in dorsal view. The cuboid facet is a near-
spherical concavity, more dorsally-facing than in most diprotodontids, bearing the closest
resemblance to that of Neohelos and Ng. tedfordi. The plantar surface is pitted and rugose,
especially towards the distal end.
Astragalus. The astragalus is damaged posterolaterally, missing the lateral part of the tibial
facet and the entire fibular facet–their contours are estimated in Fig 18B and 18C. As such, it is
not possible to discern the extent of the facet for the pyramidalis sesamoid on its dorsal surface.
Overall the remaining morphology strongly resembles Ngapakaldia tedfordi astragali in pro-
portions and joint surface shape, the principal difference besides size being that it is dorsoven-
trally deeper as in Nimbadon. Like in Ng. tedfordi and Nimbadon, the tibial facet dorsally is
very concave, much more than in Zygomaturus. In distal view, the navicular facet is ventrome-
dially sloped relative to the tibial surface, unlike the flatter facets in Zygomaturus or Nimbadon.
In ventral view, the calcaneal surface is much smaller and less anteriorly extensive than in
Zygomaturus, as well as having a much better developed medial plantar tuberosity (sensu Mun-
son [24]). The navicular surface sits separate and ventral to the calcaneal surface on the plantar
aspect and has a distinct facet for articulation with the cuboid. Anterolaterally, the sustentacu-
lar facet is very convex, dipping into a deeply concave fossa (much more concave than in Zygo-
maturus) before the plantar tuberosity arises medially. The sulcus tali is similar to that of
Zygomaturus though is more deeply excavated.
Navicular. The navicular is robust and relatively dorsoventrally expanded compared to
those of P. parvus and Ng. tedfordi. The cuboid facet is relatively larger than in Ng. tedfordi and
quite concave. The facet for the ectocuneiform is broad and only slightly concave. This articu-
lation appears relatively larger in P. azael than in Nimbadon. Its plantar tuberosity is not as
developed as in derived diprotodontids or Thylacoleo.
Cuboid. The cuboid has a well-developed plantar tuberosity, anterior to which runs a deep,
narrow sulcus for the m. peroneus longus tendon as it approaches the first digit on the plantar
surface of the pes. This sulcus is the deepest we observed among our comparative taxa.
Ectocuneiform. In dorsal view the ectocuneiform is much larger relative to the cuboid than
in Ng. tedfordi and Zygomaturus, but not as large as in Thylacoleo. Distally, it presents a concave
ovoid medial facet for metatarsal 3, two-thirds the height of its distal face, the other ventral
third being made up of a well-developed plantar tuberosity. Immediately lateral and oriented
45˚ to this ovoid facet is a sub-equally sized flat facet for the proximomedial surface of metatar-
sal 4. The similarity in size between these facets may indicate that P. azael had less reduced sec-
ond and third digits than P. parvus, or alternatively may reflect a greater degree of articulation
of metatarsal 4 with the cuboid and thus reduction in articulation with the ectocuneiform. On
the lateral edge of the ectocuneiform is a smaller surface for articulation with the cuboid, while
its proximal face is almost completely occupied by a smooth, concave facet for the navicular.
Metatarsals. The metatarsals of P. azael are thick and robust, though more elongate than
the smaller P. parvus species. Both exhibit large median keels on the plantar aspect of their
heads, suggesting the presence of flexor sesamoids.
PLOS ONE | https://doi.org/10.1371/journal.pone.0221824 September 13, 2019
27 / 69
Osteology and functional morphology of the limbs in Palorchestidae
Fig 17. Associated pedal elements of Palorchestes azael. (A) articulated partial right side pes in dorsal view; (B) right
side metatarsal 4 NMV P29620 in (left to right, top to bottom) dorsal, medial, plantar, lateral, distal and proximal
views; (C) right side metatarsal 5 NMV P29619 in dorsal, medial, plantar, lateral, distal and proximal views; (D) right
side ectocuneiform NMV P29622 in dorsal, medial, lateral, proximal and distal views; (E) right side cuboid NMV
P29621 in dorsal, medial, lateral, proximal, distal and plantar views; (F) right side navicular NMV P29623 in dorsal,
proximal, distal and plantar views; (G) left side astragalus fragment NMV P254089 in dorsal, medial, plantar, lateral
and distal views; (H) right side calcaneus NMV P30723 in dorsal, medial, lateral and distal views. Scale bar 50 mm.
https://doi.org/10.1371/journal.pone.0221824.g017
PLOS ONE | https://doi.org/10.1371/journal.pone.0221824 September 13, 2019
28 / 69
Osteology and functional morphology of the limbs in Palorchestidae
Fig 18. Labelled illustrations of associated pedal elements of Palorchestes azael. (A) articulated partial right side pes
in dorsal view; right side ectocuneiform in (B) lateral and (C) distal views; right side cuboid in (D) medial and (E)
distal views; right side navicular in (F) proximal and (G) distal views; right side calcaneus in (H) dorsal and (I) medial
views; mirrored left side astragalus in (J) dorsal and (K) plantar views. Hatching indicates surface damage to cortical
bone, dashed lines indicate inferred bone contours. Abbreviations: af, astragalar facet; caf, calcaneoastragalar facet; cal;
calcaneus; ct, calcaneal tuber; cu, cuboid; cuf, cuboid facet; dlp, distolateral process; ect, ectocuneiform; ectf,
ectocuneiform facet; entf, entocuneiform facet; ltf, lateral tibial facet; mpt, medial plantar tuberosity; Mt4, metatarsal
4; Mt4f, metatarsal 4 facet; Mt5, metatarsal 5; Mt5f, metatarsal 5 facet; mtf, medial tibial facet; nav, navicular; nf,
navicular facet; pt, plantar tuberosity; st, sulcus tali; sut, sustentaculum tali. Scale bar 50 mm.
https://doi.org/10.1371/journal.pone.0221824.g018
Metatarsal 4. The fourth metatarsal of P. azael has a smoothly convex, triangular cuboid
facet, relatively more dorsoventrally elongate than in P. parvus and lacking the horizontal sul-
cus on its articular surface seen in the latter. The proximomedial facet for the ectocuneiform is
also more dorsoventrally extensive and more sagittally oriented than in other members of its
genus, instead resembling the condition in Nimbadon. There is no accompanying facet to sug-
gest articulation with the third metatarsal, unlike in P. parvus. Proximolaterally, the facet for
metatarsal 5 is slightly lunate and proportionally broader than in P. parvus, but not as dorso-
ventrally extensive, terminating above the volar tip of the metatarsal. The metatarsal head
appears hemispherical in dorsal view, while in ventral view its large central keel is flattened
proximally and projects strongly toward the plantar surface, flanked by smaller medial and
PLOS ONE | https://doi.org/10.1371/journal.pone.0221824 September 13, 2019
29 / 69
Osteology and functional morphology of the limbs in Palorchestidae
lateral keels. In distal view the head is domed and symmetrical, lacking the lateral cant seen in
P. parvus.
Metatarsal 5. Metatarsal 5 is a stout bone. The proximal tuberosity is damaged, leaving the
bone almost cylindrical in shape with flattened ventral and medial shaft surfaces. However, it
is likely that P. azael had an extensive proximal tuberosity here based on its occurrence in both
P. parvus AM F58870 and other diprotodontoids, and the shape of the eroded metatarsal sur-
face. Proximally the cuboid facet is smoothly convex and meets the facet for metatarsal 4 medi-
ally. There is a deep notch on the proximoventral edge of the bone, bounded medially by a
ventral projection of the facet for metatarsal 4 –this would have provided passage for digital
flexor tendons. The head appears hemispherical in dorsal view, the extensive median keel only
becoming visible in distal view. The medial keel is flattened and reduced, and the lateral keel is
sloped and reduced almost to absence.
Palorchestes parvus De Vis 1895
Referred material. Measurements for all referred material below are provided in
S1 Table.
AM F58870. Associated partial skeleton including: premaxilla fragment with incisor alveoli
matches syntype QMF789 (Woods [3], Fig 4; Trusler [36], Fig 5.19K); left humerus; right os
coxa (acetabulum and partial ilium); left femur (two fragments; proximal two-fifths, some
damage to posterior femoral neck and greater trochanter; distal epiphysis); left tibia (proximal
two-fifths fused in flexion to distal femur); partial left manus (missing phalanges 5 and all car-
pals except for trapezium); partial left pes (missing calcaneus and astragalus). Collected by G.
Hope from ‘cave at Wee Jasper, Punch Bowl Hill, below and left of Signature Cave’, NSW in
1977.
NMV P159792. Associated partial skeleton including: left tibia (shaft broken and repaired
above distal epiphysis, some cortical bone missing, referred based on morphological match
with tibia from AM F58870); left ulna (proximal fragment containing humeral and radial artic-
ulations with some damage to articular surfaces); right radius (proximal two-thirds with dam-
age to the capitular rim); left radius (distal fragment with well-preserved epiphysis). These are
smaller, yet mature elements registered as NMV P159792, leading us to conclude that P. azael
and P. parvus species are both represented, with a total MNI of 3 across all NMV P159792
specimens (two P. azael and one P. parvus). Collected by F. Spry from Buchan Caves (probably
Foul Air Cave), VIC in 1907.
Humerus (Figs 19 and 20). The humerus of P. parvus is wombat-like overall, with stout
and robust proportions and thick muscle attachment crests (Figs 19 and 20).
Head. The humeral head is sub-hemispherical and larger relative to its surrounding proxi-
mal humeral structures than in other palorchestids. The posteroinferior tip of the head appears
to ‘beak’ less from the posterior diaphysis in lateral view than in P. azael.
Greater tubercle. The greater tubercle in P. parvus projects slightly proximally to the
humeral head and has two major muscle attachment fossae; the fossa for insertion of m.
supraspinatus, which is a flattened ovoid shape on the proximal surface of the tubercle, and the
broad, posterolaterally-oriented fossa for m. infraspinatus. These fossae resemble those of P.
azael in relative proportions to the tubercle and to each other. In proximal view, the greater
tubercle is slightly more anteriorly positioned than in P. azael, but not as anterior as in Propa-
lorchestes. In lateral view the greater tubercle is extensive anteroposteriorly, to the same degree
as in P. azael.
Lesser tubercle. In anterior view, the apex of the lesser tubercle sits just inferior to the
humeral head, the highest relative position of the palorchestids. In this view it has a rounded
PLOS ONE | https://doi.org/10.1371/journal.pone.0221824 September 13, 2019
30 / 69
Osteology and functional morphology of the limbs in Palorchestidae
Fig 19. Left humerus of Palorchestes parvus AM F58870. (A) anterior; (B) lateral; (C) posterior; (D) medial; (E) proximal; (F) distal views. Scale bar 50 mm.
https://doi.org/10.1371/journal.pone.0221824.g019
medial margin which appears more inflated than any other taxon studied. In medial view, the
elongate attachment scar for m. subscapularis lies obliquely along the posterior margin of the
tubercle, similar to that of Propalorchestes. Viewed proximally, the lesser tubercle resembles
that of P. azael in all respects except in the contours of the bicipital groove, which in P. parvus
is more concave/less flattened and positioned nearer to the midline of the bone.
Deltopectoral crest. The deltopectoral crest in P. parvus is thick and well developed, being
similarly shaped to vombatids overall but differing in some key respects. The pectoral insertion
crest runs subvertically from the greater tubercle down the approximate midline of the
humeral shaft, and along its entire length the pectoral crest sharply overhangs the bicipital
groove medially. This is unlike vombatids where the entire deltopectoral crest is offset laterally
from the humeral shaft, the posterior part overhanging the brachialis fossa rather than the
anteromedial part overhanging the bicipital groove. The medial overhang of this crest begins
at the proximal metaphysis and is clearly visible in proximal view–this overhang is more
PLOS ONE | https://doi.org/10.1371/journal.pone.0221824 September 13, 2019
31 / 69
Osteology and functional morphology of the limbs in Palorchestidae
Fig 20. Labelled illustrations of the Palorchestes parvus left humerus AM F58870. (A) anterior; (B) lateral; (C) posterior; (D) lateral views. Hatching indicates
surface damage to cortical bone, dashed lines indicate inferred bone contours. Abbreviations: bg, bicipital groove; brf, fossa for m. brachialis origin; ca, capitulum;
del, deltoid insertion; eta, origin for m. epitrochleoanconeus; gt, greater tubercle; hh, humeral head; inf, fossa for insertion of m. infraspinatus; ldtm, insertion for
mm. latissimus dorsi and teres major; le, lateral epicondyle; lsc, lateral supracondylar crest; lt, lesser tubercle; me, medial epicondyle; of, olecranon fossa; pec, pectoral
crest; subf, fossa for insertion of m. subscapularis; supf, fossa for insertion of m. supraspinatus; sf, supracondylar foramen; tr, trochlea; tri, origin for humeral heads
of m. triceps brachii. Scale bar 50 mm.
https://doi.org/10.1371/journal.pone.0221824.g020
proximal than in P. azael, though the crest becomes less developed than the latter as it passes
distally. Like vombatids, there is an oblique crest coursing inferoanteriorly from the lateral
shaft to converge medially with the pectoral crest at its distal tip. The lateral edge of this obli-
que crest is likely to have been the attachment site for the scapular part of m. deltoideus. Just
inferior to the greater tubercle, a faint ridge runs vertically for a short distance on the antero-
lateral aspect of the pectoral crest–this probably marks the insertion point for the clavicular
deltoid. This insertion ridge in P. parvus is weaker than the vombatid condition but is totally
absent in P. azael, suggesting the clavicular deltoid played a reduced role in Palorchestes species
(and especially in P. azael) relative to wombats. In P. parvus the terminal point on the delto-
pectoral crest is swollen into a distinct tuberosity. Vombatids lack a tuberosity here, and the
shared distal end of these crests is more laterally positioned. Propalorchestes has a similar
tuberosity in this position, although it is not associated with the scapular deltoid insertion
which is situated proximally.
Tuberosity for mm. teres major and latissimus dorsi. The insertion scar for the combined
tendon of mm. teres major and latissimus dorsi is a lachrymiform fossa situated on the medial
humeral shaft overhanging its posterior border slightly more than halfway down its length. In
relative terms it is both the largest and most distally-positioned of these tuberosities among all
the palorchestids.
PLOS ONE | https://doi.org/10.1371/journal.pone.0221824 September 13, 2019
32 / 69
Osteology and functional morphology of the limbs in Palorchestidae
Diaphysis. In anterior and lateral views, the P. parvus humeral shaft appears straight overall
though somewhat distorted by the crests and tubercles along its length. Like P. azael, in P. par-
vus attachment scars for m. triceps brachii lie on the posterior shaft surface, though the fossa
for the origin of m. brachialis below the lateral lip of the humeral head is less pronounced. Infe-
riorly, a deep triangular olecranon fossa lies above the posterior articular surface of the capitu-
lum, facilitating at least some extension of the elbow. The posteroinferior diaphysis lacks the
marked rugosity for m. epitrochleoanconeus seen in P. azael.
Lateral epicondyle. The lateral epicondyle projects slightly from the upper rim of the capit-
ulum when viewed anteriorly. In lateral view, it appears pitted and rugose, resembling that of
Propalorchestes in its shape and extent. From this epicondyle, the lateral supracondylar crest
extends proximally as a thin sheet of bone, its lateral margin damaged but the remnant charac-
teristic vombatiform hook still visible.
Trochlea. In P. parvus the trochlea faces inferiorly but due to slight convexity both medio-
laterally and anteroposteriorly its articular surface is just visible in anterior view. This repre-
sents an intermediate morphology between the curved, wombat-like condition in
Propalorchestes and the flat trochlea of P. azael. Like the latter species, in anterior view the
trochlea and capitulum project equally distally, though the angle between the two processes is
less acute in P. parvus. In distal view the trochlea is ovoid and a little larger than the capitulum
in dorsoventral depth. It is aligned with the capitulum in the dorsal (coronal) plane of the
humerus.
Supracondylar foramen. The supracondylar foramen in P. parvus is deep and ovoid,
spanned by a bridge broader and more robust than in other palorchestids.
Medial epicondyle. The medial epicondyle projects strongly medially from the vertical axis
of the humeral shaft. In anterior view, the overall shape and relative extent the P. parvus medial
epicondyle resembles that of Propalorchestes, being less wide and rounded than in P. azael. In
medial view the P. parvus epicondyle is bulbous and thick anteroposteriorly.
Ulna (Figs 21 and 22). The only ulna specimens are fragments preserving the proximal
articulations and short adjacent portion of the olecranon (Figs 21 and 22). The preserved mor-
phology suggests that the olecranon would have been consistent with other palorchestids,
aligned with the diaphysis and not posteriorly deflected as in diprotodontids, and not medially
deflected as in Propalorchestes, instead resembling the morphology in P. azael. The trochlear
surface is broad and subcircular like that of Vombatus and Ng. bonythoni, not narrowed and
elongate as in P. azael. The tallest point of the anconeal process is much more anteriorly ori-
ented than the laterally-deflected equivalent in P. azael, and creates a C-shaped trochlear notch
in medial view similar to those of Propalorchestes and extant wombats. The coronoid process
is relatively low and proximodistally thick like in P. azael, not thin and tall as in Vombatus.
The radial notch and facet for the humeral capitulum lie in the same dorsal plane, more in line
with one another than in P. azael where the radial notch is more dorsally offset than the capitu-
lar facet. The radial notch is large, approximately half the proximodistal length of the adjacent
trochlear surface. In P. azael the radial notch is relatively shorter. The pit for attachment of the
annular ligament between the trochlear and radial notches is shallower in P. parvus than in P.
azael. The ulnar tuberosity is small and less rugose than in P. azael, and in a more proximolat-
eral position. The posterior border of the ulnar shaft immediately dorsal to the coronoid pro-
cess is narrower mediolaterally than in P. azael, indicating a less robust bone overall.
Radius (Figs 23 and 24). No complete radius is known for P. parvus (Figs 23 and 24). The
overlapping morphology of the proximal fragment is similar to that of P. azael, the principal
difference being the medial and lateral borders are less curved, indicating a straighter radius
overall. The radial tuberosity for insertion of m. biceps brachii is further distal to the head than
in P. azael. The distal fragment is trapezoidal in profile like in P. azael, though in P. parvus it is
PLOS ONE | https://doi.org/10.1371/journal.pone.0221824 September 13, 2019
33 / 69
Osteology and functional morphology of the limbs in Palorchestidae
Fig 21. Left ulna fragment of Palorchestes parvus NMV P159792. (A) anterior; (B) lateral; (C) posterior; (D) medial
views. Scale bar 50 mm.
https://doi.org/10.1371/journal.pone.0221824.g021
more dorsoventrally flattened. The radiocarpal surface most resembles Ngapakaldia in form,
in particular the relative size of the styloid and extent to which it is inset from the medial bor-
der of the epiphysis. The flattened dorsal surface is notched for mm. extensor carpi radialis and
extensor digitorum communis tendons.
Fig 22. Labelled illustrations of Palorchestes parvus left ulna fragment NMV P159792. (A) anterior; (B) lateral
views. Hatching indicates surface damage to cortical bone. Abbreviations: ap, anconeal process; cf, capitular facet; cp,
coronoid process; rn, radial notch; tn, trochlear notch; ut, ulnar tuberosity. Scale bar 50 mm.
https://doi.org/10.1371/journal.pone.0221824.g022
PLOS ONE | https://doi.org/10.1371/journal.pone.0221824 September 13, 2019
34 / 69
Osteology and functional morphology of the limbs in Palorchestidae
Fig 23. Radius fragments of Palorchestes parvus NMV P159792. Right side radius in (A) dorsal; (B) medial; (C)
ventral; (D) lateral; (E) proximal views. Distal left side radius in (F) dorsal; (G) ventral; (H) distal views. Scale bar 50
mm.
https://doi.org/10.1371/journal.pone.0221824.g023
Manus (Figs 25 and 26). The P. parvus manus is represented by an associated set of stout
metacarpals and phalanges (Figs 25 and 26).
Trapezium. The trapezium is very similar to that of Ngapakaldia bonythoni overall, being
only a little larger and more proximodistally compressed. It has a saddle-shaped facet for the
first metacarpal on its palmar-distal face with slightly flatter contours than the former species. A
shallow concave facet for the trapezoid lies on its lateral face (though this is somewhat eroded),
and a well-defined facet for the palmar process of the scaphoid sits on its proximal surface. The
orientation of its articulations indicates the pollex would have been abducted to a similar degree
to that seen in Ngapakaldia and much more than in Zygomaturus or extant wombats.
Metacarpals. The metacarpals overall are stouter and more distally expanded than in Vom-
batus, zygomaturines and Ngapakaldia. They are compact, with contoured margins for adja-
cent contact over most of their length. Distal condyles are sub-hemispherical in lateral view,
with pronounced keels on the palmar surface whose contours are visible even in dorsal view.
The metacarpals are deeply pitted distolaterally and distomedially for attachment of collateral
ligaments.
PLOS ONE | https://doi.org/10.1371/journal.pone.0221824 September 13, 2019
35 / 69
Osteology and functional morphology of the limbs in Palorchestidae
Fig 24. Labelled illustration of estimated reconstruction of Palorchestes parvus right radius. (A) dorsal; (B) ventral
views. Length estimated based on Palorchestes azael radius (Fig 7), left side distal fragment mirrored and scaled to fit.
Hatching indicates surface damage to cortical bone, dashed lines indicate inferred bone contours. Abbreviations: cs,
carpal surface; ecr, notch for tendon of m. extensor carpi radialis; ioc, interosseous crest; rt, radial tuberosity; sp,
styloid process. Scale bar 50 mm.
https://doi.org/10.1371/journal.pone.0221824.g024
Metacarpal 2. The second metacarpal is considerably smaller than its neighbours, with a
sharply tapering proximal end resembling that of Ngapakaldia. The metacarpal presents a
short, oblique dorsolateral facet for the magnum/capitatum and convex medial facet where the
trapezium cups its medial edge.
Metacarpal 3. The third metacarpal has a triangular proximal facet for the magnum/capita-
tum, which in dorsal view has a chevron-shaped notch as in Ngapakaldia, and a broad, flat
facet laterally for metacarpal 4.
PLOS ONE | https://doi.org/10.1371/journal.pone.0221824 September 13, 2019
36 / 69
Osteology and functional morphology of the limbs in Palorchestidae
Fig 25. Associated partial left manus of Palorchestes parvus AM F58870. (A) articulated manus in dorsal view; (B) metacarpal 4 in (left to right) lateral view;
(C) metacarpals 4–2 (cemented together) in dorsal view; (D) metacarpal 2 in medial view; (E) metacarpals 2–4 in palmar view; (F) metacarpals 4–2 in proximal
view; (G) metacarpals 4–2 in distal view; (H) metacarpal 5 in (left to right, top to bottom) dorsal, lateral, plantar, medial, distal and proximal views; (I)
trapezium in dorsal, proximal, lateral, distal and medial views; (J-K) digit 1/pollex ungual and proximal phalanges in dorsal, lateral, plantar, medial, proximal
and distal views; (L-N) digit 2 ungual, intermediate and proximal phalanges in dorsal, lateral, plantar, medial, proximal and distal views; (O-Q) digit 3 ungual,
PLOS ONE | https://doi.org/10.1371/journal.pone.0221824 September 13, 2019
37 / 69
Osteology and functional morphology of the limbs in Palorchestidae
intermediate and proximal phalanges in dorsal, lateral, plantar, medial, proximal and distal views; (R-T) digit 4 ungual, intermediate and proximal phalanges in
dorsal, lateral, plantar, medial, proximal and distal views. Scale bar 50 mm.
https://doi.org/10.1371/journal.pone.0221824.g025
Metacarpal 4. The fourth metacarpal is the longest and most robust as is common in mar-
supials. Proximally, the facet for the unciform/hamatum is triangular, with an oblique median
groove. On the proximolateral surface, two ovoid tubercles separated by a sulcus are present
for articulation with metacarpal 5.
Metacarpal 5. Metacarpal 5 has a strongly developed tuberosity bulging proximolaterally
from the proximal quarter of the bone and extending beyond the unciform/hamatum facet.
Due to the overall shape of the metacarpal, this tuberosity is more exaggerated in relation to
the shaft than in Zygomaturus or Phascolonus but does not approach the enormous equivalent
in Diprotodon. There is a lunate dorsomedial facet for articulation with the fourth metacarpal.
Fig 26. Labelled illustration of articulated Palorchestes parvus left manus AM F58870. (A) articulated manus in
dorsal view; (B) digit 4 in lateral view. Hatching indicates surface damage to cortical bone, dashed lines indicate
inferred bone contours. Abbreviations: dij, distal interphalangeal joint; Ip1-4, intermediate phalanges 1–4; Mc2-5,
metacarpals 2–5; mk, median keel; mpj, metacarpophalangeal joint; pij, proximal interphalangeal joint; Pp1-4,
proximal phalanges 1–4; Tra, trapezium; Up1-4, ungual phalanges 1–4. Scale bar 50 mm.
https://doi.org/10.1371/journal.pone.0221824.g026
PLOS ONE | https://doi.org/10.1371/journal.pone.0221824 September 13, 2019
38 / 69
Osteology and functional morphology of the limbs in Palorchestidae
As in Ngapakaldia, the facet for the lateral process of the unciform does not extend as far lat-
erally as it does in Vombatus.
Proximal phalanges. There are well-developed pits and crests for collateral ligaments of
the digital joints throughout the hand. The proximal phalanges are short, broad and flattened
dorsoventrally with dished-in dorsal surfaces. Their bases are markedly asymmetrical in dorsal
view, with deeply concave metacarpal sockets buttressed strongly on one side (4 and 1 laterally,
2 and 3 medially). This asymmetry is presumably to resist different prevailing forces acting on
each metacarpophalangeal joint. This buttressing pattern was also noted in Nimbadon and
may be indicative of similar range of motion (ROM) and loading regimes at these joints; how-
ever, the P. parvus proximal phalanges lack the palmar tuberosities of Nimbadon, so similar
overall use of the manus seems unlikely. The heads are wide, each with two flared distal con-
dyles tilted ventrally. These create joint surfaces necessitating a flexed posture for the proximal
interphalangeal joints. Between these articular surfaces there are deep median incisurae to
accommodate the corresponding dorsomedian processes of the intermediate phalanges. The
proximal phalanx of the pollex is slightly different to the other digits, being more gracile over-
all, with a more compact and bulbous distal end allowing greater ROM in extension.
Intermediate phalanges. The intermediate phalanges are short and squat, similar in overall
proportions to those of vombatids and Diprotodon, with broader proximal than distal ends in
dorsal view. In proximal view the joint surfaces approach isosceles trapezoids in shape. Their
proximal articular surfaces are divided by dorsomedian processes into two deeply excavated fos-
sae to accommodate the condyles of the preceding phalanges. These fossae are subequal in size,
with 3 and 4 having slightly larger lateral surfaces. There is slight axial torsion in the orientation
of the proximal and distal articular ends, especially in the fourth digit where the distal end is
medially rotated relative to the proximal. The bicondylar distal articulations of these intermediate
phalanges are most strongly congruent with their unguals in a flexed distal interphalangeal joint
position. These condyles have a more reduced dorsal articular surface than in Nimbadon or Neo-
helos, indicating a reduced extension ROM. The intermediate phalanges of the manus can be dis-
tinguished from those of the pes by the shape of the proximoventral border; manual phalanges
have a distinct median excavation between the condylar fossae, making a “W” shaped ventral
contour. Pedal intermediate phalanges 4 and 5 are much flatter, with little to no concavity.
Ungual phalanges. No individual ungual from the manus of AM F58870 is complete and
undamaged, but the intact morphology can be understood in composite across the specimens
available. In shape and proportion they agree with unguals of P. azael, being only slightly
smaller than the smallest example from that species (see S1 Table). Across the digits the
unguals differ greatly in size, with the pollex being smallest and fourth ungual the largest.
Os coxa (Figs 27 and 28). The single os coxa specimen for P. parvus is missing the iliac
blade and most of the pubis and ischium (Figs 27 and 28). The remaining element appears
similar to that of P. azael, with the same positioning of the ilium, ischium and pubis relative to
the acetabulum in lateral view, though the acetabulum may be slightly smaller relative to its
surrounding morphology in P. parvus.
Ischial body. The proximal-most part of the ischial body is preserved in P. parvus. It is
thicker and more triangular in section than in P. azael, but its full extent and shape of the
ischial tuberosity is not known.
Acetabulum. The acetabulum in P. parvus is slightly more circular than in P. azael, with
relatively less overhang and narrower articular surface area posterior to the acetabular notch.
Anterior inferior iliac spine. The anterior inferior iliac spine is a broad, triangular scar for
origin of m. rectus femoris. It is relatively wider and more deeply pitted than in P. azael and lies
higher on the iliac body relative to the acetabular rim, similar to that of Neohelos though more
rugose.
PLOS ONE | https://doi.org/10.1371/journal.pone.0221824 September 13, 2019
39 / 69
Osteology and functional morphology of the limbs in Palorchestidae
Fig 27. Left os coxa fragment of Palorchestes parvus AM F58870. (A) lateral; (B) dorsal; (C) medial; (D) ventral views. Scale bar 50 mm.
https://doi.org/10.1371/journal.pone.0221824.g027
Iliopectineal eminence. The iliopectineal eminence is more diffuse and flattened compared
to that of P. azael.
Auricular surface. The auricular surface in P. parvus is relatively more superoinferiorly
extensive and more irregular than in P. azael, lacking the distinct facets seen in the latter. It is
bordered posteriorly by a more open greater sciatic notch than in the larger species.
Femur (Figs 29, 30 and 31). The femur of P. parvus is represented by associated proximal
and distal ends, missing the central femoral diaphysis (Figs 29 and 30). For this reason, the
length of the intact bone can only be estimated (378 mm predicted based on ratio of proximal
breadth to length in P. azael, see Fig 31). In shape and proportions it strongly resembles the P.
azael femur, albeit with a broader distal epiphysis relative to proximal breadth (0.86 in P. par-
vus, 0.80 in P. azael).
Femoral head. The femoral head is hemispherical, with approximately equal articular sur-
face available anteriorly, medially and posteriorly. The head is anteriorly offset from the
diaphyseal axis as in the larger P. azael, but to a slightly lesser extent. The neck is short and
does not project superomedially from the proximal shaft to the degree seen in the long-necked
femora of Zygomaturus or Diprotodon, or to a lesser extent Neohelos.
Greater trochanter. The greater trochanter is proximally tapered and anteroposteriorly
deep. In anterior view its rugose muscle attachments are superolaterally located as in P. azael,
more so than the lower and more anterior trochanter of Phascolonus. Unlike the femur of P.
azael, the P. parvus greater trochanter lacks the distinct swelling at the distal end where the
PLOS ONE | https://doi.org/10.1371/journal.pone.0221824 September 13, 2019
40 / 69
Osteology and functional morphology of the limbs in Palorchestidae
Fig 28. Labelled illustrations of Palorchestes parvus left os coxa fragment AM F58870. (A) lateral; (B) dorsal; (C) medial
views. Hatching indicates surface damage to cortical bone. Abbreviations: acf, acetabular fossa; acn, acetabular notch; aiis,
anterior inferior iliac spine; aur, auricular surface; gsn, greater sciatic notch; ipe, iliopectineal eminence; obf, obturator
foramen. Scale bar 50 mm.
https://doi.org/10.1371/journal.pone.0221824.g028
epiphysis merges with the anterolateral femoral shaft and extends a shorter distance down the
shaft. Posteriorly, the trochanteric fossa is elongate and deep as in P. azael.
Lesser trochanter. The main tuber of the lesser trochanter is eroded so its full medial extent
is unknown, however it appears to have been more robust and its flange more developed than
P. azael, while proximodistally a little shorter. At the base of the lesser trochanter on the medial
diaphysis there is a rugose area immediately proximal to the break in the shaft.
Diaphysis. The bulk of the femoral diaphysis is missing, leaving only the proximal and dis-
tal portions, however from the section profiles it appears similar in shape to that of P. azael.
Notable on the distolateral diaphysis is a deep and narrow, proximodistally elongate muscle
scar for the lateral head of m. gastrocnemius. In contrast to P. azael, this scar is deeply inset,
being similar to that of Phascolarctos in shape and proximal extent.
Medial condyle. In anterior view, as in P. azael, the distal surfaces of the medial and lateral
condyles sit approximately level. Viewed distally the medial condyle is more anteroposteriorly
extensive than the lateral, but they are more alike here in P. parvus than in any of the other
diprotodontoid femora studied. The medial condyle in this view also appears less bulbous and
presents a flattened surface posteriorly, unlike the inflated and more laterally-canted condyle
in P. azael.
Lateral condyle. The lateral condyle is short, broad and quadrangular in distal and poste-
rior views, very closely resembling the morphology of P. azael in all respects.
Patellar surface. The patellar surface of the distal femur is anteroposteriorly and proximo-
distally short. The medial condyle of P. parvus projects anteriorly even less than in P. azael.
PLOS ONE | https://doi.org/10.1371/journal.pone.0221824 September 13, 2019
41 / 69
Osteology and functional morphology of the limbs in Palorchestidae
Fig 29. Proximal left femur of Palorchestes parvus AM F58870. (A) anterior view; (B) posterior view; (C) proximal
view. Scale bar 50 mm.
https://doi.org/10.1371/journal.pone.0221824.g029
Overall this creates a distal femur in stark contrast to those of other diprotodontoids and Phas-
colonus, again resembling extant wombat morphology most closely.
Tibia (Figs 30, 32 and 33). The tibia of P. parvus is very similar morphologically to that of
P. azael, the principal difference being that for approximately the same length it is more grac-
ile, an allometric difference expected for the smaller species (Figs 30, 32 and 33). Differences to
the P. azael morphology described above are provided below.
Proximal articulations. The medial condyle, though its posteromedial border is eroded,
appears much smaller and less posteriorly extensive than in P. azael, probably representing a
weightbearing allometric difference. The intercondylar eminence is substantially narrower
mediolaterally in P. parvus than in P. azael. The fibular facet is preserved in both P. parvus
specimens. It is distinct and expanded proximally to form a more laterally-facing circular
shape than the flattened, inferiorly-directed oval facet of Vombatus, being more similar to that
of Phascolarctos. This distinct facet does not resemble the more diffuse depression for the fib-
ula in Diprotodon or Zygomaturus tibiae. This may indicate a more mobile and less weight-
bearing proximal tibiofibular articulation in P. parvus. The popliteal fossa on the posterior
surface is more deeply furrowed than in P. azael, similar to that of vombatids.
Diaphysis. In P. parvus the diaphysis is more slender overall. In mediolateral view the tibial
crest has a concave profile immediately below the tibial tuberosity before flaring out again at
the point of insertion for m. gracilis roughly a third of the way down the diaphysis, strongly
PLOS ONE | https://doi.org/10.1371/journal.pone.0221824 September 13, 2019
42 / 69
Osteology and functional morphology of the limbs in Palorchestidae
Fig 30. Cemented fragments of distal left femur and proximal left tibia of Palorchestes parvus AM F58870. Distal left femur fragment photographed in orthogonal
views with tibia fragment greyed out (A-D, I) and proximal left tibia fragment photographed in orthogonal views with femur fragment greyed out (E-H) in (A, E) anterior;
(B, F) medial; (C, G) posterior; (D, H) lateral; (I) distal views. Scale bar 50 mm.
https://doi.org/10.1371/journal.pone.0221824.g030
resembling the Vombatus condition. The interosseous border is more rounded than in vomba-
tids and diprotodontids, but not quite to the extent seen in P. azael.
Distal articulations. The medial malleolus is eroded in the only specimen preserving the
distal tibia, but from the available material it appears similar in proportion and extent to that
of P. azael. The remaining astragalar surface on the distal tibia is much more steeply inclined
and less laterally extensive than in P. azael, though similarly convex.
PLOS ONE | https://doi.org/10.1371/journal.pone.0221824 September 13, 2019
43 / 69
Osteology and functional morphology of the limbs in Palorchestidae
Fig 31. Labelled illustrations of estimated reconstruction of Palorchestes parvus left femur AM F58870. (A) anterior; (B) posterior; (C)
distal views. Length estimated based on Palorchestes azael femur (Fig 13). Hatching indicates surface damage to cortical bone, dashed lines
indicate inferred bone contours. Abbreviations: fh, femoral head; gas, origin for m. gastrocnemius lateral head; gt, greater trochanter; lc, lateral
condyle; lt, lesser trochanter; mc, medial condyle; ps, patellar surface; tf, trochanteric fossa. Scale bar 50 mm.
https://doi.org/10.1371/journal.pone.0221824.g031
Pes (Figs 34 and 35). The digital posture of the pes in P. parvus appears clearly planti-
grade, with the shape of the proximal interphalangeal joints indicating a habitually straight-
ened position unlike the flexed posture in the manual equivalent (Figs 34 and 35). The second
and third digits are extremely reduced relative to their robust lateral counterparts, more so
than in Nimbadon, Ng. tedfordi or Thylacoleo.
PLOS ONE | https://doi.org/10.1371/journal.pone.0221824 September 13, 2019
44 / 69
Osteology and functional morphology of the limbs in Palorchestidae
Fig 32. Right tibia of Palorchestes parvus NMV P159792. (A) Anterior; (B) lateral; (C) posterior; (D) medial; (E) proximal; (F) distal
views. Scale bar 50 mm.
https://doi.org/10.1371/journal.pone.0221824.g032
Ectocuneiform. The ectocuneiform is intermediate in form between those of Ng. tedfordi
and P. azael, being dorsoventrally longer and proximodistally thicker than the former, but
with relatively smaller, dorsally positioned and more circular metatarsal facets than the latter.
The facet for metatarsal 3 in particular is small, commensurate with the small proximal facet of
this highly reduced syndactylous digit. Proximally, the facet for the navicular is flattened and
dorsoventrally elongate as in P. azael.
Entocuneiform. The entocuneiform is highly eroded, with most facet edges and bony mar-
gins incomplete. Overall it is elongate, slightly inflected, and appears larger relative to the
navicular than in other diprotodontoids. Distomedially the preserved surface of the facet for
metatarsal 1 is quite flat, being less saddle-shaped than in Nimbadon.
PLOS ONE | https://doi.org/10.1371/journal.pone.0221824 September 13, 2019
45 / 69
Osteology and functional morphology of the limbs in Palorchestidae
Fig 33. Labelled illustrations of the Palorchestes parvus right tibia NMV P159792. (A) Anterior; (B) lateral; (C) posterior; (D) medial; (E) proximal views.
Hatching indicates surface damage to cortical bone, dashed lines indicate inferred bone contours. Abbreviations: af, astragalar facet; ff, fibular facet; gr,
insertion of m. gracilis; ie, intercondylar eminence; lc, lateral condylar surface; mc, medial condylar surface; mm, medial malleolus; pf, popliteal fossa; ptm,
posterior tubercle of medial malleolus; st, insertion of m. semitendinosus; iol, interosseus line; tc, tibial crest; tt, tibial tuberosity. Scale bar 50 mm.
https://doi.org/10.1371/journal.pone.0221824.g033
Navicular. The navicular is similar in mediolateral length to that of Ng. tedfordi, but is ante-
roposteriorly much thicker and dorsoventrally compressed, with a less concave facet for the
astragalus proximolaterally. Due to its increased thickness the P. parvus navicular has a much
larger convex facet surface for the ectocuneiform and mesocuneiform distally, but a similarly
elongate articular surface for the entocuneiform medially. The navicular is smaller relative to
its neighbouring entocuneiform than in Ng. tedfordi or Nimbadon pedes.
Metatarsals 2 and 3. As is typical of diprotodontian marsupials, these elements are syndac-
tylous and strongly reduced compared to the other highly robust pedal digits, with the third
being slightly larger than the second. At their base they are triangular, with convex articular
surfaces for the ecto- and mesocuneiforms. The third metatarsal has a dorsolateral facet at the
base for articulation with metatarsal 4 as in Diprotodon. The diaphyses of both bones appear
similar in dimensions. The distal ends are missing so lengths are unknown.
Metatarsal 4. The fourth metatarsal is shorter and relatively more robust than in P. azael.
The concave, triangular cuboid facet is transected horizontally by a shallow sulcus. A facet for
the ectocuneiform extends mediodistally from the medial edge of the cuboid facet and is
smaller and more dorsally positioned than in P. azael. The mediodistal tip of this ectocunei-
form facet projects from the medial shaft and on the ventral surface of this projection lies a
PLOS ONE | https://doi.org/10.1371/journal.pone.0221824 September 13, 2019
46 / 69
Osteology and functional morphology of the limbs in Palorchestidae
Fig 34. Associated left pes of Palorchestes parvus AM F58870. Articulated pes in (A) dorsal and (B) mediodorsal views; (C-E) digit IV ungual,
intermediate and proximal phalanges in (left to right) dorsal, lateral, plantar, medial, proximal and distal views; (F) metatarsal 4 in (left to right, top to
bottom) dorsal, lateral, plantar, medial, distal and proximal views; (G-H) digit 5 ungual and intermediate phalanges in dorsal, lateral, plantar, medial,
proximal and distal views; (I) metatarsal 5 in (left to right, top to bottom) dorsal, lateral, plantar, medial, distal and proximal views; (J) metatarsals 2 and
3 (cemented together) in dorsal, plantar and proximal views; (K) navicular in dorsal, proximal and distal views; (L) ectocuneiform in dorsal, lateral,
medial, proximal and distal views. Scale bar 50 mm.
https://doi.org/10.1371/journal.pone.0221824.g034
small triangular facet for articulation with the dorsolateral facet of metatarsal 3. Viewed dis-
tally, the dorsalmost articular surface head is laterally canted, with slight reduction of the lat-
eral keel as in the fifth metatarsal.
Metatarsal 5. The fifth metatarsal is slightly shorter than the fourth when measured
between the base and head. But overall the fifth is longer owing to the very large lateral tuber-
osity that tapers proximally beyond the articular surface for the cuboid. This lateral tuberosity
is similar in shape and proportion to that of Phascolonus, being a continual subtriangular
flange originating at the distal end of the metatarsal. This is distinct from the more bulbous
equivalent in diprotodontids and extant wombats. The lateral tuberosity lacks the dorsal uptick
of the flattened Ngapakaldia equivalent. This tuberosity would have provided attachment for
the m. peroneus brevis proximally, with another smaller tubercle distolaterally for the m.
PLOS ONE | https://doi.org/10.1371/journal.pone.0221824 September 13, 2019
47 / 69
Osteology and functional morphology of the limbs in Palorchestidae
Fig 35. Labelled illustration of the articulated Palorchestes parvus left pes AM F58870 in dorsal view. Hatching
indicates surface damage to cortical bone. Abbreviations: ect, ectocuneiform; ent, entocuneiform; Ip4-5, intermediate
phalanges 4–5; mes?, possible mesocuneiform; Mt2-5, metatarsals 2–5; nav, navicular; Pp2-4, proximal phalanges 2–4;
Pp5?, possible proximal phalanx 5; Up3-5, ungual phalanges 3–5. Scale bar 50 mm.
https://doi.org/10.1371/journal.pone.0221824.g035
abductor digiti V. The smooth triangular cuboid facet on the base curves medially to provide
articulation with the matching contour of the neighbouring fourth metatarsal. In ventral aspect
the central keel of the head is laterally inclined with the lateral keel reduced, in contrast to the
enlarged lateral keel described for Nimbadon, or subequal lateral keel in Ngapakaldia. Though
we lack a proximal phalanx for this digit this morphology may indicate an abducted posture
for the P. parvus fifth digit, and along with its lateral tuberosity shows high lateral loading of
the pes.
Phalanges. The proximal interphalangeal joints of the pes are less restricted in extension
due to slightly smaller dorsomedian crests. They were able to extend to a straightened position
unlike the flexed manual equivalent. When viewed mediolaterally, their proximal articular fos-
sae are less tightly concave and more open than the intermediate phalanges of the manus. In
distal view, the condyles are more flared ventrally to create a wider sub-articular angle than the
steep, more vertically-oriented distal condyles in the manus. In proximal view, the intermedi-
ate phalanges are slightly asymmetrical, with broader lateral facets.
Unguals. The non-syndactylous pedal digits 4 and 5 have unguals of very similar morphol-
ogy to those of the manus. The dorsoventral extent of the condylar fossae and flattened plantar
areas on the flexor tubercles indicate these pedal unguals were habitually held in maximal
PLOS ONE | https://doi.org/10.1371/journal.pone.0221824 September 13, 2019
48 / 69
Osteology and functional morphology of the limbs in Palorchestidae
extension with their straight dorsal border parallel to the ground. Presumably this was to
accommodate an even more ventrally recurved keratin claw sheath. The much smaller ungual
for digit 3 is poorly preserved but has the same twin lunate condylar fossae which are slightly
broader ventrally than dorsally and separated by a median keel, as those seen in other palorch-
estid unguals. Ventrally, this ungual lacks the plantar flexor tubercle seen in the manual and
large pedal unguals, likely reflecting its reduced weightbearing role and weaker flexor tendon.
The ungual process is broken off and not preserved.
Based on this associated specimen, the pedal unguals of digits 4 and 5 in P. parvus appear to
have a dorsoventrally deeper flexor tubercle, with more deeply concave articular facets lying
more dorsally on the bone than their manual equivalents.
Propalorchestes sp. Murray 1986
Referred material. Measurements for all referred material below are provided in
S1 Table.
All specimens were collected by T. H. Rich from Top Site, Bullock Creek, NT. As two Pro-
palorchestes species are known from the Bullock Creek local fauna, and no postcrania are yet
assigned to either, we refer the following elements to Propalorchestes sp. on the basis of their
morphological distinctiveness from the other diprotodontoids which are well known from this
locality, Neohelos spp. The articular surfaces of the unassociated humerus and ulna agree in
shape but are from individuals of slightly different sizes.
NTM P87115-6. Left humerus (intact but for damage to the lateral margin of the lateral
supracondylar crest). Photographed in Vickers-Rich et al. [72], pg. 170, Fig 232.
NMV P253947. Right ulna (proximal half, some damage to radial facet).
NMV P179370. Ungual phalanx.
Humerus (Figs 36 and 37). The Propalorchestes humerus is a robust bone with a straight
diaphysis and broad epiphyses, particularly the distal portion which is both laterally and medi-
ally expanded. The shaft has pronounced flanges anteriorly and laterally, for deltopectoral and
wrist extensor muscles respectively (Figs 36 and 37).
Overall the humerus of Propalorchestes bears the strongest resemblance to that of Ngapakal-
dia, though it is larger than that of the latter (see S1 Table and S1 Fig).
Head. The humeral head is sub-hemispherical, with a posteriorly offset position creating
the distinct ‘beak’ in medial view present in all vombatiforms. The head is slightly broader
mediolaterally than anteroposteriorly, but not to the same degree seen in P. azael or the vom-
batid species and totally unlike the flattened, laterally-expanded head of Neohelos.
Greater tubercle. The greater tubercle in Propalorchestes projects proximally beyond the
humeral head as in other palorchestids, unlike the in zygomaturine or diprotodontine species
where it is subequal in height to the humeral head. The tubercle has two distinct attachment
scars; a small fossa for m. supraspinatus presenting a flat surface superoanteriorly, and a larger
fossa for m. infraspinatus with a directly lateral orientation, similar to that seen in other
palorchestids. In lateral view the greater tubercle occupies two thirds of the anteroposterior
extent of the proximal humerus, like other palorchestids. In proximal view the greater tubercle
in Propalorchestes lies more anteriorly on the epiphysis than the laterally-positioned equivalent
in P. azael. The tubercle overhangs medially to create a deep notch at the proximal part of the
intertubercular sulcus, as seen in P. parvus. This contrasts with the shallow groove here seen in
diprotodontines and vombatids.
Lesser tubercle. In Propalorchestes the lesser tubercle presents a flattened anterior surface,
with margins slightly pointed proximally and rounded medially. This medial margin projects
slightly more in Propalorchestes than in the other diprotodontoids, though not as much as in
vombatids. In proximal view the tubercle is larger relative to the humeral head and more ante-
riorly situated than in other palorchestids, most closely resembling the extant wombat species
PLOS ONE | https://doi.org/10.1371/journal.pone.0221824 September 13, 2019
49 / 69
Osteology and functional morphology of the limbs in Palorchestidae
Fig 36. Left humerus of Propalorchestes sp. NTM P87115-6. (A) anterior; (B) lateral; (C) posterior; (D) medial; (E) proximal; (F) distal views. Scale bar 50 mm.
https://doi.org/10.1371/journal.pone.0221824.g036
in this respect. In medial view the well-defined insertion scan for m. subscapularis is obliquely
oriented, descending posteriorly at an angle ~25˚ from vertical. This is similar to the orienta-
tion seen in P. parvus, steeper than in P. azael (~40˚) and vombatids (~30˚) and much steeper
than in diprotodontids (> 55˚).
Pectoral crest. Descending from the broad greater tubercle, the pectoral attachment in Pro-
palorchestes narrows to a tall crest which runs sub-vertically down the approximate midline of
the anterior humeral shaft. In medial view the crest is a broad-based isosceles triangle, reaching
its apex slightly proximal to the midpoint of the humeral shaft. This makes it the shortest pec-
toral crest among the vombatiforms relative to total humeral length. Viewed anteriorly, its
medial edge curls slightly over the bicipital sulcus, but to a far lesser extent than in other
palorchestids. The apex of the crest terminates in a bulbous protuberance resembling that of P.
parvus. However, the deltoid insertion does not converge directly on this protuberance as in
the latter species. Instead, the insertion for the deltoid muscle is a poorly-defined curved scar.
PLOS ONE | https://doi.org/10.1371/journal.pone.0221824 September 13, 2019
50 / 69
Osteology and functional morphology of the limbs in Palorchestidae
Fig 37. Labelled illustrations of the Propalorchestes sp. left humerus NTM P87115-6. (A) anterior; (B) lateral; (C) posterior; (D) medial views. Hatching
indicates surface damage to cortical bone, dashed lines indicate inferred bone contours. Abbreviations: bg, bicipital groove; brf, fossa for m. brachialis origin; ca,
capitulum; del, deltoid insertion; gt, greater tubercle; hh, humeral head; inf, fossa for insertion of m. infraspinatus; ldtm, insertion for mm. latissimus dorsi and
teres major; le, lateral epicondyle; lsc, lateral supracondylar crest; lt, lesser tubercle; me, medial epicondyle; of, olecranon fossa; pec, pectoral crest; rf, radial fossa;
subf, fossa for insertion of m. subscapularis; supf, fossa for insertion of m. supraspinatus; sf, supracondylar foramen; tr, trochlea; tri, origin for humeral heads of
m. triceps brachii. Scale bar 50 mm.
https://doi.org/10.1371/journal.pone.0221824.g037
This scar is bent anteromedially from a site beneath the greater tubercle to merge with the pec-
toral crest approximately halfway down its length (a quarter of the way down the shaft). This is
distinct from the morphology in P. parvus and vombatids where the deltoid insertion ridge is
discrete and runs parallel lateral to the pectoral crest before converging on it distally. It also dif-
fers from small diprotodontids like Neohelos in which the deltoid insertion scar is a weakly
defined vertical facet that merges with the distal apex of the pectoral crest.
Tuberosity for mm. teres major and latissimus dorsi. In Propalorchestes the attachment
scar for the mm. latissimus dorsi and teres major is a sharply defined ovoid on the lateral shaft
just distal to its midpoint. Like other palorchestids, its posterior margin is recurved and over-
hangs the posterior surface of the diaphysis. Two distinct scars of subequal size are discernible
within this ovoid in the Propalorchestes humerus; an anteroproximal, slightly bulging scar, and
a posterodistal rugose portion.
Diaphysis. The shaft of the Propalorchestes humerus appears straight in anterior and medial
views. The posterior surface of the shaft is not strongly scarred, though fossae for the humeral
heads of m. triceps brachii and m. brachialis are discernible, and a deeply excavated olecranon
fossa is present distally. This olecranon fossa is much deeper than in P. parvus and instead
PLOS ONE | https://doi.org/10.1371/journal.pone.0221824 September 13, 2019
51 / 69
Osteology and functional morphology of the limbs in Palorchestidae
resembles the ulnae of Ngapakaldia species in form and positioning. As in P. parvus, the Pro-
palorchestes ulna lacks the strong origin scar for m. epitrochleoanconeus seen in P. azael. In
overall section the shaft is a laterally-expanded oval, distorted anteriorly by the large pectoral
crest and distally by the sheet-like lateral epicondylar crest arising on its distolateral margin.
Unlike in Palorchestes species, a deep radial fossa lies proximal to the capitulum on the antero-
distal shaft surface.
Lateral epicondyle. The lateral epicondyle only barely projects from the lateral edge of the
capitulum in Propalorchestes, making it the least developed of the palorchestids in this respect.
In lateral view it tapers proximally into the supinator crest almost immediately. Its surface is
strongly rugose, with sharply defined scarring present on all sides and a distinct fossa for the
attachment of the lateral collateral ligament. In distal view, the lateral epicondyle is slightly
posteriorly deflected from the transverse axis of the distal epiphysis, unlike the aligned, unde-
flected condition in Palorchestes.
Trochlea. In Propalorchestes the humeral trochlea is a domed shape presenting slightly
more articular surface on the anterior side than posteriorly. It is more convex than in P. parvus
and much more so than the flattened trochlea of P. azael. The trochlea and capitulum, though
subequal in size, are distinct and separate from one another. A ~115˚ angle separates the two
in anterior view, and there is a narrow constriction dividing their dilated surfaces when viewed
inferiorly. This is similar for most vombatiform humeri except for Neohelos in which this con-
striction is much thicker and a much shallower angle (> 150˚) separates the structures in ante-
rior view. Viewed inferiorly the Propalorchestes trochlea is ovoid, with its long axis slightly
anteriorly deflected to lie offset from the transverse axis of the medial epicondyle as in other
palorchestids. This is distinct from vombatid and Diprotodon humeri, where the medial epi-
condyle is posteriorly deflected from the trochlear axis, and from Ngapakaldia, where the
trochlea is aligned with the medial epicondyle. On the posterior aspect the trochlear surface is
barely visible, while a deeply excavated triangular olecranon fossa.
Capitulum. The Propalorchestes capitulum is large and hemispherical in anterior view,
unusual amongst vombatiforms in its marked distal offset from the transverse plane of the dis-
tal articular surface, giving the humerus a slightly lopsided appearance. In lateral view the ante-
roposterior extent of the capitulum is shallower (as in all palorchestids) than the
proportionally deep capitula of other vombatiforms. Viewed distally, the Propalorchestes capit-
ulum tapers slightly to form a small capitular tail, but not to the extent of the pronounced tail
seen in Ngapakaldia where it forms a lateral buttress for the olecranon process when the elbow
is fully extended. Such buttressing is weak in the Propalorchestes humerus, which in posterior
view shows only a small amount of capitular surface compared with other palorchestids and
vombatiforms generally.
Supracondylar foramen. The supracondylar foramen is a well-defined ovoid canal. In ori-
entation it more closely resembles the arrangement in P. parvus than the near-vertical foramen
in P. azael.
Medial epicondyle. The medial epicondyle in Propalorchestes is a rounded, medially-pro-
jecting rugose process. It is greatly expanded to create a broad distal epiphysis. In distal view
the medial epicondyle accounts for 29% of the total distal epiphyseal width, comparing closely
to P. parvus (29%), Ngapakaldia (25%) and extant wombat (27%) humeri, but not quite
approaching the proportions in P. azael (32%). The medial epicondyle lies in the same dorsal
(coronal) plane as the lateral supracondylar crest.
Ulna (Figs 38 and 39). The Propalorchestes ulna strongly resembles that of Vombatus in
proportions and morphology but is approximately 50% larger, similar in overall size to those
of Thylacoleo and Ng. bonythoni, albeit with a much longer olecranon relative to the humeral
articular surface than either of the latter (Figs 38 and 39). The ulna NMV P253947 is missing
PLOS ONE | https://doi.org/10.1371/journal.pone.0221824 September 13, 2019
52 / 69
Osteology and functional morphology of the limbs in Palorchestidae
Fig 38. Right ulna fragment of Propalorchestes sp. NMV P253947. (A) anterior; (B) medial; (C) posterior; (D) lateral
views. Scale bar 50 mm.
https://doi.org/10.1371/journal.pone.0221824.g038
its distal half, so the total length and distal epiphyseal morphology are not known (though we
approximated the length to be 276 mm based on Vombatus proportions).
Olecranon. The olecranon process is elongate, medially deflected and proximally enlarged,
nearly identical to Vombatus in all respects bar absolute size. It is less mediolaterally thick and
proximally bulbous than in Ng. bonythoni. In medial view, the ventral border of the olecranon
tapers slightly towards the bulbous proximal tip, again similar to that of Vombatus and quite
unlike the more regular, quadrangular olecranon in P. azael.
Trochlear surface. The Propalorchestes ulna has a scooped, subcircular trochlear surface. It
resembles Vombatus, Ng. bonythoni and P. parvus trochleae more than the elongate, narrow
and flat surface in P. azael, but is relatively broader than in any of the former taxa.
Coronoid process. The coronoid process appears relatively shorter dorsoventrally than in
Vombatus; however this is due to the relative increase in dorsoventral depth of the Propalorch-
estes shaft immediately distal to it, likely for increased load bearing.
Anconeal process. The anconeal process is tall and in lateral view rises roughly perpendicu-
lar to the longitudinal axis of the ulnar shaft, though not as tall or proximally curled as in wom-
bats. This curled anconeal process in wombats articulates with their pronounced capitular tail
on the humerus, a feature lacking in palorchestid humeri. The anconeal process is less laterally
deflected in the transverse plane than in P. azael or Ng. bonythoni, instead resembling the P.
parvus condition.
Radial notch. The radial notch sits in a ventral position on the lateral shaft, resembling P.
parvus, Ng. bonythoni and Vombatus ulnae in this regard rather than the dorsal positioning of
the notch in P. azael. Its perimeter is eroded in NMV P253947, but the preserved morphology
PLOS ONE | https://doi.org/10.1371/journal.pone.0221824 September 13, 2019
53 / 69
Osteology and functional morphology of the limbs in Palorchestidae
Fig 39. Labelled illustrations of the Propalorchestes sp. right ulna fragment NMV P253947. (A) anterior; (B)
medial; (C) lateral views. Hatching indicates surface damage to cortical bone. Abbreviations: ap, anconeal process; apl,
origin for m. abductor pollicis longus; cf, capitular facet; cp, coronoid process; edp, origin for m. extensor digitorum
profundus; fdp, fossa for origin of m. flexor digitorum profundus; op, olecranon process; rn, radial notch; tn, trochlear
notch; ut, ulnar tuberosity. Scale bar 50 mm.
https://doi.org/10.1371/journal.pone.0221824.g039
indicates a wide concave platform for the head of the radius, unlike the proximo-distally nar-
row, curved notch in extant wombats.
Ulnar tuberosity. In Propalorchestes, the ulnar tuberosity takes the form of a raised sub-
triangular area of rugosity as in wombats and Ng. bonythoni, rather than a distinct tubercle as
seen in Palorchestes species. However, it does resemble the morphology of Palorchestes in its
position on the ventrolateral border of the shaft, well distal of the coronoid process and radial
notch unlike the condition in wombats.
Diaphysis. In lateral view, the ulnar shaft has a posteriorly convex dorsal border. Based on
the uptick in the ventral contour immediately proximal to the break, the bone may have had
the gently sinuating shaft contour of extant wombats rather than the gradual ventral concavity
seen in P. azael. The attachment scar for m. abductor pollicis longus is strongly expressed and
lies both more dorsally on the lateral shaft and arises more distally than in Vombatus (but not
as distal as in P. azael). The attachment scar for m. extensor digitorum profundus is less rugose
than in Vombatus and is more dorsal on the lateral shaft, as in P. azael.
Ungual (Fig 40). The single known Propalorchestes ungual (NMV P179370, Fig 40) is sim-
ilar to the contemporaneous Nimbadon but is absolutely larger than all examples we observed
of that taxon. This size difference is most marked in proximal view, with the Propalorchestes
ungual also possessing a much more dorsally deflected extensor process. Compared to its
more derived palorchestid kin, the ungual is slightly shallower dorsoventrally, with a propor-
tionally thicker ungual process, but the unique palorchestid shape is evident even in this early
species. The flexor tubercle is more discrete from the rest of the proximal ungual than in
PLOS ONE | https://doi.org/10.1371/journal.pone.0221824 September 13, 2019
54 / 69
Osteology and functional morphology of the limbs in Palorchestidae
Fig 40. Ungual phalanx of Propalorchestes sp. NMV P179370. (A) Lateral; (B) medial; (C) proximal; (D) dorsal
views. Scale bar 50 mm.
https://doi.org/10.1371/journal.pone.0221824.g040
Palorchestes species, being in proximal view slightly narrower than the articular facets dorsal to
it, and in lateral view slightly smaller relative to the rest of the ungual.
Palorchestid body mass estimate results
Body mass estimates for palorchestid specimens are presented in Table 3, with values given in
kilograms representing the prediction error around each point estimate. As per the findings of
Campione and Evans [63], the equation derived from combined humeral and femoral circum-
ference data had the most predictive power, with the lowest PPE and SEE of the three models,
followed by humerus-only, with the femoral-only equation giving the lowest predictive power.
As expected, where associated stylopodia were available, estimates incorporating humeral
circumferences were heavier than those using the femur, reflecting the irregular minimum
humeral section profile typical of these animals. The maximal body mass predictions for each
species, calculated using humeral circumference only, were 155 kg for Propalorchestes, 438 kg
for Palorchestes parvus and 2059 kg for P. azael. Femur-based estimates were much smaller,
with 289 kg for P. parvus and 1160 kg for the largest example of P. azael. This large individual
yielded estimates of 1597 kg from the humerus-only and 1413 kg from the combined formulae,
demonstrating how widely the estimates differ between humeral and femoral models.
Table 3. Palorchestid body mass estimates.
Taxon
Palorchestes azael
Palorchestes
parvus
Propalorchestes sp.
Humerus
specimen #
SAMA P55199
NMV P159792
NMV P157144
NHMUK PV OR
46914
AM F58870
NTM P87115-6
CH (mm)
179
191
235
258
145
98
Femur
specimen #
-
NMV P26534
NMV
P157144
-
CF (mm) CH+F (mm)
CH mass estimate
(kg)
CF mass estimate
(kg)
CH+F mass estimate
(kg)
-
161
201
-
-
352
436
-
788 ± 125
926 ± 265
1597 ± 458
2059 ± 629
-
627 ± 212
1160 ± 392
-
791 ± 202
1413 ± 362
-
-
AM F58870
122
267
438 ± 125
289 ± 98
367 ± 94
-
-
-
155 ± 45
-
-
Model I OLS regression formulae
Humeral circumference: logBM = 2.662�logCH− 0.108, R2 = 0.986, PPE = 27.5%, SEE = 0.145
Femoral circumference: logBM = 2.809�logCF− 0.401, R2 = 0.979, PPE = 34%, SEE = 0.175
Combined humeral and femoral circumferences: logBM = 2.752�logCH+F− 1.110, R2 = 0.988, PPE = 25.5%, SEE = 0.133
Estimates are given for all individuals for which stylopodia are known, ± prediction error for each estimate in kilograms. Mass estimates for other comparative
vombatiform taxa calculated using this method are provided in S2 Table. Abbreviations: CH, minimum humeral circumference; CF, minimum femoral circumference;
CH+F, total minimum humeral and femoral circumference; PPE, percent prediction error; SEE; standard error of the estimate.
https://doi.org/10.1371/journal.pone.0221824.t003
PLOS ONE | https://doi.org/10.1371/journal.pone.0221824 September 13, 2019
55 / 69
Osteology and functional morphology of the limbs in Palorchestidae
Mass estimates for comparative vombatiform taxa are available in S2 Table. Prior predic-
tions for Diprotodon and Nimbadon (see Table 1) fell within the bounds of prediction error of
our estimates generated using the combined circumference model. Our estimates for Neohelos
and Thylacoleo were substantially lower than existing predictions (likely due to our particular
specimens), while other results were considerably greater–our Phascolonus estimates ranged
from 460–737 kg, while our estimate for a Zygomaturus specimen surpassed one tonne.
Discussion
Palorchestids were larger than previously suspected
The mass values presented in Table 3 are the first empirical estimates for body mass in
Palorchestidae. In all cases we found masses to be substantially heavier than existing predic-
tions: the early palorchestid Propalorchestes may have weighed around 150 kg, Palorchestes
parvus approximately 300–400 kg, and P. azael possibly more than one tonne.
Campione and Evans [63] found that combined circumference data from both stylopodia
provided the most robust estimates across tetrapods generally. However, they found taxa with
highly apomorphic humeral morphology (talpids) were extreme outliers, even in their com-
bined circumference model, and opted to exclude them. We posit that the highly derived mor-
phology of the P. azael humerus produces similar overestimates which should be interpreted
with caution (this may also apply to the unusual humerus of Phascolonus, see S2 Table). We
suggest that femoral circumference alone is a more conservative metric by which to estimate
body mass for P. azael, despite the lower precision of the femur-based equation overall. To
estimate mass via the femur alone is to assume that P. azael distributed its weight across the
fore- and hindlimbs in the same way as the extant taxa in the dataset from which our equations
were derived (S2 Table). Until more is known about the postcranial body form and habitual
locomotion of this species, such assumptions must be made as we cannot yet predict the place-
ment of the centre of mass. Where a femur was available, P. azael body mass estimates were
lower than for the associated humerus. No associated femur is known for the largest humerus
NHMUK PV OR 46914, which itself yielded an estimate of 2059 kg, an implausible value
approaching the mass of the objectively larger Diprotodon. However, based on humeral
dimensions that individual would certainly have been even bigger than NMV P157144 (S1
Table). So, while our body mass estimates will necessarily have wide error margins, it does
appear that Palorchestes species were heavier members of the Australian Pleistocene fauna
than previously suspected. This has deep implications for future inferences of many aspects of
their palaeobiology including life history strategy, range size and feeding ecology.
Our body mass estimates for P. azael overlapped with the size range we found for Zygoma-
turus specimens (see S2 Table). Gigantism evolved independently in both the diprotodontid
and palorchestid lineages after their divergence in the Palaeogene, as part of a broader trend
toward larger body mass among the Australian Pleistocene megafauna [7, 38]. Though P. azael
is invariably rare, both species co-occur in assemblages across the continent [30, 73, 74], so the
likelihood that P. azael converged on a similar body size to Zygomaturus provides additional
support for their special adaptation to a unique niche separate from this related taxon.
In light of the possibility that P. azael and P. parvus may have been sympatric at the Buchan
site (Table 2), the body size disparity we have shown between the species is interesting. Trusler
and Sharp [37] note that the coexistence of palorchestid species at various periods in their evo-
lutionary past–particularly during the Pleistocene–shows that smaller species were able to per-
sist in the same environmental conditions in which their close relatives achieved gigantism.
While P. azael may have been able to use its extraordinary forelimbs to exploit higher volumes
of poorer-quality browse, smaller species like P. parvus may have continued targeting the food
PLOS ONE | https://doi.org/10.1371/journal.pone.0221824 September 13, 2019
56 / 69
Osteology and functional morphology of the limbs in Palorchestidae
sources they were ancestrally adapted for. This aligns with dental evidence showing that P. azael
appears to have had morphological (and perhaps dietary) affinities divergent from the smaller
P. parvus (and P. pickeringi), with which it may have been sympatric at several localities [46].
With multiple humeri representing several individuals we were able to provide some idea
of the notable body size variation within P. azael. The two smaller P. azael humeri have open
metaphyses which suggests these animals died before reaching an asymptotic size. However,
the two larger specimens both appear to have fused metaphyses while differing in length by
approximately 30%. The size range represented here supports variability in dental dimensions
observed by Trusler [36] in P. azael individuals of similar dental wear and eruption stages.
Both indeterminate growth and sexual size dimorphism are common throughout Order
Diprotodontia [75, 76] and are suspected to have occurred in some diprotodontids [21, 42].
But, like Trusler [36], we are similarly hindered by small sample sizes and are unable to com-
ment on ontogenetic variation or the likelihood of sexual dimorphism in these species. Like-
wise, without more precise dating information it is impossible to know how close these
humeri are temporally–in principle, they could represent adult individuals two million years
apart in evolutionary time.
Specialisation of the palorchestid appendicular skeleton occurred much
later than the craniodental anatomy
The preceding descriptions depict a morphocline within the palorchestid lineage toward the
highly derived forelimb seen in the largest and latest species Palorchestes azael. Prior analyses
of the cranial morphology of the group have recognised that their apomorphic cranial char-
acteristics were already well-established in the smaller, earliest-known species and that the ori-
gin of their specialised rostrum was not associated with increasing body size [28, 36, 37, 68].
Here we demonstrate that specialisation of the postcrania was delayed relative to the skull, and
may indeed have been linked to increasing body size. That is, the forelimb morphology of the
mid-Miocene Propalorchestes is similar to other small early Miocene diprotodontoids such as
Ngapakaldia, and the truly unusual forelimb did not arise until the Late Pleistocene giant
P. azael. From this ancestral morphology, palorchestid forelimbs responded very differently to
increasing body size than their sister diprotodontids, with shortening and widening of the
humerus relative to the ulna, elongation and straightening of the olecranon process, and even-
tual ‘locking’ of the elbow in a flexed position (see discussion below). The inverse appears to
have occurred in giant diprotodontids–they are characterised by lengthened humeri, short
ulnae with posteriorly deflected olecranon processes and more columnar limb posture with
increased trochleation of the elbow (see S1–S3 Figs). Divergence in palorchestid hindlimb
morphology from diprotodontids is subtler, but is especially evident in the femur and the pes.
In many respects even the more derived palorchestids actually resemble vombatids in their
limb anatomy, however with so few fossil vombatid species represented by postcrania it is diffi-
cult to say how close this resemblance truly is.
Forelimb adaptations
Propalorchestes, with its broad distal humerus, large muscle insertion processes and a long,
medially-curved olecranon process, already had the hallmarks of a fairly powerful and special-
ised forelimb user (Figs 36–39). Palorchestes parvus was an animal with even more muscular
forelimbs and slightly reduced ROM in elbow flexion than its predecessor (Figs 19–22). But
undoubtedly, the most exaggerated palorchestid forelimb anatomy is found in Palorchestes
azael, with its flat, seemingly-immobile humeroulnar joint setting it apart from all its vombati-
form kin (Figs 3–6).
PLOS ONE | https://doi.org/10.1371/journal.pone.0221824 September 13, 2019
57 / 69
Osteology and functional morphology of the limbs in Palorchestidae
The elbow was fixed in Palorchestes azael. The shape of the humeroulnar articulation in
P. azael would have effectively fixed the elbow in a flexed posture approaching a 100˚ angle, a
condition seen in no other marsupial or placental mammal known to the authors (Figs 3–6).
This elbow appears to be a palorchestid adaptation for a particularly specialised use of the fore-
limb, likely in acquiring food. This dietary niche was apparently occupied by the earliest
known palorchestids as evidenced by their already derived craniodental and claw morphology,
both of which persisted in the lineage as body size increased. The immobilisation of the elbow
may represent a compromise that arose in later species which attained giant body size. Fixation
may have been necessary to stabilise the joint in a particular posture (see below) and resist
rotational forces at the elbow during feeding activities in a much heavier animal. Such stabili-
sation would be reinforced by the relatively sharp angle between the trochlea and capitulum,
additionally bracing the elbow against lateral stresses.
When considering the uniqueness of the humeroulnar joint in P. azael, it must be noted
that their humeroradial joint displays curiously unspecialised morphology. It does not seem to
vary sympathetically with the humerus and ulna to ‘compensate’ for the reduced mobility of
the elbow by providing increased pronation and supination. The radius is not particularly
bowed, nor does the articular circumference of the radial head appear to allow any greater
rotational movement than in extant wombats or Ngapakaldia, and the pronator muscle attach-
ment scars along its diaphysis are in all cases less rugous and pronounced than their relatives
(Figs 7 and 8D). It appears the humeroradial joint of this animal did not have especially large
rotational ROM, limiting these movements again in favour of stability. This indicates that the
freedom and positioning of the manus was limited and is further evidence for the highly spe-
cialised nature of the P. azael forelimb.
Some other (notably xenarthran) mammalian taxa show a similar flattening of the ulnar
trochlear surface (e.g., myrmecophagids in Taylor, [77] and White,[78]; Paramylodon in Stock
[79]). However, these taxa do not show flattening of the corresponding humeral trochlea, and
have nothing like the apparent loss of flexion and extension seen in P. azael. Such ulnar (but not
humeral) flattening is also expressed in the large bodied extinct vombatid Phascolonus gigas (S1D
and S2D Figs). Murray [6] suggested this morphology may help stabilise and evenly distribute
forces through the Phascolonus elbow under high loads experienced during digging or tearing
with the manus. These patterns of forelimb engagement are also seen in myrmecophagids and
(presumably) Paramylodon. That P. azael exhibits flattening not only of the ulna but also of the
humerus may suggest it was especially committed to such forelimb uses. However, we propose
that fossorial behaviour is unlikely in palorchestids (see discussion of the unguals below).
Postural differences. Could the flat humeroulnar joint have arisen to cope with a more
medially-loaded elbow in the heaviest palorchestid? All palorchestid humeri appear to transmit
weight medially over the ulna rather than centrally or laterally as in other vombatiforms. This
is evidenced by the fact that in all palorchestid humeri, the vertical midshaft axis intersects the
distal epiphysis through the trochlea, rather than through the approximate midline (as in Phas-
colonus, Neohelos and Diprotodon) or laterally through the capitulum (as in modern wombats
and Zygomaturus) (S1 Fig). This loading regime appears to have persisted across the lineage,
until in the largest species P. azael the fixation of the elbow may be a consequence of this
imbalanced loading, exacerbated at heavier body masses. That P. azael loaded its elbow on the
medial rather than lateral side could indicate a unique forelimb posture, perhaps somewhat
sprawled with the elbows abducted from the sagittal plane. Fujiwara and Hutchinson [80]
found sprawling tetrapods could be distinguished from other locomotory categories by their
longer medial epicondyles, creating a greater elbow adductor moment arm in order to resist
abducting ground reaction forces during sprawling gait. Such a posture is atypical for mam-
mals, but the unusually elongate medial epicondyle in P. azael, along with their apparently
PLOS ONE | https://doi.org/10.1371/journal.pone.0221824 September 13, 2019
58 / 69
Osteology and functional morphology of the limbs in Palorchestidae
medial loading regime and flattened humeroulnar articulation, may support postural recon-
struction of P. azael with a sprawled forelimb.
Altered muscle actions on a fixed elbow.
Immobilisation of the elbow significantly alters
the primary actions of muscles crossing this joint. The key flexor m. brachialis becomes an
important stabiliser, gaining leverage via distal positioning of its insertion on the ulnar tuber-
osity (Figs 5 and 6), and creating a deep spiral fossa and marked scars on the upper lateral and
posterior humerus (Figs 3B, 3C, 4B and 4C). M. epitrochleoanconeus, a minor extensor in
other marsupials, instead becomes hugely enlarged in its stabilising role, creating a diagnostic
muscle scar on the posterior surface of the medial epicondyle and humeral shaft seen only in
P. azael (Figs 3C and 4C).
In an immobile elbow, a biarticular muscle like m. biceps brachii becomes more significant
in shoulder flexion and supination of the manus. Similarly, m. triceps brachii, rather than an active
elbow extensor as suggested by the large olecranon [81], stabilises and maintains posture against
the force of gravity when bearing weight on the forelimb. It would also perform powerful shoulder
retraction as may be needed in raking and tearing with the manus when the elbow cannot be
extended, evidenced by the huge infraglenoid origin for its scapular head (Figs 1 and 2).
Compensating for a fixed elbow. Compensation for lack of mobility in the elbow may
instead have been available via humeral rotation, which would aid in positioning of the manus
during locomotion and also manipulation of the environment when feeding. Without a more
intact scapula such inferences are tentatively made, but many distinctive features on the P.
azael humerus suggest powerful rotational movements of the shoulder. Firstly, the deltoid
insertion is entirely separate and laterally displaced, increasing mechanical advantage for lat-
eral rotation as noted in Tamandua by McAfee [82] (Fig 4A and 4B). Similarly, the elevation
of the greater tubercle superior to the humeral head, while limiting ROM in abduction, would
increase leverage in lateral rotation as well as stabilising the actions of the rotator cuff muscles
(Fig 4A–4C). This morphology resembles that of extant wombats which perform powerful
strokes with the shoulder when digging (but see discussion of the unguals below) (S1B and
S1C Fig). Additionally, the insertion scar for mm. latissimus dorsi and teres major is very large
and very distally positioned on the medial humerus, creating significant leverage for medial
rotation (and/or resisting lateral rotation), as well as powerful retraction (Figs 4, 20 and 37).
That this feature is common to all palorchestids speaks to the importance of these actions in
their functional ecology. Finally, the pronounced medial cant peculiar to the P. azael pectoral
crest may also reflect particularly strong medial rotation. The crest, freed of opposing deltoid
muscle forces, is curled over the humerus by the pull of pectoral muscles medially rotating and
adducting the shoulder (Fig 4A and 4D). Perhaps strong medial rotation and adduction of the
shoulders allowed P. azael to exert a strong bilateral grip on the bole of a tree, grasping with
the hands and retracting the shoulder to pull the upper body toward the tree to feed.
Plesiomorphic wrist and manus anatomy. The lack of known palorchestid carpals is frus-
trating; however, based on the available associated manual material from P. parvus, the group
appear to have retained a fairly primitive plantigrade manus (Figs 25 and 26), resembling the
Ngapakaldia condition. Distal forelimb and pedal similarities support this morphological affin-
ity with Ngapakaldia, although the shortened, robust digit rays and thick metacarpals in P. par-
vus reflect increased weightbearing in the manus of these larger palorchestids.
The nature of this manual weightbearing is different to other large vombatiforms. Despite
their apparently heavy bodies, large palorchestids were not graviportally adapted. The distal
radius and ulna in P. azael and radius in P. parvus lack the dilated morphology apparent in
Zygomaturus and Diprotodon, and contrast even with the wrists of the robust but smaller Phas-
colonus (S2 Fig). Palorchestes species retain a Ngapakaldia-like, trapezoidal distal radius and
gracile, hooked styloid process. It seems that selective pressure for a relatively unmodified,
PLOS ONE | https://doi.org/10.1371/journal.pone.0221824 September 13, 2019
59 / 69
Osteology and functional morphology of the limbs in Palorchestidae
dextrous manus in Palorchestes precluded the weight-related changes seen in graviportal taxa
like Zygomaturus and Diprotodon.
Enlarged claws. The other remarkable manual morphology characteristic of palorchestids
are their deep, laterally-compressed, knife-like, penetrating ungual phalanges (Figs 9, 10, 25,
26, 34, 35 and 40). These would have been encased within keratinous claw sheaths that
extended the length and curvature of the bone contour by a great deal–around 30% based on
koalas but as much as ~80% based on large myrmecophagids and ursids (pers. obs.). Being
dorsoventrally deep they appear adapted for high stresses such as those experienced during
climbing or digging [83]. However, an arboreal habit appears extremely unlikely, both due to
their now-apparent large body size and lack of important substantiating morphological evi-
dence such as palmar tuberosities on the proximal phalanges indicative of the strong habitual
grasping characteristic of arboreal vombatiforms like Nimbadon and Phascolarctos [20]. Like-
wise, the lateral compression and sharp distal ends of the claws make them poorly shaped for
digging. Indeed, Szalay [71] notes that palorchestid claws are ‘almost at the opposite end of ter-
minal phalanx construction’ to fossorial mammals. This indicates that palorchestids interacted
very differently with the substrate than their contemporary diprotodontid sister taxa, in which
the ancestral laterally-compressed claw shape became reduced in favour of weightbearing in
larger species [11]. Despite attaining Pleistocene body masses comparable to Zygomaturus,
palorchestids retained this ancestral autopod with claws apparently adapted not for arboreal or
fossorial uses, but for slicing, clinging and raking.
Powerful movements of the wrist and digits. P. parvus has short, sturdy digits which
when extended could be partially abducted at the domed metacarpophalangeal joints to splay
the fingers and increase the spread of the manus (Figs 25 and 26). The metacarpal heads are
strongly trochleated on the palmar side, more so even than in extant ursids, which would stabi-
lise the digits when flexed and resist lateral deviations when under high loads experienced dur-
ing raking actions (Fig 25A–25H). The proximal interphalangeal joints are restricted to
dorsoventral movements by their keeled articulations, although dorsal bony stops at each joint
prevent hyperextension. The distal interphalangeal joints, while clearly mobile, certainly do
not permit the degree of ungual hyperextension seen in felids or viverrids. This eliminates the
possibility of ‘retractable claws’ suggested by Flannery and Archer [10], while their other prop-
ositions of knuckle-walking or manolateral hand postures are not supported by any morphol-
ogy of the palorchestid wrist and hand currently known.
Use of the manus in powerful clinging and tearing motions is further supported by the
hugely broad distal humerus characteristic of palorchestids (Figs 3, 4, 19, 20, 36 and 37), pro-
viding both large muscle attachment area and increased leverage for powerful flexors and
extensors of the wrist and digits. Flexors from the medial epicondyle would provide the power
to grip and embed the claws into the substrate, while strong extensors from the lateral epicon-
dyle would assist in tearing and pulling that substrate apart. Such enhanced forearm strength
and large claws are again strongly suggestive of adaptation of the limb to use during feeding.
Coombs [11] noted that few large mammals today use their clawed forelimbs to browse, excep-
tions being some bear species, and highlighted that appropriate extant analogues for such
behaviour are few. Future comparative analyses may help to resolve the appendicular biome-
chanics of feeding in the palorchestid forelimb and will be important in understanding the
overall dietary adaptations of the group.
Hindlimb adaptations
Narrow, adducted hips. Large diprotodontids are conventionally reconstructed with a
wide-set stance, columnar hindlimb posture and slightly bowed knees, owing to the obtuse
PLOS ONE | https://doi.org/10.1371/journal.pone.0221824 September 13, 2019
60 / 69
Osteology and functional morphology of the limbs in Palorchestidae
angle between their long femoral neck and adjoining shaft, the unequal form of the femoral
condyles (S3 Fig), and the morphology of the tibia and talocrural joint [84]. Fossil trackways
have supported such reconstructions, and have even been suggested to depict sexual dimor-
phism in gait width due to large and cumbersome pouch young carried by the female [85].
Palorchestid morphology described here indicates a more gracile hindlimb in a very different
posture, with hips adducted and more extended knees.
The Palorchestes femur differs strongly from diprotodontids of similar size and suggests
habitual postural differences. With a more circular diaphyseal section profile and very short
neck aligning the femoral head approximately vertically level with the medial epicondyle, this
femoral morphology resembles Phascolonus (S3 Fig). In fact, the specimen figured (Figs 13
and 14) was long referred to that genus. However, in proportions the Palorchestes femur is
much more elongate than Phascolonus and differs markedly in its distal morphology (S3D
Fig). In Palorchestes the femoral condyles lie approximately level in anterior view (Figs 13A
and 31A) and the medial patellar surface does not protrude anteriorly (Figs 13F and 30I), con-
trasting with Phascolonus and other large-bodied vombatiforms Zygomaturus and Diprotodon.
In these taxa the lateral condyle is also distally offset, reflecting a more abducted hip posture.
This posture requires marked anterior projection of the medial patellar surface to resist medial
tracking of the patelloid due to the more acute angle between the iliac origin of the quadriceps
its insertion on the tibial tuberosity. Palorchestes femora lack this morphology and instead
these animals appear to have held their femur and tibia more vertically in the dorsal plane.
The os coxae are slender, and when rearticulated using the symphyseal epiphysis produce a
relatively narrower width across the ilia than in diprotodontid or vombatid pelves (Figs 11 and
12). This indicates a leaner abdominal girth in Palorchestes compared to other browsing vom-
batiforms, which may relate to dietary differences. Coprolites or fossilised gut contents could
potentially be used to test this idea, however such specimens from any marsupial megafauna,
let alone Palorchestes, are extremely rare [86, 87]. So, the question of whether this pelvic mor-
phology is actually related to dietary ecology awaits further evidence.
Extended hindlimb posture. The limited height of the posterior femoral condylar surface
belies the morphology seen in species that bear weight on a habitually flexed knee [82, 88] and
may instead indicate a more extended posture (Figs 13C, 14C, 30C and 31C). Further to this
are the tall, quadrangular intercondylar eminence on the proximal tibia (locking into the inter-
condylar fossa on the distal femur and stabilising an extended knee, Figs 15C, 16C, 32C and
33C), and the relatively proximal origin for the lateral head of m. gastrocnemius on the femur.
In P. azael this origin is more proximal than any other vombatiform taxon studied, suggesting
a significant role of this muscle in stabilising the extended knee. Its lateral head would act to
increase lateral knee joint congruence, contracting to decrease the distance between lateral
condyle and articulating tibial surface, and preventing leg adduction/genu varus. Additionally,
the reduced and very proximal position of the tibial crest as an insertion for the m. biceps femo-
ris on the lateral tibia indicates a shorter moment arm and lower leverage of this muscle across
the knee, further evidence of habitually extended knee posture. These features point toward an
adducted, extended hindlimb posture in Palorchestes, contrasting with the wider-set stance of
diprotodontids and crouched posture in vombatids. Such postural differences would suggest
palorchestids required less hindlimb muscle bulk than equivalently-sized diprotodontids, as
more weight could be borne directly via the skeleton with the hindlimbs adducted under the
body, without the extensive muscular support needed to sustain a wider-set diprotodontid-like
posture.
A further peculiarity in the Palorchestes hindlimb is the vertical, laterally-facing and flat fib-
ular facet on the proximal tibia in P. parvus (Fig 33B and 33C), suggesting reduced loads on
the fibula than in diprotodontids or vombatids, and possibly increased tibiofibular mobility.
PLOS ONE | https://doi.org/10.1371/journal.pone.0221824 September 13, 2019
61 / 69
Osteology and functional morphology of the limbs in Palorchestidae
With damage to this area in the P. azael specimen, and lacking any fibula or complete astraga-
lus specimens, it is difficult to draw conclusions; however, this preserved morphology in two
P. parvus specimens may hint at retention of a more primitive, mobile tibiofibular joint once
important in an ancestrally arboreal lifestyle.
Plantigrade, syndactylous pes. Well-developed plantar tuberosities on the navicular (Figs
17F and 34K) and ectocuneiform (Figs 17D and 34L) in Palorchestes parvus indicate a planti-
grade foot posture. Additionally, the syndactylous second and third digits are not hugely dif-
ferent from the plesiomorphic possum-like condition from which all diprotodontian feet
evolved [71] (Figs 34A and 35). The combination of large body mass with the constraint of
a syndactylous pes necessitated inflation of the fourth and fifth digits to bear the laterally-
distributed weight. This is a less distorted pes than in the strange rotated toes in the giant
Diprotodon [84], nor is it a true ‘pedolateral’ pes as seen in some extinct ground sloths [89], as
the weight is not borne on the lateral surface of the metatarsals. Instead these thick, short,
clawed lateral digits resemble a hyper-robust Ngapakaldia [24] and bore weight on their true
plantar surface.
With additional associated tarsals for P. azael, including the cuboid and calcaneus, it is
clear they too had a plantigrade pes (Figs 17 and 18). Comparing the metatarsal facets on the
ectocuneiform between P. parvus and P. azael (Figs 17D and 34L), it appears digits 2 and 3
may have been less reduced in the latter, instead being slightly more equal in size to their
neighbouring digits as in Zygomaturus and Phascolonus. This shift toward higher digital uni-
formity in Palorchestes could be related to large body size, where P. azael is above a threshold
beyond which the ancestral disparity in digit proportions is maladaptive and therefore lost.
It is interesting to note that throughout these descriptions of the hindlimb, palorchestid
morphology frequently compares most closely to Ngapakaldia, being mostly much larger but
often similar in form. Ngapakaldia was originally described as a primitive palorchestid and
this taxonomic status remained for many years [24, 90, 91]. Eventually the genus was shifted to
Diprotodontinae after comprehensive craniodental phylogenetic analysis by Black [7]. These
historical ideas about Ngapakaldia as a primitive palorchestid, or similar to a hypothetical
common ancestor with mosaic features from zygomaturines, palorchestids and diprotodon-
tines [92], are borne out by similarities in appendicular morphology even to the most derived
Palorchestes species.
Facultative bipedality?.
In order to employ the forelimb in the manner suggested by their
anatomy, palorchestids would need to rear into a bipedal stance to free the manus from
weightbearing while in use. However, few aspects of the hindlimb in P. azael point towards
specific adaptation to bipedal posture, and several features appear to hinder it. For example,
the hindlimb overall is less robust than similarly-sized diprotodontids. The acetabulum
appears surprisingly shallow, reducing stability of the hip. In the femur, a short femoral neck
means that the greater trochanter projects above the head, limiting abduction and rotation and
restricting the hip to movements largely in the parasagittal plane. This contrasts with the proxi-
mal femoral shape of more adept facultative bipeds such as ursids [93]. Indeed, the bear-like
elevated femoral head in the Pleistocene zygomaturine Hulitherium was cited as evidence for a
facultatively bipedal stance in that species [94].
However, given their apparent forelimb strength and penetrating claws it seems likely that
palorchestids could have pulled their body upright into a bipedal position while supported
against a tree by the forelimbs, with the manual claws hooked into the bark. This would allow
the elongate face and protrusible tongue to access additional, higher browse than could be
reached in a quadrupedal position. Similarly, palorchestids may have adopted such a bipedal
posture to topple ferns or cycads by throwing their bodyweight against them with the fore-
limbs, making young fiddlehead shoots or fleshy seeds up at the crown accessible at ground
PLOS ONE | https://doi.org/10.1371/journal.pone.0221824 September 13, 2019
62 / 69
Osteology and functional morphology of the limbs in Palorchestidae
level. This possibility is further supported by the long, dorsally rotated ischium in P. azael, a
character interpreted in other mammals as adaptive for upright trunk posture through main-
taining the moment arm of the hamstrings when the femur is extended [95, 96]. Such mor-
phology would allow stronger hip extension when standing erect in this scenario, perhaps to
brace the hindlimb while pushing with the forelimb.
It is also possible that the relative gracility of the palorchestid hindlimb is evidence not of
them being facultatively bipedal, but tripodal–that is, providing additional support for the rear-
ing body with a muscular tail, behaviour well-known in macropodids [88, 97] and recently rea-
soned in the extinct vombatiform Thylacoleo [98]. Palorchestes has previously been noted to
possess a ‘kangaroo-like tail’ [47, 99] larger than that of their diprotodontid kin, potentially as
an adaptation to tripodal posture [6], though this has never been fully investigated. Axial ele-
ments including the caudal vertebrae were not considered in the current appendicular descrip-
tions but may in future provide valuable insight into any palorchestid tripodal adaptation and
potential convergence on such a ‘ground sloth’-like habit.
Conclusions
This work represents the first quantitative body mass estimates and descriptions of appendicu-
lar morphology in the Palorchestidae, collating over 60 specimens across three taxa to finally
provide a comprehensive view of their unique anatomy. The postcranial evidence presented
here reinforces existing knowledge of the extraordinary palorchestid craniodental morphology
to cement their status as one of the strangest marsupial lineages ever to have existed. Our find-
ings certainly support Flannery and Archer [10] in their notion that palorchestids were
adapted for a niche no longer occupied in modern Australian landscapes.
We are beginning to build a picture of the palorchestids as they were in life: Giant planti-
grade quadrupeds with a slender body form, muscular bent forelimbs, straighter hindlimbs
and enlarged claws. These features evoke a specialised feeder actively using its forelimbs to
acquire browse, and with restricted elbow mobility in larger species such actions must have
been driven by a powerful shoulder, perhaps in a somewhat abducted posture. They may have
adopted a bipedal stance to feed, but further details of their locomotor adaptations such as car-
riage of the manus remain unresolved. However, it is clear that palorchestids moved and
behaved in a way vastly different to their contemporaneous diprotodontid kin despite attain-
ing comparable body sizes.
The paucity of known palorchestid postcranial fossils makes more exhaustive functional
and evolutionary interpretation of the appendicular skeleton challenging, especially the lack of
scapula, fibula and carpals. It is hoped that further unidentified material held in collections
may become referable to the family, genus or species level as a result of the present work.
Future studies on the currently known material will attempt to quantify the morphological dif-
ferences described here in a broader mammalian context, and test hypotheses of limb function
using geometric morphometric and musculoskeletal modelling methods.
Supporting information
S1 Table. Measurements and museum collection registration details for appendicular
material of palorchestids and comparative taxa included in this study.
(XLSX)
S2 Table. Body size calculations for comparative vombatiform taxa included in this study,
with dataset used to generate predictive equations.
(XLSX)
PLOS ONE | https://doi.org/10.1371/journal.pone.0221824 September 13, 2019
63 / 69
Osteology and functional morphology of the limbs in Palorchestidae
S1 Fig. Illustrations of humeri from comparative vombatiform taxa used in this study. All
elements scaled to the same length to emphasise overall shape and proportion differences. (A)
Phascolarctos; (B) Vombatus; (C) Lasiorhinus; (D) Phascolonus; (E) Nimbadon; (F) Neohelos;
(G) Zygomaturus; (H) Thylacoleo; (I) Ngapakaldia; (J) Diprotodon; (K) Propalorchestes; (L)
Palorchestes parvus; (M) Palorchestes azael. Scale bar 50 mm.
(TIF)
S2 Fig. Illustrations of ulnae from comparative vombatiform taxa used in this study. All
elements scaled to the same length to emphasise overall shape and proportion differences.
Each pair shows anterior view on the left, medial view on the right. (A) Phascolarctos; (B) Vom-
batus; (C) Lasiorhinus; (D) Phascolonus; (E) Nimbadon; (F) Neohelos; (G) Zygomaturus; (H)
Thylacoleo; (I) Ngapakaldia; (J) Diprotodon; (K) Propalorchestes; (L) Palorchestes azael. Scale
bar 50 mm.
(TIF)
S3 Fig. Illustrations of femora from comparative vombatiform taxa used in this study. All
elements scaled to the same length to emphasise overall shape and proportion differences.
Each pair shows anterior view on the left, posterior view on the right. (A) Phascolarctos; (B)
Vombatus; (C) Lasiorhinus; (D) Phascolonus; (E) Nimbadon; (F) Neohelos; (G) Zygomaturus;
(H) Thylacoleo; (I) Ngapakaldia; (J) Diprotodon; (K) Palorchestes parvus; (L) Palorchestes azael.
Scale bar 50 mm.
(TIF)
S4 Fig. Cladogram of Order Diprotodontia showing the relative position of Vombati-
formes.
(PDF)
Acknowledgments
The authors gratefully acknowledge the significant foundational contribution to this project
made by Peter Murray.
Our gratitude goes to museum curatorial and collection staff Mary-Anne Binnie (SAMA),
Matt McCurry (AM), Tammy Gordon and David Maynard (QVMAG), Nikki King-Smith
and Kathryn Medlock (TMAG), Adam Yates (MAGNT), Katie Date, Karen Roberts and
Ricky-Lee Erickson (MV) and especially Tim Ziegler (MV) for collections access, specimen
loans and logistical support. Thanks also to Pip Brewer, Amy Scott-Murray and Emily Keeble
(NHMUK) for providing specimen photographs and 3D data for NHMUK PV OR 46914, and
to Michelle Quayle for 3D printing assistance.
HLR would like to thank Lisa Nink, Aaron Camens, Gavin Prideaux, Carey Burke, Jacob
Van Zoelen and the Flinders University palaeontology group for generous support and discus-
sions, as well as all members of the EvoMorph and Integrated Morphology and Palaeontology
labs at Monash University. Warm thanks to Rolan Eberhard for his hospitality and assistance,
and to Nicolas Campione for advice in applying his body mass estimation method. Apprecia-
tion also goes to Peter Trusler for his invaluable mentorship and David Hocking for his helpful
thoughts and photography instruction.
Author Contributions
Conceptualization: Hazel L. Richards, Rod T. Wells, Erich M. G. Fitzgerald.
Data curation: Hazel L. Richards.
PLOS ONE | https://doi.org/10.1371/journal.pone.0221824 September 13, 2019
64 / 69
Osteology and functional morphology of the limbs in Palorchestidae
Formal analysis: Hazel L. Richards, Justin W. Adams.
Investigation: Hazel L. Richards, Rod T. Wells.
Methodology: Hazel L. Richards, Rod T. Wells, Alistair R. Evans, Justin W. Adams.
Project administration: Hazel L. Richards.
Resources: Rod T. Wells.
Supervision: Alistair R. Evans, Erich M. G. Fitzgerald, Justin W. Adams.
Visualization: Hazel L. Richards.
Writing – original draft: Hazel L. Richards.
Writing – review & editing: Hazel L. Richards, Rod T. Wells, Alistair R. Evans, Erich M. G.
Fitzgerald, Justin W. Adams.
References
1. Wroe S, Field JH, Archer M, Grayson DK, Price GJ, Louys J, et al. Climate change frames debate over
the extinction of megafauna in Sahul (Pleistocene Australia-New Guinea). Proceedings of the National
Academy of Sciences. 2013; 110(22):8777–81.
2. Smith FA, Smith REE, Lyons SK, Payne JL. Body size downgrading of mammals over the late Quater-
nary. Science. 2018; 360(6386):310–3. https://doi.org/10.1126/science.aao5987 PMID: 29674591
3. Woods JT. The extinct marsupial genus Palorchestes Owen. Memoirs of the Queensland Museum.
1958; 13(4):177–93.
4. Archer M, Bartholomai A. Tertiary mammals of Australia: A synoptic review. Alcheringa. 1978; 2(1):1–
19.
5. Black KH, Archer M, Hand SJ, Godthelp H. The rise of Australian marsupials: A synopsis of biostrati-
graphic, phylogenetic, palaeoecologic and palaeobiogeographic understanding. In: Talent JA, editor.
Earth and Life. Netherlands: Springer; 2012. p. 983–1078.
6. Murray P. Palaeontology and palaeobiology of wombats. In: Wells RT, Pridmore PA, editors. Wombats.
Adelaide, South Australia: Surrey, Beatty and Sons; 1998. p. 1–33.
7. Black K. Diversity, phylogeny and biostratigraphy of diprotodontoids (Marsupialia: Diprotodontidae,
Palorchestidae) from the Riversleigh World Heritage Area. PhD thesis, University of New South Wales.
2008.
8. Gillespie AK, Archer M, Hand SJ. A tiny new marsupial lion (Marsupialia, Thylacoleonidae) from the
early Miocene of Australia. Palaeontologia Electronica. 2016; 19(2):1–25.
9. Wells R. Fossil mammals in the reconstruction of Quaternary environments with examples from the
Australian fauna. In: Guppy J, editor. Biology and Quaternary Environments. Canberra, ACT: Austra-
lian Academy of Science; 1978. p. 103–24.
10.
Flannery TF, Archer M. Palorchestes: Large and small palorchestids. In: Rich PV, Van Tets GF, Knight
F, editors. Kadimakara: Extinct Vertebrates of Australia. Melbourne, Victoria: Pioneer Design Studio;
1985. p. 234–9.
11. Coombs MC. Large mammalian clawed herbivores: A comparative study. Transactions of the American
Philosophical Society. 1983; 73(7):1–96.
12.
Lee AK, Carrick FN. Phascolarctidae. In: Walton DW, Richardson BJ, editors. Fauna of Australia. Can-
berra: AGPS; 1989.
13. Strahan R, editor. The Mammals of Australia. Chatswood, NSW: Reed Books; 1995.
14.
Finch M, Freedman L. Functional morphology of the limbs of Thylacoleo carnifex Owen (Thylacoleoni-
dae, Marsupialia). Australian Journal of Zoology. 1988; 36(3):251–72.
15. Wroe S, Myers T, Seebacher F, Kear B, Gillespie A, Crowther M, et al. An alternative method for pre-
dicting body mass: The case of the Pleistocene marsupial lion. Paleobiology. 2003; 29(3):403–11.
16. Anderson JF, Hall-Martin A, Russell DA. Long-bone circumference and weight in mammals, birds and
dinosaurs. Journal of Zoology. 1985; 207(1):53–61.
17. Stirling E. On the identity of Phascolomys (Phascolonus) gigas, Owen, and Sceparnodon ramsayi,
Owen: A description of some of its remains. Memoirs of the Royal Society South Australia. 1913;
1:127–78.
PLOS ONE | https://doi.org/10.1371/journal.pone.0221824 September 13, 2019
65 / 69
Osteology and functional morphology of the limbs in Palorchestidae
18. Scott GG, Richardson KC. Appendicular osteological differences between Lasiorhinus latifrons Owen
1845 and Vombatus ursinus Shaw 1800 (Marsupialia: Vombatidae). Records of the South Australian
Museum. 1988; 22:95–102.
19. Owen R. On the osteology of the Marsupialia (Part IV): Bones of the trunk and limbs, Phascolomys. The
Transactions of the Zoological Society of London. 1874; 8(8):483–500.
20. Black KH, Camens AB, Archer M, Hand SJ. Herds overhead: Nimbadon lavarackorum (Diprotodonti-
dae), heavyweight marsupial herbivores in the Miocene forests of Australia. PLOS ONE. 2012; 7(11):
e48213. https://doi.org/10.1371/journal.pone.0048213 PMID: 23185250
21. Murray P, Megirian D, Rich T, Plane M, Vickers-Rich P. Neohelos stirtoni, a new species of Zygomaturi-
nae (Diprotodonta: Marsupialia) from the mid-Tertiary of the Northern Territory, Australia. Records of
the Queen Victoria Museum and Art Gallery. 2000; 105:1–47.
22. Murray P, Megirian D, Rich T, Plane M, Black K, Archer M, et al. Morphology, systematics, and evolu-
tion of the marsupial genus Neohelos Stirton (Diprotodontidae, Zygomaturinae). Museums and Art Gal-
leries of the Northern Territory Research Report. 2000; 6(6):1–141.
23. Scott HH. A Monograph of Nototherium Tasmanicum: Genus Owen: sp. nov. Hobart, Tasmania: J.
Vail, Government Printer; 1915.
24. Munson CJ. Postcranial Descriptions of Ilaria And Ngapakaldia (Vombatiformes, Marsupialia) and the
Phylogeny of the Vombatiformes Based on Postcranial Morphology. Berkeley, California: University of
California Press; 1992. 1–99 p.
25.
Long JA, Archer M, Flannery T, Hand S. Prehistoric Mammals of Australia and New Guinea: One Hun-
dred Million Years of Evolution: Johns Hopkins University Press; 2002.
26. Owen R. On the Fossil Mammals of Australia. Part III. Diprotodon australis, Owen. Proceedings of the
Royal Society of London Series I. 1869; 18:196.
27. Wroe S, Crowther M, Dortch J, Chong J. The size of the largest marsupial and why it matters. Proceed-
ings of the Royal Society of London Series B: Biological Sciences. 2004; 271(Suppl 3):S34–6.
28. Murray P. Primitive marsupial tapirs (Propalorchestes novaculacephalus Murray and P. ponticulus sp.
nov.) from the mid-Miocene of north Australia (Marsupialia: Palorchestidae). The Beagle: Records of
the Museums and Art Galleries of the Northern Territory. 1990; 7(2):39–51.
29.
Johnson C. Australia’s Mammal Extinctions: A 50,000-Year History. Port Melbourne: Cambridge Uni-
versity Press; 2006.
30. Pledge NS. Occurrences of Palorchestes species (Marsupialia: Palorchestidae) in South Australia.
Records of the South Australian Museum. 1991; 25(2):161–74.
31. Arena DA, Travouillon KJ, Beck RM, Black KH, Gillespie AK, Myers TJ, et al. Mammalian lineages and
the biostratigraphy and biochronology of Cenozoic faunas from the Riversleigh World Heritage Area,
Australia. Lethaia. 2016; 49(1):43–60.
32. DeSantis LRG, Field JH, Wroe S, Dodson JR. Dietary responses of Sahul (Pleistocene Australia–New
Guinea) megafauna to climate and environmental change. Paleobiology. 2017; 43(2):1–15.
33. Mackness BS. Reconstructing Palorchestes (Marsupialia: Palorchestidae)—from giant kangaroo to
marsupial ’tapir’. Proceedings of the Linnean Society of New South Wales. 2008; 130:21–36.
34. Owen R. On the Fossil Mammals of Australia. Family Macropodidae. Genera Macropus, Pachysiagon,
Leptosiagon, Procoptodon, and Palorchestes—Part IX. Proceedings of the Royal Society of London.
1873:386–7.
35.
36.
37.
Fletcher H. Palorchestes—Australia’s extinct giant kangaroo. Australian Museum Magazine. 1945; 8
(11):361–5.
Trusler P. Cranial reconstruction of Palorchestes azael. PhD thesis, Monash University. 2016.
Trusler PW, Sharp AC. Description of new cranial material of Propalorchestes (Marsupialia: Palorches-
tidae) from the middle Miocene camfield beds, Northern Territory, Australia. Memoirs of Museum Victo-
ria. 2016; 74(July):291–324.
38. Murray P. The Pleistocene megafauna of Australia. In: Vickers-Rich P, Monaghan JM, Baird RF, Rich
TH, editors. Vertebrate Palaeontology of Australasia. Melbourne, Victoria: Pioneer Design Studio;
1991. p. 1071–164.
39. Davis A, Archer M. Palorchestes azael (Mammalia, Palorchestidae) from the late Pleistocene Terrace
Site Local Fauna, Riversleigh, northwestern Queensland. Memoirs of the Queensland Museum. 1997;
41:315–20.
40. Black K, Mackness B. Diversity and relationships of diprotodontoid marsupials. Australian Mammalogy.
1999; 21:20–1.
41. Black KH, Hand SJ. First crania and assessment of species boundaries in Nimbadon (Marsupialia:
Diprotodontidae) from the middle Miocene of Australia. American Museum Novitates. 2010; 3678:1–60.
PLOS ONE | https://doi.org/10.1371/journal.pone.0221824 September 13, 2019
66 / 69
Osteology and functional morphology of the limbs in Palorchestidae
42. Price GJ. Taxonomy and palaeobiology of the largest-ever marsupial, Diprotodon Owen, 1838 (Diproto-
dontidae, Marsupialia). Zoological Journal of the Linnean Society. 2008; 153(2):369–97.
43. Price GJ, Hocknull S. A small adult Palorchestes (Marsupialia, Palorchestidae) from the Pleistocene of
the Darling Downs, southeast Queensland. Memoirs of the Queensland Museum. 2005; 51(1):202–.
44. Mackness B. Palorchestes selestiae, a new species of palorchestid marsupial from the early Pliocene
Bluff Downs Local Fauna, northeastern Queensland. Memoirs of the Queensland Museum. 1995;
38:603–10.
45. Black K. A new species of Palorchestidae (Marsupialia) from the late middle to early late Miocene
Encore Local Fauna, Riversleigh, northwestern Queensland. Memoirs of the Queensland Museum.
1997; 41:181–6.
46. Piper KJ. A new species of Palorchestidae (Marsupialia) from the Pliocene and early Pleistocene of Vic-
toria. Alcheringa: An Australasian Journal of Palaeontology. 2006; 30(sup1):281–94.
47. Bartholomai A. A new species of Thylacoleo and notes on some caudal vertebrae of Palorchestes
azael. Memoirs of the Queensland Museum. 1962; 14:33–40.
48. Young GC, Laurie JR, editors. An Australian Phanerozoic Timescale. Melbourne: Oxford University
Press; 1996.
49. Shean DA. Geochronology, taxonomy and morphology of select fossils of the Buchan Caves, south-
eastern Australia. MSc thesis, Monash University. 2007.
50. Marshall LG. Late Pleistocene mammals from the Keilor cranium site, southern Victoria, Australia.
Memoirs of the National Museum of Victoria. 1974; 35:63–86.
51. Duncan J. Megafauna at Keilor and the timing of their extinction. Australian Archaeology. 2001; 53
(1):16–22.
52.
Joyce E, Anderson JR. Late Quaternary geology and environment at the Dry Creek archaeological sites
near Keilor in Victoria, Australia. The Artefact. 1976; 1(2):47–74.
53. Dunkley J, Spate A, Welch B. Wee Jasper Caves. Broadway, NSW: Australian Speleological Federa-
tion; 2010.
54. Schmidt VA, Jennings JN, Haosheng B. Dating of cave sediments at Wee Jasper, New South Wales,
by magnetostratigraphy. Australian Journal of Earth Sciences. 1984; 31(4):361–70.
55.
Theden-Ringl F, Hislop KP, Aplin K, Gru¨n R, Schurr MR. The chronology and environmental context of
a cave deposit and associated faunal assemblage including megafauna teeth near Wee Jasper, south-
eastern Australia. The Holocene. 2018; 28(9):1467–82.
56. Megirian D, Prideaux GJ, Murray PF, Smit N. An Australian land mammal age biochronological scheme.
Paleobiology. 2010; 36(4):658–71.
57. Schwartz LR. A revised faunal list and geological setting for Bullock Creek, a Camfieldian site from the
Northern Territory of Australia. Memoirs of Museum Victoria. 2016; 74:263–90.
58. Schaller O, Constantinescu GM. Illustrated Veterinary Anatomical Nomenclature. Stuttgart, Germany:
Enke Verlag; 2007.
59. Harvey KJ, Warburton N. Forelimb musculature of kangaroos with particular emphasis on the tammar
wallaby Macropus eugenii (Desmarest, 1817). Australian Mammalogy. 2010; 32(1):1–9.
60. Warburton NM, Yakovleff M, Malric A. Anatomical adaptations of the hind limb musculature of tree-kan-
garoos for arboreal locomotion (Marsupialia: Macropodinae). Australian Journal of Zoology. 2012; 60
(4):246–58.
61. Warburton NM, Marchal C-R. Forelimb myology of carnivorous marsupials (Marsupialia: Dasyuridae):
Implications for the ancestral body plan of the Australidelphia. The Anatomical Record. 2017; 300
(9):1589–608. https://doi.org/10.1002/ar.23612 PMID: 28524458
62. Ruff C. Body mass and hindlimb bone cross-sectional and articular dimensions in anthropoid primates.
In: Damuth J, MacFadden BJ, editors. Body Size on Mammalian Paleobiology: Estimation and Biologi-
cal Implications. New York: Cambridge University Press; 1990. p. 119–50.
63. Campione NE, Evans DC. A universal scaling relationship between body mass and proximal limb bone
dimensions in quadrupedal terrestrial tetrapods. BMC Biology. 2012; 10(1):60–.
64. Hopkins SS. Estimation of body size in fossil mammals. In: Croft DA, Su D, Simpson SW, editors. Meth-
ods in Paleoecology: Reconstructing Cenozoic Terrestrial Environments and Ecological Communities.
Cham, Switzerland: Springer; 2018. p. 7–22.
65. Casinos A. Bipedalism and quadrupedalism in Megatheriurn: an attempt at biomechanical reconstruc-
tion. Lethaia. 1996; 29(1):87–96.
66. Campione NE. MASSTIMATE: Body mass estimation for vertebrates. R package version 13. 1.3
ed2016.
PLOS ONE | https://doi.org/10.1371/journal.pone.0221824 September 13, 2019
67 / 69
Osteology and functional morphology of the limbs in Palorchestidae
67. R Core Team. R: A Language and Environment for Statistical Computing. R Foundation for Statistical
Computing. Vienna, Austria2018.
68. Murray P. Propalorchestes novaculacephalus gen. et sp. nov., a new palorchestid (Diprotodontoidea:
Marsupialia) from the Middle Miocene Camfield Beds, Northern Territory, Australia. The Beagle: Rec-
ords of the Museums and Art Galleries of the Northern Territory. 1986; 3(1):195–211.
69. Owen R. Researches on the Fossil Remains of the Extinct Mammals of Australia: with a Notice of the
Extinct Marsupials of England. London: J. Erxleben; 1877.
70. Wakefield NA. Naturalist’s Diary. Croydon, Victoria: Lo ngmans; 1967. 214 p.
71. Szalay FS. Evolutionary History of the Marsupials and an Analysis of Osteological Characters. Cam-
bridge: Cambridge University Press; 1994.
72. Vickers-Rich P, Rich TH, Coffa F. Wildlife of Gondwana: Dinosaurs and Other Vertebrates From the
Ancient Supercontinent. Bloomington, USA: Indiana University Press; 1999.
73.
Flannery TF, Gott B. The Spring Creek locality, southwestern Victoria: A late surviving megafaunal
assemblage. Australian Zoologist. 1984; 21(4–5):385–422.
74. Gillespie R, Camens AB, Worthy TH, Rawlence NJ, Reid C, Bertuch F, et al. Man and megafauna in
Tasmania: Closing the gap. Quaternary Science Reviews. 2012; 37:38–47.
75. Cockburn A, Johnson C. Patterns of growth. In: Tyndale-Biscoe C, Janssens P, editors. The Developing
Marsupial: Models for Biomedical Research. Berlin: Springer-Verlag; 1988. p. 28–40.
76.
77.
Tyndale-Biscoe H. Life of Marsupials. Collingwood, Victoria: CSIRO Publishing; 2005.
Taylor BK. The anatomy of the forelimb in the anteater (Tamandua) and its functional implications. Jour-
nal of Morphology. 1978; 157(3):347–67. https://doi.org/10.1002/jmor.1051570307 PMID: 30231602
78. White JL. Indicators of locomotor habits in xenarthrans: evidence for locomotor heterogeneity among
fossil sloths. Journal of Vertebrate Paleontology. 1993; 13(2):230–42.
79. Stock C. Cenozoic Gravigrade Edentates of Western North America: With Special Reference to the
Pleistocene Megalonychinae and Mylodontidae of Rancho La Brea. Washington: Carnegie Institution
of Washington; 1925. 331 p.
80.
Fujiwara S, Hutchinson JR. Elbow joint adductor moment arm as an indicator of forelimb posture in
extinct quadrupedal tetrapods. Proceedings of the Royal Society B: Biological Sciences. 2012; 279
(1738):2561–70. https://doi.org/10.1098/rspb.2012.0190 PMID: 22357261
81. Hildebrand M. Digging of quadrupeds. In: Hildebrand M, Bramble DM, Liem KF, Wake DB, editors.
Functional Vertebrate Morphology. Cambridge, MA: Harvard University Press; 1985. p. 89–109.
82. McAfee RK. Reassessing the taxonomy and affinities of the Mylodontinae sloths, Glossotherium and
Paramylodon (Mammalia: Xenarthra: Tardigrada). PhD thesis, Northern Illinois University. 2007.
83. MacLeod N, Rose KD. Inferring locomotor behavior in Paleogene mammals via eigenshape analysis.
American Journal of Science. 1993; 293(A):300.
84. Camens AB. Systematic and palaeobiological implications of postcranial morphology in the Diprotodon-
tidae (Marsupialia). PhD thesis, Flinders Univeristy. 2010.
85. Carey SP, Camens AB, Cupper ML, Gru¨ n R, Hellstrom JC, McKnight SW, et al. A diverse Pleistocene
marsupial trackway assemblage from the Victorian Volcanic Plains, Australia. Quaternary Science
Reviews. 2011; 30(5–6):591–610.
86. Megirian D, Murray P, Wells R. The late Miocene Ongeva local fauna of central Australia. The Beagle:
Records of the Museums and Art Galleries of the Northern Territory. 1996; 13:9–38.
87. Gillespie R, Fifield LK, Levchenko V, Wells R. New 14 c ages on cellulose from Diprotodon gut contents:
Explorations in oxidation chemistry and combustion. Radiocarbon. 2008; 50(1):75–81.
88. Windsor D, Dagg A. The gaits of the Macropodinae (Marsupialia). Journal of Zoology. 1971; 163
(2):165–75.
89.
Toledo N, De Iuliis G, Vizcaı´no SF, Bargo MS. The concept of a pedolateral pes revisited: The giant
sloths Megatherium and Eremotherium (Xenarthra, Folivora, Megatheriinae) as a case study. Journal
of Mammalian Evolution. 2017; 25(4):525–37.
90. Stirton R. The Diprotodontidae from the Ngapakaldi Fauna, South Australia. Bulletin Australian Bureau
of Mineralogical Resources. 1967; 85:1–44.
91. Marshall L, Case JA, Woodburne MO. Phylogenetic relationships of the families of marsupials. Current
Mammalogy. 1990; 2:433–505.
92. Archer M, Hand SJ, Godthelp H. Riversleigh. Balgowlah, NSW: Reed; 1991.
93. Adams B, Crabtree P. Comparative osteology: A laboratory and field guide of common North American
animals. Boston, Massachusetts: Academic Press; 2012.
PLOS ONE | https://doi.org/10.1371/journal.pone.0221824 September 13, 2019
68 / 69
Osteology and functional morphology of the limbs in Palorchestidae
94.
95.
96.
Flannery T, Plane M. A new late Pleistocene diprotodontid (Marsupialia) from Pureni, Southern High-
lands Province, Papua New Guinea. Bureau of Mineral Resources Journal of Australian Geology and
Geophysics. 1986; 10:65–76.
Fleagle JG, Anapol FC. The indriid ischium and the hominid hip. Journal of Human Evolution. 1992; 22
(4–5):285–305.
Janis CM, Buttrill K, Figueirido B. Locomotion in Extinct Giant Kangaroos: Were Sthenurines Hop-Less
Monsters? PLOS ONE. 2014; 9(10):e109888–e. https://doi.org/10.1371/journal.pone.0109888 PMID:
25333823
97. Dawson RS, Warburton NM, Richards HL, Milne N. Walking on five legs: Investigating tail use during
slow gait in kangaroos and wallabies. Australian Journal of Zoology. 2015; 63(3):192–200.
98. Wells RT, Camens AB. New skeletal material sheds light on the palaeobiology of the Pleistocene mar-
supial carnivore, Thylacoleo carnifex. PLOS ONE. 2018; 13(12):e0208020. https://doi.org/10.1371/
journal.pone.0208020 PMID: 30540785
99. Murray P. Extinctions downunder: A bestiary of extinct Australian late Pleistocene monotremes and
marsupials. In: Martin PS, Klein RG, editors. Quaternary Extinctions: A Prehistoric Revolution. Tuscon,
Arizona: The University of Arizona Press; 1984. p. 600–28.
PLOS ONE | https://doi.org/10.1371/journal.pone.0221824 September 13, 2019
69 / 69
|
10.1371_journal.pone.0240850 | RESEARCH ARTICLE
Assessment of knowledge and perceptions
towards diabetes mellitus and its associated
factors among people in Debre Berhan town,
northeast Ethiopia
Wondimeneh Shibabaw ShiferawID
Pammla Margaret Petrucka3, Yared Asmare Aynalem1
1*, Abel Gatew2, Getnet Afessa2, Tsedale Asebu1,
1 Department of Nursing, College of Health Sciences, Institute of Medicine and Health Science, Debre
Berhan University, Debre Berhan, Ethiopia, 2 Clinical Nurse at Debre Berhan Referral Hospital, Debre
Berhan, Ethiopia, 3 College of Nursing, University of Saskatchewan, Academics without Borders, Saskatoon,
Canada
* wshibabaw2015@gmail.com
Abstract
Background
Globally, diabetes is a major public health burden that results in more than 3.2 million adult
deaths per year. Currently, diabetes is increasingly becoming a major threat to global public
health, particularly in Sub-Saharan Africa. Although previous studies emphasized knowl-
edge and health beliefs about diabetes among patients living with diabetes, there is minimal
evidence about knowledge and perception of risk for developing diabetes at the community
level.
Objective
This study aimed to assess the knowledge and perception of diabetes mellitus and its asso-
ciated factors among people in Debre Berhan town, northeast Ethiopia.
Methods
A community-based cross-sectional study was conducted among 423 participants. The
study was carried out from 25 February to 10 March 2019. Data were collected using a
structured pretested questionnaire through face-to-face interviews. Data were entered into
Epi data V 3.1 and exported to SPSS V 24 for analysis. A variable with p< 0.2 in bivariable
analysis was entered into multivariable logistic regression. During multivariate analysis, vari-
ables with a p value of � 0.05 were considered significantly associated.
Result
A total of 237 (56.02%) participants had good general knowledge about diabetes mellitus. In
the multivariable analysis, participants who were single (AOR = 9.08, CI: 1.72–48), had a
family history of diabetes (AOR = 2.83; CI: 1.10–7.24), and had exposure to health
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: Shiferaw WS, Gatew A, Afessa G, Asebu
T, Petrucka PM, Aynalem YA (2020) Assessment
of knowledge and perceptions towards diabetes
mellitus and its associated factors among people in
Debre Berhan town, northeast Ethiopia. PLoS ONE
15(10): e0240850. https://doi.org/10.1371/journal.
pone.0240850
Editor: Kwasi Torpey, University of Ghana College
of Health Sciences, GHANA
Received: February 24, 2020
Accepted: October 3, 2020
Published: October 19, 2020
Peer Review History: PLOS recognizes the
benefits of transparency in the peer review
process; therefore, we enable the publication of
all of the content of peer review and author
responses alongside final, published articles. The
editorial history of this article is available here:
https://doi.org/10.1371/journal.pone.0240850
Copyright: © 2020 Shiferaw et al. This is an open
access article distributed under the terms of the
Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: All relevant data are
within the paper and it’s supporting information
files. There is no separate data set to share.
PLOS ONE | https://doi.org/10.1371/journal.pone.0240850 October 19, 2020
1 / 13
PLOS ONE Knowledge and perception towards diabetes mellitus and its associated factors
Funding: The authors received no specific funding
for this work.
Competing interests: The authors declare that they
have no competing interests.
education (AOR = 3.27; CI: 2.02–5.31) were associated with good knowledge. In this study,
few respondents (20.1%) felt that they had a higher risk of developing diabetes. Two-thirds
of respondents (62.4%) saw diabetes is a serious disease. On the other hand, approxi-
mately 67% agreed to the perceived benefits of screening.
Conclusion
Almost half of the Debre Berhan community was found to have inadequate knowledge of
diabetes mellitus. Married, higher educational status, exposure to health education, and
family history of diabetes mellitus were significantly associated with good knowledge. The
perceived risk of developing diabetes was low at the community level, although many
respondents felt that behavior change is important in the prevention of diabetes. Therefore,
policy makers, healthcare managers, and healthcare workers need to work cooperatively to
foster community knowledge towards diabetes mellitus.
Background
Diabetes mellitus (DM) is a common and devastating chronic disease [1]. Worldwide, the bur-
den of DM is rising dramatically and reaching epidemic proportions [2]. The World Health
Organization (WHO) estimates that DM affects at least 285 million people and causes 3.2 mil-
lion deaths, equating to 8700 deaths every day [3]. Due to sedentary lifestyles, rapidly growing
urbanization, and modified diets, tripling of the magnitude of diabetes mellitus is predicted
within the next 25 years, as well as an overall global burden escalation [4]. According to WHO
estimations globally, there are approximately 422 million adults living with diabetes mellitus
[5], while the International Diabetes Federation (IDF) estimates that 382 million people in
2013 will rise to 592 million by 2035 [6]. A more recent estimate suggests that the global preva-
lence will reach 642 million people in 2040 [7].
In Sub-Saharan Africa (SSA), more than 12 million people have diabetes, and there are
330,000 diabetes-related deaths, yet less than 1% of health expenditure is allocated for diabetes
[8]. In Ethiopia, approximately 4.36% (1.9 million) of the population was likely to live with dia-
betes, and the number of deaths attributed to diabetes reached 34,262 in 2013 [6]. Ethiopia is
among the top four countries for high adult diabetic populations in developing countries [9].
Overall, diabetes seriously increases the risk of costly problems, including emotional distress,
heart attack, stroke, kidney damage, blindness, neural damage leading to amputation, and
reduced quality of life [10].
Public knowledge about diabetes helps combat not only the disease itself but also its compli-
cations and medical and socioeconomic consequences [11]. Simple lifestyle modifications,
such as a healthy diet that includes reducing sugar intake, are considered to be essential for the
prevention and reduction of the incidence of diabetes mellitus [12, 13]. Numerous studies
have revealed that the proportion of good knowledge regarding diabetes mellitus was 27% in
Kenya [14], 49% in Debre Tabor town, Ethiopia [15], 52.5% in Bale Zone administrative, Ethi-
opia [16], 41.9% in Malaysia [17], 49.9% in India [18], 28.2% in rural Indian state [19], and
15% in Sudan [20]. Several studies undertaken in different countries have shown that age, gen-
der, educational level, socioeconomic status, and family history of diabetes mellitus were asso-
ciated with good knowledge about diabetes mellitus [15, 21–23].
This study is based on the Health Belief Model (HBM), which consists of perceived threat,
perceived susceptibility, perceived severity, perceived benefits, perceived barriers and cues for
PLOS ONE | https://doi.org/10.1371/journal.pone.0240850 October 19, 2020
2 / 13
PLOS ONE Knowledge and perception towards diabetes mellitus and its associated factors
action [24]. The aim of this model is to increase the perception of individuals about a health
threat and direct their behaviours towards health. Likewise, it focuses on a person’s health-
related behavior and belief in predicting future actions. The perceived risk of people develop-
ing diabetes mellitus is considered to be the primary motive to change within the Health Belief
Model, which assumes that the higher the perceived threat, the more likely an individual will
modify his or her behavior to circumvent that threat [25]. It may be an important motivating
factor for preventative health behaviors and control of disease [26]. In addition, according to
this model, the decision to participate in preventive and screening programs is determined by
many factors, such as awareness of the impact of disease on their health (perceived severity),
perceived benefits of undergoing preventive measures, and perceived barriers and costs of the
screening methods [24, 27]. Therefore, understanding individuals’ viewpoints and beliefs is
essential for developing strategies to prevent, control, and increase awareness of the health
risks of this disease [28].
Previous studies emphasized knowledge and attitudes towards diabetes mellitus. Local evi-
dence is limited on public knowledge and risk perceptions towards DM using the health belief
model in Ethiopia. Therefore, this study aimed to assess knowledge and risk perceptions
towards diabetes mellitus and its associated factors among Debre Berhan community mem-
bers. The results of this study may provide baseline data for future intervention programs to
promote early detection and early management of diabetes mellitus.
Methods
Study design and setting
This community-based, cross-sectional study was conducted from 25 February to 10 March
2019 among Debre Berhan community members. Debre Berhan is one of the 13 zones of the
Amhara regional state and the town of the North Shoa Zone. The town has nine kebeles,
among which seven are primarily urban populations, while the remaining two kebeles have
both urban and rural populations. The total population was 108,825, within 25,308 house-
holds. Moreover, there is one government hospital, one private hospital, three government
health centers, nine health posts and 18 private clinics [29].
Study participants
All adults living in Debre Berhan town for at least six months, aged greater than or equal to 18
years and consenting to participate were included in the study. Persons who were critically ill
during the data collection period and had complete loss of hearing and known DM patients
were excluded from the study.
Sample size estimation
Sample size was calculated for each study objective. For the first objective, the sample size was
determined based on the formula (Z1-α/2)2�(p(1-p))/d2, where (Z1-α/2) = 1.96, p = the pro-
portion with knowledge of diabetes mellitus, which is 52.5% from a previously published study
[16], and d = 0.05. We added 10% to the calculated sample size to compensate for the antici-
pated non-response rate, bringing the total sample size to 418 community members. Similarly,
for the second objective, appropriate sample size was estimated using a single proportion for-
mula, where n is the need sample size; d, marginal error (d = 0.05); Z, the required degree of
accuracy at 95% confidence level, which is 1.96; P = 0.5 (50%) level of risk perception towards
diabetes, as there was no study conducted in the study area to the best of literature search
made. Using the above formula, after adding a 10% non-response rate, the final sample size of
PLOS ONE | https://doi.org/10.1371/journal.pone.0240850 October 19, 2020
3 / 13
PLOS ONE Knowledge and perception towards diabetes mellitus and its associated factors
the study was 423 community members. Therefore, for this study, the largest sample size was
taken to include 423 eligible study participants.
Sampling procedure
The numbers of households were obtained from the community’s administrative office. Three
kebeles were selected by the lottery method out of nine kebeles in the town. Proportional allo-
cation was used to obtain the desired number of samples from each kebele. Systematic random
sampling was used to select the study unit among households. The first house was selected by
the lottery method, and the next households were selected after calculating sampling intervals
(K = 3). Then, every 3 households were taken until the required sample size was reached. The
first person to be encountered in the household meeting the age criteria was interviewed. For
those who failed to meet the inclusion criteria, a second person was interviewed, and if more
than one individual meeting the age criteria was present in the same household, a lottery
method was used. If there is no respondent in the selected households in two visits, the next
household was considered until the desired sample size was achieved.
Study variables
Knowledge and perception towards DM were the outcome variables. Sociodemographics,
source of information, and family history of DM were considered independent covariates.
Data collection tool and procedure
A data collection tool was developed from previous similar literature [19, 30, 31] that contains
socio-demographic data, knowledge, and perceptions based on the Health Belief Model. More-
over, the three-part data collection tool was designed based on study objectives. The first part
of the questionnaire focused on the sociodemographic information that included age, sex,
marital status, level of education, occupation, average family monthly income, family history
of diabetes mellitus, exposure to health education about DM, and source of health information.
The second part assessed the level of knowledge about diabetes mellitus among the study par-
ticipants and the level of participants’ understanding of various aspects of DM, including defi-
nition, causes/risk factors, signs/symptoms, control and management. Respondents answered
either “yes”, or “no”, or “I do not know” options. Correct responses were scored as one point,
while the responses “No” or “I do not know” were scored as zero points. All respondents with
below the mean score on knowledge questions were included among those with poor knowl-
edge, while respondents scoring above the mean on knowledge questions were regarded as
having good knowledge.
The third part of the tool is regarding perceptions towards DM based on the constructs of
the Health Belief Model scale; this scale has five subscales: Perceived susceptibility to having
disease was assessed by using three items. The first item inquired as to “My chances of getting
diabetes mellitus in the next few years are high”. The second regarding perceived seriousness
of DM was assessed by four items (e.g., “If I had diabetes, I would be worried and depressed”).
Perceived benefits of undergoing a preventive measure and screening were assessed by five
items (e.g., “I believe maintain a normal body weight help to control diabetes.”). Finally, per-
ceived barriers to screening and healthy lifestyle were evaluated using five items (e.g., “I think
having a regular health check-up takes too much time.”) [24, 27]. All items of the subscales use
five-point Likert-type response choices: Strongly Agree (5 points), agree (4 points), neutral (3
points), disagree (2 points) and strongly disagree (1 point). Using previous studies [32, 33], we
categorized agree, neutral and disagree as a baseline. Each of the subscales was assessed sepa-
rately, and the total score was not calculated. Subscale scores were calculated for each
PLOS ONE | https://doi.org/10.1371/journal.pone.0240850 October 19, 2020
4 / 13
PLOS ONE Knowledge and perception towards diabetes mellitus and its associated factors
construct. Higher scores indicated stronger feelings about that construct. In the original test,
Cronbach’s alpha coefficients for the five subscales were observed to fall between 0.65 and
0.87. In this study, Cronbach’s alpha coefficients of 0.83 were observed for the five subscales.
Data were collected through face-to-face interviews by six trained data collectors. The princi-
pal investigator of the study was controlling the overall activity.
Data quality assurance
Pretesting was performed among 5% of study participants two weeks prior to the actual data
collection period among non-sampled kebeles to avoid information contamination bias. The
questionnaire was modified based on the responses to the pretest. In addition, the question-
naire was initially developed in English, then translated to the local language (Amharic) by an
expert and then back to English to ensure consistency. Training on data collection was given
to data collectors and supervisors for one day before actual data collection. All data collectors
were trained on their responsibilities for the purposes of the study, how to collect the data,
how to maintain confidentiality, and how to ensure genuine replies on questions. Further-
more, the principal investigator strictly followed the overall activities of the data collection on
a daily basis to ensure the completeness of the questionnaires and to give further clarification.
Double data entry was employed.
Data processing and analysis
The collected data were checked for completeness, coded, entered into Epi Data VS 3.1 and
then exported to SPSS VS 22 for analysis. Descriptive statistics were used to describe the study
participants in relation to relevant variables. Then, bivariate and multivariate logistic regres-
sion was used to identify the possible associations between independent and outcome vari-
ables. Variables with p-values less than 0.2 during bivariable analysis were exported to
multivariable analysis to control for confounding effects. Variables with a p-value of � 0.05 in
the multivariable analysis were considered significantly associated with the outcome variable.
Assumption tests and model goodness of fit were checked. Finally, the results of the study are
presented in tables, figures, and text.
Ethics consideration
This study was approved by the ethical review committee of the Institute of Medicine and Col-
lege of Health Sciences, Debre Berhan University. An official letter of permission was provided
to the Debre Berhan town administration. After explaining the purpose of the research to the
study participants, verbal informed consent was obtained prior to data collection. The data
obtained were maintained in a manner to ensure confidentiality.
Results
Sociodemographic characteristics of the study participants
A total of 423 respondents participated, yielding a response rate of 100%. Within this sample,
231 (54.6%) were females; nearly half of the participants [195 (46.1%)] were married. More
than half [235 (55.6%)] of the study participants were educated at the college and above; nearly
two-thirds (64.3%) had an average family income greater than 1,540 Ethiopian birrs per
month. Most of the study participants, 255 (60.3%), were previously exposed to health educa-
tion about DM (Table 1).
PLOS ONE | https://doi.org/10.1371/journal.pone.0240850 October 19, 2020
5 / 13
PLOS ONE Table 1. Sociodemographic characteristics of study participants in Debre Berhan town, northeast Ethiopia.
Knowledge and perception towards diabetes mellitus and its associated factors
Variables
Sex
Marital status
Educational status
Occupation
Average monthly income
Exposure to health education about DM
Family history of DM
Source of information
N.B. ETB; Ethiopian birr.
https://doi.org/10.1371/journal.pone.0240850.t001
Category
Female
Male
Married
Single
Divorced/separated
Widowed
Unable to read and write
Grade 1–4
Grade 5–8
Grade 9–12
College and above
Housewife
Student
Merchant
Government/private employee
Day laborer
�1,539 ETB
�1540 ETB
Yes
No
Yes
No
Public media
Health care worker
Friends
Other (teacher & religious leader)
Frequency
Percent (%)
231
192
195
194
18
16
28
25
52
83
235
58
86
70
167
42
151
272
255
168
54
369
121
79
34
21
54.6
45.4
46.1
45.9
4.3
3.8
6.6
5.9
12.3
19.6
55.6
13.7
20.3
16.5
39.1
9.9
35.7
64.3
60.3
39.7
12.8
87.2
47.5
31.0
13.3
8.2
Knowledge of participants about diabetes mellitus
In general, the total mean score for correctly answered knowledge questions was 11.4 ±4.9. Two
hundred thirty-seven (56.02%) scored above the mean and were considered to have good knowl-
edge. In contrast, 186 (43.98%) scored below the mean were poorly knowledgeable. Among 423 par-
ticipants, approximately 47% responded correctly defined DM as high levels of sugar in the blood,
and approximately 40.7% said DM is incurable. In addition, regarding risk factors, the percentage of
participants correctly naming family history of DM was 54.4%; being overweight or obese was
64.1%; and sedentary lifestyle was 60.8% of participants. Significant proportions of study partici-
pants (79.7% and 79.2%) reported excessive hunger and feeling of weakness as signs and symptoms
of DM, respectively. Moreover, related to the management of DM, they described DM as being con-
trolled by insulin injection (77.1%), regular exercise (66.2%) and a healthy diet (73.8%) (Table 2).
Factors associated with participants’ knowledge about DM
In the multivariable analysis, participants who were married were almost eleven times more
likely to be knowledgeable than those who were widowed (AOR = 10.97, CI; 2.33, 51.64). Like-
wise, participants who were unable to read and write were 81% less likely to have good knowl-
edge than those who had college and above educational status (AOR = 0.19, CI; 0.53, 0.68).
PLOS ONE | https://doi.org/10.1371/journal.pone.0240850 October 19, 2020
6 / 13
PLOS ONE Knowledge and perception towards diabetes mellitus and its associated factors
Table 2. Participants’ knowledge about diabetes mellitus in Debre Berhan town, northeast Ethiopia.
Variables
What is/are DM
DM is a condition of insufficient insulin production
DM is a condition of the body which not responding for insulin
DM is a condition of high level of sugar in the blood
DM is not curable
What are the risk factors of DM
Older age
Genetic or family history of diabetes mellitus
Being overweight /Obesity
Sedentary life /Poor dietary habits
What are the signs and symptoms of DM
Frequent urination
Excessive thirst
Excessive hunger
Weight loss
High blood sugar
Blurred vision
Slow healing of cuts and wounds
Feeling of weakness
Control and management of DM
Insulin injection is available for control and management of DM
Tablets & capsule are available for control and management of DM
Regular exercise
Practices healthy diet
https://doi.org/10.1371/journal.pone.0240850.t002
Yes
number
120
115
199
251
213
230
271
257
206
214
337
223
258
233
249
335
326
226
280
312
%
28.4
27.2
47.0
59.3
50.4
54.4
64.1
60.8
48.7
50.6
79.7
52.7
61.0
55.1
58.9
79.2
77.1
53.4
66.2
73.8
No
number
303
308
224
172
210
193
152
166
217
209
86
200
165
190
174
88
97
197
143
111
%
71.6
72.8
53.0
40.7
49.6
45.6
35.9
39.2
51.3
49.9
20.3
47.3
39.0
44.9
41.1
20.8
2.9
46.6
33.2
26.2
Similarly, participants who had a family history of DM were almost three times more knowl-
edgeable than those who did not have a family history of DM (AOR = 2.83, CI: 1.11, 7.25). Par-
ticipants who had diabetes health education exposure were slightly more than three times as
knowledgeable as those who did not have diabetes health education exposure (AOR = 3.28, CI;
2.02, 5.31) (Table 3).
Perceptions towards diabetes mellitus based on the health belief model
The findings showed that 20.1% of the participants considered themselves to be at risk of
developing DM in the future (perceived susceptibility). Additionally, 62.4% of the participants
agreed to the statements related to the seriousness of the disease. Regarding the perceived ben-
efits of the screening and preventive measure, approximately two-thirds of the participants
believed that undergoing regular health care visits can help find DM early and save lives.
Twenty-seven percent of the participants had experienced barriers to undergoing screening
and implementing lifestyle changes, while 31.2% of participants never heard or read anything
encouraging them to have regular health check-up and 25.1% of participants felt having a reg-
ular health check-up takes too much time (Table 4).
Discussion
The current study showed that approximately 56.02% (95% CI: 53.82, 59.47) of the study par-
ticipants were knowledgeable about DM. This result is in line with a study performed in the
PLOS ONE | https://doi.org/10.1371/journal.pone.0240850 October 19, 2020
7 / 13
PLOS ONE Table 3. Bivariate and multivariate logistic regression analysis to characterize factors associated with knowledge of diabetes mellitus in Debre Berhan town, north-
east Ethiopia.
Knowledge and perception towards diabetes mellitus and its associated factors
Category
Knowledge
COR(95%CI)
AOR(95% CI)
Good Knowledge
Poor knowledge
Variable
Sex
Marital status
Male
Female
Single
Married
Divorced
Widowed
Educational status
Unable to read &write
Occupational status
Grade 1–4
Grade 5–8
Grade 9–12
College and above
House wife
Student
Merchant
Farmer
Exposure to health education
Family history of DM
Government employer
125
Daily labor
Driver
Yes
No
Yes
No
8
8
202
82
44
240
130
154
129
144
8
3
9
11
37
54
173
34
56
44
9
62
77
65
51
10
13
19
14
15
29
62
24
30
26
4
42
8
5
53
86
10
1.05(0.69, 1.57)
0.86(0.49, 1.49)
1
8.6(2.36, 31.25)�
12.24(3.35, 44.68)�
3.467(0.72, 16.53)
1
9.08(1.72, 48.00)�
10.97(2.33, 51.64)�
4.23(0.67, 26.58)
1
0.17(0.07, 0.39)�
0.28(0.12, 0.65)�
0.88(0.45, 1.72)
0.66(0.39, 1.14)
1
0.88(0.26, 3.04)
1.16(0.35, 3.88)
1.05(0.31, 3.57)
1.40(0.27, 0.71)
1.86(0.57, 5.99)
0.62(0.14, 2.76)
1
3.99(2.60, 6.13)�
1
2.37(1.15, 4.85)�
1
0.19(0.53, 0.68)�
0.24(0.07, 0.82)�
0.87(0.36, 2.07)
0.72(0.37, 1.39)
1
0.75(0.15, 3.65)
0.40(0.09, 1.77)
0.46(0.11, 1.95)
2.28(0.30, 17.10)
0.67(0.17, 2.64)
0.57(0.09, 3.36)
1
3.27(2.02, 5.31)�
1
2.83(1.10, 7.25)�
1
129
1
� For statistically significant variables at p<0.05 both at bivariable and multivariable analysis: 1 reference variable; COR: crude odds ratio; AOR; adjusted odds ratio.
https://doi.org/10.1371/journal.pone.0240850.t003
Bale zone, Ethiopia, which reported 52.2% [16]. Our finding was higher than a study in Debre
Tabor town at 49% [15], 41.9% in Malaysia [17], 15% in Sudan [20], 27.2% in Kenya [14],
49.9% in India [18], and 49.7% in Namibia [34]. This variation might be due to the study
period since the studies were conducted, for instance, in Sudan (2014), Kenya (2010), India
(2008), and Malaysia (2014). The other possible explanation could be that diabetes is now
emerging as an epidemic of the 21st century. To curb this scourge of diabetes, studies have
shown that knowledge is the greatest weapon in the fight against DM [35, 36]. In line with this
evidence, the Ethiopia government has designed a strategy to increase community awareness
and knowledge of diabetes mellitus through public media and incorporated noncommunic-
able disease prevention and control programs as one core component of health extension pro-
grams [37].
In the present study, approximately 40.5% of study participants knew the definition of DM,
with 60.7% having good knowledge about symptoms of DM, 57.4% of participants correctly
identifying the risk factors of DM, and 65.7% showing good knowledge on control and man-
agement of DM. This finding is in line with a study undertaken in a semi-urban Omani popu-
lation on diabetes definition and classical symptoms, which were reported as 46.5% and 57.0%,
respectively [38]. Likewise, a study conducted in Debre Tabor Town found that 60.3% knew
the definition, 39.1% had good knowledge about symptoms of DM, 49% identified the risk fac-
tors of DM, and 44% respondents had good knowledge on the control and management of
DM [15]. In contrast, our results were higher than those of a study conducted in Kenya on
PLOS ONE | https://doi.org/10.1371/journal.pone.0240850 October 19, 2020
8 / 13
PLOS ONE Knowledge and perception towards diabetes mellitus and its associated factors
Table 4. Health belief model scores of participants in Debre Berhan town, northeast Ethiopia.
Variables
Category
Perceived susceptibility
My chances of getting diabetes in next few years is great
I feel I will get diabetes sometime during my life
I believe all population are equally likely to develop diabetes
Average score (%)
Perceived severity
If I had diabetes, I would be worried and depressed
If I had diabetes, I would have to have my diabetes taken off by medication
DM can be a serious disease if you don’t prevent it.
If I had diabetes, it would cause me to die
Average score (%)
Perceived benefits
I believe diabetes can be cured easily
I believe maintain a normal body weight help to control diabetes
Regular health care visit can help finding diabetes early and save my life
I believe that eat low sugar snacks & low-fat meals prevent DM in the future
I believe that regularly physical exercise will help to prevent diabetes
Average score (%)
Perceived barriers
I don’t want to know if i have diabetes or not
I think having a regular health check-up takes too much time.
Not having enough money would keep me from having a check-up.
I never heard or read anything encouraging having regular health check-up.
I could not have enough time to exercise
Average score (%)
https://doi.org/10.1371/journal.pone.0240850.t004
Agree
N (%)
79(20.2)
70(16.5)
106(24.3)
85(20.1)
299(73.6)
356(83.6)
116(28.6)
285(66.1)
264(62.4)
160(37.8)
275(64.6)
373(88.2)
295(77.0)
298(70.5)
284(67.1)
72(16.8)
106(32.9)
92(21.5)
132(30.9)
85(19.8)
114(27.0)
Neutral
N (%)
46(11.7)
68(16.0)
45(10.3)
53(12.5)
6(1.6)
3(0.7)
38(9.3)
2295.1)
23(5.4)
51(12.1)
57(13.4)
18(4.3)
31(8.1)
64(15.1)
44(10.5)
128(30.0)
44(13.7)
84(19.7)
77(18.0)
18(4.2)
70(16.6)
Disagree
N (%)
266(68.1)
286(64.5)
285(65.4)
285(67.4)
101(24.8)
67(15.7)
252(62.1)
124(28.8)
136(32.2)
212(50.1)
94(22.0)
32(7.5)
57(14.9)
61(14.4)
95(22.4)
227(53.2)
172(53.4)
251(58.8)
218(51.1)
325(76.0)
239(56.4)
DM sign/symptom (29%) and risk factors (26%) [14] and one in Nigeria on knowledge of
respondents on risk factors of DM was 19.0% and symptoms of diabetes mellitus was 18.2%
[39]. The differences might be due to the Ethiopia government’s strategy to reduce the burden
of noncommunicable diseases, particularly diabetes, through incorporating a noncommunic-
able disease prevention program under the health extension package, which includes extensive
print media [40].
Study participants who were unable to read and write had 81% less likely good knowledge
about DM than those whose educational status was college and above. These findings were
supported by studies conducted in semi-urban Omani populations [38], in Debre Tabor [15],
and in Bangkok [23]. This finding might reflect that respondents who had higher education
would have the chance to obtain different information, such as leaflets and manuals, which
make them more aware about DM. Furthermore, this more highly educated group is more
likely able to communicate more easily with health care providers regarding questions or
concerns.
In the present study, those who had diabetes health education exposure were almost three
times more knowledgeable than those who did not have diabetes health education exposure.
This finding was consistent with the study conducted in Bangladesh [21] and the finding also
supported by the study done in Bale zone [16]. On the other hand, in the current study, partici-
pants who had a family history of DM were nearly three times more knowledgeable than those
who did not have a family history of DM, which is similar to findings in a semi-urban Omani
population [38], in Debre Tabor town [15], and in Southwest Ethiopia [41]. In this study, mar-
ried individuals were eleven times and single individuals nine times more knowledgeable than
those who were widowed, reflecting findings from a previous study performed in Southwest
PLOS ONE | https://doi.org/10.1371/journal.pone.0240850 October 19, 2020
9 / 13
PLOS ONE Knowledge and perception towards diabetes mellitus and its associated factors
Ethiopia [41]. This could be due to the proportion of participants who were married in our
study sample. Therefore, the authors suggest that future study is required to investigate the
correlation between marital status and level of knowledge.
According to the present study, approximately 20.1% of participants had perceived suscep-
tibility to developing diabetes in the future. This is consistent with the findings in Rwanda [42]
found that the majority of respondents thought that there was almost no chance or no chance
at all that they would develop diabetes. Additionally, 62.4% of the participants correctly agreed
to the statements related to the seriousness of the disease. Regarding perceived severity, our
finding is lower than a study done in Rwanda, where approximately 79.4% thought of DM as a
severe disease [43] and 71% in Namibia reported as a serious disease [34]. This may reflect that
there are other diseases that take priority and are viewed as more serious. In addition, not
knowing the seriousness of a condition could be due to a lack of education interventions that
can assist in alerting communities about these health conditions.
Concerning perceived benefits of the screening and preventive measure, approximately
67% of the participants believed that undergoing regular health care visits can help find DM
early and save lives. Twenty-seven percent of the participants had experienced barriers to
undergoing screening and implementing lifestyle change. For example, 31.2% of participants
never heard or read anything encouraging about having regular health check-up; 25.05% of
participants considered having a regular health check-up takes too much time; and 21.7% of
the participant did not have enough money, which would keep them from having a check-up.
This finding suggests that there is a need for education campaigns in the district to reduce bar-
riers to diabetes screening about all aspects of barriers, such as time consumption and misin-
formation. It is therefore important to identify interventions that reduce people’s perceived
barriers despite their levels of knowledge of diabetes mellitus.
Limitation of the study
Despite extensive efforts, the results of this study are subject to certain limitations. First, the
cross-sectional nature of the study design makes it impossible to form causal relationships
between exposure and outcome variables. Second, the study is conducted in one community
and one cultural group; therefore, the findings might not necessarily be generalized to be the
same in rural areas.
Conclusion
This community-based cross-sectional study showed that the overall knowledge about DM
was moderate. Educational status, marital status, family history of DM and exposure to health
education had significant associations with the mean knowledge of the study participants. The
majority of the sample had a low perception of the risk for developing diabetes. Therefore, pol-
icy makers, healthcare managers, and healthcare workers should plan health education inter-
ventions that can enhance public knowledge and the correct perception of the risk of
developing diabetes. The use of the HBM, as in this study, could form the framework for fur-
ther research on diabetes prevention among the Ethiopian population. Furthermore, future
nationwide studies based on this framework are recommended to help develop more effective
DM prevention interventions among the adult population in Ethiopia.
Supporting information
S1 File. English version of the questionnaire.
(DOCX)
PLOS ONE | https://doi.org/10.1371/journal.pone.0240850 October 19, 2020
10 / 13
PLOS ONE Knowledge and perception towards diabetes mellitus and its associated factors
S2 File. Amharic version of the questionnaire.
(DOCX)
Acknowledgments
The authors express their appreciation to Debre Berhan University and Debre Berhan town
administration for their kind cooperation during data collection. The authors are grateful to
the study participants and data collectors.
Author Contributions
Conceptualization: Wondimeneh Shibabaw Shiferaw.
Data curation: Yared Asmare Aynalem.
Formal analysis: Abel Gatew, Getnet Afessa, Tsedale Asebu, Yared Asmare Aynalem.
Investigation: Getnet Afessa, Tsedale Asebu, Yared Asmare Aynalem.
Methodology: Wondimeneh Shibabaw Shiferaw, Abel Gatew, Getnet Afessa, Tsedale Asebu,
Pammla Margaret Petrucka, Yared Asmare Aynalem.
Software: Wondimeneh Shibabaw Shiferaw, Pammla Margaret Petrucka.
Supervision: Abel Gatew, Getnet Afessa, Tsedale Asebu, Yared Asmare Aynalem.
Validation: Abel Gatew.
Visualization: Getnet Afessa.
Writing – original draft: Wondimeneh Shibabaw Shiferaw, Pammla Margaret Petrucka,
Yared Asmare Aynalem.
Writing – review & editing: Wondimeneh Shibabaw Shiferaw, Pammla Margaret Petrucka.
References
1. Aldasouqi SA, Alzahrani AS: Terminology in diabetes; an example of resistance to change. Saudi medi-
cal journal 2004, 25(9):1289–1291. PMID: 15448793
2. Stratton IM, Adler AI, Neil HA, Matthews DR, Manley SE, Cull CA, et al. Association of glycaemia with
macrovascular and microvascular complications of type 2 diabetes (UKPDS 35): prospective observa-
tional study. BMJ. 2000 Aug 12; 321(7258):405–12. https://doi.org/10.1136/bmj.321.7258.405 PMID:
10938048
3. Moodley LM, Rambiritch V. An assessment of the level of knowledge about diabetes mellitus among
diabetic patients in a primary healthcare setting. South African Family Practice. 2007 Nov 1; 49(10):16-
d.
4.
Fauci AS, editor Harrison’s principles of internal medicine. New York: McGraw-Hill, Medical Publishing
Division; 2008.
5. World Health Organization: Global report on Diabetes, Geneva, 2016 Accessed 30 August 2016.
6.
International Diabetic Federation: Diabetes Atlas, 6th ed; 2013.
7. Guariguata L MP, Forouhi N, Hambleton I, et al.: International diabetes federation atlas report: Global
burden of diabetes Belgium, Brussels 2013.
8. Abebe SM BY, Worku A, Assefa A: Diabetes mellitus in North West Ethiopia: a community based study.
BMC Public Health; 2014.
9. Abebe N, Kebede T, Addise D. Diabetes in Ethiopia 2000–2016 prevalence and related acute and
chronic complications; a systematic review. Afr J Diabetes Med. 2017 Nov 1; 25(2):7–12.
10. Mbanya J MA, & Sobngwi E: "Diabetes in sub-Saharan Africa." The Lancet 375: 2010.: 2254–2266.
PLOS ONE | https://doi.org/10.1371/journal.pone.0240850 October 19, 2020
11 / 13
PLOS ONE Knowledge and perception towards diabetes mellitus and its associated factors
11. Murugesan N SC, Shobhana R, Roglic G & Ramachandran A (2007) Awareness about Diabetes and its
Complications in the General and Diabetic Population in a City in Southern India. World Health Organi-
zation, Department of Non Communicable Disease Management, Geneva.
12.
Tuomilehto J LJ, Eriksson JG, Valle TT, Ha¨ ma¨la¨inen H, Ilanne-Parikka P, et al.: Prevention of type 2
diabetesmellitus by changes in lifestyle among subjects with impairedglucose tolerance. Uusitupa M:N
Engl J Med, 2001, 344:1343–1350. https://doi.org/10.1056/NEJM200105033441801 PMID: 11333990
13. American Diabetes Association. The prevention or delay of type 2 diabetes. Diabetes care. 2002 Apr 1;
25(4):742–9. https://doi.org/10.2337/diacare.25.4.742 PMID: 11919135
14. Maina WK NZ, Njenga EW, Muchemi EW: Knowledge, attitude, and practices related to diabetes
among community members in four provinces in Kenya: a cross-sectional study. African J Diabetes
Med 2011.
15. Asmamaw A, Asres G, Negese D, Fekadu A, Assefa G. Knowledge and attitude about diabetes mellitus
and its associated factors among people in Debre Tabor town, Northwest Ethiopia: cross sectional
study. Science. 2015 Feb 16; 3(2):199–209.
16. Kassahun CW, Mekonen AG: Knowledge, attitude, practices and their associated factors towards dia-
betes mellitus among non diabetes community members of Bale Zone administrative towns, South East
Ethiopia. A cross-sectional study. PloS one 2017, 12(2):e0170040. https://doi.org/10.1371/journal.
pone.0170040 PMID: 28152066
17. Minhat HS, Hamedon TR. Understanding towards diabetes mellitus among rural adult community in
Malaysia. World J Med Sci. 2014; 11(2):217–.
18. Rani PK, Raman R, Subramani S, Perumal G, Kumaramanickavel G, Sharma T. Knowledge of diabetes
and diabetic retinopathy among rural populations in India, and the influence of knowledge of diabetic ret-
inopathy on attitude and practice. Rural and Remote Health 2008; 8: 838. Available: www.rrh.org.au/
journal/article/838. PMID: 18656993
19. Sristi S, Sheeladevi S, Rani PK. Knowledge, attitude and practices on diabetes and diabetic retinopathy
of rural population from an Indian state. International Journal of Diabetes in Developing Countries. 2015
Mar 1; 35(1):33–8.
20. Balla SA, Ahmed HA, Awadelkareem MA. Prevalence of diabetes, knowledge, and attitude of rural, pop-
ulation towards diabetes and hypoglycaemic event, Sudan 2013. Am J Health Res. 2014 Nov 5; 2
(6):356–60.
21.
Islam FMA, Chakrabarti R, Dirani M, Islam MT, Ormsby G, Wahab M, et al.: Knowledge, attitudes and
practice of diabetes in rural Bangladesh: the Bangladesh Population based Diabetes and Eye Study
(BPDES). PLoS One 2014, 9(10):e110368. https://doi.org/10.1371/journal.pone.0110368 PMID:
25313643
22. Maina WK, Ndegwa ZM, Njenga EW, Muchemi EW. Knowledge, attitude and practices related to diabe-
tes among community members in four provinces in Kenya: a cross-sectional study. Pan African Medi-
cal Journal. 2010; 7(1).
23. Pongmesa T SL, Hl W: A Comparison of Diabetes Knowledge among Residents in Bangkok and other
Central Provinces of Thailand. 2007.
24. Rosenstock IM, Strecher VJ, Becker MH: The health belief model and HIV risk behavior change. In: Pre-
venting AIDS. edn.: Springer; 1994: 5–24.
25. Debarr KA. Review of current health education theories. Californian Journal of Health Promotion. 2004
Mar 1; 2(1):74–87.
26. Walter FM EJ: Perceptions of family history across common diseases: a qualitative study in primary
care. Fam Pract 2006.
27. Glanz K, Rimer BK, Viswanath K: Health behavior and health education: theory, research, and practice:
John Wiley & Sons; 2008.
28. Harvey JN, Lawson VL. The importance of health belief models in determining self-care behaviour in
diabetes. Diabetic Medicine. 2009 Jan; 26(1):5–13. https://doi.org/10.1111/j.1464-5491.2008.02628.x
PMID: 19125754
29. Debre Berhan town, Annual report of Debre Berhan town Administration plan comition, Amhara Region,
Ethiopia: 2018.
30. Mukhopadhyay P, Paul B, Das D, Sengupta N, Majumder R: Perceptions and practices of type 2 diabet-
ics: A cross-sectional study in a tertiary care hospital in Kolkata. International Journal of Diabetes in
Developing Countries 2010, 30(3):143.
31. Mgbahurike A. A, Amadi D, Onwusah D: knowledge attitude and perception of diabetes mellitus among
students of university of port harcourt, rivers state, south–south zone nigeria. World Journal of Pharma-
ceutical Research 2017, 6(10):92–107.
PLOS ONE | https://doi.org/10.1371/journal.pone.0240850 October 19, 2020
12 / 13
PLOS ONE Knowledge and perception towards diabetes mellitus and its associated factors
32. Al-Mutairi RL, Bawazir AA, Ahmed AE, Jradi H: Health beliefs related to diabetes mellitus prevention
among adolescents in Saudi Arabia. Sultan Qaboos University Medical Journal 2015, 15(3):e398.
https://doi.org/10.18295/squmj.2015.15.03.015 PMID: 26355752
33. Rasheed P, Soweilum LS. Knowledge and perception of breast cancer and practice of breast self-
examination among female patients attending primary health care centers in Al Khobar City, Saudi Ara-
bia. J Kuwait Med Assoc. 2013 Jun 1; 45:123–9.
34. Kambinda DN: Knowledge, attitudes and perceptions about diabetes mellitus among an urban adult
population in Windhoek, Namibia. 2017.
35. Ellis SE, Speroff T, Dittus RS, Brown A, Pichert JW, Elasy TA. Diabetes patient education: a meta-anal-
ysis and meta-regression. Patient education and counseling. 2004 Jan 1; 52(1):97–105. https://doi.org/
10.1016/s0738-3991(03)00016-8 PMID: 14729296
36. Adejoh SO. Diabetes knowledge hb, and diabetes management among the Igala, Nigeria. Sage Open.
2014 Jun 19; 4(2):2158244014539966.
37. Workie NW, Ramana GN. The health extension program in Ethiopia. 2013.
38. Al Shafaee MA, Al-Shukaili S, Rizvi SGA, Al Farsi Y, Khan MA, Ganguly SS, et al.: Knowledge and per-
ceptions of diabetes in a semi-urban Omani population. BMC public health 2008, 8(1):249.
39. Ndibuagu Edmund O, Arinze-Onyia Sussan U, Chime Onyinye H: Knowledge and attitude on diabetes
mellitus among residents of a rural community in Enugu state, Southeast, Nigeria. Int Res J Med Sci
2016, 4:3.
40.
FMOH: National strategic action plan (NSAP) for prevention & control of non-communicable diseases in
Ethiopia: 2014–2016.
41. Kassahun T, Gesesew H, Mwanri L, Eshetie T: Diabetes related knowledge, self-care behaviours and
adherence to medications among diabetic patients in Southwest Ethiopia: a cross-sectional survey.
BMC endocrine disorders 2016, 16(1):28. https://doi.org/10.1186/s12902-016-0114-x PMID:
27381349
42. Mukeshimana MM, Nkosi ZZ: Communities’ knowledge and perceptions of type two diabetes mellitus in
R wanda: a questionnaire survey. Journal of clinical nursing 2014, 23(3–4):541–549. https://doi.org/10.
1111/jocn.12199 PMID: 23789978
43. Mukeshimana M: Exploring knowledge and perceptions of Type Two Diabetes Mellitus in a selected
sector of Rwamagana District’s residents: Rwanda K waZulu-Natal November 2010.
PLOS ONE | https://doi.org/10.1371/journal.pone.0240850 October 19, 2020
13 / 13
PLOS ONE |
10.1371_journal.pone.0229417 | RESEARCH ARTICLE
Biochemical, genetic and transcriptional
characterization of multibacteriocin
production by the anti-pneumococcal dairy
strain Streptococcus infantarius LP90
Cristina Campanero1, Estefanı´a Muñoz-Atienza1¤, Dzung B. Diep2, Javier Feito1,
Sara Arbulu1, Rosa del Campo3, Ingolf F. Nes2, Pablo E. Herna´ ndez1, Carmen Herranz1,
Luis M. CintasID
1*
1 Grupo de Seguridad y Calidad de los Alimentos por Bacterias La´cticas, Bacteriocinas y Probio´ ticos (Grupo
SEGABALBP), Seccio´ n Departamental de Nutricio´n y Ciencia de los Alimentos (Nutricio´n, Bromatologı´a,
Higiene y Seguridad Alimentaria), Facultad de Veterinaria, Universidad Complutense de Madrid, Madrid,
Spain, 2 Department of Chemistry, Biotechnology and Food Science, Norwegian University of Life Sciences
(NMBU), Ås, Norway, 3 Servicio de Microbiologı´a, Hospital Universitario Ramo´ n y Cajal & Instituto Ramo´n y
Cajal de Investigacio´ n Sanitaria (IRYCIS), Madrid, Spain
¤ Current address: Centro de Investigacio´n en Sanidad Animal (CISA), Instituto Nacional de Investigacio´n y
Tecnologı´a Agraria y Alimentaria (INIA), Valdeolmos, Madrid, Spain
* lcintas@vet.ucm.es
Abstract
Streptococcus pneumoniae infections are one of the major causes of morbility and mortality
worldwide. Although vaccination and antibiotherapy constitute fundamental and comple-
mentary strategies against pneumococcal infections, they present some limitations includ-
ing the increase in non-vaccine serotypes and the emergence of multidrug-resistances,
respectively. Ribosomally-synthesized antimicrobial peptides (i.e. bacteriocins) produced
by Lactic Acid Bacteria (LAB) may represent an alternative or complementary strategy to
antibiotics for the control of pneumococal infections. We tested the antimicrobial activity of
37 bacteriocinogenic LAB, isolated from food and other sources, against clinical S. pneumo-
niae strains. Streptococcus infantarius subsp. infantarius LP90, isolated from Venezuelan
water-buffalo milk, was selected because of its broad and strong anti-pneumococcal spec-
trum. The in vitro safety assessment of S. infantarius LP90 revealed that it may be consid-
ered avirulent. The analysis of a 19,539-bp cluster showed the presence of 29 putative open
reading frames (ORFs), including the genes encoding 8 new class II-bacteriocins, as well as
the proteins involved in their secretion, immunity and regulation. Transcriptional analyses
evidenced that the induction factor (IF) structural gene, the bacteriocin/IF transporter genes,
the bacteriocin structural genes and most of the bacteriocin immunity genes were tran-
scribed. MALDI-TOF analyses of peptides purified using different multichromatographic pro-
cedures revealed that the dairy strain S. infantarius LP90 produces at least 6 bacteriocins,
including infantaricin A1, a novel anti-pneumococcal two-peptide bacteriocin.
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: Campanero C, Muñoz-Atienza E, Diep DB,
Feito J, Arbulu S, del Campo R, et al. (2020)
Biochemical, genetic and transcriptional
characterization of multibacteriocin production by
the anti-pneumococcal dairy strain Streptococcus
infantarius LP90. PLoS ONE 15(3): e0229417.
https://doi.org/10.1371/journal.pone.0229417
Editor: Anirudh K. Singh, All India Institute of
Medical Sciences, Bhopal, INDIA
Received: October 31, 2019
Accepted: February 5, 2020
Published: March 5, 2020
Peer Review History: PLOS recognizes the
benefits of transparency in the peer review
process; therefore, we enable the publication of
all of the content of peer review and author
responses alongside final, published articles. The
editorial history of this article is available here:
https://doi.org/10.1371/journal.pone.0229417
Copyright: © 2020 Campanero et al. This is an
open access article distributed under the terms of
the Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: All relevant data are
within the paper and its Supporting Information
files.
PLOS ONE | https://doi.org/10.1371/journal.pone.0229417 March 5, 2020
1 / 19
PLOS ONE Funding: This work was supported by projects
RTI2018-094907-B-I00 from Ministerio de Ciencia,
Innovacio´n y Universidades (MICINN, Madrid,
Spain), 07CSA007E from the Junta de Galicia,
Spain, AGL2012-34829 from Ministerio de
Economı´a y Competitividad (MINECO, Madrid,
Spain), and S-2009/AGR-1489 and S-2013/ABI-
2747 from Consejerı´a de Educacio´n, Comunidad de
Madrid, Madrid, Spain. C. Campanero and E.
Muñoz-Atienza were recipients of predoctoral
fellowships (FPU) from UCM, Spain. J. Feito holds
a contract from UCM. S. Arbulu held a predoctoral
fellowship (FPI) from MINECO.
Competing interests: The authors have declared
that no competing interests exist.
Biochemical, genetic and transcriptional characterization of multibacteriocin production by S.infantarius LP90
Introduction
Streptococcus pneumoniae, regarded as one of the most important human pathogens, is the eti-
ological agent of diseases such as acute otitis media, sinusitis and pneumonia, as well as of
potentially life-threatening invasive infections, such as bacteremia and meningitis, which con-
stitute major causes of morbility and mortality worldwide [1]. The development of vaccines
could alleviate the problems related to the spreading of the most virulent strains. Nevertheless,
several studies have revealed some drawbacks including the parallel increase of non-vaccine
serotypes [2]. Although antibiotherapy constitutes a complementary strategy to vaccination,
treatments with antibiotics may lead to dysbiosis and the spreading of antibiotic-resistant bac-
teria. The concerns about the worldwide emergence of antibiotic resistant S. pneumoniae
strains, mainly against β-lactams and macrolides, and the raise of clinical isolates showing
multidrug-resistance have increased [3]. Researchers are urged to discover and develop alter-
native or complementary strategies to antibiotherapy based on novel modes of action, such as
bacteriocins produced by Lactic Acid Bacteria (LAB) [4]. Bacteriocins are ribosomally-synthe-
sized antimicrobial peptides or proteins active against a wide variety of Gram-positive and, to
a lesser extent, Gram-negative bacteria [5]. This work describes for the first time bacteriocin
production by Streptococcus infantarius. The biochemical, genetic and transcriptional charac-
terization of multibacteriocin production by the dairy strain S. infantarius LP90 revealed that
it produces at least 6 bacteriocins, including infantaricin A1, a novel anti-pneumococcal two-
peptide bacteriocin which could be used as an alternative or complementary strategy to antibi-
otics for the treatment of human infections caused by S. pneumoniae, for which vaccination
and antibiotic programs may be non-fully effective.
Materials and methods
Bacterial strains, growth media and culture conditions
The 37 bacteriocinogenic LAB and the 23 clinical S. pneumoniae strains used as indicators are
listed in S1 Table. LAB were grown in de Man, Rogosa and Sharpe (MRS) broth (Oxoid Ltd.,
Basingstoke, United Kingdom) at 30˚C for 16 h. S. pneumoniae colonies were obtained on
Columbia agar with horse blood (5%, w/v) (BioMe´rieux, Marcy l’Etoile, France) at 37˚C over-
night. Subsequently, colonies were transferred to Brain Heart Infusion (BHI) broth (Oxoid),
and cultures were incubated at 37˚C for approximately 6 h (OD600 of 0.5).
Detection of antimicrobial activity
Direct antimicrobial activity assay. The direct antimicrobial activity of the 37 bacterioci-
nogenic LAB was screened by a stab-on-agar test (SOAT) [6] against 9 S. pneumoniae strains
(S2 Table). Briefly, the bacteriocinogenic strains were stabbed onto MRS agar (1.5% w/v) and
incubated at 30˚C for 5 h, and then 40 ml of BHI soft (0.8%, w/v) agar (Oxoid) containing
about 1×105 cfu/ml of the indicator strain was poured over the plates. After incubation, at
37˚C for 12 h, the plates were checked for inhibition zones around the colonies. Antimicrobial
activity was quantified by measuring the diameter of the inhibition halo (mm), considering as
positive results only diameters higher than 3 mm.
Extracellular antimicrobial activity assay. The extracellular antimicrobial activity activ-
ity of the 37 bacteriocinogenic LAB was determined by an agar well-diffusion test (ADT) [6]
against 23. S. pneumoniae strains (S2 Table). Firstly, the cell-free culture supernatants (CFS)
from the 37 bacteriocinogenic LAB grown in MRS broth at 30˚C until the stationary phase
(OD600 of ca. 1.0) were obtained by centrifugation at 12,000 × g at 4˚C for 10 min. After adjust-
ing their pH to 6.2 with 1M NaOH, they were filter-sterilized through 0.22 μm-pore-size filters
PLOS ONE | https://doi.org/10.1371/journal.pone.0229417 March 5, 2020
2 / 19
PLOS ONE Biochemical, genetic and transcriptional characterization of multibacteriocin production by S.infantarius LP90
(Millipore Corporation, Bedford, Massachusetts, USA), and stored at –20˚C until use. Then,
BHI soft agar (80 ml) were inoculated with approximately 1×105 cfu/ml of the S. pneumoniae
indicator strains and poured into a Petri dish. After cooling, wells (6-mm diameter) were
made in the agar, and 50 μl of CFS were added. After incubation of the plates at 37˚C for 12 h,
the antimicrobial activity was quantified by measuring the diameter (mm) of the inhibition
zones around the wells, and the bacteriocinogenic strains showing the highest antimicrobial
activity were selected for further evaluation.
These strains were grown in MRS broth at 30 and 37˚C for 16 h and their CFSs obtained at
6, 10, 12, and 16 h were tested for antimicrobial activity by an ADT against 7 S. pneumoniae
strains. CFSs showing the highest antimicrobial activity were challenged against the same
strains by a microtiter plate assay (MPA) [7]. Each well of the microtiter plate contained 50 μl
of two-fold serial dilutions (in BHI broth) of CFS and 150 μl of a diluted (in BHI broth) fresh
overnight culture of the indicator (approx., 4×105 cfu/ml). After aerobic incubation of the
plates at 37˚C for 12 h, growth inhibition was measured spectrophotometrically at 620 nm
with a microtiter plate reader (Labsystems iEMS Reader MF, Labsystems, Helsinki, Finland).
CFSs were heated at 100˚C for 10 min and treated with proteinase K (AppliChem GmbH,
Darmstadt, Germany) (10 mg/ml) at 37˚C for 1 h.
Taxonomic identification of Streptococcus sp. LP90
Streptococcus sp. LP90 was taxonomically identified by DNA sequencing of a PCR-amplified
internal fragment of the manganese-dependent superoxide dismutase gene (sodA) using the
primers d1 (5’-CCITAYICITAYGAYGCIYTIGARCC-3’) and d2 (5’-ARRTARTAIGC
RTGYTCCCAIACRTC-3’) obtained from Sigma-Genosys Ltd. (Cambridge, United King-
dom) [8]. PCR mixtures were subjected to thermal cycling under the following conditions: ini-
tial denaturation (97˚C for 3 min) followed by 35 cycles of denaturation (94˚C for 30 sec),
annealing (50˚C for 2 min), and elongation (72˚C for 1 min 30 sec), and a final 12-min exten-
sion at 72˚C, in an Eppendorf Mastercycler DNA thermal cycler (Eppendorf, Hamburg, Ger-
many). The amplified PCR products were resolved by electrophoresis through agarose (2%,
w/v) (Pronadisa, Madrid, Spain) gels stained with GelRed Nucleic Acid Gel Stain (Biotium,
USA), and analyzed with the Gel Doc 1000 documentation system (Bio-Rad, Madrid, Spain).
The molecular size marker used was HyperLadder II (Bioline London, United Kingdom). The
PCR products were purified using the NucleoSpin Extract II (Macherey-Nagel, Du¨ren, Ger-
many) and sequenced at the Genomics Service of Parque Cientı´fico-UCM (Madrid, Spain).
The nucleotide BLAST software (NCBI [http://blast.ncbi.nlm.nih.gov/Blast.cgi?PROGRAM=
blastn&PAGE_TYPE=BlastSearch&LINK_LOC=blasthome]) was used for nucleotide similar-
ity search.
Safety assessment of S. infantarius LP90
Antimicrobial susceptibility testing. The minimum inhibitory concentration (MIC) of
16 antibiotics against S. infantarius LP90 was determined using Lact-1 and Lact-2 VetMIC
microplates (National Veterinary Institute, Uppsala, Sweden) [9]. The antibiotics evaluated
were ampicillin, penicillin, vancomycin, erythromycin, virginiamycin, tetracycline, clindamy-
cin, chloramphenicol, kanamycin, gentamycin, streptomycin, neomycin, linezolid, rifampicin,
ciprofloxacin, and trimethoprim. Individual colonies of S. infantarius LP90 were suspended in
a sterile glass tube containing 5 ml of saline solution (0.85% NaCl, w/v) to a turbidity of 1 in
the McFarland scale (approx., 3×108 cfu/ml) and further diluted 1,000-fold in susceptibility
test medium (LSM medium) (Iso-Sensitest [IST]: MRS, 9:1) (Oxoid). Then, 100 μl of the
diluted S. infantarius LP90 suspension was added to each microplate well, which was sealed
PLOS ONE | https://doi.org/10.1371/journal.pone.0229417 March 5, 2020
3 / 19
PLOS ONE Biochemical, genetic and transcriptional characterization of multibacteriocin production by S.infantarius LP90
with a transparent covering tape and incubated at 37˚C for 18 h. After incubation, MICs were
defined as the lowest concentration of the antibiotic that inhibited bacterial growth. Due to the
lack of breakpoints for S. infantarius, the results were interpreted according to those estab-
lished for Streptococcus thermophilus by the European Food Safety Authority (EFSA) [10]. S.
thermophilus CECT801 was used as control.
Phenotypic detection of virulence factors. Production of gelatinase, β-hemolysin and
protease activity were determined on Todd Hewitt agar containing gelatine (3%, w/v) (Oxoid),
Columbia agar supplemented with horse blood (5%, w/v), and Tryptone Soya agar (TSA) sup-
plemented with skim milk (1.5%, w/v) (Oxoid), respectively. E. faecalis P4 and E. faecalis P36
[11] were used as positive controls. S. infantarius LP90 was grown in MRS broth at 37˚C in
aerobic conditions for 16 h and then inoculated into MRS agar to obtain single colonies that
were streaked onto the culture media cited above. After incubation at 37˚C for 24 h, the plates
containing gelatine were placed at 4˚C for 5 h to facilitate the detection of zones of turbidity
(indicating hydrolysis) around the colonies. The production of β-hemolysin and protease
activity were detected by the presence of clear zones around the colonies.
PCR-detection of genes encoding virulence factors. The presence of streptococcal and
enterococcal virulence factor genes in S. infantarius LP90 was determined using the primer
pairs listed in S3 Table. PCR mixtures were subjected to an initial cycle of denaturation (97˚C
for 2 min), followed by 35 cycles of denaturation (94˚C for 45 sec), annealing (at an appropri-
ate temperature [S3 Table] for 30 sec), and elongation (72˚C for 30 to 180 s), ending with a
final extension step at 72˚C for 7 min. PCR products were resolved and analyzed as described
above. The molecular size markers used were HyperLadder II (Bioline London, United King-
dom) and 1 Kb Plus DNA ladder (Invitrogen).
Screening of histidine, tyrosine and ornithine decarboxylase activities in differential
growth medium. Histidine, tyrosine and ornithine decarboxylase activities were evaluated
using the improved decarboxylase differential growth medium [12]. S. infantarius LP90 was
subcultured 5 times in MRS broth containing 0.1% (w/v) of the corresponding amino acid pre-
cursor (histidine, tyrosine or ornithine to check for HDC, TDC and ODC activities, respec-
tively) (Sigma-Aldrich) and 0.005% (w/v) of pyridoxal-5-phosphate at 30˚C overnight.
Subsequently, cultures were streaked in duplicate on decarboxylase differential growth
medium agar plates with or without (negative control) the corresponding amino acid precur-
sor, and incubated at 37˚C for 4 days under aerobic and anaerobic conditions. Biogenic amine
production was detected by a color change from yellow to violet in the decarboxylase medium.
Lactobacillus brevis CECT4121 was used as positive control for tyramine, and as negative con-
trol for histamine and putrescine. Likewise, Lactobacillus sp. ATCC30a was used as positive
control for histamine and putrescine, and as negative control for tyramine [13].
PCR-detection of histidine, tyrosine and ornithine decarboxylase genes. Total DNA
from S. infantarius LP90 was subjected to PCR-amplification to detect histidine, tyrosine and
ornithine decarboxylase genes (hdc, tdc and odc, respectively) using the primers CL1/JV17HC
[14], TD2/TD5 [15] and 3/16 [16], respectively. Lb. brevis CECT4121 was used as positive con-
trol for tdc, and as negative control for hdc and odc. Lactobacillus sp. ATCC30a was used as
positive control for hdc and odc, and as negative control for tdc. PCR mixtures prepared as
described above were subjected to an initial cycle of denaturation (94–97˚C for 2 to 4 min), fol-
lowed by 35 cycles of denaturation (94–95˚C for 45 to 60 s), annealing (48–52˚C for 30 sec),
and elongation (72˚C for 60 to 180 s), ending with a final extension step at 72˚C for 7 min.
PCR products were resolved and analyzed as described above.
Degradation of mucin. The ability to degrade gastric mucin was determined as previously
described [17]. Mucin from porcine stomach type III (Sigma-Aldrich) and agar were added to
medium B (minimal anaerobic culture medium) without glucose at concentrations of 0.5 and
PLOS ONE | https://doi.org/10.1371/journal.pone.0229417 March 5, 2020
4 / 19
PLOS ONE Biochemical, genetic and transcriptional characterization of multibacteriocin production by S.infantarius LP90
1.5% (w/v), respectively. A volume of 10 μl of an overnight liquid culture of S. infantarius LP90
was inoculated onto the surface of medium B. The plates were incubated anaerobically at 37˚C
for 72 h, stained with amido black (0.1%, w/v) (Merck) in 3.5 M acetic acid (Merck) for 30
min, and then washed with 1.2 M acetic acid. A discolored zone around the spotted culture
was considered as a positive result. A fresh fecal slurry from a healthy adult horse was used as
positive control.
Deconjugation of bile salts. The ability to deconjugate bile salts was determined as previ-
ously described [18]. For this purpose, plates were prepared by adding sodium salts of tauro-
cholate (TC) and taurodeoxycholate (TDC) (0.5%, w/v) (Sigma-Aldrich) to MRS agar
supplemented with L-cysteine (0.05%, w/v) (Merck). A volume of 10μl of an overnight culture
of S. infantarius LP90 was spotted onto the MRS agar plates supplemented with the bile salts,
and the plates were incubated anaerobically at 37˚C for 72 h. The presence of precipitated bile
acid around the spotted cultures was considered as a positive result. A fresh fecal slurry from a
healthy adult horse was used as positive control.
Determination of other enzymatic activities. The APIZYM kit (BioMe´rieux) was used
to determine the presence of the following enzymatic activities in S. infantarius LP90: esterase
(C4), esterase lipase (C8), trypsin, α-chymotrypsin, leucine arylamidase, valine arylamidase,
cystine arylamidase, acid phosphatase, naphthol-AS-BI phosphohydrolase, alkakine phospha-
tase, α-galactosidase, β-galactosidase, α-glucosidase, β-glucosidase, N-acetyl-β-glucosamini-
dase, β-glucuronidase, α-mannosidase, and α-fucosidase. Cells from an overnight S.
infantarius LP90 culture were harvested by centrifugation at 12,000 × g for 2 min, resuspended
in 2 ml of API Suspension Medium (BioMe´rieux) and adjusted to a turbidity of 5–6 in the
McFarland scale (approx., 1.5–1.9×109 cfu/ml). Aliquots of 65 μl of the suspensions were
added to each of the 20 reaction cupules in the APIZYM strip. The strips were incubated at
37˚C for 4.5 h and the reactions were developed by the addition of one drop of each of the API-
ZYM reagents A and B.
PCR-detection of streptococcal and enterococcal bacteriocin structural
genes in S. infantarius LP90
The presence of streptococcal and enterococcal bacteriocin structural genes in S. infantarius
LP90 was determined by PCR using the primer pairs listed in S4 Table. PCR mixtures were
subjected to an initial cycle of denaturation (97˚C for 2 min), followed by 35 cycles of denatur-
ation (94˚C for 45 sec), annealing (at an appropriate temperature for 30 sec) and elongation
(72˚C for 30 to 1 min 45 sec), ending with a final extension step at 72˚C for 7 min. PCR prod-
ucts were resolved and analyzed as described above.
Purification, mass spectrometry and amino acid sequencing of the
bacteriocins secreted by S. infantarius LP90
Bacteriocins secreted by S. infantarius LP90 were purified using modifications of the multi-
chromatographic procedure described by Cintas et al. [6]. An overnight preculture of this
strain was used to inoculate three 1 liter MRS broth bottles which were incubated at 37˚C
under aerobic conditions for 12 h. After the cells were removed by centrifugation, the proteins
present in the CFSs were subjected to precipitation with ammonium sulphate (50% [w/v]
[Merck, Farma y Quı´mica S. A., Barcelona, Spain]). The solution was then stirred at 4˚C for 3
h and subsequently centrifuged (12,000 × g at 4˚C for 30 min). The pellet and floating materi-
als obtained after centrifugation were mixed and solubilized in 200 ml of 20 mM sodium phos-
phate buffer (pH 6.0) and desalted by gel filtration (PD-10 columns; GE Healthcare Life
Sciences, Barcelona, Spain). From this point, each of the three concentrated and desalted CFSs
PLOS ONE | https://doi.org/10.1371/journal.pone.0229417 March 5, 2020
5 / 19
PLOS ONE Biochemical, genetic and transcriptional characterization of multibacteriocin production by S.infantarius LP90
was subjected to a different procedure: (i) cation-exchange (HiPrep 16/10 SP XL, GE Health-
care) and reversed-phase (PepRPC HR 5/5, GE Healthcare) chromatographies, both per-
formed in a Fast Protein Liquid Chromatography system (A¨ KTA-FPLC) (GE Healthcare)
(procedure P1); (ii) anion-exchange (CM Sepharose Fast Flow, GE Healthcare) and hydropho-
bic-interaction (Octyl Sepharose CL-4B, Sigma-Aldrich) batch chromatographies, followed by
reversed-phase chromatography in a A¨ KTA-FPLC system (procedure P2), and (iii) cation-
exchange (SP Sepharose Fast Flow, GE Healthcare) and hydrophobic-interaction batch chro-
matographies (procedure P3).
The antimicrobial activity of the fractions was determined by an MPA against S. pneumo-
niae 66405 and Pediococcus damnosus CECT4797. Purified peptides were subjected to mass
spectrometry analysis in a Matrix-Assisted Laser Desorption Ionization-Time Of Flight Voy-
ager-DE STR mass spectrometer (MALDI-TOF MS) (PerSeptive Biosystems, Foster City, Cali-
fornia, USA) at the Mass Spectrum Service of Parque Cientı´fico-UCM. N-terminal amino acid
sequencing of the purified anti-pneumococcal peptide was performed by Edman degradation
using an ABI 494 Protein Sequencing System (Applied Biosystems, Foster City, California
USA) at the Protein Chemistry Service of the Centro de Investigaciones Biolo´gicas (CIB, Con-
sejo Superior de Investigaciones Cientı´ficas, Madrid, Spain). The protein BLAST software
(National Center for Biotechnological Information [NCBI], US National Library of Medicine,
http://blast.ncbi.nlm.nih.gov) was used for amino acid similarity search. The N-terminal
extensions of the putative antimicrobial peptides and the transmembrane segments of their
putative immunity proteins were identified by the ExPASy (https://www.expasy.org/
proteomics) and SOSUI (http://harrier.nagahama-i-bio.ac.jp/sosui) programs, respectively.
Nucleotide sequencing of a multibacteriocinogenic gene cluster from S.
infantarius LP90
Total DNA from S. infantarius LP90 was isolated using the Wizard Genomic DNA Purifica-
tion kit (Promega). Specific oligonucleotide primers (Sigma-Genosys Ltd) designed according
to the nucleotide sequence of S. infantarius ATCC BAA-102 (Genbank accession no.
ABJK02000019) and to that of S. infantarius LP90 progressively determined by chromosome
walking were used to obtain the sequence of the multibacteriocinogenic cluster. PCR mixtures
were subjected to an initial cycle of denaturation (94˚C for 2 min), followed by 35 cycles of
denaturation (94˚C for 15 sec), annealing (at an appropriate temperature for 30 sec), and elon-
gation (68˚C for 30 sec to 1 min), ending with a final extension step at 68˚C for 7 min. The
resulting products were resolved and analyzed as described above. The PCR-generated frag-
ments were purified with the Nucleospin Extract II kit and sequenced at the Genomics Service
of Parque Cientı´fico-UCM. Nucleotide sequence analysis was performed using the ORF-finder
(NCBI) and the BLAST tools for the prediction of ORFs and similarity searches, respectively.
The identification of rho-independent transcription terminators was performed using the
ARNold finding terminator tool (Institut de Ge´ne´tique et Microbiologie, Universite´ Paris-
Sud11 [http://rna.igmors.u-psud.fr/toolbox/arnold]). The 19,539 bp sequence obtained was
deposited in GenBank under accession no. KJ475434.
Transcriptional analysis of a multibacteriocinogenic cluster from S.
infantarius LP90
Total RNA was isolated from exponentially growing S. infantarius LP90 cultures (OD600 of
0.6–0.9) using the RNeasy Mini kit (Qiagen, Hilden, Germany), according to the procedure
recommended by the manufacturer. The isolated RNA was treated with RNase-free DNase I
(Qiagen) to thoroughly remove residual DNA. One-step reverse transcription (RT)-PCR was
PLOS ONE | https://doi.org/10.1371/journal.pone.0229417 March 5, 2020
6 / 19
PLOS ONE Biochemical, genetic and transcriptional characterization of multibacteriocin production by S.infantarius LP90
performed using the Access RT-PCR System (Promega) with DNase-treated RNA as template
for cDNA synthesis. Specific oligonucleotide primers, obtained from Sigma-Genosys Ltd.,
were designed according to the localization of promoter and transcription termination regions
found in the nucleotide sequence of S. infantarius LP90 (19,539 bp) [region I (coordinates 1–-
3,878), sub-region IIa (coordinates 3,886–6,776), sub-region IIb (coordinates 6,829–7,877),
sub-region IIc (coordinates 7,891–9,746), sub-region IId (coordinates 9,764–11,195), sub-
region IIe (coordinates 11,212–11,953), and region III (coordinates 15,197–17,625). PCR
amplifications using DNase-treated RNA as template were also performed as negative control.
RT-PCR and PCR amplifications were carried out as recommended by the manufacturer (Pro-
mega). The resulting products were resolved and analyzed as described above.
Results
Antimicrobial activity of bacteriocinogenic LAB against clinical S.
pneumoniae strains
The 37 bacteriocinogenic LAB, isolated in previously reported works (S1 Table), displayed a
broad direct antimicrobial activity spectrum, inhibiting the growth of most of the 9 tested clin-
ical S. pneumoniae strains (S2 Table). Thirty-one strains (84.2%) exhibited extracellular anti-
microbial activity against, at least, one of the 23 tested clinical S. pneumoniae strains. The 6
bacteriocinogenic LAB showing the broadest spectrum and strongest extracellular antimicro-
bial activity were Streptococcus sp. LP90, L. lactis BB24, E. faecalis DBH18, E. faecium BNM58,
L. garvieae DCC43 and E. faecium L50. The anti-pneumococcal activity of these strains with-
stood heat treatment at 100˚C for 1 h and was sensitive to proteinase K, indicating the protein-
aceous nature of these antimicrobial compounds (i.e., bacteriocins). These bacteriocinogenic
strains were grown in MRS broth at 30 and 37˚C for 16 h to ascertain the best conditions for
maximum bacteriocin production (results not shown). Streptococcus sp. LP90 grown at 37˚C
for 12 h displayed the strongest extracellular antimicrobial activity and the broadest anti-pneu-
mococcal spectrum inhibiting the growth of the 7 S. pneumoniae strains used as indicators
(results not shown). This strain was taxonomically identified as Streptococcus infantarius
subsp. infantarius and selected for further studies.
Safety assessment of S. infantarius LP90
S. infantarius LP90 was sensitive to all the tested antibiotics for which a breakpoint has been
established by EFSA (Table 1). S. infantarius LP90 showed no hemolysin, gelatinase and prote-
ase activities (S1 Fig). None of the tested genes encoding streptococcal and enterococcal viru-
lence factors were detected in this strain (results not shown). Histidine, tyrosine and ornithine
decarboxylase activities (S2 Fig) and the respective genes hdc, tdc and odc were not detected in
S. infantarius LP90 (results not shown). Only the peptidase activities leucine and cystine aryla-
midases were found (results not shown). Mucine degradation and bile salt deconjugation abili-
ties were not detected in S. infantarius LP90 (S3 Fig).
PCR-detection of streptococcal and enterococcal bacteriocin structural
genes in S. infantarius LP90
None of the tested streptococcal and enterococcal bacteriocin genes were amplified by PCR in
this bacteriocinogenic strain (results not shown).
PLOS ONE | https://doi.org/10.1371/journal.pone.0229417 March 5, 2020
7 / 19
PLOS ONE Biochemical, genetic and transcriptional characterization of multibacteriocin production by S.infantarius LP90
Table 1. Antibiotic susceptibility of S. infantarius LP90a.
Mechanism of action/ Family
Cell wall synthesis inhibitors
Penicillins
Glycopeptides
Protein synthesis inhibitors
Macrolides
Streptogramins
Tetracyclines
Lincosamides
Amphenicols
Aminoglycosides
Oxazolidinone
Nucleic acid synthesis inhibitors
Rifamycins
Fluoroquinolones
Trimethoprim
Antibiotic
Ampicillin
Penicillin
Vancomycin
Erythromycin
Virginiamycin
Tetracycline
Clindamycin
Chloramphenicol
Kanamycin
Gentamycin
Streptomycin
Neomycin
Linezolid
Rifampicin
Ciprofloxacin
Trimethoprim
Test range (μg/ml)
MICs (μg/ml)b
FEEDAP BPs valuesc, d
Susceptibility/ Resistanced
0.03–16
0.03–16
0.25–128
0.016–8
0.016–8
0.12–64
0.03–16
0.12–64
2–1024
0.5–256
0.5–256
0.5–256
0.25–16
0.12–64
0.25–128
0.12–64
�0.06
�0.25
0,12
0,50
0,12
0,25
0,06
2,00
64,00
2,00
32,00
�8.00
�1.00
�0.50
�2.00
�32.00
2
Nav
4
2
Nav
4
2
4
64
32
64
Nav
Nav
Nav
Nav
Nav
S
Na
S
S
Na
S
S
S
S
S
S
Na
Na
Na
Na
Na
aAntibiotic susceptibility determined by a broth microdilution test. MICs were defined as the lowest concentration of the antibiotic at which no growth was detected.
bMICs determined by a VetMIC test
cBreakpoints (BPs) for S. thermophilus proposed by EFSA [10]. BPs for penicillin, virginiamycin, neomycin, linezolid, rifampicin, ciprofloxacin, and trimethoprim are
not available (Nav).
dBased on the EFSA [10] BPs, S. infantarius LP90 was categorized as susceptible (S) or resistant (R) to the corresponding antibiotic. Na, not applicable.
https://doi.org/10.1371/journal.pone.0229417.t001
Purification, mass spectrometry and amino acid sequencing of the anti-
pneumococcal bacteriocin secreted by S. infantarius LP90
Purification of the anti-pneumococcal bacteriocin produced by S. infantarius LP90 revealed a
single well-separated absorbance peak with antimicrobial activity which eluted at 32% (v/v)
2-propanol in aqueous 0.1% (v/v) TFA (results not shown). The purity and molecular mass of
the purified bacteriocin was determined by MALDI-TOF MS, which revealed a major peptide
with a molecular mass of 3,963 Da (Fig 1). An N-terminal amino acid sequence of 19 residues
of this peptide was obtained by Edman degradation (Fig 2). This anti-pneumococcal bacterio-
cin, which was termed infantaricin A1 (InfA1), is highly similar (74% of identity) to a “bacteri-
ocin-type signal sequence” from S. infantarius subsp. infantarius ATCCBAA-102 deposited in
the Genbank database (accession no. EDT47634.1) (Fig 2).
Nucleotide sequence and organization of a multibacteriocinogenic gene
cluster from S. infantarius LP90
The 19,539-bp sequence of S. infantarius LP90 (accession no. KJ475434) shows a 99% identity
to a nucleotide sequence of S. infantarius ATCCBAA-102, and includes 29 putative ORFs
(orf1-orf29) (Table 2 and Fig 3) located on the same DNA strand (except orf1, orf2, and orf29)
and preceded by putative ribosome binding sites (RBSs), except orf14 and orf21. These ORFs
can be grouped into three functional regions (Fig 3): (i) region I (orf1-orf2), related to
PLOS ONE | https://doi.org/10.1371/journal.pone.0229417 March 5, 2020
8 / 19
PLOS ONE Biochemical, genetic and transcriptional characterization of multibacteriocin production by S.infantarius LP90
Fig 1. Mass spectrometry analysis of the purified anti-pneumococcal bacteriocin infantaricin A1 secreted by S. infantarius LP90.
https://doi.org/10.1371/journal.pone.0229417.g001
bacteriocin/IF secretion and organized in one operon; (ii) region II (orf3-orf26) involved in
bacteriocin synthesis and immunity and organized in seven operons (sub-divided into six sub-
regions, IIa to IIf), and (iii) region III (orf27-orf29), related to regulation of the bacteriocin pro-
duction and organized in two operons. The region II (sub-regions IIa, IIc, and IId) includes 13
genes likely encoding bacteriocins according to their homology to known bacteriocins
(Table 2), physico-chemical characteristics (Table 3) and the presence of N-terminal exten-
sions of the GG-type of 15–30 amino acid residues (results not shown). The proteins encoded
by orf9, orf10, orf25 and orf26 contain six, three, three and two transmembrane segments,
respectively, and orf11 and orf16 seem to encode soluble proteins (results not shown).
Transcriptional analysis of a multibacteriocinogenic cluster from S.
infantarius LP90
mRNA transcripts of the expected sizes were detected in the following regions (Fig 3): (i)
region I (putative transcript T1, PT1) (orf1-orf2); (ii) region II, sub-region IIa (PT2, orf3-orf9),
sub-region IIb (PT3, orf10-orf11), sub-region IIc ([PT4, orf12-orf14] and [PT5, orf15-orf16]),
Fig 2. Alignment of the partial N-terminal sequence of infantaricin A1 and the bacteriocin-type signal sequence from S. infantarius subsp.
infantarius ATCCBAA-102. Identical residues are shaded in grey. “X” indicates an unidentified residue.
https://doi.org/10.1371/journal.pone.0229417.g002
PLOS ONE | https://doi.org/10.1371/journal.pone.0229417 March 5, 2020
9 / 19
PLOS ONE Biochemical, genetic and transcriptional characterization of multibacteriocin production by S.infantarius LP90
Table 2. Predicted Open Reading Frames (ORFs) in the 19,539-bp fragment of genome from S. infantarius LP90 and the closest relationships of the deduced
proteins.
Name
Length (amino
acids)
Molecular mass
(kDa)
pI
Closest relationshipa (number of amino acid; E-value, level of amino acid identity/
level of amino acid similarity; microorganism) (Accession no.)
Protein
ORF
Gene
Position in
nucleotide
sequence
5’
ORF1
1.563
3’
169
ORF2
3.722
1.574
ORF3
3.997
4.254
ORF4
4.272
4.466
ORF5
4.662
4.889
ORF6
4.873
5.079
ORF7
5.548
5.706
ORF8
5.734
5.886
ORF9
5.929
6.604
orf1
orf2
orf3
(infC2)
orf4
(infC1)
orf5
(infB2)
orf6
(infB1)
orf7
(infA2)
orf8
(infA1)
orf9
ORF10
6.944
7.246
orf10
ORF11
7,540
7.842
orf11
ORF12
8,000
8.233
ORF13
8.253
8.441
ORF14
8.721
8.936
ORF15
9.191
9.409
ORF16
9.409
9.705
ORF17
9.874
10.107
ORF18 10.112 10.282
ORF19 10.297 10.497
ORF20 10.523 10.705
ORF21 10.677 10.829
ORF22 10.854 11.069
ORF23 11.323 11.553
ORF24 11.642 11.951
orf12
(infD1)
orf13
(infD2)
orf14
orf15
(infE)
orf16
orf17
(infF)
orf18
orf19
(infG1)
orf20
(infG2)
orf21
orf22
orf 23
(infH)
orf24
464
715
85
64
75
68
52
51
224
100
100
77
62
71
72
98
77
56
66
60
50
71
76
51.4
5.90
80.1
6.70
Bacteriocin secretion accesory protein (464; 0.0, 99%/99%; S. gallolyticus)
(WP_074658230.1)
Peptide cleavage/export ABC transporter ATP (715; 0.0, 99%/99%; S. infantarius)
(WP_043878344.1)
8.2
6.0
7.2
6.9
5.5
5.8
25.2
11.4
4.60
Bacteriocin BlpJ (85; 1e-15, 98%/98%; S. infantarius ATCC BAA-102) (EDT47628.1)
4.50 Class IIb bacteriocin, lactobinA/cerein7B family (71; 2e-21, 100%/100%; S. infantarius
4.00
subsp. infantarius ATCC BAA-102) (EDT47629.1)
Bacteriocin BlpJ (85; 1e-05, 83%/86%; S. infantarius ATCC BAA-102)
(EDT47628.1)
4.30 Class IIb bacteriocin, lactobin A/cerein 7B family (68; 1e-23, 99%/98%; Streptococcus
9.80
caballi) (WP_040807669.1)
Hypothetical protein (52; 7e-28; 100%/100%; S. infantarius ATCC BAA-102)
(EDT47633.1)
7.70
Bacteriocin (54; 2e-08, 53%/70%; Streptococcus thoraltensis) (WP_018380431.1)
9.30 CAAX amino terminal protease family membrane protein (224; 3e-129, 100%/100%;
9.20
S. infantarius) (SUN69110.1)
Putative bacteriocin self-immunity protein (100; 1e-55, 100%/100%; S. infantarius)
(SUN69109.1)
11.6
7.90
4.90
4.70
Enterocin A immunity protein (100; 8e-64, 100%/100%; S. infantarius)
(SUN69108.1)
Bacteriocin-like peptide BlpJ (77; 1e-45, 100%/100%; S. infantarius)
(SUN69107.1)
Class IIb bacteriocin, lactobin A/cerein 7B family (41; 0.075, 100%/100%;
S. infantarius ATCC BAA-102) (EDT47640.1)
9.10 Hypothetical protein (71; 4e-12, 94%/98%; S. gallolyticus subsp. gallolyticus TX20005)
(EFM28808.1)
Bacteriocin piscicolin-126 (72; 5e-41, 93%/95%; Streptococcus equinusATCC 700338)
(EFM26697.1)
Bacteriocin immunity protein (98; 9e-62, 99%/100%;
S. infantarius CJ18) (AEZ63022.1)
Bacteriocin BlpK (77; 2e-29, 100%/100%; S. gallolyticus) (ALT80184.1)
Hypothetical protein (65; 2e-21, 98%/100%; S. caballi) (WP_018364974.1)
Bacteriocin class II (66; 2e-18, 98%/98%; S. infantarius CJ18)
(AEZ63021.1)
Bacteriocin class II (60; 9e-14, 100%/100%; S. infantarius CJ18)
(AEZ63020.1)
Hypothetical protein (48; 6e-21, 88%/93%; S. gallolyticusTX20005) (EFM28809.1)
Hypothetical protein (71; 2e-15, 100%/100%; S. gallolyticus) (SEF21267.1)
Bacteriocin class II with double-glycine leader peptide (76;6e-49,100%/100%; S.
infantarius) (SUN69099.1)
Hypothetical protein (103; 9e-69, 98%/99%; S. equinus) (EFM26700.1)
4.90
9.40
7.70
9.80
6.00
4.30
9.52
8.80
5.64
9.30
(Continued )
7.5
6.0
8.4
7.7
11.3
7.8
6.8
6.5
5.8
5.9
8.3
8.0
103
12.0
PLOS ONE | https://doi.org/10.1371/journal.pone.0229417 March 5, 2020
10 / 19
PLOS ONE Biochemical, genetic and transcriptional characterization of multibacteriocin production by S.infantarius LP90
Name
Length (amino
acids)
Molecular mass
(kDa)
pI
Closest relationshipa (number of amino acid; E-value, level of amino acid identity/
level of amino acid similarity; microorganism) (Accession no.)
Protein
Table 2. (Continued)
ORF
Gene
Position in
nucleotide
sequence
5’
3’
ORF25 12.367 12,720
orf25
ORF26 12.901 13.947
orf26
ORF27 15.304 16.029
orf27
ORF28 16.042 17.351
orf28
ORF29 17.521 17.373
orf29
117
348
241
439
21
13.0
8.79
Immunity protein BlpL (134; 6e-62,99%/99%; S. infantarius CJ18)
(AEZ63016.1)
37.6
4.16
LPXTG cell wall anchor domain-containing protein (348; 0.0, 100%/100%;
27.9
5.25
51.4
5.85
2.6
6.90
S. infantarius) (WP_006531888.1)
DNA-binding response regulator (241; 2e-177, 99%/100%; Streptococcus sp.)
(HAK39972.1)
Histidine kinase (439; 0.0, 99%/99%; Streptococcus sp.)
(HAK39971.1)
Protein of the ComC family (49; 5e-28, 100%/100%; S. infantarius CJ18)
(AEZ63012.1)
aResults of the similarity searches conducted on October of 2019.
https://doi.org/10.1371/journal.pone.0229417.t002
sub-region IId (PT6, orf17-orf22), and sub-region IIe (PT7, orf23-orf24), and (iii) region III
(PT10, orf29). No RT-PCR amplicons were detected at sub-region IIf (PT8) (orf25 and orf26)
and region III (PT9) (orf27-orf28). The following regions were found to be co-transcribed: (i)
sub-regions IIa and IIb (PT2-3, orf3-orf11); (ii) sub-regions IIb and IIc (PT3-4, orf10-orf14);
(iii) sub-region IIc (PT4-5, orf12-orf16), and (iv) sub-regions IId and IIe (PT6-7, orf17-orf24).
No co-transcriptional unit was found between sub-regions IIc and IId (PT5-6, orf15-orf22).
Purification and mass spectrometry of the non-anti-pneumococcal
bacteriocins secreted by S. infantarius LP90
Bacteriocin purification based on cation-exchange and RP-FPLC chromatographies
(procedure P1). RP-FPLC resulted in one well-separated absorbance peak eluting at 30% (v/
Fig 3. Physical map showing the genetic organization of a 19,539-bp multibacteriocinogenic cluster present in S. infantarius LP90. Putative ORFs are represented
by wide arrows indicating the direction of transcription. Small flags and loop symbols indicate putative promoters and rho-independent transcription termination
regions, respectively. Horizontal arrows below the genes refer to the transcriptional and co-transcriptional units as determined by RT-PCR. Dashed arrows refer to
genes not transcribed or not co-transcribed.
https://doi.org/10.1371/journal.pone.0229417.g003
PLOS ONE | https://doi.org/10.1371/journal.pone.0229417 March 5, 2020
11 / 19
PLOS ONE Biochemical, genetic and transcriptional characterization of multibacteriocin production by S.infantarius LP90
Table 3. Hypothetical bacteriocins encoded by S. infantarius LP90.
Gene
ORF
Name
Name
Length
orf3
orf4
orf5
orf6
orf7
orf8
orf12
orf13
orf15
orf17
orf19
orf20
orf23
infC2
infC1
infB2
infB1
infA2
infA1
infD1
infD2
infE
infF
infG1
infG2
infH
InfC2
InfC1
InfB2
InfB1
InfA2
InfA1
InfD1
InfD2
InfE
InfF
InfG1
InfG2
InfH
(aa)
62
42
52
43
29
30
54
43
49
54
43
42
53
https://doi.org/10.1371/journal.pone.0229417.t003
Size
(Da)
5.680
3.702
4.681
4.009
3.000
3.357
4.988
3.902
5.162
5.290
4.033
3.764
5.439
Protein
Amino acid sequence
YSAIDCRNAMLTGIASGIIAGGTGAGLATLGVGGLAGAFVGAHIGAIGGGLTCVGGMIGNKF
KGNLGSAIGGCIGGVILAAATGPLTGGAAALVCVGSGLSAYL
YSAGACLTDIGLGMVGGGAAGVAGGVAGIALGVALGQIGGNVRCITSYVGGK
VNWGVAAGYCVAGALIGAAGGAISAGVGCLVSGLQEVSDGLFK
pI
8.05
8.05
8.03
4.37
10.40
DNITKGAYKLGKATRQAAKAIGKAVGKLF
8.76
8.05
8.03
8.79
9.70
DDLYKLVNQGFGAICSFIDGFNGKKYRGKC
NKCVNAIFGGALTGAGSGFVGGMATLGVTSIPGAFVGAHFGAIAGGLYCVGASL
GKVGAAVGGCLGGMLLAWAGGPVTATGYAVVCGTAGAAAAYFN
KTIYYGNGLYCNANKCWVNWSQTATTIANNSVINGLTGGNAGWHSGGIA
GCNWKGAAATVAVGAVGGAIKGAVTTYSWQGAALKAVGYGIKAGVAYGVTCRWT
10.10
KSKEGRNMGCILGTAGMAGAGFLVAGPAGAAALGGATALRVCR
8.90
8.90
KVSPTCAALVAASIYGGLAVAGPAGVGLAMAVGGAAAGSFCR
GMEWIGDVLGAIGNAAHPVNPRQVVDQLNGKYPRRGPVRSCPPGGTGGTPNAC
v) 2-propanol in aqueous TFA (0.1%, v/v) (results not shown) showing antimicrobial activity
against P. damnosus CECT4797. MALDI-TOF MS of this fraction revealed three major peaks
of 4,714; 3,354 and 2,359 Da (Fig 4).
Bacteriocin purification based on anion-exchange and hydrophobic interaction batch
chromatographies followed by RP-FPLC (procedure P2). RP-FPLC resulted in three peaks
(fractions A, B and C) eluting at 34, 36 and 38% (v/v) 2-propanol in aqueous TFA (0.1%, v/v),
respectively (results not shown), showing antimicrobial activity against P. damnosus
CECT4797. MALDI-TOF MS revealed predominant peaks of 4,718; 3,968 and 4,067 Da in
fractions A (Fig 5A), fraction B (Fig 5B) and fraction C (Fig 5C), respectively.
Fig 4. Mass spectrometry analysis of the antimicrobial peptides obtained from S. infantarius LP90 using the purification procedure P1.
https://doi.org/10.1371/journal.pone.0229417.g004
PLOS ONE | https://doi.org/10.1371/journal.pone.0229417 March 5, 2020
12 / 19
PLOS ONE Biochemical, genetic and transcriptional characterization of multibacteriocin production by S.infantarius LP90
PLOS ONE | https://doi.org/10.1371/journal.pone.0229417 March 5, 2020
13 / 19
PLOS ONE Biochemical, genetic and transcriptional characterization of multibacteriocin production by S.infantarius LP90
Fig 5. Mass spectrometry analysis of the antimicrobial peptides obtained from S. infantarius LP90 using the
purification procedure P2. a. Molecular mass of the peak eluting at 34% (v/v) 2-propanol in aqueous TFA (0.1%, v/v).
b. Molecular mass of the peak eluting at 36% (v/v) 2-propanol in aqueous TFA (0.1%, v/v). c. Molecular mass of the
peak eluting at 38% (v/v) 2-propanol in aqueous TFA (0.1%, v/v). The predominant peak is pointed out with an arrow.
https://doi.org/10.1371/journal.pone.0229417.g005
Partial bacteriocin purification based on cation-exchange and hydrophobic-interaction
batch chromatographies (procedure P3). MS-analysis of a crude fraction obtained after gel
filtration, cation-exchange and hydrophobic-interaction batch chromatographies of the pro-
teins present in a S. infantarius LP90 CFS resulted in several peaks with molecular masses
included in the size range of bacteriocins (3.5–5.5 kDa), namely, 3,654; 3,709; 3,890; 4032;
4,718 and 5,439 Da (Fig 6A), and 3,963; 4,716 and 5,436 Da (Fig 6B).
Discussion
Treatment of S. pneumoniae infections represents a challenge because of the emergence of
multidrug-resistant pneumococci [3]. LAB bacteriocins might constitute an alternative or
complementary treatment to antibiotics [4]. From 37 bacteriocinogenic LAB isolated from
food and other sources, S. infantarius subsp. infantarius LP90, obtained from Venezuelan
water buffalo milk [19], displayed the broadest and strongest anti-pneumococcal spectrum. S.
infantarius belongs to the Streptococcus bovis-Streptococcus equinus group, comprising com-
mensal non-β haemolytic inhabitants of the gastrointestinal tract of animals and humans [20].
This species has been isolated from dairy products [21] where it seems to fulfil the function of
S. thermophilus [22], frozen peas, faeces of infants and clinical samples from patients suffering
from endocarditis [23]. S. infantarius LP90 does not encode the glucosyltransferase enzyme,
Fig 6. Mass spectrometry analysis of the proteins obtained from S. infantarius LP90 using the purification procedure P3. a. Molecular mass of the crude sample
obtained after hydrophobic-interaction batch chromatography. b. Molecular mass of the precipitated sample (2-D Clean-Up Kit, Sigma-Aldrich) obtained after
hydrophobic-interaction batch chromatography.
https://doi.org/10.1371/journal.pone.0229417.g006
PLOS ONE | https://doi.org/10.1371/journal.pone.0229417 March 5, 2020
14 / 19
PLOS ONE Biochemical, genetic and transcriptional characterization of multibacteriocin production by S.infantarius LP90
which might constitute a virulence factor for humans and animals [21], or any of the previ-
ously described streptococcal and enterococcal virulence factors; moreover, it does not pro-
duce any of the tested biogenic amines, nor exhibits β-glucuronidase, mucinolytic or bile salt
deconjugation activities. S. infantarius LP90 shows an antibiotic susceptibility profile similar
to S. thermophilus, a species used as dairy starter culture [24]. Therefore, this strain can be con-
sidered as safe.
Analysis of the 19,539-bp cluster from S. infantarius LP90 reveals that this is a multibacter-
iocinogenic producer whose bacteriocins would share the same secretion and regulation sys-
tem. Region II contains 13 genes (orf3-orf8, orf12-orf13, orf15, orf17, orf19-orf20 and orf23)
encoding bacteriocins (Table 2). These peptides have a relatively low size (<6 kDa) and a high
pI (Table 3), which are characteristic of class I and II bacteriocins, and contain N-terminal
extensions of the GG-type of 15–30 amino acid residues, which are typical of bacteriocins
secreted by dedicated ABC transporters [25]. Ten out of these 13 bacteriocin structural genes
might act synergistically (subclass IIb bacteriocins), as suggested by their tandem organization
and high homology to other two-peptide bacteriocins, thus constituting the two-peptide bacte-
riocins InfC, InfB, InfA, InfD and InfG. On the other hand, InfE is a subclass IIa bacteriocin,
while InfF and InfH are subclass IIe bacteriocins. Several loci displaying characteristics of class
II bacteriocin genes have been identified in the genome of different Streptococcus spp. [26].
Moreover, the presence of a high number of bacteriocin genes, although lower than that found
in this study, seems to be common in streptococci [26]. Region II also contains 4 genes display-
ing homology to bacteriocin immunity proteins, and 1 gene showing homology to CAAX
amino-terminal proteases (Abi proteins), which have been shown to be involved in self-immu-
nity against bacteriocins [27]. RT-PCR showed that bacteriocin/IF transporter genes (region I)
and all the bacteriocin structural and immunity genes (region II), except orf25 and orf26, are
transcribed. Amongst the bacteriocin regulation genes (region III), only orf29, which based on
the sequence similarity analysis likely encodes an IF, was transcribed.
The 3,963 and 3,354 Da peptides secreted by S. infantarius LP90 could be two independent
peptides or, more likely, two InfA1 variants. InfA1 (3,963 Da) shows anti-pneumococcal activ-
ity but its molecular mass does not match with the predicted one (3,357 Da), while InfA1 with
the predicted molecular mass (3,354 Da, Fig 4) does not show inhibitory activity against S.
pneumoniae. From these observations, three hypotheses are proposed: (i) InfA1 (3,963 Da) is
somehow modified or bounded to an unknown molecule which confers anti-pneumococcal
activity; (ii) the amount of InfA1 (3,357 Da) produced by S. infantarius LP90 is not high
enough to inhibit S. pneumoniae, while the amount of InfA1 (3,963 Da) would be higher and
thus it would permit to detect its anti-pneumococcal activity (in this case, the hypothetical
modification of InfA1 would not be positively involved in the anti-pneumococcal activity), and
(iii) considering that InfA1 and InfA2 seem to constitute a two-peptide bacteriocin, InfA2
could be present in the same fraction as InfA1 (3,963 Da) in an amount necessary to confer
anti-pneumococcal activity but not sufficient to be identified by MALDI-TOF MS. Similarly to
InfA1 (3,963Da), InfA2 could have a modified molecular mass which would hinder its identifi-
cation in this fraction. The use of different purification procedures allowed demonstration of
the production of five more of the 13 putative bacteriocins encoded by S. infantarius LP90.
MS-analyses showed peaks of 4,004 (Fig 5C); 3,709; 4,031; and 5,439 Da (Fig 6A), which
approximately matched the predicted molecular sizes of InfB1, InfC1, InfG1,and InfH
(Table 3). However, in the case of InfG1 two additional peaks with modified molecular masses
(4,067 and 4,051 Da; Fig 5C) were also detected. Considering that InfG1 contains two methio-
nine residues, the peak of 4,067 Da could be the result of the conversion of one Met to MetSO2
(addition of 32 Da) or the conversion of the two Met to methionine sulfoxide (MetSO) (addi-
tion of 16 Da each one). Likewise, the lower intensity peak with a molecular mass of 4,051 Da
PLOS ONE | https://doi.org/10.1371/journal.pone.0229417 March 5, 2020
15 / 19
PLOS ONE Biochemical, genetic and transcriptional characterization of multibacteriocin production by S.infantarius LP90
could also correspond to InfG1 due to the conversion of one Met to MetSO (addition of 16
Da). Moreover, peptides that could correspond to InfB2 were also detected (4,714 Da in Fig 4,
and 4,718 Da in Figs 5A and 6A). The molecular mass of these purified peptides is approxi-
mately 32 Da higher than that of the predicted one, which might be due to a spontaneous mod-
ification of the methionine residue to methionine disulfide (MetSO2) (addition of 32 Da).
Considering these results, we can not discard the possibility that other infantaricins be pro-
duced but not detected, especially considering that all their structural genes are transcribed. In
this respect, some of the bacteriocins produced by S. infantarius LP90 could have undergone
modifications which would preclude their identification by MALDI-TOF MS.
Conclusions
S. infantarius LP90, a safe dairy strain exerting anti-pneumococcal activity, harbours a
19,539-bp multibacteriocinogenic cluster consisting of 29 ORFs including the genes encoding
8 new class II-bacteriocins (InfA [InfA1-InfA2], InfB [InfB1-InfB2], InfC [InfC1-InfC2], InfD
[InfD1-InfD2], InfE, InfF, InfG [InfG1-InfG2] and InfH), as well as the proteins involved in
their secretion, immunity and transcriptional regulation. The anti-pneumococcal activity of S.
infantarius LP90 is due, at least, to a modified InfA1 (3,963Da). In addition to InfA1, biologi-
cally active InfB1, InfB2, InfC1, InfG1 and InfH, which lack anti-pneumococcal activity, are
also secreted by this strain. InfA1 (3,963Da) displays a broad anti-pneumococcal spectrum and
other interesting characteristics and could be employed as an alternative or complementary
strategy to antibiotics for the control of human infections caused by S. pneumoniae.
Supporting information
S1 Fig. Phenotypic detection of virulence factors in S. infantarius LP90. (A) β-hemolysin,
(B) gelatinase and (C) protease activity; (1) and (2) S. infantarius LP90, (3) E. faecalis P4, and
(4) E. faecalis P36 used as positive controls.
(PPT)
S2 Fig. Biogenic amine production in S. infantarius LP90. (A) Tyrosine, (B) histidine, and
(B) ornithine decarboxylase activities. (1) S. infantarius LP90, (2) Lb. brevis CECT4121 used as
positive control for tyrosine descarboxylase activity, and (3) Lactobacillus sp. ATTCC30a used
as positive control for histidine and ornithine descarboxylase activities.
(PPT)
S3 Fig. Mucine degradation (A) and bile salt deconjugation (B) abilities in S. infantarius
LP90. (1) and (2) S. infantarius LP90, and (3) fresh fecal slurry from a healthy adult horse as
positive control.
(PPT)
S4 Fig. RP-FPLC chromatography showing the peptides obtained from a S. infantarius
LP90 CFS subjected to purification procedure P1. Fractions 26 and 28 eluted at 30 and 32%
(v/v) 2-propanol in aqueous TFA (0.1%, v/v), respectively. Red and black dashed lines indicate
antimicrobial activity against S. pneumoniae 66405 and P. damnosus CECT4797, respectively.
(PPT)
S5 Fig. RP-FPLC chromatography showing the peptides obtained from a S. infantarius
LP90 CFS subjected to purification procedure P2. Fractions 15, 17, and 18 eluted at 34, 36,
and 38% (v/v) 2-propanol in aqueous TFA (0.1%, v/v), respectively. Red and black dashed
lines indicate antimicrobial activity against S. pneumoniae 66405 and P. damnosus CECT4797,
PLOS ONE | https://doi.org/10.1371/journal.pone.0229417 March 5, 2020
16 / 19
PLOS ONE Biochemical, genetic and transcriptional characterization of multibacteriocin production by S.infantarius LP90
respectively.
(PPT)
S1 Table. Bacteriocinogenic LAB and indicator microorganisms used in this study.
(XLS)
S2 Table. Antimicrobial activity of bacteriocinogenic LAB against S. pneumoniae clinical
strains.
(XLS)
S3 Table. Primers and PCR conditions for detection of virulence factor genes.
(XLS)
S4 Table. Primers and PCR conditions for detection of bacteriocin structural genes.
(XLS)
Acknowledgments
The authors express their gratitude to Dr. T. J. Eaton and Dr. M. Gasson, Institute of Food
Research, Norwich, United Kingdom, for supplying some strains used as PCR positive
controls.
Author Contributions
Conceptualization: Dzung B. Diep, Rosa del Campo, Ingolf F. Nes, Pablo E. Herna´ndez, Car-
men Herranz, Luis M. Cintas.
Data curation: Cristina Campanero, Rosa del Campo.
Formal analysis: Ingolf F. Nes, Pablo E. Herna´ndez, Carmen Herranz, Luis M. Cintas.
Funding acquisition: Pablo E. Herna´ndez, Luis M. Cintas.
Investigation: Cristina Campanero, Estefanı´a Muñoz-Atienza, Dzung B. Diep, Javier Feito,
Sara Arbulu, Rosa del Campo.
Methodology: Cristina Campanero, Estefanı´a Muñoz-Atienza, Dzung B. Diep, Javier Feito,
Sara Arbulu, Rosa del Campo.
Project administration: Pablo E. Herna´ndez, Luis M. Cintas.
Resources: Luis M. Cintas.
Supervision: Pablo E. Herna´ndez, Carmen Herranz, Luis M. Cintas.
Writing – original draft: Cristina Campanero.
Writing – review & editing: Cristina Campanero, Carmen Herranz, Luis M. Cintas.
References
1. Braido F, Bellotti M, De Maria A, Cazzola M, Canonica GW. The role of pneumococcal vaccine. Pulm
Pharmacol Ther. 2008; 21: 608–615. https://doi.org/10.1016/j.pupt.2008.04.001 PMID: 18502670
2. Rodgers GL, Arguedas A, Cohen R, Dagan R. Global serotype distribution among Streptococcus pneu-
moniae isolates causing otitis media in children: potential implications for pneumococcal conjugate vac-
cines. Vaccine. 2009; 27: 3802–3810. https://doi.org/10.1016/j.vaccine.2009.04.021 PMID: 19446378
3.
Toda H, Satoh K, Komatsu M, Fukuda S, Nakamura T, Jikimoto T, et al. Laboratory surveillance of anti-
microbial resistance and multidrug resistance among Streptococcus pneumoniae isolated in the Kinki
region of Japan, 2001–2015. J Infect Chemother. 2018; 24: 171–176. https://doi.org/10.1016/j.jiac.
2017.12.010 PMID: 29361416
PLOS ONE | https://doi.org/10.1371/journal.pone.0229417 March 5, 2020
17 / 19
PLOS ONE Biochemical, genetic and transcriptional characterization of multibacteriocin production by S.infantarius LP90
4. Cotter PD, Ross RP, Hill C. Bacteriocins- a viable alternative to antibiotics? Nat Rev Microbiol. 2013;
11: 95–105. https://doi.org/10.1038/nrmicro2937 PMID: 23268227
5. Ahmad V, Khan MS, Jamal QMS, Alzohairy MA, Al Karaawi MA, Siddiqui MU. Antimicrobial potential of
bacteriocins: in therapy, agriculture and food preservation. Int J Antimicrob Agents. 2017; 49: 1–11.
https://doi.org/10.1016/j.ijantimicag.2016.08.016 PMID: 27773497
6. Cintas LM, Rodrı´guez JM, Ferna´ ndez MF, Sletten K, Nes IF, Herna´ndez PE, et al. Isolation and charac-
terization of pediocin L50, a new bacteriocin from Pediococcus acidilactici with a broad inhibitory spec-
trum. Appl Environ Microbiol. 1995; 61: 2643–2648. PMID: 7618877
7. Cintas LM, Casaus P, Herranz C, Håvarstein LS, Holo H, Herna´ ndez PE, et al. Biochemical and genetic
evidence that Enterococcus faecium L50 produces enterocins L50A and L50B, the sec-dependent
enterocin P, and a novel bacteriocin secreted without an N-terminal extension termed enterocin Q. J
Bacteriol. 2000; 182: 6806–6814. https://doi.org/10.1128/jb.182.23.6806-6814.2000 PMID: 11073927
8. Poyart C, Quesne G, Coulon S, Berche P, Trieu-Cuot P. Identification of streptococci to species level by
sequencing the gene encoding the manganese-dependent superoxide dismutase. J Clin Microbiol.
1998; 36: 41–47. PMID: 9431917
9. Muñoz-Atienza E, Go´ mez-Sala B, Arau´jo C, Campanero C, del Campo R, Herna´ndez PE, et al. Antimi-
crobial activity, antibiotic susceptibility and virulence factors of Lactic Acid Bacteria of aquatic origin
intended for use as probiotics in aquaculture. BMC Microbiol. 2013; 13:15. https://doi.org/10.1186/
1471-2180-13-15 PMID: 23347637
10. European Food Safety Authority. Guidance on the assessment of bacterial susceptibility to antimicrobial
of human and veterinary importance. EFSA J. 2012; 10: 2740.
11. Eaton T, Gasson M. Molecular screening of Enterococcus virulence determinants and potential for
genetic exchange between food and medical isolates. Appl Environ Microbiol. 2001; 67: 1628–1635.
https://doi.org/10.1128/AEM.67.4.1628-1635.2001 PMID: 11282615
12. Bover-Cid S, Holzapfel WH. Improved screening procedure for biogenic amine production by lactic acid
bacteria. Int J Food Microbiol. 1999; 53: 33–41. https://doi.org/10.1016/s0168-1605(99)00152-x PMID:
10598112
13. Muñoz-Atienza E, Landeta G, De las Rivas B, Go´mez-Sala B, Muñoz R, Herna´ndez PE, et al. Pheno-
typic and genetic evaluations of biogenic amine production by lactic acid bacteria isolated from fish and
fish products. Int J Food Microbiol. 2011; 146: 212–216. https://doi.org/10.1016/j.ijfoodmicro.2011.02.
024 PMID: 21411165
14.
Le Jeune C, Lonvaud-Funel A, ten Brink B, Hofstra H, van der Vossen JM. Development of a detection
system for histidine decarboxylating lactic acid bacteria based on DNA probes, PCR and activity test. J
Appl Bacteriol. 1995; 78: 316–326. https://doi.org/10.1111/j.1365-2672.1995.tb05032.x PMID:
7730207
15. Coton M, Coton E, Lucas P, Lonvaud A. Identification of the gene encoding a putative tyrosine decar-
boxylase of Carnobacterium divergens 508. Development of molecular tools for the detection of tyra-
mine-producing bacteria. Food Microbiol. 2004; 21: 125–130.
16. Marcobal A, de la Rivas B, Moreno-Arribas MV, Muñoz R. Multiplex PCR method for the simultaneous
detection of histamine-, tyramine-, and putrescine-producing lactic acid bacteria in foods. J Food Prot.
2005; 68: 874–878. https://doi.org/10.4315/0362-028x-68.4.874 PMID: 15830688
17.
Zhou JS, Gopal PK, Gill HS. Potential probiotic lactic acid bacteria Lactobacillus rhamnosus (HN001),
Lactobacillus acidophilus (HN017) and Bifidobacterium lactis (HN019) do not degrade gastric mucin in
vitro. Int J Food Microbiol. 2001; 63: 81–90. https://doi.org/10.1016/s0168-1605(00)00398-6 PMID:
11205957
18. Noriega L, Cuevas I, Margolles A, de los Reyes-Gavila´n CG. Deconjugation and bile salts hydrolase
activity by Bifidobacterium strains with acquired resistance to bile. Int Dairy J. 2006; 16: 850–855.
https://doi.org/10.1016/j.idairyj.2005.09.008
19. Citti R. Aislamiento e identificacio´ n de bacterias la´ cticas bacteriocinoge´ nicas de leches y quesos de
bu´ fala de Venezuela: actividad antimicrobiana y caracterizacio´ n bioquı´mica y gene´tica de sus bacterio-
cinas. 2005. Ph.D. thesis. Universidad Complutense de Madrid, Spain.
20. Schlegel L, Grimont F, Ageron E, Grimont PAD, Bouvet A. Reappraisal of the taxonomy of the Strepto-
coccus bovis/Streptococcus equinus complex and related species: description of Streptococcus galloly-
ticus subsp. gallolyticus subsp. nov., S. gallolyticus subsp. macedonicus subsp. nov. and S. gallolyticus
subsp. pasteurianus subsp. nov. Int J Syst Evol Microbiol. 2003; 53: 631–645. https://doi.org/10.1099/
ijs.0.02361-0 PMID: 12807180
21. Abdelgadir W, Nielsen DS, Hamad S, Jakobsen M. A traditional Sudanese fermented camel’s milk prod-
uct, Gariss, as a habitat of Streptococcus infantarius subsp. infantarius. Int J Food Microbiol. 2008;
127: 215–219. https://doi.org/10.1016/j.ijfoodmicro.2008.07.008 PMID: 18774196
PLOS ONE | https://doi.org/10.1371/journal.pone.0229417 March 5, 2020
18 / 19
PLOS ONE Biochemical, genetic and transcriptional characterization of multibacteriocin production by S.infantarius LP90
22.
Jans C, Follador R, Lacroix C, Meile L, Stevens MJA. Complete genome sequence of the African dairy
isolate Streptococcus infantarius subsp. infantarius strain CJ18. J Bacteriol. 2012; 194: 2105–2106.
https://doi.org/10.1128/JB.00160-12 PMID: 22461547
23. Hoshino T, Fujiwara T, Kilian M. Use of phylogenetic and phenotypic analyses to identify nonhemolytic
streptococci isolated from bacteremic patients. J Clin Microbiol. 2005; 43: 6073–6085. https://doi.org/
10.1128/JCM.43.12.6073-6085.2005 PMID: 16333101
24. Delorme C. Safety assessment of dairy microorganisms: Streptococcus thermophilus. Int J Food Micro-
biol. 2008; 126: 274–277. https://doi.org/10.1016/j.ijfoodmicro.2007.08.014 PMID: 17822794
25. Skaugen M, Cintas LM, Nes IF. Genetics of bacteriocin production in lactic acid bacteria. In: Wood BJB,
Warner PJ, editors. Genetics of lactic acid bacteria. New York: Kluwer Academic/Plenum Publishers;
2003. pp. 225–260.
26.
Fontaine L, Boutry C, Gue´don E, Guillot A, Ibrahim M, Grossiord B, et al. Quorum-sensing regulation of
the production of Blp bacteriocins in Streptococcus thermophilus. J Bacteriol. 2007; 189: 7195–7205.
https://doi.org/10.1128/JB.00966-07 PMID: 17693498
27. Kjos M, Snipen L, Salehian Z, Nes IF, Diep DB. The abi proteins and their involvement in bacteriocin
self-immunity. J Bacteriol. 2010; 192: 2068–2076. https://doi.org/10.1128/JB.01553-09 PMID:
20154137
PLOS ONE | https://doi.org/10.1371/journal.pone.0229417 March 5, 2020
19 / 19
PLOS ONE |
10.1371_journal.pone.0232695 | RESEARCH ARTICLE
Effects of miR-103 by negatively regulating
SATB2 on proliferation and osteogenic
differentiation of human bone marrow
mesenchymal stem cells
Hao Lv, Huashan Yang, Yuanrui WangID*
Department of Trauma Center, Jinan Central Hospital Affiliated to Shandong First Medical University, Jinan,
Shandong Province, P.R. China
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
* wangyrjinan@sina.com
Abstract
Background
OPEN ACCESS
Citation: Lv H, Yang H, Wang Y (2020) Effects of
miR-103 by negatively regulating SATB2 on
proliferation and osteogenic differentiation of
human bone marrow mesenchymal stem cells.
PLoS ONE 15(5): e0232695. https://doi.org/
10.1371/journal.pone.0232695
Editor: Gianpaolo Papaccio, Università degli Studi
della Campania, ITALY
Received: December 30, 2019
Accepted: April 19, 2020
Published: May 7, 2020
Peer Review History: PLOS recognizes the
benefits of transparency in the peer review
process; therefore, we enable the publication of
all of the content of peer review and author
responses alongside final, published articles. The
editorial history of this article is available here:
https://doi.org/10.1371/journal.pone.0232695
Copyright: © 2020 Lv et al. This is an open access
article distributed under the terms of the Creative
Commons Attribution License, which permits
unrestricted use, distribution, and reproduction in
any medium, provided the original author and
source are credited.
Data Availability Statement: All relevant data are
within the manuscript and its Supporting
Information files.
The proliferation and osteogenic differentiation of human bone marrow mesenchymal stem
cells (HBMScs) are modulated by a variety of microRNAs (miRNAs). SATB homeobox 2
(SATB2) is a critical transcription factor that contributes to maintain the balance of bone
metabolism. However, it remains unclear how the regulatory relationship between miR-103
and SATB2 on HBMScs proliferation and osteogenic differentiation.
Methods
HBMScs were obtained from Cyagen Biosciences and successful induced osteogenic dif-
ferentiation. The proliferation abilities of HBMScs after treatment with agomiR-103 and
antagomiR-103 were assessed using a cell counting Kit-8 (CCK-8) assay, and osteogenic
differentiation was determined using alizarin red S staining and alkaline phosphatase (ALP)
activity assay. The expression levels of miR-103, SATB2, and associated osteogenic differ-
entiation biomarkers, including RUNX family transcription factor 2 (RUNX2), bone gamma-
carboxyglutamate protein (BGLAP), and secreted phosphoprotein 1 (SPP1), were evalu-
ated using real-time qPCR and Western blot. The regulatory sites of miR-103 on SATB2
were predicted using bioinformatics software and validated using a dual luciferase reporter
assay. The underlying mechanism of miR-103 on SATB2-medicated HBMScs proliferation
and osteogenic differentiation were confirmed by co-transfection of antagomiR-103 and
SATB2 siRNA.
Results
The expression of miR-103 in HBMScs after induction of osteogenic differentiation was
reduced in a time-dependent way. Overexpression of miR-103 by transfection of agomiR-
103 suppressed HBMScs proliferation and osteogenic differentiation, while silencing of
miR-103 by antagomiR-103 abolished these inhibitory effects. Consistently, RUNX2,
BGLAP and SPP1 mRNA and protein expression were decreased in agomiR-103 treated
PLOS ONE | https://doi.org/10.1371/journal.pone.0232695 May 7, 2020
1 / 16
PLOS ONE Role and mechanism of miR-103 in mesenchymal stem cells osteogenic differentiation
Funding: The authors received no specific funding
for this work.
Competing interests: The authors have declared
that no competing interests exist.
Abbreviations: HBMScs, human bone marrow
mesenchymal stem cells; miRNAs, microRNAs;
SATB2, SATB homeobox 2; CCK, 8cell counting
Kit-8; RUNX2, RUNX family transcription factor 2;
ALP, alkaline phosphatase; BGLAP, bone gamma-
carboxyglutamate protein; SPP1, secreted
phosphoprotein 1; UTRs, untranslated regions.
HBMScs compared with those in agomiR-NC group. Meanwhile, antagomiR-103 upregu-
lated the mRNA and protein expression levels of RUNX2, BGLAP and SPP1 in HBMScs.
Further studies revealed that SATB2 was a direct target gene of miR-103. BMSCs trans-
fected with agomiR-103 exhibited significantly downregulated protein expression level of
SATB2, whereas knockdown of miR-103 promoted it. Additionally, rescue assays confirmed
that silencing of SATB2 partially reversed the effects of antagomiR-103 induced HBMScs
proliferation and osteogenic differentiation.
Conclusions
The present results suggested that miR-103 negatively regulates SATB2 to serve an inhibi-
tory role in the proliferation and osteogenic differentiation of HBMScs, which sheds light
upon a potential therapeutic target for treating bone-related diseases.
Introduction
microRNAs (miRNAs) are a family of highly conserved, endogenously expressed, single-
stranded small non-coding RNAs of ~22 nucleotides in length [1]. miRNAs act as critical regu-
lators of gene expression through binding with the 3’-untranslated regions (UTRs) of associ-
ated mRNAs [2]. Numerous previous studies demonstrated that miRNAs are involved in a
variety of cellular biological behaviors, such as proliferation, migration, autophagy, and differ-
entiation, and aberrant expression of miRNAs can lead to the development of multiple human
diseases [3]. Notably, recent studies have indicated that miRNAs are related to the dysfunction
of bone metabolism [4]. Osteoporosis is a progressive systemic skeletal disease in aged people
characterized by low mineral density and microarchitecture deterioration of bone tissues [5].
According to previous epidemiological statistics, osteoporosis significantly increases the risk
of fractures, and causes a serious social burden worldwide [6]. Osteogenic differentiation has
been regarded as a critical issue in fracture therapy of osteoporosis; however, its mechanisms
remain largely unclear.
Stem cells therapy provides a promising novel approach for repairing defective tissues or
organs through the transplantation of cells. Dental mesenchymal stem cells have gained con-
siderable attention as an attractive source for maxillofacial regenerative therapy [7,8]. Human
bone marrow mesenchymal stem cells (HBMScs) are pluripotent stem cells that possess multi-
ple differentiation potential, including chondrocytes, osteoblasts, fibroblasts, and adipocytes.
Increasing data suggested that the HBMScs osteogenic differentiation is modulated by hor-
mones, medicines, as well as growth factors [9]. Of note, several recent studies have shown
that abnormal expression of miRNAs is relevant to HBMScs osteogenic differentiation [10].
For instance, Li et al [11] found that miR-21 facilitates osteogenesis of mouse BMScs by rgulat-
ing Smad7-Smad1/5/8-Runx2 pathway. Zhang et al [12] showed that miR-664a-5p promotes
osteogenic differentiation of HBMScs by directly downregulating high-mobility group A2
expression. Xiao et al [13] reported that miR-483-3p regulates osteogenic differentiation of
mouse BMScs by regulation of signal transducer and activator of transcription 1. Additionally,
Li et al [14] demonstrated that miR-92b-5p modulates melatonin-mediated osteogenic differ-
entiation of mouse BMScs by targeting intracellular adhesion molecule-1. Therefore, identifi-
cation of the potential mechanisms underlying osteogenic differentiation of HBMScs is a
meaningful process that developed novel therapeutic targets for osteoporosis treatment.
PLOS ONE | https://doi.org/10.1371/journal.pone.0232695 May 7, 2020
2 / 16
PLOS ONE Role and mechanism of miR-103 in mesenchymal stem cells osteogenic differentiation
miR-103 is one of the members of the miR-15/107 family [15]. Several previous reports
indicated that miR-103 is involved in various human diseases, including malignancies [16],
nervous system disease [17], as well as fatty liver disease [18]. Chen et al [19] found that miR-
103 expression was markedly downregulated in serum samples of osteoporotic patients.
Valassi et al [20] reported that circulating miR-103 is associated with bone parameters in
patients with controlled acromegaly. Additionally, evidence from microarray information
showed that miR-103 is significantly disregulated in senescent BMScs [21]. SATB homeobox 2
(SATB2), as a member of AT-rich binding proteins family, is a special transcription factor that
improved transcription by binding with nuclear matrix-attachment regions. SATB2 was found
to be a critical factor of osteoblast differentiation in bone development [22]. Nonetheless, the
regulatory relationship between miR-103 and SATB2 is still unclear. Thus, the present study
aimed to explore the role of miR-103 on the proliferation and osteogenic differentiation of
HBMScs using gain- and lose-of-function assays, and to investigate the underlying molecular
mechanism of miR-103 on SATB2. The findings suggested that miR-103 inhibits HBMScs pro-
liferation and osteogenic differentiation by directly targeting SATB2, indicating regulation of
miR-103 as a potential molecular therapeutic strategy for osteoporosis treatment.
Materials and methods
Cell culture and osteogenic differentiation
HBMScs were obtained from Cyagen Biosciences (HUXMA-01001, Guangzhou, China). Cells
were conventional cultured in OriCell1 HBMScs Complete Medium (HUXMA-90011, Cya-
gen Biosciences) supplemented with 10% of fetal bovine serum, 100 U/ml penicillin and
100 μg/ml streptomycin, and maintained at 37˚C in a humidified atmosphere under 5% CO2.
For osteogenic differentiation, cells were treated with OriCell1 Osteogenic Induction Differ-
entiation Medium (HUXMA-90021, Cyagen Biosciences) at 37˚C and 5% CO2 for 21 d incu-
bation, and the medium was replaced every 2 d.
Oligonucleotides synthesis and cell transfection
The agomiR-103, agomiR-NC, antagomiR-103, and antagomiR-NC were designed and syn-
thesized by GenePharma (Shanghai, China). SATB2 siRNA and corresponding siRNA NC
were acquired from Hanbio Biotechnology Co., Ltd (Shanghai, China). All oligonucleotides
were dissolved to suitable concentration in diethylpyrocarbonate-treated water. HBMScs in
logarithmic growth phase were trypsinized and seeded in 6-well plates. When HBMScs grew
to 60% cell confluence, cell were transfected with these oligonucleotides at final concentration
of 200 nM using Lipofectamine 2000 Transfection Reagent (Thermo Fisher Scientific, Carls-
bad, CA, USA) along with Opti-MEM Reduced Serum Medium (Thermo Fisher Scientific),
according to the manufacturer’s protocols. The cells were harvested 48 h after transfection for
further experiments.
Alkaline phosphatase (ALP) staining
The calcium deposition by osteoblasts was assessed using ALP staining. HBMSCs were placed
in 12-well plates and induced osteogenic differentiation for 21 d. Then HBMSCs were rinsed
three times with PBS for three times and fixed with 75% alcohol (Sigma, St Louis, CA, USA)
for 30 minutes at 37˚C. The fixed cells were soaked in BCIP/NBT solution (Yeasen, Shanghai,
China) and washed with PBS, and then were observed with an Olympus SP-500UZ Digital
Camera (Nikon, Japan).
PLOS ONE | https://doi.org/10.1371/journal.pone.0232695 May 7, 2020
3 / 16
PLOS ONE Role and mechanism of miR-103 in mesenchymal stem cells osteogenic differentiation
ALP activity assay
21 days after osteogenic induction, HBMSCs were lysed with 100 μl of cell lysate buffer
(21101ES60, Yeasen, Shanghai, China) for 5 min at 4˚C at 37˚C. Thereafter, the supernatant
was added into 40 μl of the ALP substrate reaction solution (Yeasen) for 30 min at 37˚C in the
dark according to the manufacturer’s protocols, and 150 μl of the stop buffer (Yeasen) was
added to stop the reaction. Finally, the optical density (OD) values were determined by mea-
suring the absorbance at a wavelength of 405 nm using a microplate reader (Molecular Devices
LLC, Sunnyvale, CA, USA). The total protein content ratio demonstrates the amounts of ALP
produced by differentiated HBMSCs.
Enzyme-linked immunosorbent assay (ELISA)
HBMSCs was collected after oligonucleotides treatment for 48 h. The SATB2 protein expres-
sion was calculated using a commercially available ELISA kit (abx383034, Abbexa, Cambridge,
UK), according to the manufacturer’s instructions.
Alizarin red S staining
HBMSCs were placed in 12-well plates and induced osteogenic differentiation induced osteo-
genic differentiation for 21 d. Mineralization of cells was detected using alizarin red S Staining.
Briefly, HBMSCs were trypsinized and collected, and fixed with 75% ethanol for 1 h at 37˚C.
Then the cells were incubated using 40 mM alizarin red S (HUXMA-90021, Cyagen Biosci-
ences) for 20 min at 37˚C. After washes with PBS twice to rinse needless unbound stains, the
stained matrix was photographed with an Olympus SP-500UZ Digital Camera (Nikon). Five
images were randomly selected and analyzed for quantification of staining with an Image Pro
plus 6.0 software (Media Cybernetics, Rockville, MD, USA).
Real-time qPCR
Total RNA was obtained from HBMScs using Trizol Reagent (Thermo Fisher Scientific), and
reverse transcribed into first-stranded cDNA sequences using a miRNA cDNA Synthesis Kit
(YB130911-25, Ybscience, Shanghai, China) or a PrimeScript™ RT reagent kit (DRR037A,
TaKaRa, Japan). The amplification of miR-103, SATB2, RUNX family transcription factor 2
(RUNX2), bone gamma-carboxyglutamate protein (BGLAP), and secreted phosphoprotein 1
(SPP1) was performed by qPCR using the cDNA as a template on the ABI 7900HT Real-Time
PCR System 7900 (Applied Biosystems, Carlsbad, CA, USA). The primers of the genes were
listed in Table 1. Thermocycling conditions of PCR amplifications were performed in dupli-
cate at 98˚C for 15 s, followed by 40 cycles of 95˚C for 10 s, 60˚C for 30 s, and 72˚C for 45 s,
finally 4˚C for 30 min. The relative levels of target gene expression were quantified using the 2-
ΔΔCq method [23]. small nuclear U6 RNA was used as an internal reference for miR-103
expression analysis, and glyceraldehyde-3-phosphate dehydrogenase (GAPDH) served as an
internal control for other genes.
Western blot
HBMScs were lysed using mammalian protein extraction RIPA reagent (Yeasen). Equivalent
amounts (50 μg) of cell protein lysates were electrophoresed on an 10% SDS-polyacrylamide
gel and transferred to PVDF membranes (Thermo Fisher Scientific). The membranes were
incubated with primary antibodies of RUNX (1:750, ab76956, Abcam, Cambridge, MA, USA),
BGLAP (1:1000, ab13421, Abcam), SSP1 (1:750, ab69498, Abcam), and GAPDH (1:1000,
60004-1-Ig, ProteinTech Group, Chicago, IL, USA). After incubating with HRP-associated
PLOS ONE | https://doi.org/10.1371/journal.pone.0232695 May 7, 2020
4 / 16
PLOS ONE Role and mechanism of miR-103 in mesenchymal stem cells osteogenic differentiation
Table 1. Forward and reverse primers listed for real-time qPCR.
Name
miR-103 forward
miR-103 reverse
SATB2 forward
SATB2 reverse
RUNX2 forward
RUNX2 reverse
BGLAP forward
BGLAP reverse
SPP1 forward
SPP1 reverse
U6 forward
U6 reverse
GAPDH forward
GAPDH reverse
Sequence (5’-3’)
AGCAGCATTGTACAGGGCTATGA
AAGGCGAGACGCACATTCTT
CCTGGCCCTGGGGTATTCT
GTGCATCTGTCACATAACTGAGG
TCTTAGAACAAATTCTGCCCTTT
GCTTTGGTCTTGAAATCACA
GCAGCTTGGTGCACACCTAG
GGAGCTGCTGTGACATCCAT
CTTTCACTCCAATCGTCCCTA
GCTCTCTTTGGAATGCTCAAGT
CTCGCTTCGGCAGCACA
AACGCTTCACGAATTTGCGT
ATTTGGTCGTATTGGGCG
TGGAAGATGGTGATGGGATT
https://doi.org/10.1371/journal.pone.0232695.t001
second antibodies, protein blots were observed using ECL Select Western Blotting Detection
System (GE Healthcare, Buckinghamshire, UK) and the results were quantified using Image
Pro Plus 6.0 (Media Cybernetics).
Cell counting Kit-8 (CCK-8) assay
To measure in vitro growth of HBMScs, a CCK-8 assay was performed. A total of 7 × 103 cells
were seeded into each well of 96-well plates and transfected with agomiR-103, agomiR-NC,
antagomiR-103, and antagomiR-NC. On the 0, 24, 48, and 72 h of the measurement, each well
was replaced with 100 μl of fresh DMEM medium containing 10 μl of the CCK-8 reagent
(Dojindo Laboratories, Kumamoto, Japan). Plates were incubated at 37˚C in a humidified
atmosphere for 4 h, then the medium was shaken for 20 min. Absorbances were then mea-
sured with a microplate reader (Molecular Devices LLC) at a wavelength of 450 nm.
Bioinformatics analysis and reporter vector construction
Bioinformatics analysis was conducted to predict the putative targets of miR-103 using Tar-
getScan (http://www.targetscan.org) and miranda (http://www.microrna.org). The wild type
(WT) 3’-UTR of SATB2 contained the predicted miR-103 binding sites and a mutant (MUT)
3’-UTR of SATB2 were obtained from Hanbio Biotechnology Co., Ltd. The two sequences
were inserted into the psiCHECK™2 vector (Promega Corporation, Madison, WI, USA) to
produce SATB2 WT and MUT reporter vectors, respectively.
Luciferase reporter assay
HBMScs cells at 60% confluence in 6-well plates were co-transfected with 50 nM agomiR-103,
agomiR-NC, antagomiR-103, and antagomiR-NC, along with 0.75 μg SATB2 WT or MUT
reporter vectors using Lipofectamine 2000 Transfection Reagent. The Firefly and Renilla lucif-
erase activities were measured 48 h after transfection using a Dual-Luciferase Reporter Assay
System (E1910, Promega Corporation), according the manufacturer’s protocols. Renilla lucif-
erase acted as a reporter gene and Firefly luciferase as a normalized internal reference for each
individual analysis.
PLOS ONE | https://doi.org/10.1371/journal.pone.0232695 May 7, 2020
5 / 16
PLOS ONE Role and mechanism of miR-103 in mesenchymal stem cells osteogenic differentiation
Statistical analysis
SPSS 19.0 software (SPSS Inc., Chicago, IL, USA) was used for the statistical analyses. Each
experiment was conducted and repeated for three times. The quantitative data were listed as
the mean ± standard deviation. Independent Student’s t-tests and one-way ANOVA analyses
were used to compare differences between groups. Differences with p-values of less than 0.05
were considered to be statistically significant.
Results
The expression of miR-103, RUNX2, BGLAP and SPP1 in HBMScs after
induction of osteogenic differentiation
To assess the association between miR-103 and osteogenic differentiation, HBMScs were
induced osteogenic differentiation for 21 d incubation, and the expression levels of miR-103 in
HBMScs after 0, 7, 14, and 21 d induction were observed. Our results showed that the associ-
ated osteogenic differentiation biomarkers (RUNX2, BGLAP, and SPP1) mRNA expression
was significantly upregulated in HBMScs on days 7, 14, and 21 as compared with undifferenti-
ated cells on 0 day (Fig 1A–1C, p<0.05 or p<0.001), indicating osteogenic differentiation of
HBMScs is successful. Notably, the results of real-time qPCR revealed that the expression level
of miR-103 was markedly reduced in HBMScs during osteogenic differentiation in a time-
dependent way (Fig 1D, p<0.05 or p<0.001). Consistently, the protein expression of RUNX,
BGLAP and SSP1 was increased in HBMScs during osteogenic differentiation (Fig 1E and 1F,
p<0.05 or p<0.01 or p<0.001). The data suggested that decreased miR-103 expression corre-
lates with HBMScs osteogenic differentiation.
The modulation of miR-103 on the proliferation of HBMScs
To determine the expression levels of miR-103 following oligonucleotides transfection, real-
time qPCR was conducted. The results revealed that the expression of miR-103 was 102.55
times higher in HBMScs treated with agomiR-103 than in cells transfected with the ago-
miR-NC group (Fig 2A, p<0.001). Also, miR-103 expression was downregulated by 76.18% in
HBMScs transfected with antagomiR-103 compared with the antagomiR-NC group (Fig 2B,
p<0.001). Further, the effects of the miR-103 overexpression and knockdown on the prolifera-
tion of HBMScs were determined. The results indicated that the proliferation abilities of
HBMScs decreased by 14.62%, 21.70% and 34.89% at 24, 48 and 72 h posttransfection of the
agomiR-103, as compared with the agomiR-NC group (Fig 2C, p<0.05 or p<0.001). Mean-
while, in HBMScs treated with antagomiR-103 the results revealed that at 24, 48 and 72 h the
cell proliferation was increased by 11.24%, 30.15% and 29.44%, compared with cells trans-
fected with the antagomiR-NC group (Fig 2D, p<0.05 or p<0.001). These data validated that
miR-103 suppresses the proliferation of HBMScs.
The effects of miR-103 overexpression and knockdown on HBMScs
osteogenic differentiation
Furthermore, to better identify the functional effects of miR-103 overexpression and knock-
down on the osteogenic differentiation of HBMScs, the expressions of RUNX2, BGLAP, and
SPP1 were detected. As presented in Fig 3A, the mRNA expression levels of RUNX2, BGLAP,
and SPP1 were markedly reduced during osteogenic differentiation in the agomiR-103 group
compared with the agomiR-NC group (p<0.05 or p<0.001). Compared with the antago-
miR-NC group, these genes expression were significantly increased in the antagomiR-103
treated HBMScs (Fig 3B, p<0.001). Western blot showed that agomiR-103 treatment
PLOS ONE | https://doi.org/10.1371/journal.pone.0232695 May 7, 2020
6 / 16
PLOS ONE Role and mechanism of miR-103 in mesenchymal stem cells osteogenic differentiation
Fig 1. The expression of miR-103, RUNX2, BGLAP and SPP1 in HBMScs after induction of osteogenic differentiation. Real-time qPCR was performed to detect the
mRNA expression levels of RUNX2 (A), BGLAP (B) and SPP1 (C) in HBMScs after 0, 7, 14, and 21 d induction. (D) The expression of miR-103 was markedly reduced in
HBMScs during osteogenic differentiation in a time-dependent way. (E and F) Western blot analysis of the protein expression of RUNX, BGLAP and SSP1 in HBMScs
during osteogenic differentiation. Columns mean of three independent experiments, and bars SD. �p<0.05, ��p<0.01, ���p<0.001.
https://doi.org/10.1371/journal.pone.0232695.g001
decreased RUNX2, BGLAP, and SPP1 protein expression, whereas antagomiR-103 induced
these proteins expression (Fig 3C and 3D, p<0.05 or p<0.001). To corroborate the roles of
miR-103 on the osteogenic process, ALP activity and alizarin red S staining were performed.
As expected, overexpression of miR-103 decreased ALP activity and mineralized-bone matrix
formation in HBMScs (Fig 3E and 3F, p<0.001). Consistently, silencing of miR-103 exhibited
higher ALP activity and mineralized-bone matrix formation after osteogenic differentiation of
21 days (Fig 3G and 3H, p<0.001). Thus, the above data confirmed that miR-103 negatively
regulates the osteogenic differentiation of HBMScs.
The negatively regulation of miR-103 on the potential target gene SATB2
SATB2 mRNA 3’-UTR has a potential complimentary binding sites to miR-203a (Fig 4A).
Real-time qPCR and ELISA were applied to investigate the relationship between the expression
PLOS ONE | https://doi.org/10.1371/journal.pone.0232695 May 7, 2020
7 / 16
PLOS ONE Role and mechanism of miR-103 in mesenchymal stem cells osteogenic differentiation
Fig 2. miR-103 suppresses the proliferation of HBMScs. (A) Elevated expression of miR-103 was determined by
real-time qPCR when HBMScs were transfected with agomiR-103. (B) The expression of miR-103 was significantly
downregulated in BMMSCs after treated with antagomiR-103. (C) A CCK-8 assay demonstrated that overexpression
of miR-103 retarded the growth of HBMScs in vitro. (D) Knockdown of miR-103 promoted HBMScs proliferation
ability. Data are shown as the mean ± SD of three independent experiments. �p<0.05, ���p<0.001.
https://doi.org/10.1371/journal.pone.0232695.g002
levels of miR-103 and SATB2. As shown in Fig 4B and 4C, enforced miR-103 expression led to
downregulate the mRNA and protein expression levels of SATB2 in HBMScs; by contrast,
application of antagomiR-103 to the HBMScs markedly upregulated SATB2 mRNA and pro-
tein expression (p<0.01 or p<0.001). Additional, a dual luciferase reporter assay was per-
formed to validated the miR-103 binding sites on the 3’-UTR of SATB2 mRNA, and WT and
MUT SATB2 reporter plasmids were constructed according to this sites. Results showed that
in WT SATB2 reporter plasmid the relative luciferase activity significantly decreased in ago-
miR-103-transfected HBMScs (Fig 4D, p<0.001). With MUT SATB2 reporter plasmid, there
was no significant difference in relative luciferase activity between the agomiR-103 and ago-
miR-NC groups. Conversely, miR-103 knockdown substantially escalated the relative lucifer-
ase activity of the SATB2 reporter plasmid that carried the WT but not MUT 3’-UTR of
PLOS ONE | https://doi.org/10.1371/journal.pone.0232695 May 7, 2020
8 / 16
PLOS ONE Role and mechanism of miR-103 in mesenchymal stem cells osteogenic differentiation
Fig 3. miR-103 represses the osteogenic differentiation of HBMScs. (A) Real-time qPCR analysis detected mRNA expression of RUNX2, BGLAP and SPP1 in
HBMScs after treatment with agomiR-103 or agomiR-NC. (B) The RUNX2, BGLAP and SPP1 mRNA expression were significantly increased in the antagomiR-103
treated HBMScs. (C and D) Western blot analysis of the protein expression of RUNX, BGLAP and SSP1 in HBMScs by miR-103 overexpression and knockdown. (E)
The ALP activity of HBMScs was significantly suppressed when miR-103 was overexpressed. (F) The osteogenic differentiation of HBMScs transfected with agomiR-103
and agomiR-NC was observed by Alizarin red S staining. (G) Knockdown of miR-103 exhibited higher ALP activity after osteogenic differentiation of 21 days. (H)
Alizarin red S staining investigated the ability of mineralized-bone matrix formation in HBMScs treated with antagomiR-103 or antagomiR-NC. Columns mean of three
independent experiments. �p<0.05, ���p<0.001.
https://doi.org/10.1371/journal.pone.0232695.g003
PLOS ONE | https://doi.org/10.1371/journal.pone.0232695 May 7, 2020
9 / 16
PLOS ONE Role and mechanism of miR-103 in mesenchymal stem cells osteogenic differentiation
Fig 4. SATB2 is a direct target of miR-103 in HBMScs. (A) miR-103 target sites in the 3’-UTR of SATB2 mRNA. (B) Real-time qPCR and ELISA were conducted to
examine the effects of miR-103 overexpression on the mRNA and protein expression of SATB2. GAPDH was also detected as a loading control. (C) Knockdown of miR-
103 markedly upregulated SATB2 mRNA and protein expression in HBMScs. (D) Relative luciferase activity of the WT or MUT SATB2 reporter plasmids along with
agomiR-103 or agomiR-NC in HBMScs was shown. Renilla luciferase activity was normalized to Firefly luciferase activity and plotted as relative luciferase activity. (E)
Luciferase activities were measured in HBMScs following cotransfection with the WT or MUT SATB2 reporter plasmids and antagomiR-103 or antagomiR-NC. Data
are shown as the mean ± SD of three independent experiments. ��p<0.01, ���p<0.001.
https://doi.org/10.1371/journal.pone.0232695.g004
PLOS ONE | https://doi.org/10.1371/journal.pone.0232695 May 7, 2020
10 / 16
PLOS ONE Role and mechanism of miR-103 in mesenchymal stem cells osteogenic differentiation
SATB2 (Fig 4E, p<0.001). These findings meant that SATB2 is a direct target of miR-103 in
HBMScs.
The regulatory mechanism of miR-103 on HBMScs proliferation and
osteogenic differentiation by directly downregulating SATB2
As SATB2 plays a critical role in osteoblast differentiation of bone development, and SATB2 is
a direct target of miR-103, we speculated that miR-103 may perform its functions by regulating
SATB2. To verify the participation of SATB2 in the effects of miR-103 on the proliferation and
osteogenic differentiation, HBMScs with miR-103 knockdown were transfected with either the
SATB2 siRNA or corresponding siRNA NC, and SATB2 expression was detected by real-time
qPCR and ELISA. The data of results confirmed that the protein and mRNA expression levels
of SATB2 in HBMScs were significantly decreased following SATB2 siRNA transfection, and
SATB2 siRNA partly reversed the upregulation of SATB2 expression by antagomiR-103(Fig
5A and 5B, p<0.01 or p<0.001). Interesting, HBMScs co-transfected with antagomiR-103 and
SATB2 siRNA exhibited a lower cell proliferation ability than cells transfected with antago-
miR-103 and siRNA NC (Fig 5C, p<0.001). Moreover, antagomiR-103 induced HBMScs oste-
ogenic differentiation could be partly abolished by SATB2 knockdown, as evidenced by
reduction in ALP activity and mineralized matrix formation (Fig 5D and 5E, p<0.05 or
p<0.001), and reduced protein expression levels of RUNX2, BGLAP, and SPP1 (Fig 5F and
5G, p<0.01 or p<0.001). Altogether, these data demonstrated miR-103 participates in the pro-
liferation and osteogenic differentiation of HBMScs by directly targeting SATB2.
Discussion
BMScs are a commonly used progenitor cell source to study osteogenic differentiation in pro-
gressive systemic skeletal diseases because of their capacity for self-renewal and differentiation
potential. During the past years, a number of miRNAs have been determined as key regulators
of osteogenic differentiation of BMScs by modulating a series of genes expression involved in
bone development [24]. Recently, study by Wang and his colleagues [25] found that steroid-
induced avascular necrosis of the femoral head exhibits reduced osteogenic differentiation and
promotes fat differentiation, and elevates miR-103 expression. Shen and his team members
[26] reported that miR-103 expression is consistently downregulated in the forkhead tran-
scription factor C1-overexpressing MC3T3-E1 cells. Given these findings, we explored the
potential function and mechanism of miR-103 in regulating proliferation and osteogenic dif-
ferentiation. We selected HBMScs to perform our experiments, and our in vitro results sup-
ported that miR-103 negatively regulates SATB2 to serve an inhibitory role in the proliferation
and osteogenic differentiation of HBMScs. Our findings provides more insights into the func-
tion and mechanism of miRNAs in modulating the osteogenesis of HBMScs.
miR-103 is a notable miRNA that it is evolutionarily conserved and involved in regulating
cell differentiation, metabolism and immune [27]. For example, Croizier et al [28] demon-
strated that miR-103/107 signal controls developmental switch of proopiomelanocortin pro-
genitors into neuropeptide Y neurons and impacts glucose homeostasis. In another study,
Holik and his team [29] reported that n(�)-carboxymethyllysine promotes the miR-103/143
expression and enhances lipid accumulation in 3T3-L1 cells. In addition, recent study reported
that shenmai injection improves the postoperative immune function by inhibiting differentia-
tion into Treg cells via miR-103/GPER1 axis [30]. However, researches concerning miR-103
on osteogenesis are still very limited. Here, we examined the effect and mechanism of miR-103
on HBMScs osteogenic differentiation because we detected decreased levels of miR-103 in a
time-dependent way after induction of osteogenic differentiation. Similar to our findings, Yoo
PLOS ONE | https://doi.org/10.1371/journal.pone.0232695 May 7, 2020
11 / 16
PLOS ONE Role and mechanism of miR-103 in mesenchymal stem cells osteogenic differentiation
Fig 5. miR-103 participates in the proliferation and osteogenic differentiation of HBMScs by directly targeting SATB2. (A) Expression levels of SATB2 mRNA and
protein were examined after transfection of SATB2 siRNA and siRNA NC. (B) Transfection of SATB2 siRNA partly reversed the upregulation of SATB2 expression by
antagomiR-103. (C) HBMScs co-transfected with antagomiR-103 and SATB2 siRNA exhibited a lower cell proliferation ability than cells transfected with antagomiR-
103 and siRNA NC. (D) SATB2 knockdown partly abolished antagomiR-103 induced ALP activity of HBMScs. (E) SATB2 knockdown partly abolished antagomiR-103
induced mineralized matrix formation of HBMScs. (F and G) HBMScs treatment with antagomiR-103 and SATB2 siRNA reduced the protein expression levels of
RUNX2, BGLAP, and SPP1 compared cells treated with antagomiR-103 and siRNA NC. Columns mean of three independent experiments. �p<0.05, ��p<0.01,
���p<0.001.
https://doi.org/10.1371/journal.pone.0232695.g005
PLOS ONE | https://doi.org/10.1371/journal.pone.0232695 May 7, 2020
12 / 16
PLOS ONE Role and mechanism of miR-103 in mesenchymal stem cells osteogenic differentiation
et al [21] recently reported expression profiles of miRNAs in senescent BMScs. It has previ-
ously been demonstrated that upregulation expression of some osteogenic markers, including
RUNX2, BGLAP, and SPP1, and accumulated mineralization of the extracellular matrix can be
observed during the osteogenic differentiation [31]. Similarly, our study revealed that overex-
pression of miR-103 led to the decrease of ALP activity and mineralized-bone matrix forma-
tion, and reduction of RUNX2, BGLAP, and SPP1 expression, and conversely, while silencing
of miR-103 elevated these phenomenon. Additional, we found HBMScs proliferation was
arrested by miR-103, which is in agreement with previous study in the cancer field, where
miR-103 has been shown to functions as a tumor suppressor gene in non-small cell lung can-
cer by directly targeting programmed cell death 10 [32]. These data confirmed that miR-103
negatively regulates the proliferation and osteogenic differentiation of HBMScs.
To further to decipher mechanism, computational bioinformatics analysis was taken to
identify target mRNAs that that could explain the phenotypic effects inhibited by miR-103 on
HBMScs proliferation and osteogenic differentiation. One target gene of interest was the
SATB2 gene (Gene ID: 23314), which located at 2q33.1 and have 18 exons. SATB2 is a member
of the highly conserved AT-rich binding proteins family, and its protein is involved in tran-
scription regulation and chromatin remodeling [33]. The mutation or deficiency of SATB2 in
humans is associated with several severe diseases, such as craniofacial deformity, craniosynos-
tosis, and mental retardation [34]. Indeed, SATB2 is documented to be a sensitive biomarker
for colorectal carcinoma [35], pancreatic cancer [36] and osteosarcoma [37]. Recently, accu-
mulating evidences confirmed that SATB2 serves a key role in bone development and osteo-
blastic differentiation [22]. Notably, SATB2 has been found to be targeted by several miRNAs,
including miR-31 [38], miR-34a/b/c [39,40], miR-383 [41], and miR-875-5p [42]. Here, by
luciferase reporter and ELISA assay, we identified that miR-103 directly bound to the SATB2
mRNA 3’-UTR and suppressed SATB2 expression in HBMScs. Moreover, rescue assays con-
firmed that silencing of SATB2 partially rescued the effects of miR-103 knockdown induced
HBMScs proliferation and osteogenic differentiation. Our study is not without limitations.
We could not determine miR-103 expression and the correlation between miR-103 and
SATB2 in osteoporosis patients. Further studies are warranted to identify the regulatory mech-
anism of miR-103 on SATB2 in inhibiting HBMScs proliferation and osteogenic differentia-
tion in vivo.
Taken together, the present study demonstrated that miR-103 suppresses the proliferation
and osteogenic differentiation of HBMScs by directly targeting SATB2. Thus, targeting miR-
103 may be a molecular therapeutic strategy for bone-related diseases treatment.
Supporting information
S1 Fig. Western blot analysis of the protein expression of RUNX, BGLAP and SSP1 in
HBMScs.
(TIF)
Author Contributions
Conceptualization: Yuanrui Wang.
Formal analysis: Hao Lv, Huashan Yang.
Methodology: Hao Lv, Huashan Yang, Yuanrui Wang.
Supervision: Hao Lv.
Writing – original draft: Hao Lv.
PLOS ONE | https://doi.org/10.1371/journal.pone.0232695 May 7, 2020
13 / 16
PLOS ONE Role and mechanism of miR-103 in mesenchymal stem cells osteogenic differentiation
Writing – review & editing: Yuanrui Wang.
References
1. Bartel DP. MicroRNAs: genomics, biogenesis, mechanism, and function. Cell. 2004; 116(2):281–297.
https://doi.org/10.1016/s0092-8674(04)00045-5 PMID: 14744438
2. Bartel DP. MicroRNAs: target recognition and regulatory functions. Cell. 2009; 136(2):215–233. https://
doi.org/10.1016/j.cell.2009.01.002 PMID: 19167326
3. Mendell JT, Olson EN. MicroRNAs in stress signaling and human disease. Cell. 2012; 148(6):1172–
1187. https://doi.org/10.1016/j.cell.2012.02.005 PMID: 22424228
4. Yuan Y, Zhang L, Tong X, Zhang M, Zhao Y, Guo J, et al. Mechanical Stress Regulates Bone Metabo-
lism Through MicroRNAs. Journal of Cellular Physiology. 2017; 232(6):1239–1245. https://doi.org/10.
1002/jcp.25688 PMID: 27861865
5.
Tella SH, Gallagher JC. Prevention and treatment of postmenopausal osteoporosis. Journal of Steroid
Biochemistry and Molecular Biology. 2014; 142(155–170. https://doi.org/10.1016/j.jsbmb.2013.09.008
PMID: 24176761
6. Black DM, Rosen CJ. Clinical Practice. Postmenopausal Osteoporosis. New England Journal of Medi-
cine. 2016; 374(3):254–262. https://doi.org/10.1056/NEJMcp1513724 PMID: 26789873
7.
La Noce M, Mele L, Laino L, Iolascon G, Pieretti G, Papaccio G, et al. Cytoplasmic Interactions between
the Glucocorticoid Receptor and HDAC2 Regulate Osteocalcin Expression in VPA-Treated MSCs.
Cells. 2019; 8(3). https://doi.org/10.3390/cells8030217 PMID: 30841579
8. Winning L, El Karim IA, Lundy FT. A Comparative Analysis of the Osteogenic Potential of Dental Mes-
enchymal Stem Cells. Stem Cells and Development. 2019; 28(15):1050–1058. https://doi.org/10.1089/
scd.2019.0023 PMID: 31169063
9. Caplan AI. Mesenchymal Stem Cells: Time to Change the Name! Stem Cells Translational Medicine.
2017; 6(6):1445–1451. https://doi.org/10.1002/sctm.17-0051 PMID: 28452204
10.
11.
12.
Li M, Zhang YL, Huang H, Xiong Y. MicroRNA-10-5p regulates differentiation of bone marrow mesen-
chymal stem cells into cardiomyocytes by targeting TBX5. European Review for Medical and Pharma-
cological Sciences. 2019; 23(2):479–485. https://doi.org/10.26355/eurrev_201901_16859 PMID:
30720154
Li X, Guo L, Liu Y, Su Y, Xie Y, Du J, et al. MicroRNA-21 promotes osteogenesis of bone marrow mes-
enchymal stem cells via the Smad7-Smad1/5/8-Runx2 pathway. Biochemical and Biophysical
Research Communications. 2017; 493(2):928–933. https://doi.org/10.1016/j.bbrc.2017.09.119 PMID:
28943430
Zhang Y, Liu Y, Wu M, Wang H, Wu L, Xu B, et al. MicroRNA-664a-5p promotes osteogenic differentia-
tion of human bone marrow-derived mesenchymal stem cells by directly downregulating HMGA2. Bio-
chemical and Biophysical Research Communications. 2020; 521(1):9–14. https://doi.org/10.1016/j.
bbrc.2019.09.122 PMID: 31630797
13. Xiao Y, Guo Q, Jiang TJ, Yuan Y, Yang L, Wang GW, et al. miR4833p regulates osteogenic differentia-
tion of bone marrow mesenchymal stem cells by targeting STAT1. Molecular Medicine Reports. 2019;
20(5):4558–4566. https://doi.org/10.3892/mmr.2019.10700 PMID: 31702021
14.
15.
Li Y, Feng C, Gao M, Jin M, Liu T, Yuan Y, et al. MicroRNA-92b-5p modulates melatonin-mediated oste-
ogenic differentiation of bone marrow mesenchymal stem cells by targeting ICAM-1. Journal of Cellular
and Molecular Medicine. 2019; 23(9):6140–6153. https://doi.org/10.1111/jcmm.14490 PMID:
31304676
Finnerty JR, Wang WX, Hebert SS, Wilfred BR, Mao G, Nelson PT. The miR-15/107 group of microRNA
genes: evolutionary biology, cellular functions, and roles in human diseases. Journal of Molecular Biol-
ogy. 2010; 402(3):491–509. https://doi.org/10.1016/j.jmb.2010.07.051 PMID: 20678503
16. Du J, Zhang F, Zhang L, Jia Y, Chen H. MicroRNA-103 regulates the progression in endometrial carci-
noma through ZO-1. International Journal of Immunopathology and Pharmacology. 2019;
33:2058738419872621. https://doi.org/10.1177/2058738419872621 PMID: 31456452
17. Moncini S, Lunghi M, Valmadre A, Grasso M, Del Vescovo V, Riva P, et al. The miR-15/107 Family of
microRNA Genes Regulates CDK5R1/p35 with Implications for Alzheimer’s Disease Pathogenesis.
Molecular Neurobiology. 2017; 54(6):4329–4342. https://doi.org/10.1007/s12035-016-0002-4 PMID:
27343180
18. Xu Q, Li Y, Shang YF, Wang HL, Yao MX. miRNA-103: molecular link between insulin resistance and
nonalcoholic fatty liver disease. World Journal of Gastroenterology. 2015; 21(2):511–516. https://doi.
org/10.3748/wjg.v21.i2.511 PMID: 25593466
PLOS ONE | https://doi.org/10.1371/journal.pone.0232695 May 7, 2020
14 / 16
PLOS ONE Role and mechanism of miR-103 in mesenchymal stem cells osteogenic differentiation
19. Chen J, Li K, Pang Q, Yang C, Zhang H, Wu F, et al. Identification of suitable reference gene and bio-
markers of serum miRNAs for osteoporosis. Scientific Reports. 2016; 6:36347. https://doi.org/10.1038/
srep36347 PMID: 27821865
20. Valassi E, Garcia-Giralt N, Malouf J, Crespo I, Llauger J, Diez-Perez A, et al. Circulating miR-103a-3p
and miR-660-5p are associated with bone parameters in patients with controlled acromegaly. Endocrine
Connections. 2019; 8(1):39–49. https://doi.org/10.1530/EC-18-0482 PMID: 30640713
21. Yoo JK, Kim CH, Jung HY, Lee DR, Kim JK. Discovery and characterization of miRNA during cellular
senescence in bone marrow-derived human mesenchymal stem cells. Experimental Gerontology.
2014; 58:139–145. https://doi.org/10.1016/j.exger.2014.07.020 PMID: 25087724
22. Conner JR, Hornick JL. SATB2 is a novel marker of osteoblastic differentiation in bone and soft tissue
tumours. Histopathology. 2013; 63(1):36–49. https://doi.org/10.1111/his.12138 PMID: 23701429
23. Schmittgen TD, Livak KJ. Analyzing real-time PCR data by the comparative C(T) method. Nature Proto-
cols. 2008; 3(6):1101–1108. https://doi.org/10.1038/nprot.2008.73 PMID: 18546601
24.
Fan L, Fan J, Liu Y, Li T, Xu H, Yang Y, et al. miR-450b Promotes Osteogenic Differentiation In Vitro
and Enhances Bone Formation In Vivo by Targeting BMP3. Stem Cells and Development. 2018; 27
(9):600–611. https://doi.org/10.1089/scd.2017.0276 PMID: 29649414
25. Wang T, Teng S, Zhang Y, Wang F, Ding H, Guo L. Role of mesenchymal stem cells on differentiation
in steroid-induced avascular necrosis of the femoral head. Experimental and Therapeutic Medicine.
2017; 13(2):669–675. https://doi.org/10.3892/etm.2016.3991 PMID: 28352349
26. Shen H, Lu C, Shi J, Li H, Si J, Shen G. Satb2 expression in Foxc1-promoted osteogenic differentiation
of MC3T3-E1 cells is negatively regulated by microRNA-103-3p. Acta Biochimica et Biophysica Sinica
(Shanghai). 2019; 51(6):588–597. https://doi.org/10.1093/abbs/gmz037 PMID: 31089719
27.
Li M, Liu Z, Zhang Z, Liu G, Sun S, Sun C. miR-103 promotes 3T3-L1 cell adipogenesis through AKT/
mTOR signal pathway with its target being MEF2D. Biological Chemistry. 2015; 396(3):235–244.
https://doi.org/10.1515/hsz-2014-0241 PMID: 25400071
28. Croizier S, Park S, Maillard J, Bouret SG. Central Dicer-miR-103/107 controls developmental switch of
POMC progenitors into NPY neurons and impacts glucose homeostasis. Elife. 2018; 7:e40429. https://
doi.org/10.7554/eLife.40429 PMID: 30311908
29. Holik AK, Lieder B, Kretschy N, Somoza MM, Held S, Somoza V. N() -Carboxymethyllysine Increases
the Expression of miR-103/143 and Enhances Lipid Accumulation in 3T3-L1 Cells. Journal of Cellular
Biochemistry. 2016; 117(10):2413–2422. https://doi.org/10.1002/jcb.25576 PMID: 27137869
30.
Fang T, Li J, Wu X. Shenmai injection improves the postoperative immune function of papillary thyroid
carcinoma patients by inhibiting differentiation into Treg cells via miR-103/GPER1 axis. Drug Develop-
ment Research. 2018; 79(7):324–331. https://doi.org/10.1002/ddr.21459 PMID: 30267584
31. Yi J, Liu D, Xiao J. LncRNA MALAT1 sponges miR-30 to promote osteoblast differentiation of adipose-
derived mesenchymal stem cells by promotion of Runx2 expression. Cell and Tissue Research. 2019;
376(1):113–121. https://doi.org/10.1007/s00441-018-2963-2 PMID: 30511267
32. Yang D, Wang JJ, Li JS, Xu QY. miR-103 Functions as a Tumor Suppressor by Directly Targeting Pro-
grammed Cell Death 10 in NSCLC. Oncology Research. 2018; 26(4):519–528. https://doi.org/10.3727/
096504017X15000757094686 PMID: 28734041
33. Chung J, Grant RI, Kaplan DR, Irwin MS. Special AT-rich binding protein-2 (SATB2) differentially affects
disease-causing p63 mutant proteins. Journal of Biological Chemistry. 2011; 286(47):40671–40680.
https://doi.org/10.1074/jbc.M111.271189 PMID: 21965674
34.
Zhao X, Qu Z, Tickner J, Xu J, Dai K, Zhang X. The role of SATB2 in skeletogenesis and human dis-
ease. Cytokine & Growth Factor Reviews. 2014; 25(1):35–44. https://doi.org/10.1016/j.cytogfr.2013.12.
010 PMID: 24411565
35. Xu M, Xu X, Pan B, Chen X, Lin K, Zeng K, et al. LncRNA SATB2-AS1 inhibits tumor metastasis and
affects the tumor immune cell microenvironment in colorectal cancer by regulating SATB2. Molecular
Cancer. 2019; 18(1):135. https://doi.org/10.1186/s12943-019-1063-6 PMID: 31492160
36. Yu W, Ma Y, Shankar S, Srivastava RK. Role of SATB2 in human pancreatic cancer: Implications in
transformation and a promising biomarker. Oncotarget. 2016; 7(36):57783–57797. https://doi.org/10.
18632/oncotarget.10860 PMID: 27472393
37. Xu HY, Fang W, Huang ZW, Lu JC, Wang YQ, Tang QL, et al. Metformin reduces SATB2-mediated
osteosarcoma stem cell-like phenotype and tumor growth via inhibition of N-cadherin/NF-kB signaling.
European Review for Medical and Pharmacological Sciences. 2017; 21(20):4516–4528. PMID:
29131265
38.
Zhen L, Jiang X, Chen Y, Fan D. MiR-31 is involved in the high glucose-suppressed osteogenic differen-
tiation of human periodontal ligament stem cells by targeting Satb2. American Journal of Translational
Research. 2017; 9(5):2384–2393. PMID: 28559988
PLOS ONE | https://doi.org/10.1371/journal.pone.0232695 May 7, 2020
15 / 16
PLOS ONE Role and mechanism of miR-103 in mesenchymal stem cells osteogenic differentiation
39. Hao J, Zhang L, Cong G, Ren L, Hao L. MicroRNA-34b/c inhibits aldosterone-induced vascular smooth
muscle cell calcification via a SATB2/Runx2 pathway. Cell and Tissue Research. 2016; 366(3):733–
746. https://doi.org/10.1007/s00441-016-2469-8 PMID: 27503378
40. Wu G, Li Z, Wang Y, Ju X, Huang R. miR-34a Inhibits Cell Proliferation by Targeting SATB2 in Hepato-
cellular Carcinoma. BioMed Research International. 2018; 2018:2863902. https://doi.org/10.1155/
2018/2863902 PMID: 30627547
41.
Tang J, Zhang Z, Jin X, Shi H. miR-383 negatively regulates osteoblastic differentiation of bone marrow
mesenchymal stem cells in rats by targeting Satb2. Bone. 2018; 114:137–143. https://doi.org/10.1016/j.
bone.2018.06.010 PMID: 29909059
42. Wang J, Lu Y, Ding H, Gu T, Gong C, Sun J, et al. The miR-875-5p inhibits SATB2 to promote the inva-
sion of lung cancer cells. Gene. 2018; 644:13–19. https://doi.org/10.1016/j.gene.2017.11.066 PMID:
29196257
PLOS ONE | https://doi.org/10.1371/journal.pone.0232695 May 7, 2020
16 / 16
PLOS ONE |
10.1371_journal.pone.0230651 | RESEARCH ARTICLE
Are dogs with congenital hearing and/or
vision impairments so different from sensory
normal dogs? A survey of demographics,
morphology, health, behaviour,
communication, and activities
Sophie SavelID
1*, Patty Sombe´ 2
1 Aix Marseille University, CNRS, Centrale Marseille, LMA UMR 7031, Marseille, France, 2 Non-profit
organisation “Blanc Comme Neige”, Pont d’Ouilly, France
* savel@lma.cnrs-mrs.fr
Abstract
The births of domestic dogs with pigment deletion and associated congenital hearing and/or
vision impairments are increasing, as a result of mutations of certain genes expressing pop-
ular coat colour patterns (Merle, piebald, Irish spotting). The future of these dogs is often
pessimistic (early euthanasia or placement in rescues/fosters, lack of interactions and activi-
ties for adults). These pessimistic scenarios result from popular assumptions predicting that
dogs with congenital hearing/vision impairments exhibit severe Merle-related health trou-
bles (cardiac, skeletal, neurological), impairment-related behavioural troubles (aggres-
siveness, anxiety), and poor capacities to communicate, to be trained, and to be engaged in
leisure or work activities. However, there is no direct scientific testing, and hence no evi-
dence or refutation, of these assumptions. We therefore addressed an online questionnaire
to owners of 223 congenitally sensory impaired (23 vision impaired, 63 hearing impaired,
137 hearing and vision impaired) and 217 sensory normal dogs from various countries. The
sensory normal cohort was matched in age, lifetime with owner, breed and sex with the sen-
sory impaired cohort, and was used as a baseline. The questionnaire assessed demograph-
ics, morphology, sensory impairments, health and behavioural troubles, activities, and dog-
owner communication. Most hearing and/or vision impaired dogs exhibited abnormal pig-
ment deletion in their coat and irises. Vision impaired dogs additionally exhibited ophthalmic
abnormalities typically related to Merle. The results are opposed to all above-listed assump-
tions, except for neurological troubles, which were more frequently reported in sensory
impaired dogs. However, we suggest that this finding could be partially accounted for by a
lack of diagnosis of breed-related drug sensitivity and impairment-related compulsive
behaviours. Results about communication and activities are particularly optimistic. The
need for future studies of numerous dogs from various breeds tested for Merle, piebald and
medical-drug-resistance genes, and the beneficial effects that present and future research
may have on the future of sensory impaired dogs, are discussed.
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: Savel S, Sombe´ P (2020) Are dogs with
congenital hearing and/or vision impairments so
different from sensory normal dogs? A survey of
demographics, morphology, health, behaviour,
communication, and activities. PLoS ONE 15(9):
e0230651. https://doi.org/10.1371/journal.
pone.0230651
Editor: Simon Clegg, University of Lincoln, UNITED
KINGDOM
Received: March 4, 2020
Accepted: August 4, 2020
Published: September 4, 2020
Copyright: © 2020 Savel, Sombe´. This is an open
access article distributed under the terms of the
Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: All relevant data are
within the manuscript and its Supporting
Information files.
Funding: The author(s) received no specific
funding for this work.
Competing interests: The authors have declared
that no competing interests exist.
PLOS ONE | https://doi.org/10.1371/journal.pone.0230651 September 4, 2020
1 / 38
PLOS ONE Comparison between hearing and/or vision impaired and sensory normal dogs
Introduction
In order to meet an increasing demand for pet dogs, most countries report growing numbers
of dog breeders, atypical phenotypes in either existing or novel dog breeds, and births of pup-
pies with various genetic defects [1]. For example, the population of dogs with congenital hear-
ing and/or vision impairments is increasing. This population results from the mutation of
specific genes expressing popular patterns of coat colour.
Demographics and genetics of congenitally sensory impaired dogs
One of the most popular coat colour patterns in dogs is Merle. The Merle coat can be described
as a patchwork randomly composed of areas of full pigmentation combined with areas of ligh-
ter, diluted pigmentation. Originally, the Merle trait was essentially produced in certain breeds,
mainly from the herding group. This trait has progressively been introduced in a growing num-
ber of, sometimes unexpected, breeds. To date, Merle can be found in the following breeds,
listed in alphabetic order: Alapaha Blue Blood Bulldog, American Cocker Spaniel, American Pit
Bull Terrier, American Staffordshire Terrier, Australian Koolie, Australian Shepherd, Border
Collie, Boxer, Chihuahua, Collie, Dachshund, French Beauceron, French Bulldog, Great Dane,
Hungarian Mudi, Labradoodle, Louisiana Catahoula, Lurcher, Miniature American Shepherd,
Norwegian Dunkerhound, Schnauzer, Shetland Sheepdog, Pomeranian, Poodle, Pyrenean
Shepherd, and Welsh Sheepdog [2–4]. The Merle coat is particularly frequent in Australian
Shepherds, followed by Border Collies, Great Danes, and Shetland Sheepdogs. To note, Merle is
not accepted for registration in kennel clubs for several of the breeds listed.
The Merle coat is expressed by the gene of the same name located on the M locus, that is
inherited in an autosomal, incomplete dominant manner [5]. Merle is, at the homozygous
“double Merle” state, one the four known pigment genes in dogs, along with piebald [2], Irish
spotting [2] and KIT [6], whose mutation has the deleterious effect of deleting pigments in
hairs, skin, nose and mucous, iris and tapetum lucidum, and stria vascularis of the inner ear.
The lack of pigments in the stria vascularis causes early death of sensory hair cells in the scala
media. As a result, dogs with mutated above-mentioned genes frequently have excessive white
coat, pink skin, nose and mucous, light blue irises, and congenital, sensorineural, irreversible
hearing impairments.
As for Merle, piebald concerns various breeds, but this trait, located on the S locus, is reces-
sive. Pigment deletion and hearing impairments mostly occur in homozygous piebalds. No
genetic testing is yet available for Irish spotting, although this gene is assumed to be present in
numerous breeds. The KIT mutation is less statistically problematic, because it exclusively con-
cerns the German Shepherd breed and is early embryonic lethal at the homozygous state.
Contrary to piebald, Irish spotting and KIT, Merle is additionally associated with various
congenital ophthalmic abnormalities, referred to as Merle Ocular Dysgenesis. The most severe
ophthalmic abnormalities are observed in double Merles. For these dogs, ophthalmic abnor-
malities can concern the eyeball (reduced size, called microphthalmia, or total absence), the
cornea (microcornea), the iris (coloboma, hypoplasia), the size, shape, position or reaction of
the pupil (starburst or misshapen pupil, dyscoria, corectopia), the pupillary membrane (persis-
tence), the lens (cataract, microphakia, luxation), the sclera (coloboma, staphyloma), the retina
(detachment, dysplasia), and the optic nerve [7, 8]. Depending on their severity, these ophthal-
mic abnormalities in double Merles can induce moderate to severe vision impairments, which
are susceptible to worsen, or even to appear, over the life course.
It has long been assumed that the M locus has two possible alleles, namely non-merle (m,
expressing solid phenotype) and Merle (M, expressing Merle phenotype). Latest research has
PLOS ONE | https://doi.org/10.1371/journal.pone.0230651 September 4, 2020
2 / 38
PLOS ONE Comparison between hearing and/or vision impaired and sensory normal dogs
identified between four and six variations of the Merle allele, as a function of the length of the
poly-A tail of its SINE insertion [9, 10; 3]. The most detailed work is that conducted by Lange-
vin and colleagues, who have tested hundreds of dogs from various breeds on the M locus and
have determined six variations of the Merle allele [3, 4]. Their goals were to accurately deter-
mine which SINE insertion lengths can express a Merle pattern, which phenotype is most typi-
cally associated with each possible genotype, which cases of mosaicism can occur, and which
breeding between genotypes are susceptible to produce excess white and sensory impaired, so-
called double Merle, puppies. However, these genetics studies of Merle are very recent. The
state-of-the-art equipment needed for precise examination of SINE insertion length and mosa-
icism is recent and expensive. Therefore, only two of the 16 laboratories that propose Merle
testing in dogs can currently provide this detailed information in their test results. Few of the
remaining laboratories provide state-of-the-art information about Merle genetics on their
public websites [11]. As a result, many dog breeders and owners are not fully aware of the com-
plexity of Merle genetics and the conditions of at-risk breeding. Many countries have not yet
strictly regulated Merle breeding. Comparable lack of information and regulation is observed
for different breeds with piebald trait.
For all the reasons mentioned above, births of excess white puppies with congenital, senso-
rineural, irreversible hearing and/or vision impairments are still frequently observed world-
wide. There are different possible scenarios regarding the future of these puppies. Non-
professional breeders and private individuals with little-or-no knowledge of the genetics of
sensory impairments sell their puppies as exotic specimens without providing any adequate
information to buyers. Numerous professional breeders with informed knowledge either have
the puppies euthanized shortly after birth or entrust them after weaning to specialised rescue
centres or foster programs that have very restrictive adoption criteria for such dogs. Dogs that
have been lucky enough to be adopted and to become adults often live in controlled–some-
times “overprotective”–environments, and do not often have access to canine activities. All
these pessimistic scenarios result from a series of popular assumptions about deaf and/or blind
dogs, that are detailed below.
Assumptions about sensory impaired dogs
It is often assumed that excess white dogs, particularly double Merles, exhibit severe, or even
lethal, health issues in their neurological, cardiac, skeletal and reproductive systems. Below, we
propose three possible origins of these potentially false assumptions.
First, the assumption of lethality in excess white, double Merle dogs may originate from the
fact that, in horses, the mutation of the Overo gene causes both abnormal white coat and early
death of the foal [12]. A popular belief has incorrectly extended this relatively well-known fact
to all mammalian animal species. Accordingly, many websites and social media relative to
canine genetics refer to excess white dogs as “lethal white dogs”. In fact, there are only four
canine genes that have proved to be–early embryonic–lethal at the homozygous state (KIT in
German Shepherds, “Natural Bobtail” in various breeds, Harlequin in Great Danes, and “Hair-
less” in Chinese, Mexican and Peruvian hairless breeds) [4]. Merle is not one of them.
Second, the assumption of neurological issues in double Merles may originate in part from
the fact that the most common neurological disorder in dogs, namely primary idiopathic epi-
lepsy, is frequent in certain breeds from the herding group, such as Australian Shepherds and
Border Collies [13]. As specified above, these two breeds are most frequently concerned by the
homozygous Merle genotype. The assumption of neurological issues may additionally, or even
above all, result from the fact that Australian Shepherds and Border Collies are also frequently
concerned by mutations of the medical drug resistance (MDR1) gene [14, 15]. As detailed
PLOS ONE | https://doi.org/10.1371/journal.pone.0230651 September 4, 2020
3 / 38
PLOS ONE Comparison between hearing and/or vision impaired and sensory normal dogs
further below, mutated MDR1 gene elicits neurotoxic, sometimes epileptiform reactions to
common chemical agents (e.g., parasite control products) that are well tolerated by dogs with
normal MDR1 gene. In other words, whether the neurological signs observed in double Merle
Australian Shepherds and Border Collies are linked to their homozygous Merle genotype, as
frequently assumed, or to breed-dependent epilepsy and MDR1 mutation, is undetermined.
Third, assumptions of cardiac, skeletal and reproductive issues in double Merles may essen-
tially originate from multiple citations of a single study, that just contained the following short
statement in the Introduction: “In all breeds, the double merle genotype can be sublethal and is
associated with multiple abnormalities of the skeletal, cardiac, and reproductive systems” [5].
However, this study was on genetic testing of Merle, not on health, and cited three studies to
support the statement [16–18]. These three studies were conducted long before genetic testing
of Merle was available, examined either small or poorly genetically diversified dog cohorts (i.e.,
total of 32 dogs or single genealogical branch), and never clearly referred to the types of health
issues that are nonetheless listed in the statement quoted above.
Moreover, it is often assumed that congenitally deaf and/or blind dogs exhibit behavioural
troubles (see foreword by Strain in [19]). Principally, their sensory impairment(s) are believed
to increase frustration, and to elicit resultant aggressiveness and anxiety troubles. Also, it is
assumed that deaf and/or blind dogs are particularly susceptible to bite, because they are easily
startled when they are approached. Abnormal brain structures, and concomitant abnormal
mental capacities, have also been assumed in congenitally deaf dogs. However, this assumption
is based on a single neuro-imagery study that just reported a reduction in size of the auditory
cortex in congenitally deaf Dalmatians [20].
Finally, for dogs as for many social species, hearing and vision are two important sensory
modalities for conspecific and interspecific communication [21]. Thus, deaf and/or blind dogs
are believed to have poorer communication capacities than sensory normal dogs, in particular
with their human caregivers. As a result, it is often assumed that deaf and/or blind dogs cannot
be trained, and cannot be safely and efficaciously engaged in any–individual or collective, con-
specific or interspecific, leisure or work–activity. Another related belief is that because they
had no possibility to benefit from auditory-based vocal learning during early ontogenesis, con-
genitally deaf dogs are less “talkative” than sensory normal ones, and are therefore less able to
produce vocalisations during interactions.
Aims and methodological choices of the study
In summary, above-mentioned assumptions predict that congenitally hearing and/or vision
impaired dogs frequently exhibit health and behavioural troubles, and are poorly capable of
communicating and practicing activities. These assumptions are so popular that they have
drastic consequences on the future of sensory impaired puppies. However, there is to date no
scientific evidence or refutation of these assumptions. Precisely, there is no study that we are
aware of that directly assessed either health, behaviour, communication or activities in congen-
itally sensory impaired dogs, or that compared sensory impaired and sensory normal dogs on
these points. One exception is the study by Farmer-Dougan and colleagues, who addressed a
survey of behavioural traits to owners of hearing/vision impaired and sensory normal dogs
[22]. They found lower scores of aggressiveness and anxiety in the former cohort, which is
opposite to the assumption.
The aims of the present study were therefore to examine health, behaviour, communication
and activities for a cohort of congenitally hearing and/or vision impaired dogs, and to compare
the results with those from a “baseline” cohort of sensory normal dogs that was matched in
breed, age, sex and lifetime with owner with the sensory impaired cohort. Additionally, we
PLOS ONE | https://doi.org/10.1371/journal.pone.0230651 September 4, 2020
4 / 38
PLOS ONE Comparison between hearing and/or vision impaired and sensory normal dogs
aimed to gain insight concerning the diagnosis of sensory, particularly hearing, impairments.
The sole way to assess unilateral hearing impairment in dogs is to conduct objective measure-
ments of brainstem auditory evoked responses (BAER). However, animal BAER testing sites
are infrequent (e.g., see list in [19]). Little is known about how exactly hearing is subjectively
evaluated by veterinarians, breeders, owners, etc, in the numerous dogs that have no access to
BAER testing. Finally, we aimed to verify whether congenital sensory impairments were fre-
quently associated with pigment deletion in the coat and irises and with ophthalmic abnormal-
ities. As such, the genetic cause of the sensory impairment(s) was indirectly explored.
To assess these different points, we chose to conduct an owner survey, a method that is fre-
quently employed to assess, for example, health [23] and behaviour [22], in dogs. Surveys of
dog owners have some disadvantages relative to the degrees of interest, understanding, recall
and impartiality of the respondents, but have the following advantages: they allow the inclu-
sion of dogs with much wider characteristics compared to surveys of veterinarians or animal
insurance companies for questions on health, dog behaviourists for questions on behaviour, or
canine clubs for questions on activities. In other words, owner surveys are not restricted to
dogs with health/behavioural issues and activities. We chose the online diffusion of the ques-
tionnaire in two languages, namely French and English, in order to expand its geographic
distribution.
Materials and methods
Questionnaire content
The questionnaire contained 30 questions about dogs, divided into 7 sections:
• Demographics: country; date of birth; date of acquisition; site of acquisition; sex; breed;
presence of other dog(s) at home.
• Morphology: surface of white coat on body and head; colour of irises; ophthalmic abnormal-
ities; for sensory impaired dogs only: is there indication, from either genetic testing or paren-
tal phenotypes, that the dog is double Merle?
• Determination of sensory impairments: sensory status (i.e., normal, partially impaired,
totally impaired) at each ear and each eye; type of diagnosis test (i.e., objective or subjective);
operator of the subjective test; for hearing impairments only: stimuli and conditions of the
subjective test.
• Health: has the dog ever suffered from neurological, heart, bones/joints, skin, digestive or
other health troubles? has the dog been tested for the MDR1 gene? if so, did the test indicate
MDR1 mutation, and hence abnormal drug sensitivity?
• Behaviour: has the dog ever suffered from aggressiveness, anxiety, attention deficit/hyperac-
tivity disorder (ADHD), obsessive compulsive disorder (OCD), or other behavioural trou-
bles? who diagnosed the behavioural trouble(s)? have drugs been prescribed for this/these
trouble(s)?
• Activities: frequency of practice of a series of leisure/sport activities; level at which each
reported activity is practiced; is the dog engaged in assistance/therapy activities with either
elderly, blind, or diabetic/epileptic persons?
• Interspecific communication: types of vocalisations produced by the dogs to communicate
with their owners; types of signs used by the owners to communicate with, and train, their
dogs.
PLOS ONE | https://doi.org/10.1371/journal.pone.0230651 September 4, 2020
5 / 38
PLOS ONE Comparison between hearing and/or vision impaired and sensory normal dogs
All respondents gave their informed consent for the anonymous use and publication of
their responses. They were proposed to send a picture of their dog to the first author by email.
We received 88 pictures, that are presented in S1 Fig as illustrative examples of coat colour
and ophthalmic abnormalities. The study was carried out in accordance with the ethical stan-
dards of the institutional review board at the Aix-Marseille University. No ethical approval
was required by the institutional review board, as the study was an online survey of pet dog
characteristics.
Survey distribution
The questionnaire was published online using Google forms in two languages: French and
English (see copy in S2 Fig). Both versions were operational online from 19th April 2019 until
30th September 2019. Calls for participation, that included a short description of the survey
and a direct link to the google form, were published on a variety of social media. The social
media dealt with various canine themes, such as breeding, genetics, sensory impairments,
training methods, activities, veterinary medicine, and behaviour. Calls for participation speci-
fied that the questionnaire was addressed to owners of dogs:
• aged between 9 months and 12 years
• with either no or congenital hearing and/or vision impairments
• that belonged to breeds for which the Merle coat is–frequently or occasionally–observed.
Dogs with acquired, late onset sensory impairments resulting from trauma, age, medica-
tion, surgery, etc, were explicitly excluded. Owners had the possibility to fill the questionnaire
several consecutive times for different individual dogs.
Size and geographic dispersion of the sample
Following data collection, responses to the French version of the questionnaire were translated
in their English equivalent. Responses to French and English versions were then gathered.
Overall, owners of 510 individual dogs completed the survey. The responses relative to 75 dogs
were excluded from data analysis because mandatory questions were inadequately responded
and/or one above-mentioned criterion was not fulfilled. For example, 33 dogs were aged less
than 9 months, 24 were aged more than 12 years, and 20 were from “out-of-subject” breeds (e.
g., Beagle, Dalmatian, Yorkshire Terrier, Golden Retriever, unidentified mix of breeds, etc.).
The final sample therefore included 440 individual dogs, whose geographic dispersion between
continent and countries is presented in Table 1. About 85% of the dogs were from either
France or United States of America. The remaining dogs were spread between 15 countries.
Constitution of groups based on sensory status
Responses about the hearing and vision sensory status of the dog (i.e., normal, partially
impaired, or totally impaired) were used to classify the 440 dogs in four groups:
• HNVI (Hearing Normal Vision Impaired) = 23 dogs: response “normal” for hearing at
both ears, and response “partially/totally impaired” for vision at either one or both eyes
• HIVN (Hearing Impaired Vision Normal) = 63 dogs: response “partially/totally impaired”
for hearing at either one or both ears, and response “normal” for vision at both eyes
• HIVI (Hearing Impaired Vision Impaired) = 137 dogs: response “partially/totally
impaired” for hearing at either one or both ears and for vision at either one or both eyes
PLOS ONE | https://doi.org/10.1371/journal.pone.0230651 September 4, 2020
6 / 38
PLOS ONE Comparison between hearing and/or vision impaired and sensory normal dogs
Table 1. List of countries and corresponding number of dogs.
Continent
Europe
Country
France
Belgium
Switzerland
Netherlands
United Kingdom
Germany
Finland
Italy
Spain
Slovakia
Number of dogs
240
16
7
4
4
3
2
2
1
1
America
United States of America
136
Canada
Mexico
Brazil
Australia
New Zealand
South Africa
Oceania
Africa
Continent and countries are sorted by descending number of dogs.
https://doi.org/10.1371/journal.pone.0230651.t001
9
2
1
5
3
4
• HNVN (Hearing Normal Vision Normal) = 217 dogs: response “normal” for hearing at
both ears and for vision at both eyes.
In summary, the study included 223 sensory impaired dogs, classified in three groups, and
217 sensory normal dogs. The first two sensory impaired groups had one–either hearing or
vision–impairment, while the third sensory impaired group had both hearing and vision
impairments. In the Figures and Tables below, the three sensory impaired groups were fre-
quently gathered in a single “IMP” (impaired) cohort. The HNVN, sensory normal group was
used as a baseline for comparison with the sensory impaired groups/cohort.
Statistical analysis
Owners were asked to report the exact dates of birth and acquisition of their dog. These two
dates were used to determine the dog’s age and lifetime with owner, respectively, in years, at
the day of participation in the survey. The normal distributions of individual age and lifetime
values for each group and for the IMP cohort were assessed using Shapiro-Wilk tests. All dis-
tributions differed significantly from normal (p < 0.05). Between-group comparisons in both
age and lifetime with owner were therefore assessed using non-parametric, Kruskal-Wallis
tests [24].
The frequency of each categorical response investigated was obtained for each group by
dividing the number of times the response was reported in the group by the total number of
dogs in that group and then multiplying by 100. The response frequencies obtained are pre-
sented below either in Figures (ordinate) or in Tables. Chi2 tests for unpaired data were used
to statistically assess two-by-two differences between groups in categorical responses. Each
Chi2 test compared the raw numbers of reported/A and non-reported/B responses obtained in
one group (e.g., numbers of “yes” and “no” responses, numbers of “male” and “female”
responses, etc.) with those obtained in the other group. The list of comparisons assessed using
Chi2 tests is provided in Table 2.
PLOS ONE | https://doi.org/10.1371/journal.pone.0230651 September 4, 2020
7 / 38
PLOS ONE Table 2. List of two-by-two comparisons between groups assessed using Chi2 tests.
Comparison between hearing and/or vision impaired and sensory normal dogs
Response category
Demographics
Breed (Australian Shepherd or Border Collie/other)
Sex (male/female)
Other dog(s) at home (yes/no)
Morphology
White on body (< 50%/� 50%) Wf body (< 50%/�
50%)
White on head (< 50%/� 50%) of body (< 50%/�
50%)
Iris colour (normal/abnormal)
Ophthalmic abnormalities (yes/no)
Microphthalmia
Misshapen pupil
Cataract
Absence of eyeball
Other than listed
None
Health troubles (yes/no)
Neurological
Heart
Bones and joints
Skin
Digestive
Other than listed
None
Tested for MDR1 gene
MDR1 test showed drug sensitivity
Behavioural troubles (yes/no)
Aggressiveness
Anxiety
ADHD
OCD
Other than listed
None
Activities (yes/no)
Canicross/bike/scootering
Agility
Dog Dancing
Sheep Herding
Tracking of objects or persons
Frisbee/flyball/treiball
Other than listed
None
Dog vocalisations (yes/no)
Barks
Growls/Grunts
https://doi.org/10.1371/journal.pone.0230651.t002
Groups
compared
HNVI HIVN
HNVI
HIVI
HIVN
HIVI
HNVN
HIVN
HNVN HIVI
HNVN IMP
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
(<50%/� 50%) of body (< 50%/�
50%)
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
PLOS ONE | https://doi.org/10.1371/journal.pone.0230651 September 4, 2020
8 / 38
PLOS ONE Comparison between hearing and/or vision impaired and sensory normal dogs
The categorical responses under comparison are given between brackets in italics, sepa-
rately by a slash, in the left column. For data on morphology, comparisons between groups
focused on sensory impaired groups (HNVI, HIVN and HIVI). For data on health, behaviour
and activities, the three sensory impaired groups were gathered in a single IMP cohort and
then compared with the HNVN group. For data on sex and dog vocalisations, comparisons
were selected from the results.
Two-tailed p values reported in text and Figures were adjusted using the Holm’s correction
for multiple comparisons [25]. In Figures, p values for significant tests are emboldened and are
followed by either four (p < 0.0001), three (p � 0.0001 and < 0.001), two (p � 0.001
and < 0.01) or one (p � 0.01 and < 0.05) asterisk(s). p values for non-significant tests are
reported in plain and are followed by “(ns)” (non significant).
Results and discussion
Demographics
Age. Box-plots of age values for each group and for the IMP cohort are presented in Fig
1A. There was no significant two-by-two difference in age between sensory impaired groups
(median ages for HNVI, HIVN and HIVI groups = 2.6, 3.0 and 3.3 years, respectively;
H < 0.25, p = 1.0) or between IMP and HNVN (median ages = 3.1 and 3.5 years, respectively;
H = 3.87, p = 0.34). The five distributions had comparable lower quartiles (between 1.4 and 2.1
years), upper quartiles (between 5.0 and 5.8 years) and inter-quartile distances (between 3.1
and 3.8 years), and had few outlier values (frequencies � 5%). Age has therefore unlikely con-
tributed to between-group differences in the data examined in the different sections of the
survey.
Lifetime with owner. Box-plots of lifetime values for each group and for the IMP cohort
are presented in Fig 1B. There was no significant two-by-two difference between sensory
impaired groups (median lifetimes with owner for HNVI, HIVN and HIVI groups = 2.2, 2.4
and 2.3 years, respectively; H < 0.24, p = 1.0), while the difference between IMP and HNVN
cohorts was significant (median lifetimes with owner = 2.3 and 3.1 years, respectively;
H = 13.50, p = 0.002). However, both cohorts had lower quartiles above one year. In beha-
vioural studies of dog-owner communication that report the lifetime of the dog-owner pair,
the smallest lifetime is one year [e.g., 26]. The difference in lifetime with owner between sen-
sory impaired and sensory normal dogs in the present survey has therefore unlikely contrib-
uted to differences between groups in owners’ responses relative to interspecific activities and
communication.
Breed. Owners were asked to report the breed of their dog using a list of purebred and
mixed breeds followed by a field for manual reporting of non-listed breeds (see S2 Fig). The
frequencies of the different responses obtained for each group are provided in Table 3A. The
four groups were essentially composed of breeds from the herding group. Australian Shep-
herds and Border Collies, either purebred or mixed, represented 79% and 84%, respectively, of
sensory impaired and sensory normal cohorts (X2 = 0.22, p = 1.0, ns). Potential effects of breed
may have therefore been equally elicited in all groups in the data examined below.
All the breeds listed have standard coat colour patterns with only minor areas of white
according to kennel clubs. Also, all breeds possibly have three of the genes (Merle, piebald,
Irish spotting) whose mutations are known to cause both pigment deletion in hairs and eyes
and congenital hearing impairments (plus vision impairments for Merle) [27; 3, 4].
Sex. The frequencies of males and females for each group are listed in Table 3B. The
group with the highest frequency of females (HNVI = 61%) did not significantly differ from
that with the lowest frequency (HIVN = 44%; X2 = 1.82, p = 1.0). Overall, males and females
PLOS ONE | https://doi.org/10.1371/journal.pone.0230651 September 4, 2020
9 / 38
PLOS ONE Comparison between hearing and/or vision impaired and sensory normal dogs
Fig 1. Box-plots of individual values of (a) age and (b) lifetime with owner for each group, and for the three sensory impaired groups gathered (IMP).
HNVI = hearing normal vision impaired (grey), HIVN = hearing impaired vision normal (orange), HIVI = hearing impaired vision impaired (red),
HNVN = hearing normal vision normal (green), IMP = impaired (purple), ns = non significant. The bold bar within the boxplot is the median, the cross is the
mean, the bottom and top of the box are the lower and upper quartiles, respectively, and the dots are outlier values above the [upper quartile + 1.5 � inter-
quartile distance] limit (top of upper vertical bars). Horizontal brackets indicate the two-by-two comparisons between groups that were statistically assessed
using Shapiro-Wilk tests.
https://doi.org/10.1371/journal.pone.0230651.g001
PLOS ONE | https://doi.org/10.1371/journal.pone.0230651 September 4, 2020
10 / 38
PLOS ONE Table 3. Frequencies of responses, in percentages, obtained for each group and for the IMP cohort for the following demographic data: (a) breed, (b) sex, (c) site of acqui-
sition, and (d) presence of other dog(s) at home.
Comparison between hearing and/or vision impaired and sensory normal dogs
Demographic data
(a) Breed
Herding group
Australian Shepherd
Australian Shepherd x Border Collie
Australian Shepherd x unknown
Border Collie
Border Collie x unknown
Shetland Sheepdog
Koolie
Miniature American Shepherd
Beauceron
Beauceron x unknown
Rough Collie
Welsh Corgi
Other breed groups
Great Dane
Catahoula
Dachshund
Chihuahua
American Cocker Spaniel
(b) Sex
male
female
(c) Site of acquisition
rescue
breeder
private
no responsea
(d) Other dog(s) at home
yes
HNVI
HIVN
HIVI
HNVN
IMP
39
4
4
26
9
--
--
--
4
4
--
--
--
--
4
4
--
39
61
57
9
< 1
34
83
49
8
5
19
3
--
--
2
--
--
--
2
8
2
2
2
--
56
44
46
5
16
33
70
41
9
11
12
4
4
1
--
--
--
1
--
8
4
2
1
2
55
45
59
2
4
34
85
51
3
2
25
3
3
2
2
3
2
--
< 1
< 1
--
1
1
--
51
49
8
47
42
3
62
43
8
9
15
4
2
< 1
< 1
< 1
< 1
1
< 1
7
3
2
2
1
53
47
55
4
7
34
80
In the list of breeds, the “x” symbols mean “mixed with”. Breed names without “x” symbol are for purebred breeds. “--” means that no dog was concerned.
a “no response” was for owners who did not respond to the optional question regarding the site of acquisition of their dog.
https://doi.org/10.1371/journal.pone.0230651.t003
were likewise balanced in all groups. Thus, differences between genders in health troubles,
aggressiveness, interspecific communication and cooperative activities (e.g., [28]) may have
been equally compensated for in all groups.
Site of acquisition. Owners were free to respond to the following optional question: “Site
of acquisition of your dog–where does your dog come from?” The response choices were:
• a rescue centre or a foster program
• a professional, registered breeder
• a private individual or a non-registered breeder.
The response frequencies obtained for each group are provided in Table 3C. No response
was given for 34% of the sensory impaired dogs and for 3% of the sensory normal ones.
PLOS ONE | https://doi.org/10.1371/journal.pone.0230651 September 4, 2020
11 / 38
PLOS ONE Comparison between hearing and/or vision impaired and sensory normal dogs
According to the responses for the 358 remaining dogs, sensory impaired dogs mostly came
from rescues/fosters, while sensory normal dogs came from either professional, registered
breeders or private individuals/non-registered breeders. The numerous missing responses and
the low frequency of “professional breeder” responses for sensory impaired dogs can easily be
explained. As described in the Introduction, both the sensory impairments and the morpho-
logical abnormalities of these dogs (see section “Morphology” below) result from inopportune,
sometimes illegal, breeding practices. Consequently, these dogs cannot officially be sold by reg-
istered breeders.
Presence of other dog(s) at home. Owners were asked to indicate whether they had other
dog(s) at home than the one concerned by their participation in the survey. This question was
asked because many rescue centres and foster programs that propose sensory impaired dogs
for adoption recommend the presence of at least one sensory normal dog at the adopter’s
home. According to these rescues/fosters, the sensory normal dog is expected to become a “ref-
erent” for the sensory impaired dog for various aspects of life, such as, for example, spatial
exploration and interactions. The responses obtained for each group are presented in
Table 3D. Significantly more sensory impaired than sensory normal dogs were reported as liv-
ing with other dog(s) (frequencies = 80% and 62%, respectively; X2 = 17.55, p = 0.002). This
difference may be explained by the above-mentioned recommendation, provided that many
sensory impaired dogs were adopted from rescues/fosters. An additional, related explanation
is that the adoption of a sensory impaired dog needs prior experience in dog-human commu-
nication and dog training. As a result, sensory impaired dogs are more “easily” adopted by per-
sons that already have had, or presently have, dogs. However, we have no hypothesis as to
whether the presence of other dog(s) at home could have differently affected the responses for
sensory impaired and sensory normal dogs for the various data compared below. This point is
therefore not analysed in further detail.
Determination of sensory impairments
Severity of the impairment. As mentioned above, owners were asked to report the sen-
sory status of their dog, at each ear and each eye separately, by choosing one of three possible
responses:
• normal
• partially impaired
• totally impaired (deaf/blind).
Table 4 shows how the responses were used to provide a “severity” score to hearing and
vision impairments. Scores 1 and 2 mean that the impairment is unilateral. Scores 3 and 4
mean that both ears/eyes are impaired but at possibly different degrees. Score 5 means that the
impairment is both total (i.e., deafness/blindness) and bilateral.
Table 4. Score of severity of hearing and vision impairments determined from owners’ responses to the sensory
status of their dog at each ear/eye.
Score of severity
Sensory status at one ear/eye
Sensory status at the other ear/eye
1
2
3
4
5
normal
normal
partially impaired
partially impaired
totally impaired
https://doi.org/10.1371/journal.pone.0230651.t004
partially impaired
totally impaired
partially impaired
totally impaired
totally impaired
PLOS ONE | https://doi.org/10.1371/journal.pone.0230651 September 4, 2020
12 / 38
PLOS ONE Comparison between hearing and/or vision impaired and sensory normal dogs
The left panel in Fig 2A shows the distributions of severity scores for the two hearing
impaired groups. Scores were equally distributed in the two groups (mean scores of severity ± 1
standard deviation for HIVN and HIVI groups = 4.6 ± 0.9 and 4.8 ± 0.7, respectively). Most
hearing impaired dogs were reported as being bilaterally deaf (frequencies of score 5 for HIVN
and HIVI groups = 81% and 87%, respectively). The right panel in Fig 2A shows the distribu-
tions of severity scores for the two vision impaired groups. Only 25% of the vision impaired
dogs were reported as being bilaterally blind. Thus, most vision impaired dogs had residual,
unilateral or bilateral, vision. The two vision impaired groups showed no clear difference in
score distributions, in spite of the trend for score 3 to be slightly more frequent for the HIVI
group (mean scores of severity ± 1 standard deviation for HNVI and HIVI groups = 3.3 ± 1.5
and 3.4 ± 1.3, respectively).
Diagnosis test. Owners who reported sensory impairment(s) in their dog were asked to
indicate whether the impairment(s) had been diagnosed using either:
• objective testing (i.e., BAER test in certified clinic for hearing; Canine Eye Registration
Foundation–CERF–or equivalent standardised ophthalmic test in certified clinic for vision)
• subjective testing (i.e., someone produced sounds/visual signals and observed the dog’s reac-
tion to these signals).
For vision impairments, an additional response choice was available:
• “diagnosis of vision impairment just based on abnormal eye(s) aspect”. This response was
for dogs having severe ophthalmic abnormalities, such as for example no eyeball, which
noticeably affect visual function (see pictures of vision impaired dogs in S1 Fig; see also sec-
tion “Ophthalmic abnormalities” below).
The responses obtained for hearing impaired and vision impaired groups are presented in
the left and right panels of Fig 2B, respectively. Sensory impairments were seldomly diagnosed
using objective testing (frequencies of BAER tested dogs for HIVN and HIVI groups = 17%
and 8%, respectively; frequencies of CERF-like tested dogs for HNVI and HIVI groups = 17%
and 29%, respectively). The finding that hearing impaired dogs were mostly diagnosed using
subjective testing can easily be explained by the small number of veterinary clinics that propose
BAER testing (see [19] for United States of America and several other countries; see https://
www.centrale-canine.fr/lofselect/actualites/la-surdite-comment-la-depister for France). Vision
impairments were almost equally diagnosed from CERF-like testing, subjective testing and
aspect of the eye(s).
Operator of the subjective test. Owners who responded that the sensory impairment(s)
of their dog had been diagnosed using subjective testing were asked to indicate who had con-
ducted that subjective test by choosing one the following responses:
• a veterinarian
• an employee or a volunteer in a rescue centre/foster
• the owner of the dog (themselves)
• the breeder of the dog.
The responses obtained for hearing impaired and vision impaired groups are presented in
the left and right panels of Fig 2C, respectively. Subjective testing was performed by a veteri-
narian in 61 to 67% of the cases. For the remaining dogs, subjective testing was more fre-
quently performed by the owner to evaluate vision than to evaluate hearing. Accordingly,
unilateral and/or partial impairments are more easily noticeable when they concern vision.
PLOS ONE | https://doi.org/10.1371/journal.pone.0230651 September 4, 2020
13 / 38
PLOS ONE Comparison between hearing and/or vision impaired and sensory normal dogs
Fig 2. Frequencies of responses, in percentages, obtained for hearing impaired groups (left panels) and for vision impaired groups (right panels) for the
following data: (a) score of severity of the impairment, (b) type of diagnosis test, and (c) operator of the subjective test. HNVI = hearing normal vision impaired
(grey), HIVN = hearing impaired vision normal (orange), HIVI = hearing impaired vision impaired (red).
https://doi.org/10.1371/journal.pone.0230651.g002
PLOS ONE | https://doi.org/10.1371/journal.pone.0230651 September 4, 2020
14 / 38
PLOS ONE Comparison between hearing and/or vision impaired and sensory normal dogs
Unilateral hearing impairments mainly affect sound source localisation while having less
noticeable effect on sound detection. Conversely, unilateral and/or partial vision impairments
affect the stereoscopic processing of space, objects, human gestures, etc, which has a visible
impact on both the motion and the posture of the dog. Moreover, subjective testing of monoc-
ular vision is much easier to conduct than subjective testing of monaural hearing. A visual
source presented on the edge of the visual field can exclusively be processed by the ipsilateral
eye. Conversely, a sound reaches the two ears regardless of the spatial position from which it is
presented. This explains why BAER testing is currently the only test of unilateral hearing
impairments. However, it should be mentioned that most clinical BAER tests use a single
sound (e.g., a click) presented at either fixed or few different level(s), which does not allow
assessing partial hearing impairments in one ear. This could possibly explain in part why con-
genital hearing impairments in dogs are so frequently reported as being total in the impacted
ear [29].
Stimuli and conditions of the subjective test. The 171 owners who indicated that the
hearing impairment of their dog had been diagnosed using subjective testing were asked to
“describe in a few words what the test consisted of”. This open question was asked in order to
get some information on the sounds, sites and conditions of the subjective tests that are per-
formed in the numerous dogs that have no access to BAER testing. In total, 109 responses
were unusable, because either the subjective test had been conducted prior to adoption of the
dog by the respondent or the response given was too vague (e.g., “my vet made different noise
to observe my dog’s reaction”, “my dog has never reacted to any sound”, etc).
According to the 62 usable responses, most sounds used in subjective testing of hearing
were natural sounds. These natural sounds were produced by either clapping/snapping/bang-
ing hands or fingers (22 responses), shaking/striking/dropping on the floor a metal object (12
responses), calling/talking to the dog out loud (11 responses), ringing a doorbell or an alarm (5
responses), producing whistles (5 responses), using a clicker (1 response) or a tuning fork (1
response), or turning on a vacuum cleaner (1 response). Seven other respondents indicated
that sounds were produced using an automated device, such as a smartphone application or an
audiometer, which allowed playing tonal or narrowband “artificial” sounds of different fre-
quencies at several levels. On the subject of the test conditions, 14 respondents indicated that
sounds were intentionally produced while the dog was sleeping. These 14 dogs were consid-
ered as deaf because the sound(s) produced did not wake them up. Regardless of dog arousal,
sounds were produced either very close to the dog’s ear (7 responses) or out of sight from sev-
eral locations and distances (18 responses). Only one respondent mentioned occlusion of one
ear during sound presentation, but without specifying how exactly the ear had been occluded.
Morphology
The survey was specifically addressed to owners of dogs with no or congenital sensory impair-
ments. As detailed in the Introduction, most congenital impairments in dogs are associated
with genetic-related deletion of pigments in hair and irises. Below, we therefore assessed the
extent to which sensory impairments were associated with discolouration of the coat and irises.
Ophthalmic abnormalities, which are consistently reported in dogs with mutation of one pig-
ment-deletion gene, namely in homozygous Merles, were also assessed.
Excess white coat. Owners were asked to indicate what surface of the dog’s coat was
white, on the body and head separately, by choosing one of the following responses:
• less than 50%
• between 50 and 75%
PLOS ONE | https://doi.org/10.1371/journal.pone.0230651 September 4, 2020
15 / 38
PLOS ONE Comparison between hearing and/or vision impaired and sensory normal dogs
• more than 75%.
Because all dogs belonged to breeds whose standard coat includes only minor areas of
white, dogs reported as having 50% or more of white were considered as “excess white”.
The results obtained for each group for the body and for the head are presented in Fig 3A
and 3B, respectively. Few sensory normal dogs (frequencies for the HNVN group � 10%) but
most sensory impaired dogs (frequencies for HNVI, HIVN and HIVI groups ranging from
74% to 97%) had excess white coat. There was a non-significant trend for higher frequencies
for the HIVI group than for HNVI and HIVN groups (X2 � 9.75, p � 0.07). Pictures of 88
dogs sent by their owners can be seen in S1 Fig as illustrative examples of the coat colours
most frequently reported for each group.
Iris colour. Owners were asked to indicate whether the colours of the left and right irises
of their dog were either:
• normal for the breed standard (e.g., brown, green, deep blue)
• discoloured to extreme light blue
• indiscernible (due to absence of eyeball, covering by eyelid or membrane, etc).
The frequencies of dogs with discoloured or indiscernible iris were assessed regardless of
whether the “discoloured” or “indiscernible” response was selected for one or for both eyes.
The results obtained for each group are provided in Fig 3C. Few sensory normal dogs (fre-
quency = 12%) but most sensory impaired dogs (frequencies > 80%) had discoloured or indis-
cernible iris(es). Frequencies were similar for the two vision impaired groups (91% and 96%,
respectively; X2 = 1.20; p = 1.0), and were slightly lower for HIVN group (81%; comparison
with HIVI group: X2 = 13.16, p = 0.01; comparison with HNVI group: X2 = 1.32, p = 1.0, ns).
Ophthalmic abnormalities. Owners were asked to indicate whether their dog had, at the
left and right eyes separately, the following ophthalmic abnormalities:
• microphthalmia
• misshapen pupil
• cataract
• absence of eyeball
• other than those mentioned above
• the dog has no, listed or “other”, ophthalmic abnormalities.
Multiple responses were allowed. The list was based on data collected by the second author
from the ophthalmologist veterinarians of 40 presumed double Merle dogs, and was followed
by a field for manual reporting of the type(s) of “other”, non-listed, ophthalmic abnormalities.
The frequencies of dogs having at least one ophthalmic abnormality, regardless of whether
this was at one or both eyes, are presented for each group in Fig 3D. Ophthalmic abnormalities
were seldom for the HNVN group (frequency = 8%) but were extremely frequent for vision
impaired groups (frequencies for HNVI and HIVI groups = 83% and 91%, respectively). There
was no statistical difference between the two vision impaired groups (X2 = 1.30, p = 1.0). Com-
pared to vision impaired groups, the HIVN group showed ophthalmic abnormalities to a sig-
nificantly smaller frequency (30%; X2 � 18.75, p � 0.0005).
Fig 4 shows the frequencies at which each ophthalmic abnormality was reported for sensory
impaired groups. There was no difference between the two vision impaired groups (HNVI vs.
HIVI: X2 � 1.45, p = 1.0). For these two groups, the most frequent abnormality was
PLOS ONE | https://doi.org/10.1371/journal.pone.0230651 September 4, 2020
16 / 38
PLOS ONE Comparison between hearing and/or vision impaired and sensory normal dogs
Fig 3. Frequencies of responses, in percentages, obtained for each group for the following morphological data: (a) excess white coat on body, (b) excess white
coat on head, (c) discoloured or indiscernible iris, (d) ophthalmic abnormalities. HNVI = hearing normal vision impaired (grey), HIVN = hearing impaired
vision normal (orange), HIVI = hearing impaired vision impaired (red), HNVN = hearing normal vision normal (green), ns = non significant. Horizontal
brackets show two-by-two comparisons assessed using Chi2 tests.
https://doi.org/10.1371/journal.pone.0230651.g003
microphthalmia (frequencies = 70% and 64%, respectively), followed from afar by misshapen
pupil, cataract, absence of eyeball, and “other” (frequencies ranging from 12% to 27%). The
HIVN group significantly differed from either one or both vision impaired groups in the
PLOS ONE | https://doi.org/10.1371/journal.pone.0230651 September 4, 2020
17 / 38
PLOS ONE Comparison between hearing and/or vision impaired and sensory normal dogs
responses relative to microphthalmia (X2 � 29.39, p � 0.0001), cataract (X2 � 8.54, p � 0.02)
and absence of eyeball (X2 = 9.10, p = 0.03), but not in those relative to misshapen pupil (X2 �
3.94, p = 1.0) and “other” ophthalmic abnormalities (X2 � 4.45, p = 1.0). Table 5 details the
types of “other” ophthalmic abnormalities manually reported by owners.
Possibly responsible genes. All the ophthalmic abnormalities that are depicted in Fig 4
and Table 5 are frequently observed in homozygous Merles [7], but are not associated with the
two other pigment-deletion genes that are possibly present in the breeds under study (piebald,
Irish spotting). Among the 160 vision impaired dogs (23 HNVI, 137 HIVI), 131 (82%) had
excess white heads, discoloured or indiscernible iris(es), and ophthalmic abnormalities. Thus,
we suggest that these 131 dogs were likely double Merles, although few of them have been
directly tested as double Merles on the M locus (9), or at least bred from two parents with
Merle phenotype according to their owners (38).
Health troubles
Owners were asked to indicate whether their dog had ever suffered from the following type(s)
of health trouble:
• neurological (e.g., seizure, epilepsy, etc.)
• heart (e.g., heart murmur, malformation, etc.)
• bones/joints (e.g., dysplasia, etc.)
• skin
• digestive
• other than those mentioned above
• the dog has never suffered from any, listed or “other”, health troubles.
Multiple responses were allowed. The list was based on both assumptions on the poor
health of double Merles (see Introduction) and unpublished data from a survey of 110 pre-
sumed double Merle’s owners collected by the second author. The list was followed by a field
for manual reporting of “other”, non-listed, troubles. To note, assumptions predict that double
Merles also have issues in their reproductive systems [5]. This point has not been investigated
in the present study because many excess white dogs with congenital sensory impairments are
neutered early so as to avoid at-risk breeding.
Fig 5A presents the frequencies of dogs, for each group and for the IMP cohort, with no
health trouble reported. These dogs are labelled below as “healthy”. Fig 5B–5G present the fre-
quencies at which the different types of heath troubles were reported for the remaining dogs.
Healthy dogs. Seventy-five percent of the sensory normal dogs showed no health trouble
according to their owners and were therefore considered as healthy (see Fig 5A). Similar
results have been previously reported for comparable breeds in a large-scale survey (e.g., 65%
to 75% of “unaffected” dogs within groups of 1,005 Border Collies, 360 Shetland Sheepdogs,
and 785 Dachshunds; frequency of unaffected dogs within the group of 71 Australian Shep-
herds not provided; [23]). Fifty-nine percent of the sensory impaired dogs were healthy, which
significantly differs from sensory normal dogs (X2 = 11.86, p = 0.02).
Neurological troubles. Fig 5B shows the frequencies of neurological troubles reported by
owners for each group and for the IMP cohort. Overall, neurological troubles were reported
for 6.8% of the entire sample. This percentage is substantially higher than those reported in a
past survey of about 43,000 dog owners for a large number of diseases of the nervous system
(prevalence � 1%, [23]). However, data comparison between the two studies is rendered
PLOS ONE | https://doi.org/10.1371/journal.pone.0230651 September 4, 2020
18 / 38
PLOS ONE Comparison between hearing and/or vision impaired and sensory normal dogs
Fig 4. Frequencies of responses, in percentages, obtained for each sensory impaired group for the following ophthalmic abnormalities: (a) microphthalmia, (b)
misshapen pupil, (c) cataract, (d) absence of eyeball, and (e) other. HNVI = hearing normal vision impaired (grey), HIVN = hearing impaired vision normal
(orange), HIVI = hearing impaired vision impaired (red), ns = non significant. Horizontal brackets show the two-by-two comparisons that were statistically
assessed using Chi2 tests.
https://doi.org/10.1371/journal.pone.0230651.g004
PLOS ONE | https://doi.org/10.1371/journal.pone.0230651 September 4, 2020
19 / 38
PLOS ONE Comparison between hearing and/or vision impaired and sensory normal dogs
Table 5. Raw number of dogs from sensory impaired groups obtained for each “other” ophthalmic abnormalities
as manually reported by owners.
HNVI
HIVN
HIVI
coloboma
corectopia
detached retina
dropped or fixated pupil
entropion or ectropion
glaucoma
strabismus
unspecified
--
--
--
--
--
--
--
3
1
1
--
--
--
--
1
3
3
3
3
5
8
3
--
10
Abnormalities are listed in alphabetic order. The “unspecified” response was for owners who responded "other
ophthalmic abnormality than those mentioned above" but without specifying the exact type of the abnormality(s).
The “--” symbol means that no dog was concerned.
https://doi.org/10.1371/journal.pone.0230651.t005
difficult by several differences. First, Australian Shepherds and Border Collies represented less
than 3% of Wiles and colleagues’ sample, while they represented 82% of the present sample.
Second, the survey by Wiles and colleagues exclusively revolved around health, and listed
more than 700 specific diseases. In the present study, the section on health troubles was only
one of a seven-section questionnaire, and listed only six main categories of health troubles.
The two studies had very distinct goals. The study by Wiles and colleagues aimed to quantify
the prevalence, across and within breeds, of a variety of specific diseases in the general popula-
tion of domestic dogs. The small health section in the present study was designed only to assess
the veracity of the following assumption: excess white dogs, particularly double Merles, suffer
from severe neurological, heart and bones/joints troubles. The present study is therefore not
further compared below to that by Wiles and colleagues.
If both the above-mentioned assumption and our suggestion that at least 131 of the sensory
impaired dogs in our sample were double Merles were true, then the frequency of neurological
troubles reported for sensory impaired dogs (HNVI = 9%, HIVN = 2%, HIVI = 15%,
IMP = 11%) should have been higher. However, neurological troubles were significantly less
frequently reported for sensory normal dogs (3%) than for sensory impaired ones (X2 = 11.07,
p = 0.04). Whether–and to the extent to which–this difference is related to the double Merle
genotype, as suggested by the assumption, is undetermined. On the one hand, reports of neu-
rological troubles mainly concerned vision impaired, possibly double Merle, dogs. Among the
131 dogs presumed above to be double Merles according to their morphological data, 19
(14.5%) had neurological troubles. On the other hand, only two of the 16 sensory impaired
dogs that have been tested on the M locus as double Merles showed neurological troubles.
Below, we propose two possible complementary explanations as to why neurological troubles
were more frequently reported for sensory impaired dogs than for sensory normal ones.
Undiagnosed MDR1-related drug sensitivity. All except three of the 31 dogs (25 sensory
impaired, 6 sensory normal) for which neurological troubles were reported were Australian
Shepherds, Border Collies or Rough Collies. As mentioned in the Introduction, both primary
idiopathic epilepsy and mutation of the MDR1 gene are frequent in these breeds [14, 15]. The
MDR1 mutation prevents the blood-brain barrier from blocking chemical agents at the
entrance of the central nervous system. As a result, commonly administered drugs (including
antibiotics, anti-diarrheal, parasite control products, pain medications, sedatives and tranquili-
sers) that rouse no deleterious reaction in dogs with normal MDR1 elicit severe neurological
symptoms (i.e., seizure, tremors, disorientation) in dogs with mutated MDR1.
PLOS ONE | https://doi.org/10.1371/journal.pone.0230651 September 4, 2020
20 / 38
PLOS ONE Comparison between hearing and/or vision impaired and sensory normal dogs
Fig 5. Frequencies of responses, in percentages, obtained for each group and for the IMP cohort for the following health troubles: (a) none, (b) neurological,
(c) heart, (d) bones/joints, (e) skin, (f) digestive, and (g) other. The ordinate width is larger in panel (a) than in panels (b) to (g). HNVI = hearing normal vision
impaired (grey), HIVN = hearing impaired vision normal (orange), HIVI = hearing impaired vision impaired (red), HNVN = hearing normal vision normal
(green), IMP = impaired (HNVI, HIVN and HIVI gathered, purple), ns = not significant. Horizontal brackets show comparisons between HNVN and IPM
assessed using Chi2 tests.
https://doi.org/10.1371/journal.pone.0230651.g005
PLOS ONE | https://doi.org/10.1371/journal.pone.0230651 September 4, 2020
21 / 38
PLOS ONE Comparison between hearing and/or vision impaired and sensory normal dogs
Owners were asked to indicate whether their dogs had been tested for the MDR1 gene, and,
if so, whether the result indicated either normal or–heterozygous or homozygous–mutated
allele(s). The frequency of dogs tested for MDR1 was low in the entire sample (26%), and was
significantly lower for sensory impaired than for sensory normal dogs (14% and 38%, respec-
tively; X2 > 50, p < 0.00001). The sensory impaired and sensory normal dogs that have been
MDR1 tested showed statistically similar frequencies of MDR1 mutation, and hence of drug
sensitivity (19% and 31%, respectively; X2 = 1.60, p = 1.0). The smaller frequency of MDR1
testing for sensory impaired dogs that possibly have ophthalmic abnormalities, sensitivity of
the skin and eyes to UVs, etc., could be explained by the numerous veterinary exams (sensory
impairment diagnosis, ophthalmological tests, etc.) and specific equipment (sunglasses, vibrat-
ing collar, etc.) that their owners and rescue centres already incur. Testing these dogs for the
MDR1 mutation could possibly be considered as being of secondary importance. Veterinari-
ans from the countries investigated have access to lists of both “highly risky” drugs and thresh-
old dosages of “moderately risky” drugs for herding dogs that have not been tested for MDR1.
However, there is some empirical evidence of unexpected neurological reactions to agents/dos-
ages that are not included in these lists.
In other words, we suggest that the greater report of neurological troubles for sensory
impaired dogs than for sensory normal ones could be partially accounted for by their lower
frequency of MDR1 testing, and hence by a greater risk of “missing” their drug sensitivity.
Accordingly, Table 6 presents summary data for the 25 sensory impaired dogs for which neu-
rological troubles were reported. Only three of them have been MDR1 tested.
Undiagnosed compulsive behaviours. During the last three years, the second author has reg-
ularly followed the 40 excess white, congenitally sensory impaired dogs that have been rescued
by and adopted from her organisation. She has observed behavioural stereotypes, often
referred to as obsessive compulsive disorders (OCDs), in the majority of the dogs followed.
For example, several dogs exhibited compulsive spinning, circling, tail chasing, star gazing,
excessive barking, etc (see examples in first part of S1 Video). We suggest that when these
types of behaviours are verbally described by owners to veterinarians or dog trainers/behav-
iourists, they can be considered, in foremost instance, as being possibly symptomatic of a neu-
rological disorder. Accordingly, the first part of S1 Video shows the compulsive behaviours of
two sensory impaired dogs. Both dogs were foremost considered as exhibiting neurological
signs, which has finally been refuted by adequate medical screening. More importantly, both
dogs showed no more compulsive behaviour after behavioural adjustments of their owners to
their sensory impairments, as instructed by the second author (see second part of S1 Video).
In other words, we suggest that OCDs are frequent in sensory impaired dogs, and that, if undi-
agnosed, these OCDs can be considered as neurological troubles. Accordingly, it can be seen
in Table 6 that, in the present study, OCDs have been diagnosed (see details in “Behavioural
troubles” section below) in only six of the 25 sensory impaired dogs for which neurological
troubles were reported.
Heart and bones/joints troubles. Excess white, double Merle dogs are assumed to also
frequently suffer from cardiac and skeletal troubles [5]. Fig 5C and 5D indicate that reports of
heart and bones/joints troubles were low, and were statistically similar for sensory impaired
and sensory normal cohorts (heart = 5% and 1%, respectively, X2 = 6.17, p = 0.45; bones/
joints = 4% and 8%, respectively, X2 = 2.85, p = 1.0). Results for the vision impaired groups,
that include 131 presumed double Merles, are much lower than those expected from the
assumption (frequencies of heart and bones/joints troubles ranging from 0 to 7%).
Skin, digestive and other troubles. Frequencies of skin, digestive and “other” health trou-
bles reported are presented in Fig 5E, 5f and 5g, respectively. These frequencies, ranging from
4 to 14%, were statistically similar for sensory impaired and sensory normal cohorts (skin: X2
PLOS ONE | https://doi.org/10.1371/journal.pone.0230651 September 4, 2020
22 / 38
PLOS ONE Table 6. Summary data for the 25 sensory impaired dogs for which neurological troubles were reported by their owners.
Group
Breed Age (yrs) Excess white body/head Abn. iris(es) colour Opht. abn. Objective indication for double Merle MDR1 tested OCD diagnosed
Comparison between hearing and/or vision impaired and sensory normal dogs
HNVI
HNVI
HIVN
HIVN
HIVI
HIVI
HIVI
HIVI
HIVI
HIVI
HIVI
HIVI
HIVI
HIVI
HIVI
HIVI
HIVI
HIVI
HIVI
HIVI
AS
CHI
AS
BC
AS
GD
RC
AS
AS
BC
AS
AS
AS
AS
BC
AS
AS x?
AS
AS
AS
HIVI AS x BC
HIVI
CAT
HIVI AS x BC
HIVI
HIVI
BC x?
BC
5.9
2.4
6.6
1.1
6.7
3.8
7.6
2.0
9.6
4.1
6.3
9.5
1.6
1.0
2.5
3.2
1.3
1.0
1.9
7.7
1.2
2.9
4.0
2.2
3.0
Y/Y
Y/Y
Y/Y
Y/Y
Y/Y
Y/Y
Y/Y
Y/Y
Y/Y
Y/Y
N/N
Y/N
Y/Y
Y/Y
Y/Y
Y/Y
Y/Y
Y/Y
Y/Y
Y/Y
N/Y
Y/Y
Y/Y
Y/Y
Y/Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
N
N
N
Y
Y
Y
Y
Y
Y
Y
Y
Y
N
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
?
2 parents are Merle
?
Dog is M/M
?
2 parents are Merle
?
?
Dog is M/M
?
?
?
?
?
?
?
?
?
?
?
?
?
2 parents are Merle
?
?
N
N
N
Y neg
N
N
N
N
N
N
N
N
N
N
N
N
N
Y neg
N
N
N
N
N
N
Y neg
N
N
N
Y
N
N
N
N
N
N
Y
Y
Y
N
Y
N
N
N
Y
N
N
N
N
N
N
Abn = abnormal/abnormalities. Oph = ophthalmic. AS = Australian Shepherd. BC = Border Collie. CHI = Chihuahua. GD = Great Dane. RC = Rough Collie. ? =
unknown. Y = yes. N = no. neg = negative (i.e., no drug sensitivity related to MDR1 mutation).
Data that are incompatible with the hypothesis that the dog is double Merle are underlined. Data that are incompatible with our hypothesis that neurological troubles
have been confounded with undiagnosed MDR1 drug sensitivity or undiagnosed OCDs are embolded.
https://doi.org/10.1371/journal.pone.0230651.t006
= 8.03, p = 0.17; digestive: X2 = 4.15, p = 1.0; other health troubles: X2 = 2.89, p = 1.0), and con-
firmed the unpublished results from a survey of 110 owners of excess white, sensory impaired
dogs. Table 7 details the “other” troubles as manually reported by owners. Sensory impaired
and sensory normal dogs mainly differed in allergies. However, neither the causes nor the
symptoms of these allergies were specified by owners.
Behavioural troubles
Owners were asked to indicate whether their dog had ever suffered from the following beha-
vioural troubles:
• aggressiveness
• anxiety, including separation anxiety
• attention deficit/hyperactivity disorder (ADHD)
• obsessive compulsive disorder (OCD)
PLOS ONE | https://doi.org/10.1371/journal.pone.0230651 September 4, 2020
23 / 38
PLOS ONE Comparison between hearing and/or vision impaired and sensory normal dogs
Table 7. Raw numbers of IMP and HNVN dogs for each “other” (non-listed) health trouble as manually reported
by owners.
allergy
breath trouble
hormonal trouble
keratoconjunctivitis
leishmaniasis
runt
splenectomy
urinary incontinence
urinary stones
uveitis
vaccinosis
IMP
24
3
1
--
--
1
1
--
--
1
1
HNVN
12
--
1
1
1
--
--
2
1
--
--
Troubles are sorted in alphabetic order. The “--” symbol means that no dog was concerned.
https://doi.org/10.1371/journal.pone.0230651.t007
• other than those mentioned above
• the dog has never suffered from any, listed or “other”, behavioural troubles.
Multiple responses were allowed. This list was based on:
• the common assumption that deaf and/or blind dogs frequently exhibit aggressiveness and
anxiety (see Introduction)
• observations of 40 sensory impaired dogs by the second author during three years (see
“Undiagnosed compulsive behaviours” section above)
• informal discussions between the two authors and dog trainers, veterinarians and behaviour-
ists about the behavioural troubles that are frequently observed in Australian Shepherds and
Border Collies with insufficient or inadequate activities and interactions.
The list was followed by a field for manual reporting of “other”, non-listed, troubles.
Fig 6A presents the frequencies of dogs, for each group and for the IMP cohort, with no
behavioural trouble reported. Significantly more sensory normal than sensory impaired dogs
were reported as having no behavioural troubles (frequencies = 65% and 48%, respectively; X2
= 13.64, p = 0.01). Fig 6B–6F present the frequencies at which the different behavioural trou-
bles were reported for the remaining dogs.
Aggressiveness was likewise seldom for sensory normal and sensory impaired cohorts (fre-
quencies = 7% and 12%, respectively; X2 = 2.93, p = 1.0; see Fig 6B), which is opposite to the
above-mentioned assumption. There is only one past study that we are aware of that compared
behavioural troubles in sensory impaired and sensory normal dogs [22]. As for the present
study, the authors conducted an owner survey. However, there are four main differences
between the study by Farmer-Dougan and colleagues and the present one. First, the authors
used a previously existing questionnaire (i.e., Canine Behavioural Assessment and Research
Questionnaire, C-BARQ [30]). Second, their respondents had to quantify the severity or fre-
quency of each behavioural trouble listed using 0–4 scales. Third, the authors had no inclusion
criteria regarding the type of sensory impairment (i.e., congenital or late onset, hereditary or
acquired, sensorineural or conductive). Fourth, their study investigated a much larger variety
of dog breeds (see Table 3 in [22]) than ours. Farmer-Dougan and colleagues found smaller
scores of aggressiveness for sensory impaired than for sensory normal dogs, which differs
PLOS ONE | https://doi.org/10.1371/journal.pone.0230651 September 4, 2020
24 / 38
PLOS ONE Comparison between hearing and/or vision impaired and sensory normal dogs
Fig 6. Frequencies of responses, in percentages, obtained for each group and for the IMP cohort for the following behavioural troubles: (a) none, (b)
aggressiveness, (c) anxiety, (d) obsessive-compulsive disorder (OCD), (e) attention-deficit/hyperactivity (ADHD), and (f) other. The ordinate width is larger in
panel (a) than in panels (b) to (f). HNVI = hearing normal vision impaired (grey), HIVN = hearing impaired vision normal (orange), HIVI = hearing impaired
vision impaired (red), HNVN = hearing normal vision normal (green), IMP = impaired (HNVI, HIVN and HIVI gathered, purple), ns = non significant.
Horizontal brackets show comparisons between HNVN and IPM assessed using Chi2 tests.
https://doi.org/10.1371/journal.pone.0230651.g006
from the present finding of similar frequencies of aggressiveness for both cohorts. However,
both studies refute the above-mentioned assumption.
Anxiety was likewise frequent for sensory normal and sensory impaired cohorts (frequen-
cies = 23% and 31%, respectively; X2 = 3.48, p = 1.0; see Fig 6C). Farmer-Dougan and col-
leagues found lower anxiety scores for sensory impaired than for sensory normal dogs [22].
Both studies thus refute the above-mentioned assumption. However, Farmer-Dougan and col-
leagues assessed the behavioural traits that are listed in the C-BARQ, while we have deter-
mined our list of behavioural troubles from common assumptions, pilot observations, and
PLOS ONE | https://doi.org/10.1371/journal.pone.0230651 September 4, 2020
25 / 38
PLOS ONE Comparison between hearing and/or vision impaired and sensory normal dogs
informal discussions with professionals. The behavioural data of the two studies are therefore
not further compared below. The high prevalence of anxiety in our sensory normal cohort
(23%) is similar to that previously reported in various breeds for three items relative to anxiety
(i.e., separation anxiety, fearfulness and noise sensitivity, see [31]).
Reports of OCDs were seldom for sensory normal dogs (frequency = 4%), but were five
times more frequent for sensory impaired dogs (frequency = 19%; X2 = 26.10, p < 0.0001, see
Fig 6D). This finding is in agreement with both past observations by the second author (see
examples of OCDs in the first part of S1 Video) and our hypothesis that part of the neurologi-
cal troubles reported for sensory impaired dogs could have been confounded with
impairment-related, undiagnosed OCDs.
ADHDs (frequencies = 10% and 13%, respectively; X2 = 64, p = 1.0; see Fig 6E) and “other”
behavioural troubles (frequencies = 2% and 6%, respectively; X2 = 5.51, p = 0.62; see Fig 6F)
were reported at similar frequencies for sensory normal and sensory impaired cohorts.
Table 8 details the “other” behavioural troubles as manually reported by owners. Excessive
barking, as well as certain eating disorders (e.g., pica), can be parts of compulsive behaviours.
Owners who reported behavioural troubles in their dog were asked to indicate who had
“diagnosed” the trouble(s) by choosing one of the following responses:
• a veterinarian specialised in behaviour
• a general veterinarian
• a dog trainer or a dog behaviourist
• the owner of the dog (themselves).
They were also asked whether drugs had been prescribed for this/these trouble(s). The
responses obtained for sensory impaired and sensory normal cohorts are presented in Table 9.
Almost 60% of the behavioural troubles have been “diagnosed” by owners. Behavioural trou-
bles have been otherwise diagnosed by a general veterinarian (27% of sensory impaired dogs)
or a dog trainer/behaviourist (33% of sensory normal dogs). Drugs have been prescribed to
only 15% of the dogs with diagnosed behavioural troubles. In other words, most behavioural
troubles reported were not considered as severe enough to require professional consultation
and/or chemical treatment. Thus, these troubles may have been either treated using a beha-
vioural approach or untreated. To note, the behavioural troubles of many sensory normal dogs
that have been prescribed drugs have not been diagnosed by a veterinarian.
Activities
Leisure and sport activities. Owners were asked to indicate how frequently their dog was
practicing each of the following leisure/sport activities:
Table 8. Raw numbers of IMP and HNVN dogs for each “other” (non-listed) behavioural trouble as manually
reported by owners.
depression
eating disorder
excessive barking
sleep disorder
IMP
--
7
2
5
HNVN
1
3
--
--
Troubles are sorted in alphabetic order. The “--” symbol means that no dog was concerned.
https://doi.org/10.1371/journal.pone.0230651.t008
PLOS ONE | https://doi.org/10.1371/journal.pone.0230651 September 4, 2020
26 / 38
PLOS ONE Comparison between hearing and/or vision impaired and sensory normal dogs
Table 9. Frequencies of responses, in percentage, obtained for IMP and HNVN dogs concerning behavioural trou-
bles: (a) operator of the diagnosis and (b) drugs prescription.
IMP
HNVN
a. Operator of the diagnosis
veterinarian specialised in behaviour
general veterinarian
dog trainer or behaviourist
owner
b. Drugs prescription
yes
5
27
12
56
16
1
7
33
58
15
Frequencies were assessed from the number of dogs for which behavioural troubles were reported by owners.
https://doi.org/10.1371/journal.pone.0230651.t009
• canicross, bikejoring, scootering
• agility
• sheep herding
• dog dancing
• tracking of objects or persons
• frisbee, flyball, treiball
This list included the activities that are mostly practiced worldwide by the breeds under
study, and was followed by an open question that allowed reporting all non-listed activities. To
provide their responses, owners had to select one of the following response choices:
• several times a day
• once a day
• several times a week
• once a week
• every two weeks
• once a month
• less frequently than once a month
• never.
We considered that the dog was practicing the activity under examination for all responses
except “less frequently than once a month” and “never”.
Fig 7A presents the frequencies of dogs, for each group and for the sensory impaired groups
gathered (IMP), for which no–listed or “other”–activity was reported. Fig 7B–7H present the
response frequencies obtained for each activity. Table 10 details the “other” activities as manu-
ally reported by owners.
The frequency of dogs involved in absolutely no canine activity was twice greater for the
sensory impaired cohort than for the sensory normal one (40% vs. 20%, respectively; X2 =
20.10, p = 0.0004). This large difference can easily be explained by the assumption that sensory
impaired dogs are poorly capable of practicing activities, as well as by the fact that many official
competitions in the countries under study have long been inaccessible to dogs that are sensory
PLOS ONE | https://doi.org/10.1371/journal.pone.0230651 September 4, 2020
27 / 38
PLOS ONE Comparison between hearing and/or vision impaired and sensory normal dogs
Fig 7. Frequencies of responses, in percentages, obtained for each group and for the IMP cohort for the following leisure/sport activities: (a) none, (b)
canicross/bikejoring/scootering, (c) agility, (d) sheep herding, (e) dog dancing, (f) tracking of objects or persons, (g) Frisbee/flyball/treiball, and (h) other. The
ordinate width is larger in panel (a) than in panels (b) to (h). HNVI = hearing normal vision impaired (grey), HIVN = hearing impaired vision normal
(orange), HIVI = hearing impaired vision impaired (red), HNVN = hearing normal vision normal (green), IMP = impaired (HNVI, HIVN and HIVI gathered,
purple), ns = non significant. Horizontal brackets show comparisons between HNVN and IPM assessed using Chi2 tests.
https://doi.org/10.1371/journal.pone.0230651.g007
PLOS ONE | https://doi.org/10.1371/journal.pone.0230651 September 4, 2020
28 / 38
PLOS ONE Comparison between hearing and/or vision impaired and sensory normal dogs
Table 10. Raw numbers of IMP and HNVN dogs for each “other” (non-listed) activity as manually reported by
owners.
IMP
HNVN
Barn hunt
Cani-roller
Cani-walk
Dog diving
Hiking
Hoopers
Jumping
Kayak
Lure course
Nosework�
Obedience
Paddle
Paragliding
Parkour
Rally-O
Retrieving�
Ring
Seek�
Skijoring
Sled
Swimming
Trail running
Tricks
2
--
--
1
8
--
1
--
2
5
2
--
--
1
3
1
--
2
1
1
6
4
9
1
1
1
--
--
3
--
1
1
1
13
4
1
1
2
2
--
--
--
--
8
3
2
Activities are sorted in alphabetic order. Activities followed by an asterisk strongly rely on olfactory capacities. The
“--” symbol means that no dog was concerned.
https://doi.org/10.1371/journal.pone.0230651.t010
impaired and/or unregistered in kennel clubs. Accordingly, the following three activities were
significantly more frequently practiced by sensory normal dogs than by sensory impaired
ones: canicross/bikejoring/scootering (frequencies = 24% and 9%, respectively; X2 = 18.45,
p = 0.001), agility (frequencies = 30% and 15%, respectively; X2 = 14.59, p = 0.006) and sheep
herding (frequencies = 13% and 3%, respectively; X2 = 15.31, p = 0.004). Importantly, 58% of
the sensory impaired dogs that practiced no activity, against 36% of the sensory normal dogs
that practiced no activity, exhibited behavioural troubles.
On the other hand, the following four activities were reported at statistically comparable
frequencies for sensory normal and sensory impaired dogs: dog dancing (frequencies = 12%
and 8%, respectively; X2 = 1.48, p = 1.0), tracking (frequencies = 23% and 21%, respectively; X2
= 0.25, p = 1.0), frisbee/flyball/treiball (frequencies = 25% and 16%, respectively; X2 = 5.76,
p = 0.56) and “other” (frequencies = 22% and 17%, respectively; X2 = 1.22, p = 1.0). It is note-
worthy that tracking, the activity that sensory impaired dogs practiced the most, as well as
three other activities listed in Table 10 (i.e., nosework, retrieving, seek), essentially rely on
olfactory capacities.
For each above-listed activity, owners were also asked to indicate the dog’s level in that
activity by choosing one of the following responses:
• not concerned, because response “never” or “less frequently than once a month” given above
PLOS ONE | https://doi.org/10.1371/journal.pone.0230651 September 4, 2020
29 / 38
PLOS ONE Comparison between hearing and/or vision impaired and sensory normal dogs
• just for fun, at home or during walks
• beginner in a club
• intermediate in a club
• experienced in a club
• competition/championship.
Table 11 shows the frequencies of “high level” responses (i.e., responses “experienced” and
“competition/championship” gathered) obtained for sensory impaired and sensory normal
cohorts. No general pattern emerges from these data. Compared to those for sensory normal
dogs, frequencies of high-level responses for sensory impaired dogs were lower for agility, dog
dancing and tracking, were conversely higher for frisbee, and were similar for canicross and
sheep herding.
Assistance and therapy activities. Owners were asked to indicate whether their dog was
involved in assistance/therapy activities with:
• elderly persons or groups
• a blind person
• a diabetic or epileptic person, with the role of detecting crises and alerting.
Responses indicated that no dog was engaged in activities with blind persons. Eight percent
of the sensory impaired dogs, against 4% of the sensory normal dogs, were involved in ther-
apy/assistance activities with elderly persons or groups. Only two dogs in the entire sample,
both being hearing and vision impaired (HIVI), were involved with diabetic/epileptic persons.
Accordingly, in the study by Farmer and colleagues, about 3% of 183 hearing and/or vision
impaired dogs had therapy or working–rather than family pet–roles at home [22]. It is note-
worthy that the ability of assistance dogs to detect epileptic and diabetic crises is based on their
ability to perceive small variations in the chemical signals produced by the human’s body, and
thus on their olfactory capacities.
Interspecific communication
Dogs with congenital hearing and/or vision impairments are often believed to have poorer
abilities to communicate with congeners and humans, and to be less “talkative”, compared to
sensory normal dogs. The present study focused on communication with humans, provided
the various social and medical roles that dogs are acknowledged to play in working activities
with humans. We investigated two aspects of dog-human communication: vocalisations
Table 11. Frequencies, in percentages, of HNVN and IMP dogs for which the level in the activity practiced was
reported as being either “experienced” or “competition/championship”.
Canicross
Agility
Sheep herding
Dog dancing
Tracking
Frisbee/flyball/treiball
HNVN
6
30
14
12
10
8
IMP
5
18
14
6
6
12
Frequencies were assessed from the number of dogs from each group that practice the activity at a minimum
frequency of once a month.
https://doi.org/10.1371/journal.pone.0230651.t011
PLOS ONE | https://doi.org/10.1371/journal.pone.0230651 September 4, 2020
30 / 38
PLOS ONE Comparison between hearing and/or vision impaired and sensory normal dogs
addressed by the dog to the owner during interactions, and communication/training signs
addressed by the owner to the dog.
Dog vocalisations. Owners were asked to answer to the following question: “Is your dog
talkative with you? In other words, which of the following vocalisations does your dog fre-
quently produce in order to communicate with you?
• barks
• whines, whimpers, moans
• yelps, yaps
• growls, grunts
• other than those mentioned above
• “your dog never produces any vocalisation during interactions with you”.”
Multiple responses were allowed. This list of canine vocalisations was based on literature
(see review in chapter on communication in [21]), and was followed by a field for manual
reporting of “other”, non-listed, vocalisations.
The responses “no vocalisation” (9 and 4%, respectively, of sensory normal and sensory
impaired cohorts) and “other” (4% of both sensory normal and sensory impaired cohorts)
were infrequently chosen. Fig 8 shows the response frequencies obtained for each vocalisation
listed. Whines/whimpers/moans (frequencies = 57 to 61%) and yelps/yaps (frequencies = 39 to
48%) were reported at similar frequencies for all groups. However, barks (frequencies for
HNVI, HIVN, HIVI and HNVN groups = 74, 90, 85 and 62%, respectively) and growls/grunts
(frequencies = 43, 60, 46 and 30%, respectively) were significantly more frequently reported
for hearing impaired dogs than for sensory normal ones (X2 � 18.58, p � 0.001). One excep-
tion to this is noted for the non-significant difference between HIVI and HNVN groups in
growls/grunts (X2 = 8.79 p = 0.12). The two hearing impaired groups did not statistically differ
from the HNVI group (X2 � 3.85, p = 1.0). Thus, the present results are partially opposite to
the assumption that congenitally deaf dogs are less “talkative” than normal hearing ones just
because they could not benefit from auditory-based vocal learning during early ontogenesis.
Human signs. There are four main types of signs that humans can use to communicate
with, and train, dogs:
• Gesture, which includes arm, hand, finger or object position and movement, as well as hand
sign language
• Sounds, which includes natural and artificial sounds, such as voice, whistle, clicker, etc
• Touch, which includes direct touch of the dog’s body with the hand or a stick, remote-con-
trolled vibrating collars, etc
• Odours, which includes all odour sources that are manipulated by owners for interactions
with their dogs, such as smelling boxes, food pieces, clothes, etc.
Owners were asked to indicate which sign(s) they used with their dogs by choosing one
response within a long list of unique signs, and combinations of two, three and four above-
listed signs (see S2 Fig). Fig 9 shows the responses obtained for each group. For HNVN dogs,
the most frequent response was for the “classical” combination of gesture and sounds (fre-
quency = 62%), followed from afar by the combination of all four signs (frequency = 32%). For
HNVI dogs, the most frequent response was for sounds only (frequency = 48%), followed by
the combination of all four signs (frequency = 26%). For HIVN dogs, the most frequent
PLOS ONE | https://doi.org/10.1371/journal.pone.0230651 September 4, 2020
31 / 38
PLOS ONE Comparison between hearing and/or vision impaired and sensory normal dogs
Fig 8. Frequencies of responses, in percentages, obtained for each group for the following dog vocalisations: (a) barks, (b) whines, whimpers, moans, (c) yelps,
yaps, and (d) growls, grunts. HNVI = hearing normal vision impaired (grey), HIVN = hearing impaired vision normal (orange), HIVI = hearing impaired
vision impaired (red), HNVN = hearing normal vision normal (green), ns = non significant. Brackets show the two-by-two comparisons that were assessed
using Chi2 tests following visual inspection of the data.
https://doi.org/10.1371/journal.pone.0230651.g008
response was for gesture only (frequency = 63%), followed from afar by the combination of
gesture and touch (frequency = 22%). For HIVI dogs, responses were distributed between the
touch and odour combination (frequency = 38%), gesture only (23%), touch only (13%), the
PLOS ONE | https://doi.org/10.1371/journal.pone.0230651 September 4, 2020
32 / 38
PLOS ONE Comparison between hearing and/or vision impaired and sensory normal dogs
Fig 9. For each group (panels), frequencies at which different unique signs (left) and combinations of signs (right) were used by
owners to communicate with their dogs. HNVN = hearing normal vision normal (green), HNVI = hearing normal vision impaired
(grey), HIVN = hearing impaired vision normal (orange), HIVI = hearing impaired vision impaired (red).
https://doi.org/10.1371/journal.pone.0230651.g009
gesture and touch combination (12%), and the combination of all four signs (9%). In sum-
mary, “preferred” signs clearly emerged for dogs with either no or one sensory impairment,
PLOS ONE | https://doi.org/10.1371/journal.pone.0230651 September 4, 2020
33 / 38
PLOS ONE Comparison between hearing and/or vision impaired and sensory normal dogs
but not for dogs with both hearing and vision impairments. Gesture, either alone or in combina-
tion with another sign, was almost never used by owners of HNVI dogs, in spite of the large num-
ber of dogs with residual vision (see right panel in Fig 2A). Odours were almost exclusively used
by owners of HIVI dogs, in combination with touch. Thus, odours were almost never used by
owners of HNVI and HIVN dogs as communication/training signals, in spite of the different
olfaction-based activities in which many of these sensory impaired dogs were involved.
In other words, owners are able to adapt their communication and training signs to the sen-
sory impairment(s) of their dogs. Similar conclusions have previously been drawn from the
results of an owner survey [22]. Overall, these findings are opposite to the assumption that
bilateral communication between sensory impaired dogs and their human caregivers is poor.
Summary and conclusions
In this study, we addressed online an international questionnaire to owners of dogs with either
no or congenital hearing and/or vision impairments. Our main goal was to gain insight on the
veracity of various popular assumptions concerning congenitally sensory impaired dogs, that
often have dramatic consequences on the future of these dogs. According to these assump-
tions, congenitally hearing and/or vision impaired dogs frequently exhibit health and beha-
vioural troubles, and are poorly capable of communicating and practicing activities. In
addition, we aimed to examine both the tools used for determination, and the possible genetic
causes, of the sensory impairments. The main findings are summarized below.
Demographics, morphology and sensory impairments
The two dog cohorts (223 sensory impaired vs. 217 sensory normal) were well matched in age,
lifetime with owner, breed, and sex. Most breeds were from the herding group. All breeds were
possibly concerned by three genes (Merle, piebald and Irish spotting) whose mutations are
known to produce pigment deletion in hairs and irises and congenital hearing impairments
(plus, for Merle, ophthalmic abnormalities associated with vision impairments). Most sensory
impaired dogs showed excess white coat and discoloured iris(es). Ophthalmic abnormalities
were mainly reported for vision impaired dogs. Thus, most sensory impairments investigated
had a pigment-deletion genetic basis, and most vision impaired dogs were likely double
Merles.
Hearing impairments were seldomly diagnosed using objective testing (BAER). The num-
ber of BAER testing sites seems insufficient compared to the growing number of hearing
impaired dogs. Also, most BAER tests are not designed to evaluate partial impairments in one
ear. For the numerous dogs that had no access to BAER testing, subjective testing of hearing
never fulfilled the following three criteria: monaural testing with total occlusion of one ear,
presentation of different sounds with various spectral characteristics and levels, and absence of
non-auditory (visual and nearfield/floor vibration) cues. We conclude that currently available
tests of hearing in dogs have poor capacities to accurately distinguish unilateral from bilateral
and partial from total hearing impairments in the meantime, in reliable “laboratory” condi-
tions. Vision impairments were almost equally diagnosed using objective (CERF-like) testing,
subjective testing, and abnormal aspect of the eye. Most vision impaired dogs were reported as
having residual vision. However, many congenital ophthalmic abnormalities in double Merles
are susceptible to worsen over age, and hence to result in a growing, or even late onset, impact
on vision.
Further research is needed to quantify the exact prevalence of excess-white coat, ophthalmic
abnormalities, and objectively-assessed hearing and vision impairments, within a large sample
PLOS ONE | https://doi.org/10.1371/journal.pone.0230651 September 4, 2020
34 / 38
PLOS ONE Comparison between hearing and/or vision impaired and sensory normal dogs
of dogs of various breeds and ages that have all been tested as homozygous for Merle and non-
carriers for piebald (as no genetic test is yet available for Irish spotting).
Health and behavioural troubles
Contrary to the common assumption [5], the sensory impaired cohort, including 131 “pre-
sumed” double Merles, did not exhibit more heart, skeletal, skin, digestive or other health trou-
bles than the sensory normal cohort. Fewer sensory impaired than sensory normal dogs have
been tested for the MDR1 gene, although the mutation of this gene, frequent in the herding
breeds under study, can elicit epileptiform signs resulting from abnormal sensitivity to com-
mon drugs. Contrary to another popular assumption, but in agreement with a past study [22],
sensory impaired dogs did not exhibit more aggressiveness, anxiety, ADHD or other beha-
vioural troubles than sensory normal dogs. However, OCDs, that can mimic neurological
signs, were much more frequently reported in sensory impaired dogs.
The only present finding that is not opposite to popular assumptions is the greater report of
neurological troubles for sensory impaired dogs, including many presumed double Merles,
than for sensory normal ones. We suggest that this difference may be partially accounted for
by a lack of diagnosis of both MRD1-related drug sensitivity and impairment-related compul-
sive behaviours in sensory impaired dogs. Additionally, primary idiopathic epilepsy is frequent
in the breeds investigated. As a result, the present data to not allow to separate neurological
troubles directly related to the double Merle genotype from those related to the breed (MDR1
mutation, primary idiopathic epilepsy) and from neurological-like behaviours related to the
sensory impairments (OCDs).
Further research is needed to either refute or confirm assumptions on the poor health of
double Merles. The best manner to proceed would be to assess a detailed list of various diseases
in a large number of dogs of various breeds–including breeds with low prevalence of primary
idiopathic epilepsy–and various ages that have all been tested as homozygous for expressing
alleles of Merle, non-carriers for piebald, homozygous normal for MDR1, as well as exam-
ined–and, if diagnosed, treated–for compulsive behaviours.
Activities
It is generally assumed that sensory impaired dogs cannot be safely and efficaciously engaged in
any activity. Accordingly, a total lack of activity was twice more frequently reported for sensory
impaired dogs than for sensory normal ones. Among the “inactive” dogs, behavioural troubles
were more frequently reported in sensory impaired dogs than in sensory normal ones. How-
ever, specific leisure activities were practiced at either smaller or equivalent frequencies/levels
by the two cohorts. Assistance/therapy activities were even more frequently practiced by sensory
impaired dogs. In other words, sensory impaired dogs may be as capable as sensory normal
ones of both practicing and achieving good levels of competence in the activities in which their
owners engage them. Accordingly, an increasing number of competitions, non-competitive
activities and certifications are rendered open to deaf dogs in USA (see list in [19]).
Interspecific communication
The results showed a trend for hearing and vision impaired dogs to produce more barks and
growls/grunts than sensory normal ones during interactions with their owners. This finding is
opposite to the assumptions that congenitally deaf dogs are less “talkative”. However, the pres-
ent study is, to our knowledge, the first attempt to investigate vocalisations in sensory impaired
dogs. We cannot determine whether respondents to our survey actually understood the vocali-
sation terminology used in the questionnaire, whether the vocalisations reported actually had
PLOS ONE | https://doi.org/10.1371/journal.pone.0230651 September 4, 2020
35 / 38
PLOS ONE Comparison between hearing and/or vision impaired and sensory normal dogs
interspecific communication functions, and what emotional valence and arousal had the dif-
ferent vocalisations reported. Also, whether greater barking for sensory impaired dogs is
related to compulsive behavioural troubles is undetermined.
Responses concerning communication and training human signs to dogs showed that own-
ers are capable of adapting their behaviours to the sensory status of their dogs so as to effi-
ciently communicate with, and train, them. Olfaction of sensory impaired dogs was however
less frequently used in communication/training signs than in canine activities.
Implications
Although further research is needed, the results on health and behavioural troubles do not jus-
tify that congenitally sensory impaired puppies are either euthanized right after birth or placed
in “poor” environments. This does not mean, of course, that we encourage at-risk breeding.
Instead, we expect that present and future research will ultimately have beneficial effects on
the future of the numerous sensory impaired puppies that are still born despite the recent
developments of knowledge on canine genetics.
The positive outcomes on canine activities may hopefully encourage more owners to engage
their sensory impaired dogs in activities. Most dogs belonged to herding breeds, for which the
need for regular physical and mental activities to prevent behavioural troubles related to frus-
tration or boredom (e.g., anxiety, ADHD, OCD) has largely been proven. For example, recent
studies have demonstrated the inverse relationship between engagement in activities and beha-
vioural troubles in sensory normal dogs [32, 33]. Thus, greater involvement of sensory
impaired dogs in activities may therefore have the beneficial effect of reducing their beha-
vioural troubles.
Finally, the results indicated that sensory impaired dogs can actually be engaged in both lei-
sure/sport and therapy/assistance cooperative activities that rely on olfactory capacities. There are
numerous studies of olfactory capacities in dogs, due to the important social and medical roles
that these capacities can play for humans (e.g., rescue of missing or enshrouded persons, detection
of cancer cells, explosives and toxic fumes, etc., see review in [21]). However, there is no data on
olfactory capacities in dogs with congenital hearing and/or vision impairments. Brain plasticity
during early ontogenesis could possibly have resulted in overdeveloping their olfactory capacities.
We suggest that sensory impaired dogs should not be excluded from olfaction-based cooperative
activities with humans, because they may exhibit super normal capabilities.
Supporting information
S1 Fig. Pictures of 55 sensory impaired and 33 sensory normal dogs from the present study
illustrating the most typical coat colour patterns. Pictures are sorted by group (HNVI,
HIVN, HIVI, HNVN). Pictures of sensory impaired dogs with lesser white in the coat are
framed in red. Pictures of sensory normal dogs with excess white coat are framed in green.
Each individual picture has been sent by one dog owner to the first author on a voluntary
basis. The portfolio has then been created by the first author for this paper.
(PDF)
S2 Fig. Copy of the online questionnaire.
(JPG)
S1 Video. Compulsive behaviours of two sensory impaired dogs filmed before and after
behavioural adjustments by owners to the sensory impairments of their dogs. The “initial”
compulsive behaviours of these two dogs had been foremost considered as neurological signs
PLOS ONE | https://doi.org/10.1371/journal.pone.0230651 September 4, 2020
36 / 38
PLOS ONE Comparison between hearing and/or vision impaired and sensory normal dogs
prior to medical screening.
(AVI)
Acknowledgments
The authors are grateful to the numerous owners who took time to fill the questionnaire, to
the administrators of social media who shared the calls for participation in the survey, and to
Thierry Legou for his intensive reading of this manuscript.
Author Contributions
Conceptualization: Sophie Savel, Patty Sombe´.
Data curation: Sophie Savel.
Formal analysis: Sophie Savel.
Investigation: Sophie Savel, Patty Sombe´.
Methodology: Sophie Savel.
Software: Sophie Savel.
Supervision: Sophie Savel.
Writing – original draft: Sophie Savel.
Writing – review & editing: Sophie Savel.
References
1. Asher L, Diesel G, Summers JF, McGrevy PD, Collins LM. Inherited defects in pedigree dogs. Part 1:
disorders related to breed standards. Vet J. 2009; 182(3):402–411. https://doi.org/10.1016/j.tvjl.2009.
08.033 PMID: 19836981
2. Strain GM. The genetics of deafness in domestic animals. Front Vet Sci. 2015. 2(29). https://doi.org/10.
13389/fvets.2015.00029.
3.
4.
Langevin M, Synkova H, Jancuskova T, Pekova S. Merle phenotypes in dogs–SILV SINE insertions
from Mc to Mh. PLoS ONE. 2018; 13(9): https://doi.org/10.11371/journal.pone.0198536
Langevin M. Merle—SINE Insertion from Mc—Mh “The Incredible Story of Merle”. 2018. Library and
Archives Canada Cataloguing in Publication. 130 pages. Available from: https://www.merle-sine-
insertion-from-mc-mh.com/order/.
5. Clark LA, Wahl JM, Rees CA, Murphy KE. Retrotransposon insertion in SILV is responsible for merle
patterning of the domestic dog. Proceedings of the National Academy of Sciences of the United States
of America. 2006; 103: 1376–1381. https://doi.org/10.1073/pnas.0506940103 PMID: 16407134
6. Wong AK, Ruhe AL, Robertson KR, Loew ER, Williams DC, Neff MW. A de novo mutation in KIT causes
white spotting in a subpopulation of German Shepherd dogs. Anim Genet. 2012; 44, 305–31 https://
doi.org/10.1111/age.12006 PMID: 23134432
7. Bauer BS, Sandmayer LS, Grahn BH. Diagnostic Ophthalmology. Can Vet J. 2015. 56(7): 767–768.
PMID: 26130844
8. Gwin RM, Wyman M, Lim DJ, Ketring K, Werling K. Multiple ocular defects associated with partial albi-
nism and deafness in the dog. J Am Anim Hosp Assoc. 1980. 17:401–408.
9. Murphy SC, Evans JM, Tsai KL, Clark LA. Length variations within the Merle retrotransposon of canine
PMEL: correlating genotype with phenotype. Mobile DNA. 2018. 9:26. https://doi.org/10.1186/s13100-
018-0131-6 PMID: 30123327.
10. Ballif BC, Ramirez CJ, Carl CR, Sundin K, Krug M, Zahand A, Shaffer LG, Flores-Smith H. The PMEL
Gene and Merle in the Domestic Dog: A Continuum of Insertion Lengths Leads to a Spectrum of Coat
Color Variations in Australian Shepherds and Related Breeds. Cytogenet Genome Res. 2018. https://
doi.org/10.1159/000491408 PMID: 30071510
PLOS ONE | https://doi.org/10.1371/journal.pone.0230651 September 4, 2020
37 / 38
PLOS ONE Comparison between hearing and/or vision impaired and sensory normal dogs
11.
Langevin M. Unraveling the mysteries of merle. Australian Shepherd Club of America. 2019. 12 pages.
Available from: https://merle-sine-insertion-from-mc-mh.com/_files/200000312-8a1128a114/
Unraveling%20the%20Mysteries%20of%20Merle.pdf.
12. McCabe L, Griffin LD, Kinzer A, Chandler M, Beckwith JB, McCabe RB. Overo lethal white foal syn-
drome; equine model of aganglionic megacolon (Hirschsprung disease). Am J Med Genet. 1990.
36:336–340. https://doi.org/10.1002/ajmg.1320360319 PMID: 2363434
13. Hu¨lsmeyer V-I, Fischer A, Mandigers PJJ, DeRisio L, Berendt M, Rusbridge C, Bhatti SFM, Pakozdy A,
Patterson EE, Platt S, Packer RMA, Volk HA. International Veterinary Epilepsy Task Force’s current
understanding of idiopathic epilepsy of genetic or suspected genetic origin in purebred dogs. BMC Vet.
Res. 2015; 11:175. https://doi.org/10.1186/s12917-015-0463-0 PMID: 26316206
14. Geyer J, Do¨ ring B, Godoy JR, Leidolf R, Moritz A, Petzinger E. Frequency of the nt230 (del4) MDR1
mutation in Collies and related dog breeds in Germany. J. Vet. Pharmacology and Therapeutics. 2005;
28(6):545–51. https://doi.org/10.11111/j.1439-0531.1976.tb00313.x
15. Gramer I, Leidolf R, Do¨ring B, Klintzsch S, Kra¨ mer E-M, Yalcin E, et al. Breed distribution of the nt230
(del4) MDR1 mutation in dogs. Vet. J. 2011. 189(1): 67–71. https://doi.org/10.1016/j.tvjl.2010.06.012
PMID: 20655253
16. Sorsby A, Davey JB. Ophthalmic associations of dappling or merling in the coat color of dogs. J. Genet.
1954.; 54, 425–44.
17.
Little CC. The Inheritance of Coat Color in Dogs. Howell Book House, New York. 1957.
18. Sponenberg DP, Bowling, AT. Heritable Syndrome of Skeletal Defects in a Family of Australian Shep-
herd dogs. J. Hered. 1985; 76, 393–394. PMID: 4056374
19. Cope Becker S. Living with a deaf dog: A book of training advice, facts and resources about canine
deafness caused by genetics, aging, illness. 2nd ed. Vonore (TN): Susan Cope. 2017. 151 pages.
20.
Ferrara M, Alnan CRE. Congenital structural brain defects in the deaf dalmatian. Vet record. 1983. 112:
344–6. https://doi.org/10.1136/vr.112.15.344 PMID: 6857975
21. Miklo´si A´ . 2015. Dog behaviour, evolution, and cognition. 2nd ed. Oxford (UK): Oxford University
Press.
22.
Farmer-Dougan V, Quick A, Harper K, Schmidt K. Behaviour of hearing or vision impaired and normal
hearing and vision dogs (Canis lupus familiaris): Not the same, but not that different. J Vet Behav. 2014;
9(6); 316–323.
23. Wiles BM, Llewellyn-Zaidi AM, Evans KM, O’Neill DG, Lewis TW. Large-scale survey to estimate the
prevalence of disorders for 192 Kennel Club registered breeds. Canine Genetics and Epidemiology.
2017; 4:8. https://doi.org/10.1186/s40575-017-0047-3 PMID: 28932406
24. Siegel S, Castellan NJ. Non parametric statistics for the behavioral sciences. McGraw-Hill, New York,
1988.
25. Aickin M, Gensler H. Adjusting for multiple testing when reporting research results: the Bonferroni vs Holm
methods. Am J Public Health. 1996; 86:726–8. https://doi.org/10.2105/ajph.86.5.726 PMID: 8629727
26. Gaunet F. How do guide dogs of blind owners and pet dogs of sighted owners (Canis familiaris) ask
their owners for food? Anim Cogn. 2008; 11:475–483. https://doi.org/10.1007/s10071-008-0138-3
PMID: 18278527
27. Strain GM, Clark LA, Wahl JM, Turner AE, Murphy KE. Prevalence of deafness in dogs heterozygous
or homozygous for the Merle allele. J Vet Intern Med. 2009; 23:282–286. https://doi.org/10.1111/j.
1939-1676.2008.0257.x PMID: 19192156
28. Scandurra A, Alterisio A, Di Cosmo A, D’Aniollo B. Behavioural and perceptual differences between
sexes in dogs: An overview. Animals. 2018; 8(151). https://doi.org/10.13390/ani8090151
29. Strain GM. Aetiology, prevalence and diagnosis of deafness in dogs and cats. Br Vet J. 1996. 1523:
17–36.
30. Hsu Y. Serpell, J.A. Development and validation of a questionnaire for measuring behavior and temper-
ament traits in pet dogs. J Am Vet Med Assoc. 2003. 223: 1293–1300. https://doi.org/10.2460/javma.
2003.223.1293 PMID: 14621216
31.
Tiira K, Sulkama S, Lohi H. Prevalence, comorbidity, and behavioral variation in canine anxiety. J. Vet.
Behav. 2016; 36–44. https://doi.org/11016/j.jveb.2016.06.008.
32. Puurunen J, Hakanen E, Salonen MK, Mikkola S, Sulkama S, Araujo C, et al. inadequate socialisation,
inactivity, and urban living environment are associated with social fearfulness in pet dogs. Nature. 2020.
10:3527. https://doi.org/10.1038/s41598-020-60546-w
33. Bennet PC, Rohlf VI. Owner-companion dog interactions: relationships between demographic vari-
ables, potentially problematic behaviours, training engagement and shared activities. Appl. Anim.
Behav. Sci. 2007. 102: 65–84.
PLOS ONE | https://doi.org/10.1371/journal.pone.0230651 September 4, 2020
38 / 38
PLOS ONE |
10.1371_journal.pone.0228975 | RESEARCH ARTICLE
Effect of altitude on wing metric variation of
Aedes aegypti (Diptera: Culicidae) in a region of
the Colombian Central Andes
Luis Mı´guel Leyton Ramos1,2☯, Oscar Alexander Aguirre Obando2☯, Jonny
Edward DuqueID
3, Vı´ctor Hugo Garcı´a-Mercha´ nID
1☯*
1 Grupo de Evolucio´n, Ecologı´a y Conservacio´ n (EECO), Universidad del Quindı´o, Armenia, Quindı´o,
Colombia, 2 Escuela de Investigacio´n en Biomatema´ticas, Universidad del Quindı´o, Armenia, Quindı´o,
Colombia, 3 Centro de Investigaciones en Enfermedades Tropicales – CINTROP, Facultad de Salud,
Escuela de Medicina, Departamento de Ciencias Ba´sicas, Universidad Industrial de Santander, Piedecuesta,
Santander, Colombia
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
☯ These authors contributed equally to this work.
* victorhgarcia@uniquindio.edu.co
Abstract
OPEN ACCESS
Citation: Leyton Ramos LM, Aguirre Obando OA,
Duque JE, Garcı´a-Mercha´n VH (2020) Effect of
altitude on wing metric variation of Aedes aegypti
(Diptera: Culicidae) in a region of the Colombian
Central Andes. PLoS ONE 15(8): e0228975. https://
doi.org/10.1371/journal.pone.0228975
Editor: Olle Terenius, Swedish University of
Agricultural Sciences, SWEDEN
Received: January 24, 2020
Accepted: August 2, 2020
Published: August 20, 2020
Copyright: © 2020 Leyton Ramos et al. This is an
open access article distributed under the terms of
the Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: All relevant data are
within the manuscript and its Supporting
Information files.
Funding: The authors thank office of the vice-
president for research, Universidad del Quindı´o for
funding the Project (Grant 828).
Competing interests: The authors have declared
that no competing interests exist.
In mosquitoes of medical importance, wing shape and size can vary with altitude, an aspect
that can influence dispersion and, consequently, their vector capacity. Using geometric mor-
phometry analysis, Aedes aegypti wing size and shape variation of males and females was
studied in four altitudes in the second-smallest department in Colombia: 1,200 m (Tebaida),
1,400 m (Armenia), 1,500 m (Calarca´ ), and 1,700 m (Filandia). Wing shape in males (P <
0.001) and females (P < 0.001) was significantly different through the altitudinal gradient; in
turn, wing size in males followed the altitudinal gradient males (R2 = 0.04946, P = 0.0002),
females (R2 = 0.0011, P = 0.46). Wing allometry for males (P < 0.001) and females (P <
0.001) was significant. Likewise, the shape and size of the wings of males (P < 0.001) and
females (P < 0.001) had significant fluctuating asymmetry. It is concluded that, in a small
scale with an altitudinal variation of 500 meters, it is detected that the size and shape of the
wings varied in A. aegypti, main vector the agents that cause dengue, chikungunya, and
Zika. The fluctuating asymmetry is present in the individuals studied and could be associ-
ated with environmental effects caused by vector control campaigns present in some sam-
pling locations.
1. Introduction
Aedes (Stegomyia) aegypti (Linnaeus, 1762) is an urban anthropophilic mosquito from Africa,
distributed in the world’s tropical and sub-tropical regions [1]. In the Americas, this mosquito
is present in almost every country, considered the main vector the agents that cause dengue
(DENV), Zika fever (ZIKV), and chikungunya (CHIKV) [2–4]. In Colombia, A. aegypti is reg-
istered in 80% of the country up to 2,300 m [5]. Nevertheless, still unknown is the epidemio-
logical impact altitude exerts on the population dynamics of A. aegypti in areas, like the
PLOS ONE | https://doi.org/10.1371/journal.pone.0228975 August 20, 2020
1 / 15
PLOS ONE Wing metric variation in Aedes aegypti
Andean region. It has been observed that the altitude in the zones where the mosquito inhabits
has a direct impact on the abundance, geographic distribution, vector capacity, epidemiology,
and pathogenicity of the mosquitoes [6]. Additionally, in culicids, the range of altitudinal dis-
tribution may be modified by increased global temperature [7], a phenomenon observed in
the natural populations of A. aegypti of the Americas, including Colombia [5].
In A. aegypti, the size of the individuals has been associated with components of the repro-
ductive success [8, 9]. Bigger A. aegypti individuals (per se, bigger wingspan) could be more
involved in the transmission of arthropod-borne virus (arbovirus), like dengue, than smaller
ones [10]. In addition, bigger individuals have been associated with a higher frequency of feed-
ing from blood in human hosts [11], greater survival, and fertility [12]. On the contrary,
smaller mosquitoes (hence, with smaller wingspan) may have a higher number of feeding
events throughout their lives, which can increase infection levels and arbovirus dissemination
[9, 13, 14]. Furthermore, the biological shape is an outstanding aspect of the phenotype of an
organism and provides a link between the genotype and the environment [15]. The biological
shape has been studied in insects, such as butterflies and fruit flies, with an emphasis on wings,
which are a trait that is associated with carrying capacity and dispersal [16, 17]. In mosquitoes,
wing shape is associated with dispersion capacity [18]. Additionally, wing flapping produces
vibrations that generate sounds [19, 20], which are different and are related with the precopu-
latory behavior [21].
In mosquitoes, it has been noted that temperature (climatic variable inversely proportional
to altitude) causes changes in the life cycle, affecting the body size and shape [22]. An inverse
relationship has been observed in some cases between temperature and the duration of devel-
opment [23]. Consequently, at lower temperatures, the transmission of arbovirus may—in
some cases—be impeded [24]. Hence, knowing how wing shape and size vary in A. aegypti,
with relation to altitude, could contribute useful information for its vector control.
The geometric morphometry permits detecting information patterns on the type, ecological
relationships, and environmental factors associated to populations present in the area [25–27].
In A. aegypti, morphometric analyses on wings have been widely studied to investigate hetero-
geneity and structuring in natural populations [28–31]. However, very few prior studies have
related altitude and wing metric variation in mosquitoes. Studies in northeastern Turkey on
Aedes vexans between 808 and 1,620 m and on Culex theileri between 808 and 2,130 m showed
variation in wing size and shape. Besides, in Culex theileri, a positive correlation was observed
between wing size and altitude [18, 32]. Recently, in Aedes albopictus of Albania, the region
where this Asian mosquito was first registered in Europe, it was observed between 154 and
1,559 m shape, size and sex variations among altitudinal populations of these species [33].
Nevertheless, in Colombia and the rest of the world, the metric variation of A. aegypti and its
relation with the altitudinal gradient has not been studied much [34]. In Colombia, the depart-
ment of Quindı´o is on the central mountain range of the Andes and it is the second smallest in
geographic extension with altitudes in its urban area ranging from 1,200 to 1,917 m in 1,961
km2 [35–37]. Due to the aforementioned, this study sought to explore the wing size and shape
variation of A. aegypti males and females from an altitudinal gradient of the central mountain
range of the Colombian Andes.
2. Methodology
2.1 Field work
The department of Quindı´o is the second smallest regarding land area in Colombia, with an
extension of 1,961 km2, with an altitude range in the urban area from 1.200 (Tebaida) to 1,917
m (Filandia) [35, 36]. Here, for six months, in the rainy season, between August (2017) and
PLOS ONE | https://doi.org/10.1371/journal.pone.0228975 August 20, 2020
2 / 15
PLOS ONE Table 1. Information regarding mosquitoes collected by altitude, locality, geographic coordinates and individual number for sex, as well as some climatic data.
Altitude (m)
Municipality
Mean temperature (˚C)
Precipitation (mm)
Latitude
Longitud
No. Females
No. Males
Wing metric variation in Aedes aegypti
1200
Tebaida
1400
Armenia
1500
Calarca´
1700
Filandia
https://doi.org/10.1371/journal.pone.0228975.t001
21.24
21.24
21.24
21.24
21.24
21.24
21.24
21.24
21.24
20.96
20.96
20.96
20.96
20.96
15.68
20.96
20.96
20.96
15.68
15.68
15.68
15.68
11.90
21.13
21.13
1880
1880
1880
1880
1880
1880
1880
1880
1880
1993
1993
1993
1993
1993
1971
1993
1993
1993
1971
1971
1971
1971
1663
2036
2036
4˚27’22.38"N
75˚47’10.05"W
4˚27’1.40"N
75˚47’5.92"W
4˚26’58.48"N
75˚46’39.23"W
4˚26’55.96"N
75˚47’3.72"W
4˚27’10.83"N
75˚47’24.69"W
4˚26’56.89"N
75˚47’42.89"W
4˚27’18.69"N
75˚47’2.42"W
4˚27’7.48"N
75˚46’16.92"W
4˚27’14.50"N
75˚47’23.59"W
4˚31’4.65"N
4˚32’1.73"N
75˚42’3.98"W
75˚40’46.18"W
4˚31’50.78"N
75˚40’38.28"W
4˚31’4.39"N
75˚41’38.82"W
4˚31’12.39"N
75˚42’28.19"W
4˚34’20.97"N
75˚38’37.14"W
4˚31’47.04"N
75˚41’52.94"W
4˚32’25.20"N
75˚42’18.01"W
4˚31’20.22"N
75˚41’53.29"W
4˚33’7.94"N
4˚31’7.56"N
75˚39’33.72"W
75˚38’39.77"W
4˚30’46.28"N
75˚39’19.90"W
4˚32’21.42"N
75˚38’16.08"W
4˚29’28.54"N
75˚38’39.16"W
4˚42’24.38"N
75˚42’29.10"W
4˚42’25.33"N
75˚42’25.50"W
17
4
6
37
1
12
32
56
23
7
8
7
11
1
40
10
2
10
10
30
45
11
1
14
3
9
5
6
26
8
4
7
22
20
6
13
1
3
2
40
17
1
15
7
10
17
6
0
10
6
398
261
February (2018), adult individuals of A. aegypti were collected in four altitudes of the urban
zone [The taxonomic keys used are described at the end of this item]. The sampling sites were
the following: 1,200 m (Tebaida), 1,400 m (Armenia), 1,500 m (Calarca´), and 1,700 m (Filan-
dia). In Filandia, one of the municipalities with the highest altitude urban settlement in the
department, no mosquitoes were found at altitudes above 1,700 m. Table 1 shows information
regarding mosquitoes collected by altitude, locality, geographic coordinates and sex. For each
altitude and for the rainy season, from historical data available in the Wolrdclim-2 library on
temperature from 1970 to 2000 [38], the mean historic temperature was extracted using the
geographic coordinates. This variable is recognized in the literature as one of the most influen-
tial in the life cycle of A. aegypti [39, 40]. Thereafter, the layers of each altitude were overlapped
with each of the raster layers for temperature. Then, for each altitude and each raster layer of
temperature, the median, minimum, and maximum historical average temperatures were
extracted. The R software version 4.0.1, and raster library [41] were used for these purpose. To
select the sampling sites by altitude, 10 points were randomly selected. Each point corre-
sponded to a neighborhood and in each neighborhood, homes were visited where their dwell-
ers permitted. Each home was visited four times per month and each visit lasted from 45 to 60
minutes. In each home sampled, an informed consent was delivered on the objectives of the
research and the authorization to conduct the sampling.
PLOS ONE | https://doi.org/10.1371/journal.pone.0228975 August 20, 2020
3 / 15
PLOS ONE Wing metric variation in Aedes aegypti
Fig 1. Map with sampling altitudes and total number of males and females of A. aegypti by altitude.
https://doi.org/10.1371/journal.pone.0228975.g001
The adults were collected through mechanical aspiration through an electric aspirator.
After collection, the individuals were sedated and sacrificed with acetone. All the collections
were conducted under the framework permit from the Corporacio´n Autonoma Regional del
Quindı´o (CRQ) N˚ 240 issued for the department of Quindı´o, Colombia. Finally, the speci-
mens were identified at species level by using the taxonomic keys of Forattini [42] and Rueda
[43]. Fig 1 shows the location, as well as the total number of mosquitoes by altitude and sex
used in this work.
2.2 Laboratory work
The left and right wings were removed from each male and female mosquito collected;
removal was from the base, following the protocol described in [44] and [45]. Each wing was
submerged in NaClO solution at 5% to remove scales and rinse them. Thereafter, each wing
was submerged in ethanol solution at 99.5% to remove excess NaClO, to be mounted on a
slide with ethanol at 70%. The photographs were taken on a stereomicroscope (Zeiss Stemi
DV4) with integrated camera (Canon EOS REBEL T3i) with 32X magnification, according to
specifications for taking landmarks (LM) for two dimensions (2D) [46].
Each wing was marked with 22 LM [47] type I (Fig 2A), according to the method by Rohlf
& Slice [26, 48]. The photos were organized randomly with the tpsUtil software version 7.0 to
reduce LM marking bias [49]. All the LM of the wings were marked twice by using the tpsDig2
program [50]. The LM were located by the same operator to reduce human error in taking
points. To guarantee the reproducibility of the experiment, the marking of the LM was carried
out twice. To detect atypical LM, the Morphoj program was used [51].
PLOS ONE | https://doi.org/10.1371/journal.pone.0228975 August 20, 2020
4 / 15
PLOS ONE Wing metric variation in Aedes aegypti
Fig 2. Location of LM, general analysis of procrustes, and comparison among replicas. (A). Anatomical frameworks used. (B) Procrustes coordinates after the
analysis of atypical data. (C) PC1 set of original data. (D) PC1 set of replica data.
https://doi.org/10.1371/journal.pone.0228975.g002
2.3 Data analysis
From the array of data with the coordinates of each LM, the effect of the scale, translation, and
rotation was eliminated through a general Procrustes analysis [48, 52]. Thereafter, the Procrus-
tes coordinates were obtained as representative variable of the wing shape and centroid size
(CS) of the wing size, which were used in all the analyses performed. To guarantee the repro-
ducibility of the data used, the measurement error rate (% ME) was estimated according to
[53, 54], using the variables CS and individual factor to both markings (original and replica).
These variables were obtained the variance components, one through one-way ANOVA. The
individual factor was used as a source of variation between original and replica marking. To
visualize the differences of each LM among the original data and their replica, from the princi-
pal components obtained from Procrustes coordinates, deformation grids were used [55].
Information from the right wing was used to analyze the shape and size variation in the alti-
tudinal gradient for both sexes. The CS for each sex and altitude was evaluated by using the
Kruskal-Wallis non-parametric test. When differences were significant, a pair-wise Mann–
Whitney U-test was used, and it was visualized by using a box diagram and a chart to indicate
the significant differences among the comparisons. To determine the relationship between
size and the altitudinal gradient, a simple linear regression was performed. The allometric
influence of wing size within the shape was analyzed through a multivariate regression of Pro-
crustes coordinates [56] in function of the CS, using a permutations test with 10,000 randomi-
zations [57].
The wing shape variation patterns for each sex were visualized through an analysis of prin-
cipal components analysis (PCA). For each sex, wing shape and its variation in the altitudinal
gradient, a Canonical Variable Analysis (CVA) was conducted. In addition, the PCA used
deformation grids to determine in what LM did the wing shape variation originate. The shape
variation in function of the altitudinal gradient for each sex was evaluated through an analysis
of variance (ANOVA) with 10,000 randomizations [58].
PLOS ONE | https://doi.org/10.1371/journal.pone.0228975 August 20, 2020
5 / 15
PLOS ONE Wing metric variation in Aedes aegypti
With the mosquitoes (Males = 223 and Females = 385) that had both wings in good condi-
tion, symmetry or asymmetry were determined. For this, the study used values of the CS and
the Procrustes coordinates of the original data and their replica, which were analyzed through
a Procrustes ANOVA with 1,000 iterations [59, 60]. All the analyses were performed in the R
programming software, version 4.0.1 [61], using the Geomorph package, version 3.3.1 [62],
dplyr version 0.8.0.1 [63], rcompanion version 2.3.25 [64, 65], and multcompView 0.1–8 [66],
except for the CVA, Mahalanobis distances, and the allometric regression, which were carried
out in the MorphoJ program, version 2.6 [51].
3. Results
Fig 2B shows that the distribution of the LM used is located continuously. Fig 2C and 2D indi-
cate that the set of original data and their replica had no variations by the user to mark the 22
LM (% ME = 1.92). The CS of the wings of males and females had significant differences
(Df = 1, P < 0.05), being higher in females (4.19 CS ± 0.38) than in males (3.29 CS± 0.30); (Fig
3). In turn, the CS variation in function of the altitudinal gradient for wings from males
(Df = 3, P < 0.05) and females (Df = 3, P < 0.05) was statistically significant, indicating the CS
for males, based on the Mann–Whitney test, statistical differences between the altitudes from
1,400 to 1,500 m, 1,200–1,400 m, 1,500–1,700 m and 1,200–1,700 m. The CS of the wings for
females was different between the altitudes 1,400–1,500 m and 1,200–1,400 m Table 2. Fig 4A
suggests that wings of males have a slight tendency to being bigger at higher altitudes,
Fig 3. Box diagram for the centroid size and altitude for wings from both sexes of A. aegypti.
https://doi.org/10.1371/journal.pone.0228975.g003
Table 2. Significant differences for size by sex in the altitudinal gradient, according to the Pairwise Mann–Whit-
ney U-tests.
Sex
Males
Females
1,200 m
a (3.18)
a (4.17)
1,400 m
b (3.37)
b (4.41)
1,500 m
a (3.22)
a (4.10)
1,700 m
b (3.40)
ab (4.24)
Different letters indicate significant differences (p < 0.05), Mean size in parentheses.
https://doi.org/10.1371/journal.pone.0228975.t002
PLOS ONE | https://doi.org/10.1371/journal.pone.0228975 August 20, 2020
6 / 15
PLOS ONE Wing metric variation in Aedes aegypti
Fig 4. Linear regression for centroid size and altitude for both sexes of A. aegypti, Males (A) and Females (B).
https://doi.org/10.1371/journal.pone.0228975.g004
nevertheless, other variables associated with the mosquito’s development must be analyzed to
have better resolution (R2 = 0.04946, P = 0.0002), a pattern not observed for wings of females
(R2 = 0.0011, P = 0.46; (Fig 4B)).
The allometry test indicated that the contribution of CS wing shape variation was signifi-
cant for both sexes (Males P < 0.001 and Females P < 0.001), where the percentage of wing
shape variance explained by the size was 4.9% for males and 2.5% for females.
The PCA in total explained 53.1% of data variation (PC1 = 43.8%, PC2 = 9.3%). The PC1
separated two groupings corresponding to each sex. The wing shape of males was present in
the negative part of PC1 and that of the females in the positive part. In males, wing shape varia-
tion was observed for LM from the wing contour (LM: 1–11, 21, 22) in negative sense to PC1
and internal LM (LM 12–20) in positive sense to PC1. For females, the LM from the wing con-
tour were present in positive sense to PC1 and in contrary sense, the internal LM (Fig 5).
The CVA for each sex showed differences in wing shape among mosquitoes located at dif-
ferent altitudes. In males, CV1 and 2 explain in total 83.2% of the wing shape variation
(CV1 = 62.2%, CV2 = 21%), while in females, this explained 82.8% of the wing shape variation
(CV1 = 60.4%, CV2 = 22.4%). These differences in wing shape were supported by the permuta-
tions test for distances by Mahalanobis for males (P < 0.001) and females (P < 0.001) (Fig 6).
Wing shape in function of the altitudinal gradient for males (F = 3.8251; Df = 3; P = 0.0001)
and females (F = 3.5457; Df = 3 P = 0.0001) were significant Table 3.
The bilateral symmetry test for wing shape indicated no significant variation between the
left and right sides for males (Side P = 0.377) and females (Side P = 0.207) of A. aegypti. On the
contrary, the variation among individuals, side and its interaction, was indeed significant for
males (Individuals P = 0.001, Side�Individual P = 0.001; Table 4 and females (Individuals
P = 0.001, Side�Individual P = 0.001; Table 5. For the CS, the bilateral symmetry test indicated
PLOS ONE | https://doi.org/10.1371/journal.pone.0228975 August 20, 2020
7 / 15
PLOS ONE Wing metric variation in Aedes aegypti
Fig 5. PCA and deformation grid for wing shape between both sexes of A. aegypti.
https://doi.org/10.1371/journal.pone.0228975.g005
variation among individuals, between sides and its interaction for males (Side P = 0.001, Indi-
viduals P = 0.001, Side�Individual P = 0.001) and females (Side P = 0.001, Individuals
P = 0.001, Side�Individual P = 0.001).
4. Discussion
To our knowledge, this is the first work on an altitudinal gradient in the Andean region identi-
fying differences for wing size and shape of A. aegypti males and females in an altitudinal gra-
dient. This is a pattern previously observed in females from Culex theileri [18] and Aedes
vexans [32], West Nile virus and Valley fever vectors, respectively. In C. theileri, it was found
PLOS ONE | https://doi.org/10.1371/journal.pone.0228975 August 20, 2020
8 / 15
PLOS ONE Wing metric variation in Aedes aegypti
Fig 6. CVA for wing shape in the altitudinal gradient: (A) Females. (B) Males.
https://doi.org/10.1371/journal.pone.0228975.g006
that wing size and altitude are correlated positively, while in A. vexans, these differences were
observed for wing size and shape through the altitudinal gradient. Additionally, in C. theileri
and A. vexans, these differences were noted in altitudinal gradients from 808 to 2,130 m (with
a difference of 1,322 m) and 808 to 1,620 m (with a difference of 812 m), respectively. Curi-
ously in our case, said difference was observed within an altitudinal range from 1,200 to 1,700
m with a difference of 500 m. For both species, the differences observed regarding wing shape
or size could be attributed to variables, like relative humidity and temperature. In turn, in A.
aegypti, the wing shape and size variation observed may be due to the influence of temperature.
In this species, under laboratory conditions, it has been noted that larvae subjected to tempera-
tures between 24 and 35 ˚C have generated males and females with larger wing size in temper-
atures from 24 to 25 ˚C, while at temperatures from 34 to 35 ˚C, males and females have been
obtained with smaller wing size [67]. According to Instituto de Hidrología, Meteorología y Estu-
dios Ambientales (IDEAM) and Unidad de Planeación Minero Energética (UPME)) [68], his-
torical means annual temperatures between 1981 and 2010 for altitudes in our study of 1,200,
1,400, 1,500, and 1,700 m are 20–22 ˚C, 20–22 ˚C, 16–20 ˚C, and 20–24 ˚C, respectively. How-
ever, for our case, smaller wings were observed at altitudes between 1,200 and 1,500 m for
males and females. Bigger wings were found females at 1,400 m and for males between 1,400
and 1,700 m, which differs from the study already mentioned. Probably, this disparity may be
Table 3. ANOVA of Procrustes for wing shape in function of the altitudinal gradient for males and females of A. aegypti.
Males
Females
Altitude
Residuals
Total
Altitude
Residuals
Total
Df
3
257
260
3
394
397
SS
0.009286
0.207972
0.217259
0.007739
0.286646
0.294385
MS
0.00309542
0.00080923
0.00257961
0.00072753
Rsq
0.04274
0.95726
0.02629
0.97371
Degrees of freedom (Df); Sum of squared (SS); Mean squared (MS); F distribution (F); Effect sizes (Z);
��� Significant differences (p < 0.001).
https://doi.org/10.1371/journal.pone.0228975.t003
F
3.8251
Z
5.5797
p-value
0.0001���
3.5457
5.2838
0.0001���
PLOS ONE | https://doi.org/10.1371/journal.pone.0228975 August 20, 2020
9 / 15
PLOS ONE Table 4. ANOVA of Procrustes for the shape and centroid size of wings in males of A. aegypti.
I
S
I � S
I � S � R
Total
I
S
I � S
I � S � R
Total
Df
222
1
222
446
891
Df
222
1
222
446
891
SS
0.55978
0.00049
0.10224
0.05460
0.71711
SS
64.251
0.010
0.984
0.035
0.0717
Shape
Rsq
0.78060
0.00068
0.14257
0.07614
Centroid Size
Rsq
89.598
0.014
1.372
0.049
MS
0.00252153
0.00048901
0.00046055
0.00012242
MS
0.289421
0.010222
0.004432
0.000078
Wing metric variation in Aedes aegypti
F
5.4751
1.0618
0.5632
F
3704.368
130.828
56.731
Z
5.2847
0.3232
13.1852
Z
28.7815
2.7983
23.7410
p-value
0.001��
0.377
0.002��
p-value
0.001��
0.001��
0.001��
Individual (I); Side (S); Replica (R); Degrees of freedom (Df); Sum of squared (SS); Mean squared (MS); F distribution (F); Effect sizes (Z);
�� Significant differences (p < 0.01).
https://doi.org/10.1371/journal.pone.0228975.t004
attributed to other variables not measured in this study, like larval density and availability of
food. In A. aegypti males and females, it was experimentally observed that wing size is corre-
lated negatively in function of the larval density, and positively in function of the availability of
food, [69], hence, expecting that with a higher larval density, mosquitoes will have a smaller
wing size, and with greater availability of food, there will be bigger individuals.
Previously, in A. aegypti, it was noted that females with bigger wings have higher survival
and greater amount of feeding events through blood [70] and, thereby, increase the probability
of transmitting some type of arbovirus [14]. This study did not measure wing length; nonethe-
less, it has been evidenced that CS has a linear relation with traditional wing length measure-
ments [28], hence, higher CS values would indicate longer wings. In our case, higher CS were
Table 5. ANOVA of procrustes for shape and centroid size on wings of females of A. aegypti.
Df
384
1
384
770
Shape Anova
SS
0.92854
0.00044
0.13376
0.06408
MS
0.00241808
0.00043688
0.00034833
0.00008322
Rsq
0.82404
0.00039
0.11871
0.05687
1539
1.12682
Df
384
1
384
770
1539
SS
189.936
0.020
1.518
0.081
1.127
Centroid size Anova
MS
0.49463
0.01985
0.00395
0.00011
Rsq
168.560
0.018
1.347
0.072
I
S
I � S
I � S � R
Total
I
S
I � S
I � S � R
Total
F
6.9419
1.2542
0.4710
F
4689.132
188.140
37.482
Z
7.4911
0.8178
12.6084
Z
29.8657
2.8827
25.2517
p-value)
0.001��
0.207
0.001��
p-value
0.001��
0.001��
0.001��
Individual (I); Side (S); Replica (R); Degrees of freedom (Df); Sum of squared (SS); Mean squared (MS); F distribution (F); Effect sizes (Z);
�� Significant differences (p < 0.01).
https://doi.org/10.1371/journal.pone.0228975.t005
PLOS ONE | https://doi.org/10.1371/journal.pone.0228975 August 20, 2020
10 / 15
PLOS ONE Wing metric variation in Aedes aegypti
found at 1,400 m (Armenia), a city that has historically reported a higher number of dengue
cases for the department of Quindı´o [71].
Furthermore, sexual dimorphism, allometry, and fluctuating asymmetry were found
between both sexes for wing size and shape of A. aegypti. Sexual dimorphism, is a pattern pre-
viously observed in morphometric studies for wing shape and size in other Culicidae species
[72–75]. Our results suggest allometry between wing size and shape of both sexes, which has
been observed in A. aegypti [76], as well as in other Culicidae species [27, 77]. Results obtained
of fluctuating asymmetry indicate that the wings of both sexes of A. aegypti do not have direc-
tionality. The fluctuating asymmetry in mosquitoes can be attributed to environmental pres-
sures [78], among them, vector control in urban zones could be one of them, given that
previously resistance was detected to organophosphorus compounds, a class of insecticide
commonly used for larval control in some of the sampling zones evaluated herein [79].
5. Conclusion
In a small scale and in an altitudinal gradient of the Colombian Andes, we found that geo-
metric morphometry permits identifying phenotypic variation for A. aegypti wing size and
shape. Geometric morphometry studies on wing variation could be used by vector control
programs as a diagnostic tool to quantify the dispersion and vector capacity of A. aegypti.
Future studies must be carried out to test if wing size is related with the vector capacity in
this species.
Supporting information
S1 File.
(XLSX)
Acknowledgments
Gratitude is expressed to the Biology Program and the Center of Studies and Research on Bio-
diversity and Biotechnology at Universidad del Quindı´o (CIBUQ) for providing reagents and
equipment, as well as the Health Secretary of Armenia for its support and suggestions during
the sampling.
We also thank the members of the Center of Research on Tropical Diseases (CINTROP) for
their help and training in identifying the mosquitoes.
This work is dedicated to the memory of the grandmother (La Plita) of Vı´ctor Hugo, who
reached the age of 100 years by the time this research ended.
Author Contributions
Conceptualization: Oscar Alexander Aguirre Obando, Vı´ctor Hugo Garcı´a-Mercha´n.
Data curation: Luis Mı´guel Leyton Ramos, Jonny Edward Duque, Vı´ctor Hugo Garcı´a-
Mercha´n.
Formal analysis: Luis Mı´guel Leyton Ramos, Oscar Alexander Aguirre Obando, Jonny
Edward Duque, Vı´ctor Hugo Garcı´a-Mercha´n.
Funding acquisition: Vı´ctor Hugo Garcı´a-Mercha´n.
Investigation: Oscar Alexander Aguirre Obando, Vı´ctor Hugo Garcı´a-Mercha´n.
Methodology: Luis Mı´guel Leyton Ramos, Oscar Alexander Aguirre Obando, Jonny Edward
Duque, Vı´ctor Hugo Garcı´a-Mercha´n.
PLOS ONE | https://doi.org/10.1371/journal.pone.0228975 August 20, 2020
11 / 15
PLOS ONE Wing metric variation in Aedes aegypti
Project administration: Vı´ctor Hugo Garcı´a-Mercha´n.
Resources: Vı´ctor Hugo Garcı´a-Mercha´n.
Software: Vı´ctor Hugo Garcı´a-Mercha´n.
Supervision: Vı´ctor Hugo Garcı´a-Mercha´n.
Validation: Oscar Alexander Aguirre Obando, Vı´ctor Hugo Garcı´a-Mercha´n.
Visualization: Oscar Alexander Aguirre Obando, Vı´ctor Hugo Garcı´a-Mercha´n.
Writing – original draft: Luis Mı´guel Leyton Ramos, Oscar Alexander Aguirre Obando, Vı´c-
tor Hugo Garcı´a-Mercha´n.
Writing – review & editing: Luis Mı´guel Leyton Ramos, Oscar Alexander Aguirre Obando,
Jonny Edward Duque, Vı´ctor Hugo Garcı´a-Mercha´n.
References
1. Khormi HM, Kumar L. Climate change and the potential global distribution of Aedes aegypti: spatial
modelling using GIS and CLIMEX. Geospatial health. 2014; 8(2):405–15. https://doi.org/10.4081/gh.
2014.29. PMID: 24893017
2. World Health Organization. Global strategy for dengue prevention and control 2012–2020. Geneva:
WHO. 2012.
3. Marchette NJ, Garcia R, Rudnick A. Isolation of Zika virus from Aedes aegypti mosquitoes in Malaysia.
The American journal of tropical medicine and hygiene. 1969; 18(3):411–5. https://doi.org/10.4269/
ajtmh.1969.18.411 PMID: 4976739
4. Pialoux G, Gau¨zère BA, Jaure´guiberry S, Strobel M. Chikungunya, an epidemic arbovirosis. The Lancet
infectious diseases. 2007; 7(5):319–27. https://doi.org/10.1016/S1473-3099(07)70107-X. PMID:
17448935
5. Ruiz-Lo´pez F, Gonza´ lez-Mazo A, Ve´ lez-Mira A, Go´ mez GF, Zuleta L, Uribe S, et al. Presencia de
Aedes (Stegomyia) aegypti (Linnaeus, 1762) y su infeccio´n natural con el virus del dengue en alturas
no registradas para Colombia. Biome´ dica. 2016; 36(2):303–8. https://doi.org/10.7705/biomedica.v36i2.
3301.
6. Rodrı´guez Diego JG, Olivares JL, Sa´nchez Castilleja Y, Alema´ n Y, Arece J. Cambios clima´ ticos y su
efecto sobre algunos grupos de para´sitos. Revista de Salud Animal. 2013; 35(3):145–50.
7. Hopp MJ, Foley JA. Global-scale relationships between climate and the dengue fever vector, Aedes
aegypti. Climatic change. 2001; 48(2–3):441–63.
8.
Leisnham PT, Sala LM, Juliano SA. Geographic variation in adult survival and reproductive tactics of
the mosquito Aedes albopictus. Journal of medical entomology. 2014; 45(2):210–21.
9. Maciel-de-Freitas R, Codeco CT, Lourenc¸o-de-Oliveira R. Body size-associated survival and dispersal
rates of Aedes aegypti in Rio de Janeiro. Medical and Veterinary Entomology. 2007; 21(3):284–92.
https://doi.org/10.1111/j.1365-2915.2007.00694.x. PMID: 17897370
10. Sumanochitrapon W, Strickman D, Sithiprasasna R, Kittayapong P, Innis BL. Effect of size and geo-
graphic origin of Aedes aegypti on oral infection with dengue-2 virus. The American journal of tropical
medicine and hygiene. 1998; 58(3):283–6. https://doi.org/10.4269/ajtmh.1998.58.283. PMID: 9546404
11. Xue RD, Barnard DR, Schreck CE. Influence of body size and age of Aedes albopictus on human host
attack rates and the repellency of deet. Journal of the American Mosquito Control Association. 1995; 11
(1):50–3. PMID: 7616190
12. Briegel H, Timmermann SE. Aedes albopictus (Diptera: Culicidae): physiological aspects of develop-
ment and reproduction. Journal of medical entomology. 2001; 38(4):566–71. https://doi.org/10.1603/
0022-2585-38.4.566. PMID: 11476337
13. Alto BW, Lounibos LP, Mores CN, Reiskind MH. Larval competition alters susceptibility of adult Aedes
mosquitoes to dengue infection. Proceedings of the Royal Society of London B: Biological Sciences.
2008; 275(1633):463–71. http://dx.doi.org/10.1098/rspb.2007.1497.
14. Alto BW, Reiskind MH, Lounibos LP, hygiene. Size alters susceptibility of vectors to dengue virus infec-
tion and dissemination. The American journal of tropical medicine and hygiene. 2008; 79(5):688–95.
15. Abul-Hab J. Larvae of culicine mosquitos in north Iraq (Diptera, Culicidae). Bulletin of entomological
research. 1967; 57(2):279–84. https://doi.org/10.1017/S0007485300049981. PMID: 4382531
PLOS ONE | https://doi.org/10.1371/journal.pone.0228975 August 20, 2020
12 / 15
PLOS ONE Wing metric variation in Aedes aegypti
16. Hoffmann AA, Ratna E, Sgro CM, Barton M, Blacket M, Hallas R, et al. Antagonistic selection between
adult thorax and wing size in field released Drosophila melanogaster independent of thermal conditions.
Journal of evolutionary biology. 2007; 20(6):2219–27. https://doi.org/10.1111/j.1420-9101.2007.01422.x.
PMID: 17887974
17. Dapporto L, Bruschini C, Baracchi D, Cini A, Gayubo SF, Gonzalez JA, et al. Phylogeography and
counter-intuitive inferences in island biogeography: evidence from morphometric markers in the mobile
butterfly Maniola jurtina (Linnaeus)(Lepidoptera, Nymphalidae). Biological Journal of the Linnean Soci-
ety. 2009; 98(3):677–92. https://doi.org/10.1111/j.1095-8312.2009.01311.x.
18. Demirci B, Lee Y, Lanzaro GC, Alten B. Altitudinal genetic and morphometric variation among popula-
tions of Culex theileri Theobald (Diptera: Culicidae) from northeastern Turkey. Journal of Vector Ecol-
ogy. 2012; 37(1):197–209. https://doi.org/10.1111/j.1948-7134.2012.00217.x. PMID: 22548554
19. Alexander RD. Sound production and associated behavior in insects. The Ohio Journal of Science.
1957; 57(2):101–13.
20. Belton P. Attractton of male mosquitoes to sound. J Am Mosq Control Assoc. 1994; 10:297–301.
21. Roth LM. A study of mosquito behavior. An experimental laboratory study of the sexual behavior of
Aedes aegypti (Linnaeus). J The American Midland Naturalist. 1948; 40(2):265–352.
22. Mohammed A, Chadee DD. Effects of different temperature regimens on the development of Aedes
aegypti (L.) (Diptera: Culicidae) mosquitoes. Acta tropica. 2011; 119(1):38–43. https://doi.org/10.1016/
j.actatropica.2011.04.004. PMID: 21549680
23.
24.
Tun-Lin W, Burkot T, Kay B. Effects of temperature and larval diet on development rates and survival of
the dengue vector Aedes aegypti in north Queensland, Australia. Medical and veterinary entomology.
2000; 14(1):31–7. https://doi.org/10.1046/j.1365-2915.2000.00207.x. PMID: 10759309
Lambrechts L, Paaijmans KP, Fansiri T, Carrington LB, Kramer LD, Thomas MB, et al. Impact of daily
temperature fluctuations on dengue virus transmission by Aedes aegypti. Proceedings of the National
Academy of Sciences. 2011; 108(18):7460–5. https://doi.org/10.1073/pnas.1101377108.
25. Wilke ABB, de Oliveira Christe R, Multini LC, Vidal PO, Wilk-da-Silva R, de Carvalho GC, et al. Morpho-
metric wing characters as a tool for mosquito identification. PloS one. 2016; 11(8):e0161643. https://
doi.org/10.1371/journal.pone.0161643. PMID: 27551777
26. Bookstein FL. Morphometric tools for landmark data: geometry and biology: Cambridge University
Press; 1991.
27. Vidal PO, Carvalho E, Suesdek L. Temporal variation of wing geometry in Aedes albopictus. Memo´rias
do Instituto Oswaldo Cruz. 2012; 107(8):1030–4. https://dx.doi.org/10.1590/S0074-
02762012000800011. PMID: 23295754
28. Vargas REM, Ya-umphan P, Phumala-Morales N, Komalamisra N, Dujardin J-P. Climate associated
size and shape changes in Aedes aegypti (Diptera: Culicidae) populations from Thailand. Infection,
Genetics and Evolution. 2010; 10(4):580–5. https://doi.org/10.1016/j.meegid.2010.01.004. PMID:
20123039
29. Carvajal JJ, Moncada Ll, Rodrı´guez MH, L. PdP, Olano VA. Caracterizacio´n preliminar de los sitios de
crı´a de Aedes (Stegomyia) albopictus (Skuse, 1894)(Diptera: Culicidae) en el municipio de Leticia,
Amazonas, Colombia. Biome´ dica. 2009; 29(3). https://doi.org/10.7705/biomedica.v29i3.13.
30. Sendaydiego JP, Torres MAJ, Demayo CG. Describing wing geometry of Aedes aegypti using land-
mark-based geometric morphometrics. International Journal of Bioscience, Biochemistry and Bioinfor-
matics. 2013; 3(4):379.
31. Oliveira Paloma, S Lincoln. Comparison of wing geometry data and genetic data for assessing the pop-
ulation structure of Aedes aegypti. Infection, Genetics and Evolution. 2012; 12(3):591–6. https://doi.org/
10.1016/j.meegid.2011.11.013. PMID: 22178147
32. Kuclu O, Aldemir A, Demirci B. Altitudinal variation in the morphometric characteristics of Aedes vexans
Meigen from northeastern Turkey. Journal of Vector Ecology. 2011; 36(1):30–41. https://doi.org/10.
1111/j.1948-7134.2011.00138.x. PMID: 21635639
33. Prudhomme J, Velo E, Bino S, Kadriaj P, Mersini K, Gunay F, et al. Altitudinal variations in wing mor-
phology of Aedes albopictus (Diptera, Culicidae) in Albania, the region where it was first recorded in
Europe. Parasite. 2019; 26.
34. Cassab A, Morales V, Mattar S. Climatic factors and cases of dengue in Monteria, Colombia: 2003–
2008. Revista de Salud Pu´blica. 2011; 13(1):115–28.
35. Gobernacio´ n del Quindio. Ficha basica municipal La Tebada 2017 [cited 2019 September 09]. https://
www.quindio.gov.co/observatorio-departamental-del-quindio/fichas-basicas-municipales/fichas-la-
tebaida.
36. Gobernacio´ n del Quindio. Ficha basica municipal Filandia, 2017 [cited 2019 September 09]. https://
www.quindio.gov.co/observatorio-departamental-del-quindio/fichas-basicas-municipales/filandia.
PLOS ONE | https://doi.org/10.1371/journal.pone.0228975 August 20, 2020
13 / 15
PLOS ONE Wing metric variation in Aedes aegypti
37. Gobernacio´ n del Quindio. Ana´ lisis de la situacio´n del departamento del Quindı´o 2012 [cited 2019 Sep-
tember 09]. https://www.minsalud.gov.co/plandecenal/mapa/Analisis-de-situacion-de-salud-del-
departamento-del-Quindio.pdf.
38.
Fick S HJ. Worldclim 2: New 1-km spatial resolution climate surfaces for global land areas. International
Journal of Climatology. International journal of climatology. 2017; 37(12):4302–15.
39. Ewing A CA, Purse V, Nunn M, White S. Modelling the effect of temperature on the seasonal population
dynamics of temperate mosquitoes. Journal of theoretical biology. 2016;(400):65–79. https://doi.org/10.
1016/j.jtbi.2016.04.008 PMID: 27084359
40. Huber J CM, Caldwell J, Mordecai E. Seasonal temperature variation influences climate suitability for
dengue, chikungunya, and Zika transmission. Plos Negl Trop Dis. 2018; 12(5). https://doi.org/10.1371/
journal.pntd.0006451 PMID: 29746468
41. Hijmans RJ, Van Etten J, Cheng J, Mattiuzzi M, Sumner M, Greenberg JA, et al. Package ‘raster’. R
package 2015; 734.
42.
Forattini OP. Culicidologia me´dica: identificac¸ão, biologia, epidemiologia Vol. 2: Edusp; 1996.
43. Rueda LM. Pictorial keys for the identification of mosquitoes (Diptera: Culicidae) associated with Den-
gue Virus Transmission. Zootaxa. 2004; 589(1):1–60.
44.
Lorenz C, Suesdek L. Evaluation of chemical preparation on insect wing shape for geometric morpho-
metrics. The American journal of tropical medicine and hygiene. 2013; 89(5):928–31. https://doi.org/10.
4269/ajtmh.13-0359. PMID: 24019438
45. Dos-Santos EB. Estrutura da comunidade de culicidae (Insecta: Diptera) em a´rea de Mata Atlaˆ ntica do
Parana´ [M.Sc.]: Universidade federal do Parana´; 2016.
46.
Zelditch ML, Swiderski DL, Sheets HD. Geometric morphometrics for biologists: a primer: Academic
Press; 2012.
47. Guedes MLP. Ana´lise de fatores regulato´rios na variac¸ão interpopulacional e na composic¸ão das
comunidades de culicidae (Diptera) [PhD]: Universidade Federal do parana´; 2014.
48. Rohlf J, Slice D. Extensions of the Procrustes method for the optimal superimposition of landmarks.
Systematic Biology. 1990; 39(1):40–59.
49. Rohlf FJ. tpsUtil, file utility program, version 1.7. Department of Ecology and Evolution, State University
of New York at Stony Brook. 2016.
50. Rohlf F. TPSDig2: a program for landmark development and analysis. Department of Ecology and Evo-
lution, State University of New York, Stony Brook. 2001.
51. Klingenberg CP. MorphoJ: an integrated software package for geometric morphometrics. Molecular
ecology resources. 2011; 11(2):353–7. https://doi.org/10.1111/j.1755-0998.2010.02924.x PMID:
21429143
52. Gower JC. Generalized procrustes analysis. Psychometrika. 1975; 40(1):33–51.
53. Yezerinac SM, Lougheed SC, Handford P. Measurement error and morphometric studies: statistical
power and observer experience. Systematic Biology. 1992; 41(4):471–82. https://doi.org/10.1093/
sysbio/41.4.471.
54. Claude J. Morphometrics with R: Springer Science & Business Media; 2008. 317 p.
55.
Fruciano C. Measurement error in geometric morphometrics. Development genes and evolution. 2016;
226(3):139–58. https://doi.org/10.1007/s00427-016-0537-4 PMID: 27038025
56. Drake AG, Klingenberg CP. The pace of morphological change: historical transformation of skull shape
in St Bernard dogs. Proceedings of the Royal Society B: Biological Sciences. 2007; 275(1630):71–6.
https://doi.org/10.1098/rspb.2007.1169.
57. Good P. Permutation tests: a practical guide to resampling methods for testing hypotheses: Springer
Science & Business Media; 2013.
58. Goodall C. Procrustes methods in the statistical analysis of shape. Journal of the Royal Statistical Soci-
ety Series B (Methodological). 1991:285–339. https://doi.org/10.1111/j.2517-6161.1991.tb01825.x.
59. Klingenberg CP, Leamy LJ. Quantitative genetics of geometric shape in the mouse mandible. Evolution.
2001; 55(11):2342–52. https://doi.org/10.1111/j.0014-3820.2001.tb00747.x PMID: 11794792
60. Mardia KV, Bookstein FL, Moreton IJJB. Statistical assessment of bilateral symmetry of shapes. Biome-
trika. 2000; 87(2):285–300.
61. R Development CoreTeam. R: A language and environment for statistical computing. R Foundation for
Statistical Computing, Vienna, Austria. 2018.
62. Adams DC, Ota´ rola-Castillo E. geomorph: an R package for the collection and analysis of geometric
morphometric shape data. Methods in Ecology and Evolution. 2013; 4(4):393–9. https://doi.org/10.
1111/2041-210X.12035.
PLOS ONE | https://doi.org/10.1371/journal.pone.0228975 August 20, 2020
14 / 15
PLOS ONE Wing metric variation in Aedes aegypti
63. Wickham H, Franc¸ois R, Henry L, Mu¨ ller K. dplyr: A Grammar of Data Manipulation. R package version
0.8. 0.1. ed; 2019.
64. Mangiafico S, Mangiafico MS. Package ‘rcompanion’. CRAN package repository. 2017:1–71.
65. Mangiafico SS. An R companion for the handbook of biological statistics. rcompanion org/documents/
RCompanionBioStatistics pdf. 2015.
66. Graves S, Piepho H-P, Selzer L, Dorai-Raj S. multcompView: visualizations of paired comparisons. R
package version 01–7. 2015.
67. Mohammed A, Chadee D. Effects of different temperature regimens on the development of Aedes
aegypti (L.) (Diptera: Culicidae) mosquitoes. Acta tropica. 2011; 119(1):38–43. https://doi.org/10.1016/
j.actatropica.2011.04.004 PMID: 21549680
68.
69.
Instituto de Hidrologı´a MyEAIUdPMEU. Atlas Climatolo´gico de Colombia 2020 [cited 2020 May 22].
http://atlas.ideam.gov.co/visorAtlasClimatologico.html.
Jirakanjanakit N, Leemingsawat S, Thongrungkiat S, Apiwathnasorn C, Singhaniyom S, Bellec C, et al.
Influence of larval density or food variation on the geometry of the wing of Aedes (Stegomyia) aegypti.
Tropical Medicine & International Health. 2007; 12(11):1354–60.
70. Nasci RS. The size of emerging and host-seeking Aedes aegypti and the relation of size to blood-feed-
ing success in the field. J Am Mosq Control Assoc. 1986; 2(1):61–2. PMID: 3507471
71. Muñoz LGD. Incidencia del Dengue en el departamento del Quindio 1999–2010. Hospitium. 2015; 2.
72. Virginio F, Vidal PO, Suesdek L. Wing sexual dimorphism of pathogen-vector culicids. Parasites & Vec-
tors. 2015; 8(1):159.
73. Sanchez E, Castillo D, Liria J. Pupal shape and size dimorphism in Aedes albopictus (Skuse, 1894)
(Diptera: Culicidae). Journal of Threatened Taxa. 2017; 9(6):10314–9.
74. Oliveira Christe R, Wilke ABB, Vidal PO, Marrelli MT. Wing sexual dimorphism in Aedes fluviatilis (Dip-
tera: Culicidae). Infection, Genetics and Evolution. 2016; 45:434–6. https://doi.org/10.1016/j.meegid.
2016.10.007. PMID: 27746293
75. Devicari M, Lopes AR, Suesdek L. Wing sexual dimorphism in Aedes scapularis (Diptera: Culicidae).
Biota Neotropica. 2011; 11(2):165–9.
76.
Jirakanjanakit N, Leemingsawat S, Dujardin JP. The geometry of the wing of Aedes (Stegomyia)
aegypti in isofemale lines through successive generations. Infection, Genetics Evolution. 2008; 8
(4):414–21. https://doi.org/10.1016/j.meegid.2007.05.004 PMID: 17600773
77. Garzo´ n MJ, Schweigmann N. Morphometric Variation of the Aedes albifasciatus (Diptera: Culicidae)
Wings in Three Populations From Different Ecoregions of Argentina. Journal of medical entomology.
2018; 55(6):1602–6. https://doi.org/10.1093/jme/tjy096 PMID: 29939291
78. Galbo KR, Tabugo SRM. Fluctuating asymmetry in the wings of Culex quinquefasciatus (Say) (Diptera:
Culicidae) from selected barangays in Iligan City, Philippines. Aquaculture, Aquarium, ConservationLe-
gislation. 2014; 7(5):357–64.
79. Aguirre-Obando OA, Bona ACD, Duque L, Jonny E, Navarro-Silva MA. Insecticide resistance and
genetic variability in natural populations of Aedes (Stegomyia) aegypti (Diptera: Culicidae) from Colom-
bia. Zoologia. 2015; 32(1):14–22.
PLOS ONE | https://doi.org/10.1371/journal.pone.0228975 August 20, 2020
15 / 15
PLOS ONE |
10.1371_journal.pone.0229861 | RESEARCH ARTICLE
Postnatal pediatric systemic antibiotic
episodes during the first three years of life are
not associated with mode of delivery
2, Nicole Cacho3,
Dominick J. LemasID
Elizabeth Plasencia4, Alice S. Rhoton-Vlasak4, Josef Neu3, Lindsay Thompson3,
Magda Francois1, Keval Patel1, William R. HoganID
1*, Jasmine A. Mack1, Jennifer J. SchochID
1, Gloria P. Lipori5, Matthew J. Gurka1
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
1 Department of Health Outcomes and Biomedical Informatics, University of Florida College of Medicine,
Gainesville, Florida, United States of America, 2 Department of Dermatology, University of Florida College of
Medicine, Gainesville, Florida, United States of America, 3 Department of Pediatrics, University of Florida
College of Medicine, Gainesville, Florida, United States of America, 4 Department of Obstetrics and
Gynecology, University of Florida College of Medicine, Gainesville, Florida, United States of America,
5 University of Florida Health Shands Hospital, Gainesville, Florida, United States of America
OPEN ACCESS
Citation: Lemas DJ, Mack JA, Schoch JJ, Cacho N,
Plasencia E, Rhoton-Vlasak AS, et al. (2020)
Postnatal pediatric systemic antibiotic episodes
during the first three years of life are not associated
with mode of delivery. PLoS ONE 15(3): e0229861.
https://doi.org/10.1371/journal.pone.0229861
Editor: Ju Lee Oei, University of New South Wales,
AUSTRALIA
Received: July 3, 2019
Accepted: February 16, 2020
Published: March 4, 2020
Copyright: © 2020 Lemas et al. This is an open
access article distributed under the terms of the
Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: Data cannot be
shared publicly in order to protect participant
confidentiality. All requests to review data that may
include potential subject and/or patient identifiers
will be reviewed by our University’s Privacy Office
and Institutional Review Board to ensure adequate
protection of subjects. Any dissemination approved
by the University will be reviewed by a University
honest data broker to ensure appropriate legal and
ethical compliance with regulatory agencies.
Requests for data should be addressed to Chris
Harle at the University of Florida (charle@ufl.edu).
* djlemas@ufl.edu
Abstract
Background
Delivery by cesarean section (C-section) is associated with adverse short-term and long-
term infant outcomes. Given that antibiotics during early life are prescribed for infant out-
comes that are more likely among c-section deliveries, we hypothesized that postnatal
antibiotic exposure will be greater among c-section infants compared to vaginally delivered
infants.
Objective
The aim of this paper was to evaluate if mode of infant delivery was associated with patterns
of systemic antibiotic exposure in children during their first three years.
Methods
Pediatric electronic health records from UFHealth, 2011 to 2017 were reviewed. We
included singleton, term infants (37–42 weeks gestation) with a birth weight � 2500 grams,
with documented mode of delivery and well visits on record. Infants with a neonatal intensive
care unit stay were excluded. Both oral and intravenous antibiotics for a 10-day duration
were classified as a single episode. The primary outcome was antibiotic episodes in the
first three years of life, and a sub-analysis was performed to compare broad-spectrum ver-
sus narrow-spectrum antibiotic exposures.
Results
The mean number of antibiotic episodes in 4,024 full-term infants was 0.34 (SD = 0.79) and
24.1% of infants had at least one antibiotic episode. Penicillins were the most prescribed
antibiotic in children 0–1 years (66.9%) and cephalosporins were the most common
PLOS ONE | https://doi.org/10.1371/journal.pone.0229861 March 4, 2020
1 / 12
PLOS ONE Funding: The Research reported in this publication
was supported by the National Institute of Diabetes
and Digestive and Kidney Diseases
(K01DK115632), National Center for Advancing
Translational Sciences of the National Institutes of
Health under University of Florida Clinical and
Translational Science Awards UL1 TR000064 and
UL1TR001427. All authors - designed research
(project conception, development of overall
research plan, and study oversight); JAM and MJG
analyzed data or performed statistical analysis;
DJL, JS, NC, JN - interpreted data findings; DJL
wrote paper and had primary responsibility for final
content. The funders had no role in study design,
data collection and analysis, decision to publish, or
preparation of the manuscript.
Competing interests: The authors have declared
that no competing interests exist.
Pediatric antibiotics according to mode of delivery
antibiotic prescribed for children 1–3 years (56.2%). We did not detect a meaningful or sig-
nificant rate ratio (RR) between mode of delivery and overall antibiotic episodes 1.14 (95%
CI 0.99, 1.31), broad-spectrum episodes 1.19 (95% CI 0.93, 1.52, or narrow-spectrum epi-
sodes 1.14 (95% CI 0.97, 1.34).
Conclusion
Our results do not support the hypothesis that postnatal antibiotic exposure was greater
among infants delivered by cesarean section compare to infants delivered vaginally during
the first three years of life.
Introduction
Cesarean section (CS) is the most common surgical procedure in the United States[1] and
accounts for approximately 30% of infant deliveries[2]. Delivery by cesarean section has been
associated with adverse health outcomes for both infant and mother,[3,4] and CS rates beyond
10–15% may be potentially avoidable[3–5]. Short-term health outcomes for infants delivered via
CS include increased respiratory distress[4,6–8], delayed breastfeeding, hypoglycemia[3,9], sup-
pressed immune function[3,10], blood pressure abnormalities[3], altered thermogenic response
[3], and increased admissions to the NICU[8,11]. Long-term adverse health outcomes for CS
deliveries include increased risk of immune-related conditions such as asthma,[3,4,10,12–14]
Type 1 diabetes,[3,12] food allergies,[3,4,10] allergic rhinitis[3,4,10] and celiac disease,[9,15] as
well as outcomes related to obesity [3,12,16–21]. The human microbiome has emerged as a
potential therapeutic mechanism that links CS with adverse infant health outcomes [22]. One
mechanism believed to link CS and later disease is the hygiene hypothesis and the differential
exposure to maternal microbiota experienced by infants born via CS versus vaginal delivery
[22]. Despite these observations, population-level data has produced conflicting results in impli-
cating mode of delivery as a critical factor contributing to infant health outcomes [16,23]. A
potential explanation for this inconsistency in outcomes may be residual confounding by early
life exposures such as postnatal antibiotics [24].
Antibiotics are among the most common drug class prescribed to children [25] with the
highest incidence rates in children under the age of five [26–30]. Overutilization of antibiotics,
especially those that are broad-spectrum, remain a public health concern as they promote
greater levels of resistance [31]. Postnatal antibiotic exposure in children has been shown to
have a direct influence on the infant microbiome [32] and health outcomes such as obesity
[16]. Epidemiological data has demonstrated CS with subsequent antibiotic exposure during
the first six months of life was associated with increased risk of obesity among boys at age
seven [16]. Notably, antibiotics are frequently prescribed for symptoms (asthma, respiratory,
and gastrointestinal infections) that are more likely among cesarean delivered infants[33].
Exposure to antibiotics during the first year of life is associated with subsequent antibiotic
exposure throughout the life span[34]. Population-based investigation of antibiotics frequently
report the effect size of the exposure as antibiotic episodes (1–4+) associated with a change in a
given health outcome [35]. Taken together, we hypothesized that postnatal antibiotic exposure
will be greater among c-section infants compared to vaginally delivered infants. To address
this hypothesis, we leveraged electronic health records (EHRs) to characterize the prevalence
of pediatric antibiotic presecriptions during the first three years of life and tested for significant
associations between mode of delivery and antibiotic episodes. EHRs are a low-cost, large-
PLOS ONE | https://doi.org/10.1371/journal.pone.0229861 March 4, 2020
2 / 12
PLOS ONE Pediatric antibiotics according to mode of delivery
scale data source with relatively low error rates [36], especially when compared to manual data
extractions, that can be utilized to track population trends in pediatric antibiotic utilization
[35].
Methods
Study design
De-identified linked maternal-infant electronic health records (EHRs) were collected from
UFHealth Shands Hospital and UFHealth Jacksonville between June 1, 2011 and April 30,
2017. Briefly, an honest broker within UF Health Integrated Data Repository (IDR) used fully
identified data to perform the linkage, then removed identifiers from the linked data set before
releasing the de-identified data to the research team. Demographic information such as insur-
ance type at delivery, age, race, and ethnicity were extracted from the maternal chart and infant
demographics including sex, race and ethnicity were extracted from the infant chart. Notably,
all infants delivered with the UF Health system are linked to maternal records as a standard
of care. The inclusion criteria for maternal-infant pairs included in the analysis is presented
in S1 Fig We included term singleton infants (37–42 weeks gestational age) with a birth
weight � 2500 grams, at least one documented well visit during the first 36 months of life, and
mode of delivery classified as either “low transverse cesarean section” or “spontaneous vaginal
delivery”. To account for the possibility of confounding from premature infants requiring anti-
biotics, we excluded infants that were premature (<37 weeks gestational age), low birth weight
(<2500 grams), or admitted to the neonatal intensive care unit for any reason. We included
infants with sequential measurements, usually occurring in the setting of a well visit, during
the exposure period in an effort to define a conservative cohort of children with an established
connection to UFHealth care services that likely increases the probability of capturing com-
plete antibiotic prescribing data. Taken together, our analysis includes children within the
sample that have variable follow-up times within the 36-month time period. This study was
approved by the institutional review board at the University of Florida.
Primary outcome
The primary outcome of interest was systemic antibiotic episodes in the first 36 months of life.
We extracted antibiotic prescription information from all clinical encounters, including inpa-
tient, emergency or ambulatory visits at �36 months of age. All antibiotics were categorized
according to the Anatomic Theraputic Chemical (ATC) index, a hierarchical classification
system for the purposes of drug statistics that was developed by the World Health Organiza-
tion[37]. Systemic antibiotics were defined as all drugs within the ATC-group J01 (antibiotics
for systemic use). An antibiotic episode was defined as any systemic antibiotic that was admin-
istered or prescribed more than 10 days from any prior or subsequent antibiotic prescription,
from birth to the last wellness visit on record. Cumulative antibiotic episodes were analyzed as
a count variable through the last antibiotic prescription within the first three years. We also
separately examined narrow-spectrum and broad-spectrum antibiotic episodes for each child.
Narrow spectrum antibiotics were defined as those that act on one major group of bacteria.
Broad spectrum antibiotics were defined as those that act on two or more major bacterial
groups (S1 Table). Within the analytical sample, the antibiotic episodes ranged from 0 to 12
occurrences (narrow-spectrum: 0–11, broad-spectrum: 0–6). Due to the nature of multiple
prescriptions and exposure windows, children may have simultaneous exposure to narrow-
and broad-spectrum antibiotics, thus these drug categories are not mutually exclusive. For
example, if a child receives narrow and broad-spectrum antibiotics within the same 10-day
window, the occurrence will count as one antibiotic episode. However, it will also count as one
PLOS ONE | https://doi.org/10.1371/journal.pone.0229861 March 4, 2020
3 / 12
PLOS ONE Pediatric antibiotics according to mode of delivery
episode each under the narrow and broad-spectrum antibiotic categories. We excluded topical
antibiotics as there is no evidence linking topical antibiotics to the gut microbiome or infant
outcomes such as weight gain.
Statistical analysis
We evaluated the association between mode of delivery and the rate of antibiotic episodes
over the first three years via negative binomial regression, which accounted for differential
follow-up time per child (S1 Fig). We also modeled separately by antibiotic: narrow-spec-
trum and broad-spectrum classes. To assess for possible differences over time, cumulative
antibiotic counts over the first three years were modeled and compared between the two
delivery groups via recurrent events analysis within a proportional hazards regression
framework. Unadjusted rate ratios and hazard ratios (and 95% confidence intervals (CI))
comparing the delivery groups were estimated and reported for the two model types, respec-
tively. We also adjusted for infant gender and covariates found to be different between the
two delivery groups; these adjusted rate and hazard ratios (and 95% CI’s) were also reported.
We assessed for collinearity of these covariates with variance inflation factors (VIF) before
adding covariates to the adjusted model, where VIF>5 signifying collinearity. Chi-squared
analyses and two-sample t-tests were used to compare demographic characteristics by deliv-
ery group. Analyses were performed with a significance level of α = 05 using SAS software
version 9.4 (SAS Institute, Cary, NC).
Results
Study cohort
We obtained data on 16,684 infants born between June 1, 2011 and April 30, 2017. Of these,
4,024 (24.1%) children with at least one well visit met inclusion criteria (S1 Fig). Mother and
child pairs were categorized by mode of delivery, where 1,211 (30.1%) infants were delivered
by cesarean section and 2,813 (69.9%) were delivered vaginally. Infants in either the cesarean
section or vaginal delivery group were equally distributed by gender and there was no statisti-
cally significant difference in race by group, as shown in Table 1. Infants in the vaginal delivery
group were gestationally older (39.5 weeks, SD = 1.1 vs 39.3 weeks, SD = 1.1, p < .0001) and
had smaller birthweight (3315 grams, SD = 411 vs 3375 grams, SD = 460, p = 0.0001) com-
pared to children in the cesarean section group. Mothers who delivered either by cesarean
section or vaginally had similar race/ethnicity and health insurance profiles. Those who gave
birth by cesarean section were older (28.4 years, SD = 5.7 vs 27.4 years, SD = 5.7, p < .0001).
Of 1,375 antibiotic prescriptions in 970 infants, penicillins, which includes Ampicillin,
(66.9%) were the most prescribed antibiotic class in children 0–1 years (Table 2). We also
found that cephalosporins were the most common antibiotic class for children 1–2 years
(56.2%) and 2–3 years (56.2%). The mean number of antibiotic episodes over 3 years in all
children was 0.34 (SD = 0.79), and 24.1% of infants had at least one antibiotic episode during
this period. For those children that had at least one episode, the mean number of antibiotic
episodes was 1.42 (SD = 1.02). Fig 1 shows that 75.9% of infants in this study did not receive
any antibiotics and 18.1% of children received only a single course of antibiotics. S2 Table
shows that 18.8% of children received at least one episode of narrow spectrum antibiotics
and 8.1% of children received at least one broad spectrum antibiotic episode during the first 3
years of life. As shown in Table 3, babies born by cesarean section had 0.37 episodes after three
years (95% CI: 0.32, 0.43) while those born vaginally had 0.33 episodes (95% CI: 0.30, 0.35).
Fig 2 shows the recurrent postnatal antibiotic episodes by mode of delivery. After adjusting for
covariates, there was no statistically significant difference between the number of total
PLOS ONE | https://doi.org/10.1371/journal.pone.0229861 March 4, 2020
4 / 12
PLOS ONE Table 1. Characteristics of the infant and maternal population by mode of delivery1,2.
Total
n = 4,024
Cesarean Section
n = 1,211
Vaginal Delivery
n = 2,813
p-value
Pediatric antibiotics according to mode of delivery
Infant Characteristics
Gender
Female
Male
Race/Ethnicity
Non-Hispanic White
Non-Hispanic Black
Non-Hispanic Other
Hispanic
49.3
50.7
41.3
33.4
16.9
8.4
47.8
52.2
42.1
34.5
15.6
7.8
49.9
50.1
40.9
32.9
17.5
8.7
Gestational Age (weeks), M (SD)
Baby Birth Weight (grams), M (SD)
39.4 (1.1)
3333 (427)
39.3 (1.1)
3375 (460)
39.5 (1.1)
3315 (411)
Maternal Characteristics
Maternal Age, M(SD)
Race/Ethnicity
Non-Hispanic White
Non-Hispanic Black
Non-Hispanic Other
Hispanic
Insurance
Private
Public (Medicare/Medicaid/CMS)
Managed Care/Other
27.7 (5.7)
28.4 (5.7)
27.4 (5.7)
45.2
32.8
13.9
8.0
31.8
53.9
14.3
44.7
34.1
13.5
7.8
32.0
54.8
13.3
45.5
32.2
14.2
8.1
31.7
53.5
14.8
0.2297
0.3139
< .0001
0.0001
< .0001
0.6937
0.4714
1Singleton infants included in the analysis possessed a gestational age greater than or equal to 37 weeks and less than or equal to 42 weeks, birth weight greater than 2500
grams, no time in the NICU, well visits on record and systemic antibiotic data only within the first three years of life and have demographics, a date of birth, and mode
of delivery on record.1Percentage unless otherwise noted.
2Chi-squared analyses were used to compute p-values for categorical covariates and two-sample t-test for continuous covariates.
https://doi.org/10.1371/journal.pone.0229861.t001
antibiotics episodes (Table 3, Rate Ratio (RR) = 1.14, 95% CI: (0.99, 1.31), p = 0.0758; S3 Table,
RR = 1.15 (0.98, 1.34), p = 0.0796), including narrow-spectrum (Table 4, RR = 1.14, 95% CI:
(0.97, 1.34), p = 0.1031) and broad-spectrum episodes (Table 5, RR = 1.19, 95% CI: (0.93,
1.52), p = 0.1654). No collinearity was observed in any of the statistical models (VIF < 2).
Table 2. Antibiotic prescriptions by ATC code among infants during first 3 years of life1.
ATC-code
J01A
J01C
J01D
J01E
J01F
J01X
J04A
ATC Name
Tetracyclines
Beta-lactam antibacterials, penicillins
Other beta-lactam antibacterials
Sulfonamides and trimethoprim
Macrolides, lincosamid and streptogramins
Other antibacterials
Drugs for treatment of tuberculosis
Total
1Antibiotic count and (%).
https://doi.org/10.1371/journal.pone.0229861.t002
Age Group 0<1 years
Age Group 1–2 years
Age Group 2–3 years
0 (0.0%)
588 (66.9%)
260 (29.6%)
3 (0.3%)
9 (1.0%)
19 (2.2%)
1 (0.1%)
880
1 (0.3%)
136 (39.0%)
196 (56.2%)
2 (0.6%)
6 (1.7%)
8 (2.3%)
0 (0.0%)
349
0 (0)
59 (40.4%)
82 (56.2%)
1 (0.7%)
1 (0.7%)
3 (2.1%)
0 (0.0%)
146
All
1 (0.1%)
783 (57.0%)
538 (39.1%)
6 (0.4%)
16 (1.2%)
30 (2.2%)
1 (0.1%)
1,375
PLOS ONE | https://doi.org/10.1371/journal.pone.0229861 March 4, 2020
5 / 12
PLOS ONE Pediatric antibiotics according to mode of delivery
Fig 1. Postnatal pediatric antibiotic episodes per child. Antibiotic episodes were defined as any antibiotic that was administered or prescribed more than
10 days from any prior or subsequent antibiotic prescription, from birth to the last wellness visit on record. Cumulative antibiotic episodes for each child
were analyzed as a count variable and shown here as the percentage of infants that were prescribed 0 to 4 or more antibiotic episodes during the first three
years.
https://doi.org/10.1371/journal.pone.0229861.g001
Discussion
We evaluated the association between mode of delivery and pediatric postnatal antibiotic
exposure in the first three years among term infants using linked mom-baby EHRs. Our analy-
sis revealed that nearly 1 in 4 infants in our study had at least one postnatal antibiotic episode
within the first three years of life. Penicillins (J01C, 66.9%) were the most prescribed antibiotic
in children 0–1 years and cephalosporins were the most common antibiotic for children 1–3
years (J01D, 56.2% for 1–2 years and 56.2% for 2–3 years). The primary finding of this study
demonstrates that there was no statistical difference in the rate of prescribed antibiotics in chil-
dren delivered via CS compared to children delivered vaginally in the first three years of life.
Although we adjusted for potential confounding factors, the possibility that residual
Table 3. Total Antibiotic episodes by mode of delivery.
Mode of Delivery
Cesarean Section
Vaginal Delivery
N
1,211
2,813
Total Antibiotics Episode (95% CI)
0.37 (0.32, 0.43)
0.33 (0.30, 0.35)
1Adjusted for infant race, sex, birth weight, gestational age and maternal age.
https://doi.org/10.1371/journal.pone.0229861.t003
Unadjusted
Negative Binomial Rate Ratio (95% CI)
Adjusted1
1.14 (0.99, 1.31)
1.13 (0.98, 1.30)
p = 0.1004
Ref
p = 0.0758
Ref
PLOS ONE | https://doi.org/10.1371/journal.pone.0229861 March 4, 2020
6 / 12
PLOS ONE Pediatric antibiotics according to mode of delivery
Fig 2. Recurrent postnatal antibiotic episodes by mode of delivery. After adjusting for covariates, there was no statistically significant
difference between the number of total antibiotics episodes (p = 0.07). In this line chart, the mean antibiotic episode was estimated over
the first three years of life by mode of delivery. Blue represents those infants delivered via cesarean section and the orange represents
vaginally delivered infants. Counts of infants according to mode of delivery and postnatal age are listed below the line chart.
https://doi.org/10.1371/journal.pone.0229861.g002
confounding exists within our results is plausible. Nevertheless, the precision of the effect esti-
mates and confidence intervals from our analysis indicates that even if the association between
mode of delivery and postnatal pediatric antibiotics at age three were significant, it would not
represent a clinically meaningful association, as the 95% CI’s did not contain values that one
would consider to represent clinically meaningful differences between the modes of delivery
Table 4. Narrow spectrum antibiotics by mode of delivery.
Mode of Delivery
N
Narrow Spectrum Antibiotics Episode (95% CI)
Cesarean Section
Vaginal Delivery
1,211
2,813
0.27 (0.23, 0.31)
0.23 (0.21, 0.25)
1Adjusted for infant race, sex, birth weight, gestational age and maternal age.
https://doi.org/10.1371/journal.pone.0229861.t004
Negative Binomial Rate Ratio (95% CI)
Adjusted1
Unadjusted
1.14 (0.97, 1.34)
1.13 (0.96, 1.32)
p = 0.1373
Ref
p = 0.1031
Ref
PLOS ONE | https://doi.org/10.1371/journal.pone.0229861 March 4, 2020
7 / 12
PLOS ONE Table 5. Broad spectrum antibiotics by mode of delivery.
Mode of Delivery
N
Broad Spectrum Antibiotics Episode (95% CI)
Cesarean Section
Vaginal Delivery
1,211
2,813
0.13 (0.10, 0.15)
0.10 (0.08, 0.12)
1Adjusted for infant race, sex, birth weight, gestational age and maternal age.
https://doi.org/10.1371/journal.pone.0229861.t005
Pediatric antibiotics according to mode of delivery
Negative Binomial Rate Ratio (95% CI)
Adjusted1
Unadjusted
1.19 (0.93, 1.52)
1.20 (0.94, 1.53)
p = 0.1425
Ref
p = 0.1654
Ref
[38]. Taken together, our primary results do not support the hypothesis that otherwise healthy
full-term infants delivered via cesarean section are more likely to be prescribed antibiotics
than vaginally delivered infants.
The prevalence of pediatric antibiotic episodes in our study are generally consistent with
population-based estimates. Holstiege et al 2014, demonstrated that among five European
countries, the rates of prescriptions were highest in young children (�4 years) in all countries,
predominantly due to high use of broad spectrum penicillins [39]. In a cohort of US children
(1–17 years), Sarpong et al. reported more than a quarter (27.3%) of children used antibiotics
with 12.8% using broad2010030spectrum antibiotics and 18.5 percent using narrow-spectrum
antibiotics [27]. Notably, our analysis in children 0–3 years of age revealed that nearly 1 in 5
(18.8%) children received at least one episode of narrow spectrum antibiotics and 8.1% of chil-
dren received at least one broad spectrum antibiotic episode during the first 3 years of life. Our
results also revealed that more than 65% of pediatric antibiotics were prescribed to children
less than one year of age and that antibiotics classified as beta-lactams (J01C and J01D)
accounted for more than 95% of all postnatal prescriptions during the first three years of life. It
is worth noting that cephalosporins (J01D) in our analysis were the most common antibiotic
for children 1–3 years of age. Consistent with our results, Resi et al. reported that cephalospo-
rins were the most commonly prescribed antibiotic among children under the age of 7 years
[40]. Collectively, our results suppport the importance of electronic health records as an accu-
rate and accessible information source to track population trends in antimicrobial stewardship.
The impact of antibiotic exposure and child-maternal health outcomes has largely focused
on intrapartum antibiotic prophylaxis (IAP) to reduce neonatal early-onset sepsis (EOS) [41].
As a standard of care, most women receive IAP either for Group B Streptococcus (GBS) colo-
nization during labor[42], or for unknown GBS status during preterm labor. IAP is adminis-
tered as a form of prophylaxis for infection prior to cesarean delivery[43,44] and for preterm
premature rupture of membranes (PPROM) prior or during labor and delivery [45]. Our anal-
ysis did not formally address IAP or perinatal antibiotics; however we were able to adjust for
early-onset sepsis (EOS) risk factors and excluded children with birth weight < 2500 grams
and admission to the neonatal intensive care unit (NICU). Additionally, our analysis likely
included infants that were treated with empiric antibiotics for presumed sepsis during the first
48 hours, receiving antibiotics without NICU admission. Previous work has demonstrated
~6% of term infants treated with empiric antibiotics for presumed sepsis were exposed to
ampicillin and/or gentamicin immediately after birth [46,47]. Additional limitations of this
study included limited verification of patient antibiotic compliance and bias associated with
lack of follow-up/missing data. EHR data. There is also potential for misclassification of pedi-
atric antibiotic prescriptions and the possibility that children are receiving pediatric care (and
antibiotics) outside the UFHealth system. We would anticipate that any misclassification is
nondifferential with respect to delivery type, which admittedly would bias comparisons in
rates between the delivery groups towards the null. There is also potential for antibiotic
PLOS ONE | https://doi.org/10.1371/journal.pone.0229861 March 4, 2020
8 / 12
PLOS ONE Pediatric antibiotics according to mode of delivery
episodes calculated from inpatient and outpatient prescriptions to include antibiotics that
were prescribed but may not have been taken. As described previously[35,48], we computed
antibiotic episodes using a 10-day window and it is possible a different window of time in
between antibiotic prescriptions may result in differences by mode of delivery. Notably, we
found that two percent of the infants were administered narrow-spectrum and broad-spec-
trum within the same 10-day window, thus the occurrences of narrow and broad episodes are
not mutually exclusive.
Cesarean sections delivery and pediatric antibiotic exposure have each independently been
associated with changes in pediatric outcomes such as obesity [49] and asthma [50]. Despite
these observations, there is limited population-based data that describes how mode of delivery
informs the patterning of postnatal pediatric antibiotic prescriptions. In a large Canadian
cohort (n = 251,817) focused on antibiotics and the risk of pediatric asthma, infants that were
delivered by CS accounted for 24% of births and 43% of these children were prescribed at least
one antibiotic in their first year of life [51]. An analysis of antibiotics and pediatric obesity in
28,354 Danish mother-child dyads revealed that exposure to antibiotics during the first six
months of life was the same between children delivered vaginally and children delivered by CS
[16]. Our results extend these observations by reporting mode of delivery, in healthy term
infants, was not associated with overall antibiotic episodes, narrow-spectrum or broad-spec-
trum episodes during the first three years of life.
Our study has several strengths. It is the largest population-based study to examine this
question with more than 4,000 children followed for up to three years. We also adjusted for
potential confounding factors that have not been previously examined including method of
delivery and the presence of birth complications. In addition, our birth cohort was large
enough to adequately examine multiple courses of antibiotics and the impact of mode of deliv-
ery on broad and narrow-spectrum antibiotics. This study indicates that mode of delivery in
healthy full term infants should not be influenced by concern for future antibiotic use of infant.
Although providers should continue to strive for lower cesarean section rates in the United
States [5]; our results demonstrate that pediatric systemic antibiotic prescriptions are not sig-
nificantly different between infants born by cesarean section and those born vaginally during
the first three years of life.
Supporting information
S1 Fig. Flow chart of pediatric participants.
(DOCX)
S1 Table. Classification of narrow spectrum and broad spectrum antibiotics.
(DOCX)
S2 Table. Pediatric antibiotic episodes per child during the first three years of life.
(DOCX)
S3 Table. Cox proportional hazard model of recurrent events.
(DOCX)
Author Contributions
Conceptualization: Dominick J. Lemas.
Formal analysis: Jasmine A. Mack, Matthew J. Gurka.
Funding acquisition: Dominick J. Lemas.
PLOS ONE | https://doi.org/10.1371/journal.pone.0229861 March 4, 2020
9 / 12
PLOS ONE Pediatric antibiotics according to mode of delivery
Investigation: Dominick J. Lemas.
Methodology: Dominick J. Lemas, Jennifer J. Schoch, Nicole Cacho, Josef Neu.
Project administration: Magda Francois.
Resources: Dominick J. Lemas.
Software: Dominick J. Lemas.
Supervision: Dominick J. Lemas.
Validation: Jennifer J. Schoch, Nicole Cacho, Josef Neu.
Visualization: Dominick J. Lemas.
Writing – original draft: Dominick J. Lemas.
Writing – review & editing: Dominick J. Lemas, Elizabeth Plasencia, Alice S. Rhoton-Vlasak,
Lindsay Thompson, Magda Francois, Keval Patel, William R. Hogan, Gloria P. Lipori.
References
1. Agency for Healthcare Research and Quality. HCUP Fast Stats. 2017.
2. Hamilton BE, Martin J a, Osterman MJK. Births: Preliminary Data for 2015. Natl Vital Stat Rep. 2016;
65: 1–15.
3. Hyde MJ, Mostyn A, Modi N, Kemp PR. The health implications of birth by Caesarean section. Biol Rev.
2012; 87: 229–243. https://doi.org/10.1111/j.1469-185X.2011.00195.x PMID: 21815988
4. O’Shea TM, Klebanoff MA, Signore C. Delivery after previous cesarean: long-term outcomes in the
child. Semin Perinatol. 2010; 34: 281–292. https://doi.org/10.1053/j.semperi.2010.03.008 PMID:
20654779
5. Montoya-Williams D, Lemas DJ, Spiryda L, Patel K, Neu J, Carson TL. What Are Optimal Cesarean
Section Rates in the U.S. and How Do We Get There? A Review of Evidence-Based Recommendations
and Interventions. J Womens Health (Larchmt). 2017. https://doi.org/10.1089/jwh.2016.6188 PMID:
28825512
6. Doan E, Gibbons K, Tudehope D. The timing of elective caesarean deliveries and early neonatal out-
comes in singleton infants born 37–41 weeks’ gestation. Aust N Z J Obstet Gynaecol. 2014; 54: 340–
347. https://doi.org/10.1111/ajo.12220 PMID: 24836174
7. Natile M, Ventura ML, Colombo M, Bernasconi D, Locatelli A, Plevani C, et al. Short-term respiratory
outcomes in late preterm infants. Ital J Pediatr. 2014; 40: 52. https://doi.org/10.1186/1824-7288-40-52
PMID: 24893787
8.
Tzur T, Weintraub AY, Sheiner E, Wiznitzer A, Mazor M, Holcberg G. Timing of elective repeat caesar-
ean section: maternal and neonatal morbidity and mortality. J Matern Neonatal Med. 2011; 24: 58–64.
https://doi.org/10.3109/14767051003678267 PMID: 20230324
9. Decker E, Engelmann G, Findeisen A, Gerner P, Laass M, Ney D, et al. Cesarean delivery is associated
with celiac disease but not inflammatory bowel disease in children. Pediatrics. 2010; 125: e1433–40.
https://doi.org/10.1542/peds.2009-2260 PMID: 20478942
10. Cho CE, Norman M. Cesarean section and development of the immune system in the offspring. Am J
Obstet Gynecol. 2013; 208: 249–54. https://doi.org/10.1016/j.ajog.2012.08.009 PMID: 22939691
11. Riskin A, Gonen R, Kugelman A, Maroun E, Ekhilevitch G. Does cesarean section before the scheduled
date increase the risk of neonatal morbidity? Isr Med Assoc J. 2014; 16: 559–563. PMID: 25351013
12. Kulas T, Bursac D, Zegarac Z, Planinic-Rados G, Hrgovic Z. New Views on Cesarean Section, its Possi-
ble Complications and Long-Term Consequences for Children’s Health. Med Arch. 2013; 67: 460–463.
https://doi.org/10.5455/medarh.2013.67.460-463 PMID: 25568522
13. Magnus MC, Håberg SE, Stigum H, Nafstad P, London SJ, Vangen S, et al. Delivery by Cesarean sec-
tion and early childhood respiratory symptoms and disorders: the Norwegian mother and child cohort
study. Am J Epidemiol. 2011; 174: 1275–85. https://doi.org/10.1093/aje/kwr242 PMID: 22038100
14. Pennington AF, Strickland MJ, Klein M, Drews-Botsch C, Hansen C, Darrow LA. Caesarean delivery,
childhood asthma, and effect modification by sex: An observational study and meta-analysis. Paediatr
Perinat Epidemiol. 2018; 32: 495–503. https://doi.org/10.1111/ppe.12510 PMID: 30266042
PLOS ONE | https://doi.org/10.1371/journal.pone.0229861 March 4, 2020
10 / 12
PLOS ONE Pediatric antibiotics according to mode of delivery
15. Mårild K, Stephansson O, Montgomery S, Murray JA, Ludvigsson JF. Pregnancy outcome and risk of
celiac disease in offspring: a nationwide case-control study. Gastroenterology. 2012; 142: 39–45.e3.
https://doi.org/10.1053/j.gastro.2011.09.047 PMID: 21995948
16. Ajslev TA, Andersen CS, Gamborg M, Sørensen TIA, Jess T. Childhood overweight after establish-
ment of the gut microbiota: the role of delivery mode, pre-pregnancy weight and early administration
of antibiotics. Int J Obes (Lond). 2011; 35: 522–529. https://doi.org/10.1038/ijo.2011.27 PMID:
21386800
17. Goldani H a S, Bettiol H, Barbieri M a, Silva A a M, Agranonik M, Morais MB, et al. Cesarean delivery is
associated with an increased risk of obesity in adulthood in a Brazilian birth cohort study. Am J Clin
Nutr. 2011; 93: 1344–1347. https://doi.org/10.3945/ajcn.110.010033 PMID: 21508088
18. Barros FC, Matijasevich A, Hallal PC, Horta BL, Barros AJ, Menezes AB, et al. Cesarean section and
risk of obesity in childhood, adolescence, and early adulthood: evidence from 3 Brazilian birth cohorts.
Am J Clin Nutr. 2012; 95: 465–470. https://doi.org/10.3945/ajcn.111.026401 PMID: 22237058
19. Huang X-Z, Zhu L-B, Li Z-R, Lin J. Bacterial colonization and intestinal mucosal barrier development.
World J Clin Pediatr. 2013; 2: 46–53. https://doi.org/10.5409/wjcp.v2.i4.46 PMID: 25254174
20. Mesquita DN, Barbieri MA, Goldani HA, Cardoso VC, Goldani MZ, Kac G, et al. Cesarean Section Is
Associated with Increased Peripheral and Central Adiposity in Young Adulthood: Cohort Study. PLoS
One. 2013; 8: e66827. https://doi.org/10.1371/journal.pone.0066827 PMID: 23826150
21. Wang L, Alamian A, Southerland J, Wang K, Anderson J, Stevens M. Cesarean section and the risk of
overweight in grade 6 children. Eur J Pediatr. 2013; 172: 1341–1347. https://doi.org/10.1007/s00431-
013-2043-2 PMID: 23708216
22. Montoya-Williams D, Lemas DJDJ, Spiryda L, Patel K, Carney OOOO, Neu J, et al. The Neonatal
Microbiome and Its Partial Role in Mediating the Association between Birth by Cesarean Section and
Adverse Pediatric Outcomes. Neonatology. 2018; 114: 103–111. https://doi.org/10.1159/000487102
PMID: 29788027
23. Huh SY, Rifas-Shiman SL, Zera CA, Edwards JWR, Oken E, Weiss ST, et al. Delivery by caesarean
section and risk of obesity in preschool age children: a prospective cohort study. Arch Dis Child. 2012;
97: 610–616. https://doi.org/10.1136/archdischild-2011-301141 PMID: 22623615
24. Mueller NT, Bakacs E, Combellick J, Grigoryan Z, Dominguez-Bello MG. The infant microbiome devel-
opment : mom matters. Trends Mol Med. 2015; 21: 1–9.
25. Chai G, Governale L, McMahon AW, Trinidad JP, Staffa J, Murphy D. Trends of Outpatient Prescription
Drug Utilization in US Children, 2002–2010. Pediatrics. 2012; 130: 23–31. https://doi.org/10.1542/peds.
2011-2879 PMID: 22711728
26. Di Martino M, Lallo A, Kirchmayer U, Davoli M, Fusco D. Prevalence of antibiotic prescription in pediatric
outpatients in Italy: the role of local health districts and primary care physicians in determining variation.
A multilevel design for healthcare decision support. BMC Public Health. 2017; 17: 886. https://doi.org/
10.1186/s12889-017-4905-4 PMID: 29149875
27. Sarpong EM, Miller GE. Narrow- and Broad-Spectrum Antibiotic Use among U.S. Children. Health Serv
Res. 2015; 50: 830–846. https://doi.org/10.1111/1475-6773.12260 PMID: 25424240
28. Hayward GN, Fisher RFR, Spence GT, Lasserson DS. Increase in antibiotic prescriptions in out-of-
hours primary care in contrast to in-hours primary care prescriptions: service evaluation in a population
of 600000 patients. J Antimicrob Chemother. 2016; 71: 2612–2619. https://doi.org/10.1093/jac/dkw189
PMID: 27287234
29. Majeed A, Moser K. Age- and sex-specific antibiotic prescribing patterns in general practice in England
and Wales in 1996. Br J Gen Pract. 1999; 49: 735–6. PMID: 10756619
30. Marra F, Patrick DM, Chong M, Bowie WR. Antibiotic use among children in British Columbia, Can-
ada. J Antimicrob Chemother. 2006; 58: 830–839. https://doi.org/10.1093/jac/dkl275 PMID:
16921182
31. Rao GG. Risk Factors for the Spread of Antibiotic-Resistant Bacteria. Drugs. 1998; 55: 323–330.
https://doi.org/10.2165/00003495-199855030-00001 PMID: 9530540
32.
Langdon A, Crook N, Dantas G. The effects of antibiotics on the microbiome throughout development
and alternative approaches for therapeutic modulation. 2016. https://doi.org/10.1186/s13073-016-
0294-z PMID: 27074706
33. Sharma S, Bowman C, Alladin-Karan B, Singh N. Antibiotic prescribing patterns in the pediatric emer-
gency department at Georgetown Public Hospital Corporation: a retrospective chart review. BMC Infect
Dis. 2016; 16: 170. https://doi.org/10.1186/s12879-016-1512-4 PMID: 27094391
34. Miller JE, Wu C, Pedersen LH, de Klerk N, Olsen J, Burgner DP. Maternal antibiotic exposure during
pregnancy and hospitalization with infection in offspring: a population-based cohort study. Int J Epide-
miol. 2018; 47: 561–571. https://doi.org/10.1093/ije/dyx272 PMID: 29415232
PLOS ONE | https://doi.org/10.1371/journal.pone.0229861 March 4, 2020
11 / 12
PLOS ONE Pediatric antibiotics according to mode of delivery
35. Block JP, Bailey LC, Gillman MW, Lunsford D, Daley MF, Eneli I, et al. Early Antibiotic Exposure and
Weight Outcomes in Young Children. Pediatrics. 2018; 142: e20180290. https://doi.org/10.1542/peds.
2018-0290 PMID: 30381474
36. Smith N, Coleman KJ, Lawrence JM, Quinn VP, Getahun D, Reynolds K, et al. Body weight and height
data in electronic medical records of children. Int J Pediatr Obes. 2010; 5: 237–42. https://doi.org/10.
3109/17477160903268308 PMID: 19961272
37.
2019. WHO Collaborating Centre for Drug Statistics Methodology, Guidelines for ATC classification
and DDD assignment. Oslo;
38. Nordahl-Hansen A, Øien RA, Volkmar F, Shic F, Cicchetti DV. Enhancing the understanding of clinically
meaningful results: A clinical research perspective. Psychiatry Res. 2018; 270: 801–806. https://doi.
org/10.1016/j.psychres.2018.10.069 PMID: 30551328
39. Holstiege J, Schink T, Molokhia M, Mazzaglia G, Innocenti F, Oteri A, et al. Systemic antibiotic prescrib-
ing to paediatric outpatients in 5 European countries: a population-based cohort study. BMC Pediatr.
2014; 14: 174. https://doi.org/10.1186/1471-2431-14-174 PMID: 24997585
40. Resi D, Milandri M, Moro ML, Emilia Romagna Study Group On The Use Of Antibiotics In Children. Anti-
biotic prescriptions in children. J Antimicrob Chemother. 2003; 52: 282–286. https://doi.org/10.1093/jac/
dkg302 PMID: 12865400
41. Mukhopadhyay S, Puopolo KM. Risk Assessment in Neonatal Early Onset Sepsis. Semin Perinatol.
2012; 36: 408–415. https://doi.org/10.1053/j.semperi.2012.06.002 PMID: 23177799
42. Verani JR, McGee L, Schrag SJ, Division of Bacterial Diseases, National Center for Immunization and
Respiratory Diseases, Centers for Disease Control and Prevention (CDC). Prevention of perinatal
group B streptococcal disease—revised guidelines from CDC, 2010. MMWR Recomm reports Morb
Mortal Wkly report Recomm reports. 2010; 59: 1–36.
43.
Tita ATN, Szychowski JM, Boggess K, Saade G, Longo S, Clark E, et al. Adjunctive Azithromycin Pro-
phylaxis for Cesarean Delivery. N Engl J Med. 2016; 375: 1231–41. https://doi.org/10.1056/
NEJMoa1602044 PMID: 27682034
44. Ward E, Duff P. A comparison of 3 antibiotic regimens for prevention of postcesarean endometritis: an
historical cohort study. Am J Obstet Gynecol. 2016; 214: 751.e1–4. https://doi.org/10.1016/j.ajog.2016.
02.037 PMID: 26901276
45. Mercer BM, Miodovnik M, Thurnau GR, Goldenberg RL, Das AF, Ramsey RD, et al. Antibiotic therapy
for reduction of infant morbidity after preterm premature rupture of the membranes. A randomized con-
trolled trial. National Institute of Child Health and Human Development Maternal-Fetal Medicine Units
Network. JAMA. 1997; 278: 989–95. PMID: 9307346
46. Persaud RR, Azad MB, Chari RS, Sears MR, Becker AB, Kozyrskyj AL, et al. Perinatal antibiotic expo-
sure of neonates in Canada and associated risk factors: a population-based study. J Matern Neonatal
Med. 2015; 28: 1190–1195. https://doi.org/10.3109/14767058.2014.947578 PMID: 25053193
47. Glasgow TS, Speakman M, Firth S, James B, Byington CL, Young PC. Clinical and economic outcomes
for term infants associated with increasing administration of antibiotics to their mothers. Paediatr Perinat
Epidemiol. 2007; 21: 338–346. https://doi.org/10.1111/j.1365-3016.2007.00811.x PMID: 17564591
48. Heerman WJ, Daley MF, Boone-Heinonen J, Rifas-Shiman SL, Bailey LC, Forrest CB, et al. Maternal
antibiotic use during pregnancy and childhood obesity at age 5 years. Int J Obes. 2019; 1. https://doi.
org/10.1038/s41366-018-0316-6 PMID: 30670848
49. Mueller NT, Whyatt R, Hoepner L, Oberfield S, Dominguez-Bello MG, Widen EM, et al. Prenatal expo-
sure to antibiotics, cesarean section and risk of childhood obesity. Int J Obes (Lond). 2015; 39: 665–70.
https://doi.org/10.1038/ijo.2014.180 PMID: 25298276
50. Wu P, Feldman AS, Rosas-Salazar C, James K, Escobar G, Gebretsadik T, et al. Relative Importance
and Additive Effects of Maternal and Infant Risk Factors on Childhood Asthma. PLoS One. 2016; 11:
e0151705. https://doi.org/10.1371/journal.pone.0151705 PMID: 27002979
51. Marra F, Marra CA, Richardson K, Lynd LD, Fitzgerald MJ. Antibiotic consumption in children prior to
diagnosis of asthma. 2011; 11: 32. https://doi.org/10.1186/1471-2466-11-32 PMID: 21627795
PLOS ONE | https://doi.org/10.1371/journal.pone.0229861 March 4, 2020
12 / 12
PLOS ONE |
10.1371_journal.pone.0227132 | RESEARCH ARTICLE
Defining pharmacists’ roles in disasters: A
Delphi study
Kaitlyn E. WatsonID
1,2*, Judith A. Singleton1,2, Vivienne Tippett1,2, Lisa M. Nissen2
1 School of Clinical Sciences, Queensland University of Technology, Brisbane, Australia, 2 Institute of Health
and Biomedical Innovation, Queensland University of Technology, Brisbane, Australia
* kaitlyn.watson@connect.qut.edu.au
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
Abstract
Introduction
OPEN ACCESS
Citation: Watson KE, Singleton JA, Tippett V,
Nissen LM (2019) Defining pharmacists’ roles in
disasters: A Delphi study. PLoS ONE 14(12):
e0227132. https://doi.org/10.1371/journal.
pone.0227132
Editor: Michio Murakami, Fukushima Medical
University School of Medicine, JAPAN
Received: October 9, 2019
Accepted: December 11, 2019
Published: December 26, 2019
Copyright: © 2019 Watson et al. This is an open
access article distributed under the terms of the
Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: All relevant data are
within the paper and its Supporting Information
files.
Funding: The authors received no specific funding
for this work.
Competing interests: The authors have declared
that no competing interests exist.
Abbreviations: CBRN, chemical, biological,
radiological and nuclear; CPR, cardiopulmonary
resuscitation; IGEM, Inspector-General for
Emergency Management; PPRR, prevention,
Pharmacists are uniquely placed in the community to be of assistance to disaster-affected
patients. However, the roles undertaken by pharmacists in disasters are identified based on
their own experiences and networks. There is currently no definition or acknowledgment of
pharmacists’ roles in disasters.
Objective
To acquire consensus from an expert panel of key opinion leaders within the field of disaster
health on pharmacists’ roles in disasters throughout the four disaster phases—prevention,
preparedness, response, and recovery.
Methods
A Delphi study consisting of three rounds of online surveys was utilised. Twenty-four key
opinion leaders were contacted, with 15 completing all three rounds. The 15 expert panel-
lists were presented with 46 roles identified in the literature and asked to rank their opinions
on a 5-point Likert scale. This study used an international, all-hazard, and multijurisdictional
approach. Consensus was benchmarked at 80% and any role which did not reach consen-
sus was re-queried in the subsequent round. The third round provided the results of the Del-
phi study and sought commentary on the acceptance or rejection of the roles.
Results
Of the 46 roles provided to the expert panel, 43 roles were accepted as roles pharmacists
are capable of undertaking in a disaster. There were five roles for the prevention phase,
nine for the preparedness phase, 21 for the response phase, and eight for the recovery
phase. The experts were asked to prioritise the top five roles for each of the disaster phases.
The three roles which did not make consensus were deemed to be specialised roles for
disaster pharmacists and not generalisable to the broader pharmacy profession.
PLOS ONE | https://doi.org/10.1371/journal.pone.0227132 December 26, 2019
1 / 13
preparedness, response, and recovery; US, United
States.
Defining pharmacists’ roles in disasters: A Delphi study
Conclusion
This study identifies pharmacists’ roles in disasters which have been accepted by the inter-
national disaster health community. The international key opinion leaders recommended
that pharmacists could be undertaking 43 roles in a disaster, however, this is dependent on
individual jurisdiction considerations. Pharmacy professional associations need to advocate
to policymakers for legislative support and to ensure pharmacists are equipped with the
training and education required to undertake these roles within specific jurisdictions.
Introduction
Pharmacists are uniquely positioned during disasters to provide healthcare continuity and
medication management to affected communities. It has been acknowledged that pharmacists
are the most widely distributed healthcare professional, being more accessible than supermar-
kets, banks, or medical centres.[1] However, pharmacists’ roles in disasters both formal and
informal are currently undefined or acknowledged.
Anecdotally, pharmacists have been undertaking various tasks in disasters in an ad-hoc
fashion to assist their communities.[2–5] However, pharmacists’ roles in disasters did not
become of significant interest in the literature until the events of September 11, 2001 in the
United States (US).[6] Prior to 2001, accepted roles for pharmacists generally focused on their
contributions to logistics and supply chain management.[7] Post 2001, pharmacists began to
be recognised for their informal contributions in patient care and public health during disas-
ters, although their primary purpose and formal role was still logistics.[6, 8]
The strength that pharmacists provide during disasters, is their positioning in the commu-
nity. They are considered community landmarks or communication hubs for patients and are
typically the first healthcare service to resume operations after a disaster has occurred.[9, 10]
Roles undertaken by pharmacists in disasters are within their scope of practice and are typi-
cally an extension of their daily duties (i.e. continuity of medication management). For exam-
ple, patients tend to evacuate their homes without taking their medications, prescriptions,
money or identification.[8] In this circumstance during a disaster, with the collapse of com-
munity healthcare services, pharmacists can provide a limited emergency supply of the
patient’s chronic disease medications. [9–11] This can alleviate the healthcare burden on over-
crowded emergency rooms and free up doctors and nurses to treat higher-acuity patients.
Hurricane Katrina, a Category Four hurricane, significantly affected parts of the US in 2005
[12] and highlighted the impact pharmacists can have in the response and recovery of a com-
munity following a disaster. It was found pharmacists performed many duties in the wake of
Hurricane Katrina and fulfilled additional roles in the absence of other healthcare profession-
als.[13] Some of the duties pharmacists performed following Hurricane Katrina included:[13–
17]
• triaging services within evacuation centres (separating those patients needing to see a doctor
from those who simply needed a prescription refill, and identifying and referring individuals
to allied health professionals),
• taking medication histories,
• providing vaccinations,
• performing basic medical checks,
PLOS ONE | https://doi.org/10.1371/journal.pone.0227132 December 26, 2019
2 / 13
Defining pharmacists’ roles in disasters: A Delphi study
• mixing intravenous medications,
• providing consultations on wound infections,
• assisting with major traumas,
• assessing the contamination risk of medications brought in, and,
• pill identification.
Most these roles identified during Hurricane Katrina are services pharmacists provide daily
but become significantly more important during disasters when there is a potential collapse of
community healthcare services, or affected communities become cut off from outside assis-
tance. In recognition of the crucial role pharmacists could play in alleviating the burden of low-
acuity non-communicable disease patients, Alabama State (US) temporarily extended (during
Hurricane Katrina) its ‘emergency supply’ rule. This allowed pharmacists to prescribe a 30-day
emergency supply (increased from the everyday 3-day emergency supply rule) of chronic dis-
ease medications for disaster-affected patients without a prescription. Alabama State (US) also
permitted the use of out-of-state volunteer pharmacists to assist in the relief efforts.[18, 19]
In November 2016, the Thunderstorm Asthma event affected the city of Melbourne, Aus-
tralia. The Inspector-General for Emergency Management (IGEM) review coordinated by the
Victorian State Government of the thunderstorm asthma emergency response, found commu-
nity pharmacies and pharmacists were essential in the response and should be better integrated
into disaster management teams for effective coordinated responses.[20] Pharmacists were
integral in providing medication management for this event, due to the volume and accessibil-
ity of pharmacies in the communities, patients were able to get assistance quickly. This is
highlighted in Finding Six of the IGEM report:
“IGEM finds that on 21–22 November 2016 community pharmacies played a central role in
meeting community needs during the thunderstorm asthma event. Given their community
focus and their geographic coverage, community pharmacies can provide valuable support to
the management of health emergencies or emergencies with health impacts.”[20] (p.36)
The literature has highlighted the valuable assistance pharmacists can provide in ensuring
continuity of medication care. However, the formal acknowledgment or translation of these
pharmacists’ roles into disaster health management has yet to be achieved beyond the estab-
lished pharmacists’ role in logistics. To date, pharmacists have been operating their pharmacy
disaster management response with limited formal guidelines to follow, separately to the coor-
dinated health response. The purpose of this study was to convene an expert panel of key opin-
ion leaders to discuss and begin to define and acknowledge pharmacists’ roles in disasters.
Specifically, the aim was to obtain consensus on pharmacists’ roles within each of the disaster
phases–prevention, preparedness, response, and recovery (PPRR). This study begins the pro-
cess of formally acknowledging and providing acceptance from the international disaster
health community on the roles undertaken by pharmacists in disasters. The potential pharma-
cists’ roles were identified from the literature[13–17] and a larger study conducted by the
research team on pharmacists’ roles in disasters which included surveys and interviews with
international disaster health professionals.[9]
Methods
Data collection and participant recruitment
A Delphi study is a process of gathering information through a series of survey rounds and is
commonly used to obtain consensus on matters.[21] There are currently no universally
PLOS ONE | https://doi.org/10.1371/journal.pone.0227132 December 26, 2019
3 / 13
Defining pharmacists’ roles in disasters: A Delphi study
accepted parameters for completing a Delphi study. However, it is generally agreed that,
rounds should continue until an adequate consensus level is reached or the survey results
reach stability with panellists no longer revising their rankings–typically achieved after three
rounds.[21] The Delphi procedure should involve ‘controlled feedback’ providing the panel-
lists a summary of the comments made and a simple statistical summary of the entire panel’s
position.[21, 22] The purpose of feedback and including both the statistical measure and quali-
tative comments, allows for the individual panellists to determine where their response lies rel-
ative to the overall group’s and can assist in their revision of their ranking in future Delphi
rounds.[21] The Delphi technique used should be systematic and transparent in its findings of
reaching consensus.
This Delphi study asked the participants to take an international, all-hazard, and multijuris-
dictional approach to disasters and evaluate pharmacists at the level of the profession rather
than referring to a specific employment context. Since, the study was international in scope,
panellists were asked not to restrict the pharmacist’s roles based on their current jurisdiction’s
legalisation framework. The researchers sought to investigate, whether the panellists perceived
that pharmacists have the expertise, skills, and knowledge to undertake the potential roles in
all-types of disasters.
Twenty-four national and international opinion leaders were identified by the research
team as experts on health aspects of disaster management, advancing practice of pharmacy,
and knowledge of pharmacists’ roles in disasters. These experts were identified from various
international and Australian non-governmental organisations, government, pharmacy, mili-
tary, public health, and disaster management agencies and were recruited through snowball
and purposive sampling. The utilisation of snowball sampling techniques was performed to
recruit additional potential panel members and account for the potential inherent bias of
investigator selection. The experts were identified as leaders of their respective agencies and
influential in matters of disaster management and advancing pharmacy practice. The countries
included in this study were New Zealand, United Kingdom, Pan America, Europe, and Austra-
lia. The experts were recruited for their expertise on the subject matter and not specifically for
their country of origin as many of the panellists have international interests. Initially, 24 opin-
ion leaders were contacted twice to invite them to participate on the panel. Potential partici-
pants were provided with the participant information sheet. Of the 24 opinion leaders
contacted, 15 consented and completed the three rounds of surveys. Consent was implied with
submission of the completed surveys. Ethics was obtained from the Queensland University of
Technology human research ethics committee, approval number 1700000106.
The Delphi panel were presented with 45 initial roles in Round 1 of this study. These 45
roles were identified from the literature as roles pharmacists were already undertaking in vari-
ous jurisdictions[13–17] and roles identified from a larger study conducted by the research
team which included surveys and interviews with international disaster health professionals.
[9] The panellists were invited to suggest any additional roles they deemed appropriate which
were presented to the expert panel in the subsequent rounds. There was one additional role
added after Round One by a panellist, bringing the total number of roles assessed by the
experts to 46 roles. Each of the 46 potential roles were categorised according to where in the
PPRR cycle they would fit operationally. The consensus benchmark was set at 80%, as has been
used in previous Delphi Studies on health outcomes and disaster management.[23, 24] For
Round One, roles which received an ‘agree or strongly agree’ rating by at least 80% of the
expert panel were accepted as having reached consensus and indicated agreement that phar-
macists were perceived to be capable of undertaking these roles in a disaster. Roles which fell
below 69% were required to confirm removal using a dichotomous ‘yes’ or ‘no’ question in the
second round. Panellists were asked to provide a reason why pharmacists could not undertake
PLOS ONE | https://doi.org/10.1371/journal.pone.0227132 December 26, 2019
4 / 13
Defining pharmacists’ roles in disasters: A Delphi study
this role. Roles which fell in between this range of 70%-79% were presented back to the panel
in the second round with a four-point Likert scale, asking for panellists to agree or disagree
with the role. Based on feedback from the panellists in round one, one new role was added in
for subsequent rounds increasing the 45 potential roles to 46. The results from Round One
were presented back to the panellists as ‘controlled feedback’[21, 22] and the experts were
asked to revise their rankings of the 46 roles and to provide commentary. The second round
preceded in the same manner as the first with all 15 participants providing their revised rank-
ings and comments. For the final round, as the results had not changed between the first two
rounds, the participants were provided with the final results and asked to confirm/accept them
and provide any final comments.
Data analysis
The survey Likert scale responses were exported from KeySurvey1 into IBM1 SPSS1 Statisti-
cal software version 25. Frequencies of each role in the Likert scale responses were analysed.
The five-point Likert scales were trichotomized into a three-point Likert scale, with ‘disagree’
representing ‘1–2’, ‘neutral’ remaining ‘3’ and ‘agree’ representing ‘4–5’. Consensus was
reached if 80% of the participants scored ‘4–5’ representing ‘agree’. This was similar in the sec-
ond round which utilised a four-point Likert scale, however, the option of a neutral ‘3’
response was removed.
The qualitative comments provided by the experts were presented back to the panellists as
arguments for and against pharmacists undertaking that specific role in disasters in the pro-
ceeding rounds and in the final results. This allowed the panellists to revise their rankings
based on the ‘controlled feedback’ provided. It is suggested the survey feedback process should
continue until either consensus is reached or the panellists stop revising their answers.[22] Sta-
bility was reached for this Delphi Study within the three rounds, as the experts no longer con-
tinued to revise their rankings and 43 roles reached consensus.
Results
The expert panel consisted of key opinion leaders from international organisations considered
experts in disaster health management and/or pharmacy. The original makeup of the potential
panellist group contacted was evenly distributed: 33% pharmacy background, 33% disaster
management, government, and medical background, and 33% experts with extensive experi-
ence in both backgrounds. Table 1 outlines the composition of the expert panel and self-identi-
fied background of the 15 panellists.
The panellists were then asked to identify the different perspectives in which they were
approaching the Delphi study. They were able to tick all perspectives that applied to them, thus
increasing the denominator from the 15 panellists to 20 differing perspectives. The breakdown
for the panellists’ perspectives were 15% (3/20) from disaster and emergency services, 30% (6/
20) from government and policy perspective, 10% (2/20) were ‘other’ (disaster management
Table 1. Delphi study expert panel composition.
Self-identified Background
Disaster Management, Government and Medical
Pharmacy
Both
�Total N = 15
https://doi.org/10.1371/journal.pone.0227132.t001
n (%) �
6 (40%)
5 (33%)
4 (27%)
PLOS ONE | https://doi.org/10.1371/journal.pone.0227132 December 26, 2019
5 / 13
Defining pharmacists’ roles in disasters: A Delphi study
and consultancy), and 45% (9/20) were from pharmacy. Of 14 experts who provided their age,
the median age of the expert panel was 50.5 years (IQR 16.25 years), with the oldest participant
being 72 years and the youngest 36 years old. There was seven females and eight males.
Out of 46 roles presented to the expert panel, consensus was reached on 43 roles. The final
43 roles deemed appropriate for pharmacists to undertake in a disaster are listed below, catego-
rised according to each disaster PPRR phase. The five roles in the prevention phase are listed
in Table 2, the nine roles in the preparedness phase are listed in Table 3, the 21 roles in the
response phase are provided in Table 4, and the eight roles in the recovery phase are listed in
Table 5. Each Delphi round results are presented in the Supporting Information (S1–S4
Tables).
Consensus was not reached on the following three roles and therefore they were removed
from the overall list of accepted roles pharmacists can formally undertake in disasters. The
roles removed were:
1. Develop educational tools for health professionals on preparedness, signs and symptoms
and drug treatments for CBRN (chemical, biological, radiological and nuclear) weapons
2. Making dose adjustments to existing therapeutic regimens where clinically necessary
3. Pharmacists providing behavioural and mental health support to their patients, customers,
and staff following a disaster
The expert panel were asked to prioritise the top five pharmacists’ roles for each of the
PPRR phases. Their prioritisation of pharmacists’ roles is provided in Table 6.
Discussion
This Delphi study followed the guidelines outlined in a systematic review by Sinha, Smyth and
Williamson.[25] This method provided a systematic and transparent approach resulting in rig-
orous findings. Panellists recommended 43 out of 46 possible roles that they believe pharma-
cists are capable of undertaking across the PPRR cycle. Utilising a Delphi study technique to
evaluate pharmacists’ capabilities is not a new endeavour. In 2012, Kennie-Kaulbach et al. uti-
lised a modified-Delphi study to appraise pharmacists’ competencies in their provision of pri-
mary health care.[26] Delphi studies have also been widely used across the disaster
management research space.[24, 27, 28]
This study utilised an international, all-hazard, multijurisdictional approach. Disaster liter-
ature has identified that disasters are transcending boundaries (i.e. geographical, organisa-
tional, academic), making them complex in their impact on the communities, the
organisations involved, and in the response required.[29–31] Healthcare systems, infrastruc-
ture, and resources including pharmacy need to be prepared for any type of emergency situa-
tion. This Delphi study discussed pharmacy at the international profession level as often
disasters are multijurisdictional and require the assistance of pharmacists from different
Table 2. Final Consensus on five roles pharmacists are capable of undertaking in the prevention phase of a
disaster.
Prevention/Mitigation—reduce the health risks posed by hazards
Administer vaccinations
Educate the public on reducing the spread of communicable diseases/infections
Tailored ’point of care’ messaging to chronic disease patients
Ensuring patients are aware of their increased risk of adverse health outcomes in a disaster
Optimising medication supplies for chronic disease management
https://doi.org/10.1371/journal.pone.0227132.t002
PLOS ONE | https://doi.org/10.1371/journal.pone.0227132 December 26, 2019
6 / 13
Defining pharmacists’ roles in disasters: A Delphi study
Table 3. Final Consensus on nine roles pharmacists are capable of undertaking in the preparedness phase of a
disaster.
Preparedness—ensure timely and effective response systems are in place
Ensuring uninterrupted supply of medications in a disaster
Knowing how to access national pharmaceutical stockpiles and who to contact if necessary
Develop business continuity plans that include disaster management to ensure sustainability of service
Developing drug algorithms and treatment guidelines to determine drug choice based on co-morbidities in the
event of bio terrorism (e.g. Anthrax, Plague, Tularaemia—requiring antibiotics/prophylaxis measures)
Being a part of local/state/national disaster preparedness health meetings—providing medication management
advice
Being a part of the local community disaster management teams to involve pharmacy in coordinated response
Maintain systems and process for the reconciliation and security of controlled drugs (e.g. morphine,
oxycodone, etc.)
Have systems in place to secure cold chain lines
Develop a list of at-risk patients in their community
https://doi.org/10.1371/journal.pone.0227132.t003
employment contexts and jurisdictions. Thus, this study sought to begin to outline pharma-
cists’ scope of practice in disasters and disaster health management.
Despite reaching some consensus on the scope of potential roles for pharmacists, not all
countries have supportive legislative frameworks and polices to facilitate all 43 roles when the
next disaster impacts their community. A review of legislation as well as support from
Table 4. Final Consensus on 21 roles pharmacists are capable of undertaking in the response phase of a disaster.
Response—action in disaster/emergency
Coordinating logistics of medications and medical supplies for patients with chronic diseases
Rationing limited supplies of medications
Assisting with the release and allocation of national stockpiles if required in pandemic or emergency
Triage of low-acuity patients. (e.g. medication reconciliation, patient medical history, referring to physician for
further assessment or to pharmacist for refill of lost medications)
Institute cardiopulmonary resuscitation (CPR)
Provide wound care and first aid for minor ailments
Providing one off medication emergency supply refills for up to 30 days during the declared disaster
Continue provision of chronic disease medications
Dispense medications and other necessary medication-related items to affected members of the community
(prescription, over-the-counter medications, inhalers, etc.)
Dispense general health pharmacy items to affected members of the community (toiletries, nappies, bandages,
incontinence pads, water, etc.)
Making therapeutic substitutions for drugs available on limited formularies without prior authorisation
Counselling patients on how to use and take medications
Prescribing and administering vaccinations (e.g. tetanus, antidote/prophylaxis to bio-terrorism agent following
state public health disaster protocols)
Attend clinical ward rounds to provide pharmacist expertise on medical patients
Prescribe medication needs of low-acuity patients in hospital
Medication identification and safety assessment
Monitoring the chronic disease(s) of at-risk individuals to minimise exacerbation
Advocate pharmacy’s role during an event
Maintain media liaison on medication issues
Decide on the appropriateness of donated medications and other supplies
Pharmacists should engage the pharmacy student workforce to backfill duties (dispensing, inventory, etc.),
freeing up pharmacists to perform more clinical roles in a disaster.
https://doi.org/10.1371/journal.pone.0227132.t004
PLOS ONE | https://doi.org/10.1371/journal.pone.0227132 December 26, 2019
7 / 13
Defining pharmacists’ roles in disasters: A Delphi study
Table 5. Final Consensus on eight roles pharmacists are capable of undertaking in the recovery phase of a
disaster.
Recovery—returning to ’normal’ business and beyond
Check on the health needs of the local community
Re-establish normal stock levels, destroy contaminated stock appropriately
Restock emergency/ disaster kits for next disaster event
Identify and prioritise vulnerable patients in local community
Restore order to patient records and drug records, if manually written due to power outages
Document what worked and what didn’t in the disaster response and change disaster plans accordingly
Participate in post-disaster research/reports
Inform local disaster management reports on pharmacy response improvements
https://doi.org/10.1371/journal.pone.0227132.t005
pharmacy organisations and disaster management organisations is required to collaboratively
instigate these roles across the PPRR cycle. The acknowledged pharmacists’ roles are designed
to be undertaken in a cooperative team dynamic in partnership with other stakeholders. This
could be facilitated by better integration and utilisation of pharmacists into disaster teams.
This study provides evidence to policymakers on the acknowledgment by international key
opinion leaders in the disaster health community of the accepted roles pharmacists are capable
of undertaking in disasters. However, it is up to the pharmacy profession to instigate the capa-
bility of these pharmacists’ roles through training for specialist roles and expanding pharmacy
operational environments. It is up to the pharmacy professional bodies to take a disaster phar-
macy proposal to the government about how more extensive involvement of pharmacists in
communities and on deployment could be achieved. Together, pharmacy professional bodies
and policymakers can ensure the gaps identified in pharmacy legislation and knowledge are
met, fully utilising pharmacists’ scope of practice and better integrating primary healthcare
providers within the disaster management structure.
There were three posited roles for pharmacists which generated considerable debate in this
study. The potential role for pharmacists in CBRN education was highlighted by the panellists
as perhaps a specialist role for pharmacists requiring additional training and therefore was not
adopted in the definition and consensus. This is corroborated by pharmacists working within
the Office of Public Health Preparedness teams in developing medication management guid-
ance for CBRN disasters for the US in the last 10 years.[32] However, in Canada, community
pharmacists are identified as one of the first professionals to be potentially faced with a CBRN
incident and therefore they are included in the CBRN first responder basic training program.
[33] The training outlines that being a CBRN first responder is not intervening in the incident
but recognising the signs and symptoms and responding appropriately.[33] For pharmacists
specifically, the signs and symptoms of biological attacks may present in pharmacies as an
unusual acceleration of people coming in with similar ‘flu-like’ symptoms. Terriff and Tee
reported in 2001 that pharmacists have a role in the rapid dispersal of antidotes and informa-
tion on both treatment and prophylaxis measures in a bioterrorism event.[34] They suggest
many of the medications used in the treatment and prophylaxis of bioterrorism agents are
stocked in pharmacies.[34] It has been suggested pharmacists do have a role in the planning
for these events by keeping the information on the medications and vaccines used in a CBRN
disaster up-to-date for other health professionals.[34, 35]
The potential pharmacists’ role in providing dose adjustment in disasters did not reach con-
sensus. The panellists identified this role requires close collaboration with the prescriber which
they identified not all current pharmacy models in a disaster are capable of achieving.
PLOS ONE | https://doi.org/10.1371/journal.pone.0227132 December 26, 2019
8 / 13
Defining pharmacists’ roles in disasters: A Delphi study
Table 6. Role prioritisation of pharmacists’ roles for the disaster PPRR phases obtained in round 1 of the Delphi study by the expert panel.
Priority
1st
2nd
3rd
4th
5th
Prevention
Preparedness
Response
Recovery
Optimising medication supplies
for chronic disease management
Ensuring uninterrupted supply of
essential medications in a disaster
Dispense medications and other necessary
medication-related items to affected members
of the community (prescription, over-the-
counter medications, inhalers)
Re-establish normal stock
levels, destroy contaminated
stock appropriately
Administer vaccinations
Have systems in place to secure cold
chain lines
Counselling patients on how to use and take
medications�
Restock emergency/ disaster
kits for next disaster event
Educate the public on reducing
the spread of communicable
diseases/infections
Being a part of the local community
disaster management teams to involve
pharmacy in coordinated response
Coordinating logistics of medications and
medical supplies for patients with chronic
diseases�
Check on the health needs of
the local community��
Ensuring patients are aware of
their increased risk of adverse
health outcomes in a disaster
Tailored ’point of care’
messaging to chronic disease
patients
Knowing how to access national
stockpiles if necessary
Providing one off medication emergency supply
refills for up to 30 days during the declared
disaster
Identify and prioritise
vulnerable patients in local
community��
Being a part of local/state/national
disaster preparedness health meetings
—providing medication management
advice
Assisting with the release and allocation of
national stockpiles if required in pandemic or
emergency
Restore order to patient records
and drug records, if manually
written due to power outages
�Tied for second prioritisation
��Tied for third prioritisation
https://doi.org/10.1371/journal.pone.0227132.t006
However, some countries have integrated dose-adjustment as part of the everyday responsibili-
ties of pharmacists.[36]
Pharmacists’ potential role in providing mental health and behaviour first-aid support did
not reach consensus and therefore was not adopted in the definition. The discussion of this
role by the panellists was very split. Some of the panellists believed mental health support fol-
lowing a disaster should be left up to the professionals. Whereas, the other half of the panellists
believed this role was an extension of the mental health support role pharmacists fulfil in the
community daily. In disasters, the principle of psychological first-aid or behavioural support is
similar to the daily mental health support role in the recognition of warning signs and referral
to appropriate healthcare services but is specific to the type of trauma experienced.[37] Both
sides of the panellist’s debate are accurate and reflect the need for referral onto appropriate
healthcare services. Research has suggested pharmacists are ideally placed as potential first
responders for the screening of those requiring professional assistance for mental health crises,
as they assist with mental health conditions as part of their daily practice.[38] In Australia, the
Mental Health First Aid1 training program has begun to be incorporated into undergraduate
pharmacy degree curricula. The aim of this program is to educate and upskill pharmacists in
the recognition of a mental health crisis and to identify the appropriate health services and
professionals to which to refer their patients.[39]
The expert panel revised their ranking and came to consensus on the pharmacists’ role in
CPR. Pharmacists as members on CPR teams are often not the professional doing the com-
pressions, with the skill sets of other health professionals better equipped to perform that task.
Pharmacists can provide important drug information, perform dose calculations, prepare
drugs for administration, and record the drugs used during a code blue.[40] However, there
may be instances where pharmacists are the most qualified healthcare professional available to
instigate CPR and thus should be competent to perform it until assistance arrives. In Australia,
to become a registered pharmacist it is a requirement to have a current certification in CPR
and first aid. This certificate then must be maintained for any pharmacist qualified to adminis-
ter vaccinations.[41–43] The CPR and vaccination roles in Australia are not disaster-specific,
referring to everyday situations.
PLOS ONE | https://doi.org/10.1371/journal.pone.0227132 December 26, 2019
9 / 13
Defining pharmacists’ roles in disasters: A Delphi study
The expert panellists defined the roles pharmacists should be undertaking within the four
disaster phases–PPRR. They acknowledged the principle purpose of pharmacists in disasters is
to ensure continuity of care which encompasses many different facets, from the logistics of
obtaining medications to providing direct patient care and expert advice to other healthcare
professionals. Not every jurisdiction can implement all 43 roles immediately; it requires a col-
laborative effort from all stakeholders. To undertake these roles effectively pharmacists should
work within the coordinated disaster management and public welfare team structure.
Limitations
A limitation of this Delphi study was the time commitment of key opinion leaders. Many of
these key opinion leaders lead very busy lives and jobs and therefore some of them could not
commit to partaking in a three-round Delphi panel. However, of the key opinion leaders
approached, 62.5% (15/24) completed all three rounds. Another limitation of this study is that
the opinion leaders were identified and selected by the investigators with the majority being
Australian and so there may have been potential inherent bias. To reduce this possibility of
inherent bias the investigators contacted national and international experts and utilised snow-
ball sampling techniques for the identification of additional panel members by the experts.
This research has highlighted the acknowledgment of pharmacists’ roles and has begun to
identify their scope of practice in disasters. However, due to the multijurisdictional study
design utilised, the limitations within individual jurisdictions was not explored. Further
research identifying the different legislative considerations between jurisdictions and explor-
ing potential need for additional education or trainings within individual jurisdictions could
be investigated.
Conclusion
The international key opinion leaders have acknowledged pharmacists’ roles in disasters
throughout the disaster PPRR cycle. They have formally accepted 43 roles which extend
beyond the well-established logistics role. Pharmacy professional bodies need to provide train-
ings for specialised disaster pharmacists’ roles and advocate for supportive disaster pharmacy
legislation. Together, pharmacy professional bodies and policymakers can provide better inte-
gration of pharmacist roles in disaster management teams, whether assisting in the community
or on deployment.
Supporting information
S1 Table. Round 1 Delphi study survey results utilising a five-point Likert scale, roles were
divided into the PPRR disaster phases taking an all-hazard approach.
(DOCX)
S2 Table. Round 2 Delphi survey utilising a four-point Likert scale and the panellists’ com-
ments, roles were divided into the PPRR disaster phases taking an all-hazard approach
and consensus is indicated against each role.
(DOCX)
S3 Table. Final round Delphi study qualitative comments on roles which had reached con-
sensus for pharmacists’ roles across the disaster PPRR phases utilising an all-hazard
approach.
(DOCX)
PLOS ONE | https://doi.org/10.1371/journal.pone.0227132 December 26, 2019
10 / 13
Defining pharmacists’ roles in disasters: A Delphi study
S4 Table. Final round Delphi study survey on roles which had not yet reached consensus in
the previous two rounds and included the comments from the panellists.
(DOCX)
Author Contributions
Conceptualization: Kaitlyn E. Watson, Judith A. Singleton, Vivienne Tippett, Lisa M. Nissen.
Data curation: Kaitlyn E. Watson.
Formal analysis: Kaitlyn E. Watson.
Investigation: Kaitlyn E. Watson.
Methodology: Kaitlyn E. Watson, Judith A. Singleton, Vivienne Tippett, Lisa M. Nissen.
Project administration: Kaitlyn E. Watson.
Resources: Kaitlyn E. Watson.
Software: Kaitlyn E. Watson.
Supervision: Judith A. Singleton, Vivienne Tippett, Lisa M. Nissen.
Validation: Judith A. Singleton, Vivienne Tippett, Lisa M. Nissen.
Visualization: Kaitlyn E. Watson.
Writing – original draft: Kaitlyn E. Watson.
Writing – review & editing: Kaitlyn E. Watson, Judith A. Singleton, Vivienne Tippett, Lisa M.
Nissen.
References
1.
The Pharmacy Guild of Australia. Submission in Response to the Competition Policy Review Draft
Report. National Secretariat, Barton, ACT, 2014.
2. Scott S, Constantine LM. When natural disaster strikes. Am Pharm. 1990; NS30(11):27–8, 31. https://
doi.org/10.1016/s0160-3450(16)33636-4 PMID: 2260522
3. Guynn J Jr. Disaster imminent—Hurricane Hugo. Hosp Pharm. 1990; 25(4):325–8. PMID: 10104388
4. U. S. Department Of Health Education and Welfare. The Role of the Pharmacist in National Disaster. In:
Office of Civil Defense DotA, editor. 1964.
5. Berod T, Chan-ou-Teung F. Pharmacist’s Role in Rescue Efforts After Plane Crash in Indian Ocean.
Am J Health Syst Pharm. 1997; 54(9):1110–. https://doi.org/10.1093/ajhp/54.9.1110 PMID: 9143661
6.
7.
Ford H, von Waldner T, Perri M. Pharmacists’ Roles in Post-September 11th Disasters: A Content Anal-
ysis of Pharmacy Literature. J Pharm Pract. 2014; 27(4):350–7. https://doi.org/10.1177/
0897190013507081 PMID: 24128786
Ford H, Dallas CE, Harris C. Examining Roles Pharmacists Assume in Disasters: A Content Analytic
Approach. Disaster Med Public Health Prep. 2013; 7(06):563–72.
8. Ochi S HS, Landeg O, Mayner L, Murray V. Disaster-Driven Evacuation and Medication Loss: A Sys-
tematic Literature Review. PLoS Currents Disasters. 2014. https://doi.org/10.1371/currents.dis.
fa417630b566a0c7dfdbf945910edd96 PMID: 25642363
9. Watson KE. The Roles of Pharmacists in Disaster Health Management in Natural and Anthropogenic
Disasters. [PhD Thesis] [Thesis]. QUT ePrints: Queensland University of Technology; 2019.
10. Mullen E. The Essential Role of Pharmacists in Disasters [Internet] QS/1 News2014 [cited 2018-05-09].
Available from: http://www.webcitation.org/76KqdD62T.
11. Watson KE, Singleton JA, Tippett V, Nissen LM. Do disasters predict international pharmacy legisla-
tion?. Aust Health Rev. 2019:-. https://doi.org/10.1071/AH19093.
12. Currier M, King DS, Wofford MR, Daniel BJ. A Katrina Experience: Lessons Learned. Am J Med. 2006;
119(11):986–92. https://doi.org/10.1016/j.amjmed.2006.08.021 PMID: 17071168
PLOS ONE | https://doi.org/10.1371/journal.pone.0227132 December 26, 2019
11 / 13
Defining pharmacists’ roles in disasters: A Delphi study
13. Young D. Pharmacists Play Vital Roles in Katrina Response More Disaster-Response Participation
Urged. Am J Health Syst Pharm. 2005; 62(21):2202–16. https://doi.org/10.2146/news050025 PMID:
16239405
14. Hogue MD, Hogue HB, Lander RD, Avent K, Fleenor M. The Non Traditional Role of Pharmacists After
Hurricane Katrina: Process Description and Lessons Learned. Public Health Rep. 2009:217–23. https://
doi.org/10.1177/003335490912400209 PMID: 19320363
15. Velazquez L, Dallas S, Rose L, Eva KS, Saville R, Wang J, et al. A PHS Pharmacist Team’s Response
to Hurricane Katrina. Am J Health Syst Pharm. 2006; 63(14).
16. Angelo LB, Maffeo CM. Local and Global Volunteer Opportunities for Pharmacists to Contribute to Pub-
lic Health. Int J Pharm Pract. 2011; 19(3):206–13. https://doi.org/10.1111/j.2042-7174.2011.00123.x
PMID: 21554447
17. Pedersen CA, Canaday BR, Ellis WM, Keyes EK, Pietrantoni A, Rothholz MC, et al. Pharmacists’ Opin-
ions Regarding Level of Involvement in Emergency Preparedness and Response. J Am Pharm Assoc
(2003). 2003; 43(6):694–701.
18. Pharmacy Traynor K., Public Health Intersect in Alabama Disaster Plans. Am J Health Syst Pharm.
2007; 64(19):1998–9. https://doi.org/10.2146/news070087 PMID: 17893407
19.
20.
Ford JH. Pharmacists in Disasters [PhD Thesis]. Athens, Gerogia: University of Georgia; 2013.
Inspector-General for Emergency Management (IGEM), Victorian Government. Review of Response to
the Thunderstorm Asthma Event of 21–22 November 2016: Final Report. 2017.
21. Boulkedid R, Abdoul H, Loustau M, Sibony O, Alberti C. Using and Reporting the Delphi Method for
Selecting Healthcare Quality Indicators: A Systematic Review. PLoS One. 2011; 6(6):e20476. https://
doi.org/10.1371/journal.pone.0020476 PMID: 21694759
22. Sekayi D, Kennedy A. Qualitative Delphi Method: A Four Round Process with a Worked Example. The
Qualitative Report. 2017; 22(10):2755–63.
23. Beattie E, Mackway-Jones K. A Delphi Study to Identify Performance Indicators for Emergency Medi-
cine. Emerg Med J. 2004; 21(1):47–50. https://doi.org/10.1136/emj.2003.001123 PMID: 14734375
24. Rådestad M, Jirwe M, Castre´ n M, Svensson L, Gryth D, Ru¨ter A. Essential Key Indicators for Disaster
Medical Response Suggested to be Included in a National Uniform Protocol for Documentation of Major
Incidents: A Delphi Study. Scand J Trauma Resusc Emerg Med. 2013; 21(1):68.
25. Sinha IP, Smyth RL, Williamson PR. Using the Delphi Technique to Determine which Outcomes to Mea-
sure in Clinical Trials: Recommendations for the Future Based on a Systematic Review of Existing Stud-
ies. PLoS Med. 2011; 8(1):e1000393. https://doi.org/10.1371/journal.pmed.1000393 PMC3026691.
PMID: 21283604
26. Kennie-Kaulbach N, Farrell B, Ward N, Johnston S, Gubbels A, Eguale T, et al. Pharmacist Provision of
Primary Health Care: A Modified Delphi Validation of Pharmacists’ Competencies. BMC Fam Pract.
2012; 13(1):27. https://doi.org/10.1186/1471-2296-13-27 PMID: 22455482
27. Bisson JI, Tavakoly B, Witteveen AB, Ajdukovic D, Jehel L, Johansen VJ, et al. TENTS Guidelines:
Development of Post-Disaster Psychosocial Care Guidelines Through A Delphi Process. Br J Psychia-
try. 2018; 196(1):69–74. Epub 01/02. https://doi.org/10.1192/bjp.bp.109.066266 PMID: 20044665
28.
Zhong S, Clark M, Hou X-Y, Zang Y, Fitzgerald G. Development of Key Indicators of Hospital Resil-
ience: A Modified Delphi Study. J Health Serv Res Policy. 2015; 20(2):74–82. https://doi.org/10.1177/
1355819614561537 PMID: 25504827
29. Cavallo A, Ireland V. Preparing for Complex Interdependent Risks: A System of Systems Approach to
Building Disaster Resilience. Int J Disaster Risk Reduct. 2014; 9:181–93. https://doi.org/10.1016/j.ijdrr.
2014.05.001
30. McGuire B. Global Disaster Paves Way for Global Thinking: In the Wake of the Indian Ocean Tsunami,
Disaster Expert and Geophysicist Bill Mcguire Explains why Future Disaster Management Must Place a
Greater Emphasis on Preparedness as well as Response. Geographical. 2005; 77(3):14.
31.
Leischow SJ, Milstein B. Systems Thinking and Modeling for Public Health Practice. Am J Public Health.
2006; 96(3):403–5. https://doi.org/10.2105/AJPH.2005.082842 PMID: 16449572
32. Pratt G. Showcase Essential Pharmacy Practice Functions through Emergency Preparedness Efforts.
Pharmacy Insights, A Look At: Emergency Preparedness. 2013:3–4.
33. Canadian Emergency Management College, Public Safety and Emergency Preparedness Canada.
CBRN First Responder Training Program: Basic Level Pre-Course Reading. Ottawa, Ontario,: Govern-
ment of Canada; 2008.
34.
Terriff CM, Tee AM. Citywide Pharmaceutical Preparation for Bioterrorism. Am J Health Syst Pharm.
2001; 58(3):233–7. https://doi.org/10.1093/ajhp/58.3.233 PMID: 11217178
PLOS ONE | https://doi.org/10.1371/journal.pone.0227132 December 26, 2019
12 / 13
Defining pharmacists’ roles in disasters: A Delphi study
35.
Idemyor V, Smith SW. Biological Warfare—A Growing Healthcare Concern. Ann Pharmacother. 2002;
36:1282–6. https://doi.org/10.1345/aph.1A462 PMID: 12125675
36. Canadian Pharmacists Association. Pharmacists’ Scope of Practice in Canada [Internet] 2016 [cited
2018-08-07]. Available from: http://www.webcitation.org/76Kr06Aeg.
37. Australian Red Cross, Australian Psychological Society. Psychological First Aid: An Australian Guide to
Supporting People Affected by Disaster. Melbourne: 2013.
38. Rubio-Valera M, Chen T, #039, Reilly C. New Roles for Pharmacists in Community Mental Health Care:
A Narrative Review. Int J Environ Res Public Health. 2014; 11(10):10967. https://doi.org/10.3390/
ijerph111010967 PMID: 25337943
39. O’Reilly CL, Bell JS, Kelly PJ, Chen TF. Impact of Mental Health First Aid Training on Pharmacy Stu-
dents’ Knowledge, Attitudes and Self-Reported Behaviour: A Controlled Trial. Aust N Z J Psychiatry.
2011; 45(7):549–57. https://doi.org/10.3109/00048674.2011.585454 PMID: 21718124
40. Machado C, Barlows TG, Marsh WA, Coto-Depani Y, Dalin G. Pharmacists on the Emergency Cardio-
pulmonary Resuscitation Team: Their Responsibilities, Training, and Attitudes. Hosp Pharm. 2003; 38
(1):40–9. https://doi.org/10.1177/001857870303800113
41. Veltri KT, Yaghdjian V, Morgan-Joseph T, Prlesi L, Rudnick E. Hospital Emergency Preparedness:
Push-POD Operation and Pharmacists as Immunizers. J Am Pharm Assoc (2003). 2012; 52(1):81–5.
42. Hope D, Yeates G, King M. Vaccinations: Fragmented Federation: Inconsistent Interstate Require-
ments for Pharmacist Vaccine Administration. AJP: The Australian Journal of Pharmacy. 2016; 97
(1157):18–22.
43.
Lau ET, Rochin ME, DelDot M, Glass BD, Nissen LM. “There’s No Touching in Pharmacy”: Training
Pharmacists for Australia’s First Pharmacist Immunization Pilot. Can J Hosp Pharm. 2017; 70(4):281.
PMID: 28894312
PLOS ONE | https://doi.org/10.1371/journal.pone.0227132 December 26, 2019
13 / 13
|
10.1371_journal.pone.0241692 | RESEARCH ARTICLE
Understanding growth and age of red tree
corals (Primnoa pacifica) in the North Pacific
Ocean
Emma Choy1, Kelly Watanabe1, Branwen WilliamsID
Ellen Druffel4, Thomas Lorenson5, Mary Knaak1
1*, Robert Stone2, Peter Etnoyer3,
1 W.M. Keck Science Department of Claremont McKenna, Pitzer, and Scripps Colleges, Claremont, CA,
United States of America, 2 Alaska Fisheries Science Center, National Marine Fisheries Service, NOAA,
Juneau, AK, United States of America, 3 NOAA National Centers for Coastal Ocean Science, Charleston,
SC, United States of America, 4 Department of Earth System Science, University of California Irvine, Irvine,
CA, United States of America, 5 USGS Pacific Coastal and Marine Science Center, Santa Cruz, CA, United
States of America
* bwilliams@kecksci.claremont.edu
Abstract
Massive, long-lived deep-sea red tree corals (Primnoa pacifica) form a solid, layered axis
comprised of calcite and gorgonin skeleton. They are abundant on the outer continental
shelf and upper slope of the Northeast Pacific, providing habitat for fish and invertebrates.
Yet, their large size and arborescent morphology makes them susceptible to disturbance
from fishing activities. A better understanding of their growth patterns will facilitate in-situ
estimates of population age structure and biomass. Here, we evaluated relationships
between ages, growth rates, gross morphological characteristics, and banding patterns in
11 colonies collected from depths of ~141–335 m off the Alaskan coast. These corals ran-
ged in age from 12 to 80 years old. They grew faster radially (0.33–0.74 mm year-1) and axi-
ally (2.41–6.39 cm year-1) than in previously measured older colonies, suggesting that
growth in P. pacifica declines slowly with age, and that basal diameter and axial height even-
tually plateau. However, since coral morphology correlated with age in younger colonies (<
century), we developed an in-situ age estimation technique for corals from the Northeast
Pacific Ocean providing a non-invasive method for evaluating coral age without removing
colonies from the population. Furthermore, we determined that annual bands provided the
most accurate means for determining coral age in live-collected corals, relative to radiomet-
ric dating. Taken together, this work provides insight into P. pacifica growth patterns to
inform coastal managers about the demographics of this ecologically important species.
With this new ability to estimate the age of red tree corals in-situ, we can readily determine
the age-class structure and consequently, the maturity status of thickets, using non-invasive
video survey techniques when coupled with mensuration systems such as lasers or stereo-
cameras. Enhanced surveys could identify which populations are most vulnerable to distur-
bance from human activities, and which should be highlighted for protection.
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: Choy E, Watanabe K, Williams B, Stone R,
Etnoyer P, Druffel E, et al. (2020) Understanding
growth and age of red tree corals (Primnoa
pacifica) in the North Pacific Ocean. PLoS ONE
15(12): e0241692. https://doi.org/10.1371/journal.
pone.0241692
Editor: Erik Caroselli, University of Bologna, ITALY
Received: January 10, 2020
Accepted: October 20, 2020
Published: December 1, 2020
Copyright: This is an open access article, free of all
copyright, and may be freely reproduced,
distributed, transmitted, modified, built upon, or
otherwise used by anyone for any lawful purpose.
The work is made available under the Creative
Commons CC0 public domain dedication.
Data Availability Statement: Data are contained
within the paper.
Funding: BW & PE NA08OAR4300817 NOAA’s
West Coast & Polar Regions Undersea Research
Center. Program no longer running. The funders
had no role in study design, data collection and
analysis, decision to publish, or preparation of the
manuscript.
Competing interests: The authors have declared
that no competing interests exist.
PLOS ONE | https://doi.org/10.1371/journal.pone.0241692 December 1, 2020
1 / 18
PLOS ONE Primnoidae age models
Introduction
Fishing practices, including bottom trawling and long lining, can disturb benthic ecosystems,
particularly those where the seafloor is highly structured with large sedentary invertebrates
such as corals and sponges [1–4]. Gorgonian corals are especially vulnerable to fishing prac-
tices due to easy ensnarement of this large arborescent sea-fan, which adheres to the seafloor
with a single holdfast. Some gorgonians are very long-lived and communities of older colonies
compromised by fishing gear may take decades to centuries to recover [5–9]. Understanding
how fast these corals grow and the relationship between size and age can provide estimates of
recovery times of these communities [2, 10, 11]. This is critically important in areas vulnerable
to fishing disturbance such as those in the Northeast Pacific Ocean where the importance of
these habitats to fisheries has been documented [3, 4, 12].
Gorgonian coral communities are important components of the sea floor because they pro-
vide habitat for a diversity of invertebrates and fishes [13, 14]. In fact, these corals can serve as
habitat engineers [15]: in their absence, shallow-water assemblages shift from predominately
corals and sponges to algae and turf-forming species [16]. Furthermore, some gorgonian corals
are indicator species, and in addition to disturbance from fishing activities are sensitive to
warming ocean temperatures [16, 17] and oil pollution [18, 19]. The conservation of these cor-
als is thus critical for maintaining ecological diversity and community resilience.
The gorgonian Primnoa pacifica [20], also known as the red tree coral, are ecologically
important deep-sea corals in the North Pacific Ocean [4]. They have been referred to as “key-
stone species,” “foundation species,” and “ecosystem engineers” [4, 21]. These animals are com-
prised of an internal skeleton arising from the holdfast attached directly to hard substrate. Their
skeleton is largely comprised of protein-rich organic gorgonin sometimes interspersed with cal-
cite. The source of elements to the gorgonin skeleton is organic material produced in surface
waters and transported to depth to be fed upon by the corals. In contrast, the calcite elements
are sourced from ambient seawater at depth [22–24]. A thin layer of coenenchyme with polyps
covers the entirety of the skeleton. The skeletal central axis grows axially and radially, such that
through time the coral grows taller along its axial axis and adds layers to the outside of its skele-
tal trunk increasing the trunk diameter. They can grow to massive size (greater than 2 m in
height [25]), in part because of their long lifespans that can exceed a century or more [26].
In the central skeletal axis of some gorgonian corals, concentric couplets of gorgonin-calcite
bands form annually, providing a means to determine the age of a colony; however, fine-scale
bands of unknown periodicity are also present, indicating possible drivers of skeletal banding
[9, 26–29]. The finer bands may reflect variations in the color of the organic skeleton (which in
turn reflects the degree of protein cross-linkages during skeletal formation) and/or alternations
of the gorgonin with calcite skeleton [9, 30, 31]. In addition to annual growth band counts,
radiometric dating (14C and 210Pb) can provide estimates of coral age, albeit with potential
uncertainties depending on the age and collection date of the coral [24, 26, 32, 33].
Previous studies using a combination of annual growth band counts and radiometric dating
in P. pacifica yielded different estimates of radial growth rates ranging from 0.14 to 0.57 mm
yr-1 [26, 34, 35]. Axial growth rates for this species have only been reported for two specimens,
and those estimates ranged from 1.60 to 2.32 cm yr-1 [26]. Obtaining axial growth rates in a
much larger sample of corals is key to determine the rate of recovery of disturbed
communities.
Therefore, the objectives of this study were to 1) determine the age and growth (radial and
axial) for a suite of colonies collected in the Northeast Pacific Ocean with morphological data;
2) develop and evaluate age estimate calculations converting morphological data into age; 3)
evaluate annual growth band counts and radiometric dating as age determination techniques;
PLOS ONE | https://doi.org/10.1371/journal.pone.0241692 December 1, 2020
2 / 18
PLOS ONE Primnoidae age models
and 4) examine the periodicity and potential drivers of sub-annual banding in the axial skele-
ton. As a whole, this work provides critical insight into recovery times of P. pacifica which
helps inform management policies of important deep-sea coral habitats.
Materials and methods
Sample sites and collection
Three intact colonies were collected in 2013 using the remotely operated vehicle (ROV) H2000
deployed from the F/V Alaska Provider (Table 1; Figs 1 and 2). Seven intact colonies were col-
lected in 2015 using the ROV Zeus II deployed from the R/V Dorado Discovery (Table 1, Fig
1). Additionally, a single specimen (GOA 004; Table 1) was collected with a research bottom
trawl deployed from the F/V Alaska Provider just prior to the 2013 cruise. Fieldwork research
was performed by NOAA’s Alaska Fisheries Science Center and Deep Sea Coral Research and
Technology Program under the authority of the U.S. Department of Commerce. Colonies
were air dried on board the vessels and morphological data were recorded, including maxi-
mum height, basal diameter, wet weight, and distance from the base to first branch.
Sample preparation
In the laboratory, a diamond-edge saw was used to cut three adjacent 0.25–0.50 cm cross-sec-
tion discs from the basal portion (holdfast) of each coral. One cross-section from each coral
was mounted on a glass slide and polished for annual band counts and digital imaging, one
cross-section was used for dissected band counts and radiocarbon dating, and one cross-sec-
tion was used for 210Pb dating. Although growth rings were only counted in two of the three
cross sections, it is assumed that all cross sections contain the same number of growth bands
since they were cut from adjacent parts of the coral. The two cross-sections used for (1) dis-
sected band counts and radiocarbon dating and (2) 210Pb dating were bathed in 100 ml of 5%
HCl solution for a minimum of 10 days (up to four weeks). The HCl solution was refreshed
every other week for four weeks so that all the calcite bands layered between the organic
Table 1. Sample ID, collection information, and morphological data for the 11 coral colonies included in this study. Age estimates for all specimens are derived from
annual growth bands. Dates are mm/dd/yyyy.
Sample
ID
Collection
date
WPA 001
6/4/2015
WPA 002
6/5/2015
WPA 003
6/5/2015
WPA 004
6/7/2015
Locality
Latitude Longitude Depth
58.2457
-138.9045
(m)
141
58.2458
-138.9044
142
58.2377
-138.9894
147
58.2047
-138.8175
163
Fairweather
Grand
Fairweather
Grand
Fairweather
Grand
Fairweather
Grand
WPB 005
6/9/2015
Dixon Entrance
54.6252
-132.8828
WPB 006
6/8/2015
Dixon Entrance
54.6345
-132.8510
WPB 007
6/8/2015
Dixon Entrance
54.6345
-132.8510
GOA 004
7/19/2013
Portlock Bank
58.3102
-149.5087
GOA 011
8/13/2013
Shutter Ridge
56.1749
-135.1165
GOA 022
8/13/2013
Shutter Ridge
56.1722
-135.1160
GOA 067
8/15/2013
Shutter Ridge
56.1784
-135.1180
335
164
164
147
191
203
214
Height
(cm)
Width
(cm)
123
171
110
223
160
99
87
78
65
72
67
60
70
101
98
26
23
50
24
32
153
121
BaseD
(mm)
32.5
25.5
45.0
53.0
49.0
19.0
15.0
15.0
10.0
16.0
45.0
DistToBr�
(cm)
Wet weight
(kg)
Age estimate
(years)
7
62
30
2.24
1.25
n/a
29.5
20.45
29
20
17
16
8.5
12.3
32
16.6
1.14
0.64
1.08
0.62
1.13
n/a
26 ± 2
28 ± 2
19 ± 2
80 ± 1
67 ± 4
16 ± 2
15 ± 1
14 ± 2
12 ± 2
18 ± 1
31 ± 2
�Distance to first branch.
https://doi.org/10.1371/journal.pone.0241692.t001
PLOS ONE | https://doi.org/10.1371/journal.pone.0241692 December 1, 2020
3 / 18
PLOS ONE Primnoidae age models
Fig 1. Map showing the collection sites of Primnoa pacifica from the current project relative to previous studies
(Andrews et al., 2002 [26]; Aranha et al., 2014 [34]; Williams et al., 2007 [35]). Map created by Michele Masuda.
https://doi.org/10.1371/journal.pone.0241692.g001
skeleton had dissolved. After acidification, the sample was transferred to soak in Milli’Q water
for band peeling.
Photographed band counts
Using the mounted and polished cross-sections (Fig 2), the number of annual bands along the
longest radial transect were counted by two researchers under a light microscope for all speci-
mens in this study. The radial growth rate (mm year-1) of each coral was obtained by dividing
the base diameter of the coral by the age of the coral, as determined by annual growth bands.
Similarly, axial growth rates (cm year-1) were calculated by dividing the maximum height of
the coral by the age of the coral, as determined by annual growth bands.
The cross sections were then imaged using a Nikon digital microscope with NIS-Elements,
a Nikon microscope software package. Using these high-resolution images, the total number
of sub-annual bands were counted along the same longest radial transect for four specimens
(GOA 011, WPA 002, WPA 004, and WPB 005) by two researchers (S1 File). Four larger speci-
mens were chosen to encompass a range in sizes and the ability to cut adjacent cross sections.
The bands were sometimes difficult to distinguish; when major discrepancies in band counts
were evident, researchers re-counted bands collaboratively, discussing the presence or absence
of a band when uncertainties arose.
Dissected band counts
Using one of the HCl-bathed cross-sections, sub-annual growth bands were peeled and
counted in sections from the outside to the center of the coral cross-section using forceps and
PLOS ONE | https://doi.org/10.1371/journal.pone.0241692 December 1, 2020
4 / 18
PLOS ONE Primnoidae age models
Fig 2. (A) Primnoa pacifica colonies on the Shutter Ridge at a depth of 180 m, (B) sub-section from specimen WPA
005 with a white line showing region of axis used to count growth bands, (C) image showing annual growth bands and
(D) sub-annual growth bands.
https://doi.org/10.1371/journal.pone.0241692.g002
working under a microscope at 30x magnification. Without tearing the bands, sections were
peeled with the least number of bands possible, with each section containing approximately
1–20 bands depending on the ease of band separation. If possible, sections were peeled all the
way around the circumference of the sample. The separated sections were air dried and pack-
aged into labeled weigh paper packets.
Radiocarbon analysis
For radiocarbon sample preparation, a laminar flow hood workspace was cleaned with deion-
ized water. All glassware was soaked in 10% HCl for a minimum of 1 hour. The glass vials used
for radiocarbon dating were additionally combusted at 540 ˚C for 2 hours, while the vial caps
were washed with soap and water, and then acid washed in 10% HCl for 30 seconds and rinsed
with water. After sub-annual bands were dissected and counted, three of the four colonies
(WPA 002, WPB 005, and WPA 004) were selected for radiocarbon analysis. Based on annual
band counts, colony GOA 011 was too young for radiocarbon dating to be effective. Five milli-
gram sub-samples reflecting no more than 10% of the entire sample and equidistant from each
other based on dissected band count numbers were pulverized into acid washed glass vials.
The outermost bands of each sample’s cross-section were analyzed to determine the Δ14C val-
ues at the time of collection. The 14C was measured at the Keck Carbon Cycle Accelerator
Mass Spectrometry at the University of California, Irvine. Five milligrams of a National Insti-
tute of Standards and Technology (NIST) wood standard (Firi H) and a coal standard were
also prepared as a reference for radiocarbon analysis and sample preparation backgrounds
were subtracted based on these measurements. All results were corrected for isotopic fraction-
ation according to the conventions of Stuiver and Polach (1977) [36]. Δ14C values were
assigned a year using the following equation:
Year ¼
ðD14C (cid:0) DClo AÞðYearhi A (cid:0) Yearlo AÞ
DChi A (cid:0) D14Clo A
þ Yearlo A
PLOS ONE | https://doi.org/10.1371/journal.pone.0241692 December 1, 2020
ð1Þ
5 / 18
PLOS ONE Primnoidae age models
where Year is the associated year calculated for this study, Δ14C is the measured value of the
coral sub-section being dated, Δ14Clo A and Yearlo A are the lower estimates of Δ14C and calcu-
lated year, and Δ14Chi A and Yearhi A are the upper estimates of Δ14C and calculated year (data
from Andrews et al., 2013 [37]). Radiocarbon records from Andrews et al. (2013) [37] were
constructed using Δ14C data from otoliths of Northeast Pacific yelloweye rockfish (Sebastes
ruberrimus), Pacific halibut, and known-age abalone shell samples. A loess curve, including
95% confidence intervals, was fit to the Δ14C otolith data of Northeast Pacific yelloweye rock-
fish to produce a radiocarbon bomb curve (S1 Fig in S1 File). The yelloweye rockfish are bot-
tom dwelling fish that were collected in waters off southeast Alaska [38]. The rockfish
chronology identified the initial rise in 14C in the late 1950s and peak values in the late 1960s/
early 1970s, which agrees with the 14C reconstruction by Roark et al. (2005) [23] for a bamboo
coral from the northwestern coast of Canada.
Lead-210 dating
Measurements of 210Pb activity provided a third method for evaluating coral age for specimens
WPA 002, WPA 004, and WPB 005. Specimens were rinsed thoroughly in Milli-Q water and
then dried on a clean watch glass in a laminar flow hood. While still malleable enough to be
manipulated, the specimens were split into three sections for sub-sampling (exterior, middle,
and interior). Subsamples were dried, cooled in liquid nitrogen, and pulverized in a Genogrin-
der. Subsamples were weighed (~ 1 g) and placed in clean, new 4-ml vials. Due to the smaller
size of the cross-sections initially cut, it was necessary to cut a second cross-section for WPA
002 and WPA 004, such that the sub-samples for these specimens were a combination of two
cross-sections. All prepared 210Pb samples were sent to the United States Geological Survey at
the Pacific Coastal and Marine Science Center for 210Pb dating using a germanium gamma ray
detector. 210Pb decays at a constant rate yet is also in secular equilibrium with 226Ra. To deter-
mine the specimen age from 210Pb values, the excess amount of 210Pb (210Pbex), ultimately
derived from deposition of atmospheric 222Rn, which in turn decays to 210Pb, is determined by
subtracting measured 226Ra activity. 210Pbex decays according to the law of radioactive decay:
Aex ¼ Aoe(cid:0) lt
ð2Þ
where Aex is the measured 210Pbex at time t, Ao is the initial 210Pbex at time 0, and λ is 0.0311 or
the natural log of 2 divided by the half-life of 210Pb (22.3 years). Inputting the calculated
210Pbex values for each specimen into the equation above yields the age of each sample.
Data analysis
The relationship between specimen age derived from annual band counts and the morphologi-
cal data was quantified by comparing linear and logarithmic fits to find the best fit model
using R software. For height, width, and basal diameters, these equations represent a means to
convert a measurement of a morphological parameter into specimen age. We primarily focus
on basal diameter because this metric has most often been reported in corals that also have
their ages determined. We validated these age estimation calculations in three ways: 1) We
applied the calculations to ten previously collected and aged corals with available radii data
that were less than a century old from the northeast Pacific Ocean ([34, 35]; Table 3). Basal
diameter was determined from the radius in these corals (assumed diameter = 2�radius,
although see discussion on this assumption below). 2) We developed and validated a second
calculation relating basal diameter to age using all available colonies (< century in age) for the
northeast Pacific Ocean. 3) We apply this regional age estimation to corals collected from the
western Pacific Ocean [39]. For each of these comparisons, we evaluated how well the ages
PLOS ONE | https://doi.org/10.1371/journal.pone.0241692 December 1, 2020
6 / 18
PLOS ONE Primnoidae age models
derived from the basal estimates correspond to the ages derived from annual band counts or
210Pb dating.
In the three specimens with 14C-dating (WPA 002, WPA 004, WPB 005), we calculated
banding frequencies and sub-annual radial and axial growth rates. Banding frequencies
(growth bands year-1) were calculated through time for each colony according to the following
equation:
Banding frequency ¼
Band countn (cid:0) Band countnþ1
Agen (cid:0) Agenþ1
;
n ¼ 1; 2; 3 . . .
ð3Þ
Where Band count is the average associated growth band count of the sub-section used for
radiocarbon dating, Agen is the radiocarbon age of the older n sub-section of the coral, and n is
the sub-section of the coral sent in for radiocarbon dating (a smaller n is closer to the center of
the coral representing older skeleton).
Results
Morphological data
Eleven intact specimens were collected during two research cruises in 2013 and 2015. Coral
height (axial length) ranged from 65 to 223 cm, width ranged from 23 to 121 cm, and basal
diameter ranged from 1 to 5.3 cm (Table 1). Many of the morphological parameters signifi-
cantly correlated with each. Height and width were significantly correlated, such that taller
specimens were wider (p = 0.004, r2 = 0.62, N = 11) (Table 2). Basal diameter significantly cor-
related with height (p = 0.0003, r2 = 0.79, N = 11) and width (p<0.0001, r2 = 0.85, N = 11)
(Tables 1 and 2). Height correlated with the distance to first branch (p = 0.03, r2 = 0.41,
N = 11) (Table 2). The weight of the specimen significantly correlated with age, height, weight,
and axial and radial growth rates (Table 2).
Table 2. Statistics comparing morphological data, ages, and growth rates of Primnoa pacifica.
Age estimate
(years)
Age estimate (years)
Height (cm)
Width (cm)
BaseD (cm)
DistToBr (cm)
Wet weight (kg)
(N = 9)
Radial growth rates
(mm yr-1)
Axial growth rates (cm
yr-1)
Height
(cm)
0.0001�
0.83
Width
(cm)
0.0028�
0.65
BaseD
(mm)
0.0000�
0.90
0.0041
0.0003
0.62
0.79
0.0001
0.85
DistToBr
(cm)
Wet weight (kg)
(N = 9)
0.3475
0.10
0.0333
0.41
0.2020
0.17
0.2821
0.13
0.0000
0.97
0.0100
0.64
0.0021
0.76
0.0003
0.86
0.5543
0.05
Radial growth rate
(mm yr-1)
0.0748
Axial growth rate
(cm yr-1)
0.0008
0.31
0.3930
0.08
0.8160
0.01
0.6523
0.02
0.7496
0.01
0.0296
0.51
0.73
0.0866
0.29
0.0631
0.33
0.0089
0.55
0.8717
0.00
0.0022
0.76
0.0689
0.32
Age estimates for all specimens are derived from visual counts of annual growth bands. Reported is the p-value and r2 for each comparison. All statistics are based on a
linear model, except those indicated with an asterisk, which are based on a logarithmic model. N = 11 unless otherwise noted. Significant relationships are in bold.
https://doi.org/10.1371/journal.pone.0241692.t002
PLOS ONE | https://doi.org/10.1371/journal.pone.0241692 December 1, 2020
7 / 18
PLOS ONE Primnoidae age models
Age estimates based on annual band counts ranged from 12 ± 2 to 80 ± 1 years. These age
estimates correlated with all of the morphological data, except distance to first branch
(Table 2). Using a logarithmic model, older specimens were taller (p<0.0001, r2 = 0.83,
N = 11), wider (p = 0.0028, r2 = 0.65, N = 11), and had a larger basal diameter (p<0.0001, r2 =
0.90, N = 11) (Table 2; Fig 3). A logarithmic model provided the best fit between age and mor-
phological data, excluding weight (Fig 3). Thus height, width, and basal diameter can be used
to determine specimen age in years (± standard error) using the following equations:
ð
Age �21
Þ ¼ e
Heightþ109:8
72:6
ð
Age �21
Þ ¼ e
Widthþ79:4
44:1
Age �5ð
Þ ¼ e
Basal Diameterþ45:8
23:0
ð4Þ
ð5Þ
ð6Þ
Axial growth rates varied from 2.41 to 6.39 cm year-1 and radial growth rates varied from
0.33 to 0.74 mm year-1 (Table 3). Using a linear model, axial growth rates significantly
inversely correlated with age (p<0.001, r2 = 0.73, N = 11) and basal diameter (p = 0.0089, r2 =
0.55, N = 11), and positively correlated with specimen weight (p = 0.0022, r2 = 0.76, N = 9)
(Table 2). Radial growth rates only significantly varied with specimen weight (p = 0.030, r2 =
0.51, N = 9) (Table 2).
Band count comparisons
In four specimens ranging in age from 12 to 80 years, based on annual band counts, the num-
ber of sub-annual dissected bands ranged from 290 ± 28 to 1589 ± 124, while the number of
sub-annual photographed bands ranged from 152 ± 10 to 1131 ± 45 (Table 4). The number of
sub-annual photographed growth bands positively correlated with the number of annual
growth bands, such that there is an average of 14 ± 0.8 sub-annual bands for every annual
Fig 3. Height (black circles), width (white circles), and basal diameter (grey circles) of Primnoa pacifica decrease
logarithmically with age. Statistics are presented in Table 2, Eqs (4)–(6). Dashed lines represent 95% confidence
intervals.
https://doi.org/10.1371/journal.pone.0241692.g003
PLOS ONE | https://doi.org/10.1371/journal.pone.0241692 December 1, 2020
8 / 18
PLOS ONE Primnoidae age models
Table 3. Reported ages, radial and axial growth rates of Primnoa pacifica. Radiometric age estimates for Andrews et al. (2002) [26] and Williams et al. (2007) [35] were
derived from 210Pb. Radiometric age estimates for Aranha et al. (2014) [34] were derived from 14C. OCNMS is the Olympic Coast National Marine Sanctuary. Shiribeshi
seamount is located in the Sea of Japan off the western coast of Hokkaido, Japan. Annual bands were ambiguous in specimens missing age estimates from growth bands
(column 3).
Location
Colony name Age estimate
Fairweather Ground
Fairweather Ground
Fairweather Ground
WPA 001
WPA 002�
WPA 003
Fairweather Ground
WPA 004
Dixon Entrance
Dixon Entrance
Dixon Entrance
Portlock Bank
Shutter Ridge
Shutter Ridge
Shutter Ridge
Dixon Entrance
Dixon Entrance
Dixon Entrance
Dixon Entrance
Dixon Entrance
OCNMS
OCNMS
OCNMS
OCNMS
Dixon Entrance
Dixon Entrance
WPB 005
WPB 006
WPB 007
GOA 004
GOA 011
GOA 022
GOA 067
R1153-0003
R1155-0012
R1155-0013
R1156-0004
R1156-0016
R1162-0015
R1162-0016
R1162-0005
R1165-0002
Colony 1
Colony 2
Portlock Bank, Alaska
Southeast Alaska
Eastern Aleutian Islands
PAL
P88
P26
Shiribeshi Seamount
Shiribeshi Seamount
Shiribeshi Seamount
Shiribeshi Seamount
Shiribeshi Seamount
Shiribeshi Seamount
Shiribeshi Seamount
Shiribeshi Seamount
Shiribeshi Seamount
1
2
3
4
5
6
7
8
9
Shiribeshi Seamount
10
(annual
bands)
26
28
19
80
67
16
15
14
12
18
31
46
34
30
22
16
45
31
40
21
124
5
8
5
9
12
14
15
18
29
40
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
±
##
——
±
——
——
±
±
±
±
±
±
±
±
±
±
2
2
2
1
4
2
1
2
2
1
2
4
2
3
2
1
2
2
2
4
3
3
2
2
2
2
3
2
6
2
2
Age estimate
(radiometric)
Radius (mm)
Radial growth
rate (mm year-1)
Height (cm)
Axial growth
rate (cm year-1)
Reference
——
42 ± 0.1
——
66 ± 1
38 ± 1
——
26
11
47
6
13
15
11
10
4
5
5
0.6
46
34
119
22
49
75
51
62
87
195
123
43
——
——
——
——
——
±
±
±
±
±
±
±
±
±
112
——
±
±
±
——
——
——
——
——
——
——
——
——
——
16.3
12.8
12.0
26.5
24.5
9.5
7.5
7.5
5.0
8.0
22.5
10.00
19.50
9.00
5.00
4.60
18.75
16.50
13.00
8.10
——
——
17
23
16
0.7
0.9
1.0
1.2
1.3
1.7
2.0
2.2
2.8
5.4
0.64 ± 0.05
0.46 ± 0.04
0.65 ± 0.07
0.33 ± 0.00
0.37 ± 0.02
0.61 ± 0.08
0.52 ± 0.04
0.56 ± 0.08
0.43 ± 0.08
0.46 ± 0.03
0.74 ± 0.05
0.08
0.22
±
±
±
±
±
±
±
±
±
0.13
0.07
0.05
0.08
0.06
0.06
0.03
0.08
0.002
±
——
0.57
0.30
0.23
0.29
0.42
0.53
0.33
0.39
0.18
0.14
0.19
0.37
0.13
0.11
0.19
0.13
0.11
0.12
0.14
0.12
0.10
0.13
123
171
110
223
160
99
87
78
65
72
153
——
——
——
——
——
——
——
——
——
——
——
——
——
——
——
——
——
——
——
——
——
——
——
——
4.82 ± 0.37 This study
6.22 ± 0.44 This study
5.95 ± 0.62 This study
2.79 ± 0.03 This study
2.41 ± 0.14 This study
6.39 ± 0.79 This study
6.00 ± 0.39 This study
5.78 ± 0.81 This study
5.65 ± 0.93 This study
4.11 ± 0.22 This study
5.02 ± 0.32 This study
1.74
2.32
——
——
——
——
——
——
——
——
——
±
±
——
——
——
——
——
——
——
——
——
——
——
——
——
Aranha et al., 2014
Aranha et al., 2014
Aranha et al., 2014
Aranha et al., 2014
Aranha et al., 2014
Aranha et al., 2014
Aranha et al., 2014
Aranha et al., 2014
Aranha et al., 2014
0.19 Andrews et al., 2002
0.09 Andrews et al., 2002
Williams et al., 2007
Williams et al., 2007
Williams et al., 2007
Matsumoto, 2007
Matsumoto, 2007
Matsumoto, 2007
Matsumoto, 2007
Matsumoto, 2007
Matsumoto, 2007
Matsumoto, 2007
Matsumoto, 2007
Matsumoto, 2007
Matsumoto, 2007
�14C results for WPA 002 could also yield an age of 49.1 years with radial growth rates of 0.31 mm yr-1 and axial growth rate of 4.12 cm yr-1.
https://doi.org/10.1371/journal.pone.0241692.t003
band (p = 0.0002, r2 = 1.0, N = 4). Conversely, there is an average of 18 ± 5. Sub-annual dis-
sected growth bands for every annual growth band, although this relationship was not statisti-
cally significant (p = 0.096, r2 = 0.82, N = 4). The number of dissected band counts increased
linearly with sub-annual photographed bands and in most cases was greater than the number
of sub-annual photographed band counts, although this relationship was also not significant
(p = 0.103, r2 = 0.81, N = 4) (Table 4).
PLOS ONE | https://doi.org/10.1371/journal.pone.0241692 December 1, 2020
9 / 18
PLOS ONE Primnoidae age models
Table 4. Annual, sub-annual dissected, and sub-annual photographed band counts for four Primnoa pacifica col-
onies in this study.
Sample ID
Annual Band Count
Sub-annual Dissected Band
Count
Sub-annual Photographed
Band Count
WPA 002
WPA 004
WPB 005
GOA 011
28
80
67
12
±
±
±
±
2
1
4
2
477
1589
790
290
±
±
±
±
64
124
93
28
399
1131
960
152
±
±
±
±
36
45
72
10
https://doi.org/10.1371/journal.pone.0241692.t004
Radiometric age determination
Measured Δ14C values ranged from -93.7 ± 1.3 to 89.4 ± 2.0 (Table 5) and were consistent with
previously published regional Δ14C records from the North Pacific Ocean [23, 37] (S1 Fig in
S1 File). The ages of the corals determined from the timing of the bomb-curve radiocarbon
ranged in age from 38 ± 1 (WPB 005) to 66 ± 1 (WPA 004) years old. Thus, based on ages and
collection year, the corals started growing between 1949 and 1977.
The radiocarbon-derived coral ages were similar to, but consistently older, than the ages
derived from the lead-210 dating (Table 5). WPA 002 has two data points for the oldest value
(S1 Fig in S1 File) because the innermost Δ14C value in this specimen could reflect either rising
or declining bomb Δ14C values. Based on growth rates and comparisons with the other corals,
Table 5. Δ14C and 210Pbex values for radiocarbon and lead-210 dating, respectively, and resulting ages from each dating method for specimens WPA 002, WPA 004,
and WPB 005. The banding frequency and growth rate for each specimen was calculated from the radiocarbon ages. The brackets around the 210Pbex values represent the
possible range of bands associated with this measurement.
Sample ID &
sub-section
Sub-annual
dissected band
count
Δ14C (‰)
Age from
radiocarbon (years)
Frequency (growth
band year-1)
Radial growth
rate (mm yr-1)
Vertical growth
rate (cm yr-1)
210Pbex
(dpm/g)
Age from
210Pb (years)
WPA 002
(Outer) 1
2
(Center) 3
(Center)� 3
WPA 004
(Outer) 1
2
3
4
5
6
7
(Center) 8
WPB 005
(Outer) 1
2
3
4
5
(Center) 6
6.5 ± 2
241 ± 56
477 ± 64
477 ± 64
13.5 ± 2
534 ± 38
712 ± 47
869 ± 19
1070 ± 81
1210 ± 6
1332 ± 33
1589 ± 124
23 ± 10
265 ± 52
348 ± 11
416 ± 28
534 ± 106
790 ± 93
-5.8 ± 2.0
31.5 ± 1.5
43.6 ± 1.7
43.6 ± 1.7
-12.2 ± 1.8
68.1 ± 1.6
80.0 ± 2.2
89.4 ± 2.0
75.3 ± 1.8
-22.5 ± 1.9
-56.1 ± 1.8
-93.7 ± 1.3
5.2 ± 1.8
87.2 ± 1.6
27.5 ± 1.9
-58.3 ± 1.7
-91.0 ± 1.4
-93.5 ± 1.6
0
38.2 ± 0.6
41.5 ± 0.1
49.1 ± 0.4
0
50.0 ± 0.3
47.7 ± 0.3
47.0
48.1 ± 0.1
51.7 ± 1.0
53.8 ± 0.2
65.7 ± 1.0
0 ± 0.7
19.5
22.4 ± 0.1
26.6 ± 0.2
35.4 ± 1.4
38.1 ± 1.3
0
6.1
16.1
53.6
0
11.6
81.5
203.5
177.1
39.5
56.4
17.5
0
12.4
29.3
16.1
13.3
83.6
—
—
0.26
0.307
—
—
—
—
—
—
—
0.404
—
—
—
—
—
0.644
—
—
3.49
4.12
—
—
—
—
—
—
—
3.4
—
—
—
—
—
4.2
2.52 ± 0.96
5.46 ± 1.28
2.82 ± 1.38
1.51 ± 0.73
—
—
—
0
0.90 ± 1.76
16.5
0.24 ± 0.69
59.7
3.57 ± 0.96
3.54 ± 1.28
0
0.3
1.58 ± 1.23
26.2
�WPA 002 center Δ14C value yielded two ages depending if the sample was placed on the rise or the decline of the bomb carbon.
https://doi.org/10.1371/journal.pone.0241692.t005
PLOS ONE | https://doi.org/10.1371/journal.pone.0241692 December 1, 2020
10 / 18
PLOS ONE Primnoidae age models
we estimate that the older value is more likely accurate and estimate the 14C-derived coral age
as 42 ± 0.1 years old.
210Pbex values ranged from 0.24 ± 0.69 to 3.57 ± 0.96 dpm/g (desentigrations per minute /
gram) and decreased as expected from the outside of the colony toward the inside for corals
WPA 004 and WPB 005 (Table 5; S2 Fig in S1 File). Eq (2) calculated the age of WPA 004 to
between 28.4 and 59.7 and WPB 005 to be 26.2 (+ 38.4/– 10.9) years old. 210Pb dating for sam-
ple WPA 002 was inconclusive because 210Pbex activity increased from the outside of the coral
toward the inside (Table 5; S2 Fig in S1 File).
Age estimation validation
In previously collected and aged corals (Table 2; corals from Aranha et al., 2104 [34] and Wil-
liams et al., 2007 [35]), we find that the calculated ages using Eq (6) were similar in magnitude
but underestimated age (on average ~ 10 years) relative to the reported ages (Fig 4).
Thus, we revised the age estimate equation to include the corals with reported basal diame-
ters from the Aranha et al. (2014) [34] and Williams et al. (2007) [35] for corals < century old:
Age �8ð
Þ ¼ e
Basal Diameterþ41:9
20:5
ð7Þ
that is inclusive of corals from the northeast Pacific Ocean (Fig 5; N = 21). Using this new
regional equation, the average difference between age estimation from annual band counts or
210Pb dating versus basal diameter declined to only three years (Fig 6).
We evaluated Eq (7) by calculating age in ten P. pacifica colonies collected from the Sea of
Japan [39]. We find that the age estimated from basal diameter underestimated age compared
to the annual band counts by an average of seven years, although this was largely driven by
older corals with wider diameters (Fig 7). In coral IDs 1 through 7, the basal diameter underes-
timated age by three years.
Discussion
The results of this analysis indicate a strong correlation between age and size (height, width,
and basal diameter) of Primnoa pacifica (Table 2; Fig 3), and thus provide support for develop-
ment of non-destructive techniques to estimate the age of younger (< century) specimens in
situ. In this species of gorgonian coral, the growth rates declined logarithmically with age such
Fig 4. Age estimations based on basal diameter (light grey bar) compared to age (annual bands; black line) of
Primnoa pacifica <100 years old from previous studies (Aranha et. al. 2014 [34]; Williams et al. 2007 [35]). Age
estimations were calculated using Eq 6 which translates basal diameter into a calculated age. Error bars show standard
deviation. �Age based on lead-210 dating not annual growth bands.
https://doi.org/10.1371/journal.pone.0241692.g004
PLOS ONE | https://doi.org/10.1371/journal.pone.0241692 December 1, 2020
11 / 18
PLOS ONE Primnoidae age models
Fig 5. Basal diameter of Primnoa pacifica corals from this study (white circles), Aranha et al., 2014 [34] (black
circles), and Williams et al., 2007 [35] (grey circles) decrease logarithmically with age (Eq (7), r2 = 0.66, p<0.001,
N = 21). Corals over a century in age were excluded from this analysis.
https://doi.org/10.1371/journal.pone.0241692.g005
that younger corals grew faster than older corals. This is consistent with other species of gorgo-
nians [32, 40, 41].
Radial growth rates were similar to those reported in previous studies, albeit on the higher
end of the reported ranges (Table 3). Axial growth rates were higher than previously reported
for two colonies, both of which were older than a century (Table 3). Thus, there is significant
variability in radial and axial growth rates, depending the age of the coral. The faster growth
rates reported here likely reflect the younger age of the corals. Because growth logarithmically
declines with age (Fig 3), eventually reaching a plateau–the age estimate techniques developed
here are less useful for older corals and may underestimate the age of very old corals.
Fig 6. Age estimations based on basal diameter (grey bar) compared to age (annual growth bands; black line) for
21 Primnoa pacifica colonies <100 years old from three studies (This study; Aranha et al. 2014 [34]; Williams
et al. 2007[35]). Age estimations were calculated using Eq (7). Colonies are ordered from left to right based on
increasing basal diameter measurements; diamond markers indicate a basal diameter of 39 mm or more. �Age based
on lead-210 dating not annual growth bands.
https://doi.org/10.1371/journal.pone.0241692.g006
PLOS ONE | https://doi.org/10.1371/journal.pone.0241692 December 1, 2020
12 / 18
PLOS ONE Primnoidae age models
Fig 7. Age estimations based on basal diameter (grey bar) compared to age (annual growth bands; black line) for
ten (n = 10) Primnoa pacifica colonies collected from the Sea of Japan [39]. Age estimations were calculated using
Eq (7). Colonies are ordered from left to right based on increasing basal diameter measurements.
https://doi.org/10.1371/journal.pone.0241692.g007
Age estimation validation
Age estimated from basal diameter largely agreed with age from annual band counts in a suite
of corals from the northeast Pacific (Fig 6). Discrepancies between the age estimation and
annual band counts was the greatest for the oldest specimens with the largest diameters. We
hypothesize that smaller corals represent colonies with minimal branching and symmetric
radii. In contrast, the larger diameter colonies have more asymmetric basal growth (in which
the basal diameter may not reflect 2 x radius) and often with many branches (e.g., specimen
WPB 005, Fig 2). Thus Eq 7 is most effective for colonies with basal diameter < 39 mm to pro-
vide approximate age estimates for P. pacifica corals in the North Pacific Ocean.
When the northeast Pacific regional age estimation (Eq 7) was applied to corals from the
western Pacific Ocean, basal diameter underestimated age in the oldest colonies. These colo-
nies were, on average, collected from colder, deeper waters (350–505 m; 0.6–0.7˚C) and exhib-
ited slower radial growth rates (0.1–0.19 mm year-1) than those from the northeast Pacific
Ocean [39]. Thus, growth rates may plateau earlier in these corals than in the northeast Pacific
Ocean.
Some coral colonies in this study were collected remotely with fishing techniques, which
tends to result in fragments, or portions, of a full colony, such that accurate height and width
measurements are unavailable, so in-situ age estimations were not possible based on these
properties. However, the equations can also be retroactively applied to previous studies that
collected morphological data via video analysis but that did not collect corals for age analyses.
A Stone (2014) [3] study used video with scaling lasers to categorize red tree corals into four
size (height) categories: <0.5 m (small), 0.5–1 m (medium), 1–2 m (large), and >2 m (very
large). The height categories can be transformed into the following approximate age ranges
using Eq (4): 0–9 years, 9–18 years, 18–71 years, and 71+ years. Equs (4), (5) and (7) provide
approximate age estimations for corals based solely on morphological data derived from in
situ observations. The age estimates are most accurate for the smaller, younger corals
(< century) because of the plateau in growth with age. Using this technique, it is possible to
identify the locations of gorgonian thickets where recruitment is particularly high (many
younger colonies), or colonies may serve as a source of the recruits (many larger colonies), and
to locate climax communities where all age classes are well represented.
PLOS ONE | https://doi.org/10.1371/journal.pone.0241692 December 1, 2020
13 / 18
PLOS ONE Primnoidae age models
Insights into coral dating
Three techniques determined the ages of a subset of the collected corals: annual growth bands
were counted in 11 corals, and 14C and 210Pb dating techniques were performed on three cor-
als. In some cases, results were ambiguous. In P. pacifica, growth bands form annually but are
sometimes difficult to distinguish. Furthermore, radial growth does not always form concen-
trically, meaning one direction of the coral can form more growth bands than other directions
(Fig 2). Thus, most researchers count along the axes of maximum growth yielding the highest
number of growth bands. In P. pacifica, annual growth bands form as clumps of gorgonin
bands with some contributions of calcite. Older P. pacifica corals form significant calcite build-
ups that mask growth bands; consequently, radiometric estimates of coral age typically exceed
those of ages derived from growth bands for older corals [34].
The modern use of radiocarbon dating relies on the identification of the 14C bomb curve
signature from thermonuclear bomb detonations in the 1950-60s. In the three corals analyzed
for 14C content, identification of bomb-derived carbon provided age constraints on the timing
of skeletal growth. The coral sample sizes analyzed for 14C typically included multiple years of
growth, smoothing any seasonal variability in ocean 14C so that the coral 14C values should
align with the regional 14C bomb curves. However, exact placement on that curve is subject to
uncertainty, which is compounded by the analytical uncertainty of the 14C measurement.
There are two instances in this study where these uncertainties impact age estimates. In WPA
002, the earliest 14C value measured from the core could either fall on the rising or declining
limb of the bomb carbon curve (S1 Fig in S1 File). We assign it to the rise of the curve because
otherwise radial growth rates are unreasonably high for the earliest part of this coral’s growth.
In WPB 005, the coral was collected with no living tissue at the base of the specimen; thus, the
outer layers of the skeleton were likely not formed immediately prior to collection. The 14C
value for the outer layers is higher than the outer layers of the live collected WPA 002 and
WPA 004, indicating it died some time prior to collection. However, because the slope of the
decline of the 14C bomb curve flattens toward recent time, the placement of this sample on the
curve has large uncertainty with time. Aligning the point exactly on the regional bomb curve
gives an age of 38 years for the coral, which yields potentially unrealistically high growth rates.
In this colony, annual band counts suggest a much older coral, with a death of only a few years
prior to collection. As a result, the strength of 14C dating is most evident for specimens with
known collection date, and with an age range that encompass the full bomb curve (extending
prior to the mid-1950s).
The 210Pb dating technique provides only rough age estimates. For example, 210Pb of a P.
pacifica colony yielded an age range of 78 to 193 years with 95% confidence [26]. 210Pb dating
can also yield inconclusive results, potentially in corals that are collected dead, although a
mechanism of why is unknown [26]. Here, 210Pb yields an age for the dead colony WPB 005,
albeit this age may be unrealistically young when viewed in the context of specimen size. In
contrast, 210Pb yielded inconclusive results for the small live-collected coral WPA 002. Due to
the large uncertainty around 210Pb, we propose that 14C and growth band counting are pre-
ferred methods for determining coral age in this species.
Sub-annual growth bands
We closely examined the sub-annual banding using two techniques: growth bands counted
through physical dissection of a cross-section and those visible in photographs of polished
cross-sections. The number of dissected growth bands counted was in most cases much greater
than the number counted in photographs, although the opposite was true for specimen WPB
004 (Table 4). The skeletal bands were difficult to count, which likely contributed to some of
PLOS ONE | https://doi.org/10.1371/journal.pone.0241692 December 1, 2020
14 / 18
PLOS ONE Primnoidae age models
the variation within each counting method–particularly with the photographed bands. Fused
growth bands could, in some cases, be carefully peeled apart and counted, although more com-
monly a peeled fused section included numerous bands that could not be separated. These
bands were likely not individually visible in the photographed sections; thus, they were not
included in the photographed band counts.
We compared the number of sub-annual bands with the age of each coral to determine
banding frequency. Overall, photographed bands average 14 ± 0.80 sub-annual bands year-1 and
dissected bands average 18 ± 5.2 sub-annual bands year-1. We did not count the number of sub-
annual bands per year directly because the annual groupings of sub-annual bands were not visible
when physically dissecting the cross-sections nor at the lower magnification needed to count the
annual bands in the photographs. However, when compared to ages derived from 14C-dating, the
sub-annual banding frequency ranged from 200 bands year-1 to less than 15 bands year-1 for spec-
imen WPA 004, 80 bands year-1 to less than 15 bands year-1 for specimen WPB 005, and over 15
bands year-1 to less than 10 bands year-1 for specimen WPA 002 (Table 5). Thus, the banding fre-
quency can be highly variable through time. Spring tides (26 per year) have been proposed as a
primary driver of sub-annual banding due to the influx of sedimentary organic layer [9]. How-
ever, the variability in frequency and the number of bands per year is not consistent with only
spring tides. We instead hypothesize that high seasonal and interannual variability of primary pro-
ductivity [42] and/or energy allocation to reproduction [43, 44] combined with spring tides can
influence the variable banding and growth pattern. Additional collections of colonies spanning
the size range of the species coupled with time series in situ measurements of primary productiv-
ity, specifically flux (POC or particulate organic carbon) delivery to the seafloor, and studies on
reproductive seasonality of P. pacifica in the eastern Gulf of Alaska could elucidate the relationship
between these factors and the sub-annual skeletal development.
Growth rates
Reported radial growth rates of Primnoa pacifica range from 0.14 to 0.74 mm year-1 (Table 3),
with the corals in this study on the higher end of that range 0.33 to 0.74 mm year-1, compared
with 0.22 to 0.57 mm year-1 (Aranha et al., 2014 [34]), 0.36 mm year-1 (Andrews et al., 2002
[26]), and 0.14 to 0.37 mm year-1 (Williams et al., 2007 [35]). Expanding the age–growth com-
parisons to include all of these studies shows that radial growth rates are lower in older speci-
mens (linear regression; p = 0.004, r2 = 0.35, N = 22). The decrease in radial growth rates
overtime is also observed in North Atlantic Primnoa resedaeformis, a slower growing congener
of P. pacifica [9, 32, 41]. Additionally, reported radial growth rates of P. resedaeformis (0.083 to
0.215 mm year-1) varied based on colony location; colonies in areas with a stronger tidal cur-
rent exhibited faster radial growth than colonies in areas of weaker current [32]. There was no
statistical difference between location and radial growth rates of the P. pacifica included in this
study (one-way ANOVA; F4,17 = 1.15, p>0.05).
Axial growth rates ranged from 2.41 to 6.39 cm year-1 (Table 3). This is substantially faster
than previously reported axial growth rates of 1.74 and 2.32 cm year-1 (Andrews et al., 2002
[26]). This difference is easily explained by the decline in growth rates with age, and that the
specimens measured by Andrews et al. (2002) [26] were substantially older than those in the
present study (age of 114 years is reported for the specimen growing 1.74 cm year-1, with no
age for the second colony). Age continues to have a strong, significant explanation of axial
growth rates when we add this one specimen with both age and axial growth rates to our age–
growth comparisons (linear regression; p<0.0001, r2 = 0.80, N = 12). A similar decrease in
axial growth rates with age was reported in North Atlantic P. resedaeformis, where it was pro-
posed that young corals (<30 years) grew four times as fast as older corals (>30 years) [41].
PLOS ONE | https://doi.org/10.1371/journal.pone.0241692 December 1, 2020
15 / 18
PLOS ONE Primnoidae age models
Axial growth rates in these P. resedaeformis studies ranged from 1.00 to 2.61 cm year-1 (coral
ages ranged from 18 years to 100 years) and were on the lower end compared to P. pacifica in
this study [32, 41].
Implications for coral conservation
With the ability to now estimate the age of red tree corals in situ we can readily determine how
old corals are using non-invasive video survey techniques coupled with mensuration systems
such as lasers or stereo-cameras [4, 6]. Such enhanced surveys could quickly determine the
age-class structure and consequently maturity status of coral habitats. This information could
be used by coastal managers to identify which aggregations are most vulnerable to disturbance
from human activities, and which should be highlighted for protection. If age and growth char-
acteristics are phylogenetically constrained, as has been suggested for some taxa [40], then the
techniques developed and insights gained in this study could have broader application in the
North Atlantic Ocean where another Primnoa species (e. g. P. resedaeformis; [13, 45]) also
forms ecologically important habitats.
Supporting information
S1 File.
(PDF)
Acknowledgments
We thank Ocean Science Services and the captain and crew of the FV Alaska Provider (2013) and
Pelagic Research Services and the captain and crew of the RV Dorado Discovery (2015) for their
support. Many thanks to Enrique Salgado and Robert McGuinn for assistance with specimen col-
lection and field preparation. The findings and conclusions in this report are those of the author
(s) and do not necessarily represent the views of the National Marine Fisheries Service (NOAA).
The work was performed under the authority of the U.S. Department of Commerce.
Author Contributions
Conceptualization: Branwen Williams, Peter Etnoyer.
Data curation: Emma Choy, Kelly Watanabe, Robert Stone, Thomas Lorenson, Mary Knaak.
Formal analysis: Emma Choy, Kelly Watanabe, Ellen Druffel, Mary Knaak.
Funding acquisition: Branwen Williams, Peter Etnoyer.
Methodology: Branwen Williams.
Project administration: Branwen Williams.
Supervision: Robert Stone.
Writing – original draft: Emma Choy, Kelly Watanabe, Branwen Williams, Robert Stone.
Writing – review & editing: Emma Choy, Kelly Watanabe, Branwen Williams, Robert Stone,
Ellen Druffel.
References
1. Althaus F., Williams A., Schlacher T.A., Kloser R.J., Green M.A., Barker B.A., et al. 2009. Impacts of
bottom trawling on deep-coral ecosystems of seamounts are long-lasting. Marine Ecology Progress
Series 397, 279–294.
PLOS ONE | https://doi.org/10.1371/journal.pone.0241692 December 1, 2020
16 / 18
PLOS ONE Primnoidae age models
2. Davies A.J., Roberts J.M. and Hall-Spencer J. 2007. Preserving Deep-Sea Natural Heritage: Emerging
Issues in Offshore Conservation and Management. Biological Conservation 138, 299–312.
3. Stone R., 2014. The ecology of deep-sea coral and sponge habitats of the Aleutian Islands of Alaska.
NOAA Profissional Paper NMFS 16, 52 pp, https://doi.org/10.7755/PP.7719
4. Stone R., Masuda M., Karinen J., 2014. Assessing the ecological importance of red tree coral thickets
in the eastern Gulf of Alaska. ICES Journal of Marine Sciences, https://doi.org/10.1093/icesjms/
fsu1190
5. Clark M.R., Althaus F., Schlacher T., Williams A., Bowden D.A., and Rowden A.A. 2016. The Impacts of
Deep-Sea Fisheries on Benthic Communities: A Review. ICES Journal of Marine Science 73, 51–69.
6. Etnoyer P., Wagner D., Fowle H., Poti M., Kinlan B., Georgian S., et al. 2018. Models of habitat suitabil-
ity, size, and age-class structure for the deep-sea black coral Leiopathes glaberrima in the Gulf of
Mexico. Deep Sea Research Part II: Topical Studies in Oceanography 150, 218–228.
7. Krieger K., Wing B., 2002. Megafaunal associations with deepwater corals (Primnoa sp.) in the Gulf of
Alaska. Hydrobiologia 471, 83–90.
8. Morgan L., Etnoyer P., Scholz A., Mertens M., Powell M., 2005. Conservation and management impli-
cations of deep-sea coral and fishing effort distributions in the Northeast Pacific Ocean. Cold-Water
Corals and Ecosystems, 1171–1187.
9. Risk M., Heikoop J., Snow M., Beukens R., 2002. Lifespans and growth patterns of two deep-sea cor-
als: Primnoa resedaeformis and Desmophyllum cristigalli. Hydrobiologia 471, 125–131.
10. Hapsari K.A., Biagioni S.A., Jennerjahn T.C., Reimer P., Saad A., Sabiham S., et al. 2018. Resilience of
a Peatland in Central Sumatra, Indonesia to Past Anthropogenic Disturbance: Improving Conservation
and Restoration Designs Using Palaeoecology. Journal of Ecology 106, 2473–2490.
11.
Lovich J.E., and Bainbridge D., 1999. Anthropogenic Degradation of the Southern California Desert
Ecosystem and Prospects for Natural Recovery and Restoration. Environmental Management 24,
309–326. https://doi.org/10.1007/s002679900235 PMID: 10486042
12. Du Preez C.D., and Tunnicliffe V., 2011. Shortspine thornyhead and rockfish (Scorpaenidae) distribu-
tion in response to substratum, biogenic structures and trawling. Marine Ecology Progress Series 425,
217–231.
13. Risk M., McAllister D., Behnken L., 1998. Conservation of cold- and warm-water seafans: threatened
ancient gorgonian groves. Sea Wind 12, 2–21.
14. Roberts S., Hirshfield M., 2004. Deep-sea corals: out of sight but no longer out of mind. Frontiers in
Ecology and the Environment 2, 123–130.
15. Stone RP, Shotwell SK (2007) State of the deep coral ecosystems of the Alaska Region: Gulf of Alaska,
Bering Sea, and the Aleutian Islands. In: Lumsden SE, Hourigan TF, Bruckner AW, Dorr G, editors. The
State of Deep Coral Ecosystems of the United States, pp. 65–108. United States Department of Com-
merce, NOAA Technical Memorandum CRCP-3. Silver Spring, MD.
16. Verdura J., Linares C., Ballesteros E., Coma R., Uriz M.J., Bensoussan N., et al. 2019. Biodiversity loss
in a Mediterranean ecosystem due to an extreme warming event unveils the role of an engineering gor-
gonian species. Scientific Reports 9, 5911. https://doi.org/10.1038/s41598-019-41929-0 PMID:
30976028
17. Gugliotti E.F., DeLorenzo M.E., Etnoyer P.J., 2019. Depth-dependent temperature variability in the
Southern California bight with implications for the cold-water gorgonian octocoral Adelogorgia phyllo-
sclera. Journal of Experimental Marine Biology and Ecology 514– 515, 118–126.
18. DeLeo D., Ruiz-Ramos D., Baums I., Cordes E., 2016. Response of deep-water corals to oil and chemi-
cal dispersant exposure. Deep Sea Research Part II: Topical Studies in Oceanography 129, 137–147.
19.
Frometa J., DeLorenzo M., Pisarski E., Etnoyer P., 2017. Toxicity of oil and dispersant on the deep
water gorgonian octocoral Swiftia exserta, with implications for the effects of the Deepwater Horizon oil
spill. Marine Pollution Bulletin 122, 91–99. https://doi.org/10.1016/j.marpolbul.2017.06.009 PMID:
28666594
20. Kinoshita K., 1907. Vorlaufige Mitteilung uber einige neue japanische Primnoidkorallen. Annotationes
zoologicae Japonenses 6, 229–237.
21. Rossin A.H., Waller R.G., Stone R.P. 2019. The effects of in-vitro pH decrease on the gametogenesis
of the red tree coral, Primnoa pacifica. PLoS ONE 14, e0203976. https://doi.org/10.1371/journal.pone.
0203976 PMID: 30998686
22. Griffin S., & Druffel E. (1989). Sources of Carbon to Deep-Sea Corals. Radiocarbon, 31(3), 533–543.
23. Roark E. B., Guilderson T. P., Flood-Page S., Dunbar R. B., Ingram B. L., Fallon S. J., et al. (2005),
Radiocarbon-based ages and growth rates of bamboo corals from the Gulf of Alaska, Geophys. Res.
Lett., 32, L04606, https://doi.org/10.1029/2004GL021919
PLOS ONE | https://doi.org/10.1371/journal.pone.0241692 December 1, 2020
17 / 18
PLOS ONE Primnoidae age models
24. Sherwood OA., Scott DB., Risk MJ., Guilderson, TP. (2005) Radiocarbon evidence for annual growth
rings in the deep-sea octocoral Primnoa resedaeformis. Marine Ecology Progress Series. 301:129–
134.
25. Rooper C., Stone R., Etnoyer P., Conrath C., Reynolds J., Greene H.G., et al. 2017. Deep-Sea Coral
Research and Technology Program: Alaska Deep-Sea Coral and Sponge Initiative Final Report.
NOAA Tech. Memo. NMFS-OHC-2, 65 p.
26. Andrews A., Cordes E., Mahoney M., Munk K., Coale K., Cailliet G., et al. 2002. Age, growth and radio-
metric age validation of a deep-sea, habitat-forming gorgonian (Primnoa resedaeformis) from the Gulf
of Alaska, in: Watling L., Risk M. (Eds.), Hydrobiologia. Kluwer Academic Publishers, Netherlands.
27. Grigg R., 1974. Growth rings: Annual periodicity in two gorgonian corals. Ecology 55, 876–888.
28. Mistri M., Ceccherelli V., 1994. Growth and secondary production of the Mediterranean gorgonian Para-
muricea clavata. Marine Ecology Progress Series 103, 291–296.
29. Mitchell N., Dardeau M., Schroeder W., 1993. Colony morphology, age structure, and relative growth of
two gorgonian corals, Leptogorgia hebes (Verrill) and Leptogorgia virgulata (Lamarck), from the north-
erm Gulf of Mexico. Coral Reefs 12, 65–70.
30. Goldberg W., 1974. Evidence of a sclerotized collagen from the skeleton of a gorgonian coral. Compar-
ative Biochemistry and Physiology Part B: Comparative Biochemistry 49, 525–526. https://doi.org/10.
1016/0305-0491(74)90188-6 PMID: 4154179
31. Szmant-Froelich A., 1974. Structure, iodination and growth of the axial skeletons of Muricea californica
and M. fruticosa (Coelenterata: Gorgonacea). Marine Biology 27, 299–306.
32. Sherwood OA, and Edinger E. (2009) Ages and Growth Rates of Some Deep-Sea Gorgonian and Anti-
patharian Corals of Newfoundland and Labrador, Canadian Journal of Fisheries and Aquatic Sciences,
2009, 66(1): 142–152.
33. Noe´ S.U., Lembke-Jene L. & Dullo W. (2008) Varying growth rates in bamboo corals: sclerochronology
and radiocarbon dating of a mid-Holocene deep-water gorgonian skeleton (Keratoisis sp.: Octocorallia)
from Chatham Rise (New Zealand). Facies 54, 151–166 (2008).
34. Aranha R., Edinger E., Layne G., Piercey G., 2014. Growth rate variation and potential paleoceano-
graphic proxies in Primnoa pacifica: Insights from high-resolution trace element microanalysis. Deep-
Sea Research II 99, 213–226.
35. Williams B., Risk M., Stone R., Sinclair D., Ghaleb B., 2007. Oceanographic changes in the North
Pacific Ocean over the past century recorded in deep-water gorgonian corals. Marine Ecology Progress
Series 335, 85–94.
36. Stuiver M. and Polach H.A. (1977) Discussion Reporting of 14C Data. Radiocarbon. 19:355–363.
37. Andrews A., Leaf R., Rogers-Bennett L., Neuman M., Hawk H., Cailliet G. 2013. Bomb radiocarbon dat-
ing of the endangered white abalone (Haliotis sorenseni): investigations of age, growth and lifespan.
Marine and Freshwater Research 64, 1029–1039.
38. Kerr L.A., Andrews A.H., Frantz B.R., Coale K.H., Brown T.A., and Cailliet G.M. 2004. Radiocarbon in
otoliths of yelloweye rockfish (Sebastes ruberrimus): a reference time series for the coastal waters of
southeast Alaska. Canadian Journal of Fisheries and Aquatic Sciences 61, 443–451.
39. Matsumoto A., 2007. Effects of low water temperature on growth and magnesium carbonate concentra-
tions in the cold-water gorgonian Primnoa pacifica. Bulletin of Marine Science 81: 423–435.
40. Stone R., Malecha P., Masuda M., 2017. A Five-Year, In Situ Growth Study on Shallow-Water Popula-
tions of the Gorgonian Octocoral Calcigorgia spiculifera in the Gulf of Alaska. PLos ONE 12, e0169470.
https://doi.org/10.1371/journal.pone.0169470 PMID: 28068374
41. Mortensen P.B. and Buhl-Mortensen L. 2005. Morphology and growth of the deep-water gorgonians
Primnoa resedaeformis and Paragorgia arborea. Marine Biology. 147: 775–788.
42. Balcom B., Biggs D., Hu C., Montaga P., Stockwell D., 2011. A comparison of marine productivity
among Outer Continental Shelf planning areas, in: Prepared by CSA International Inc. for the U. S.
Dept. of the Interior, Bureau of Ocean Energy Management, Regulation and Enforcement (Eds.), Hern-
don, VA., OCS Study BOEMRE 2011– 019, 195 pp.
43. Waller R.W., Stone R.P., Johnstone J., Mondragon J., 2014. Sexual reproduction and seasonality of
the Alaskan red tree coral, Primnoa pacifica. PLos ONE 9, e90893. 908 https://doi.org/10.1371/journal.
pone.0090893 PMID: 24770675
44. Waller R.W., Stone R. P., Rice L.N., Johnstone J., Rossin A.M., Hartill E., et al. 2019. Phenotypic plas-
ticity or a reproductive dead end? Primnoa pacifica (Cnidaria: Alcyonacea) in the southeastern Alaska
region. Frontiers in Marine Science 6, https://doi.org/103389/fmars.2019.00709
45. Auster P.J., Kilgour M., Packer D., Waller R., Auscavitch S. Watling L. 2013. Octocoral gardens in the
Gulf of Maine (NW Atlantic). Biodiversity 14, 193–194.
PLOS ONE | https://doi.org/10.1371/journal.pone.0241692 December 1, 2020
18 / 18
PLOS ONE |
10.1371_journal.pone.0229744 | RESEARCH ARTICLE
KLF4 is required for suppression of histamine
synthesis by polyamines during bone
marrow-derived mast cell differentiation
Kazuhiro NishimuraID
Toru Shibayama1, Tetsuhiko Yoshino1, Teruki Murakami1, Masashi Yamaguchi2,
Satoshi Tanaka3, Toshihiko Toida1, Kazuei Igarashi1,4
1*, Moemi Okamoto1, Rina Shibue1, Toshio Mizuta1,
1 Graduate School of Pharmaceutical Sciences, Chiba University, Chiba, Japan, 2 Medical Mycology
Research Center, Chiba University, Chiba, Japan, 3 Department of Pharmacology, Division of Pathological
Sciences, Kyoto Pharmaceutical University, Kyoto, Japan, 4 Amine Pharma Research Institutes, Chiba,
Japan
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
* kaznishi@faculty.chiba-u.jp
Abstract
OPEN ACCESS
Citation: Nishimura K, Okamoto M, Shibue R,
Mizuta T, Shibayama T, Yoshino T, et al. (2020)
KLF4 is required for suppression of histamine
synthesis by polyamines during bone marrow-
derived mast cell differentiation. PLoS ONE 15(2):
e0229744. https://doi.org/10.1371/journal.
pone.0229744
Editor: Sakamuri V. Reddy, Charles P. Darby
Children’s Research Institute, 173 Ashley Avenue,
Charleston, SC 29425, USA, UNITED STATES
Received: November 26, 2019
Accepted: February 13, 2020
Published: February 26, 2020
Peer Review History: PLOS recognizes the
benefits of transparency in the peer review
process; therefore, we enable the publication of
all of the content of peer review and author
responses alongside final, published articles. The
editorial history of this article is available here:
https://doi.org/10.1371/journal.pone.0229744
Copyright: © 2020 Nishimura et al. This is an open
access article distributed under the terms of the
Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Mast cells have secretory granules containing chemical mediators such as histamine and
play important roles in the immune system. Polyamines are essential factors for cellular pro-
cesses such as gene expression and translation. It has been reported that secretory gran-
ules contain both histamine and polyamines, which have similar chemical structures and are
produced from the metabolism of cationic amino acids. We investigated the effect of poly-
amine depletion on mast cells using bone marrow-derived mast cells (BMMCs). Polyamine
depletion was induced using α-difluoromethylornithine (DFMO), an irreversible inhibitor of
ornithine decarboxylase. DFMO treatment resulted in a significant reduction of cell number
and abnormal secretory granules in BMMCs. Moreover, the cells showed a 2.3-fold increase
in intracellular histamine and up-regulation of histidine decarboxylase (HDC) at the tran-
scriptional level during BMMC differentiation. Levels of the transcription factor kruppel-like
factor 4 (KLF4) greatly decreased upon DFMO treatment; however, Klf4 mRNA was
expressed at levels similar to controls. We determined the translational regulation of KLF4
using reporter genes encoding Klf4-luc2 fusion mRNA, for transfecting NIH3T3 cells, and
performed in vitro translation. We found that the efficiency of KLF4 synthesis in response to
DFMO treatment was enhanced by the existence of a GC-rich 50-untranslated region (50-
UTR) on Klf4 mRNA, regardless of the recognition of the initiation codon. Taken together,
these results indicate that the enhancement of histamine synthesis by DFMO depends on
the up-regulation of Hdc expression, achieved by removal of transcriptional suppression of
KLF4, during differentiation.
Introduction
Polyamines are small basic molecules with multiple amino groups and are involved in cell pro-
liferation and differentiation [1, 2]. Three polyamines (putrescine, spermidine and spermine)
PLOS ONE | https://doi.org/10.1371/journal.pone.0229744 February 26, 2020
1 / 16
Data Availability Statement: All relevant data are
within the manuscript and its Supporting
Information files.
Funding: This work was supported by a Grant-in-
Aid for Scientific Research 15K07921 from the
Ministry of Education, Culture, Sports, Science and
Technology, Japan.
Competing interests: The authors have declared
that no competing interests exist.
Translational regulation of KLF4 by polyamines
are present in mammalian cells at the millimolar level, and are stringently controlled by bio-
synthesis, degradation, and transport [3]. Putrescine is synthesized from ornithine by orni-
thine decarboxylase (ODC), and spermidine is synthesized from putrescine by the addition of
an aminopropyl group donated from decarboxylated S-adenosylmethionine, which is synthe-
sized from S-adenosylmethionine by S-adenosylmethionine decarboxylase (AdoMetDC). Sim-
ilarly, spermine is synthesized from spermidine by the addition of another aminopropyl
group. These enzymes catalyze rate-limiting steps in polyamine biosynthesis. Studies on mice
with disruption of ODC- and AdocMetDC-encoding genes revealed that these deficient mice
were embryonic lethal [4, 5]. It has been demonstrated that cellular polyamines mostly interact
with RNA and consequently affect translation efficiency of several kinds of proteins [6]. The
mRNAs encoding these proteins have characteristic nucleotide sequences in their 50-UTRs
that are difficult to translate. In addition, the term ‘polyamine modulon’, is used to describe
genes encoding proteins regulated by polyamines at the translational level [7]. We have previ-
ously reported three kinds of genes (Cct2, Hnrpl, and Pgam1) as the polyamine modulon and
clarified one of the mechanisms of polyamine stimulate protein synthesis in eukaryotes [8].
Thus, polyamine mediated stimulation of protein synthesis contributes to cell proliferation
and differentiation.
Mast cells are present in connective and submucosal tissues of mammals and are known
to be involved in immunological processes contributing to both innate and adaptive immu-
nity in defense against pathogens [9, 10]. They contain many secretory granules with sergly-
cin proteoglycans (PGs) and important chemical mediators [11]. Cutaneous mast cells
preferentially express serglycin PGs with heparin chains in humans and mice, whereas jejunal
mast cells and BMMCs preferentially express serglycin PGs with chondroitin sulfate-E chains
[12]. The absence of serglycin was shown to result in absence of metachromatic staining and
resulted in severely defective storage of granule proteases and biogenic amines [13, 14].
Therefore, serglycin PGs manifest strong electronic interactions with secretory granule com-
pounds and play an essential role in maintaining the morphological characteristics of mast
cells [11].
It was previously found that blood coagulation in the presence of heparin was enhanced by
polyamines due to interactions between heparin and polyamines [15]. Moreover, glycosami-
noglycan has been shown to be involved in polyamine transport: heparan sulfate on glypican-1
acted as a vehicle for polyamine uptake [16]. Garcı´a-Faroldi et al reported that polyamine
depletion resulted in the upregulation of Hdc expression and activity, accompanied by
increased histamine levels during early stages of BMMC differentiation [17]. However, the
mechanism of upregulation of Hdc expression was unclear. A study in 2004 showed that KLF4
suppresses expression of Hdc in gastric cancer [18]. There is little known about the role of
KLF4 in mast cells.
Here, we investigated the effects of polyamine depletion, using a low concentration of
DFMO, on BMMC differentiation and function. We demonstrated that KLF4 synthesis is reg-
ulated at the translational level by polyamines and is involved in histamine synthesis in mast
cells.
Materials and methods
Animals
All animal experiments were approved by the Institutional Animal Care and Use Committee
of Chiba University and carried out according to the Guidelines for Animal Research of Chiba
University. Female C57BL/6 mice were obtained from Japan SLC, Inc.
PLOS ONE | https://doi.org/10.1371/journal.pone.0229744 February 26, 2020
2 / 16
Translational regulation of KLF4 by polyamines
Materials
Calcium ionophore A23187 was purchased from Sigma-Aldrich. Purified mouse IgE κ Isotype
Control (clone C38-2) and purified rat anti-mouse IgE (clone R35-72) were from BD Biosci-
ences Pharmingen. Recombinant murine IL-3 was purchased from PeproTech, Inc.
Preparation of BMMCs. BMMCs wer prepared as previously described [19]. Female,
8–12 week-old, C57BL/6 mice were euthanized by cervical dislocation, during isoflurane anes-
thesia, tibia bone marrow cells were isolated, and cultured in RPMI-1640 containing 10% fetal
bovine serum, 10 ng/mL IL-3, 50 μM β-mercaptoethanol, 0.1 mM non-essential amino acids,
50 U/mL penicillin G and 50 U/mL streptomycin. Cells were subcultured at 5 x 105 cells/mL
with 3–4 d intervals for 28–34 d to obtain mature BMMCs. Maturation of BMMCs was con-
firmed by toluidine blue staining.
Transmission electron microscopy. Cells were fixed with 2.5% glutaraldehyde and kept
on ice. They were post-fixed with 1% osmium tetroxide, dehydrated with a graded series of
ethanol, and embedded in an epoxy resin [20]. Ultrathin sections were cut, stained with uranyl
acetate and lead citrate, and observed under a JEM-1400 electron microscope (JEOL Ltd.)
[21].
Degranulation assay. Cells (1 x 106) were stimulated by either the calcium ionophore
A23187 or IgE/anti-IgE for this assay. In the A23187 stimulation method, the cells were
washed and suspended in Tyrode’s buffer (10 mM HEPES/NaOH (pH 7.4), 130 mM NaCl, 5
mM KCl, 1.4 mM CaCl2, 1 mM MgCl2, 5.6 mM glucose, and 0.1% BSA). After preincubation
for 10 min at 37 ˚C, the cells in 200 μL of Tyrode’s buffer containing 2 μM A23187 were incu-
bated for 15 min at 37 ˚C and placed on ice for 10 min. In the IgE/anti-IgE stimulation
method, the cells were washed with phosphate buffer saline (PBS), suspended with 1 μg/mL
anti-TNP mouse IgE in conditioned medium and incubated for sensitization for 6 h at 37 ˚C.
The sensitized cells were washed with PBS, suspended in 200 μL of Tyrode’s buffer containing
2 μg/mL anti-mouse IgE, incubated for 1 h at 37 ˚C and placed on ice for 10 min. To evaluate
the degranulation rate of the two methods, β-hexosaminidase activity in supernatants and cell
pellets (solubilized with 1% Triton-X-100 in Tyrode’s buffer) was determined. The samples,
with 1 mM p-nitrophenyl N-acetyl-β-D-glucosaminide in 50 mM citrate buffer (pH 4.5), were
incubated for 1 h at 37 ˚C; 2 volumes of 0.1 M Na2CO3/NaHCO3 (pH 10) was added to stop
the reaction and the release of p-nitrophenol was measured spectrophotometrically at 405 nm.
The degranulation rate was calculated using the absorbance of the supernatant/the total absor-
bance (supernatant + cell pellet).
Measurement of amines.
Intracellular amine (polyamines and histamine) levels in
BMMCs were determined according to the method described previously [22] with minor
modifications. Cells (1 x 106) or culture medium were treated with final 5% trichloroacetic
acid and centrifuged at 10,000 x g for 5 min. Isocratic single or tandem-mode HPLC system,
with one or two columns (TSKgel polyaminepak; Tosoh Corporation) were used to analyze
the supernatants. Eluent (pH 5.6) was used to separate amines for 30 min in the single-mode,
and eluent (pH 5.3) for 120 min in the tandem-mode. The amount of protein present in
trichloroacetic acid -precipitates was evaluated using Bio-Rad Protein Assay (Bio-Rad
Laboratories).
Measurement of HDC activity. HDC assay was performed following the method
described [23], with minor modifications. Cells (5 x 106) were washed with PBS and suspended
with 320 μL of HDC assay buffer (100 mM potassium phosphate (pH 6.8), 0.2 mM dithiothrei-
tol, 10 μM pyridoxal 50-phosphate, 2% polyethylene glycol 300, 20 μM FUT-175, and 1% Tri-
tonX-100). Thereafter 150 μL of cell lysate was centrifuged at 16,000 x g for 15 min, incubated
with 50 μL of HDC buffer containing 3.2 mM L-histidine, for 2 h, and treated with 5%
PLOS ONE | https://doi.org/10.1371/journal.pone.0229744 February 26, 2020
3 / 16
Translational regulation of KLF4 by polyamines
trichloroacetic acid (final concentration) for stopping the enzymatic reaction. Histamines pro-
duced were analyzed by the single-mode HPLC system described above.
Quantitative RT-PCR analysis. Total RNA was isolated from 1 x 106 cells using the
RNeasy Micro Kit (Qiagen) and used for first-strand cDNA synthesis using SuperScript™ II
Reverse Transcriptase (Invitrogen). Real-time PCR was performed using KAPA SYBR FAST1
qPCR kit (Kapa Biosystems Inc.) and Eco Real-Time PCR System (Illumina Inc.) The primer
pairs (forward, reverse) used for PCR were as follows; Hdc, 50-ATC TCT GGT CAG AAG
CGA CCC TTC-30, 50-GCT CCT GGC TGC TTG ATG ATC TTC-30, Klf4, 50-GCA GGC
TGT GGC AAA ACC TAT AC-30, 50-CTG ACT TGC TGG GAA CTT GAC-30, Gapdh,
50-GGT ATC GTG GAA GGA CTC ATG AC-30, 50-ATG CCA GTG AGC TTC CCG
TTC AGC-30.
Western blotting. Western blot analysis was performed as described previously [24]. Pri-
mary antibodies used were rabbit polyclonal KLF4 antibody (1:2,000; GTX101508; GeneTex)
and mouse monoclonal β-actin antibody (1:2,000; ab8226; Abcam). Secondary antibodies used
were rabbit IgG HRP Linked Whole Ab (1:10,000; NA943; GE Healthcare) and mouse IgG
HRP Linked Whole Ab (1:10,000; NA941; GE Healthcare).
Reporter plasmid construction. The pGL4.13[luc2/SV40] vector was purchased from
Promega. For the ΔE2 plasmid, the Klf4 E1 fragment (605 bp) was obtained by PCR using
primers (forward, 50-CAA AAA GCT TAG TTC CCC GGC CAA GAG AGC GAG-30,
reverse, 50-GAT CAA GCT TAA TGT GGG GGC CCA GAA GG-30), digested with HindIII
and inserted into compatible sites on pGL4.13. For the WT plasmid, the partial Klf4 E1-E2
fragment (179 bp) was obtained by PCR using the primers (forward, 50-CAA AAA GCT TAG
TTC CCC GGC CAA GAG AGC GAG-30, reverse, 50-CGC TGG GCC CTT CTT AAT
GTT TTT GGC ATC TTC CAT AGA CTC GCC AG-30), digested with ApaI, and inserted
into compatible sites on the ΔE2 plasmid. For the ΔE1 plasmid, the WT plasmid was digested
with HindIII, to delete the E1 fragment (660 bp), and self-ligated. Three mutant plasmids
(uORFmut, 1st ATGmut, and 2nd ATGmut) were generated by making a single nucleotide sub-
stitution (A!G) in the WT plasmid. Mutant strands were synthesized by PCR using primers
(forward, reverse): uORFmut, 50-CGT GAC CCG CGC CCG TGG CCG CGC GCA CCC-
30, 50-GGG TGC GCG CGG CCA CGG GCG CGG GTC ACG-30; 1st ATGmut, 50-GGC
CCC CAC ATT AGT GAG GTA GGT GAG ATG-30, 50-CAT CTC ACC TAC CTC ACT
AAT GTG GGG GCC-30; 2nd ATGmut, 50-CCT GGC GAG TCT GAC GTG GAA GAT
GCC AAA AAC-30, 5’-GTT TTT GGC ATC TTC CAC GTC AGA CTC GCC AGG-30,
digested with DpnI and transformed in Escherichia coli. Mutations were confirmed by DNA
sequencing.
Transfection and reporter assay. Reporter plasmids were transfected into NIH3T3 cells
using polyethylenimine (PEI), as previously described [25]. The cells were maintained in
DMEM, containing 10% fetal bovine serum, 50 U/mL penicillin G, and 50 U/mL streptomy-
cin. Cells (6 x 104) were plated on 12-well plates for 12 h, cultured in the absence or presence
of 5 mM DFMO for 12 h, and replaced with 250 μL of Opti-MEMI before transfection. To
form PEI/DNA polyplexes, 3 μg of PEI, 0.5 μg of reporter plasmid, and 0.05 μg of pGL4.75
[hRluc/CMV] (used as an internal control for transfection efficiency) were mixed in 50 μL of
lactate-buffered saline (20 mM sodium lactate (pH 4.0), 150 mM NaCl) and added to each
well. After 8 h, the cells were further cultured with DMEM in the absence or presence of 5 mM
DFMO for 16 h. Luciferase activities were measured using the Dual-luciferase Reporter Assay
System (Promega) with luminometer GL-210A (Microtec Co., Ltd.), and firefly luciferase
activity of reporter plasmid was normalized with Renilla luciferase activity.
In vitro translation. To prepare luc2 fused with the T3 promoter for in vitro transcrip-
tion, T3-luc2 and T3-Klf4-luc2 templates were obtained by PCR using the pGL4.13 and ΔE2
PLOS ONE | https://doi.org/10.1371/journal.pone.0229744 February 26, 2020
4 / 16
Translational regulation of KLF4 by polyamines
plasmids as template and the following primers (forward, reverse); 50-GAA ATA TTA ACC
CTC ACT AAA GGG CTT TTG CAA AAA GCT T-30, 50-TGT TAA CTT GTT TAT
TGC AG-30. The luc2 and Klf4-luc2 mRNAs were synthesized by using mMESSAGE mMA-
CHINE™ T3 Transcription Kit (Invitrogen), according to the manufacturer’s protocol. FM3A
nuclease-treated cell lysates, for in vitro translation in FM3A cell-free systems, was prepared as
described previously [26]. The reaction mixture (10 μL) contained 15 mM Hepes-KOH (pH
7.6), 10 μM hemin, 75 mM potassium acetate, 2 mM dithiothreitol, 0.5 mM glucose 6-phos-
phate, 1 mM ATP, 0.4 mM GTP, 8 mM creatine phosphate, 150 μg/mL of creatine kinase,
10 μM FUT-175, 30 μM of each of the 20 amino acids, 1x protease inhibitor cocktail, 1.5 mM
magnesium acetate, 0.5 μg of synthesized mRNA, different concentrations of polyamines
(spermidine or spermine), and 0.1 mg of FM3A nuclease-treated cell lysate. After incubation
at 37 ˚C for 1 h, the synthesized proteins were detected by measuring luciferase activity, using
the Luciferase Assay System (Promega).
Statistical analysis. The data were analyzed for statistical significance using the two-tailed
student’s t-test. P < 0.05 was considered as statistically significant.
Results
Effect of low concentration-DFMO treatment during BMMC
differentiation
To evaluate the effect of DFMO treatment during BMMC differentiation, bone marrow cells
were cultured in the absence or presence of 5 mM DFMO, as shown in previous reports [17,
27], however, all cells died before maturation. We then modified the culture conditions and
grew the cells in the absence or presence of 0.1 mM DFMO. The cells were subcultured for
four weeks under this condition. As shown in Fig 1A, relative cell numbers during differentia-
tion significantly decreased at a low concentration of DFMO, and no dead cells appeared to
suggest apoptosis. To confirm the correlation between decrease in cell number and cellular
polyamine concentration, the polyamine levels were analyzed at several time points during dif-
ferentiation. Levels of putrescine and spermidine were efficiently reduced during differentia-
tion, and spermine levels increased 2-fold only on day 28 (Fig 2A). A previous study reported
altered ultrastructure of secretary granules in DFMO-treated BMMCs [27]. Therefore, we per-
formed transmission electron microscopy analysis. Control BMMCs showed a distinctive
dense core structure in secretary granules (Fig 1B and 1C), however, DFMO-treated cells were
filled with amorphous structures (Fig 1D and 1E). We assayed degranulation by using two
independent stimuli; the calcium ionophore, A23187 or IgE/anti-IgE. The degranulation rates
of BMMCs were unaffected by DFMO treatment (Fig 1F). These results indicated that the
effect of 0.1 mM DFMO treatment during BMMC differentiation was similar to that of 5 mM
DFMO treatment.
Increase of intracellular histamine in DFMO-treated BMMCs
To determine the effects of DFMO treatment on intracellular histamine levels during BMMC
differentiation, we performed simultaneous analysis of polyamines and histamine. Although
intracellular histamine levels were similar in response to DFMO treatment during differentia-
tion, until day 21, there was a 2.3-fold increase on day 28 (Fig 2B).
We then investigated whether histamine levels were affected by polyamine levels after
BMMC differentiation. As shown in Fig 3, BMMCs in the absence or presence of DFMO, were
placed in four dishes on day 28 and further cultured until day 34. Reduced putrescine and
spermidine levels, in the presence of DFMO, were recovered to control levels by changing the
PLOS ONE | https://doi.org/10.1371/journal.pone.0229744 February 26, 2020
5 / 16
Translational regulation of KLF4 by polyamines
Fig 1. Cell number and morphology of BMMCs in the absence or presence of DFMO. (A) Starting with 0.1 mM
DFMO treatment from day 3, the cells were cultured until day 28. The relative cell number on day 0 is shown as 1.
Standard deviation (s.d.) was calculated; n = 3 replicates. �p < 0.05 compared with control. (B) Control and (D)
DFMO-treated cells on day 28, observed with transmission electron microscopy. (C) and (E) Boxes with white dots
from (B) and (D) in larger magnification, respectively. (F) β-hexosaminidase activity of supernatant and cell pellet after
degranulation stimuli (calcium ionophore A23187 or IgE/anti-IgE) was given on day 28. Degranulation rate (%) ± s.d.
is shown as [supernatant]/[supernatant + cell pellet], n = 3 replicates.
https://doi.org/10.1371/journal.pone.0229744.g001
culture condition (without DFMO) on day 34, and spermine levels were slightly decreased.
Under this condition, histamine levels were still high. DFMO treatment from day 28 did not
affect histamine levels. These results indicated that lower levels of putrescine and spermidine
may be required to store higher levels of histamine during BMMC differentiation, but not
after their differentiation.
Histidine decarboxylase activity is regulated by KLF4 during BMMC
differentiation
To determine whether higher histamine levels, produced due to DFMO treatment, correlated
with HDC activity, cell lysates were prepared at different time points during BMMC differenti-
ation and used for HDC assay. As shown in Fig 4A, activity of DFMO-treated cells between
day 14 and 21 were 2-fold higher than control cells. We investigated the expression levels of
Hdc mRNA on day 7, 14, and 21 by qRT-PCR. Hdc mRNA levels of DFMO-treated cells
between day 14 and 21 were significantly higher, compared to control cells (Fig 4B). It has
been previously reported that the transcription factor, KLF4, suppresses expression of Hdc
[18]. We hypothesized that the decrease of KLF4 level by DFMO treatment increased Hdc
PLOS ONE | https://doi.org/10.1371/journal.pone.0229744 February 26, 2020
6 / 16
Translational regulation of KLF4 by polyamines
Fig 2. Polyamine and histamine levels during BMMC differentiation in the absence or presence of DFMO. During BMMC differentiation,
cells on day 7, 14, 21, and 28 were used for analysis of intracellular polyamine (A) and histamine (B) content by HPLC. Mean ± s.d. is shown,
n = 3 replicates. �p < 0.05, compared with control.
https://doi.org/10.1371/journal.pone.0229744.g002
mRNA level. To confirm this hypothesis, we determined the amount of KLF4 (483 amino
acids) during differentiation by western blotting. As shown in Fig 4C, S1 and S2 Figs, the
amount of KLF4 in DFMO-treated cells, on day 14 and 21, was lower compared to control
cells. qRT-PCR was further performed to determine Klf4 mRNA levels on day 7, 14, and 21;
Klf4 mRNA levels did not decrease in response to DFMO treatment (Fig 4D). The in vivo bind-
ing of KLF4 on the Hdc promoter was examined by ChIP assay; a fragment of the proximal
promoter region was amplified using anti-KLF4 precipitated-DNA (S3 Fig). Moreover, to
exclude non-specific effects of DFMO, we carried out a recovery experiment by adding
100 μM putrescine. During differentiation, cell number and intracellular amines recovered to
levels similar to control cells (S4 Fig). Importantly, Hdc and Klf4 mRNA and KLF4 levels also
recovered by the addition of putrescine (Fig 4B–4D). These results suggested that Klf4 expres-
sion is translationally regulated by polyamines, and decreased KLF4 due to reduced poly-
amines resulted in the upregulation of Hdc expression.
Involvement of the 50-UTR of Klf4 mRNA in the translational regulation by
polyamines
We have named the genes encoding proteins whose syntheses are enhanced by polyamines at
the translational level as the ‘polyamine modulon’ [7, 28, 29]. They have a characteristic nucle-
otide sequence and structure at the 50-UTR end of the mRNA. Therefore, we investigated the
PLOS ONE | https://doi.org/10.1371/journal.pone.0229744 February 26, 2020
7 / 16
Translational regulation of KLF4 by polyamines
Fig 3. Effect of DFMO on the level of polyamines and histamine after BMMC differentiation. After 28 d of BMMC differentiation, in the absence
or presence of DFMO, the cells were continuously cultured for 6 d under four different conditions, as shown in the upper panel. At each time point on
day 28, 31, and 34, the level of polyamines and histamine was analyzed by HPLC. Mean ± s.d. is shown, n = 3 replicates.
https://doi.org/10.1371/journal.pone.0229744.g003
nucleotide sequence of mouse Klf4 (Gene ID: 16,600) in NCBI (https://www.ncbi.nlm.nih.gov/
gene/). Klf4 consists of 5 exons and 4 introns. There were two candidates for the initiation
codon ATG in Klf4; the 1st and 2nd ATGs were located on exon1 (E1) and exon2 (E2), respec-
tively (Fig 5A). The length of the 50-UTR of Klf4 mRNA is 604 nucleotides (nt) if translation
starts from the 1st ATG. The GC content, from the transcriptional start site to 140 nt, on E1
was 81.4% (114/140). To investigate whether the 50-UTR of Klf4 mRNA is involved in transla-
tional regulation by polyamines, reporter plasmids were prepared and used for transfection
into NIH3T3 cells (Fig 5A). In comparison with the pGL4.13 vector, relative luciferase activity
of the ΔE1 plasmid was similar, however, the activities of WT and ΔE2 plasmids, containing
the long E1 fragment, were decreased (left panel, Fig 5B). We tested the effect of DFMO
PLOS ONE | https://doi.org/10.1371/journal.pone.0229744 February 26, 2020
8 / 16
Translational regulation of KLF4 by polyamines
Fig 4. Hdc and Klf4 expression during BMMC differentiation in the absence or presence of DFMO. (A) During BMMC differentiation,
cells on day 7, 14, 21, and 28 were collected and used to prepare cell lysates. HDC assays were performed after dialysis of the cell lysate.
Mean ± s.d. is shown, n = 3 replicates. �p < 0.05 compared with control. (B) Total RNA extracted from cells on day 7, 14, and 21 was
isolated and used for qRT-PCR. Relative expression of Hdc mRNA was normalized to Gapdh and control expression is shown as 1.
Mean ± s.d. is shown, n = 3 replicates. �p < 0.05 compared with control. (C) During BMMC differentiation, cells on day 7, 14, 21, and 28
were collected and analyzed for KLF4 and β-actin by western blotting. Lanes (C, D, and DP) represent control, DFMO, and DFMO + PUT,
respectively. (D) Total RNA from cells on day 7, 14, and 21 was isolated and used for qRT-PCR. Relative expression of Klf4 mRNA was
normalized to Gapdh; control expression is shown as 1. Mean ± s.d. is shown, n = 3 replicates. �p < 0.05, ��p < 0.01, and ���p < 0.001,
compared with control.
https://doi.org/10.1371/journal.pone.0229744.g004
treatment; values are presented as the ratio of relative luciferase activity of DFMO (+/−). The
ratio of WT and ΔE2 plasmids was significantly reduced compared with the pGL4.13 vector,
and that of the ΔE1 plasmid was equal (right panel, Fig 5B). As shown in Fig 5A, an upstream
open reading frame (uORF, 180 nt) is present on E1 of Klf4. It is known that uORFs may nega-
tively regulate translation of the downstream ORFs; therefore, uORF may be involved in trans-
lational regulation by polyamines. The uORFmut plasmid was prepared by a single nucleotide
substitution (A!G) in the WT plasmid (uORFmut, Fig 5A). As shown in Fig 5C, relative lucif-
erase activity of the uORFmut plasmid was 2-fold higher than the WT plasmid (left panel, Fig
5C); however, DFMO treatment resulted in no difference between WT and uORFmut (right
panel, Fig 5C). Furthermore, we determined initiation codon usage efficiency by using 1st
ATGmut and 2nd ATGmut plasmids (Fig 5A). The relative luciferase activity of 1st ATGmut
PLOS ONE | https://doi.org/10.1371/journal.pone.0229744 February 26, 2020
9 / 16
Translational regulation of KLF4 by polyamines
Fig 5. Effect of polyamines on luciferase activity from normal and modified Klf4-luc2 reporter plasmids in NIH3T3 cells. (A) Schematic illustration of
mouse Klf4 (E1-E3) and various Klf4-luc2 reporter plasmids are shown. Nucleotide sequences for exons and introns are represented as uppercase and
lowercase, respectively. The two initiation codons (ATG) on Klf4 are underlined as 1st and 2nd. The dotted line on Klf4-luc2 reporter plasmids means that the
region was deleted from WT plasmid. The site of mutated initiation codon, by site-directed mutagenesis, is indicated by an arrowhead (A!G). The green
box in the plasmid illustration means the case of translation from 1st initiation codon on E1. The yellow box shows luc2 ORF. (B) and (C); NIH3T3 cells co-
transfected with the reporter plasmids and pGL4.75[hRluc/CMV] in the absence or presence of 5 mM DFMO, respectively. The graph on the left shows
relative luciferase activities measured as the ratio of firefly and Renilla luciferase activity. The graph on the right shows data calculated as the ratio of relative
luciferase activity of DFMO (+/−). The values of vector are shown as 1. Mean ± s.d. is shown, n = 3 replicates. ��p < 0.01 compared with vector. n.d., not
determined.
https://doi.org/10.1371/journal.pone.0229744.g005
PLOS ONE | https://doi.org/10.1371/journal.pone.0229744 February 26, 2020
10 / 16
Translational regulation of KLF4 by polyamines
Fig 6. Effect of polyamines on Klf4-luc2 translation in an FM3A cell-free system. In vitro translation of (A) luc2 and (B)
Klf4-luc2 mRNAs were performed as described in MATERIALS AND METHODS. The level of luciferase synthesized from luc2
was measured using Luciferase Assay System as relative luminescence unit (RLU). Mean ± s.d. is shown, n = 3 replicates.
https://doi.org/10.1371/journal.pone.0229744.g006
plasmid was equal to the WT plasmid; however, the activity of the 2nd ATGmut plasmid almost
ceased to be visible, regardless of DFMO treatment (Fig 5C). This result suggests that the 1st
ATG does not operate as an initiation codon for KLF4. Taken together, we conclude that the
5’-UTR of Klf4 mRNA is involved in translational regulation by polyamines, but the presence
of uORF and two initiation codons is unrelated.
Finally, the effects of spermidine and spermine on translation of Klf4-luc2 mRNA were
examined using in vitro synthesized mRNAs in an FM3A cell-free system. The Klf4-luc2
mRNA was synthesized from T3-Klf4-luc2 of the ΔE2 plasmid (see Fig 5A) fused with a T3
promoter for in vitro transcription. Similarly, the luc2 mRNA was synthesized from T3-luc2 of
the pGL4.13 vector as control. The translation of luc2 mRNA occurred in the presence of 0.1
mM spermidine; however, translation of Klf4-luc2 mRNA required a 2-fold higher spermidine
concentration (circles and solid line, Fig 6A and 6B, respectively). On the other hand, the opti-
mal concentration of spermine required for the stimulation of translation was 2-fold higher
for Klf4-luc2 mRNA (0.03 to 0.06 mM) compared to that for luc2 mRNA (0.03 mM) (triangle
and dashed line, Fig 6A and 6B, respectively). These results indicate that the 50-UTR of Klf4
mRNA influenced the sensitivity of polyamine concentration for translation.
Discussion
Mast cells, derived from hematopoietic stem cells, are major players in IgE-mediated allergic
reactions, and contain abundant chemical mediators such as histamine. Polyamines, which are
essential for cell proliferation and differentiation, are abundant in mast cells. In this study, we
demonstrated that intracellular histamine levels increased with an increase in Hdc expression
and a decrease in KLF4 level during BMMC differentiation. Therefore, we propose a model
that suggests that translational regulation of KLF4 by DFMO-induced polyamine depletion is
required for the suppression of histamine synthesis during BMMC differentiation (Fig 7).
In 2009, Garcı´a-Faroldi et al reported that DFMO-induced Hdc mRNA up-regulation was
observed in early bone marrow cell cultures [17]. The reason that we could not reproduce
BMMC differentiation under their DFMO treatment condition is unknown. However, low
DFMO concentration treatment showed a similar effect, such as morphological change, on
PLOS ONE | https://doi.org/10.1371/journal.pone.0229744 February 26, 2020
11 / 16
Translational regulation of KLF4 by polyamines
Fig 7. Model of suppression of histamine synthesis by polyamines during BMMC differentiation. The regulation of KLF4 synthesis
depends on cellular polyamines at the translational level. Upon polyamine depletion, an upregulation of Hdc expression is caused due to the
disappearance of KLF4. Higher levels of histamine is synthesized during differentiation.
https://doi.org/10.1371/journal.pone.0229744.g007
secretary granules and degranulation response (Fig 1B–1F). Our data also indicated the pres-
ence of an inverse relationship between histamine and polyamine metabolism. A point of dif-
ference was that the intracellular histamine levels of our BMMCs increased more than 2-fold
in response to DFMO treatment (Figs 2B and 3 and S1B Fig), although intracellular histamine
levels of their BMMCs did not increase [17, 27]. This may be due to the difference in intracel-
lular spermidine levels. It is known that only spermidine is required as a substrate for post-
translational modification of the eukaryotic translation initiation factor 5A (eIF5A). This mod-
ification, hypusination, is a two-step enzymatic reaction that only occurs in eIF5A, which is
essential for cell proliferation and life [30–32]. A previous study indicated that a dramatic
decrease of intracellular spermidine (below 5 nmol/mg protein) caused inhibition of hypusina-
tion [33]. The spermidine concentration was halved in our DFMO-treated BMMCs on day 21,
however the level in their BMMCs was significantly reduced to 1.87 nmol/mg protein [17].
Therefore, inhibition of eIF5A may affect protein synthesis involved in histamine storage in
granules. Recently, Carlos Acosta-Andrade et al reported that slowly decreasing intracellular
polyamine levels in Azin2 hypomorphic BMMCs showed an increase in histamine levels and
HDC activity [34]. Our hypothesis is supported by this report.
An abnormal morphology of secretory granules in BMMCs was observed in response to
treatment with low concentrations of DFMO. Serglycin (SG) exists as a core protein of proteo-
glycans in mast cell granules [12]. As seen in transmission electron micrographs of BMMCs of
SG−/− mice, dense core formation was defective in SG−/− mast cell granules [35]. This was simi-
lar to the morphological changes observed in granules of DFMO-treated BMMC. It is known
that the SG in BMMCs mainly attaches chondroitin sulfate. We suggest that morphologic
abnormality of secretory granules by DFMO treatment may involve a decrease in SG and
chondroitin sulfate due to polyamine depletion.
PLOS ONE | https://doi.org/10.1371/journal.pone.0229744 February 26, 2020
12 / 16
Translational regulation of KLF4 by polyamines
Here we showed that polyamine depletion maintains high HDC activity at the transcrip-
tional level and increases of intracellular histamine during BMMC differentiation (Figs 2B, 4A
and 4B). Previously, KLF4 has been reported to suppresses expression of Hdc in a gastric can-
cer cell [18]; we found that it is also involved in differentiation. The level of KLF4 decreased,
without any change in mRNA levels, in response to polyamine depletion (Fig 4C and 4D). We
classified the Klf4 as a member of the ‘polyamine modulon’. Klf4 has a long GC-rich 50-UTR,
one of the characteristics of genes regulated by polyamines at the translational level. Several
mechanisms of polyamine mediated stimulation of translation have been reported, such as
ribosome shunting on the 50-UTR [8], distant positioning of the CR sequence from the AUG
[29], frameshifting at the termination codon (UGA) on the ORF [36], and miRNA mediated
suppression of initiation codon recognition [37]. In this study, to clarify the mechanism of
translational stimulation of KLF4 by polyamines, we used luciferase reporter plasmids contain-
ing several 50-UTR versions of Klf4 mRNA. First, Klf4 had two initiation codons in the same
frame, however, the long form from 1st ATG on E1 was untranslated, and only the short form
from 2nd ATG on E2 was translated (Fig 5C). It has been reported that KLF4 N-terminal vari-
ance modulates reprogramming efficiency of induced pluripotent stem (iPS) cells [38]. These
differences between long and short forms may depend on different tissue and cell types.
Next, since the 50-UTR on Klf4 mRNA significantly inhibited luciferase activity in response to
polyamine depletion, we focused on the uORF of Klf4 mRNA. Upstream ORFs abound in
mammalian mRNAs, their uORFs mostly suppress the translational efficiency of the down-
stream ORF [39]. Amd1 mRNA, which encodes the polyamine biosynthesis rate-limiting
enzyme AdoMetDC, has a unique uORF encoding MAGDIS, and polyamine depletion causes
uORF-dependent translation stimulation [40]. Although the uORF of Klf4 mRNA is 59 codons
longer than MAGDIS, analysis of an uORF mutant of the reporter plasmid uORFmut did not
detect translation stimulation by polyamines (Fig 5C). Recently, it has been reported that poly-
amines stimulate CHSY1 synthesis through the unfolding of the RNA G-quadruplex at the
50-untrasnslated region [41]. A similar 12-nucleotide guanine quartet (CGG)4 motif that can
form RNA G-quadruplex structures is also present in the 50-UTR of Klf4 mRNA. Therefore,
we hypothesize that this (CGG)4 motif plays a key role in translational regulation by poly-
amines. However, enrichment of the (CGG)4 motif in the 50-UTRs of eIF4A-dependent
mRNAs has been reported to be due to the formation of stable hairpin structures rather than
G-quadruplexes [42, 43]. It is controversial whether RNA G-quadruplex structures are actually
formed in cells.
Although Klf4 is important as one of the Yamanaka factors of iPS cells [44], it also plays a
role as a zinc finger type transcription factor that is involved in various processes of cell prolif-
eration and differentiation [45]. Our finding, that translational regulation of KLF4 is mediated
by polyamines, may represent a common mechanism regulating various processes occurring
in Klf4-expressing cells.
Supporting information
S1 Fig. Original blot data_KLF4.
(PDF)
S2 Fig. Original blot data_Actin.
(PDF)
S3 Fig. Chromatin immunoprecipitation assay during BMMC differentiation in the
absence or presence of DFMO.
(PDF)
PLOS ONE | https://doi.org/10.1371/journal.pone.0229744 February 26, 2020
13 / 16
Translational regulation of KLF4 by polyamines
S4 Fig. Cell number and intracellular amines in BMMCs in the absence or presence of
DFMO and putrescine.
(PDF)
Acknowledgments
We thank Dr. A. J. Michael for his help in preparing this manuscript. We also thank Aventis
Pharma and Torii Pharmaceutical Co. for providing DFMO and FUT-175, respectively.
Author Contributions
Conceptualization: Kazuhiro Nishimura.
Data curation: Kazuhiro Nishimura.
Formal analysis: Kazuhiro Nishimura.
Funding acquisition: Kazuhiro Nishimura.
Investigation: Kazuhiro Nishimura, Moemi Okamoto, Rina Shibue, Toshio Mizuta, Toru Shi-
bayama, Tetsuhiko Yoshino, Teruki Murakami, Masashi Yamaguchi, Satoshi Tanaka,
Toshihiko Toida.
Methodology: Kazuhiro Nishimura, Satoshi Tanaka.
Project administration: Kazuhiro Nishimura.
Resources: Kazuhiro Nishimura.
Supervision: Kazuhiro Nishimura, Toshihiko Toida, Kazuei Igarashi.
Validation: Kazuhiro Nishimura.
Visualization: Kazuhiro Nishimura.
Writing – original draft: Kazuhiro Nishimura.
Writing – review & editing: Kazuei Igarashi.
References
1.
Tabor CW, Tabor H. Polyamines. Annu Rev Biochem. 1984; 53: 749–90. https://doi.org/10.1146/
annurev.bi.53.070184.003533 PMID: 6206782
2. Cohen S. A Guide to the Polyamines. 1998. Oxford: Oxford University Press
3. Pegg AE. Mammalian polyamine metabolism and function. IUBMB Life. 2009; 61(9): 880–94. https://
doi.org/10.1002/iub.230 PMID: 19603518
4. Pendeville H, Carpino N, Marine JC, Takahashi Y, Muller M, Martial JA, et al. The ornithine decarboxyl-
ase gene is essential for cell survival during early murine development. Mol Cell Biol. 2001; 21(19):
6549–58. https://doi.org/10.1128/MCB.21.19.6549-6558.2001 PMID: 11533243
5. Nishimura K, Nakatsu F, Kashiwagi K, Ohno H, Saito T, Igarashi K. Essential role of S-adenosylmethio-
nine decarboxylase in mouse embryonic development. Genes Cells. 2002; 7(1): 41–7. https://doi.org/
10.1046/j.1356-9597.2001.00494.x PMID: 11856372
6. Watanabe S, Kusama-Eguchi K, Kobayashi H, Igarashi K. Estimation of polyamine binding to macro-
molecules and ATP in bovine lymphocytes and rat liver. J Biol Chem. 1991; 266(31): 20803–9. PMID:
1718969
7.
Igarashi K, Kashiwagi K. Modulation of protein synthesis by polyamines. IUBMB Life. 2015; 67(3):
160–9. https://doi.org/10.1002/iub.1363 PMID: 25906835
8. Nishimura K, Okudaira H, Ochiai E, Higashi K, Kaneko M, Ishii I, et al. Identification of proteins whose
synthesis is preferentially enhanced by polyamines at the level of translation in mammalian cells. Int J
Biochem Cell Biol. 2009; 41(11): 2251–61. https://doi.org/10.1016/j.biocel.2009.04.021 PMID:
19427401
PLOS ONE | https://doi.org/10.1371/journal.pone.0229744 February 26, 2020
14 / 16
Translational regulation of KLF4 by polyamines
9. Metcalfe DD, Baram D, Mekori YA. Mast cells. Physiol Rev. 1997; 77(4): 1033–79. https://doi.org/10.
1152/physrev.1997.77.4.1033 PMID: 9354811
10. Galli SJ, Nakae S, Tsai M. Mast cells in the development of adaptive immune responses. Nat Immunol.
2005; 6(2): 135–42. https://doi.org/10.1038/ni1158 PMID: 15662442
11. Ro¨nnberg E, Melo FR, Pejler G. Mast Cell Proteoglycans. J Histochem Cytochem. 2012; 60(12): 950–
62. https://doi.org/10.1369/0022155412458927 PMID: 22899859
12. Stevens RL, Adachi R. Protease–proteoglycan complexes of mouse and human mast cells and impor-
tance of their β-tryptase–heparin complexes in inflammation and innate immunity. Immunol Rev. 2007;
217: 155–67. https://doi.org/10.1111/j.1600-065X.2007.00525.x PMID: 17498058
13. Åbrink M, Grujic M, Pejler G. Serglycin is essential for maturation of mast cell secretory granule. J Biol
Chem. 2004; 279(39): 40897–905. https://doi.org/10.1074/jbc.M405856200 PMID: 15231821
14. Ringvall M, Ro¨ nnberg E, Wernersson S, Duelli A, Henningsson F, Åbrink M, et al. Serotonin and hista-
mine storage in mast cell secretory granules is dependent on serglycin proteoglycan. J Allergy Clin
Immunol. 2008; 121(4): 1020–6. https://doi.org/10.1016/j.jaci.2007.11.031 PMID: 18234316
15. Hommma R, Mase A, Toida T, Kashiwagi K, Igarashi K. Modulation of blood coagulation and fibrinolysis
by polyamines in the presence of glycosaminoglycans. Int J Biochem Cell Biol. 2005; 37(9): 1911–20.
https://doi.org/10.1016/j.biocel.2005.04.014 PMID: 15936241
16. Belting M, Borsig L, Fuster MM, Brown JR, Persson L, Fransson LA, et al. Tumor attenuation by com-
bined heparan sulfate and polyamine depletion. Proc Natl Acad Sci U S A. 2002; 99(1): 371–6. https://
doi.org/10.1073/pnas.012346499 PMID: 11752393
17. Garcı´a-Faroldi G, Correa-Fiz F, Abrighach H, Berdasco M, Fraga MF, Esteller M, et al. Polyamines
affect histamine synthesis during early stages of IL-3-induced bone marrow cell differentiation. J Cell
Biochem. 2009; 108(1): 261–71. https://doi.org/10.1002/jcb.22246 PMID: 19562674
18. Ai W, Liu Y, Langlois M, Wang TC. Kruppel-like factor 4 (KLF4) represses histidine decarboxylase gene
expression through an upstream Sp1 site and downstream gastrin responsive elements. J Biol Chem.
2004; 279(10): 8684–93. https://doi.org/10.1074/jbc.M308278200 PMID: 14670968
19.
Takano H, Nakazawa S, Okuno Y, Shirata N, Tsuchiya S, Kainoh T, et al. Establishment of the culture
model system that reflects the process of terminal differentiation of connective tissue-type mast cells.
FEBS Lett. 2008; 582(10): 1444–50. https://doi.org/10.1016/j.febslet.2008.03.033 PMID: 18381075
20. Yamaguchi M, Namiki Y, Okada H, Uematsu K, Tame A, Maruyama T, et al. Improved preservation of
fine structure of deep-sea microorganisms by freeze-substitution after glutaraldehyde fixation. J Elec-
tron Microsc (Tokyo). 2011; 60(4): 283–7. https://doi.org/10.1093/jmicro/dfr032 PMID: 21571752
21. Yamaguchi M, Okada H, Namiki Y. Smart specimen preparation for freeze substitution and serial ultra-
thin sectioning of yeast cells. J Electron Microsc (Tokyo). 2009; 58(4):61–6. https://doi.org/10.1093/
jmicro/dfp013 PMID: 19289851
22. Nishimura K, Shiina R, Kashiwagi K, Igarashi K. Decrease in polyamines with aging and their ingestion
from food and drink. J Biochem. 2006; 139(1): 81–90. https://doi.org/10.1093/jb/mvj003 PMID:
16428322
23.
Furuta K, Nakayama K, Sugimoto Y, Ichikawa A, Tanaka S. Activation of histidine decarboxylase
through post-translational cleavage by caspase-9 in a mouse mastocytoma P-815. J Biol Chem. 2007;
282(18): 13438–46. https://doi.org/10.1074/jbc.M609943200 PMID: 17360717
24. Yamashita T, Nishimura K, Saiki R, Okudaira H, Tome M, Higashi K, et al. Role of polyamines at the
G1/S boundary and G2/M phase of the cell cycle. Int J Biochem Cell Biol. 2013; 45(6): 1042–50. https://
doi.org/10.1016/j.biocel.2013.02.021 PMID: 23500523
25.
Fukumoto Y, Obata Y, Ishibashi K, Tamura N, Kikuchi I, Aoyama K, et al. Cost-effective gene transfec-
tion by DNA compaction at pH 4.0 using acidified, long shelf-life polyethylenimine. Cytotechnology.
2010; 62(1): 73–82. https://doi.org/10.1007/s10616-010-9259-z PMID: 20309632
26. Nishimura K, Sakuma A, Yamashita T, Hirokawa G, Imataka H, Kashiwagi K, et al. Minor contribution of
an internal ribosome entry site in the 5’-UTR of ornithine decarboxylase mRNA on its translation. Bio-
chem Biophys Res Commun. 2007; 364(1): 124–30. https://doi.org/10.1016/j.bbrc.2007.09.112 PMID:
17927956
27. Garcı´a-Faroldi G, Rodrı´guez CE, Urdiales JL, Pe´ rez-Pomares JM, Da´ vila JC, Pejler G, et al. Poly-
amines are present in mast cell secretory granules and are important for granule homeostasis. PLoS
One. 2010; 5(11): e15071. https://doi.org/10.1371/journal.pone.0015071 PMID: 21151498
28.
29.
Igarashi K, Kashiwagi K. Characterization of genes for polyamine modulon. Methods Mol Biol. 2011;
720:51–65. https://doi.org/10.1007/978-1-61779-034-8_3 PMID: 21318866
Terui Y, Sakamoto A, Yoshida T, Kasahara T, Tomitori H, Higashi K, et al. Polyamine stimulation of
eEF1A synthesis based on the unusual position of a complementary sequence to 18S rRNA in eEF1A
PLOS ONE | https://doi.org/10.1371/journal.pone.0229744 February 26, 2020
15 / 16
Translational regulation of KLF4 by polyamines
mRNA. Amino Acids. 2015; 47(2): 345–56. https://doi.org/10.1007/s00726-014-1867-z PMID:
25425115
30. Park MH, Nishimura K, Zanelli CF, Valentini SR. Functional significance of eIF5A and its hypusine mod-
ification in eukaryotes. Amino Acids. 2010; 38(2): 491–500. https://doi.org/10.1007/s00726-009-0408-
7 PMID: 19997760
31. Nishimura K, Lee SB, Park JH, Park MH. Essential role of eIF5A-1 and deoxyhypusine synthase in
mouse embryonic development. Amino Acids. 2012; 42(2–3): 703–10. https://doi.org/10.1007/s00726-
011-0986-z PMID: 21850436
32. Sievert H, Pa¨llmann N, Miller KK, Hermans-Borgmeyer I, Venz S, Sendoel A, et al. A novel mouse
model for inhibition of DOHH-mediated hypusine modification reveals a crucial function in embryonic
development, proliferation and oncogenic transformation. Dis Model Mech. 2014; 7(8): 963–76. https://
doi.org/10.1242/dmm.014449 PMID: 24832488
33. Nishimura K, Murozumi K, Shirahata A, Park MH, Kashiwagi K, Igarashi K. Independent roles of eIF5A
and polyamines in cell proliferation. Biochem J. 2005; 385(Pt 3): 779–85. https://doi.org/10.1042/
BJ20041477 PMID: 15377278
34. Acosta-Andrade C, Lambertos A, Urdiales JL, Sa´nchez-Jime´ nez F, Peñafiel R, Fajardo I. A novel role
for antizyme inhibitor 2 as a regulator of serotonin and histamine biosynthesis and content in mouse
mast cells. Amino Acids. 2016; 48(10): 2411–21. https://doi.org/10.1007/s00726-016-2230-3 PMID:
27084713
35. Henningsson F, Hergeth S, Cortelius R, Åbrink M, Pejler G. A role for serglycin proteoglycan in granular
retention and processing of mast cell secretory granule components. FEBS J. 2006; 273(21): 4901–12.
https://doi.org/10.1111/j.1742-4658.2006.05489.x PMID: 17010166
36. Matsufuji S, Matsufuji T, Miyazaki Y, Murakami Y, Atkins JF, Gesteland RF, et al. Autoregulatory frame-
shifting in decoding mammalian ornithine decarboxylase antizyme. Cell. 1995; 80(1): 51–60. https://
doi.org/10.1016/0092-8674(95)90450-6 PMID: 7813017
37.
Imamura M, Higashi K, Yamaguchi K, Asakura K, Furihata T, Terui Y, et al. Polyamines release the let-
7b-mediated suppression of initiation codon recognition during the protein synthesis of EXT2. Sci Rep.
2016; 6: 33549. https://doi.org/10.1038/srep33549 PMID: 27650265
38. Kim SI, Oceguera-Yanez F, Hirohata R, Linker S, Okita K, Yamada Y, et al. KLF4 N-terminal variance
modulates induced reprogramming to pluripotency. Stem Cell Reports. 2015; 4(4): 727–43. https://doi.
org/10.1016/j.stemcr.2015.02.004 PMID: 25772473
39. Hinnebusch AG, Ivanov IP, Sonenberg N. Translational control by 5’-untranslated regions of eukaryotic
mRNAs. Science. 2016; 352(6292): 1413–6. https://doi.org/10.1126/science.aad9868 PMID:
27313038
40. Raney A, Law GL, Mize GJ, Morris DR. Regulated translation termination at the upstream open reading
frame in S-adenosylmethionine decarboxylase mRNA. J Biol Chem. 2002; 277(8): 5988–94. https://
doi.org/10.1074/jbc.M108375200 PMID: 11741992
41. Yamaguchi K, Asakura K, Imamura M, Kawai G, Sakamoto T, Furihata T, et al. Polyamines stimulate
the CHSY1 synthesis through the unfolding of the RNA G-quadruplex at the 5’-untraslated region. Bio-
chem J. 2018; 475(23): 3797–3812. https://doi.org/10.1042/BCJ20180672 PMID: 30401686
42. Guo JU, Bartel DP. RNA G-quadruplexes are globally unfolded in eukaryotic cells and depleted in bac-
teria. Science. 2016; 353(6306). pii: aaf5371. https://doi.org/10.1126/science.aaf5371 PMID:
27708011
43. Waldron JA, Raza F, Le Quesne J. eIF4A alleviates the translational repression mediated by classical
secondary structures more than by G-quadruplexes. Nucleic Acids Res. 2018; 46(6): 3075–3087.
https://doi.org/10.1093/nar/gky108 PMID: 29471358
44.
Takahashi K, Yamanaka S. Induction of pluripotent stem cells from mouse embryonic and adult fibro-
blast cultures by defined factors. Cell. 2006; 126(4): 663–76. https://doi.org/10.1016/j.cell.2006.07.024
PMID: 16904174
45. Ghaleb AM, Yang VW. Kru¨ppel-like factor 4 (KLF4): What we currently know. Gene. 2017; 611: 27–37.
https://doi.org/10.1016/j.gene.2017.02.025 PMID: 28237823
PLOS ONE | https://doi.org/10.1371/journal.pone.0229744 February 26, 2020
16 / 16
|
10.1371_journal.pone.0242591 | RESEARCH ARTICLE
Complete chloroplast genomes of three
important species, Abelmoschus moschatus, A.
manihot and A. sagittifolius: Genome
structures, mutational hotspots, comparative
and phylogenetic analysis in Malvaceae
Jie Li☯, Guang-ying Ye☯, Hai-lin Liu, Zai-hua WangID*
Guangdong Provincial Key Lab of Ornamental Plant Germplasm Innovation and Utilization, Environmental
Horticulture Research Institute, Guangdong Academy of Agricultural Sciences, Guangzhou, China
☯ These authors contributed equally to this work.
* wang211210@163.com
Abstract
Abelmoschus is an economically and phylogenetically valuable genus in the family Malva-
ceae. Owing to coexistence of wild and cultivated form and interspecific hybridization, this
genus is controversial in systematics and taxonomy and requires detailed investigation.
Here, we present whole chloroplast genome sequences and annotation of three important
species: A. moschatus, A. manihot and A. sagittifolius, and compared with A. esculentus
published previously. These chloroplast genome sequences ranged from 163121 bp to
163453 bp in length and contained 132 genes with 87 protein-coding genes, 37 transfer
RNA and 8 ribosomal RNA genes. Comparative analyses revealed that amino acid fre-
quency and codon usage had similarity among four species, while the number of repeat
sequences in A. esculentus were much lower than other three species. Six categories of
simple sequence repeats (SSRs) were detected, but A. moschatus and A. manihot did not
contain hexanucleotide SSRs. Single nucleotide polymorphisms (SNPs) of A/T, T/A and C/
T were the largest number type, and the ratio of transition to transversion was from 0.37 to
0.55. Abelmoschus species showed relatively independent inverted-repeats (IR) boundary
traits with different boundary genes compared with the other related Malvaceae species.
The intergenic spacer regions had more polymorphic than protein-coding regions and intro-
nic regions, and thirty mutational hotpots (�200 bp) were identified in Abelmoschus, such
as start-psbA, atpB-rbcL, petD-exon2-rpoA, clpP-intron1 and clpP-exon2.These mutational
hotpots could be used as polymorphic markers to resolve taxonomic discrepancies and bio-
geographical origin in genus Abelmoschus. Moreover, phylogenetic analysis of 33 Malva-
ceae species indicated that they were well divided into six subfamilies, and genus
Abelmoschus was a well-supported clade within genus Hibiscus.
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: Li J, Ye G-y, Liu H-l, Wang Z-h (2020)
Complete chloroplast genomes of three important
species, Abelmoschus moschatus, A. manihot and
A. sagittifolius: Genome structures, mutational
hotspots, comparative and phylogenetic analysis in
Malvaceae. PLoS ONE 15(11): e0242591. https://
doi.org/10.1371/journal.pone.0242591
Editor: Tzen-Yuh Chiang, National Cheng Kung
University, TAIWAN
Received: September 22, 2020
Accepted: November 6, 2020
Published: November 25, 2020
Copyright: © 2020 Li et al. This is an open access
article distributed under the terms of the Creative
Commons Attribution License, which permits
unrestricted use, distribution, and reproduction in
any medium, provided the original author and
source are credited.
Data Availability Statement: The complete
chloroplast genomes have been submitted to NCBI
under GenBank accession numbers of MT890968,
MT898000 and MT898001. All other relevant data
are within the paper and its Supporting Information
files.
Funding: This work was supported by the Science
and Technology Plan Project of Guangzhou (No.
201904010264), the special fund for Scientific
Innovation Strategy Construction of High Level
PLOS ONE | https://doi.org/10.1371/journal.pone.0242591 November 25, 2020
1 / 19
PLOS ONE Academy of Agriculture Science (No. R2019PY-
QY003), Emerging Discipline Team Project of
Guangdong Academy of Agricultural Sciences (No.
201806xx) and Provincial Rural Revitalization
Project of Guangdong Province (2020KJ148).
Competing interests: The authors have declared
that no competing interests exist.
Complete chloroplast genomes of Abelmoschus moschatus, A. manihot and A. sagittifolius
Introduction
Family Malvaceae consists of 244 genera and over 4200 species, and most of them are widely
distributed in tropics and temperate regions [1]. According to the diverse morphological char-
acteristics, this family could be divided into nine subfamilies, including Sterculioideae, Tilioi-
deae, Malvoideae, Helicteroideae, Grewioideae, Dombeyoideae, Byttnerioideae,
Brownlowioideae and Bombacoideae [2]. Abelmoschus is one of important genera in subfamily
Malvoideae of family Malvaceae. This genus was previously placed within Hibiscus, and subse-
quently isolated by taxonomists due to genetic differences [3]. As currently defined, genus
Abelmoschus contains 11 species, 4 subspecies and 5 varieties [4], and displays a variable habit,
from annual to perennial, herbs to shrubs, and is distributed in Asia, Australia and southwest-
ern Africa [5]. Most members of this genus are economically important plants, and used in
agriculture, food and medicines. A. esculentus (okra) and A. caillei are widely cultivated as veg-
etables due to their tender pods [6–8]. A. manihot is a popular green leafy vegetable and its
flowers have been applied in clinical treatment of burns, chronic kidney disease and oral ulcers
owing to the flavonoids [9, 10]. A. moschatus, as an aromatic plant, could be suitable for medi-
cal or food uses to improve insulin sensitivity [11]. A. sagittifolius also has a long history of
medicinal usage, and cadinane sesquiterpenoid glucoside extracted from the stem tubers
exhibited antitumor activity [12]. Moreover, antioxidant, antimicrobial, wound healing, anti-
inflammatory and immunomodulatory activities were also found in Abelmoschus species [13–
16]. Seed oil and levels of oleic acid have also been reported in Abelmoschus [3].
Due to coexistence of wild and cultivated form and interspecific hybridization, genus Abel-
moschus is controversial in systematics and taxonomy, such as taxonomic position of some
Abelmoschus species and the relationships between Abelmoschus species and part of Hibiscus
species [8]. In terms of morphological and cytological features, highly variable root, flower and
fruit characters of Abelmoschus have been used extensively in classification system [17, 18].
Patil et al. found seed coat sculpturing and seed trichomes could be used as the diagnostic
characters for many morphologically closely related species of Abelmoschus [5]. Fluoro-
chrome-binding pattern of nine Abelmoschus species showed polyploidy was an important fac-
tor in the chromosome number variation and evolution in this genus [4]. Some researchers
also used molecular markers to analyze genetic relationships of Abelmoschus, but most studied
focused on genetic diversity within A. esculentus and A. manihot [2, 7, 9, 18, 19], molecular
markers were relatively lacking in other species. Thus, new molecular tools were necessary to
study the accurate phylogeny in Abelmoschus.
Chloroplast is characteristic organelle in plant cells, and crucial in the photosynthesis and
biosynthesis of pigments, amino acids, starch and fatty acids [20, 21]. The chloroplast genome
generally has a circular structure with a pair of inverted-repeats (IR) regions (further called
IRa and IRb), a large single copy (LSC) region and a small single copy (SSC) region. Due to the
small size, conserved structure and gene content, it has been applied for resolving phyloge-
netics, evolution, taxonomic issues, population genetics and environmental adaptability [22].
Although chloroplast genome sequences of Abelmoschus esculentus has been deposited in Gen-
Bank (NC_035234.1) [23], there are no systematic, comprehensive and comparative studies of
chloroplast genome in Abelmoschus.
In this study, three chloroplast genomes of A. moschatus, A. manihot and A. sagittifolius
were sequenced and compared with the chloroplast genomes of A. esculentus (NC_035234.1)
and related species in Malvaceae. Apart from gene content and structure organization, com-
parative studies were conducted to identify mutational hotspots in Abelmoschus, and a phylo-
genetic tree of 33 species in family Malvaceae were constructed. These results will be useful in
PLOS ONE | https://doi.org/10.1371/journal.pone.0242591 November 25, 2020
2 / 19
PLOS ONE Complete chloroplast genomes of Abelmoschus moschatus, A. manihot and A. sagittifolius
developing molecular markers for resolving taxonomic issues of Abelmoschus, and elucidating
the evolutionary and phylogenetic relationships in the family of Malvaceae.
Materials and methods
Plant material, DNA isolation and sequencing
The fresh leaves of A. moschatus, A. manihot and A. sagittifolius were collected from the exper-
imental field of Guangdong academy of agricultural sciences (Guangzhou, China). All samples
were frozen in liquid nitrogen immediately and stored at −80˚C. Total DNA was extracted by
Plant DNA Isolation Kit (Tiangen, Beijing, China). Paired-end (PE) library was constructed
according to protocol of Illumina manual (San Diego, CA, USA), and then it was run on an
Illumina NovaSeq platform (Genepioneer Biotechnologies, Nanjing, China) with PE150
sequencing strategy and 350 bp insert size.
Chloroplast genome assembly and annotation
Raw reads of three Abelmoschus species were filtered using the software NGSQCToolkit
V2.3.3. In order to reduce the complexity of sequence assembly, filtered reads were compared
with the chloroplast genome database built by Genepioneer Biotechnologies (Nanjing, China)
using Bowtie2 V2.2.4, and sequences on the alignment was used as the chloroplast genome
sequence of samples [24]. Seed sequence was obtained by software SPAdes v3.10.1, and contigs
was acquired by kmer iterative extend seed. Then, the contigs were connected as scaffolds by
SSPACE v 2.0, and gaps were filled using Gapfiller v2.1.1 until the complete chloroplast
genome sequence was recovered. Finally, quality control was adopted to ensure the accuracy
of assembly results with the reference genome of A. esculentus (NC_035234.1).
The coding sequences and ribosomal RNA (rRNA) were obtained using software BLAST V
2.2.25 and HMMER V3.1 b2 after compared with the chloroplast genome database in National
Center of Biotechnology Information (NCBI). Aragorn V1.2.38 software was used for transfer
RNA (tRNA) prediction, then tRNA annotation information of chloroplast genome was
obtained. Chloroplast genome maps were made by OrganellarGenomeDRAW (OGDRAW).
Relative Synonymous Codon Usage (RSCU) and RNA editing sites
RSCU analysis of A. moschatus, A. manihot, A. sagittifolius and A. esculentus (NC_035234.1)
was determined using MEGA v7.0, and value of RSCU greater than one was considered to be a
higher codon frequency. The putative RNA editing sites were analyzed by PREP-cp (http://
prep.unl.edu/) with default parameters [25].
Simple Sequence Repeats (SSRs) and repeat sequences
The comparison of SSRs within four Abelmoschus species were identified using MISA
(MIcroSAtellite identification tool) v1.0 with 8 for mononucleotide repeats, 5 for di- and 3
each for tri-, tetra-, penta- and hexanucleotide repeats. Software vmatch v2.3.0 was used to
identify forward (F), reverse (R), palindromic (P), and complementary (C) repeats with mini-
mum repeats size �30 bp and sequence similarity of 90%.
Genetic divergence, substitutions and insertion/deletions (Indels) analysis
MAFFT ((Multiple Alignments using Fast Fourier Transform) V7.427 was used to perform
global alignment of protein-coding genes, intergenic spacer (IGS) regions, and intron regions
of complete chloroplast genome among A. moschatus, A. manihot, A. sagittifolius and A. escu-
lentus (NC_035234.1), and the value of genetic divergence (π) was calculated using DNAsp5.
PLOS ONE | https://doi.org/10.1371/journal.pone.0242591 November 25, 2020
3 / 19
PLOS ONE Complete chloroplast genomes of Abelmoschus moschatus, A. manihot and A. sagittifolius
With the reference genome of A. moschatus, different types of single nucleotide polymor-
phisms (SNPs) and Indels were determined in Abelmoschus using MAFFT program.
Analysis of non-synonymous (Ka)/synonymous (Ks), IR scope and
collinearity
In order to analyze substitution rates of Ka/Ks, the protein-coding genes of A. moschatus (as
reference) was compared with A. manihot, A. sagittifolius, A. esculentus, and three related spe-
cies in Malvoideae: Hibiscus rosa-sinensis (NC_042239.1), Althaea officinalis (NC_034701.2)
and Gossypium hirsutum (NC_007944.1). Protein-coding genes of all this species were aligned
with A. moschatus and analysed by MAFFT V7.427, and the Ka/Ks value was calculated by the
KaKs_calculator 2.0 [26].
The contraction and expansion of the IR boundaries among the above seven species in Mal-
voideae were visualized between the four regions of the chloroplast genome (LSC/IRb/SSC/
IRa) by Geneious R8.1. Meanwhile, the analysis of chloroplast sequence homology and collin-
earity was performed by Mauve software.
Phylogenetic analysis
Phylogenetic analysis was performed using the chloroplast genomes of A. moschatus, A. mani-
hot, A. sagittifolius and A. esculentus, along with related 29 species within the same family of
Malvaceae. Their accession numbers were listed in S1 Table. All chloroplast genome sequences
were aligned through MAFFT V7.427, and Indels were removed by TrimAl (V1.4.rev15), then
phylogenetic tree was constructed under maximum composite likelihood method
(GTRGAMMA model and bootstrap = 1000) using RAxML v8.2.10.
Results
Characterization of chloroplast genomes in Abelmoschus species
Illumina Novaseq 6000 produced a total of 25,192,038, 19,864,607 and 21,300,029 paired-end
(150bp) clean reads for A. moschatus, A. manihot and A. sagittifolius, with average organelle cov-
erage 4470, 1888 and 3194, respectively. Chloroplast genome size was ~163 kb in Abelmoschus
species, including a pair of IR regions separated by a LSC region and a SSC region (Fig 1 and
Table 1). The GC content of Abelmoschus chloroplast genomes was ~36%, and the LSC, SSC and
IR regions had similar content in four species, with ~34%, ~31% and ~41%, respectively.
The chloroplast genome of Abelmoschus species contained 132 genes (112 unique genes),
including 87 protein-coding, 37 tRNA, and 8 rRNA genes (Table 2). Gene trnH-GUG was not
annotated in the original annotation of A. esculentus (NC_035234.1). There are 20 duplicated
genes, including four rRNA genes and 16 other genes (ndhB, rpl2, rpl22, rpl23, rps12, rps19,
rps3, rps7, trnA-UGC, trnI-CAU, trnI-GAU, trnL-CAA, trnN-GUU, trnR-ACG, trnV-GAC and
ycf2), and all of them repeats once. Moreover, 18 intron-containing genes were found
(Table 3), fifteen of which contained one intron and three of which (ycf3, trnV-UAC and clpP)
contained two introns. Except the genes of trnA-UGC, trnI-GAU, ndhB, petD and petB, thir-
teen other genes had different fragment sizes of intron. The complete chloroplast genome has
been submitted to NCBI under GenBank accession numbers MT890968 for A. moschatus,
MT898000 for A. manihot, and MT898001 for A. sagittifolius.
Amino acid frequency, codon usage and RNA editing sites
Four Abelmoschus species showed similarity in amino acids frequency and codon usage. Pro-
tein-coding genes comprised 26713, 26705, 26714 and 26717 codons in A. moschatus, A.
PLOS ONE | https://doi.org/10.1371/journal.pone.0242591 November 25, 2020
4 / 19
PLOS ONE Complete chloroplast genomes of Abelmoschus moschatus, A. manihot and A. sagittifolius
Fig 1. Chloroplast genome map of three Abelmoschus species. Genes shown outside the circle are transcribed clockwise and those inside
counterclockwise. Genes belonging to different functional groups are color-coded.
https://doi.org/10.1371/journal.pone.0242591.g001
manihot, A. sagittifolius and A. esculentus, respectively (S2 Table and S1 Fig). Among those
amino acids, Leucine was the most encoded amino acid followed by Isoleucine and Serine,
while the Cysteine was the least abundant in chloroplast genomes. The use of the codons ATG
and TGG, encoding Methionine and Tryptophan, exhibited no bias (RSCU = 1.00) in Abel-
moschus. The findings also revealed that most of the amino acids preferred synonymous
codons (RSCU >1.00) having A/T at 30 end, except ATA and CTA encoding for Isoleucine
and Leucine, respectively.
Putative RNA editing sites were also determined in four Abelmoschus species. PREP pre-
dicted 55 putative RNA editing sites in 24 genes of A. moschatus and A. sagittifolius, 56 puta-
tive RNA editing sites in 24 genes of A. manihot, and 62 putative RNA editing sites in 24 genes
PLOS ONE | https://doi.org/10.1371/journal.pone.0242591 November 25, 2020
5 / 19
PLOS ONE Complete chloroplast genomes of Abelmoschus moschatus, A. manihot and A. sagittifolius
Table 1. Summary statistics for the chloroplast genomes of Abelmoschus species.
Genome features
Genome size (bp)
LSC size (bp)
SSC size (bp)
IR size (bp)
Number of genes
Protein genes [unique]
tRNA genes [unique]
rRNA genes [unique]
Duplicated genes in IR
GC content (%)
GC content in LSC (%)
GC content in SSC (%)
GC content in IR (%)
A. moschatus
A. manihot
A. sagittifolius
A. esculentus
163430
88243
18931
28128
132(112)
87(78)
37(30)
8(4)
20
36.71
34.47
31.55
41.95
163428
88194
18934
28150
132(112)
87(78)
37(30)
8(4)
20
36.70
34.48
31.55
41.93
163453
88314
18815
28162
132(112)
87(78)
37(30)
8(4)
20
36.69
34.45
31.59
41.90
163121
88071
19032
28009
131(111) a
87(78)
36(29) a
8(4)
20
36.74
34.55
31.48
41.97
a Data was from the A. esculentus chloroplast genome (NC_035234.1), and the number of genes should add one because gene trnH-GUG was not annotated.
https://doi.org/10.1371/journal.pone.0242591.t001
of A. esculentus (S3 Table). Similar RNA editing sites were found in most genes, however, gene
ycf3 was unique to A. esculentus and gene clpP was unique to A. moschatus, A. manihot and A.
sagittifolius. The highest number of editing sites were determined in ndhB (12), ndhD (7),
Table 2. List of annotated genes in the chloroplast genomes of A. moschatus, A. manihot, A. sagittifolius and A. esculentus.
Category
Gene group
Photosynthesis
Subunits of photosystem I (5)
Self-replication
Subunits of photosystem II (15)
Subunits of NADH
dehydrogenase (12)
Subunits of cytochrome b/f
complex (6)
Subunits of ATP synthase (6)
Large subunit of rubisco (1)
Proteins of large ribosomal
subunit (12)
Proteins of small ribosomal
subunit (16)
Subunits of RNA polymerase (4)
Ribosomal RNAs (8)
Transfer RNAs (37)
Other genes
Maturase (1)
Protease (1)
Envelope membrane protein (1)
Acetyl-CoA carboxylase (1)
c-type cytochrome synthesis gene
(1)
psaA, psaB, psaC, psaI, psaJ
psbA, psbB, psbC, psbD, psbE, psbF, psbH, psbI, psbJ, psbK, psbL, psbM, psbN, psbT, psbZ
ndhA�, ndhB�(×2), ndhC, ndhD, ndhE, ndhF, ndhG, ndhH, ndhI, ndhJ, ndhK
Gene name
petA, petB�, petD�, petG, petL, petN
atpA, atpB, atpE, atpF�, atpH, atpI
rbcL
rpl14, rpl16�, rpl2�(×2), rpl20, rpl22(×2), rpl23(×2), rpl32, rpl33, rpl36
rps11, rps12��(×2), rps14, rps15, rps16�, rps18, rps19(×2), rps2, rps3(×2), rps4, rps7(×2), rps8
rpoA, rpoB, rpoC1�, rpoC2
rrn16(×2), rrn23(×2), rrn4.5(×2), rrn5(×2)
trnA-UGC�(×2), trnC-GCA, trnD-GUC, trnE-UUC, trnF-GAA, trnG-GCC, trnG-UCC�, trnH-GUG, trnI-CAU
(×2), trnI-GAU�(×2), trnK-UUU�, trnL-CAA(×2), trnL-UAA�, trnL-UAG, trnM-CAU, trnN-GUU(×2),
trnP-UGG, trnQ-UUG, trnR-ACG(×2), trnR-UCU, trnS-GCU, trnS-GGA, trnS-UGA, trnT-GGU, trnT-UGU,
trnV-GAC(×2), trnV-UAC�, trnW-CCA, trnY-GUA, trnfM-CAU
matK
clpP��
cemA
accD
ccsA
unknown
function
Conserved hypothetical
chloroplast ORF (5)
ycf1, ycf2(×2), ycf3��, ycf4
Gene�: Gene with one intron; Gene��: Gene with two introns; Gene (×2): Number of copies of multi-copy gene.
https://doi.org/10.1371/journal.pone.0242591.t002
PLOS ONE | https://doi.org/10.1371/journal.pone.0242591 November 25, 2020
6 / 19
PLOS ONE Table 3. Information on 18 intron-containing genes in the chloroplast genomes of Abelmoschus species.
Complete chloroplast genomes of Abelmoschus moschatus, A. manihot and A. sagittifolius
Gene
trnK-UUU
rps16
trnG-UCC
atpF
rpoC1
ycf3
trnL-UAA
trnV-UAC
rps12 b
clpP
petB
petD
rpl16
rpl2
ndhB
trnI-GAU
trnA-UGC
ndhA
Location
Exon I (bp)
LSC
LSC
LSC
LSC
LSC
LSC
LSC
LSC
LSC/IRb
LSC
LSC
LSC
IR
IR
IR
IR
IR
SSC
37
40
23
145
432
124
35
38
114
71
6
8
9
391
777
37
38
553
Intron I (bp) a
2571/2563/2573/2576
862/865/870/ 856
803/809/804/ 811
815/815/816/814
773/774/781/780
791/790/791/790
557/558/559/556
590/590/590/608
-
676 / 679/ 677/676
812/812/812/821
757/757/757/757
1148/1147/1147/ 1142
698/698/698/696
683/683/683/683
957/957/957/957
794/794/794/794
1119/1120/1119/1119
Exon II (bp)
Intron II (bp) a
Exon III (bp)
809/804/813/ 834
536/536/536/536
943/942/943/948
153
26
228
35
227
48
410
1626
230
50
35
232
292
642
475
399
434
756
35
35
539
a The fragment size of intron is in the order of A. moschatus / A. manihot / A. sagittifolius / A.esculentus.
b The rps12 gene is divided into 5'-rps12 in the LSC region and 3'-rps12 in the IR region.
https://doi.org/10.1371/journal.pone.0242591.t003
matK(5) and petB (5). Genes of ndhD, ndhA and matK varied widely variations among species:
In A. moschatus, A. manihot and A. sagittifolius five and one RNA editing sites were found for
ndhD and ndhA, while in A. esculentus seven and one RNA editing sites were present, respec-
tively. A. manihot contained one more RNA editing sites in matK gene than other three spe-
cies. Most conversion occurred at the first and second nucleotides of the codons, and mainly
were C/G to A/T conversion. Change of RNA editing sites would produce abundant hydro-
phobic amino acids, especially Leucine, which was 29 in A. moschatus, A. manihot and A. sagit-
tifolius, and 28 in A. esculentus.
SSRs and repeat sequences
SSRs were detected by MISA software in Abelmoschus (Fig 2A and 2B). A. moschatus con-
tained 350, A. manihot (351), A. sagittifolius (350) and A. esculentus (344) SSRs. The maximum
SSRs were mononucleotide and accounted for about 60% of total SSRs, varying in size from 8
to 18 nucleotides. Trinucleotide and dinucleotide SSRs were also abundant and accounted for
about 33% of the total SSRs. A. moschatus and A. manihot did not contain hexanucleotides.
The A/T and AT/TA were the most abundant mononucleotide and dinucleotide SSRs, respec-
tively. The number of repeats units was also determined for all types of SSRs repeats (S4
Table). About 67% SSRs repeats were found in LSC, 13% in SSC, and 19% in IR. The IGS
regions contained the most SSRs, and comprised approximately 58% of the total SSRs.
Four categories of repeat sequences were also found in Abelmoschus, and there were 486
repeats were present in the chloroplast genomes of four species, 122 in A. moschatus, 137 in A.
manihot, 136 in A. sagittifolius and 91 in A. esculentus (Fig 2C–2F). Types of repeats (P, F and
R) had similar numbers in each species, but the number of type C is relatively small. The size
of repeats was mainly 30–54 bp in four Abelmoschus, and all contained one repeats above 55
PLOS ONE | https://doi.org/10.1371/journal.pone.0242591 November 25, 2020
7 / 19
PLOS ONE Complete chloroplast genomes of Abelmoschus moschatus, A. manihot and A. sagittifolius
Fig 2. Comparison of SSRs and repeat sequences among four Abelmoschus species. (a) Numbers of different types of SSRs; Mono-: mononucleotide, Di-:
dinucleotide, Tri-: trinucleotide, Tetra-: tetranucleotide, Penta-: pentanucleotide, Hexa-: hexanucleotide; (b) Location of SSRs in different chloroplast genome regions.
LSC: large single copy, SSC: small single copy, IR: inverted-repeat region. IGS: Intergenic spacer regions, CDS: coding DNA sequences, Intron: intronic regions; (c)
Different types of repeat sequences. Total: total numbers of all repeats. F: forward repeats, P: palindromic repeats, R: reverse repeats, C: complementary repeats; (d)
Number of repeats present in different locations of chloroplast genomes. LSC/IR: one copy of repeat present in LSC and another in IR, LSC/SSC: one copy of repeat
present in LSC and another in SSC, SSC/IR: one copy present in SSC and another in IR; (e) Number of repeats in different size. For example, 30–34 represent the
numbers of repeats with the size from 30 to 34 bp; (f) Number of repeats in different regions of chloroplast genomes. IGS/Intron: one copy of repeat present in
intergenic spacer regions and another in intronic regions. IGS/CDS: one copy of repeat present in intergenic spacer regions and another in coding regions.
https://doi.org/10.1371/journal.pone.0242591.g002
bp. Abundant repeats were found in the IGS regions, followed by IGS/Intron regions. Mean-
while, most of the repeats were located in LSC (92, 107, 105, 69), followed by IR (21, 22, 21, 14)
and lowest were in SSC (1, 2, 2, 2) in A. moschatus, A. manihot, A. sagittifolius and A.
PLOS ONE | https://doi.org/10.1371/journal.pone.0242591 November 25, 2020
8 / 19
PLOS ONE Complete chloroplast genomes of Abelmoschus moschatus, A. manihot and A. sagittifolius
esculentus, respectively. We also found some shared sequences in LSC/SSC (all 1), SSC/IR (all
2), and LSC/IR (2–4) in four species. The complete details of repeat sequences in four Abel-
moschus species were also listed in S5 Table.
SNPs and Indels in Abelmoschus
Diverse types of SNPs were determined in four Abelmoschus species using A. moschatus as refer-
ence. A. manihot, A. sagittifolius and A. esculentus showed 166, 79 and 262 SNPs in complete
chloroplast genome, respectively. SNPs of A/T, T/A and C/T were the largest number type
among the 12 substitutions in Abelmoschus (Fig 3A and 3B), and most SNPs were located in LSC
regions followed by SSC regions. The ratio of transition to transversion was 0.37 for A. manihot,
0.55 for A. sagittifolius and 0.51 for A. esculentus. Furthermore, Indels were also detected in differ-
ent regions of chloroplast genomes. A total of 120, 84 and 177 Indels were found in A. manihot,
A. sagittifolius and A. esculentus, and most of them existed in the LSC regions, but IR regions had
the longest Indel average length in A. manihot and A. sagittifolius (Fig 3C and 3E).
Mutational hotspots in Abelmoschus
Comparative analysis was conducted to identify mutational hotspots of protein-coding genes,
IGS and intron regions of chloroplast genome among four Abelmoschus species. The IGS
regions had more polymorphic (average π = 0.00432) compared to protein-coding regions
(average π = 0.00285) and intronic regions (average π = 0.00269). The nucleotide diversity was
ranged from 0.00013 (rps12-exon2-ndhF) to 0.02113 (clpP-exon3) in all the polymorphic con-
taining regions (Fig 4). A total of thirty highly diverse regions (region length �200 bp) were
Fig 3. Comparison of SNPs and Indels in four Abelmoschus species. A. moschatus was used as reference for SNPs and Indels detection. (a) The number of different
types of SNPs. (b) The number of SNPs in LSC, IR and SSC regions. (c) The number of Indels in LSC, IR and SSC regions. (d) Total length of Indels in LSC, IR and SSC
regions. (e) Average length of Indels in LSC, IR and SSC regions. LSC: large single copy, SSC: small single copy, IR: inverted-repeat region.
https://doi.org/10.1371/journal.pone.0242591.g003
PLOS ONE | https://doi.org/10.1371/journal.pone.0242591 November 25, 2020
9 / 19
PLOS ONE Complete chloroplast genomes of Abelmoschus moschatus, A. manihot and A. sagittifolius
Fig 4. Nucleotide diversity (π) of different regions among A. moschatus, A. manihot, A. sagittifolius and A. esculentus. Regions with 0 nucleotide diversity
were ignored. The x-axis represents chloroplast genome regions, and the y-axis represents nucleotide diversity.
https://doi.org/10.1371/journal.pone.0242591.g004
listed in Table 4. Most of these mutational hotspots belong to IGS regions, such as start-psbA,
atpB-rbcL and petD-exon2-rpoA. Higher nucleotide diversity was also observed for protein
coding genes, including 1st and 2nd exon of clpP, 1st intron of clpP, 2nd intron of ycf3, matK,
ndhF and 1st exon of atpF.
PLOS ONE | https://doi.org/10.1371/journal.pone.0242591 November 25, 2020
10 / 19
PLOS ONE Table 4. Mutational hotspots in four Abelmoschus species.
Complete chloroplast genomes of Abelmoschus moschatus, A. manihot and A. sagittifolius
Region
start-psbA
atpB-rbcL
petD-exon2-rpoA
clpP-intron1
clpP-exon2
accD-psaI
ndhC-atpE
ndhF-rpl32
clpP-exon1
psaJ-rpl33
psbI-atpA
rpoB-petN
rpl36-rps8
matK-rps16-exon1
psaI-ycf4
ndhE-ndhG
ycf3-intron2
atpF-exon2-atpH
petA-psbJ
psbA-matK
rpl32-ccsA
atpH-atpI
matK
rpl16-intron1
ndhF
psaA-ycf3-exon1
rps16-exon2-psbK
atpF-exon1
petB-exon2-petD-exon1
psbZ-rps14
https://doi.org/10.1371/journal.pone.0242591.t004
Genetic divergence
Total Number of mutations
Region length
0.0165
0.0111
0.0097
0.0089
0.0078
0.0063
0.0063
0.0058
0.0056
0.0045
0.0044
0.0040
0.0037
0.0035
0.0032
0.0032
0.0032
0.0031
0.0028
0.0027
0.0026
0.0026
0.0025
0.0025
0.0023
0.0023
0.0022
0.0021
0.0021
0.0021
23
30
6
14
5
11
33
11
3
5
24
16
4
13
3
2
6
4
6
4
7
7
9
6
12
5
5
2
1
5
642
1154
266
671
292
743
2343
862
228
475
2380
1907
464
1653
396
267
802
597
1155
631
1267
1169
1515
1132
2196
948
1023
410
207
1043
IR boundary and collinearity
The IR regions were compared among four Abelmoschus species and three closely related spe-
cies in family Malvaceae (Fig 5). The trnH gene of A. esculentus was reannotated in the junc-
tion of IRa/LSC. In four Abelmoschus species, the LSC/IRb boundary was located within the
coding region of rpl16 gene, with 66 to 68 bp in the IRb region. The ycf1 gene spanned the
boundary of the SSC/IRa region, with 959–1097 bp in the IRa region. The IRb/SSC and IRa/
LSC boundaries were crossed by the ndhF gene and trnH gene. However, ndhF gene was all
located in SSC region, and 7 bp from the boundary in A. esculentus. The trnH gene had the
same fragment size of 64 bp in LSC region of four Abelmoschus. Moreover, the genes of rpl16,
ndhF and ycf1 showed different fragment sizes of 1550-1556bp, 2196-2202bp and 5655-5712bp
in four Abelmoschus species, respectively. Based on LSC/IRb/SSC/IRa boundaries, the relation-
ships among A. moschatus, A. manihot and A. sagittifolius were closer than A. esculentus. In
addition, the pseudogene fragment of ycf1 was 123 bp in the junction of SSC/IRb in Hibiscus
rosa-sinensis. The trnH gene was all located in LSC region, with 0-13bp from the boundary in
H. rosa-sinensis, Althaea officinalis and G. hirsutum. The rps19 gene was in junction of LSC/
IRb in A. officinalis. The chloroplast genomes of 7 species were relatively conserved after
PLOS ONE | https://doi.org/10.1371/journal.pone.0242591 November 25, 2020
11 / 19
PLOS ONE Complete chloroplast genomes of Abelmoschus moschatus, A. manihot and A. sagittifolius
aligned by Mauve software, and no rearrangement occurred in gene organization (Fig 6), but
the gene layouts within SSC regions of A. officinalis and G. hirsutum were in the opposite ori-
entations compared with H. rosa-sinensis and four Abelmoschus species.
Ka/Ks substitution rate
In this study, we analyzed Ka/Ks rate of A. moschatus compared with to three species in the
same genus and three closely related species in family Malvaceae (S6 Table). Eighty-five pro-
tein-coding genes were analyzed and thirty-eight of them had an average Ka/Ks rate between 0
to 0.1 in seven species, which indicated these genes were under strong purifying selection pres-
sure in family Malvaceae. In contrast, three genes showed Ka/Ks>1.0, included gene rpl23 in
H. rosa-sinensis (1.12), A. officinalis (2.80) and G. hirsutum (1.79), gene clpP in A. esculentus
(6.01) and H. rosa-sinensis (3.67), gene ycf1 in A. manihot (1.50) and A. esculentus (2.98). In
addition, gene matK had Ka/Ks = 1.0 in A. esculentus, and seven genes (ndhA, ccsA, psbT,
rps15, rbcL, accD and ycf2) had Ka/Ks rate between 0.5 and 1.0 in at least one species.
Phylogenetic analysis
Maximum likelihood phylogenetic tree of 33 species in family Malvaceae were constructed based
on complete chloroplast genomes after removing the Indels. Phylogenetic analysis indicated that
A. moschatus sister to A. sagittifolius, four Abelmoschus species shared a common node with H.
taiwanensis and H. mutabilis, and then they came together with other Hibiscus species to form a
large group. The species of six different subfamilies were well distinguished with bootstrap values
about 100. However, Sterculioideae subfamily was divided into two groups because Heritiera
elata did not share a same node with other species in the same genus (Fig 7).
Discussion
Genome characteristics of Abelmoschus species and comparison with other
species in Malvaceae
Most species in Abelmoschus were economically important plants, but the chloroplast genomes
remained relatively limited, with only A. esculentus was sequenced [23]. In this study, three
chloroplast genomes of A. moschatus, A. manihot and A. sagittifolius were sequenced and com-
pares with A. esculentus. The sizes of chloroplast genomes ranged narrowly from 163121 to
163453 in four Abelmoschus species, and comparative analyses revealed highly conserved
structure and gene. Most angiosperms typically contained 74 to 79 protein-coding genes in
chloroplast genomes [27]. In this study, four Abelmoschus species all encoded 78 unique pro-
tein-coding genes, and was different with previously reported species of Hibiscus cannabinus
and three Firmiana speies in Malvaceae, which contain 79 protein-coding genes [22, 28].
Rabah et al. [23] reported A. esculentus had 29 unique tRNA genes, but the gene trnH-GUG,
located at LSC/IRa boundary, was not annotated, so we reannotated this gene for this species.
Within the same subfamily, previous studies reported 17, 19 and 18 intron-containing genes
in H. cannabinus, H. rosa-sinensis and 12 species of Gossypium, respectively [2, 28, 29], while
Abelmoschus harbored 18 intron-containing genes, thirteen out of them had intron length dif-
ferences among 4 species, and gene trnK-UUU had the longest intron with 2563–2576 bp.
The chloroplast genomes had well collinearity relationship among four Abelmoschus species
and three closely related species in Malvaceae, but some differences were detected in terms of
the direction of SSC, gene miss and IR expansion and contraction. Gene layouts within SSC
region had the same orientations between four Abelmoschus species and H. rosa-sinensis, but
A. officinalis and G. hirsutum had the opposite orientations compared with them, and similar
PLOS ONE | https://doi.org/10.1371/journal.pone.0242591 November 25, 2020
12 / 19
PLOS ONE Complete chloroplast genomes of Abelmoschus moschatus, A. manihot and A. sagittifolius
Fig 5. Comparative analysis of boundary regions: IR, SSC and LSC among four Abelmoschus species and three related species in Malvaceae.
https://doi.org/10.1371/journal.pone.0242591.g005
PLOS ONE | https://doi.org/10.1371/journal.pone.0242591 November 25, 2020
13 / 19
PLOS ONE Complete chloroplast genomes of Abelmoschus moschatus, A. manihot and A. sagittifolius
Fig 6. Co-linear analysis of seven Malvaceae chloroplast genomes. The Abelmoschus moschatus genome is shown at top as the reference. Within each of the alignment,
local collinear blocks are represented by blocks of the same color connected by lines.
https://doi.org/10.1371/journal.pone.0242591.g006
phenomenon with different inversions in the LSC region was also found in Chenopodium qui-
noa and Mangifera indica [23]. The infA gene as a translation initiation factor has been inde-
pendently lost many times during the evolution of land plants [27, 30], and it also missed in
Abelmoschus, but infA showed functional or non-functional in different Malvaceae species,
such as H. rosa-sinensis [2].
The border of IR was highly variable region with many nucleotide changes in chloroplast
genomes of closely related species. Among four Abelmoschus species, the genes of rpl16, ndhF
and ycf1 showed different fragment sizes in the IR boundaries, and the IRb/SSC border was
crossed by the ndhF except A. esculentus, in which ndhF had larger gene size and all located in
SSC region, this indicated that the relationships among A.moschatus, A. manihot and A. sagitti-
folius were closer than A. esculentus. Moreover, Abelmoschus species showed relatively inde-
pendent boundary traits compared with the other Malvaceae species. Gene rpl16 was located
at the junction of IRb/LSC in Abelmoschus, whereas A. officinalis and G. hirsutum presented
rps19 gene crossing the boundary or locating in LSC region, and ten species in different genus
of Malvaceae also showed rps19 gene in IRb/LSC [2]. In the IRb region, rps3 was the closest
gene to the IRb/LSC boundary in Abelmoschus, but this gene was replaced by the rpl2 in other
Malvaceae species. Durio zibethinus was a Malvaceae species with another boundary character-
istic, rpl23 (in LSC) and trnI-CAU (in IRb) were the closest genes to the IRb/LSC boundary,
and rpl23 and rpl2 had only one copy due to IR expansion and contraction [31]. These results
seem to be line with phylogenetic analysis, which indicated that species with more similar
boundary traits had closer phylogenetic relationship in Malvaceae.
SSRs and repeat sequences in Abelmoschus
Owing to the advantages of non-recombination, haploidy, uniparental inheritance and low
nucleotide substitution rate, chloroplast SSRs markers can be considered as an excellent tool in
PLOS ONE | https://doi.org/10.1371/journal.pone.0242591 November 25, 2020
14 / 19
PLOS ONE Complete chloroplast genomes of Abelmoschus moschatus, A. manihot and A. sagittifolius
Fig 7. Maximum likelihood phylogenetics tree of 33 species in family Malvaceae based on chloroplast genomes (Indels removed).
https://doi.org/10.1371/journal.pone.0242591.g007
population genetics and phylogeny analysis [32]. In the current study, mononucleotide SSR in
four Abelmoschus species varied in size from 8 to 18 nucleotides, which was different from
related species in Hibiscus (7 to 15 nucleotides) and Firmiana (7 to 22 nucleotides) [2, 22].
Both of A. sagittifolius and A. esculentus had six types SSRs, but A. moschatus and A. manihot
did not contain hexanucleotides. Most SSRs were distributed in LSC region and intergenic
region, and the identified SSRs in Abelmoschus revealed that A/T and AT/TA were the most
abundant in mononucleotide and dinucleotide SSRs respectively, which agreed with the
majority of plant family [24]. Moreover, repeat sequences was lower in A. esculentus compared
PLOS ONE | https://doi.org/10.1371/journal.pone.0242591 November 25, 2020
15 / 19
PLOS ONE Complete chloroplast genomes of Abelmoschus moschatus, A. manihot and A. sagittifolius
with A. moschatus, A. manihot and A. sagittifolius, but they shared similar distribution regions.
Abundant repeats were found in the intergenic spacer regions (IGS), followed by intronic
region and coding sequences, and the same distribution pattern of repeat sequences were also
reported in Hibiscus [2]. These repeat sequences were also crucial in chloroplast genome
arrangement and sequence variation of Abelmoschus.
Taxonomic discrepancies and hotspots in Abelmoschus
Previous studies reported that seed shape and trichome structure had major taxonomic impor-
tance and proved to be valuable characters for separating taxa of Abelmoschus [5]. SSR markers
(mainly in A. esculentus) were also developed from transcriptome data and genomic DNA to
investigate genetic relatedness and cross-species transferability [7, 19]. Pfeil et al. analyzed the
phylogeny of Hibiscus and the Tribe Hibisceae using chloroplast DNA sequences of ndhF and
rp116 intron, and found two tested Abelmoschus species were embedded within Hibiscus [33].
Werner et al. [8] used nuclear internal transcribed spacer (ITS) and chloroplast rpl16
sequences to construct phylogenetic relationships within Abelmoschus, and its relationship
with the genus Hibiscus and other related species in Malvaceae, but A. esculentus and A. caillei
cannot be distinguished from each other, and genetic diversity within A. esculentus and A. cail-
lei was low. In this study, we listed thirty highly mutational hotspots (�200 bp) after compar-
ing nucleotide diversity of protein-coding genes, IGS, and intron regions among Abelmoschus,
and these hotspots could be used to solve taxonomic discrepancies for genus Abelmoschus.
Most mutational hotspots belong to IGS regions, and some hotspots in protein-coding genes
had also been commonly used for barcoding markers in related genera, such as matK, rbcL
and ndhF [2, 33]. The nucleotide diversity of rpl16 intron was 0.0029, while the thirty hotspots
identified in Abelmoschus had nucleotide diversity from 0.0024 to 0.0142, and 23 regions had
higher polymorphic than previously reported sequence of rpl16. Interestingly, three exons and
one intron of clpP gene all showed high nucleotide diversity, especially the third exon (π =
0.02113, region length = 71bp), and polymorphic region of this gene had been proved to be
effective in evaluating the crop types and biogeographical origin of Cannabis sativa [34].
Therefore, all these mutational hotspots provided useful information for subsequent develop-
ment of chloroplast markers, evolutionary relationships and biogeographical origin.
Phylogenetic relationship in Malvaceae
Phylogenetic tree of 33 species in family Malvaceae were reconstructed using chloroplast
genomes without Indels in this study, and they were well divided into six subfamilies, except
Heritiera elata which did not share a node with other species in the same genus. Few previous
studies referred to the taxonomic position of A. sagittifolius, our phylogenetic tree suggested
that A. sagittifolius was closer to A. moschatus than to A. esculentus, which was consistent with
the morphological characteristics of pod and flower [12]. Furthermore, genus Abelmoschus
was previously included in the genus Hibiscus and later isolated from it [3]. Four Abelmoschus
species shared a common node with H. taiwanensis and H. mutabilis, and then they formed a
large group with other Hibiscus species located in different branches. These results indicated
that Abelmoschus was a well-supported clade within Hibiscus, and agreed with the viewpoint of
Werner et al. [8]. Thus, taxonomic treatment of Abelmoschus is an issue that required further
discussion. Abelmoschus could be merged with Hibiscus to form a broad genus Hibiscus, or it
maintains the taxonomic position of Abelmoschus, but some Hibiscus species need to change
their taxonomic position. As more complete chloroplast genomes are sequenced, the chloro-
plast genome data could be expected to help resolve the deeper branches of phylogeny and
complex evolutionary histories in Malvaceae [35].
PLOS ONE | https://doi.org/10.1371/journal.pone.0242591 November 25, 2020
16 / 19
PLOS ONE Complete chloroplast genomes of Abelmoschus moschatus, A. manihot and A. sagittifolius
Conclusions
Three chloroplast genomes of A. moschatus, A. manihot and A. sagittifolius were sequenced
and annotated in the present study, and compared with the chloroplast genomes of A. esculen-
tus and related species in Malvaceae. The results revealed the gene number and order, amino
acid frequency, and codon usage were similar in Abelmoschus. However, the differences were
also found in IR boundaries, intron-containing genes and the number of repeat sequences and
SNPs. Abelmoschus species also showed relatively independent IR boundary traits compared
with related species in Malvaceae, and identified thirty mutational hotpots might be useful for
developing molecular markers and resolving taxonomic discrepancies and biogeographical
origin both at genus Abelmoschus and family Malvaceae levels.
Supporting information
S1 Fig. Amino acids frequency in A. moschatus, A. manihot, A. sagittifolius and A. esculen-
tus.
(TIF)
S1 Table. Accessions of 33 species used in phylogenetic tree.
(DOCX)
S2 Table. Comparison of Relative Synonymous Codon Usage (RSCU) among A. moschatus,
A. manihot, A. sagittifolius and A. esculentus.
(DOCX)
S3 Table. RNA editing sites in A. moschatus, A. manihot, A. sagittifolius and A. esculentus.
(XLSX)
S4 Table. Details of SSRs in A. moschatus, A. manihot, A. sagittifolius and A. esculentus.
(XLSX)
S5 Table. Details of repeat sequences in A. moschatus, A. manihot, A. sagittifolius and A.
esculentus.
(XLSX)
S6 Table. Rate of synonymous and non-synonymous substitutions. Abelmoschus moschatus
(as reference genome) was compared with A. manihot, A. sagittifolius, A. esculentus, and three
closely related species in Malvoideae: Hibiscus rosa-sinensis (NC_042239.1), Althaea officinalis
(NC_034701.2) and Gossypium hirsutum (NC_007944.1). Eighty-five protein-coding genes
were analyzed.
(XLSX)
Author Contributions
Conceptualization: Jie Li, Guang-ying Ye.
Formal analysis: Jie Li, Hai-lin Liu.
Methodology: Jie Li.
Resources: Hai-lin Liu, Zai-hua Wang.
Supervision: Hai-lin Liu, Zai-hua Wang.
Validation: Guang-ying Ye.
Visualization: Jie Li.
PLOS ONE | https://doi.org/10.1371/journal.pone.0242591 November 25, 2020
17 / 19
PLOS ONE Complete chloroplast genomes of Abelmoschus moschatus, A. manihot and A. sagittifolius
Writing – original draft: Jie Li, Guang-ying Ye.
Writing – review & editing: Zai-hua Wang.
References
1. Christenhusz MJM, Byng JW. The number of known plants species in the world and its annual increase.
Phytotaxa. 2016; 261:201–217.
2. Abdullah, Mehmood F, Shahzadi I, Waseem S, Mirza B, Ahmed, et al. Chloroplast genome of Hibiscus
rosa-sinensis (Malvaceae): Comparative analyses and identification of mutational hotspots. Genomics.
2020; 112:581–591. https://doi.org/10.1016/j.ygeno.2019.04.010 PMID: 30998967
3.
Jarret RL, Wang ML, Levy IJ. Seed oil and fatty acid content in okra (Abelmoschus esculentus) and
related species. J Agric Food Chem. 2011; 59:4019–4024. https://doi.org/10.1021/jf104590u PMID:
21413797
4. Merita K, Kattukunnel JJ, Yadav SR, Bhat KV, Rao SR. Comparative analysis of heterochromatin distri-
bution in wild and cultivated Abelmoschus species based on fluorescent staining methods. Proto-
plasma. 2015; 252:657–664. https://doi.org/10.1007/s00709-014-0712-2 PMID: 25300590
5. Patil P, Malik S, Sutar S, Yadav S, Kattukunnel JJ, Bhat KV. Taxonomic importance of seed macro-and
micro-morphology in Abelmoschus (Malvaceae). Nord J Bot. 2015, 33:696–707.
6. Abd El-Fattah BES, Haridy AG, Abbas HS. Response to planting date, stress tolerance and genetic
diversity analysis among okra (Abelmoschus esculentus (L.) Moench.) varieties. Genet Resour Crop
Evol. 2020; 67:831–851.
7. Ravishankar KV, Muthaiah G, Mottaiyan P, Gundale SK. Identification of novel microsatellite markers in
okra (Abelmoschus esculentus (L.) Moench) through next-generation sequencing and their utilization in
analysis of genetic relatedness studies and cross-species transferability. J Genet. 2018; 97(Suppl
1):39–47. PMID: 29700273
8. Werner O, Magdy M, Ros RM. Molecular systematics of Abelmoschus (Malvaceae) and genetic diver-
sity within the cultivated species of this genus based on nuclear ITS and chloroplast rpL16 sequence
data. Genet Resour Crop Evol. 2016; 63:429–445.
9. Rubiang-Yalambing L, Arcot J, Greenfield H, Holford P. Aibika (Abelmoschus manihot L.): Genetic vari-
ation, morphology and relationships to micronutrient composition. Food Chem. 2016; 193:62–68.
https://doi.org/10.1016/j.foodchem.2014.08.058 PMID: 26433288
10. Yang Z, Tang H, Shao Q, Bilia A, Wang Y, Zhao X. Enrichment and purification of the bioactive flavo-
noids from flower of Abelmoschus manihot (L.) Medic using macroporous resins. Molecules. 2018;
23:2649. https://doi.org/10.3390/molecules23102649 PMID: 30332764
11.
Liu IM, Tzeng TF, Liou SS. Abelmoschus moschatus (malvaceae), an aromatic plant, suitable for medi-
cal or food uses to improve insulin sensitivity, Phytother Res. 2010; 24:233–239. https://doi.org/10.
1002/ptr.2918 PMID: 19610024
12. Chen DL, Li G, Liu YY, Ma GX, Zheng W, Sun XB. A new cadinane sesquiterpenoid glucoside with cyto-
toxicity from Abelmoschus sagittifolius. Nat Prod Res. 2018; 33:1–6. https://doi.org/10.1080/14786419.
2018.1437427 PMID: 29417849
13. Pritam SJ, Amol AT, Sanjay BB, Sanjay JS. Analgesic activity of Abelmoschus manihot extracts. Int J
Pharm. 2011; 7:716–720.
14. Patel K, Kumar V, Rahman M, Verma A, Patel DK. New insights into the medicinal importance, physio-
logical functions and bioanalytical aspects of an important bioactive compound of foods ‘Hyperin’:
Health benefits of the past, the present, the future. Beni-Suef Univ J Basic Appl Sci. 2018; 7:31–42.
15.
Jain PS, Bari SB. Evaluation of wound healing effect of petroleum ether and methanolic extract of Abel-
moschus manihot (L.) Medik., Malvaceae, and Wightiatinctoria R. Br., Apocynaceae, in rats. Rev Bras
Farmacogn. 2010; 30:756–761.
16. Pan XX, Tao JH, Jiang S, Zhu Y, Qian DW, Duan JA. Characterization and immunomodulatory activity
of polysaccharides from the stems and leaves of Abelmoschus manihot and a sulfated derivative. Int J
Biol Macromol. 2018; 107:9–16. https://doi.org/10.1016/j.ijbiomac.2017.08.130 PMID: 28860057
17. Sutar SP, Patil P, Aitawade M, John J, Malik S, Rao S, et al. A new species of Abelmoschus Medik.
(Malvaceae) from Chhattisgarh, India. Genet Resour Crop Evol. 2013; 60:1953–1958.
18. Pravin P, Shrikant S, Venkataraman BK. Species relationships among wild and cultivated Abelmoschus
Medik., (Malvaceae) species as reveled by molecular markers. Int J Life Sci. 2018; 6 (1):49–59.
19. Schafleitner R, Kumar S, Lin CY, Hedge SG, Ebert A. The okra (Abelmoschus esculentus) transcrip-
tome as a source of gene sequence information and molecular markers for diversity analysis. Gene.
2013; 517:27–36. https://doi.org/10.1016/j.gene.2012.12.098 PMID: 23299025
PLOS ONE | https://doi.org/10.1371/journal.pone.0242591 November 25, 2020
18 / 19
PLOS ONE Complete chloroplast genomes of Abelmoschus moschatus, A. manihot and A. sagittifolius
20.
Li W, Zhang C, Guo X, Liu Q, Wang K. Complete chloroplast genome of Camellia japonica genome
structures, comparative and phylogenetic analysis. PLOS ONE. 2019; 14(5):e0216645. https://doi.org/
10.1371/journal.pone.0216645 PMID: 31071159
21. Yan C, Du J, Gao L, Li Y, Hou X. The complete chloroplast genome sequence of watercress (Nastur-
tium officinale R. Br.): Genome organization, adaptive evolution and phylogenetic relationships in Car-
damineae. Gene. 2019; 699:24–36. https://doi.org/10.1016/j.gene.2019.02.075 PMID: 30849538
22. Abdullah, Shahzadi I, Mehmood F, Ali Z, Malik MS, Waseem S, et al. Comparative analyses of chloro-
plast genomes among three Firmiana species: Identification of mutational hotspots and phylogenetic
relationship with other species of Malvaceae. Plant Gene. 2019; 19:100199. https://doi.org/10.1016/j.
plgene.2019.100199
23. Rabah SO, Lee C, Hajrah NH, Makki RM, Alharby HF, Alhebshi AM, et al. Plastome Sequencing of Ten
Nonmodel Crop Species Uncovers a Large Insertion of Mitochondrial DNA in Cashew. Plant Genome.
2017; 10:1–14. https://doi.org/10.3835/plantgenome2017.03.0020 PMID: 29293812
24. Xiong Y, Xiong Y, He J, Yu Q, Zhao J, Lei X, et al. The complete chloroplast genome of two important
annual clover species, Trifolium alexandrinum and T. resupinatum: genome structure, comparative
analyses and phylogenetic relationships with relatives in Leguminosae. Plant. 2020; 9:478. https://doi.
org/10.3390/plants9040478 PMID: 32283660
25. Mower JP. The PREP suite: predictive RNA editors for plant mitochondrial genes, chloroplast genes
and user-defined alignments, Nucleic Acids Res. 2009; 37:253–259. https://doi.org/10.1093/nar/
gkp337 PMID: 19433507
26.
Zhang Z, Li J, Zhao XQ, Wang J, Wong GKS, Yu J. KaKs_Calculator: Calculating Ka and Ks through
model selection and model averaging. Genom Proteom Bioinf. 2006; 4:259–263. https://doi.org/10.
1016/S1672-0229(07)60007-2 PMID: 17531802
27. Millen RS, Olmstead R, Adams K, Palmer J, Lao N, Heggie L, et al. Many parallel losses of infA from
chloroplast DNA during angiosperm evolution with multiple independent transfers to the nucleus. Plant
Cell. 2001; 13:645–658. https://doi.org/10.1105/tpc.13.3.645 PMID: 11251102
28. Cheng Y, Zhang L, Qi J, Zhang L. Complete chloroplast genome sequence of Hibiscus cannabinus and
comparative analysis of the Malvaceae family. Front Genet. 2020; 11:227. https://doi.org/10.3389/
fgene.2020.00227 PMID: 32256523
29. Xu Q, Xiong G, Li P, He F, Huang Y, Wang K, et al. Analysis of complete nucleotide sequences of 12
Gossypium chloroplast genomes: origin and evolution of Allotetraploids. PLoS ONE. 2012; 7(8):
e37128. https://doi.org/10.1371/journal.pone.0037128 PMID: 22876273
30. Wicke S, Schneeweiss GM, dePamphilis CW, Mu¨ller KF, Quandt D. The evolution of the plastid chro-
mosome in land plants: gene content, gene order, gene function. Plant Mol Biol. 2011; 76: 273–297.
https://doi.org/10.1007/s11103-011-9762-4 PMID: 21424877
31. Cheon SH, Jo S, Kim HW, Kim YK, Sohn JY, Kim KJ. The complete plastome sequence of Durian,
Durio zibethinus L. (Malvaceae). Mitochondrial DNA Part B. 2017; 2:763–764.
32.
Liu L, Wang Y, He P, Li P, Lee J, Soltis DE. Chloroplast genome analyses and genomic resource devel-
opment for epilithic sister genera Oresitrophe and Mukdenia (Saxifragaceae), using genome skimming
data. BMC Genomics. 2018; 19:235. https://doi.org/10.1186/s12864-018-4633-x PMID: 29618324
33. Pfeil BE, Brubaker CL, Craven LA, Crisp MD. Phylogeny of Hibiscus and the Tribe Hibisceae (Malva-
ceae) using chloroplast DNA sequences of ndhF and the rp116 intron. Syst Bot. 2002; 27(2):333–350.
34. Roman MG, Houston R. Investigation of chloroplast regions rps16 and clpP for determination of Canna-
bis sativa crop type and biogeographical origin. Legal Med-Tokyo. 2020; 47:101759. https://doi.org/10.
1016/j.legalmed.2020.101759 PMID: 32711370
35. Conover JL, Karimi N, Stenz N, Ane´ C, Grover CE, Skema C, et al. A Malvaceae mystery: A mallow
maelstrom of genome multiplications and maybe misleading methods? J Integr Plant Biol. 2019;
61:12–31. https://doi.org/10.1111/jipb.12746 PMID: 30474311
PLOS ONE | https://doi.org/10.1371/journal.pone.0242591 November 25, 2020
19 / 19
PLOS ONE |
10.1371_journal.pone.0239797 | RESEARCH ARTICLE
The unequal impact of the coronavirus
pandemic: Evidence from seventeen
developing countries
Nicolas BottanID
1☯*, Bridget Hoffmann2☯, Diego Vera-Cossio2☯
1 Policy Analysis and Management, Cornell University, Ithaca, NY, United States of America, 2 Research
Department, Inter-American Development Bank, Washington, DC, United States of America
☯ These authors contributed equally to this work.
* nicolas.bottan@cornell.edu
Abstract
The current coronavirus pandemic is an unprecedented public health challenge that is hav-
ing a devastating economic impact on households. Using a sample of 230,540 respondents
to an online survey from 17 countries in Latin America and the Caribbean, the study shows
that the economic impacts are large and unequal: 45 percent of respondents report that a
household member has lost their job and, among households owning small businesses, 59
percent of respondents report that a household member has closed their business. Among
households with the lowest income prior to the pandemic, 71 percent report that a house-
hold member lost their job and 61 percent report that a household member has closed their
business. Declines in food security and health are among the disproportionate impacts. The
findings provide evidence that the current public health crisis will exacerbate economic
inequality and provides some of the first estimates of the impact of the pandemic on the
labor market and well-being in developing countries.
Introduction
Economic inequality is one of the leading economic issues of our era [1–4]. Recent economic
downturns, such as the Great Recession of 2008-2009, significantly increased economic
inequality [5–7]. Compared to other economic recessions, however, the COVID-19 pan-
demic is changing economic activity through different channels and on a substantially faster
timeline.
To slow the spread of COVID-19, governments have implemented regulations that require
social distancing, the closing of non-essential businesses, travel restrictions and, in many
cases, stay-at-home orders [8]. Human interactions that drive the economy, such as working
together in enclosed areas and enjoying entertainment activities, have been discouraged,
restricted, or banned altogether. Residents are complying with these measures, report that they
strongly support them [9–11], and actively seek information [12]. Although these measures
are necessary for public health, recent evidence from developed countries suggests that they
have negative economic impacts in the short-run [13–15] and can potentially deepen the pre-
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: Bottan N, Hoffmann B, Vera-Cossio D
(2020) The unequal impact of the coronavirus
pandemic: Evidence from seventeen developing
countries. PLoS ONE 15(10): e0239797. https://
doi.org/10.1371/journal.pone.0239797
Editor: William Joe, Institute of Economic Growth,
INDIA
Received: May 19, 2020
Accepted: September 15, 2020
Published: October 7, 2020
Peer Review History: PLOS recognizes the
benefits of transparency in the peer review
process; therefore, we enable the publication of
all of the content of peer review and author
responses alongside final, published articles. The
editorial history of this article is available here:
https://doi.org/10.1371/journal.pone.0239797
Copyright: © 2020 Bottan et al. This is an open
access article distributed under the terms of the
Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: Replication codes
and data for this study have been uploaded to the
Harvard Dataverse and can be accessed at: https://
doi.org/10.7910/DVN/7WX5UU.
PLOS ONE | https://doi.org/10.1371/journal.pone.0239797 October 7, 2020
1 / 10
PLOS ONE Funding: This project was funded by the Inter-
American Development Bank’s Coronavirus
research funds (RG-E1700 - Coronavirus survey).
The funder provided support in the form of salaries
for authors Bridget Hoffmann and Diego Vera-
Cossio, and provided funds to cover the cost of
recruiting subjects via social media advertising, but
did not have any additional role in the study design,
data collection and analysis, decision to publish, or
preparation of the manuscript. The specific roles of
these authors are articulated in the ‘author
contributions’ section.”
Competing interests: The authors Bridget
Hoffmann and Diego A. Vera-Cossio are employees
in the Research Department at the Inter-American
Development Bank. The author Nicolas Bottan is
employed at Cornell University. The authors have
no relevant financial or non-financial competing
interests to declare. Our affiliations do not alter our
adherence to PLOS ONE policies on sharing data
and materials.
The unequal impacts of the coronavirus pandemic
existing gaps between rich and poor [16]. These negative impacts could be exacerbated in
developing countries because firms and the workforce may be more vulnerable due to high lev-
els of informality and weaker governmental capacity to alleviate the pandemic.
This study employs a large-scale online household survey to examine how the COVID-19
pandemic resulted in differential economic impacts for households across the income distribu-
tion in Latin America and the Caribbean. The data show large and unequal job losses and busi-
ness closures, and the effects are strongest for the lowest income households. These negative
consequences also translate into declines in food security and support for policies to manage
the COVID-19 pandemic. Research on economic inequality is particularly relevant in Latin
America and the Caribbean. Although inequality and poverty declined over the most recent
decade [17, 18], prior to the pandemic, the region still had the highest income inequality in the
world [19], and a large share of citizens were vulnerable to falling back into poverty due to eco-
nomic shocks [20].
It is important to measure the economic impacts of the COVID-19 pandemic on house-
holds in Latin America and the Caribbean. [21], for example, describes the potential disruptive
effects of the pandemic across a wide domain of the global economy, including on labor supply
and the risks of small business closure and unemployment. This study complements this work
by quantifying the short-term implications of the pandemic on job losses and business closures
and shows that these impacts further aggravate inequality in the region. Furthermore, [22]
describes how macroeconomic spillovers can amplify the adverse economic effects of the pan-
demic, which would suggest that the estimates obtained are a lower bound. The results pre-
sented in this study may be useful to inform pandemic mitigation policy; by indicating where
the economic impacts of large-scale lockdowns are particularly large [23].
Methods
Design, setting and participants
For the purpose of this study, the questionnaire was mostly standardized across the countries
to allow for pooling the responses from all of the countries surveyed. The primary objective of
the survey was to measure the economic and well-being impacts that the current pandemic is
having on households in Latin America and the Caribbean. For this reason, the questionnaire
focused on collecting data on labor market outcomes, financial situation, and social program
enrollment. The survey also collected information on hunger, shortages of key goods, and
agreement with policies aimed at slowing the spread of COVID-19. A copy of the question-
naire can be found in the Supplementary Information section.
Households in 17 countries were surveyed: 8 South American countries (Chile, Colombia,
Bolivia, Ecuador, Guyana, Peru, Suriname, and Uruguay), 4 North and Central American
countries (Costa Rica, El Salvador, Mexico, and Panama), and 5 Caribbean countries (Domini-
can Republic, Bahamas, Barbados, Jamaica, and Trinidad and Tobago). The survey was first
launched on March 27, 2020 in Chile, and progressively rolled out to all other countries in our
sample by April 17, 2020. With the exception of Costa Rica, data collection continued until
April 30, 2020.
Sampling and validation
The same recruitment methods were followed in all countries. The study recruited participants
who were 18 years of age and above using paid advertisements on social media. Participation
in the survey was purely voluntary. The advertisements used keywords with broad appeal,
such as fútbol (soccer) or the names of local celebrities, to avoid selecting participants based on
COVID-19 knowledge or interest (details are available in the Supplementary Information
PLOS ONE | https://doi.org/10.1371/journal.pone.0239797 October 7, 2020
2 / 10
PLOS ONE The unequal impacts of the coronavirus pandemic
section). For each country, we exclude incomplete surveys, surveys in which the respondent’s
IP address did not belong to that country, surveys flagged as repeated, surveys with invalid
responses, and surveys unreasonable completion time from the sample.
The final sample consists of a total of 230,540 completed responses. S1 Fig in S1 File depicts
the geographic coverage of the sample. It shows the number of observations as a share of popu-
lation (in percent) by sub-national region for each country. The sample achieved broad geo-
graphic coverage, with observations in 92 percent of the sub-national regions (see S1 Table in
S1 File).
To validate the representativeness of the data, demographic characteristics from the online
survey were compared to nationally representative household surveys. Columns (1) and (2) in
S3 Table in S1 File shows that although the respondents of the online survey are more educated
and more likely to be females, they do not differ substantially in terms of household structure
or income levels. Columns (3) and (4) conduct an out-of-sample validation exercise and show
that by re-weighting the online survey responses by the inverse probability of being in the
nationally representative sample, the differences in demographic characteristics vanish. The
Supplementary Information section provides details on the steps taken to estimate weights for
the online survey and the validation exercise.
In order to document the economic and well-being impacts of the current pandemic is
having on households in Latin America and the Caribbean, most of the analysis presented in
this study re-weights observations to achieve national representativeness. These estimates
also weight observations according to country population to account for differences in sam-
ple size across countries. The exception is Fig 3, which re-weights observations to account
for temporal changes in the sample. See the Supplementary Information section for estima-
tion details. All results are robust to not using weights as shown in S2 and S3 Figs and S1
Table in S1 File.
Statistical analysis
The study presents descriptive statistics for the relevant outcomes aggregated across countries
(e.g., rates of job loss and business closure). Details on the analyses conducted for each figure
are available in the Supplementary Information. The study further examines how the loss of
livelihood relates to changes in household nutrition and policy support by estimating linear
regressions. These linear regressions include various controls to isolate time-varying locality
shocks and prevent differences in industry sectors from driving results. Refer to the Supple-
mentary Information section for more details on the estimations.
Results
COVID-19 and loss of livelihood
The data shows that 45 percent of respondents report that a household member lost a job and,
among households owning small family businesses, 58 percent of respondents report that a
household member closed their business. The recall period for these questions was random-
ized between one week, two weeks, and one month. The job-loss rates ranges from 42 percent
for a recall period of one week to 47 percent for a recall period of one month. The results sug-
gest that the rates of business closures remain constant at 58 percent across recall periods.
Compared to similar statistics in the United States (where 43 percent of small businesses closed
[15]), this study finds greater rates of business closure, implying that the economic impacts of
the pandemic may be stronger in developing countries.
PLOS ONE | https://doi.org/10.1371/journal.pone.0239797 October 7, 2020
3 / 10
PLOS ONE The unequal impacts of the coronavirus pandemic
Loss of livelihood and inequality
The overall effects obscure highly unequal impacts across income levels prior to the pandemic.
Fig 1 shows that after accounting for fixed factors by country, the percentage of households
reporting job losses declines monotonically with January 2020 income, prior to the onset of
the pandemic. In the case of business closures, the decline is similar though less dramatic.
Households reporting income of less than the national monthly minimum wage for January
2020 experienced the largest impacts, with nearly 71 percent reporting that a household mem-
ber lost their job and 61 percent reporting that a household member closed their business.
This contrasts sharply with the impacts reported by respondents with the highest household
incomes. Among the highest income respondents, only 14 percent report that a household
member lost their job and 54 percent report that a household member closed their business.
One potential explanation for these patterns is that high levels of informality in the region
may limit the ability of the most-vulnerable households to maintain their income source.
Using the share of self-employed workers as a proxy for the share of informal workers in the
labor market, Fig 2 shows that labor market informality is positively correlated with loss of
livelihood (job loss or business closure). The slope coefficient suggests that a percentage point
increase in the share of self-employed workers in a country increases the likelihood that a
respondent lost their livelihood by 0.54 percent (p-value = 0.005) with an R-squared of 42 per-
cent. Because informality rates are high in most developing countries, this result provides a
novel explanation for why labor markets in developing countries are particularly hard hit dur-
ing the crisis.
Fig 1. Higher rates of job loss and business closure among households in the lowest income group. Point estimates
and 95 percent confidence intervals for regressing the labor market outcome on income bin indicators and country
fixed effects. Data is weighted using within- and cross-country weights. See Empirical Methods in the Supplementary
Information section for details.
https://doi.org/10.1371/journal.pone.0239797.g001
PLOS ONE | https://doi.org/10.1371/journal.pone.0239797 October 7, 2020
4 / 10
PLOS ONE The unequal impacts of the coronavirus pandemic
Fig 2. Higher rates of livelihood loss in countries with higher informality. Each dot represents the share of
respondents who report that a household member lost a job or closed a business. Data is weighted using within- and
cross-country weights. See Empirical Methods in the Supplementary Information section for details.
https://doi.org/10.1371/journal.pone.0239797.g002
The impacts on job losses and business closures translate into reductions in income.
Table 1 reports within-locality changes in outcomes as a response to job losses or business clo-
sures. Column 1 of Panel A shows that respondents who report a job loss or business closure
are 24 percentage points (p-value < 0.01) more likely to report a reduction in income. Overall,
71 percent of respondents report that they expect their household income in April 2020 to be
lower than their January 2020 household income. Thirty-one percent of respondents report
household income of less than the national monthly minimum wage for January 2020 and
56% of households report that they expect their household income to be less than the national
minimum wage in April 2020. Fig 3 shows that the distribution of household income expected
in April 2020 is a leftward shift of the distribution of January 2020 household income. In par-
ticular, the share of households with incomes marginally above the national minimum wage
declines between January and April 2020, suggesting that many vulnerable households expect
to fall into poverty.
Nutrition and policy support
The data suggest that job losses and business closures lead to reductions in health and food
security. Columns 2 and 3 of Panel A of Table 1 shows that households with a job loss or busi-
ness closure are 13 percentage points (p-value < 0.01) more likely to suffer from hunger and 8
percentage points (p-value < 0.01) more likely to have a less healthy diet relative to their diet
prior to the pandemic. These magnitudes represent around a 25 percent increase from the
adjusted averages reported by households that report not having lost their livelihood (40
percent).
Column 1 of Panel B of Table 1 shows that households that lost their livelihoods during the
crises are 22 percentage points (p-value < 0.01) more likely to receive transfers from relatives
or friends. This finding suggests that households cooperate across income levels to smooth the
negative economic impacts of the pandemic. Seventy percent of respondents who report
household income less than the national minimum wage for January 2020 also report that a
household member received a gift or loan from a friend or relative. In contrast, only 26 percent
of respondents with the highest incomes in January 2020 report that a household member
received a gift or loan from a friend or relative. This pattern reverses for providing a gift or
PLOS ONE | https://doi.org/10.1371/journal.pone.0239797 October 7, 2020
5 / 10
PLOS ONE The unequal impacts of the coronavirus pandemic
Fig 3. The share of households in the bottom part of the income distribution is expected to increase. Shares of
households in each income bin for incomes reported for January 2020 and April 2020. Data counts weighted using
within- and cross-country weights. See Empirical Methods in the Supplementary Information section for details.
https://doi.org/10.1371/journal.pone.0239797.g003
Table 1. The loss of livelihood during the pandemic is linked to changes in nutrition and policy support.
Lost job or closed business
Observations
Adjusted R2
Lost job or closed business
Observations
Adjusted R2
Panel A: Impacts on income, food security, and health
(1)
(2)
Decreased income
Went hungry
Eats less healthy
0.241���
(0.008)
186,058
0.487
0.127���
(0.008)
198,190
0.602
Panel B: Impacts on transfers and policy support
(3)
0.085���
(0.008)
173,956
0.430
(3)
Gift/Loan
(1)
0.225���
(0.008)
198,017
0.479
(2)
Gov. Priority
Lockdown (> = month)
-0.027���
(0.008)
196,076
0.482
-0.042���
(0.010)
125,359
0.540
�p < 0.1,
��p < 0.05,
���p < 0.01.
The table reports regression coefficients capturing the relationship between loss of livelihood (job loss or business closure) and outcomes during the pandemic. Each
column reports results of a regression of the dependent variable on an indicator of whether any household member either lost a job or closed a business and a vector of
covariates. In addition, all regressions control for locality × day of survey completion fixed effects (18,764), as well as economic-sector fixed effects. Standard errors are
clustered at the locality level (3,165). Data is weighted using within- and cross-country weights. See Empirical Methods in the Supplementary Information section for
details.
https://doi.org/10.1371/journal.pone.0239797.t001
PLOS ONE | https://doi.org/10.1371/journal.pone.0239797 October 7, 2020
6 / 10
PLOS ONE The unequal impacts of the coronavirus pandemic
Fig 4. Support for extending lockdown policies declines more among households that lost their livelihoods. Point
estimates and 95 percent confidence intervals for the share of respondents supporting extending lockdown policies in
relation to the number of days since the first COVID-19 case in the country. Data is weighted using within- and cross-
country weights. See the Empirical Methods in the Supplementary Information section for details.
https://doi.org/10.1371/journal.pone.0239797.g004
loan. Thirty percent of households with January 2020 income less than the national minimum
wage also report that a household member provided a gift or loan, compared to 57 percent of
households with the highest incomes in January 2020 who report the same. Across all income
categories, the receipt of a gift or transfer is concentrated in households that report losing a job
or closing a business. This highlights the importance of informal social protection networks as
a tool for coping with the negative impacts of the pandemic.
Despite these substantial economic impacts, respondents strongly support measures to slow
the spread of the coronavirus in these early stages. Overall, 77 percent of respondents agree
with the statement that the top priority of the national government should be to stop the pan-
demic and 54 percent of respondents think that non-essential businesses should remain closed
for an additional month. Although the support for these policies is broad, the support is likely
to decrease as more households lose their livelihoods. Column 2 of Panel B from Table 1
shows that the probability of agreeing with the statement that the government’s priority should
be fighting the pandemic is 2 percentage points lower (p-value < 0.01) among households that
experienced a job loss or business closure. Column 3 of Panel B shows that the loss of liveli-
hoods during the pandemic is linked to a 4 percentage-point decline (p-value < 0.01) in the
probability of agreeing with the idea of keeping non-essential business closed for an additional
month. The support for policies that aim to slow the spread of the coronavirus is thus fragile.
Further, Fig 4 shows that, as days go by, the decline in support for keeping businesses closed
for an additional month declines faster among households that lose their livelihoods. One
important implication is that, without further assistance to impacted households, compliance
with mobility restriction policies is likely to decline.
PLOS ONE | https://doi.org/10.1371/journal.pone.0239797 October 7, 2020
7 / 10
PLOS ONE The unequal impacts of the coronavirus pandemic
Discussion
The data suggest two important explanations for the higher vulnerability of households in
developing countries with high levels of informality. First, the type of policies that aim to pre-
vent the spread of the virus is likely to affect informal workers more than formal workers (Fig
2). The survey captures data both from countries without enforced mobility-restriction poli-
cies or curfews, such as Uruguay, and also from countries with more-stringent, mandatory
quarantines and closures of non-essential businesses, as is the case in Bolivia and Peru. Because
most of the informal and self-employed workers tend to work in jobs that make them prone to
contact with other people (such as those in the retail or services sectors, as opposed to office or
industry jobs), the latter set of policies may lead to larger disruptions in labor markets. Indeed,
the data indicates that the share of respondents reporting job losses in their households during
April (69 percent) is substantially higher in countries with national or local mandatory quaran-
tines, relative to those in countries that did not implement mobility-restriction measures (34
percent) or only curfews (54 percent). See the Supplementary Information sections for a list of
countries by type of policy. This finding of smaller economic costs in countries that did not
implement national mobility restrictions at the onset of the pandemic than countries with
more stringent policies is consistent with evidence from modelling various quarantine regimes
gershon2020managing.
Second, differences in the ability to telework could be another reason the negative impacts
of the pandemic are concentrated in households with lower incomes. Among respondents that
are still employed, the share of respondents that report working from home during the past
week increases monotonically with January 2020 household income. Thirty percent of workers
from households with incomes below the national minimum wage report working from home,
while 76 percent of workers from the highest-income households report working from home.
This study’s findings on the inequality of the pandemic’s effects across the income distribu-
tion suggests that the pandemic may have long-lasting consequences linked to declines in the
stock of human capital. Consistent with the results presented in Fig 2, the consequences of the
loss of livelihoods on food security are stronger in countries with higher levels of informality
(see S5 Table in S1 File). This suggests that the structure of labor markets is not only magnify-
ing exposure to job losses and business closures, but is also magnifying the impacts of the loss
of livelihoods on household welfare because informal workers may have less access to formal
safety nets.
Taken together, these results show that the negative economic impacts of the COVID-19
pandemic have been concentrated among those who had lower incomes prior to the pan-
demic. This finding is important from both social and economic perspectives. Inequality is an
important social outcome in itself and as well as having important economic implications.
Although further research is needed, several studies have found that current inequality is nega-
tively correlated with future economic growth [24, 25], and in particular, inequality driven by
the lower tail of the income distribution stunts economic growth [26]. This implies that the
unequal economic impacts of this short-term public health pandemic could have long-term
implications for economic growth. The results of this study further indicate that country-level
rates of informality in labor markets are linked to stronger negative impacts and lower resil-
ience, suggesting that implementing policies to protect informal workers are needed.
Supporting information
S1 File.
(PDF)
PLOS ONE | https://doi.org/10.1371/journal.pone.0239797 October 7, 2020
8 / 10
PLOS ONE The unequal impacts of the coronavirus pandemic
Acknowledgments
We would like to thank Sebastian Espinoza and Maria Paula Medina for superb research assis-
tance. We would like to thank Julia´n Cristia for his encouragement and advice with the project.
We also want to thank German Reyes, Tom Sarrazin, Sebastia´n Oliva and Pablo Bachelet for
their tremendous support in the dissemination of the survey.
Author Contributions
Conceptualization: Nicolas Bottan, Bridget Hoffmann, Diego Vera-Cossio.
Data curation: Diego Vera-Cossio.
Formal analysis: Nicolas Bottan, Diego Vera-Cossio.
Funding acquisition: Bridget Hoffmann, Diego Vera-Cossio.
Investigation: Nicolas Bottan, Bridget Hoffmann, Diego Vera-Cossio.
Methodology: Nicolas Bottan, Bridget Hoffmann, Diego Vera-Cossio.
Project administration: Nicolas Bottan, Bridget Hoffmann, Diego Vera-Cossio.
Resources: Nicolas Bottan, Bridget Hoffmann, Diego Vera-Cossio.
Supervision: Nicolas Bottan, Bridget Hoffmann, Diego Vera-Cossio.
Validation: Nicolas Bottan, Bridget Hoffmann, Diego Vera-Cossio.
Visualization: Nicolas Bottan, Bridget Hoffmann, Diego Vera-Cossio.
Writing – original draft: Nicolas Bottan, Bridget Hoffmann, Diego Vera-Cossio.
Writing – review & editing: Nicolas Bottan, Bridget Hoffmann, Diego Vera-Cossio.
References
1. Stiglitz JE. Why It Matters. In: The priceof inequality: How today’s divided society endangers our future.
W.W. Norton and Company; 2013.
2. Milanovic B. Global Inequality in This Century and the Next. In: Global inequality: A new approach for
the age of globalization. Harvard University Press; 2016.
3. Piketty T, Yang L, Zucman G. Capital accumulation, private property, and rising inequality in China,
1978-2015. American Economic Review. 2019; 109(7):2469–2496. https://doi.org/10.1257/aer.
20170973
4.
Zucman G. Global wealth inequality; 2020.
5. Smeeding T. Income, wealth, and debt and the Great Recession; 2012.
6. Heathcote J, Perri F, Violante GI. Unequal we stand: An empirical analysis of economic inequality in the
United States, 1967-2006. Review of Economics Dynamics. 2010; 13(1):15–51. https://doi.org/10.
1016/j.red.2009.10.010
7. Brzezinski M. Income inequality and the Great Recession in Central and Eastern Europe. Economic
Systems. 2018; 42(2):219–247. https://doi.org/10.1016/j.ecosys.2017.07.003
8. Blackman A, Iba´ ñez AM, Izquierdo A, Keefer P, Moreira MM, Schady N, et al. La polı´tica publica frente
al Covid-10; 2020. IDB Policy Note.
9.
Fetzer T, Witte M, Hensel L, Jachimowicz JM, Haushofer J, Ivchenko A, et al. Global behaviors and per-
ceptions at the onset of the COVID-19 pandemic; 2020.
10. Allcott H, Boxell L, Conway JC, Gentzkow M, Thaler M, Yang DY. Polarization and public health: Parti-
san differences in social distancing during the coronavirus pandemic; 2020. NBER Working Paper No.
26946.
11. Briscese G, Lacetera N, Macis M, Tonin M. Compliance with COVID-19 social-distancing measure in
Italy: The role of expectations and duration; 2020. NBER Working Paper No. 26916.
PLOS ONE | https://doi.org/10.1371/journal.pone.0239797 October 7, 2020
9 / 10
PLOS ONE The unequal impacts of the coronavirus pandemic
12. Bento AI, Nguyen T, Wing C, Lozano-Rojas F, Ahn YY, Simon K. Evidence from internet search data
shows information-seeking responses to news of local COVID-19 cases. Proceedings of the National
Academy of Sciences. 2020; Available from: https://www.pnas.org/content/early/2020/05/01/
2005335117.
13. Adams-Prassl A, Boneva T, Golin M, Rauh C. Inequality in the impact of the coronavirus shock: Evi-
dence from real time surveys; 2020. https://doi.org/10.13140/RG.2.2.11867.72482
14.
Fetzer T.,Hensel L., Hermle J., Roth C. Coronavirus perceptions and economic anxiety. Review of Eco-
nomics and Statistics, 1–36; 2020. https://doi.org/10.1162/rest_a_00946
15. Bartik AW, Bertrand M, Cullen ZB, Glaeser EL, Luca M, Stanton CT. How Are Small Businesses Adjust-
ing to COVID-19? Early Evidence from a Survey. National Bureau of Economic Research; 2020. 26989.
16. Alon TM, Doepke M, Olmstead-Rumsey J, Tertilt M. The Impact of COVID-19 on Gender Equality.
National Bureau of Economic Research; 2020. 26947. https://doi.org/10.3386/w26947
17. Messina J, Silva J. Wage inequaity in Latin America. Understanding the past to prepare the future;
2018.
18.
Levy S, Schady N. Latin America’s social policy challenge: Education, social insurance, and redistribu-
tion. Journal of Economic Perspectives. 2013; 27(2):193–218. https://doi.org/10.1257/jep.27.2.193
19. Amarante V, Galva´n M, Mancero X. Inequality in Latin America: A global measurement; 2016. https://
doi.org/10.18356/bf11809a-en
20. Ba´ ez JE, Fuchs A, Rodrı´guez-Castela´ n C. Shaking up economic progress. Aggregate shocks in Latin
America and the Caribbean; 2017.
21.
Zaman K Taeen and Islam Hiya and Khan Asir Newaz and Shweta Dipita Saha and Rahman Ahsab
and Masud Jaasia and Araf Yusha and Sarkar Bishajit and Ullah Md Asad COVID-19 Pandemic Burden
on Global Economy: A Paradigm Shift; 2020. https://doi.org/10.20944/preprints202005.0461.v2
22. Kohlscheen Emanuel and Mojon Benoit and Rees Daniel The macroeconomic spillover effects of the
pandemic on the global economy. SSRN 3569554. 2020. https://doi.org/10.2139/ssrn.3569554
23. Gershon, David and Lipton, Alexander and Levine, Hagai Managing covid-19 pandemic without
destructing the economy. arXiv preprint arXiv:2004.10324; 2020.
24. Berg AG, Ostry JD. Inequality and Unsustainable growth: Two sides of the same coin? 2011.
25. Ostry J, Berg A, Tsangarides CG. Redistribution, inequality, and growth; 2014.
26. Cingano F. Trends in income inequality and its impact on economic growth; 2014.
PLOS ONE | https://doi.org/10.1371/journal.pone.0239797 October 7, 2020
10 / 10
PLOS ONE |
10.1371_journal.pone.0241407 | RESEARCH ARTICLE
Low-dose chest CT for diagnosing and
assessing the extent of lung involvement of
SARS-CoV-2 pneumonia using a semi
quantitative score
Thomas Leger1, Alexis Jacquier1,2, Pierre-Antoine Barral1, Maxime Castelli1,
Julie FinanceID
Philippe Brouqui3,4, Didier Raoult3,4, Axel Bartoli1,2, Jean-Yves Gaubert1,6,7,
Paul HabertID
3, Jean-Christophe Lagier3,4, Matthieu Million3,4, Philippe Parola3,5,
1,6,7*
1 Department of Radiology, La Timone Hospital, Assistance Publique des Hoˆpitaux de Marseille, Marseille,
France, 2 Aix-Marseille Universite´, UMR 7339, CNRS, Centre de Re´ sonance Magne´tique Biologique et
Me´dicale–Centre d’Exploration Me´taboliques par Re´sonance Magne´ tique, Assistance Publique—Hoˆ pitaux
de Marseille, Marseille, France, 3 IHU-Me´ diterrane´ e Infection, Marseille, France, 4 Aix Marseille Univ, IRD,
APHM, MEPHI, Marseille, France, 5 Aix Marseille Univ, IRD, APHM, SSA, VITROME, Marseille, France,
6 Aix Marseille Univ, LIIE, Marseille, France, 7 Aix Marseille Univ, CERIMED, Marseille, France
* Paul.habert@ap-hm.fr
Abstract
Objectives
The purpose is to assess the ability of low-dose CT (LDCT) to determine lung involvement in
SARS-CoV-2 pneumonia and to describe a COVID19-LDCT severity score.
Materials and methods
Patients with SARS-CoV-2 infection confirmed by RT-PCR were retrospectively analysed.
Clinical data, the National Early Warning Score (NEWS) and imaging features were
recorded. Lung features included ground-glass opacities (GGO), areas of consolidation and
crazy paving patterns. The COVID19-LDCT score was calculated by summing the score of
each segment from 0 (no involvement) to 10 (severe impairment). Univariate analysis was
performed to explore predictive factor of high COVID19-LDCT score. The nonparametric
Mann-Whitney test was used to compare groups and a Spearman correlation used with
p<0.05 for significance.
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: Leger T, Jacquier A, Barral P-A, Castelli
M, Finance J, Lagier J-C, et al. (2020) Low-dose
chest CT for diagnosing and assessing the extent
of lung involvement of SARS-CoV-2 pneumonia
using a semi quantitative score. PLoS ONE 15(11):
e0241407. https://doi.org/10.1371/journal.
pone.0241407
Editor: Wenbin Tan, University of South Carolina,
UNITED STATES
Received: June 16, 2020
Accepted: October 12, 2020
Published: November 3, 2020
Copyright: © 2020 Leger et al. This is an open
access article distributed under the terms of the
Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: All relevant data are
within the paper and its Supporting Information
files.
Results
Funding: The author(s) received no specific
funding for this work.
Competing interests: The authors have declared
that no competing interests exist.
Abbreviations: ARDS, acute respiratory distress
syndrome; COVID19, coronavirus disease 2019;
Eighty patients with positive RT-PCR were analysed. The mean age was 55 years ± 16, with
42 males (53%). The most frequent symptoms were fever (60/80, 75%) and cough (59/80,
74%), the mean NEWS was 1.7±2.3. All LDCT could be analysed and 23/80 (28%) were
normal. The major imaging finding was GGOs in 56 cases (67%). The COVID19-LDCT
score (mean value = 19±29) was correlated with NEWS (r = 0.48, p<0.0001). No symptoms
were risk factor to have pulmonary involvement. Univariate analysis shown that dyspnea,
PLOS ONE | https://doi.org/10.1371/journal.pone.0241407 November 3, 2020
1 / 12
PLOS ONE Low-dose chest CT scoring for SARS-CoV-2 pneumonia
GGO, ground-glass opacity; LDCT, low-dose chest
CT; NEWS, national early warning score; RT-PCR,
real-time reverse transcription-polymerase chain
reaction; SARS-CoV-2, severe acute respiratory
syndrome coronavirus 2.
high respiratory rate, hypertension and diabetes are associated to a COVID19-LDCT score
superior to 50.
Conclusions
COVID19-LDCT score did correlate with NEWS. It was significantly different in the clinical
low-risk and high-risk groups. Further work is needed to validate the COVID19-LDCT score
against patient prognosis.
Introduction
The outbreak of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), that began
in Wuhan, China, spread worldwide, and is now in the ascending phase of the epidemic in
Europe [1]. SARS-CoV-2 is responsible for coronavirus disease 2019 (COVID-19) pneumonia.
Respiratory involvement has a wide variety of clinical features, ranging from simple nasal con-
gestion to pulmonary failure [2]. Moreover, many patients remain asymptomatic but a vector
of the disease, allowing the epidemic to spread easily.
The main problem of COVID-19 pneumonia is the risk of saturation of our health system
due to an uncommonly massive inflow of patients requiring intensive care. Patient selection is
therefore essential in order for health care practitioners to focus on patients with comorbidities
and on severe cases. This selection is currently essentially based on clinical criteria.
Reverse-transcription polymerase chain reaction (RT-PCR) is currently the standard of ref-
erence for diagnosis of COVID 19 pneumonia. Chest x-ray might be replaced by low-dose
computed tomography (LDCT) in COVID-19 to assess lung involvement. Thus, computed
tomography (CT) has an important role in the management of patients, both for early screen-
ing and diagnosis and for establishing disease severity [3–5].
COVID-19 pneumonia lung lesions are now well characterized: they are typically asymmet-
ric, with ground-glass opacity (GGO) lesions or, less commonly, pulmonary areas of consoli-
dation, they have a peripheral distribution, and they preferentially affect the lower territories
[6]. It seems that there is a continuum in the lung lesions visible on CT scans, first with GGO
lesions; then reticulations appear with crazy-paving patterns, parenchymatous areas of consoli-
dation and finally, healing [7, 8].
The objective is to evaluate the ability of (LDCT) to analyze well-known imaging abnormal-
ities as well as to establish a COVID19-LDCT score reflecting disease severity and correlate it
with clinical risk scores to allow better selection and follow-up of patients.
Material and methods
Study design
The first 80 patients with a diagnosis of COVID-19 pneumonia confirmed by RT-PCR method
[9] were included in this single-centre retrospective study conducted from the 13th to the 20th
of March 2020. The inclusion criteria and patient enrolment included all consecutive patients
presenting to the department of infectious disease for 12 consecutive days with a diagnosis of
COVID-19 confirmed by RT-PCR. All patients underwent LDCT. The protocol was approved
by the local institutional review board: institut hospitalo-universitaire me´dite´rane´e-infection
N˚:2020–0012. The exclusion criteria was protocol refusal. For each patient, the following clin-
ical parameters were recorded: age, sex, date of first symptoms, date of chest CT scan, delay
between the first symptom and chest CT scan, fever, cough, dyspnea, diarrhea, myalgia,
PLOS ONE | https://doi.org/10.1371/journal.pone.0241407 November 3, 2020
2 / 12
PLOS ONE Low-dose chest CT scoring for SARS-CoV-2 pneumonia
rhinorrhea, abnormalities at lung auscultation, temperature, heart rate, blood pressure, respi-
ratory rate, oxygen saturation, and oxygen needed. Medical history parameters were recorded
as follows: heart disease, tobacco use, COPD, asthma, diabetes, obesity, sleep apnea syndrome,
oncologic status and immunosuppression status. The National Early Warning Score (NEWS)
was calculated for each patient using respiratory rate, oxygen saturation, supplemental oxygen
needed, temperature, systolic blood pressure, heart rate, and AVPU score, as described previ-
ously [10]. The NEWS determines the degree of illness of a patient and prompts critical care
intervention. According to previously published data, patients were divided into the following
two groups based on the NEWS in order to compare the radiological data in these populations:
low NEWS (NEWS � 4) and high NEWS (NEWS > 4) groups [10].
Equipment
All departments had to be reorganized in response to the increase in COVID-19 pneumonia,
and CT scans were dedicated for this activity with a specific patient circuit. Technicians and
workers in contact with COVID19 patients took all preventative means to protect themselves
against transmission of the virus following the local institutional recommendations. A video
was performed in our department to highlight the work of radiology technicians and available
on https://youtu.be/mI-L_ZrL__U.
LDCT
All patients underwent LDCT on the same system (Revolution EVO—GE Healthcare). All
LDCT scans were unenhanced in profound and maximal inspiration with the following
parameters: detector collimation: 0.625 mm; field of view: 500 mm; matrix: 512x512; pitch:
1.375; gantry speed 0.35s; 120 KV; 45 mAs; and reconstructed slice thickness 1.2mm. All imag-
ing data were reconstructed using high resolution and standard algorithms. LDCT data were
sent directly to a picture archiving and communicating system. Monitors were used to view
both mediastinal (width, 400 HU; level, 20 HU) and lung (width, 1500 HU; level, -2700 HU)
windows. The pre-established top anatomic border was the lower part of the neck. The pre-
established anatomic bottom boundary was the estimated location of the adrenal glands below
the costophrenic angle.
All LDCT scans were analysed by a single chest radiologist with more than 25 years of expe-
rience in chest imaging (JYG). All abnormalities were described according to the Fleischner
glossary [11]. The main features encountered during SARS-CoV-2 have been described else-
where as GGOs, areas of consolidation, crazy paving patterns, and extension of these lesions
might evolve dramatically to acute respiratory distress syndrome (ARDS). The type and distri-
bution of the lesions were analysed [12, 13].
COVID19-LDCT score
LDCT scores were used to quantify the extent of lung abnormalities. The score was obtained
by summing the notes attributed to each segment. The extent of the lesions of COVID-19
pneumonia was visually classified into 4 types for each segment: lack of lesions and mini-
mal, intermediate and severe involvement. Lack of involvement was defined as a normal
pattern and was equivalent to 0. Minimal involvement was defined as the presence of maxi-
mum 10 secondary lobules of any features and was equivalent to 1. Intermediate involve-
ment was defined as less than 50% involvement of the segment by any features and was
equivalent to 4. Severe involvement was defined as more than 50% involvement of the seg-
ment by any feature and was equivalent to 10. The score was obtained by summing the
score of all segments for the right and left lungs, with the result ranking between 0
PLOS ONE | https://doi.org/10.1371/journal.pone.0241407 November 3, 2020
3 / 12
PLOS ONE Low-dose chest CT scoring for SARS-CoV-2 pneumonia
Fig 1. Illustration of how to calculate the COVID19-LDCT score. A. The triangle with red borders shows the 6th
segment of the left lung with a unique GGO corresponding to minimal impairment. B. The triangle with red borders
shows the 6th segment of the right lung with GGO inside involving 50% of the segment corresponding to intermediate
impairment. C. Two triangles showing the 10th and the 9th segments of the left lung, where the extent of GGO
was > 50% and 50%, corresponding to severe and intermediate impairment, respectively.
https://doi.org/10.1371/journal.pone.0241407.g001
corresponding to a normal chest CT and 200 if all segments had involvement of more than
50% of their volume (Fig 1). Chest-CT was considered with high impairment if the global
score was superior to 50, we compared two groups to assess if clinical data could be predic-
tor of severe involvement of lung parenchyma.
PCR assay
Virological diagnosis of SARS-CoV-2 infection was performed using a sample nasopharyngeal
swabs with an hydrolysis probe-based real-time reverse transcription-PCR system that targets
the envelope (E) protein-encoding gene [9, 14].
Statistical analysis
Quantitative data were expressed as mean, standard deviation and range. Qualitative data were
expressed raw numbers, proportions and percentages. The differences between the variables
were compared by the Chi 2 or Fischer test for qualitative variables and by U Mann-Whitney
for quantitative variables. The relationship between quantitative parameters was examined
with Spearman correlation test. Binary logistic regression analysis was used to test the relation-
ships between COVID-LDCT score and clinically relevant variables. A value of p <0.05 indi-
cated a statistically significant difference. Statistical analysis was performed on XLSTAT v19.1
(Addinsoft, New York, USA).
Results
Patients
Between the 13th and the 20th of March 2020 the first 80 patients with positive RT-PCR results
for COVID-19 were enrolled in a single center. All data could be recorded for all patients. The
mean age of the population was 55 years ± 16, with 42 males and 38 females (47%). For 75 of
the patients (94%), the medical consultation was carried out following a symptom; for the oth-
ers, it followed close contact with an affected case.
PLOS ONE | https://doi.org/10.1371/journal.pone.0241407 November 3, 2020
4 / 12
PLOS ONE Low-dose chest CT scoring for SARS-CoV-2 pneumonia
Clinical and LDCT Data
Initially the most frequent symptoms were fever in 60 patients (75%) and cough in 59 patients
(74%). Myalgia, rhinorrhea, dyspnea and diarrhea were less frequent and were found in 35
(44%), 18 (23%), 11 (14%) and 11 (14%) patients, respectively. Lung auscultation abnormalities
were found in only 14 patients (18%). Five patients needed resuscitation for an ARDS 5/80 (6%).
At the first examination before LDCT, the mean temperature was 37.5˚C ± 0.9, heart rate
was 83 bpm ± 13, systolic blood pressure was 134 mmHg ± 21, diastolic blood pressure was 76
mmHg ± 14, respiratory rate was 19/min ± 5 and oxygen saturation was 97% ± 2. The mean
NEWS was 1.7 ± 2.3 (Table 1).
The delay between the first symptom and the LDCT scan was 7 ± 4 days. The dose-length
product mean was 41.7 mGy.cm ± 15.5. All patients could have their images analysed 80/80
Table 1. Characteristics of the population and clinical data.
Number of patients
Demographic data
Male
Female
Age (years)
Delay between symptoms and CT (days)
Symptomatic patients
Asymptomatic patients
Clinical data
Fever, chills
Cough
Dyspnea
Myalgia
Diarrhea
Rhinorrhea
Auscultation abnormalities
Temperature (˚C)
Heart rate (bpm)
Systolic blood pressure (mmHg)
Diastolic blood pressure (mmHg)
Respiratory rate (/min)
Oxygen saturation (%)
Oxygen needed
Heart disease
Tobacco use
COPD
Asthma
Sleep apnea syndrome
Oncologic disease
Diabetes
Total Population
80
42/80; 53%
38/80; 47%
55 ± 16 [17–89]
7 ± 4 [2–21]
75/80; 94%
5/80; 6%
60/80; 75%
59/80; 74%
11/80; 14%
35/80; 44%
11/80; 14%
18/80; 23%
14/80; 18%
37.5 ± 0.9 [36–40]
83 ± 13 [53–110]
134 ± 21 [100–219]
76 ± 14 [11–120]
19 ± 5 [8–36]
97 ± 2 [91–100]
10/80; 13%
9/80; 11%
10/80; 12%
3/80; 4%
8/80; 10%
6/80; 7%
6/80; 7%
8/80; 10%
Note: The qualitative variables are expressed as figures with percentages, and the continuous variables are expressed
as the as mean values ± SDs. ˚C: Celsuis degree; bpm: beat per minute; COPD: chronic obstructive pulmonary
disease; CT: computed tomography; mmHg: millimeter of mercury.
https://doi.org/10.1371/journal.pone.0241407.t001
PLOS ONE | https://doi.org/10.1371/journal.pone.0241407 November 3, 2020
5 / 12
PLOS ONE Low-dose chest CT scoring for SARS-CoV-2 pneumonia
Fig 2. Chest CT of patients with laboratory-proven COVID-19 pneumonia and paucisymptomatic patients.
Minimal form (a-c): Axial (a,b) and frontal reconstructions (c) of patchy ground-glass opacities (white arrows) with
segmental parenchyma involvement of 1 to 3 patchy lesions. Intermediate form (d-f): Axial (d,e) and frontal
reconstructions (f) of patchy ground-glass opacities (white arrows), band-like appearance (dotted arrow), and areas of
consolidation (arrowheads) with involvement less than 50%. Severe form (g-i): Axial (g,h) and frontal reconstructions
(i) of patchy ground-glass opacities (white arrows), band-like appearance (dotted arrow), and areas of consolidation
(arrowheads) with an involvement greater than 50%.
https://doi.org/10.1371/journal.pone.0241407.g002
(100%). On LDCT 23 patients showed no abnormalities (28%), 56 patients demonstrated GGOs
(67%), 21 patients had areas of consolidation (25%). An exclusive peripheral distribution was
found in 33 patients (40%). Only 9 patients showed crazy paving patterns (11%) (Fig 2).
COVID19-LDCT-score
The mean COVID19-LDCT score was 19 ± 28 with a mean right lung score of 10 ± 15 and a
mean left lung score of 8 ± 15 (Table 2). A significant correlation was found between the
Table 2. Imaging data.
Imaging data
Dose x length product (mGy.cm)
GGOs
Consolidation
Exclusively peripheral lesions
Crazy paving patterns
Normal imaging
Pleural effusions
Mediastinal nodes
COVID19-LDCT score
Lung score
Total Population
41.7 ± 15.5 [30.1–100.1]
56/80; 67%
21/80; 25%
33/80; 40%
9/80; 11%
23/80; 28%
4/80; 5%
4/80; 5%
19 ± 29 [0–161]
Note: The qualitative variables are expressed as figures with percentages, and the continuous variables are expressed
as the mean values ± SDs. GGOs: ground-glass opacities; LDCT: Low-dose computed tomography.
https://doi.org/10.1371/journal.pone.0241407.t002
PLOS ONE | https://doi.org/10.1371/journal.pone.0241407 November 3, 2020
6 / 12
PLOS ONE Low-dose chest CT scoring for SARS-CoV-2 pneumonia
Fig 3. Correlation between the NEWS and COVID19-LDCT score. A significant correlation was found. Pearson
coefficient = 0.48; p < 0.001.
https://doi.org/10.1371/journal.pone.0241407.g003
NEWS and the COVID19-LDCT score (r = 0.48, p<0.001). The correlation was also signifi-
cant for the right lung and the left lung (Fig 3).
No clinical symptoms were significant risk factors to find abnormalities on LDCT and espe-
cially cough was not a relevant clinical symptom to determine if patients had risk to have lung
involvement on LDCT.
Mean pulse oxymetry show a tendency to decrease in patient with pneumonia on LDCT
than without pneumonia: 97.6% ± 1.9 [92.0–100.0] and 96.9% ± 2.1 [91.0–100.0] respectively
(p = 0.07). Dyspnea was found in 11/80 cases. Over 69 patients without dyspnea 47 have an
abnormal LDCT and 22 were normal. Patients with normal LDCT showed a higher pulse oxy-
metry 98.0% ± 1.4 [95.0–100.0] compared to patient with pneumonia on LDCT 97.2% ± 1.8
[91.0–100.0] (p = 0.028) (Fig 4).
Score NEWS
Sixty-nine patients were classified as having low NEWS, and 11 patients had high NEWS. The
COVID-LDCT score was 13 ± 19 in the low-NEWS group and was significantly higher in the
high-NEWS group (52 ± 52; p = 0.001).
There was no significant difference between groups according to their symptoms or blood
pressure at the time of admittance. Patients with a NEWS > 4 had a significantly higher tem-
perature, heart rate, and respiratory rate and a lower oxygen saturation than patients with a
low NEWS (38.3˚C ± 1.0 versus 37.3˚C ± 0.8; p < 0.0001, 95 bpm ± 8 versus 81 bpm ± 12;
p<0.0001, 25/min ± 9 versus 18/min ± 4; p < 0.0001, 94% ± 2 versus 98% ± 2; p < 0.0001;
respectively). Oxygen was needed in 7 of 11 patients (64%) in the high-NEWS group and in 3
of 69 patients (7%) in the low-NEWS group (Table 3). The clinical factors associated with high
COVID19-LDCT Score in univariable analysis were diabetes (OR = 6.9 [1.32–36.2], p = 0.02),
PLOS ONE | https://doi.org/10.1371/journal.pone.0241407 November 3, 2020
7 / 12
PLOS ONE Low-dose chest CT scoring for SARS-CoV-2 pneumonia
Fig 4. Pulse oxymetry according to LDCT and dyspnea. A significant difference was found between pneumonia and non pneumonia on LDCT concerning
the oxymetry (p = 0.07) and between pneumonia and non pneumonia on LDCT among patient without dyspnea (p = 0.028). LDCT: low dose computed
tomography.
https://doi.org/10.1371/journal.pone.0241407.g004
hypertension (OR = 7.20 [1.22–42.5], p = 0.03), dyspnoea (OR = 7.65 [1.66–35.3], p = 0.01)
and respiratory rate (OR = 1.15 [1.03–1.30], p = 0.02). Because of the limited number of
patients with high COVID19-LDCT score and the sample size, no multivariable analysis was
performed.
Discussion
The main findings of the presented work are that 1) LDCT scans can depict the typical features
of SARS-CoV-2 pneumonia with limited irradiation; and 2) the COVID19-LDCT score is cor-
related with the NEWS used routinely to assess disease severity and patient prognosis.
Table 3. Low- and high-risk groups.
Clinical data’s
Symptoms
Temperature (˚C)
Heart rate (bpm)
Systolic blood pressure (mmHg)
Diastolic blood pressure (mmHg)
Respiratory rate (/min)
Oxygen saturation (%)
Oxygen need
COVID19-LDCT score
Lung score
Low-risk patients
NEWS 0–4 N = 69
64/69; 93%
37.3 ± 0.8 [35.8–39.5]
81 ± 12 [53–110]
134 ± 22 [100–219]
77 ± 14 [11–120]
18 ± 4 [10–28]
98 ± 2 [92–100]
3/69; 4%
High-risk patients
NEWS >4 N = 11
11/11; 100%
38.3 ± 1.0 [36.5–40.0]
95 ± 8 [80.0–105.0]
128 ± 18 [107–160]
72 ± 13 [50–90]
25 ± 9 [8–36]
94 ± 2 [91–98]
7/11; 64%
13 ± 19 [0–80]
52 ± 52 [3–161]
p < 0.001
p < 0.001
p = 0.501
p = 0.440
p < 0.001
p < 0.001
p < 0.001
P = 0.001
Note: The qualitative variables are expressed as figures with percentages, and the continuous variables are expressed as the mean values ± SDs. ˚C: Celsuis degree; bpm:
beat per minute; LDCT: low-dose computed tomography; mmHg: millimeter of mercury.
https://doi.org/10.1371/journal.pone.0241407.t003
PLOS ONE | https://doi.org/10.1371/journal.pone.0241407 November 3, 2020
8 / 12
PLOS ONE Low-dose chest CT scoring for SARS-CoV-2 pneumonia
Unenhanced chest CT was performed using a low-dose acquisition protocol. All the CT
scans performed allowed optimal analysis of parenchyma involvement. Recently, recommen-
dations have been made regarding the use of LDCT for COVID19 that indicate preference for
LDCT [15]. They reported a mean dose-length product of 14.4 mGy.cm. A patient with
COVID-19 pneumonia could have 3 to 6 chest CT scans in a short period of time, and a
healthy patient could have 1 or 2 chest CT scans to ensure that they did not have COVID-19
[16]. Irradiation awareness is needed, especially for younger patients [17]. The present study
showed that CT image quality using LDCT is sufficient for depicting all the typical features of
COVID-19 pneumopathy. Acquisition parameters settings should be adapted to each patient
to maintain good contrast-to-noise ratio and signal-to-noise ratio, allowing adequate quality
of the lung parenchyma images [18]. Debray et al. have already shown in patients with lung
transplantation that the use of LDCT could be interesting to decrease global irradiation
because of repeated imaging for follow-up [19]. However, in patients with COVID-19 an
enhanced chest CT scan might be required for patients with clinical worsening considering an
over-risk of pulmonary embolism caused by inflammatory syndrome and bed rest.
As recently described, the clinico-radiological correlation is good, and diffuse lung involve-
ment is frequently observed in severe forms [20, 21]. Lesions extent is higher in patients with a
need for intensive care units than for ordinary units. In particular, lymph nodes and pleural
effusion are highly frequent in severe patients. The present paper is in favor of a correlation
between disease extent and clinical severity. There are some patients with a RT-PCR positive
result and a normal chest CT 23 of 80 (28%), that is not consistent with the whole studies on
CT scan sensitivity. We explained that because patients came from a screening center and
some patients are tested because they are contact case with infected patients. The sensitivity of
the CT scan depends on the level of pre-test risk of the population to present lung involvement.
Disappointingly, we did not highlight any symptoms that could predict pulmonary involve-
ment, particularly cough, which is generally one of the main symptoms of manifestations of
alveolar involvement. It is in line with a recent publication from Yang et al. showing that
patients could be COVID-19 infected and had few or no clinical symptoms with a normal
chest CT and/or negative RT-PCR test [22].
The COVID19-LDCT score may provide the quantitative measurement of COVID-19
injury. Patients can be easily classified in various degrees of the disease. This type of CT score
was first described in 2004 for severe acute respiratory syndrome (SARS). Investigators divided
each lung into 3 parts (upper, middle and lower) and evaluated the involvement on a scale of
0–4: 0, <25%, 25–50 >50% and >75% involvement: the maximum possible score was 24 [23].
The present approach uses the simplest scale, comprising 3 kinds of impairment: minimal,
intermediate, and severe. It focuses on the anatomical distribution of lesions based on segmen-
tal analysis. This scoring system is based on the decreased in lung function applayed in tho-
racic surgery. It has been shown that wedge resection spare parenchyma loss more than
segmentectomy and segmentectomy more than lobectomy [24]. We guess that a minimal
involvement will have only few consequences on hematosis and more than 50% of segment
involvement seems lead to a total shunt effect and we grade it 10. The score over 50/200 had
been arbitrary chosen for severe involvement because that mean involvement of 50% of the
lung corresponding to a lobectomy. Patients with comorbidities such as emphysema or former
thoracic surgery have a higher risk of death in case of lobectomy [25]. We considered that a
score > 50/200 could lead to death according to the cormobidities and need to classify patients
in the severe group.
There are a few limitations in the present study. Our population was mainly paucisympto-
matic patients which did not allow us to have a wide repartition of NEWS and CT scores for
correlation, especially for high scores. This could be explained because we enrolled the 80 first
PLOS ONE | https://doi.org/10.1371/journal.pone.0241407 November 3, 2020
9 / 12
PLOS ONE Low-dose chest CT scoring for SARS-CoV-2 pneumonia
patients with COVID-19 pneumonia in our centre and the severe forms are not the most fre-
quent. Another limit is the low number of patient that does not allowed us to perform a multi-
variate analysis and further studies are needed to establish this score a clinical prognosis tool.
Low-dose CT scan could depict the typical features of COVID-19 pneumonia. In this popu-
lation of 80 COVID-19 patients detected positive by RT-PCR, there was a significant correla-
tion between the clinical NEWS and the COVID19-LDCT score, showing the close
relationship between signs and the extent of disease in the lung. More work is required to
assess whether this score could add clinically relevant information to assess the prognosis of
patients with COVID-19 pneumonia.
Supporting information
S1 File. IRB.
(PDF)
S1 Data.
(XLSX)
Acknowledgments
We would like to thank all the technicians for their professionalism during this crisis. Thanks
to health managers to the optimal organization of workflow in the department.
Author Contributions
Conceptualization: Alexis Jacquier, Jean-Yves Gaubert.
Data curation: Maxime Castelli, Julie Finance, Axel Bartoli.
Formal analysis: Pierre-Antoine Barral, Paul Habert.
Investigation: Thomas Leger, Alexis Jacquier, Jean-Christophe Lagier, Matthieu Million, Phi-
lippe Parola, Philippe Brouqui, Axel Bartoli.
Methodology: Philippe Parola.
Project administration: Alexis Jacquier, Jean-Yves Gaubert, Paul Habert.
Resources: Jean-Christophe Lagier, Matthieu Million, Philippe Parola, Philippe Brouqui,
Didier Raoult, Axel Bartoli.
Validation: Alexis Jacquier, Jean-Yves Gaubert.
Visualization: Thomas Leger, Alexis Jacquier, Pierre-Antoine Barral, Jean-Christophe Lagier,
Matthieu Million, Philippe Parola, Philippe Brouqui, Didier Raoult, Paul Habert.
Writing – original draft: Thomas Leger, Paul Habert.
Writing – review & editing: Thomas Leger, Alexis Jacquier, Didier Raoult, Jean-Yves Gaubert,
Paul Habert.
References
1. Munster VJ, Koopmans M, van Doremalen N, van Riel D, de Wit E. A novel coronavirus emerging in
China—key questions for impact assessment. N Engl J Med. 2020; 382(8):692–694. https://doi.org/10.
1056/NEJMp2000929 PMID: 31978293
2. Huang C, Wang Y, Li X, Ren L, Zhao J, Hu Y, et al. Clinical features of patients infected with 2019 novel
coronavirus in Wuhan, China. Lancet Lond Engl. 15 2020; 395(10223):497–506.
PLOS ONE | https://doi.org/10.1371/journal.pone.0241407 November 3, 2020
10 / 12
PLOS ONE Low-dose chest CT scoring for SARS-CoV-2 pneumonia
3. Ai T, Yang Z, Hou H, Zhan C, Chen C, Lv W, et al. Correlation of chest CT and RT-PCR testing in coro-
navirus disease 2019 (COVID-19) in China: a report of 1014 cases. Radiology. 26 fe´vr 2020;200642.
https://doi.org/10.1148/radiol.2020200642 PMID: 32101510
4. Chen Z, Li Y, Wu B, Hou Y, Bao J, Deng X. A patient with COVID-19 presenting a false-negative reverse
transcriptase polymerase chain reaction result. Korean J Radiol [Internet]. 2020 [cite´ 25 mars 2020];21.
Disponible sur: https://www.kjronline.org/DOIx.php?id=10.3348/kjr.2020.0195 PMID: 32207257
5.
Fang Y, Zhang H, Xie J, Lin M, Ying L, Pang P, et al. Sensitivity of chest CT for COVID-19: comparison
to RT-PCR. Radiology. 2020;200432. https://doi.org/10.1148/radiol.2020200432 PMID: 32073353
6. Chung M, Bernheim A, Mei X, Zhang N, Huang M, Zeng X, et al. CT imaging features of 2019 novel
coronavirus (2019-nCoV). Radiology. avr 2020; 295(1):202–7. https://doi.org/10.1148/radiol.
2020200230 PMID: 32017661
7. Xiong Y, Sun D, Liu Y, Fan Y, Zhao L, Li X, et al. Clinical and high-resolution CT features of the COVID-
19 infection: comparison of the initial and follow-up changes. Invest Radiol. 3 mars 2020. https://doi.org/
10.1097/RLI.0000000000000674 PMID: 32134800
8. Wang Y, Dong C, Hu Y, Li C, Ren Q, Zhang X, et al. Temporal changes of CT findings in 90 patients
with COVID-19 pneumonia: a longitudinal study. Radiology. 19 mars 2020;200843. https://doi.org/10.
1148/radiol.2020200843 PMID: 32191587
9. Amrane S, Tissot-Dupont H, Doudier B, Eldin C, Hocquart M, Mailhe M, et al. Rapid viral diagnosis and
ambulatory management of suspected COVID-19 cases presenting at the infectious diseases referral
hospital in Marseille, France,—January 31st to March 1st, 2020: A respiratory virus snapshot. Travel
Med Infect Dis. mars 2020;101632. https://doi.org/10.1016/j.tmaid.2020.101632 PMID: 32205269
10. Smith GB, Prytherch DR, Meredith P, Schmidt PE, Featherstone PI. The ability of the National Early
Warning Score (NEWS) to discriminate patients at risk of early cardiac arrest, unanticipated intensive
care unit admission, and death. Resuscitation. avr 2013; 84(4):465–70. https://doi.org/10.1016/j.
resuscitation.2012.12.016 PMID: 23295778
11. Hansell DM, Bankier AA, MacMahon H, McLoud TC, Mu¨ ller NL, Remy J. Fleischner Society: glossary
of terms for thoracic imaging. Radiology. mars 2008; 246(3):697–722. https://doi.org/10.1148/radiol.
2462070712 PMID: 18195376
12. Cheng Z, Lu Y, Cao Q, Qin L, Pan Z, Yan F, et al. Clinical features and chest CT manifestations of coro-
navirus disease 2019 (COVID-19) in a single-center study in Shanghai, China. Am J Roentgenol. 14
mars 2020;1–6.
13. Ye Z, Zhang Y, Wang Y, Huang Z, Song B. Chest CT manifestations of new coronavirus disease 2019
(COVID-19): a pictorial review. Eur Radiol [Internet]. 19 mars 2020; Disponible sur: https://doi.org/10.
1007/s00330-020-06801-0 PMID: 32193638
14.
Lagier JC, Colson P, Tissot Dupont H, Salomon J, Doudier B, Aubry C, et al. Testing the repatriated for
SARS-Cov2: Should laboratory-based quarantine replace traditional quarantine? Travel Med Infect Dis.
14 mars 2020;101624. https://doi.org/10.1016/j.tmaid.2020.101624 PMID: 32179125
15. Kang Z, Li X, Zhou S. Recommendation of low-dose CT in the detection and management of COVID-
2019. Eur Radiol [Internet]. 19 mars 2020 [cite´ 25 mars 2020]; Disponible sur: https://doi.org/10.1007/
s00330-020-06809-6 PMID: 32193637
16.
Lei DP. The progression of computed tomographic (CT) images in patients with coronavirus disease
(COVID-19) pneumonia: the CT progression of COVID-19 pneumonia. J Infect [Internet]. 20 mars 2020
[cite´ 25 mars 2020];0(0). Disponible sur: https://www.journalofinfection.com/article/S0163-4453(20)
30151-1/abstract PMID: 32205140
17.
Taqi AH, Faraj KA, Zaynal SA. The Effect of Long-term x-ray exposure on human lymphocyte. J Biomed
Phys Eng. 1 fe´ vr 2019; 9(1):127–32. PMID: 30881942
18. Martini K, Moon JW, Revel MP, Dangeard S, Ruan C, Chassagnon G. Optimization of acquisition
parameters for reduced-dose thoracic CT: A phantom study. Diagn Interv Imaging. 24 fe´ vr 2020. https://
doi.org/10.1016/j.diii.2020.01.012 PMID: 32107196
19. Debray M-P, Dauriat G, Khalil A, Leygnac S, Tubiana S, Grandjean A, et al. Diagnostic accuracy of low-
mA chest CT reconstructed with Model Based Iterative Reconstruction in the detection of early pleuro-
pulmonary complications following a lung transplantation. Eur Radiol. sept 2016; 26(9):3138–46.
https://doi.org/10.1007/s00330-015-4126-0 PMID: 26645864
20.
21.
Zhao W, Zhong Z, Xie X, Yu Q, Liu J. Relation between chest CT findings and clinical conditions of coro-
navirus disease (COVID-19) pneumonia: a multicenter study. Am J Roentgenol. 3 mars 2020;1–6.
https://doi.org/10.2214/AJR.20.22976 PMID: 32125873
Li K, Wu J, Wu F, Guo D, Chen L, Fang Z, et al. The clinical and chest CT features associated with
severe and critical COVID-19 pneumonia. Invest Radiol. 29 fe´ vr 2020. https://doi.org/10.1097/RLI.
0000000000000672 PMID: 32118615
PLOS ONE | https://doi.org/10.1371/journal.pone.0241407 November 3, 2020
11 / 12
PLOS ONE Low-dose chest CT scoring for SARS-CoV-2 pneumonia
22. Yang W, Yan F. Patients with RT-PCR confirmed COVID-19 and normal chest CT | Radiology [Internet].
2020 [cite´ 31 mars 2020]. Disponible sur: https://pubs.rsna.org/doi/10.1148/radiol.2020200702?url_
ver=Z39.88-2003&rfr_id=ori:rid:crossref.org&rfr_dat=cr_pub%3dpubmed
23. Ooi GC, Khong PL, Mu¨ ller NL, Yiu WC, Zhou LJ, Ho JCM, et al. Severe acute respiratory syndrome:
temporal lung changes at thin-section CT in 30 patients. Radiology. mars 2004; 230(3):836–44. https://
doi.org/10.1148/radiol.2303030853 PMID: 14990845
24. Dziedzic R, Rzyman W. Lobectomy versus segmentectomy and wedge resection in the treatment of
stage I non-small cell lung cancer. J Thorac Dis. mars 2018; 10(3):E234–5. https://doi.org/10.21037/jtd.
2018.03.30 PMID: 29708140
25. Nakada T, Noda Y, Kato D, Shibasaki T, Mori S, Asano H, et al. Risk factors and cancer recurrence
associated with postoperative complications after thoracoscopic lobectomy for clinical stage I non-small
cell lung cancer. Thorac Cancer. oct 2019; 10(10):1945–52. https://doi.org/10.1111/1759-7714.13173
PMID: 31436042
PLOS ONE | https://doi.org/10.1371/journal.pone.0241407 November 3, 2020
12 / 12
PLOS ONE |
10.1371_journal.pone.0238693 | RESEARCH ARTICLE
Correlates of older adult inpatients’ personal
care provision to people with functional
difficulties in Ghana
Kofi Awuviry-NewtonID
1*, Kwadwo Ofori-Dua2, Abraham Newton3
1 Priority Research Centre for Generational Health and Ageing, School of Medicine and Health, The
University of Newcastle, Newcastle, Australia, 2 Department of Sociology and Social Work, Kwame Nkrumah
University of Science and Technology, Kumasi, Ghana, 3 Department of Social Studies and Education,
University of Education, Winneba, Ghana
* newscous@gmail.com
Abstract
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Introduction
Citation: Awuviry-Newton K, Ofori-Dua K, Newton
A (2020) Correlates of older adult inpatients’
personal care provision to people with functional
difficulties in Ghana. PLoS ONE 15(10): e0238693.
https://doi.org/10.1371/journal.pone.0238693
Editor: Ellen L. Idler, Emory University, School of
Public Health, UNITED STATES
Received: October 15, 2019
Accepted: August 21, 2020
Published: October 2, 2020
Peer Review History: PLOS recognizes the
benefits of transparency in the peer review
process; therefore, we enable the publication of
all of the content of peer review and author
responses alongside final, published articles. The
editorial history of this article is available here:
https://doi.org/10.1371/journal.pone.0238693
Copyright: © 2020 Awuviry-Newton et al. This is an
open access article distributed under the terms of
the Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: All relevant data are
within the manuscript and its Supporting
Information files.
Funding: This study was funded by The University
of Newcastle International Postgraduate Research
Most research depicts older adults as needing personal care, with limited research on older
adults’ contributions in the lives of others in developing countries like Ghana. The purpose
of the study was to examine the personal care provision and socio-demographic correlates
of personal care provision by older adult inpatients in Ghana.
Materials and methods
A hospital-based survey was conducted among 400 consecutively surveyed older adult
inpatients at Komfo Anokye Teaching Hospital in southern Ghana. Stata 15 was used to
analyse the data. The relationship between personal care and older adult inpatients’ socio-
demographic characteristics were analysed using the chi-square test. Multivariate logistic
regression analysis was employed.
Results
Overall, 28% of older adult inpatients provided personal care. Participants were mostly
females, married or cohabiting, completed at most junior high school, Christians, urban resi-
dents, not working and living with their immediate family. Primarily, most participants pro-
vided personal care to one person, once a week, and one-hour duration. Nearly three-
quarters of participants provided personal care to someone who lives with them. After
adjustment, male older adults were 50% less likely to and urban residents were 83% more
likely to provide personal care. Being single, separated or divorced was statistically signifi-
cantly associated with personal care provision, however, were not statistically significant
after adjusting for sex and residence. A post hoc analysis testing for interaction revealed no
relationship existing between sex and marital status concerning personal care provision (p =
0.106).
PLOS ONE | https://doi.org/10.1371/journal.pone.0238693 October 2, 2020
1 / 12
PLOS ONE Older adults as community resources: Understanding their contributions
Scholarship (UNIPRS) and The University of
Newcastle Research Scholarship Central 50:50
(UNRSC50:50). This research was supported by
the Australian Research Council Centre of
Excellence in Population Ageing Research (project
number CE170100005).
Competing interests: No author have competing
interest
Conclusion
Female and urban resident older adult inpatients in Ghana are not just passive receivers of
care but also provide personal care to others with functional difficulties, independent on age.
It further draws attention to the need for policies and programs that can support older adults,
particularly females and urban residents, to be productive in the later life.
Introduction
The older population in Ghana, those aged 60 years or older, is growing rapidly due to decreas-
ing birth rates and increased longevity [1]. Ghana’s older population increased more than
seven-fold from 213,477 (4.5%) in 1960 to 1,643,381 (6.7%) in 2010 [2]. The percentage of
older adults is further expected to increase to 9.6% by 2050 [3]. The growth in Ghana’s propor-
tion of older adults has sparked an increase in research depicting older adults as dependent on
others for personal care [4–12].
Several authors have defined personal care with different emphasis. Shiel [13] defines it as
assistance in the basic activities a person may need to achieve the best out of life. Personal care
includes assistance in bathing, grooming, toileting and medication [13]. It also involves instru-
mental activities such as managing money, household chores shopping, telephone use and pre-
paring meals [9,14], as well as advice, informational, psychological and emotional assistance in
dealing with life challenges [15,16]. A lot of research on personal care for older adults in
Ghana has been fueled by evidence in Ghana that some older adults experience some levels of
functional limitation, participation restrictions in voluntary work and therefore require the
support of others [4,10–12,17–19]. Whiles, it is relevant for research to focus on the provision
of personal care for older adults in Ghana; a more positive perspective is that many older
adults may be productive irrespective of their poor health. These older adults experiencing
functional limitations can contribute to the wellbeing of others, particularly those experiencing
difficulties meeting their daily needs in society by providing personal care such as assistance
with toileting. It is in line with this that a shift in Gerontological research from total depen-
dence to productive ageing gained prominence in the 1980s [20], and it is still relevant today.
Productive ageing refers to a state where older adults use their knowledge, experiences and
skills to consistently contribute to their family and community to increase their self-worth and
value [21]. It can take varying forms including engaging in either paid or unpaid work; provid-
ing care for significant others including family, relatives and friends; volunteering; and pursu-
ing education aimed to enhance their human capital and productivity [21].
In support of this perspective is a study that examined the profiles of caregivers reporting
the mean age of caregivers as 61 years [22], depicting that caregivers in Ghana were mostly
older adults. These caregivers were providing personal care, emotional and financial assistance
to their dependents. In sub-Saharan African countries, studies abound about older adults’
involvement in the caring of their grandchildren because of the changing household structure.
A study that used data from demographic and health surveys to examine the composition of
household structure with older adults from 24 sub-Saharan African countries reported that
older adults living with grandchildren with no adult children are a common phenomenon.
Although most of the adult children live elsewhere, about 8% of grandchildren have lost one
parent. Moreover, in countries with a high prevalence of Human Immunodeficiency Virus
(HIV) mortality, older adults are more likely to live with grandchildren who have lost both
parents to death [23]. In this family structure, older adults may take the role of parents respon-
sible for childcare and personal care for their grandchildren.
PLOS ONE | https://doi.org/10.1371/journal.pone.0238693 October 2, 2020
2 / 12
PLOS ONE Older adults as community resources: Understanding their contributions
Several reasons exist on why people, including older adults receiving healthcare themselves,
may provide personal care to those who need it. According to Schulz [24], people provide care
for an expectation of rewards ranging from gaining social approval to fear of disapproval for
not providing care. Another explanation is based on altruistic motivation suggesting that the
primary aim of caring is sacrificing self-desires to helping the cared-for achieve his or her goal
[24,25]. According to Schulz [24], the benefit to the caregiver is just a consequence of provid-
ing care. Others may also offer care for reciprocity reasons. People may give care to others as a
payback. A longitudinal study conducted in Amsterdam reported that older adult’s assistance
in childcare for grandchildren for their adult children determine support they receive later in
life. This assistance, however, varies according to the gender of their adult children. Older
adults who provide care for their sons’ children are more likely to receive instrumental and
emotional support from their sons. Providing care for their daughters’ children does not pay
off [26]. Moreover, activity theory suggests that maintenance of activity in old age compensates
for role-loss and promotes moral wellness and thus should be encouraged. As a result, provid-
ing care to others promote the health of both the cared-for and the caregiver [27].
Studies on the profiles of older adults who assist people with functional difficulties exist in
advanced countries. In China, for instance, variation exists among older adults living in urban
and rural areas in terms of the amount and kind of assistance they provide. In the study, 41%
of older adults living in the urban area provide care to grandchildren compared to 31% by
rural-dwellers. In terms of housework, rural older dwellers participate more than urban dwell-
ers (32% vs 42%) [28]. Moreover, older adults, mostly women aged 60–64, used more hours to
provide unpaid assistance to either family members or friend [29].
In Ghana, the socio-demographic characteristics of older adults, particularly those hospital-
ised who provide personal care to those with functional difficulties, the extent of care provision
and beneficiaries are unknown. The attributes of caregivers revealed by Sanuade and Boatemaa
[22] included those in the middle ages (50–59 years) and population from the community
obscuring our understanding of those older adults inpatients who provide personal care.
Awareness of the socio-demographic correlates, the extent of care provision and beneficiaries
of older adult inpatients’ personal care they offer may provide evidence for further research
and policymakers for enacting policies and programs that can support older adults in their
quest to be productive in their later lives. The present study seeks to fill the gap in knowledge
by addressing these research questions: What are the socio-demographic characteristics of
older adults inpatients who provide personal care? To what extent do older adults’ inpatients
provide personal care? Who benefits from the personal care older adults provide? What are
the correlates of older adults inpatients who provide personal care?
Material and methods
Study area
The survey was conducted in Komfo Anokye Teaching Hospital (KATH) located in Kumasi,
the Regional Capital of Ashanti Region with a population of 4,780,249 as of 2010 [2]. Ashanti
Region shares boundaries with some regions of Ghana including Brong Ahafo Region to the
north, Eastern Region to the east, and Central Region to the south and Western Region to the
west [30]. KATH is accessible to a wide range of people from 13 of the 16 administrative
regions of Ghana. The hospital receives referrals from these regions. KATH provides medical
services to the majority of older adults.
Study design
Survey descriptive design was employed.
PLOS ONE | https://doi.org/10.1371/journal.pone.0238693 October 2, 2020
3 / 12
PLOS ONE Older adults as community resources: Understanding their contributions
Study sampling
This study is a portion of the primary author’s PhD research conducted at KATH to examine
the functional difficulties among older adult inpatients. As part of the objectives of the overall
PhD study, participants’ involvement in personal care for others in their various homes before
their admission to the hospital were also assessed. The population of the present survey was
older adults (60 years or older) admitted to the Komfo Anokye Teaching Hospital within the
defined sampling period. Random sampling technique was used to select four days per week to
recruit participants. All older adults admitted to the hospital at the time of the study were eligi-
ble to participate if they met the inclusion criteria. The inclusion criteria were; 1) They must be
older adults (60 years and over), 2) They must be admitted to the hospital for any health rea-
son, 3) They must have stayed a minimum of one night, 4) They must not be seriously sick
and, 5) They must be willing to participate in the study. On the other hand, older adults were
excluded if 1) older adults aged 60 years and over but whose health status is critical and, 2)
older adults who did not express willingness to take part in the study [9]. The inclusion criteria
were purposeful to determine whether older adults who receive care were also providers of
care. However, no prior research identified employing this methodology in the recruitment of
participants.
The study used consecutive sampling techniques to recruit individuals who met the eligibil-
ity criteria. The study recruited 400 participants during the data collection period. All data was
collected at the hospital in the time and day chosen by the participants.
Before recruitment, nurses were asked to seek the consent of older adult inpatients who sat-
isfy the inclusion criteria. The researcher allowed 48 hours for older adult inpatients to give
their consent. When a patient agreed, the researcher and a research assistant approached the
patient and conducted the structured interviews. The researcher read and translated for partic-
ipants who could not read and write. All communication was conducted in “Twi”, the domi-
nant language of Ghana. The detail of the methodology is explained in Awuviry-Newton et al.
[9].
Data collection
A survey questionnaire was used to collect information from older adults inpatients. The
researcher read the questions to participants who could not read to solicit quantitative data.
The structured interview was composed of closed-ended questions. Data collection with the
participants were completed between August and December 2018.
Dependent variable: Personal care
Personal care provision was assessed among older adults using a variable with a nominal
response category. The question asked was, “Do you regularly provide care or assistance (e.g.
bathing, transport) to any other person because of their long-term illness, disability or frailty
before you were hospitalised?” The original nominal response categories were; 1) yes, for
someone who lives with me, 2) yes, for someone who lives elsewhere and 3) no, I do not pro-
vide care. For this analysis the response categories; yes, I do provide care for someone who
lives with me; and, yes, I do provide care for someone who lives elsewhere were classified as
“yes” and no, I do not provide care was classified as “no”. This categorisation is necessary to
determine those who provide care and those who do not rather than to identify to whom care
is provided as elsewhere [12]. “Yes” was assigned “1” whereas, “No was assigned “0” for multi-
variate logistic regression.
The extent by which older adults provide personal care has been provided. This variable
measuring the extent of personal care provision included; 1) the number of care recipients, 2)
PLOS ONE | https://doi.org/10.1371/journal.pone.0238693 October 2, 2020
4 / 12
PLOS ONE Older adults as community resources: Understanding their contributions
the number of times older adults provide care, and 3) the number of hours used in providing
personal care. Moreover, the pictorial description of the beneficiaries of personal care provided
by older adults has been presented.
Independent variables
The variables included in the analysis were age, sex, marital status, education, residence,
employment status, living arrangement, and religion. Age was measured as a continuous vari-
able. Marital status was categorised as never married, married or cohabiting, separated or
divorced and widowed. For the purpose of analysis, marital status was categorised into three
responses as single/separated/divorced, married/cohabiting and widowed. Education was cate-
gorised as no education, less than primary school, primary school, secondary school, high
school, and college/pre-university, undergraduate and postgraduate degree. Education
response categories were further categorised as no education, at maximum junior high com-
pleted and at least senior high completed. A residence was measured as living in either rural or
urban areas. Concerning the living arrangement, this variable was initially measured as
“alone”, “with husband/wife”, “with children/with husband/wife and children”, and “extended
family”. The responses for living arrangement were then categorised as “alone” for alone, “liv-
ing as a couple” for living with husband/wife, and “with partner and children” for with chil-
dren/with husband/wife and children, and extended family. Employment status was measured
as “currently working” and “currently not working”.
Data analysis
All analysis was facilitated by Stata 15. Descriptive statistics were used to describe essential var-
iables in the study population. Chi-square test was used to determine the relationship between
independent variables and the dependent variable. Univariate and multivariable logistic
regression was performed to assess any significant relationship between independent variables
and the dependent variable. Crude and adjusted odds ratios were used to ascertain any associa-
tions between the dependent and independent variables using a 95% confidence interval to
determine the level of significance. Any significant association was determined at a p-value of
less than 0.05 in the multivariable logistic regression model to control for potential confound-
ing variables. Logistic regression analysis was conducted to determine the correlates of per-
sonal care provision among older adults inpatients in Ghana.
Ethical consideration
Ethical approval for this analysis was obtained from The University of Newcastle (H-2018-
0163), New South Wales, Australia and Kwame Nkrumah University of Science and Technol-
ogy (CHRPE/AP/112/18), in Ghana in keeping with the Declaration of Helsinki. Written
Informed consent was obtained from the director of Komfo Anokye Teaching Hospital and
study participants. Anonymity and confidentiality were ensured.
Results
Twenty-eight per cent of participants provided care. The average age of the 400 participants
was approximately 71 years, and most were females (51.0%). Majority of participants were
married or cohabiting (53.0%), completed at least junior high school (52.3%), mostly Chris-
tians (82.8%) and living in the rural area (56.8%). Moreover, most participants were living
with their immediate family (62.5%) and were not working (61.0%) (see Table 1).
PLOS ONE | https://doi.org/10.1371/journal.pone.0238693 October 2, 2020
5 / 12
PLOS ONE Older adults as community resources: Understanding their contributions
Table 1. Socio-demographic characteristics of participants.
Demographic Characteristics (N = 400)
Age (mean, SD)
Sex
Male
Female
Marital status
Single/separated/divorced
Currently married/cohabiting
Widowed
Education
No education
At maximum junior high completed
At least senior high completed
Religion
None
Christianity
Islam
Residence
Rural
Urban
Living arrangement
Alone
With couple
With couple and children
Employment status
Currently working
Currently not working
Provide personal care?
Yes
No
N (%)
71.3±8.42
196 (49.0
204 (51.0)
58 (14.5)
212 (53.0)
130 (32.5)
128 (32.0)
209 (52.3)
63 (15.8)
27 (6.75)
331 (82.8)
42 (10.0)
227 (56.8)
173 (43.3)
50 (12.5)
100 (25.0)
250 (62.5)
156 (39.0)
244 (61.0)
112 (28)
288 (72)
https://doi.org/10.1371/journal.pone.0238693.t001
Older adult inpatients who provide personal care
The univariate analysis shows that the average age of participants who provide personal care was
approximately 72 years. Participants who provided personal care were mostly females (60.7%),
and married or cohabiting (50%). Furthermore, more than half had completed at most junior
high (58.0%), mostly Christians (80.4%), living in the urban areas (55.4%), with the majority liv-
ing with their immediate family (61.6%) and mostly not working (66.1%) (see Table 2).
The extent by which older adult inpatients provide personal care to
others
Most of the personal care was provided to one person (15.3%), often offered once a week, once
every few weeks (12.0%) and usually lasted for about an hour (6.75%) (see Table 3).
Who receives the personal care older adult inpatients provide?
Nearly three-quarters of older adults offers personal care to someone who lives with them
(73.2%) (see Fig 1).
PLOS ONE | https://doi.org/10.1371/journal.pone.0238693 October 2, 2020
6 / 12
PLOS ONE Table 2. Univariate analysis of older adult inpatients who provide personal care.
Older adults as community resources: Understanding their contributions
Characteristics
Age
Sex
Male
Female
Marital status
Single/separated/divorced
Married/cohabiting
Widowed
Education
No education
At most junior high completed
At least senior high completed
Religion
None
Christianity
Islam
Residence
Rural
Urban
Living arrangement
Alone
With couple
With couple and children
Employment status
Currently working
Currently not working
Significant at �p-value < 0.05; ��p-value < 0.01.
https://doi.org/10.1371/journal.pone.0238693.t002
Personal Care
Total N (%)
71.3±8.42
196 (100)
204 (100)
58 (100)
212 (100)
130 (100)
128 (100)
209 (100)
63 (100)
27 (100)
331 (100)
42 (100)
227 (100)
173 (100)
50 (100)
100 (100)
250 (100)
156 (100)
244 (100)
Yes N (%)
71.9±8.91
44 (22.5)
68 (33.3)
26 (44.8)
56 (26.4)
30 (23.1)
34 (26.6)
65 (31.1)
13 (20.6)
8 (29.6)
90 (27.2)
14 (33.3)
50 (22.0)
62 (35.8)
13 (26.0)
30 (30.0)
69 (27.0)
38 (24.4)
74 (30.3)
No N (%)
71.0±8.23
152 (77.6)
136 (66.7)
32 (55.2)
156 (73.6)
100 (76.9)
94 (73.4)
144 (68.9)
50 (79.4)
19 (70.4)
241 (72.8)
28 (66.7)
177 (78.0)
111 (64.2)
37 (74.0)
70 (70.0)
181 (72.4)
118 (75.6)
170 (69.7)
p-value
0.370
0.016
<0.01
0.244
0.692
<0.01
0.850
0.195
Correlates of personal care provision by older adults inpatients
After adjustment, the multivariate analysis revealed that male older adults were 50% less likely
to provide personal care compared to female older adults (AOR (95% C.I): 0.50 (0.31, 0.81)).
Compared to the unadjusted odds ratio, the odds ratio reduced by 8%, when adjusted for mari-
tal status and residential location. Being single, separated or divorced was statistically signifi-
cantly associated with personal care provision (COR (95% C.I): 2.26 (1.24, 4.13), however,
became dependently statistically not significant after adjusting for sex and residence. Concern-
ing the widow marital status, it was independently not significantly related to the personal care
provision before and after adjustment (see Table 4).
Living in an urban area was independently statistically significantly associated with per-
sonal care provision. Being an urban resident was 83% more likely to provide personal care to
someone with functional difficulties (AOR (95% C.I): 1.83 (1.16, 2.89)) compared to rural
dwellers. There was a 15% reduction in the odds ratio when adjusted for sex and marital status.
A post hoc analysis to test for interaction revealed no relationship existing between sex and
marital status related to personal care provision (p = 0.106).
PLOS ONE | https://doi.org/10.1371/journal.pone.0238693 October 2, 2020
7 / 12
PLOS ONE Older adults as community resources: Understanding their contributions
Table 3. Extent of which older adults provides personal care.
The extent of personal care provision (N = 112)
Number (N)
Percentage (%)
Number of people you regularly care
None
One person
Two people
More than two people
Often assistance is provided
None
Every day, several times a week
Once a week, once every few weeks
Less often
Number of times spent on care
None
All day and night
All night
All day
Several hours
About an hour
https://doi.org/10.1371/journal.pone.0238693.t003
Discussion
288
61
32
19
288
30
48
34
288
24
23
18
20
27
72.0
15.3
8.00
4.75
72.0
7.50
12.0
8.50
72.0
6.00
5.75
4.50
5.00
6.75
Our findings indicate that older adults inpatients contribute to the wellbeing of others through
the provision of personal care. The finding by this current study that older adults provide per-
sonal care corroborates with the finding of a study conducted in the UK that older adults
Fig 1. Who receives the personal care older adults provide?
https://doi.org/10.1371/journal.pone.0238693.g001
PLOS ONE | https://doi.org/10.1371/journal.pone.0238693 October 2, 2020
8 / 12
PLOS ONE Older adults as community resources: Understanding their contributions
Table 4. Multivariate analysis of the correlates of personal care provided by older adult inpatients.
Variables
Sex
Male
Female
Marital status
Single/separated/divorced
Currently Married/cohabiting
Widowed
Residence
Rural
Urban
COR 95%CI
AOR 95%CI
0.58 (0.37, 0.90)�
0.50 (0.31, 0.81) ��
1
1
2.26 (1.24, 4.13)��
1
0.84 (0.50, 1.39)
1
1.98 (1.27, 3.08)��
1.83 (0.98, 3.41)
1
0.65 (0.38, 1.12)
1
1.83 (1.16, 1.2.89)��
Significant at �p-value < 0.05
��p-value < 0.01.
https://doi.org/10.1371/journal.pone.0238693.t004
provide a huge contribution to society [31]. This affirmation concludes that older adults are
active providers of care irrespective of their poor health and therefore, government institu-
tions, friends, or families should nurture this potential. The smaller proportion of the older
adults contributions compared to the study in the UK could be that the present study sampled
older adult inpatients. In contrast, the UK study sampled community-dwelling older adults.
After adjustment and controlling for confounding, female older adults and urban-dwelling
older adults are more likely to provide personal care to someone who needed it.
The finding that female older adults provide personal care could be understood in the cul-
tural context throughout the life course in Ghanaian society. Culturally and traditionally,
women are more engaged in care, including personal care whereas men are breadwinners and
decision-makers in the family [6]. In the old age, women find their identity unchanged and so
are more inclined into caring for people needing care compared to older men. This finding
corroborates other findings that older women provide more assistance to family members or
friend [29], which may include personal care. A significant concern is how people with func-
tional difficulties will be cared for, in this era where women in Ghana are attaining higher edu-
cation and migrating following urbanization and modernization. We recommend that health
and welfare service that can support older adults’ especially older women to be productive and
contribute to wellbeing of others as far as possible will be helpful.
The previous study conducted in China reported that most of the older adults who provide
care to grandchildren are urban dwellers [28]. Consistent with this previous study, the current
study report that older adult urban dwellers are 16% more likely to provide personal care than
older adults’ rural dwellers. This practice is common because of the existence of at least three-
generational households in Ghana [32]. Most of the time, adult children in the urban cities
invite their parents to live with them in the cities and by so doing; older adults may be provid-
ing personal care to a person with functional difficulties or providing childcare.
Age was not significantly associated with personal care provision in this study. This finding
is contrary to previous mainstream studies which report an association between age and per-
sonal care [27,33]. The variation in findings may be attributed to the study settings. In this cur-
rent study, older adult inpatients were recruited from a hospital setting, where all participants
irrespective of the age may have been experiencing poor health. However, the current study
asked participants to reflect on care provision at home, before they were admitted to the hospi-
tal. Further research recruiting community-dwelling older adults in Ghana on personal care
provision is relevant to understanding any variation that may be existing. This research will
PLOS ONE | https://doi.org/10.1371/journal.pone.0238693 October 2, 2020
9 / 12
PLOS ONE Older adults as community resources: Understanding their contributions
help in understanding the health and social needs of these older adults who provide care and
support.
As evident in the current study, participants provided personal care to one person, once a
week, and mostly on-hour duration. Although participants mostly offer care for one hour, the
percentages of the length of categories of care provision are quite similar, implying that older
adults could provide care to any length. Moreover, the finding that most participants provided
personal care to someone who lives with them infers that participants provide most of the per-
sonal care to either their spouse, own children or family members who may stay with partici-
pants due to their illness. The number of time participants put in caregiving and how often
demonstrate their interest in enhancing the wellbeing of others. Participants may also be
improving their emotional wellbeing and happiness as posited by the activity theory [27]. Due
to the effort older adults put in to contribute to the wellbeing of others, we suggest that there
should be community support for older adults in Ghana to nurture their potentials, especially
when older adults inpatients are discharged from the hospital setting.
The findings from this current study have implications for research, policy and practice. In
terms of research, the health and social care needs of older adults who provide personal care
should be explored. Research is also needed to understand older adults ‘contributions in vol-
unteering services and child care to complement the existing evidence in this study. Moreover,
more research needs to explore how the social and physical environment can nurture the con-
tributions of older adults in the Ghanaian setting. This understanding is essential to facilitate
practices to ensure productive ageing. Concerning policy, the existing policy on older adults
developed in Ghana in 2010 [34] should be implemented to help Ghana amass the unique con-
tributions of older adults in various avenues. Specifically, the policy seeks to reduce poverty
and improve healthcare among older adults; however, evidence in Ghana shows that some
older adults do not have good health and feel neglect owing to the deprivation from their own
family as well as the government institutions [10]. In terms of practice, health and social care
professionals should promote older adults ability to be productive through healthcare and
counselling. The finding in this study offers social workers the opportunity to capitalise on
older adults’ contribution to increase older adults’ economic wellbeing to reduce the impact of
poverty on the overall functioning of older adults.
This study has some strengths that need to be acknowledged. First, this study using older
adult inpatients from the hospital and considering the finding that older adults contribute to
the wellbeing of others echoes our understanding that older adults are not just passive receiv-
ers but also caregivers. Sampling participants from one hospital may have missed other partici-
pants who did not come to the hospital; however, the hospital receives patients from about 13
regions of the 16 regions in Ghana. Moreover, probability sampling was employed to recruit
participants and may have reduced some bias. Another potential limitation for this study is
that though the activity theory applied in this study is of the assumption that the healthy older
adults provide care; however, the current study sampled hospitalised patients. Although, the
literature search did not identify any study on the health status of older adults’ caregivers, the
current study provides an opportunity for new research exploring the health status of older
adults who care for others. A study sampling a large sample size of older adults’ caregivers will
help to understand their health and social care needs.
Conclusion
Though a small proportion of older adults receiving health care from the hospital setting pro-
vide personal care, the findings in this study demonstrate that older adults in the community
may be resourceful to the community and their family members due to their relatively better
PLOS ONE | https://doi.org/10.1371/journal.pone.0238693 October 2, 2020
10 / 12
PLOS ONE Older adults as community resources: Understanding their contributions
health. Moreover, the study’s findings draw attention to the need to provide financial and
health support to older adults receiving healthcare who do not provide personal care to enable
them to be productive. In this period where Ghana’s population is reportedly growing old, the
proportion and number of people aged 60 years or older will increase. These populations are
valuable resources to their families, communities and the country as a whole. The health and
social care needs of all older adults should be ensured so that older adults can remain produc-
tive as far as possible. The findings have drawn attention to the policy and research initiatives
to facilitate productive ageing.
Supporting information
S1 Appendix.
(DOCX)
Acknowledgments
Thanks to the participants for their time in participating in the survey.
Author Contributions
Conceptualization: Kofi Awuviry-Newton.
Formal analysis: Kofi Awuviry-Newton.
Methodology: Kofi Awuviry-Newton, Kwadwo Ofori-Dua, Abraham Newton.
Software: Kofi Awuviry-Newton.
Supervision: Kwadwo Ofori-Dua, Abraham Newton.
Visualization: Kofi Awuviry-Newton.
Writing – original draft: Kofi Awuviry-Newton.
Writing – review & editing: Kofi Awuviry-Newton, Kwadwo Ofori-Dua, Abraham Newton.
References
1. Biritwum RB, Mensah G, Minicuci N, Yawson AE, Naidoo N, Chatterji S, et al. Household characteristics
for older adults and study background from SAGE Ghana Wave 1. Global Health Action. 2013;
6:20096. https://doi.org/10.3402/gha.v6i0.20096 PMID: 23759325
2. Population and Housing Census Report [Internet]. Ghana Statistical Service, Accra. 2013 [cited 26
March, 2017]. Available from: www.statsghana.gov.gh/docfiles/2010phc/National_Analytical_Report.
pdf.
3. Global AgeWatch Index 2015 [Internet]. 2015 [cited 16 January, 2017]. Available from: http://www.
helpage.org/global-agewatch/population-ageing-data/country-ageing-data/?country=Ghana.
4. Awuviry-Newton K, Byles J, Tavener M. Evolution of familial care and support for older people in
Ghana: A literature review. Emerging Researchers in Ageing; Western Australia: Emerging Research-
ers in Ageing; 2017. p. 1–11.
5. Brown CK. Caring for the elderly: perspectives from Ghana and Japan. Cape Coast: Catholic Mission
Press; 1999.
6. Nukunya GK. Tradition and Change in Ghana: An Introduction to Sociology. Accra: University Press;
2003.
7. Ofori-Dua K. Extended Family Support and Elderly Care in Bamang, Ashanti Region of Ghana: Univer-
sity of Ghana; 2014.
8. Sackey H.A. Female labour force participation in Ghana: The effects of education. 2005.
9. Awuviry-Newton K, Wales K, Tavener M, Byles J. Do factors across the World Health Organisation’s
International Classification of Functioning, Disability and Health framework relate to caregiver
PLOS ONE | https://doi.org/10.1371/journal.pone.0238693 October 2, 2020
11 / 12
PLOS ONE Older adults as community resources: Understanding their contributions
availability for community-dwelling older adults in Ghana? PloS One. 2020; 15(5):e0233541. https://doi.
org/10.1371/journal.pone.0233541 PMID: 32469915
10. Awuviry-Newton K, Nkansah JO, Ofori-Dua K. Attributions of elder neglect: A phenomenological study
of older people in Ghana. Health Soc Care Community. 2020;n/a(n/a).
11. Awuviry-Newton K, Tavener M, Wales K, Byles J. Interpretative Phenomenological Analysis of the
Lived Experiences of Older Adults Regarding Their Functional Activities in Ghana. Journal of Primary
Care & Community Health. 2020; 11:2150132720931110.
12. Awuviry-Newton K, Tavener M, Wales K, Kowal P, Byles J. Activities of daily living difficulties and toilet-
ing among older Ghanaians: an application of WHO-ICF framework. Innovation in Aging. 2019; 3(Sup-
plement_1): S520-S.
13. Shiel W. Common Medical Abbreviations List (Acronyms and Definitions): MedicineNet; 2018 [Available
from https://www.medicinenet.com/common_medical_abbreviations_and_terms/article.htm.
14. Roberts CE, Phillips LH, Cooper CL, Gray S, Allan JL. Effect of different types of physical activity on
activities of daily living in older adults: systematic review and meta-analysis. Journal of Aging and Physi-
cal Activity. 2017; 25(4):653–70. https://doi.org/10.1123/japa.2016-0201 PMID: 28181837
15. Mui A, Shibusawa T. Asian American elders in the 21st century: Key indicators of psychosocial well-
being. New York: Columbia University Press; 2008.
16. Mui AC. Productive ageing in China: a human capital perspective. China Journal of Social Work. 2010;
3(2–3):111–23.
17. Amosun S, Nyante G, Wiredu E. Perceived and experienced restrictions in participation and autonomy
among adult survivors of stroke in Ghana. African Health Sciences. 2013; 13(1):24–31. https://doi.org/
10.4314/ahs.v13i1.4 PMID: 23658564
18. Biritwum R, Minicuci N, Yawson A, Theou O, Mensah G, Naidoo N, et al. Prevalence of and factors
associated with frailty and disability in older adults from China, Ghana, India, Mexico, Russia and South
Africa. Maturitas. 2016; 91:8–18. https://doi.org/10.1016/j.maturitas.2016.05.012 PMID: 27451316
19. Debpuur C, Welaga P, Wak G, Hodgson A. Self-reported health and functional limitations among older
people in the Kassena-Nankana District, Ghana. Global Health Action. 2010; 3(1):2151.
20. Butler RN, Gleason HP. Productive aging: Springer Pub. Co.; 1985.
21. Morrow-Howell N, Hinterlong J, Sherraden M. Productive aging: Concepts and challenges: JHU Press;
2001.
22. Sanuade O, Boatemaa S. Caregiver profiles and determinants of caregiving burden in Ghana. Public
Health. 2015; 129(7):941–7. https://doi.org/10.1016/j.puhe.2015.05.016 PMID: 26115592
23.
Zimmer Z, Dayton J. Older adults in sub-Saharan Africa living with children and grandchildren. Popula-
tion studies. 2005; 59(3):295–312. https://doi.org/10.1080/00324720500212255 PMID: 16249151
24. Schulz R. Theoretical perspectives on caregiving: Concepts, variables, and methods. 1990.
25. Noddings N. Caring: A Feminine Approach to Moral Education. Berkeley: University of California
Press.; 1984.
26. Geurts T, Poortman AR, Van Tilburg TG. Older parents providing child care for adult children: Does it
pay off? Journal of Marriage and Family. 2012; 74(2):239–50.
27. Cnaan RA, Cwikel J. Elderly volunteers: Assessing their potential as an untapped resource. Journal of
Aging & Social Policy. 1992; 4(1–2):125–47.
28.
Zhang K. Study on social activity and mental state of the elderly in urban/rural China. Beijing: China
Social Sciences Press.; 2009.
29. Sun J. Chinese older adults taking care of grandchildren: practices and policies for productive aging.
Ageing International. 2013; 38(1):58–70.
30.
31.
2010 Population & Housing Census. [Internet]. 2014 [cited 6 June, 2017]. Available from: www.
statsghana.gov.gh/docfiles/2010_District_Report/Ashanti/KMA.pdf.
Fenton S-J, Draper H. Older people make a huge contribution to society. Some communities and faith
groups draw on this contribution in responding to the needs of all their members. Birmingham Policy
Commission. 2014.
32. Biritwum RB, Mensah G, Minicuci N, Yawson AE, Naidoo N, Chatterji S, et al. Household characteristics
for older adults and study background from SAGE Ghana Wave 1. Global Health Action. 2013; 6(1):20096.
33. Kincade JE, Rabiner DJ, Bernard SL, Woomert A, Konrad TR, Defriese GH, et al. Older adults as a
community resource: Results from the national survey of self-care and aging. The Gerontologist. 1996;
36(4):474–82. https://doi.org/10.1093/geront/36.4.474 PMID: 8771975
34. Government of Ghana. National Ageing Policy: Ageing with Security and Dignity. Ministry of Employ-
ment and Social Welfare, Ghana; 2010.
PLOS ONE | https://doi.org/10.1371/journal.pone.0238693 October 2, 2020
12 / 12
PLOS ONE |
10.1371_journal.pone.0237585 | RESEARCH ARTICLE
Telemedicine in the OECD: An umbrella
review of clinical and cost-effectiveness,
patient experience and implementation
Nkiruka D. EzeID
1*, Ce´ u Mateus1, Tiago Cravo Oliveira Hashiguchi2
1 Division of Health Research, Health Economics at Lancaster, Lancaster University, Lancaster, United
Kingdom, 2 Health Division Organisation for Economic Co-operation and Development, Directorate for
Employment, Labour and Social Affairs, Paris, France
* nkiru.eze@outlook.com
Abstract
Introduction
Patients and policy makers alike have high expectations for the use of digital technologies
as tools to improve health care service quality at a sustainable cost. Many countries within
the Organisation for Economic Co-operation and Development (OECD) are investing in tele-
medicine initiatives, and a large and growing body of peer-reviewed studies on the topic has
developed, as a consequence. Nonetheless, telemedicine is still not used at scale within the
OECD. Seeking to provide a snapshot of the evidence on the use of telemedicine in the
OECD, this umbrella review of systematic reviews summarizes findings on four areas of pol-
icy relevance: clinical and cost-effectiveness, patient experience, and implementation.
Methods
This review followed a prior written, unregistered protocol. Four databases (PubMed/Med-
line, CRD, and Cochrane Library) were searched for systematic reviews or meta-analyses
published between January 2014 and February 2019. Based on the inclusion criteria, 98
systematic reviews were selected for analysis. Due to substantial heterogeneity, a meta-
analysis was not conducted. The quality of included reviews was assessed using the
AMSTAR 2 tool.
Results
Most reviews (n = 53) focused on effectiveness, followed by cost-effectiveness (n = 18),
implementation (n = 17) and patient experience (n = 15). Eighty-three percent of clinical
effectiveness reviews found telemedicine at least as effective as face-to-face care, and
thirty-nine percent of cost-effectivenss reviews found telemedicine to be cost saving or cost-
effective. Patients reported high acceptance of telemedicine and the most common barriers
to implementation were usability and lack of reimbursement. However, the methodological
quality of most reviews was low to critically low which limits generalizability and applicability
of findings.
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: Eze ND, Mateus C, Cravo Oliveira
Hashiguchi T (2020) Telemedicine in the OECD: An
umbrella review of clinical and cost-effectiveness,
patient experience and implementation. PLoS ONE
15(8): e0237585. https://doi.org/10.1371/journal.
pone.0237585
Editor: Hannah E. Carter, Queensland University of
Technology, AUSTRALIA
Received: January 27, 2020
Accepted: July 29, 2020
Published: August 13, 2020
Copyright: © 2020 Eze et al. This is an open access
article distributed under the terms of the Creative
Commons Attribution License, which permits
unrestricted use, distribution, and reproduction in
any medium, provided the original author and
source are credited.
Data Availability Statement: All relevant data are
within the manuscript and its Supporting
Information files.
Funding: No funding was received for this study.
Competing interests: We have no competing
interests to declare.
PLOS ONE | https://doi.org/10.1371/journal.pone.0237585 August 13, 2020
1 / 24
PLOS ONE Telemedicine in the OECD: An umbrella review
Conclusion
This umbrella review finds that telemedicine interventions can improve glycemic control in
diabetic patients; reduce mortality and hospitalization due to chronic heart failure; help
patients manage pain and increase their physical activity; improve mental health, diet quality
and nutrition; and reduce exacerbations associated with respiratory diseases like asthma. In
certain disease and specialty areas, telemedicine may be a less effective way to deliver
care. While there is evidence that telemedicine can be cost-effective, generalizability is hin-
dered by poor quality and reporting standards. This umbrella review also finds that patients
report high levels of acceptance and satisfaction with telemedicine interventions, but that
important barriers to wider use remain.
Introduction
In the last decade, several factors have contributed to a growth in the use of telemedicine.
These factors include significant advances in information and communication technology, the
prevalent use of high speed internet, and the increasing adoption of electronic health records.
More recently, the SARS-CoV-2 pandemic has significantly accelerated telemedicine adoption
and use, as patients and health care workers (HCWs) alike seek new ways to stay healthy with
minimal physical contact. Telemedicine is used to encourage self-care through remote and
chronic disease monitoring, to provide consultations to patients who are unable to attend in-
person (face-to-face) appointments, and within hospitals and clinics to improve patient care.
A key advantage of telemedicine is its ability to increase access to health care by offering
patients the opportunity to receive care in their homes and communities. This umbrella review
aims to provide a broad overview of how telemedicine is being used across OECD countries,
focusing specifically on clinical and cost-effectiveness, patient experience and implementation.
Objectives
To conduct a systematic review of the current literature on telemedicine and provide summary
evidence of its effectiveness, cost-effective, patient experiences and implementation.
Methods
Eligibility criteria
We included only peer-reviewed published systematic reviews of primary studies that focused
on one or more types of telemedicine (remote monitoring, real-time, and store-and-forward).
Telemedicine is defined as the use of telecommunication systems to deliver health care at a dis-
tance [1]. Telemedicine is often used as a synonym for the boarder term, telehealth which
encompasses both remote health care delivery and the delivery of education or training to
patients and/or HCWs [2]. Other terms used to describe telemedicine include e-health, tele-
care, or m-health. Irrespective of the term used, we included all interventions that involved
health care delivery through telecommunication systems that included either synchronous or
asynchronous communication between a patient and their HCW. Therefore, reviews that
focused on interventions such as automated text message reminders to improve medication
adherence, or self-help websites with no direct interaction with a HCW were excluded.
PLOS ONE | https://doi.org/10.1371/journal.pone.0237585 August 13, 2020
2 / 24
PLOS ONE Telemedicine in the OECD: An umbrella review
We excluded non-English language reviews, reviews that focused on non-OECD countries,
umbrella reviews, and reviews that did not clearly report on one or more of the predefined out-
comes of this umbrella review: clinical effectiveness, cost-effectiveness, patient experience and
implementation. For clinical effectiveness reviews, we further excluded reviews that did not
report a pooled quantitative outcome (meta-analysis). For cost-effectiveness reviews, we
included reviews of primary studies of economic evaluations or cost analyses.
Search methods for the identification of reviews
Four online databases, PubMed/Medline, Centre for Reviews and Dissemination (CRD) and
the Cochrane library were searched for systematic reviews or meta-analyses published between
January 2014 and February 2019. Excluded citations and reasons for exclusion at full text eligi-
bility stage are included in the supporting information (S1 File). Two review authors (NE) and
(TCOH) designed the search strategy, data extraction template and conducted preliminary
searches. One author (NE) performed the initial screening all titles and abstracts identified
from the searches for inclusion based on the inclusion criteria and extracted initial data using
a piloted data extraction form. Citations were coded as ‘included’, ‘excluded’ or ‘unclear’ as
appropriate. Data was checked by (TCOH) and both authors agreed or disagreed on what
reviews to include, and data to extract based on the protocol, and inclusion criteria. All dis-
agreements were resolved by discussion. Due to heterogeneity of populations, interventions
and outcomes in the included systematic reviews, no attempt was made to compare telemedi-
cine interventions across reviews or across review populations.
Assessment of methodological quality of included reviews
All three authors assessed the methodological quality of included reviews using the AMSTAR
2 tool [3]. The AMSTAR 2 is an update of the previous AMSTAR tool and it was chosen for its
content validity, and its ability to assess systematic reviews of both randomised and non-ran-
domised studies. Unlike the previous tool, the AMSTAR 2 is not designed to produce an over-
all score, rather reviews are graded according to methodological flaws in seven critical and
nine non-critical domains.
Data synthesis and unit of analysis
A meta-analysis was precluded due to heterogeneity of telemedicine interventions, interven-
tion intensity, face-to-face comparators, study duration, HCW feedback methods and possible
overlap of primary studies in the clinical effectiveness reviews. We presented results narra-
tively, with tables and graphs for illustration where appropriate. The unit of analysis in this
umbrella review is the systematic review and not the included primary studies, therefore only
review level evidence was synthesised and included in findings.
Results
The review selection process is summarized in a PRISMA flow diagram (Fig 1). The combined
searches yielded 320 citations, and after exclusions 98 reviews were included in this umbrella
review.
Characteristics of included reviews
Characteristics of the 98 included reviews are summarized in the supporting information S2
File. All reviews were categorized according to the predefined outcomes of this review Table 1.
Fifty-three reviews focused on clinical effectiveness, 18 on cost-effectiveness, 15 addressed
PLOS ONE | https://doi.org/10.1371/journal.pone.0237585 August 13, 2020
3 / 24
PLOS ONE Telemedicine in the OECD: An umbrella review
Fig 1. Prisma flow diagram.
https://doi.org/10.1371/journal.pone.0237585.g001
PLOS ONE | https://doi.org/10.1371/journal.pone.0237585 August 13, 2020
4 / 24
PLOS ONE Telemedicine in the OECD: An umbrella review
Table 1. Classification of telemedicine reviews according to review outcome.
Review Outcome
Review [References]
n
Effectiveness
Cost-effectiveness
Patient experience
Implementation
�Double listings
[4–53] [54]� [55]�
[51,55–69] [70]�[71]�[54]�[55]�
[72–84][85]�[71]�
[86–100][85]�[70]�
53
18
15
17
https://doi.org/10.1371/journal.pone.0237585.t001
patient experiences and 17 were on implementation of telemedicine. Five reviews addressed
more than one outcome area and are marked as double listings.
Origin of primary studies included in reviews. Sixty-four reviews (65%) of included
details of the origin of primary studies (Fig 2). The US had the highest proportion of primary
studies 37% (534/1,456) followed by Australia 17% (251/1,456), UK 10% (135/1,456) and
Japan 5% (75/1,456). The full list of countries and the number of studies from each country is
included in the supporting information S2 File.
Description of telemedicine interventions. Telemedicine interventions included in this
review are categorized as remote monitoring, interactive real-time telemedicine and store-
and-forward systems. Remote monitoring involves asynchronous communication with moni-
toring devices, such as glucometers, blood pressure monitors, and mobile apps, which have the
ability to transmit patient results remotely to HCW. Store-and-forward systems allow for the
secure electronic transmission of patient data, and often take place between HCWs to aid diag-
noses [4]. Real-time interventions use synchronous communication methods e.g. video and
telephone consultations. These categories may be compared with usual care, an alternative
telemedicine intervention, in addition to usual care, or no intervention. Table 2 includes the
speciality area and type of telemedicine intervention used across the fifty three clinical-effec-
tiveness reviews. The most frequently reported intervention used was device or app based
remote monitoring in 75% (40/53) of reviews, while store-and-forward was the least frequent
intervention reported in 4% (2/53) of reviews (Table 2). Three reviews included less conven-
tional interventions such as virtual reality gaming [5] and social media [6,7]. A number of
reviews included an element of patient education or training, however the primary aim of
these interventions was health care delivery. All reviews except one addressed one or more spe-
cialty or disease areas. The sole review that didn’t focus on a speciality described the effect of
telemedicine interventions delivered by allied health professionals and nurses in rural and
remote areas [8].
Methodological quality of included systematic reviews. Overall, 8% (8/98) of reviews
were graded as high quality, while the remaining 92% (90/98) were of low to critically low qual-
ity, S3 File. The most common critical weakness across reviews was the failure to include the
full citation list of excluded studies with reasons for exclusion, followed by an explicit state-
ment that the review was based on a prior protocol. Most reviews however performed adequate
searches, used satisfactory techniques to assess methodological quality, and appropriate meth-
ods for meta-analysis where applicable.
In the non-critical domains, most reviews adequately followed and reported PICO ele-
ments, performed comprehensive literature searches, and reduced bias by selecting studies
and extracting data in duplicate. The most common non-critical flaw across reviews was the
failure to report sources of funding for included primary studies.
PLOS ONE | https://doi.org/10.1371/journal.pone.0237585 August 13, 2020
5 / 24
PLOS ONE Telemedicine in the OECD: An umbrella review
Fig 2. Origin of primary studies included in reviews.
https://doi.org/10.1371/journal.pone.0237585.g002
Clinical effectiveness
Endocrinology. Eleven reviews evaluated the effectiveness of telemedicine interventions
for diabetes management and all reviews concluded that telemedicine was superior or compa-
rable to face-to-face interventions. Reviews compared telemedicine interventions to usual
face-to-face care or no intervention and focused mainly on remote self-monitoring. One
review focused on healing time for diabetic foot ulcers, using real-time video consultation and
the second on the prevention or delay of Type 2 diabetes for at risk populations through
remote monitoring weight loss interventions. [9,10]. The first review concluded that the tele-
medicine group had similar ulcer healing times as the face-to-face care group, while the second
review found promising evidence of the efficacy of telemedicine with remote behavioral sup-
port for weight loss to prevent or delay Type 2 diabetes. Three out of the remaining ten reviews
assessed the effectiveness of remote self-monitoring interventions on gestational diabetes. Two
out of the three views were graded as high quality [11,12]. All three reviews found that tele-
medicine had similar clinical outcomes to face-to-face care for this population including glyce-
mic control, cesarean delivery rates, pre-eclampsia, induction of labor and neonatal
hypoglycemia [11–13]. The remaining seven reviews focused on glycemic control. One high-
quality review found that remote monitoring can improve the control of blood glucose in dia-
betic patients [1], while the other six reviews concluded that remote monitoring was more
effective than face-to-face care [14–19].
Cardiovascular disease. Eleven reviews assessed the effectiveness of telemedicine in car-
diovascular disease management compared to usual care. In nine reviews, primary outcomes
of interest across reviews included mortality, hospital readmissions, emergency room visits
and risk factors measured by body mass index (BMI), systolic and diastolic blood pressure and
low-density lipoprotein (LDL) cholesterol. Two reviews found that remote monitoring, and
structured telephone support was superior to usual care in reducing the odds of mortality and
hospitalization related to heart failure, improve survival rates and reduce cardiovascular dis-
ease risk factors such as weight, BMI and blood pressure [20,21]. Four reviews, including two
high-quality reviews [1, 22] found no difference between remote monitoring with telephone
support and face-to-face care to reduce planned hospital visits and to improve modifiable
PLOS ONE | https://doi.org/10.1371/journal.pone.0237585 August 13, 2020
6 / 24
PLOS ONE Table 2. Effectiveness reviews by speciality area and type of telemedicine intervention.
First Author (Year)
Speciality Area
Remote Monitoring
Device or App based Web-based�
Interactive Real-time
Store-and-forward
Telephone Consults Video Consults
Telemedicine in the OECD: An umbrella review
Adamse (2018)
Rehabilitation
Agostini (2015)
Rehabilitation
Cottrell (2017)
Rehabilitation
Cruz (2014)
Dario (2017)
Deady (2017)
Direito (2017)
Feltner (2014)
Respiratory Medicine
General Medicine
Mental Health
General Medicine
Cardiology
Flodgren (2015)
Cardiology Endocrinology
Hakala (2017)
Huang (2015)
Huang (2014)
Huang (2015)
Hui (2017)
Rehabilitation
Cardiology
Gastroenterology
Endocrinology
Respiratory Medicine
Hutchesson (2015)
Bariatric Medicine
Jeon (2015)
Joiner (2017)
Kelly (2016)
Not specified
Endocrinology
General Medicine
Kepplinger (2016)
Neurology
Klersy (2016)
Kotb (2015)
Lee (2016)
Lee (2018)
Linde (2015)
Liu (2017)
Cardiology
Cardiology
Paediatrics
Neurology
Mental Health
Endocrinology Respiratory Medicine
Lundell (2015)
Respiratory Medicine
Marx (2018)
Geriatric Medicine
McLean (2016)
Respiratory Medicine
Merriel (2014)
Cardiology
Ming (2016)
Nair (2018)
Obstetrics Endocrinology
Obstetrics Mental Health
Oosterveen (2017)
Mental Health
Raman (2017)
Obstetrics Endocrinology
Rasekaba (2015)
Obstetrics Endocrinology
Rawstorn (2016)
Cardiology
Seiler (2017)
Oncology
Seyffert (2016)
Mental Health
Sherifali (2017)
Obstetrics
Speyer (2018)
Not specified
Stratton (2017)
Mental Health
Su (2016)
Tchero (2017)
Endocrinology
Endocrinology
Thabrew (2018)
Mental Health
Thomas (2014)
Ophthalmology
Toma (2014)
Endocrinology
van Beugen (2014)
Mental Health
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
(Continued )
PLOS ONE | https://doi.org/10.1371/journal.pone.0237585 August 13, 2020
7 / 24
PLOS ONE Table 2. (Continued)
First Author (Year)
Speciality Area
van Egmond (2018)
Rehabilitation
Van Spall (2017)
Cardiology
Vigerland (2016)
Paediatrics Mental Health
Widmer (2015)
Cardiology
Wootton (2016)
Mental Health
Zhai (2014)
Zhao (2015)
Endocrinology
Respiratory Medicine
�includes emails and online chat
https://doi.org/10.1371/journal.pone.0237585.t002
Telemedicine in the OECD: An umbrella review
Remote Monitoring
Device or App based Web-based�
Interactive Real-time
Store-and-forward
Telephone Consults Video Consults
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
cardiovascular risk factors through cardiac rehabilitation [22–24]. One review produced
mixed results, and found that telemedicine delivered through structured telephone support
reduced heart failure specific readmissions but not all cause mortality readmissions [25]. Of
the remaining two reviews, one concluded that telephone or remote monitoring interventions
compared to nurse home visits had no significant improvement on readmission or mortality
in heart failure patients [26], while the other found insufficient evidence that remote monitor-
ing and counselling reduced overall cardiovascular disease risk [27].
The remaining two reviews assessed the effectiveness of telemedicine in stroke manage-
ment. The first review investigated the safety and efficacy of treatment delivered through tele-
stroke networks (real-time telemedicine) for patients with acute ischemic stroke compared to
face-to-face care and concluded that telemedicine was safe and effective with no difference in
mortality or functional independence between intervention and comparison groups [28]. The
second review concluded that telephone consultations used in the management of oral anti-
coagulation may lower the risk of major thromboembolic events, but not other clinically rele-
vant outcomes [29].
Rehabilitation. Seven reviews assessed the effectiveness of telemedicine in rehabilitation
(tele-rehabilitation) compared to usual care. The results were varied. For musculoskeletal con-
ditions, video and telephone consultations were effective in the improvement of physical func-
tion compared to conventional care [30]. In surgical populations, wireless monitored exercise,
web-based support and telephone consultations were at least as clinically effective as face-to-
face care, with greater improvements in quality of life compared to face-to-face care [31]. For
patients with chronic obstructive pulmonary disease (COPD), limited evidence suggested that
remote monitoring with and telephone support could lead to improvements in physical activ-
ity [32]. For pain and disability, exercise monitoring, web-based programs and telephone con-
sultations had no significant effect compared with usual care, but interventions that combined
usual care and telemedicine were more effective than usual care alone [33]. In motor function,
remote monitoring of exercise and training was effective for cardiac and orthopedic (total
knee arthroscopy) patients, but the evidence was inconclusive for neurological patients [34].
For chronic pain, remote monitoring of excercise compared to no intervention, was effective
in reducing pain and, furthermore, no difference was found in effectiveness between telemedi-
cine compared to usual care for increased physical activity, or activities of daily living [35]. In
cancer patients, therapist and self-guided web-based programs, and remote monitoring were
effective for managing cancer-related fatigue [36].
Mental health. Ten reviews assessed the effectiveness of telemedicine interventions in
mental health and behavioral conditions compared to usual care or no intervention. Nine
PLOS ONE | https://doi.org/10.1371/journal.pone.0237585 August 13, 2020
8 / 24
PLOS ONE Telemedicine in the OECD: An umbrella review
studies concluded that telemedicine was at least as effective as face-to-face interventions, while
one study was inconclusive. Eight studies had depression and/or anxiety as a primary outcome,
while the remaining three focused on symptom reduction for obsessive compulsive disorder
(OCD), insomnia, and reduction of alcohol consumption. From the eight depression studies,
one focused on employees, one on maternal depression and two on children. The review that
focused on employees observed a small positive effect for web-based congnitive behavioural
therapy (CBT) with remote monitoring in reducing stress and anxiety in employees, but con-
cluded that interventions targeted to individual employees were more effective than universal
interventions [37]. For maternal depression, remote monitoring, and telephone consultations
were effective in improving maternal depression symptoms [38]. For depression in children
and adolescents, one high quality review found that the effect of multi-modal web-based CBT
with support was inconclusive [39], while another review concluded that web-based CBT with
structured telephone calls was effective in the treatment of psychiatric and somatic conditions
[40]. Similarly, three other reviews found web-based CBT with remote monitoring, and
remote therapist support was as effective as usual care in reducing symptoms of depression
[41–43]. For adults with insomnia, web-based CBT with remote monitoring and real-time sup-
port was an effective therapy for improving sleep [44]. For OCD and behavioral interventions
to reduce alcohol consumption in young adults, the effect of telephone counselling and remote
monitoring was comparable to conventional face-to-face interventions [45,46].
Physical activity and diet. Four reviews assessed the effect of telemedicine on physical
activity and nutrient intake. All four reviews found that telemedicine interventions were at
least as effective as usual care. Targeted remote monitoring interventions were more effective
than face-to-face care to increase physical activity, and universal interventions were compara-
ble to face-to-face interventions in reducing sedentary behavior [47,48]. Remote monitoring,
video and telephone consultations could improve diet quality, as well as fruit, vegetable, and
dietary sodium intake for people with chronic conditions and malnourished community
dwelling older adults [49,50].
Weight loss. Four reviews investigated the effect of telemedicine interventions on weight
loss. The first review found that remote monitoring alone achieved modest weight loss com-
pared to no intervention, but achieved significantly greater weight loss when combined with
behavioral features [51]. The second review found that app-based remote monitoring inter-
ventions in elementary school children had no significant effect for weight control, exercise, or
sugar sweetened beverage intake [52]. The third review found a small positive effect of app-
based remote monitoring for weight loss [19]. Finally, for weight loss during pregnancy and
the postpartum period, remote monitoring interventions were only effective in postpartum
women [53].
Respiratory disease. Four reviews investigated the effectiveness of telemedicine for
asthma control and COPD. Two reviews found telemedicine was at least as effective as face-to-
face care, and two were uncertain about the effectiveness of telemedicine. The first review con-
cluded that mobile app-based remote monitoring interventions which facilitated professional
support improved asthma control and exacerbation rates [54]. The second review found no
difference in asthma symptom scores between the intervention groups, (remote monitoring
and telephone consults) and face-to-face care groups [55]. The third review concluded, using
low quality evidence, that remote monitoring interventions had small beneficial effects on
asthma control [56], while the fourth found limited evidence of the effect of home based
remote monitoring to reduce health care utilization, quality of life and respiratory exacerba-
tions in COPD patients [57].
Other specialties. One review assessed the effectiveness of telemedicine in the manage-
ment of inflammatory bowel disease (IBD). The authors concluded that remote monitoring
PLOS ONE | https://doi.org/10.1371/journal.pone.0237585 August 13, 2020
9 / 24
PLOS ONE Telemedicine in the OECD: An umbrella review
interventions reduced clinic visit utilization compared to face-to-face care, and were as effec-
tive as face-to-face care in other outcome areas including relapse rates, hospital admission
rates and quality of life [58].
Remote monitoring, video and telephone consultations delivered by allied health profes-
sionals to patients in rural areas were at least as effective as face-to-face interventions [59]. For
glaucoma, tele-glaucoma via video consultations was more effective than face-to-face examina-
tions in detecting glaucoma besides reducing waiting times, offering services to rural areas,
and reducing travel [60].
Cost-effectiveness
Eighteen reviews assessed the costs or cost-effectiveness of telemedicine interventions. Seven
reviews found that telemedicine was cost-effective or cost saving, Table 3.
For the management of rheumatoid arthritis video and telephone consultations were found
to be cost-effective compared to usual face-to-face care [61]. In mental health, a review
assessed computerized cognitive behavioral therapy (cCBT) delivered through remote moni-
toring and video consultations from a provider or societal perspective. They found that cCBT
was cost-effective compared to face-to-face care mainly because the time spent on cCBT inter-
ventions were up to six times lower compared to face-to-face therapy, thereby reducing the
workload of mental health practitioners [62]. For tele-glaucoma, the time spent by the physi-
cian compared to time spent in face-to-face examination was shorter, and as a result tele-glau-
coma was cost-effective, offering a potential saving of $175.9 USD per patient compared to
conventional face-to-face diagnosis [60]. In Japan, remote monitoring and video consultations
were found to be generally cost-effective for teledermatology and teleradiology from a provider
or societal perspective. However tele-homecare produced mixed results due to lower willing-
ness to pay and higher system costs in areas with public subsidies, which led to overutilization
[63]. In one high-quality review, 74% of included studies found that remote monitoring
through mobile devices to support data collection was cost-effective, economically beneficial,
or cost saving at base case from a provider or societal perspective [64]. Compared to hospital
monitoring, pacemaker remote monitoring interventions were cost-effective from a provider
or societal perspective with no significant difference between groups in quality of life or
adverse events [65]. Store-and-forward teledermatology, as an adjunct to conventional face-to-
face care may be cost-effective from a provider or societal perspective especially when used as
a triage tool to reduce waiting times as well as the costs and time associated with travel to face-
to-face appointment requirements [66].
Five reviews concluded that although telemedicine interventions could be cost-effective, or
cost saving, heterogeneity, poor quality and paucity of cost data limited the ability to arrive at a
more robust conclusion. In one review of home-based monitoring, costs varied substantially
depending on specific chronic conditions, and ranged from $1,352 USD for heart failure to
$206,718 USD for congestive heart failure, COPD and diabetes as a whole. While cost data was
limited, the available data indicated that remote monitoring and video consultations were cost
saving from a provider perspective [67]. The second review concluded that remote monitoring
and video consultations could potentially be cost-effective from a provider perspective, how-
ever insufficient high-quality evidence precluded a more certain conclusion [68]. The remain-
ing three reviews found that cost-effectiveness of video consultations and remote monitoring
was uncertain due to low quality and insufficient evidence. The first review included studies
that performed analysis from a provider or patient perspective for tele-dentistry [69], while the
other two reviews were from a provider perspective for chronic heart failure [70] and COPD
[71].
PLOS ONE | https://doi.org/10.1371/journal.pone.0237585 August 13, 2020
10 / 24
PLOS ONE Telemedicine in the OECD: An umbrella review
Table 3. Summary charactheristics of cost-effectiveness reviews.
Author (Yr.) Country
a. Speciality
a. Intervention
b. Perspective
Conclusions
Akiyama (2016) Japan
a. Geriatrics
a. RM, VC urban and
rural homes
TM programs in Japan appear to have a favourable level of
economic efficiency.
b. Included Studies
c. Time horizon (Range)
b. 11(1 CMA, 2 CBA, 4 WTPE, 4 BE) b. Provider, Societal
c. 30 days to 11 yrs.
De la Torre-Diez (2015)
Denmark
a. Dermatology, Cardiology,
Endocrinology, Mental Health,
Ophthalmology
b. 35(CUA, CEA)
a. RM, VC rural areas,
inmates
Variation across studies made it unrealistic to draw an overall
conclusion about cost-effectiveness of TM
b. Provider
c. n/s
Estai (2018) Australia, UK
a. Dentistry
a. VC screening for oral
diseases
VC can be cost-saving compared to a F2F care, however there is
inconclusive evidence to support evidence-based policy decisions
Fuertes-Guiro (2017) UK,
Austria, Italy
b. 2 CMAs
a. Dermatology
b. 8 RCTs
Grustam (2014) US, Australia,
Italy, China, Taiwan
a. Cardiology
b. Provider, Patient
c. n/s
a. Teledermatology
(TD): VC, S&F
b. Provider, Patient
c. n/s
a. RM, TC to manage
CHF patients
Hammed (2014) UK, USA,
Germany, Brazil, New Zealand,
Taiwan
b. 32(21 RCTs, 2 QRCTs,9 NRs)
b. Provider
a. Heart Failure
c. 78% < 1 yr.
a. RM, TC for heart
failure patients
b. 9 (2 RCTs, 1 QRCT, 2 CEA, 3NRs)
b. n/s
Irribarren (2017) US, UK,
Africa, Europe, China, Thailand,
Australia
a. Cardiology, Diabetes
b. 39 (25 CEA, 12 CUA, 1 CMA, 1
CBA)
Jackson (2017) UK, Denmark,
Ireland
a. Gastroenterology
b. 2 (1 RCT, 1 NRs)
c. n/s
a. RM with mobile
devices
b. Provider, Societal
c. 30 days to Lifetime
a. RM, VC
b. n/s
c. n/s
TD lasted 7.5 mins more on average than F2F consultation,
increasing costs but there is no unamity due to incomplete cost
reporting
60% of high-quality studies found TM cost-effective, but cost-
effectiveness is uncertain due to poor and scarce economic
evidence
There is currently no evidence that increased patient adherence,
supported by TM, has led to a reduction in treatment costs for
HF patients.
Findings support the cost-effectiveness of Mhealth interventions,
but this result must be considered cautiously, and evaluated on a
case-by case basis
Studies described cost savings, but did not account for indirect
intervention costs. Further work is required to appreciate
whether or not TM is cost-effective
Liddy (2016) Spain, Netherlands,
USA
a. Dermatology, Paediatrics,
Multispeciality
a. Teledermatology
(TD): VC
TD has the potential for large cost savings, but gaps remain due
to variation in reporting which limits comparability.
b. 7(2 CA, 3 CMA, 2 CEA)
b. Provider, Societal
Lopez-Villegas (2015) UK, US,
Canada, Italy, Austria, France
a. Cardiology
c. n/s
a. RM, pacemaker
monitoring
b. 7 (4 CEA, 1 CBA, 1 CUA, 1 CMA) b. Provider, Societal
McDougall (2017) Australia, UK,
Canada, Italy, US
a. Internal Medicine
b. (6 CA)
c. 10 months to 7 yrs
a. Telerheumatology:
VC, TC
b. n/s
c. n/s
Pacemaker RM detects cardiovascular events earlier reduces
hospitalizations, the number of hospital visits, and associated
follow-up costs.
All included studies found within a limited body of evidence that
TM was cost-effective for the management of rheumatic disease
Michaud (2018)
a. Diabetes, Heart Failure, Cancer,
COPD
a. Home-based
monitoring: VC, RM
b. 12 (6 RCT, 2 QRCTs, 4NRs)
b. Provider
Although all studies found that TM reduced costs, a
comprehensive analysis of costs and their determinants is still
required for wider adoption and reimbursement of TM
c. n/s
(Continued )
PLOS ONE | https://doi.org/10.1371/journal.pone.0237585 August 13, 2020
11 / 24
PLOS ONE Telemedicine in the OECD: An umbrella review
Table 3. (Continued)
Author (Yr.) Country
a. Speciality
Musiat (2014) n/s
b. Included Studies
a. Mental Health
b. 13 (5 CUA, 8 CEA)
a. Intervention
b. Perspective
c. Time horizon (Range)
a. cCBT: RM, VC
b. Societal, Provider
c. n/s
Conclusions
Overall results suggest that cCBT is a cost-effective alternative to
usual care with similar or superior results at lower direct costs.
Sanyal (2018) Australia, Canada,
Netherlands, Spain, US, UK
a. Cardiology, Diabetes, Mental
Health
a. RM, TC, VC, for older
adults
b. 11 (CEA, CUA)
b. Provider
c. 4 months to lifetime
There is a lack of convincing evidence to conclude whether TM
to deliver health care to older adults will demonstrate value at any
acceptable level of investment.
Snowswell (2016) US, UK, New
Zealand, Netherlands, Spain
Thomas (2014) Germany,
Canada, Finland, Indonesia,
Taiwan, US, Spain
a. Dermatology
a. Teledermatology: S&F Current evidence is sparse, but suggests that TD can be cost-
b. 11 (4 CMA, 4 CEA, 2 CUA, 1 CA)
b. Provider, Societal
effective
a. Opthalomologyb. 9 (3 CEA, 6
NRs)
a. Teleglaucoma (TG):
remote areas
c. n/s
b. n/s
c. n/s
TG considerably reduces patient access time, and physician
consultation time, thereby saves costs to patients and the
healthcare system as a whole.
Udsen (2014) Canada, US, Italy,
Spain, Denmark
a. Respiratory Medicine
a. RM: COPD patients
b. 6 (1 CUA, 3 CCA, 2 CMA)
b. Provider
c. 3–12 months
Limited evidence showed a potential for cost savings with lower
average cost per patient with TM plus usual care vs. usual care
alone.
Zhai (2014) n/s
a. Endocrinology
a. RM: Type 2 diabetes
patients
The two studies reported wide ranging costs, confounding
attempts to draw meaningful conclusions about cost-effectiveness
b. 2 (1 RCT, 1 CEA)
b. n/s
c. n/s
TM = Telemedicine, CUA = Cost Utility Analysis, CCA = Cost Consequence Analysis, CMA = Cost Minimization Analysis, CEA = Cost Effectiveness Analysis,
RCT = Randomized Controlled Trial, S&F = Store-and-forward, RM = Remote Monitoring, TC = Telephone Consultation CA = Cost Analysis, NHS = National Health
Service, WTPE—Willingness to Pay Estimation, BE = Benefit Estimation, RM = Remote Monitoring, VC = Video Consultation, F2F = Face to face, QRCTs = Quasi-
randomised Controlled Trials, NRS = Non-randomized Studies, CHF = Chronic Heart Failure, N/S = Not Stated, cCBT = computerized cognitive behavioural therapy
https://doi.org/10.1371/journal.pone.0237585.t003
Two reviews were unable to arrive at a conclusion regarding cost-effectiveness due to varia-
tion in outcomes within included primary studies. The first review included studies that per-
formed analysis from a provider and societal perspective and concluded that heterogeneity in
outcome reporting limited the ability to draw conclusions on the cost-effectiveness of teleder-
matology. [72] Included studies in the second review produced very different incremental
cost-effectiveness ratio (ICER) values of $491 and $29,869 for remote monitoring in patients
with Type 2 diabetes [18].
Four reviews concluded that based on the available evidence, telemedicine interventions
were not cost-effective. The first review concluded that it was doubtful, based on current evi-
dence, that the use of remote monitoring, video or telephone consultations to deliver health to
older adults could demonstrate value at any acceptable level of investment from a providers
perspective [73], while the second found using limited evidence that remote monitoring and
video consultations may increase costs [74]. The third review included studies that performed
analysis from a provider and patient perspective and concluded that teledermatology takes
more time than face-to-face consultations presenting higher costs than face-to-face care, and
thereby increasing the total cost of consultations [75]. Finally, the fourth review found no evi-
dence that increased patient adherence supported by remote monitoring and telephone calls
led to a reduction in treatment costs for heart failure patients [76].
PLOS ONE | https://doi.org/10.1371/journal.pone.0237585 August 13, 2020
12 / 24
PLOS ONE Telemedicine in the OECD: An umbrella review
Patient experience
Fifteen reviews focused on patient experiences with telemedicine. Three main themes were
identified: patient acceptability, patient satisfaction and patient barriers to the use of
telemedicine.
Patient acceptability. Six reviews addressed patient acceptability of telemedicine. All
reviews found that telemedicine interventions were acceptable to patients. In primary care set-
tings, remote monitoring and real-time telemedicine interventions are generally acceptable,
however the level of acceptability varied by population demographics such as gender, age and
socio-economic status. Furthermore, these interventions were more acceptable to patients
than to health care providers [77]. Mobile-app based remote monitoring interventions were
acceptable to children and adolescents with mental health conditions [78]. For people with
severe mental illness, web or mobile-phone based remote-monitoring was acceptable, with
higher acceptability levels when participants were provided with remote online support [79].
For the secondary prevention of cardiovascular disease, app-based remote monitoring inter-
ventions were feasible and acceptable to cardiac patients with high levels of engagement, usage
and adherence [80]. Finally, remote-monitoring, real-time telemedicine, and store-and-for-
ward interventions were acceptable to COPD and dermatology patients [81,82].
Patient satisfaction. Six reviews addressed patient satisfaction and enablers to the use of
telemedicine interventions. For patients with mental health conditions, text-messaging inter-
ventions improved treatment adherence and symptom surveillance, and increased patient sat-
isfaction with management and health care services [83]. Mental health patients who had
undergone computerized cognitive behavioural therapy (cCBT) reported high treatment satis-
faction rates, however more research was advised to address attrition due to dissatisfaction
[62]. COPD patients reported that they were satisfied with remote monitoring, and that they
found interventions useful to help manage their condition [84]. Cancer patients and survivors
had a positive experience with remote monitoring and telephone consultations. They reported
that interventions were convenient, acceptable, fostered independence and remote reassur-
ance, reduced burden and provided the safety net of a professional health care connection
[85,86]. Telemedicine interventions incuding remote monitoring and real-time telemedicine
improved social and emotional wellbeing of indigenous people receiving care in the commu-
nity, improved patient outcomes and access to specialist services, and resulted in greater
patient empowerment due to improved health literacy [87].
Barriers to patient uptake and use. Four reviews reported patient barriers to the use of
telemedicine. The first review concluded that lack of training, poor patient tolerance with faulty
systems, and technical problems all negatively affected uptake and sustained use of home-based
remote monitoring [88]. The second review, was graded as high-quality and focused on the
experiences of young people with chronic non-communicable diseases. The authors concluded
that lack of collaboration between implementers and end users, as well as an inability to tailor
and adapt technology to meet person-centered needs and preferences, were barriers to the
wider use of mobile or app-based remote monitoring interventions [89]. For COPD patients,
the third review found that remote monitoring devices with reported technological and usability
problems had lower intervention compliance rates [84]. In heart failure patients, low patient
motivation, lack of confidence and a preference for face-to-face interventions were all barriers
to the use of remote monitoring interventions, resulting in poor uptake of interventions [90].
Implementation
Seventeen reviews reported on the implementation of telemedicine. The main themes identi-
fied were barriers, enablers, risk factors, feasibility and sustainability.
PLOS ONE | https://doi.org/10.1371/journal.pone.0237585 August 13, 2020
13 / 24
PLOS ONE Telemedicine in the OECD: An umbrella review
Barriers to implementation. Three reviews reported on barriers to the implementation
of telemedicine. Three reviews identified barriers to remote monitoring in the management of
diabetes, including poorly designed interfaces, manual data input, transmission with delayed
feedback, limitations on scalability, and technological illiteracy [91,92]. These factors contrib-
uted to recall bias or human error which impeded patient engagement, resulting in large drop-
out rates [92]. In indigenous communities, barriers to telediabetes included the lack of techni-
cal skills and shortage of local staff [93]. Cost factors included start-up costs, ongoing costs,
and costs related to loss of revenue. The absence of or inadequate legislation and policies, and
liability cover might constitute a barrier to implementation at organization and health profes-
sional level. Furthermore, incompatibility with health systems, work practices or daily clinical
work leading to disruptions to workflow and delivery of care were common barriers to imple-
mentation. Finally, lack of strategic planning and engagement of key stakeholders in develop-
ment, selection and adoption of electronic health (eHealth) systems, were also problematic.
Enablers of successful implementation. Five reviews reported on enablers of telemedi-
cine implementation. The first review concluded that interventions should be targeted, and
systems designed to solve clinical or behavioral problems identified by patients as priorities.
Furthermore, telemedicine technology should be automated, streamlined, mobile and low-
cost to limit the burden of usage on patients, and maximize clinical utility [91]. The second
review identified tailoring home-based remote monitoring interventions to patient character-
istics and needs, relationship and communication between patient and health care profession-
als, technology usability and quality, and home health care organization, as enablers of
telemedicine [94]. For indigenous populations, the involvement of indigenous health workers
was essential in delivering telediabetes interventions to these communities because of their
role in communication in local language and the potential to help clinicians understand the
local community and facilitate integration [93]. In mental health, understanding treatment
preferences, facilitating uptake and governing mechanisms were identified as enablers to
implementation [95]. For diabetic patients, interventions incorporating elements of structured
monitoring of blood glucose identified by the International Diabetes Foundation as essential
for improving A1c (average blood glucose level over the past 3 months) were associated with
comparatively significant improvement in clinical outcomes. These elements included patient
and provider education, the use of pre-approved structured monitoring, goal setting, shared
decision making and interactive communication [96].
Risks associated with using telemedicine. One review reported on patient risk factors
associated with home-based remote monitoring and found that patient safety risks were
mainly attributable to a lack of patient and or staff knowledge and understanding, changes in
the nature of clinical work, technological issues, and patient dependency or anxiety [97]. Prob-
lems with telemedicine technology and devices, organizational issues and environmental fac-
tors could also contribute to patient safety risks but to a lesser extent. Risk concerns were
mainly raised by health care professionals [97].
Sustainability of telemedicine initiatives. Four reviews addressed the sustainability of
telemedicine. In rural and remote Australia, vision, ownership, adaptability, and economic
efficiency aided sustainability while contributing to equity of access to health care [98]. For
older adults, the success of remote monitoring interventions to monitor and improve their
health greatly depended on the support and training they received [99]. Staff, technical and
service factors also had an impact on the sustainability of remote monitoring interventions
[90]. Staff factors associated with unsustainability included the absence of champions, dislike
of new clinical routines or interactions, perception of no value or compromised clinical exper-
tise. Technical factors that made telemedicine service likely unsustainable include unreliable
or difficult technology and inadequate helpdesk support, while service factors included a lack
PLOS ONE | https://doi.org/10.1371/journal.pone.0237585 August 13, 2020
14 / 24
PLOS ONE Telemedicine in the OECD: An umbrella review
of clarity on who was responsible for interpreting or actioning remote monitoring data. The
fourth review found that the renumeration of participating clinicians was crucial to the sus-
tainability of real-time and store-and-forward telemedicine interventions [100].
Feasibility of telemedicine interventions. Seven reviews reported on the feasibility of
telemedicine interventions. Feasibility includes clinical, organizational or technical factors
determining the practicality of implementing telemedicine services. Remote monitoring and
real time telemedicine was feasible in patients with neurological diseases [101], in primary care
settings [77]. Real-time and store-and-forward intervenitons were feasible in dentistry [100]
and for remote monitoring was feasible in Japan [102]. A higher level evidence was required to
support clinical application of teledermoscopy for diagnostic measurement in the treatment of
precancerous and cancerous skin lesions [103]. Feasibility of telemedicine depended on part-
nerships between researchers, health care planners and policymakers to help align implemen-
tation research with policy development and funding [104]. The final review concluded that
for older adults, usability and costs are key to feasibility [105].
Discussion
This umbrella review identified a diverse and growing body of literature on telemedicine, and
provides a broad summary of the use of telemedicine within the OECD. Eighty-three percent
(44/53) of effectiveness reviews found that telemedicine interventions were at least as effective
as face-to-face care. In disease areas such as diabetes management, all included reviews found
that telemedicine interventions were effective. Two (4%) of reviews found that telemedicine
was ineffective compared to face-to-face care, and seven (13%) reviews were uncertain about
the effect of telemedicine. Although most reviews found positive effects for the use of telemedi-
cine, these findings should be interepreted cautiously, given heterogenetiy across populations,
interventions, settings and the overall low quality of incuded reviews.
A wide range of terms were used to describe telemedicine interventions such as telehealth,
mhealth, and ehealth, confounded by terms specific to disciplines and specialities such as tele-
nursing, teledermatology and telerehabilitation. This inconsistent use of terms may impede
knowledge translation and dissemination. Categorizing interventions by type of telemedicine
used (e.g. video consultations vs remote monitoring) may be useful to provide a more targeted
summary. It may also be helpful when pooling results, to conduct sub-group analysis accord-
ing to intervention type, and intensitiy.
The impact of interventions on utilization of health care was not examined. Although tele-
medicine may reduce the demand for face-to-face care, some evidence suggests that telemedi-
cine, especially direct-to-consumer (DTC) telemedicine is often used as an adjunct to face-to-
face care and may increase health care utilization and costs [106].
Seven out of the 18 (39%) included reviews on cost-effectiveness concluded that telemedi-
cine interventions were cost-effective or cost saving compared to usual care. Five out of 18
reviews (28%) found that telemedicine might be cost-effective or cost saving, two out of 18
reviews (11%) were unable to arrive at a robust conclusion, and four out of 18 reviews (22%)
found that telemedicine was not cost-effective compared to usual care.
Most of the reviews on cost-effectiveness reported shortcomings in the economic literature
on telemedicine including a general paucity of economic studies, poor quality of available pri-
mary studies, poor outcome reporting, and lack of cost data. Some reviews included partial
economic evaluations (e.g. cost anaylsis) which fails to provide the requisite detail for cost-
effectiveness, therefore the evidence from these reviews were limited.
Six reviews included economic evaluations that performed analysis from a societal perspec-
tive. This perspective is considered high quality because it includes both direct and indirect
PLOS ONE | https://doi.org/10.1371/journal.pone.0237585 August 13, 2020
15 / 24
PLOS ONE Telemedicine in the OECD: An umbrella review
costs that may be saved or incurred by patients and providers irrespective of who bears the
costs or receives the benefits. Ideally, a societal perspective should be combined with a budget
impact assessment, which will help improve the feasibility of implementing interventions [71].
Most reviews included analysis from a provider perspective, where only costs to the health care
system were included and costs that may be incurred by patients and carers were excluded,
such as for example, travel time, productivity or wage loss, accommodation costs or co-pay-
ments [66]. For the reviews that included cost analysis, some studies failed to take into account
the cost of treatments or devices already in place, and implementation costs [61,74]. Five
reviews included time horizons from economic evaluations. Time horizons ranged from thirty
days to lifetime, see Table 3. Longer time horizons are often useful, especially when evaluating
chronic conditions where costs and outcomes are likely to incur later. However, telemedicine
is a rapidly evolving field, and longer time horizons may become meaningless as the telemedi-
cal technology becomes obsolete. Therefore, infrastructure costs such as bandwith and security
may be more relevant than the cost of equipment when evaluating the costs and cost-effective-
ness of telemedicine.
There was some discordant findings in reviews. One review found that the average consul-
tation time for telemedicine was longer than for a face-to-face consultation, whilst another
review found that telemedicine consultations were quicker [75, 60]. This is an important con-
sideration in the context of physician reimbursement and emphasizes the context specificity of
cost-effectiveness studies, which varies depending on the accuracy of cost information, spe-
cialty area, setting, and health care systems. It is also important to consider how much societal
value is placed on the improvement of a particular health status or population e.g. the cost-
effectiveness threshold. In addition, as observed in one review, health care systems may con-
sider equitable factors, such as initial additional costs for delivering healthcare to remote
regions, to reduce health inequalities with potentially larger future cost savings [69]. These fac-
tors all limit the comparablilty and generalizability of findings and make it difficult to reach a
robust conclusion of whether telemedicine services are cost-effective or not.
This umbrella review finds that patients demonstrate high acceptability and satisfaction
with telemedicine. Telemedicine affords patients convenience, and independence to manage
their conditions at home or within their communities. The cost of devices and technological
illiteracy may present a barrier to patient uptake of telemedicine especially in low-income pop-
ulations and low resource settings. Other identified patient barriers to the wider use of tele-
medicine associated with high dropout rates and attrition are modifiable from the
implementers perspective, and include technical and usability challenges. This umbrella review
also finds that telemedicine is feasible and identifies several factors that may affect its
sustainability.
Limitations
This review intended to provide a summary of the current evidence on telemedicine within
the OECD, however results within this review are limited by several factors. First, by synthesiz-
ing evidence from reviews, and relying on review author interpretations, some important
details that may have been included in primary studies was lost. Furthermore, a small number
of primary studies (5%) from non-OECD countries were included in reviews. Second, the
overall methodological quality of the included reviews was low. In particular, several economic
evaluations were limited by poor reporting quality or unavailable data. Third, publication bias
was noted in some reviews, meaning that positive results on the effect of telemedicine were
more likely to be published. Fourth, language bias may have been introduced by the exclusion
of non-English language studies. Finally, telemedicine is a rapidly evolving field, and although
PLOS ONE | https://doi.org/10.1371/journal.pone.0237585 August 13, 2020
16 / 24
PLOS ONE Telemedicine in the OECD: An umbrella review
we attempted to include the most current review level evidence, many systematic reviews
included much older primary studies. The inclusion of evidence from older primary studies
may lead to conclusions regarding clinical and cost-effectiveness that are no longer timely. For
example, the costs associated with telemedicine may be lower today compared to a decade ago
due the lower costs of the technology that contribute to the ubiquitous use of smart phones, fit-
ness monitors, and high-speed broadband. Improved clinical practice and revised guidelines
within this period may also narrow the gap between the observed effect of telemedicine and
face-to-face care. These factors may limit the generalizability of the evidence presented in this
review.
Conclusion
The literature on telemedicine is diverse and rapidly growing. Most of the included systematic
reviews, and primary studies focus on the effectiveness of telemedicine interventions, however
decision makers are also concerned with cost-effectiveness, patient-centredness and imple-
mentation. This umbrella review finds that telemedicine is comparable to face-to-face care
across several disease and specialtiy areas. Patients find interventions acceptable and report
high satisfaction levels, although barriers to uptake and use, such as lack of training and usabil-
ity problems remain. The overall quality of the evidence included in this umbrella review was
low. Future systematic reviews may be improved by reporting protocols, including a list of
excluded full-text studies, and sources of funding of primary studies in supporting materials.
Interventions should be clearly defined, and stratifed by the type of telemedicine used (e.g
remote monitoring vs. video consultation) to provide results that are more useful for practice,
research and decision making.
The current evidence on cost-effectiveness is limited, both in availability and generalizabil-
ity. Policy intervention may be required through regulation to ensure that DTC telemedicine
interventions are user-friendly, clinically beneficial, and can integrate into existing health care
systems. Health care reimbursement models may need to be adapted, acknowledging that tele-
medicine may take as much time as face-to-face care. Initial set-up costs could be an important
barrier to telemedicine uptake, although these costs are likely to be offset in the medium to
long-term. All these factors are important if telemedicine interventions are to deliver person-
centred high-quality cost-effective care.
The provision of health care for sicker patients and those who live in rural areas is also rele-
vant in the context of telemedicine. There is often a trade-off between efficiency and equity
when delivering care to remote populations. An unintended consequence of telemedicine is
that it is likely to reach relatively healthier and tech savvy patients in urban areas, rather than
sicker digitally excluded patients in rural areas. Again, policy oversight may be required to
improve access to health care, and to ensure that telemedicine is inclusive in reaching those
patients who already experience limitations in accessing conventional models of care.
Supporting information
S1 File. Search strategy and reviews excluded at full text stage.
(DOCX)
S2 File. Characteristics of included systematic reviews.
(DOCX)
S3 File. Methodological quality of included systematic reviews.
(DOCX)
PLOS ONE | https://doi.org/10.1371/journal.pone.0237585 August 13, 2020
17 / 24
PLOS ONE Telemedicine in the OECD: An umbrella review
Acknowledgments
The views expressed in this Article are the authors’ alone and do not necessarily represent the
views, decisions, or policies of the Organisation for Economic Co-operation and Development
(OECD), nor the views of its member countries.
Author Contributions
Conceptualization: Nkiruka D. Eze, Tiago Cravo Oliveira Hashiguchi.
Data curation: Nkiruka D. Eze.
Formal analysis: Nkiruka D. Eze, Tiago Cravo Oliveira Hashiguchi.
Funding acquisition: Ce´u Mateus.
Investigation: Nkiruka D. Eze, Tiago Cravo Oliveira Hashiguchi.
Methodology: Nkiruka D. Eze, Ce´u Mateus, Tiago Cravo Oliveira Hashiguchi.
Project administration: Nkiruka D. Eze.
Resources: Nkiruka D. Eze.
Software: Nkiruka D. Eze.
Supervision: Ce´u Mateus, Tiago Cravo Oliveira Hashiguchi.
Validation: Nkiruka D. Eze.
Visualization: Nkiruka D. Eze.
Writing – original draft: Nkiruka D. Eze, Tiago Cravo Oliveira Hashiguchi.
Writing – review & editing: Nkiruka D. Eze, Ce´u Mateus, Tiago Cravo Oliveira Hashiguchi.
References
1.
Flodgren G, Rachas A, Farmer AJ, Inzitari M, Shepperd S. Interactive telemedicine: effects on profes-
sional practice and health care outcomes. Cochrane Database Syst Rev 2015:CD002098. https://doi.
org/10.1002/14651858.CD002098.pub2 PMID: 26343551
2. World Health Organisation. Health and Sustainable Development: Telehealth 2020. https://www.who.
int/sustainable-development/health-sector/strategies/telehealth/en/.
3. Shea BJ, Reeves BC, Wells G, Thuku M, Hamel C, Moran J, et al. AMSTAR 2: a critical appraisal tool
for systematic reviews that include randomised or non-randomised studies of healthcare interventions,
or both. BMJ 2017; 358:j4008. https://doi.org/10.1136/bmj.j4008 PMID: 28935701
4. Wallace S, Wyatt J, Taylor P. Telemedicine in the NHS for the millennium and beyond. Postgrad Med
J 1998; 74:721–8. https://doi.org/10.1136/pgmj.74.878.721 PMID: 10320886
5. Agostini M, Moja L, Banzi R, Pistotti V, Tonin P, Venneri A, et al. Telerehabilitation and recovery of
motor function: a systematic review and meta-analysis. J Telemed Telecare 2015; 21:202–13. https://
doi.org/10.1177/1357633X15572201 PMID: 25712109
6.
Toma T, Athanasiou T, Harling L, Darzi A, Ashrafian H, T. T, et al. Online social networking services in
the management of patients with diabetes mellitus: Systematic review and meta-analysis of rando-
mised controlled trials. Diabetes Res Clin Pract 2014; 106:200–11. https://doi.org/10.1016/j.diabres.
2014.06.008 PMID: 25043399
7. Sherifali D, Nerenberg KA, Wilson S, Semeniuk K, Ali MU, Redman LM, et al. The Effectiveness of
eHealth Technologies on Weight Management in Pregnant and Postpartum Women: Systematic
Review and Meta-Analysis. J Med Internet Res 2017; 19:e337. https://doi.org/10.2196/jmir.8006
PMID: 29030327
8. Speyer RR, Denman D, Hons B, Wilkes-Gillan S, Chen Y-WW, Bogaardt H, et al. Effects of telehealth
by allied health professionals and nurses in rural and remote areas: A systematic review and meta-
Analysis. J Rehabil Med 2018; 50:225–35. https://doi.org/10.2340/16501977-2297 PMID: 29257195
PLOS ONE | https://doi.org/10.1371/journal.pone.0237585 August 13, 2020
18 / 24
PLOS ONE Telemedicine in the OECD: An umbrella review
9.
10.
Tchero H, Noubou L, Becsangele B, Mukisi-Mukaza M, Retali G-R, Rusch E. Telemedicine in Diabetic
Foot Care: A Systematic Literature Review of Interventions and Meta-analysis of Controlled Trials. Int
J Low Extrem Wounds 2017; 16:274–83. https://doi.org/10.1177/1534734617739195 PMID:
29168418
Joiner KL, Nam S, Whittemore R. Lifestyle interventions based on the diabetes prevention program
delivered via eHealth: A systematic review and meta-analysis. Prev Med (Baltim) 2017; 100:194–207.
https://doi.org/10.1016/j.ypmed.2017.04.033.
11. Raman P, Shepherd E, Dowswell T, Middleton P, Crowther CA. Different methods and settings for glu-
cose monitoring for gestational diabetes during pregnancy. Cochrane Database Syst Rev 2017; 10:
CD011069. https://doi.org/10.1002/14651858.CD011069.pub2 PMID: 29081069
12. Rasekaba TM, Furler J, Blackberry I, Tacey M, Gray K, Lim K. Telemedicine interventions for gesta-
tional diabetes mellitus: A systematic review and meta-analysis. Diabetes Res Clin Pract 2015; 110:1–
9. https://doi.org/10.1016/j.diabres.2015.07.007 PMID: 26264410
13. Ming W-K, Mackillop LH, Farmer AJ, Loerup L, Bartlett K, Levy JC, et al. Telemedicine Technologies
for Diabetes in Pregnancy: A Systematic Review and Meta-Analysis. J Med Internet Res 2016; 18:
e290. https://doi.org/10.2196/jmir.6556 PMID: 27829574
14. Huang Z, Tao H, Meng Q, Jing L. Management of endocrine disease. Effects of telecare intervention
on glycemic control in type 2 diabetes: a systematic review and meta-analysis of randomized con-
trolled trials. Eur J Endocrinol 2015; 172:R93–101. https://doi.org/10.1530/EJE-14-0441 PMID:
25227131
15.
Liu S, Feng W, Chhatbar PY, Liu Y, Ji X, Ovbiagele B. Mobile health as a viable strategy to enhance
stroke risk factor control: A systematic review and meta-analysis. J Neurol Sci 2017; 378:140–5.
https://doi.org/10.1016/j.jns.2017.04.050 PMID: 28566151
16. Su D, Zhou J, Kelley MS, Michaud TL, Siahpush M, Kim J, et al. Does telemedicine improve treatment
outcomes for diabetes? A meta-analysis of results from 55 randomized controlled trials. Diabetes Res
Clin Pract 2016; 116:136–48. https://doi.org/10.1016/j.diabres.2016.04.019 PMID: 27321329
17.
18.
19.
Toma T, Athanasiou T, Harling L, Darzi A, Ashrafian H. Online social networking services in the man-
agement of patients with diabetes mellitus: systematic review and meta-analysis of randomised con-
trolled trials. Diabetes Res Clin Pract 2014; 106:200–11. https://doi.org/10.1016/j.diabres.2014.06.
008 PMID: 25043399
Zhai Y, Zhu W, Cai Y, Sun D, Zhao J. Clinical- and cost-effectiveness of telemedicine in type 2 diabe-
tes mellitus: a systematic review and meta-analysis. Medicine (Baltimore) 2014; 93:e312. https://doi.
org/10.1097/MD.0000000000000312.
Jeon E, Park H-A. Nursing Intervention using smartphone technologies; a systematic review and
meta-analysis. Stud Health Technol Inform 2015; 210:321–5. PMID: 25991158
20. Kotb A, Cameron C, Hsieh S, Wells G. Comparative effectiveness of different forms of telemedicine
for individuals with heart failure (HF): a systematic review and network meta-analysis. PLoS One
2015; 10:e0118681. https://doi.org/10.1371/journal.pone.0118681 PMID: 25714962
21. Widmer RJ, Collins NM, Collins CS, West CP, Lerman LO, Lerman A. Digital health interventions for
the prevention of cardiovascular disease: a systematic review and meta-analysis. Mayo Clin Proc
2015; 90:469–80. https://doi.org/10.1016/j.mayocp.2014.12.026 PMID: 25841251
22. Klersy C, Boriani G, De Silvestri A, Mairesse GH, Braunschweig F, Scotti V, et al. Effect of telemonitor-
ing of cardiac implantable electronic devices on healthcare utilization: a meta-analysis of randomized
controlled trials in patients with heart failure. Eur J Heart Fail 2016; 18:195–204. https://doi.org/10.
1002/ejhf.470 PMID: 26817628
23. Huang K, Liu W, He D, Huang B, Xiao D, Peng Y, et al. Telehealth interventions versus center-based
cardiac rehabilitation of coronary artery disease: A systematic review and meta-analysis. Eur J Prev
Cardiol 2015; 22:959–71. https://doi.org/10.1177/2047487314561168 PMID: 25488550
24. Rawstorn JC, Gant N, Direito A, Beckmann C, Maddison R. Telehealth exercise-based cardiac reha-
bilitation: a systematic review and meta-analysis. Heart 2016; 102:1183–92. https://doi.org/10.1136/
heartjnl-2015-308966 PMID: 26936337
25.
Feltner C, Jones CD, Cene´ CW, Zheng Z-J, Sueta CA, Coker-Schwimmer EJL, et al. Transitional care
interventions to prevent readmissions for persons with heart failure: a systematic review and meta-
analysis. Ann Intern Med 2014; 160:774–84. https://doi.org/10.7326/M14-0083 PMID: 24862840
26. Van Spall HGC, Rahman T, Mytton O, Ramasundarahettige C, Ibrahim Q, Kabali C, et al. Comparative
effectiveness of transitional care services in patients discharged from the hospital with heart failure: a
systematic review and network meta-analysis. Eur J Heart Fail 2017; 19:1427–43. https://doi.org/10.
1002/ejhf.765 PMID: 28233442
PLOS ONE | https://doi.org/10.1371/journal.pone.0237585 August 13, 2020
19 / 24
PLOS ONE Telemedicine in the OECD: An umbrella review
27. Merriel SWD, Andrews V, Salisbury C. Telehealth interventions for primary prevention of cardiovascu-
lar disease: a systematic review and meta-analysis. Prev Med (Baltim) 2014; 64:88–95. https://doi.
org/10.1016/j.ypmed.2014.04.001.
28. Kepplinger J, Barlinn K, Deckert S, Scheibe M, Bodechtel U, Schmitt J. Safety and efficacy of throm-
bolysis in telestroke: A systematic review and meta-analysis. Neurology 2016; 87:1344–51. https://
doi.org/10.1212/WNL.0000000000003148 PMID: 27566746
29.
Lee M, Wang M, Liu J, Holbrook A. Do telehealth interventions improve oral anticoagulation manage-
ment? A systematic review and meta-analysis. J Thromb Thrombolysis 2018; 45:325–36. https://doi.
org/10.1007/s11239-018-1609-2 PMID: 29350322
30. Cottrell MA, Galea OA, O’Leary SP, Hill AJ, Russell TG. Real-time telerehabilitation for the treatment
of musculoskeletal conditions is effective and comparable to standard practice: a systematic review
and meta-analysis. Clin Rehabil 2017; 31:625–38. https://doi.org/10.1177/0269215516645148 PMID:
27141087
31.
32.
van Egmond MA, van der Schaaf M, Vredeveld T, Vollenbroek-Hutten MMR, van Berge Henegouwen
MI, Klinkenbijl JHG, et al. Effectiveness of physiotherapy with telerehabilitation in surgical patients: a
systematic review and meta-analysis. Physiotherapy 2018; 104:277–98. https://doi.org/10.1016/j.
physio.2018.04.004 PMID: 30030037
Lundell S, Holmner Å, Rehn B, Nyberg A, Wadell K. Telehealthcare in COPD: a systematic review and
meta-analysis on physical outcomes and dyspnea. Respir Med 2015; 109:11–26. https://doi.org/10.
1016/j.rmed.2014.10.008 PMID: 25464906
33. Dario AB, Moreti Cabral A, Almeida L, Ferreira ML, Refshauge K, Simic M, et al. Effectiveness of tele-
health-based interventions in the management of non-specific low back pain: a systematic review with
meta-analysis. Spine J Off J North Am Spine Soc 2017; 17:1342–51. https://doi.org/10.1016/j.spinee.
2017.04.008.
34. Agostini M, Moja L, Banzi R, Pistotti V, Tonin P, Venneri A, et al. Telerehabilitation and recovery of
motor function: a systematic review and meta-analysis. J Telemed Telecare 2015; 21:202–13. https://
doi.org/10.1177/1357633X15572201 PMID: 25712109
35. Adamse C, Dekker-Van Weering MG, van Etten-Jamaludin FS, Stuiver MM. The effectiveness of
exercise-based telemedicine on pain, physical activity and quality of life in the treatment of chronic
pain: A systematic review. J Telemed Telecare 2018; 24:511–26. https://doi.org/10.1177/
1357633X17716576 PMID: 28696152
36. Seiler A, Klaas V, Tro¨ster G, Fagundes CP. eHealth and mHealth interventions in the treatment of
fatigued cancer survivors: A systematic review and meta-analysis. Psychooncology 2017; 26:1239–
53. https://doi.org/10.1002/pon.4489 PMID: 28665554
37. Stratton E, Lampit A, Choi I, Calvo RA, Harvey SB, Glozier N. Effectiveness of eHealth interventions
for reducing mental health conditions in employees: A systematic review and meta-analysis. PLoS
One 2017; 12:e0189904. https://doi.org/10.1371/journal.pone.0189904 PMID: 29267334
38. Nair U, Armfield NR, Chatfield MD, Edirippulige S. The effectiveness of telemedicine interventions to
address maternal depression: A systematic review and meta-analysis. J Telemed Telecare 2018;
24:639–50. https://doi.org/10.1177/1357633X18794332 PMID: 30343660
39.
Thabrew H, Stasiak K, Hetrick SE, Wong S, Huss JH, Merry SN. E-Health interventions for anxiety
and depression in children and adolescents with long-term physical conditions. Cochrane Database
Syst Rev 2018; 8:CD012489. https://doi.org/10.1002/14651858.CD012489.pub2 PMID: 30110718
40. Vigerland S, Lenhard F, Bonnert M, Lalouni M, Hedman E, Ahlen J, et al. Internet-delivered cognitive
behavior therapy for children and adolescents: A systematic review and meta-analysis. Clin Psychol
Rev 2016; 50:1–10. https://doi.org/10.1016/j.cpr.2016.09.005 PMID: 27668988
41. Deady M, Choi I, Calvo RA, Glozier N, Christensen H, Harvey SB. eHealth interventions for the pre-
vention of depression and anxiety in the general population: a systematic review and meta-analysis.
BMC Psychiatry 2017; 17:310. https://doi.org/10.1186/s12888-017-1473-1 PMID: 28851342
42.
43.
van Beugen S, Ferwerda M, Hoeve D, Rovers MM, Spillekom-van Koulil S, van Middendorp H, et al.
Internet-Based Cognitive Behavioral Therapy for Patients With Chronic Somatic Conditions: A Meta-
Analytic Review. J Med Internet Res 2014; 16:e88. https://doi.org/10.2196/jmir.2777 PMID: 24675372
Linde K, Sigterman K, Kriston L, Ru¨ cker G, Jamil S, Meissner K, et al. Effectiveness of psychological
treatments for depressive disorders in primary care: systematic review and meta-analysis. Ann Fam
Med 2015; 13:56–68. https://doi.org/10.1370/afm.1719 PMID: 25583894
44. Seyffert M, Lagisetty P, Landgraf J, Chopra V, Pfeiffer PN, Conte ML, et al. Internet-Delivered Cogni-
tive Behavioral Therapy to Treat Insomnia: A Systematic Review and Meta-Analysis. PLoS One 2016;
11:e0149139. https://doi.org/10.1371/journal.pone.0149139 PMID: 26867139
45. Wootton BM. Remote cognitive-behavior therapy for obsessive-compulsive symptoms: A meta-analy-
sis. Clin Psychol Rev 2016; 43:103–13. https://doi.org/10.1016/j.cpr.2015.10.001 PMID: 26494179
PLOS ONE | https://doi.org/10.1371/journal.pone.0237585 August 13, 2020
20 / 24
PLOS ONE Telemedicine in the OECD: An umbrella review
46. Oosterveen E, Tzelepis F, Ashton L, Hutchesson MJ. A systematic review of eHealth behavioral inter-
ventions targeting smoking, nutrition, alcohol, physical activity and/or obesity for young adults. Prev
Med (Baltim) 2017; 99:197–206. https://doi.org/10.1016/j.ypmed.2017.01.009.
47. Hakala S, Rintala A, Immonen J, Karvanen J, Heinonen A, Sjo¨gren T. Effectiveness of physical activity
promoting technology-based distance interventions compared to usual care. Systematic review, meta-
analysis and meta-regression. Eur J Phys Rehabil Med 2017; 53:953–67. https://doi.org/10.23736/
S1973-9087.17.04585-3 PMID: 28466628
48. Direito A, Carrac¸a E, Rawstorn J, Whittaker R, Maddison R. mHealth Technologies to Influence Physi-
cal Activity and Sedentary Behaviors: Behavior Change Techniques, Systematic Review and Meta-
Analysis of Randomized Controlled Trials. Ann Behav Med A Publ Soc Behav Med 2017; 51:226–39.
https://doi.org/10.1007/s12160-016-9846-0.
49. Kelly JT, Reidlinger DP, Hoffmann TC, Campbell KL. Telehealth methods to deliver dietary interven-
tions in adults with chronic disease: a systematic review and meta-analysis. Am J Clin Nutr 2016;
104:1693–702. https://doi.org/10.3945/ajcn.116.136333 PMID: 27935523
50. Marx W, Kelly JT, Crichton M, Craven D, Collins J, Mackay H, et al. Is telehealth effective in managing
malnutrition in community-dwelling older adults? A systematic review and meta-analysis. Maturitas
2018; 111:31–46. https://doi.org/10.1016/j.maturitas.2018.02.012 PMID: 29673830
51. Hutchesson MJ, Rollo ME, Krukowski R, Ells L, Harvey J, Morgan PJ, et al. eHealth interventions for
the prevention and treatment of overweight and obesity in adults: a systematic review with meta-analy-
sis. Obes Rev An Off J Int Assoc Study Obes 2015; 16:376–92. https://doi.org/10.1111/obr.12268.
52.
Lee J, Piao M, Byun A, Kim J. A Systematic Review and Meta-Analysis of Intervention for Pediatric
Obesity Using Mobile Technology. Stud Health Technol Inform 2016; 225:491–4. PMID: 27332249
53. Sherifali D, Nerenberg KA, Wilson S, Semeniuk K, Ali MU, Redman LM, et al. The Effectiveness of
eHealth Technologies on Weight Management in Pregnant and Postpartum Women: Systematic
Review and Meta-Analysis. J Med Internet Res 2017; 19:e337. https://doi.org/10.2196/jmir.8006
PMID: 29030327
54. Hui CY, Walton R, McKinstry B, Jackson T, Parker R, Pinnock H. The use of mobile applications to
support self-management for people with asthma: a systematic review of controlled studies to identify
features associated with clinical effectiveness and adherence. J Am Med Informatics Assoc JAMIA
2017; 24:619–32. https://doi.org/10.1093/jamia/ocw143.
55.
Zhao J, Zhai Y-K, Zhu W-J, Sun D-X. Effectiveness of Telemedicine for Controlling Asthma Symp-
toms: A Systematic Review and Meta-analysis. Telemed J E-Health Off J Am Telemed Assoc 2015;
21:484–92. https://doi.org/10.1089/tmj.2014.0119.
56. McLean G, Murray E, Band R, Moffat KR, Hanlon P, Bruton A, et al. Interactive digital interventions to
promote self-management in adults with asthma: systematic review and meta-analysis. BMC Pulm
Med 2016; 16:83. https://doi.org/10.1186/s12890-016-0248-7 PMID: 27215329
57. Cruz J, Brooks D, Marques A. Home telemonitoring effectiveness in COPD: a systematic review. Int J
Clin Pract 2014; 68:369–78. https://doi.org/10.1111/ijcp.12345 PMID: 24472009
58. Huang VW, Reich KM, Fedorak RN. Distance management of inflammatory bowel disease: system-
atic review and meta-analysis. World J Gastroenterol 2014; 20:829–42. https://doi.org/10.3748/wjg.
v20.i3.829 PMID: 24574756
59. Speyer R, Denman D, Wilkes-Gillan S, Chen Y-W, Bogaardt H, Kim J-H, et al. Effects of telehealth by
allied health professionals and nurses in rural and remote areas: A systematic review and meta-analy-
sis. J Rehabil Med 2018; 50:225–35. https://doi.org/10.2340/16501977-2297 PMID: 29257195
60.
Thomas S-M, Jeyaraman MM, Jeyaraman M, Hodge WG, Hutnik C, Costella J, et al. The effective-
ness of teleglaucoma versus in-patient examination for glaucoma screening: a systematic review and
meta-analysis. PLoS One 2014; 9:e113779. https://doi.org/10.1371/journal.pone.0113779 PMID:
25479593
61. McDougall JA, Ferucci ED, Glover J, Fraenkel L. Telerheumatology: A Systematic Review. Arthritis
Care Res (Hoboken) 2017; 69:1546–57. https://doi.org/10.1002/acr.23153.
62. Musiat P, Tarrier N. Collateral outcomes in e-mental health: a systematic review of the evidence for
added benefits of computerized cognitive behavior therapy interventions for mental health. Psychol
Med 2014; 44:3137–50. https://doi.org/10.1017/S0033291714000245 PMID: 25065947
63. Akiyama M, Yoo B-K. A Systematic Review of the Economic Evaluation of Telemedicine in Japan. J
Prev Med Public Health 2016; 49:183–96. https://doi.org/10.3961/jpmph.16.043 PMID: 27499161
64.
Iribarren SJ, Cato K, Falzon L, Stone PW. What is the economic evidence for mHealth? A systematic
review of economic evaluations of mHealth solutions. PLoS One 2017; 12:e0170581. https://doi.org/
10.1371/journal.pone.0170581 PMID: 28152012
PLOS ONE | https://doi.org/10.1371/journal.pone.0237585 August 13, 2020
21 / 24
PLOS ONE Telemedicine in the OECD: An umbrella review
65.
Lo´pez-Villegas A, Catala´ n-Matamoros D, Martı´n-Saborido C, Villegas-Tripiana I, Robles-Musso E. A
Systematic Review of Economic Evaluations of Pacemaker Telemonitoring Systems. Rev Esp Cardiol
(Engl Ed) 2016; 69:125–33. https://doi.org/10.1016/j.rec.2015.06.020.
66. Snoswell C, Finnane A, Janda M, Soyer HP, Whitty JA. Cost-effectiveness of Store-and-Forward Tele-
dermatology: A Systematic Review. JAMA Dermatology 2016; 152:702–8. https://doi.org/10.1001/
jamadermatol.2016.0525 PMID: 27074289
67. Michaud TL, Zhou J, McCarthy MA, Siahpush M, Su D. COSTS OF HOME-BASED TELEMEDICINE
PROGRAMS: A SYSTEMATIC REVIEW. Int J Technol Assess Health Care 2018; 34:410–8. https://
doi.org/10.1017/S0266462318000454 PMID: 30058505
68.
de la Torre-Dı´ez I, Lo´pez-Coronado M, Vaca C, Aguado JS, de Castro C. Cost-utility and cost-effec-
tiveness studies of telemedicine, electronic, and mobile health systems in the literature: a systematic
review. Telemed J E-Health Off J Am Telemed Assoc 2015; 21:81–5. https://doi.org/10.1089/tmj.
2014.0053.
69. Estai M, Kanagasingam Y, Tennant M, Bunt S. A systematic review of the research evidence for the
benefits of teledentistry. J Telemed Telecare 2018; 24:147–56. https://doi.org/10.1177/
1357633X16689433 PMID: 28118778
70. Grustam AS, Severens JL, van Nijnatten J, Koymans R, Vrijhoef HJM. COST-EFFECTIVENESS OF
TELEHEALTH INTERVENTIONS FOR CHRONIC HEART FAILURE PATIENTS: A LITERATURE
REVIEW. Int J Technol Assess Health Care 2014; 30:59–68. https://doi.org/10.1017/
S0266462313000779 PMID: 24495581
71. Udsen FW, Hejlesen O, Ehlers LH. A systematic review of the cost and cost-effectiveness of telehealth
for patients suffering from chronic obstructive pulmonary disease. J Telemed Telecare 2014; 20:212–
20. https://doi.org/10.1177/1357633X14533896 PMID: 24803277
72.
Liddy C, Drosinis P, Keely E. Electronic consultation systems: worldwide prevalence and their impact
on patient care-a systematic review. Fam Pract 2016; 33:274–85. https://doi.org/10.1093/fampra/
cmw024 PMID: 27075028
73. Sanyal C, Stolee P, Juzwishin D, Husereau D. Economic evaluations of eHealth technologies: A sys-
tematic review. PLoS One 2018; 13:e0198112. https://doi.org/10.1371/journal.pone.0198112 PMID:
29897921
74.
75.
Jackson BD, Gray K, Knowles SR, De Cruz P. EHealth Technologies in Inflammatory Bowel Disease:
A Systematic Review. J Crohns Colitis 2016; 10:1103–21. https://doi.org/10.1093/ecco-jcc/jjw059
PMID: 26928960
Fuertes-Guiro´ F, Girabent-Farre´ s M. Opportunity cost of the dermatologist’s consulting time in the eco-
nomic evaluation of teledermatology. J Telemed Telecare 2017; 23:657–64. https://doi.org/10.1177/
1357633X16660876 PMID: 27450572
76. Hameed AS, Sauermann S, Schreier G. The impact of adherence on costs and effectiveness of tele-
medical patient management in heart failure: a systematic review. Appl Clin Inform 2014; 5:612–20.
https://doi.org/10.4338/ACI-2014-04-RA-0037 PMID: 25298802
77. Bashshur RL, Howell JD, Krupinski EA, Harms KM, Bashshur N, Doarn CR. The Empirical Founda-
tions of Telemedicine Interventions in Primary Care. Telemed J E-Health Off J Am Telemed Assoc
2016; 22:342–75. https://doi.org/10.1089/tmj.2016.0045.
78. Grist R, Porter J, Stallard P. Mental Health Mobile Apps for Preadolescents and Adolescents: A Sys-
tematic Review. J Med Internet Res 2017; 19:e176. https://doi.org/10.2196/jmir.7332 PMID:
28546138
79. Berry N, Lobban F, Emsley R, Bucci S. Acceptability of Interventions Delivered Online and Through
Mobile Phones for People Who Experience Severe Mental Health Problems: A Systematic Review. J
Med Internet Res 2016; 18:e121. https://doi.org/10.2196/jmir.5250 PMID: 27245693
80. Hamilton SJ, Mills B, Birch EM, Thompson SC. Smartphones in the secondary prevention of cardio-
vascular disease: a systematic review. BMC Cardiovasc Disord 2018; 18:25. https://doi.org/10.1186/
s12872-018-0764-x PMID: 29415680
81. Brunton L, Bower P, Sanders C. The Contradictions of Telehealth User Experience in Chronic
Obstructive Pulmonary Disease (COPD): A Qualitative Meta-Synthesis. PLoS One 2015; 10:
e0139561. https://doi.org/10.1371/journal.pone.0139561 PMID: 26465333
82.
Trettel A, Eissing L, Augustin M. Telemedicine in dermatology: findings and experiences worldwide—
a systematic literature review. J Eur Acad Dermatology Venereol JEADV 2018; 32:215–24. https://doi.
org/10.1111/jdv.14341.
83. Berrouiguet S, Baca-Garcı´a E, Brandt S, Walter M, Courtet P. Fundamentals for Future Mobile-Health
(mHealth): A Systematic Review of Mobile Phone and Web-Based Text Messaging in Mental Health. J
Med Internet Res 2016; 18:e135. https://doi.org/10.2196/jmir.5066 PMID: 27287668
PLOS ONE | https://doi.org/10.1371/journal.pone.0237585 August 13, 2020
22 / 24
PLOS ONE Telemedicine in the OECD: An umbrella review
84. Cruz J, Brooks D, Marques A. Home telemonitoring in COPD: a systematic review of methodologies
and patients’ adherence. Int J Med Inform 2014; 83:249–63. https://doi.org/10.1016/j.ijmedinf.2014.
01.008 PMID: 24529402
85.
Liptrott S, Bee P, Lovell K. Acceptability of telephone support as perceived by patients with cancer: A
systematic review. Eur J Cancer Care (Engl) 2018; 27. https://doi.org/10.1111/ecc.12643.
86. Cox A, Lucas G, Marcu A, Piano M, Grosvenor W, Mold F, et al. Cancer Survivors’ Experience With
Telehealth: A Systematic Review and Thematic Synthesis. J Med Internet Res 2017; 19:e11. https://
doi.org/10.2196/jmir.6575 PMID: 28069561
87. Caffery LJ, Bradford NK, Wickramasinghe SI, Hayman N, Smith AC. Outcomes of using telehealth for
the provision of healthcare to Aboriginal and Torres Strait Islander people: a systematic review. Aust N
Z J Public Health 2017; 41:48–53. https://doi.org/10.1111/1753-6405.12600 PMID: 27868300
88. Gorst SL, Armitage CJ, Brownsell S, Hawley MS. Home telehealth uptake and continued use among
heart failure and chronic obstructive pulmonary disease patients: a systematic review. Ann Behav
Med A Publ Soc Behav Med 2014; 48:323–36. https://doi.org/10.1007/s12160-014-9607-x.
89. Slater H, Briggs A, Stinson J, Campbell JM. End user and implementer experiences of mHealth tech-
nologies for noncommunicable chronic disease management in young adults: a qualitative systematic
review protocol. JBI Database Syst Rev Implement Reports 2017; 15:2047–54. https://doi.org/10.
11124/JBISRIR-2016-003299.
90. Greenhalgh T, A’Court C, Shaw S. Understanding heart failure; explaining telehealth—a hermeneutic
systematic review. BMC Cardiovasc Disord 2017; 17:156. https://doi.org/10.1186/s12872-017-0594-2
PMID: 28615004
91. Macdonald EM, Perrin BM, Kingsley MI. Enablers and barriers to using two-way information technol-
ogy in the management of adults with diabetes: A descriptive systematic review. J Telemed Telecare
2018; 24:319–40. https://doi.org/10.1177/1357633X17699990 PMID: 28347218
92. Alvarado MM, Kum H-C, Gonzalez Coronado K, Foster MJ, Ortega P, Lawley MA. Barriers to Remote
Health Interventions for Type 2 Diabetes: A Systematic Review and Proposed Classification Scheme.
J Med Internet Res 2017; 19:e28. https://doi.org/10.2196/jmir.6382 PMID: 28193598
93. Wickramasinghe SI, Caffery LJ, Bradford NK, Smith AC. Enablers and barriers in providing teledia-
betes services for Indigenous communities: A systematic review. J Telemed Telecare 2016; 22:465–
71. https://doi.org/10.1177/1357633X16673267 PMID: 27799449
94. Radhakrishnan K, Xie B, Berkley A, Kim M. Barriers and Facilitators for Sustainability of Tele-Home-
care Programs: A Systematic Review. Health Serv Res 2016; 51:48–75. https://doi.org/10.1111/1475-
6773.12327 PMID: 26119048
95. Meurk C, Leung J, Hall W, Head BW, Whiteford H. Establishing and Governing e-Mental Health Care
in Australia: A Systematic Review of Challenges and A Call For Policy-Focussed Research. J Med
Internet Res 2016; 18:e10. https://doi.org/10.2196/jmir.4827 PMID: 26764181
96. Greenwood DA, Young HM, Quinn CC. Telehealth Remote Monitoring Systematic Review: Structured
Self-monitoring of Blood Glucose and Impact on A1C. J Diabetes Sci Technol 2014; 8:378–89. https://
doi.org/10.1177/1932296813519311 PMID: 24876591
97. Guise V, Anderson J, Wiig S. Patient safety risks associated with telecare: a systematic review and
narrative synthesis of the literature. BMC Health Serv Res 2014; 14:588. https://doi.org/10.1186/
s12913-014-0588-z PMID: 25421823
98. Bradford NK, Caffery LJ, Smith AC. Telehealth services in rural and remote Australia: a systematic
review of models of care and factors influencing success and sustainability. Rural Remote Health
2016; 16:4268. PMID: 27817199
99. Kampmeijer R, Pavlova M, Tambor M, Golinowska S, Groot W. The use of e-health and m-health tools
in health promotion and primary prevention among older adults: a systematic literature review. BMC
Health Serv Res 2016; 16 Suppl 5:290. https://doi.org/10.1186/s12913-016-1522-3.
100.
Irving M, Stewart R, Spallek H, Blinkhorn A. Using teledentistry in clinical practice as an enabler to
improve access to clinical care: A qualitative systematic review. J Telemed Telecare 2018; 24:129–46.
https://doi.org/10.1177/1357633X16686776 PMID: 28092220
101. Block VAJ, Pitsch E, Tahir P, Cree BAC, Allen DD, Gelfand JM. Remote Physical Activity Monitoring in
Neurological Disease: A Systematic Review. PLoS One 2016; 11:e0154335. https://doi.org/10.1371/
journal.pone.0154335 PMID: 27124611
102.
Ito J, Edirippulige S, Aono T, Armfield NR. The use of telemedicine for delivering healthcare in Japan:
Systematic review of literature published in Japanese and English languages. J Telemed Telecare
2017; 23:828–34. https://doi.org/10.1177/1357633X17732801 PMID: 29081269
PLOS ONE | https://doi.org/10.1371/journal.pone.0237585 August 13, 2020
23 / 24
PLOS ONE Telemedicine in the OECD: An umbrella review
103. Bruce AF, Mallow JA, Theeke LA. The use of teledermoscopy in the accurate identification of cancer-
ous skin lesions in the adult population: A systematic review. J Telemed Telecare 2018; 24:75–83.
https://doi.org/10.1177/1357633X16686770 PMID: 28056600
104. Gehring ND, McGrath P, Wozney L, Soleimani A, Bennett K, Hartling L, et al. Pediatric eMental health-
care technologies: a systematic review of implementation foci in research studies, and government
and organizational documents. Implement Sci IS 2017; 12:76. https://doi.org/10.1186/s13012-017-
0608-6 PMID: 28637479
105. Kapadia V, Ariani A, Li J, Ray PK. Emerging ICT implementation issues in aged care. Int J Med Inform
2015; 84:892–900. https://doi.org/10.1016/j.ijmedinf.2015.07.002 PMID: 26250987
106. Shigekawa E, Fix M, Corbett G, Roby DH, Coffman J. The Current State Of Telehealth Evidence: A
Rapid Review. Health Aff 2018; 37:1975–82. https://doi.org/10.1377/hlthaff.2018.05132.
PLOS ONE | https://doi.org/10.1371/journal.pone.0237585 August 13, 2020
24 / 24
PLOS ONE |
10.1371_journal.pone.0241200 | RESEARCH ARTICLE
Automated classification of bacterial cell sub-
populations with convolutional neural
networks
Denis Tamiev1, Paige E. Furman1, Nigel F. ReuelID
2*
1 Department of Biochemistry Biophysics and Molecular Biology, Iowa State University, Ames, Iowa, United
States of America, 2 Department of Chemical and Biological Engineering, Iowa State University, Ames, Iowa,
United States of America
* reuel@iastate.edu
Abstract
Quantification of phenotypic heterogeneity present amongst bacterial cells can be a chal-
lenging task. Conventionally, classification and counting of bacteria sub-populations is
achieved with manual microscopy, due to the lack of alternative, high-throughput, autono-
mous approaches. In this work, we apply classification-type convolutional neural networks
(cCNN) to classify and enumerate bacterial cell sub-populations (B. subtilis clusters). Here,
we demonstrate that the accuracy of the cCNN developed in this study can be as high as
86% when trained on a relatively small dataset (81 images). We also developed a new
image preprocessing algorithm, specific to fluorescent microscope images, which increases
the amount of training data available for the neural network by 72 times. By summing the
classified cells together, the algorithm provides a total cell count which is on parity with man-
ual counting, but is 10.2 times more consistent and 3.8 times faster. Finally, this work pres-
ents a complete solution framework for those wishing to learn and implement cCNN in their
synthetic biology work.
Introduction
Control of cell morphology, cycle state, and cell-to-cell interactions are some of the key design
goals in synthetic biology to develop ‘living materials’ [1–4]. As one example, bacteria endo-
spores can be incorporated into materials for revival and functional activation [5]. The pheno-
typic heterogeneity of cells needs to be controlled for precise functional materials such as
sensors or bacterial therapies [6, 7]. Classification of heterogeneity is currently assessed via
manual inspection of microscope images, as we did before with B. subtilis endospore activated
as fluorescent reporters (Fig 1A) [8]. Accurate, higher throughput quantification of cell sub-
populations is necessary for improved design of cell-based materials, sensors, and therapies.
Such tools would enable better understanding of underlying genetic mechanisms of the cell
sub-populations.
Current methods of classification work are insufficient to match the throughput of syn-
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: Tamiev D, Furman PE, Reuel NF (2020)
Automated classification of bacterial cell sub-
populations with convolutional neural networks.
PLoS ONE 15(10): e0241200. https://doi.org/
10.1371/journal.pone.0241200
Editor: Yan Chai Hum, University Tunku Abdul
Rahman, MALAYSIA
Received: August 6, 2020
Accepted: October 10, 2020
Published: October 26, 2020
Peer Review History: PLOS recognizes the
benefits of transparency in the peer review
process; therefore, we enable the publication of
all of the content of peer review and author
responses alongside final, published articles. The
editorial history of this article is available here:
https://doi.org/10.1371/journal.pone.0241200
Copyright: This is an open access article, free of all
copyright, and may be freely reproduced,
distributed, transmitted, modified, built upon, or
otherwise used by anyone for any lawful purpose.
The work is made available under the Creative
Commons CC0 public domain dedication.
Data Availability Statement: Data is available from
Figshare (doi: 10.6084/m9.figshare.13100225).
Funding: Internal Iowa State University funding.
thetic design in bacteria. While flow-assisted cell counting (FACS) can address larger,
PLOS ONE | https://doi.org/10.1371/journal.pone.0241200 October 26, 2020
1 / 13
PLOS ONE Automated classification of bacterial cell sub-populations with convolutional neural networks
Competing interests: The authors have declared
that no competing interests exist.
Fig 1. Classifying and counting bacterial sub-populations. (A) Examples of B. subtilis cells, forespores and spores
captured on microscope images, identified manually in prior work. (B) “Raw” or unprocessed fluorescence microscopy
image of B. subtilis cells during vegetative growth. (C) Examples of various cell clusters found on fluorescent
microscope images (i. Artifact (cell debris, impurity), ii. single cell, iii. four-cell cluster, and iv. ten-cell clusters.). (D)
Flow cytometry output fluorescence data for viability assay of B. subtilis cells. (i) Gated region contains signal from live
cells. Data presented in log (RFU) scale. (ii) Clusters of data on flow cytometry output from (Di) that likely associate
with different cell states (vegetative, spore) but are difficult to determine from the FACS data.
https://doi.org/10.1371/journal.pone.0241200.g001
eukaryotic, cells the method cannot be as easily applied to sorting bacterial cells which can be
up to 1,000 time smaller than eukaryotes. Moreover, bacteria cells can rapidly transition
through development states, such as B. subtilis which repeatedly transitions between vegetative
cell, forespore, and spore states (Fig 1A) [8]. Moreover, bacteria can form clusters of cells, espe-
cially biofilm forming B. subtilis (Fig 1B), which further confounds FACS based analysis which
can misrepresent a cluster of cells (or cluster of various cell types) as a single cell (Fig 1C) [9].
However, FACS can be used for some bacteria classification where there are clear fluorescent
reporters and distinct populations, such as classifying viability of B. subtilis (Fig 1Di) [10]
using the permeability to green (Syto 9) dye but not red (Propidium Iodide) as an indication
of cell viability [11]. However, when a mixture of cells, spores and forespores of B. subtilis are
subjected to this method more than two clusters of fluorescence signal are present in the flow
cytometry data (Fig 1Dii). Correlating these clusters of data to bacterial sub-types can be a
challenging task; thus quantification of bacterial cell populations is typically performed with
manual microscopy. However, there has been many advances in automated approaches to
analysis of microscope images including the use of deep learning.
Deep learning tools such as Convolutional Neural Networks (CNN) are well suited for auto-
mating classification and quantification of bacterial phenotypes present in microscope images;
they have been applied in other microbiology tasks such as classification of coccoliths formed
by various coccolithophores, stalked protozoa identification, and bacterial plankton classifica-
tion [12–15]. CNN transform an image volume to a linear output volume (holding the class
scores) using a stack of interconnected convolutional, pooling, and fully-connected layers. The
convolutional and fully-connected layers transform the images using weights and biases that
are tuned via gradient descent to match the CNN output (class scores) to the annotated train-
ing data set [16]. The architecture of the CNN is based on the number of these layers and how
they are stacked (see Supplement 3 in S1 File for summary of CNN framework used in this
work and solution steps). A successful demonstration of the capability of modern CNN during
the ImageNet Large Scale Visual Recognition Challenge (ILSVRC) in 2012 attracted interest to
this field and spiked the development of a variety of more advanced, dense convolutional neu-
ral network architectures (DCNN) [17, 18]. The main disadvantage in applying these dense
PLOS ONE | https://doi.org/10.1371/journal.pone.0241200 October 26, 2020
2 / 13
PLOS ONE Automated classification of bacterial cell sub-populations with convolutional neural networks
neural networks is that a large amount of data is required to achieve sufficient performance
due to the density of these networks (e.g. large number of weights to optimize) [19]. Simpler
CNN architectures such as LeNet-5, which has a 7 layer structure, is better suited for simple
images and sparse training sets [20, 21]. Fluorescent microscope images are typically single
channel (greyscale, unlike 3-channel RGB images) and relative low resolution; also, bacterial
cells lack high level features that would necessitate a dense CNN for proper generalization. In
addition, simple architectures can require significantly fewer images to achieve proper general-
ization [22]. One example of a CNN based on LeNet-5, modified to reflect recent progress in
CNN design such as implementation of non-saturating non-linearities (ReLU), local response
normalization, non-overlapping max-pooling and random dropout layers, was successfully
implemented in bacterial colony counting [23–27]. Large colonies of cells present on agar
plates have simple circular structure and lack high level features, similar to smaller bacterial
cell clusters. Nonetheless, there are unique challenges associated with efficient pre-processing
of more limited fluorescent microscope images and adapting CNN to analyze bacterial cell
clusters, the scope of this work.
In this work, we demonstrate classification and enumeration of bacterial (B. subtilis) sub-
populations present in fluorescent microscope images with a classification-type convolutional
neural network (cCNN) adapted from the previous, larger colony counting work [23]. To
achieve this, we present a more efficient method for microscope image preprocessing and aug-
mentation that improves the training efficiency from a limited set of annotated images. We
assess the accuracy and confidence of the algorithm for each sub-population type, and discuss
how the quality of the algorithm can be further improved. We benchmark this algorithm
against one established tool, ImageJ, for total count of cells. We compare the performance of
our cCNN algorithm, in terms of accuracy, speed and repeatability, to manual processing. By
archiving all our cCNN code and training data, this work can also serve as a clear example to
others wishing to implement cCNN for other classification tasks that are encountered when
trying to identify and count the effects of synthetic biology design.
Results and discussion
Image preprocessing
While the applications for convolutional neural networks are becoming more widespread and
the tools more freely available, these networks are typically designed to work with macroscopic
objects with many distinct features (cars, people, animals etc.) [28]. Microorganisms captured
on fluorescent microscope images are significantly different in appearance compared to
images of common objects (Fig 2A) as they have no real internal features, just the unique size
and outline of the cells. For this work we acquired 1,000 fluorescent microscope images (single
channel) of B. subtilis cells, and evaluated image preprocessing algorithms commonly used
with cCNNs. The objective was to identify data preparation methods that resulted in highest
performance of common neural networks.
The raw microscope images were segmented using an adaptive binary thresholding algo-
rithm to identify individual clusters of cells that were spatially segregated from each other (Fig
2A, Supplement 2 in S1 File). Some of these clusters were single cells, some were artifacts
(dust, debris, etc.) and others were multicellular clusters of cells (nascent biofilm) with a drop
off in number as the size of the cluster increased (i.e. rarer sub-population, see Supplement 2
in S1 File for population numbers of the segmented data). The varied sizes of the segmented
data presented a challenge as the majority of classic neural networks operate based on fixed
resolution images. As such, all images must be adjusted to a standard size prior to training the
network.
PLOS ONE | https://doi.org/10.1371/journal.pone.0241200 October 26, 2020
3 / 13
PLOS ONE Automated classification of bacterial cell sub-populations with convolutional neural networks
Fig 2. Image processing workflow. Raw fluorescent microscope images (a) were processed with a binary
segmentation algorithm, and clusters of bacterial cells were manually annotated. All image segments of cell clusters
were standardized to the same size with either (b) Null Bumper, (b) Blended or (d) Masked methods. These annotated
training images were passed to the cCNN to determine optimal network weights (e). The output of the network (from
image depicted in panel c) is a confidence value for each sub-class (A–artifact; I—X–single through ten cell cluster),
here presented in a radar chart (F). Major tics– 20% confidence increments, minor– 10%.
https://doi.org/10.1371/journal.pone.0241200.g002
Generally, there are three ways to standardize sizing of the image data. Images can be (1)
distorted (shrunk or stretched), (2) masked or (3) framed [29]. We chose to standardize
cropped images to a 200 by 200-pixel resolution, since most of the cell cluster segments did
not exceed these dimensions. Distortion of the cropped images changed the appearance of bac-
terial cells. Also, application of a mask on cropped images (i.e. applying a filter to remove back-
ground pixel data) altered the edges of clusters (Fig 2D). Therefore, we settled on framing
approaches with two different methods of filling in the frame. In the first framed approach
(referred to as “null bumper” or NB), the cropped images were placed in the center of a
200x200 canvas, and the void space was filled with 0 intensity pixels (Fig 2B). In the second, we
created a novel “blended bumper” (referred to as BB) where the frames of cropped images
were blended with local background pixels, which resulted in a much more consistent appear-
ance of microscope images (Fig 2C, Supplement 2 in S1 File).
The NB and BB image datasets were created by rotating captured images at right angles and
in a mirrored dimension, resulting in 8x times as many images. During training of NB and BB
networks, we observed that the latter was underfitting (Supplement 5 in S1 File). To reduce the
magnitude of underfitting, we trained a third network with images created by rotation at finer
angle which generated more training data (referred to as “advanced rotation” or AR). The AR
image dataset was created by rotating images by 10 degrees across both dimensions, then
blending those images with the same algorithm described for BB, resulting in 72x increase in
training images.
As mentioned above, the NB, BB, and AR data were used to train cCNNs (see Supplement 3
in S1 File for code and steps). For this study we selected a network architecture that was dem-
onstrated to work well in a similar classification tasks and used it with all three image datasets
[23]. The structure of this cCNN is discussed in Supplement 4 in S1 File. Briefly, the input of
the cCNN is an image (2D array of pixel intensities) and the output of the network are confi-
dence values (0%-100%) for each output class (artifact or one, two, three, and so on, through
ten cell-cluster). The top confidence was selected to indicate the predicted class (Fig 2F).
To evaluate the quality of the network, annotated images that were not used for training of
the network were passed through the network for classification. Accuracy (comparison of pre-
diction to annotated truth value) and confidence (% of how certain the cCNN is of its predic-
tion) were used as metrics to identify best methods of data preparation (NB, BB, vs. AR) and
overall performance of the network at cell cluster classification tasks.
PLOS ONE | https://doi.org/10.1371/journal.pone.0241200 October 26, 2020
4 / 13
PLOS ONE Automated classification of bacterial cell sub-populations with convolutional neural networks
Evaluation of networks
Accuracies of CNNs trained on NB, BB, and AR datasets were then evaluated with new images.
These results can be tabulated in common confusion matrices (Supplements 7 and 8 in S1
File) that count the precise predictions and number of each off-target predictions. Alterna-
tively, we can visualize this same data using a box and whisker plot (Fig 3), to show the most
frequent classifications and the directionality of variance. This plot maps the true (manually
annotated count, X- axis) against the algorithm predictions (Y-axis). It is observed that the
NB-trained network was much more likely to misclassify A, I and II classes compared to BB-
and AR-trained networks. Availability of more training data (AR-trained network), improved
the network’s performance in classifying III, IV, V-cell clusters.
The average (across all cluster types) accuracies of the blended bumper and null bumper
cCNNs were similar, 344% and 40% respectively (Supplement 8 in S1 File). The average accuracy
of the AR cCNN was 58%. Upon closer examination, the networks performed best at differentiat-
ing between an artifact, single cells and two cell clusters (cells joined together), with accuracies
reaching as high as 86%, 73% and 75% for the AR network (Fig 3, Supplement 8 in S1 File).
As another measure of performance, we assess the relative confidences of the NB, BB, and
AR trained cCNN on their predictions for each population sub-class (Fig 4). When examining
mean confidences for each individual output class, it becomes evident that the NB network
classified all cell clusters with lower confidence than BB and AR networks (Fig 4); the average
confidences of both BB and AR networks were 12% higher than that of the NB network across
all classes (Supplement 9 in S1 File). The most apparent improvement in prediction confidence
can be observed for smaller objects (artifacts, single cells and II-cell clusters), and small cell
clusters (III-cell cluster). At this small scale, there is a dramatic improvement in network’s per-
formance when more data is provided for training using the AR network. We attribute this to
the asymmetric features of these smaller clusters that present dramatically new images when
rotated. This effect is apparently diminished as the cluster size increases, as the NB and AR
Fig 3. Accuracy of NB, BB and AR trained cCNNs across all output classes (A, I-X cell clusters). Standard box and whisker levels are
used with the center mark (dotted line) indicating the median, the bottom and top edges of the box indicating the 25th and 75th
percentiles, respectively, the whiskers extending to the most extreme data points not considered outliers.
https://doi.org/10.1371/journal.pone.0241200.g003
PLOS ONE | https://doi.org/10.1371/journal.pone.0241200 October 26, 2020
5 / 13
PLOS ONE Automated classification of bacterial cell sub-populations with convolutional neural networks
Fig 4. Confidence of NB, BB and AR trained cCNN across all output classes (A, I-X). Standard box and whisker levels are used
with the center mark indicating the median, the bottom and top edges of the box indicating the 25th and 75th percentiles,
respectively, the whiskers extending to the most extreme data points not considered outliers, and the outliers plotted individually.
https://doi.org/10.1371/journal.pone.0241200.g004
perform at the same relative level. These larger clusters become more symmetric and thus rota-
tion does not present dramatically new images on which to train.
The apparent increase in performance at the higher cluster count may also be an artifact
due to the limitation of classes imposed by the network. There are only 11 outputs available
and thus anything larger would get placed in the largest feature class. However, this could also
be an effect of more features present on the larger cluster; additional images of large cell clus-
ters would have to be obtained to determine if this is a real effect or artifact.
The variability in accuracy and confidence between different classes can be explained by the
fact that the networks were trained on a limited number of training images which did not cap-
ture all spatial and rotational orientations of multicellular clusters. These could be further
improved using more training data, although at this scale the cCNN approach provides suit-
able confidence in distinguishing smaller cell clusters and artifacts especially when using the
AR training data. Additionally, the variability in sub-population count is less important when
each class is summed together to obtain a total count of cells, which is another common appli-
cation in synthetic biology applications, which we present next.
Comparing cCNN total cell count to ImageJ
As mentioned in the introduction, total count of bacterial cells, especially those prone to clus-
tering, is difficult to do with FACS and is an important metric in quantifying effects of syn-
thetic design. Here we benchmark our cCNN data (summing counts across classes) against
two existing methods: 1) a popular open source tool ImageJ that relies on a watershed algo-
rithm and 2) manual counting [30].
For benchmarking the automated approaches, 24 new microscope images were segmented
with a binary thresholding algorithm, and cell clusters were either classified and summed with a
cCNN (NB, BB and AR trained networks) or sub-segmented with a watershed algorithm and
counted with ImageJ (Supplement 10 in S1 File). For truth data we performed manual counting
(one person, averaging three independent counts) and compare the automated findings in a
parity chart (Fig 5). In this representation a slope equal to unity would indicate that the software
PLOS ONE | https://doi.org/10.1371/journal.pone.0241200 October 26, 2020
6 / 13
PLOS ONE Automated classification of bacterial cell sub-populations with convolutional neural networks
Fig 5. Comparing accuracy of manual (same person, average of 3 counts) to software (cCNN or ImageJ) counting
methods (Supplement 13 in S1 File). Each point represents an evaluated image. The units are the number of cells per
image.
https://doi.org/10.1371/journal.pone.0241200.g005
approach has the same evaluation as manual counting. NB-trained networks appear to over-
count cells on microscope images (slope of 0.75), BB-trained networks perform significantly
better (slope of 0.93) but some over counting is still present, and AR-trained networks count
equally well as person (slope of 1). ImageJ undercounts images, as evident by the slope of 1.2.
These results clearly demonstrate a significant improvement in average accuracy when
comparing NB and BB networks. This improvement is attributed to the quality of the data
used to train the network, further supporting that blending the background of images leads to
a more accurate network. Some of the neural networks created for this study were overcount-
ing cells on images that contained high cell counts. This can be explained by the fact that high
cell count images also contained more multi-cell clusters compared to low cell count images.
High cell count images resulted from cell cultures with high cell density, which leads to biofilm
formation in B. subtilis. As observed previously, the networks developed in this study exhibited
lower accuracy at predicting multi-cell clusters, compared to single or two-cell clusters. The
undercounting of ImageJ is attributed to its poor capacity to accurately sub-segment images of
biofilms with a watershed algorithm (Supplement 11 in S1 File). Surprisingly, while the AR
trained network had an average class prediction accuracy of 58%, when these sub-population
numbers are summed together the overall count is on par with manual counting.
Comparing consistency, repeatability, and speed of cCNN total cell count
to human performance
While the accuracy of the AR-trained network prepared in this study is similar to that of man-
ual for total cell counting, consistency, repeatability, and speed should vary. To quantify these
PLOS ONE | https://doi.org/10.1371/journal.pone.0241200 October 26, 2020
7 / 13
PLOS ONE Automated classification of bacterial cell sub-populations with convolutional neural networks
metrics and benchmark manual counting against the cCNN approach, five people were
recruited to count bacterial cells on printed microscope images that were used for this study.
Test subjects were given 25 images, 15 unique and 5 that appeared twice in random order
(Supplement 12 in S1 File). The study evaluated two parameters–(1) how the total bacterial cell
count on all 25 images compared between five people, and (2) how the cell counts compared
when test subjects were presented with duplicate images.
Variability of human test subjects that counted cells on 25 images was, on average, as high
as 39.5 cells per image (Fig 6A). As expected, larger variability was observed with images
that depicted more cells (Supplement 12 in S1 File). In comparison, variability of the cCNN
based algorithm on average did not exceed 3.7 cells per image. Average variability across all
images for human test subjects was 10.2 times greater than for cCNN based algorithm (Fig 6A,
inset).
Fig 6. Comparing consistency and repeatability of manual cell counting. (A) Standard deviation (STD) as a
measure of variance of cell counts across 25 microscope images (ordered in increasing number of cells per image)
using manual counting from five separate human counters (interoperator variability) benchmarked against the
advanced rotation (AR) trained cCNN algorithm. Average variance of human subjects was 10.2 times larger than that
of the cCNN algorithm (insert). (B) Standard deviation of each test subject when counting a duplicate image
randomized in their set (intraoperator variability). Images 1 through 5 presented in the order of increasing cell
population (Supplement 13 in S1 File). Total time required to count cells on all images for each operator is presented in
the pie chart (inset). Full circle is 1000 seconds, tick marks are set to 100 seconds. Each individual is represented with a
different color (Blue–Person 1; Orange–Person 2; Yellow–Person 3; Purple–Person 4; Green–Person 5). If the data
from a specific person is not present, then their variability was 0 cells/image.
https://doi.org/10.1371/journal.pone.0241200.g006
PLOS ONE | https://doi.org/10.1371/journal.pone.0241200 October 26, 2020
8 / 13
PLOS ONE Automated classification of bacterial cell sub-populations with convolutional neural networks
Variability due to human error is not uncommon in manual tasks. In some cases, differ-
ences between individuals can be normalized if there are patterns in human behavior. For
example, if one person is systematically under or over-counting, a correction factor (bias) can
be introduced to correct the results of a cohort. However, it is not always possible to correct
results, if the inconsistencies are random. In this study we evaluated inconsistency of specific
individuals on duplicate images, and compared it to that of the cCNN based algorithm. Results
show that individuals tend to count a different number of cells on the same image that was
duplicated (Fig 6B). The difference is reported as standard deviation between the two instances
for each individual (Supplement 12 in S1 File). The variability of individuals in some cases was
as high as 12.0 cells per image, or 10% (Fig 6B). In contrast, the variability of the algorithm
tested on duplicate images was 2.26 cells per image or 2% on average, and it ranged from 0.7 to
4.24, or 1–4% (Supplement 12 in S1 File). It is important to note that the network performs
with no variability when it is loaded in the memory of the computer. We traced algorithm vari-
ability to the process of loading the network with a computer vision library. It is possible that
the global system variables of the device impact how the network is loaded (rounding error).
In addition, we observed that the processing speed of the cCNN based algorithm is consid-
erably faster than that of human test subjects. Specifically, when the algorithm was tested on a
computer with an NVIDIA Quadro K620 GPU, the algorithm processed all 25 images up to
3.8 times faster than human test subjects (Supplement 12 in S1 File, Fig 6A inset). Consider-
ably higher processing speeds can be achieved with more computational power.
Conclusions and summary
Convolutional neural networks have played a significant role in automating many complex
image processing tasks in biomedical research such as histology, cellular localization, embry-
onic cell counting and many others [31–33]. In this paper we explored the prospect of develop-
ing a cCNN algorithm that can classify bacteria sub populations (in this case stage of biofilm
development) from fluorescent microscope images. This algorithm can then be used to readily
quantify the effects of synthetic genetic circuits installed in cell-based materials, therapies, and
sensors.
In this paper we tested the classification performance of cCNN trained on three different
types of annotated training data sets: null bumper (NB), blended bumper (BB), and advanced
rotation (AR). Of these we find that advanced rotation is best suited for improving accuracy
and confidence of the CNN for smaller clusters (debris artifacts, single, double, and triple
cells) with accuracies of 86%. The larger clusters (4 to 10 cell clusters) would require additional
training data to improve the accuracy (now at 50–66%).A significant finding was that although
the individual classification accuracies were lower than desired, when the total count of cells is
summed, these inconsistencies balance out and near parity is found using our algorithm com-
pared to manual counting when assessed over 24 images. Established algorithms, such as Ima-
geJ, undercounts especially with higher cell count images. We also benchmarked this AR-
trained cCNN algorithm against multiple manual counts and found 10.2x reduction in intero-
perator variability and 3.8X increase in processing speed.
It is important to note that these results were achieved with a modest training data set
(annotated segments from 81 images that had representative clusters from 1–10 cells per clus-
ter). Acquiring additional images would be the simplest method to improve the cCNN perfor-
mance (through additional training) but in this application, this does take significant time
with microscopy to search for the rarer, sub-populations (high cell count clusters). Data acqui-
sition of rare events has always been one of the central challenges of the AI industry, but some
level of autonomous data acquisition can be achieved with unsupervised learning. Then, this
PLOS ONE | https://doi.org/10.1371/journal.pone.0241200 October 26, 2020
9 / 13
PLOS ONE Automated classification of bacterial cell sub-populations with convolutional neural networks
tool could be used for synthetic biology applications, such as tracking the effect of genetic
changes on the timing and heterogeneity of sporulation development stages. We anticipate
there will be many other applications for automated classification in this field, and thus have
provided all our data and architecture code to serve as another learning example (freely avail-
able at http://www.reuelgroup.org/resources.html).
Methods
Cell culture prep for microscope imaging
Cells of B. subtilis 168 were grown overnight until saturation in LB media using standard
methods. The overnight was used as a seed culture to infect sterile LB in a microplate. The
growth in the microplate reader was monitored as a measure of absorbance at 600 nm. Cells
were grown at 37C, and with reasonable agitation to prevent excess clumping. Cells were har-
vested at various stages of growth–early growth, log phase, saturation. When necessary, the
culture was diluted with fresh LB prior to microscope imaging. More detailed instructions on
sample preparation are outlined in the Supplement 1 in S1 File.
Flow cytometry
Cells were diluted to the appropriate density using LB with 5% sodium azide solution. Then,
cell samples were used with the BacLight bacterial viability and Counting Kit (ThermoFisher).
All flow cytometry experiments were performed on the factory-direct unmodified BD FACS-
Canto flow cytometer (San Jose, CA). We used 488 nm laser for extinction and detected fluo-
rescence with a 525/550 nm as well as 610/620 nm bandpass filters as described previously [8].
Fluorescent microscopy and image preprocessing
Images of B. subtilis cell culture were acquired on the Nikon Eclipse E800 microscope
equipped with a FITC filter set. Cells were stained with Ethidium Bromide to promote fluores-
cence. Raw images (unaltered pixel intensity) were segmented with an adaptive binary thresh-
olding algorithm. Images of individual cells or cell clusters were cropped, saved, and annotated
using a custom Matlab GUI (Supplement 13 in S1 File). All cell clusters were examined, and
we determined that the image segments did not exceed 200 by 200 pixel resolution. As such,
we set the input layer of the neural network to accept images of 200 by 200 pixel resolution
(Supplement 3 in S1 File). In the null bumper approach, cropped images of cell clusters were
placed in the middle of the 200 by 200 matrix, and the pixels around the cropped image were
set to 0 intensity. In the blended bumper approach, the space around the cropped image was
filled with pixel intensities that matched those in the local background of the cropped image
(Supplement 2 in S1 File).
For this study, we acquired 1,000 fluorescent microscope images, and then manually anno-
tated these (20,855 segments in total) and 10% of the data was used for training the network
and the remainder was reserved for evaluation of performance (see Supplement 2 in S1 File for
population numbers of the segmented data).
Convolutional neural network and training
The structure of the convolutional neural network used in this study is described in the Sup-
plement section 3 [23]. Briefly, it is a feed-forward network with 4 convolutional layers. Image
data was preprocessed on local computers, while the training procedures were executed with
Amazon Web Service resources (Supplement 4 in S1 File).
PLOS ONE | https://doi.org/10.1371/journal.pone.0241200 October 26, 2020
10 / 13
PLOS ONE Automated classification of bacterial cell sub-populations with convolutional neural networks
Evaluating network
Trained networks were evaluated with reserved data (10% of the original data set). The output
of the network from these annotated, evaluation images was scored with a confusion matrix
(Supplement 6 in S1 File).
Counting cells
The networks were also used to perform cell counting. To achieve that, raw fluorescence
microscope images were preprocessed with an adaptive binary thresholding algorithm.
Cropped images of cells and cell clusters were normalized to the 200 by 200 pixel resolution,
and supplied to the classification neural network. The output of the network corresponded to
the number of cells in a given cell clusters. The individual outputs were then added to find the
total cell count.
For ImageJ adaptive binary thresholding is first used to find clusters and then these cell
clusters were further sub-segmented with a watershed algorithm. The image is treated as a
topological map, with highest pixel intensities representing the bottom of the “basin”. The
basins are then filled with pixels until pixels from neighboring basins come in contact. The
boundary is drawn in the place of the contact. The total number of sub-segmented cells was
then reported as a total cell count.
Manual cell counting
Fluorescent microscope images that were used to test the efficiency of software-based cell
counting were also counted manually. These human cell counting experiments were con-
ducted under the IRB’s oversight (IRB ID 19–566) and is described in more detail in the Sup-
plement sections 10 (for Fig 5), and 12 (for Fig 6).
Supporting information
S1 File.
(DOCX)
Acknowledgments
We would like to thank Shawn Rigby for assistance with acquiring, reviewing and discussing
the flow cytometry data.
Author Contributions
Conceptualization: Denis Tamiev, Paige E. Furman, Nigel F. Reuel.
Data curation: Denis Tamiev, Paige E. Furman, Nigel F. Reuel.
Formal analysis: Denis Tamiev, Paige E. Furman, Nigel F. Reuel.
Funding acquisition: Denis Tamiev, Paige E. Furman, Nigel F. Reuel.
Investigation: Denis Tamiev, Paige E. Furman, Nigel F. Reuel.
Methodology: Denis Tamiev, Paige E. Furman, Nigel F. Reuel.
Project administration: Denis Tamiev, Paige E. Furman, Nigel F. Reuel.
Resources: Denis Tamiev, Paige E. Furman, Nigel F. Reuel.
Software: Denis Tamiev, Paige E. Furman, Nigel F. Reuel.
PLOS ONE | https://doi.org/10.1371/journal.pone.0241200 October 26, 2020
11 / 13
PLOS ONE Automated classification of bacterial cell sub-populations with convolutional neural networks
Supervision: Denis Tamiev, Paige E. Furman, Nigel F. Reuel.
Validation: Denis Tamiev, Paige E. Furman, Nigel F. Reuel.
Visualization: Denis Tamiev, Paige E. Furman, Nigel F. Reuel.
Writing – original draft: Denis Tamiev, Paige E. Furman, Nigel F. Reuel.
Writing – review & editing: Denis Tamiev, Paige E. Furman, Nigel F. Reuel.
References
1. Wei S.-P. et al., “Formation and functionalization of membraneless compartments in Escherichia coli,”
Nat. Chem. Biol., pp. 1–6, Jun. 2020, https://doi.org/10.1038/s41589-020-0579-9 PMID: 32601486
2.
3.
Leventhal D. S. et al., “Immunotherapy with engineered bacteria by targeting the STING pathway for
anti-tumor immunity,” Nat. Commun., vol. 11, no. 1, Art. no. 1, Jun. 2020, https://doi.org/10.1038/
s41467-020-16602-0 PMID: 32483165
Jha R. K. and Strauss C. E. M., “Smart Microbial Cells Couple Catalysis and Sensing to Provide High-
Throughput Selection of an Organophosphate Hydrolase,” ACS Synth. Biol., vol. 9, no. 6, pp. 1234–
1239, Jun. 2020, https://doi.org/10.1021/acssynbio.0c00025 PMID: 32369698
4. Siebert D. et al., “Genetic Engineering of Oligotropha carboxidovorans Strain OM5—A Promising Can-
didate for the Aerobic Utilization of Synthesis Gas,” ACS Synth. Biol., vol. 9, no. 6, pp. 1426–1440, Jun.
2020, https://doi.org/10.1021/acssynbio.0c00098 PMID: 32379961
5. Gonza´ lez L. M., Mukhitov N., and Voigt C. A., “Resilient living materials built by printing bacterial
spores,” Nat. Chem. Biol., vol. 16, no. 2, Art. no. 2, Feb. 2020, https://doi.org/10.1038/s41589-019-
0412-5 PMID: 31792444
6.
Lin C., Jair Y.-C., Chou Y.-C., Chen P.-S., and Yeh Y.-C., “Transcription factor-based biosensor for
detection of phenylalanine and tyrosine in urine for diagnosis of phenylketonuria,” Anal. Chim. Acta, vol.
1041, pp. 108–113, Dec. 2018, https://doi.org/10.1016/j.aca.2018.08.053 PMID: 30340682
7. Wu M.-R., Jusiak B., and Lu T. K., “Engineering advanced cancer therapies with synthetic biology,” Nat.
Rev. Cancer, vol. 19, no. 4, Art. no. 4, Apr. 2019, https://doi.org/10.1038/s41568-019-0121-0 PMID:
30837696
8.
9.
Tamiev D., Lantz A., Vezeau G., Salis H., and Reuel N. F., “Controlling Heterogeneity and Increasing
Titer from Riboswitch-Regulated Bacillus subtilis Spores for Time-Delayed Protein Expression Applica-
tions,” ACS Synth. Biol., vol. 8, no. 10, pp. 2336–2346, Oct. 2019, https://doi.org/10.1021/acssynbio.
9b00163 PMID: 31490060
Flemming H.-C., Wingender J., Szewzyk U., Steinberg P., Rice S. A., and Kjelleberg S., “Biofilms: an
emergent form of bacterial life,” Nat. Rev. Microbiol., vol. 14, no. 9, Art. no. 9, Sep. 2016, https://doi.org/
10.1038/nrmicro.2016.94 PMID: 27510863
10.
“LIVE/DEAD BacLight Bacterial Viability Kit, for microscopy—Thermo Fisher Scientific.” https://www.
thermofisher.com/order/catalog/product/L7007?SID=srch-srp-L7007 (accessed Nov. 04, 2019).
11. Rosenberg M., Azevedo N. F., and Ivask A., “Propidium iodide staining underestimates viability of
adherent bacterial cells,” Sci. Rep., vol. 9, no. 1, Art. no. 1, Apr. 2019, https://doi.org/10.1038/s41598-
019-42906-3 PMID: 31019274
12.
Li C. et al., “A Brief Review for Content-Based Microorganism Image Analysis Using Classical and
Deep Neural Networks,” in Information Technology in Biomedicine, Cham, 2019, pp. 3–14, https://doi.
org/10.1007/978-3-319-91211-0_1
13. Beaufort L. and Dollfus D., “Automatic recognition of coccoliths by dynamical neural networks,” Mar.
Micropaleontol., vol. 51, no. 1, pp. 57–73, Apr. 2004, https://doi.org/10.1016/j.marmicro.2003.09.003
14. Amaral A. L., Ginoris Y. P., Nicolau A., Coelho M. A. Z., and Ferreira E. C., “Stalked protozoa identifica-
tion by image analysis and multivariable statistical techniques,” Anal. Bioanal. Chem., vol. 391, no.
4, pp. 1321–1325, Jun. 2008, https://doi.org/10.1007/s00216-008-1845-y PMID: 18327573
15. Blackburn N., Hagstro¨ m Å., Wikner J., Cuadros-Hansson R., and Bjørnsen P. K., “Rapid Determination
of Bacterial Abundance, Biovolume, Morphology, and Growth by Neural Network-Based Image Analy-
sis,” Appl. Environ. Microbiol., vol. 64, no. 9, pp. 3246–3255, Sep. 1998, https://doi.org/10.1128/AEM.
64.9.3246-3255.1998 PMID: 9726867
16. A. Geron Hands-On Machine Learning with Scikit-Learn, Keras, and TensorFlow: Concepts, Tools, and
Techniques to Build Intelligent Systems.
PLOS ONE | https://doi.org/10.1371/journal.pone.0241200 October 26, 2020
12 / 13
PLOS ONE Automated classification of bacterial cell sub-populations with convolutional neural networks
17. Rawat W. and Wang Z., “Deep Convolutional Neural Networks for Image Classification: A Comprehen-
sive Review,” Neural Comput., vol. 29, no. 9, pp. 2352–2449, Jun. 2017, https://doi.org/10.1162/
NECO_a_00990 PMID: 28599112
18. Russakovsky O. et al., “ImageNet Large Scale Visual Recognition Challenge,” Int. J. Comput. Vis., vol.
115, no. 3, pp. 211–252, Dec. 2015, https://doi.org/10.1007/s11263-015-0816-y
19. Sultana F., Sufian A., and Dutta P., “A Review of Object Detection Models Based on Convolutional Neu-
ral Network,” in Intelligent Computing: Image Processing Based Applications, Mandal J. K. and Bane-
rjee S., Eds. Singapore: Springer, 2020, pp. 1–16.
20.
21.
LeCun Y. et al., “Backpropagation Applied to Handwritten Zip Code Recognition,” Neural Comput., vol.
1, no. 4, pp. 541–551, Dec. 1989, https://doi.org/10.1162/neco.1989.1.4.541.
Lecun Y., Bottou L., Bengio Y., and Haffner P., “Gradient-based learning applied to document recogni-
tion,” Proc. IEEE, vol. 86, no. 11, pp. 2278–2324, Nov. 1998, https://doi.org/10.1109/5.726791
22. Shorten C. and Khoshgoftaar T. M., “A survey on Image Data Augmentation for Deep Learning,” J. Big
Data, vol. 6, no. 1, p. 60, Jul. 2019, https://doi.org/10.1186/s40537-019-0197-0
23.
Ferrari A., Lombardi S., and Signoroni A., “Bacterial colony counting with Convolutional Neural Net-
works in Digital Microbiology Imaging,” Pattern Recognit., vol. 61, pp. 629–640, Jan. 2017, https://doi.
org/10.1016/j.patcog.2016.07.016 PMID: 31243076
24. M. Ranzato, F. J. Huang, Y.-L. Boureau, and Y. LeCun, “Unsupervised Learning of Invariant Feature
Hierarchies with Applications to Object Recognition,” in 2007 IEEE Conference on Computer Vision and
Pattern Recognition, Jun. 2007, pp. 1–8, https://doi.org/10.1109/CVPR.2007.383157
25. V. Nair and G. E. Hinton, “Rectified Linear Units Improve Restricted Boltzmann Machines,” presented at
the ICML, Jan. 2010, Accessed: Sep. 08, 2020. [Online]. Available: https://openreview.net/forum?id=
rkb15iZdZB.
26. Krizhevsky A., Sutskever I., and Hinton G. E., “ImageNet classification with deep convolutional neural
networks,” 2012, vol. 2, pp. 1097–1105.
27. G. E. Hinton, N. Srivastava, A. Krizhevsky, I. Sutskever, and R. R. Salakhutdinov, “Improving neural
networks by preventing co-adaptation of feature detectors,” ArXiv12070580 Cs, Jul. 2012, Accessed:
Sep. 08, 2020. [Online]. Available: http://arxiv.org/abs/1207.0580.
28.
Lin T.-Y. et al., “Microsoft COCO: Common Objects in Context,” in Computer Vision–ECCV 2014,
Cham, 2014, pp. 740–755, https://doi.org/10.1007/978-3-319-10602-1_48
29. Wiesmann V., Bergler M., Palmisano R., Prinzen M., Franz D., and Wittenberg T., “Using simulated
fluorescence cell micrographs for the evaluation of cell image segmentation algorithms,” BMC Bioinfor-
matics, vol. 18, no. 1, p. 176, Mar. 2017, https://doi.org/10.1186/s12859-017-1591-2 PMID: 28315633
30. Kornilov A. S. and Safonov I. V., “An Overview of Watershed Algorithm Implementations in Open
Source Libraries,” J. Imaging, vol. 4, no. 10, p. 123, Oct. 2018, https://doi.org/10.3390/
jimaging4100123
31. Gertych A. et al., “Convolutional neural networks can accurately distinguish four histologic growth pat-
terns of lung adenocarcinoma in digital slides,” Sci. Rep., vol. 9, no. 1, pp. 1–12, Feb. 2019, https://doi.
org/10.1038/s41598-018-37186-2 PMID: 30626917
32. A. Khan, S. Gould, and M. Salzmann, “Deep Convolutional Neural Networks for Human Embryonic Cell
Counting,” in Computer Vision–ECCV 2016 Workshops, Oct. 2016, pp. 339–348, https://doi.org/10.
1007/978-3-319-46604-0_25
33.
“Automated Identification and Localization of Hematopoietic Stem Cells in 3D Intravital Microscopy
Data,” Stem Cell Rep., vol. 5, no. 1, pp. 139–153, Jul. 2015, https://doi.org/10.1016/j.stemcr.2015.05.
017 PMID: 26120058
PLOS ONE | https://doi.org/10.1371/journal.pone.0241200 October 26, 2020
13 / 13
PLOS ONE |
10.1371_journal.pone.0226037 | RESEARCH ARTICLE
The relation between local and distal muscle
fat infiltration in chronic whiplash using
magnetic resonance imaging
Anette KarlssonID
Helena Ljunggren3, Magnus Borga1,2, Olof Dahlqvist Leinhard2,6
1,2*, Anneli Peolsson2,3, James Elliott4,5, Thobias Romu1,2,
1 Department of Biomedical Engineering, Linko¨ping University, Linko¨ ping, Sweden, 2 Center for Medical
Image Science and Visualization, Linko¨ ping University, Linko¨ ping, Sweden, 3 Department of Medical and
Health Sciences, Physiotherapy, Linko¨ ping University, Linko¨ ping, Sweden, 4 Faculty of Health Sciences, The
University of Sydney, Northern Sydney Local Health District, The Kolling Institute, St Leonards, NSW,
Australia, 5 Department of Physical Therapy and Human Movement Sciences, Feinberg School of Medicine,
Northwestern University, Chicago, IL, United States of America, 6 Department of Medical and Health
Sciences, Linko¨ping University, Linko¨ ping, Sweden
* anette.k.karlsson@liu.se
Abstract
The objective of this study was to investigate the relationship between fat infiltration in the
cervical multifidi and fat infiltration measured in the lower extremities to move further into
understanding the complex signs and symptoms arising from a whiplash trauma. Thirty-one
individuals with chronic whiplash associated disorders, stratified into a mild/moderate group
and a severe group, together with 31 age- and gender matched controls were enrolled in
this study. Magnetic resonance imaging was used to acquire a 3D volume of the neck and of
the whole-body. Cervical multifidi was used to represent muscles local to the whiplash
trauma and all muscles below the hip joint, the lower extremities, were representing wide-
spread muscles distal to the site of the trauma. The fat infiltration was determined by fat frac-
tion in the segmented images. There was a linear correlation between local and distal
muscle fat infiltration (p<0.001, r2 = 0.28). The correlation remained significant when adjust-
ing for age and WAD group (p = 0.009) as well as when correcting for age, WAD group and
BMI (p = 0.002). There was a correlation between local and distal muscle fat infiltration
within the severe WAD group (p = 0.0016, r2 = 0.69) and in the healthy group (p = 0.022, r2 =
0.17) but not in the mild/moderate group (p = 0.29, r2 = 0.06). No significant differences (p =
0.11) in the lower extremities’ MFI between the different groups were found. The absence of
differences between the groups in terms of lower extremities’ muscle fat infiltration indicates
that, in this particular population, the whiplash trauma has a local effect on muscle fat infiltra-
tion rather than a generalized.
Introduction
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: Karlsson A, Peolsson A, Elliott J, Romu T,
Ljunggren H, Borga M, et al. (2019) The relation
between local and distal muscle fat infiltration in
chronic whiplash using magnetic resonance
imaging. PLoS ONE 14(12): e0226037. https://doi.
org/10.1371/journal.pone.0226037
Editor: Xi Chen, McLean Hospital, UNITED STATES
Received: July 12, 2019
Accepted: November 17, 2019
Published: December 5, 2019
Peer Review History: PLOS recognizes the
benefits of transparency in the peer review
process; therefore, we enable the publication of
all of the content of peer review and author
responses alongside final, published articles. The
editorial history of this article is available here:
https://doi.org/10.1371/journal.pone.0226037
Copyright: © 2019 Karlsson et al. This is an open
access article distributed under the terms of the
Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: All relevant data are
within the manuscript and its Supporting
Information files.
Funding: The study received funding from the
Swedish Research Council and the Medical
Approximately 50% of individuals involved in a motor vehicle collision should expect to dem-
onstrate signs of recovery within the first 6–12 weeks following the collision event [1]. Others
PLOS ONE | https://doi.org/10.1371/journal.pone.0226037 December 5, 2019
1 / 13
Research Council of South-East Sweden (FORSS).
The funders had no role in study design, data
collection and analysis, decision to publish, or
preparation of the manuscript. Furthermore, the
commercial company AMRA Medical AB provided
support in the form of salaries for authors TR, MB
and ODL, but did not have any additional role in the
study design, data collection and analysis, decision
to publish, or preparation of the manuscript. The
specific roles of these authors are articulated in the
‘author contributions’ section.
Competing interests: We have read the journal’s
policy and the authors of this manuscript have the
following competing interests: TR, MB, and ODL
receive salaries and are stockholders of AMRA
Medical AB. AK is a stockholder of AMRA Medical
AB. This does not alter our adherence to PLOS
ONE’s policies on sharing data and material.
Distribution of muscle fat infiltration in chronic whiplash
will transition from acute to chronic trauma-related disability, presenting with physical and
psychological symptoms including neck pain, radiating arm pain, headache, anxiety and
depression [1–3]. However, there exists no standard by which to objectively diagnose whiplash
associated disorders (WAD), rather the grading is reliant on physical examination findings [4]
and a history of a whiplash from a motor vehicle collision.
Recently, cross-sectional and longitudinal studies across three different countries (and
insurance schemas) have qualified and quantified larger magnitudes of muscle fat infiltration
(MFI) in the cervical multifidi of participants with persistent WAD compared to those report-
ing lower levels of pain-related disability, those nominating full recovery, idiopathic neck pain,
and healthy controls [5–7]. The cervical multifidi may be directly involved during the whiplash
trauma as the muscles are inserted directly to the facet capsules of the cervical vertebrae [8]. A
neck injury due to the whiplash trauma may explain the higher amount of MFI in multifidi of
patients with chronic whiplash [9], although, the mechanism behind MFI reported in cervical
multifidi [5–7] is unclear. A number of hypotheses around injury severity, pain intensity and/
or related disability [10, 11], heightened stress-responses [12–14] central/peripheral neuronal
interference [15–17], spinal cord injury [18] and/or physical inactivity [19, 20] may be offered
to explain the rapid expression and larger magnitude of MFI in patients with more severe self-
reported symptoms. Factors such as heightened stress response and physical inactivity are sug-
gested to have general impact on muscle fat infiltration [21], but local injuries would only have
an impact on the neck.
Furthermore, a previous review of WAD [21] highlighted the potential influence of and
main actions exerted by the sympathetic nervous system on widespread motor function with a
focus on the underlying mechanisms for the onset and maintenance of chronic pain. While
much of the subsequent research has provided breakthrough knowledge about WAD related
pain-processing deficits and psychological distress, the landscape of available literature focus-
ing on the muscle system and motor output following whiplash continues to expand, but not
yet in parallel. Albeit preliminary, a case-study reported MFI in both the neck region and the
lower extremities [15] that could point to a mechanical injury involving descending white mat-
ter pathways of the cervical spinal cord. A higher value of MFI detected in both the lower
extremity and neck musculature corresponded to altered spinal cord anatomy and reductions
in the ability to maximally activate plantar flexor torques [15]. The preliminary nature of the
case-based study [15] supports the need for a larger cohort to further investigate the potential
link between MFI in muscles directly associated with the trauma likely affecting the head and
neck and in muscles distal to the site of potential injury.
Contrary to the findings by the case study above, a study by Pedler et al, on a larger cohort,
found no differences in fat infiltration in the right soleus muscle on group level where the groups
were divided into moderate/severe WAD, mild WAD and healthy controls [9]. These contrasting
findings calls for further investigations of local and distal MFI in WAD patients. The study by
Pedler et al. only investigated one muscle in the lower extremities. In addition, no regression mod-
els were used to compare the associations between the local and distal fat compartments.
The aim of this study was to investigate the potential pathophysiological link between local
MFI (that may have been injured in the trauma) and generalized MFI distal to the whiplash
trauma in patients with severe WAD, in those with mild/moderate WAD, and in healthy controls.
Method
Participants
Thirty-one individuals with mild/moderate (N = 20) or severe (N = 11) chronic WAD (at least
6 months duration) and 31 healthy matched controls were included. The Neck Disability
PLOS ONE | https://doi.org/10.1371/journal.pone.0226037 December 5, 2019
2 / 13
Distribution of muscle fat infiltration in chronic whiplash
Table 1. Descriptive data in format: [mean] ± [standard deviation] ([Range]).
Participants, n
Age, y
Body mass index, kg/m2
NDI, %
Time since injury, mo
Healthy Controls
Total
31
41.5 ± 0.6 (22–61)
24.4 ± 3.2 (19.7–34.5)
N/A
N/A
Total
31
41.5 ± 0.9 (20–62)
25.6 ± 4.1 (19.1–33.8)
35.8 ± 14.1 (10–68)
18.1 ± 9.2 (6–36)
WAD
NDI<40%
20
39.2 ± 11.5 (20–62)
25.5 ± 4.1 (19.1–33.8)
27.3 ± 6.8 (10–38)
20.1 ± 9.8 (7–36)
NDI�40%
11
45.7 ± 8.5 (34–58)
25.8 ± 3.4 (20.3–32.3)
51.3 ± 10.2 (40–68)
14.5 ± 7.2 (6–32)
WAD: Whiplash Associated Disorders, NDI: Neck Disability IndexBMI: Body Mass Index kg/m2
https://doi.org/10.1371/journal.pone.0226037.t001
Index (NDI) was used for stratification (mild/moderate: 20% < NDI < 40% and severe:
NDI � 40%)[22]. See Table 1 for descriptive statistics.
All participants provided written informed consent prior participation in the study and the
study was approved by the Regional Ethical Review Board in Linko¨ping (DNR 2011/262-32).
All experiments were performed with the relevant guidelines and regulations.
Inclusion and exclusion criteria
The WAD cohort was included from an ongoing randomized controlled trial (RCT) compar-
ing three different exercise strategies; neck specific exercises, neck specific exercises in combi-
nation with a behavioral approach, or prescribed general physical activity [23, 24]. All
participants in this study underwent the MR-scan before entering the RCT. The inclusion cri-
teria for entering the RCT were: Age 18–62 years old, right handed with dominant right-sided
pain, WAD > 6 months and < 3 years, Quebec Task Force WAD of grade II or III (grade II
means neck complaints and musculoskeletal signs; grade III means grade II plus neurological
signs [4]) and pain intensity of greater than 20 mm on a 100-mm visual analog scale [25] and/
or a score greater than 20% on the NDI scale).
Exclusion criteria from the WAD cohort were known or suspected serious physical pathol-
ogy, earlier fracture or luxation of the cervical column, neck trauma with persistent symptoms
from previous injury, surgery on the cervical column; neck pain that caused absence from
work >1 month in the year prior to the WAD trauma, signs of traumatic brain injury from or
before the whiplash injury, generalized or more dominant pain elsewhere in the body, diseases
or other injuries that might prevent full participation in the study, diagnosis of a severe psychi-
atric disorder, known drug abuse, contradiction for MRI, or insufficient knowledge of the
Swedish language to answer the questionnaires.
The age- and sex-matched healthy cohort was included for comparative purposes to the
WAD cohort. The exclusion criteria were present or past neck pain, dysfunction, or related
disability, history of neck trauma, or lower back pain, rheumatologic or neurological disease,
generalized myalgia and institutional contraindications for undergoing an MRI exam.
Magnetic resonance imaging
All images were acquired with a Philips Ingenia 3.0T scanner (Royal Philips, Amsterdam, the
Netherlands). The coils used for imaging was the built-in phased array posterior coil, a
32-channel head coil and two anterior flexible coils.
The neck images were acquired using a 2-point Dixon 3D gradient-echo sequence with out-
of-phase echo time of 3.66 milliseconds, and in-phase echo time of 7.24 milliseconds. TR was
10 milliseconds; the flip angle was 10˚ and the acquisition time was 9 minutes. Later echoes
PLOS ONE | https://doi.org/10.1371/journal.pone.0226037 December 5, 2019
3 / 13
Distribution of muscle fat infiltration in chronic whiplash
enabled the acquisition of a high resolution of 0.75�0.75�0.75 mm3. A high resolution is needed
to distinguish the small muscles in the deep neck musculature. Literature values of T2� relaxa-
tion (23.9 milliseconds for water and fat) was used for T2� correction (23,24). One participant
was excluded due to a fat-water swap artifact.
The lower extremities were imaged using a 2-point Dixon 3D gradient-echo sequence with
out of phase echo time of 1.15 milliseconds, and in-phase echo time of 2.3 milliseconds. TR
was 3.78 milliseconds and the flip angle was 10˚. The acquisition time was 5 minutes and the
voxel size was 1.75�1.75�1.75 mm3. Water- and fat-separated images were acquired using a
phase-based reconstruction method [26]. In the lower extremities, an intensity inhomogeneity
correction was performed using fat signal referencing [27]. The fat-referenced imaging tech-
nique has only been validated for short echo-times and was therefore not applied to the neck
images. As for the neck images, literature values for T2� correction were applied (23, 24).
Measurement of muscle fat infiltration
Left and right multifidi were segmented in the 3D volume from the level of cervical vertebra
3–7 using the semi-automatic image foresting transform technique [28]. Shortly described, the
technique uses an algorithm that calculates a segmentation based on a few manually defined
foreground seeds (pixels within the region of interest) and background seeds (outside the
region of interest). This process is iterated by adding foreground and background seeds to
help the algorithm until the operator is satisfied. The segmentation was done on the water
image, and large fatty streaks adjacent to the muscle were excluded. The segmentation was per-
formed by a musculoskeletal physiotherapist with more than 6 months experience doing this
particular analysis. Fourteen randomly selected participants were segmented twice for investi-
gating intra-rater reliability. The intra class correlation was 0.90 (CI 0.76–0.99) using two-way
random-effects, single-measure intra-class correlation coefficient with absolute agreement. A
physiotherapist researcher with more than 10 years of medical imaging experience also con-
firmed the segmentations. The MFI within the neck was acquired by calculating the fat signal
fraction: MFI %ð
Þ ¼ 100 �
.
fat
fatþwater
The lower extremities were automatically segmented using a multi-atlas segmentation tech-
nique presented by Karlsson et al. [29]. The method can handle different field strengths (1.5 T
and 3.0 T) and image resolutions [29, 30] and has also been validated using test-retest with
lean, overweight and obese participants [31]. All muscles below the hip joint were included.
Fifteen atlases with annotated muscles were non-rigidly registered onto the subject followed
by majority voting, i.e. more than half of the atlases needed to agree to classify the tissue as
muscle. MFI was defined as the average fat in muscle tissue where the fat concentration was
less than 50%. This way, contributions of fat signal from large fatty streaks and potential leak-
age in subcutaneous adipose tissue were avoided. The range of limits of agreement for the pre-
cision of this MFI measure is <±2% for leg muscles [30].
Statistics
The Shapiro-Wilk’s test together with visual inspection of histograms were used for investigat-
ing if MFI in the neck and lower extremities were approximately normally distributed.
Although the Schapiro-Wilk’s test could not exclude non-normal distribution, the visual
inspection of the histograms and box plots showed that MFI in multifidi and the lower extrem-
ities respectively still could be considered as normally distributed.
The analysis was performed using a mixed linear model with (A)—the association between
MFI in multifidi with MFI in the lower extremities, (B)–model A with WAD-group and age as
fixed factors, and (C)–model B with BMI as an additional factor. A linear regression between
PLOS ONE | https://doi.org/10.1371/journal.pone.0226037 December 5, 2019
4 / 13
Distribution of muscle fat infiltration in chronic whiplash
multifidi and duration since injury was also made on the two WAD groups. Analysis of vari-
ance (ANOVA) was used to investigate if there are any differences in MFI in the lower extrem-
ities between severe WAD, mild/moderate WAD and healthy controls. A Bonferroni
correction of multiple comparisons was chosen as post-hoc test to investigate potential
between-group differences. A p-value < 0.05 was considered significant.
Results
A typical result for the semi-automatic segmentation of the neck is showed in Fig 1 and Fig 2
shows a typical example of the fully automatic segmentation of the lower extremities.
There was a significant linear association (p<0.001) between MFI in multifidus and MFI in
the lower extremities (Model A) with r2 = 0.28. This relation remained significant when the
model was adjusted for age, and WAD-group (Model B, p = 0.009) as well as when adjusted
for age, WAD-group and BMI (Model C, p = 0.002). The severe WAD group had a significant
(p = 0.0016) and high correlation (r2 = 0.69) between MFI in multifidi and MFI in the lower
extremities. The healthy controls also showed a significant (p = 0.022) correlation (r2 = 0.17).
However, mild/moderate WAD showed no significant (p = 0.29) correlation between MFI in
multifidus and MFI in the lower extremities (r2 = 0.06). Fig 3 shows the MFI in multifidi plot-
ted against the MFI in the lower extremities color coded for the different groups. Observe that
caution should be made for comparing the r2 from the different subgroups due to different
within-group variances. WAD group was a significant factor both in model B (p = 0.017) and
in model C (p = 0.024) when comparing multifidus MFI to MFI in the lower extremities.
Severe WAD had significantly higher slope than both mild/moderate WAD(model B, p =
0.005, model C, p = 0.01) and healthy controls (model B, p = 0.016, and model C, p = 0.015).
Also, age was a significant factor both in model B (p = 0.009) and in model C (p = 0.008). BMI
did not show as a significant factor in model C (p = 0.065). There was no significant (p = 0.9)
association between MFI in multifidi and duration since injury.
The mean MFI in the lower extremities was 5.04% for the healthy controls, 5.06% for the
mild/moderate WAD and 6.25% for the severe WAD. The ANOVA showed no statistically sig-
nificant differences (p = 0.11) in the lower extremities’ MFI between the different groups. Fig 4
shows the mean MFI in the lower extremities for the three different groups.
Fig 1. Segmentation of m. multifidi. An axial (left), a coronal (middle), and a sagittal (right) cut of a high-resolution (0.75�0.75�0.75 mm3) 3D
image volume acquired using 2-point Dixon imaging followed by water-fat separation highlighting the segmentation of the left cervical
multifidii in one of the healthy controls. The segmented mask is overlaid on the fat + water image volume.
https://doi.org/10.1371/journal.pone.0226037.g001
PLOS ONE | https://doi.org/10.1371/journal.pone.0226037 December 5, 2019
5 / 13
Distribution of muscle fat infiltration in chronic whiplash
PLOS ONE | https://doi.org/10.1371/journal.pone.0226037 December 5, 2019
6 / 13
Distribution of muscle fat infiltration in chronic whiplash
Fig 2. Segmentation of lower extremities. A coronal slice of the selected distal muscles (lower extremities) that was
included in this study. The distal muscles were automatically defined using multi-atlas segmentation and is overlaid
the water-fat images in a different color.
https://doi.org/10.1371/journal.pone.0226037.g002
Fig 3. Association between MFI in multifidi and lower extremities. The MFI of the m. multifidi plotted against MFI in the lower extremities using
different colors and shapes for the different WAD groups. The MFI is reported as the fraction of fat in the muscle (%). Regression lines are fitted to the
data: Total (Black, dot dashed), y = -2.6 + 1.8x, r2 = 0.28, p < 0.001; Healthy Controls (Red, solid), y = -0.35 + 1.3x, r2 = 0.17, p = 0.022; Mild/Moderate
WAD (green, short dashed), y = 1.0 + 0.77x, r2 = 0.29, p = 0.29; Severe WAD (blue, long dashed), y = -5.0+2.7x, r2 = 0.69, p = 0.0016.
https://doi.org/10.1371/journal.pone.0226037.g003
PLOS ONE | https://doi.org/10.1371/journal.pone.0226037 December 5, 2019
7 / 13
0102001864MFI Lower Extremities (%)MFI m. Multifidi (%)Healthy ControlsMild/Moderate WADSevere WADTotal Distribution of muscle fat infiltration in chronic whiplash
Fig 4. Mean MFI in the three different groups. The figure shows the MFI based on the volumetric analysis of the lower extremities. The top of the bars
represents the mean MFI value (in %) and the whiskers show the 95% confidence intervals.
https://doi.org/10.1371/journal.pone.0226037.g004
Discussion
There was a significant linear association between the MFI in the multifidi muscles and in the
lower extremities with r2 = 0.28 (model A). This association remained significant when adding
age and WAD group (model B) as well as BMI (model C) as potential confounding factors,
suggesting that self-reported disability after WAD injury and MFI distal to the whiplash
trauma might be linked. This potential link is further implied when looking at the correlations
of each group. Severe WAD shows a strong correlation between multifidi MFI and MFI in the
lower extremities, with a r2 = 0.69. Furthermore, mild/moderate WAD shows no significant
correlation between local and distal MFI.
The duration since injury is not correlated to the MFI in multifidi. Patients with severe
WAD had significantly higher MFI in multifidi both in model B and model C, implying that
self-reported disability and MFI are connected although the time since injury does not affect
the result. One hypothesis is that a higher fat infiltration (or degenerative pathologies) prior
the trauma might affect the recovery of the patient groups [32]. While it is largely unknown if
and how changes in MFI is associated with poor functional recovery from whiplash, the results
PLOS ONE | https://doi.org/10.1371/journal.pone.0226037 December 5, 2019
8 / 13
0246Heatlhy ControlsMild/Moderate WADSevere WADMean MFI Lower Extremities (%) Distribution of muscle fat infiltration in chronic whiplash
of this study support that higher amounts of MFI could contribute to predictive models
towards determining whiplash recovery or other common conditions. A previous study inves-
tigating fibromyalgia, found significantly higher MFI values in the quadriceps muscles com-
pared to healthy controls [33]. However, note that widespread pain was an exclusion criterion
in this study.
The difference in MFI in the lower extremities was not statistically significant between the
groups according to the ANOVA. This is similar to the findings by Pedler et al where there
were no significant differences between MFI in multifidi and the right soleus muscle [9]. These
findings indicate complex mechanistic pathways influence MFI development and that physical
inactivity may not play a significant role in the recovery after a whiplash trauma since general-
ized MFI was not significantly different between the groups. Nevertheless, the potential that
higher MFI prior the collision or larger expressions following the collision may influence a sys-
temic response needs to be studied in a larger inception cohort in longitudinal fashion. How-
ever, to draw such conclusions, prognostic studies are needed. Our findings do indicate that
the whiplash trauma might have a direct or indirect effect on the muscles close to the cervical
spine may be at play but not as featured to occur in distal muscles. Further mechanistic work
involving a larger sample of participants with varying levels of signs and symptoms and across
different muscles is required before definitive conclusions can be drawn.
A case-study also reported high MFI in both the neck region and the lower extremities
compared to one recovered participant [15], supporting the heterogeneity of the whiplash
injury and recovery thereof. There are some differences between the designs comparing that
study from this. The inclusion criteria for the studies are not in alignment. Also, the case study
[15] involved participants with a more complex chronic WAD and one recovered participant.
While the severity of injury is largely unknown, future quantitative work investigating spinal
cord pathways and heightened molecular–neuroimmune—responses across a number of
patient populations is warranted, available [34–38] and underway to fully understand the
mechanisms underlying the cause and progression of compositional whole-body muscle
changes in WAD, and other conditions. In the present study there were no clinical or radiolog-
ical signs of a known spinal cord injury (e.g. jumped facet joints), however minor insults
involving the spinal cord cannot be excluded.
While the possibilities to measure MFI in the small muscles in the neck with high-resolu-
tion MRI continues to evolve, the wider literature would benefit from a clear and broadly
accepted definition of MFI. Suggestions on those definitions are, for example, proposed in the
study by Crawford et al. [39] and the study by Elliott et al. [40]. The measures should be repro-
ducible and insensitive to different scanners, different laboratories, and performed using con-
sensus driven methodologies. In addition, future research should investigate MFI across a
number of common conditions to compare and contrast diagnosis-dependent changes in local
and whole-body MFI. The prospects of training and using deep learning neural networks also
increases the feasibility of translating such measures to clinical practice [41].
The Shapiro-Wilks test indicated that the distribution was slightly non-normal. A non-
parametric analysis was therefore performed and confirmed the results from the parametric
test. For future studies, larger sample sizes could permit the development of larger, more com-
plex, models without losing statistical power. However, this study establishes a potentially new
link with MRI whole-body acquisition, which is not a normal clinical routine for patients with
suspected spine trauma following motor vehicle collision. These links between local and distal
MFI are intriguing and are opening up for further studies of the distal MFI’s influence to high
self-reported disability in chronic WAD. With the evolved MRI technique, a head-to-knee
protocol with enough resolution to automatically analyze the composition of the thigh muscles
is possible using a six minute scan [42].
PLOS ONE | https://doi.org/10.1371/journal.pone.0226037 December 5, 2019
9 / 13
Distribution of muscle fat infiltration in chronic whiplash
One limitation with this study was that no functional analysis was made on the lower
extremities. Furthermore, no exclusion was made for previous trauma or injury in the lower
extremities. However, an exclusion criterion for this study was dominating or generalized pain
(other than for the neck region for WAD participants) minimizing the potential influences of
previous trauma/injury.
Another limitation with this study was that the methods for measuring MFI in multifidi
and lower extremities differs slightly. In the lower extremities, lower resolution is sufficient,
which enables a shorter MR scan. Furthermore, the muscles can be analyzed with automatic
methods with high reported precision [29, 30]. With evolving MRI technology better possibili-
ties for scanning the smallest muscles (e.g. cervical multifidi) with enough resolution for seg-
mentation and analysis have been possible [41]. In this manuscript the whole multifidi muscle
volume has been analyzed instead of a few cross-sectional slices, which has been considered
the previous gold standard. When the entire muscle is analyzed the result is not sensitive to the
placement of single image slices. This increased precision makes it easier to perform longitudi-
nal studies investigating e.g. different rehabilitation programs or to follow the progression of
the WAD condition. However, one limitation with our approach is that a single peak lipid
model with a theoretical value for T2� was used. This could induce a T2� bias in the investiga-
tion. Another limitation is that no calibration of the fat signal was applied to the high-resolu-
tion neck images. This may have introduced T1-bias in calculating the fat content of the m.
multifidi. Future works including technical development of quantitative MFI measurement in
the small deep neck muscles are of high interest to further investigate the local changes in MFI
after, not only a whiplash injury, but other common, yet equally enigmatic, degenerative and
pathological conditions of the spine (e.g. myelopathy and radiculopathy).
To conclude, since there was no significant difference in MFI distal to the trauma in the
severe WAD group compared to either the mild/moderate WAD group, or to the healthy
group, the results indicate the response to the whiplash trauma reflects a local physiological
process in patients with severe chronic WAD. Furthermore, time since injury did not have a
significant correlation with MFI in the multifidi, which indicate that slow development of gen-
eralized MFI due to e.g. physical inactivity after the trauma appears to have no major effect.
Finally, the strong association between multifidi MFI and MFI in the lower extremities found
to be unique to the group with severe WAD (r2 = 0.69) but not in the mild/minor WAD group
(r2 = 0.06) does not exclude the possibility that high generalized MFI prior the accident could
play a role in the development of severe chronic WAD. The associations between MFI and
WAD are intriguing and may in the future contribute to better understanding regarding onset
and progress of WAD after a whiplash trauma.
Supporting information
S1 Table. Complete measurement results for all participants included in this study.
(XLSX)
Author Contributions
Conceptualization: Anette Karlsson, Anneli Peolsson, Olof Dahlqvist Leinhard.
Data curation: Anette Karlsson, Anneli Peolsson, Thobias Romu, Helena Ljunggren, Olof
Dahlqvist Leinhard.
Formal analysis: Anette Karlsson, James Elliott, Magnus Borga, Olof Dahlqvist Leinhard.
Funding acquisition: Anneli Peolsson, Olof Dahlqvist Leinhard.
PLOS ONE | https://doi.org/10.1371/journal.pone.0226037 December 5, 2019
10 / 13
Distribution of muscle fat infiltration in chronic whiplash
Investigation: Anette Karlsson.
Methodology: Anette Karlsson, Thobias Romu, Olof Dahlqvist Leinhard.
Project administration: Anneli Peolsson.
Software: Anette Karlsson.
Supervision: Anneli Peolsson, Magnus Borga.
Validation: James Elliott.
Visualization: Anette Karlsson.
Writing – original draft: Anette Karlsson.
Writing – review & editing: Anette Karlsson, Anneli Peolsson, James Elliott, Thobias Romu,
Helena Ljunggren, Magnus Borga, Olof Dahlqvist Leinhard.
References
1. Carroll LJ, Holm LW, Hogg-Johnson S, Cote P, Cassidy JD, Haldeman S, et al. Course and prognostic
factors for neck pain in whiplash-associated disorders (WAD): results of the Bone and Joint Decade
2000–2010 Task Force on Neck Pain and Its Associated Disorders. Spine. 2008; 33(4 Suppl):S83–92.
https://doi.org/10.1097/BRS.0b013e3181643eb8 PMID: 18204405
2.
Leth-Petersen S, Rotger GP. Long-term labour-market performance of whiplash claimants. Journal of
health economics. 2009; 28(5):996–1011. https://doi.org/10.1016/j.jhealeco.2009.06.013 PMID:
19683817
3. Rebbeck T, Sindhusake D, Cameron ID, Rubin G, Feyer AM, Walsh J, et al. A prospective cohort study
of health outcomes following whiplash associated disorders in an Australian population. Injury preven-
tion: journal of the International Society for Child and Adolescent Injury Prevention. 2006; 12(2):93–8.
4. Spitzer WO, Skovron ML, Salmi LR, Cassidy JD, Duranceau J, Suissa S, et al. Scientific monograph of
the Quebec Task Force on Whiplash-Associated Disorders: redefining "whiplash" and its management.
Spine. 1995; 20(8 Suppl):1S–73S. PMID: 7604354
5. Elliott J, Jull G, Noteboom JT, Darnell R, Galloway G, Gibbon WW. Fatty infiltration in the cervical exten-
sor muscles in persistent whiplash-associated disorders: a magnetic resonance imaging analysis.
Spine. 2006; 31(22):E847–55. https://doi.org/10.1097/01.brs.0000240841.07050.34 PMID: 17047533
6. Elliott J, Sterling M, Noteboom JT, Darnell R, Galloway G, Jull G. Fatty infiltrate in the cervical extensor
muscles is not a feature of chronic, insidious-onset neck pain. Clinical radiology. 2008; 63(6):681–7.
https://doi.org/10.1016/j.crad.2007.11.011 PMID: 18455560
7. Karlsson A, Leinhard OD, Aslund U, West J, Romu T, Smedby O, et al. An Investigation of Fat Infiltra-
tion of the Multifidus Muscle in Patients With Severe Neck Symptoms Associated With Chronic Whip-
lash-Associated Disorder. The Journal of orthopaedic and sports physical therapy. 2016; 46(10):886–
93. https://doi.org/10.2519/jospt.2016.6553 PMID: 27590177
8. Anderson JS, Hsu AW, Vasavada AN. Morphology, architecture, and biomechanics of human cervical
multifidus. Spine (Phila Pa 1976). 2005; 30(4):E86–91.
9. Pedler A, McMahon K, Galloway G, Durbridge G, Sterling M. Intramuscular fat is present in cervical mul-
tifidus but not soleus in patients with chronic whiplash associated disorders. PloS one. 2018; 13(5):
e0197438. https://doi.org/10.1371/journal.pone.0197438 PMID: 29795590
10. Elliott J, Pedler A, Kenardy J, Galloway G, Jull G, Sterling M. The temporal development of fatty infil-
trates in the neck muscles following whiplash injury: an association with pain and posttraumatic stress.
PloS one. 2011; 6(6):e21194. https://doi.org/10.1371/journal.pone.0021194 PMID: 21698170
11. Elliott JM, Courtney DM, Rademaker A, Pinto D, Sterling MM, Parrish TB. The Rapid and Progressive
Degeneration of the Cervical Multifidus in Whiplash: An MRI Study of Fatty Infiltration. Spine. 2015; 40
(12):E694–700. https://doi.org/10.1097/BRS.0000000000000891 PMID: 25785961
12.
Linnstaedt SD, Riker KD, Walker MG, Nyland JE, Zimny E, Lewandowski C, et al. MicroRNA 320a Pre-
dicts Chronic Axial and Widespread Pain Development Following Motor Vehicle Collision in a Stress-
Dependent Manner. The Journal of orthopaedic and sports physical therapy. 2016; 46(10):911–9.
https://doi.org/10.2519/jospt.2016.6944 PMID: 27690835
13.
Linnstaedt SD, Walker MG, Parker JS, Yeh E, Sons RL, Zimny E, et al. MicroRNA circulating in the
early aftermath of motor vehicle collision predict persistent pain development and suggest a role for
PLOS ONE | https://doi.org/10.1371/journal.pone.0226037 December 5, 2019
11 / 13
Distribution of muscle fat infiltration in chronic whiplash
microRNA in sex-specific pain differences. Mol Pain. 2015; 11:66. https://doi.org/10.1186/s12990-015-
0069-3 PMID: 26498117
14. McLean SA. The potential contribution of stress systems to the transition to chronic whiplash-associ-
ated disorders. Spine. 2011; 36(25 Suppl):S226–32. https://doi.org/10.1097/BRS.0b013e3182387fb4
PMID: 22020617
15. Elliott JM, Dewald JP, Hornby TG, Walton DM, Parrish TB. Mechanisms underlying chronic whiplash:
contributions from an incomplete spinal cord injury? Pain Med. 2014; 15(11):1938–44. https://doi.org/
10.1111/pme.12518 PMID: 25139822
16. Guez M, Hildingsson C, Rosengren L, Karlsson K, Toolanen G. Nervous tissue damage markers in
cerebrospinal fluid after cervical spine injuries and whiplash trauma. J Neurotrauma. 2003; 20(9):853–
8. https://doi.org/10.1089/089771503322385782 PMID: 14577863
17. Smith AC, Parrish TB, Hoggarth MA, McPherson JG, Tysseling VM, Wasielewski M, et al. Potential
associations between chronic whiplash and incomplete spinal cord injury. Spinal Cord Ser Cases.
2015; 1.
18. Elliott JM, Pedler AR, Cowin G, Sterling M, McMahon K. Spinal cord metabolism and muscle water diffu-
sion in whiplash. Spinal Cord. 2012; 50(6):474–6. https://doi.org/10.1038/sc.2011.17 PMID: 21383759
19. Clark BC. In vivo alterations in skeletal muscle form and function after disuse atrophy. Med Sci Sports
Exerc. 2009; 41(10):1869–75. https://doi.org/10.1249/MSS.0b013e3181a645a6 PMID: 19727027
20. Manini TM, Clark BC, Nalls MA, Goodpaster BH, Ploutz-Snyder LL, Harris TB. Reduced physical activ-
ity increases intermuscular adipose tissue in healthy young adults. Am J Clin Nutr. 2007; 85(2):377–84.
https://doi.org/10.1093/ajcn/85.2.377 PMID: 17284732
21. Passatore M, Roatta S. Influence of sympathetic nervous system on sensorimotor function: whiplash
associated disorders (WAD) as a model. European journal of applied physiology. 2006; 98(5):423–49.
https://doi.org/10.1007/s00421-006-0312-8 PMID: 17036216
22. Vernon H, Mior S. The Neck Disability Index: a study of reliability and validity. Journal of manipulative
and physiological therapeutics. 1991; 14(7):409–15. PMID: 1834753
23.
Ludvigsson ML, Peterson G, O’Leary S, Dedering A, Peolsson A. The Effect of Neck-specific Exercise
With, or Without a Behavioral Approach, on Pain, Disability, and Self-Efficacy in Chronic Whiplash-
associated Disorders: A Randomized Clinical Trial. The Clinical journal of pain. 2015; 31(4):294–303.
https://doi.org/10.1097/AJP.0000000000000123 PMID: 24918474
24. Peolsson A, Landen Ludvigsson M, Overmeer T, Dedering A, Bernfort L, Johansson G, et al. Effects of
neck-specific exercise with or without a behavioural approach in addition to prescribed physical activity
for individuals with chronic whiplash-associated disorders: a prospective randomised study. BMC mus-
culoskeletal disorders. 2013; 14:311. https://doi.org/10.1186/1471-2474-14-311 PMID: 24171699
25. Harms-Ringdahl K, Carlsson AM, Ekholm J, Raustorp A, Svensson T, Toresson HG. Pain assessment
with different intensity scales in response to loading of joint structures. Pain. 1986; 27(3):401–11.
https://doi.org/10.1016/0304-3959(86)90163-6 PMID: 3808744
26. Romu T, Dahlstrom N, Leinhard OD, Borga M. Robust water fat separated dual-echo MRI by phase-
sensitive reconstruction. Magn Reson Med. 2017; 78(3):1208–16. https://doi.org/10.1002/mrm.26488
PMID: 27775180
27. Romu T, Borga M, Dahlqvist Leinhard O. MANA—Multi scale adaptive normalized averaging. Proceed-
ings—International Symposium on Biomedical Imaging (ISBI). 2011;art.no. 5872424:361–4.
28. Malmberg F, Lindblad J, Nystro¨ m I, editors. Sub-Pixel Segmentation with the Image Foresting Trans-
form. 13th International Workshop on Combinatorial Image Analysis (IWCIA’09); 2009; Playa del Car-
men, Mexico.
29. Karlsson A, Rosander J, Romu T, Tallberg J, Gronqvist A, Borga M, et al. Automatic and quantitative
assessment of regional muscle volume by multi-atlas segmentation using whole-body water-fat MRI. J
Magn Reson Imaging. 2015; 41(6):1558–69. https://doi.org/10.1002/jmri.24726 PMID: 25111561
30. West J, Romu T, Thorell S, Lindblom H, Berin E, Holm AS, et al. Precision of MRI-based body composi-
tion measurements of postmenopausal women. PloS one. 2018; 13(2):e0192495. https://doi.org/10.
1371/journal.pone.0192495 PMID: 29415060
31.
Thomas MS, Newman D, Leinhard OD, Kasmai B, Greenwood R, Malcolm PN, et al. Test-retest reliabil-
ity of automated whole body and compartmental muscle volume measurements on a wide bore 3T MR
system. Eur Radiol. 2014; 24(9):2279–91. https://doi.org/10.1007/s00330-014-3226-6 PMID:
24871333
32. Elliott JMP T.B.; Walton D.M.; Vassallo A.J.; Fundaun J.; Wasielewski M.; Courtney D.M. Does Overall
Cervical Spine Pathology REalte to the Clinical Heterogeneity of Chronic Whiplash?. American Journal
of Emergency Medicince 2019.
PLOS ONE | https://doi.org/10.1371/journal.pone.0226037 December 5, 2019
12 / 13
Distribution of muscle fat infiltration in chronic whiplash
33. Gerdle B, Forsgren MF, Bengtsson A, Leinhard OD, Soren B, Karlsson A, et al. Decreased muscle con-
centrations of ATP and PCR in the quadriceps muscle of fibromyalgia patients—a 31P-MRS study.
European journal of pain. 2013; 17(8):1205–15. https://doi.org/10.1002/j.1532-2149.2013.00284.x
PMID: 23364928
34. Cohen-Adad J, El Mendili MM, Lehericy S, Pradat PF, Blancho S, Rossignol S, et al. Demyelination and
degeneration in the injured human spinal cord detected with diffusion and magnetization transfer MRI.
Neuroimage. 2011; 55(3):1024–33. https://doi.org/10.1016/j.neuroimage.2010.11.089 PMID:
21232610
35. De Leener B, Levy S, Dupont SM, Fonov VS, Stikov N, Louis Collins D, et al. SCT: Spinal Cord Toolbox,
an open-source software for processing spinal cord MRI data. Neuroimage. 2017; 145(Pt A):24–43.
https://doi.org/10.1016/j.neuroimage.2016.10.009 PMID: 27720818
36. Martin AR, De Leener B, Cohen-Adad J, Aleksanderek I, Cadotte DW, Kalsi-Ryan S, et al. 163 Micro-
structural MRI Quantifies Tract-Specific Injury and Correlates With Global Disability and Focal Neuro-
logical Deficits in Degenerative Cervical Myelopathy. Neurosurgery. 2016; 63 Suppl 1:165.
37. Smith AC, Weber KA 2nd, O’Dell DR, Parrish TB, Wasielewski M, Elliott JM. Lateral corticospinal tract
damage correlates with motor output in incomplete spinal cord injury. Archives of physical medicine and
rehabilitation. 2017.
38. Smith AC, Weber KA, Parrish TB, Hornby TG, Tysseling VM, McPherson JG, et al. Ambulatory function
in motor incomplete spinal cord injury: a magnetic resonance imaging study of spinal cord edema and
lower extremity muscle morphometry. Spinal Cord. 2017; 55(7):672–8. https://doi.org/10.1038/sc.2017.
18 PMID: 28244504
39. Crawford RJ, Cornwall J, Abbott R, Elliott JM. Manually defining regions of interest when quantifying
paravertebral muscles fatty infiltration from axial magnetic resonance imaging: a proposed method for
the lumbar spine with anatomical cross-reference. BMC musculoskeletal disorders. 2017; 18(1):25.
https://doi.org/10.1186/s12891-016-1378-z PMID: 28103921
40. Elliott JM, Cornwall J, Kennedy E, Abbott R, Crawford RJ. Towards defining muscular regions of interest
from axial magnetic resonance imaging with anatomical cross-reference: part II—cervical spine muscu-
lature. BMC Musculoskelet Disord. 2018; 19(1):171. https://doi.org/10.1186/s12891-018-2074-y PMID:
29807530
41. Weber KA, Smith AC, Wasielewski M, Eghtesad K, Upadhyayula PA, Wintermark M, et al. Deep Learn-
ing Convolutional Neural Networks for the Automatic Quantification of Muscle Fat Infiltration Following
Whiplash Injury. Sci Rep. 2019; 9(1):7973. https://doi.org/10.1038/s41598-019-44416-8 PMID:
31138878
42. West J, Dahlqvist Leinhard O, Romu T, Collins R, Garratt S, Bell JD, et al. Feasibility of MR-Based
Body Composition Analysis in Large Scale Population Studies. PloS one. 2016; 11(9):e0163332.
https://doi.org/10.1371/journal.pone.0163332 PMID: 27662190
PLOS ONE | https://doi.org/10.1371/journal.pone.0226037 December 5, 2019
13 / 13
|
10.1371_journal.pone.0239702 | RESEARCH ARTICLE
SAXS studies of X-ray induced disulfide bond
damage: Engineering high-resolution insight
from a low-resolution technique
Timothy R. StachowskiID
1,2¤, Mary E. Snell1, Edward H. SnellID
1,3*
1 Hauptman-Woodward Medical Research Institute, Buffalo, New York, United States of America,
2 Department of Cell Stress Biology, Roswell Park Comprehensive Cancer Center, Buffalo, New York, United
States of America, 3 Department of Materials Design and Innovation, State University at New York at Buffalo,
Buffalo, New York, United States of America
¤ Current address: Department of Chemical Biology & Therapeutics, St. Jude Children’s Research Hospital,
Memphis, Tennessee, United States of America
* esnell@hwi.buffalo.edu
Abstract
A significant problem in biological X-ray crystallography is the radiation chemistry caused by
the incident X-ray beam. This produces both global and site-specific damage. Site specific
damage can misdirect the biological interpretation of the structural models produced. Cryo-
cooling crystals has been successful in mitigating damage but not eliminating it altogether;
however, cryo-cooling can be difficult in some cases and has also been shown to limit func-
tionally relevant protein conformations. The doses used for X-ray crystallography are typi-
cally in the kilo-gray to mega-gray range. While disulfide bonds are among the most
significantly affected species in proteins in the crystalline state at both cryogenic and higher
temperatures, there is limited information on their response to low X-ray doses in solution,
the details of which might inform biomedical applications of X-rays. In this work we engi-
neered a protein that dimerizes through a susceptible disulfide bond to relate the radiation
damage processes seen in cryo-cooled crystals to those closer to physiologic conditions.
This approach enables a low-resolution technique, small angle X-ray scattering (SAXS), to
detect and monitor a residue specific process. A dose dependent fragmentation of the engi-
neered protein was seen that can be explained by a dimer to monomer transition through
disulfide bond cleavage. This supports the crystallographically derived mechanism and
demonstrates that results obtained crystallographically can be usefully extrapolated to phys-
iologic conditions. Fragmentation was influenced by pH and the conformation of the dimer,
providing information on mechanism and pointing to future routes for investigation and
potential mitigation. The novel engineered protein approach to generate a large-scale
change through a site-specific interaction represents a promising tool for advancing radia-
tion damage studies under solution conditions.
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: Stachowski TR, Snell ME, Snell EH (2020)
SAXS studies of X-ray induced disulfide bond
damage: Engineering high-resolution insight from a
low-resolution technique. PLoS ONE 15(11):
e0239702. https://doi.org/10.1371/journal.
pone.0239702
Editor: Titus J. Boggon, Yale University School of
Medicine, UNITED STATES
Received: May 21, 2020
Accepted: September 12, 2020
Published: November 17, 2020
Copyright: © 2020 Stachowski et al. This is an
open access article distributed under the terms of
the Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: All SAXS data used in
the analysis are available as supplementary data.
The data averaged series with a 5.0 mg/ml protein
concentration at pH 7.5 have been deposited in
SASBDB (the Small Angle Scattering Biological
Data Bank) as SASDHG6. The crystal structure has
been deposited in the Protein Data Bank (PDB) as
6VE1.
Funding: EHS acknowledges support from the
Louis Sklarow Charitable Trust, and National
Science Foundation Grant no. 1231306. This work
PLOS ONE | https://doi.org/10.1371/journal.pone.0239702 November 17, 2020
1 / 23
PLOS ONE was based in part upon research conducted at the
Advanced Light Source (ALS), a national user
facility operated by Lawrence Berkeley National
Laboratory on behalf the Department of Energy,
Office of Basic Energy Sciences through the
Integrated Diffraction Analysis Technologies (IDAT)
program, supported by DOE Office of Biological
and Environmental Research. Additional support
comes from the National Institute of Health project
ALS-ENABLE (P30 GM124169) and a High-End
Instrumentation Grant S10OD018483. This
research also used resources at the Industrial
Macromolecular Crystallography Association
Collaborative Access Team (IMCA-CAT) beamline
17-ID, supported by the companies of the
Industrial Macromolecular Crystallography
Association through a contract with Hauptman–
Woodward Medical Research Institute. The funders
had no role in study design, data collection, and
analysis, decision to publish, or preparation of the
manuscript.
Competing interests: The authors have declared
that no competing interests exist.
Engineering a method for site-specific damage studies with SAXS
Introduction
Radiation chemistry in macromolecular X-ray crystallography is a significant issue. When X-
rays interact with water, the major component of a crystal or the cellular environment, free
radicals are produced. These products consist of solvated electrons (e-), hydroxyl radicals
(HO•), and hydronium ions (H3O+). The ions can further react with the solvated electrons to
produce hydrogen radicals (H•) and water [1]. All the radicals cause damage seen in global
indicators and as specific structural disruption of individual residues in the resultant model. In
the latter case, glutamates and aspartates are decarboxylated, and disulfide bonds and the ter-
minal methyl group on methionines are cleaved. Disulfide bonds and S-methyl thioether
groups are the most sensitive to X-ray damage due to the high photo-absorption cross section
and electron-affinity of sulfur [2]. This specific damage has garnered much attention due to its
potential to misdirect the biological interpretation of protein mechanisms.
Many structural studies have focused on understanding damage mechanisms [3] and miti-
gating damage through cryo-cooling [4] but also with the use of radical scavenging [5]. Cryo-
cooling crystals to ~100 K prevents the diffusion of most solvent generated radicals and only
the electrons remain mobile [6]. Damage processes are slowed because the larger radicals are
trapped, which can extend the effective crystal lifetime by one to two orders of magnitude [7].
However, cryo-cooling can also limit biologically meaningful conformations [8, 9] and pro-
duce structural artefacts [10, 11]. For these reasons there has been interest in a return to near
physiological temperature data collection and the corresponding development of serial crystal-
lography methods to limit radiation damage [12].
The X-ray doses used for structural studies are on the order of tens of kilo-grays up to
mega-grays and different damage processes have been observed in cryogenic and near physio-
logical temperature crystallographic experiments [7, 12–14]. Damage mechanisms at cryogenic
temperature have been well studied at large X-ray doses, but how relevant they are in physio-
logic conditions (i.e. lower doses, higher temperatures, and in solution) is less well understood.
SAXS is a low-resolution solution technique that can be used to validate the arrangement of
proteins observed in crystal structures and develop low resolution models of large protein
complexes. It can be used in a more physiologic setting to study large scale conformational
changes and dynamics [15, 16]. The radiation dose required to get meaningful data in SAXS
experiments is orders of magnitude lower than those used in crystallography. This allows
SAXS to be used to study physiologically relevant radiation-protein interactions that are chal-
lenging to understand with current biochemical approaches [17].
By engineering a protein that dimerized through a highly solvent exposed disulfide bond, a sys-
tem is created where disulfide bond breakage leading to monomerization is easily measured by
SAXS. In this way, SAXS was able to follow a residue specific rather than a global damage process
and link damage pathways seen crystallographically to those occurring in solution with much
lower X-ray doses. The X-ray radiation induced dimer to monomer transition was pH dependent,
suggesting mechanistic processes, routes for further interrogation, and mitigation strategies.
Materials and methods
Protein engineering, expression, and purification
Endoglycosidase-H (endoH) is a 27 kDa globular monomeric protein that is commonly used
to deglycosylate proteins to promote crystallization [18] and monitor protein trafficking [19].
It does not contain any native cysteines, is moderately sized, natively monomeric [20, 21] sta-
ble under a range of pH values and physiologic temperature [22], and crystallizable [20, 21]. It
provides a suitable system to introduce a disulfide to study radiation induced cleavage.
PLOS ONE | https://doi.org/10.1371/journal.pone.0239702 November 17, 2020
2 / 23
PLOS ONE Engineering a method for site-specific damage studies with SAXS
An expression construct was created by fusing endoH (P04067, aa 47–313) to the C-termi-
nus of the maltose-binding protein (MBP) in the pMAL-p5X vector (New England Biolabs;
Ipswich, MA) and inserting a tobacco etch virus protease (TEV) cleavage site between the two
domains. Previous reports suggested that introducing a disulfide bond is more successful
when placed in a flexible region [23] and creates a large loop [24]. Considering this and that
the bond should be highly solvent exposed to promote damage and not alter the native folding,
we introduced a 7 aa fragment with a single free cysteine (SLSTGCY, ’FRAGCYS’) attached to
the flexible N-terminus of endoH. Cloning was performed by GenScript (Piscataway, NJ). For
purification, a His6-tag was added to the C-terminus of endoH so that the final construct was
MBP-TEV-FRAGCYS-endoH-His6. For expression, BL21(DE3) cells were transformed with
the construct and grown in LB overnight at 30˚C. Cells were diluted to an OD600 of 0.05 and
once the OD600 reached 0.3–0.4 the temperature was reduced to 22˚C. At OD600 0.7–0.8
expression was induced with 0.05 mM IPTG and cells continued growing at 22˚C for 24 hours
shaking at 250 rpm. Harvested cells were pelleted and lysed in 200 mM NaCl, 20 mM Tris-
HCl, pH 7.5, 5 mM imidazole with a microfluidizer. Lysed cells were centrifuged at 60,000 x g
for 45 minutes and the supernatant was combined with Ni-NTA resin (Marvelgent; Canton,
MA) and incubated overnight at 4˚C. The resin was washed with 10 column volumes of 200
mM NaCl, 20 mM Tris-HCl, pH 7.5, 20 mM imidazole and the protein eluted in 8 column vol-
umes of 200 mM NaCl, 20 mM Tris-HCl, pH 7.5, 250 mM imidazole. The eluted protein was
incubated with TEV protease (10% w/w) for 60 h at 4˚C. The cleaved protein was separated
with a second Ni-affinity purification. Size-exclusion chromatography (SEC) was used as a
final purification step and to exchange the protein into 50 mM NaCl, 20 mM Tris-HCl, pH
7.5, 5 mM EDTA. The final dimeric protein is referred to as endoHCYS. Disulfide bond forma-
tion occurred spontaneously and was monitored by SDS-PAGE and size exclusion chromatog-
raphy (SEC). Briefly, SEC was performed by equilibrating a Superdex 200 10/300 SEC column
(GE Healthcare; Chicago, IL) in 20 mM Tris-HCl, pH 7.5, 50 mM NaCl, 5 mM EDTA. 100 μL
of protein at approximately 5 mg/ml was injected for data collection. Data were analyzed using
the software Unicorn (GE Healthcare) to obtain the abundance of dimer and monomeric
forms. The elution times were compared to a standard curve to determine the corresponding
molecular weights. The final protein yield was approximately 65 mg/L.
Crystallization, data collection, and refinement
Conditions for the crystallization of endoHCYS (10 mg/ml in 20 mM Tris-HCl, pH 7.5, 50 mM
NaCl, 5 mM EDTA) were initially determined using a high-throughput microbatch-under-oil
method at the Hauptman Woodward Institute High Throughput Crystallization Screening
Center [25]. The resulting conditions were optimized, and crystals grown in microbatch under
oil by mixing 1 μL protein with 1 μL mother liquor (100 mM TAPS, pH 9, 200 mM magnesium
nitrate, and 20% PEG-20,000 (% w/v)) at room temperature. Before cryo-cooling in liquid
nitrogen, four rounds of increasing glycerol cryoprotection (30 seconds for each increase of
10%) were carried out in mother liquor.
Cryoprotected crystals were shipped to the Advanced Photon Source at Argonne National
Laboratory and diffraction data were collected on a single crystal on beam line 17-ID (IMCA--
CAT). The photon energy used was 12.4 keV (1 Å) with data collected on a PILATUS 6M
detector at 350 mm distance using oscillations of 0.25˚ over a 90˚ rotation range. The data
were integrated with MOSFLM [26, 27] and scaling was performed with AIMLESS [28].
Phases were determined using molecular replacement with the structure of monomeric
endoH [20] BALBES [29]. The structural model was built using AUTOBUILD [30] and manu-
ally extended in Coot [31]. Refinement was performed using an iterative process with PHENIX
PLOS ONE | https://doi.org/10.1371/journal.pone.0239702 November 17, 2020
3 / 23
PLOS ONE Engineering a method for site-specific damage studies with SAXS
[32] and Coot. Validation was carried out with MolProbity [33]. The coordinates were depos-
ited as PDB ID 6VE1. Detailed statistics for data collection, processing, and refinement are
shown in S1 Table. Structural figures were prepared using PyMOL (Schro¨dinger; New York,
NY) and UCSF Chimera [34].
SAXS data collection
SAXS data were collected using beamline 12.3.1 (SIBYLS) at the Advanced Light Source [35].
The photon energy used throughout was 11.0 keV (1.127 Å). Momentum-transfer values were
calculated as q = 4πsinθ/λ, where 2θ is the scattering angle and λ is the X-ray wavelength in Å.
Data were recorded using a PILATUS 2M detector (Dectris; Philadelphia, PA). A volume of
25 μl of each sample was loaded into the sample chamber. The exposure time for each frame was
0.3 s and a total of 33 frames were collected for each sample in a static position. Buffer from
desalting column flow-through was used for matched controls and buffer subtraction. The beam-
line staff experimentally determined the beam profile (top-hat) and that the flux at the sample
was 2.04x1012 ph/s was derived based on the scattering of water [36]. RADDOSE-3D modified
for SAXS experiments [37] was used for calculating the dose rate (121 Gy/s), taking into account
the attenuation (10%) by the sample container (20 μm mica), the beam type (top-hat), and the
beam dimensions (3.4 mm2). The cell path length was 1.3 mm +/- 0.1 mm. The parameters used
for calculating the X-ray dose rate are available in S2 Table and the data collection parameters
are summarized in S3 Table. Standard tests were performed to determine how accurately volume
fractions could be predicted using the calculated scattering of the dimer and monomer crystal
structures. Molecular weight (MW) estimates with SAXS carry an error of about 10% [38]. How-
ever, here the predicted molecular weight of the dimer (48.7 kDa) is ~20% lower than the actual
MW (60 kDa), while the calculated scattering of the monomer (26.0 kDa) reasonably matches
the actual MW (30 kDa). The discrepancy of the dimer MW is explained by the flexibility of the
dimer from the engineered linker. MW determination of flexible proteins is a common challenge
with SAXS [39] and makes determining experimental volume fractions (VF) difficult. The dimer
was designed to be susceptible to X-ray radiation by forming a monomer due to radiation
induced cleavage of the covalently bonded disulfide forming the dimer. An experimentally deter-
mined dimer and monomer mixture (Fig 1) was used as a starting point to monitor the change
in ratio of the two species already known to be present as a function of dose. This reduced uncer-
tainty in comparison between non-irradiated (SEC) and initial irradiated (SAXS) VFs.
The rate of formation of the disulfide radical (a precursor to cleavage) [2] and the type of
radicals generated in water-cysteine solutions [40] are pH dependent. To explore the influence
of pH on radiation damage, endoHCYS (pI 5.65) was irradiated across a range of pH (5.0, 6.0,
7.5, and 9.0) values while monitoring for structural changes with SAXS. endoHCYS, initially
purified at pH 7.5, was exchanged into three additional buffers with varying pH values ((1) 20
mM NaC2H3O2, pH 5, 50 mM NaCl, 5 mM EDTA, (2) 20 mM MES, pH 6, 50 mM NaCl, 5
mM EDTA, (3) 20 mM Tris-HCl, pH 7.5, 50 mM NaCl, 5 mM EDTA, or (4) 20 mM Bis-tris
propane, pH 9, 50 mM NaCl, 5 mM EDTA) using Zeba desalting columns (Thermo Fisher;
Waltham, MA). The samples equilibrated at each pH for approximately one week at 4˚C prior
to data collection. Protein concentration was determined from the absorbance at 280 nm and
three concentrations (5.0, 2.5, and 1.25 mg/ml) for each pH value were analyzed to determine
concentration-dependent effects. Two independent dose series (replicates) at a protein con-
centration of 5.0 mg/ml for each pH were collected and averaged to improve signal prior to
modelling. Scattering of the monomer alone as a control was collected by reducing the disul-
fide bonds with 2 mM DTT and data was collected similarly to the dimer. All buffer subtracted
scattering curves used in the analysis are shown in S1 and S2 Figs.
PLOS ONE | https://doi.org/10.1371/journal.pone.0239702 November 17, 2020
4 / 23
PLOS ONE Engineering a method for site-specific damage studies with SAXS
Fig 1. Introducing a cysteine containing fragment leads to disulfide linked dimerization of endoH. In (a) SDS-PAGE analysis shows that the
apparent MW of endoHCYS is reduced in the presence of a reducing agent. In (b) SEC analysis shows that about 75% of endoHCYS forms a dimer
in solution (b, inset). Comparison of the elution times for each peak with a standard curve shows that the dominant peak has a MW
corresponding to the dimer (53 kDa) and the minority peak has a MW corresponding the monomer (28 kDa).
https://doi.org/10.1371/journal.pone.0239702.g001
Analysis of SAXS profiles
The ATSAS program suite (EMBL) was used for all data analysis [41] except where otherwise
noted. I(q) error bars and P(r) functions were calculated using the GNOM program from
ATSAS [41]. Radius of gyration (Rg) values were calculated from the Guinier region with
�qmax ~1.3. To account for the flexible liner, molecular weight (MW)
ranges according to Rg
was calculated using the volume of correlation (Vc) method [16], which is more accurate for
conformationally dynamic systems than other approaches [15]. Total integrated intensity, sin-
gular value decomposition (SVD), elongation ratio (ER), and residuals between experimental
and theoretical scattering curves were calculated using custom Python scripts. Volume fraction
(vk) was calculated with a Mathematica script according to MW ¼
k¼1 vkmwk and vk ¼ 1 ¼
þ vkmonomer where the MW is 48.7 kDa and 26.0 kDa for dimeric and monomeric forms,
vkdimer
respectively, based on the MW estimates for each component from SAXS. Fits of VF trajecto-
ries to first order descriptions were performed with a one-phase exponential equation,
vf ðdÞ ¼ vf0 e(cid:0) kd, where vf(d) is final dose dependent volume fraction, vf0
is the initial volume
fraction, k is the rate constant (Gy-1), and d is the X-ray dose (Gy). This was performed in
Mathematica using the NonlinearModelFit function. Calculated scattering from crystal struc-
tures was calculated with CRYSOL [42]. To follow the fragmentation trajectory, OLIGOMER
[43] was used to determine the fit and volume fraction of dimer and monomer components to
a series of experimental scattering data. The monomer component was based on the first expo-
sure (36.3 Gy) of the experimental monomer at pH 7.5 and 5.0 mg/ml protein concentration
and the dimer component was developed by subtracting the volume fraction weighted
PK
PLOS ONE | https://doi.org/10.1371/journal.pone.0239702 November 17, 2020
5 / 23
PLOS ONE Engineering a method for site-specific damage studies with SAXS
monomer component from the first exposure (36.3 Gy) of experimental scattering of the
dimer-monomer mixture also at pH 7.5 and 5.0 mg/ml concentration. Ab initio electron den-
sity reconstruction was performed with DENSS, an algorithm particularly suited to potential
dynamics present in the system using the dimer (monomer subtracted) scattering curve [44].
The characteristics calculated for the data series used for modelling are summarized in S4
Table.
Ensemble optimization
The disulfide forming fragment was designed to be flexible, and likely to sample many confor-
mations in solution. Therefore, the ensemble optimization method (EOM) [45] was employed.
This approach can generate models with different conformations between the monomer
domains, can find an ensemble of conformations that together best explain the experimental
scattering data, and is sensitive to any conformational distribution of the dimer over the course
of irradiation.
Data used for modelling was collected close to physiologic conditions at pH 7.5. To improve
signal-to-noise, two identical replicate dose series at pH 7.5 and 5.0 mg/ml concentration were
averaged and used for modelling. EOM was performed with monomeric endoH [20] as a rigid
body. To simulate a disulfide bond between two endoH monomers a protocol was adapted
from Tian et al. [46] where a Cys-Cys fragment was extracted from PDB entry 1HZH and
treated as another rigid body. The program RANCH from the EOM package [47] was then
used to generate 10,000 random conformations of a linker between each endoH and the Cys-
Cys fragment, thereby simulating a hinge motion. The final pool contained 10,000 monomeric
and 10,000 dimeric structures. The optimized ensemble was selected using the genetic algo-
rithm (GA) in the program GAJOE [48] and was repeated 100 times.
Results
Engineering and crystal structure of an X-ray cleavable disulfide linked
dimer
Soluble endoHCYS was produced. Analysis of the mobility shift on SDS-PAGE in the presence
of a reducing agent showed that the protein migrated to a lower MW, indicating the presence
of inter-molecular disulfides (Fig 1A and S3 Fig). Comparison of the elution profile of
endoHCYS in size-exclusion (SEC) chromatography shows that transient disulfide bond forma-
tion leads to dimerization in approximately 75% of endoHCYS (60 kDa theoretical, 53 kDa
experimental) in solution (Fig 1B).
A structural model was determined to a resolution of 2.1 Å by X-ray crystallography to
determine the orientation of the dimer (PDB 6VE1) (Fig 2). The protein crystallized in space
group P21221 with four monomers in the asymmetric unit (Fig 2A). The monomers have an
average RMSD 0.289 +/- 0.016 Å to the original monomeric structure [20], indicating that the
cysteine containing fragment did not affect folding. A PISA [49] analysis of the interfaces
within the asymmetric unit suggests that there are too few interactions between crystal packing
oligomers to for those interactions to form in solution. The cysteine containing fragment and
disulfide bond are un-resolved in the electron density but is not unexpected due to the deliber-
ate placement of the disulfides in an accessible and already flexible region. The position of the
N-terminus indicates that the disulfide linked dimer forms between monomers across adjacent
asymmetric units, yielding dimers with chains AA’, BB’, CD’, and C’D (Fig 2B). This arrange-
ment with the N-terminus of adjacent monomers positioned in close proximity is not seen in
crystal structures of monomeric endoH [20, 21]. The distance between the modeled N-
PLOS ONE | https://doi.org/10.1371/journal.pone.0239702 November 17, 2020
6 / 23
PLOS ONE Engineering a method for site-specific damage studies with SAXS
Fig 2. Crystal structure of dimeric endoHCYS suggests a highly exposed and flexible disulfide bond linkage. In (a) the crystal packing diagram showing that the
asymmetric unit contains four monomers with the suggested dimers illustrated with chains that have matching color. (b) A front view of an alignment of one chain from
each dimer is shown with the (c) top view of the alignment indicating the position of the N-terminus and disulfide linkage. The slight differences in structure of the
individual dimers are seen by the non-ideal alignment.
https://doi.org/10.1371/journal.pone.0239702.g002
terminal Val-8 on partnered monomers (13 Å for BB’ and AA’, 7 Å for both CD dimers) is
such that the length of the introduced fragment can satisfy the gap and connect the monomers
(Fig 2C). Aligning the dimers shows slight differences in the inter-monomer orientation,
which is most likely due to crystal packing (Fig 2B).
X-ray radiation drives protein fragmentation in solution experiments
The quality of SAXS data from globular monomodal samples can be evaluated by various qual-
ity criteria [50]. In our case we expect a mixture of dimers and cleaved monomer formation so
these criteria are not suitable as quality indicators but can be explored to monitor the dimer to
monomer progression. The fundamental evidence of radiation damage in SAXS is seen
through changes in the scattered intensity. An increase in scattering at low-q indicates aggre-
gation and a decrease indicates fragmentation [51]. For the endoHCYS irradiated at four con-
centrations at pH 7.5, there was no evidence in the data indicating buildup of damaged protein
on the sample cell windows over time. There was a dose-dependent decrease in intensity at
low-q (q ~ 0.01–0.07 Å-1) and an increase beginning at mid-q (q > ~ 0.07 Å-1) that was notice-
able between the first (36.3 Gy) and second exposure (72.6 Gy). This change in the shape of the
scattering curve is characteristic of fragmentation (Fig 3A and 3C). An isoscattering point was
observed at q ~ 0.07 Å-1 indicating a transition within a system consisting of only two compo-
nents [52]. This was predicted based on the theoretical scattering of the monomer and dimer
crystal structures (S4 Fig) and confirms fragmentation of the dimer. The difference in the loca-
tion of isoscattering point in calculated and experimental scattering curves suggests that the
conformation of the dimer is different in solution, perhaps through freedom crystal lattice
restrains to rotate around the flexible linker. The intensity in the Guinier region exhibited a
dose dependent decrease while remaining linear (Fig 3D). The slope of the Guinier region is
related to the radius of gyration (Rg). Although dose does not affect the linearity of the Guinier
region, the initial Rg (36.3 Gy) was 30.6 Å and decreased after the total accumulated dose (1.2
kGy) to 27.2 Å (S4 Table). The Rg calculated from the theoretical scattering of monomeric
endoH and the average of dimers from the crystal structure (Fig 2 and S4 Fig) are 19.4 and
31.8 +/- 1.65 Å, respectively. A dimensionless Kratky plot provides a semi-quantitative
approach to assessing protein change that is normalized for differences in particle mass [53].
As endoHCYS receives more radiation, the bell-shaped intensity curve in the Kratky plot
becomes taller and narrower indicating a change in sample shape (Fig 3C). The elongation
PLOS ONE | https://doi.org/10.1371/journal.pone.0239702 November 17, 2020
7 / 23
PLOS ONE Engineering a method for site-specific damage studies with SAXS
Fig 3. X-ray solution scattering analysis indicating that fragmentation is radiation dose dependent. The color gradient
corresponds to the magnitude of the absorbed dose (36.3 Gy-1.2 kGy) delivered across 33 x 0.3 sec exposures where yellow is
low dose and black is high dose. EndoHCYS was irradiated at pH 7.5 and the results shown are averages of two replicates at a
concentration of 5.0 mg/ml. In (a) buffer subtracted scattering curves exhibit a dose-dependent decrease in intensity. In (b)
residuals calculated between the first exposure and each sequential exposure show that intensity decreases at low-q and increases
at high-q with increasing absorbed dose. The arrow identifies an isoscattering point at q ~ 0.07 Å-1, characteristic of a transition
between two states. In (c) a pairwise distance distribution plot, P(r), normalized to MW, of the first, last exposure, and difference
between first and last exposure shows that the fraction of longer distances in the protein decreased. The slope of the Guinier
region remains linear (d) but decreases, indicating that the size of the protein is decreasing. A Dimensionless Kratky plot (e) also
indicates that the protein is changing shape during irradiation.
https://doi.org/10.1371/journal.pone.0239702.g003
PLOS ONE | https://doi.org/10.1371/journal.pone.0239702 November 17, 2020
8 / 23
PLOS ONE Engineering a method for site-specific damage studies with SAXS
ratio (ER), which is calculated from the P(r) curve, estimates protein compactness where ER ~
1 is compact and ER >> 1 is elongated [54]. The ER for endoHCYS becomes smaller at higher
doses (ER36.3 Gy = 2.17 and ER1.2 kGy = 1.91) also indicating the protein is shifting to a more
compact state. This change in shape is also reflected in the loss of longer distances in the 30–80
Å range (Fig 3C). Controls with the monomer (S2 Fig) and buffer (S5 Fig) did not yield dose-
dependent changes. Together, these results reveal a two-component system with a dose depen-
dent transition from the flexible disulfide linked dimer to a compact monomer. Fragmentation
occurs without any evidence of aggregation or inter-particle interactions in the Guinier region
(Fig 3D).
Damage pathway is pH dependent
The fragmentation observed in SAXS (Fig 3) indicates disulfide bond cleavage occurred from
radiation exposure. To explore the role of pH on radiation damage, additional SAXS experi-
ments were performed where endoHCYS (pI 5.65) was irradiated with SAXS at three additional
pH values (5.0, 6.0, and 9.0) and three concentrations and compared to the previous data that
was collected at pH 7.5 (Fig 3). The results indicate that the magnitude and direction of the
change in the intensity at zero scattering angle, I(0), were strongly pH dependent (Fig 4). Until
254 Gy, samples at each pH exhibited a decrease in I(0) agreeing with fragmentation, but the
magnitude of the change increased with increasing pH values. Specifically, at the total accumu-
lated dose (1.2 kGy), the overall decrease in I(0) at pH 6.0 was much less pronounced than at
Fig 4. Fragmentation and aggregation damage pathways are pH and concentration dependent. Samples were irradiated at four pH values (5.0, 6.0, 7.5 and 9.0) and
three concentrations each (5.0, 2.5, and 1.25 mg/ml). Experiments at 5.0 mg/ml were repeated twice with fresh sample. In (a-d) I(0) and (e-h) Rg values for each
experimental condition were calculated with AUTORG [55]. I(0) values were normalized by dividing by the concentration.
https://doi.org/10.1371/journal.pone.0239702.g004
PLOS ONE | https://doi.org/10.1371/journal.pone.0239702 November 17, 2020
9 / 23
PLOS ONE Engineering a method for site-specific damage studies with SAXS
pH 7.5 and 9.0, while pH 5.0 showed an overall increase in I(0) (Fig 4A–4D). The increase at
pH 5.0 indicates aggregation and was not apparent at the other pH values. Similar results were
seen in Rg (Fig 4E–4H). At a particular pH, a similar process was seen across concentrations
but the magnitude varied. At pH values where fragmentation occurred (i.e. pH 6.0, 7.5, and
9.0) fragmentation appeared greatest at low concentrations. At pH 5, which predominately
aggregated, the increase in I(0) and Rg was greater at higher concentrations. Different struc-
tural trajectories at high and low pH are also indicated by opposite shifts in respective Kratky
plots (S6 Fig). The initial scattering of the dimer component at each pH were overall similar
indicating these changes were in part radiation driven (S7 Fig). The scattering of the monomer
component alone showed a slight dose dependent increase in I(0) and Rg at low pH but was
largely unaffected by radiation (S8 Fig). These results indicate that different damage mecha-
nisms (fragmentation and aggregation) occur in the same sample with the same radiation dose
but at different solution pH values. Altering the solution pH will affect the surface charge of
the protein and the composition of the solvent. The aggregation at pH 5.0 and fragmentation
at pH 6.0 span the isoelectric point, 5.65. The different results suggest a residue specific effect
with the consequence that aggregation caused by radiation effects might be mitigated by
adjusting the pH of the buffer.
Single value decomposition (SVD) allows the number of distinct meaningful components
that contribute to a series of data to be estimated [56]. An SVD analysis of the dose series of
scattering data shows that endoHCYS pH 5.0 and 6.0 contain multiple components (with values
greater than zero) that contribute to the overall scattering (Fig 5A and 5B). This is characteris-
tic of aggregation where many particles of different sizes are generated through non-specific
radical induced cross-linking [57]. Interestingly, while pH 6.0 exhibited an initial intensity
decay indicating fragmentation, the fragmentation is greatly reduced compared to pH 7.5 and
9.0 (Fig 4B). The SVD analysis shows multiple scattering components at pH 6.0 while only two
at pH 7.5 and 9.0 (Fig 5B and 5C). This suggests that the reduced fragmentation at pH 6.0 is
Fig 5. Fragmentation includes two components. The horizontal units are arbitrary. Singular value decomposition analysis (SVD) was conducted on data at 5.0 mg/
ml for each pH value. In (a) pH 5.0 and (b) 6.0 shows the scattering is from multiple components, which is characteristic of aggregation. SVD at (c) pH 7.5 and (d)
pH 9.0 shows the scattering is from two components, suggesting a process with only monomer and dimer components.
https://doi.org/10.1371/journal.pone.0239702.g005
PLOS ONE | https://doi.org/10.1371/journal.pone.0239702 November 17, 2020
10 / 23
PLOS ONE Engineering a method for site-specific damage studies with SAXS
caused by underlying and simultaneously occurring aggregation that results in a less substan-
tial decay in Rg and I(0) as a function of X-ray dose (Fig 3). SVD estimated that the scattering
of the monomer alone (S2 Fig) at each pH contained an equal number of components (S9 Fig)
and validates that the differences observed in the dimer-monomer mixture between high and
low pH are due to different damage processes specific to the dimer. Importantly, the two com-
ponents present at high pH agree with a dimer to monomer transition through a disulfide
bond cleavage.
A dimer to monomer transition explains fragmentation
The samples at pH 7.5 and pH 9.0 at 5.0 mg/ml concentration included two components based
on the SVD analysis (Fig 5) and enabled a volume fraction (VF) analysis to be performed. The
multiple components present in samples irradiated at pH 5.0 and 6.0 prevented a VF analysis.
The average molecular weight of the monomer and dimer components were calculated (vol-
ume of correlation, Vc) [16] using the experimental monomer scattering (S10 Fig) and calcu-
lated scattering of the dimer structure (S4 Fig). The VF analysis shows that at both pH 7.5 and
9.0 the protein is roughly 25% monomer and 75% dimer after the first exposure of 36.3 Gy,
which is in agreement with the experimental SEC analysis (Figs 1B and 6B). The VF trajecto-
ries suggests that the rate of fragmentation decreases as dose accumulates (Fig 6A and 6B).
This indicates that the rate is proportional to the amount of dimer remaining in solution,
which reflects a first order process and is a common description for many chemical and bio-
logical processes. The fit of the VF trajectories to a first order exponential equation indicates
that the total amount of fragmentation was greater as concentration decreased (Fig 6C and S5
Table). The rate of fragmentation was consistent across concentration and slightly reduced at
pH 9.0 compared to pH 7.5 (Fig 6D). However, this difference in the rate of fragmentation did
not lead to a systematic change in the magnitude of fragmentation between the two pH values
(Fig 6C). Interestingly, the sample never transitions to an entirely monomeric system but
reaches an equilibrium at approximately 600 Gy in spite of additional X-ray doses.
Collecting experimental data for the dimer alone is challenging as a mixed dimer/monomer
system is created from the initial exposure (Fig 4). While a calculated curve from the crystallo-
graphic model could be used, the dimers in the asymmetric unit differ through small inter-
monomer rotations, and that difference is distinguishable in their calculated scattering curves
(S4 Fig). To overcome this, an experimental curve for the dimer was obtained by subtracting
the volume fraction weighted contribution (based on MW) of the experimental scattering of
the monomer from the experimental scattering of the mixture at the first exposure (36.3 Gy).
Comparison of the monomer subtracted scattering (dimer) to the theoretical scattering of the
crystal structure dimers yielded reasonable agreement at low-q, validating the approach (S11A
Fig). The resulting electron density reconstruction supports a dimer but in a slightly different
orientation to that seen in the crystallographic model (Fig 2 and S11B and S11C Fig). This elec-
tron density model was consistent with one produced from a conventional bead modelling
approach (S12 Fig). That the solution structure is distinct from the crystal structure is not
unexpected given the removal of crystal packing constraints and the intended flexibility of the
designed linker.
An additional VF analysis was performed for the data closest to physiological conditions,
pH 7.5, using the scattering curve of the monomer subtracted (dimer) and the experimental
monomer scattering as additional restraints. This analysis was in good agreement with the pre-
vious VF analysis based on MW alone (Figs 6A and 7A). While the fits to the experimental
data were reasonable, they worsened at higher doses, as evidenced by the χ2 values (Fig 7B).
The monomer subtracted scattering represents the dimer well at low doses, but it becomes less
PLOS ONE | https://doi.org/10.1371/journal.pone.0239702 November 17, 2020
11 / 23
PLOS ONE Engineering a method for site-specific damage studies with SAXS
Fig 6. The fragmentation processes can be modelled by a dimer to monomer transition and follows first-order kinetics. The
volume fraction (VF) of monomer and dimer components were estimated based on the MW using the volume of correlation, Vc,
method. The MW of monomer and dimer components were based on the experimental scattering of the monomer and calculated
scattering of the dimer. VF analyses were performed at pH 7.5 (a) and 9.0 (b). In (c) The total percent change in monomer was
calculated using the VF of monomer at the first exposure (36.3 Gy) and the last exposure (1.2 kGy). In (d) The rate constants, k1,
were calculated by fitting the data to one-phase exponential decays. Error bars represent the standard error in approximating the
rate constants.
https://doi.org/10.1371/journal.pone.0239702.g006
accurate at higher doses (Fig 6A). The experimental scattering of the monomer component
alone did not change during irradiation (S8 Fig) and suggests that the conformation of the
dimer changes during irradiation.
To account for possible changes in conformations, the fragmentation series at pH 7.5 were
analyzed using the ensemble optimization method (EOM) [45]. The resulting EOM models
(w2 � 1:65) to the experimental data was greatly improved compared to fits with the crystal
structures (w2 � 4:00), especially at high-q (Fig 7B and 7C and S13 Fig). The volume fraction
analysis from EOM is also in reasonable agreement with the previous approaches, showing
that the monomer fraction becomes the dominant species at higher doses (Figs 6, 7A and 7C).
Most ensembles chosen by the genetic algorithm contained on average 15 models, with the
largest containing 20, validating the assumption about conformationally flexibility. This is also
confirmed by analysis of the resulting ensemble distributions. The ensemble of the first expo-
sure (36.3 Gy) is more flexible than the random pool (Rflex-ens/Rflex-pool = 84.04% / 72.14%).
After the total accumulated dose (1.2 kGy) the ensemble is as flexible as the random pool
(Rflex-ens = 72.14%). This agrees with the Kratky plot and P(r) distribution that indicated that
the protein became more compact with increasing doses of X-rays (Fig 3). This is most likely
PLOS ONE | https://doi.org/10.1371/journal.pone.0239702 November 17, 2020
12 / 23
PLOS ONE Engineering a method for site-specific damage studies with SAXS
Fig 7. Radiation drives a change in the magnitude and conformation distribution in the engineered dimer. In (a) the program OLIGOMER [43] was used to
calculate the volume fraction of monomer and dimer components considering the scattering curve of each component. The first exposure (36.3 Gy) of the
experimental scattering of the monomer at pH 7.5 and 5.0 mg/ml served as the monomer component and the dimer component was developed by subtracting the VF
weighted contribution of the monomer component from the first exposure (36.3 Gy) of experimental scattering of the mixture. In (b) χ2 values show that the (pink)
EOM approach explains the data better than the OLIGOMER approach (purple). In (c) the volume fractions calculated from the EOM results that follow the
fragmentation process. Population results from the EOM analysis are shown in (d) with the distribution of Dmax, (e) Rg, and (f) volume of the selected ensembles.
Dotted line represents the random pool. The color gradient corresponds to the magnitude of the absorbed dose (36.3 Gy-1.2 kGy) delivered across 33 x 0.3 sec
exposures where yellow is low dose and black is high dose.
https://doi.org/10.1371/journal.pone.0239702.g007
due to the generation of monomers that are less flexible than the linked dimer and are pre-
dominate at higher doses. However, as the sample receives more dose, the distribution of Rg
and Dmax of the dimer population also changes suggesting that radiation drives a change in
both the magnitude and conformational distribution of the dimer. Specifically, at low doses
the distributions of both Rg and Dmax are somewhat monomodal and centered relative to the
random pool but as dose increases both transition to a bimodal distribution. The sample loses
moderate (~33 Å and ~115 Å) conformations and becomes enriched with both compact
(~27.5 Å and ~87.5 Å) and extended conformations (~37 Å and ~130 Å) (Fig 7D–7F and S14
Fig). Fitting the dose-dependent changes at these positions to a logarithmic regression shows
that for Rg the middle portion of the peak (33 Å) decreases more quickly (-0.237 density (arbi-
trary units)/Log(Gy)) than the extended (37 Å; -0.029 density/Log(Gy)) and compact portions
(27.5 Å; -0.116 density/Log(Gy)) (S14A Fig and S1 Movie). This was also observed for Dmax
where mid-length distances (115 Å) decreased faster at a rate of -0.292 density/Log(Gy)) com-
pared to extended (130 Å; 0.090 density/Log(Gy)) and compact (87.5 Å; -0.091 density/Log
PLOS ONE | https://doi.org/10.1371/journal.pone.0239702 November 17, 2020
13 / 23
PLOS ONE Engineering a method for site-specific damage studies with SAXS
(Gy)) distances (S14B Fig and S2 Movie). These results suggest that in addition to changes in
the magnitude of the dimer population, the conformation distribution of the dimer also
changes over the course of irradiation which could be interpreted as a conformation depen-
dence on damage susceptibility. These dose driven changes in Rg and Dmax distributions of the
dimer were similar at pH 7.5 and 9.0 (S15 Fig).
Discussion and conclusion
The overall goal of our study was to relate damage seen in the crystallographic state to that
observed in solution, specifically X-ray induced damage to the sensitive disulfide bond (cys-
tine). Disulfides are particularly susceptible to radiation damage due to the high electron affin-
ity of sulfur. They are attacked by solvated electrons generated as part of the photolysis of
water and these electrons are mobile even at cryogenic temperatures [2]. Under cryogenic con-
ditions and crystallographic doses as low as 5 kGy, the disulfide bond suffers reductive damage
[58]. Weik et al. [59] have explored disulfide damage below (100 K) and above (155 K) temper-
atures where the larger free radicals are mobile. Structural changes to the disulfide position
were seen at 155 K but not at 100 K, and the disulfides were highly radiation sensitive at 155 K.
This extends to the solution case where in addition to the solvated electrons, disulfides can also
be impacted by radicals that are otherwise cryogenically trapped. Radiation damage to cysteine
has been studied in solution for decades but the reaction mechanism for the diverse array of
products are not fully understood [60]. The results presented here cannot specifically deter-
mine or quantitate the reactions with the larger radicals trapped in the cryogenic state but do
provide qualitative evidence that they are occurring.
In our case changes in the scattering curve that are characteristic of fragmentation are
observed with doses less than 100 Gy, fifty times less than the crystallographic studies [59].
Fragmentation is confirmed by changes in basic parameters such as decreases in Rg and I(0)
(Fig 4). SVD analysis shows fragmentation includes a two-component system, which supports
fragmentation through disulfide bond cleavage rather than main chain breakage (Fig 5). These
results from SAXS are compatible with data from the large number of solution studies on cys-
teine and damage seen in the crystallographic case. Here, no sample reached a completely
monomeric state and the rate of monomerization decreased as the absorbed dose accumulated
(Fig 6). Irreversible bond breakage is favored where the bond rupture causes a positional shift
[61]. Incomplete monomerization could be explained by the presence of repair pathways,
which can prevent bond breakage leading to fragmentation but is unlikely to lead to re-associa-
tion of previously fragmented monomers. A repair pathway occurs in crystallographic studies,
but it was not seen to have a noticeable impact [58]. Damage to disulfide bonds is not uniform
and can vary with the immediate environment in the protein. The linker connecting the dimer
was designed to be flexible and therefore the dimer is most likely sampling conformations in
solution (Fig 2). Analysis of ensemble modelling reveals that the conformation of the dimer
changes as dose accumulates (Fig 7). This indicates that the local bond environment differs
within the conformational ensemble and despite the solvent accessibility, may have some
impact on the incomplete fragmentation observed. This is consistent with crystallographic
studies where disulfide orientation and solvent exposure influence radiation damage suscepti-
bility [62]. The EOM results are not precluded by the SVD analysis that suggests that there are
only two components in solution (Fig 5). While SVD is a powerful method it has weaknesses.
SVD describes the minimum number of components necessary to explain the experimental
data. As others have noted, it might not identify all states that exist. This is particularly prob-
lematic for studies on disordered systems or kinetics experiments with short exposures [63–
65], which is the focus of this study.
PLOS ONE | https://doi.org/10.1371/journal.pone.0239702 November 17, 2020
14 / 23
PLOS ONE Engineering a method for site-specific damage studies with SAXS
The rate of formation of the disulfide radical (a precursor to cleavage) [2] and the type of
radicals generated in water-cysteine solutions [40] are pH dependent. The susceptibility of
disulfides to cleavage is also influenced by local conditions such as pH [2, 4, 66]. Our results
with pH values spanning the isoelectric point (pI ~ 5.65) all yielded initial changes in the I(0),
Rg, and MW that are indicative of fragmentation, which accelerated as pH increased (Fig 4).
Fragmentation appeared inversely related to concentration where fragmentation was greater
at lower concentrations (Fig 6). However, in contrast to global damage [57], fragmentation
was not entirely dose dependent as it reached an equilibrium prior to complete monomeriza-
tion (~ 600 Gy) in spite of additional X-ray doses (Fig 6). It is unclear if this plateau effect,
which could be described by first order kinetics, reflects the concentration dependence of the
underlying solution chemistry or the protein.
At pH 6.0 (near the isoelectric point), the change in intensity was small and remained
within 10% of the initial values (Fig 4) and could be due to the presence of aggregating species
as suggested by the SVD analysis (Fig 5). The magnitude of fragmentation was similar at pH
7.5 and 9.0 but much greater than at pH 6.0 (Fig 6). By necessity each pH condition was pre-
pared with different buffering agents since there are no single biologic buffers that function
over the wide range of pH studied. These factors add additional layers of complexity to the
analysis but shows that the solution pH whether through hydrogen ion activity, chemical prop-
erties, or both, influences both global and a specific radiation damage mechanism, aggregation
and fragmentation, in SAXS.
Since thiol groups have a pKa of 8.3 and 8.5 disulfides are be more stable at higher pH values
[66]. The protein was initial purified at pH 7.5 so that each sample should at least 75–80%
dimer based on the SEC analysis (Fig 1) and as covalent bonds, they are unlikely to be reversed
by subsequent changes in the pH of the buffer. However, initial estimates of the dimer VF
decreased with concentration (Fig 6). It is possible that a small population of dimers formed
through electrostatic interactions, which are known to be influenced by both pH and protein
concentration. The presence of non-covalently associated dimers might also explain the
increased propensity for aggregation that was observed at low pH as well as the fragmentation
resistant population of dimers (Fig 7). However, another explanation is that the discrepancy in
initial VF estimates between concentrations is due to differing degrees of damage that occur
within the first exposure. Lower protein concentrations exhibited greater total fragmentation
than higher protein concentrations at the same pH (Fig 6). Since fragmentation was apparent
between the first and second exposures (Fig 3), it is likely that fragmentation also occurs at
doses lower than the initial dose. How additional factors such as X-ray dose-rate or diffusion
rate influence disulfide bond cleavage in solution are not examined by this study. The path-
ways for damage are more complex in solution due to the free motion of radical species that
are otherwise cryogenically trapped. Serial crystallography studies of lysozyme show that global
damage occurs at five times the rate of specific damage for crystals grown at pH 4.5 [12]. Other
studies at slightly higher pH and different experimental conditions suggest a lower rate [14].
Comparing the crystallographic results to the solution case for disulfide damage is difficult.
Typically, SAXS is only sensitive to global damage through changes in the scattered intensity.
Here, our system is engineered so that the reduction in scattered intensity is from the dimer to
monomer transition through disulfide bond cleavage, making a global damage indicator into a
residue specific one. This approach allows opens the way for to a detailed understanding of
how crystallographic details of disulfide cleavage can be related to those occurring at physio-
logical conditions.
At the lowest pH studied (pH 5.0), aggregation became dominant at higher absorbed doses.
This was not observed at high pH conditions. The mechanism underlying the aggregation is
unclear but suggests that it is charge-dependent and that X-ray damage in SAXS can be more
PLOS ONE | https://doi.org/10.1371/journal.pone.0239702 November 17, 2020
15 / 23
PLOS ONE Engineering a method for site-specific damage studies with SAXS
nuanced than a completely non-specific process as previously noted in studies of lysozyme
where main chain cleavage was not detected [57]. In the lysozyme case, the aggregation
increased as a function of dose rate and also increased inversely to concentration. Crystallo-
graphic studies suggest that disulfide orientation and solvent exposure influence susceptibility
[62]. In the lysozyme case, and for other proteins, although it is highly likely disulfide bonds
are being broken, they may be buried, or other parts of the structure may constrain the overall
envelope so that any structural change is not detected at the resolution of the technique [67].
Our case was designed with an easily accessible disulfide bond that when broken, gave a clear
signal measurable by SAXS. For systems that have domains natively connected by exposed
disulfide bonds (e.g. antibodies) [68, 69], the same process that induces a dimer monomer
transition could be misinterpreted as functional dynamics in the system, underscoring a note
of caution in interpretation of results and the importance of monitoring radiation damage
effects in preliminary SAXS data analysis.
The diffusion present in a solution study allows transport of damaged and undamaged
components which can impact the measured intensity [51]. Time also has an influence, not
only in diffusion processes, but also in secondary damage due to the cascade of radicals
formed. These experiments were purposely performed with a large beam size (3.4 mm2) to
reduce the impact of diffusion. Transport and time will have an increasingly significant impact
as the beam profile is reduced. Solution studies also lack crystallographic constraints, e.g. fixed
lattices, close packing, and defined solvent channels, which may impede any damage processes.
Our results are a snapshot of a more complex process that the crystallographic case but by the
method of engineering residue specific impact that can be detected by this low-resolution tech-
nique, we open a door to the study of complex radiation chemistry phenomena in biological
solution systems.
The doses used for X-ray crystallography are typically in the kilo-gray to mega-gray range.
While doses used in biomedical applications such as radiotherapy and imaging are 1–2 orders
of magnitude lower, structural mechanisms that occur in these settings can be studied with
SAXS by utilizing low-dose collection strategies [17]. Here, the engineered protein approach
allowed us to link disulfide breakage seen in crystallographic studies to damage in solution
with X-ray doses closer to those that impact human health. There is considerable potential to
extend our approach and look at specific damage to other residues and develop X-ray con-
trolled protein switches. For example, visible light-based control of engineered proteins has
been successfully used to modulate protein localization [70], enzyme activity [71], and gene
expression [72]. These same tools, but controlled by X-rays, could augment investigations of
cellular response to radiation exposure and develop treatment and mitigation strategies. While
not providing the same fidelity as crystallographic approaches, and still ignoring the impact of
radical formation of non-water species, this novel combination of the crystallographic and
SAXS opens up a means to study and understand low dose effects, harnessing the decades of
research in understanding X-ray radiation damage in a structural context.
Supporting information
S1 Fig. All buffer subtracted scattering patterns of the dimer used in the analysis. Four pH
values were tested (a: pH 5, b: pH 6, c: pH 7.5, d: pH 9) at three concentrations each (from top
to bottom: 5 mg/ml (replicate 1), 5 mg/ml (replicate 2), 2.5 mg/ml, and 1.25 mg/ml) across 33
exposures. The color gradient corresponds to the magnitude of the absorbed dose (36.3 Gy-1.2
kGy) delivered across 33 x 0.3 sec exposures where yellow is low dose and black is high dose.
(PNG)
PLOS ONE | https://doi.org/10.1371/journal.pone.0239702 November 17, 2020
16 / 23
PLOS ONE Engineering a method for site-specific damage studies with SAXS
S2 Fig. All buffer subtracted scattering patterns of the monomer used in the analysis. Four
pH values were tested (a: pH 5, b: pH 6, c: pH 7.5, d: pH 9) at three concentrations each (from
top to bottom: 5 mg/ml, 5 mg/ml, 2.5 mg/ml, and 1.25 mg/ml) across 33 exposures. The color
gradient corresponds to the magnitude of the absorbed dose (36.3 Gy-1.2 kGy) delivered
across 33 x 0.3 sec exposures where yellow is low dose and black is high dose.
(PNG)
S3 Fig. Uncropped images of the purified protein resolved on SDS-PAGE. In (a) the protein
was run in the absence of a reducing agent and in (b) the protein was incubated with DTT to
reduce the disulfides prior to performing PAGE. Odd numbered lanes correspond to a molec-
ular weight ladder and even numbered lanes correspond to increasing amounts of protein
sample. For the non-reducing the amount of protein was loaded in order: 1.25, 2.50, 5.0, and
10 ug, and for the reducing: 0.63 ug, 1.25 ug, 2.5 ug, and 5.0 ug. Lanes with a red X above them
correspond to portions of the gel not included in Fig 1.
(PDF)
S4 Fig. Calculated scattering of crystal structure models. The arrow indicates an isoscatter-
ing point that is predicted between the monomer and dimer structures at q ~ 0.025 Å-1. The
small difference in rotation observed in the dimer crystal structures is distinguishable through-
out the scattering curve.
(PNG)
S5 Fig. Buffer intensity does not change during irradiation. The total integrated intensity
was calculated for the buffer alone at each pH value tested. Error bars represent the standard
deviation from integrating across the scattering curve across the upper and lower bounds of
errors in I(q).
(PNG)
S6 Fig. Dimensionless Kratky plots of endoHCYS at different pH values. Arrows indicate the
direction of dose dependent shifts in the shape of the plot. Protein was at 5 mg/ml.
(PNG)
S7 Fig. pH does not affect the initial scattering of the dimer component. The dimer compo-
nent alone at each pH was developed by subtracting the VF-weighted contribution of the
experimental scattering of the monomer alone collected at the same dose (36.3 Gy), pH value,
and concentration (5.0 mg/ml).
(PNG)
S8 Fig. Rg and I(0) analysis of the monomer alone indicate no systematic changes from
radiation exposure. Experimental scattering data of the monomer (reduced disulfides) was
collected at four pH values (a: pH 5, b: pH 6, c: pH 7.5, d: pH 9) and three concentrations each
(from top to bottom: 5 mg/ml, 2.5 mg/ml, and 1.25 mg/ml) across 33 exposures.
(PNG)
S9 Fig. SVD analysis of the scattering of the monomer alone shows that the sample con-
tains the same number of components at each pH value. Singular value decomposition anal-
ysis (SVD) was conducted on data collected with a 5.0 mg/ml protein concentration for each
pH value: (a) 5.0, (b) 6.0, (c), 7.5, (d), 9.0.
(PNG)
S10 Fig. The experimental scattering of the monomer is in good agreement with the crystal
structure. The first exposure (36.3 Gy) of the monomer at each pH and 5.0 mg/ml was com-
pared the calculated scattering of the monomer crystal structure (PDB 1EDT) using CRYSOL.
PLOS ONE | https://doi.org/10.1371/journal.pone.0239702 November 17, 2020
17 / 23
PLOS ONE Engineering a method for site-specific damage studies with SAXS
χ2 represents the goodness of fit between the experimental and calculated scattering.
(PNG)
S11 Fig. The dimer component at low doses is distinct from individual orientations in the
crystal structure. In (a) the dimer component at 36.3 Gy was isolated by subtracting the vol-
ume fraction weighted scattering of the monomer crystal structure. Comparison of the scatter-
ing of this dimer component and the dimers from the crystal structure show good agreement
at high-q but poor fitting at low-q suggesting that the orientation is slightly different in solu-
tion. In (a) ab initio modelling of SAXS data for the dimer component was performed using
DENSS. The model as shown was contoured so that the volume of the envelope was close to
the Porod volume as measured from the scattering data (8.0 x 103 Å3). The final model aver-
aged from 100 independent reconstructions with P2 symmetry (blue line) has a resolution of
30.7 Å as determined from a 0.5 FSC cutoff (red line). Front (b) and side (c) views show that
the electron density can accommodate two monomeric units.
(PNG)
S12 Fig. Visual comparison between ab initio electron density reconstruction and bead
modelling from solution scattering data. The same data was used for both modelling
approaches: the first exposure (36.3 Gy) collected at pH 7.5 and 5.0 mg/ml protein concentra-
tion with the volume fraction weighted monomer component subtracted. In (a) the bead
model presented was the filtered model from averaging fifteen independent reconstructions
assuming P2 symmetry and produced with DAMMIN [73]. In (b) an electron density model
was produced with DENSS as described in S11 Fig. Two copies of the monomeric crystal struc-
ture (PDB 1EDT) were docked into the respective models using the sequential fitting tool in
UCSF Chimera [34].
(PNG)
S13 Fig. EOM modelling explains the experimental data better than OLIGOMER model-
ling at pH 7.5. In (a) the fits from modelling with OLIGOMER [43] are shown where the vol-
ume fraction weighted contributions of the monomer and dimer components are determined.
The first exposure (36.3 Gy) of the monomer at pH 7.5 and 5.0 mg/ml was used as the mono-
mer component. The dimer component was developed by subtracting the VF weighted (MW
based) monomer component from the scattering of the mixture at 36.3 Gy. Fits of the VF
weighted components (orange) are compared to the experimental scattering (black). Residuals
are shown in (b) where the black lines are references for perfect fits between the experimental
scattering and fitted models (where residuals are ~0). The fits from EOM modelling are shown
in (c) and residuals are shown in (d). Ensemble models are able to satisfy the experimental
scattering at all q values. The EOM modelling was also done using the average of two indepen-
dent series of exposures at pH 7.5 at 5 mg/ml.
(PNG)
S14 Fig. The rate of fragmentation is both dose and conformation dependent. Three points
along (a) Dmax and (b) Rg distributions from EOM were monitored for dose-dependent
changes in magnitude and fit (solid black, dotted and dashed) to a logarithmic regression to
determine rates of change.
(PNG)
S15 Fig. The dose driven change in the conformational ensemble of the dimer was similar
at pH 7.5 and 9.0. Differences in the Rg (a-c) and Dmax (e-f) distributions between pH 7.5
(blue) and 9.0 (red) were monitored at three dose points: 36.3 Gy, 181.3 Gy, and 1.2 kGy. The
distribution of the random pool of structures is shown in black. For both pH values, the data
PLOS ONE | https://doi.org/10.1371/journal.pone.0239702 November 17, 2020
18 / 23
PLOS ONE Engineering a method for site-specific damage studies with SAXS
used for modelling was averaged from the two replicates collected with a 5.0 mg/ml protein
concentration.
(PNG)
S1 Movie. Movie of dose-dependent radius of gyration distribution progression from EOM
modelling.
(GIF)
S2 Movie. Movie of dose dependent Dmax distribution progression from EOM modelling.
(GIF)
S1 Table. Data-collection and refinement statistics.
(DOCX)
S2 Table. Parameters for X-ray diffraction weighted dose calculations for SAXS experi-
ments using RADDOSE-3D.
(DOCX)
S3 Table. SAXS data collection parameters, analysis software employed and deposition.
(DOCX)
S4 Table. Experimental determined parameters from SAXS analysis endoHCYS at pH 7.5
and 5 mg/ml. Values represent averages and errors are the standard deviation between two
identical replicates.
(DOCX)
S5 Table. Coefficients from fits of an exponential (first order) function to the VF trajecto-
ries.
(DOCX)
S1 Data.
(GZ)
Acknowledgments
Drs. James Holton, Gregory Hura, and Elspeth Garman are acknowledged for useful
discussions.
Author Contributions
Conceptualization: Timothy R. Stachowski.
Data curation: Timothy R. Stachowski.
Formal analysis: Timothy R. Stachowski, Edward H. Snell.
Funding acquisition: Edward H. Snell.
Investigation: Timothy R. Stachowski, Mary E. Snell, Edward H. Snell.
Methodology: Timothy R. Stachowski, Mary E. Snell.
Project administration: Timothy R. Stachowski, Edward H. Snell.
Resources: Edward H. Snell.
Supervision: Edward H. Snell.
Visualization: Timothy R. Stachowski.
PLOS ONE | https://doi.org/10.1371/journal.pone.0239702 November 17, 2020
19 / 23
PLOS ONE Engineering a method for site-specific damage studies with SAXS
Writing – original draft: Timothy R. Stachowski, Edward H. Snell.
Writing – review & editing: Timothy R. Stachowski, Edward H. Snell.
References
1. Singh A, Singh H. Time-scale and nature of radiation-biological damage: Approaches to radiation pro-
tection and post-irradiation therapy. Prog Biophys Mol Biol. 1982; 39: 69–107. https://doi.org/10.1016/
0079-6107(83)90014-7 PMID: 7048420
2. Ravelli RB, McSweeney SM. The “fingerprint” that X-rays can leave on structures. Structure. 2000; 8:
315–328. https://doi.org/10.1016/s0969-2126(00)00109-x PMID: 10745008
3. Garman EF, Weik M. Radiation Damage in Macromolecular Crystallography BT—Protein Crystallogra-
phy: Methods and Protocols. In: Wlodawer A, Dauter Z, Jaskolski M, editors. New York, NY: Springer
New York; 2017. pp. 467–489. https://doi.org/10.1007/978-1-4939-7000-1_20 PMID: 28573586
4. Rodgers DW. Cryocrystallography. Structure. 1994; 2: 1135–1140. https://doi.org/10.1016/s0969-2126
(94)00116-2 PMID: 7704524
5. Barker AI, Southworth-Davies RJ, Paithankar KS, Carmichael I, Garman EF. Room-temperature scav-
engers for macromolecular crystallography: increased lifetimes and modified dose dependence of the
intensity decay. J Synchrotron Radiat. 2009; 16: 205–216. https://doi.org/10.1107/
S0909049509003343 PMID: 19240332
6. Symons MCR. Mechanism of Radiation Damage to Proteins and Dna–an EPR Perspective. Prog React
Kinet. 1999; 24: 139–164. https://doi.org/10.3184/007967499103165076
7. Southworth-Davies RJ, Medina MA, Carmichael I, Garman EF. Observation of Decreased Radiation
Damage at Higher Dose Rates in Room Temperature Protein Crystallography. Structure. 2007; 15:
1531–1541. https://doi.org/10.1016/j.str.2007.10.013 PMID: 18073104
8.
9.
10.
11.
12.
Fraser JS, Clarkson MW, Degnan SC, Erion R, Kern D, Alber T. Hidden alternative structures of proline
isomerase essential for catalysis. Nature. 2009; 462: 669–673. https://doi.org/10.1038/nature08615
PMID: 19956261
Fraser JS, van den Bedem H, Samelson AJ, Lang PT, Holton JM, Echols N, et al. Accessing protein
conformational ensembles using room-temperature X-ray crystallography. Proc Natl Acad Sci. 2011;
108: 16247–16252. https://doi.org/10.1073/pnas.1111325108 PMID: 21918110
Frauenfelder H, Hartmann H, Karplus M, Kuntz ID, Kuriyan J, Parak F, et al. Thermal expansion of a
protein. Biochemistry. 1987; 26: 254–261. https://doi.org/10.1021/bi00375a035 PMID: 3828301
Juers DH, Matthews BW. Reversible lattice repacking illustrates the temperature dependence of macro-
molecular interactions. J Mol Biol. 2001; 311: 851–862. https://doi.org/10.1006/jmbi.2001.4891 PMID:
11518535
de la Mora E, Coquelle N, Bury CS, Rosenthal M, Holton JM, Carmichael I, et al. Radiation damage and
dose limits in serial synchrotron crystallography at cryo- and room temperatures. Proc Natl Acad Sci.
2020; 117: 4142–4151. https://doi.org/10.1073/pnas.1821522117 PMID: 32047034
13. Russi S, Gonza´lez A, Kenner LR, Keedy DA, Fraser JS, van den Bedem H. Conformational variation of
proteins at room temperature is not dominated by radiation damage. J Synchrotron Radiat. 2017; 24:
73–82. https://doi.org/10.1107/S1600577516017343 PMID: 28009548
14. Gotthard G, Aumonier S, De Sanctis D, Leonard G, von Stetten D, Royant A. Specific radiation damage
is a lesser concern at room temperature. IUCrJ. 2019; 6: 665–680. https://doi.org/10.1107/
S205225251900616X PMID: 31316810
15. Brosey CA, Tainer JA. Evolving SAXS versatility: solution X-ray scattering for macromolecular architec-
ture, functional landscapes, and integrative structural biology. Curr Opin Struct Biol. 2019; 58: 197–213.
https://doi.org/10.1016/j.sbi.2019.04.004 PMID: 31204190
16. Rambo RP, Tainer JA. Accurate assessment of mass, models and resolution by small-angle scattering.
Nature. 2013; 496: 477–481. doi:10.1007/978-1-62703-691-7_18 https://doi.org/10.1038/nature12070
PMID: 23619693
17. Stachowski T, Grant TD, Snell EH. Structural consequences of transforming growth factor beta-1 acti-
vation from near-therapeutic X-ray doses. J Synchrotron Radiat. 2019; 26. https://doi.org/10.1107/
S1600577519005113 PMID: 31274418
18. Chang VT, Crispin M, Aricescu AR, Harvey DJ, Nettleship JE, Fennelly JA, et al. Glycoprotein structural
genomics: solving the glycosylation problem. Structure. 2007; 15: 267–273. https://doi.org/10.1016/j.
str.2007.01.011 PMID: 17355862
19.
Freeze HH, Kranz C. Endoglycosidase and glycoamidase release of N-linked glycans. Curr Protoc Mol
Biol. 2010;Chapter 17. https://doi.org/10.1002/0471142727.mb1713as89 PMID: 20069534
PLOS ONE | https://doi.org/10.1371/journal.pone.0239702 November 17, 2020
20 / 23
PLOS ONE Engineering a method for site-specific damage studies with SAXS
20. Rao V, Guan C, Roey P Van. Crystal structure of endo-β-N-acetylglucosaminidase H at 1.9 Å resolu-
tion: active-site geometry and substrate recognition. Structure. 1995; 3: 449–457. https://doi.org/10.
1016/s0969-2126(01)00178-2 PMID: 7663942
21. Rao V, Roey P Van, Cui T, Guan C. Mutations of endo-β-N-acetylglucosaminidase H active site resi-
dues Asp 130 and Glu 132: Activities and conformations. Protein Sci. 1999; 8: 2338–2346. https://doi.
org/10.1110/ps.8.11.2338 PMID: 10595536
22. Wang F, Wang X, Yu X, Fu L, Liu Y, Ma L, et al. High-level expression of endo-β-N-acetylglucosamini-
dase H from Streptomyces plicatus in Pichia pastoris and its application for the deglycosylation of glyco-
proteins. PLoS One. 2015; 10: e0120458–e0120458. https://doi.org/10.1371/journal.pone.0120458
PMID: 25781897
23. Matsumura M, Becktel WJ, Levitt M, Matthews BW. Stabilization of phage T4 lysozyme by engineered
disulfide bonds. Proc Natl Acad Sci. 1989; 86: 6562–6566. https://doi.org/10.1073/pnas.86.17.6562
PMID: 2671995
24. Baase WA, Liu L, Tronrud DE, Matthews BW. Lessons from the lysozyme of phage T4. Protein Sci.
2010; 19: 631–641. https://doi.org/10.1002/pro.344 PMID: 20095051
25.
Luft JR, Collins RJ, Fehrman NA, Lauricella AM, Veatch CK, DeTitta GT. A deliberate approach to
screening for initial crystallization conditions of biological macromolecules. J Struct Biol. 2003; 142:
170–179. https://doi.org/10.1016/s1047-8477(03)00048-0 PMID: 12718929
26. Battye TGG, Kontogiannis L, Johnson O, Powell HR, Leslie AGW. iMOSFLM: a new graphical interface
for diffraction-image processing with MOSFLM. Acta Crystallogr D Biol Crystallogr. 2011; 67: 271–281.
https://doi.org/10.1107/S0907444910048675 PMID: 21460445
27. Powell HR. The Rossmann Fourier autoindexing algorithm in MOSFLM. Acta Crystallogr Sect D. 1999;
55: 1690–1695. https://doi.org/10.1107/s0907444999009506 PMID: 10531518
28. Evans PR, Murshudov GN. How good are my data and what is the resolution? Acta Crystallogr D Biol
Crystallogr. 2013; 69: 1204–1214. https://doi.org/10.1107/S0907444913000061 PMID: 23793146
29.
30.
Long F, Vagin AA, Young P, Murshudov GN. BALBES: a molecular-replacement pipeline. Acta Crystal-
logr D Biol Crystallogr. 2008; 64: 125–132. https://doi.org/10.1107/S0907444907050172 PMID:
18094476
Terwilliger TC, Grosse-Kunstleve RW, Afonine P V, Moriarty NW, Zwart PH, Hung LW, et al. Iterative
model building, structure refinement and density modification with the PHENIX AutoBuild wizard. Acta
Crystallogr D Biol Crystallogr. 2008; 64: 61–69. https://doi.org/10.1107/S090744490705024X PMID:
18094468
31. Emsley P, Cowtan K. Coot: model-building tools for molecular graphics. Acta Crystallogr D Biol Crystal-
logr. 2004; 60: 2126–2132. https://doi.org/10.1107/S0907444904019158 PMID: 15572765
32. Adams PD, Afonine P V, Bunko´ czi G, Chen VB, Davis IW, Echols N, et al. PHENIX: a comprehensive
Python-based system for macromolecular structure solution. Acta Crystallogr D Biol Crystallogr. 2010;
66: 213–221. https://doi.org/10.1107/S0907444909052925 PMID: 20124702
33. Chen VB, Arendall 3rd WB, Headd JJ, Keedy DA, Immormino RM, Kapral GJ, et al. MolProbity: all-
atom structure validation for macromolecular crystallography. Acta Crystallogr D Biol Crystallogr. 2010;
66: 12–21. https://doi.org/10.1107/S0907444909042073 PMID: 20057044
34. Pettersen EF, Goddard TD, Huang CC, Couch GS, Greenblatt DM, Meng EC, et al. UCSF Chimera—A
visualization system for exploratory research and analysis. J Comput Chem. 2004; 25: 1605–1612.
https://doi.org/10.1002/jcc.20084 PMID: 15264254
35. Dyer KN, Hammel M, Rambo RP, Tsutakawa SE, Rodic I, Classen S, et al. High-throughput SAXS for
the characterization of biomolecules in solution: a practical approach. Methods Mol Biol. 2014; 1091:
245–258. https://doi.org/10.1007/978-1-62703-691-7_18 PMID: 24203338
36. Clark GNI, Hura GL, Teixeira J, Soper AK, Head-Gordon T. Small-angle scattering and the structure of
ambient liquid water. Proc Natl Acad Sci. 2010; 107: 14003–14007. https://doi.org/10.1073/pnas.
1006599107 PMID: 20647388
37. Brooks-Bartlett JC, Batters RA, Bury CS, Lowe ED, Ginn HM, Round A, et al. Development of tools to
automate quantitative analysis of radiation damage in SAXS experiments. J Synchrotron Radiat. 2017;
24: 63–72. https://doi.org/10.1107/S1600577516015083 PMID: 28009547
38. Orthaber D, Bergmann A, Glatter O. SAXS experiments on absolute scale with Kratky systems using
water as a secondary standard. J Appl Crystallogr. 2000; 33: 218–225. https://doi.org/10.1107/
S0021889899015216
39. Hajizadeh NR, Franke D, Jeffries CM, Svergun DI. Consensus Bayesian assessment of protein molecu-
lar mass from solution X-ray scattering data. Sci Rep. 2018; 8: 7204. https://doi.org/10.1038/s41598-
018-25355-2 PMID: 29739979
PLOS ONE | https://doi.org/10.1371/journal.pone.0239702 November 17, 2020
21 / 23
PLOS ONE Engineering a method for site-specific damage studies with SAXS
40. Armstrong DA, Wilkening VG. Effects of pH in the γ-radiolysis of aqueous solutions of cysteine and
methyl mercaptan. Can J Chem. 1964; 42: 2631–2635. https://doi.org/10.1139/v64-389
41.
Franke D, Petoukhov M V, Konarev P V, Panjkovich A, Tuukkanen A, Mertens HDT, et al. ATSAS 2.8:
a comprehensive data analysis suite for small-angle scattering from macromolecular solutions. J Appl
Crystallogr. 2017; 50: 1212–1225. https://doi.org/10.1107/S1600576717007786 PMID: 28808438
42. Svergun D, Barberato C, Koch MHJ. CRYSOL a Program to Evaluate X-ray Solution Scattering of Bio-
logical Macromolecules from Atomic Coordinates. J Appl Crystallogr. 1995; 28: 768–773. https://doi.
org/10.1107/S0021889895007047
43. Konarev P V, Volkov V, Sokolova A V, Koch MHJ, Svergun DI. PRIMUS: a Windows PC-based system
for small-angle scattering data analysis. J Appl Crystallogr. 2003; 36: 1277–1282. https://doi.org/10.
1107/S0021889803012779
44. Grant TD. Ab initio electron density determination directly from solution scattering data. Nat Methods.
2018; 15: 191–193. https://doi.org/10.1038/nmeth.4581 PMID: 29377013
45.
46.
Tria G, Mertens HDT, Kachala M, Svergun DI. Advanced ensemble modelling of flexible macromole-
cules using X-ray solution scattering. IUCrJ. 2015; 2: 207–217. https://doi.org/10.1107/
S205225251500202X PMID: 25866658
Tian X, Vestergaard B, Thorolfsson M, Yang Z, Rasmussen HB, Langkilde AE. In-depth analysis of sub-
class-specific conformational preferences of IgG antibodies. IUCrJ. 2015; 2: 9–18. https://doi.org/10.
1107/S205225251402209X PMID: 25610623
47. Petoukhov M V, Franke D, Shkumatov A V, Tria G, Kikhney AG, Gajda M, et al. New developments in
the ATSAS program package for small-angle scattering data analysis. J Appl Crystallogr. 2012; 45:
342–350. https://doi.org/10.1107/S0021889812007662 PMID: 25484842
48. Bernado´ P, Mylonas E, Petoukhov M V, Blackledge M, Svergun DI. Structural Characterization of Flexi-
ble Proteins Using Small-Angle X-ray Scattering. J Am Chem Soc. 2007; 129: 5656–5664. https://doi.
org/10.1021/ja069124n PMID: 17411046
49. Krissinel E, Henrick K. Inference of Macromolecular Assemblies from Crystalline State. J Mol Biol.
2007; 372: 774–797. https://doi.org/10.1016/j.jmb.2007.05.022 PMID: 17681537
50.
Trewhella J, Duff AP, Durand D, Gabel F, Guss JM, Hendrickson WA, et al. 2017 publication guidelines
for structural modelling of small-angle scattering data from biomolecules in solution: an update. Acta
Crystallogr D Biol Crystallogr. 2017; 73: 710–728. https://doi.org/10.1107/S2059798317011597 PMID:
28876235
51. Hopkins JB, Thorne RE. Quantifying radiation damage in biomolecular small-angle X-ray scattering. J
Appl Crystallogr. 2016; 49: 880–890. https://doi.org/10.1107/S1600576716005136 PMID: 27275138
52. Sauter A, Zhang F, Szekely NK, Pipich V, Sztucki M, Schreiber F. Structural Evolution of Metastable
Protein Aggregates in the Presence of Trivalent Salt Studied by (V)SANS and SAXS. J Phys Chem B.
2016; 120: 5564–5571. https://doi.org/10.1021/acs.jpcb.6b03559 PMID: 27285548
53. Durand D, Vivès C, Cannella D, Pe´ rez J, Pebay-Peyroula E, Vachette P, et al. NADPH oxidase activator
p67phox behaves in solution as a multidomain protein with semi-flexible linkers. J Struct Biol. 2010;
169: 45–53. https://doi.org/10.1016/j.jsb.2009.08.009 PMID: 19723583
54. Putnam CD. Guinier peak analysis for visual and automated inspection of small-angle X-ray scattering
data. J Appl Crystallogr. 2016; 49: 1412–1419. https://doi.org/10.1107/S1600576716010906 PMID:
27738411
55. Petoukhov M V, Konarev P V, Kikhney AG, Svergun DI. ATSAS 2.1 towards automated and web-sup-
ported small-angle scattering data analysis. J Appl Crystallogr. 2007; 40: s223–s228. https://doi.org/10.
1107/S0021889807002853
56. Okamoto S, Sakurai S. Singular value decomposition analysis of time-resolved small-angle X-ray scat-
tering during morphological transition in a block copolymer. J Appl Crystallogr. 2003; 36: 982–985.
https://doi.org/10.1107/S0021889803008483
57. Kuwamoto S, Akiyama S, Fujisawa T. Radiation damage to a protein solution, detected by synchrotron
X-ray small-angle scattering: dose-related considerations and suppression by cryoprotectants. J Syn-
chrotron Radiat. 2004; 11: 462–468. https://doi.org/10.1107/S0909049504019272 PMID: 15496733
58. Sutton KA, Black PJ, Mercer KR, Garman EF, Owen RL, Snell EH, et al. Insights into the mechanism of
X-ray-induced disulfide-bond cleavage in lysozyme crystals based on EPR, optical absorption and X-
ray diffraction studies. Acta Crystallogr D Biol Crystallogr. 2013; 69: 2381–2394. https://doi.org/10.
1107/S0907444913022117 PMID: 24311579
59. Weik M, Kryger G, Schreurs AMM, Bouma B, Silman I, Sussman JL, et al. Solvent behaviour in flash-
cooled protein crystals at cryogenic temperatures. Acta Crystallogr D Biol Crystallogr. 2001; 57: 566–
573. https://doi.org/10.1107/s0907444901001196 PMID: 11264586
PLOS ONE | https://doi.org/10.1371/journal.pone.0239702 November 17, 2020
22 / 23
PLOS ONE Engineering a method for site-specific damage studies with SAXS
60. Xu G, Chance MR. Hydroxyl Radical-Mediated Modification of Proteins as Probes for Structural Proteo-
mics. Chem Rev. 2007; 107: 3514–3543. https://doi.org/10.1021/cr0682047 PMID: 17683160
61. Weik M, Ravelli RBG, Kryger G, McSweeney S, Raves ML, Harel M, et al. Specific chemical and struc-
tural damage to proteins produced by synchrotron radiation. Proc Natl Acad Sci. 2000; 97: 623–628.
https://doi.org/10.1073/pnas.97.2.623 PMID: 10639129
62. Gerstel M, Deane CM, Garman EF. Identifying and quantifying radiation damage at the atomic level. J
Synchrotron Radiat. 2015; 22: 201–212. https://doi.org/10.1107/S1600577515002131 PMID:
25723922
63. Putnam CD, Hammel M, Hura GL, Tainer JA. X-ray solution scattering (SAXS) combined with crystal-
lography and computation: defining accurate macromolecular structures, conformations and assem-
blies in solution. Q Rev Biophys. 2007; 40: 191–285. https://doi.org/10.1017/S0033583507004635
PMID: 18078545
64. Koch MHJ, Vachette P, Svergun DI. Small-angle scattering: a view on the properties, structures and
structural changes of biological macromolecules in solution. Q Rev Biophys. 2003; 36: 147–227. https://
doi.org/10.1017/s0033583503003871 PMID: 14686102
65.
Fischetti RF, Rodi DJ, Mirza A, Irving TC, Kondrashkina E, Makowski L. High-resolution wide-angle X-
ray scattering of protein solutions: effect of beam dose on protein integrity. J Synchrotron Radiat. 2003;
10: 398–404. https://doi.org/10.1107/s0909049503016583 PMID: 12944630
66. Cremers CM, Jakob U. Oxidant sensing by reversible disulfide bond formation. J Biol Chem. 2013; 288:
26489–26496. https://doi.org/10.1074/jbc.R113.462929 PMID: 23861395
67. Koenig S, Svergun D, Koch MHJ, Huebner G, Schellenberger A. Synchrotron radiation solution x-ray
scattering study of the pH-dependence of the quarternary structure of yeast pyruvate decarboxylase.
Biochemistry. 1992; 31: 8726–8731. https://doi.org/10.1021/bi00152a007 PMID: 1390658
68.
Liu H, May K. Disulfide bond structures of IgG molecules: structural variations, chemical modifications
and possible impacts to stability and biological function. MAbs. 2012; 4: 17–23. https://doi.org/10.4161/
mabs.4.1.18347 PMID: 22327427
69. Remesh SG, Armstrong AA, Mahan AD, Luo J, Hammel M. Conformational Plasticity of the Immuno-
globulin Fc Domain in Solution. Structure. 2018; 26: 1007–1014.e2. https://doi.org/10.1016/j.str.2018.
03.017 PMID: 29731233
70.
Lerner AM, Yumerefendi H, Goudy OJ, Strahl BD, Kuhlman B. Engineering Improved Photoswitches
for the Control of Nucleocytoplasmic Distribution. ACS Synth Biol. 2018; 7: 2898–2907. https://doi.org/
10.1021/acssynbio.8b00368 PMID: 30441907
71. Kapp GT, Liu S, Stein A, Wong DT, Reme´nyi A, Yeh BJ, et al. Control of protein signaling using a com-
putationally designed GTPase/GEF orthogonal pair. Proc Natl Acad Sci. 2012; 109: 5277–5282. https://
doi.org/10.1073/pnas.1114487109 PMID: 22403064
72. Yumerefendi H, Lerner AM, Zimmerman SP, Hahn K, Bear JE, Strahl BD, et al. Light-induced nuclear
export reveals rapid dynamics of epigenetic modifications. Nat Chem Biol. 2016; 12: 399–401. https://
doi.org/10.1038/nchembio.2068 PMID: 27089030
73. Svergun D. I. (1999). Biophys. J. 76, 2879–2886. https://doi.org/10.1016/S0006-3495(99)77443-6
PMID: 10354416
PLOS ONE | https://doi.org/10.1371/journal.pone.0239702 November 17, 2020
23 / 23
PLOS ONE |
10.1371_journal.pone.0241981 | RESEARCH ARTICLE
Better coverage, better outcomes? Mapping
mobile network data to official statistics using
satellite imagery and radio propagation
modelling
Till KoebeID*
Department of Economics, Freie Universita¨t, Berlin, Germany
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
* till.koebe@fu-berlin.de
Abstract
OPEN ACCESS
Citation: Koebe T (2020) Better coverage, better
outcomes? Mapping mobile network data to official
statistics using satellite imagery and radio
propagation modelling. PLoS ONE 15(11):
e0241981. https://doi.org/10.1371/journal.
pone.0241981
Editor: Jacinto Estima, Instituto de Engenharia de
Sistemas e Computadores Investigacao e
Desenvolvimento em Lisboa, PORTUGAL
Received: February 25, 2020
Accepted: October 23, 2020
Published: November 9, 2020
Peer Review History: PLOS recognizes the
benefits of transparency in the peer review
process; therefore, we enable the publication of
all of the content of peer review and author
responses alongside final, published articles. The
editorial history of this article is available here:
https://doi.org/10.1371/journal.pone.0241981
Copyright: © 2020 Till Koebe. This is an open
access article distributed under the terms of the
Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: The mobile phone
data at the antenna- and commune-level
aggregated to the year 2013 including noisy
Mobile sensing data has become a popular data source for geo-spatial analysis, however,
mapping it accurately to other sources of information such as statistical data remains a chal-
lenge. Popular mapping approaches such as point allocation or voronoi tessellation provide
only crude approximations of the mobile network coverage as they do not consider holes,
overlaps and within-cell heterogeneity. More elaborate mapping schemes often require
additional proprietary data operators are highly reluctant to share. In this paper, I use human
settlement information extracted from publicly available satellite imagery in combination
with stochastic radio propagation modelling techniques to account for that. I show in a simu-
lation study and a real-world application on unemployment estimates in Senegal that better
coverage approximations do not necessarily lead to better outcome predictions.
Introduction
Mobile phone metadata has become a popular data source to complement official statistics.
When an individual makes a call, sends a message or uses the mobile internet, meta informa-
tion about this interaction, such as the time stamp and the location, are stored in a database of
the mobile network operator (MNO). Researchers exploit those spatio-temporal references for
geo-located analysis. One string of research in this field investigates the question whether a
certain characteristic such as poverty, literacy or food insecurity is reflected in mobile phone
behaviour. Matching this behaviour accurately to a ‘groundtruth’—often statistical data from
surveys or censuses provided for statistical areas—however, poses a major challenge as the two
data sources lack a common reference. In the case of call detail records (CDRs), the geographic
reference is provided by the antenna location, often stored as a point coordinate of the physical
location of the corresponding base transmitter station (BTS). Due to its simplicity, some scien-
tific literature treat antennas as point coordinates [1]. However, the interactions captured by
the antenna do not happen entirely at this exact coordinate, but within the coverage area of the
antenna—the cell. While an antenna may be located in one statistical area, most of the cell may
lie within the neighboring area. The state-of-the-art attempt to address this is to use spatial
weights based on the overlapping area size of statistical areas and cells approximated via voro-
noi tessellation [2, 3]. This approach has three major drawbacks: First, voronoi tessellation
PLOS ONE | https://doi.org/10.1371/journal.pone.0241981 November 9, 2020
1 / 28
PLOS ONE antenna locations as well as instructions for
replicating the study results have been added as
part of the Supporting information. In order to
access record-level mobile phone data and exact
antenna locations, one would need to contact
Sonatel directly and present the research project
that would require the data (contact: Mr El Hadji
Birahim Gueye, Direction des Systèmes
d’information Sonatel, ebgueye@orange-sonatel.
com or post mail: Orange-Sonatel, 46 Boulevard de
la Re´publique, BP 69 Dakar, Senegal). GUF data
cannot be shared publicly because third-party
access conditions apply (for scientific, non-
commercial use). However, it is available for
research purposes under a data user agreement.
For data access, please contact the German
Aerospace Agency under guf@dlr.de (https://www.
dlr.de/eoc/en/PortalData/60/Resources/
dokumente/guf/DLR-GUF_LicenseAgreement-and-
OrderForm.pdf). Census data used in the study
cannot be shared publicly because third-party
access conditions apply. However, it is available for
research purposes under a data user agreement.
For data access, please visit the microdata
catalogue of the statistical office in Senegal (http://
anads.ansd.sn/index.php/catalog/51) or send the
inquiry to statsenegal@ansd.sn. All code required
for replicating the findings of this study is fully
available in the Supporting information of this
submission (S1 and S2 Files) and under https://
github.com/tilluz/geomatching_open.
Funding: The author received no specific funding
for this work.
Competing interests: The author have declared
that no competing interests exist.
Mapping mobile network data using satellite imagery and radio propagation modelling
perfectly divides the space around BTS locations depending on the distance to the surrounding
BTS. This represents a naïve approximation of the true coverage areas as it does not take over-
laps, areas without coverage and additional network complexities (multiple antennas per site/
BTS, directionality of antennas, varying frequency bands etc.) into account [4]. For example,
roughly 90 million people in Africa in 2019 were still not connected to any mobile network
hinting at major holes in the coverage [5]. Second, even though the concept of ‘home-locating’
subscribers to specific BTS offers a network-based alternative to the statistical concept of ‘usual
place of residence’, it is not reflected within cells. As the weights are based on area sizes, the
voronoi tessellation implicitly assumes that individuals/households are homogeneously dis-
tributed within cells, which in most cases does not hold true. For example, a lake would receive
the same importance in the creation of area-level mobile phone metadata aggregates as an
equally sized built-up area. Third, as mobile stations (MS, generally defined as a combination
of device and SIM card) and antennas communicate via modulated radio signals whose propa-
gation paths depend on a range of factors such as the weather, coverage areas are stochastic by
nature. More elaborate approaches to model coverage ranges of mobile networks exist [4, 6],
especially in the field of radio propagation modelling native to electrical engineering, however,
they often require detailed information on the area’s topology, a number of technical details
concerning the network infrastructure and additional information from passive monitoring
systems, which mobile network operators are generally highly reluctant to share and in the lat-
ter case often not capable to collect.
Contributions
Acknowledging this, I divide my methodological contribution in this paper in two parts: First,
I propose the use of settlement information extracted from publicly available satellite imagery
to account for within-cell heterogeneity within the mobile network when linking statistical
data with mobile phone metadata. Building on this, the second part of the methodology takes
advantage of scenarios where additional technical specifications are available in order to
address the issues for holes, overlaps and non-linearities within the mobile network using
propagation-based modelling. My main contributions are as follows:
1. The idea of using settlements retrieved from publicly available satellite imagery as a common
reference for statistical units such as households and ‘home-located’ MS in order to calculate
weights for mapping mobile phone metadata and statistical data based on settlement counts in
scenarios where MS counts are not available. This way, within-cell heterogeneity is addressed.
2. A propagation-based approach to account for overlaps, holes and non-linearities in coverage
service provision—in case additional information on the network infrastructure are available.
3. A large-scale simulation study on a synthetic population grid to systematically compare the
accuracy of different mapping approaches and their effects on predictive performance.
4. A real-world application that demonstrates the impact of the mapping choice on outcomes
in later analysis.
Datasets
In the application, I revisit the simulation study of Schmid et al. [1] published in 2017 in the
Journal of the Royal Statistical Society Series A on fine-granular unemployment estimates from
mobile phone metadata in Senegal in order to investigate the effects of different mapping
schemes on the unemployment outcomes. Therefore, I re-run the original simulation with the
difference that I implement multiple mapping schemes to derive area-level covariates from
PLOS ONE | https://doi.org/10.1371/journal.pone.0241981 November 9, 2020
2 / 28
PLOS ONE Mapping mobile network data using satellite imagery and radio propagation modelling
Fig 1. Settlements in Senegal provided as b/w image by the GUF project. Lower resolution built-settlements extents
data reprinted from [10] under a CC BY license, with permission from WorldPop, original copyright 2018, are used in
this figure for illustrative purposes.
https://doi.org/10.1371/journal.pone.0241981.g001
CDRs. Specifically, I use behavioural indicators and SIM card counts extracted from CDRs
provided by the major Senegalese MNO Sonatel in the context of the D4D 2014 challenge for
the whole year of 2013 and aggregated on the level of BTS, for which the exact geo-coordinates
are also provided [7]. The behavioural indicators are generated using the popular open-source
Python module Bandicoot [8]. Further, I use population counts from the full 2013 general pop-
ulation and housing census (RGPHAE 2013) available for the NUTS 4-level of Senegal—the
communes—on the website of ANSD, the National Statistical Office of Senegal. Commune-
level unemployment information are generated from a 10% sample of RGPHAE 2013. Unem-
ployment information in RGPHAE 2013 are self-reported.
Geographic information on the administrative boundaries are available for communes and
above. The settlement-based weights I present in this paper use data on human settlement
areas in Senegal extracted from the Global Urban Footprint (GUF) project [9] of the German
Aerospace Center (DLR) at a resolution of 0.4 arc seconds, which is approximately 12m x
12m. The GUF project used 180,000 TerraSAR-X and TanDEM-X images collected during the
period of 2011—2012 (with some data from 2013/14 to fill gaps) to create black and white
abstractions where white pixels represent human settlements with a true positive rate (accu-
racy to correctly detect human settlements) of 85% on average, with 68% at lowest and 98% at
heighest. GUF data for Senegal is provided as a single black and white.tif-file with a resolution
of 55568 x 39459 pixels (see Fig 1). All datasets used in this study are available for research pur-
poses under the conditions of the respective data use agreements.
Related work
Increasing processing capabilities have propelled the use of satellite imagery in official statis-
tics. The UN [11] recommends using satellite imagery to prioritize and check geospatial
PLOS ONE | https://doi.org/10.1371/journal.pone.0241981 November 9, 2020
3 / 28
PLOS ONE Mapping mobile network data using satellite imagery and radio propagation modelling
processes such as the delineation of enumeration areas during census preparation. It further
supports the construction of population grids as a common spatial reference system as pro-
posed by [12, 13]. Various studies have used remote sensing, sometimes in combination with
mobile phone metadata, to estimate key statistical indicators such as economic growth [14–
16], population density [17–20] or poverty [2, 21, 22]. Work in that field most closely related
to this study uses settlement information extracted from satellite imagery in combination with
radio propagation models for application in cost-benefit analysis concerning additional infra-
structure investments [23]. While [23] also uses population counts from official statistics to
estimate the latent demand for mobile services, the author neither investigates the effects of
different coverage mapping techniques on the results nor does he use mobile phone metadata
for statistical purposes.
In addition, the last decade has seen an impressive amount of research on proposing the
use of mobile phone metadata for official statistics foremost in the hope to overcome the limit-
ing relationship of sample size and data collection costs. [24] provides an excellent overview
on the use of mobile phone metadata that also covers its application for statistical purposes.
Use cases to produce more frequent, more granular and/or more timely data on a wide range
of statistical topics have been identified. For example, [4, 25–28] use mobile phone metadata to
investigate population dynamics for more frequent population and tourism statistics. [29, 30]
apply the question on the whereabouts of a population to the post disaster setting. Mobility
aspects such as commuting and travelling routines have been looked at in more detail by [31–
36]. By exploiting both mobility and (social) network characteristics of mobile phone meta-
data, [37–42] and [43, 44] use mobile phone metadata to model disease spreading and integra-
tion, respectively. Mobile usage patterns have been explored to provide fine granular insights
on socio-demographic indicators such as multi-dimensional poverty [2, 3], literacy [1, 45] and
economic vulnerability [46, 47]. While most of these studies have mapped mobile phone meta-
data and groundtruth data using point-to-polygon allocation or voronoi tessellation, very few
studies have applied more elaborate approximation schemes. [4] propose a methodology based
on maximum likelihood estimation that uses cell footprints provided by one or multiple
MNOs in combination with location data from passive monitoring systems to acquire more
accurate measures on the density of MS. The authors run a simulation study on a 100x100m
synthetic population grid to compare the proposed methodology against voronoi-based cover-
age maps. However, the methodology requires very detailed information from the involved
MNOs, e.g. on the cell footprints and the signalling data that may prove difficult to acquire in
practice (see Section Mobile phone metadata). Further, while the authors rightly assume a
multinomial distribution of the MS counts, finding appropriate distributions for the wide
range of behavioural covariates appears less trivial. In order to simplify and improve the cover-
age mapping process, members of the European Statistical System as part of the ESSnet Big
Data project are currently developing mobloc [48]—an R package that implements the free
space path loss propagation model using technical specifications of antennas as input parame-
ters. However, neither [4] nor [48] systematically evaluate different coverage mapping tech-
niques on statistical modelling approaches using real-world data.
Background
Mobile phone metadata
Mobile networks not only transport data for communication purposes, they also generate data
for reasons such as network auditing, billing, maintenance and service provision. Some of this
meta information is created in interaction with user equipment such as MS. There are four
main caveats of using mobile phone metadata for population statistics in general. All of them
PLOS ONE | https://doi.org/10.1371/journal.pone.0241981 November 9, 2020
4 / 28
PLOS ONE Mapping mobile network data using satellite imagery and radio propagation modelling
have in common that they are active areas of current research. First, the customer base of an
MNO constitutes a non-representative population sample with unknown sampling design. The
consequences are varying sampling rates, i.e. locally changing market shares and parts of the
population being structurally excluded from the sample such as children, elderly and the very
poor. Second, the unit of observation—i.e. the MS, device, the SIM card and/or the subscriber—
does not perfectly match the unit of interest, which is the individual or household, as phone
sharing schemes or multi-SIM uses illustrate. Common approaches to account for these two
caveats are calibration and/or reconstructing the sampling design empirically. Third, mobile
phone metadata lacks the statistical concept of usual residence—a concept frequently used in
official statistics to determine the geo-location of an individual/household defined as the place
where an individual has lived or intends to live for a period of at least 6 or 12 months [49]. Dif-
ferent approaches to approximate the home location of an MS exists (e.g. night-time home loca-
tion defined as the most frequently used cell by an MS between 7pm and 7am during a certain
time window), however, the definitions do not map perfectly introducing uncertainty in further
analysis [50]. Fourth, coverage areas cannot be pinpointed as radio propagation is dynamic and
stochastic by nature. Propagation models of various complexity exist to provide approximations
as coverage ranges can generally vary from couple of hundred meters to over 40km.
Most scientific studies in the context of international development and official statistics use
CDRs—logs of interactions such as calls, text messages or internet use containing attributes of
the MS, the network and the connection—as a basis for further analysis. The advantages of
CDRs compared to other mobile phone metadata such as Visitor Location Registers (VLRs) or
other signalling data are threefold: First, they provide fine-grained geographical resolution
through cell-level identifiers. Second, they provide information both on the mobility and the
(social) network of the MS. Third, CDRs are fairly easy to access and to use in analysis as the
storage of essential attributes adheres to global standards such as 3gpp 32.295. However, in addi-
tion to the aforementioned general caveats of mobile phone metadata there are important cave-
ats specific to CDRs: Social network information extracted from CDRs are increasingly
incomplete due to a shift towards app-based communication (e.g. Whatsapp and Facebook mes-
senger). Mobility patterns are fragmented as locations are logged only during active MS use—
again a case of non-random sampling. Some MNOs are able to extract more detailed informa-
tion on the location of an MS and its app usage e.g. for geo-fencing purposes or app-based pric-
ing schemes through trilateration of signalling data and deep packet inspection, respectively.
This, however, requires specific hardware equipment and software capabilities, which not every
MNO has. Consequently, these type of information are rarely available to researchers.
Radio propagation modelling
Radio propagation modelling has been subject to research for decades. Coverage mappings in
mobile networks are generally used for network planning purposes [23, 51]. Looking at Phil-
lips et al. [6] is highly recommended as they provide an excellent overview on coverage map-
ping methods. In general, radio propagation modelling techniques in mobile networks largely
focus on estimating the path loss Lp a radio signal incurs en route between a transmitter tx and
a receiver rx. Together with the output power of the transmitter Ptx, the gains through directiv-
ity and efficiency of the involved antennas Gtx and Grx and their respective technically-
incurred losses Ltx and Lrx, it defines the link budget—the received power Prx usually expressed
logarithmically in decibel per milliwatt (dBm).
Prx ¼ Ptx þ Gtx þ Grx (cid:0) Ltx (cid:0) Lrx (cid:0) Lp
PLOS ONE | https://doi.org/10.1371/journal.pone.0241981 November 9, 2020
ð1Þ
5 / 28
PLOS ONE Mapping mobile network data using satellite imagery and radio propagation modelling
Since all RHS parameters except Lp are either known in advance due to the choice of the
technical equipment (i.e. Gtx and Ltx) or hardly observable (i.e. Grx and Lrx), I assume Gtx + Grx
− Ltx − Lrx = 0 in the following, leading to a simplified link budget defined as:
Prx ¼ Ptx (cid:0) Lp
ð2Þ
Intuitively, Eq 2 thus states that the signal strength observed on a MS solely depends on the
output power of the connected antenna and the loss in signal strength that occurs along the
way between antenna and MS. Given the abundance of available models, I follow the guidance
of the European Conference of Postal and Telecommunications Administrations (CEPT) on
radio propagation simulation for mobile services and opt for the widely popular extended
HATA model [52], named after Masaharu Hata, the author of the 1980 landmark study on the
“Empirical Formula for Propagation Loss in Land Mobile Radio Services” [53]. It is derived
from the COST-231 HATA model [54], which in turn builds on the original HATA [53] and
Okumura model [55]. They all have in common that they are empirical models to estimate the
median path loss between a transmitter and a receiver based on real-world measurements. The
HATA model extends the Okumura model by distinguishing between urban, suburban and
rural settings, thus accounting for different levels of mean attenuation due to obstacles and
changes in terrain. The COST-231 HATA model increases the frequency range of the original
HATA model. The extended HATA model is applicable for settings with frequencies f between
30-3000 MHz, distances d between 0-100km, transmitter heights htx between 30-200m and
receiver heights hrx between 1-10m. The general form of the extended HATA model LEH
p con-
sists of a loss function L for the median path loss and a path loss variation term V drawn from
a log-normal distribution that accounts for the stochastic nature of radio propagation Since
model parameters vary depending on the distance, the expected environment env (indoor/out-
door and rural/suburban/urban) and the frequency, the full extended HATA model is not
spelled out in this paper, but can be accessed here: https://ecocfl.cept.org/display/SH/A17.3.1
+Outdoor-outdoor+propagation.
p ðf ; d; htx; hrx; envÞ ¼ Lðf ; d; htx; hrx; envÞ þ Vðm; s; dÞ
LEH
ð3Þ
As an example, I provide the path loss function of the extended HATA model LEH
p
tances above 0.1km outdoor in rural areas for frequencies between 150 and 1500 MHz:
for dis-
p ¼ 69:6þ
LEH
46:09 � log 10f (cid:0)
13:82 � log 10htxþ
ð44:9 (cid:0) 6:55 � log 10htxÞ � log 10d(cid:0)
ð1:1 � log 10f (cid:0) 0:7Þ � hrx(cid:0)
20 � log 10ðhrx=10Þ(cid:0)
20 � log 10ðhtx=30Þ(cid:0)
4:78 � ð log 10f Þ2(cid:0)
40:14þ
Vð12; 12Þ
ð4Þ
PLOS ONE | https://doi.org/10.1371/journal.pone.0241981 November 9, 2020
6 / 28
PLOS ONE Mapping mobile network data using satellite imagery and radio propagation modelling
So, for example, an MS 1m above the ground at a line-of-sight distance of 3km in a rural
area to an omnidirectional antenna that is 30m above the ground transmitting at the 900 MHz
frequency band would experience a path loss of LEH
p � 118dBm. Assuming a GSM macro-cell
with an output power Ptx = 43 dBm using Eq 2 yields a budget for that link, also known as
received signal strength (RSS), of Prx ¼ Ptx (cid:0) LEH
p � (cid:0) 75dBm. As a rule of thumb, signals with
RSS values above −80 dBm are considered excellent, RSS values below −110 dBm point to very
poor signals.
Methodology
Usually, statistical data on individuals or households are geo-located to statistical areas via
their respective places of residence. Further, unit-level data is aggregated to area-level aggre-
gates using some form of weighting factor such as survey weights. For example, the poverty
rate of a region can either be calculated as the share of units classified as poor among the inter-
viewed residents of the region multiplied by their sampling weight or via sub-regional poverty
rates weighted with the respective sub-regional population counts. However, neither the places
of residences nor the weights are generally available on the cell-level of a mobile network (as
an equivalent to the sub-region). Hence, they need to be estimated.
In mobile phone metadata analysis, the place of residence of an individual/household is
usually approximated with the night-time home location of an MS recorded at the cell-level.
To derive survey weight proxies, for example, point-to-polygon allocation assumes equal
weights for all cells point-located within a statistical area. Voronoi tessellation uses the area
size of the intersection of voronoi tile and statistical area as weighting factor, i.e. 1 km2 always
conveys the same importance in aggregation, no matter whether it is 1 km2 of sparsely-inhab-
ited desert or 1 km2 of a densely-populated city.
In most cases, the place of residence of an individual/household (thus is approximation
alike) is linked to some form of settlement. However, neither the statistical area nor the cover-
age area of a cell account for that fact. Consequently, the underlying idea behind the proposed
methodology is to use human settlement information extracted from publicly available satellite
imagery as common geographic reference level for both statistical units such as households
and home-located MS. This allows to a) construct weights based on settlement counts and b)
refine weights in cases where MS counts, often regarded as highly sensitive information by the
MNO, are available. Further, in combination with technical information on the antenna, it
allows for an efficient coverage estimation to address the issues of holes and overlaps in a
mobile network.
In the following, settlements are denoted as i, BTS as j, statistical areas as t, the number of
home-located MS as d, the population count as p, the number of settlements as n and metadata
covariates as R. To illustrate the value added of the proposed methodologies, Fig 2a and
Table 1 showcase a typical setup faced when one seeks to augment official statistics with mobile
phone metadata: statistical indicators are provided for statistical areas A, B and C. Mobile
phone metadata is provided as BTS-level aggregates with the corresponding point locations 1
and 2. To account for that, I treat each cell site that may host multiple antennas as single omni-
directional antenna, calling it BTS subsequently. This constitutes a simplification of real
mobile networks where usually multiple directional antennas serving on various frequency
bands are co-located at the same site that does not necessarily have to be an actual (cell) tower.
Although accounting for directionality of antennas as done by e.g. [4] is likely to affect the
overall outcome of later analysis by increasing the number of network tiles available for map-
ping, the challenges for allocating them correctly (holes, non-linearities, overlaps, within-cell
PLOS ONE | https://doi.org/10.1371/journal.pone.0241981 November 9, 2020
7 / 28
PLOS ONE Mapping mobile network data using satellite imagery and radio propagation modelling
Fig 2. Popular and proposed mapping schemes. Three statistical areas (A-C), two BTS (1-2) and numerous dots
representing built-up areas illustrate how different mapping schemes affect the allocation of BTS-level data to statistical
data.
https://doi.org/10.1371/journal.pone.0241981.g002
heterogeneity) remain. Consequently, it is expected that results from this study also apply to a
setup based on directional antennas, thereby justifying the simplifying assumption. Further
details on Fig 2b–2f are provided in the following subsections.
Point-to-polygon allocation
For purposes such as model fitting one approach to combine statistical data and mobile phone
metadata is to aggregate metadata covariates onto the same geographical level, e.g. statistical
areas. To do so, the point-to-polygon approach (p2p) treats BTS point locations as such and
allocates BTS-level metadata covariates using a binary weighting scheme (see Fig 2b and Eq 5).
wp2p
j;t ≔
(
1 if j � t
0 otherwise
ð5Þ
Consequently, all network traffic handled by a BTS is attributed to one statistical area exclu-
sively, no matter whether it was generated by a home-located MS actually ‘residing’ in this area
or not. In the toy example, but also in the real-world application presented in Section Applica-
tion this leads to a situation where no metadata covariates are available for certain area, e.g.
area C—with negative effects on the final sample size in model fitting.
Voronoi tessellation
In contrast, voronoi tessellation (denoted by superscript v) divides the total space of interest
into perfectly disjunct tiles along the equidistant lines between points, in this case the BTS
Table 1. Example of statistical data and mobile phone metadata.
area_id
poverty_rate
1
2
0.23
0.11
bts_id
6453
8348
# of calls
34050
1023
lon
43.2344
50.0988
lat
23.2342
18.84217
https://doi.org/10.1371/journal.pone.0241981.t001
PLOS ONE | https://doi.org/10.1371/journal.pone.0241981 November 9, 2020
8 / 28
PLOS ONE Mapping mobile network data using satellite imagery and radio propagation modelling
point locations (see Fig 2c). The current state-of-the-art procedure is to intersect these tiles—
representing approximated coverage areas of BTS—with the statistical areas. The weights to
aggregate BTS-level metadata covariates to the respective statistical area are derived from the
size of the intersection of tiles aj and at of BTS j and statistical area t, respectively, in relation to
the total size of at, also expressed as
wv
j;t≔
aj \ at
at
ð6Þ
In the toy example of Fig 2c, this would reduce to be the intersection of e.g. statistical area
A and the voronoi tile of BTS 1 divided by the total area of A. However, as mentioned above,
area sizes are used in that approach to approximate the (usually) unknown population counts
per intersection by implicitly assuming homogeneous distribution of the population within a
given statistical area.
Augmented voronoi tessellation
The proposed settlement-based mapping schemes relax this obviously strong assumption by
assuming a homogeneous housing structure instead, i.e. a constant population density per set-
tlement area within a given statistical area. Applied to voronoi tessellation, Fig 2c and 2d—
with settlement areas represented as dots—illustrate the difference. Instead of using the area
sizes aj and at to calculate the weights, the “augmented” voronoi tessellation (av) uses the num-
ber of settlements per area, denoted as nj and nt, respectively.
wav
j;t ≔
nj \ nt
nt
Consequently, statistical area-level covariates can easily be acquired for both approaches
using a weighted average (or a weighted median) on BTS-level data.
^R t ¼
XJ
wj;tRj
j¼1
ð7Þ
ð8Þ
Going back to the toy example, while BTS 1 covers the smaller part of C in Fig 2c, thus
receives a smaller weight in the calculation of area-level metadata aggregates, it looks different
in Fig 2d when comparing the number of settlements, represented by green and purple dots.
This way, the proposed methodology accounts for within-cell heterogeneity of the population
distribution.
Both voronoi tessellation and augmented voronoi tessellation splits the full space of interest
into disjunct tiles. Applied to a mobile network this means ubiquituous coverage and zero
redundancies, i.e. all dots are uniquely associated to a specific BTS in the toy example. Again
this is a strong assumption that most likely does not hold true in any real-world application.
To relax this assumption by introducing holes and overlaps in the network coverage, addi-
tional information are necessary that allow for the estimation of coverage measures such as the
received signal strength (RSS) at any given point in space. Fig 2e exemplifies the consequences:
Some settlements are not covered (black dots) and some settlements, even though closer to
one BTS, receive a stronger signal from a more distant BTS. Assuming coverages are correctly
estimated in Fig 2e and 2f, it demonstrates that point-to-polygon allocation tends to underesti-
mate the coverage of statistical areas while voronoi tessellation tends to overestimate it.
PLOS ONE | https://doi.org/10.1371/journal.pone.0241981 November 9, 2020
9 / 28
PLOS ONE Mapping mobile network data using satellite imagery and radio propagation modelling
Propagation-based mapping schemes
Previously presented schemes follow a ‘BTS-centric’ approach by first determining the respec-
tive coverage area of a BTS and then analyzing potential overlaps with other places of interest
such as settlements. In contrast, propagation-based schemes follow an ‘MS-centric’ approach
by looking at the connectivity at the place of interest, i.e. the place of usual residence or the
home location first and then estimating which (group of) BTS it most likely serves. As outlined
in Section Radio propagation modelling, multiple ways exist to estimate the ‘connectivity’ of
an MS, but all require at least information on the distance to the surrounding BTS and addi-
tional technical specifications. With that, the serving BTS can be determined at each place of
interest, thus allowing for a more nuanced coverage mapping. Here, settlements can provide a
common geographic reference for the place usual residence and the home location alike.
Best server area (BSA).
In mobile networks, an MS usually connects to the antenna that
offers the strongest signal. Thus, the settlement-level weight is 1 for the BTS with the strongest
signal and 0 otherwise.
wbsa
i;j ≔
(
1
if Prx;i;j ¼ maxðPrx;i;�Þ
0 otherwise;
ð9Þ
Links weaker than a certain threshold (e.g. a Prx value below—110 dBm) can be discarded
as they represent ‘dead’ links. This way the approach accounts for holes in the network cover-
age. The weights wi,j express the importance of a BTS for a pixel. Similarly to Eq 8, they can be
used to determine the statistical area-level covariate estimates ^Rt using a weighted average:
^R t ¼
Xnt
i¼1
wi;jPnt
i¼1 wi;j
Rj
ð10Þ
Pnt
Due to the binary nature of the weight,
i¼1 wi;j represents the number of settlements with
mobile coverage within a given statistical area. In areas with homogeneous network infrastruc-
ture and full coverage, the best server approach closely resembles the augmented voronoi tes-
sellation with the difference that path loss increases non-linearly with the distance, i.e.
locations very close to the location of a BTS may be served by another, more distant one.
Inverse signal strength. Radio propagation is stochastic by nature. Changing environ-
mental conditions and varying network loads affect the RSS at a given location across time.
Consequently, the strongest signal is not always provided by the same BTS. In order to assure
quality of service, mobile networks usually exhibit a certain number of overlaps. To account
for that, I calculate inverse distance weights (IDW) for each pixel i using the median link bud-
get Prx,i,j as non-linear distance measure (see Eq 11) to the k-nearest antennas. s denotes a tun-
ing parameter, where s = 0 reduces widw
i;j
approximate the best server approach.
to a fixed weight per BTS and a large s can be used to
widw
i;j ≔
vi;jPki
j¼1 vi;j
with vi;j≔
1
jPrx;i;jjs
8j 2 ki
ð11Þ
Here again, widw
i;j can be used to calculate statistical area-level weighted averages of BTS-level
mobile phone metadata covariates as presented in Eq 10.
PLOS ONE | https://doi.org/10.1371/journal.pone.0241981 November 9, 2020
10 / 28
PLOS ONE Mapping mobile network data using satellite imagery and radio propagation modelling
Potential extensions
Depending on data availability, the methodology can further be extended. While MNOs often
regard MS counts as highly sensitive information since they reveal a detailed picture of local
market shares, they can be used to further refine the weights towards more accurate population
counts. [4] presents elaborate approaches to use MS counts and advanced technical network
specifications to derive high-resolution population density estimates from signalling data.
Further, high-resolution population grid estimates such as provided by WorldPop at
100x100m [12] can be used as an alternative to binary settlement data. Here, ^wi;j can be substi-
tuted with the estimated population count ^pi per pixel directly extracted from the image.
Simulation
In order to evaluate the underlying motivation behind this methodology, i.e. more accurate
mapping schemes produce more accurate outcomes, I test the performance of the different
mapping approaches in terms of their overlap with the true coverage area and the accuracy of
the predictions in a controlled setting with groundtruth information. Therefore, I run a simu-
lation T = 1000 times on a synthetic population grid in which I re-distribute individuals, their
poverty status, BTS locations and technical BTS specifications randomly. I observe the geo-
graphical overlap of the true and the estimated coverage areas, the overlap in home-located set-
tlements and the correlation between the true and the estimated variable of interest (in this
case the poverty rate). The main challenge in this simulation is to create “true” coverage areas
for each BTS that provide a realistic, but simplified benchmark for this study. Consequently, I
opt for the extended HATA model. The choice is motivated by a series of propagation model
evaluations using real-world measurements, notably [56–58]. The stochastic component
within the HATA model is disabled in order to isolate the effect of interest.
Setup
I simulate a country including a major city, an uninhabited area such as a large lake or a
national park and rural area otherwise using a 1000 x 1000 grid where each quadratic pixel rep-
resents an edge length of 100m. The urban area is divided into 16 equally-sized (50 x 50 pixel)
small statistical areas, whereas the rural area is divided into 24 larger ones (200 x 200 pixel). I
randomly distribute one million individuals across the grid using a multivariate normal distri-
butions with μx = 10, μy = 10, Sx = [50, 0] and Sy = [0, 50] for the urban area (1/2 of the total
population) and varying parameter values for the rural centers and a uniform distribution for
the remaining rural area. Pixel-level population counts are calculated from individual-level
data. Fig 3 shows an example of the settlement distribution across space and the corresponding
population density.
In the next step, I randomly assign a poverty rate to each pixel. First, I generate a 4x4-pixel
poverty grid for which I calculate the population density (see Fig 4b). In order to account for
differences in the poverty rate between urban and rural areas, I randomly draw from a uniform
distribution with values between 0 and 1 and multiply it with the inverted normalized popula-
tion density. This poverty rate serves as the mean μ for randomly assigning poverty rates to set-
tlements within the respective grid area using a normal distribution N(μ, σ) with σ = 0.5.
Values below 0 and above 1 are windsorized. This two-step procedure tries to limit good pre-
dictive performances for areas not actually covered due to inference facilitated by the same
underlying data generating process. Further, I assume that every inhabitant has one and only
one MS and that there exists an indicator derived from mobile phone metadata that perfectly
correlates with the true poverty rate of a given set of MS. Consequently, deviations in the
PLOS ONE | https://doi.org/10.1371/journal.pone.0241981 November 9, 2020
11 / 28
PLOS ONE Mapping mobile network data using satellite imagery and radio propagation modelling
Fig 3. Simulation setup—Settlements. (a) shows locations of the built-up areas in a hypothetical country, while (b)
shows the corresponding population density in these areas (the brighter the colour, the higher the population density).
https://doi.org/10.1371/journal.pone.0241981.g003
correlation between the poverty rate captured via the “true” coverage area and the poverty rate
captured via the estimated coverage area exclusively originate in their coverage mismatch.
In order to create a mobile network on top of that structure, I use a clustering algorithm
based on the population density (see Fig 5b). BTS are distributed across the country at a ratio
of roughly 1 BTS per 5,000 inhabitants in urban areas and 1 BTS per 10,000 inhabitants in
rural areas. This results in 100 urban and 50 rural BTS in this simulation. BTS are interpreted
as omnidirectional antennas and assigned specific heights, frequencies and output powers.
The specifications vary more strongly in the urban area in order to reflect the greater complex-
ity of network topology generally found in metropolitan areas. Since the HATA model
requires a classification of areas into urban, suburban and rural, I use those 50% of BTS with
the smallest number of pixels associated to them by the clustering algorithm used above as
urban and those 5% of BTS with the largest number of pixels as rural, suburban otherwise. At
the end, BTS heights are between 15—60 m with frequencies at 900 MHz and 2100 MHz and
output power between 40 and 47 dBm. The MS height is fixed at 1m above ground level.
Based on these technical specifications, the true coverage areas and the true home locations
of the settlements using the extended HATA model are calculated and used to create bench-
mark estimates of the true poverty rate. The results are then compared against estimates from
point-to-polygon allocation, voronoi tessellation, augmented voronoi tessellation and BSA
and IDW approaches of a naïve (’simple’) version of the extended HATA model that does not
know the exact technical BTS specifications, but makes an educated guess based on publicly
available information such as the frequencies used in the country and the location of urban
centers. Fig 6 exemplifies how the approaches differ in terms of geographical coverage.
The results are compared in three different ways: How much do they overlap geographi-
cally? How much do they overlap in terms of home-located settlements? How well do they pre-
dict the true poverty rate of a given statistical area?
Results
Table 2 shows the best performing approach in each round across round for all five perfor-
mance indicators. Performance differences between voronoi tessellation versus the augmented
voronoi tessellation and the augmented voronoi tessellation versus the HATA (BSA) approach
PLOS ONE | https://doi.org/10.1371/journal.pone.0241981 November 9, 2020
12 / 28
PLOS ONE Mapping mobile network data using satellite imagery and radio propagation modelling
Fig 4. Simulation setup—True poverty rate.
https://doi.org/10.1371/journal.pone.0241981.g004
PLOS ONE | https://doi.org/10.1371/journal.pone.0241981 November 9, 2020
13 / 28
PLOS ONE Mapping mobile network data using satellite imagery and radio propagation modelling
Fig 5. Simulation setup—BTS locations.
https://doi.org/10.1371/journal.pone.0241981.g005
showcase the relative contribution of settlement weighting and radio-propagation modelling,
respectively. As expected, the simple HATA model clearly outperforms the other mapping
approaches in terms of overlap, both geographically with the true coverage area (see Table 3)
as well as concerning the home-located settlements (see Table 4). As the settlement-based
approaches do only affect the calculation of weights and not of the coverage area, the coverage
results are identical for voronoi tessellation and augmented voronoi tessellation and for the
two HATA approaches, respectively. However, this advantage is not reflected to a similar
extent in the predictive performance.
Interestingly, the HATA (IDW) approach performs poorly in prediction in contrast to the
HATA (BSA) approach. This is due to the fact that the poverty rate in the true coverage area is
calculated based on a deterministic home location, i.e. it is calculated from a constant set of set-
tlements. This coincides directly with the mode-based HATA (BSA) approach, however, it
does not reflect most real-world settings, in which stochastic radio propagation and overlap-
ping coverage areas lead to situations where the captured poverty rate by the BTS is sourced
from varying sets of settlements. The HATA (IDW) approach addresses this setup. Conse-
quently, it is expected that the differences between these two approaches at least diminish in
the application with real-world data in Section Application. Also, deviations of the HATA
(BSA) approach from the benchmark exclusively originate in the technical misspecifications as
the true coverage area is calculated from a correctly specified HATA model. The network com-
plexity faced in real-world settings is expected to further undermine the accuracy of propaga-
tion-based mapping schemes.
Looking at the performance of the two voronoi approaches in Table 2 the value added of
using settlement information becomes apparent. Recalling the setup, the simulation assumes
error-free human settlement identification. This, again, may not hold true in a real-world
application as some buildings may not be detected while some detected buildings may not be
inhabited. Consequently, it is expected that the difference between thee two voronoi
approaches will be less stark in the application.
Fig 7 shows the distribution of the three performance indicators across rounds for those sta-
tistical areas for which every mapping scheme can provide estimates. On average, this reduces
the underlying set of observations from 40 to 32 (see the sample sizes in Table 5). The result
PLOS ONE | https://doi.org/10.1371/journal.pone.0241981 November 9, 2020
14 / 28
PLOS ONE Mapping mobile network data using satellite imagery and radio propagation modelling
Fig 6. Coverage areas exemplified.
https://doi.org/10.1371/journal.pone.0241981.g006
for the true coverage area are represented as benchmark for the other approaches as it estimates
the settlement-level poverty rates actually captured by the respective BTS. Consequently, the
benchmark should provide the upper bound for the R2 and the lower bound for the bias and
the RMSE in each round. Deviations thereof may only be due to spurious correlation.
The sample size difference also explains the difference between the performance of the
point-to-polygon approach in terms of correlation in Table 5 vis-à-vis the performance met-
rics, especially in rural areas. Point-to-polygon allocation does not provide poverty estimates
for 8 out of 40 statistical areas, on average, as they do not host a BTS (cf. Fig 6b). As both
PLOS ONE | https://doi.org/10.1371/journal.pone.0241981 November 9, 2020
15 / 28
PLOS ONE Mapping mobile network data using satellite imagery and radio propagation modelling
Table 2. Best performing approach by round across rounds (in%).
Mapping
Point
Voronoi
Aug. Voronoi (GUF)
HATA (GUF, BSA)
HATA (GUF, IDW)
Coverage
Prediction
Geography
Settlements
0.0
0.0
100.0
0.0
0.07
99.3
R2
27.5
2.6
35.5
29.7
4.7
Bias
28.2
9.9
33.1
13.7
15.1
RMSE
28.6
2.1
36.8
27.7
4.8
https://doi.org/10.1371/journal.pone.0241981.t002
poverty rate and BTS allocation is linked to the population density by design, it can be
expected that the predictive performance for rural areas not hosting a BTS are poor as they are
generated from different underlying distributions.
However, this does not fully explain the performance differences between the approaches.
On one hand, statistical areas are quite large, thus most of the BTS experience little overlaps in
their true coverage area with other statistical areas. Consequently, the statistical area provides a
decent approximation for the coverage. In contrast, simple voronoi tessellation with geograph-
ical weights tends to overemphasize the importance of remote areas as a) it assumes to cover
areas for which data is actually not captured and b) BTS are usually located in close proximity
to populated areas while serving remote areas further away as a side effect of it. This may be
especially relevant in situations with large between-variation among statistical areas, strong
population clusters and imperfect mobile network coverage. While b) is accounted for in the
simulation, only approx. 0.1% of the settlements are not covered by the network. Although this
in line with the mobile network coverage in most countries, it can be expected that propaga-
tion-based schemes that account for holes in the mobile network outperform established
approaches in setups with poor coverage.
Application
In their 2017 study on estimating literacy rates in Senegal published in the Journal of the Royal
Statistical Society Series A, Schmid et al. [1] use point-to-polygon allocation to map BTS point
locations to statistical areas (communes). I revisit the design-based simulation of the study and
extend it with four alternative mapping schemes, notably voronoi tessellation, satellite-aug-
mented voronoi tessellation and the herein presented propagation-based coverage estimation
methods using the best server area approach and the inverse signal strength weights. I compare
the outcomes of all five schemes in terms of bias, root mean squared error (RMSE) and
adjusted R2.
Situation in Senegal
The application draws on real-world data from Orange-Sonatel for the year of 2013 [7]. Dur-
ing that time, the MNO operated mainly on the GSM 900 (2G) band with some UMTS 2100
(3G) deployments in urban centers. A large share of on-net traffic (approx. 91% of overall
Table 3. Geographical overlap with true coverage area (in%).
Mapping
Point
Voronoi
Simple HATA
Total
25.8
30.7
55.3
Rural
15.3
14.1
80.1
Suburban
30.9
25.5
62.1
Urban
22.3
37.0
46.7
https://doi.org/10.1371/journal.pone.0241981.t003
PLOS ONE | https://doi.org/10.1371/journal.pone.0241981 November 9, 2020
16 / 28
PLOS ONE Mapping mobile network data using satellite imagery and radio propagation modelling
Table 4. Overlap with true home-located settlements (in%).
Mapping
Point
Voronoi
Simple HATA
Total
16.9
54.2
59.7
Rural
44.3
56.8
87.6
Suburban
15.9
60.7
66.5
Urban
14.9
48.0
50.6
https://doi.org/10.1371/journal.pone.0241981.t004
traffic vis-à-vis a market share of approx. 57%) during that year suggests a high prevalence of
dual SIM use. It is expected that in this setting a negligible share of SIM cards are used by IoT
devices others than MS. Coverage advantages in rural areas suggest dual-SIM use to be a phe-
nomenon of more densely populated areas. The country exhibits little irregularities in the ter-
rain: The highest point of Senegal being approx. 648 m above sea level is located at its southern
border. The lowest point constitutes the sea level. Urban built-up areas with multi-storey
buildings are predominantly limited to downtown Dakar. Most of the country is dominated
by savanna with sparse high-grown vegetation.
Original study
In their design-based simulation, Schmid et al. [1] implement a stratified two-stage cluster
sample design similar to the one used in large-scale household surveys such as the Demo-
graphic and Health Survey (DHS) using a 10% random sample of a pseudo-population as sam-
pling frame, the 431 communes of Senegal as primary sampling units (PSUs) and the 14
regions of Senegal as strata. The authors combine the constructed ‘survey’ data with covariates
extracted from mobile phone metadata on the level of communes in order to evaluate different
small area estimation techniques using the unemployment rate as target variable of choice. The
72 available covariates are calculated on the subscriber-level using the Python library Bandi-
coot [8]. The subscriber-level covariates are allocated and aggregated to a BTS using the most
frequently used BTS by a subscriber between 7pm and 7am as the home location. The BTS-
level covariates are then allocated and aggregated using point-in-polygon allocation. Variable
selection is performed backwards on large communes using the Bayesian Information Crite-
rion. The covariates are used to generate small area unemployment rate estimates using a
transformed Fay-Herriot model. Finally, Schmid et al. evaluate the small area estimates against
the ‘true’ pseudo-population aggregates in 500 simulation runs using bias and RSME for a)
communes covered by the survey (in-sample) b) communes not covered by the survey (out-of-
sample) and c) communes without covariates from mobile phone metadata. For additional
details on the setup of the original study, I refer to [1].
Extensions
I re-run the simulation of the original study five times thereby only varying the commune-
level matrix of covariates as inputs. Specifically, I create five distinct sets of commune-level
covariates beforehand by applying different mapping schemes during the aggregation process
of the BTS-level data of the original study. First, I use the point-to-polygon allocation used in
the original study. Second, I apply a standard voronoi tessellation to extract spatial weights
proportional to the geographical overlap of tile and statistical area as described in Voronoi tes-
sellation since it is used in most other studies in this field. Third, I augment the voronoi tessel-
lation with settlement information from GUF by taking the number of white pixels
(representing (part of) a settlement) within each section as a weight for commune-level aggre-
gates to account for within-cell heterogeneity. Fourth, I implement the extended HATA (BSA)
PLOS ONE | https://doi.org/10.1371/journal.pone.0241981 November 9, 2020
17 / 28
PLOS ONE Mapping mobile network data using satellite imagery and radio propagation modelling
Fig 7. Estimating the true poverty rate for statistical areas. Distribution of the three performance metrics adjusted R2, bias
and RMSE with the estimated poverty rate using the true coverage area, i.e. built-up areas perfectly allocated to BTS, as
‘Benchmark’ across 1000 simulation runs.
https://doi.org/10.1371/journal.pone.0241981.g007
PLOS ONE | https://doi.org/10.1371/journal.pone.0241981 November 9, 2020
18 / 28
PLOS ONE Mapping mobile network data using satellite imagery and radio propagation modelling
Table 5. Area-level correlation of estimated and true poverty rate & sample size.
Mapping
Benchmark
Point
Voronoi
Aug. Voronoi
Simple HATA (BSA)
Simple HATA (IDW)
ρ
0.905
0.930
0.873
0.896
0.897
0.885
n
40
36
40
40
40
40
ρRural
0.734
0.828
0.622
0.715
0.717
0.670
nRural
24
20
24
24
24
24
ρUrban
0.971
0.940
0.966
0.966
0.957
0.962
nUrban
16
16
16
16
16
16
https://doi.org/10.1371/journal.pone.0241981.t005
model as presented in Methodology and GUF data. In densely populated areas, this approach
closely resembles voronoi tessellation, however, it allows for holes in the network and for non-
linear relationships between signal strength and distance. Fifth, I use inverse signal strength
weights—HATA (IDW)—to capture the stochastic nature of a link.
Comparing Fig 8c and 8d to the direct estimator (Fig 8b) shows the benefits of augmenting
survey data with mobile phone metadata: providing estimates for small areas not originally
covered by the survey. Looking at settlements in Fig 8a, it is noteworthy that one commune—
Thietty in the region Kolda—does not appear to host any settlement identified as such in GUF
data. While official population numbers do not support this view, it underlines the fact that
information extracted from satellite imagery, e.g. settlement classifications, are subject to some
degree of uncertainty.
Assumptions
In contrast to point-to-polygon allocation and voronoi tessellation, the extended HATA
model requires additional technical antenna specifications, notably the antenna and receiver
height, the frequency and the transmitter power. As additional information are not available
in the original study, I make following assumptions: I fix both the antenna height htx and the
receiver height hrx at the lower bound of the extended HATA model, which is 30 m and 1 m,
respectively, both located outdoors with line-of-sight and a transceiver installed above the
roof. As most of Senegal is flat without high multi-storey buildings except in downtown Dakar
and in large parts no high-grown vegetation this assumption appears reasonable. Further, I fix
the frequency in rural areas at 900 MHz and in urban centers at 2100 MHz and I interpret BTS
as omnidirectional antennas with an output power of 45 dBm. This is clearly a simplification
of the actual network topology, especially in urban areas with a mix of directed micro and
macro cells. However, in Senegal in 2013, 4G has not yet been introduced and Orange-Sonatel
was operating 3G (on the 2100 MHz frequency band) only in urban areas. The remaining
country was served with 2G technology on the 900 MHz band. Comparing own estimates with
coverage area estimates for 2G in 2017 published by Sonatel [59] allows for a rough sanity
check for the assumptions.
While Senegal offers an official classification of rural and urban on the commune-level, it is
imperfect for the purposes of this study, as it takes a wide variety of non-network-specific fac-
tors into account. This leads to a situation where places with a high population density, e.g.
Touba Mosque, are classified as commune rurale. Instead, I use BTS density per km2 as a proxy
for urbanity with a threshold of 1. Communes with more than one BTS per km2 are classified
as urban, those 50% of the communes with the lowest site density are classified as rural, the
remaining communes are classified as suburban. This represents a more network-oriented
measure of urbanity and is also in line with the area type classification of the HATA model.
PLOS ONE | https://doi.org/10.1371/journal.pone.0241981 November 9, 2020
19 / 28
PLOS ONE Mapping mobile network data using satellite imagery and radio propagation modelling
Fig 8. Commune-level coverage areas in Senegal. Areas for which estimates of indicators of interest are available are coloured in red. Lower resolution
built-settlements extents data reprinted from [10] under a CC BY license, with permission from WorldPop, original copyright 2018, are used in (a) for
illustrative purposes.
https://doi.org/10.1371/journal.pone.0241981.g008
Results
Similar to Table 2 in the simulation, Table 6 shows which mapping scheme performed best
across the 500 evaluation rounds. Confirming initial findings of Section Simulation, there is
no clear winner. While point-to-polygon allocation performs best in out-of-sample predictions
in terms of RMSE (54.0% of the rounds), it performs poorest in in-sample predictions. One
possible explanation is that the lower average number of predictors used across rounds reduces
the effects of overfitting. While HATA (IDW), HATA (BSA) and the augmented voronoi
PLOS ONE | https://doi.org/10.1371/journal.pone.0241981 November 9, 2020
20 / 28
PLOS ONE Mapping mobile network data using satellite imagery and radio propagation modelling
Table 6. Best performing approach by round across rounds (in%).
Mapping
Point
Voronoi
Aug. Voronoi (GUF)
HATA (GUF, BSA)
HATA (GUF, IDW)
Adj. R2
in
6.0
10.2
27.2
27.0
29.6
Bias
RMSE
Avg. # of predictors
in
16.4
21.0
22.6
21.8
18.2
out
23.2
16.6
18.0
16.4
25.8
in
12.6
21.6
33.2
18.0
14.6
out
54.0
22.8
5.2
7.8
10.2
4.2
5.0
6.5
6.4
6.2
https://doi.org/10.1371/journal.pone.0241981.t006
approach perform well across performance metrics, the overall difference between the
approaches is limited (see Fig 9 and Table 7).
In contrast, urban communes do not perform significantly better than rural ones as sug-
gested by the simulation results. Table 8 shows, similar to Table 5 for the simulation, the corre-
lation between the actual and predicted commune-level unemployment rates. Fig 10 shows an
orientation along the diagonal signalling overall good fit. A possible explanation is that the
structural relationship of mobile phone metadata covariates and the unemployment rate is
captured more robustly for rural areas as they constitute 385 out of 431 communes in Senegal.
To test this explanation, Tables 1 and 2 in S1 Appendix show the results for in-sample and
out-of-sample predictions by commune status, respectively. While urban communes outper-
form rural ones in in-sample prediction they fare worse for in the out-of-sample setting, thus
supporting the aforementioned hypothesis.
While settlement-based mapping schemes exhibit improvements in the model fit compared
to point allocation or voronoi tessellation, they do not translate into major efficiency gains in
terms of bias and rmse (see Fig 9b and 9c). Possible reasons are threefold: There is a significant
classification error in the settlement data. The complete absence of settlements in Thietty,
Kolda, support this assumption. As a cross-check, I re-run the analysis with an alternative
source of settlement information. Specifically, I use high-resolution population density esti-
mates from WorldPop [12], however, it does not lead to gains in efficiency (cf. Table 3 in S1
Appendix). Second, there is high spatial auto-correlation, thus little structural difference
between the densely and sparsely populated areas in terms of the variable of interest—here
unemployment—so even though latter are overemphasized in the calculations, it does not
affect the outcome predictions. Here, I re-run the application with alternative variables of
interest, i.e. the literacy rate and the population count (cf. Tables 4 and 5 in S1 Appendix);
again, without significant efficiency gains versa point allocation and voronoi tessellation.
Third, there is little within-area variation of the population density so that geographic weights
and settlement-based weights are very similar. The correlation coefficient between the weights
of the two voronoi approaches confirm that with ρ = 0.98. Also, I use the 100 meters x 100
meters population estimates from WorldPop to extract commune-specific variation coeffi-
cients. For 76.8% of the communes, the within-commune variance is below 1, for 4% it is
above 100 with a maximum at 3553.4.
In general, the value added of using propagation-based mapping schemes appears to be
negligible in this application, even though official coverage area estimates by Sonatel [59] hint
at the abundant presence of both overlaps and holes in the mobile network. A potential expla-
nation is that the simplified HATA model is misspecified to an extent where the introduced
errors cancel out the potential benefits. Looking at the specifications used in the application,
this is most likely due to an underestimation of the coverage as the augmented voronoi
PLOS ONE | https://doi.org/10.1371/journal.pone.0241981 November 9, 2020
21 / 28
PLOS ONE Mapping mobile network data using satellite imagery and radio propagation modelling
Fig 9. Evaluation of poverty rate estimates for in-sample communes. Distribution of the three performance metrics adjusted
R2, bias and RMSE across 500 simulation runs on a comparable set of communes. The typical trade-off between the bias and the
variance of a small area estimator vis-à-vis the direct survey estimator becomes apparent.
https://doi.org/10.1371/journal.pone.0241981.g009
PLOS ONE | https://doi.org/10.1371/journal.pone.0241981 November 9, 2020
22 / 28
PLOS ONE Mapping mobile network data using satellite imagery and radio propagation modelling
Table 7. Correlation with true unemployment rate and sample size in Senegal.
Mapping
Point
Voronoi
Aug. Voronoi (GUF)
HATA (GUF, BSA)
HATA (GUF, IDW)
ρ
0.535
0.542
0.519
0.511
0.527
n
431
431
431
431
431
ρin
0.765
0.778
0.780
0.770
0.781
nin
192
196
195
194
196
ρout
0.320
0.313
0.280
0.269
0.308
nout
210
235
233
232
234
ρooc
0.355
-
0.586
0.670
-
nooc
29
0
3
5
1
https://doi.org/10.1371/journal.pone.0241981.t007
approach closely resembles the upper bound for an overestimation using the HATA (BSA)
within a—by assumption—largely homogeneous network.
Conclusion
Augmenting official statistics with mobile phone metadata still faces multiple methodological
challenges, one of them is finding a common reference unit. As record-linkage on the individ-
ual-level presents considerable privacy risks a common procedure is to combine aggregates of
these two disparate data sources on a geographical level. However, the stochastic nature of
radio propagation makes it difficult to pin down coverage areas of the mobile network. Based
on this study the good news is that it does not have to be complicated if supervised learning /
prediction is the goal. While propagation-based models can help to refine the accuracy of cov-
erage area estimation, it does not greatly impact the quality of the outcome predictions. One
reason is that usually cells are located in a way that they provide a good service to as many MS
as possible. As radio signals fade over distance, this means they are in close proximity to areas
with high demand, i.e. densely populated places. Mapping schemes, in turn, mainly differ
from each other when looking at the limits of a cell. However, most of the traffic which is cor-
related with statistical data for training/prediction is generated nearby, so the differences
between mapping schemes become less relevant. Also, while geographical weights as used in
most applications in this field ignore heterogeneity occurring within the cells, the correspond-
ing statistical areas are often significantly larger. Therefore, cross-border cells, which could
actually profit from weighting schemes that take within-cell heterogeneity into account, occur
less frequent. In addition, cells and administrative (thus often statistical) areas are intimately
linked via population clusters as both tend to be centered around them.
However, this study just provided initial evidence to inform future mapping choices and
could be extended in multiple ways: First, both in the simulation and the application direc-
tional antennas are combined to omnidirectional antennas. While this is motivated by the typ-
ical data availability in real-world applications, it is of course a strong simplification of the
actual network topology. As the lower bound of spatial heterogeneity captured is given by the
number of unique areas resulting from intersecting coverage areas and statistical areas, studies
Table 8. Area-level correlation of estimated and true unemployment rate & sample size.
Mapping
Point
Voronoi
Aug. Voronoi
Simple HATA (BSA)
Simple HATA (IDW)
ρ
0.535
0.542
0.519
0.511
0.527
n
431
431
431
431
431
ρRural
0.507
0.519
0.495
0.487
0.510
nRural
385
385
385
385
385
ρUrban
0.527
0.469
0.411
0.374
0.369
nUrban
46
46
46
46
46
https://doi.org/10.1371/journal.pone.0241981.t008
PLOS ONE | https://doi.org/10.1371/journal.pone.0241981 November 9, 2020
23 / 28
PLOS ONE Mapping mobile network data using satellite imagery and radio propagation modelling
Fig 10. True vs. estimated unemployment rate by commune status for a single simulation run.
https://doi.org/10.1371/journal.pone.0241981.g010
such as [4] have shown that moving from an BTS-oriented to a cell-oriented analysis could
greatly affect analysis, especially via potential increases in sample size. However, it needs fur-
ther investigation how refined mapping schemes can add further value, particularly in the
presence of measurement uncertainty, to supervised learning setups in cell-level analysis. Sec-
ond, the study used comparatively simple empirical propagation models based on real-world
measurements largely ignoring actual environments. More advanced propagation models
exist, however, they require significantly more computing resources that could limit their
applicability as they take the physical surrounding via digital surface models into account.
Nevertheless, investigating this constitutes an interesting path for further research.
Supporting information
S1 Appendix. Results and instructions. Results from the cross-checks of the application and
instructions for replicating the findings of this study.
(PDF)
S1 File. Simulation. Code for replicating the simulation study.
(ZIP)
PLOS ONE | https://doi.org/10.1371/journal.pone.0241981 November 9, 2020
24 / 28
PLOS ONE Mapping mobile network data using satellite imagery and radio propagation modelling
S2 File. Application. Code and data for replicating the application study. See S1 Appendix for
further details.
(ZIP)
Acknowledgments
The author would like to thank Damien Jacques, Emmanuel Letouze´, Edward Oughton, So¨ren
Pannier, Neeti Pokhriyal and Timo Schmid for excellent comments and helpful discussions.
Author Contributions
Conceptualization: Till Koebe.
Data curation: Till Koebe.
Formal analysis: Till Koebe.
Funding acquisition: Till Koebe.
Investigation: Till Koebe.
Methodology: Till Koebe.
Project administration: Till Koebe.
Resources: Till Koebe.
Software: Till Koebe.
Supervision: Till Koebe.
Validation: Till Koebe.
Visualization: Till Koebe.
Writing – original draft: Till Koebe.
Writing – review & editing: Till Koebe.
References
1. Schmid T, Bruckschen F, Salvati N, Zbiranski T. Constructing sociodemographic indicators for national
statistical institutes by using mobile phone data: estimating literacy rates in Senegal. Journal of the
Royal Statistical Society Series A: Statistics in Society. 2017; 180(4):1163–1190. https://doi.org/10.
1111/rssa.12305
2. Pokhriyal N, Jacques DC. Combining disparate data sources for improved poverty prediction and map-
ping. Proceedings of the National Academy of Sciences of the United States of America. 2017; 114(46):
E9783–E9792. https://doi.org/10.1073/pnas.1700319114 PMID: 29087949
3. Blumenstock J, Cadamuro G, On R. Predicting poverty and wealth from mobile phone metadata. Sci-
ence. 2015; 350:1073–1076. https://doi.org/10.1126/science.aac4420 PMID: 26612950
4. Ricciato F, Widhalm P, Pantisano F, Craglia M. Beyond the’single-operator, CDR-only’ paradigm: An
interoperable framework for mobile phone network data analyses and population density estimation.
Pervasive and Mobile Computing. 2017; 35:65–82. https://doi.org/10.1016/j.pmcj.2016.04.009
5.
The Economist Intelligence Unit. The Inclusive Internet Index 2019; 2019. Available from: https://
theinclusiveinternet.eiu.com/.
6. Phillips C, Sicker D, Grunwald D. A survey of wireless path loss prediction and coverage mapping meth-
ods. IEEE Communications Surveys and Tutorials. 2013; 15(1):255–270. https://doi.org/10.1109/
SURV.2012.022412.00172
7.
de Montjoye YA, Smoreda Z, Trinquart R, Ziemlicki C, Blondel VD. D4D-Senegal: The Second Mobile
Phone Data for Development Challenge. arXiv preprint arXiv:14074885. 2014;.
8. De Montjoye YA, Rocher L, Pentland AS. Bandicoot: A python toolbox for mobile phone metadata. Jour-
nal of Machine Learning Research. 2016; 17:1–5.
PLOS ONE | https://doi.org/10.1371/journal.pone.0241981 November 9, 2020
25 / 28
PLOS ONE Mapping mobile network data using satellite imagery and radio propagation modelling
9. Esch T, Heldens W, Hirner A, Keil M, Marconcini M, Roth A, et al. Breaking new ground in mapping
human settlements from space—The Global Urban Footprint. ISPRS Journal of Photogrammetry and
Remote Sensing. 2017; 134:30–42. https://doi.org/10.1016/j.isprsjprs.2017.10.012
10. WorldPop. (www.worldpop.org—School of Geography and Environmental Science, University of South-
ampton; Department of Geography and Geosciences, University of Louisville; Departement de Geogra-
phie, Universite de Namur) and Center for International Earth Science Information Network (CIESIN),
Columbia University; 2018. Global High Resolution Population Denominators Project—Funded by The
Bill and Melinda Gates Foundation (OPP1134076). https://dx.doi.org/10.5258/SOTON/WP00649
11. Department of Economic and Social Affairs UN. Handbook on geospatial infrastructure in support of
census activities. St/esa/sta ed. New York, USA: United Nations Publication; 2009.
12. Stevens FR, Gaughan AE, Linard C, Tatem AJ. Disaggregating census data for population mapping
using Random forests with remotely-sensed and ancillary data. PLoS ONE. 2015; 10(2):e0107042.
https://doi.org/10.1371/journal.pone.0107042 PMID: 25689585
13.
Freire S, MacManus K, Pesaresi M, Doxsey-Whitfield E, Mills J. Development of new open and free
multi-temporal global population grids at 250 m resolution. AGILE. 2016; p. 6.
14. Henderson JV, Storeygard A, Weil DN. Measuring economic growth from outer space; 2012. Available
from: http://pubs.aeaweb.org/doi/10.1257/aer.102.2.994.
15. Chen X, Nordhaus WD. Using luminosity data as a proxy for economic statistics. Proceedings of the
National Academy of Sciences of the United States of America. 2011; 108(21):8589–8594. https://doi.
org/10.1073/pnas.1017031108 PMID: 21576474
16. Pinkovskiy M, Sala-i Martin X. Lights, Camera. . . Income! Illuminating the National Accounts-Household
Surveys Debate. The Quarterly Journal of Economics. 2016; 131(2):579–631. https://doi.org/10.1093/
qje/qjw003
17.
Leyk S, Gaughan AE, Adamo SB, de Sherbinin A, Balk D, Freire S, et al. Allocating people to pixels: A
review of large-scale gridded population data products and their fitness for use. Earth System Science
Data Discussions. 2019; 11(3):1–30. https://doi.org/10.5194/essd-2019-82
18. Bonafilia D, Gill J, Kirsanov D, Sundram J. Mapping for humanitarian aid and development with weakly-
and semi-supervised learning. Facebook; 2019. Available from: https://bit.ly/2PxK5dx.
19. Harvey JT. Estimating census district populations from satellite imagery: Some approaches and limita-
tions. International Journal of Remote Sensing. 2002; 23(10):2071–2095. https://doi.org/10.1080/
01431160110075901
20. Steinnocher K, De Bono A, Chatenoux B, Tiede D, Wendt L. Estimating urban population patterns from
stereo-satellite imagery. European Journal of Remote Sensing. 2019; 52(sup2):12–25. https://doi.org/
10.1080/22797254.2019.1604081
21.
Jean N, Burke M, Xie M, Davis WM, Lobell DB, Ermon S. Combining satellite imagery and machine
learning to predict poverty. Science. 2016; 353(6301):790–794. https://doi.org/10.1126/science.
aaf7894 PMID: 27540167
22. Weidmann NB, Schutte S. Using night light emissions for the prediction of local wealth. Journal of
Peace Research. 2017; 54(2):125–140. https://doi.org/10.1177/0022343316630359
23. Oughton E. Quantified Global Broadband Strategies for Connecting Unconnected Communities. SSRN
Electronic Journal. 2019;. https://doi.org/10.2139/ssrn.3427492
24. Blondel VD, Decuyper A, Krings G. A survey of results on mobile phone datasets analysis; 2015. Avail-
able from: http://www.epjdatascience.com/content/4/1/10.
25. Deville P, Linard C, Martin S, Gilbert M, Stevens FR, Gaughan AE, et al. Dynamic population mapping
using mobile phone data. Proceedings of the National Academy of Sciences. 2014; 111 (45):15888–
15893. https://doi.org/10.1073/pnas.1408439111 PMID: 25349388
26. Khodabandelou G, Gauthier V, Fiore M, El Yacoubi AM. Estimation of Static and Dynamic Urban Popu-
lations with Mobile Network Metadata. IEEE Transactions on Mobile Computing. 2018;. https://doi.org/
10.1109/TMC.2018.2871156
27. Botta F, Moat HS, Preis T. Quantifying crowd size with mobile phone and Twitter data. Royal Society
Open Science. 2015; 2(5):150162. https://doi.org/10.1098/rsos.150162 PMID: 26064667
28. Douglass RW, Meyer DA, Ram M, Rideout D, Song D. High resolution population estimates from tele-
communications data. EPJ Data Science. 2014; 4(1):1–13. https://doi.org/10.1140/epjds/s13688-015-
0040-6
29.
Lu X, Bengtsson L, Holme P. Predictability of population displacement after the 2010 Haiti earthquake.
Proceedings of the National Academy of Sciences of the United States of America. 2012; 109
(29):11576–11581. https://doi.org/10.1073/pnas.1203882109 PMID: 22711804
30. Gundogdu D, Incel OD, Salah AA, Lepri B. Countrywide arrhythmia: emergency event detection using
mobile phone data. EPJ Data Science. 2016; 5(1):25. https://doi.org/10.1140/epjds/s13688-016-0086-0
PLOS ONE | https://doi.org/10.1371/journal.pone.0241981 November 9, 2020
26 / 28
PLOS ONE Mapping mobile network data using satellite imagery and radio propagation modelling
31. Schneider CM, Belik V, Couronne´ T, Smoreda Z, Gonza´lez MC. Unravelling daily human mobility
motifs. Journal of the Royal Society Interface. 2013; 10(84). https://doi.org/10.1098/rsif.2013.0246
PMID: 23658117
32. Wesolowski A, Eagle N, Noor AM, Snow RW, Buckee CO. The impact of biases in mobile phone owner-
ship on estimates of human mobility. Journal of the Royal Society Interface. 2013; 10(81):20120986.
https://doi.org/10.1098/rsif.2012.0986
33. Matamalas JT, De Domenico M, Arenas A. Assessing reliable human mobility patterns from higher
order memory in mobile communications. Journal of the Royal Society Interface. 2016; 13
(121):20160203. https://doi.org/10.1098/rsif.2016.0203 PMID: 27581479
34.
35.
36.
Iovan C, Olteanu-Raimond AM, Couronne´ T, Smoreda Z. Moving and calling: Mobile phone data quality
measurements and spatiotemporal uncertainty in human mobility studies. In: Lecture Notes in Geoinfor-
mation and Cartography. vol. 2013-Janua. Springer, Cham; 2013. p. 247–265. Available from: http://
link.springer.com/10.1007/978-3-319-00615-4_14.
Janzen M, Vanhoof M, Smoreda Z, Axhausen KW. Closer to the total? Long-distance travel of French
mobile phone users. Travel Behaviour and Society. 2018; 11:31–42. https://doi.org/10.1016/j.tbs.2017.
12.001
Taylor L. No place to hide? The ethics and analytics of tracking mobility using mobile phone data. Envi-
ronment and Planning D: Society and Space. 2016; 34(2):319–336. https://doi.org/10.1177/
0263775815608851
37. Wesolowski A, Eagle N, Tatem AJ, Smith DL, Noor AM, Snow RW, et al. Quantifying the impact of
human mobility on malaria. Science. 2012; 338(6104):267–270. https://doi.org/10.1126/science.
1223467 PMID: 23066082
38. Rubrichi S, Smoreda Z, Musolesi M. A comparison of spatial-based targeted disease mitigation strate-
gies using mobile phone data. EPJ Data Science. 2018; 7(1):17. https://doi.org/10.1140/epjds/s13688-
018-0145-9
39.
40.
41.
Tizzoni M, Bajardi P, Decuyper A, Kon Kam King G, Schneider CM, Blondel V, et al. On the Use of
Human Mobility Proxies for Modeling Epidemics. PLoS Computational Biology. 2014; 10(7):e1003716.
https://doi.org/10.1371/journal.pcbi.1003716 PMID: 25010676
Le Menach A, Tatem AJ, Cohen JM, Hay SI, Randell H, Patil AP, et al. Travel risk, malaria importation
and malaria transmission in Zanzibar. Scientific Reports. 2011; 1(1):93. https://doi.org/10.1038/
srep00093 PMID: 22355611
Frı´as-Martı´nez E, Williamson G, Frı´as-Martı´nez V. An agent-based model of epidemic spread using
human mobility and social network information. In: Proceedings—2011 IEEE International Conference
on Privacy, Security, Risk and Trust and IEEE International Conference on Social Computing, PAS-
SAT/SocialCom 2011. IEEE; 2011. p. 57–64. Available from: http://ieeexplore.ieee.org/document/
6113095/.
42.
Lima A, Domenico MD, Pejovic V, Musolesi M. Exploiting Cellular Data for Disease Containment and
Information Campaigns Strategies in Country-Wide Epidemics. CoRR. 2013;abs/1306.4.
43. Park PS, Blumenstock JE, Macy MW. The strength of long-range ties in population-scale social net-
works. Science. 2018; 362(6421):1410–1413. https://doi.org/10.1126/science.aau9735
44. Bakker MA, Piracha DA, Lu PJ, Bejgo K, Bahrami M, Leng Y, et al. Measuring Fine-Grained Multidimen-
sional Integration Using Mobile Phone Metadata: The Case of Syrian Refugees in Turkey. In: Guide to
Mobile Data Analytics in Refugee Scenarios. Cham: Springer International Publishing; 2019. p. 123–
140. Available from: http://link.springer.com/10.1007/978-3-030-12554-7_7
45. Sundsøy P. Can mobile usage predict illiteracy in a developing country? arXiv preprint
arXiv:160701337. 2016;.
46. Blumenstock J, Callen M, Ghani T. Why do defaults affect behavior? Experimental evidence from
Afghanistan. American Economic Review. 2018; 108(10):2868–2901. https://doi.org/10.1257/aer.
20171676
47. Bruckschen F, Koebe T, Ludolph M, Marino MF, Schmid T. Refugees in Undeclared Employment—A
Case Study in Turkey. In: Guide to Mobile Data Analytics in Refugee Scenarios. Cham: Springer Inter-
national Publishing; 2019. p. 329–346. Available from: http://link.springer.com/10.1007/978-3-030-
12554-7_17
48.
Tennekes M. mobloc: Mobile phone location algorithms and tools; 2018. Available from: https://github.
com/MobilePhoneESSnetBigData/mobloc_v0.1.
49. OECD. Household definitions in other statistical standards. In: OECD Guidelines for Micro Statistics on
Household Wealth. OECD Publishing; 2013. p. 275–277. Available from: https://www.oecd-ilibrary.org/
docserver/9789264194878-18-en.pdf?expires=1570525698&id=id&accname=guest&checksum=
FE901313FB9732B831D7F32703E8569C.
PLOS ONE | https://doi.org/10.1371/journal.pone.0241981 November 9, 2020
27 / 28
PLOS ONE Mapping mobile network data using satellite imagery and radio propagation modelling
50. Vanhoof M, Lee C, Smoreda Z. Performance and sensitivities of home detection from mobile phone
data. arXiv preprint arXiv:180909911. 2018;.
51. Oughton EJ, Frias Z, van der Gaast S, van der Berg R. Assessing the capacity, coverage and cost of
5G infrastructure strategies: Analysis of the Netherlands. Telematics and Informatics. 2019; 37:50–69.
https://doi.org/10.1016/j.tele.2019.01.003
52. Green MP, Wang SS. Signal propagation model used to predict location accuracy of GSM mobile
phones for emergency applications. In: Proceedings—RAWCON 2002: 2002 IEEE Radio and Wireless
Conference. Institute of Electrical and Electronics Engineers Inc.; 2002. p. 119–122.
53. Hata M. Empirical Formula for Propagation Loss in Land Mobile Radio Services. IEEE Transactions on
Vehicular Technology. 1980; 29(3):317–325. https://doi.org/10.1109/T-VT.1980.23859
54. Damasso E L M Correia. Digital Mobile Radio Towards Future Generation. Luxembourg: European
Commission; 1999. 11. Available from: https://publications.europa.eu/en/publication-detail/-/
publication/f2f42003-4028-4496-af95-beaa38fd475f/language-en.
55. Okumura Y, Ohmori E, Kawano T, Fukuda K. Field Strength and Its Variability in UHF and VHF Land-
Mobile Radio Service. Review of the Electrical Communication Laboratory, September-October, 1968.
1968; 16:825–873.
56. Sharma RK Singh PK. Comparative Analysis of Propagation Path loss Models with Field Measured
Data. International Journal of Engineering Science and Technology. 2010; 2(6):2008–2013.
57. Abhayawardhana VS, Wassellt IJ, Crosby D, Sellars MP, Brown MG. Comparison of empirical propaga-
tion path loss models for fixed wireless access systems. In: IEEE Vehicular Technology Conference.
vol. 61; 2005. p. 73–77.
58. Phillips C, Sicker D, Grunwald D. Bounding the error of path loss models. In: 2011 IEEE International
Symposium on Dynamic Spectrum Access Networks, DySPAN 2011; 2011. p. 71–82.
59. Sonatel. Coverage Map Sonatel 2019; 2019. Available from: https://bit.ly/2uJplYk.
PLOS ONE | https://doi.org/10.1371/journal.pone.0241981 November 9, 2020
28 / 28
PLOS ONE |
10.1371_journal.pone.0239694 | RESEARCH ARTICLE
Analysis of COVID-19 clinical trials: A data-
driven, ontology-based, and natural language
processing approach
Shray AlagID*
The Harker School, San Jose, CA, United States of America
* 21shraya@students.harker.org
Abstract
With the novel COVID-19 pandemic disrupting and threatening the lives of millions,
researchers and clinicians have been recently conducting clinical trials at an unprecedented
rate to learn more about the virus and potential drugs/treatments/vaccines to treat its infec-
tion. As a result of the influx of clinical trials, researchers, clinicians, and the lay public, now
more than ever, face a significant challenge in keeping up-to-date with the rapid rate of dis-
coveries and advances. To remedy this problem, this research mined the ClinicalTrials.gov
corpus to extract COVID-19 related clinical trials, produce unique reports to summarize find-
ings and make the meta-data available via Application Programming Interfaces (APIs).
Unique reports were created for each drug/intervention, Medical Subject Heading (MeSH)
term, and Human Phenotype Ontology (HPO) term. These reports, which have been run
over multiple time points, along with APIs to access meta-data, are freely available at http://
covidresearchtrials.com. The pipeline, reports, association of COVID-19 clinical trials with
MeSH and HPO terms, insights, public repository, APIs, and correlations produced are all
novel in this work. The freely available, novel resources present up-to-date relevant biologi-
cal information and insights in a robust, accessible manner, illustrating their invaluable
potential to aid researchers overcome COVID-19 and save hundreds of thousands of lives.
Introduction
Since the onset of the Coronavirus disease 2019 (COVID-19) pandemic, researchers and clini-
cians have been swiftly conducting clinical trials to better understand the virus, its transmis-
sion, and potential drugs and vaccines to counter its rapid spread. Such COVID-19 related
clinical trials can be found at ClinicalTrials.gov, a database for globally-conducted clinical tri-
als run by the United States National Library of Medicine.
The number of COVID-19 related clinical trials is dramatically increasing: There were
approximately 500 clinical trials in mid-late April, more than 1000 in early May, over 2000 in
early June, and over 3000 in mid-July [1]. By late August, this number has increased to nearly
3,500 clinical trials. As a result of the unprecedented volume of new clinical trials, the task of
staying informed about crucial developments and ongoing research is not only arduous but
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: Alag S (2020) Analysis of COVID-19
clinical trials: A data-driven, ontology-based, and
natural language processing approach. PLoS ONE
15(9): e0239694. https://doi.org/10.1371/journal.
pone.0239694
Editor: Sunghwan Kim, National Library of
Medicine, UNITED STATES
Received: June 17, 2020
Accepted: September 13, 2020
Published: September 30, 2020
Copyright: © 2020 Shray Alag. This is an open
access article distributed under the terms of the
Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: All of the data used in
this research is publically available at http://
ClinicalTrials.gov. All of the reports generated from
this research can be found at http://
CovidResearchTrials.com.
Funding: The author received no specific funding
for this work.
Competing interests: The author has declared that
no competing interests exist.
PLOS ONE | https://doi.org/10.1371/journal.pone.0239694 September 30, 2020
1 / 13
PLOS ONE Analysis of COVID-19 clinical trials: A data-driven, ontology-based, and natural language processing approach
also extremely time-consuming. Ignorance of recent advancements could result in misconcep-
tions, the misuse of time, and non-optimal allocation of funds/resources. Therefore, improved
accessibility of information about COVID-19 related clinical trials would aid clinicians,
researchers, and the lay public alike.
This research followed seven principal steps:
1. Obtain all COVID-19 related clinical trials from ClincialTrials.gov.
2. Extract relevant Interventions, Drugs, Outcomes, Location, Medical Subject Heading
(MeSH [2]) tags, and more information from the identified clinical trials.
3. Associate the clinical trial with a Human Phenotype Ontology (HPO [3]) term, if applicable,
through using the MeSH tags.
4. Correlate Drugs, MeSH terms, and HPO terms computationally by examining the fre-
quency of these elements in COVID-19 clinical trials. In essence, for a given term of interest
compute other correlated terms [4, 5].
5. Generate unique publicly-accessible, informative—yet concise—reports for each of the
Intervention/Drug terms, MeSH terms, and HPO terms.
6. Create a freely-available public repository detailing associations between Interventions/
Drugs, MeSH terms, and HPO terms while additionally providing access to Application
Programming Interfaces (APIs), which enable the user to interact with the data using a pro-
gramming language, such as Java or Python.
7. Analyze the clinical trials at multiple time points, enabling future meta-analyses.
Materials and methods
This work builds upon the pipeline developed in Alag 2020 [6]. The core of the pipeline and
methodology is publicly available at protocols.io (dx.doi.org/10.17504/protocols.io.bfacjiaw).
As detailed in the protocols, the following online repositories/vocabularies were used:
1. ClinicalTrials.gov [7]: The complete repository of clinical trials displayed at ClinicalTrials.
gov is available in extensible Markup Language (XML) format with a well-defined schema
[8].
2. Human Phenotype Ontology (HPO [3]): HPO is a standardized vocabulary of phenotype
abnormalities that are seen in humans [3].
3. Medical Subject Headings (MeSH [2]): The ClinicalTrials.gov XML contains information
about MeSH terms, and, using the MeSH online tool [9], MeSH ids were retrieved from
their corresponding MeSH terms.
The predominant differences between the methodology employed here and that followed
in Alag 2020 [6] are detailed in the following subsections. Readers are encouraged to refer to
Alag 2020 for more in-depth details. It is relevant to note that MeSH and HPO ontologies were
chosen over multiple alternative ontologies as these were easily compatible with the clinical
trails XML format, enabling higher-level correlations across related genes, SNPs, protein muta-
tions, and even clinical trials. Additionally, the additional feature of correlating COVID-19
clinical trials to SNPs and protein mutations is provided through utilizing the HPO ontology
and previous work in Alag 2020 [6].
PLOS ONE | https://doi.org/10.1371/journal.pone.0239694 September 30, 2020
2 / 13
PLOS ONE Analysis of COVID-19 clinical trials: A data-driven, ontology-based, and natural language processing approach
Identifying COVID-19 clinical trials
To accurately and efficiently identify COVID-19 trials, sections of the text of the trial—the
title, brief summary, outcomes sections, clinical trial criteria, conditions, MeSH terms, and
detailed description—were seen to have a case-insensitive match to any of the following terms:
COVID 19, COVID-19, SARS-CoV-2, 2019-ncov, coronavirus, severe acute respiratory syn-
drome coronavirus 2, 2019 novel coronavirus, and wuhan coronavirus. These terms were
selected after reviewing various sites, including clinicaltrials.gov [10], that have kept a running
list of COVID-19-related publications and clinical trials. The research attempts to ensure no
COVID-19 clinical trials are overlooked, and, as a result, the terms list is more extensive than
the parameters clinicaltrials.gov uses in its COVID-19 search. These terms were kept constant
to be able to analyze longitudinal trends without bias.
Analysis pipeline
To analyze the clinical trials, a methodology similar to that of Alag 2020 [6] was used with the
following additions. It is important to note that interventions are the focus of a clinical trial:
often in a clinical trial, the response of patients who are given an intervention (drug, test, pro-
cedure, etc.) are compared to patients who do not receive that intervention. Interventions can
be drugs, medical devices, vaccines, procedures, genetic tests, noninvasive techniques, such as
diet, education, or exercise, and are sorted into eleven different categories (e.g., genetic, radia-
tion, etc.).
1. Interventions/Drugs: A list of unique interventions/drugs that appear in COVID-19 related
clinical trials were created. This dictionary of terms was essential to later formulate reports
and correlations. The data was retrieved at multiple time points to get insights about the
rate at which these trials are occurring.
2. Correlations: Through using co-occurrences of a specific term, correlations were noted
between drugs/interventions, MeSH terms, and HPO terms, as further described in the fol-
lowing subsection.
3. Reports: Unique hypertext markup language (HTML)-based reports were created for each
of the interventions/drugs, MeSH tags, and HPO terms. These reports have associated clini-
cal trials and related HPO terms, each of which also has associated genes, SNPs, and protein
mutations. A more in-depth discussion of the reports is provided in the Results section.
Computing correlations
The co-occurrence of a term across different clinical trials was used to compute correlations
between terms using the following procedure [4]:
1. Create an incidence matrix where each row is a term of interest (drug, MeSH, or HPO). By
doing so, there are m such terms and n clinical trials. A value of one is marked each time
the mi term is correlated with the nj clinical trial. All other non-correlated positions should
have a value of zero.
2. Normalize the data by creating a unit vector for each term. Unit vectors are obtained by
dividing each element of a row by the magnitude of that row.
3. For each term, compute the pair-wise dot product between its vector and all other vectors.
The resulting number is a measure of normalized correlation.
4. Sort the results to create a prioritized list of related terms.
PLOS ONE | https://doi.org/10.1371/journal.pone.0239694 September 30, 2020
3 / 13
PLOS ONE Analysis of COVID-19 clinical trials: A data-driven, ontology-based, and natural language processing approach
Hierarchical Clustering [11] or K-Means [12] could also be used to find clusters of related
terms. Additionally, the rows and columns can be switched to cluster similar clinical trials by
their associated terms [5].
Results
The “Results” section is comprised of discussions about the following three main areas:
1. Details on the created public repository to provide access to the data used, reports created,
correlations mapped, and APIs produced.
2. Insights from analyzing COVID-19 clinical trials.
3. Findings related to correlations.
Public repository
Web page to access longitudinal analysis data, reports, and APIs. All analysis results are
accessible via the Covid Research Trials home page, available at http://CovidResearchTrials.
com. A view of the home page is seen in S1 File. The web page provides access to data and
reports from multiple time frames: At the date of publication, information from May 2nd,
2020, May 23rd, 2020, June 6th, July 18th, and August 16th, 2020 are illustrated. Although sub-
ject to change, the Covid Research Trials home page provides the latest analysis results. Addi-
tionally, the home page has links to numerous Java APIs and a Google Colab page, which
facilitate easy local access to this research’s insights and results. The functionalities of the vari-
ous APIs are to retrieve information about the following:
1. A list of all COVID-19 related clinical trials along with the description of each trial
2. Potential vaccines and their corresponding clinical trials
3. Drugs and their corresponding clinical trials
4. COVID-19 related HPO terms and their correlated clinical trials
5. COVID-19 related MeSH terms along with their corresponding clinical trials
6. COVID-19 related clinical trials and the outcomes of each of the trials
Each Java class is a stand-alone program and does not require any other package beyond
the Java core classes: Users can simply download a Java IDE, install Java, and run the class on
that IDE. S2 File contains screenshots of the documentation of each of the six APIs mentioned
above. The main function in each of the classes demonstrates how each of the public methods
can be called. Additionally, the Google Colab Notebook, which uses Python, reads in and
details information about all relevant clinical trials, the tested drugs, and the potential
vaccines.
Insights from analyzing COVID-19 clinical trials
The following subsections detail trends gleaned from analyzing the longitudinal data, meta-
level information about COVID-19 related clinical trials, key intervention/drug, MeSH, and
HPO terms, and provide information about the generated reports.
Longitudinal information associated with COVID-19. Fig 1 illustrates the trends of
COVID-19 clinical trials, with the key takeaways being the following:
PLOS ONE | https://doi.org/10.1371/journal.pone.0239694 September 30, 2020
4 / 13
PLOS ONE Analysis of COVID-19 clinical trials: A data-driven, ontology-based, and natural language processing approach
Fig 1. COVID-19 clinical trial trends: Longitudinal trends from COVID-19 related clinical trials. The data is
plotted across five time points: May, 3rd 2020, May, 23rd, 2020, June 6th, 2020, July 18th, 2020, and August 16, 2020.
(a) The total number of clinical trials at each time points; (b) Percent of new clinical trials (trials published between
time segments) that are COVID-19 related; (c) Percent of all clinical trials that are COVID-19 related; (d) The total
number of COVID-19 related clinical trials.
https://doi.org/10.1371/journal.pone.0239694.g001
1. The number of COVID related trials has been linearly increasing with approximately 120
new clinical trials per week. As of late August 2020 there were nearly 3,500 COVID-19
related clinical trials.
2. Out of the more recent trials, an increasing percentage of all trials are COVID-19 related: in
the first period approximately 5% of all new trials were COVID related while around a
quarter of new clinical trials from May 23rd to June 6th are COVID-19 related. In July-
August 2020, the rate of change of new COVID-19 clinical trials appears to be slowing.
3. The percent of COVID-19 clinical trials, as a percentage of all clinical trials, is close to 1%
as of late August 2020.
The methodology employed in this research will continue to be performed at future time
points, and the future changes in trends can provide interesting insights.
Key interventions/drugs associated with COVID-19. At the last data point, on August
16th, 2020, there were 3,523 unique interventions/drugs that were associated with COVID-19
clinical trials. Fig 2 shows the most-frequently occurring interventions/drugs for COVID-19
related clinical trials. The most popular interventions being tested are Hydroxychloroquine,
Azithromycin, Tocilizumab, Standard of Care, Placebo, Convalescent Plasma, Ivermectin, and
Remdesivir, as seen in Fig 2. Additionally, it is important to note that Hydroxychloroquine,
with 102 trials, is the most common drug that is being tested in clinical trials.
Fig 3 depicts the frequency of intervention types across COVID-19 related trials, illustrating
that the most popular intervention category is Drug, followed by Other, Behavioral, Biological,
and Diagnostic test. As most COVID-19 clinical trials are either experimenting with the effi-
cacy of drugs, observing behavioral changes, creating/validating diagnostic tests, or discover-
ing the biological effects of the virus, the high intervention frequency occurrences of drug,
behavioral, biological, and diagnostic test validate the procedure employed. It is interesting to
see the emergence of clinical trials associated with genetic information (though just eight), and
future changes in the frequencies of interventions will be insightful, especially as researchers
PLOS ONE | https://doi.org/10.1371/journal.pone.0239694 September 30, 2020
5 / 13
PLOS ONE Analysis of COVID-19 clinical trials: A data-driven, ontology-based, and natural language processing approach
Fig 2. Intervention/Drug information. The graph shows the relative frequency of the different intervention/drugs
that are referenced in COVID-19 related clinical trials (August 2020). A minimum occurrence of 8 clinical trials was
used to create the graph.
https://doi.org/10.1371/journal.pone.0239694.g002
begin looking to better understand the role of genomic factors in symptoms and recovery of
patients with COVID-19.
Outcomes, phase, and status of COVID-19 clinical trials. Analyzing the COVID-19
related clinical trials provides the following insights:
1. Outcomes are events (e.g., patient death or discharge from the hospital) that are actively
being monitored in a clinical trial [13]. As depicted in Fig 4, the majority of outcomes either
deal with acute lung disease, time to clinical improvement, antibodies, or, unfortunately,
mortality.
Fig 3. Intervention-categories. The majority of the interventions used in clinical trials are drugs, other, behavioral,
biological, and diagnostic tests (August 2020).
https://doi.org/10.1371/journal.pone.0239694.g003
PLOS ONE | https://doi.org/10.1371/journal.pone.0239694 September 30, 2020
6 / 13
PLOS ONE Analysis of COVID-19 clinical trials: A data-driven, ontology-based, and natural language processing approach
Fig 4. Outcomes are events that are being monitored in a clinical trial. This graph shows the most commonly
occurring outcomes that are being measured in COVID-19 related clinical trials. Here a threshold of six was used to
list the outcome.
https://doi.org/10.1371/journal.pone.0239694.g004
2. Including Phase 0, the Food and Drug Administration (FDA [14]) defines a five-phase [15]
approach for ensuring the safety and efficacy of an intervention. Fig 5 illustrates that the
majority of the COVID-19 related clinical trials are in Phases 2 and 3, indicating that they
have made considerable headway but are still far from completion. Notably, there are many
very early phase trials indicating the recent interest in performing COVID-19 related stud-
ies. Encouragingly, there are quite a few clinical trials that are in Phase 4.
3. Recruitment status [16] indicates the degree to which a trial may need to enroll subjects. Fig
6 shows that the vast majority of COVID-19 clinical trials are either still recruiting partici-
pants or have not yet begun recruiting, illustrating that most of these trials are still far from
complete. Further, the preliminary stages of these clinical trials indicates that they have
been recently undertaken, portraying the growing interest in COVID-19 clinical trials.
MeSH and HPO terms associated with COVID-19 clinical trials. Across COVID-19
related clinical trials, the most prevalent MeSH terms, as illustrated in Fig 7, are Infections,
Severe Acute Respiratory Syndrome, and Pneumonia. The frequencies of MeSH terms shed
light on the areas researchers are currently examining and possibly illuminate under-
researched topics.
As illustrated in Fig 8, Respiratory tract infection and Abnormality of the cardiovascular
system are the leading HPO nodes that were identified with COVID-19 trials. The validity of
Fig 5. Phase information including phase 0, the FDA defines a five phase process for approving new drugs [14].
This graph shows the phases for COVID-19 related clinical trials.
https://doi.org/10.1371/journal.pone.0239694.g005
PLOS ONE | https://doi.org/10.1371/journal.pone.0239694 September 30, 2020
7 / 13
PLOS ONE Analysis of COVID-19 clinical trials: A data-driven, ontology-based, and natural language processing approach
Fig 6. Recruitment status indicates where the clinical trialis with respect to recruiting new patients. This graph
shows the number of clinical trials in each status state.
https://doi.org/10.1371/journal.pone.0239694.g006
our methodology is upheld by the occurrence of HPO nodes like Abnormal lung morphology,
Abnormality of the cardiovascular system, Acute kidney injury, and Respiratory HPO terms
since these terms have proven to be associated with COVID-19.
Reports for drugs, MeSH, and HPO terms associated with COVID-19 clinical trials.
An HTML report was created for each of the unique drugs, MeSH, and HPO terms associated
with COVID-19 clinical trials. Longitudinal access (reports over multiple time periods) is
freely available via the home page (http://CovidResearchTrials.com). As shown in S1 File, each
report contains a list of either the drug, the MeSH terms, or the HPO terms. All of the terms in
a category are displayed on the left-hand side of the report to enable easy navigation, and the
reports contain a list of correlated drugs, MeSH, and HPO terms. Further, all reports contain
the details of the clinical trials in which the term is referenced. Every clinical trial report shows
the mapped HPO and MeSH terms, which are also hyperlinked. Related HPO terms, with
their associated genes, protein mutations, and SNPs are also referenced in the report.
Fig 7. MeSH information details the most prevalent MeSH terms across COVID-19 related clinical trials. Some
notable and expected prominent terms are Infections, Severe Acute Respiratory Syndrome, and Pneumonia.
https://doi.org/10.1371/journal.pone.0239694.g007
PLOS ONE | https://doi.org/10.1371/journal.pone.0239694 September 30, 2020
8 / 13
PLOS ONE Analysis of COVID-19 clinical trials: A data-driven, ontology-based, and natural language processing approach
Fig 8. HPO information portrays the most widely noted HPO terms. Note that HPO terms are normalized as
detailed in Alag 2020 [6]. The most frequent terms were Respiratory tract infection, Abnormality of the cardiovascular
system, Acute kidney injury, Abnormal lung morphology, Depressivity, Diabetes mellitus, and more.
https://doi.org/10.1371/journal.pone.0239694.g008
Correlations across drugs, MeSH, and HPO terms
In this section, two representative terms are selected to demonstrate the results of the cluster-
ing across drugs, MeSH, and HPO terms. Terms related to the MeSH term “D018352: Corona-
virus Infections” are shown in Table 1 and discussed below:
1. Drugs: As expected, the most common drugs include comparisons to Placebo, Hydroxy-
chloroquine [17], Standard of care, No Intervention, Ivermectin [18], Nitazoxanide [19],
Favipiravir [20], Remdesivir [21], Vitamin C, etc. Since these drugs are all related to
COVID-19 efforts, the methodology for correlation is validated, suggesting the potential of
using more advanced machine learning techniques.
2. MeSH: Related MeSH terms are D045169 Severe Acute Respiratory Syndrome, D003141
Communicable Disease, D007239 Infection, D011014 Pneumonia, D013577 Syndrome,
D012128 Respiratory Distress Syndrome, Adult, D012127 Respiratory Distress Syndrome,
Newborn, D014777 Virus Diseases, D055371 Acute Lung Injury, D011024 Pneumonia,
Viral, D003333 Coronaviridae Infections, D012141 Respiratory Tract Infections, D012327
RNA Virus Infections, D016638 Critical Illness, etc. As COVID-19 is a communicable dis-
ease, caused by a virus that causes an infection associated with the respiratory system, the
appearance of as a top MeSH term only increases the validity of the methdology [22].
3. HPO Terms: Related HPO terms include: HP:0002090 Pneumonia, HP:0011947 Respira-
tory tract infection, HP:0000819 Diabetes mellitus, HP:0010444 Pulmonary insufficiency,
HP:0012418 Hypoxemia, HP:0002905 Hyperphosphatemia, HP:0006802 Abnormal ante-
rior horn cell morphology, HP:0001871 Abnormality of blood and blood-forming tissues,
HP:0002900 Hypokalemia, HP:0000846 Adrenal insufficiency, HP:0000132 Menorrhagia,
PLOS ONE | https://doi.org/10.1371/journal.pone.0239694 September 30, 2020
9 / 13
PLOS ONE Analysis of COVID-19 clinical trials: A data-driven, ontology-based, and natural language processing approach
Table 1. Related drugs, MeSH, and HPO terms using co-occurrences with D018352: Coronavirus infection.
Term Type
Related Term
Score
1
Drug
Placebo [multiple]
Hydroxychloroquine
No intervention
Ivermectin
Nitazoxanide
Favipiravir
Remdesivir
Vitamin C
Zinc
Rivaroxaban
2
MeSH
D045169 Severe Acute Respiratory Syndrome
3
HPO
D003141 Communicable Disease
D007239 Infection
D011014 Pneumonia
D013577 Syndrome
D012128 Respiratory Distress Syndrome, Adult
D012127 Respiratory Distress Syndrome, Newborn
D014777 Virus Diseases
D055371 Acute Lung Injury
D011024 Pneumonia, Viral
D003333 Coronaviridae Infections
D012141 Respiratory Tract Infections
D012327 RNA Virus Infections
D016638 Critical Illness
HP:0002090 Pneumonia
HP:0011947 Respiratory tract infection
HP:0000819 Diabetes mellitus
HP:0010444 Pulmonary insufficiency
HP:0012418 Hypoxemia
HP:0002905 Hyperphosphatemia
HP:0006802 Abnormal anterior horn cell morphology
HP:0001871 Abnormality of blood and blood-forming tissues
HP:0002900 Hypokalemia
HP:0000846 Adrenal insufficiency
HP:0000132 Menorrhagia
HP:0002088 Abnormal lung morphology
HP:0005978 Type II diabetes mellitus
0.62
0.19
0.12
0.11
0.10
0.10
0.10
0.09
0.09
0.09
0.8
0.28
0.27
0.18
0.17
0.13
0.13
0.13
0.12
0.11
0.09
0.07
0.07
0.06
0.18
0.08
0.05
0.04
0.04
0.04
0.04
0.04
0.04
0.04
0.04
0.04
0.04
Results from finding similar Drugs, MeSH, and HPO terms for the MeSH term Coronavirus Infection (from
August 2020). For full details and updated results readers are referred to the report that is available at http://
CovidResearchTrials.com.
https://doi.org/10.1371/journal.pone.0239694.t001
HP:0002088 Abnormal lung morphology, HP:0005978 Type II diabetes mellitus, etc. The
correlation algorithm can even discern diabetes as related to COVID-19, which is accurate
as individuals with diabetes are at an increased risk of developing severe illness from
COVID-19 [23].
PLOS ONE | https://doi.org/10.1371/journal.pone.0239694 September 30, 2020
10 / 13
PLOS ONE Analysis of COVID-19 clinical trials: A data-driven, ontology-based, and natural language processing approach
Table 2. Related drugs, MeSH, and HPO terms using co-occurrences with drug: Hydroxychloroquine.
Term Type
1
Drug
Related Term
Azithromycin
Placebo [multiple]
Vitamin C
Interferon Beta-1A
Therapeutic anticoagulation
Routine care for COVID-19 patients
Lopinavir-Ritonavir
Favipiravir
Interferon Beta-1B
Zinc
Vitamin D
2
3
MeSH
D003141 Communicable Diseases
D007239 Infection
D045169 Severe Acute Respiratory Syndrome
D055371 Acute Lung Injury
D018352 Coronavirus Infections
D012127 Respiratory Distress Syndrome, Newborn
D012128 Respiratory Distress Syndrome, Adult
HPO
HP:0003326 Myalgia
HP:0002090 Pneumonia
HP:0100526 Neoplasm of the lung
HP:0002487 Hyperkinetic movements
HP:0100598 Pulmonary edema
HP:0030358 Non-small cell lung carcinoma
Score
0.37
0.22
0.17
0.15
0.14
0.14
0.14
0.17
0.11
0.11
0.11
0.27
0.26
0.21
0.19
0.19
0.17
0.16
0.10
0.09
0.09
0.07
0.07
0.07
Results from finding similar Drugs, MeSH, and HPO terms for the Drug: Hydroxychloroquine (August 2020). For
full details and latest results readers are referred to the report that is available at http://CovidResearchTrials.com
https://doi.org/10.1371/journal.pone.0239694.t002
Similarly, terms related to the drug Hydroxychloroquine [17] are shown in Table 2 and dis-
cussed below:
1. Drugs: Other drugs correlated to Hydroxychloroquine are: Azithromycin, Placebo, Inter-
feron Beta-1A, Routine care for COVID-19 patients, Lopinavir-Ritonavir, Favipiravir,
Interferon Beta-1B, Zinc, Vitamin D, etc.
2. MeSH: Related MeSH terms are D003141 Communicable Diseases, D007239 Infection,
D045169 Severe Acute Respiratory Syndrome, D055371 Acute Lung Injury, D018352 Coro-
navirus Infections, D012127 Respiratory Distress Syndrome, Newborn, D012128 Respira-
tory Distress Syndrome, Adult, etc.
3. HPO Terms: Related HPO terms include: HP:0003326 Myalgia, HP:0002090 Pneumonia,
HP:0100526 Neoplasm of the lung, HP:0002487 Hyperkinetic movements, HP:0100598
Pulmonary edema, HP:0030358 Non-small cell lung carcinoma, etc.
Readers are encouraged to see the full results and many other correlations/insights from the
project home page http://CovidResearchTrials.com.
PLOS ONE | https://doi.org/10.1371/journal.pone.0239694 September 30, 2020
11 / 13
PLOS ONE Analysis of COVID-19 clinical trials: A data-driven, ontology-based, and natural language processing approach
Conclusion and future work
In this work, COVID-19 related clinical trials were not only successfully mined from Clinical-
Trials.gov but also associated with Drugs, HPO, and MeSH terms. Unique reports for inter-
vention/drugs, MeSH, and HPO terms were created and are freely available on the web, along
with APIs (Java and Google Colab notebooks) for programmatic access. Further, the publicly-
available site (http://CovidResearchTrials.com) contains analysis at multiple time points, fur-
ther providing researchers with longitudinal information about clinical trials and associated
entities, as well as demonstrating the reproducibility of the methods. The programmatic access
of the data connecting COVID-19 with MeSH and HPO terms can also be useful for machine
learning and other insights. This methodology and the generated reports provide a succinct
summary of COVID-19 related Interventions/Drugs, MeSH terms, HPO terms, clinical trials,
genes, SNPs, and protein mutations all in one place. Overall, the insights and resources gener-
ated could potentially be an invaluable, time-saving resource to researchers, clinicians, and the
lay public.
In the future, this framework can additionally be applied to other scientific corpora, such as
PubMed [24] and PubMed Central [25]. Further, as evaluations will be done at future time-
points, the changes in trends over the coming months will be noteworthy and may provide
insight onto the global community united response to fight the COVID-19 pandemic. The
analysis results can be furthered enhanced by normalizing drug/intervention terms across the
clinical trials.
Supporting information
The following supporting figures are available for this article:
S1 File. Screen shots of Covid Research Trials homepage, various reports, and API tool
kits.
(PDF)
S2 File. Screen shots of Covid Research Trials Java API documentation.
(PDF)
Acknowledgments
The author would like to thank Ayush Alag, Princeton University, for his valuable feedback on
the manuscript and guidance during the project. Additionally, the author thanks Dr. Eric Nel-
son, the Computer Science Department Chair at The Harker School, for his encouragement of
the author’s scientific endeavors.
Author Contributions
Conceptualization: Shray Alag.
Data curation: Shray Alag.
Formal analysis: Shray Alag.
Investigation: Shray Alag.
Methodology: Shray Alag.
Resources: Shray Alag.
Software: Shray Alag.
PLOS ONE | https://doi.org/10.1371/journal.pone.0239694 September 30, 2020
12 / 13
PLOS ONE Analysis of COVID-19 clinical trials: A data-driven, ontology-based, and natural language processing approach
Validation: Shray Alag.
Visualization: Shray Alag.
Writing – original draft: Shray Alag.
Writing – review & editing: Shray Alag.
References
1.
Thorlund K, Dron L, Park J, Hsu G, Forrest JI, Mills EJ. A real-time dashboard of clinical trials for
COVID-19. Lancet Digit Health. 2020 Apr 24; 2(6):e286–7. Epub ahead of print. https://doi.org/10.1016/
S2589-7500(20)30086-8 PMID: 32363333
2. NIH MeSH https://meshb.nlm.nih.gov/search.
3.
4.
The Human Phenotype Ontology Available at https://hpo.jax.org/app/.
Ingersoll G, Morton T, Farris A. Taming Text: How to Find, Organize, and Manipulate It. New York:
Manning Publications Co; 2013.
5. Alag, S. Collective Intelligence in Action, 2008 ISBN: 1933988312, Manning Publications Co.
6. Alag S (2020). Unique insights from ClinicalTrials.gov by mining protein mutations and RSids in addition
to applying the Human Phenotype Ontology Unique insights from ClinicalTrials.gov by mining protein
mutations and RSids in addition to applying the Human Phenotype Ontology. PLOS ONE 15(5):
e0233438. https://doi.org/10.1371/journal.pone.0233438 PMID: 32459809
7. Clinicaltrials.gov Available at: https://clinicaltrials.gov/. Accessed August 2019
8. Clinical trials XML schema https://clinicaltrials.gov/ct2/html/images/info/public.xsd
9. NCBI MeSH https://www.ncbi.nlm.nih.gov/mesh
10. COVID-19 Clinical Trials https://clinicaltrials.gov/ct2/results?cond=COVID-19
11. What is Hierarchical Clustering? https://www.kdnuggets.com/2019/09/hierarchical-clustering.html
12. K Means https://stanford.edu/cpiech/cs221/handouts/kmeans.html
13.
14.
Ferreira JC, Patino CM. Types of outcomes in clinical research. J Bras Pneumol. 2017 Jan-Feb; 43
(1):5. https://doi.org/10.1590/S1806-37562017000000021 PMID: 28380183
The FDA’s Drug Review Process: Ensuring Drugs Are Safe and Effective [accessed June 13, 2020]
https://www.fda.gov/drugs/drug-information-consumers/fdas-drug-review-process-ensuring-drugs-are-
safe-and-effective
15. ClinicalTrials.gov Phases [accessed June 13, 2020] https://www.clinicaltrials.gov/ct2/help/glossary/
phase
16. ClinicalTrials.gov RECRUITMENT STATUS [accessed June 13, 2020] https://clinicaltrials.gov/ct2/help/
glossary/recruitment-status
17. Das S, Bhowmick S, Tiwari S, Sen S. An Updated Systematic Review of the Therapeutic Role of Hydro-
xychloroquine in Coronavirus Disease-19 (COVID-19) [published online ahead of print, 2020 May 28].
Clin Drug Investig. 2020;1-11. https://doi.org/10.1007/s40261-020-00927-1 PMID: 32468425
18. Heidary F., Gharebaghi R. Ivermectin: a systematic review from antiviral effects to COVID-19 comple-
mentary regimen. J Antibiot 73, 593–602 (2020). https://doi.org/10.1038/s41429-020-0336-z PMID:
32533071
19. Kelleni MT. Nitazoxanide/azithromycin combination for COVID-19: A suggested new protocol for early
management. Pharmacol Res. 2020 Jul; 157:104874. Epub 2020 Apr 30. https://doi.org/10.1016/j.phrs.
2020.104874 PMID: 32360581
20.
21.
Favipiravir. In: Drugs and Lactation Database (LactMed). Bethesda (MD): National Library of Medicine
(US); 2006.
Li Z, Wang X, Cao D, Sun R, Li C, Li G. Rapid review for the anti-coronavirus effect of remdesivir. Drug
Discov Ther. 2020; 14(2):73-76. https://doi.org/10.5582/ddt.2020.01015 PMID: 32378648
22. Singhal T. A Review of Coronavirus Disease-2019 (COVID-19). Indian J Pediatr. 2020 Apr; 87(4):281–
286. Epub 2020 Mar 13. https://doi.org/10.1007/s12098-020-03263-6 PMID: 32166607
23. Drucker DJ. Coronavirus Infections and Type 2 Diabetes-Shared Pathways with Therapeutic Implica-
tions. Endocr Rev. 2020; 41(3):bnaa011. https://doi.org/10.1210/endrev/bnaa011 PMID: 32294179
24. PubMed https://www.ncbi.nlm.nih.gov/pubmed/
25. PubMed Central https://www.ncbi.nlm.nih.gov/pmc/
PLOS ONE | https://doi.org/10.1371/journal.pone.0239694 September 30, 2020
13 / 13
PLOS ONE |
10.1371_journal.pone.0250041 | RESEARCH ARTICLE
An observational study of engineering online
education during the COVID-19 pandemic
Shadnaz AsgariID
Antonella Sciortino3,5
1,2*, Jelena Trajkovic2, Mehran Rahmani3, Wenlu Zhang2, Roger C. Lo4,
1 Department of Biomedical Engineering, California State University, Long Beach, California, United States
of America, 2 Department of Computer Engineering and Computer Science, California State University, Long
Beach, California, United States of America, 3 Department of Civil Engineering and Construction Engineering
Management, California State University, Long Beach, California, United States of America, 4 Department of
Chemical Engineering, California State University, Long Beach, California, United States of America,
5 College of Engineering, California State University, Long Beach, California, United States of America
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
* Shadnaz.Asgari@csulb.edu
Abstract
The COVID-19 pandemic compelled the global and abrupt conversion of conventional face-
to-face instruction to the online format in many educational institutions. Urgent and careful
planning is needed to mitigate negative effects of pandemic on engineering education that
has been traditionally content-centered, hands-on and design-oriented. To enhance engi-
neering online education during the pandemic, we conducted an observational study at Cali-
fornia State University, Long Beach (one of the largest and most diverse four-year university
in the U.S.). A total of 110 faculty members and 627 students from six engineering depart-
ments participated in surveys and answered quantitative and qualitative questions to high-
light the challenges they experienced during the online instruction in Spring 2020. Our
results identified various issues that negatively influenced the online engineering education
including logistical/technical problems, learning/teaching challenges, privacy and security
concerns and lack of sufficient hands-on training. For example, more than half of the stu-
dents indicated lack of engagement in class, difficulty in maintaining their focus and Zoom
fatigue after attending multiple online sessions. A correlation analysis showed that while
semi-online asynchronous exams were associated with an increase in the perceived cheat-
ing by the instructors, a fully online or open-book/open-note exams had an association with
a decrease in instructor’s perception of cheating. To address various identified challenges,
we recommended strategies for educational stakeholders (students, faculty and administra-
tion) to fill the tools and technology gap and improve online engineering education. These
recommendations are practical approaches for many similar institutions around the world
and would help improve the learning outcomes of online educations in various engineering
subfields. As the pandemic continues, sharing the results of this study with other educators
can help with more effective planning and choice of best practices to enhance the efficacy of
online engineering education during COVID-19 and post-pandemic.
OPEN ACCESS
Citation: Asgari S, Trajkovic J, Rahmani M, Zhang
W, Lo RC, Sciortino A (2021) An observational
study of engineering online education during the
COVID-19 pandemic. PLoS ONE 16(4): e0250041.
https://doi.org/10.1371/journal.pone.0250041
Editor: Mohammed Saqr, KTH Royal Institute of
Technology, SWEDEN
Received: November 22, 2020
Accepted: March 30, 2021
Published: April 15, 2021
Copyright: © 2021 Asgari et al. This is an open
access article distributed under the terms of the
Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: All relevant data are
within the manuscript and its Supporting
Information files.
Funding: This research is partially supported by
CSULB Champions program through Coronavirus
Aid, Relief, and Economic Security (CARES) Act
funding.
Competing interests: The authors have declared
that no competing interests exist.
PLOS ONE | https://doi.org/10.1371/journal.pone.0250041 April 15, 2021
1 / 17
PLOS ONE Engineering online education during COVID-19 pandemic
1. Introduction
Engineering education has been traditionally content-centered, hands-on, design-oriented,
and focused on the development of critical thinking or problem-solving skills [1]. Various ped-
agogical methodologies have shown efficacy in enhancement of engineering education includ-
ing active learning [2], flipped classroom [3] and project-based learning [4–6]. Over the last
decade, online education has become a viable component of higher education in engineering
subfields such as electrical and computer engineering, computer science and information tech-
nology especially at the master’s or post-graduate level [7].
Although the online education has not been a new concept to educators in general, the
COVID-19 pandemic introduced an unprecedented and global need to explore online teach-
ing/learning opportunities within the entire spectrum of educational levels and majors.
According to the UNESCO, since the onset of pandemic, more than 1.5 billion students world-
wide (90.1% of total enrolled learners) have been affected by the COVID-19 closures and sub-
sequent educational changes [8]. The sudden closure of most educational institutions around
the world compelled the conversion of the face-to-face instruction into a fully online (or
blended/hybrid) format in a short transitional time. As a result, academic institutions that
were mainly focused on traditional face-to-face instructions encountered various challenges in
this transition [9].
Urgent, careful and evidence-based planning is needed to mitigate the impact of pandemic
on engineering education especially for vulnerable, disadvantaged and underrepresented stu-
dents facing substantial challenges beyond their academic responsibilities, including family
obligations, financial burden and additional employments [10–12]. Additional efforts need to
be taken to guarantee that the online instruction of engineering courses still meets the rigorous
requirements of the program accreditations such as Accreditation Board for Engineering and
Technology (ABET).
Despite the existing literature on online engineering education, to the best of our knowl-
edge, there has been no thorough (quantitative and qualitive) analysis of challenges and factors
affecting the pandemic online engineering education in the universities that were mainly offer-
ing face-to-face classes pre-pandemic. This work is aimed for addressing this gap by consider-
ing the following two questions:
1. What are the main challenges influencing online engineering education during COVID-19
pandemic for institutions that were mainly focused on traditional face-to-face instruction
pre-COVID?
2. What are the empirical insight and recommendations to address these challenges?
Sloan’s online learning consortium has defined the five pillars of high-quality online educa-
tion as: learning effectiveness, student satisfaction, faculty satisfaction, access, scale, and cost
[1]. Given these factors, we designed and conducted surveys among engineering faculty mem-
bers and students at California State University, Long Beach (CSULB) to systematically investi-
gate the challenges encountered during the abrupt transition from face-to-face to the online
mode of instruction in Spring 2020. This paper presents the results of the conducted surveys
and propose solutions for the improvement of online engineering education. Sharing the
results of this observational study with other educators can facilitate a more robust continuity
of engineering education during ongoing pandemic. It can also aid with overall improvement
and consequently further promotion of online engineering education in the post-pandemic
era especially for universities that were previously focused on traditional face-to-face instruc-
tion. CSULB is one of the most diverse universities in the U.S. in terms of race/ethnicity, gen-
der, financial and cultural characteristics (e.g. with a large percentage of first-generation or
PLOS ONE | https://doi.org/10.1371/journal.pone.0250041 April 15, 2021
2 / 17
PLOS ONE Engineering online education during COVID-19 pandemic
low-income students). Thus, the results of this study can especially help the institutions with
similar demographics to enhance their online engineering education during and post-
pandemic.
1.1. Related work
The existing literature has identified several challenges that need to be considered for the effec-
tive design and offering of online courses:
• Converting a course from conventional face-to-face to the online format is time consuming
and requires the instructor’s familiarity with (or willingness to learn about) online learning
pedagogy and instructional tools, including the learning management system (LMS) [13].
• Some students prefer to learn difficult concepts face-to-face [14] and believe that face-to-face
instructions provide deeper level of learning compared to the online [15].
• Designing a fair, equitable, and rigorous assessment to minimize cheating and plagiarism is
difficult in online environment [16].
• A successful education requires creating and maintaining a reliable and robust infrastructure
that supports both faculty and students [7, 17–19].
• Hands-on training to work with equipment, instruments, and materials in a controlled labo-
ratory setting is an inherent and necessary aspect of a successful engineering education [1,
10]. Addressing this essential aspect within a fully online teaching platform is challenging
particularly at the undergraduate level.
Recently, several studies have tried to identify the major factors and best practices that con-
tribute to the acceptance, assimilation and success of online education including course design,
course content support, instructor’s personal characteristics and students’ familiarity with and
access to technical resources [20–22]. Due to sudden conversion to online instructions, caused
be COVID-19, faculty and students at academic institutions, mainly focused on traditional
face-to-face instruction, encountered various challenges. As the pandemic continues, a small
body of literature on educational impact of COVID-19 is starting to emerge. A group of inves-
tigators conducted a U.S. nationwide survey study among faculty and students of STEM fields
in June 2020. Their results highlighted the gender disparities in online learning during pan-
demic: female faculty and students reported more challenges in technological issues and adapt-
ing to remote learning compared with their male peers [12].They also found out that 35.5% of
doctoral students, 18.0% of master’s students and 7.6% of undergraduate students would have
a delayed graduation due to pandemic [11]. Hispanic and Black undergraduates were two
times and 1.7 times more likely, respectively, to delay graduation relative to Whites.
Dhawan presented a comprehensive literature review on the existing pedagogical approaches
for the online instruction while identifying the strengths, weaknesses and challenges of adopting
each approach for the online education during the COVID-19 pandemic [9].
Vielma and Brey conducted a qualitative surveying from 170 students who took asynchro-
nous classes within two engineering departments (biomedical engineering and chemical engi-
neering) at a U.S. Hispanic-serving institution [10]. The goal was to assess the effectiveness of
their online education during pandemic. Their results indicated the students’ need in having
synchronous instructional content (in addition to asynchronous content) to enhance the social
component of learning.
Almaiah et al. conducted a semi-structured interview (using a list of general topics as inter-
view guideline instead of a structured list of questions) with 30 students and 31 experts in the
PLOS ONE | https://doi.org/10.1371/journal.pone.0250041 April 15, 2021
3 / 17
PLOS ONE Engineering online education during COVID-19 pandemic
field of information technology from six universities in Jordan and Saudi Arabia. Their goal
was to identify the challenges that impede the successful employment of online education dur-
ing pandemic in developing countries and provide educational stakeholders with useful guide-
lines to enhance education efficacy.
Our work conducts a thorough (quantitative and qualitive) analysis of challenges and fac-
tors affecting the online education of engineering courses by conducting surveys among stu-
dents and faculty members from various engineering subfields at one of the largest and most
diverse four-year U.S universities (CSULB). Thus, the presented work has several unique
aspects that distinguish it from the few existing studies focused on online education during
pandemic, such as the use of both quantitative and qualitative survey questions, and participa-
tion of large number of engineering students and faculty from various subfields and diverse
backgrounds. Our observational study provides empirical evidence for various solutions we
propose to enhance online engineering education during and post-pandemic, especially for
those universities with limited resources, or with a large population of minority, first-genera-
tion and low-income students.
2. Materials and methods
2.1. Engineering education at CSULB
California State University, Long Beach (CSULB) is one of the largest and most diverse four-
year universities in the U.S. Approximately 52% of CSULB student body are NSF-defined
underrepresented minority including 59.2% female, 46.9% Hispanic, 4.5% African American
and 1% Native American [23]. As a result, CSULB is recognized as a minority serving institu-
tion: namely Hispanic, Asian American, Native American, and Pacific Islander-Serving Insti-
tution. Also, more than half of our students are low-income or first-generation college
students. CSULB College of Engineering (COE) currently has more than 250 faculty and 5000
students (undergraduate and graduate). COE offers a total of 11 programs that are hosted by
six departments: Biomedical Engineering (BME), Chemical Engineering (CHE), Civil Engi-
neering & Construction Engineering Management (CECEM), Computer Engineering & Com-
puter Science (CECS), Electrical Engineering (EE), and Mechanical & Aerospace Engineering
(MAE). The majority of the courses in COE were offered face-to-face prior to pandemic. Since
2010, CSULB has been using an LMS called BeachBoard (BB) ̶ a customized version of "Bright-
space" platform developed and supported by "Desire 2 Learn" company. BB provides various
features to facilitate the course instruction, including a robust platform for communication
between the instructor and students, sharing course materials and recorded lectures with stu-
dents, discussion forums, design and management of assessments, assignments and grades.
Prior to pandemic, while some CSULB faculty members had been employing (at least some of)
BB features (e.g. gradebook) for their instruction on a regular basis, many others had opted
out as its usage has not been mandatory.
The unprecedented circumstances of global COVID-19 pandemic forced the swift conver-
sion of the mode of instruction from face-to-face to fully online for all CSULB engineering
programs (including 349 courses for a total of 1004 sections) within a transitional period of 10
days in March 2020. Hence, the teaching materials and assessment methods had to be devel-
oped “on the fly”. CSULB advised instructors to mainly focus on learning/using BB (and
Zoom videoconferencing) to convert their instructions to the online format. This recommen-
dation seemed reasonable given the availability and practicality of BB features. However, both
our students and faculty encountered various challenges during the online instruction in
Spring 2020. By the end of the semester in May 2020, CSULB announced that Fall 2020 semes-
ter was also going to be in the alternative mode of instructions. Thus, 313 engineering courses
PLOS ONE | https://doi.org/10.1371/journal.pone.0250041 April 15, 2021
4 / 17
PLOS ONE Engineering online education during COVID-19 pandemic
were scheduled to be offered in synchronous fully online format. 18 additional classes were
exempted and offered in hybrid/blended format. These were the classes where the face-to-face
component is considered essential to meet the course learning outcomes and therefore could
not be conducted fully online, (e.g. laboratories and senior design capstone projects).
2.2. Surveys
Our goal was to identify and study the magnitude of various issues that our faculty and stu-
dents encountered during the six weeks of online instruction in Spring 2020 (March 23-May
8) and plan for an enhanced online instruction in Fall 2020. The faculty and student surveys
were designed holistically considering the overall verbal feedback received from stakeholders
during the Spring 2020 online instruction. The faculty survey consisted of 10 multiple-choice
and 2 free-response questions, while student survey included 8 multiple-choice questions with
fill-in or additional comment options for each question.
The faculty survey questions covered a variety of online teaching issues including, but not
limited to, the lack of access to necessary hardware (e.g. computer/tablet, stylus, scanner/
printer, microphone/headset, camera), software and reliable internet connection. Some ques-
tions focused on various learning assessment methods that instructors used in Spring 2020 (or
the ones they were planning to use in Fall 2020) including open-book or closed-book exams,
synchronous or asynchronous exams, fully-online exam (using randomized questions on BB)
or semi-online exams (where students solve the assigned problems on a paper, then scan and
upload their solutions on BB). Some questions focused on proctoring exams and the instruc-
tors’ perceived prevalence of cheating/plagiarism. Faculty were also asked to indicate the topics
that they were interested to enhance their skills on, e.g., basic or advanced BB features, Zoom
features, automatic grading, etc. The two open-ended questions provided instructors addi-
tional opportunities to comment about their online teaching experience and make any sugges-
tion or request to COE that could help with improvement of online instruction in Fall 2020.
The student survey was designed to identify the challenges students confronted during
online instruction in Spring 2020, including lack of access to hardware, software, reliable inter-
net connection, quiet/private space to study, potential issues of balancing study with work and
family duties, and stress management. The students were also asked about the difficulties they
had during the synchronous classes on Zoom (e.g., lack of focus or engagement, instructor’s
lack of familiarity with technology) or during the online exams (e.g., time management, issues
with methods of proctoring using camera). Copies of faculty and student surveys are enclosed
in the S1 Appendix for the readers’ further reference.
3. Results
The faculty survey was conducted using Qualtrics over a three-week period (June 20-July 10).
Similarly, the student survey was designed and conducted in Qualtrics afterwards (July
27-August 12). This later timeframe was decided based on the assumption that more students
(including the incoming students) might be available to participate in the survey closer to the
beginning of the Fall 2020 semester (August 21). Participation in both surveys were
anonymous.
A total of 110 instructors took the survey where 43% of them were full-time including ten-
ured/tenure track faculties and the rest were part-time lecturers. Also, 627 students partici-
pated in the survey: First-year students (4%), Sophomore (14%), Junior (30%), Seniors (35%)
and graduate students (17%). Fig 1 shows the distribution of survey participants among vari-
ous departments within the COE (question #1 on both surveys). We observe that all depart-
ments have relatively similar representations in terms of percentage of faculty and student
PLOS ONE | https://doi.org/10.1371/journal.pone.0250041 April 15, 2021
5 / 17
PLOS ONE Engineering online education during COVID-19 pandemic
Fig 1. Distribution of the survey participants among various departments within the college of Engineering: (A) Faculty participants; (B)
Student participants.
https://doi.org/10.1371/journal.pone.0250041.g001
participants in respective surveys (9% BME, 5–10% CHE, 15–23% CECEM, 19–22% CECS,
18–22% EE, and 21–26% MAE).
These percentages are consistent with the size of our departments in terms of total number
of faculty and students. Therefore, our survey sample population could be a good representa-
tive of the general COE populations in terms of existing majors.
3.1. Logistical challenges for both students and faculty
Fig 2 shows the percentages of survey respondents who indicated various logistical challenges
they had during the online instruction period of Spring 2020 (question #3 on the faculty survey
and question #3 on the student survey). Close to 15% of the faculty had issues with software
license or no access to personal computer/tablet. About 20% of the faculty did not have access
to microphone/headset or printer/scanner. 23% of faculty had no reliable internet connection,
while 32% had no access to webcam or camera for the online instruction. Finally, 47% of the
faculty indicated that they had no access to or had technical difficulties with online writing
tools. Among the student respondents, 1% had no access to any computer/tablet, while close
to 5% had only access to a shared computer at home. 3% had no internet connection, while
26% had issues with reliability of their internet. 28% indicated having issues with software
access, while 26% had no printer/scanner at home.
3.2. Students challenges with online instruction
Fig 3 summarizes the prevalence of challenges students had with online instruction during
Spring 2020 (questions # 3–6 on the student survey). About 70% of students indicated diffi-
culty in maintaining their focus or experiencing Zoom fatigue after attending multiple online
sessions. 55% of students felt social disconnection from their classmates/peers, while 64% did
not feel engaged during the online classes. 60% of the students felt there was a lack of clear
guidance or communication from the instructors. Also, a quarter of students had issues with
online submission of assignments and exams, mainly due to the lack of access to printer/scan-
ner as we learned from students’ optional comments. About 40% of students had technical
PLOS ONE | https://doi.org/10.1371/journal.pone.0250041 April 15, 2021
6 / 17
PLOS ONE Engineering online education during COVID-19 pandemic
Fig 2. Logistical challenges of online instruction from perspectives of faculty and students. The horizontal access represents the percentage of survey participants who
indicated the corresponding challenge. (A) Faculty respondents; (B) Student respondents.
https://doi.org/10.1371/journal.pone.0250041.g002
difficulty and ineptness issues with using or navigating through Zoom or BB. 48% of the stu-
dents experienced time management issues during the online exams. In optional comments,
some students expressed their frustration with not being able to go back to previous questions
(a BB feature for the instructors to limit cheating). 23% of the students indicated that the
unavailability of the instructor during the online exam (in contrast to in-person exam) caused
challenges.
48% of the students specified that they either do not have camera or feel uncomfortable
turning the camera/microphone on during the class or online exams (question #7 on the stu-
dent survey). Optional comments revealed that many participants have privacy concerns with
Fig 3. Prevalence of challenges students encountered during online instruction in Spring 2020.
https://doi.org/10.1371/journal.pone.0250041.g003
PLOS ONE | https://doi.org/10.1371/journal.pone.0250041 April 15, 2021
7 / 17
PLOS ONE Engineering online education during COVID-19 pandemic
usage of camera/microphone or being recorded, especially if they were living in a crowded
home or shared space. Furthermore, some students experienced an increased level of anxiety
being watched on camera that hindered their focus and lowered their performance during the
online exams. 28% of the students indicated that they had difficulty with balancing work and
study. From the optional comments, we understood that the latter issue has been escalated for
many during pandemic. Some parents had lost their jobs and consequently the whole family
was relying on the part-time jobs of the younger adults (students) to survive financially.
Our survey also indicated that more than 50% of our students did not have access to a pri-
vate or quiet space to attend the online classes or to study. 55% of students also lacked motiva-
tion to study (question #3 on the student survey). The optional comments shed further light
onto the lack of motivation: the uncertainty of the COVID-19 pandemic and loss of peer inter-
action/support were identified as the major contributing factors. Finally, 24% of the students
rated their overall experience of online instruction (question #8 on the student survey) as satis-
fying, 37% found it dissatisfying, while the rest (39%) were neutral.
3.3. Assessment methods used during emergency online instruction
Table 1 shows the prevalence of various methods that the faculty used to assess students’ learn-
ing during the online instruction of Spring 2020. Semi-online refers to an exam where students
solve the assigned problems on a paper, then scan and upload their solutions. Asynchronous
exam refers to a take-home exam while a synchronous exam is the one conducted during the
scheduled class or exam time. The survey allowed respondents to choose more than one assess-
ment method per question (because faculty might have taught multiple classes, held more than
one exam during the semester or applied multiple assessment methods in the same class), thus
the sum of the percentages would not equal to 100.
We observe that the fully online exams such as the BB quizzes were used by 63% of the fac-
ulty. BB quizzes provides the faculty with the convenient option of randomizing the order
and/or the parameter values of the questions. The instructor can also limit the view to one
question per page for students and prevent them from going back to previous questions. The
effectiveness of these options in limiting cheating/ plagiarism, and consequently the reduced
need for further proctoring, might have contributed to the high popularity of this assessment
method among the faculty.
The remaining assessment methods in the decreasing order of their prevalence were proj-
ect/term paper (50%), semi-online synchronous exam (40%), oral presentation/exam (33%),
and semi-online asynchronous exam (28%). Our survey also revealed that 70% of the faculty
used the open-book/open-note exam while 33% tried closed-book/closed note exams. The
preference of open-book/open-note exam among faculty could be also justified by the
Table 1. Learning assessment methods faculty used during the online instruction in Spring 2020. The respondents
could choose more than one option for each question depending on the number of exams administered during the
semester.
Survey Question #
Assessment method
Percentage of faculty who employed the method
Question #4
Fully online exam (e.g., BB quiz)
Semi-online Asynchronous exam
Semi-online Synchronous exam
Project/term paper
Oral presentation/demo
Question #7
Open-book/Open-note exam
Closed-book/Closed-note exam
https://doi.org/10.1371/journal.pone.0250041.t001
63%
28%
40%
50%
33%
70%
33%
PLOS ONE | https://doi.org/10.1371/journal.pone.0250041 April 15, 2021
8 / 17
PLOS ONE Engineering online education during COVID-19 pandemic
decreased need for proctoring tools. In fact, our data (faculty survey question #7) revealed that
among those faculty who employed open-book/open-note exam, only 27% used Zoom camera
and microphone for proctoring of the exam. 21% used lockdown browsers (e.g. respondus),
while 61% did not have any proctoring. However, when the exams were closed-book/closed-
note, 56% of the faculty decided to proctor the exam using Zoom camera and microphone,
18% chose to use the lockdown browsers and 35% did not proctor. We also evaluated the asso-
ciation of instructors’ perception of cheating/plagiarism with various assessment methods by
calculating the Pearson correlation of faculty’s assessment methods with their trichotomized
perception of online cheating (less cheating, the same, more cheating) relative to that of face-
to-face (faculty survey question #9). The results revealed no statistically significant correlation
between perception of cheating and assessment methods except for the following: Semi-online
asynchronous exam (correlation = 0.23, p-value = 0.01) and Closed-note/Closed-book (corre-
lation = 0.21, p-value = 0.03). This data analysis shows that semi-online asynchronous and
closed-book exams were associated with an increase in the perceived cheating,
3.4. Perceived faculty skills that needed enhancement
Faculty indicated various topics that they were interested to enhance their skills in, as summa-
rized in Table 2.
About 60% of the faculty needed to learn about the advanced features of BB (e.g. how to create
online surveys or make quizzes with randomized questions or personalized parameter values).
Also, more than half of the faculty were interested in learning about semi-automatic grading tools
(e.g. Gradescope). Close to 40% of the faculty needed to learn how to create a syllabus for an online
class or become more competent with using Zoom features. A similar percentage of participants
indicated interest in enhancing their multimedia skills (e.g. working with Kaltura Capture, Camta-
sia or Snagit). Finally, 26% of the faculty needed more training to become familiar with basic fea-
tures of BB. In the optional comments (faculty survey questions #11–12), some faculty members
expressed their concerns about the delivery of the hands-on components of their courses and
requested some general guideline on how to address this issue for an online instruction.
4. Discussion
In this section, we will discuss the challenges we identified and propose relevant interventions
to improve the online delivery of engineering courses during pandemic.
Table 2. A list of topics identified by faculty for further skill enhancement. Respondents could choose as many top-
ics as they were interested to learn.
Survey
Question #
Topics
Percentage of faculty
interested
Question #10
The major requirements of syllabus for an online course
Basic BB features: How to create/modify/improve BB for my course
More advanced BB features: How to create online surveys /discussion
groups/quizzes that reduce the potential of cheating, how to
automatically export grades to BB gradebook, how to use Master Shell in
BB, etc.
Zoom features (basic and advanced): How to schedule/record a meeting,
how to use Zoom’s Whiteboard or OneNote, how to do breakout rooms,
etc.
Multimedia skills: How to create interactive multimedia files using
Kaltura Capture, Camtasia or Snagit, how to use Alt captions in media
you generate (Word, PPT, page in BB) to facilitate accessibility
38%
26%
58%
39%
39%
Assessment: How to use automatic grading tools (e.g. Gradescope)
54%
https://doi.org/10.1371/journal.pone.0250041.t002
PLOS ONE | https://doi.org/10.1371/journal.pone.0250041 April 15, 2021
9 / 17
PLOS ONE Engineering online education during COVID-19 pandemic
4.1. Student challenges
Our results revealed that a quarter of our students did not have access to reliable internet con-
nection, triggering a concern about widening of the digital equity gap among students due to
COVID-19 pandemic. With COVID-19 and the abrupt transition to online teaching, access to
reliable internet connection and personal computer/tablet have become major factors affecting
the learning outcomes for students. To address this issue, institution can provide WiFi access
on campus’s open areas and well-ventilated buildings while monitoring for social distancing
and sanitizing the surfaces frequently. For those who require computing devices, a loaner pro-
gram can be implemented where students can borrow laptops for a certain period of time to
access the course materials and complete the course requirements. The institution can also
provide a virtual desktop environment for students to access all necessary software. Using free
scanning applications on smartphones or tablets can address the lack of access to scanners.
Our survey also indicated that about 30% of engineering students had work-life balance
issues, while 55% of students lacked motivation, and 50% did not have access to a private space
to attend classes. These results are consistent with those reported in a recent study conducted
at Biomedical and Chemical Engineering departments of a Hispanic-serving institution [10].
While the percentage of our students who had issues with lack of motivation or private space
seemed to be higher, both studies highlight the necessity of providing more socio-emotional
support for students during the difficult times of pandemic.
Students identified various challenges they experienced in online synchronous instruction
of courses through Zoom including lack of peer-support/interaction, focus, engagement, and
clear guideline from instructors. They also indicated difficulties with time management and
Zoom fatigue. Peer-support/interaction has shown to improve the success rate of students
especially those from underrepresented groups [24]. Lack of peer-support during the online
instruction in the COVID-19 era negatively affects the motivation of the students. However,
the remaining raised issues could be addressed in part by employing appropriate teaching
techniques by faculty as follows: breaking down a long lecture into shorter segments with
more frequent breaks, encouraging group discussion among students, making themselves
available during the exams, providing students with a clear roadmap for the online course,
making the recordings of the live lectures available after the lecture is over. The latest would
help struggling students to learn at their own pace [10]. To assist with the time management
issue during the exams, faculty can design practice exams to allow students to familiarize
themselves with the questions’ setup and adapt with the exam’s style before the actual exam.
Pandemic has caused educational loss, delayed graduations, cancelled internships and lost
job offers. The new generation of students who have been away from face-to-face instructions
may lack certain learning experiences. For example, there might be a generation of engineering
students who performed the majority of their lab activities virtually and thus, lacks true hands-
on skills. While the pandemic educational gap will affect everyone, it is likely to impact under-
privileged students (e.g. first generation, low income or care givers) more profoundly [25]. As
a result, the socioeconomic factors would constitute key mediators in explaining the poten-
tially large and heterogeneous educational gap. This gap may have long-lasting implications
for income inequality and health disparities [26].
To reduce the educational gap, universities could adopt the practice of developing and
implementing diagnostic tools to learn where and how large the deficiencies are. Based on the
acquired knowledge, they could offer short remediation programs with long-term reorienta-
tion of curriculum to align with student’s learning levels [27]. For example, a summer session
that deals with hands-on aspects of lab safety or experimentations could be implemented. In
some cases, close coordination between the instructors who teach the courses in a sequence
PLOS ONE | https://doi.org/10.1371/journal.pone.0250041 April 15, 2021
10 / 17
PLOS ONE Engineering online education during COVID-19 pandemic
may be required, so they can develop extracurricular materials or propose activities that would
help students bridge a gap in a specific topic. As the pandemic progresses, the flexibility of uni-
versity policies could also help with narrowing down the educational gap especially for those
students with lower socioeconomical status. Allowing students to adjust their course load, tim-
ing of assignments and tuition payment schedule would enable them to make reactive deci-
sions to mitigate the educational loss [25]. A need for further research on this top is
undeniable.
4.2. Faculty challenges
Establishment of institutional quality standards related to online education is of paramount
importance in online education. Effective communication is the key factor in bridging the
divide and reconciling administrator and faculty for an enhanced online education [28]. A
considerable number of our faculty reported lack of access to hardware, software and necessary
tools for online instruction. Especially, in the absence of traditional in-class whiteboard, many
faculty members indicated lacking an online writing tool. This issue can be addressed by insti-
tution’s budget allocation to acquire necessary hardware and tools (e.g. personal computer/
tablet with web camera, digital pen for touch screen devices, digital clipboard, document
camera).
Development of online learning assessment methods as rigorous as in conventional face-to-
face setting to prevent cheating/plagiarism is not straightforward [16, 29]. While one cannot
propose a single assessment method that would work ideally for all engineering courses and
classroom sizes, it would still be interesting to study how various online exams and assessment
methods (e.g. online quiz tools within the LMS, open-book or take-home examinations, stu-
dent presentations, peer-reviewed activities, cooperative quizzes [30], oral assessments [31],
course summary papers or online portfolios) stack up against each other. Since the onset of
pandemic, a limited number of studies (mainly within the fields outside the engineering) have
been conducted to evaluate the successes and challenges of the online assessments. The study
in [32] revealed that although the majority of undergraduate Management students required
more time and effort to prepare for the online exams (compared to the traditional exams),
they regarded the clarity and prompt grading and feedback features of the online exams sub-
stantially advantageous. Another recent study revealed that cheating remains one of the major
concerns for the online examinations and needs to be addressed using available techniques
including online proctoring and randomizations of the exam questions [33]. Few other studies
showed that the online examinations increased the level of stress and anxiety among medical
students [34, 35]. The added stress was in part caused by the lack of a robust examination plat-
form (i.e., reliable LMS) as well as not providing students with sample online practice exams.
Finally, a survey conducted among civil engineering students showed that high-achieving stu-
dents performed significantly better than low-achieving students in online examinations and
there was a significant increase in the students’ dropout rate in the 2020–2021 academic year
relative to the previous ones [36].
Our student survey results indicated that the use of camera/microphone to proctor the
online exams can raise equity concerns (for those who do not have access to camera and can-
not afford it) and privacy concern (for monitoring students’ private space). To address these
valid concerns, faculty are advised to choose alternative methods for reducing cheating during
online exams. Randomizing the exam questions by shuffling both the problem statements and
the multiple choices, and randomly selecting a subset of questions from a question library with
individualized/randomized input variables are viable practical solutions. Fortunately, most
LMS provide these options. However, although 99% of postsecondary US institutions have an
PLOS ONE | https://doi.org/10.1371/journal.pone.0250041 April 15, 2021
11 / 17
PLOS ONE Engineering online education during COVID-19 pandemic
LMS in use, only approximately half of faculty at those institutions have been using it on a regu-
lar basis [37]. As a result, many faculty members were not familiar with the basic or advances
features of the LMS or other tools for effective online instruction. Our survey result confirmed
this observation. In fact, our faculty identified a broad range of topics related to BB or other
online teaching tools that they felt the need to enhance their skills in. Institutions could address
this issue by organizing training workshops, webinars, short-courses, and discussion panels for
the faculty to enhance their online teaching skills. At CSULB, stipends were offered in summer
2020 to further incentivize faculty participation in these professional development programs.
Hands-on training is an integral component of engineering education. Following the
abrupt conversion of classes to the online format in Spring 2020, many instructors adopted
simulations or processing of already acquired data for engineering students to complete their
course projects. Our survey indicated the faculty’s need to learn about additional effective
ways for providing hands-on training/experience. Depending on the content of the course,
employment of “home lab kits” and recording of the lab experiments could partially help.
However, design, preparation, distribution/collection of the lab kits or recording of the experi-
ments can be extremely time consuming for faculty especially given all the access restrictions
to on-campus labs and additional safety precautions imposed by COVID-19 pandemic. Virtual
labs might be a more effective solution. Additionally, remotely accessible labs where the exper-
iment setup is on campus and students use tools for remote control and managing of the setup
can be employed, whenever possible [10].
4.3. Summary of proposed interventions
From the analysis of the survey results we propose several intervention strategies that can be
employed by stakeholders at different levels to improve the online instruction of engineering
courses. The proposed strategies (the targeted issues and the survey questions that identified
them) are summarized as follows:
➢ Strategies for institution/engineering administration
• Budget allocation to provide basic equipment for the online instruction to both faculty and
students in need. Examples of such equipment include personal computer/tablet preferably
with webcam/camera, online writing tool, reliable internet connection (to address the logisti-
cal challenges indicated by students and faculty in response to question # 3 of both surveys)
• Creating a virtual desktop environment and allowing faculty and students to access neces-
sary software (addressing technical access challenges of online instruction indicated in
response to questions # 3, #7 and # 11 from the faculty survey, and question #5 from the
student survey)
• Organizing training workshops for faculty/students to further familiarize with online
teaching/learning technology and tools (addressing technical skills that were indicated in
response to question #10 of the faculty survey and question #5 of the student survey)
• Providing a syllabus template for online courses including all the important information
needed for ABET accreditation (addressing lack of clear communication or instruction
indicated in response to question #10 of the faculty survey and question #5 of the student
survey)
• Development and organization of systematic repository of resources pertinent to engineer-
ing online instruction (to enhance the faculty’s online teaching skills as the need was indi-
cated in response to questions #10–12 of the faculty survey)
PLOS ONE | https://doi.org/10.1371/journal.pone.0250041 April 15, 2021
12 / 17
PLOS ONE Engineering online education during COVID-19 pandemic
➢ Strategies for engineering faculty
• Leveraging on the institution’s LMS to manage the course, grades, forum discussions and
exams (to enhance the faculty’s online teaching skills as the need was indicated in response
to questions #10–12 of the faculty survey)
• Breaking down a long lecture into shorter segments with more frequent breaks (addressing
Zoom fatigue indicated in response to question #4 of the student survey)
• Encouraging group discussion or problem-solving activities among students such as Zoom
breakout rooms (addressing the lack of social interactions with peers as indicated in
response to question # 4 of the student survey).
• Being available during the exams (e.g. on Zoom) to answer students’ questions (addressing
the lack of access to the instructors during exams as indicated in response to question # 4
of the student survey).
• Providing students with a clear roadmap and instruction for the online course (addressing
lack of clear communication or instruction indicated in response to question #5 of the stu-
dent survey)
• Making the recordings of the live lectures available after the lecture (addressing online
instruction challenges and lack of access to reliable internet indicated in response to ques-
tion #4 and question #3 of the student surveys, respectively)
• Administering practice exams for students (addressing issues with the online exams indi-
cated in response to question #6 of the student survey)
• Using open-book/open-note and synchronous assessment methods that support academic
integrity. Examples include randomized questions/ restricted time/ question pools on
LMS. (addressing the challenges with online assessment methods indicated in response to
questions # 4, #7–9 of the faculty survey)
• Avoiding using camera/microphone to proctor exams (addressing privacy issues with the
indicated in response to question #7 of the student survey)
• Employment of “home lab kits”, recording of the hands-on experiments and virtual labs to
partially address the hands-on training aspect of the course (enhancing online instruction
as indicated in response to questions # 11–12 of the faculty survey)
➢ Strategies for engineering students
• Using free scanning applications on their smartphones (addressing lack of access to scan-
ner as indicated in response to questions # 6 of the student survey).
Most of the proposed solutions were implemented at the CSULB college of Engineering in
preparation for Fall 2020 semester. Our future work will include evaluation of the efficacy of
the implemented interventions by conducting a post-intervention survey at the end of Spring
2021 semester.
This work contributes to the developing body of knowledge about the effect of pandemic
on engineering education by investigating the challenges and obstacles faced by a large group
of engineering students and faculty at CSULB which exemplifies an institution that previously
taught face-to-face engineering classes (predominantly), with a large minority population and
socio-economic gap. The recommended strategies for various educational stakeholders
(including students, faculty and administration) aims to fill the tools and technology gap,
PLOS ONE | https://doi.org/10.1371/journal.pone.0250041 April 15, 2021
13 / 17
PLOS ONE Engineering online education during COVID-19 pandemic
enhance faculty skills in teaching online courses by taking full advantage of online learning
management tools, and finally, propose effective assessment methods for online courses while
considering the potential equity and privacy issues. These recommendations are practical
approaches for many similar institutions around the world and would help improve the learn-
ing outcomes of online educations in all fields of engineering.
4.4. Potential limitations of the study
Some limitations should be addressed in this study. We investigated the challenges of engi-
neering online education during Spring 2020 when the pandemic started, and a global emer-
gency occurred. Thus, the reported experiences and perceptions might have been affected by
confounding factors related to the onset of pandemic. As the pandemic continues and various
academic stakeholders explore and learn about new strategies to better adjust to the new nor-
mal, subsequent studies conducted in the near future might provide a more accurate picture of
the online engineering education.
We advertised the surveys to all faculty and students of the CSULB college of Engineering
by sending announcement emails to their university email accounts in summer 2020. While
the faculty survey’s response rate was 44%, the student survey’s response rate was 12%. The
low response rate of the students might have introduced some participation bias to the results.
Our main goal of conducting the surveys was to identify the urgent needs and challenges of
the general body of our students and faculty without focusing on any specific underrepre-
sented groups. Our assumption was that the demographics of survey participants are likely
proportional to those of the college of Engineering. Further studies with inclusion of race, gen-
der and socioeconomics demographics are needed to investigate the magnitude of educational
challenges that underrepresented groups experienced during the pandemic in comparison
with other groups. Consideration of some institutional data (e.g. grades, faculty/ student per-
ception of learning, financial aid requests) from both pre- and during pandemic would
enhance the study, as well.
The current work did not evaluate the degree of effectiveness and sustainability of each con-
ducted intervention. It also did not compare the efficacy of various alternative assessment
methods for engineering online education. A follow-up study is needed to address these
limitations.
5. Conclusion
We conducted an observational study to identify challenges encountered due to abrupt transi-
tion to online instruction of engineering courses during COVID-19 pandemic by surveying
(quantitively and qualitatively) students and faculty at our minority-serving institution. Vari-
ous logistical, technical and learning/teaching issues were identified, and several interventions
were proposed to address them. The results of this study add to the developing body of knowl-
edge about the effect of pandemic on engineering education. This study also provides empiri-
cal evidence for the proposed strategies to enhance (and consequently further promote) the
online engineering education during and post-pandemic. Our future work will include a thor-
ough study on evaluating the efficacy and sustainability of each proposed intervention.
Supporting information
S1 Appendix. Questionnaire for both student and faculty surveys.
(DOCX)
PLOS ONE | https://doi.org/10.1371/journal.pone.0250041 April 15, 2021
14 / 17
PLOS ONE Engineering online education during COVID-19 pandemic
S1 Data. Students survey data in response to multiple choice questions.
(XLSX)
S2 Data. Faculty survey data in response to multiple choice questions.
(XLSX)
Acknowledgments
The authors would like to thank Dr. Daniel Whisler, Dr. Shabnam Sodagari and Ms. Asieh
Jalali-Farahani for their help with designing the surveys.
Author Contributions
Conceptualization: Shadnaz Asgari, Jelena Trajkovic, Mehran Rahmani, Wenlu Zhang, Roger
C. Lo, Antonella Sciortino.
Formal analysis: Shadnaz Asgari.
Investigation: Shadnaz Asgari, Jelena Trajkovic, Mehran Rahmani, Wenlu Zhang, Roger C.
Lo.
Methodology: Shadnaz Asgari.
Supervision: Antonella Sciortino.
Visualization: Shadnaz Asgari.
Writing – original draft: Shadnaz Asgari, Jelena Trajkovic.
Writing – review & editing: Shadnaz Asgari, Jelena Trajkovic, Mehran Rahmani, Wenlu
Zhang, Roger C. Lo, Antonella Sciortino.
References
1. Bourne J, Harris D, Mayadas F (2005) Online engineering education: Learning anywhere, anytime.
Journal of Engineering Education 94: 131–146.
2.
Lima RM, Andersson PH, Saalman E (2017) Active Learning in Engineering Education: a (re)introduc-
tion. European Journal of Engineering Education 42: 1–4.
3. Bishop JL, Verleger MA. The flipped classroom: A survey of the research; 2013. pp. 1–18.
4. Mills JE, Treagust DF (2003) Engineering education—Is problem-based or project-based learning the
answer. Australasian journal of engineering education 3: 2–16.
5. Asgari S, Englert B (2014) Teaching Pattern Recognition: A Multidisciplinary Experience. American
Society of Engineering Education (ASEE) Conference- Zone IV. Long Beach, CA. pp. 44–52.
6. Asgari S, Penzenstadler B, Monge A, Richardson D. Computing to Change the World for the Better: A
Research-Focused Workshop for Women; 2020; Portland, Oregon, USA. IEEE.
7. Martı´nez PJ, Aguilar FJ, Ortiz M (2019) Transitioning from face-to-face to blended and full online learn-
ing engineering master’s program. IEEE Transactions on Education 63: 2–9.
8. COVID-19 Impact on Education. https://en.unesco.org/covid19/educationresponse United Nations
Education, Scientific and Cultural Organization (UNESCO).
9. Dhawan S (2020) Online learning: A panacea in the time of COVID-19 crisis. Journal of Educational
Technology Systems 49: 5–22.
10. Vielma K, Brey EM (2020) Using Evaluative Data to Assess Virtual Learning Experiences for Students
During COVID-19. Biomedical Engineering Education: 1–6.
11. Saw GK, Chang C, Lomelı´ U, Zhi M (June 2020) Fall Enrollment and Delayed Graduation Among
STEM Students during the COVID-19 Pandemic. https://nreeducation.wordpress.com Network for
Research and Evaluation in Education (NREED) Data brief. pp. 1–5 p.
PLOS ONE | https://doi.org/10.1371/journal.pone.0250041 April 15, 2021
15 / 17
PLOS ONE Engineering online education during COVID-19 pandemic
12. Saw GK, Chang C, Lomelı´ U, Zhi M (August 2020) Gender Disparities in Remote Learning during the
COVID-19 Pandemic: A National Survey of STEM Faculty and Students. https://nreeducation.
wordpress.com Network for Research and Evaluation in Education (NREED) Data brief. 1–5 p.
13. Ryan TG, Toye M, Charron K, Park G (2012) Learning management system migration: An analysis of
stakeholder perspectives. The International Review of Research in Open and Distributed Learning 13:
220–237.
14.
Jaggars SS (2014) Choosing between online and face-to-face courses: Community college student
voices. American Journal of Distance Education 28: 27–38.
15. Holzweiss PC, Joyner SA, Fuller MB, Henderson S, Young R (2014) Online graduate students’ percep-
tions of best learning experiences. Distance education 35: 311–323.
16.
Lee-Post A, Hapke H (2017) Online learning integrity approaches: Current practices and future solu-
tions. Online Learning 21: 135–145.
17. Moskal P, Dziuban C, Hartman J (2013) Blended learning: A dangerous idea? The Internet and Higher
Education 18: 15–23.
18. Gold S (2001) A constructivist approach to online training for online teachers. Journal of Asynchronous
Learning Networks 5: 35–57.
19. Kukulska-Hulme A (2012) How should the higher education workforce adapt to advancements in tech-
nology for teaching and learning? The Internet and Higher Education 15: 247–254.
20. Almaiah MA, Alyoussef IY (2019) Analysis of the effect of course design, course content support,
course assessment and instructor characteristics on the actual use of E-learning system. IEEE Access
7: 171907–171922.
21. Al-Gahtani SS (2016) Empirical investigation of e-learning acceptance and assimilation: A structural
equation model. Applied Computing and Informatics 12: 27–50.
22. Almaiah MA, Alamri MM, Al-Rahmi WM (2019) Analysis the effect of different factors on the develop-
ment of Mobile learning applications at different stages of usage. IEEE Access 8: 16139–16154.
23.
2019 CSULB Institutional Data. https://www.csulb.edu/institutional-research-analytics California State
University, Long Beach.
24. Williams SN, Thakore BK, McGee R (2017) Providing social support for underrepresented racial and
ethnic minority PhD students in the biomedical sciences: a career coaching model. CBE—Life Sciences
Education 16: ar64. https://doi.org/10.1187/cbe.17-01-0021 PMID: 29196425
25. Aucejo EM, French J, Ugalde Araya MP, Zafar B (2020) The impact of COVID-19 on student experi-
ences and expectations: Evidence from a survey. Journal of Public Economics 191: 104271. https://
doi.org/10.1016/j.jpubeco.2020.104271 PMID: 32873994
26. Kuhfeld M, Soland J, Tarasawa B, Johnson A, Ruzek E, et al. (2020) Projecting the potential impact of
COVID-19 school closures on academic achievement. Educational Researcher 49: 549–565.
27. Kaffenberger M (2021) Modelling the long-run learning impact of the Covid-19 learning shock: Actions
to (more than) mitigate loss. International Journal of Educational Development 81: 102326. https://doi.
org/10.1016/j.ijedudev.2020.102326 PMID: 33716394
28. Wickersham LE, McElhany JA (2010) Bridging the divide: Reconciling administrator and faculty con-
cerns regarding online education. Quarterly Review of Distance Education 11: 1.
29. Baron J, Crooks SM (2005) Academic integrity in web based distance education. TechTrends 49: 40–
45.
30. Slusser SR, Erickson RJ (2006) Group quizzes: an extension of the collaborative learning process.
Teaching Sociology 34: 249–262.
31.
32.
Jensen BB (2010) Oral assessment in engineering education. International Journal of Electrical Engi-
neering Education 47: 375–379.
Tilak P, Deshmukh M, Phadake S (2020) A survey on online examination during COVID 19 pandemic:
Perception of Management Students. Mukt Shabd Journal 9: 284–290.
33. Hassan B, Shati AA, Alamri A, Patel A, Asseri AA, et al. (2020) Online assessment for the final year
medical students during COVID-19 pandemics; the exam quality and students’ performance. Onkologia
i Radioterapia 14: 1–6.
34. Elsalem L, Al-Azzam N, Jum’ah AA, Obeidat N, Sindiani AM, et al. (2020) Stress and behavioral
changes with remote E-exams during the Covid-19 pandemic: A cross-sectional study among under-
graduates of medical sciences. Annals of Medicine and Surgery 60: 271–279. https://doi.org/10.1016/j.
amsu.2020.10.058 PMID: 33163179
35. Mastour H, Ghalibaf AM, Niroumand S (2021) Remote Online Exams Anxiety during the COVID-19 Cri-
sis: A Cross-Sectional Study among Medical Students.
PLOS ONE | https://doi.org/10.1371/journal.pone.0250041 April 15, 2021
16 / 17
PLOS ONE Engineering online education during COVID-19 pandemic
36. Garcı´a-Alberti M, Sua´ rez F, Chiyo´ n I, Mosquera Feijoo JC (2021) Challenges and Experiences of
Online Evaluation in Courses of Civil Engineering during the Lockdown Learning Due to the COVID-19
Pandemic. Education Sciences 11: 59.
37. Dahlberg T, Barnes T, Buch K, Rorrer A (2011) The STARS alliance: Viable strategies for broadening
participation in computing. ACM Transactions on Computing Education (TOCE) 11: 18.
PLOS ONE | https://doi.org/10.1371/journal.pone.0250041 April 15, 2021
17 / 17
PLOS ONE |
10.1371_journal.pone.0240995 | RESEARCH ARTICLE
Risk factors for surgical site infections using a
data-driven approach
J. M. van NiekerkID
holt3, J. E. W. C. van Gemert-Pijnen1
1,2,3, M. C. Vos3, A. Stein2, L. M. A. Braakman-Jansen1*, A. F. Voor in ‘t
1 Department of Psychology, Health and Technology/Centre for eHealth Research and Disease
Management, Faculty of Behavioural Sciences, University of Twente, Enschede, The Netherlands,
2 Department of Earth Observation Sciences, Faculty of Geo-Information Science and Earth Observation
(ITC), University of Twente, Enschede, The Netherlands, 3 Department of Medical Microbiology and
Infectious Diseases, Erasmus MC University Medical Centre, Rotterdam, The Netherlands
* l.m.a.braakman-jansen@utwente.nl
Abstract
Objective
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: van Niekerk JM, Vos MC, Stein A,
Braakman-Jansen LMA, Voor in ‘t holt AF, van
Gemert-Pijnen JEWC (2020) Risk factors for
surgical site infections using a data-driven
approach. PLoS ONE 15(10): e0240995. https://
doi.org/10.1371/journal.pone.0240995
Editor: Francesco Di Gennaro, National Institute for
Infectious Diseases Lazzaro Spallanzani-IRCCS,
ITALY
Received: May 15, 2020
Accepted: October 6, 2020
Published: October 28, 2020
Peer Review History: PLOS recognizes the
benefits of transparency in the peer review
process; therefore, we enable the publication of
all of the content of peer review and author
responses alongside final, published articles. The
editorial history of this article is available here:
https://doi.org/10.1371/journal.pone.0240995
Copyright: © 2020 van Niekerk et al. This is an
open access article distributed under the terms of
the Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: The medical ethical
technical committee of Erasmus MC did not grant
permission to publish these data due to ethical
The objective of this study was to identify risk factors for surgical site infection from diges-
tive, thoracic and orthopaedic system surgeries using clinical and data-driven cut-off values.
A second objective was to compare the identified risk factors in this study to risk factors iden-
tified in literature.
Summary background data
Retrospective data of 3 250 surgical procedures performed in large tertiary care hospital in
The Netherlands during January 2013 to June 2014 were used.
Methods
Potential risk factors were identified using a literature scan and univariate analysis. A multi-
variate forward-step logistic regression model was used to identify risk factors. Standard
medical cut-off values were compared with cut-offs determined from the data.
Results
For digestive, orthopaedic and thoracic system surgical procedures, the risk factors identi-
fied were preoperative temperature of �38˚C and antibiotics used at the time of surgery. C-
reactive protein and the duration of the surgery were identified as a risk factors for digestive
surgical procedures. Being an adult (age �18) was identified as a protective effect for tho-
racic surgical procedures. Data-driven cut-off values were identified for temperature, age
and CRP which can explain the SSI outcome up to 19.5% better than generic cut-off values.
Conclusions
This study identified risk factors for digestive, orthopaedic and thoracic system surgical pro-
cedures and illustrated how data-driven cut-offs can add value in the process. Future
PLOS ONE | https://doi.org/10.1371/journal.pone.0240995 October 28, 2020
1 / 14
PLOS ONE considerations and the sensitivity of the data. Data
are however available from the Erasmus MC upon
reasonable request. Contact person: Dr. Julie¨tte
Severin, Infection prevention and control (IPC) and
antimicrobial resistance (AMR) (Data Access)
E-mail: info.microbiologie.
infectieziekten@erasmusmc.nl.
Funding: This research was supported by the
INTERREG V A (202085) funded project EurHealth-
1Health (http://www.eurhealth1health.eu), part of a
Dutch-German cross-border network supported by
the European Commission, the Dutch Ministry of
Health, Welfare and Sport, the Ministry of
Economy, Innovation, Digitalisation and Energy of
the German Federal State of North Rhine-
Westphalia and the Ministry for National and
European Affairs and Regional Development of
Lower Saxony. The funders had no role in study
design, data collection and analysis, decision to
publish, or preparation of the manuscript.
Competing interests: The authors have declared
that no competing interests exist.
Risk factors for surgical site infections using a data-driven approach
studies should investigate if data-driven cut-offs can add value to explain the outcome being
modelled and not solely rely on standard medical cut-off values to identify risk factors.
Introduction
Surgical site infections (SSI), as defined by the European Centre for Disease Prevention and
Control (ECDC) [1], make up 19.6% of the total number of healthcare-associated infections
(HAIs) in Europe. With an estimated 81 089 patients in Europe having an HAI on any given
day, almost 16 000 people in Europe are suffering from some form of SSI at any given time [2].
The burden of SSI can be measured in terms of increased length of stay in hospital, additional
(surgical) procedures required, increased morbidity and mortality, as well as in economic
terms [3].
Risk factors relating to the patient, procedure and the environment alter the odds of an SSI
occurring. Research has been done to identify risk factors for SSI with the aim to identify pre-
ventative actions to reduce the incidence rate of SSI [4–10]. Patient-related risk factors for SSI,
such as obesity, diabetes, surgery duration and the American Society of Anaesthesiologists
(ASA) score are risk factors for digestive system, thoracic and orthopaedic surgical procedures
[11–22]. Risk factors in low-income countries also include unemployment and level of educa-
tion due to the disparity in socioeconomic status [14]. Risk factors can be modifiable or non-
modifiable [23]. Modifiable risk factors are most interesting of the two since they can be
changed preoperatively to reduce the risk of SSI.
The Segmentation of surgical procedures into homogenous groups makes it possible to find
useful and relevant risk factors unique to each segment. Digestive system surgical procedures
are more prone to SSI as they are generally clean-contaminated or dirty surgeries which make
deep space SSI more likely. The occurrence of SSI after thoracic and orthopaedic surgeries are
both relatively low because they are both typically clean surgeries, but the probability of attract-
ing a deep space SSI after thoracic surgery is much higher compared to orthopaedic surgeries
[15]. Because of these differences, we focus on digestive system, thoracic and orthopaedic sur-
gical procedures for this study.
Multivariate logistic regression is the most common statistical model used to identify risk
factors in longitudinal study design data [16]. Not all studies report the discriminatory power
of the multivariate logistic regression model fitted. Risk factor identification studies do not
usually specify how continuous variables cut-offs are determined. Cut-off values for variables
such as age (�18) or patient temperature (37˚C) may seem intuitive or standard for clinical
practice, but they may not statistically be the best cut-offs values determined by the data [17].
The objective of this study is to identify risk factors for SSI from digestive, thoracic and
orthopaedic system surgeries using clinical and data-driven cut-off values. A second objective
is to compare the identified risk factors in this study to risk factors identified in the literature.
Materials and methods
Literature search
A literature search was performed to identify known risk factors for SSI associated with diges-
tive system surgical procedures, thoracic surgery and orthopaedic procedures using the corre-
sponding medical subject headings (MeSH) linked data representation and the MEDLINE
database.
Search strings used for MEDLINE literature search:
PLOS ONE | https://doi.org/10.1371/journal.pone.0240995 October 28, 2020
2 / 14
PLOS ONE Risk factors for surgical site infections using a data-driven approach
1. “Surgical Wound Infection”[Mesh] AND “Risk Factors”[Mesh] AND “Digestive System
Surgical Procedures”[Mesh]
2. “Surgical Wound Infection”[Mesh] AND “Risk Factors”[Mesh] AND “Orthopaedic
Procedures”[Mesh]
3. “Surgical Wound Infection”[Mesh] AND “Risk Factors”[Mesh] AND “Thoracic
Surgery”[Mesh]
The search results were sorted, using the Best Match algorithm [18] developed by PubMed.
Search results were deemed relevant using title and abstract screening. Risk factors were
extracted if they were significant in a multivariable analysis until data saturation was achieved
[19]. Risk factors identified, which were common to all three groups of surgeries, were defined
as “general risk factors” in this study.
Setting and data collection
The Erasmus MC University Medical Centre in Rotterdam is the largest university medical
hospital in the Netherlands with more than 1 300 beds [15]. The data used for this study were
anonymised in accordance with the Dutch Personal Data Protection Act (WBP). Approval
from the Medical Ethical Research Committee was obtained (MEC-2018-1185).
A weekly prevalence survey was performed by infection control practitioners (ICP) from
January 2013 until December 2013 and two-weekly until June 2014 using a semi-automated
algorithm proposed by Streefkerk et al. [20, 21]. This algorithm was used to calculate a nosoco-
mial infection index (NII) which was then verified by ICP in case of a positive outcome to
determine whenever an HAI was present or not. An ICP verified all patients with an NII > 7,
and a definite SSI outcome was concluded by the ICP using the electronic patient data system.
This outcome was used in this study as the occurrence of SSI outcome variable.
Data were extracted from a centralised database, containing cross-departmental data, clini-
cal synopsis reports, infectious disease consultation reports, laboratory results and imaging
reports. Data regarding the prescription of antimicrobials, in the J01 class of the Anatomical
Therapeutic Chemical (ATC) classification system [22], were also included. Surgeries were
included if they were part of the three groups of surgeries under investigation in this study and
had a point prevalence measurement within 30 days after the surgery took place. If a second
surgery took place within 30 days after an included surgery, then the recent surgery was
excluded. All emergency surgeries were excluded to avoid possible undesirable confounding
effects relating to the urgency and necessity of the surgeries.
Statistical analysis
The differences in the averages of variables with missing values and those without were evalu-
ated using t-tests and were found statistically significant. These tests, together with Little’s
MCAR test, convinced us that the missing values were not completely randomly missing and
that we could not make use of more simple imputation methods. Therefore, we chose to use
conditional Markov chain Monte Carlo (MCMC) with multiple imputations for the imputa-
tion process [24, 25].
Two methods were used to discretise continuous measurement variables: 1) standard medi-
cal cut-offs as used by Erasmus MC and 2) recursive partitioning [17]. Recursive partitioning
is a data-driven, supervised discretisation method, used to group continuous values with simi-
lar outcomes optimally. The data-driven method was used to test and confirm if the standard
medical cut-offs were the best way to explain the outcome variable for the groups of surgical
procedures considered.
PLOS ONE | https://doi.org/10.1371/journal.pone.0240995 October 28, 2020
3 / 14
PLOS ONE Risk factors for surgical site infections using a data-driven approach
To build a prognostic prediction model for SSI, Hosmer et al. suggest fitting a univariate
logistic regression model to each variable separately and if the p-value is less than a specific p-
value, 0.1 is this case, then consider the variable good enough to include in the multivariate
logistic regression model [26]. A univariate analysis was performed for each of the three
groups of surgeries using the variables identified from the literature search. Significant vari-
ables (p<0.1) in the univariate analysis were added to the list of variables associated with each
group of surgery, together with the variables identified from the literature search. This resulted
in an extended list of general risk factors as more risk factors were common across the three
groups of surgeries.
A multivariate logistic regression model was built using a forward stepwise approach for
each of the three groups of surgeries [27]. The general risk factors were first added to the
model and then the risk factors unique to each surgery group in the order of the Akaike infor-
mation criterion (AIC) until convergence was reached. In this case, we chose the conversion of
the model to imply that there are no additional variables which can be added which will be sta-
tistically significant with a p-value of less than 0.05 or an AIC of 3.8415. Model performance
was determined using the Gini coefficient after each step of the multivariate model, and the
difference is reported as the marginal contribution of surgery group-specific risk factors for
this study [19, 28]. Model performance was cross-validated using 5-fold cross-validation to
estimate how the model would perform on new data [29]. R [30] was used in this study
together with packages mice (multiple imputation) [31], smbinning (recursive partitioning)
[32], dplyr (data wrangling) [33], finalfit (formatting of tables) [34] and scorecard (cross-vali-
dation) [35].
Approval was obtained from the Medical Ethical Committee of Erasmus MC (MEC-2018-
1185) to perform this study. Data were analysed anonymously, and thus no further consent
was obtained.
Results
Literature search
The literature search resulted in 1 422 research papers (as at 5 March 2020) using the MeSH
headings in the PubMed search engine. We identified 24 research papers, published from 2008
until 2019, which contained statistically significant results from a multivariate analysis. A total
of 79 risk factors were identified for the three groups of surgical procedures [11–13, 16, 23, 36–
54] (S1 Table). Age, ASA class, body mass index (BMI), preoperative length of stay and diabe-
tes were identified as general risk factors from the literature search. In total, 29 risk factors for
digestive system surgical procedures, 31 for orthopaedic procedures and 19 for thoracic sur-
geries were identified. This amounted to 59 unique risk factors, of which 15 were present in
more than one group of surgeries.
Risk factor identification
A total of 21 of the 59 unique risk factors could be replicated using our own data. The variable
describing the type of surgery was used to create three homogenous groups of surgical proce-
dures. The emergency classification variable was used to exclude emergency surgeries from the
study such that 19 risk factors remained (Table 1). We observed 3 250 surgeries over the study
period and excluded 526 (16.2%) emergency surgeries to be left with 2 724 surgical observa-
tions. CRP and temperature data were available for 52.55% (60.47% for in-patients) and
96.88% of all surgeries respectively.
The significant univariate results of digestive system, orthopaedic and thoracic surgical pro-
cedures are shown in Table 2. Antibiotic use, CRP and temperature were added to the list of
PLOS ONE | https://doi.org/10.1371/journal.pone.0240995 October 28, 2020
4 / 14
PLOS ONE Risk factors for surgical site infections using a data-driven approach
Table 1. Variable names and definitions used to investigate the occurrence of SSI in this study.
Variable
Demographic
Gender
Age
ASA class
BMI
Behavioural
Alcohol use
Smoking
Comorbidities
Heart disease
Liver disease
Hypertension
Diabetes
Measurement
Temperature
CRP
Leukocyte
Serum total protein
Glucose
Haemoglobin
Operative
Surgery
group
Definition
D,O
D,O,T
D,O,T
D,O,T
O
D,O
Gender of patient (Male/Female)
Age of patient on the day of surgery (Years)
ASA class of patient (I-V)
BMI of patient at the time of surgery.
Alcohol use of patient at the time of surgery (Current/Never/Past).
Smoking status of patient at the time of surgery (Current/Never/Past).
O,T
Patient has a history of heart disease at the time of surgery (Yes/No).
D
O
Patient has a history of liver disease at the time of surgery (Yes/No).
Patient has a history of hypertension (Yes/No).
D,O,T
Patient has diabetes Type I or II at the time of surgery (Yes/No).
D
O
D
D
D
D
Highest temperature of patient in the past 7 days before surgery.
Highest CRP of patient in the 7 days before surgery.
Highest leukocyte level of patient in the 7 days before surgery.
Highest serum total protein of patient in the 7 days before surgery.
Highest glucose level of patient in the 7 days before surgery.
Highest haemoglobin level of patient in the 7 days before surgery.
Preoperative length of
D,O,T
stay
Antibiotic use
T
Preoperative length of hospital stay of patient at the time of surgery
(Days).
Antibiotic (WHO ATC code J01 [22]) use of patient at the time of
surgery (Yes/No).
Duration of surgery
D,O
Duration of the surgical procedure (Minutes).
D, Digestive system surgical procedures; O, Orthopaedic system surgical procedures; T, Thoracic system surgical
procedures; ASA, American Society of Anaesthesiologists; CRP, C-reactive protein; BMI, Body Mass Index; SSI,
Surgical Site Infection; ATC, Anatomical Therapeutic Chemical; WHO, World Health Organization.
https://doi.org/10.1371/journal.pone.0240995.t001
general risk factors after being found statistically significant in the univariate analysis–increas-
ing the number of general risk factors to 8. Diabetes was identified as a general risk factor from
our literature search but was not found significant in any of the three univariate analyses in
our own study. For digestive system surgical procedure and thoracic procedures, the data-
driven cut-off for age was obtained as 23 years and both the standard cut-off (18 years) and the
data-driven cut-off were statistically significant with p-values of less than 0.001 which resulted
in rejecting the null hypothesis that the coefficient associated with the age of the patient is
zero. For orthopaedic procedures, the data-driven cut-off for the temperature (39 degrees) was
found statistically significant, but the standard medical cut-off not. A data-driven CRP cut-off
of 8.1 was identified for orthopaedic surgical procedures as opposed to a standard medical
CRP cut-off of 10; both cut-offs are statistically significant.
The multivariate results using standard medical cut-offs and data-driven cut-offs are shown
in Tables 3 and 4, respectively. The temperature variable was statistically significant in the mul-
tivariate analysis using the data-driven cut-offs for all three groups of surgeries, but not in one
of the multivariate analysis using the medical standard cut-offs. The duration of the surgery
was the only statistically significant variable in the multivariate analyses which was not
PLOS ONE | https://doi.org/10.1371/journal.pone.0240995 October 28, 2020
5 / 14
PLOS ONE Table 2. Digestive system surgical procedures: univariate analysis of risk factors for the future occurrence of SSI.
Variable
SSI = No (2 600)
SSI = Yes (124)
Univariate OR (95%CI, P-value)
Digestive System Surgical Procedures
Risk factors for surgical site infections using a data-driven approach
Gender
Age1
Age (data-driven)
Antibiotic use
Temperature1
Temperature (data-driven)
CRP1
CRP (data-driven)
Female
Male
�18
>18
�23
>23
No
Yes
�36.5
(36.5,37.5]
>37.5
�38
(38,39]
>39
�10
>10
�8.1
>8.1
Preoperative length of stay (Days)
Mean Days (SD)
359 (43.9)2
458 (56.1)
246 (30.1)
571 (69.9)
258 (31.6)
559 (68.4)
496 (60.7)
321 (39.3)
0 (0.0)
98 (12.0)
719 (88.0)
535 (65.5)
187 (22.9)
95 (11.6)
397 (48.6)
420 (51.4)
365 (44.7)
452 (55.3)
6.6 (24.1)
Duration of surgery
Mean Minutes (SD)
243.6 (143)
Orthopaedic Procedures
ASA class
Alcohol use
Antibiotic use
Temperature (data-driven)
Age1
Age (data-driven)
BMI
Alcohol use
Antibiotic use
Temperature1
ASA CLASS I
ASA CLASS II
ASA CLASS III
ASA CLASS � IV
Current
Never
Past
No
Yes
�39
>39
�18
>18
�23
>23
Mean (SD)
Current
Never
Past
No
Yes
�36.5
(36.5,37.5]
>37.5
196 (26.8)
339 (46.4)
182 (24.9)
13 (1.8)
327 (44.8)
339 (46.4)
64 (8.8)
591 (81.0)
139 (19.0)
695 (95.2)
35 (4.8)
Thoracic Surgery
232 (22.0)
821 (78.0)
226 (21.5)
827 (78.5)
24.5 (5.3)
534 (50.7)
422 (40.1)
97 (9.2)
705 (67.0)
348 (33.0)
0 (0.0)
302 (28.7)
751 (71.3)
24 (33.8)
47 (66.2)
8 (11.3)
63 (88.7)
8 (11.3)
63 (88.7)
17 (23.9)
54 (76.1)
0 (0.0)
2 (2.8)
69 (97.2)
20 (28.2)
25 (35.2)
26 (36.6)
21 (29.6)
50 (70.4)
18 (25.4)
53 (74.6)
12.1 (37.3)
330.4 (190.8)
6 (33.3)
6 (33.3)
4 (22.2)
2 (11.1)
6 (33.3)
8 (44.4)
4 (22.2)
8 (44.4)
10 (55.6)
14 (77.8)
4 (22.2)
16 (45.7)
19 (54.3)
16 (45.7)
19 (54.3)
22.1 (4.2)
11 (31.4)
18 (51.4)
6 (17.1)
18 (51.4)
17 (48.6)
0 (0.0)
3 (8.6)
32 (91.4)
Reference
1.54 (0.93–2.60, p = 0.099)
Reference
3.39 (1.70–7.77, p<0.001)
Reference
3.63 (1.82–8.32, p<0.001)
Reference
4.91 (2.85–8.86, p<0.001)
NA
Reference
4.70 (1.44–28.91, p = 0.033)
Reference
3.58 (1.95–6.66, p<0.001)
7.32 (3.94–13.79, p<0.001)
Reference
2.25 (1.35–3.89, p = 0.003)
Reference
2.38 (1.39–4.24, p = 0.002)
1.01 (1.00–1.01, p = 0.092)
1.00 (1.00–1.01, p<0.001)
0.58 (0.18–1.87, p = 0.348)
0.72 (0.18–2.55, p = 0.612)
5.03 (0.69–24.47, p = 0.062)
Reference
1.29 (0.44–3.94, p = 0.645)
3.41 (0.85–12.26, p = 0.063)
Reference
5.31 (2.06–14.16, p<0.001)
Reference
5.67 (1.55–16.79, p = 0.003)
Reference
0.34 (0.17–0.67, p = 0.002)
Reference
0.32 (0.16–0.65, p = 0.001)
0.91 (0.85–0.98, p = 0.010)
Reference
2.07 (0.98–4.57, p = 0.061)
3.00 (1.01–8.09, p = 0.034)
Reference
1.91 (0.97–3.77, p = 0.060)
NA
Reference
4.29 (1.52–17.94, p = 0.017)
(Continued )
PLOS ONE | https://doi.org/10.1371/journal.pone.0240995 October 28, 2020
6 / 14
PLOS ONE Risk factors for surgical site infections using a data-driven approach
Table 2. (Continued)
Temperature (data-driven)
CRP1
Haemoglobin1
Variable
SSI = No (2 600)
SSI = Yes (124)
Univariate OR (95%CI, P-value)
�38
>38
�10
>10
�8.6
(8.6,10.5]
>10.5
882 (83.8)
171 (16.2)
684 (65.0)
369 (35.0)
665 (63.2)
358 (34.0)
30 (2.8)
20 (57.1)
15 (42.9)
17 (48.6)
18 (51.4)
21 (60.0)
11 (31.4)
3 (8.6)
Reference
3.87 (1.91–7.67, p<0.001)
Reference
1.96 (1.00–3.88, p = 0.050)
Reference
0.97 (0.45–2.00, p = 0.942)
3.17 (0.72–9.85, p = 0.074)
CRP, C-reactive protein; OR, Odds Ratio; BMI, Body Mass Index; NA, Not Applicable; CI, Confidence Interval; SSI, Surgical Site Infection; OR, Odds ratio; Data-
driven, cut-off values determined using recursive partitioning.
1Standard Erasmus MC clinical cut-offs.
2The percentage distribution of the SSI outcome is provided in brackets next to the frequency for each variable.
https://doi.org/10.1371/journal.pone.0240995.t002
identified as a general risk factor to increase the odds of SSI by approximately 6% for every 30
minutes spent in surgery. For digestive surgical procedures, the addition of duration of surgery
to the multivariate model increased the Gini coefficient from 0.46 to 0.52 based on standard
medical cut-offs and from 0.57 to 0.62 for the multivariate model based on the data-driven
cut-offs. This increase translates into a 12.5% and 8.8% increase in the Gini coefficient, respec-
tively. Neither the orthopaedic nor the thoracic group of surgical procedures had any statisti-
cally significant risk factors which are not part of the general risk factors group of surgeries.
The Gini coefficient of the data-driven multivariate model is 19.5% (0.62 vs 0.52) higher than
the multivariate model based on the standard medical cut-offs. The 5-fold cross-validated 95%
confidence intervals for the Gini coefficients based on the validation samples of the data-
driven models are (0.49, 0.72) for digestive procedures, (0.21, 0.86) for orthopaedic procedures
and (0.21,0.70) for thoracic procedures.
An overview of the study results (Table 5) shows that 10 of the 19 risk factors, identified
during the literature search, were not statistically significant in the univariate or multivariate
analysis for any of the surgery groups. BMI and diabetes were identified across all three groups
of surgeries and multiple studies as risk factors for SSI but were not statistically significant in
this study. Temperature and the duration of the surgery were confirmed as risk factors for
digestive system surgeries, and similarly, antibiotic use and age were confirmed as risk factors
Table 3. Multivariate analysis risk factors for the occurrence of SSI by group of surgeries using standard medical cut-offs.
Risk factor by surgery group1
Digestive System Surgical Procedures
Antibiotic use
Duration of surgery (Minutes)
CRP >10
Orthopaedic Surgical Procedures
Antibiotic use
Thoracic Surgical Procedures
Age >18
Antibiotic use
Coefficient
Multivariate OR (95%CI)
P-value
1.240
0.003
0.803
1.670
-4.195
1.311
3.455 (1.951–6.384)
1.003 (1.001–1.004)
2.232 (1.302–3.951)
5.315 (2.059–14.158)
0.146 (0.058–0.351)
4.849 (2.035–12.266)
<0.001
<0.001
0.004
<0.001
<0.001
<0.001
CRP, C-reactive protein; CI, Confidence Interval; OR, Odds ratio.
1The multivariate analysis was performed using Erasmus MC clinical cut-offs.
https://doi.org/10.1371/journal.pone.0240995.t003
PLOS ONE | https://doi.org/10.1371/journal.pone.0240995 October 28, 2020
7 / 14
PLOS ONE Risk factors for surgical site infections using a data-driven approach
Table 4. Multivariate analysis risk factors for the occurrence of SSI by group of surgeries using data-driven cut-offs.
Risk factor by surgery group1
Digestive System Surgical Procedures
Temperature [38,39]
Temperature >39
Antibiotic use
Duration of surgery (Minutes)
CRP >8.1
Orthopaedic Surgical Procedures
Antibiotic use
Temperature >39
Thoracic Surgical Procedures
Age >17
Antibiotic use
Temperature >38
Coefficient
Multivariate OR (95%CI)
P-value
1.067
1.732
1.201
0.002
0.639
1.552
1.224
-1.847
1.597
0.824
2.907 (1.556–5.497)
5.650 (2.952–10.947)
3.322 (1.856–6.200)
1.002 (1.001–1.004)
1.894 (1.062–3.510)
3.665 (1.370–10.006)
5.120 (1.316–16.387)
0.158 (0.055–0.426)
4.939 (1.896–14.043)
2.280 (1.098–4.653)
<0.001
<0.001
<0.001
0.003
0.035
0.009
0.009
<0.001
0.002
0.024
Data-driven, cut-off values determined using recursive partitioning; CRP, C-reactive protein; CI, Confidence Interval; OR, Odds ratio.
1The multivariate analysis was performed using data-driven cut-offs.
https://doi.org/10.1371/journal.pone.0240995.t004
for thoracic surgeries. Antibiotic use and CRP were identified as risk factors for digestive sur-
geries from the multivariate analysis, which were identified during the literature search for
thoracic and orthopaedic surgeries, respectively. Antibiotic use and temperature were
Table 5. Statistical significance of risk factors and the source which lead them to be considered by surgical procedure.
Risk Factor
Age
Alcohol use
Antibiotic use
ASA Class
BMI
CRP
Diabetes
Duration of surgery
Gender
Glucose
Haemoglobin
Heart Disease
Hypertension
Leukocyte
Liver disease
Preoperative length of stay
Serum total protein
Smoking
Temperature
Significance1
DU,TM
OU,TU
DM,OM,TM
OU
None
DM
None
DM
DU
None
None
None
None
None
None
DU
None
None
DM,OM,TM
Digestive System2
[38, 11, 43, 47]
[37, 39, 41, 43, 54]
[44]
[38, 47, 50]
[36, 38, 41, 43, 44, 49, 54]
[38, 11, 43]
[47]
[11, 44, 54]
[55]
[54]
[41, 50]
[36, 49]
[49]
[55]
Orthopaedic2
[16]
[51]
[16, 51, 53]
[51–53]
[16]
[16, 45, 51, 53]
[16, 45, 51, 53]
[16, 51]
[51]
[51]
[16, 52]
[51–53]
Thoracic2
[12]
[40]
[16]
[42]
[13]
[12]
[12, 13, 40]
D, Digestive system surgical procedures; O, Orthopaedic system surgical procedures; U, Significant in univariate analysis; M, Significant in multivariate analysis; T,
Thoracic system surgical procedures; ASA, American Society of Anaesthesiologists; CRP, C-reactive protein; SSI, Surgical Site Infection; BMI, Body Mass Index.
1During which part of the analysis the risk factor was found statistically significant.
2References to the literature which had the risk factor as a multivariate result for each group of surgeries.
https://doi.org/10.1371/journal.pone.0240995.t005
PLOS ONE | https://doi.org/10.1371/journal.pone.0240995 October 28, 2020
8 / 14
PLOS ONE Risk factors for surgical site infections using a data-driven approach
statistically significant for all three groups of surgeries and were included because of two stud-
ies regarding thoracic and digestive system surgeries, respectively [40, 55].
Discussion
We identified temperature and antibiotics used at the time of surgery as risk factors for diges-
tive, orthopaedic and thoracic system surgical procedures in this study. The duration of the
surgery was identified as a risk factor for digestive surgical procedures. Being an adult
(age � 18) was identified as a protective effect for thoracic surgical procedures. Data-driven
cut-offs were identified for temperature, CRP and age, which differ from the standard medical
cut-offs. Temperature would not have been identified as a risk factor if only standard medical
cut-offs were considered. From our literature search, we identified age, ASA class, BMI, preop-
erative length of stay and diabetes as general risk factors, while CRP, temperature and antibi-
otic use were identified as general risk factors because of this study.
The identified risk factors may be classified as modifiable or non-modifiable, depending
upon the circumstances of the patient like the complexity of his condition. For instance, the
temperature of a patient may be high because of an existing infection, which is why the surgery
is needed in the first place and may not be modifiable before surgery. Age, on the other hand,
may be a modifiable risk factor if the surgery can be postponed for several years, e.g. due to a
heart defect. This study revealed that children are more likely to be diagnosed with an SSI after
thoracic surgery than adults. There are studies which identify risk factors for children after
thoracic surgeries, but none found that being a child is a risk factor for SSI [42, 48] after under-
going thoracic surgery. We segmented the thoracic surgeries between adults and children and
obtained multivariate results for children and adults separately. The multivariate model based
only on children (age � 18) did not reveal any significant results, contrary to the results of the
thoracic study which found age to be a risk factor for children [12]. This absence could be
partly due to the small study population size of 248. Antibiotic usage was the only significant
factor in the multivariate analysis of thoracic surgeries based on adults. The other two groups
of surgical procedures were consistent in terms of their statistical significance of risk factors
based on adults.
The data-driven cut-offs confirmed the existing standard medical cut-offs. On average the
clinical cut-off for temperature was one degree Celsius lower, while for digestive system surgi-
cal procedures, the clinical cut-off for CRP (10) was just less than two units more than the
data-driven cut-off of 8.1. This means that there is a greater difference between the occurrence
of SSI for patients with a CRP below and above 8.1 than below and above 10. The data-driven
cut-offs improved the ability of the statistical model to explain the occurrence of SSI. The per-
formance of the digestive system surgical procedure prediction model increased by 19.5% due
to using data-driven cut-offs rather than the standard medical cut-offs. Using data-driven cut-
offs, we were able to identify temperature as a risk factor for all three groups of surgical proce-
dures. If standard clinical cut-offs were used, temperature would not have been significant
from the multivariate analysis. This potential oversight illustrates the importance of evaluating
the cut-offs used for continuous variables against the data before identifying risk factors.
Antibiotic use, temperature and CRP were added to the list of general risk factors by incor-
porating the statistically significant results of the univariate analysis. These risk factors might
have been overlooked when the focus was on only one type of surgery. Temperature was iden-
tified as a risk factor in the multivariate results for all three groups of surgical procedures,
whereas the literature search identified it only for digestive surgeries. Antibiotic use was not
found during our literature search for digestive or orthopaedic surgical procedures but was
found significant for both groups of surgeries in the multivariate analysis of our study.
PLOS ONE | https://doi.org/10.1371/journal.pone.0240995 October 28, 2020
9 / 14
PLOS ONE Risk factors for surgical site infections using a data-driven approach
The Centres for Disease Control and Prevention (CDC), the European centre for disease
prevention and control (ECDC), World Health Organisation (WHO) and Netherlands
National Institute for Public Health and the Environment (RIVM) suggest maintaining nor-
mothermia intraoperatively to prevent undesirable hypothermia (during some thoracic and
neurosurgeries, hypothermia may be desirable). [56–58] A lower intraoperative bound for
temperature of 35.5˚C to 36˚C is explicitly mentioned, and only the RIVM mention an upper
bound of 38˚C which is consistent with the risk factors identified in our study. An upper limit
for preoperative temperature should, therefore, be investigated instead of only the lower limit.
The four health organisations refer to the proper administration and timing of surgical antimi-
crobial prophylaxis, but not to the proper preoperative use of standard prescription antibiotics.
Systemic antibiotics are typically prescribed to stabilise patients before undergoing surgery. A
possible explanation for the increased occurrence of SSI associated with antimicrobials pre-
scribed before surgery could be that these patients were not completely stabilised before sur-
gery which increased their risk of SSI. The proper preoperative use of antibiotics should be
well defined, and the reason why antibiotic-use was identified as a risk factor for SSI should be
further investigated.
Limitations
This is a retrospective, single-centre study, and therefore the data were not collected for the
purpose of this study. Even though cross-validation was performed to estimate model perfor-
mance on new data, the models were not externally validated. Surgeries were aggregated into
three broad groups of surgical procedures which serve as a proxy for the reason for surgery but
leads to the loss of information regarding the exact reasons for the surgery. Some measure-
ments, like temperature and CRP, were not always present and was partly overcome using
imputation. Patient information concerning smoking and drinking habits may be understated
due to incomplete medical records. The literature search used for this study was not exhaustive
but rather based on the principal on data saturation. A comprehensive list of variables related
to the nutritional and immunological alterations of the patients was not included in the analy-
ses as they were not available from the data. We used a 30-day outcome period in which we
observe if an SSI was present or not, but according to the CDC definition, this outcome period
should be one year for surgical implantation procedures. Since our data only spans over 18
months, it was not possible to use a 12-month outcome window for all surgical implantation
procedures, which is a limitation of this study. The administration of prophylaxis and the opti-
mal timing thereof is an important risk factor for the occurrence of SSI. However, these data
were not available.
Future work
Future work will investigate the modifiability of the risk factors identified in this study in more
detail, as the circumstances under which this occurs are hitherto unclear. The exact purpose of
the use of antibiotics over the time of surgery was not investigated in depth, which can be done
in future studies. Future research can also investigate differences between adults and children,
which lead to the occurrence of SSI among children. Another opportunity for future research
is to investigate which risk factors are predictive for the occurrence of SSI over different peri-
ods. Doing this will enable healthcare workers to identify which risk factors explain the occur-
rence of SSI soon after surgery, towards the end of the 30 days and even later for implantation
surgeries. These insights can help set guidelines to determine the vigilance necessary to miti-
gate the risk of SSI on a patient level.
PLOS ONE | https://doi.org/10.1371/journal.pone.0240995 October 28, 2020
10 / 14
PLOS ONE Risk factors for surgical site infections using a data-driven approach
Conclusion
This study shows that data-driven cut-offs can be used to identify risk factors which would not
have been identified by only using standard medical cut-offs. Preoperative temperature and
antibiotic use were identified as risk factors for digestive, orthopaedic, thoracic system surger-
ies, while the duration of surgery and age were identified as risk factors for orthopaedic and
thoracic system surgeries, respectively. In contrast with literature, this study found that an SSI
is more likely to occur in children (age < 18) than in adults after thoracic system surgeries. Sta-
tistical modelling has been important to quantify important risk factors and indicate their sig-
nificance. Clinical studies using retrospective data are important to carry out, despite
limitations in the data sets. To this end, future studies should use both standard medical cut-
offs and data-driven cut-offs to investigate risk factors.
Supporting information
S1 Table. Risk factors identified from multivariate analysis during literature search.
(DOCX)
S1 Formulae. The multivariate logistic regression equations based on the data-driven cut-offs.
(DOCX)
Acknowledgments
We would like to thank C. P. (Conrad) van der Hoeven, A.G.D (Arnim) Mulder and M. (Mar-
ius) Vogel for their help and constant willingness to help with questions regarding the data
used for this study. Also, thank you to R. H. (Roel) Streefkerk for the work done to organise
and combine the data as well as producing the SSI outcome variable used in this study.
Author Contributions
Conceptualization: J. M. van Niekerk, M. C. Vos, A. Stein, L. M. A. Braakman-Jansen, A. F.
Voor in ‘t holt, J. E. W. C. van Gemert-Pijnen.
Data curation: J. M. van Niekerk, M. C. Vos.
Formal analysis: J. M. van Niekerk, A. F. Voor in ‘t holt.
Investigation: J. M. van Niekerk, M. C. Vos, A. F. Voor in ‘t holt.
Methodology: J. M. van Niekerk, M. C. Vos, A. Stein, A. F. Voor in ‘t holt.
Supervision: A. Stein, L. M. A. Braakman-Jansen, A. F. Voor in ‘t holt, J. E. W. C. van Gemert-
Pijnen.
Validation: J. M. van Niekerk.
Writing – original draft: J. M. van Niekerk, M. C. Vos, A. Stein, L. M. A. Braakman-Jansen,
A. F. Voor in ‘t holt.
Writing – review & editing: J. M. van Niekerk, M. C. Vos, A. Stein, L. M. A. Braakman-Jan-
sen, A. F. Voor in ‘t holt, J. E. W. C. van Gemert-Pijnen.
References
1. European Centre for Disease Prevention and Control. Surveillance of surgical site infections in Euro-
pean hospitals–HAISSI protocol [Internet]. 2012. 1–47 p. Available from: http://www.ecdc.europa.eu/
en/publications/Publications/120215_TED_SSI_protocol.pdf
PLOS ONE | https://doi.org/10.1371/journal.pone.0240995 October 28, 2020
11 / 14
PLOS ONE Risk factors for surgical site infections using a data-driven approach
2.
Zarb P, Coignard B, Griskeviciene J, Muller A, Vankerckhoven V, Weist K, et al. point prevalence survey
of healthcare-associated infections and antimicrobial use in European acute care hospitals. Vol. 17,
Eurosurveillance. 2012. 1–16 p.
3. Roy S, Patkar A, Daskiran M, Levine R, Hinoul P, Nigam S. Clinical and Economic Burden of Surgical
Site Infection in Hysterectomy. Surg Infect (Larchmt). 2014; 15(3):266–73. https://doi.org/10.1089/sur.
2012.163 PMID: 24801549
4.
5.
Leung Wai Sang S, Chaturvedi R, Alam A, Samoukovic G, de Varennes B, Lachapelle K. Preoperative
hospital length of stay as a modifiable risk factor for mediastinitis after cardiac surgery. J Cardiothorac
Surg [Internet]. 2013; 8(1):45. Available from: Journal of Cardiothoracic Surgery https://doi.org/10.
1186/1749-8090-8-45 PMID: 23497663
Triantafyllopoulos G, Stundner O, Memtsoudis S, Poultsides LA. Patient, surgery, and hospital related
risk factors for surgical site infections following total hip arthroplasty. Sci World J. 2015; 2015:1–9.
6. Abuzaid A, Zaki M, Al Tarief H. Potential risk factors for surgical site infection after isolated coronary
artery bypass grafting in a Bahrain Cardiac Centre: A retrospective, case-controlled study. Hear Views.
2015; 16(3):79.
7. Schimmel JJP, Horsting PP, De Kleuver M, Wonders G, Van Limbeek J. Risk factors for deep surgical
site infections after spinal fusion. Eur Spine J. 2010; 19(10):1711–9. https://doi.org/10.1007/s00586-
010-1421-y PMID: 20445999
8. Guohua X, cheng keping, Li J, Kong Q, Wang C, Nanyuan Y. Risk factors for surgical site infection in a
teaching hospital: a prospective study of 1,138 patients. Patient Prefer Adherence. 2015; 9:1171.
https://doi.org/10.2147/PPA.S86153 PMID: 26316722
9. Gomila A, Carratalà J, Biondo S, Badia JM, Fraccalvieri D, Shaw E, et al. Predictive factors for early-
and late-onset surgical site infections in patients undergoing elective colorectal surgery. A multicentre,
prospective, cohort study. J Hosp Infect. 2018; 99(1):24–30. https://doi.org/10.1016/j.jhin.2017.12.017
PMID: 29288776
10. Martin ET, Kaye KS, Knott C, Nguyen H, Santarossa M, Evans R, et al. Diabetes and risk of surgical
site infection: A systematic review and meta-analysis. Infect Control Hosp Epidemiol. 2016; 37(1):88–
99. https://doi.org/10.1017/ice.2015.249 PMID: 26503187
11.
Isik O, Kaya E, Dundar HZ, Sarkut P. Surgical Site Infection: Re-assessment of the Risk Factors. Chir-
urgia (Bucur) [Internet]. 2015; 110(5):457–61. Available from: http://www.ncbi.nlm.nih.gov/pubmed/
26531790 PMID: 26531790
12. Chen LF, Arduino JM, Sheng S, Muhlbaier LH, Kanafani ZA, Harris AD, et al. Epidemiology and out-
come of major postoperative infections following cardiac surgery: Risk factors and impact of pathogen
type. Am J Infect Control. 2012; 40(10):963–8. https://doi.org/10.1016/j.ajic.2012.01.012 PMID:
22609237
13.
Jolivet S, Lescure FX, Armand-Lefevre L, Raffoul R, Dilly MP, Ghodbane W, et al. Surgical site infection
with extended-spectrum β-lactamase-producing Enterobacteriaceae after cardiac surgery: incidence
and risk factors. Clin Microbiol Infect [Internet]. 2018; 24(3):283–8. Available from: https://doi.org/10.
1016/j.cmi.2017.07.004 PMID: 28698036
14. Di Gennaro F, Marotta C, Pisani L, Veronese N, Pisani V, Lippolis V, et al. Maternal caesarean section
infection (MACSI) in Sierra Leone: a case-control study. Epidemiol Infect. 2020;1–6. https://doi.org/10.
1017/S0950268820000370 PMID: 32102721
15. European Centre for Disease Prevention and Control. Healthcare-associated infections: surgical site
infections. ECDC Annu Epidemiol Rep 2016 Stock ECDC. 2018;(May).
16. KUNUTSOR SK, WHITEHOUSE MR, BLOM AW, BESWICK AD. Systematic review of risk prediction
scores for surgical site infection or periprosthetic joint infection following joint arthroplasty. Epidemiol
Infect. 2017; 145(9):1738–49. https://doi.org/10.1017/S0950268817000486 PMID: 28264756
17. Strobl C, Malley J, Gerhard Tutz. Characteristics of Classification and Regression Trees, Bagging and
Random Forests. Psychol Methods. 2009; 14(4):323–48. https://doi.org/10.1037/a0016973 PMID:
19968396
18.
Fiorini N, Canese K, Starchenko G, Kireev E, Kim W, Miller V, et al. Best Match: New relevance search
for PubMed. PLoS Biol. 2018; 16(8):1–12. https://doi.org/10.1371/journal.pbio.2005343 PMID:
30153250
19. Aldiabat KM, Le Navenec CL. Data saturation: The mysterious step in grounded theory methodology.
Qual Rep. 2018; 23(1):245–61.
20. Streefkerk RHRA, Moorman PW, Parlevliet GA, van der Hoeven C, Verbrugh HA, Vos MC, et al. An
Automated Algorithm to Preselect Patients to Be Assessed Individually in Point Prevalence Surveys for
Hospital-Acquired Infections in Surgery. Infect Control Hosp Epidemiol [Internet]. 2014; 35(7):886–7.
Available from: http://www.jstor.org/stable/info/10.1086/676868 https://doi.org/10.1086/676868 PMID:
24915221
PLOS ONE | https://doi.org/10.1371/journal.pone.0240995 October 28, 2020
12 / 14
PLOS ONE Risk factors for surgical site infections using a data-driven approach
21. Streefkerk RHRA, Borsboom GJJM, van der Hoeven CP, Vos MC, Verkooijen RP, Verbrugh HA. Evalu-
ation of an Algorithm for Electronic Surveillance of Hospital-Acquired Infections Yielding Serial Weekly
Point Prevalence Scores. Infect Control Hosp Epidemiol [Internet]. 2014 [cited 2017 Jul 18]; 35
(07):888–90. Available from: https://www.cambridge.org/core/product/identifier/S0899823X00192372/
type/journal_article https://doi.org/10.1086/676869 PMID: 24915222
22. World Health Organization. Guidelines for ATC classification and DDD assignment 2020 [Internet].
2020. Available from: https://www.whocc.no/filearchive/publications/2020_guidelines_web.pdf
23. Silvestri M, Dobrinja C, Scomersi S, Giudici F, Turoldo A, Princic E, et al. Modifiable and non-modifiable
risk factors for surgical site infection after colorectal surgery: a single-center experience. Surg Today.
2018; 48(3):338–45. https://doi.org/10.1007/s00595-017-1590-y PMID: 28948367
24. Asendorpf JB, Van De Schoot R, Denissen JJA, Hutteman R. Reducing bias due to systematic attrition
in longitudinal studies: The benefits of multiple imputation. Int J Behav Dev. 2014; 38(5):453–60.
25.
van Buuren S. Multiple imputation of discrete and continuous. Stat Methods Med Res. 2007; 16(3):219–
42. https://doi.org/10.1177/0962280206074463 PMID: 17621469
26. Hosmer DW, Lemeshow S. Applied logistic regression [Internet]. Wiley. 2000. 118–128 p. Available
from: http://onlinelibrary.wiley.com/book/10.1002/9781118548387
27.
In Lee K, Koval JJ. Determination of the best significance level in forward stepwise logistic regression.
Commun Stat—Simul Comput. 2007; 26(2):559–75.
28. Hastie T, Tibshirani R, Friedman J. The Elements of Statistical Learning. Math Intell [Internet]. 2009
[cited 2017 Aug 23]; 27(2):764. Available from: http://www.springerlink.com/index/10.1007/b94608
PMID: 19443179
29. Wong T-T. Performance evaluation of classification algorithms by k-fold and leave-one-out cross valida-
tion. Pattern Recognit. 2015; 48(9):2839–46.
30. R Core Team. R: A Language and Environment for Statistical Computing [Internet]. Vienna, Austria;
2020. Available from: http://www.r-project.org/
31.
32.
van Buuren S, Groothuis-Oudshoorn K. {mice}: Multivariate Imputation by Chained Equations in R. J
Stat Softw [Internet]. 2011; 45(3):1–67. Available from: https://www.jstatsoft.org/v45/i03/
Jopia H. smbinning: Scoring Modeling and Optimal Binning [Internet]. 2019. Available from: https://cran.
r-project.org/package=smbinning
33. Wickham H, Franc¸ois R, Henry L, Mu¨ ller K. dplyr: A Grammar of Data Manipulation [Internet]. 2019.
Available from: https://cran.r-project.org/package=dplyr
34. Harrison E, Drake T, Ots R. finalfit: Quickly Create Elegant Regression Results Tables and Plots when
Modelling [Internet]. 2019. Available from: https://cran.r-project.org/package=finalfit
35. Xie S. scorecard: Credit Risk Scorecard [Internet]. 2020. Available from: https://cran.r-project.org/
package=scorecard
36.
37.
Takahashi Y, Takesue Y, Fujiwara M, Tatsumi S, Ichiki K, Fujimoto J, et al. Risk factors for surgical site
infection after major hepatobiliary and pancreatic surgery. J Infect Chemother [Internet]. 2018; 24
(9):739–43. Available from: https://doi.org/10.1016/j.jiac.2018.05.007 PMID: 30001844
Fukuda H. Patient-related risk factors for surgical site infection following eight types of gastrointestinal
surgery. J Hosp Infect [Internet]. 2016; 93(4):347–54. Available from: https://doi.org/10.1016/j.jhin.
2016.04.005 PMID: 27209057
38. Kokudo T, Uldry E, Demartines N, Halkic N. Risk factors for incisional and organ space surgical site
infections after liver resection are different. World J Surg. 2015; 39(5):1185–92. https://doi.org/10.1007/
s00268-014-2922-3 PMID: 25561190
39. Moreno Elola-Olaso A, Davenport DL, Hundley JC, Daily MF, Gedaly R. Predictors of surgical site infec-
tion after liver resection: A multicentre analysis using National Surgical Quality Improvement Program
data. Hpb. 2012; 14(2):136–41. https://doi.org/10.1111/j.1477-2574.2011.00417.x PMID: 22221576
40.
Lola I, Levidiotou S, Petrou A, Arnaoutoglou H, Apostolakis E, Papadopoulos GS. Are there indepen-
dent predisposing factors for postoperative infections following open heart surgery? J Cardiothorac
Surg [Internet]. 2011; 6(1):151. Available from: https://doi.org/10.1186/1749-8090-6-151 PMID:
22082355
41. Araki T, Okita Y, Uchino M, Ikeuchi H, Sasaki I, Funayama Y, et al. Risk factors for surgical site infection
in Japanese patients with ulcerative colitis: A multicenter prospective study. Surg Today. 2014; 44
(6):1072–8. https://doi.org/10.1007/s00595-013-0809-9 PMID: 24337501
42. Ben-Ami E, Levy I, Katz J, Dagan O, Shalit I. Risk factors for sternal wound infection in children under-
going cardiac surgery: a case-control study. J Hosp Infect [Internet]. 2008; 70(4):335–40. Available
from: https://doi.org/10.1016/j.jhin.2008.08.010 PMID: 18951662
PLOS ONE | https://doi.org/10.1371/journal.pone.0240995 October 28, 2020
13 / 14
PLOS ONE Risk factors for surgical site infections using a data-driven approach
43. Ejaz A, Schmidt C, Johnston FM, Frank SM, Pawlik TM. Risk factors and prediction model for inpatient
surgical site infection after major abdominal surgery. J Surg Res [Internet]. 2017; 217:153–9. Available
from: https://doi.org/10.1016/j.jss.2017.05.018 PMID: 28595819
44. Giri S, Kandel BP, Pant S, Lakhey PJ, Singh YP, Vaidya P. Risk factors for surgical site infections in
abdominal surgery: A study in Nepal. Surg Infect (Larchmt). 2013; 14(3):313–8. https://doi.org/10.1089/
sur.2012.108 PMID: 23672239
45. Hijas-Go´mez AI, Egea-Ga´ mez RM, Martı´nez-Martı´n J, Gonza´lez-Dı´az R, Losada-Viñas JI, Rodrı´guez-
Caravaca G. Surgical Wound Infection Rates and Risk Factors in Spinal Fusion in a University Teaching
Hospital in Madrid, Spain. Spine (Phila Pa 1976). 2017; 42(10):748–54.
46.
Liu S, Miao J, Wang G, Wang M, Wu X, Guo K, et al. Risk factors for postoperative surgical site infec-
tions in patients with Crohn’s disease receiving definitive bowel resection. Sci Rep [Internet]. 2017; 7
(1):1–6. Available from: https://doi.org/10.1038/s41598-016-0028-x PMID: 28127051
47. McKenzie Stancu S, Iordache F. Thrombocytosis Is a Risk Factor for Surgical Site Infections after
Colon Resection: A Prospective Observational Study. Surg Infect (Larchmt) [Internet]. 2019; 20(1):39–
44. Available from: http://www.ncbi.nlm.nih.gov/pubmed/30256728 https://doi.org/10.1089/sur.2018.
146 PMID: 30256728
48. Mehta PA, Cunningham CK, Colella CB, Alferis G, Weiner LB. Risk factors for sternal wound and other
infections in pediatric cardiac surgery patients. Pediatr Infect Dis J. 2000; 19(10):1000–4. https://doi.
org/10.1097/00006454-200010000-00012 PMID: 11055604
49. Morikane K, Honda H, Suzuki S. Factors associated with surgical site infection following gastric surgery
in Japan. Infect Control Hosp Epidemiol. 2016; 37(10):1167–72. https://doi.org/10.1017/ice.2016.155
PMID: 27430979
50. Perkins JD. Techniques to ensure adequate portal flow in the presence of splenorenal shunts. Liver
Transplant. 2007; 13(5):767–8. PMID: 17563933
51. Shao J, Zhang H, Yin B, Li J, Zhu Y, Zhang Y. Risk factors for surgical site infection following operative
treatment of ankle fractures: A systematic review and meta-analysis. Int J Surg [Internet]. 2018; 56
(May):124–32. Available from: https://doi.org/10.1016/j.ijsu.2018.06.018 PMID: 29929022
52. Su J, Cao X. Risk factors of wound infection after open reduction and internal fixation of calcaneal frac-
tures. Med (United States). 2017; 96(44):e8411. https://doi.org/10.1097/MD.0000000000008411
PMID: 29095273
53. Xing D, Ma JX, Ma XL, Song DH, Wang J, Chen Y, et al. A methodological, systematic review of evi-
dence-based independent risk factors for surgical site infections after spinal surgery. Eur Spine J. 2013;
22(3):605–15. https://doi.org/10.1007/s00586-012-2514-6 PMID: 23001381
54.
55.
Zhang JF, Zhu HY, Sun YW, Liu W, Huo YM, Liu DJ, et al. Pseudomonas aeruginosa infection after
pancreatoduodenectomy: Risk factors and clinic impacts. Surg Infect (Larchmt). 2015; 16(6):769–74.
https://doi.org/10.1089/sur.2015.041 PMID: 26237502
Isik O, Kaya E, Sarkut P, Dundar HZ. Factors affecting surgical site infection rates in hepatobiliary sur-
gery. Surg Infect (Larchmt). 2015; 16(3):281–6. https://doi.org/10.1089/sur.2013.195 PMID: 25830815
56. Berrı´os-Torres SI, Umscheid CA, Bratzler DW, Leas B, Stone EC, Kelz RR, et al. Centers for Disease
Control and Prevention Guideline for the Prevention of Surgical Site Infection, 2017. JAMA Surg [Inter-
net]. 2017; 152(8):784–91. Available from: https://doi.org/10.1001/jamasurg.2017.0904 PMID:
28467526
57.
Leaper DJ, Edmiston CE. World Health Organization: global guidelines for the prevention of surgical
site infection. J Hosp Infect [Internet]. 2017; 95(2):135–6. Available from: http://linkinghub.elsevier.com/
retrieve/pii/S0195670116305874 https://doi.org/10.1016/j.jhin.2016.12.016 PMID: 28139389
58. Werkgroep infectiepreventie. Preventie van postoperatieve wondinfecties. 2011;1–26. Available from:
http://www.rivm.nl/Documenten_en_publicaties/Professioneel_Praktisch/Richtlijnen/Infectieziekten/
WIP_Richtlijnen/Actuele_WIP_Richtlijnen/Ziekenhuizen/WIP_richtlijn_Postoperatieve_wondinfecties_
ZKH
PLOS ONE | https://doi.org/10.1371/journal.pone.0240995 October 28, 2020
14 / 14
PLOS ONE |
10.1371_journal.pone.0240176 | RESEARCH ARTICLE
Behavioral and corticosterone responses to
carbon dioxide exposure in reptiles
Daniel J. D. NatuschID
Ain Isa5, Che Ku Zamzuri5, Andre GanswindtID
6,7, Dale F. DeNardo8
1,2☯*, Patrick W. Aust3,4☯, Syarifah Khadiejah5, Hartini Ithnin5,
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: Natusch DJD, Aust PW, Khadiejah S,
Ithnin H, Isa A, Zamzuri CK, et al. (2020) Behavioral
and corticosterone responses to carbon dioxide
exposure in reptiles. PLoS ONE 15(10): e0240176.
https://doi.org/10.1371/journal.pone.0240176
Editor: Todd Adam Castoe, University of Texas at
Arlington, UNITED STATES
Received: July 2, 2020
Accepted: September 21, 2020
Published: October 6, 2020
Copyright: © 2020 Natusch et al. This is an open
access article distributed under the terms of the
Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: All relevant data are
within the manuscript and its Supporting
Information files.
Funding: Daniel Natusch recieved funding from the
Southeast Asian Reptile Conservation Alliance and
the Swiss Federal Veterinary Office. The funders
had no role in study design, data collection and
analysis, decision to publish, or preparation of the
manuscript.
Competing interests: The authors have declared
that no competing interests exist.
1 Department of Biological Sciences, Macquarie University, North Ryde, NSW, Australia, 2 EPIC
Biodiversity, Frogs Hollow, NSW, Australia, 3 Department of Zoology, University of Oxford, Oxford, United
Kingdom, 4 Bushtick Environmental Services, Grantham, Lincolnshire, United Kingdom, 5 Department of
Wildlife and National Parks, Peninsular Malaysia, Kuala Lumpur, Malaysia, 6 Endocrine Research
Laboratory, Mammal Research Institute, Department of Zoology and Entomology, Faculty of Natural and
Agricultural Sciences, University of Pretoria, Pretoria, South Africa, 7 Centre of Veterinary Wildlife Studies,
Faculty of Veterinary Science, University of Pretoria, Pretoria, Onderstepoort, South Africa, 8 School of Life
Sciences, Arizona State University, Tempe, Arizona, United States of America
☯ These authors contributed equally to this work.
* d.natusch@epicbiodiversity.com
Abstract
The use of carbon dioxide (CO2) exposure as a means of animal euthanasia has received
considerable attention in mammals and birds but remains virtually untested in reptiles. We
measured the behavioral responses of four squamate reptile species (Homalopsis buccata,
Malayopython reticulatus, Python bivitattus, and Varanus salvator) to exposure to 99.5%
CO2 for durations of 15, 30, or 90 minutes. We also examined alterations in plasma cortico-
sterone levels of M. reticulatus and V. salvator before and after 15 minutes of CO2 exposure
relative to control individuals. The four reptile taxa showed consistent behavioral responses
to CO2 exposure characterized by gaping and minor movements. The time taken to lose
responsiveness to stimuli and cessation of movements varied between 240–4260 seconds
(4–71 minutes), with considerable intra- and inter-specific variation. Duration of CO2 expo-
sure influenced the likelihood of recovery, which also varied among species (e.g., from
0–100% recovery after 30-min exposure). Plasma corticosterone concentrations increased
after CO2 exposure in both V. salvator (18%) and M. reticulatus (14%), but only significantly
in the former species. Based on our results, CO2 appears to be a mild stressor for reptiles,
but the relatively minor responses to CO2 suggest it may not cause considerable distress or
pain. However, our results are preliminary, and further testing is required to understand opti-
mal CO2 delivery mechanisms and interspecific responses to CO2 exposure before endors-
ing this method for reptile euthanasia.
Introduction
Ensuring the humane euthanasia of animals used by humans is critically important to fulfil
our ethical obligation for compassion towards other species. In addition, a painless and dis-
tress-free death can, in some contexts, result in a higher quality meat product for human
PLOS ONE | https://doi.org/10.1371/journal.pone.0240176 October 6, 2020
1 / 14
PLOS ONE CO2 exposure in reptiles
consumption [1]. In pursuit of these goals, methodologies, guidelines, and regulations for
humane euthanasia have been developed and implemented for animal use ranging from meat
production to scientific research [2].
However, a severe taxonomic bias currently exists. Although humane treatment protocols
are well established for mammals and birds, the welfare needs of reptiles and the methodolo-
gies considered humane and acceptable for euthanasia, especially in instances where human
consumption of part of the carcass occurs, remain in their infancy [2]. For example, debate
continues about the appropriateness of hypothermia (freezing) as an euthanasia method [3–5],
and humane killing methods for reptiles used in the meat and skin industries were only
adopted by the World Organization for Animal Health (OIE) in 2019 [see 6, 7].
Chemical agents offer an effective and humane way to euthanize reptiles, but their useful-
ness is sometimes limited. Access and use restrictions, and situations where large numbers of
animals are slaughtered for human consumption in short periods, often prohibit their use.
With the possible exception of hypothermia, all recommended non-chemical methods of rep-
tile euthanasia involve destruction of the brain (e.g., captive bolt, pithing). However, the effec-
tiveness of brain destruction is vulnerable to operator error and may be impractical in
situations where large numbers of animals need to be killed at one time.
Carbon dioxide (CO2) is widely used as a euthanizing agent in the livestock industry and
for scientific research [2, 8–10]. The guidelines of the American Veterinary Medical Associa-
tion cite 86 studies on the effectiveness and suitability of CO2 as a humane means of euthanasia
for mammals and birds [2]. Mammalian and avian responses to CO2 exposure vary consider-
ably by species, and are dependent on CO2 concentration and delivery method [2, 8–10].
Mice, rats, cats, dogs, pigs, rabbits, chickens, and turkeys lose consciousness after 20–120 sec-
onds of CO2 exposure, but may require exposures of 5–50 minutes to ensure death [2, 9, 10].
Exposure to CO2 has been shown to increase plasma corticosterone levels in rats and dogs and
results in mouth gaping in mice, rats, and chickens [2, 9]. Rats and mink will actively avoid
CO2 exposure if given the opportunity, but goats and chickens will not (despite the latter gap-
ing when exposed; [2, 8]).
The use of CO2 to euthanize reptiles has generally been discouraged by veterinary guidance,
animals ethics committees, and by the OIE based on physiological considerations [2, 6, 11, 12].
The rationale implies that because reptiles have a variable metabolic rate and can potentially
tolerate long periods without breathing or oxygen, they are vulnerable to the distressful effects
of suffocation. However, to the best of our knowledge the argumentation against using CO2 to
euthanize reptiles lacks empirical data and rests instead upon untested hypotheses and theoret-
ical inference.
Here, we examine the efficacy of CO2 to humanely euthanize squamate reptiles (lizards and
snakes). Specifically, we tested the potential value of CO2 in (1) creating a low-stress, tempo-
rary unconscious state to make physical methods of euthanasia safer and more efficient and
(2) killing squamates outright. We used both behavioral responses and blood corticosterone
concentrations (the primary glucocorticoid associated with stress in reptiles) to determine
whether CO2 exposure provides a humane transition to unconsciousness and examined how
duration of CO2 exposure influences the post-exposure duration of unconsciousness and like-
lihood of death.
Materials and methods
Study species and locations
Behavioral responses to CO2 exposure were examined in four species of reptile: reticulated
pythons (Malayopython reticulatus); Burmese pythons (Python bivittatus); masked water
PLOS ONE | https://doi.org/10.1371/journal.pone.0240176 October 6, 2020
2 / 14
PLOS ONE CO2 exposure in reptiles
snakes (Homalopsis buccata); and Asian water monitors (Varanus salvator). These species are
semi-aquatic to varying degrees and wide-ranging in Southeast Asia. The two python species
grow to be large (> 5 m), while masked water snakes are relatively small (< 1.2 m). Asian
water monitors are the world’s second largest lizard, growing to 3 metres in length and weigh-
ing as much as 25 kg. In many instances, these species are commensal with humans and are
regularly harvested and traded for their meat, skin, and medicinal value.
In May 2019, we examined responses to CO2 in these reptiles in Malaysia (2˚14’N, 103˚
03’E) and Thailand (17˚38’N, 100˚07’E) at two commercial facilities producing meat for
human consumption and skins for the exotic leather trade. In Malaysia, free-roaming M. reti-
culatus and V. salvator are legally collected from the wild by licensed hunters and brought to
abattoirs for processing [13, 14]. Animals are kept alive at the facility for up to a week before
being killed using a strong blow to the head followed by decapitation. No individual-based his-
tory was available for the animals used in our study, and animals were held according to stan-
dard commercial protocols (i.e., maintained individually in mesh bags with water provided
intermittently). In Thailand, we examined specimens of M. reticulatus, P. bivittatus, and H.
buccata. The two python species were captive-bred for commercial purposes following proto-
cols described in Natusch and Lyons [15]. The H. buccata were wild-caught and temporarily
held in large outdoor ponds with food provided. This research was undertaken with approval
from the Animal Institutional Care and Use Committee of Arizona State University (protocol
# 10-1689R).
Experimental design—behavioral monitoring
To assess behavioral responses of reptiles to CO2 exposure, we placed study animals individu-
ally into 100 micron 375 mm x 500 mm clear plastic bags. Very large animals were double-
bagged as a precaution. CO2 was supplied via 47 litre steel cylinders containing 99.5% CO2
and fitted with single-stage CO2 regulators. A 5 mm inside diameter CO2 supply hose was
placed in the bag through the opening at the top, and the bag was sealed with an elastic band
to limit but not eliminate the escape of gas. Bags were gently compressed around the body of
the animal prior to CO2 admission to minimize residual air pockets. This design enabled CO2
to rapidly displace the limited amount of air present in the bag and thus minimized gas equili-
bration time [16]. By using plastic bags instead of a rigid container, we were able to closely
evaluate the animal during its exposure to CO2 (e.g., examine the animal’s righting response
and its response to touch stimulation). CO2 flow was set to rapidly replace any existing air and
then reduced to maintain positive CO2 pressure in the bag. For the longer exposure times,
once the animal was unconscious, the flow of CO2 was stopped and the bagged was completely
sealed. The process was similar for water monitors except that the bag was secured over their
head rather than placing the entire body inside the bag (to minimize damage to the plastic bag
by the lizard’s claws). We prevented monitors from perforating the bag during movements by
gently placing a hand around the animal’s neck and preventing the forelimbs from contacting
the bag. For some individuals this was not necessary and did not prevent observation of gen-
eral body movements in response to CO2 exposure. For all individuals, the response of the ani-
mal to CO2 exposure was recorded via direct visual examination until the animal was removed
from the bag after the duration of CO2 exposure dictated by its assigned treatment group.
For each animal, we recorded signs of consciousness and all behavioral responses to CO2,
including movement, tongue flicking, and gaping. The animal’s behavior and body move-
ments at the time of removal were recorded, as were changes in behavior over time and the
eventual outcome (i.e., recovery or confirmed death). It was difficult to determine conscious-
ness in many specimens. Although several individuals continued to respond to deep-touch
PLOS ONE | https://doi.org/10.1371/journal.pone.0240176 October 6, 2020
3 / 14
PLOS ONE CO2 exposure in reptiles
stimuli (e.g., a deep pinch of the tail), a lack of righting reflex (failure to turnover when placed
upside down), corneal reflex in lizards, and cessation of breathing, strongly indicated that indi-
viduals were unconscious despite exhibiting a muscular response to deep stimuli. Animals that
reached a state indicative of imminent recovery of consciousness (i.e., voluntary movement
often associated with tongue flicking) were euthanized using standard commercial practices
(i.e., forceful blunt trauma to the dorsal surface of the head at the location of the brain case).
Animals were deemed dead if no heartbeat and/or movements were detected (visually or via
palpation) or by a lack of response to all stimuli (most notably a deep tail pinch) for up to one
hour after removal from CO2 exposure.
To test the effect of CO2 exposure duration on reptile responses, we first conducted a pre-
liminary assessment using different exposure durations on five M. reticulatus (30 min, 60 min,
90 min, 120 min, or 180 min; n = 1 per duration). Based on related observations, we selected
three CO2 exposure durations (15 min, 30 min, and 90 min) for the primary study. We used
the results from the reticulated pythons to select exposure durations for the other species. As
our results from M. reticulatus showed that 15 min was an insufficient duration, we began
studies of other species with the 30 min exposure duration to minimise the number of animals
used and to streamline efforts. If all specimens of the species failed to recover at this exposure
duration, we assumed longer durations would achieve the same result, so did not conduct lon-
ger duration trials. This was not true for H. buccata for which we did not complete the 90 min
exposure treatment due to specimen availability and logistic constraints. We measured snout-
vent length (SVL; using a steel tape measure) and body mass (using a digital scale) of each
specimen while unconscious or dead, and then determined sex via direct inspection of the
gonads upon dissection. Sample sizes for each species and their CO2 exposure times are pre-
sented in Table 1. Air temperature was recorded to confirm constant temperatures throughout
the course of study.
Experimental design–sample collection for hormone monitoring
We measured the effect of the CO2 euthanasia process on circulating corticosterone by collect-
ing blood from seven M. reticulatus and seven V. salvator before and after CO2 exposure. Spec-
imens were brought to the National Wildlife Forensic Laboratory, Department of Wildlife and
National Parks Peninsular Malaysia. Sexes and body sizes are reported in Table 2. Each animal
was kept individually within a mesh bag and secured within a plastic crate at ambient tempera-
ture for two days before trials began. We collected 2 ml of blood from each individual within
Table 1. Means, standard errors and ranges for snout-vent length (SVL) and body mass for reptile specimens used to examine behavioral responses to CO2
exposure.
Species
Thailand
Malayopython reticulatus
Python bivittatus
Homalopsis buccata
Malaysia
Malayopython reticulatus
Varanus salvator
Sex
M
F
M
M
M
F
M
F
N
1
3
18
11
12
14
5
5
https://doi.org/10.1371/journal.pone.0240176.t001
SVL (cm)
Mass (g)
N per exposure duration
Mean
Range
Mean
Range
15 min
30 min
90 min
273
265.3 ± 8.9
241.5 ± 2.7
104 ± 2.2
272.8 ± 8.6
297.4 ±8.3
63 ± 3.3
59 ± 3.8
-
255–283
220–263
93–116
238–331
255–374
50–68
52–71
8200
7200 ± 1790
6941 ± 545
686 ± 36
7335 ± 728
7878 ± 608
4990 ± 708
4000 ± 714
-
4200–10400
3900–11800
530–850
4550–13450
4050–12850
2250–6350
2550–6000
0
0
0
0
3
5
0
0
1
3
9
8
4
6
5
5
0
0
9
0
4
4
0
0
PLOS ONE | https://doi.org/10.1371/journal.pone.0240176 October 6, 2020
4 / 14
PLOS ONE CO2 exposure in reptiles
Table 2. Means, standard errors and ranges for snout-vent length (SVL) and body mass for reptile specimens used to examine plasma corticosterone responses to
CO2 exposure.
Species
Treatment
Sex
Malayopython reticulatus
CO2
Varanus salvator
Control
CO2
Control
https://doi.org/10.1371/journal.pone.0240176.t002
M
F
M
F
M
F
M
F
N
3
4
2
2
2
5
2
1
SVL (cm)
Mass (g)
Mean
246 ± 5.6
253.5 ± 4.6
295 ± 55
375 ± 25
53.7 ± 1.8
56.2 ± 2.9
79 ± 10
69
Range
235–255
240–260
240–350
350–400
51–57
47–63
69–89
-
Mean
4720 ± 204
5280 ± 225
8500 ± 3500
35000 ± 0
2830 ± 233
2900 ± 370
7850 ± 2350
6500
Range
4400–5100
4720–5800
5000–12000
35000
2600–3300
1500–3750
5500–10200
-
90 seconds of removal from the mesh bag using a 22 gauge needle and 5 ml syringe inserted
into the caudal vein at the base of the tail. The blood sample was then placed in a tube contain-
ing lithium heparin (Vacuette #454084, Greiner Bio-One, Kremsmu¨nster, Austria). After
blood collection, the same specimens were immediately exposed to CO2. A second blood sam-
ple was collected from the same specimen after 15 minutes of CO2 exposure when the animal
was unconscious. We did this by amputating the lower third of the tail and collecting the
blood directly into a heparinized tube. The animal was then immediately euthanized following
standard methods as described above. Blood samples were placed on ice until centrifugation to
separate the plasma. We stored the isolated plasma samples at -20˚C until they were assayed.
As confinement in the mesh bag may in itself result in elevated levels of corticosterone, we col-
lected blood samples from several ‘control’ animals for comparison. The control water moni-
tors (n = 3) were freshly killed wild animals harvested during a government sanctioned control
program in Ladang Eng Tai, Malaysia (4˚57’N 100˚27’E). Animals were harvested using a
12-gauge shotgun at close range, with head shots resulting in near-instantaneous death. We
collected blood from the severed tail of each animal within 90 seconds using the same method
described above. Control reticulated python (n = 4) samples were obtained from captive-bred
animals at a commercial reptile breeding facility outside Kuala Lumpur, Malaysia (2˚56’N
101˚53’E). The farm breeds high-value pythons for the pet trade, and general husbandry and
welfare standards are high. Animals were selected based on size and relative docility (i.e., ease
of handling), and blood samples were collected from the caudal vein within 90 seconds of
removal from their enclosures using the same method described above. We recorded tempera-
tures (27–30˚C) and kept all animals at approximately the same temperature both before and
after exposure to CO2. This was not possible for control specimens sampled in the wild, but
plasma corticosterone levels are not highly sensitive to body temperature in reptiles [17]. We
obtained all blood samples over several hours on the same day to avoid diel and seasonal varia-
tion in plasma hormone levels.
Hormone analysis
Immunoreactive plasma corticosterone concentrations were determined via an enzyme-linked
immunosorbent assay (ELISA; ADI-900-097, Enzo Life Sciences, Farmingdale, NY) following
the manufacturer’s instructions. This kit has been used in previous studies assessing plasma
corticosterone concentrations in a variety of animal species, including alligators [18], birds
[19], lizards [20] and turtles [21], but had not been previously documented for pythons or
monitor lizards. Based on results from other species, we used a dilution ratio of 40:1. All sam-
ples were run in duplicate format on a single assay plate. Results confirmed an average
PLOS ONE | https://doi.org/10.1371/journal.pone.0240176 October 6, 2020
5 / 14
PLOS ONE CO2 exposure in reptiles
difference between duplicates of less than 1.8% (mean: 1.73 ± 1.18%), and duplicate means
were thus used in the analysis.
Data analysis
Our behavioral analysis measured the binary dependent variable of whether reptiles recovered
after CO2 exposure or not. This metric was evaluated after different CO2 exposure durations
for each species. For our corticosterone study we used a paired sample t-test to test for signifi-
cant differences in plasma corticosterone concentrations before and after CO2 exposure. We
used a one-way analysis of variance to test for differences in corticosterone level between the
control animals and the pre-CO2 exposure samples from the study animals. Data were ln-
transformed where needed to meet the normality and homogeneity of variance assumptions
required for our parametric tests. All analyses were conducted in JMP Pro 14 (SAS Institute,
Cary, NC).
Results
Behavioral observations
Reticulated pythons (Malayopython reticulatus). After exposure to CO2, reticulated
pythons remained still for 60–300 secs (1–5 mins) before tongue flicking and gaping (Fig 1).
These responses eventually proceeded to slow and controlled whole-body movements; at this
time snakes were responsive to touch through the bag. It was difficult to determine the point at
which snakes lost full consciousness. However, we suspect that snakes lost consciousness, but
continued to undergo unconscious movements including a response to touch stimuli. Between
240–1380 secs (4–23 mins) after CO2 exposure the snakes ceased all movements and lost
responsiveness to stimuli (Fig 1). After the cessation of movement, but sometimes before, 18
of the 30 snakes exhibited mild muscle twitching of parts of their body. This twitching was
unique to the reticulated pythons.
All Malaysian reticulated pythons that were exposed to CO2 for 15 and 30 min eventually
recovered (Fig 2). At the time of removal from the bag, none of these snakes had voluntary
movements, but 7 of 8 snakes in the 15-min exposure group and 1 of 10 snakes in the 30-min
group responded to a deep tail pinch with local movement. First voluntary movements
occurred 4.9 ± 0.9 (mean ± SE) and 23.8 ± 4.7 min after removal from CO2 for the 15 min and
30-min exposure groups, respectively. In contrast, all reticulated pythons exposed to 90-min of
CO2 did not recover, never having any reflex or voluntary movements (Fig 2). Reticulated
pythons tested in Thailand that were exposed to CO2 for 30 min responded similarly to those
in Malaysia, but one of the four snakes did not recover and, for those that did, recovery took
13.7 ± 3.7 min (42% faster than the 30-min exposure snakes in Malaysia).
Burmese pythons (Python bivittatus). Burmese pythons showed similar behavioral
responses to reticulated pythons, but took slighter longer to gape and lose responsiveness to
stimuli (Fig 1). Burmese pythons also did not undergo muscle twitching and late-stage non-
responsive (likely unconscious) movements were greater. All 8 snakes in the 30-min group
responded to a deep tail pinch upon removal from the CO2, while none of the 90-min snakes
responded. Two of the 8 snakes exposed to CO2 for 30 min and all of the snakes exposed to
CO2 for 90 min did not recover (Fig 2). For the six 30-min snakes that did recover, it took
17.4 ± 2.5 min until they showed their first voluntary movements.
Masked water snakes (Homalopsis buccata). The water snakes exposed to CO2 for 30
min showed behavioral responses that were very similar to those of the Burmese pythons, with
no twitching but a considerable amount of unconscious movements. Mean time of first gape
was about 120 secs (range: 60–420 secs, 1–7 min) and complete loss of consciousness was 300–
PLOS ONE | https://doi.org/10.1371/journal.pone.0240176 October 6, 2020
6 / 14
PLOS ONE CO2 exposure in reptiles
Fig 1. Variation in timing (in minutes) of key behavioural changes in (a) Malayopython reticulatus, (b) Python bivittatus, (c) Homalopsis buccata, and (d) Varanus
salvator subject to carbon dioxide (CO2) exposure. Gaping: the time at which the mouth of the specimen opened. Unresponsive: the time the specimen had ceased
movement and became unresponsive to stimuli. Thicker parts of the violin plots represent CO2 exposure times where the behaviour was most often observed. Note the
different time scales represented on the x-axes of each panel.
https://doi.org/10.1371/journal.pone.0240176.g001
840 secs (5–14 mins) after the onset of exposure (Fig 1). While all eight water snakes had a tail
pinch reflex upon removal from the CO2, only two of the eight snakes recovered after 10 and
20 min, respectively.
Water monitors (Varanus salvator). The water monitors showed the least behavioral
response to exposure to CO2. The lizards exhibited no tongue flicking and no muscle twitching
during the 30 min exposure. All monitors gaped within 240 secs (4 mins) of the onset of CO2
exposure (Fig 1) Both conscious and unconscious movements were limited in number and
intensity with the last detected movements occurring 930 ± 66 secs (range: 720–1560 seconds)
after the onset of exposure (Fig 1). All monitors lacked a tail pinch reflex when removed from
the CO2, and they all failed to recover (Fig 2).
Plasma corticosterone concentrations
Corticosterone concentrations for the animals that did not go through the capture and con-
finement associated with the trade prior to killing (i.e., ‘controls’) were significantly lower than
those of the CO2-euthanized animals prior to CO2 exposure (pythons: 7.2 ± 1.3 ng/ml; F1,10 =
9.01, P = 0.015; monitors: 3.1 ± 0.7 ng/ml; F1,10 = 24.4, P < 0.001; Fig 3). Reticulated python
PLOS ONE | https://doi.org/10.1371/journal.pone.0240176 October 6, 2020
7 / 14
PLOS ONE CO2 exposure in reptiles
Fig 2. Percentage of Malayopython reticulatus, Python bivittatus, Homalopsis buccata, and Varanus salvator that recovered from different durations of CO2
exposure. X denotes treatments where no individuals recovered from CO2 exposure. Sample sizes appear above each column.
https://doi.org/10.1371/journal.pone.0240176.g002
plasma corticosterone concentrations increased by 14% after CO2 exposure, (t0 = 11.8 ± 0.9
ng/ml vs t15 = 13.2 ± 0.4 ng/ml). However, this increasing trend was not statistically significant
(matched pairs t-test: t6 = 2.23, P = 0.065; Fig 3). In contrast, CO2 exposure significantly
increased plasma corticosterone concentrations in water monitors (by 18%; t0 = 9.6 ± 0.9 ng/
ml; t15 = 11.7 ± 0.8 ng/ml; t6 = 5.03, P = 0.02; Fig 3). Individual immunoreactive plasma corti-
costerone concentrations before and after CO2 exposure were significantly correlated
(pythons: n = 7; r2 = 0.61; P = 0.037; lizards: n = 8; r2 = 0.77; P = 0.009).
Discussion
Although available euthanasia methods for commercial reptile processing (e.g., brain destruc-
tion) are humane, they can be vulnerable to operator error, are aesthetically displeasing, and
are inefficient for rapidly processing numerous individuals. Our study provides initial results
supporting the potential for carbon dioxide asphyxiation as an effective option for euthanizing
PLOS ONE | https://doi.org/10.1371/journal.pone.0240176 October 6, 2020
8 / 14
PLOS ONE CO2 exposure in reptiles
Fig 3. Mean plasma corticosterone concentrations (ng/ml) before and after 15 minutes of CO2 exposure and in control specimens (free-ranging or
farmed; see text) of (a) Malayopython reticulatus and (b) Varanus salvator. Differences between corticosterone concentrations before and after CO2
exposure were not statistically significant for M. reticulatus, but were for V. salvator. Corticosterone concentrations between control specimens not
subject to capture and handling are significantly lower than those captured from the wild for trade (although sample sizes were low; see text for details).
Sample sizes for each group are reported directly above the x-axis.
https://doi.org/10.1371/journal.pone.0240176.g003
reptiles in a variety of settings. Exposure to CO2 was effective for creating a temporary uncon-
scious state at all exposure durations that was sufficient to safely and humanely employ a phys-
ical method of euthanasia. Longer but still logistically practical exposures to CO2 were able to
kill reptiles.
The different taxa in our study varied subtly in their responses to CO2 exposure, both while
conscious and after losing consciousness. For example, despite the similar body size of the two
python species, the CO2 exposure duration required to induce unconsciousness in P. bivittatus
was greater than M. reticulatus (Fig 1). The only lizard species in our study was rapidly ren-
dered unconscious and did not recover from CO2 exposure durations that were unable to kill
most of the snakes (Fig 2). Taxonomic differences and variation in metabolic rates may both
be responsible for this difference [22–24]. The species we studied also differed in the effects
PLOS ONE | https://doi.org/10.1371/journal.pone.0240176 October 6, 2020
9 / 14
PLOS ONE CO2 exposure in reptiles
that a given duration of CO2 exposure had once the animal was removed from CO2, including
the extent of involuntary/reflex muscle activity and the likelihood of death. Unfortunately, we
did not have a sufficient sample size to examine sexual differences in species’ responses to CO2
exposure. Plausibly, CO2 may affect males and females differently, especially in those species
with strong sexual dimorphism. Related to this, our study was undertaken on several of the
world’s largest reptiles, all of which are semi-aquatic and can remain submerged under water
for considerable periods. Application of CO2 exposure to the myriad of smaller-bodied rep-
tiles, and to strictly terrestrial species, may yield different results.
We made the assumption that because the density of CO2 is greater than air, complete
(100%) CO2 saturation would occur as air was expelled from the small opening positioned at
the top of the bag [25]. However, we did not directly measure the concentration of CO2 within
the bag and whether the concentration was homogenous. Layering of CO2 could enable speci-
mens to avoid exposure [2]. The variation in responses to CO2 exposure in our study may be
related to minor but functionally significant difference in CO2 distribution [see 26]. In order
to more broadly apply CO2 as a euthanasia method in reptiles, there needs to be a better under-
standing of interspecific difference among taxa as well as a delivery system with established
displacement parameters and sufficient holding capacity.
Regardless of species, our behavioral observations suggest the reptiles used in our study do
not suffer significant distress from CO2 exposure. Although our observational assessments
were subjective, the body movements made by conscious reptiles were minor and appeared
considerably less vigorous than the escape behavior displayed by these same animals when first
removed from their holding bags. In the case of V. salvator, some specimens went unconscious
without showing any signs of movement. Nevertheless, it is challenging to accurately deter-
mine if reptiles are indeed dead, let alone feeling pain, based solely on behavioral responses
[27, 28]. For example, an active heartbeat, involuntary movements, and response to touch sti-
muli can continue for hours after complete destruction, pithing, and removal of the brain
[Natusch unpubl. data 2020, 2]. Similarly, our data on the time reptiles take to lose responsive-
ness are difficult to interpret. It was often unknown if specimens were consciously responsive,
or unconscious and merely exhibiting involuntary muscular reflex. Importantly, the difficulty
of assuring death, and the high but less than 100% effectiveness at killing at some CO2 expo-
sure durations, may warrant the use of a secondary method to ensure death as is commonly
used for chemical-induced euthanasia of research animals [see 2].
The most consistent behavioral response to CO2 exposure was the non-violent gaping dis-
played by most (90%) individuals. Gaping is common in mammals and birds subject to CO2
exposure, and in birds does not appear to be a sign of distress when exposed to CO2 [29]. It is
unknown whether gaping is a sign of significant distress in reptiles. Gaping occurred within 30
seconds to 16 minutes of initiating CO2 exposure and the timing varied among taxa (Fig 1). The
short duration between initial exposure and gaping, and then unconsciousness, suggests that
suffocation may not be the cause of death in reptiles exposed to CO2. All species used in our
study are semi-aquatic, and capable of spending significant time underwater (>20 minutes),
suggesting another physiological response is taking place. Despite the lack of behavioral indica-
tors for stress and pain, reptiles take considerably longer to lose consciousness than mammals
and birds [30–32]. Some consider a gentle death that takes longer is preferable to a rapid but
more distressing death [26, 33]. In the context of CO2 and reptiles, further research is needed.
Our additional approach to investigate the impact of CO2 exposure in our study species, by
monitoring plasma corticosterone concentrations, also suggests that reptiles experience rela-
tively minor distress from CO2 exposure. Comparison to our control (wild or farmed) speci-
mens suggests the relative increase in stress involved in restraint and transportation of
specimens to the laboratory was greater than the distress induced by CO2 exposure [2, 34].
PLOS ONE | https://doi.org/10.1371/journal.pone.0240176 October 6, 2020
10 / 14
PLOS ONE CO2 exposure in reptiles
Brown tree snakes (Boiga irregularis) and red-sided garter snakes (Thamnophis sirtalis) cap-
tured and placed in bags for 2–4 hours increased plasma corticosterone levels by 280–1200%
[35, 36], but resulted in no appreciable increase in corticosterone concentrations in bearded
dragons (Pogona barbata) [37]. Several studies reveal a lack of adverse impacts of corticoste-
rone increase on survival, feeding behavior, and reproduction [38–40]. Other studies docu-
ment invasive procedures (e.g., toe clipping, microchipping) inducing smaller corticosterone
increases than did natural stresses experienced in the wild [27]. The relatively small increases
in plasma corticosterone concentrations observed in pythons (14%) and lizards (18%) in our
study may suggest that the functional relevance (distress or pain) of CO2 exposure-induced
increases in corticosterone may be negligible. It is possible that the small increases in cortico-
sterone levels we observed were related mostly to the stress caused by restraining and collect-
ing an initial (T0) blood sample from each specimen, rather than by exposure to the CO2 itself.
Alternatively, a post-CO2 exposure increase in corticosterone may have been suppressed
because the recent capture, confinement, and handling had already maximized the hypotha-
lamic-pituitary-adrenal (HPA) axis response.
Intriguingly, exposure to CO2 may have additional benefits beyond the possibility of a pain-
less death. After death, animals can have spinal cord induced muscle activity, and this can last
for an extended duration in reptiles due to their tissue’s high tolerance of hypoxia. This phe-
nomenon can lead to the impression that the animals is still alive [2], and thus has been capi-
talized on by activists who oppose the consumption of animals, claiming they are being
processed while still alive. In addition to being aesthetically displeasing, continued muscle
movements after death force staff in commercial facilities to delay the harvesting of tissues for
up to two hours after death [41]. When killed via CO2 exposure, we recorded no involuntary
muscle movements after the presumed point of death, including during the processing of the
reptiles. The physiological cause of this lack of muscle tone is unknown but, given its func-
tional and cosmetic advantages, warrants further investigation.
In conclusion, our study presents some of the first results on the effects of CO2 exposure in
reptiles. We stress that our results are preliminary and therefore are reluctant to recommend
CO2 as a humane method of reptile euthanasia at this time. Despite our results being generally
positive, we identified some interspecific differences and methodological variables that may
influence the effectiveness of CO2 exposure. Future studies could usefully disentangle the
influence of these variables and employ alternative methods for assessing stress, pain, and
death in reptiles (e.g., electroencephalography).
Supporting information
S1 Data. CO analyses.
(XLSX)
Acknowledgments
We thank Yuan Wai Lek reptile trading company, Sisatchanalai python farm, and Lim Maju
Jaya Trading for providing the animals used in this study. We also thank the Malaysian
Department of Wildlife and National Parks Peninsular Malaysia for providing access to their
forensic laboratory and equipment. We thank anonymous reviewers for comments that
improved an earlier draft of this manuscript.
Author Contributions
Conceptualization: Daniel J. D. Natusch, Patrick W. Aust, Dale F. DeNardo.
PLOS ONE | https://doi.org/10.1371/journal.pone.0240176 October 6, 2020
11 / 14
PLOS ONE CO2 exposure in reptiles
Data curation: Daniel J. D. Natusch, Dale F. DeNardo.
Formal analysis: Daniel J. D. Natusch, Andre Ganswindt, Dale F. DeNardo.
Funding acquisition: Daniel J. D. Natusch.
Investigation: Daniel J. D. Natusch, Patrick W. Aust, Syarifah Khadiejah, Hartini Ithnin, Ain
Isa, Che Ku Zamzuri, Dale F. DeNardo.
Methodology: Daniel J. D. Natusch, Patrick W. Aust, Syarifah Khadiejah, Hartini Ithnin,
Andre Ganswindt, Dale F. DeNardo.
Project administration: Daniel J. D. Natusch, Patrick W. Aust.
Resources: Daniel J. D. Natusch, Hartini Ithnin, Ain Isa, Che Ku Zamzuri.
Software: Daniel J. D. Natusch.
Supervision: Daniel J. D. Natusch, Patrick W. Aust, Dale F. DeNardo.
Validation: Daniel J. D. Natusch.
Visualization: Daniel J. D. Natusch.
Writing – original draft: Daniel J. D. Natusch, Patrick W. Aust, Andre Ganswindt, Dale F.
DeNardo.
Writing – review & editing: Daniel J. D. Natusch, Patrick W. Aust, Syarifah Khadiejah, Har-
tini Ithnin, Ain Isa, Che Ku Zamzuri, Andre Ganswindt, Dale F. DeNardo.
References
1. Chambers PG, Grandin T. Guidelines for humane handling, transport and slaughter of livestock. Heinz
G, Srisuvan T, editors. Humane Society and International 2001. [accessed on 28 May 2020]. pp. 1–17.
Available from: http://www.fao.org/3/a-x6909e.pdf.
2.
Leary SL, Underwood W, Anthony R, Cartner S, Corey D, Grandin T, et al. AVMA Guidelines for the
Euthanasia of Animals: 2020 Edition. American Veterinary Medical Association\. Schaumburg, IL.
[(accessed on 30 May 2020)]. Available from: https://www.avma.org/sites/default/files/2020-01/2020-
Euthanasia-Final-1-17-20.pdf.
3. Shine R, Amiel J, Munn AJ, Stewart M, Vyssotski AL, Lesku JA. Is “cooling then freezing” a humane
way to kill amphibians and reptiles? Biol Open 2015; 4: 760–763. https://doi.org/10.1242/bio.012179
PMID: 26015533
4. Shine R, Lesku JA, Lillywhite HB. Assessment of the cooling-then-freezing method for euthanasia of
amphibians and reptiles. J Am Vet Med Assoc 2019; 255: 48–50. https://doi.org/10.2460/javma.255.1.
48 PMID: 31194656.
5.
Lillywhite HB, Shine R, Jacobson E, DeNardo DF, Gordon MS, Navas CA, et al. Anesthesia and eutha-
nasia of amphibians and reptiles used in scientific research: should hypothermia and freezing be prohib-
ited? Bioscience 2017; 67:53–61. https://doi.org/10.1093/biosci/biw143
6. Swiss Expert Panel. Analysis of humane killing methods for reptiles in the skin trade. Swiss Federal Vet-
erinary Office. 2013. [(accessed on 30 May 2020)]. Available from: https://recht.pogona.ch/data/_
uploaded/file/3.0%20F%C3%BCtterung/5.2_BVET_Analysis%20on%20humane%20killing%
20methods%20for%20reptiles%20in%20the%20skin%20trade%20frm%5B1%5D.pdf.
7. World Organisation for Animal Health. Terrestrial Animal Health Code, Chapter 7.14. 2019 [accessed
on 29 June 2020)]. Available from: https://www.oie.int/international-standard-setting/terrestrial-code/.
8. Withrock IC. The use of carbon dioxide (CO2) as an alternative euthanasia method for goat kids. M.Sc.
Thesis, Iowa State University. 2015. Available from: https://lib.dr.iastate.edu/etd/14718.
9. Boivin GP, Hickman DL, Creamer-Hente MA, Pritchett-Corning KR, Bratcher NA. Review of CO2 as a
euthanasia agent for laboratory rats and mice. J Am Assoc Lab Anim Sci. 2017; 56: 491–499. PMID:
28903819.
10. C¸ avuşoğlu E, Rault JL, Gates R, Lay DC. Behavioral Response of Weaned Pigs during Gas Euthanasia
with CO2, CO2 with Butorphanol, or Nitrous Oxide. Animals 2020; 10: 787. https://doi.org/10.3390/
ani10050787 PMID: 32370086
PLOS ONE | https://doi.org/10.1371/journal.pone.0240176 October 6, 2020
12 / 14
PLOS ONE CO2 exposure in reptiles
11. Close B, Banister K, Baumans V, Bernoth EM, Bromage N, Bunyan J, et al. Recommendations for
euthanasia of experimental animals: Part 2. Lab anim 1997; 31: 1–32. https://doi.org/10.1258/
002367797780600297 PMID: 9121105.
12. Warren K. Reptile Euthanasia—No Easy Solution? Pac Conserv Biol 2014; 20: 25–27. https://doi.org/
10.1071/PC140025
13. Khadiejah S, Razak N, Ward-Fear G, Shine R, Natusch DJD. Asian water monitors (Varanus salvator)
remain common in Peninsular Malaysia, despite intense harvesting. Wildl Res 2019; 46: 265–275.
https://doi.org/10.1071/WR18166
14. Natusch DJD, Lyons JA, Riyanto A, Mumpuni, Khadiejah S, Shine R. Detailed biological data are infor-
mative, but robust trends are needed for informing sustainability of wildlife harvesting: A case study of
reptile offtake in Southeast Asia. Biol Conserv 2019; 233: 83–92. https://doi.org/10.1016/j.biocon.2019.
02.016
15. Natusch DJD, Lyons JA. Assessment of python breeding farms supplying the international high-end
leather industry. A report under the ‘Python Conservation Partnership’ programme of research. Occa-
sional Paper of the IUCN Species Survival Commission 2014; 50. [(accessed on 15 May 2020)]. Avail-
able from: https://portals.iucn.org/library/sites/library/files/documents/SSC-OP-050.PDF.
16.
Lasiewski RC, Acosta AL, Bern-stein MH. Evaporative water loss in birds. 1. Characteristics of the open
flow method of determination, and their relation to estimates of thermoregulatory ability. Comp Biochem
Physiol 1966; 19:445–457.
17. Sykes KL, Klukowski M. Effects of acute temperature change, confinement and housing on plasma cor-
ticosterone in water snakes, Nerodia sipedon (Colubridae: Natricinae). J Exp Zool 2009; 311A: 172–
181. https://doi.org/10.1002/jez.515 PMID: 19051318.
18.
Finger JW Jr, Hamilton MT, Kelley MD, Stacy NI, Glenn TC, Tuberville TD. Examining the Effects of
Chronic Selenium Exposure on Traditionally Used Stress Parameters in Juvenile American Alligators,
Alligator mississippiensis. Arch Environ Contam Toxicol. 2019; 77(1):14–21. https://doi.org/10.1007/
s00244-019-00626-9 PMID: 30976886
19. Xie S, Romero LM, Htut ZW, McWhorter TJ. Stress responses to heat exposure in three species of Aus-
tralian desert birds. Physiol Biochem Zool. 2017; 90(3):348–358. https://doi.org/10.1086/690484 PMID:
28384428
20. Seddon RJ, Hews DK. Populations of the Lizard, Sceloporus occidentalis, that differ in melanization
have different rates of wound healing. J Exp Zool A Ecol Genet Physiol. 2016; 325(8):491–500. https://
doi.org/10.1002/jez.2033 PMID: 27597293
21. West JM, Klukowski M. Seasonal changes in baseline corticosterone, association with innate immunity,
and effects of confinement in free-ranging Eastern Box Turtles, Terrapene carolina carolina. Gen Comp
Endocrinol. 2018; 262:71–80. https://doi.org/10.1016/j.ygcen.2018.03.016 PMID: 29548757
22. Chappell MA, Ellis TM. Resting metabolic rates in boid snakes: allometric relationships and temperature
effects. J Comp Physiol B 1987; 157: 227–235. https://doi.org/10.1007/BF00692367 PMID: 3106432.
23.
Thompson G, Heger N, Heger T, Withers P. Standard metabolic rate of the largest Australian lizard,
Varanus giganteus. Comp Biochem Physiol A Mol Integr Physiol 1995; 111: 603–608.
24. Hopkins W, Roe J, Philippi T, Congdon J. Standard and digestive metabolism in the banded water
snake, Nerodia fasciata fasciata. Comp Biochem Physiol A Mol Integr Physiol 2004; 137: 141–149.
https://doi.org/10.1016/j.cbpb.2003.09.017 PMID: 14720599.
25. Murray D. (2009). CO2 euthanasia methods for neonatal piglets. Allen D. Lemon Swine Conference;
2009 University of Minnesota: Minneapolis, MN United States of America. p. 114–116. [(accessed on
15 May 2020)]. Available from: http://hdl.handle.net/11299/139762.
26. Coenen AM, Drinkenburg WH, Hoenderken R, van Luijtelaar EL. Carbon dioxide euthanasia in rats:
oxygen supplementation minimizes signs of agitation and asphyxia. Lab Anim 1995; 29: 262–268.
https://doi.org/10.1258/002367795781088289 PMID: 7564209.
27.
Langkilde T, Shine R. How much stress do researchers inflict on their study animals? A case study
using a scincid lizard, Eulamprus heatwolei. J Exp Biol 2006; 209:1035– 1043. https://doi.org/10.1242/
jeb.02112 PMID: 16513929.
28. Eatwell K. Options for analgesia and anaesthesia in reptiles. In Pract 2010; 32: 306–311. https://doi.org/
10.1136/inp.c3917
29. McKeegan DEF, McIntyre J, Demmers TGM, Wathes CM, Bryan Jones R. Behavioural responses of
broiler chickens during acute exposure to gaseous stimulation. Appl Anim Behav Sci 2006; 99: 271–
286. https://doi.org/10.1016/j.applanim.2005.11.002
30. Hewett TA, Kovacs MS, Artwohl JE, Bennett BT. A comparison of euthanasia methods in rats, using
carbon dioxide in prefilled and fixed flow-rate filled chambers. Lab Anim Sci 1993; 43: 579–582. PMID:
8158983.
PLOS ONE | https://doi.org/10.1371/journal.pone.0240176 October 6, 2020
13 / 14
PLOS ONE CO2 exposure in reptiles
31. Shea SA, Harty HR, Banzett RB. Self-control of level of mechanical ventilation to minimize CO2-
induced air hunger. Respir Physiol 1996; 103: 113–125. https://doi.org/10.1016/0034-5687(95)00086-0
PMID: 8833543.
32. Reed B, Varon J, Chait BT, Kreek MJ. Carbon dioxide-induced anaesthesia result in a rapid increase in
plasma levels of vasopressin. Endocrinology 2009; 150: 2934–2939. https://doi.org/10.1210/en.2008-
1408 PMID: 19213839
33. Hawkins P, Playle L, Golledge H, Leach M, Banzett R, Coenen A, et al. Newcastle Consensus Meeting
on Carbon Dioxide Euthanasia of Laboratory Animals; Newcastle upon Tyne, UK. 27–28 February
2006; [(accessed on 28 May 2020)]. pp. 1–17. Available from: www.nc3rs.org.uk/downloaddoc.asp?id=
416&page=292&skin=0.
34. Sharp J, Azar T, Lawson D. Comparison of carbon dioxide, argon, and nitrogen for inducing uncon-
sciousness or euthanasia of rats. J Am Assoc Lab Anim Sci 2006; 45: 21–25. PMID: 16542038.
35. Moore IT, Lemaster MP, Mason RT. Behavioural and hormonal responses to capture stress in the male
red-sided garter snake, Thamnophis sirtalis parietalis. Anim Behav 2000; 59: 529e534. https://doi.org/
10.1006/anbe.1999.1344 PMID: 10715174.
36. Mathies T, Felix TA, Lance VA. Effects of trapping and subsequent short-term confinement stress on
plasma corticosterone in the brown treesnake, Boiga irregularis on Guam. Gen Comp Endocrinol 2001;
124(1):106–114. https://doi.org/10.1006/gcen.2001.7694 PMID: 11703076.
37. Cree A, Amey AP, Whittier JM. Lack of consistent hormonal responses to capture during the breeding
season of the bearded dragon, Pogona barbata. Comp Biochem Physiol 2000; 126: 275–285. https://
doi.org/10.1016/j.yhbeh.2005.08.004 PMID: 16153645.
38. Cote J, Clobert J, Meylan S, Fitze P. Experimental enhancement of corticosterone levels positively
affects subsequent male survival. Horm Behav 2006; 49: 320–327. https://doi.org/10.1016/j.yhbeh.
2005.08.004 PMID: 16153645.
39.
Thaker M, Vanak AT, Lima SL, Hews DK. Stress and aversive learning in a wild vertebrate: the role of
corticosterone in mediating escape from a novel stressor. Am Nat 2010; 175: 50–60. https://doi.org/10.
1086/648558 PMID: 19922261.
40. Dupoue´ A, Angelier F, Brischoux F, DeNardo DF, Trouve´ C, Parenteau C, et al. Water deprivation
increases maternal corticosterone levels and enhances offspring growth in the snake, Vipera aspis. J
Exp Biol 2016; 219: 658–667. https://doi.org/10.1242/jeb.132639 PMID: 26747902.
41. Aust PW, Webb GJW, DeNardo DF, Natusch DJD. Welfare Principles for Snakes and Monitor Lizards
in the Southeast Asian Skin Trade–a guide for stakeholders. Swiss Federal Veterinary Office, Switzer-
land; Southeast Asian Reptile Conservation Alliance, France 2019.
PLOS ONE | https://doi.org/10.1371/journal.pone.0240176 October 6, 2020
14 / 14
PLOS ONE |
10.1371_journal.pone.0240269 | RESEARCH ARTICLE
Using association rule mining to jointly detect
clinical features and differentially expressed
genes related to chronic inflammatory
diseases
Rosana VeronezeID
RochaID
Fernando J. Von Zuben1, Raquel Mantuaneli Scarel-Caminaga2
3, Cla´ udia V. Maurer-Morelli3, Silvana Regina Perez Orrico4,5, Joni A. CirelliID
1*, Saˆ mia Cruz Tfaile Corbi2, Ba´ rbara Roque da Silva2, Cristiane de S.
4,
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: Veroneze R, Cruz Tfaile Corbi S, Roque da
Silva B, de S. Rocha C, V. Maurer-Morelli C, Perez
Orrico SR, et al. (2020) Using association rule
mining to jointly detect clinical features and
differentially expressed genes related to chronic
inflammatory diseases. PLoS ONE 15(10):
e0240269. https://doi.org/10.1371/journal.
pone.0240269
Editor: Paolo Magni, Università degli Studi di
Milano, ITALY
Received: June 25, 2020
Accepted: September 23, 2020
Published: October 2, 2020
Peer Review History: PLOS recognizes the
benefits of transparency in the peer review
process; therefore, we enable the publication of
all of the content of peer review and author
responses alongside final, published articles. The
editorial history of this article is available here:
https://doi.org/10.1371/journal.pone.0240269
Copyright: © 2020 Veroneze et al. This is an open
access article distributed under the terms of the
Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: Genomics data are
now available at the NCBI repository: https://www.
1 Department of Computer Engineering and Industrial Automation, School of Electrical and Computer
Engineering, University of Campinas (UNICAMP), Campinas, SP, Brazil, 2 Department of Morphology,
Genetics, Orthodontics and Pediatric Dentistry, School of Dentistry at Araraquara, São Paulo State
University (UNESP), Araraquara, SP, Brazil, 3 Department of Medical Genetics and Genomic Medicine,
University of Campinas (UNICAMP), Campinas, SP, Brazil, 4 Department of Diagnosis and Surgery, School
of Dentistry at Araraquara, São Paulo State University (UNESP), Araraquara, SP, Brazil, 5 Advanced
Research Center in Medicine, Union of the Colleges of the Great Lakes (UNILAGO), São Jose´ do Rio Preto,
SP, Brazil
* veroneze@dca.fee.unicamp.br, rveroneze@gmail.com
Abstract
Objective
It is increasingly common to find patients affected by a combination of type 2 diabetes melli-
tus (T2DM), dyslipidemia (DLP) and periodontitis (PD), which are chronic inflammatory dis-
eases. More studies able to capture unknown relationships among these diseases will
contribute to raise biological and clinical evidence. The aim of this study was to apply associ-
ation rule mining (ARM) to discover whether there are consistent patterns of clinical features
(CFs) and differentially expressed genes (DEGs) relevant to these diseases. We intend to
reinforce the evidence of the T2DM-DLP-PD-interplay and demonstrate the ARM ability to
provide new insights into multivariate pattern discovery.
Methods
We utilized 29 clinical glycemic, lipid and periodontal parameters from 143 patients divided
into five groups based upon diabetic, dyslipidemic and periodontal conditions (including a
healthy-control group). At least 5 patients from each group were selected to assess the tran-
scriptome by microarray. ARM was utilized to assess relevant association rules considering:
(i) only CFs; and (ii) CFs+DEGs, such that the identified DEGs, specific to each group of
patients, were submitted to gene expression validation by quantitative polymerase chain
reaction (qPCR).
PLOS ONE | https://doi.org/10.1371/journal.pone.0240269 October 2, 2020
1 / 22
PLOS ONE Association rule mining to detect clinical features and genes related to chronic inflammatory diseases
Results
We obtained 78 CF-rules and 161 CF+DEG-rules. Based on their clinical significance, Peri-
odontists and Geneticist experts selected 11 CF-rules, and 5 CF+DEG-rules. From the five
DEGs prospected by the rules, four of them were validated by qPCR as significantly different
from the control group; and two of them validated the previous microarray findings.
Conclusions
ARM was a powerful data analysis technique to identify multivariate patterns involving clini-
cal and molecular profiles of patients affected by specific pathological panels. ARM proved
to be an effective mining approach to analyze gene expression with the advantage of includ-
ing patient’s CFs. A combination of CFs and DEGs might be employed in modeling the
patient’s chance to develop complex diseases, such as those studied here.
ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=
GSE156993.
Funding: RV and FJVZ are supported by São Paulo
Research Foundation (FAPESP - http://www.
fapesp.br/) Grant 2017/21174-8, Coordination of
Superior Level Staff Improvement (CAPES - https://
www.capes.gov.br/) and Brazilian National Council
for Scientific and Technological Development
(CNPq - http://www.cnpq.br/) Grant 307228/2018-
5. RMSC is supported by FAPESP Grants 2007/
08362-8, 2009/16233-9, 2010/10882-2, 2014/
16148-0 and 2016/25418-6, CAPES and CNPq
Grant 304570/2017-6. The funders had no role in
study design, data collection and analysis, decision
to publish, or preparation of the manuscript.
Competing interests: The authors have declared
that no competing interests exist.
Introduction
As a metabolic disorder, diabetes mellitus (DM) is caused either by a deficiency of insulin’s
mechanism of action, by an insulin secretion deficit, or by both [1]. As recently reported by
Jeong et al. [2], the prevalence of DM has increased exponentially in recent decades, being
expected to affect 693 million patients within 25 years. Of all adults newly diagnosed with DM,
more than 90% are affected by type 2 diabetes mellitus (T2DM) [3]. According to Jeong et al.
[2], in 2017 the estimated total global healthcare expenditure considering DM was USD 850
billion, with a relevant proportion of these costs arising from the treatment of various compli-
cations associated with the progression of DM. Over a period of years most T2DM patients
progress to three major groups of complications: microvascular, macrovascular, and miscella-
neous [4]. Regarding miscellaneous T2DM complications, Jeong et al. [2] recently reported
that dyslipidemia had the highest relative incidence risk of comorbidities that evolved after a
diagnosis of T2DM in Koreans. In 2010, the third cause of premature deaths (before the age of
70 years) in Brazilian subjects was regarded as diabetes, with high fasting plasma glucose and
high body mass index (BMI) being some of the major risk factors related to diabetes mortality
(53,353 individuals, or 12%) [5].
Dyslipidemia (DLP) is a metabolic dysfunction that results from an increased level of lipo-
proteins in the blood [6, 7]. Some studies have revealed that DLP could be one factor associ-
ated with DM-induced immune cell alterations [7–9]. It is believed that pro-inflammatory
cytokines produce an insulin resistance syndrome similar to that observed in DM [7, 9]. Find-
ings concerning chronically elevated levels of inflammatory markers suggest that poor glyce-
mic control of T2DM patients could increase risk for cardiovascular disease and infectious
diseases, including periodontitis [8, 10].
Periodontitis (PD) is a common chronic inflammatory disease characterized by destruction
of the periodontium, which is the supporting structures of the teeth, such as gingiva, periodon-
tal ligament and alveolar bone [11]. PD is a microbially induced oral disease, in which the bac-
terial biofilm is formed on the surfaces of teeth providing a chronic microbial stimulus that
elicits a local inflammatory response in the gingival tissues [12]. PD is also considered an
inflammatory disorder influenced by factors such as genetics [13], immune system reactions,
smoking [14] and the occurrence of systemic diseases, including DM [15]. Periodontal infec-
tion and DM have a two-way relationship [16] and PD can be recognized as the sixth largest
PLOS ONE | https://doi.org/10.1371/journal.pone.0240269 October 2, 2020
2 / 22
PLOS ONE Association rule mining to detect clinical features and genes related to chronic inflammatory diseases
complication associated with DM [17]. In response to bacterial products after periodontium
infection, there are local and systemic elevations of pro-inflammatory cytokines [18], which
may induce alterations in the metabolism of lipids, contributing to DLP in these patients [7,
9]. Some studies indicate an association between elevation in blood lipoproteins and alter-
ations in the periodontal condition [6, 19–21].
Currently, the interplay of T2DM, DLP and PD has been increasingly affecting patients
worldwide. Those are chronic inflammatory diseases, including systemic T2DM and DLP,
while PD is localized at the periodontium of the patient. Growing evidence indicates a biologi-
cal connection among T2DM-DLP-PD, demonstrated by the finding that these patients pres-
ent a hyperinflammatory state promoted by systemically increased levels of pro-inflammatory
molecules, as reviewed by Soory et al. [22]. Moreover, all of them are considered chronic and
complex diseases, since they are caused by a combination of genetic, environmental and life-
style factors [23]. Therefore, more studies focused on detecting unknown relationships in data-
sets of diseased patients will contribute to a better understanding of the interplay of T2DM,
DLP and PD.
Association rule mining (ARM) has been widely used to discover hidden relationships
established by multiple attributes that characterize a complex process under investigation. It
has several applications in the medical domain (for instance, see [24–26]) promoting highly
interpretable explanations without requiring data mining expertise [27]. In addition to
interpretability, another reason that makes ARM a widely used data mining technique is that
the obtained rules are capable of summarizing the joint impact of several factors [27, 28].
Thus, ARM is a powerful technique to assess the supposed interplay of T2DM, DLP and PD.
The ARM was previously used to assess the T2DM survival risk [29], and to determine the
T2DM comorbidities in large amounts of clinical data [30]. Ramezankhani et al. [31] showed
that ARM is a useful approach to determine the most frequent subsets of attributes in people
who will develop diabetes. However, this is the first study using ARM to simultaneously iden-
tify the potential clinical patterns and genetic markers of this group of diseases, thus revealing
clinical features and differentially expressed genes capable of properly characterizing these
chronic inflammatory diseases.
The outline of this paper is as follows. Section Materials and Methods presents the literature
review and our proposed methodology. Section Results and Discussion presents the experi-
mental results and an analytical explanation of their implications, followed by concluding
remarks in Section Conclusion.
Materials and methods
Datasets
Studied population. This research was approved by the Ethics in Human Research Com-
mittee of School of Dentistry at Araraquara (UNESP; Protocol number 50/06). Patients who
voluntarily sought dental treatment at the School of Dentistry at Araraquara (UNESP), Brazil,
were informed about the aims and methods of the study, providing their written consent to
participate; therefore, the whole study was conducted according to the ethical principles of the
Declaration of Helsinki.
The patients were characterized by the following criteria: age from 35 to 60 years, presence
of at least 15 natural teeth and similar socioeconomic level. Pre-selected patients, according to
their medical history, had their glycemic and lipid profiles investigated by biochemical blood
analysis, and were submitted to full periodontal examination. Then, 143 patients were divided
into five groups based upon diabetic, dyslipidemic and periodontal conditions:
PLOS ONE | https://doi.org/10.1371/journal.pone.0240269 October 2, 2020
3 / 22
PLOS ONE Association rule mining to detect clinical features and genes related to chronic inflammatory diseases
1. Group 1: poorly controlled T2DM with DLP and PD. Number of subjects = 28.
2. Group 2: well-controlled T2DM with DLP and PD. Number of subjects = 29.
3. Group 3: DLP and PD. Number of subjects = 29.
4. Group 4: systemically healthy individuals with PD. Number of subjects = 29.
5. Group 5: systemically and periodontally healthy individuals (control group). Number of
subjects = 28.
No patient in those five groups presented: history of antibiotic therapy in the previous 3
months and/or nonsteroidal anti-inflammatory drug therapy in the previous 6 months, preg-
nancy or use of contraceptives or any other hormone, current or former smoking addiction,
history of anemia, periodontal treatment or surgery in the preceding 6 months, use of hypoli-
pidemic drugs such as statins or fibrates, and history of diseases that interfere with lipid metab-
olism, such as hypothyroidism and hypopituitarism.
Additionally, patients enrolled in this study were previously investigated regarding malo-
naldehyde (MDA) quantification and some inflammatory cytokine levels [32], micronuclei
frequency (DNA damage evaluation) [33] and lipid peroxidation [32]. In these previous stud-
ies, power analysis based on a pilot study determined that at least 20 patients in each group
would be sufficient to assess differences in those molecules with 90% power and 95% confi-
dence interval.
Biochemical, physical and periodontal evaluations. Clinical criteria to include each
patient in the studied group are presented in what follows. Subjects were submitted to physical
and anthropometric examination for evaluating obesity such as abdominal circumference
(cm), height (m), weight (kg), waist (cm), hip (cm) and body mass index [33].
After a 12-hour overnight fast, each subject was referred to a clinical analysis laboratory
that collected a blood sample for evaluating: glycated haemoglobin (HbA1c) by enzymatic
immunoturbidimetry, fasting plasma glucose (mg/dL) by the modified Bondar & Mead
method, high-sensitivity C-reactive protein by the nephelometric method and insulin levels
by the chemiluminescence method (U/L). The homeostasis model assessment (HOMA) was
evaluated to calculate insulin resistance (IR). The diagnosis of T2DM was made by an endocri-
nologist who monitored the glycemic levels of each patient by evaluation of HbA1c; being
patients considered poorly controlled (HbA1c �8.0%) or well-controlled (HbA1c �7.0%).
Normoglycemic (nondiabetic) individuals presented fasting glucose levels <100 mg/dL and
HbA1c <5.7% [34–36].
The lipid profile [triglycerides (TG), total cholesterol (TC), and high density lipoprotein
(HDL)] was performed by enzymatic methods. Low density lipoprotein (LDL) was determined
by the Friedewald formula. Individuals with transitory DLP were not included here by consid-
ering the highest cutoff values: TC �240 mg/dL, LDL �160 mg/dL, HDL <40 mg/dL, and
TGs �200 mg/dL, according to the 2018 AHA / ACC / AACVPR / AAPA / ABC / ACPM /
ADA / AGS / APhA / ASPC / NLA / PCNA Guideline on the Management of Blood Choles-
terol [37]. It was also considered in this analysis the non-HDL-cholesterol (N-HDL-C), given
by N-HDL-C = TC—HDL, being the abnormal cutoff value �130 mg/dL, which is considered
to be a good predictor of cardiovascular disease (CVD) risk [38].
Diagnosis of periodontitis in at least 4 non-adjacent teeth, including local signs of inflam-
mation, loss of the connective tissue attachment of gingiva to teeth (clinical attachment loss,
CAL �4mm), and tissue destruction (presence of deep periodontal pockets �6mm) was
adopted according to the American Academy of Periodontology [39]. Each subject underwent
a periodontal clinical examination performed at 6 sites per tooth. The presence of deep
PLOS ONE | https://doi.org/10.1371/journal.pone.0240269 October 2, 2020
4 / 22
PLOS ONE Association rule mining to detect clinical features and genes related to chronic inflammatory diseases
periodontal pockets �6mm with CAL �5mm and bleeding on probing in at least 8 sites dis-
tributed in different quadrants of the dentition were the criteria of severe periodontitis [40].
Regarding the mutagenesis analysis, the description of the peripheral blood sampling, cell
culture and cytokinesis-block micronucleus (CBMN) assay can be found in Corbi et al. [33].
Table 1 summarizes the clinical features collected from the 143 investigated subjects. The
clinical feature dataset is available in S1 File.
Isolation of peripheral blood mononuclear cells, RNA extraction and microarray analy-
sis. Patients with greater glycemic, lipid and periodontal homogeneity parameters had their
transcriptome investigated (30 subjects in total) from peripheral blood mononuclear cells
(PBMCs), divided into: Group 1 (number of subjects = 5), Group 2 (number of subjects = 7),
Group 3 (number of subjects = 6), Group 4 (number of subjects = 6) and Group 5 (number of
subjects = 6). PBMCs were isolated, and total RNA was extracted using TRizol (Invitrogen,
Rockville, MD, USA) and purified by an RNeasy Protection Mini Kit (Qiagen, Hilden, Ger-
many) according to the manufacturer’s instructions. RNA was quantified by a NanoVue Spec-
trophotometer (GE Healthcare Life Sciences, Oslo, Norway), and its integrity was assessed by
agarose gel electrophoresis (1%). Only RNA samples in the λ(260/280) and λ(260/230) reasons
between 1.8 and 2.2 were used for microarray and quantitative real-time PCR analyses. Micro-
array data were generated from 500 nanograms of RNA as the initial input of each sample in
the GeneChip IVT Labeling Kit and hybridized to the U133 Plus 2.0 (Affymetrix Inc., Santa
Clara, CA, USA) arrays, which comprise 54,675 human transcripts. The U133 Plus 2.0 arrays
were scanned twice using the GeneChip Scanner 3000 7G (Affymetrix Inc., Santa Clara, CA,
USA). The Robust Multichip Average (RMA) strategy was used to preprocess raw .CEL files
[41, 42]. This strategy performs background correction through a normal-exponential convo-
lution model, quantile normalizes the probe intensities and summarizes them into probeset-
level quantities using an additive model fit through the median-polish strategy [43]. The gene
expression dataset is available in S2 File.
Association rule mining
Let An×m be a binary data matrix with the row index set X = {1, 2, . . ., n} and the column index
set Y = {1, 2, . . ., m}. Each row represents a transaction, and each column represents an item.
Each element aij 2 A holds the binary relationship between transaction i and item j. Let (X, Y)
denote the entire matrix A and (I, J) denote a submatrix of A with I � X and J � Y.
Definition 1 A subset J = {j1, . . ., js} � Y is called an itemset.
For a subset J � Y, we define J# = {x 2 X|axj = 1, 8j 2 J} as the set of transactions common to
all the items in J. The support of an itemset J is given by σ(J) = |J#|.
The problem of mining all frequent itemsets can be described as follows: determine all sub-
sets J � Y such that σ(J)�minSup, where minSup is a user-defined parameter.
To reduce the computational cost of the frequent itemset (pattern) mining problem, some
algorithms mine only the maximal frequent itemsets, i.e., those frequent itemsets from which
all supersets are infrequent and all subsets are frequent. The problem of this approach is that it
leads to loss of information since the supports of the subsets of the maximal frequent itemsets
are not available. An option to reduce the computational cost of the frequent pattern mining
problem without loss of information is to mine only the closed frequent itemsets. A frequent
itemset J is called closed if there exists no superset H aˆSˇƒ J with H# = J#. Remarkably, the set of
closed frequent itemsets uniquely determines the exact frequency of all frequent itemsets, and
it can be orders of magnitude smaller than the set of all frequent itemsets [44]. Therefore, this
approach drastically reduces the number of rules that have to be presented to the user, without
any information loss [45].
PLOS ONE | https://doi.org/10.1371/journal.pone.0240269 October 2, 2020
5 / 22
PLOS ONE Association rule mining to detect clinical features and genes related to chronic inflammatory diseases
Table 1. Description of the clinical features of the 143 subjects enrolled in this study (%ts stands for % of tooth sites).
Characteristic
Demographic
# Attribute
1
Sex
2 Age
Cardiovascular and obesity risk
3
Body Mass Index
4 Waist / Hip Ratio
5 Abdominal Circumference
Type 2 Diabetes Mellitus
6
Fasting Plasma Glucose
Alias
Sex
Age
BMI
WHR
AC
FPG
Unit Domain
1. Female
2. Male
1. �50
2. >50
yr.
m/kg2
1. Underweight: <18.5
2. Normal weight: [18.5, 25)
3. Overweight: [25, 30)
4. Obesity class I: [30, 35)
5. Obesity class II e III: �35
cm/cm (see Table 2)
cm
(see Table 3)
mg/dL
1. Normoglycemic: <100
2. Prediabetes or high-risk: [100, 126)
3. Established diabetes: �126
7
Insulin
INS
U/L
1. Normal: �25
2. Altered: >25
8 Glycated Haemoglobin
HbA1c
%
1. Normoglycemic: <5.7
9 HOMA-IR
HOMA-IR
2. Prediabetes or high-risk: [5.7, 6.5)
3. Decompensation (transitory): [6.5, 8)
4. Decompensation (defined): �8
1. Normal: �2.15
2. Altered: >2.15
Dyslipidemia
10 Total Cholesterol
TC
mg/dL
1.<150 (Optimal)
11 HDL cholesterol
HDL
mg/dL
1. <40 (Low)
2. [40, 60]
3. >60
12
LDL cholesterol
LDL
mg/dL
1. <100 (Optimal)
2. [150, 200)
3. [200, 240)
4. �240
2. [100, 130)
3. [130, 160)
4. [160, 190)
5. �190
13 Triglycerides
TG
mg/dL
1. <150 (Optimal)
2. [150, 200)
3. �200
14 Non-HDL-Cholesterol
N-HDL-C
mg/dL
1. <130 (Optimal)
2. [130, 160)
3. [160, 190)
4. [190, 220)
5. �220
(Continued )
PLOS ONE | https://doi.org/10.1371/journal.pone.0240269 October 2, 2020
6 / 22
PLOS ONE Association rule mining to detect clinical features and genes related to chronic inflammatory diseases
Table 1. (Continued)
Characteristic
Periodontal
# Attribute
15 Visible Plaque
Alias
VP
Unit Domain
%ts
1. Low: <30
2. Medium: [30, 50]
3. High: >50
16 Gingival Index bleeding
GI
%ts
1. Low: <30
2. Medium: [30, 50]
3. High: >50
17 Bleeding on probing
BOP
%ts
1. Low: <30
18 Total Number of Teeth
TNT
2. Medium: [30, 50]
3. High: >50
1. low number teeth: �20
2. high number teeth: >20
19
Interproximal periodontal pocket depth (PPDi) �3mm PPDi3mm
%ts
1. Low: <30
2. Medium: [30, 50]
3. High: >50
20 PPDi = 4—5mm
PPDi4-5mm
%ts
1. Low: <30
2. Medium: [30, 50]
3. High: >50
21 PPDi � 6mm
PPDi6mm
%ts
1. Low: <30
2. Medium and High: �30
22
Interproximal clinical attachment loss (CALi) �2mm
CALi2mm
%ts
1. Low: <30
2. Medium: [30, 50]
3. High: >50
23 CALi = 3-4mm
CALi3-4mm
%ts
1. Low: <30
2. Medium: [30, 50]
3. High: >50
24 CALi � 5mm
CALi5mm
%ts
1. Low: <30
2. Medium: [30, 50]
3. High: >50
25
Suppuration
SUPP
%ts
1. Abscence: <1
Mutagenesis
26 Nuclear Division Index
NDI
2. Moderate: [1, 16)
3. Severe: �16
1. Low: <1.87
2. Moderate: [1.87, 2.08)
3. High: �2.08
27
Frequency of Binucleated cells with Micronuclei
MNCF
%
1. Low: <3.05
2. Moderate: [3.05, 7.2)
3. High: �7.2
28 Micronucleus Frequency
MNF
%
1. Low: <3.5
29
Frequency of Nucleoplasmic Bridges
FNB
%
1. Low: <1.21
2. Moderate: [3.5, 6.1)
3. High: �6.1
2. Moderate: [1.21, 2.7)
3. High: �2.7
https://doi.org/10.1371/journal.pone.0240269.t001
PLOS ONE | https://doi.org/10.1371/journal.pone.0240269 October 2, 2020
7 / 22
PLOS ONE Association rule mining to detect clinical features and genes related to chronic inflammatory diseases
Definition 2 An association rule (AR) is an expression of the form J ) H, where J and H are
itemsets, H \ J = ;. J is called antecedent (or head) and H is called consequent (or tail) of the
rule.
The support of an association rule J ) H is the number of transactions that contain the
itemset J[H: σ(J ) H) = σ(J[H). The confidence of an association rule J ) H measures its pre-
dictive accuracy and is given by conf(J ) H) = σ(J ) H)/σ(J). A rule is considered a strong rule
if conf(J ) H)�minConf, where minConf is a user-defined parameter. The completeness (or
recall) is given by comp(J ) H) = σ(J ) H)/σ(H). Remark that confidence and completeness
are not symmetric measures because by definition they are conditional on the antecedent and
consequent, respectively. The metric lift measures the degree of surprise of a rule and is given
by lift(J ) H) = σ(J ) H)/(σ(J) × σ(H)).
A user can be interested in a more specific set of association rules, where the consequents of
the rules describe a target attribute. These rules are known as class association rules (CARs).
Definition 3 A class association rule (CAR) is an expression of the form J ) c, where J is an
itemset and c is a class label (a target item).
In this work, each item is given by an attribute-value pair. Thus, for instance, FPG = 3 is an
item; {AC = 3, FPG = 3, HbA1c = 4} is an itemset; and {AC = 3, FPG = 3, HbA1c = 4} )
{GI = 3, BOP = 3} is an association rule.
Given that the result to be presented to the user is more parsimonious, we will focus on
closed frequent itemsets here. The patterns will be mined using the RIn-Close_CVCP algo-
rithm [46, 47], which is a fast algorithm and avoids the necessity of the itemization step [47].
Its implementation is available at https://github.com/rveroneze/rinclose.
Association rule mining from the clinical features alone. T2DM, DLP, and PD have
their own specific characteristics (features or attributes) generally taken as decision variables
to perform a diagnosis. However, given the increasing incidence of patients affected by differ-
ent interplays of T2DM-DLP-PD, we originally used ARM to assess whether there are joint
attributes present in patients with these comorbidities that might indicate the biological inter-
relationship among them.
Fig 1 shows a flowchart that summarizes the process of association rule mining from the
dataset containing solely clinical features. From the clinical features collected from the investi-
gated patients (presented in Table 1), we selected the most clinically relevant to diagnose
T2DM, DLP and PD diseases isolated. We did not use the mutagenesis attributes because they
are not applied in a clinical routine for disease diagnosis. The following 17 clinical features
were selected for this analysis: BMI, WHR, AC, FPG, HbA1c, HOMA-IR, TC, HDL, LDL, TG,
N-HDL-C, GI, BOP, PPDi6mm, CALi34mm, CALi5mm and SUPP. Thus, the dataset to be
analyzed has 143 subjects and 17 attributes. BMI, WHR and AC attributes represent character-
istics that confer cardiovascular and obesity risk, according to the World Health Organization
[19, 48]. The N-HDL-C attribute is considered a good predictor of CVD risk [38]. The glyce-
mic parameters: FPG, HbA1c and HOMA-IR (Homeostasis Model Assessment to calculate
the insulin resistance) are considered essential for the diagnosis of T2DM and its metabolic
control [35, 36]. TC, HDL, LDL and TG are important lipid parameters to diagnose DLP [37].
Regarding periodontitis, the American Academy of Periodontology (AAP) utilizes the clinical
periodontal parameters: GI, BOP, PPDi6mm, CALi3-4mm, CALi5mm and SUPP [39, 40].
The parameters used in ARM were: minSup = 14 and minConf = 70%. A rule was consid-
ered interesting whenever at least one of the following attributes is present: PPDi6mm = 2; GI,
BOP, CALi34mm, CALi5mm, SUPP 2{2, 3}. We followed those clinical periodontal parame-
ters, as recommended by the AAP, because they indicate periodontal disease activity. Those
selected attributes are considered relevant to identify individuals undoubtedly affected by
moderate or severe periodontitis, allowing us to check if there is an evident association
PLOS ONE | https://doi.org/10.1371/journal.pone.0240269 October 2, 2020
8 / 22
PLOS ONE Association rule mining to detect clinical features and genes related to chronic inflammatory diseases
Fig 1. Flowchart that summarizes the process of association rule mining from the dataset containing solely clinical features.
https://doi.org/10.1371/journal.pone.0240269.g001
between both systemic diseases (T2DM and DLP) and PD. In this way, we corroborate the
existence of a T2DM-DLP-PD biological interrelationship.
In addition, we performed an analysis focusing on the cardiovascular and obesity risk attri-
butes to determine whether they are associated with periodontal disease. Therefore, we per-
formed an analysis with only the cardiovascular and obesity risk attributes in the antecedent
part of the rule (BMI, WHR, AC, FPG, N-HDL-C), and the same attributes in the consequent
part of the rule. We also performed an analysis comprising only T2DM patients presenting
diabetic dyslipidemia, which are the 10 patients from Groups 1 and 2 having TG �204 mg/dL
and HDL <38 mg/dL [49, 50].
The results of these analysis will be presented and discussed in Section Results and
Discussion.
Association rule mining from the clinical features and gene expression datasets in con-
junction. The transcriptome of the patients studied here obtained from PBMCs by microar-
ray was analyzed utilizing bioinformatics and statistical tools, as described in topic Isolation of
peripheral blood mononuclear cells, RNA extraction and microarray analysis. Those analyses,
developed as regularly, produced a list of differentially expressed genes (DEGs). However, in
that kind of analysis the gene expression profile obtained by the probesets did not consider the
patient’s clinical features (CFs). In conventional bioinformatics and statistical tools, adequate
clinical diagnosis of each group of patients is used to determine whether a DEG is related to a
specific pathological condition. Here, we used ARM to identify the joint interplay of CFs and
DEGs, having the advantage of taking together CFs and genetic markers to identify each com-
bination of T2DM-DLP-PD complex diseases. This approach might contribute to better iden-
tifying new targets for the diagnosis of each combination of those complex diseases, as well as
for modeling the patient’s chance to develop them.
Fig 2 shows a flowchart that summarizes the process of class association rule mining from
the dataset containing both CFs and DEGs. First, we performed the preprocessing of the
PLOS ONE | https://doi.org/10.1371/journal.pone.0240269 October 2, 2020
9 / 22
PLOS ONE Association rule mining to detect clinical features and genes related to chronic inflammatory diseases
Fig 2. Flowchart that summarizes the process of class association rule mining from the dataset containing both clinical features (CFs) and
differentially expressed genes (DEGs).
https://doi.org/10.1371/journal.pone.0240269.g002
original gene expression dataset (GED), which has the gene expression profile of 54,675 genes
obtained from the transcriptome of the 30 subjects, in the following three steps:
1. Gene selection: we filtered out genes with small profile variance, in specific we filtered
out gene expression profiles with variation less than 0.1 when considering the difference
between its maximum and minimum values. It was done because gene profiling experi-
ments typically include genes that exhibit little variation in their profile and these genes are
usually uninteresting. Thus, these genes are commonly removed from the analysis. With
this filter, 50.441 genes were removed, leaving 4.234 genes for the subsequent analysis.
2. Normalization: we used zero-mean normalization to adjust the values measured on different
scales to a common scale. Let g be the gene expression profile of a gene g for the 30 subjects
of our study. The normalized gene expression profile ^g is given by ^g ¼ ðg (cid:0) avgðgÞÞ=stdðgÞ,
where avg(g) and std(g) are, respectively, the sample average and the sample standard devia-
tion of g.
3. Discretization: if a normalized gene expression value was above 1.0, it was considered
over-expressed (and it is represented by the value 1 in our results); if a normalized gene
expression value was below -1.0, it was considered under-expressed (and it is represented
by the value -1 in our results); otherwise the gene expression value was considered uninter-
esting and was ignored.
PLOS ONE | https://doi.org/10.1371/journal.pone.0240269 October 2, 2020
10 / 22
PLOS ONE Association rule mining to detect clinical features and genes related to chronic inflammatory diseases
We performed the mining of CARs in this preprocessed GED with the following parame-
ters: minSup = 3 and minConf = 90%. The group of each individual (Groups 1 to 5) is the tar-
get attribute. The result, containing 118 CARs, was used for a new phase of gene selection as
described in what follows. The 118 CARs have a coverage of 1081 genes (this means that
1081 genes are presented in these rules). Of these 1081 genes, 17 genes are present in con-
flicting rules, exhibiting the same value for the control group (Group 5) and for the other
groups (Groups 1 to 4). Therefore, these 17 genes were discarded. Thus, 1081 − 17 = 1064
genes were selected for the new phase of analysis, together with the 29 CFs listed in Table 1.
In this new phase of analysis, we performed the mining of CARs with the same parameters,
i.e., minSup = 3 and minConf = 90%. The results will be presented and discussed in Section
Results and Discussion.
Reverse transcription-quantitative polymerase chain reaction (RT-qPCR)
Real-Time Analysis
To biologically validate the genes selected from the CARs considering the CFs+DEGs, we con-
ducted RT-qPCR analyses in all 143 patients (including the 30 patients who were analyzed
by microarray) distributed into the 5 groups, according to the subitem Studied population.
Reverse transcription reactions were performed utilizing the High Capacity Kit (Thermo
Fisher Scientific). Complementary DNA (cDNA) was used to perform qPCR reactions for the
selected DEGs, which are represented as probe sets in Table 7. To investigate the expression of
the probe (or gene) identified by the rule selected for each group of patients, the TaqMan1
gene expression assay specific for each of these “target” genes was utilized. Each target gene is
normalized by a gene considered an endogenous control of the qPCR reactions, in this case,
we utilized the GAPDH -Glyceraldehyde-3-Phosphate Dehydrogenase gene (Hs02758991_g1),
due to its housekeeping expression pattern.
All reactions were performed in duplicate utilizing the 7500 Real-Time PCR-System
(Thermo Fisher Scientific, Foster City, CA, USA). To calculate gene expression, Expression
Suite Software was used (Thermo Fisher Scientific, Foster City, CA, USA), which employs the
comparative Cycle Threshold (ΔCt) method for multivariate data analysis. Statistical analysis
to find differences in the gene expression by the values of 2−ΔCt between the groups was per-
formed by the Mann-Whitney test, utilizing GraphPad Prism software, version 5.0, and con-
sidering a significance level of 0.05 [51].
Results and discussion
Association rules for the dataset of clinical features (CFs)
It was obtained 78 rules comprising the CF dataset, which are presented in S1 Table. The peri-
odontists and geneticist experts analyzed those rules to select examples of rules of high clinical
relevance to demonstrate the T2DM-DLP-PD interrelationship. To select the rules, the follow-
ing requirements were established in decreasing order of relevance:
1. In the antecedent part of the rule, the joint presence of attributes with altered values in
these characteristics of Tables 1, 2 and 3: cardiovascular and obesity risk; T2DM; and DLP;
2. The highest confidence value.
The rules of Table 4 present, in general, WHR = 4 and AC = 3, which represent very high
cardiovascular and obesity risk for all ages of both male and female (see Tables 2 and 3);
FPG = 3, HbA1c = 4 and HOMA-IR = 2 represent the worst glycemic parameters, evidencing
that those patients have established T2DM with defined metabolic decompensation and
PLOS ONE | https://doi.org/10.1371/journal.pone.0240269 October 2, 2020
11 / 22
PLOS ONE Association rule mining to detect clinical features and genes related to chronic inflammatory diseases
Table 2. Waist / hip ratio domain.
1. Low
2. Moderate
3. High
4. Very High
1. Low
2. Moderate
3. High
4. Very High
Age � 39
<0.72
[0.72, 0.79)
[0.79, 0.84]
> 0.84
Age � 39
<0.84
[0.84, 0.92)
[0.92, 0.96]
> 0.96
https://doi.org/10.1371/journal.pone.0240269.t002
Female
39 < Age � 49
<0.73
[0.73, 0.80)
[0.80, 0.87]
> 0.87
Male
39 < Age � 49
<0.88
[0.88, 0.96)
[0.96, 1]
> 1
Age > 49
<0.74
[0.74, 0.82)
[0.82, 0.88]
> 0.88
Age > 49
<0.90
[0.90, 0.97)
[0.97, 1.02]
> 1.02
Table 3. Table caption Nulla mi mi, venenatis sed ipsum varius, volutpat euismod diam.
1. Low risk
2. High risk
3. Very high risk
Female
< 80
[80, 88)
� 88
Male
< 94
[94, 102)
� 102
https://doi.org/10.1371/journal.pone.0240269.t003
insulin resistance; the patients are also dyslipidemic as demonstrated by the highest levels of
total cholesterol (TC = 4) and triglycerides (TG = 3). The consequent part of those rules is
BOP = 3, which means that more than 50% of tooth sites bleed during the periodontal exam,
demonstrating wide and active inflammation of the periodontal tissues including the gingiva.
There are 4 rules showing as consequent SUPP = 2, meaning that those patients have a moder-
ate suppuration, since it affects 1% to 16% of tooth sites, indicating the presence of an estab-
lished periodontitis. The seventh and eighth rules of Table 4 show TC = 4 and N-HDL-C = 5,
meaning that individuals with the highest levels of TC and N-HDL-C have 78% of confidence
of presenting BOP = 3 or SUPP = 2, demonstrating wide and active inflammation of the peri-
odontal tissues and an established periodontitis.
Table 4. Association rules for the clinical feature dataset.
Rule
WHR = 4, FPG = 3, HbA1c = 4, TG = 3 ) BOP = 3
AC = 3, FPG = 3, HbA1c = 4, HOMA-IR = 2, TG = 3 ) BOP = 3
AC = 3, FPG = 3, HOMA-IR = 2, TG = 3 ) BOP = 3
WHR = 4, AC = 3, FPG = 3, HOMA-IR = 2, TG = 3 ) BOP = 3
WHR = 4, FPG = 3, TG = 3 ) BOP = 3
WHR = 4, FPG = 3, HOMA-IR = 2, TG = 3 ) BOP = 3
AC = 3, HOMA-IR = 2, TC = 4, TG = 3 ) SUPP = 2
TC = 4, N-HDL-C = 5 ) BOP = 3
TC = 4, N-HDL-C = 5 ) SUPP = 2
AC = 3, HOMA-IR = 2, TC = 4 ) SUPP = 2
WHR = 4, AC = 3, HOMA-IR = 2, TC = 4 ) SUPP = 2
https://doi.org/10.1371/journal.pone.0240269.t004
σrule
14
15
22
19
21
20
15
18
18
23
18
σhead
14
15
26
23
26
25
19
23
23
31
25
σtail
74
74
74
74
74
74
67
74
67
67
67
%Conf.
100.00
100.00
84.62
82.61
80.77
80.00
78.95
78.26
78.26
74.19
72.00
Lift
1.93
1.93
1.64
1.60
1.56
1.55
1.68
1.51
1.67
1.58
1.54
PLOS ONE | https://doi.org/10.1371/journal.pone.0240269 October 2, 2020
12 / 22
PLOS ONE Association rule mining to detect clinical features and genes related to chronic inflammatory diseases
Table 5. Association rules for the clinical feature dataset—Cardiovascular risk.
Rule
WHR = 4, AC = 3, FPG = 3 ) BOP = 3
FPG = 3 ) BOP = 3
AC = 3, FPG = 3 ) BOP = 3
WHR = 4, FPG = 3 ) BOP = 3
N-HDL-C = 5 ) BOP = 3
N-HDL-C = 5 ) SUPP = 2
BMI = 3, WHR = 4, AC = 3 ) SUPP = 2
BMI = 3, WHR = 4 ) SUPP = 2
https://doi.org/10.1371/journal.pone.0240269.t005
σrule
24
35
28
26
18
18
18
20
σhead
28
41
33
31
23
23
24
28
σtail
74
74
74
74
74
67
67
67
%Conf.
85.71
85.37
84.85
83.87
78.26
78.26
75.00
71.43
Lift
1.66
1.65
1.64
1.62
1.51
1.67
1.60
1.52
There was interest in verifying the association of cardiovascular and obesity parameters
with the presence of periodontitis. In that analysis we also included the N-HDL-C attribute,
which predicts CVD risk even better than LDL [52]. The rules obtained by focusing on only
those 11 attributes are presented in Table 5. We highlighted the rules: BMI = 3, WHR = 4,
AC = 3 ) SUPP = 2 and N-HDL-C = 5 ) BOP = 3, as supporting the evidence of an associa-
tion between cardiovascular risk factors and periodontitis. The obtained rules support the
clear association between N-HDL-C and parameters of periodontitis. The N-HDL-C was the
best predictor among all cholesterol measures, both for coronary artery disease events and for
strokes [53]. More recently, this was confirmed, since the highest N-HDL-C concentrations in
blood (�220 mg/dL, which is equivalent to �5.7 mmol/L) were associated with the highest
long-term risk of atherosclerotic cardiovascular disease [54]. Here we observed exactly this
highest level of N-HDL-C in the rules of Table 5. Interestingly, there are good reasons for
the usefulness of N-HDL-C in monitoring patients, since unlike LDL, N-HDL-C does not
require the triglyceride concentration to be 4.5 mmol/L (400 mg/dL), and has an additional
advantage of not requiring patients to fast before blood sampling. Therefore, it is certainly a
better measure than calculated LDL for patients with increased plasma triglyceride concentra-
tions [38, 53].
In general, these rules demonstrate the interplay between cardiovascular and obesity risk,
T2DM, DLP and PD, which is in line with some studies as reviewed by Soory [22] and Khu-
maedi et al. [8]. These diseases manifest persistent elevation of systemic inflammatory media-
tors, characterizing chronic inflammation [8]. It is known to be one of the atherosclerosis non-
traditional risk factors and has a role in every phase of atherogenesis [8]. Atherogenic dyslipi-
demia is expressive among T2DM individuals, for example, in 10 − 15% of the European popu-
lation [49, 50]. Therefore, we performed an analysis comprising only our 10 T2DM patients
presenting diabetic dyslipidemia [49, 50]. The rules found for this pathologic condition are
presented in Table 6. We highlighted the rule: FPG = 3, HOMA-IR = 2, TC = 2, HDL = 1,
TG = 3 ) BOP = 3, as it demonstrated that diabetic dyslipidemia was associated with
more than 50% of tooth sites bleeding, one of the main significant signals of periodontium
Table 6. Association rules for the clinical feature dataset—Diabetic dyslipidemia.
Rule
AC = 3, FPG = 3, HOMA-IR = 2, HDL = 1, TG = 3 ) GI = 3
AC = 3, FPG = 3, HOMA-IR = 2, TC = 2, HDL = 1, TG = 3 ) GI = 3, BOP = 3
FPG = 3, HOMA-IR = 2, TC = 2, HDL = 1, TG = 3 ) BOP = 3
AC = 3, FPG = 3, HOMA-IR = 2, HDL = 1, TG = 3 ) GI = 3, PPDi6mm = 1
https://doi.org/10.1371/journal.pone.0240269.t006
σrule
6
5
6
5
σhead
6
5
6
6
σtail
6
5
6
5
%Conf.
100.00
100.00
100.00
83.33
Lift
1.67
2.00
1.67
1.67
PLOS ONE | https://doi.org/10.1371/journal.pone.0240269 October 2, 2020
13 / 22
PLOS ONE Association rule mining to detect clinical features and genes related to chronic inflammatory diseases
inflammation. Periodontitis is the most common cause of chronic inflammation in diabetic
patients. Both periodontitis and diabetes have detrimental effects on each other in terms of
alveolar bone destruction and poor metabolic control, by continuous inflammatory mediator
activation [8].
Association rules for the datasets of clinical features and differentially
expressed genes in conjunction
Remark that we used ARM to obtain rules with joint patterns of CFs and DEGs, having the
advantage of taking together the clinical characteristics and the genetic markers to identify
each T2DM-DLP-PD combination of complex diseases. Also different from the rules consider-
ing only CFs (Table 4), the CF+DEG-rules were obtained for identifying specifically a group of
patients. Therefore, both CFs and DEGs were considered in the antecedent part of the rules,
and the consequent part of the rules is given by the number representing the groups (Groups 1
to 5). It was obtained 161 CF+DEG-rules, which are presented in S2 Table.
Because of the importance of biologically validating the CF+DEG-rules, Periodontists and
Geneticist experts selected only one discriminant rule for each of the five groups, as presented
in Table 7. The Periodontists and Geneticist experts make the decision of the CF+DEG-rules’s
choice following these criteria in decreasing order of relevance:
1. The joint presence of attributes showing values as altered as possible (according to the
reference values presented in Tables 1, 2 and 3) referring to the cardiovascular and obe-
sity risk, T2DM, DLP, PD, and also, at lower relevance, mutagenesis and demographic
characteristics;
2. The presence of one probe representing an over-expressed gene, such as ‘229026_at = 1’;
3. The highest confidence value;
4. The highest completeness value.
All the selected rules in Table 7 have 100% of confidence, which means that all subjects who
give support to a rule are from the same group.
Specifically to Group 1 of patients (poorly controlled T2DM with DLP and PD), the selected
rule means that 80% of the patients of Group 1 have high abdominal circumference (AC = 3),
meaning high CHD risk; altered glycemic parameters (FPG = 3, HbA1c = 4, HOMA-IR = 2),
evidencing that those patients have established T2DM with defined metabolic decompensation
Table 7. Association rules for the clinical feature and gene expression datasets in conjunction.
Rule
AC = 3, FPG = 3, INS = 1, HbA1c = 4, HOMA-IR = 2, HDL = 2, TG = 3, VP = 3, BOP = 3,
PPDi6mm = 1, CALi2mm = 1, SUPP = 2, 223130_s_at = -1, 229026_at = 1 ) 1
HOMA-IR = 2, TC = 4, TG = 3, N-HDL-C = 5, 208485_x_at = 1, 212386_at = -1 ) 2
FPG = 1, HDL = 2, PPDi3mm = 3, PPDi6mm = 1, MNCF = 2, 223422_s_at = 1,
224902_at = 1 ) 3
BMI = 2, FPG = 1, INS = 1, HbA1c = 1, HOMA-IR = 1, HDL = 2, TG = 1, TNT = 2,
PPDi6mm = 1, CALi2mm = 1, CALi3-4mm = 3, N-HDL-C = 1, 1560999_a_at = 1,
228766_at = -1, 244413_at = 1 ) 4
Age = 1, FPG = 1, INS = 1, HbA1c = 1, TG = 1, VP = 1, GI = 1, BOP = 1, TNT = 2,
PPDi3mm = 3, PPDi4-5mm = 1, PPDi6mm = 1, CALi5mm = 1, SUPP = 1, NDI = 2,
MNCF = 1, MNF = 1, FNB = 1, 236395_at = 1 ) 5
https://doi.org/10.1371/journal.pone.0240269.t007
%Comp. %Conf. Lift
80.00
100.00
6.00
71.00
67.00
100.00
100.00
4.29
5.00
67.00
100.00
5.00
67.00
100.00
5.00
PLOS ONE | https://doi.org/10.1371/journal.pone.0240269 October 2, 2020
14 / 22
PLOS ONE Association rule mining to detect clinical features and genes related to chronic inflammatory diseases
and insulin resistance; high triglyceride level (TG = 3); established severe periodontitis as
denoted by VP = 3 (more than 50% of tooth sites showing poor oral hygiene), BOP = 3 (more
than 50% of tooth sites bleeding), PPDi6mm = 1 (up to 30% of tooth sites with deep periodon-
tal pockets), and SUPP = 2 (suppuration at maximum of 16% of tooth sites). Though the fol-
lowing attributes did not contribute to the identification of Group 1, they also did not disturb
it: INS = 1, HDL = 2 and CALi2mm = 1.
The rule selected for Group 2 (well-controlled T2DM with DLP and PD) means that 71%
of the patients of Group 2 have insulin resistance demonstrated by HOMA-IR = 2; and the
highest levels of total cholesterol (TC = 4), triglycerides (TG = 3) and non-HDL-cholesterol
(N-HDL-C = 5). Surprisingly, considering the first criterion for selecting these 5 rules, for
identifying Group 2 of patients, a few rules were obtained. Because of this, in the selected rule
there were no attributes regarding the cardiovascular and obesity risk and PD. Moreover, it
should be taken into account that the rules obtained for Group 2 of patients should reflect
the clinical criteria defined to select the patients. For example, in comparison with Group 1,
Group 2 of patients differs only by the better metabolic control of T2DM.
The rule selected for Group 3 (DLP and PD) means that 67% of the patients have normal
fasting plasma glucose (FPG = 1) which is expected since they are not affected by T2DM; they
present altered HDL levels (HDL = 2), and they are affected by PD, since up to 30% of tooth
sites present very deep periodontal pockets (PPDi6mm = 1). Moreover, in this rule the moder-
ate frequency of binucleated cells with micronuclei (MNCF = 2) means that the circulating
blood of the patients is affected by a moderate level of mutagenesis, probably as a consequence
of the altered lipid metabolism of the patients. Indeed, a previous study of our research group
enrolling the same patients showed significantly higher mRNA levels of leptin in dyslipidemic
individuals (Groups 1, 2 and 3). Moreover, those leptin mRNA levels were significantly corre-
lated with periodontal parameters such as BOP, suppuration and mainly CALi � 5 mm [55].
Regarding Group 4 (systemically healthy individuals with PD), the selected rule means that
67% of the patients of this group are not obese, diabetic or dyslipidemic, as expected by the
underlined clinical criteria for selecting them. Those patients are only affected by generalized
periodontitis with pronounced alveolar bone loss, since they present more than 50% of tooth
sites with 3 to 4 mm of clinical attachment loss (CALi34mm = 3), and up to 30% of tooth sites
with very deep periodontal pockets (PPDi6mm = 1).
The rule selected for Group 5 (systemically and periodontally healthy individuals, or control
group) means that 67% of the patients of this group are not characterized by obesity, T2DM or
DLP, as expected by the underlined clinical criteria for selecting them. In addition, they did
not present active PD because it was not present in the rule any domain of bleeding or inflam-
mation, and the presence of the shallow periodontal pockets (PPDi3mm = 3) in at least 50% of
tooth sites is not an indicator of periodontal disease. Conversely, the occurrence of up to 30%
of tooth sites with PPDi45mm, PPDi6mm = 1, and clinical attachment loss (CALi5mm = 1)
suggests that those patients were previously affected by localized PD. Moreover, although the
rule includes the mutagenic parameters, their values are not altered.
To proceed to the biological validation of DEGs, we chose to validate by RT-qPCR (see Sub-
section Reverse transcription-quantitative polymerase chain reaction (RT-qPCR) Real-Time
Analysis) one highly expressed gene in each of the five rules. Certainly, more rules with more
probes/DEGs could be selected for validation, but we had limitations in the volume of the bio-
logical sample of the patients (RNA obtained from PBMCs).
For Group 1, we selected the probe 229026_at = 1, whose gene is CDC42SE2 (Cell Division
Cycle 42 Small Effector 2), detected by the TaqMan assay Hs00184113_m1. Although there is
another gene in the rule of Group 1 (23130_s_at), this gene was down-regulated, and therefore
did not meet the criteria of choice. The CDC42SE2 gene has diverse biological functions, such
PLOS ONE | https://doi.org/10.1371/journal.pone.0240269 October 2, 2020
15 / 22
PLOS ONE Association rule mining to detect clinical features and genes related to chronic inflammatory diseases
Fig 3. Validation results by RT-qPCR of the genes considering the different Group (G) comparisons. All mRNA
levels of the investigated genes were normalized to the GAPDH endogenous control gene. (A) CDC42SE2 gene
expression, �p � 0.0001; (B) CFLAR gene expression, no statistical difference among the groups; (C) PDPR gene
expression, �p � 0.0002; (D) Validation of the CLECL1 gene expression, �p � 0.0064; (E) Validation of the MEF2C
gene expression, �p � 0.0425. Data represent the mean ± SEM 2−ΔCt of all patients in that group (Mann–Whitney U
test; α = 5%).
https://doi.org/10.1371/journal.pone.0240269.g003
as the organization of the actin cytoskeleton by acting downstream of CDC42SE2, inducing
actin filament assembly, and it may play a role in early contractile events in phagocytosis in
macrophages. Accordingly, the CDC42SE2 gene alters CDC42-induced cell shape changes. In
activated T-cells, the CDC42SE2 gene may play a role in CDC42-mediated F-actin accumula-
tion at the immunological synapse [56]. The CDC42 (Cell Division Cycle 42) gene encodes a
small GTPase protein belonging to the Rho-subfamily, which regulates signaling pathways that
control diverse cellular functions including cell morphology, migration, endocytosis and cell
cycle progression [56].
In Fig 3(A), it can be observed that the CDC42SE2 gene was down-regulated in the decom-
pensated T2DM, dyslipidemic and PD patients (Group 1) (p-value � 0.0001) in comparison
to the healthy patients (Group 5). Actually, this finding obtained by qPCR is contrary to the
expected by the rule based on the microarray data (denoted by the positive 1 value of the
‘229026_at’). Therefore, the qPCR method showed discordant gene expression levels from
those detected by the microarray. Actually, it is not uncommon to find discrepant results of
gene expression between qPCR and microarray, either because the gene expression between
the diseased and control groups did not reach statistical difference or because conflicting
results were found between the qPCR and microarray methods [51]. The discordant
CDC42SE2 gene expression between qPCR and microarray (not validation) means more a lim-
itation of the method for identification of gene expression levels than a limitation of CAR min-
ing. In addition, considering that Group 2 of patients only differs from Group 1 in patients’
metabolic control, we also investigated the CDC42SE2 gene expression in the well-controlled
T2DM-DLP-PD (Group 2) patients, and we observed significantly lower levels in Group 1 but
no significant difference in Group 2 in comparison to the control Group 5. Therefore, when
we performed the CDC42SE2 gene expression comparison involving Groups 1, 2 and 5, we
observed the lowest expression in the worst metabolic condition of patients (Group 1), while
the patients with adequate metabolic control (Group 2) had similar CDC42SE2 expression
when compared with the healthy patients of Group 5.
For Group 2, the selected probe is 208485_x_at = 1, which is the CFLAR (CASP8 and FADD
Like Apoptosis Regulator) gene, detected by the TaqMan assay Hs01117851_m1. The protein
encoded by the CFLAR gene is a regulator of apoptosis which may function as a crucial link
PLOS ONE | https://doi.org/10.1371/journal.pone.0240269 October 2, 2020
16 / 22
PLOS ONE Association rule mining to detect clinical features and genes related to chronic inflammatory diseases
between cell survival and cell death pathways. Additionally, this protein acts as an inhibitor of
TNF receptor superfamily member 6 (TNFRSF6) mediated apoptosis [56]. Considering the
rule, an over-expression of the CFLAR gene was expected in Group 2 compared to Group 5.
However, there was a similarly high expression of the CFLAR gene in both Groups 2 and 5 (see
Fig 3(B)). We also performed the analysis of the CFLAR gene expression for Groups 1, 2 and 5,
observing no significant difference among them, although a lower gene expression can be
found in the patients with the worst metabolic condition (Group 1).
For Group 3, the rule has 2 highly expressed genes/probes, and we selected the 224902_at
probe for further analysis, which is the PDPR (Pyruvate Dehydrogenase Phosphatase Regulatory
Subunit) gene, detected by the TaqMan assay Hs01663324_m1, because it takes part in a more
interesting metabolic pathway. This gene acts on the pyruvate dehydrogenase complex by cata-
lyzing the oxidative decarboxylation of pyruvate and linking glycolysis to the tricarboxylic acid
cycle and to the synthesis of fatty acids [56]. The observed significant down-regulation of the
PDPR gene in Group 3 (DLP-PD) in comparison with the healthy Group 5 (p-value � 0.0002)
by qPCR was discordant from those detected by the microarray, as shown in Fig 3(C).
Regarding Group 4 (patients affected by only PD), the rule also has 2 highly expressed
genes/probes: the IL12RB2 gene (1560999_a_at), and the CLECL1 gene (244413_at), which
was chosen to validate the gene expression by using the TaqMan assay Hs00416849_m1. The
CLECL1 (C-Type Lectin Like 1) gene acts as a co-stimulating molecule of T cells and plays a
role in the interaction of dendritic cells with T cells and the cells of the adaptive immune
response [56]. In the comparison between Group 4 and Group 5, there was a highly statistically
significant (p-value � 0.0064) expression of the CLECL1 gene in Group 4, validating the DEG
detected by microarray, as shown in Fig 3(D).
For Group 5 (healthy patients), the only highly expressed gene is the MEF2C (Myocyte
Enhancer Factor 2C) gene (identified by the 236395_at probe), and detected by the TaqMan
assay Hs00231149_m1. The MEF2C gene is involved in several normal pathways of muscular,
vascular, neural, megakaryocyte and platelet development, bone marrow B lymphopoiesis, B
cell survival and proliferation in response to BCR stimulation, efficient responses of IgG1 anti-
bodies to T cell dependent antigens and normal induction of B cells from the germinal center
[56]. The MEF2C gene expression by qPCR validated the DEG detected by microarray, as sig-
nificantly highly expressed in Group 5 when compared with Group 1 (p-value � 0.0425) (see
Fig 3(E)). It is interesting to compare PBMC gene expression between patients with the most
opposite healthy conditions, such as Groups 1, 2 and 5, in which the worst metabolic condition
(Group 1) showed the lowest level of MEF2C gene expression.
To our knowledge, this is the first initiative to investigate the expression of CDC42SE2 and
CLECL1 genes in the context of T2DM, DLP and PD, demonstrating the innovative character
of this study. Regarding CFLAR gene expression, only one study was reported in the literature
investigating the relationship between body composition and BMI in children and DNA meth-
ylation. CFLAR gene expression was positively regulated in PBMCs of obese children [57].
Similarly, only one study investigated the PDPR gene with the genetic risk for DM, but the
authors focused on type 1 DM, not allowing direct comparison with the T2DM results [58].
Two previous studies reported changes in the function of the MEF2C gene: Yuasa et al. [59]
found MEF2C transcriptional repression in patients with T2DM, and Davegårdh et al. [60]
verified a down-regulation of MEF2C related to obesity. Such results are in agreement with the
findings of our study, with MEF2C being more highly expressed in patients in Group 5 (sys-
temically and periodontally healthy individuals) than in Groups 1 and 2 (individuals with met-
abolic and periodontal involvement).
Although we originally utilized the ARM to investigate CFs and DEGs relevant in the con-
text of T2DM, DLP and PD, it is important to attest that:
PLOS ONE | https://doi.org/10.1371/journal.pone.0240269 October 2, 2020
17 / 22
PLOS ONE Association rule mining to detect clinical features and genes related to chronic inflammatory diseases
1. We just considered the periodontitis parameters as the consequent part of the rules because
the literature demands more evidences regarding the association between systemic diseases
like T2DM and DLP, with PD;
2. Regarding the CF+DEG rules, more rules could be selected for each patient group, permit-
ting biological validation of up- or down-regulated probesets/genes, but we had limitations
in the volume of biological samples of the patients (RNA obtained from PBMCs) necessary
for the RT-qPCR technique.
Conclusion
We demonstrated that ARM is a powerful data analysis technique to identify consistent pat-
terns between the clinical and molecular profiles of patients affected by specific pathological
panels. In addition, ARM was able to evidence relevant associations among important parame-
ters of the periodontal, glycemic, lipid, cardiovascular and obesity risk conditions of the
patients. Considering the qPCR validation results of the DEGs prospected by the CARs of each
group of patients, four of the five genes revealed significant differences in comparison to the
control group; two of them CLECL1 and MEF2C genes validated the previous microarray find-
ings. These last genes were referred to groups without systemic metabolic impairment (Group
4 and Group 5). Further studies will investigate other DEGs and other rules. Additionally, as
an alternative to other commonly used techniques, ARM can be applied as a highly-interpret-
able mining approach to analyze the gene expression signal, with the advantage of including
the patient’s clinical features. Moreover, the combination of CFs and DEGs can be utilized to
further estimate the patient’s chance of developing complex diseases, such as those studied
here.
Supporting information
S1 File. Clinical feature dataset.
(CSV)
S2 File. Gene expression dataset.
(TXT)
S1 Table. Association rules mined from the clinical feature dataset.
(XLS)
S2 Table. Class association rules mined from clinical feature and gene expression datasets
in conjunction.
(XLS)
Author Contributions
Conceptualization: Rosana Veroneze, Fernando J. Von Zuben, Raquel Mantuaneli Scarel-
Caminaga.
Data curation: Rosana Veroneze, Saˆmia Cruz Tfaile Corbi, Cristiane de S. Rocha, Cla´udia V.
Maurer-Morelli, Silvana Regina Perez Orrico, Joni A. Cirelli, Raquel Mantuaneli Scarel-
Caminaga.
Formal analysis: Rosana Veroneze, Saˆmia Cruz Tfaile Corbi, Ba´rbara Roque da Silva, Cris-
tiane de S. Rocha, Cla´udia V. Maurer-Morelli, Silvana Regina Perez Orrico, Joni A. Cirelli.
PLOS ONE | https://doi.org/10.1371/journal.pone.0240269 October 2, 2020
18 / 22
PLOS ONE Association rule mining to detect clinical features and genes related to chronic inflammatory diseases
Funding acquisition: Rosana Veroneze, Silvana Regina Perez Orrico, Fernando J. Von Zuben,
Raquel Mantuaneli Scarel-Caminaga.
Investigation: Rosana Veroneze, Saˆmia Cruz Tfaile Corbi, Ba´rbara Roque da Silva, Cristiane
de S. Rocha.
Methodology: Rosana Veroneze, Cla´udia V. Maurer-Morelli, Silvana Regina Perez Orrico,
Fernando J. Von Zuben, Raquel Mantuaneli Scarel-Caminaga.
Project administration: Rosana Veroneze, Silvana Regina Perez Orrico, Fernando J. Von
Zuben, Raquel Mantuaneli Scarel-Caminaga.
Resources: Rosana Veroneze, Cla´udia V. Maurer-Morelli, Silvana Regina Perez Orrico, Fer-
nando J. Von Zuben, Raquel Mantuaneli Scarel-Caminaga.
Software: Rosana Veroneze, Fernando J. Von Zuben.
Supervision: Cla´udia V. Maurer-Morelli, Silvana Regina Perez Orrico, Joni A. Cirelli, Fer-
nando J. Von Zuben, Raquel Mantuaneli Scarel-Caminaga.
Validation: Saˆmia Cruz Tfaile Corbi, Cla´udia V. Maurer-Morelli, Silvana Regina Perez Orrico,
Joni A. Cirelli, Raquel Mantuaneli Scarel-Caminaga.
Visualization: Rosana Veroneze, Ba´rbara Roque da Silva, Silvana Regina Perez Orrico, Joni A.
Cirelli, Raquel Mantuaneli Scarel-Caminaga.
Writing – original draft: Rosana Veroneze, Ba´rbara Roque da Silva, Fernando J. Von Zuben,
Raquel Mantuaneli Scarel-Caminaga.
Writing – review & editing: Rosana Veroneze, Saˆmia Cruz Tfaile Corbi, Ba´rbara Roque da
Silva, Cristiane de S. Rocha, Cla´udia V. Maurer-Morelli, Silvana Regina Perez Orrico, Joni
A. Cirelli, Fernando J. Von Zuben, Raquel Mantuaneli Scarel-Caminaga.
References
1. American Diabetes Association and others. Diagnosis and classification of diabetes mellitus. Diabetes
care. 2014; 37(Supplement 1):S81–S90. https://doi.org/10.2337/dc14-S081 PMID: 24357215
2.
Jeong E, Park N, Kim Y, Jeon JY, Chung WY, Yoon D. Temporal trajectories of accompanying comor-
bidities in patients with type 2 diabetes: a Korean nationwide observational study. Scientific Reports.
2020; 10(1):1–10. https://doi.org/10.1038/s41598-020-62482-1 PMID: 32218498
3. Holman N, Young B, Gadsby R. Current prevalence of Type 1 and Type 2 diabetes in adults and chil-
dren in the UK. Diabetic Medicine. 2015; 32(9):1119–1120. https://doi.org/10.1111/dme.12791 PMID:
25962518
4. Papatheodorou K, Papanas N, Banach M, Papazoglou D, Edmonds M. Complications of diabetes
2016. Journal of diabetes research. 2016; 2016.
5.
de Oliveira Otto MC, Afshin A, Micha R, Khatibzadeh S, Fahimi S, Singh G, et al. The impact of dietary
and metabolic risk factors on cardiovascular diseases and type 2 diabetes mortality in Brazil. PLoS
One. 2016; 11(3):e0151503. https://doi.org/10.1371/journal.pone.0151503 PMID: 26990765
6. Almeida Abdo J, Cirano FR, Casati MZ, Ribeiro FV, Giampaoli V, Viana Casarin RC, et al. Influence of
dyslipidemia and diabetes mellitus on chronic periodontal disease. Journal of periodontology. 2013; 84
(10):1401–1408. https://doi.org/10.1902/jop.2012.120366 PMID: 23136946
7. Nassar PO, Walker CS, Salvador CS, Felipetti FA, Orrico SRP, Nassar CA. Lipid profile of people with
diabetes mellitus type 2 and periodontal disease. Diabetes research and clinical practice. 2012; 96
(1):35–39. https://doi.org/10.1016/j.diabres.2011.11.017 PMID: 22154377
8. Khumaedi AI, Purnamasari D, Wijaya IP, Soeroso Y. The relationship of diabetes, periodontitis and car-
diovascular disease. Diabetes & Metabolic Syndrome: Clinical Research & Reviews. 2019; 13(2):1675–
1678. https://doi.org/10.1016/j.dsx.2019.03.023
9.
Zhou X, Zhang W, Liu X, Zhang W, Li Y. Interrelationship between diabetes and periodontitis: role of
hyperlipidemia. Archives of Oral Biology. 2015; 60(4):667–674. https://doi.org/10.1016/j.archoralbio.
2014.11.008 PMID: 25443979
PLOS ONE | https://doi.org/10.1371/journal.pone.0240269 October 2, 2020
19 / 22
PLOS ONE Association rule mining to detect clinical features and genes related to chronic inflammatory diseases
10. Prattichizzo F, De Nigris V, Spiga R, Mancuso E, La Sala L, Antonicelli R, et al. Inflammageing and
metaflammation: the yin and yang of type 2 diabetes. Ageing Research Reviews. 2018; 41:1–17.
https://doi.org/10.1016/j.arr.2017.10.003 PMID: 29081381
11. Preshaw P, Alba A, Herrera D, Jepsen S, Konstantinidis A, Makrilakis K, et al. Periodontitis and diabe-
tes: a two-way relationship. Diabetologia. 2012; 55(1):21–31. https://doi.org/10.1007/s00125-011-
2342-y PMID: 22057194
12.
Lalla E, Papapanou PN. Diabetes mellitus and periodontitis: a tale of two common interrelated diseases.
Nature Reviews Endocrinology. 2011; 7(12):738. https://doi.org/10.1038/nrendo.2011.106 PMID:
21709707
13. Michalowicz BS, Diehl SR, Gunsolley JC, Sparks BS, Brooks CN, Koertge TE, et al. Evidence of a sub-
stantial genetic basis for risk of adult periodontitis. Journal of periodontology. 2000; 71(11):1699–1707.
https://doi.org/10.1902/jop.2000.71.11.1699 PMID: 11128917
14.
Leite FR, Nascimento GG, Scheutz F, Lopez R. Effect of smoking on periodontitis: a systematic review
and meta-regression. American Journal of Preventive Medicine. 2018; 54(6):831–841. https://doi.org/
10.1016/j.amepre.2018.02.014 PMID: 29656920
15. Mealey BL, Oates TW. Diabetes mellitus and periodontal diseases. Journal of periodontology. 2006; 77
(8):1289–1303. https://doi.org/10.1902/jop.2006.050459 PMID: 16881798
16.
17.
Lalla E, Lamster IB. Assessment and management of patients with diabetes mellitus in the dental office.
Dental Clinics. 2012; 56(4):819–829. PMID: 23017553
Lo¨ e H. Periodontal disease: the sixth complication of diabetes mellitus. Diabetes care. 1993; 16
(1):329–334. https://doi.org/10.2337/diacare.16.1.329 PMID: 8422804
18. Carrizales-Sepu´lveda EF, Ordaz-Farı´as A, Vera-Pineda R, Flores-Ramı´rez R. Periodontal disease,
systemic inflammation and the risk of cardiovascular disease. Heart, Lung and Circulation. 2018; 27
(11):1327–1334. https://doi.org/10.1016/j.hlc.2018.05.102 PMID: 29903685
19. Arboleda S, Vargas M, Losada S, Pinto A. Review of obesity and periodontitis: an epidemiological view.
20.
British dental journal. 2019; 227(3):235–239. https://doi.org/10.1038/s41415-019-0611-1
Fentoğlu O¨ , Kırzıoğlu F, O¨ zdem M, Koc¸ak H, Su¨ tc¸u¨ R, Sert T. Proinflammatory cytokine levels in hyper-
lipidemic patients with periodontitis after periodontal treatment. Oral diseases. 2012; 18(3):299–306.
https://doi.org/10.1111/j.1601-0825.2011.01880.x PMID: 22151458
21. Nepomuceno R, Pigossi SC, Finoti LS, Orrico SR, Cirelli JA, Barros SP, et al. Serum lipid levels in
patients with periodontal disease: A meta-analysis and meta-regression. Journal of Clinical Periodontol-
ogy. 2017; 44(12):1192–1207. https://doi.org/10.1111/jcpe.12792 PMID: 28782128
22. Soory M. Inflammatory mechanisms and redox status in periodontal and cardiometabolic diseases:
effects of adjunctive nutritional antioxidants and statins. Infectious Disorders-Drug Targets (Formerly
Current Drug Targets-Infectious Disorders). 2012; 12(4):301–315. PMID: 22697128
23. McAllister K, Mechanic LE, Amos C, Aschard H, Blair IA, Chatterjee N, et al. Current challenges and
new opportunities for gene-environment interaction studies of complex diseases. American journal of
epidemiology. 2017; 186(7):753–761. https://doi.org/10.1093/aje/kwx227 PMID: 28978193
24. Buxton EK, Vohra S, Guo Y, Fogleman A, Patel R. Pediatric population health analysis of southern and
central Illinois region: A cross sectional retrospective study using association rule mining and multiple
logistic regression. Computer methods and programs in biomedicine. 2019; 178:145–153. https://doi.
org/10.1016/j.cmpb.2019.06.020 PMID: 31416543
25.
Ivančević V, Tusˇek I, Tusˇek J, Knezˇević M, Elheshk S, Luković I. Using association rule mining to iden-
tify risk factors for early childhood caries. Computer Methods and programs in Biomedicine. 2015; 122
(2):175–181. https://doi.org/10.1016/j.cmpb.2015.07.008 PMID: 26271408
26. Kalgotra P, Sharda R. BIARAM: A process for analyzing correlated brain regions using association rule
mining. Computer methods and programs in biomedicine. 2018; 162:99–108. https://doi.org/10.1016/j.
cmpb.2018.05.001 PMID: 29903499
27.
28.
Toti G, Vilalta R, Lindner P, Lefer B, Macias C, Price D. Analysis of correlation between pediatric
asthma exacerbation and exposure to pollutant mixtures with association rule mining. Artificial intelli-
gence in medicine. 2016; 74:44–52. PMID: 27964802
Lin Y, Qian X, Krischer J, Vehik K, Lee HS, Huang S. A rule-based prognostic model for type 1 diabetes
by identifying and synthesizing baseline profile patterns. PloS one. 2014; 9(6):e91095. https://doi.org/
10.1371/journal.pone.0091095 PMID: 24926781
29. Simon GJ, Schrom J, Castro MR, Li PW, Caraballo PJ. Survival association rule mining towards type 2
diabetes risk assessment. In: AMIA annual symposium proceedings. vol. 2013. American Medical Infor-
matics Association; 2013. p. 1293.
30. Kim HS, Shin AM, Kim MK, Kim YN. Comorbidity study on type 2 diabetes mellitus using data mining.
The Korean journal of internal medicine. 2012; 27(2):197. https://doi.org/10.3904/kjim.2012.27.2.197
PLOS ONE | https://doi.org/10.1371/journal.pone.0240269 October 2, 2020
20 / 22
PLOS ONE Association rule mining to detect clinical features and genes related to chronic inflammatory diseases
31. Ramezankhani A, Pournik O, Shahrabi J, Azizi F, Hadaegh F. An application of association rule mining
to extract risk pattern for type 2 diabetes using tehran lipid and glucose study database. International
journal of endocrinology and metabolism. 2015; 13(2). https://doi.org/10.5812/ijem.25389 PMID:
25926855
32. Bastos AS, Graves DT, Loureiro APdM, Rossa Ju´ nior C, Abdalla DSP, Faulin TdES, et al. Lipid peroxi-
dation is associated with the severity of periodontal disease and local inflammatory markers in patients
with type 2 diabetes. The Journal of Clinical Endocrinology & Metabolism. 2012; 97(8):E1353–E1362.
https://doi.org/10.1210/jc.2011-3397 PMID: 22564665
33. Corbi SC, Bastos AS, Orrico SR, Secolin R, Dos Santos RA, Takahashi CS, et al. Elevated micronu-
cleus frequency in patients with type 2 diabetes, dyslipidemia and periodontitis. Mutagenesis. 2014; 29
(6):433–439. https://doi.org/10.1093/mutage/geu043 PMID: 25239120
34.
de Souza Bastos A, Graves DT, de Melo Loureiro AP, Ju´ nior CR, Corbi SCT, Frizzera F, et al. Diabetes
and increased lipid peroxidation are associated with systemic inflammation even in well-controlled
patients. Journal of Diabetes and its Complications. 2016; 30(8):1593–1599. https://doi.org/10.1016/j.
jdiacomp.2016.07.011 PMID: 27497685
35. Association AD, et al. 2. Classification and diagnosis of diabetes: standards of medical care in diabetes
—2019. Diabetes Care. 2019; 42(Supplement 1):S13–S28. https://doi.org/10.2337/dc19-S002
36. Association AD, et al. 6. Glycemic Targets: Standards of Medical Care in Diabetes-2019. Diabetes
Care. 2019; 42(Suppl 1):S61. PMID: 30559232
37. Grundy SM, Stone NJ, Bailey AL, Beam C, Birtcher KK, Blumenthal RS, et al. 2018 AHA/ACC/
AACVPR/AAPA/ABC/ACPM/ADA/AGS/APhA/ASPC/NLA/PCNA guideline on the management of
blood cholesterol: executive summary: a report of the American College of Cardiology/American Heart
Association Task Force on Clinical Practice Guidelines. Journal of the American College of Cardiology.
2019; 73(24):3168–3209. https://doi.org/10.1016/j.jacc.2018.11.002 PMID: 30423391
38. Piepoli MF, Hoes AW, Agewall S, Albus C, Brotons C, Catapano AL, et al. 2016 European Guidelines
on cardiovascular disease prevention in clinical practice: The Sixth Joint Task Force of the European
Society of Cardiology and Other Societies on Cardiovascular Disease Prevention in Clinical Practice
(constituted by representatives of 10 societies and by invited experts) Developed with the special contri-
bution of the European Association for Cardiovascular Prevention & Rehabilitation (EACPR). European
heart journal. 2016; 37(29):2315–2381. https://doi.org/10.1093/eurheartj/ehw106 PMID: 27222591
39. Periodontology AA. International workshop for a classification of periodontal diseases and conditions.
Ann Periodontol. 1999; 4(1).
40. Koromantzos PA, Makrilakis K, Dereka X, Katsilambros N, Vrotsos IA, Madianos PN. A randomized,
controlled trial on the effect of non-surgical periodontal therapy in patients with type 2 diabetes. Part I:
effect on periodontal status and glycaemic control. Journal of clinical periodontology. 2011; 38(2):142–
147. https://doi.org/10.1111/j.1600-051X.2010.01652.x PMID: 21114680
41. Gautier L, Cope L, Bolstad BM, Irizarry RA. affy—analysis of Affymetrix GeneChip data at the probe
level. Bioinformatics. 2004; 20(3):307–315. https://doi.org/10.1093/bioinformatics/btg405
42. Breitling R, Armengaud P, Amtmann A, Herzyk P. Rank products: a simple, yet powerful, new method
to detect differentially regulated genes in replicated microarray experiments. FEBS letters. 2004; 573(1-
3):83–92. https://doi.org/10.1016/j.febslet.2004.07.055 PMID: 15327980
43. Corbi SC, de Vasconcellos JF, Bastos AS, Bussaneli DG, da Silva BR, Santos RA, et al. circulating lym-
phocytes and monocytes transcriptomic analysis of patients with type 2 diabetes mellitus, dyslipidemia
and periodontitis. Scientific Reports. 2020; 10(1):1–14. https://doi.org/10.1038/s41598-020-65042-9
PMID: 32424199
44.
45.
Zaki MJ, Hsiao CJ. CHARM: An Efficient Algorithm for Closed Itemset Mining. In: Proceedings of the
2002 SIAM International Conference on Data Mining. vol. 2; 2002. p. 457–473.
Lakhal L, Stumme G. Efficient mining of association rules based on formal concept analysis. In: Formal
concept analysis: Foundations and Applications. Springer; 2005. p. 180–195.
46. Veroneze R, Banerjee A, Von Zuben FJ. Enumerating all maximal biclusters in numerical datasets.
Information Sciences. 2017; 379:288–309. https://doi.org/10.1016/j.ins.2016.10.029
47. Veroneze R, Von Zuben FJ. New advances in enumerative biclustering algorithms with online partition-
ing. arXiv preprint arXiv:200304726. 2020;.
48. Organization WH. Obesity: preventing and managing the global epidemic. World Health Organization;
2000.
49.
Fruchart JC, Santos RD, Aguilar-Salinas C, Aikawa M, Al Rasadi K, Amarenco P, et al. The selective
peroxisome proliferator-activated receptor alpha modulator (SPPARMα) paradigm: conceptual frame-
work and therapeutic potential. Cardiovascular diabetology. 2019; 18(1):71. https://doi.org/10.1186/
s12933-019-0864-7 PMID: 31164165
PLOS ONE | https://doi.org/10.1371/journal.pone.0240269 October 2, 2020
21 / 22
PLOS ONE Association rule mining to detect clinical features and genes related to chronic inflammatory diseases
50. Halcox JP, Banegas JR, Roy C, Dallongeville J, De Backer G, Guallar E, et al. Prevalence and treat-
ment of atherogenic dyslipidemia in the primary prevention of cardiovascular disease in Europe: EUR-
IKA, a cross-sectional observational study. BMC cardiovascular disorders. 2017; 17(1):160. https://doi.
org/10.1186/s12872-017-0591-5 PMID: 28623902
51. Corbi SCT, Bastos AS, Nepomuceno R, Cirelli T, Santos RAd, Takahashi CS, et al. Expression profile
of genes potentially associated with adequate glycemic control in patients with type 2 diabetes mellitus.
Journal of diabetes research. 2017; 2017. https://doi.org/10.1155/2017/2180819 PMID: 28812028
52. Robinson JG, Wang S, Smith BJ, Jacobson TA. Meta-analysis of the relationship between non–high-
density lipoprotein cholesterol reduction and coronary heart disease risk. Journal of the American Col-
lege of Cardiology. 2009; 53(4):316–322. https://doi.org/10.1016/j.jacc.2008.10.024
53. Virani SS. Non-HDL cholesterol as a metric of good quality of care: opportunities and challenges. Texas
Heart Institute Journal. 2011; 38(2):160. PMID: 21494527
54. Brunner FJ, Waldeyer C, Ojeda F, Salomaa V, Kee F, Sans S, et al. Application of non-HDL cholesterol
for population-based cardiovascular risk stratification: results from the Multinational Cardiovascular
Risk Consortium. The Lancet. 2019; 394(10215):2173–2183. https://doi.org/10.1016/S0140-6736(19)
32519-X PMID: 31810609
55. Nepomuceno R, Vallerini BdF, da Silva RL, Corbi SC, Bastos AdS, dos Santos RA, et al. Systemic
expression of genes related to inflammation and lipid metabolism in patients with dyslipidemia, type 2
diabetes mellitus and chronic periodontitis. Diabetes & Metabolic Syndrome: Clinical Research &
Reviews. 2019; 13(4):2715–2722. https://doi.org/10.1016/j.dsx.2019.07.003
56. Weizmann Institute of Science. GeneCards: The Human Data Base; 2020. https://www.genecards.org.
57. Rzehak P, Covic M, Saffery R, Reischl E, Wahl S, Grote V, et al. DNA-methylation and body composi-
tion in preschool children: epigenome-wide-analysis in the European Childhood Obesity Project
(CHOP)-Study. Scientific reports. 2017; 7(1):1–13. https://doi.org/10.1038/s41598-017-13099-4 PMID:
29084944
58. Wallace C, Rotival M, Cooper JD, Rice CM, Yang JH, McNeill M, et al. Statistical colocalization of mono-
cyte gene expression and genetic risk variants for type 1 diabetes. Human molecular genetics. 2012; 21
(12):2815–2824. https://doi.org/10.1093/hmg/dds098 PMID: 22403184
59. Yuasa K, Aoki N, Hijikata T. JAZF1 promotes proliferation of C2C12 cells, but retards their myogenic dif-
ferentiation through transcriptional repression of MEF2C and MRF4—Implications for the role of Jazf1
variants in oncogenesis and type 2 diabetes. Experimental cell research. 2015; 336(2):287–297. https://
doi.org/10.1016/j.yexcr.2015.06.009 PMID: 26101156
60. Davegårdh C, Broholm C, Perfilyev A, Henriksen T, Garcı´a-Calzo´ n S, Peijs L, et al. Abnormal epige-
netic changes during differentiation of human skeletal muscle stem cells from obese subjects. BMC
medicine. 2017; 15(1):39. https://doi.org/10.1186/s12916-017-0792-x PMID: 28222718
PLOS ONE | https://doi.org/10.1371/journal.pone.0240269 October 2, 2020
22 / 22
PLOS ONE |
10.1371_journal.pone.0240987 | RESEARCH ARTICLE
Association between serum uric acid and
spirometric pulmonary function in Korean
adults: The 2016 Korea National Health and
Nutrition Examination Survey
Jae Won Hong, Jung Hyun Noh, Dong-Jun KimID*
Department of Internal Medicine, Ilsan-Paik Hospital, College of Medicine, Inje University, Koyang, Gyeonggi-
do, Republic of Korea
* djkim@paik.ac.kr
Abstract
Background
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: Hong JW, Noh JH, Kim D-J (2020)
Association between serum uric acid and
spirometric pulmonary function in Korean adults:
The 2016 Korea National Health and Nutrition
Examination Survey. PLoS ONE 15(10): e0240987.
https://doi.org/10.1371/journal.pone.0240987
Editor: Sung Kweon Cho, National Institutes of
Health, UNITED STATES
Received: December 12, 2019
Accepted: October 6, 2020
Published: October 22, 2020
Peer Review History: PLOS recognizes the
benefits of transparency in the peer review
process; therefore, we enable the publication of
all of the content of peer review and author
responses alongside final, published articles. The
editorial history of this article is available here:
https://doi.org/10.1371/journal.pone.0240987
Copyright: © 2020 Hong et al. This is an open
access article distributed under the terms of the
Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: Readers can access
the dataset by registering an account with the
Korean CDC website (https://knhanes.cdc.go.kr/
knhanes/index.do). There is a blue bar on the top of
A limited number of epidemiological studies have investigated the association between
serum uric acid and pulmonary function in the general population. However, the results
have been inconclusive.
Objectives
This study was performed to investigate the association between serum uric acid and spiro-
metric pulmonary function in general population.
Methods
Among the 8,150 participants who participated in the 2016 Korea National Health and Nutri-
tion Examination Survey, 2,901 participants were analyzed in this study. Subjects were
divided into four groups according to forced vital capacity (FVC)% predicted or forced expi-
ratory volume in 1 second (FEV1) % predicted quartiles. Participants in the lowest quartile of
FVC % predicted and FEV1% predicted were compared to those in the remaining quartiles
according to age, education level, household income, smoking status, alcohol consumption,
aerobic exercise, obesity, hypertension, diabetes, renal impairment, serum uric acid, and
hyperuricemia. Multivariable logistic regression analyses were used to calculate the odds
ratio (OR) of hyperuricemia for participants in the lowest quartile of FVC% and FEV1 pre-
dicted, with above covariates.
Results
In women, hyperuricemia was associated with lowest quartile of FVC% predicted (OR 1.71,
95% CI 1.06–2.75, p = 0.027) and FEV1 predicted (OR 1.70, 95% CI 1.06–2.74, p = 0.028)
respectively, serving as above confounding variables. In men, hyperuricemia (OR 1.54,
95% CI 1.07–2.22, p = 0.021) was associated with the lowest quartile of FEV1% predicted,
not FVC% predicted.
PLOS ONE | https://doi.org/10.1371/journal.pone.0240987 October 22, 2020
1 / 16
PLOS ONE the website. Click the third menu, written in Korean,
“원시자료” (The content about the raw data) on
the blue bar. There is a second submenu below the
blue bar, written in Korean “원시자료 다운로드
(Download the raw data)”. Once readers click this,
an e-mail address for log-in is required. Once
logged in, readers can download the raw data from
1998-2016 Korea National Health and Nutrition
Examination Survey database using SAS or SPSS.
The authors do not have any special access
privileges to the data. For other data related
inquiries, please contact the corresponding author.
Funding: The author(s) received no specific
funding for this work.
Competing interests: The authors have declared
that no competing interests exist.
Association between serum uric acid and pulmonary function
According to median age, in women, age � 56 years old with hyperuricemia was associ-
ated with lowest quartile of FVC% predicted (OR 1.85, 95% CI 1.04–3.28, p = 0.037) and
FEV1% predicted (OR 1.99, 95% CI 1.11–3.75, p = 0.021), respectively. In men, age � 56
years old with hyperuricemia was associated with lowest quartile of FEV1% predicted (OR
1.75, 95% CI 1.05–2.94, p = 0.033), not FCV% predicted.
Conclusions
Hyperuricemia was associated with lowest quartile of FEV1% or FVC% predicted in Korean
general population. This correlation between hyperuricemia and low pulmonary function
was more pronounced in women and older age.
Introduction
Uric acid is primarily synthesized in the liver, intestines, muscles, kidneys and vascular endo-
thelium, as an end product of the action of xanthine oxidase on exogenous purines [1]. Hyper-
uricemia is well known to be associated with increased risk of incident and recurrent gout [2].
However, recent research in uric acid has focused primarily on its potential as a mediator of
human diseases other than gout.
Most epidemiologic studies have suggested that elevated serum uric acid levels are associ-
ated with cardiovascular diseases, including coronary heart disease, stroke, congestive heart
failure, and hypertension [3–6]. Furthermore, traditional cardiovascular risk factors, including
metabolic syndrome, insulin resistance, obesity, non-alcoholic fatty liver disease, and chronic
kidney disease, are also related to elevated serum uric acid levels [7–11]. Several experimental
and clinical studies support a role of uric acid as a causal factor in these conditions, via its con-
tributions to systemic inflammation, endothelial dysfunction, and oxidative stress [1, 8].
In this context, increased levels of uric acid have also been observed in a variety of respira-
tory disorders, including obstructive sleep apnea, pulmonary hypertension and chronic
obstructive pulmonary disease (COPD) [12–14]. A limited number of epidemiological studies
have investigated the association between serum uric acid levels and pulmonary function in
the general population [15, 16]. However, the results have been inconclusive.
In this study, we investigated the association between serum uric acid and pulmonary func-
tion in Korean population using data from the 2016 Korea National Health and Nutrition
Examination Survey (KNHANES).
Methods
Study population and data collection
This study was based on data from the 2016 KNHANES, a cross-sectional and nationally rep-
resentative survey conducted by the Korean Center for Disease Control for Health Statistics.
The KNHANES has been conducted periodically since 1998 to assess the health and nutri-
tional status of the noninstitutionalized civilian population of Korea. Participants were selected
using a systemic proportional allocation sampling method with multistage stratification. A
standardized interview was conducted in the homes of all participants to collect information
on demographic variables, family history, medical history, medications used, and a variety of
other health-related variables. The health interview was based on an established questionnaire
to determine the demographic and socioeconomic characteristics of the subjects including
PLOS ONE | https://doi.org/10.1371/journal.pone.0240987 October 22, 2020
2 / 16
PLOS ONE Association between serum uric acid and pulmonary function
age, education level, occupation, income, marital status, smoking status, alcohol consumption,
aerobic exercise, previous and current diseases, and family disease history.
As part of the interview, subjects were asked whether they exercised with an intensity that
caused a slight increase in breathing rate and sweating; those who exercised regularly at mod-
erate intensity were asked about the frequency with which they exercised per week, and the
length of time per exercise session. Participants who performed aerobic exercise were defined
as those who spent at least 150 min performing moderate physical activity, or 75 min perform-
ing high-intensity physical activity, per week. Alcohol consumption was assessed based on
responses to questions regarding drinking behavior, including average alcohol consumption
and drinking frequency, during the month prior to the interview. Diabetes was defined as a
fasting plasma glucose (FPG) concentration � 126 mg/dL (7.0 mmol/L), current use of anti-
diabetes medication, or a previous diagnosis of diabetes by a physician. Obesity was defined as
a body mass index (BMI) �25 kg/m2 according to the Asia-Pacific obesity classification [17].
Height and weight were obtained using standardized techniques and equipment. Height was
measured to the nearest 0.1 cm using a portable stadiometer (Seriter, Bismarck, ND, USA).
Weight was measured to the nearest 0.1 kg using a Giant-150N calibrated balance-beam scale
(Hana, Seoul, Korea). BMI was calculated by dividing the weight by the height squared (kg/
m2). Systolic and diastolic blood pressure were measured by standard methods using a stan-
dard mercury sphygmomanometer (Baumanometer, WA Baum Co. Inc., Copiague, NY,
USA), while the patient was seated. Three measurements were obtained for all subjects at
5-min intervals, and the average of the second and third measurements was used in the analy-
sis. Hypertension was defined as systolic blood pressure � 140 mmHg, or diastolic blood
pressure � 90 mmHg, or use of antihypertensive medications irrespective of blood pressure.
Of the 8,150 participants in the 2016 KNHANES, the number of individuals aged � 19
years were 6,382. Among these subjects, 3,341 were assessed for both serum uric acid levels
and pulmonary function. Pulmonary function tests were performed only in individuals
aged � 40 years.
A total of 282 subjects were excluded from the analysis due to pre-existing diseases, includ-
ing liver cirrhosis (n = 13), renal failure (n = 9), lung cancer (n = 9), asthma (n = 94), heart fail-
ure (n = 117), and cerebrovascular accident (n = 60). A total of 158 examinees who had
baseline FEV1 or FVC quality attribute of “D (questionable results, use with caution)” or
“F”(results not valid)” were also exclude [18, 19]. Finally, 2,901 participants were analyzed in
this study (Fig 1).
Pulmonary function test
Well-trained technicians measured forced vital capacity (FVC) and forced expiratory volume
in 1 second (FEV1) using a spirometer (Vyntus Spiro; Care Fusion, San Diego, CA, USA) and
SentrySuite (Care Fusion) according to the American Thoracic Society criteria [20]. All spi-
rometry values were prebronchodilator results.
The FVC% predicted and FEV1% predicted were calculated by dividing the FVC and FEV1
by the predicted FVC and FEV1, respectively. The predicted FEV1 and FVC for each subject
were calculated using published equations [21]. Subjects were divided into four groups accord-
ing to FVC% predicted or FEV1% predicted quartiles, as follows:
FVC% predicted:
Men: Q1 <82.4%; Q2 82.–89.9%; Q3 90.0–97.2%; Q4 � 97.3%
Women: Q1, <84.1%; Q2 84.2–91.8%; Q3 91.9–100.0%; Q4 � 100.1%
FEV1% predicted:
PLOS ONE | https://doi.org/10.1371/journal.pone.0240987 October 22, 2020
3 / 16
PLOS ONE Association between serum uric acid and pulmonary function
Fig 1. Participants disposition to be included in this study.
https://doi.org/10.1371/journal.pone.0240987.g001
Men: Q1 <81.8%; Q2 81.9–89.9%; Q3 90.0–97.6%; Q4 � 97.7%
Women: Q1 <84.7%; Q2 84.8–92.7%; Q3 92.8–100.0%; Q4 � 100.1%
Obstructive pattern was defined as FEV1/FVC<0.70, and restrictive pattern was defined as
FEV1/FVC�0.7 and FVC <80%, While normal lung function was defined as FEV1/
FVC�0.70 and FVC �80% [22, 23].
Laboratory methods
Serum uric acid levels were measured using a uricase colorimetry method with an autoanalyzer
(7600-210; Hitachi, Tokyo, Japan), with serum uric acid levels > 7.0 mg/dL in men and > 6.0
mg/dL in women defined as hyperuricemia.
The estimated glomerular filtration rate (eGFR) was calculated based on serum creatinine
levels using the Chronic Kidney Disease Epidemiology Collaboration equation. Renal
impairment was defined as an eGFR < 60 mL/min/1.73 m2 [24].
Ethics statement
This study was approved by the institutional review board of Ilsan Paik Hospital, Republic of
Korea (IRB 2019-10-012). After obtaining approval for the study protocol, the KNHANES
dataset was made available at the request of the investigator. Because the dataset did not
include any personal information and participants’ consent had already been given for the
KNHANES, our study was exempt from the requirement for informed consent.
PLOS ONE | https://doi.org/10.1371/journal.pone.0240987 October 22, 2020
4 / 16
PLOS ONE Association between serum uric acid and pulmonary function
Statistical analyses
Statistical analyses were performed using SPSS for Windows software (ver. 21.0; SPSS Inc.,
Chicago, IL, USA). Demographic and clinical characteristics, including gender, age, education
level (elementary school/junior high school/senior high school/college graduated), smoking
status (non-/ex-/current smoker), alcohol consumption (none/1-3/week/�4/week), aerobic
exercise, obesity, hypertension, diabetes, and hyperuricemia were analyzed using chi-squared
tests. Household income, FVC% predicted, FEV1% predicted, and serum uric acid were ana-
lyzed by independent samples t-tests. Age, education level, household income, smoking status,
alcohol consumption, aerobic exercise, obesity, hypertension, diabetes, renal impairment,
serum uric acid, and hyperuricemia of participants in the lowest quartile of FEV1 or FVC%
predicted (Q1 group) were compared to those in the remaining quartiles (Q2–4 group) using
independent samples t-tests or chi-squared tests. Comparisons of serum uric acid and the
prevalence of hyperuricemia between Q1 group and Q2–4 group were done after adjusting for
confounding factors in model 1 (Age) and 2 (all variables) (ANCOVA test). Serum uric acid
and the prevalence of hyperuricemia were adjusted for confounding factors in model 1 (Age)
and 2 (all variables). Spearman’s correlation was used in order to determine the association
between serum uric acid level and FVC% predicted, FEV1% predicted or FEV1/FVC. Age was
adjusted in Model 1. In Model 2, age, education, household income (KRW), smoking (non-,
ex-, and current smoker), alcohol consumption (none, 1-3/week, and � 4/week), aerobic exer-
cise, BMI (kg/m2), systolic BP (mmHg), fasting plasma glucose (mg/dL), and eGFR (mL/min/
1.73 m2) were adjusted.
Multivariable logistic regression analyses were used to calculate the odds ratio (OR) of
hyperuricemia for participants in the lowest quartile of FVC% predicted and FEV1% pre-
dicted, with age, education, household income, alcohol consumption, smoking, aerobic exer-
cise, obesity, hypertension, diabetes, and renal impairment serving as covariates. All tests were
two sided, with p < 0.05 considered indicative of statistical significance.
Results
Clinical characteristics of the study population
The demographic and clinical characteristics of the study population according to gender are
shown in Table 1. The median age was 57 years (range: 40–80 years), and 57% of the partici-
pants were female. The mean FVC% predicted and FEV1% predicted in the general population
were 91.1 ± 0.2 and 91.2 ± 0.2, respectively. The mean serum uric acid level was 4.9 ± 0.1 mg/
dL, with an overall prevalence of hyperuricemia of 9.7%.
Men were found to have higher levels of education and household income, and were also
more likely to smoke and engage in heavy alcohol consumption compared to women. The
rates of aerobic exercise, obesity, hypertension, and diabetes were also higher in men than
women. The mean serum uric acid level was 5.7 ± 0.1 mg/dL in men and 4.4 ± 0.1 mg/dL in
women. The overall prevalence of hyperuricemia was 14.8% in men and 5.7% in women. Both
the mean FVC% predicted and mean FEV1% predicted were higher in women than in men
(89.6 ± 0.3 vs. 92.1 ± 0.3; p < 0.001, and 89.0 ± 0.4 vs. 92.9 ± 0.3; p < 0.001, respectively).
Serum uric acid and the prevalence of hyperuricemia according to the
FEV1/FVC 0.7
Overall, 74.7% of participants showed normal spirometric values. 12.1% and 13.2% of subjects
have restrictive pattern and obstructive pattern in pulmonary function test, respectively. We
compared the values of serum uric acid and prevalence of hyperuricemia in those with FEV1/
PLOS ONE | https://doi.org/10.1371/journal.pone.0240987 October 22, 2020
5 / 16
PLOS ONE Association between serum uric acid and pulmonary function
Table 1. Clinical characteristics of study population.
Age (years)
Education (%)
Men (n = 1,266)
Women (n = 1,635)
56 (40–80)
57 (40–80)
Elementary school graduated
Junior high school graduated
Senior high school graduated
College graduated
20.3
12.1
30.6
37.0
32.2
14.0
31.3
22.5
Household Income (x10,000 KRW)
423.0 ± 8.8
392.3 ± 7.9
Smoking (%)
Non-smoker
Ex- smoker
Current smoker
None
1-3/week
� 4/week
Alcohol consumption (%)
Aerobic exercise (%)
Obesity (%)
Hypertension (%)
Diabetes (%)
Renal impairment (%)
FVC% predicted
FEV1% predicted
Serum uric acid (mg/dL)
Hyperuricemia (%)
19.1
49.0
31.9
19.7
65.1
15.2
45.7
41.4
43.6
17.5
4.3
89.6 ± 0.3
89.0 ± 0.4
5.7 ± 0.1
14.8
92.3
3.0
4.7
40.0
57.7
2.3
40.4
36.8
35.4
12.7
3.5
92.1 ± 0.3
92.9 ± 0.3
4.4 ± 0.1
5.7
Data are expressed as mean ± SEM or %, except age. Age are expressed as median (minimum-maximum).
FVC, forced vital capacity; FEV1, forced expiratory volume in 1 second.
https://doi.org/10.1371/journal.pone.0240987.t001
p
0.870
<0.001
0.009
<0.001
<0.001
0.005
0.011
<0.001
<0.001
0.245
<0.001
<0.001
<0.001
<0.001
Total (n = 2,901)
57 (40–80)
27.0
13.2
31.0
28.8
405.7 ± 5.9
60.4
23.1
16.6
31.2
60.9
7.9
42.7
38.8
39.0
14.8
3.9
91.1 ± 0.2
91.2 ± 0.2
4.9 ± 0.1
9.7
FVC � 0.7 to those with < 0.7. The prevalence of hyperuricemia in those with FEV1/
FVC � 0.7 was 14.2% in men and 5.6% in women. The prevalence of hyperuricemia in those
with FEV1/FVC < 0.7 was 17.2% in men and 8.6% in women. There is no significant differ-
ence in the prevalence of hyperuricemia according to the obstructive pattern in pulmonary
function test, using FEV1/FVC 0.7 criteria. The mean serum uric acid level in those with
FEV1/FVC � 0.7 was 5.7 ± 0.1 mg/dL in men and 4.4 ± 0.1 mg/dL in women. The mean
serum uric acid level in those with FEV1/FVC < 0.7 was 5.8 ± 0.1 mg/dL in men and 4.4 ± 0.1
mg/dL in women. There is also no significant difference in the mean serum uric acid levels
according to the obstructive pattern in pulmonary function test.
Clinical characteristics according to gender and FVC % predicted
We compared clinical characteristics between the participants in the lowest FVC% predicted
quartile (Q1 group) and those in the remaining quartiles (Q2–4 group) (Table 2). In men, the
median FVC% predicted of the Q1 and Q2–4 groups were 77.1 and 93.4, respectively. The
Q2–4 group was younger, and had a higher education level and higher household income,
compared to the Q1 group. The rates of obesity, hypertension, diabetes, and renal impairment
were higher in the Q1 group. Serum uric acid levels, and the prevalence of hyperuricemia,
were not statistically different between groups.
PLOS ONE | https://doi.org/10.1371/journal.pone.0240987 October 22, 2020
6 / 16
PLOS ONE p
<0.001
<0.001
<0.001
0.001
0.523
0.004
0.010
<0.001
<0.001
<0.001
0.002
<0.001
<0.001
0.006
<0.001
0.001
0.025
FVC% predicted
Age (years)
Education (%)
Elementary school graduated
Junior high school graduated
Senior high school graduated
College graduated
Household Income (x10,000
KRW)
Smoking (%)
Non-smoker
Ex- smoker
Current smoker
Alcohol consumption (%)
None
1-3/week
� 4/week
Aerobic exercise (%)
Obesity (%)
Hypertension (%)
Diabetes (%)
Renal impairment (%)
Serum uric acid (mg/dL)
Table 2. Clinical characteristics according to gender and FVC% predicted.
Association between serum uric acid and pulmonary function
Men (n = 1,.266)
Women (n = 1,635)
FVC% predicted Q1
(n = 313)
FVC% predicted Q2-4
(n = 953)
p
FVC% predicted Q1
(n = 406)
FVC% predicted Q2-4
(n = 1,229)
77.1 (33.1–82.3)
93.4 (82.5–121.6)
64 (40–80)
55 (40–80)
<0.001
<0.001
0.001
79.4 (48.2–84.1)
95.8 (84.1–138.6)
61 (40–80)
55 (40–80)
26.5
14.4
29.1
30.0
18.3
11.3
31.2
39.2
40.6
14.5
28.6
16.3
29.5
13.8
32.1
24.6
353.2 ± 17.8
446.0 ± 10.0
<0.001
348.2 ± 15.1
406.7 ± 9.2
19.2
53.0
27.8
26.2
58.1
15.7
44.1
52.1
56.5
28.4
7.0
19.1
47.6
33.3
17.6
67.4
15.0
46.3
37.9
39.3
13.9
3.5
0.165
0.003
0.514
<0.001
<0.001
<0.001
0.010
0.645
0.175
0.159
0.233
0.077
0.136
93.1
3.2
3.7
46.8
50.7
2.5
35.0
50.7
48.3
20.2
6.2
92.0
2.9
5.0
37.8
60.0
2.2
42.2
32.1
31.2
10.3
2.6
4.6 (4.5–4.7)
4.5 (4.5–4.6)
4.5 (4.4–4.6)
9.9
9.1 (6.8–11.4)
7.9 (5.7–10.1)
4.3 (4.2–4.3)
4.3 (4.2–4.4)
4.3 (4.3–4.4)
4.4
4.6 (3.4–5.9)
5.0 (3.8–6.2)
Unadjusted
Model 1
Model 2
5.7 (5.6–5.9)
5.8 (5.6–5.9)
5.8 (5.6–5.9)
Hyperuricemia (%)
Unadjusted
16.9
Model 1
Model 2
18.0 (14.0–22.1)
17.6 (13.5–21.6)
5.7 (5.6–5.8)
5.7 (5.6–5.8)
5.7 (5.6–5.8)
14.2
13.8 (11.5–16.2)
14.0 (11.7–16.2)
Data are expressed as mean ± SEM or %, except FVC% predicted, age, serum uric acid, and hyperuricemia. FVC% predicted and age are expressed as median
(minimum-maximum). Serum uric acid, and hyperuricemia are expressed as mean (95% CI). FVC, forced vital capacity.
Model 1: adjusted for age Model 2: adjusted for all variables
https://doi.org/10.1371/journal.pone.0240987.t002
In women, the median FVC% predicted of the Q1 and Q2–4 groups were 79.4 and 95.8,
respectively. The Q2–4 group was younger, had a higher education level and household income,
and was more likely to engage in aerobic exercise compared to the Q1 group. The rates of fre-
quent alcohol consumption, obesity, hypertension, diabetes, and renal impairment were all
higher in the Q1 group. Serum uric acid levels (4.6 mg/dL vs. 4.3 mg/dL) and the prevalence of
hyperuricemia (9.9% vs. 4.4%) were higher in the Q1 group compared to the Q2–4 group. These
differences remained statistically significant after adjusting for all variables (model 1 and 2).
Clinical characteristics according to gender and FEV1% predicted
Next, we compared clinical characteristics between participants in the lowest quartile of
FEV1% predicted (Q1 group) and those in the remaining quartiles (Q2–4 group) (Table 3).
PLOS ONE | https://doi.org/10.1371/journal.pone.0240987 October 22, 2020
7 / 16
PLOS ONE Association between serum uric acid and pulmonary function
Table 3. Clinical characteristics according to gender and FEV1% predicted.
Men (n = 1,.266)
Women (n = 1,635)
FEV1% predicted Q1
(n = 315)
FEV1% predicted Q2-4
(n = 951)
p
FEV1% predicted Q1
(n = 411)
FEV1% predicted Q2-4
(n = 1,224)
p
FEV1% predicted
Age (years)
Education (%)
Elementary school graduated
Junior high school graduated
Senior high school graduated
College graduated
Household Income (x10,000
KRW)
Smoking (%)
Non- smoker
Ex-smoker
Current smoker
Alcohol consumption (%)
None
1-3/week
� 4/week
Aerobic exercise (%)
Obesity (%)
Hypertension (%)
Diabetes (%)
Renal impairment (%)
Serum uric acid (mg/dL)
74.9 (28.1–81.7)
93.5 (81.8–129.8)
62 (40–80)
55 (40–80)
27.9
16.8
27.0
28.3
17.8
10.5
31.9
39.9
375.8 ± 18.1
438.7 ± 10.0
15.2
49.8
34.9
24.4
53.7
21.9
42.2
41.0
51.1
24.8
5.4
20.4
48.7
30.9
18.2
68.9
12.9
46.9
41.5
41.1
15.0
4.0
Unadjusted
Model 1
Model 2
5.7 (5.6–5.9)
5.8 (5.6–5.9)
5.8 (5.7–6.0)
Hyperuricemia (%)
Unadjusted
18.4
Model 1
Model 2
19.1 (15.2–23.1)
19.2 (15.3–23.2)
5.7 (5.6–5.8)
5.7 (5.6–5.8)
5.7 (5.6–5.8)
13.7
13.4 (11.2–15.7)
13.4 (11.2–15.7)
<0.001
<0.001
<0.001
0.002
0.103
<0.001
0.152
0.895
0.002
<0.001
0.298
0.508
0.212
0.105
0.040
0.015
0.014
78.9 (47.0–84.7)
96.9 (84.7–158.0)
<0.001
57 (40–80)
56 (40–80)
33.1
12.4
32.8
21.7
31.9
14.5
30.7
22.8
386.2 ± 15.3
394.3 ± 9.2
93.2
2.7
4.1
45.7
51.8
2.4
41.1
36.0
37.7
14.8
4.6
92.0
3.1
4.9
38.1
59.7
2.2
40.2
37.0
34.6
12.0
3.1
4.4 (4.3–4.5)
4.4 (4.3–4.5)
4.4 (4.3–4.5)
8.3
8.1 (5.8–10.3)
7.8 (5.6–10.0)
4.3 (4.3–4.4)
4.3 (4.3–4.4)
4.3 (4.3–4.4)
4.9
5.0 (3.7–6.3)
5.0 (3.7–6.3)
0.044
0.626
0.651
0.734
0.019
0.772
0.723
0.283
0.146
0.162
0.095
0.158
0.226
0.011
0.020
0.034
Data are expressed as mean ± SEM or %, except FEV1% predicted, age, serum uric acid, and hyperuricemia. FEV1% predicted and age are expressed as median
(minimum-maximum). Serum uric acid, and hyperuricemia are expressed as mean (95% CI). FEV1, forced expiratory volume in 1 second.
Model 1: adjusted for age Model 2: adjusted for age, education, household income, smoking, alcohol consumption, hypertension, and diabetes
https://doi.org/10.1371/journal.pone.0240987.t003
In men, the median FEV1% predicted of the Q1 and Q2–4 groups were 74.9 and 93.5,
respectively. The Q2–4 group was younger age, and had higher education household income
levels compared to the Q1 group. The rates of frequent alcohol consumption, hypertension,
and diabetes were all higher in the Q1 group. Serum uric acid levels were not statistically dif-
ferent between groups (5.7 mg/dL vs. 5.7 mg/dL for the Q1 and Q2–4 groups, respectively).
However, the prevalence of hyperuricemia was significantly higher in the Q1 group, even after
adjusting for age, education, household income, smoking, alcohol consumption, hypertension,
and diabetes (model 1 and 2).
In women, the median FEV1% predicted of the Q1 and Q2–4 groups were 78.9 and 96.9,
respectively. The Q2–4 group was younger and consumed less alcohol compared to the Q1
group. Serum uric acid levels were not significantly different between the groups (4.4 mg/dL
PLOS ONE | https://doi.org/10.1371/journal.pone.0240987 October 22, 2020
8 / 16
PLOS ONE Association between serum uric acid and pulmonary function
vs. 4.3 mg/dL for the Q1 and Q2–4 groups, respectively). The prevalence of hyperuricemia
(8.3% vs. 4.9%) was significantly higher in the Q1 group, even after adjusting for above vari-
ables (model 1 and 2).
Correlations of serum uric acid with FVC% predicted, FEV1% predicted,
and FEV1/FVC
No association was observed between serum uric acid levels and spirometric pulmonary func-
tion in men (Table 4). In contrast, unadjusted or adjusted FVC% predicted was negatively cor-
related with serum uric acid levels in women (Spearman correlation coefficient -0.145,
p < 0.001). Unadjusted or adjusted FEV1% predicted was also negatively associated with serum
uric acid levels (Spearman correlation coefficient -0.058, p = 0.019) in women, irrespective of
age. There is no significant association between serum uric acid levels and FEV1/FVC (Table 4).
Multivariable logistic regression analysis of the lowest quartile of FVC%
predicted
Logistic regression analyses were performed to identify factors independently associated with
the lowest quartile of FVC% predicted. Age, education level, household income, alcohol con-
sumption, smoking, aerobic exercise, obesity, hypertension, diabetes, renal impairment, and
hyperuricemia were included as potential confounders (Table 5).
Older age, obesity, and diabetes were associated with the lowest quartile of FVC% predicted
for both men and women. In only women, hyperuricemia (OR 1.71, 95% CI 1.06–2.75,
p = 0.027) was associated with lowest quartile of FVC% predicted.
Multivariable logistic regression analyses of lowest quartile of FEV1%
predicted
Logistic regression analyses were performed to identify factors independently associated with
the lowest quartile of FEV1% predicted. Age, education level, household income, alcohol con-
sumption, smoking, aerobic exercise, obesity, hypertension, diabetes, renal impairment, and
hyperuricemia were included as potential confounders (Table 6). In men, age, current smoker,
diabetes, and hyperuricemia (OR 1.54, 95% CI 1.07–2.22, p = 0.021) were associated with the
lowest quartile of FEV1% predicted. In women, age and hyperuricemia (OR 1.70, 95% CI
1.06–2.74, p = 0.028) were associated with the lowest quartile of FEV1% predicted.
Odds ratio of hyperuricemia for the lowest quartile of FCV% predicted and
the lowest quartile of FEV1% predicted according to age � 56 years old
and < 56 years old
Additionally, we showed odds ratios of hyperuricemia for the lowest quartile of FCV% pre-
dicted and the lowest quartile of FEV1% predicted according to age � 56 years old and < 56
years old (median age) in Table 7. In women, age � 56 years old with hyperuricemia was asso-
ciated with lowest quartile of FVC% predicted (OR 1.85, 95% CI 1.04–3.28, p = 0.037) and
FEV1% predicted (OR 1.99, 95% CI 1.11–3.75, p = 0.021), respectively. In men, age � 56 years
old with hyperuricemia was associated with lowest quartile of FEV1% predicted (OR 1.75, 95%
CI 1.05–2.94, p = 0.033), not FCV% predicted. Neither man nor woman younger than 56 years
old was associated with lowest quartile of FCV% predicted or FEV1% predicted.
PLOS ONE | https://doi.org/10.1371/journal.pone.0240987 October 22, 2020
9 / 16
PLOS ONE Association between serum uric acid and pulmonary function
Table 4. Correlation between serum uric acid and FVC% predicted or FEV1% predicted or FEV1/FVC.
Men
FVC% predicted
FVC% predicted model 1
FVC% predicted model 2
FEV1% predicted
FEV1% predicted model 1
FEV1% predicted model 2
FEV1/FVC
FEV1/FVC model 1
FEV1/FVC model 2
Women
FVC% predicted
FVC% predicted model 1
FVC% predicted model 2
FEV1% predicted
FEV1% predicted model 1
FEV1% predicted model 2
FEV1/FVC
FEV1/FVC model 1
FEV1/FVC model 2
Total
Age < 56 years old
Age � 56 years old
Total
Age < 56 years old
Age � 56 years old
Total
Age < 56 years old
Age � 56 years old
Total
Age < 56 years old
Age � 56 years old
Total
Age < 56 years old
Age � 56 years old
Total
Age < 56 years old
Age � 56 years old
r
-0.028
-0.053
-0.059
-0.059
-0.058
-0.022
-0.032
-0.053
-0.036
-0.067
-0.029
-0.019
-0.009
-0.034
-0.030
r
-0.145
-0.122
-0.095
-0.100
-0.087
-0.058
-0.064
-0.092
-0.078
-0.099
-0.018
-0.029
-0.015
-0.026
-0.047
p
0.323
0.059
0.037
0.153
0.140
0.424
0.259
0.063
0.387
0.087
0.306
0.510
0.740
0.407
0.445
p
<0.001
<0.001
<0.001
0.006
0.011
0.019
0.010
<0.001
0.032
0.004
0.476
0.240
0.556
0.481
0.171
r, Spearman correlation coefficient. FVC, forced vital capacity; FEV1, forced expiratory volume in 1 second.
Model 1: adjusted for age
Model 2: adjusted for age, education, household income (KRW), smoking (non-, ex-, and current smoker), alcohol
consumption (none, 1-3/week, and � 4/week), aerobic exercise, BMI (kg/m2), systolic BP (mmHg), fasting plasma
glucose (mg/dL), and eGFR (mL/min/1.73 m2)
https://doi.org/10.1371/journal.pone.0240987.t004
PLOS ONE | https://doi.org/10.1371/journal.pone.0240987 October 22, 2020
10 / 16
PLOS ONE Association between serum uric acid and pulmonary function
Table 5. Multivariable logistic regression analyses for lowest quartile of FVC% predicted.
Men
Women
Odds ratio (95% CI)
p
Odds ratio (95% CI)
p
Age (years)
19–44
Reference
Reference
45–64
1.98 (1.19–3.32)
0.009
1.29 (0.86–1.96)
0.228
65-
4.71 (2.67–8.33)
<0.001
2.70 (1.63–4.46)
<0.001
Education
Household income
Elementary school graduated
Reference
Junior high school graduated
1.04 (0.65–1.57)
Senior high school graduated
0.98 (0.66–1.45)
College graduated
1.03 (0.67–1.57)
Quartile 1 (the lowest quartile)
Reference
Quartile 2
0.71 (0.47–1.06)
Quartile 3
0.62 (0.40–0.95)
Quartile 4 (the highest quartile)
0.63 (0.40–0.98)
Alcohol consumption
None
Reference
1-3/week
0.81 (0.57–1.15)
� 4/week
0.74 (0.47–1.18)
Non-smoker
Reference
Ex-smoker
1.03 (0.71–1.50)
Current smoker
1.08 (0.71–1.64)
1.03 (0.78–1.36)
0.907
0.857
0.908
0.904
0.061
0.093
0.028
0.042
0.392
0.238
0.201
0.879
0.722
0.843
Reference
1.10 (0.75–1.60)
1.19 (0.85–1.68)
1.07 (0.71–1.61)
Reference
1.31 (0.93–1.86)
1.48 (1.02–2.14)
1.14 (0.76–1.69)
Reference
0.87 (0.68–1.12)
1.05 (0.48–2.27)
Reference
1.09 (0.55–2.14)
0.77 (0.42–1.40)
0.82 (0.64–1.05)
0.704
0.640
0.314
0.744
0.128
0.041
0.526
0.283
0.910
0.805
0.385
0.124
1.94 (1.46–2.57)
<0.001
1.74 (1.36–2.22)
<0.001
1.42 (1.07–1.89)
1.75 (1.25–2.45)
0.84 (0.45–1.58)
1.32 (0.90–1.93)
0.016
0.001
0.589
0.151
1.29 (0.99–1.68)
1.58 (1.14–2.21)
0.95 (0.52–1.75)
1.71 (1.06–2.75)
0.065
0.007
0.867
0.027
Smoking
Aerobic exercise
Obesity
Hypertension
Diabetes
Renal impairment
Hyperuricemia
FVC, forced vital capacity.
https://doi.org/10.1371/journal.pone.0240987.t005
Discussion
Using data from the 2016 KNHANES, the prevalence of hyperuricemia in the representative
sample of Korean adults included in this study was 9.7% (14.8% in men, 5.7% in women). We
found that hyperuricemia was inversely associated with spirometric pulmonary function in
Korean population. This reverse correlation between hyperuricemia and pulmonary function
was more pronounced in women and older age (�56 years old).
Previous epidemiological studies have investigated the association between serum uric acid
and pulmonary function in the general population, however, the results have been inconclu-
sive. A recent Japanese study showed similar results to those presented here, including signifi-
cant inverse correlations of serum uric acid levels with spirometric parameters in female
subjects who participated in an annual health check [15].
In contrast, a second study found a significant positive correlation between pulmonary
function and serum uric acid levels, in both males and females recruited from Kangbuk Sam-
sung Hospital Health Screening Center, Seoul, Korea [16]. While that study was similar to the
analysis presented herein, in that it used a cross-sectional design, there are some significant
PLOS ONE | https://doi.org/10.1371/journal.pone.0240987 October 22, 2020
11 / 16
PLOS ONE Association between serum uric acid and pulmonary function
Table 6. Multivariable logistic regression analyses for lowest quartile of FEV1% predicted.
Men
Women
Odds ratio (95% CI)
p
Odds ratio (95% CI)
p
Age (years)
19–44
Reference
45–64
1.32 (0.83–2.09)
65-
2.26 (1.33–3.83)
0.240
0.002
Education
Household income
Elementary school graduated
Reference
Junior high school graduated
1.10 (0.71–1.71)
Senior high school graduated
0.68 (0.46–0.99)
College graduated
0.70 (0.46–1.05)
Quartile 1 (the lowest quartile)
Reference
Quartile 2
0.72 (0.48–1.07)
Quartile 3
0.70 (0.46–1.07)
Quartile 4 (the highest quartile)
0.79 (0.51–1.23)
Alcohol consumption
0.670
0.046
0.087
0.099
0.100
0.298
None
Reference
1-3/week
0.71 (0.50–1.00)
� 4/week
1.23 (0.80–1.89)
0.051
0.353
Non-smoker
Reference
Ex-smoker
1.35 (0.92–1.98)
Current smoker
1.80 (1.19–2.72)
0.95 (0.72–1.24)
1.00 (0.76–1.32)
1.14 (0.86–1.51)
1.50 (1.07–2.10)
0.67 (0.351.28)
1.54 (1.07–2.22)
0.127
0.006
0.680
0.984
0.356
0.018
0.222
0.021
Smoking
Aerobic exercise
Obesity
Hypertension
Diabetes
Renal impairment
Hyperuricemia
FEV1, forced expiratory volume in 1 second.
https://doi.org/10.1371/journal.pone.0240987.t006
Reference
1.50 (1.02–2.21)
1.74 (1.07–2.83)
Reference
0.90 (0.61–1.32)
1.22 (0.87–1.70)
1.15 (0.78–1.69)
Reference
1.09 (0.76–1.54)
1.23 (0.85–1.78)
1.14 (0.78–1.67)
Reference
0.74 (0.58–0.94)
0.98 (0.46–2.08)
Reference
0.96 (0.48–1.92)
0.90 (0.51–1.60)
1.04 (0.82–1.31)
0.89 (0.69–1.14)
1.02 (0.78–1.33)
1.22 (0.87–1.72)
1.04 (0.55–1.96)
1.70 (1.06–2.74)
0.039
0.025
0.584
0.249
0.486
0.650
0.270
0.510
0.014
0.958
0.909
0.728
0.742
0.346
0.886
0.250
0.907
0.028
differences between the two studies. First, the subjects who participated in the Kangbuk Sam-
sung Health Study were significantly younger than those included in our study (mean age
were 40 and 56 years, respectively). Because age might be a moderator for the reverse
Table 7. Odds ratio of hyperuricemia for the lowest quartile of FCV% predicted and FEV1% predicted according to age � 56 years old and < 56 years old.
Men (n = 1266)
Odds ratio (95% CI)
for lowest quartile of FCV% predicted
< 56 years old
0.36 (0.77–2.43) (n = 606)
� 56 years old
1.32 (0.79–2.22) (n = 660)
for lowest quartile of FEV1% predicted
< 56 years old
� 56 years old
1.36 (0.79–2.33)
1.75 (1.05–2.94)
p
0.291
0.293
0.269
0.033
Women (n = 1635)
Odds ratio (95% CI)
1.54 (0.63–3.76) (n = 761)
1.85 (1.04–3.28) (n = 874)
1.20 (0.49–2.96)
1.99 (1.11–3.57)
p
0.347
0.037
0.692
0.021
Covariates: age, education level, household income, smoking status, alcohol consumption, aerobic exercise, obesity, hypertension, diabetes, and renal impairment.
FVC, forced vital capacity; FEV1, forced expiratory volume in 1 second.
https://doi.org/10.1371/journal.pone.0240987.t007
PLOS ONE | https://doi.org/10.1371/journal.pone.0240987 October 22, 2020
12 / 16
PLOS ONE Association between serum uric acid and pulmonary function
association between pulmonary function and serum uric acid, we performed sub-analysis
about odds ratio of hyperuricemia for the lowest quartile of FCV% predicted and the lowest
quartile of FEV1% predicted according to age � 56 years old and < 56 years old. Actually, cor-
relation between lowest quartile of FCV% and/or FEV1% predicted and hyperuricemia was
apparent only in women � 56 years old, not in women <56 years old. There was also signifi-
cant association between hyperuricemia and lowest quartile of FEV1% predicted in men � 56
years old, not in men <56 years old. Furthermore, the Kangbuk Samsung Health Study was
performed in Seoul, where the higher average socioeconomic status of that region could have
constituted a sampling bias. Using a definition of hyperuricemia of > 7.0 mg/dL uric acid in
men and > 6.0 mg/dL in women, the prevalence rates of hyperuricemia in the Kangbuk Sam-
sung Health study (25.5% in men and 8.5% in women) were much higher than those in both
our study (14.8% in men and 5.7% in women) and a previous study of the Korean population
(14.3% in men and 2.2% in women) [25]. The analytical methods were also different between
the studies. We compared serum uric acid level and the prevalence of hyperuricemia between
participants in the lowest quartile of pulmonary function and those in the other quartiles,
while the Kangbuk Samsung Health Study compared pulmonary function between hyperurice-
mia and normouricemia groups. In this study, we included age, education level, household
income, smoking status, alcohol consumption, aerobic exercise, obesity, hypertension, and
diabetes as covariates; only some of these variables overlapped with those adjusted for in the
Kangbuk Samsung Health Study, which included age, BMI, smoking status, alcohol consump-
tion, liver and renal function, lipid profiles, C-reactive protein level, mean blood pressure, and
glycosylated hemoglobin (HbA1c) level. Therefore, we believe that differences in study popula-
tions, potential confounders including age and sex, and analytical methods may be responsible
for the discrepancies in results seen among otherwise similar cross-sectional studies.
In this study, there was sex-difference in the association between serum uric acid and low
pulmonary function as well as age. No association was observed between serum uric acid levels
and spirometric pulmonary function in men. In contrast, unadjusted or adjusted FVC% pre-
dicted and FEV1% predicted were negatively correlated with serum uric acid levels in women,
irrespective of age. Regarding to hyperuricemia, in women, hyperuricemia was associated with
both lowest quartile of FVC% predicted and FEV1 predicted. However, in men, hyperuricemia
was associated with only the lowest quartile of FEV1% predicted, not FVC% predicted. These
results suggested that reverse correlation between serum uric acid and pulmonary function was
more pronounced in women. However, the mechanism of the age or sex-specific difference
between them is still unclear. Previously, although several studies showed that sex differences in
the association between hyperuricemia and cardiometabolic risk, there has been no studies
about sex differences in the relation between hyperuricemia and lung function [26–29].
One possibility is that there is an inherent difference in uric acid metabolism between the
sexes, including a higher fractional urate excretion and lower serum urate concentration in
women compared to men [26, 30]. We speculated that estrogen change might play a role in
mediation of the association between serum uric acid and lung function. Another possibility is
that other cofactors influencing pulmonary function not considered in this study may also be
affected by gender.
Oxidative stress is thought to play a mediating role in the association between uric acid level
and impaired lung function. The final reaction necessary for uric acid production involves the
conversion of xanthine to uric acid, which is catalyzed by the enzyme xanthine oxidoreductase
(XOR) [31]. During this process, XOR uses molecular oxygen as an electron acceptor, produc-
ing superoxide anions and other ROS [32]. The lungs could be a major target organ for exoge-
nous oxidants and endogenous ROS generated by inflammatory cells [33], with evidence of
higher uric acid concentrations within the epithelial lining fluid of the airways [33, 34]. ROS
PLOS ONE | https://doi.org/10.1371/journal.pone.0240987 October 22, 2020
13 / 16
PLOS ONE Association between serum uric acid and pulmonary function
and cellular injury have also been implicated in a variety of pulmonary diseases, including
asthma, COPD, acute respiratory distress syndrome (ARDS), and cystic fibrosis [33].
However, because high level of uric acid could be viewed as a ‘double-edge sword’; it may
have beneficial antioxidant effects, paradoxically, any potential association between hyperuri-
cemia and lung function in a cross-sectional study must be interpreted carefully, putting con-
founding factors aside [35–37].
Other mechanisms, such as hypoxia and systemic inflammation, may also modulate the
association between hyperuricemia and poor pulmonary function [38]. However, this study
excluded subjects with overt clinical disease, such as asthma, lung cancer, or heart failure diag-
nosed by a doctor.
The major strength of this study was the inclusion of a large, representative sample of adult
Koreans. To the best of our knowledge, this is the large scale study to demonstrate an inverse
association between hyperuricemia and pulmonary function in general population using
nationally representative data. However, our study also had some limitations. First, we did not
take into account the use of uric acid-lowering drugs, which reduce serum uric acid levels. Sec-
ond, we used a single measurement of uric acid. Third, although we adjusted for many con-
founding factors, there may be other potentially important variables that were not accounted
for, similar to other cross-sectional studies. Finally, we could not establish a causal relationship
between hyperuricemia and poor pulmonary function due to the inherent limitations of the
retrospective design.
Lastly, we could not identify a definitive mechanism underlying the age- and sex-specific
association between pulmonary function and hyperuricemia.
In conclusion, hyperuricemia was associated with lowest quartile of FEV1% or FVC% pre-
dicted using data taken from the 2016 KNHANES. This correlation between hyperuricemia
and low pulmonary function was more pronounced in women and older age. Further epidemi-
ologic studies will be required to confirm this reverse association between serum uric acid and
pulmonary function.
Author Contributions
Conceptualization: Dong-Jun Kim.
Data curation: Dong-Jun Kim.
Formal analysis: Dong-Jun Kim.
Investigation: Dong-Jun Kim.
Methodology: Dong-Jun Kim.
Software: Dong-Jun Kim.
Supervision: Dong-Jun Kim.
Writing – original draft: Jae Won Hong.
Writing – review & editing: Jung Hyun Noh.
References
1. El Ridi R, Tallima H. Physiological functions and pathogenic potential of uric acid: A review. Journal of
advanced research. 2017; 8(5):487–93. Epub 2017/07/28. https://doi.org/10.1016/j.jare.2017.03.003
PMID: 28748115; PubMed Central PMCID: PMC5512149.
2. Shiozawa A, Szabo SM, Bolzani A, Cheung A, Choi HK. Serum Uric Acid and the Risk of Incident and
Recurrent Gout: A Systematic Review. The Journal of rheumatology. 2017; 44(3):388–96. Epub 2017/
02/06. https://doi.org/10.3899/jrheum.160452 PMID: 28148699.
PLOS ONE | https://doi.org/10.1371/journal.pone.0240987 October 22, 2020
14 / 16
PLOS ONE Association between serum uric acid and pulmonary function
3. Ndrepepa G. Uric acid and cardiovascular disease. Clinica chimica acta; international journal of clinical
chemistry. 2018; 484:150–63. Epub 2018/05/29. https://doi.org/10.1016/j.cca.2018.05.046 PMID:
29803897.
4.
Li M, Hou W, Zhang X, Hu L, Tang Z. Hyperuricemia and risk of stroke: a systematic review and meta-
analysis of prospective studies. Atherosclerosis. 2014; 232(2):265–70. Epub 2014/01/29. https://doi.
org/10.1016/j.atherosclerosis.2013.11.051 PMID: 24468137.
5. Huang H, Huang B, Li Y, Huang Y, Li J, Yao H, et al. Uric acid and risk of heart failure: a systematic
review and meta-analysis. European journal of heart failure. 2014; 16(1):15–24. Epub 2013/08/13.
https://doi.org/10.1093/eurjhf/hft132 PMID: 23933579.
6.
Zuo T, Liu X, Jiang L, Mao S, Yin X, Guo L. Hyperuricemia and coronary heart disease mortality: a
meta-analysis of prospective cohort studies. BMC cardiovascular disorders. 2016; 16(1):207. Epub
2016/10/30. https://doi.org/10.1186/s12872-016-0379-z PMID: 27793095; PubMed Central PMCID:
PMC5084405.
7. Dehghan A, van Hoek M, Sijbrands EJ, Hofman A, Witteman JC. High serum uric acid as a novel risk
factor for type 2 diabetes. Diabetes care. 2008; 31(2):361–2. Epub 2007/11/06. https://doi.org/10.2337/
dc07-1276 PMID: 17977935.
8. Kanbay M, Jensen T, Solak Y, Le M, Roncal-Jimenez C, Rivard C, et al. Uric acid in metabolic syn-
drome: From an innocent bystander to a central player. European journal of internal medicine. 2016;
29:3–8. Epub 2015/12/26. https://doi.org/10.1016/j.ejim.2015.11.026 PMID: 26703429; PubMed Cen-
tral PMCID: PMC4826346.
9.
Zheng R, Chen C, Yang T, Chen Q, Lu R, Mao Y. Serum Uric Acid Levels and the Risk of Obesity: a
Longitudinal Population-Based Epidemiological Study. Clinical laboratory. 2017; 63(10):1581–7. Epub
2017/10/17. https://doi.org/10.7754/Clin.Lab.2017.170311 PMID: 29035437.
10. Darmawan G, Hamijoyo L, Hasan I. Association between Serum Uric Acid and Non-Alcoholic Fatty
Liver Disease: A Meta-Analysis. Acta medica Indonesiana. 2017; 49(2):136–47. Epub 2017/08/10.
PMID: 28790228.
11. Giordano C, Karasik O, King-Morris K, Asmar A. Uric Acid as a Marker of Kidney Disease: Review of
the Current Literature. Disease markers. 2015; 2015:382918. Epub 2015/06/25. https://doi.org/10.
1155/2015/382918 PMID: 26106252; PubMed Central PMCID: PMC4461768.
12. Wan YF, Zheng YL, Niu HY, Xu CQ, He YQ, Wang Y, et al. Uric acid levels in obstructive sleep apnea
patients with atrial fibrillation. Archives of medical research. 2014; 45(2):132–7. Epub 2013/12/10.
https://doi.org/10.1016/j.arcmed.2013.11.006 PMID: 24316394.
13. Bartziokas K, Papaioannou AI, Loukides S, Papadopoulos A, Haniotou A, Papiris S, et al. Serum uric
acid as a predictor of mortality and future exacerbations of COPD. The European respiratory journal.
2014; 43(1):43–53. Epub 2013/05/07. https://doi.org/10.1183/09031936.00209212 PMID: 23645404.
14. Simpson CE, Damico RL, Hummers L, Khair RM, Kolb TM, Hassoun PM, et al. Serum uric acid as a
marker of disease risk, severity, and survival in systemic sclerosis-related pulmonary arterial hyperten-
sion. Pulmonary circulation. 2019; 9(3):2045894019859477. Epub 2019/08/07. https://doi.org/10.1177/
2045894019859477 PMID: 31384431; PubMed Central PMCID: PMC6664664.
15. Aida Y, Shibata Y, Osaka D, Abe S, Inoue S, Fukuzaki K, et al. The relationship between serum uric
acid and spirometric values in participants in a health check: the Takahata study. International journal of
medical sciences. 2011; 8(6):470–8. Epub 2011/08/19. https://doi.org/10.7150/ijms.8.470 PMID:
21850198; PubMed Central PMCID: PMC3156995.
16. Song JU, Hwang J, Ahn JK. Serum uric acid is positively associated with pulmonary function in Korean
health screening examinees. Modern rheumatology. 2017; 27(6):1057–65. Epub 2017/07/12. https://
doi.org/10.1080/14397595.2017.1285981 PMID: 28693364.
17. Appropriate body-mass index for Asian populations and its implications for policy and intervention strat-
egies. Lancet. 2004; 363(9403):157–63. Epub 2004/01/17. https://doi.org/10.1016/S0140-6736(03)
15268-3 PMID: 14726171.
18. Miller MR, Hankinson J, Brusasco V, Burgos F, Casaburi R, Coates A, et al. Standardisation of spirome-
try. The European respiratory journal. 2005; 26(2):319–38. Epub 2005/08/02. https://doi.org/10.1183/
09031936.05.00034805 PMID: 16055882.
19. Hankinson JL, Eschenbacher B, Townsend M, Stocks J, Quanjer PH. Use of forced vital capacity and
forced expiratory volume in 1 second quality criteria for determining a valid test. The European respira-
tory journal. 2015; 45(5):1283–92. Epub 2014/12/30. https://doi.org/10.1183/09031936.00116814
PMID: 25537554; PubMed Central PMCID: PMC4416111 article at erj.ersjournals.com.
20. Standardization of Spirometry, 1994 Update. American Thoracic Society. American journal of respira-
tory and critical care medicine. 1995; 152(3):1107–36. Epub 1995/09/01. https://doi.org/10.1164/
ajrccm.152.3.7663792 PMID: 7663792.
PLOS ONE | https://doi.org/10.1371/journal.pone.0240987 October 22, 2020
15 / 16
PLOS ONE Association between serum uric acid and pulmonary function
21. Choi JK, Paek D, Lee JO. Normal Predictive Values of Spirometry in Korean Population. Tuberc Respir
Dis. 2005; 58(3):230–42.
22. Kim HY, Sohn TS, Seok H, Yeo CD, Kim YS, Song JY, et al. Prevalence and risk factors for reduced pul-
monary function in diabetic patients: The Korea National Health and Nutrition Examination Survey.
Korean J Intern Med. 2017; 32(4):682–9. Epub 2017/02/01. https://doi.org/10.3904/kjim.2016.045
PMID: 28142231; PubMed Central PMCID: PMC5511936.
23. Mannino DM, Doherty DE, Sonia Buist A. Global Initiative on Obstructive Lung Disease (GOLD) classifi-
cation of lung disease and mortality: findings from the Atherosclerosis Risk in Communities (ARIC)
study. Respir Med. 2006; 100(1):115–22. Epub 2005/05/17. https://doi.org/10.1016/j.rmed.2005.03.
035 PMID: 15893923.
24.
Levey AS, Stevens LA, Schmid CH, Zhang YL, Castro AF, 3rd, Feldman HI, et al. A new equation to
estimate glomerular filtration rate. Annals of internal medicine. 2009; 150(9):604–12. Epub 2009/05/06.
https://doi.org/10.7326/0003-4819-150-9-200905050-00006 PMID: 19414839; PubMed Central
PMCID: PMC2763564.
25. Eung Ho K, Kyeongman J, Keon Woo P, et al. The Prevalence of Gout among Hyperuricemic Popula-
tion. J Rheum Dis. 2004; 11(1):7–13. Epub 03/30.
26.
27.
Lin YK, Lin YP, Lee JT, Lin CS, Wu TJ, Tsai KZ, et al. Sex-specific association of hyperuricemia with
cardiometabolic abnormalities in a military cohort: The CHIEF study. Medicine (Baltimore). 2020; 99
(12):e19535. Epub 2020/03/21. https://doi.org/10.1097/md.0000000000019535 PMID: 32195957;
PubMed Central PMCID: PMC7220045.
Lin JW, Tsai KZ, Chen KW, Su FY, Li YH, Lin YP, et al. Sex-Specific Association Between Serum Uric
Acid and Elevated Alanine Aminotransferase in a Military Cohort: The CHIEF Study. Endocr Metab
Immune Disord Drug Targets. 2019; 19(3):333–40. Epub 2018/12/01. https://doi.org/10.2174/
1871530319666181129163802 PMID: 30499423.
28. Rodrigues SL, Baldo MP, Capingana P, Magalhães P, Dantas EM, Molina Mdel C, et al. Gender distri-
bution of serum uric acid and cardiovascular risk factors: population based study. Arq Bras Cardiol.
2012; 98(1):13–21. Epub 2011/12/08. https://doi.org/10.1590/s0066-782x2011005000116 PMID:
22146989.
29. Kamei K, Konta T, Ichikawa K, Sato H, Suzuki N, Kabasawa A, et al. Serum uric acid levels and mortal-
ity in the Japanese population: the Yamagata (Takahata) study. Clin Exp Nephrol. 2016; 20(6):904–9.
Epub 2016/01/19. https://doi.org/10.1007/s10157-016-1228-1 PMID: 26779905.
30. Anton FM, Garcia Puig J, Ramos T, Gonzalez P, Ordas J. Sex differences in uric acid metabolism in
adults: evidence for a lack of influence of estradiol-17 beta (E2) on the renal handling of urate. Metabo-
lism: clinical and experimental. 1986; 35(4):343–8. Epub 1986/04/01. https://doi.org/10.1016/0026-
0495(86)90152-6 PMID: 3959904.
31. Chen C, Lu JM, Yao Q. Hyperuricemia-Related Diseases and Xanthine Oxidoreductase (XOR) Inhibi-
tors: An Overview. Medical science monitor: international medical journal of experimental and clinical
research. 2016; 22:2501–12. Epub 2016/07/18. https://doi.org/10.12659/msm.899852 PMID:
27423335; PubMed Central PMCID: PMC4961276.
32. Glantzounis GK, Tsimoyiannis EC, Kappas AM, Galaris DA. Uric acid and oxidative stress. Current
pharmaceutical design. 2005; 11(32):4145–51. Epub 2005/12/27. https://doi.org/10.2174/
138161205774913255 PMID: 16375736.
33. Ciencewicki J, Trivedi S, Kleeberger SR. Oxidants and the pathogenesis of lung diseases. The Journal
of allergy and clinical immunology. 2008; 122(3):456–68; quiz 69–70. Epub 2008/09/09. https://doi.org/
10.1016/j.jaci.2008.08.004 PMID: 18774381; PubMed Central PMCID: PMC2693323.
34.
van der Vliet A, O’Neill CA, Cross CE, Koostra JM, Volz WG, Halliwell B, et al. Determination of low-
molecular-mass antioxidant concentrations in human respiratory tract lining fluids. The American jour-
nal of physiology. 1999; 276(2):L289–96. Epub 1999/02/10. https://doi.org/10.1152/ajplung.1999.276.
2.L289 PMID: 9950891.
35. Shaheen SO. Uric acid, lung function and COPD: a causal link is unlikely. Thorax. 2018; 73(8):697–8.
Epub 2018/03/25. https://doi.org/10.1136/thoraxjnl-2017-211230 PMID: 29572272.
36. So A, Thorens B. Uric acid transport and disease. The Journal of clinical investigation. 2010; 120
(6):1791–9. Epub 2010/06/03. https://doi.org/10.1172/JCI42344 PMID: 20516647; PubMed Central
PMCID: PMC2877959.
37. Shaheen SO. Antioxidants and respiratory disease: the uric acid paradox. Thorax. 2014; 69(11):978–9.
Epub 2014/08/15. https://doi.org/10.1136/thoraxjnl-2014-205751 PMID: 25122087.
38. Elsayed NM, Nakashima JM, Postlethwait EM. Measurement of uric acid as a marker of oxygen tension
in the lung. Arch Biochem Biophys. 1993; 302(1):228–32. Epub 1993/04/01. https://doi.org/10.1006/
abbi.1993.1204 PMID: 8470901.
PLOS ONE | https://doi.org/10.1371/journal.pone.0240987 October 22, 2020
16 / 16
PLOS ONE |
10.1371_journal.pone.0237815 | RESEARCH ARTICLE
Mothers’ perceptions of universal newborn
hearing screening in transitional Albania
Birkena QirjaziID
1,2☯, Ervin Toc¸ iID
1,2☯*, Eduard Tushe3‡, Genc Burazeri1‡, Enver Roshi1‡
1 Faculty of Medicine, University of Medicine, Tirana, Tirana, Albania, 2 Institute of Public Health, Tirana,
Albania, 3 University Maternity Hospital “Koc¸o Gliozheni”, Tirana, Albania
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
☯ These authors contributed equally to this work.
‡ These authors also contributed equally to this work.
* ervintoci@yahoo.com
Abstract
OPEN ACCESS
Citation: Qirjazi B, Toc¸i E, Tushe E, Burazeri G,
Roshi E (2020) Mothers’ perceptions of universal
newborn hearing screening in transitional Albania.
PLoS ONE 15(8): e0237815. https://doi.org/
10.1371/journal.pone.0237815
Editor: Bolajoko O. Olusanya, Center for Healthy
Start Initiative, NIGERIA
Received: March 25, 2020
Accepted: August 3, 2020
Published: August 21, 2020
Peer Review History: PLOS recognizes the
benefits of transparency in the peer review
process; therefore, we enable the publication of
all of the content of peer review and author
responses alongside final, published articles. The
editorial history of this article is available here:
https://doi.org/10.1371/journal.pone.0237815
Copyright: © 2020 Qirjazi et al. This is an open
access article distributed under the terms of the
Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: All relevant data are
within the manuscript and its Supporting
Information file.
Funding: This project has received funding from
the European Union’s Horizon 2020 research and
The universal newborn hearing screening (UNHS) component of the multi-center
EUSCREEN project is being piloted in Albania since January 1st 2018. The aim of this study
was to explore mothers’ perceptions about various elements of UNHS in Albania. A cross-
sectional study was carried out in the three sites of UNHS in Albania, namely in Tirana,
Kuke¨ s and Pogradec during May-June 2019. During this period 512 consecutively
approached mothers giving birth to included maternity hospitals were interviewed face-to-
face about different aspects of UNHS. Basic socio-demographic and socioeconomic infor-
mation was also collected. Mean age of participating mothers was 28.6 years ± 5.5 years.
The overwhelming majority (93%) of mothers knew what their baby was being tested for,
33% were aware that hearing screening was offered in maternity hospital, 94% were very
satisfied/satisfied with UNHS and about 62% were very stressed/stressed waiting for
screening results, with significant sociodemographic and socioeconomic differences. The
main information source about UNHS was screening staff in the maternity hospitals where
mothers gave birth, reported in 67% of cases. All mothers (100%) agreed on the importance
of early detection of newborn hearing problems, all mothers were willing to be informed
early if their newborn baby had a hearing problem and all mothers were willing to contribute
financially for testing the hearing of their newborn baby. These findings should guide infor-
mation and education campaigns about UNHS in Albania. The public willingness to finan-
cially support neonatal hearing testing should be considered as an opportunity to achieve
universal newborn hearing screening in the country.
Introduction
Screening newborns for hearing impairment or hearing loss is important as this is one of the
most frequent congenital conditions, affecting about 1–6 newborns per 1000 live births [1,2]
with much higher rates among high-risk infants admitted to neonatal intensive care unit
(NICU) [3]. In addition, about half of hearing impaired newborns have no known risk factors
[1] highlighting the necessity to screen all newborns. Furthermore, early detection and
PLOS ONE | https://doi.org/10.1371/journal.pone.0237815 August 21, 2020
1 / 12
PLOS ONE innovation program under grant agreement No
733352. This study sponsor had no role in the
study design, or the collection, analysis and
interpretation of the data or in the writing of the
report or in the decision to submit the article for
publication.
Competing interests: The authors have declared
that no competing interests exist
Perceptions of universal newborn hearing screening in Albania
treatment of hearing loss is important for ensuring the normal development of speech and lan-
guage and avoiding negative effects on cognitive, academic and social-emotional skills of
affected children [4].
Even though universal newborn hearing screening (UNHS) is mainly implemented in
developed countries [5], the practice is expanding in developing countries as well [6]. The
main challenge with UNHS in developed countries is the low rate of screening follow-up. For
example, in USA in 2012 about one third of newborns that did not pass the first hearing
screening test failed to receive timely follow-up [7]. On the other hand, implementation of
UNHS in developing countries faces challenges of different natures, including limited cover-
age, high unaffordable cost, lack of health personnel, poor infrastructure, poor access to rural
areas, socio-economic differences, etc. [8].
Besides these factors, people’s perceptions of health services constitutes an important deter-
minant of utilization of available health services and a key element for ensuring that health
interventions and services are successful [9]. Various factors contribute to shape one’s percep-
tions about health services being offered in a certain area including culture, traditions and dif-
ferent inherent characteristics of the health system/facility, the quality of interactions with the
health personnel, etc. [9] with health staff behavior potentially moderating the effect of health
services on patient satisfaction level [10].
There is a direct link between perception of a health service and users’ satisfaction. Research
shows that the perception about the quality of a health service is strongly and directly linked
with the satisfaction users get from that service and the future use of such services [11].
In this context, parents’ perception about newborn hearing screening could play an impor-
tant role in improving screening program participation rates, decrease of the refusal rates and
ensure that more newborns will be tested as well [12].
In Albania, a small country in South East Europe, the EUSCREEN project, universal new-
born hearing screening component, started to be implemented on January 1st 2018 on three
country sites and namely in Tirana, the capital city of Albania, Kuke¨s and Pogradec municipal-
ities [13]. Parents’ opinion on certain aspects of hearing screening is important and the
"acceptability" of the screening instrument by the target population is one of the criteria to be
taken into consideration when applying a particular screening procedure to a community [14].
Therefore, in the context of the EUSCREEN project, a survey was carried out in order to shed
light on mothers’ perceptions of various aspects of universal newborn hearing screening in
Albania.
Methodology
Type of study
This is a cross-sectional study conducted during May-June 2019.
Population study and sampling
The target population of the study comprised the mothers of newborn babies in the maternity
hospitals where newborn hearing screening is being implemented in Albania, and more specif-
ically in Tirana, Kuke¨s and Pogradec.
Since there are approximately a total of 11500 births each year in these 3 districts (about
80% in Tirana) and due to the inability to include the entire birth cohort in the study over a
year, we decided to draw a sample of mothers who gave birth to infants in these districts’
maternity hospitals. WINPEPI statistical program was used to calculate the sample size based
on the following parameters: alpha error = 5%, study power = 80%, assumed proportion of
mothers in favor of hearing screening = 50% (this level enables maximization of sample size);
PLOS ONE | https://doi.org/10.1371/journal.pone.0237815 August 21, 2020
2 / 12
PLOS ONE Perceptions of universal newborn hearing screening in Albania
acceptable difference = 4.3%; source population = 11500. The application of these parameters
generated a sample size of 497 individuals.
In total, 512 mothers were interviewed (200 at Koco Gliozheni Maternity Hospital in
Tirana, 200 at Queen Geraldine Maternity Hospital in Tirana as well, 57 at Kuke¨s Maternity
Hospital and 55 at Pogradec Maternity Hospital). It should be noted that among the 400 births
included in the two maternity hospitals in Tirana, in 76 of cases the mothers did not reside in
Tirana but in other districts, as many future mothers choose to give birth in one of the Tirana
maternities. Mothers who gave birth to babies in Tirana but did not reside in Tirana lived in a
large variety of cities and other areas of Albania, from south to north.
Data collection and data collection instrument
In each of the maternities included in the study, the survey instrument (questionnaire) was
applied consecutively to the mothers who gave birth to one or more living children in these
maternity hospitals during May-June 2019, until the preset number of questionnaires (at least)
was reached.
Information on mothers’ opinions on the acceptability of newborn hearing screening was
collected through a face-to-face administered questionnaire.
The questionnaire contained key socio-demographic data of mothers, such as: residence
(urban vs. rural); age, marital status, education level; employment status, social class, economic
situation, and overall health status.
The next section of the questionnaire contained specific questions regarding the acceptabil-
ity of newborn hearing screening. Questions covered issues such as: being aware that the baby
is being tested for hearing; being informed on the fact that screening is provided at this mater-
nity hospital and the sources of this information; mothers’ impression of hearing screening
process (on a 1–5 scale where 1 means "very satisfactory" and 5 "not at all satisfactory"); the
level of stress or anxiety they experienced while waiting for the hearing screening response (on
a 1–5 scale where 1 means "extremely" and 5 means "not at all"); mothers’ preference to be
informed as soon as their child has hearing problems (yes vs. no); importance of newborn
hearing screening to parents (yes vs. no); willingness to pay themselves for screening their
newborn baby (yes vs. no).
The questionnaire used for this survey has been built from the local experts (the authors of
this paper), based on browsing of the appropriate literature listed in the references and picking
up those items considered as relevant for the study objectives. Therefore, the questionnaire has
face validity, construct validity and content validity.
Ethical issues
All mothers included in this study were approached by EUSCREEN staff screening newborns’
hearing at respective maternity hospitals.
In order to be part of newborn hearing screening, each mother was required to give a writ-
ten informed consent. If such consent was not retrieved, then the newborn was not screened,
the mother’s interviewing was not done and this particular baby was not part of UNHS
anymore.
All mothers accepting to screen their baby were required to give a formal oral consent for
participating in the actual study.
All mothers agreed to participate after being thoroughly informed about the survey.
EUSCREEN project in Albania has been approved by the project counterpart in Albania,
i.e. the University of Medicine, and the Ministry of Health and Social Protection of Albania.
PLOS ONE | https://doi.org/10.1371/journal.pone.0237815 August 21, 2020
3 / 12
PLOS ONE Perceptions of universal newborn hearing screening in Albania
Data analysis
Absolute numbers and the corresponding percentages were reported, in order to describe the
data. Chi-square test was used to detect any statistically significant difference when comparing
categorical variables.
All data analyzes were performed using the Statistical Package for Social Sciences (SPSS)
statistical software, version 19.
Results
General information
A total of 512 mothers who had a live birth in one of the four selected maternity hospitals
participated in the survey. The maternal mean age was 28.6 years ± 5.5 years, ranging from a
minimum of 16 years (1 case) to a maximum of 47 years (1 case). Table 1 shows the general
characteristics of participating mothers by birth survey site. More than one fourth (27.3%)
of mothers were between 16 and 25 years old, about one quarter (24.4%) had secondary edu-
cation or less, 56% resided in urban areas, 98% were married at the time of the survey, about
half (50.2%) were employed, 91.2% declared to belong to middle social class, 80.9% had
average economic status and 92.4% declared to have very good or good general health. Sig-
nificantly higher proportions of mothers in Tirana had university education, resided in
urban areas and had very good or good economic status, compared to mothers from other
survey sites (Table 1). On the other hand, the proportion of employment was significantly
higher among Pogradec mothers whereas the proportion of mothers with very good or good
health status was higher among Kuke¨s and Pogradec mothers compared to Tirana (Table 1).
There were no significant age, marital status and social class differences by survey site
(Table 1).
Mother’s perceptions about various aspects of newborn hearing screening
Table 2 shows mothers’ perceptions about different aspects of newborn hearing screening.
More than 9 in 10 mothers (93%) knew what their baby was being tested for, only one third
(33%) were informed beforehand that newborn hearing screening was being offered at the
respective maternity hospital where they gave birth, 94% were very satisfied or satisfied with
newborn hearing screening and about 62% thought that waiting for the hearing screening
result was very stressful or stressful (Table 2).
Significantly higher proportions of highly educated, employed, high social class, very good
or good economic or health status mothers knew what their newborn baby was being tested
for compared to their respective counterparts (Table 2).
Significantly higher proportions of mothers in Kuke¨s (70.2%) and Pogradec (87.3%) were
informed that hearing screening was offered in the maternity hospital where they gave birth
compared to Tirana mothers (20.3%) and such proportion was higher among rural, high
school and university, low and middle social class mothers and those with very good/good
self-reported health status compared to their respective counterparts.
Significantly higher proportions of mothers in Pogradec and Kuke¨s, and those belonging to
higher social class and economic status were very satisfied/satisfied with the newborn hearing
screening.
Lastly, waiting for the results of newborn hearing screening was very stressful/stressful for
significantly higher proportions of Kuke¨s mothers, those belonging to high social class and
those experiencing poor/very poor health status (Table 2).
PLOS ONE | https://doi.org/10.1371/journal.pone.0237815 August 21, 2020
4 / 12
PLOS ONE Table 1. General characteristics of participating mothers.
Perceptions of universal newborn hearing screening in Albania
Variable
Age-group
16–25 years
26–30 years
>30 year
Education level
Secondary or les
High school
University
Residence
Urban
Rural
Marital status
Single ††
Married
Employment status
Employed
Not employed
Student
Social class
Low
Middle
High
Economic status
Very good, good
Average
Poor, very poor
General health status
Very good, goo
Average
Poor, very poor
Total N = 512
Tirana n = 400
Survey site
Kuke¨s n = 57
Pogradec n = 55
140 (27.3)�
193 (37.7)
179 (35.0)
125 (24.4)
203 (39.6)
184 (35.9)
288 (56.3)
224 (43.8)
10 (2.0
502 (98.0)
257 (50.2)
245 (47.9)
10 (2.0)
20 (3.9)
467 (91.2)
25 (4.9)
82 (16.0)
414 (80.9)
16 (3.1)
473 (92.4)
37 (7.2)
2 (0.4)
100 (25.0)
149 (37.3)
151 (37.8)
105 (26.3)
131 (32.8)
164 (41.0)
256 (64.0)
144 (36.0)
6 (1.5)
394 (98.5)
202 (50.5)
190 (47.5)
8 (2.0)
18 (4.5)
358 (89.5)
24 (6.0)
81 (20.3)
307 (76.8)
12 (3.0)
361 (90.3)
37 (9.3)
2 (0.5)
21 (36.8)
21 (36.8)
13 (22.8)
11 (19.3)
36 (63.2)
10 (17.5)
15 (26.3)
42 (73.7)
2 (3.6)
55 (96.4)
20 (35.1)
37 (64.9)
0 (0.0)
1 (1.8)
55 (96.5)
1 (1.8)
1 (1.8)
53 (93.0)
3 (5.3)
57 (100.0)
0 (0.0)
0 (0.0)
19 (34.5)
21 (38.2)
15 (27.3)
9 (16.4)
36 (65.5)
10 (18.2)
17 (30.9)
38 (69.1)
2 (3.6)
53 (96.4)
35 (63.6)
18 (32.7)
2 (3.6)
1 (1.8)
54 (98.2)
0 (0.0)
0 (0.0)
54 (98.2)
1 (1.8)
55 (100.0)
0 (0.0)
0 (0.0)
P-value
0.088 †
<0.001
<0.001
0.375
0.013
0.131
<0.001
0.019
� Absolute number and column percentage.
† P-value according to chi-square test.
†† Including divorced and widowed.
https://doi.org/10.1371/journal.pone.0237815.t001
Source of information about newborn hearing screening
Mothers reporting to have been informed that newborn hearing screening is offered in the
maternity where they gave birth (n = 169 or 33% of participating mothers, see Table 2) were
asked to mention the source of such information (Table 3). The majority of informed mothers
(67.3%) were informed that newborn hearing screening was offered at this maternity by the
medical staff at the time they were admitted to the maternity hospital, followed by information
from family members, relatives or friends (in 8.6% of cases), local television programs talking
about newborn hearing screening (4.9%), Internet (3.7%), posters in maternity hospital prem-
ises (3.7%) whereas other sources of information were less frequent.
PLOS ONE | https://doi.org/10.1371/journal.pone.0237815 August 21, 2020
5 / 12
PLOS ONE Table 2. Mothers’ perception about various aspects of newborn hearing screening.
Perceptions of universal newborn hearing screening in Albania
Knowing what the newborn is
being tested for, n (%)
Informed that hearing screening is
offered in this maternity, n (%)
Very satisfied or satisfied with
newborn hearing screening, n (%)
Waiting for screening result is
very stressful or stressful, n (%)
476 (93.0)
169 (33.0)
482 (94.1)
315 (61.5)
NS
NS
���
NS
NS
�
���
���
�
369 (92.3)
57 (100.0)
50 (90.9)
125 (89.3)
182 (94.3)
169 (94.4)
101 (80.8)
192 (94.6)
183 (99.5)
269 (93.4)
207 (92.4)
8 (80.0)
468 (93.2)
246 (95.7)
220 (89.8)
10 (100.0)
11 (55.5)
440 (94.2)
25 (100.0)
79 (96.3)
390 (94.2)
7 (43.8)
439 (92.8)
36 (97.3)
1 (50.0)
���
NS
��
�
NS
NS
�
NS
��
81 (20.3)
40 (70.2)
48 (87.3)
47 (33.6)
72 (37.3)
50 (27.9)
31 (24.8)
86 (42.4)
52 (28.3)
79 (27.4)
90 (40.2)
5 (50.0)
164 (32.7)
88 (34.2)
78 (31.8)
3 (30.0)
7 (35.0)
160 (34.3)
2 (8.0)
19 (23.2)
146 (35.3)
4 (25.0)
166 (35.1)
3 (8.1)
0 (0.0)
�
NS
NS
NS
NS
NS
��
���
NS
371 (92.8)
56 (98.2)
55 (100.0)
127 (90.7)
185 (95.9)
170 (95.0)
114 (91.2)
194 (95.6)
174 (94.6)
267 (92.7)
215 (96.0)
9 (90.0)
473 (94.2)
241 (93.8)
231 (94.3)
10 (100.0)
15 (75.0)
442 (94.6)
25 (100.0)
77 (93.9)
395 (95.4)
10 (62.5)
444 (93.9)
36 (97.3)
2 (100.0)
���
NS
NS
NS
NS
NS
��
NS
��
235 (58.8)
48 (84.2)
32 (58.2)
95 (67.9)
118 (61.1)
102 (57.0)
74 (59.2)
134 (66.0)
107 (58.2)
172 (59.7)
143 (63.8)
6 (60.0)
309 (61.6)
159 (61.9)
152 (62.0)
4 (40.0)
11 (55.0)
280 (60.0)
24 (96.0)
45 (54.9)
259 (62.6)
11 (68.8)
282 (59.6)
31 (83.8)
2 (100.0)
Total
Survey site
Tirana
Kuke¨s
Pogradec
Age-group
16–25 years
26–30 years
>30 year
Education level
Secondary or les
High school
University
Residence
Urban
Rural
Marital status
Single ††
Married
Employment
status
Employed
Not employed
Student
Social class
Low
Middle
High
Economic status
Very good, good
Average
Poor, very poor
General health
status
Very good, goo
Average
Poor, very poor
NS–not significant.
� P<0.05
�� P<0.01
��� P<0.001 (P-value according to chi-square test).
†† Including divorced and widowed.
https://doi.org/10.1371/journal.pone.0237815.t002
Other aspects related to the newborn hearing screening
Participating mothers were also asked the following questions: "If your child has hearing prob-
lems, would you prefer to know this early?"; "Do you think it’s important for all babies to be
PLOS ONE | https://doi.org/10.1371/journal.pone.0237815 August 21, 2020
6 / 12
PLOS ONE Perceptions of universal newborn hearing screening in Albania
Table 3. Source of information among mothers who know that newborn hearing screening is provided at the
maternity hospital where they gave birth.
Information source
Various relatives or friends
Internet
Work colleagues who had previously giving birth in this maternity hospital
From previous birth
At the women’s consultory
In the hospital when I was hospitalized (from staff working here)
Poster placed in maternity hospital premises
Relatives who work at the maternity hospital
Relatives undergoing hearing test in maternity
Relatives previously giving birth
Television (in general)
TV at the maternity waiting hall
Total
Absolute number
Percentage
14
6
1
2
5
109
6
3
2
8
3
3
162�
8.6
3.7
0.6
1.2
3.1
67.3
3.7
1.9
1.2
4.9
1.9
1.9
100.0
� Any discrepancy with the total number (in this case n = 169) is due to missing information.
https://doi.org/10.1371/journal.pone.0237815.t003
tested for hearing in the early days of life?" and "Would you be willing to pay for a hearing
test?" Analysis of their responses revealed that all mothers (100%) included in the study
answered “Yes” to each of these questions (Table 4).
Discussion
This study generated novel and interesting information that sheds light on various aspects
related to the acceptability of universal newborn hearing screening program, an activity being
piloted in three regions of Albania, in the framework of the EUSCREEN project.
The overwhelming majority (93%) of mothers who have given birth to children in the
maternities included in the EUSCREEN project in Albania were aware of what their baby was
being tested for. This suggests that the staff involved in hearing screening in all 4 participating
maternity hospitals has adequately and appropriately explained to mothers the details of hear-
ing test.
Knowledge of what the baby was being screened for was significantly and positively related
to educational level, social class, economic status and overall health status. This finding has
important implications for the future, suggesting that hearing screening explanation and edu-
cation may need to be adapted to various educational, social, economic, and health groups for
greater effectiveness.
Two-thirds of mothers who gave birth to babies in the maternities included in the
EUSCREEN project in Albania were not informed that hearing screening was provided at
these maternities. This finding is similar to previous research in developing countries. For
Table 4. Opinions related to other aspects of newborn hearing screening.
Question
If your child has hearing problems, would you prefer to know this early?—Yes
Do you think it’s important for all babies to be tested for hearing in the early days
of life?—Yes
Would you be willing to pay for a hearing screening?—Yes
https://doi.org/10.1371/journal.pone.0237815.t004
Absolute
number
Percentage
512
512
512
100.0
100.0
100.0
PLOS ONE | https://doi.org/10.1371/journal.pone.0237815 August 21, 2020
7 / 12
PLOS ONE Perceptions of universal newborn hearing screening in Albania
example a study from Nigeria, where there is no national universal newborn hearing screening
but neonatal hearing screening is offered in the frame of research projects, suggested that
about 63% of mothers were not aware of neonatal hearing screening as well [15]. This finding
suggests that greater work needs to be done so that prospective mothers and the general public
are informed about the screening of newborn hearing and the places where this testing is
offered. Research shows that better informed mothers have more positive attitudes toward uni-
versal newborn hearing screening programs [16].
A significantly higher percentage of babies born to mothers residing in Kuke¨s and Pogradec,
in rural areas, mothers with secondary or tertiary education, mothers with low or middle social
status and mothers with very good/good health status were informed that newborn hearing
screening was provided in the maternity hospital compared to their respective counterparts.
These findings suggest that screening information efforts should target less informed groups.
On the other hand, about 80% of mothers in Tirana were not informed that hearing screen-
ing was provided in the two Tirana maternity hospitals. This result seems to be partly influ-
enced by the fact that about 20% of births here occurred among mothers residing in other
cities not being specifically covered by awareness campaigns for newborn hearing screening.
Another explanation can be found in the much more dynamic, active and very heterogeneous
population in Tirana compared to Kuke¨s and Pogradec, where the latter are relatively small
communities and the information is disseminated faster and more effectively than in Tirana.
However, the finding regarding Tirana suggests that information campaigns in big cities
should be different from those employed in small cities in terms of the aggressiveness of the
coverage and the ways in which information is communicated.
About 67% of mothers informed about newborn hearing screening had received this infor-
mation at the maternity ward after giving birth, by the relevant medical staff there, just as it
occurs in most countries [16]. In Nigeria mothers learned about the availability of newborn
hearing screening from antenatal clinics (28%), media (11%), family members or friends
(17%), Internet (44%), etc. [15]. In Hong Kong about 69% of mothers learned about neonatal
hearing screening at the postnatal ward premises and other information sources included pre-
vious screening of older children (16.9%), friends or relatives (5.9%) etc. [16]. It seems that in
Albania, for the time being, the information sources other than the maternity hospitals where
UNHS is taking place play a rather small role in informing the population about this screening
program. On the other hand, international research suggests that most mothers would like to
be informed about newborn hearing screening well in advance and probably not in maternity
hospital premises [17]. The implication for the future is that there is need to improve and
expand UNHS awareness campaigns in Albania to cover also mass media, social media and
the Internet.
The overwhelming majority (94.1%) of mothers who gave birth to children in the materni-
ties included in the EUSCREEN project in Albania were either very satisfied or satisfied with
the hearing screening provided to their babies. This data suggests that the staff involved in
hearing screening in all 4 maternities in Albania has performed the procedure properly, gain-
ing the confidence of mothers. Parents’ positive attitudes towards UNHS has also been
reported in literature [16, 18]. Also, there was a positive relationship between satisfaction with
newborn hearing screening with social class and economic status, in accordance with previous
research [15].
On the other hand, a smaller percentage of mothers in Tirana are more satisfied with hear-
ing screening compared to mothers who gave birth in Kuke¨s or Pogradec. The implication is
that measures should also be taken in Tirana to increase the percentage of mothers who are
very satisfied/satisfied with the screening of their newborn baby. In this regard, efforts should
be focused on mothers with low social status and disadvantaged economic status.
PLOS ONE | https://doi.org/10.1371/journal.pone.0237815 August 21, 2020
8 / 12
PLOS ONE Perceptions of universal newborn hearing screening in Albania
In our study about two-thirds of mothers (61.5%) were very stressful or stressful waiting for
the hearing screening result. Being anxious about the screening results is also reported by pre-
vious research and it seems to be mainly associated with the inconclusive message (the result
of the screening test is not definitive and there is need for further testing) at the end of each
screening stage [19]. In addition, the role of screeners seems to be fundamental for reassuring
parents about the need to be screened again [19]. The implications are that a better job should
be done by the relevant hearing screening staff for supporting mothers and informing them
about the screening procedure, stages and interpretation of results and intervention
opportunities.
All mothers included in the current study (100%) stated that they would prefer to know
early if their child has hearing problems. This finding best supports the value and importance
of hearing screening in newborns as this procedure provides the fastest possible responses
(meeting future mothers’ expectations) and ensures early detection of the problem, greatly
increasing the chances of early treatment, and potential full rehabilitation of infants affected by
these conditions. The willingness to know early if own child has a hearing problem is sup-
ported by international research [16, 20]. Knowing early does not take away the grief that
comes with the unfavorable diagnosis but it could give parents the needed time to somehow
getting used to this situation and it can give them a sense of being able to take action quickly
[20].
The importance of early detection and early treatment is also supported by the opinions of
the mothers included in the study, where all mothers (100%) stated that it is very important for
all babies to be tested for hearing in the early days of life, similar to international research
results showing a very high level of parental agreement with regard to implementation and
importance of newborn hearing screening [15, 18]. This is an additional argument in favor of
universal hearing screening of all newborns, as this procedure is not only of medical and gen-
eral population health benefit but it also enables the achievement of an aspiration, expectation
or ideal of future mothers as an integral part of the universal right to good health.
As literature review shows, there are evidence-based arguments to support the implementa-
tion of UNHS [5]. However, early identification needs to be followed by timely fitting, amplifi-
cation and family support for the affected babies; this requires effective follow-up of babies
with hearing loss being detected by UNHS and often this represents the real challenge of the
UNHS [7]. Research shows that on the one hand parents do support early identification of
hearing loss but on the other hand they are rightly concerned about the effectiveness of follow-
up and appropriate treatment for the affected babies [21]. Indeed, early identification [within
the first 6 months of life] of hearing loss or impairment is accompanied by many different
challenges for the parents, including timing, early amplification issues, information and sup-
port, daily management issues, etc. [21], issues that call for the development of services that
appropriately address early identification of hearing impairment [21]. In addition, it should be
kept in mind that most mothers have rather a limited understanding of their baby’s hearing
development milestones [16], which means that passing the newborn screening test is not a
full assurance that future hearing problems will not occur and therefore mothers have to be
able to continually monitor the baby’s growth and detect any warning signs that something is
not right with baby’s hearing [16]. The relevant implication is that prospective mothers need
to be better educated about child hearing development milestones as well.
All mothers included in the study (100%) stated that they would be willing to pay for their
child’s hearing medical control. This finding assures us once again about the public perception
that universal newborn hearing screening is important, and that is why all possible efforts
should be taken to achieve full coverage of all babies born in Albania. The finding specifically
suggests that such activity is so positively perceived and highly appreciated by mothers as there
PLOS ONE | https://doi.org/10.1371/journal.pone.0237815 August 21, 2020
9 / 12
PLOS ONE Perceptions of universal newborn hearing screening in Albania
may be room for discussions about potential parental involvement in universal hearing screen-
ing funding schemes (in the event of an absolute inability to be 100% covered by public funds),
economic contribution which undoubtedly needs to be tailored to the socio-economic level
and other factors and based on the results of extensive consultations with all stakeholders.
Study limitations
Our study has several limitations. The interviewing of consecutive mothers showing up to give
birth in the selected maternity hospitals does not exclude the selection bias. However, we cov-
ered all the UNHS pilot sites in Albania and also selected the mothers proportional to size of
respective maternities, in order to reduce the selection bias. On the other hand, the study sam-
ple is rather small and does not allow the unconditional generalization of the results. Finally,
the cross-sectional nature of the study does not allow us to draw any causal conclusions.
The strong point of the present study is that it offers for the first time a glimpse on mothers’
perception about newborn hearing screening, a largely under researched topic in Albania.
Conclusions
In Albania mothers in general had positive attitudes toward universal newborn hearing screen-
ing program. The overwhelming majority of mothers giving birth to maternity hospitals where
UNHS is being implemented were aware what their newborn is tested for, meaning the screen-
ing staff has done its job quite well. On the other hand, about two-thirds of mothers didn’t
know the hearing screening was offered in these maternity hospitals with considerable differ-
ences between the capital and two other regions, implying further targeted information efforts.
About two-thirds of mothers were stressed or anxious waiting for the testing results. All moth-
ers were in favor of early detection of hearing problems among newborn babies, they wished
to know early if their baby has a hearing problem and they are willing to contribute financially
for testing the hearing of their newborn babies. The various significant socio-demographic dif-
ferences noticed regarding mother’s awareness, information and satisfaction rates about vari-
ous UNHS elements and perceived stress of screening results should guide information and
education campaigns. The public willingness to financially contribute in order to screen the
hearing of newborns might be an opportunity for scaling-up the UNHS in the country.
Supporting information
S1 Database.
(SAV)
Author Contributions
Conceptualization: Birkena Qirjazi, Ervin Toc¸i.
Data curation: Birkena Qirjazi, Ervin Toc¸i.
Formal analysis: Birkena Qirjazi, Ervin Toc¸i.
Investigation: Birkena Qirjazi, Ervin Toc¸i.
Methodology: Genc Burazeri, Enver Roshi.
Writing – original draft: Birkena Qirjazi, Ervin Toc¸i.
Writing – review & editing: Birkena Qirjazi, Ervin Toc¸i, Eduard Tushe, Genc Burazeri, Enver
Roshi.
PLOS ONE | https://doi.org/10.1371/journal.pone.0237815 August 21, 2020
10 / 12
PLOS ONE Perceptions of universal newborn hearing screening in Albania
References
1. Nelson HD, Bougatsos C, Nygren P. 2001 US Preventive Services Task Force. Universal newborn
hearing screening: systematic review to update the 2001 US Preventive Services Task Force Recom-
mendation. Pediatrics. 2008; 122(1):e266–76. https://doi.org/10.1542/peds.2007-1422 PMID:
18595973
2. Kushalnagar P, Mathur G, Moreland CJ, Napoli DJ, Osterling W, Padden C, et al. Infants and children
with hearing loss need early language access. J Clin Ethics. 2010 Summer; 21(2):143–54. PMID:
20866021
3. Khairy MA, Abuelhamed WA, Ahmed RS, El Fouly HES, Elhawary IM. Hearing loss among high-risk
newborns admitted to a tertiary Neonatal Intensive Care Unit. J Matern Fetal Neonatal Med. 2018; 31
(13):1756–61. https://doi.org/10.1080/14767058.2017.1326902 PMID: 28482715
4. Vos B, Noll D, Pigeon M, Bagatto M, Fitzpatrick EM. Risk factors for hearing loss in children: a system-
atic literature review and meta-analysis protocol. Syst Rev. 2019; 8(1):172. https://doi.org/10.1186/
s13643-019-1073-x PMID: 31315672
5. Patel H, Feldman M. Universal newborn hearing screening. Paediatr Child Health. 2011; 16(5):301–10.
https://doi.org/10.1093/pch/16.5.301 PMID: 22547950
6. WHO. Newborn and infant hearing screening. Current issues and guiding principles for action. Outcome
of a WHO informal consultation held at who headquarters, Geneva, Switzerland, 09–10 November
200. World Health Organization, 2010.
7. Hunter LL, Meinzen-Derr J, Wiley S, Horvath CL, Kothari R, Wexelblatt S. Influence of the WIC Program
on Loss to Follow-up for Newborn Hearing Screening. Pediatrics. 2016; 138(1). pii: e20154301. https://
doi.org/10.1542/peds.2015-4301 PMID: 27307144
8. Krishnan LA, Donaldson LK. Newborn Hearing Screening in Developing Countries: Understanding the
Challenges and Complexities of Implementation. Perspectives on Global Issues in Communication Sci-
ences and Related Disorders. 2013; 3(2):54–61.
9. Onyeneho NG, Amazigo UV, Njepuome NA, Nwaorgu OC, Okeibunor JC. Perception and utilization of
public health services in Southeast Nigeria: Implication for health care in communities with different
degrees of urbanization. Int J Equity Health. 2016; 15:12. https://doi.org/10.1186/s12939-016-0294-z
PMID: 26791575
10. Manzoor F, Wei L, Hussain A, Asif M, Shah SIA. Patient Satisfaction with Health Care Services; An
Application of Physician’s Behavior as a Moderator. Int J Environ Res Public Health. 2019; 16(18): pii:
E3318. https://doi.org/10.3390/ijerph16183318 PMID: 31505840
11. Aljaberi MA, Juni MH, Al-Maqtari RA, Lye MS, Saeed MA, Al-Dubai SAR, et al. Relationships among
perceived quality of healthcare services, satisfaction and behavioural intentions of international stu-
dents in Kuala Lumpur, Malaysia: a cross-sectional study. BMJ Open. 2018; 8(9):e021180. https://doi.
org/10.1136/bmjopen-2017-021180 PMID: 30209153
12. Pynnonen MA, Handelsman JA, King EF, Singer DC, Davis MM, Lesperance MM. Parent Perception of
Newborn Hearing Screening: Results of a US National Survey. JAMA Otolaryngol Head Neck Surg.
2016; 142(6):538–43. https://doi.org/10.1001/jamaoto.2015.3948 PMID: 26967534
13. Qirjazi B, Toc¸i E, Roshi E. Preliminary steps towards implementation of universal neonatal hearing
screening in Albania. Management in Health, 2019; 23(1):14–16.
14. Dobrow MJ, Hagens V, Chafe R, Sullivan T, Rabeneck L. Consolidated principles for screening based
on a systematic review and consensus process. CMAJ. 2018; 190(14):E422–E429. https://doi.org/10.
1503/cmaj.171154 PMID: 29632037
15.
16.
Jatto ME, Ogunkeyede SA, Adeyemo AA, Adeagbo K, Saiki O. Mothers’ perspectives of newborn hear-
ing screening programme. Ghana Med J. 2018; 52(3):158–62. https://doi.org/10.4314/gmj.v52i3.9
PMID: 30602802
Lam MYY, Wong ECM, Law CW, Lee HHL, McPherson B. Maternal knowledge and attitudes to univer-
sal newborn hearing screening: Reviewing an established program. Int J Pediatr Otorhinolaryngol.
2018; 105:146–53. https://doi.org/10.1016/j.ijporl.2017.12.021 PMID: 29447803
17. Arnold CL, Davis TC, Humiston SG, et al. Infant hearing screening: stakeholder recommendations for
parent-centered communication. Pediatrics. 2006; 117(5 Pt 2):S341–54.
18. Hergils L, Hergils A. Universal neonatal hearing screening—parental attitudes and concern. Br J Audiol.
2000; 34(6):321–7. https://doi.org/10.3109/03005364000000147 PMID: 11201318
19. Young A, Tattersall H. Parents’ of deaf children evaluative accounts of the process and practice of uni-
versal newborn hearing screening. J Deaf Stud Deaf Educ. 2005; 10(2):134–45. https://doi.org/10.
1093/deafed/eni014 PMID: 15778210
PLOS ONE | https://doi.org/10.1371/journal.pone.0237815 August 21, 2020
11 / 12
PLOS ONE Perceptions of universal newborn hearing screening in Albania
20. Young A, Tattersall H. Universal newborn hearing screening and early identification of deafness:
parents’ responses to knowing early and their expectations of child communication development. J
Deaf Stud Deaf Educ. 2007; 12(2):209–20. https://doi.org/10.1093/deafed/enl033 PMID: 17277310
21. McCracken W, Young A, Tattersall H. Universal newborn hearing screening: parental reflections on
very early audiological management. Ear Hear. 2008; 29(1):54–64. https://doi.org/10.1097/AUD.
0b013e31815ed8d0 PMID: 18091098
PLOS ONE | https://doi.org/10.1371/journal.pone.0237815 August 21, 2020
12 / 12
PLOS ONE |
10.1371_journal.pone.0250576 | RESEARCH ARTICLE
Marmoset angiography just by percutaneous
puncture of the caudal ventral artery
Hiroki Ohta1,2☯, Teppei Komatsu3☯, Kanako Muta1, Makoto Koizumi4, Yasuyuki Iguchi3,
Hirotaka James OkanoID
1*
1 Division of Regenerative Medicine, Research Center for Medical Sciences, The Jikei University School of
Medicine, Tokyo, Japan, 2 Division of Vascular Surgery, Department of Surgery, The Jikei University School
of Medicine, Tokyo, Japan, 3 Department of Neurology, The Jikei University School of Medicine, Tokyo,
Japan, 4 Laboratory of Animal Facilities, Research Center for Medical Sciences, The Jikei University School
of Medicine, Tokyo, Japan
☯ These authors contributed equally to this work.
* hjokano@jikei.ac.jp
Abstract
Surgery in humans is continuously evolving and promoted minimally invasive treatment. On
the other hand, despite the importance of the 3Rs principles for experimental animals is well
documented, no reports describe specific methodologies for implementing "refinement" in
practice. Here, we describe a new technique, the "Ohta Method" for caudal arthrocentesis in
the pursuit of the 3Rs for animal experiments and the development of innovative methods
for investigating systemic organ arteries through minimally invasive procedures. This proce-
dure requires only a percutaneous puncture of the caudal artery without any injury to the
limb or body trunk. In addition, it does not cut down the artery, making hemostasis easier
and recovering arterial damage easier. We will show multiple organ artery angiographies in
marmoset for the first time in the world. The principle described in this paper could also be
applied to many other small animals, such as rats. Moreover, using this method, multiple
doses of the drug or cells can be administered to the target organ at the time of therapeutic
intervention, thereby enabling the establishment of more sophisticated and complex thera-
peutic intervention studies as translational research.
Introduction
Surgery in humans is continuously evolving, and advances in endoscopic surgery and robotic
surgery have promoted minimally invasive treatment. Endovascular treatment techniques also
play a significant role in minimally invasive treatment. Many cases in which surgery will not
be tolerated because of advanced age or low cardiorespiratory fitness, and operating is there-
fore judged to be impossible. Endovascular treatment has become safe for such cases [1, 2]. On
the other hand, despite the 3Rs principles for using experimental animals [3], no reports
describe specific methodologies for implementing “refinement” in practice. We believe that
there are several reasons for this gap:
1. Rodents commonly used in the laboratory are much smaller than humans.
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: Ohta H, Komatsu T, Muta K, Koizumi M,
Iguchi Y, Okano HJ (2021) Marmoset angiography
just by percutaneous puncture of the caudal ventral
artery. PLoS ONE 16(4): e0250576. https://doi.org/
10.1371/journal.pone.0250576
Editor: Jinglu Ai, Barrow Neurological Institute,
UNITED STATES
Received: February 25, 2021
Accepted: April 9, 2021
Published: April 28, 2021
Copyright: © 2021 Ohta et al. This is an open
access article distributed under the terms of the
Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: All relevant data are
within the paper and its Supporting information
files.
Funding: This work was supported by MEXT
KAKENHI Grant Number 18K16404 (to Hiroki
Ohta), 17K18130 (to Teppei Komatsu), 15K15044
(to Hirotaka James Okano), The Jikei University
Strategic Prioritizing Research Fund (to Hirotaka
James Okano) The Jikei University Research Fund
(to Teppei Komatsu), The Jikei University
Department of Surgery Division of Vascular
Surgery Fund.
PLOS ONE | https://doi.org/10.1371/journal.pone.0250576 April 28, 2021
1 / 10
PLOS ONE The ultra-minimally invasive technique allows the 3Rs to be achieved in small primates
Competing interests: The authors have declared
that no competing interests exist.
2. Sufficiently small laboratory instruments for use in small laboratory animals have not been
optimized.
3. Rather sophisticated techniques are required to perform minimally invasive procedures on
small laboratory animals.
Animal experiments usually involve a thoracotomy or laparotomy [4] and are highly inva-
sive. Even the most peripheral arterial approach among endovascular approaches that do not
manipulate the trunk is limited by manipulating the carotid and femoral arteries. These
approach vessels are ligated without reconstruction, and a decline in peripheral blood flow
from the approach site cannot be avoided. For example, in a conventional cerebral infarction
model, a large incision is made in the neck, and a nylon thread or something similar is inserted
through the carotid artery. Then, after the procedure is completed, the carotid artery is ligated
[5, 6].
Previous experimental animal models’ design forced intra-arterial approaches from the
extremities or trunk at the expense of peripheral arteries [7]. However, in our approach, we do
not sacrifice any arteries in the extremities or trunk and are able to perform a minimally inva-
sive transarterial approach. That is, we established a way to approach almost all organs only by
puncturing the caudal artery. In addition to this very minimally invasive approach, we aimed
to establish a method that enables establishing small laboratory animal models and therapeutic
interventions that are equally minimally invasive as those in humans by applying endovascular
therapeutic procedures commonly used in humans. To this end, thin wires and microcatheters
must be made and, in some cases, custom-made for the use of high-resolution angiography
devices to refine the technique. No matter what factors are missing, it will not be successful.
Here, we describe a new technique for caudal ventral arteriocentesis to pursue the 3Rs for
laboratory experiments on animals (Replacement, Reduction, and Refinement) through mini-
mally invasive procedures and multiple therapeutic interventions. Because of the ethics of ani-
mal experiments, the establishment of a less invasive experimental method is urgent. In
addition, developing minimally invasive approaches not only practices the principles of 3Rs in
laboratory animals. Multiple doses of the drug or cells can be administered to the target organ
at the time of therapeutic intervention, thereby enabling more sophisticated and complex ther-
apeutic intervention studies.
Materials and methods
Ethical statement
This study was approved by the Institutional Animal Care and Use Committee of the Jikei
University (No. 2015-123C3, 2016–105, 2018–024). All procedures were conducted according
to the Fundamental Guidelines for Proper Conduct of Animal Experiments and Related Activ-
ities in Academic Research Institutions issued by the Japanese Ministry of Education, Culture,
Sports, Science, and Technology.
Animals
Marmosets acquired from Nihon Clea (CLEA Japan, Inc., Gifu, Japan) or self-propagated on
the day of surgery (300~350 g) were used. The marmosets were housed in family cages
(W800mm, D650mm, H1500mm), paired cages (W800mm, D650mm, H720mm), or individ-
ual/intensive care cages (W400mm, D650mm, H720mm). The room temperature was
maintained at 26–30˚C. The marmosets were fed 40 g of pellets (LabDiet, PMI Nutrition Inter-
national LLC, U.S.A.) twice a day supplemented with honey, Vitamin C, and Vitamin D. Envi-
ronmental enrichment such as wooden toys, climbing structures, and swings were provided
PLOS ONE | https://doi.org/10.1371/journal.pone.0250576 April 28, 2021
2 / 10
PLOS ONE The ultra-minimally invasive technique allows the 3Rs to be achieved in small primates
depending on the housing conditions. The animal care staff observed the animals’ health and
well-being daily using criteria such as fecal condition, appetite, hair condition, and movement.
If the staff suspected any problems, they would consult with veterinarians. Each animal was
kept in a pain-free state during surgery by appropriately regulating the inhalational anesthetic
based on the SpO2 and heart rate. Animals were maintained at a body temperature of 36–38˚C
during surgery. After the completion of this experiment, they were not sacrificed, but were
allowed to survive as breeding stock.
Animal anaesthesia
Procedures were performed in the hybrid operating room with digital subtraction angiography
(Artis Zee, Siemens, Germany) in the Jikei University School of Medicine laboratory animal
facilities. Marmosets were anesthetized by injecting 12 mg/kg of alfaxalone into the thigh mus-
cle. After that, a 14-gauge catheter (Terumo, Tokyo, Japan) was visually intubated into the tra-
chea, and controlled breathing was started. Anesthesia was maintained through ventilation
using an animal ventilator and 1%-3% isoflurane. The body temperature was maintained
using a multipanel heater (Vivaria, Osaka, Japan).
Caudal Ventral Artery (CVA) puncture
The microcatheter was first flushed with heparinized physiological saline. The marmoset was
placed in a supine position with a shaved tail, and the tail skin was sterilized with povidone-
iodine. A 22-gauge sheath (Terumo, Tokyo, Japan) was inserted into the CVA. A microcath-
eter for angiography was inserted into the CVA. After confirmation of arterial blood backflow
from the sheath, a wire and microcatheter were inserted as long as no resistance was felt. Upon
resistance, the insertion of the microcatheter was stopped immediately, and the position of the
microcatheter was confirmed by fluoroscopic imaging.
Selective cannulation and angiography
The microcatheter was inserted into the abdominal aorta from the CVA. The wire and micro-
catheter were guided from the abdominal aorta to the center and were raised to the ascending
aorta using a digital subtraction angiographic unit (Artis Zee, Siemens, Germany). Only the
wire was removed while taking care not to change the position of the microcatheter. The con-
trast media (Iohexol, Daiichi Sankyo, Tokyo, Japan) at half concentration was rapidly injected
from the microcatheter, and arch aortography was carried out, in which 0.15 ml contrast
media was injected each time. For subsequent branch cannulation, the contrast images were
overlaid for navigation. This procedure can reduce the volume of contrast media needed. We
used only approximately 0.1–0.15 ml of half-diluted contrast (approximately 0.2 ml of aortog-
raphy). The maximum contrast material dose was up to 3 ml.
At the time of cannulation, wire orientation was controlled by slightly bending the wire tip
and twisting the hand. The bifurcation position and angle were recognized by aortic angiogra-
phy described earlier, and the target blood vessel was cannulated. During cannulation of the
cervical branch, the C-arm has tilted to view the aortic arch at a right angle. The left internal
carotid artery was cannulated, and angiography was performed. For selective cannulation of
the superior mesenteric artery (SMA), abdominal aortography was performed at the level 3–4
vertebral bodies caudal from the diaphragm to confirm the SMA’s location direction bifurca-
tion. The same procedure was carried out in other abdominal visceral branches. If the total
amount of contrast agent exceeded 3 ml, the inspection was terminated at that time. The
sheath was removed to complete the procedure.
PLOS ONE | https://doi.org/10.1371/journal.pone.0250576 April 28, 2021
3 / 10
PLOS ONE The ultra-minimally invasive technique allows the 3Rs to be achieved in small primates
In some cases, the sheath punctured in the tail artery was not removed, and the treatment
was carried out so that it could not be removed but rather served as a route for blood collection
on the following day. Needless aspiration was avoided, following the goal for "refinement" for
the welfare of animals. By injecting embolism material through this procedure after selective
organ cannulation was performed as described, a minimally invasive organ ischemia model
was produced. The microcatheter tip was positioned at the origin of the abdominal visceral
branches. The renal artery (RA) was cannulated selectively based on the previous angiography.
After the RA was cannulated, angiography confirmed the tip of the microcatheter again.
Results
Percutaneous cannulation of the Caudal Ventral Artery (CVA) "Ohta
Method"
The “Ohta Method” is an ultra-minimally invasive technique in which a 22-gauge sheath (ID:
0.6 mm, OD: 0.9 mm, Terumo, Tokyo, Japan) by percutaneous puncture alone without incis-
ing the tail, and a microcatheter or wire is inserted through the sheath and cannulated into ves-
sels of the desired organ using X-ray fluoroscopy. Just by performing percutaneous puncture
of the CVA of common marmosets (Callithrix jacchus) was allowed placement of a 22-gauge
sheath. This sheath, whose inner tube is a puncture needle, is a highly versatile medical instru-
ment when administering an infusion to humans. The sheath was inserted through the ventral
midline, approximately 5 cm from the root of the animal body (Fig 1a), with a sharp angle. As
we have already reported [8], when puncturing the CVA, it was necessary to keep the tail and
the whole body warm. The sheath could be inserted into the CVA after shaving. We showed
these tips as a S1 Fig. Compared with the procedure in the rat, puncturing the CVA in the mar-
moset was difficult because the skin was hard, and the subcutaneous tissue was loose. Although
these data are not presented, the distal approach was also possible after the previous proximal
site was punctured using this puncture method. A 0.016" guidewire (Toray Medical Co., Ltd.,
Tokyo, Japan) was carefully inserted through the sheath inserted into the caudal artery, along
which a small-diameter microcatheter (ID: 0.42 mm, OD: 0.55 mm, Kaneko Code Co., Ltd.,
Tokyo, Japan) could be guided. This wire technique under X-ray fluoroscopy is similar to that
performed in humans to safely advance wires through blood vessels (Fig 1b). However, inser-
tion from the caudal artery root into the abdominal aorta was technically challenging and
needed much training. It was important that, as the tail artery root is easily spastic, the wire
and microcatheter are through there as soon as possible after the insertion. The microcatheters
used are thinner than those commercially available and are custom-made. Since the micro-
catheter is custom-made, it fits the 22-gauge sheath and will not be pushed out by blood flow
or other factors. Therefore, there was no need for fixation.
Small animal angiography
After the wire was slightly advanced from the tip of the microcatheter and guided into the
ascending/abdominal aorta, the wire was removed, and arch/abdominal aortography was per-
formed (Fig 1c and 1d). In arch angiography, coronary arteries could also be imaged simulta-
neously. The amount of half-concentrated contrast medium (Iohexol, Daiichi Sankyo, Tokyo,
Japan) consumed for a single aortography was approximately 0.15 ml, and sufficiently detailed
vascular information was obtained using digital subtraction angiography equipment (Artis
Zee, Siemens, Germany). In these angiographies, the artificial respiration management for
controlling respiration was suitable; the endotracheal intubation was indispensable for this
purpose.
PLOS ONE | https://doi.org/10.1371/journal.pone.0250576 April 28, 2021
4 / 10
PLOS ONE The ultra-minimally invasive technique allows the 3Rs to be achieved in small primates
Fig 1. The most minimally invasive aortic approach. (a) Puncture of caudal ventral artery (CVA) in marmosets. A sheath was inserted through the
ventral midline, approximately 5 cm from the root of the tail of the marmoset, at a sharp angle after tail shaving. (b) Marmoset fluoroscopic images of
the wire inserted carefully through the sheath inserted into the CVA and guided into the abdominal aorta. If the wire is lost in the abdominal visceral
branch, it can be withdrawn immediately because it can be confirmed in real time by the fluoroscope. (c,d) A contrast agent was injected into a
microcatheter indwelling in the arch/abdominal aorta, and arch/abdominal angiography was performed. There were individual differences in the
position and direction of the branch of visceral arteries. Detailed vascular information was obtained in the aortic arch (c) and abdominal aorta (d) of
marmosets.
https://doi.org/10.1371/journal.pone.0250576.g001
Selective organ cannulation
We could cannulate various vessels by endovascular operation using wires selectively and
microcatheters advanced from the caudal artery. For cannulation of the cervical branch, the
cervical branch’s bifurcation was visualized by tilting the C-arm angle to view the aortic arch
at an approximately left anterior oblique (LAO) 20 degrees. Although this procedure is funda-
mental to endovascular treatment, it has been more challenging to perform in laboratory ani-
mals than in humans because of the small vessels and thin microcatheters. Care was always
taken not to exert excessive force on the tip of the wire to avoid vascular damage.
Middle Carotid Artery (MCA). Cerebral angiography was successfully performed by
selective cannulation of the left common carotid artery (Fig 2a). The cerebrovascular
PLOS ONE | https://doi.org/10.1371/journal.pone.0250576 April 28, 2021
5 / 10
PLOS ONE The ultra-minimally invasive technique allows the 3Rs to be achieved in small primates
Fig 2. Marmoset angiography. (a) CCA angiography of a marmoset. (b) Angiographic images of selective cannulation into the left internal carotid artery (ICA)
with frontal view in the marmoset. (c) The SMA was cannulated with wires and microcatheters, and angiography was performed by injecting a contrast agent
through the tip of the microcatheter located at the origin of the SMA in the marmoset. Marmosets showed good visualization of the first bifurcation up to the
periphery of the SMA. (d) Wires and microcatheters were cannulated in the RA based on the results of abdominal aortography, as well as the SMA procedure,
and selective angiography was performed in the marmoset. Detailed vascular information was obtained up to the interlobar RA. Using a thin microcatheter,
more peripheral cannulation and super-selective angiography were successfully performed.
https://doi.org/10.1371/journal.pone.0250576.g002
PLOS ONE | https://doi.org/10.1371/journal.pone.0250576 April 28, 2021
6 / 10
PLOS ONE The ultra-minimally invasive technique allows the 3Rs to be achieved in small primates
structures of marmosets are more similar to those of humans than those of rodents. Angiogra-
phy was performed to confirm the course of the left internal carotid artery (ICA). The selection
of the angle at which the bifurcation of the target vessel is well separated was outstanding
because the skull’s blood vessel runs three-dimensionally in a complicated pattern. The angle
that best demonstrated the course of the ICA in marmosets was LAO 55 to 80 degrees. The
wire and microcatheter were raised from the caudal artery to the aortic arch retrogradely, and
angiography was carried out. After insertion of the cannula into the left common carotid artery
was confirmed, the left ICA was cannulated, and cerebral angiography was carried out (Fig
2b). The angle that best showed the bifurcation of the anterior cerebral artery (ACA) and
MCA in marmosets was 0 degrees in front. Insertion of a wire or microcatheter into the ICA
periphery should be performed as carefully as possible, and unnecessary stress on the blood
vessel wall should not be applied. Small arteries easily spasm, which is troublesome and did
not improve much with the use of drugs.
Superior Mesenteric Artery (SMA). The SMA was cannulated by referring to the abdom-
inal aortography image. Then, angiography was performed by injecting a contrast agent
through the tip of the microcatheter located at the SMA’s origin (Fig 2c). Although there were
rarely problems in the frontal view (LAO 0 degrees) when cannulating the SMA from the
abdominal aorta, a tilt of the viewing angle to LAO 80 degrees when the bifurcation was
approaching the right renal artery (RA) was sometimes beneficial. Because the SMA is the
most critical blood vessel in the visceral branch in both laboratory animals and humans, the
cannulation must be carried out carefully. Notably, the wire must not accidentally penetrate
the periphery. Researchers need to be careful as the force applied to the vessel wall increases
arterial dissection risk.
Renal Artery (RA). Abdominal angiography was also usually performed during selective
cannulation of the RAs to determine the position and direction of the bifurcation of the RAs.
However, the left RA is often branched almost at right angles and is easy to cannulate without
performing abdominal angiography. The RA was the most easily cannulated vessel among the
abdominal visceral branches (Fig 2d). However, there are many individual differences in the
position and angle of the bifurcation were observed. Abdominal angiography reduced contrast
enhancements, which is essential because contrast material load and circulating volume load
apply a more significant burden on small animals. Besides, caution should be exercised when
blind wire manipulation is performed because if the wire is mistakenly misplaced in the RA,
and manipulation is continued without checking the fluoroscopic image, a renal injury may be
caused.
Discussion
Using this technique, which has been applied in the clinic in human, minimally invasive proce-
dures can lead to a “reduction” in suffering and “refinement” of the welfare of laboratory ani-
mals, innovative change of veterinary care. In common marmosets, which have attracted
attention recently as non-human primates [9], which is an alternative to the rats commonly
used in animal experiments, high-resolution imaging equipment and minute intravascular
treatment technology can also be utilized in new minimally invasive animal experimental pro-
tocols with high safety and high versatility possible. The development of these protocols also
helps prevent the high costs of investing in unnecessary clinical trials, laboratory animals’
death, and ineffective drug discovery.
Thin catheters can be inserted into an organ of interest with fluoroscopy and a small
amount of contrast material through a method similar to that used in endovascular treatment
techniques currently in clinical practice. In both marmosets and rats, when puncturing
PLOS ONE | https://doi.org/10.1371/journal.pone.0250576 April 28, 2021
7 / 10
PLOS ONE The ultra-minimally invasive technique allows the 3Rs to be achieved in small primates
percutaneously, the CVA runs ventrally in the midline, aiming for the middle to result in a suc-
cessful puncture. The tail bone of the marmoset projects ventrally around the CVA, making it
easy to palpate. However, this procedure requires hard training when performed in small ani-
mals. A well-trained researcher can super-selectively place microcatheters to perform imaging
and administer substances through the microcatheters. Administering an embolic material
creates an animal model of organ ischemia, whereas administering a drug allows drug inter-
vention studies on the models created. Further, the administration of cells would allow cell
therapy studies on the model produced. The method described here is a highly versatile
method that can be applied in various animal experiments. There is no problem if the needle
for blood collection is thin, but the thicker the sheath, the better. There are a wide variety of
catheters that can enter the thicker sheath. Therefore, it is also useful in regenerative medicine
research, which is currently beginning to show its clinical practice effectiveness [10, 11]. The
insertion of a microcatheter through the caudal artery approach allows for selective drug
administration and cellular therapies targeting specific organs.
In the previous animal experiments, drug administration and cell therapy have limitations
to be performed only once or twice due to technical difficulties. The same caudal approach
termed the cut-down method allows only three procedures (maximum) because at least 5–10
mm of avulsion areas are required to expose, cut down the caudal arteries, and hemostasis liga-
tion is essential [12]. With the puncture technique described here, no dissection is required;
therefore, a slight shift in the puncture site can be tried, and approaches from the periphery
rather than the previous puncture site can be made. Although not shown here, the puncture
technique can be employed in one animal 5–6 times. CVA puncture is minimally invasive, and
there is no damage to the trunk or extremities. Because this method can be performed just by
puncture, it is as invasive as typical intravascular experiments, with only temporary pain. The
procedure time is also only a matter of seconds. The cut-down method requires ligation to
stop the bleeding, but the puncture method readily achieves hemostasis with compression
hemostasis (S1 Table). The major organs of small animals, such as marmosets or rats, can also
be cannulated without surgical damage to the extremities. Also, the target organ and the
peripheral artery of the organ can be selectively cannulated. Furthermore, securing the CVA
route is possible. In this study, we described the method as cannulation to organs, but there
are also applications for an arterial blood collection route [13].
The biggest bottlenecks of this method are technical skills and some pitfalls. Experiments
should be carried out with a clear understanding of the procedure. We want to refer readers to
the S1 Movie shown in the supplementary material regarding proficiencies in the procedure.
The animal’s body (including the entire tail) must be kept warm, as we have already reported
[8]. Besides, laboratory animals’ unnecessary injury can be avoided by switching to the cut-
down technique if puncture fails even after a few trials or by attempting a retry puncture one
week later. A contrast medium is essential to guide intravascular microcatheters to the
intended site during angiography. Therefore, attention should be paid to allergy development
concerns to contrast media in experimental animals and contrast-induced nephropathy. The
risk of heart failure due to volume overload and contrast media dose is a critical problem in
small animals with low body weight, and efforts should be made to minimize contrast media
used in a single administration. With these points in mind, we used only half-diluted contrast.
Moreover, approximately 0.1–0.15 ml of half-diluted contrast (approximately 0.2 ml of aor-
tography) was injected. The maximum contrast media dose was up to 3 ml. The most signifi-
cant impact on the number of angiographic examinations is radiopaque of microcatheters.
Human clinical microcatheters always have radiopaque markers, but similar microcatheters
have not been developed for small animals, such as rats or marmosets, and require angiogra-
phy to locate the microcatheter tip. In addition, because our described technique is very
PLOS ONE | https://doi.org/10.1371/journal.pone.0250576 April 28, 2021
8 / 10
PLOS ONE The ultra-minimally invasive technique allows the 3Rs to be achieved in small primates
minimally invasive, it can be performed with only local anesthesia in humans but requires gen-
eral anesthesia in experimental animals in all cases because they are unable to remain at rest
and still. Of course, there is also a need for angiography fluoroscopy equipment in laboratory
animals, X-ray control areas, and other settings to reduce excess radiation exposure to
researchers and laboratory animals. Knowledge of radiation exposure, the angiographic fluo-
roscopy system itself, and the experimental instruments and devices’ characteristics are also
crucial.
This technique is fundamental and revolutionary in animal experiments. We hope that this
approach will be carried out by many researchers, leading to a “reduction” in suffering and
“refinement” of laboratory animals’ welfare. This technique will also contribute to small ani-
mal medicine such as dogs and cats.
Supporting information
S1 Fig. Tips and schema of artery cannulation “Ohta Method”. If performed with these
points in mind, it is possible to reach any target organ’s vessels in an ultra-minimally invasive
procedure.
(PDF)
S1 Table. Comparison of approach methods at the CVA. This table compares the approach
methods of the puncture method and the cut-down method. The puncture method is less inva-
sive, less time consuming and is can be performed multiple times. CVA; caudal ventral artery.
(DOCX)
S1 Movie. A movie of actual “Ohta Method”. The trick is to do it carefully and slowly until
the outer sheath is inserted and then quickly inserted the microcatheter.
(MOV)
S1 File.
(DOCX)
Acknowledgments
We thank the following staff members for their contributions to this study: Yasuo Kimura,
Hazuki Sato, Saori Ogatsu, Yuki Minowa(Laboratory Animal Facilities, The Jikei University
School of Medicine).
Author Contributions
Conceptualization: Hiroki Ohta.
Data curation: Hiroki Ohta, Teppei Komatsu.
Formal analysis: Hiroki Ohta, Teppei Komatsu.
Funding acquisition: Hiroki Ohta, Teppei Komatsu.
Investigation: Hiroki Ohta, Teppei Komatsu, Kanako Muta, Makoto Koizumi.
Methodology: Hiroki Ohta, Makoto Koizumi.
Project administration: Hiroki Ohta.
Supervision: Yasuyuki Iguchi, Hirotaka James Okano.
Visualization: Kanako Muta.
PLOS ONE | https://doi.org/10.1371/journal.pone.0250576 April 28, 2021
9 / 10
PLOS ONE The ultra-minimally invasive technique allows the 3Rs to be achieved in small primates
Writing – original draft: Hiroki Ohta, Teppei Komatsu.
Writing – review & editing: Hiroki Ohta, Teppei Komatsu, Hirotaka James Okano.
References
1. Kanaoka Y, Ohki T, Toya N, Ishida A, Tachihara H, Hirayama S, et al. Technical challenges in endovas-
cular repair of complex thoracic aortic aneurysms. Ann Vasc Dis. 2012; 5:21–29. https://doi.org/10.
3400/avd.oa.11.01011 PMID: 23555482
2. Ohta H, Ohki T, Kanaoka Y, Kurosawa K, Fukushima S, Toya N. Hybrid approach to complex arch and
thoracoabdominal aneurysms in a case of high-risk patient. Clin Surg. 2016; 1:1011.
3. Carvalho C, Gaspar A, Knight A, Vicente L. Ethical and scientific pitfalls concerning laboratory research
with non-human primates, and possible solutions. Animals (Basel). 2018; 9.
4. Yang HC, Zuo Y, Fogo AB. Models of chronic kidney disease. Drug Discov Today Dis Models. 2010;
7:13–19. https://doi.org/10.1016/j.ddmod.2010.08.002 PMID: 21286234
5.
Jin-ichi Koizumi Y Y, Teiji Nakazawa, Genju Ooneda Experimental studies of ischemic brain edema. I.
A new experimental model of cerebral embolism in rats in which recirculation can be introduced in the
ischemic area. Jpn J Stroke. 1986; 8:1–8.
6. Belayev L, Alonso OF, Busto R, Zhao W, Ginsberg MD. Middle cerebral artery occlusion in the rat by
intraluminal suture. Neurological and pathological evaluation of an improved model. Stroke. 1996;
27:1616–1622; discussion 1623. https://doi.org/10.1161/01.str.27.9.1616 PMID: 8784138
7. Buhalog A, Yasuda R, Consigny D, Maurer K, Strother CM. A method for serial selective arterial cathe-
terization and digital subtraction angiography in rodents. AJNR Am J Neuroradiol. 2010; 31:1508–1511.
https://doi.org/10.3174/ajnr.A2134 PMID: 20488903
8. Ohta H, Ohki T, Kanaoka Y, Koizumi M, Okano HJ. Pitfalls of invasive blood pressure monitoring using
the caudal ventral artery in rats. Sci Rep. 2017; 7:41907. https://doi.org/10.1038/srep41907 PMID:
28198822
9. Mansfield K. Marmoset models commonly used in biomedical research. Comp Med. 2003; 53:383–392.
PMID: 14524414
10. Ohta H, Liu X, Maeda M. Autologous adipose mesenchymal stem cell administration in arteriosclerosis
and potential for anti-aging application: A retrospective cohort study. Stem Cell Res Ther. 2020; 11:538.
https://doi.org/10.1186/s13287-020-02067-x PMID: 33308301
11. Chu DT, Nguyen Thi Phuong T, Tien NLB, Tran DK, Minh LB, Thanh VV, et al. Adipose tissue stem
cells for therapy: An update on the progress of isolation, culture, storage, and clinical application. J Clin
Med. 2019; 8.
12. Kumagai K, Horikawa M, Yamada K, Uchida BT, Farsad K. Transtail Artery Access in Rats: A New
Technique for Repeatable Selective Angiography. J Vasc Interv Radiol. 2020; 31(4):678–681.e4.
https://doi.org/10.1016/j.jvir.2019.07.027 PMID: 31706884
13. Ohta H, Vo N-MV, Hata J, Terawaki K, Shirakawa T, Okano HJ. Utilizing Dynamic Phosphorous-31
Magnetic Resonance Spectroscopy for the Early Detection of Acute Compartment Syndrome: A Pilot
Study on Rats. Diagnostics. 2021; 11(4):586. https://doi.org/10.3390/diagnostics11040586 PMID:
33805144
PLOS ONE | https://doi.org/10.1371/journal.pone.0250576 April 28, 2021
10 / 10
PLOS ONE |
10.1371_journal.pone.0240348 | RESEARCH ARTICLE
Spatiotemporal analysis of medical resource
deficiencies in the U.S. under COVID-19
pandemic
Dexuan ShaID
Yuyang Tian5, Chaowei Yang1,2
1,2, Xin MiaoID
3*, Hai Lan1,4, Kathleen Stewart4, Shiyang Ruan2, Yifei Tian1,
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: Sha D, Miao X, Lan H, Stewart K, Ruan S,
Tian Y, et al. (2020) Spatiotemporal analysis of
medical resource deficiencies in the U.S. under
COVID-19 pandemic. PLoS ONE 15(10):
e0240348. https://doi.org/10.1371/journal.
pone.0240348
Editor: Wenbin Tan, University of South Carolina,
UNITED STATES
Received: June 16, 2020
Accepted: September 24, 2020
Published: October 14, 2020
Peer Review History: PLOS recognizes the
benefits of transparency in the peer review
process; therefore, we enable the publication of
all of the content of peer review and author
responses alongside final, published articles. The
editorial history of this article is available here:
https://doi.org/10.1371/journal.pone.0240348
Copyright: © 2020 Sha et al. This is an open access
article distributed under the terms of the Creative
Commons Attribution License, which permits
unrestricted use, distribution, and reproduction in
any medium, provided the original author and
source are credited.
Data Availability Statement: All relevant data are
available from GitHub (https://github.com/
stccenter/COVID-19-Data/tree/master/US).
1 NSF Spatiotemporal Innovation Center, George Mason University, Fairfax, VA, United States of America,
2 Department of Geography and GeoInformation Science, George Mason University, Fairfax, VA, United
States of America, 3 Department of Geography, Geology and Planning, Missouri State University,
Springfield, MO, United States of America, 4 Department of Geographical Sciences, University of Maryland,
College Park, MD, United States of America, 5 Mercy Clinic Family Medicine, Springfield, MO, United States
of America
* xinxiao@missouristate.edu
Abstract
Coronavirus disease 2019 (COVID-19) was first identified in December 2019 in Wuhan,
China as an infectious disease, and has quickly resulted in an ongoing pandemic. A data-
driven approach was developed to estimate medical resource deficiencies due to medical
burdens at county level during the COVID-19 pandemic. The study duration was mainly
from February 15, 2020 to May 1, 2020 in the U.S. Multiple data sources were used to
extract local population, hospital beds, critical care staff, COVID-19 confirmed case num-
bers, and hospitalization data at county level. We estimated the average length of stay from
hospitalization data at state level, and calculated the hospitalized rate at both state and
county level. Then, we developed two medical resource deficiency indices that measured
the local medical burden based on the number of accumulated active confirmed cases nor-
malized by local maximum potential medical resources, and the number of hospitalized
patients that can be supported per ICU bed per critical care staff, respectively. Data on med-
ical resources, and the two medical resource deficiency indices are illustrated in a dynamic
spatiotemporal visualization platform based on ArcGIS Pro Dashboards. Our results pro-
vided new insights into the U.S. pandemic preparedness and local dynamics relating to
medical burdens in response to the COVID-19 pandemic.
1. Introduction
Coronavirus disease 2019 (COVID-19) was first identified in December 2019 in Wuhan,
China as an infectious disease, and has quickly resulted in an ongoing pandemic. Just before
the global pandemic COVID-19, a report by the Global Health Security Index was released,
which is the first-ever comprehensive ranking of 195 countries based on their pandemic pre-
paredness, with six categories of 140 questions and 34 indicators [1]. Although national health
security is fundamentally weak across the globe, the U.S. scored 83.5/100 and ranked No.1 in
PLOS ONE | https://doi.org/10.1371/journal.pone.0240348 October 14, 2020
1 / 19
PLOS ONE Funding: X.M. acknowledges support from NSF
CSSI-1835512. C.Y. is supported by NSF CNS-
1841520 and CSSI-1835507. (www.nsf.gov/) The
funders had no role in study design, data collection
and analysis, decision to publish, or preparation of
the manuscript.
Competing interests: The authors have declared
that no competing interests exist.
Spatiotemporal analysis of COVID-19 medical resource deficiencies
the report. As evidence, there were 34.7 critical care beds per 100,000 inhabitants in the U.S.
by 2009, which is higher than that of any other country [2, 3]. However, the U.S. has fewer hos-
pital beds (2.8), and practicing physicians (2.6) per 1,000 capita compared to other similar
large and wealthy countries [4].
Since the COVID-19 outbreak, it has been estimated that a significant percentage of the U.
S. population would test positive for COVID-19 even given a conservative estimation [5]. For
example, a recent AHA (American Hospital Association) webinar on COVID-19 projected
that 30% (96 million) of the U.S. population would test positive, with 5% (4.8 million) being
hospitalized, 2% (1.9 million) would be admitted to the intensive care unit (ICU), and 1%
(960,000) would require ventilators [6]. This projection is generally compatible with the char-
acteristics of COVID-19 in Wuhan, China, where 5% of patients required the intensive care
unit and 2.3% required a ventilator [7]. Based on a recent CDC survey, the actual weekly hospi-
talization rate in April 2020 was around 5.8–7.5% for 100 counties across 14 states [8], which
means a large number of infected patients will swarm into hospitals and ICUs. As a matter of
fact, the U.S. had the highest number of confirmed cases of COVID-19 (82,404) in the world
on March 26, 2020, and surpassed Italy for the highest national death toll (20,413) on April 11,
2020 [9, 10].
Are U.S. medical resources enough to handle the worst scenario during this crisis? The
Society of Critical Care Medicine (SCCM) released a report regarding the medical resources
both available and needed for a potentially overwhelming number of critically ill patients [6].
In this report, three fundamental elements or features, i.e. ventilators, ICU beds, and critical
care staff (CCS) were identified as medical resources to plan for or manage a COVID-19 pan-
demic, and it would be wise to consider the interconnections among these factors in a spatio-
temporal data analysis framework. Specifically, the medical resource distribution should be
correlated with COVID-19 pandemic statistics in space (2D) and time (1D). So medical
resource burden or deficiency can be identified through feature selection, visualization, moni-
toring, and cluster analysis [11].
Among the three elements mentioned above, an inventory of ventilators is difficult to quan-
tify for estimating critical supply shortages. Based on a 2009 AHA survey, a total of 5,752 U.S.
acute care hospitals were estimated to have 62,188 full-featured mechanical ventilators and
98,738 ventilators with limited features [12]. The Strategic National Stockpile (SNS) had an
estimated 8,900 ventilators for emergency deployment in 2010, and between 12,000 and
13,000 ventilators by March 13, 2020 [13–15]. Based on these numbers, the ventilator inven-
tory was approximately 173,000–174,000 in the U.S. A model-based analysis suggested that US
hospitals could absorb between 26,200 to 56,300 additional ventilators at the peak of a national
pandemic with robust pre-pandemic planning [16]. Since SNS can deliver ventilators within
24–36 hours after being requested by states and approved by federal organizations, and no reli-
able database for ventilator inventory exists at county or state level, we will not consider this
factor in our spatiotemporal analysis. A recent model-driven study simply assumes one venti-
lator per critical care bed [17] and we use this same assumption in our analysis.
Hospital beds, especially ICU beds, are an important factor in evaluating medical resource
deficiency during the COVID-19 pandemic, and quantity of beds has been used as a major fac-
tor in model-driven predictions of local critical care capacity limit [17, 18]. However, safe use
of ventilators in ICU requires trained personnel. In a previous study, the number of trained
medical personnel is assumed to correlate with the number of staffed beds maintained by hos-
pitals [16]. This assumption is perhaps unrealistic at county level without considering the geo-
graphic disparity.
For this research, we assumed that a realistic measurement of the medical burden at county
level should consider both ICU beds and critical care staff (CCS), which will provide
PLOS ONE | https://doi.org/10.1371/journal.pone.0240348 October 14, 2020
2 / 19
PLOS ONE Spatiotemporal analysis of COVID-19 medical resource deficiencies
reasonable evidence for stakeholder (e.g., hospital, county and State governments policy and
decision-making). In this study, we (1) conduct a medical data analysis, and re-evaluate the
spatial distribution of medical resource features (hospital beds, ICU beds, and CCS) at county
level; (2) develop two Medical Resource Deficiency Indices (MRDI and MRDId) by linking
positive COVID-19 infections and local medical resources to measure local medical burden;
and (3) develop a data-driven dynamic spatiotemporal framework to visualize and analyze the
MRDI /MRDId trends at the county level. Our results provided a new dimension of insight
into the U.S. pandemic preparedness and local dynamic medical burden during COVID-19
pandemic. The dataset is open sourced and hosted on GitHub (https://github.com/stccenter/
COVID-19-Data/tree/master/US), and are visualized through ArcGIS Dashboards at: http://
mrd-dashboard.stcenter.net/.
2. Data
2.1. Base map and unit of analysis
A total of 3,143 counties and county-equivalents in the U.S. are used as the primary unit of this
study, since they are manageable in a GIS system and small enough to reflect local geographic
discrepancies. The base map was downloaded from the 2019 TIGER/LINE products from the
U.S. Census Bureau, which is the most comprehensive spatial dataset designed for GIS plat-
forms [19]. The county vector layer delineates the administrative boundary with land/water
area without any demographic data, but it provides geographic entity codes (GEOIDs) for
joining with other socio-economic data such as Census data. Based on the attributes of our col-
lected medical-related datasets, we also prepared state and ZIP code boundaries for data fusion
and integration at county level.
2.2. Medical resource feature extraction
In this study, two fundamental features of medical resources in the U.S. were extracted, i.e.,
hospital beds and critical care staff. Besides, the population and 60+ senior population data
was extracted at county level from KHN online database [20], which is used to normalize the
local medical data in the subsequent analysis.
2.2.1. Hospital beds. National public and private online datasets were used to prepare
county-level hospital bed counts. Hospital data were collected from Definitive Healthcare [21].
Definitive Healthcare consulting services share their hospital dataset to the entire health
research community through ArcGIS online, which cover information of nationwide bed
capacity and average yearly bed utilization of hospitals. Although it is not a real-time dataset
that reflects each hospital’s bed capacity during COVID-19, it can be used as a baseline to esti-
mate the geographic disparity of local health resources.
A hospital is defined as a healthcare institution providing inpatient, therapeutic, or rehabili-
tation services under the supervision of physicians with the capability of inpatient care [21].
All types of hospitals are included in our study. Five types of hospital beds are clearly identified
in the Definitive Healthcare dataset. In our study, two hospital bed capacities were selected
and used in the analysis. The first one is the number of licensed beds, which is the potential or
maximum number of beds for which a hospital holds a license to operate. The second type of
capacity refers to the number of adult ICU beds that could be used for COVID-19. During this
crisis, hospitals could use additional intensive care beds to supplement an influx of patients.
Therefore, adult ICU beds include not only internal medical ICU beds, but also burn, surgical,
and trauma ICU beds. However, pediatric, premature or neonatal ICU beds are not included
because they are mainly for a different target patient population, which has a much lower inci-
dence rate of COVID-19.
PLOS ONE | https://doi.org/10.1371/journal.pone.0240348 October 14, 2020
3 / 19
PLOS ONE Spatiotemporal analysis of COVID-19 medical resource deficiencies
Two other independent data sources of hospital beds are compared with the data from Defini-
tive Healthcare. One is from Kaiser Health News (KHN) based on reports of ICU beds in 2018–
2019 [20], and the other is from Homeland Infrastructure Foundation-Level Data (HIFLD) for
licensed hospital beds updated on October 7, 2019 [22]. We conducted a regression analysis com-
paring KHN with Definitive Healthcare in terms of ICU beds, and comparing HIFLD with Defin-
itive Healthcare in terms of licensed beds, and the coefficients of determination (r2) are 0.94 and
0.97, respectively. The results validated the quality of the Definitive Healthcare dataset.
2.2.2. Critical care staff. A dataset of critical care staff (CCS) was extracted from the weekly
updated National Provider Identifier Registry (NPI) database (~7.1 GB) through structured query
language (SQL) [23]. The NPI is a unique 10-digit identification number for each health-care pro-
vider issued by the Centers for Medicare Medicaid Services through the National Plan and Pro-
vider Enumeration System. Each health-care provider could have multiple taxonomy codes, which
indicate areas of specialization. Through consulting with medical researchers and front-line physi-
cians, we extracted detailed CCS data from the NPI database released on April 15, 2020 as a medi-
cal resource feature (Table 1). Our study identifies 197,061 health care providers by searching
unique NPI records and removing duplicate records. With the development of COVID-19 in the
U.S., all these ICU-related staff (emergency medicine physician, critical care physicians, anesthesi-
ologists, hospitalists, pulmonologist, infectious disease physician, surgery, anesthesiologist assis-
tant, critical care nurses, nurse anesthetist, and respiratory therapists trained in mechanical
ventilation) would become valuable but limited asset for critically ill ventilated patients [6].
2.3. COVID-19 patients
The U.S. Centers for Disease Control and Prevention (CDC) published daily COVID-19 con-
firmed cases on February 25, 2020. Each state got involved soon after and began to report
COVID-19 data, including the daily and accumulated test and confirmed case numbers, hospi-
talization data, and death numbers at state level. However, numbers of discharged or released
patients from hospitals are less widely available, e.g., only a few states, such as Maryland, Colo-
rado, and New York provide some (incomplete) statistics on recovered patients from both
hospital and home. This study mainly uses the data collected by the NSF Spatiotemporal Inno-
vation Center (STC) at George Mason University. This dataset uses a datacube structure for
Table 1. Critical care staff extracted from NPI database.
Physician
Critical Care Staff (CCS)
Emergency Medicine
Anesthesiology (Critical Care Medicine)
Hospitalist
Internal Medicine (Infectious Disease)
Internal Medicine (Critical Care Medicine)
Internal Medicine (Pulmonary Disease)
Surgery (Surgical Critical Care)
Physician Assistant
Anesthesiologist Assistant
Nurse
Certified & registered Nurse Anesthetist
Technician
Total
Nurse Practitioner (Critical Care Medicine)
Certified Respiratory Therapist
Registered Respiratory Therapist
-
Taxonomy Code
Number
207P00000X
207LC0200X
208M00000X
207RI0200X
207RC0200X
207RP1001X
2086S0102X
367H00000X
367500000X
363LC0200X
2278C0205X
2279C0205X
-
67591
1871
27827
11299
10976
19990
2392
2953
61585
1040
164
379
-
Total Number�
131519
2953
62589
538
197,061
� Duplicate records were removed since one health care provider may have multiple Taxonomy Codes.
https://doi.org/10.1371/journal.pone.0240348.t001
PLOS ONE | https://doi.org/10.1371/journal.pone.0240348 October 14, 2020
4 / 19
PLOS ONE Spatiotemporal analysis of COVID-19 medical resource deficiencies
spatiotemporal data aggregation from multiple sources. The data is cleaned, standardized, and
updated daily to solve any data conflicts, and a time-series summary at state and county level is
provided for the U.S. [10, 24].
The numbers of county-level confirmed positive cases as well as deaths were originally
extracted from USA Facts based on CDC data [25], and compared with local public health
agencies for verification. The confirmed and death cases reflect cumulative statistics since Jan-
uary 22, 2020, the day after the first confirmed cases were reported in Washington State. Fur-
thermore, state level test and hospitalization data were extracted from the COVID Tracking
Project [26]. However, the current and accumulated hospitalization cases from state health
departments are largely incomplete. By April 29, 2020, a total of 22 states reported both current
and accumulated hospitalized patient numbers, 17 states reported only current hospitalized
numbers, and 10 states only reported accumulated hospitalized numbers, while Washington,
D.C., Nevada and Nebraska did not provide information on the number of hospitalized cases.
3. Methods
Our analysis was mainly based on the publicly available data of the new confirmed daily cases
reported for the U.S. from the 25th of February until the 1st of May, 2020. All data were fully
anonymized.
3.1. Medical feature extraction and aggregation
Raw datasets in this study were collected from multiple sources with heterogeneous formats
and structures. All data were processed and aggregated at county level based on County Fed-
eral Information Processing Standard (FIPS). Several aggregation methods were used for each
raw dataset, as summarized in Fig 1.
First, the hospital data was originally presented as a point location in a coordinate format,
and its attribute table includes five types of hospital beds. The spatial point aggregation algo-
rithm was used to integrate the numbers of licensed beds and adult ICU beds at county level.
The bed numbers per 1,000 residents were also calculated at county level.
The primary practice addresses of CCS were imported from the NPI database, and 5-digital
ZIP codes were extracted. The total number of CCS within a county was counted based on the
county’s ZIP codes through geocoding and the point/ polygon aggregation algorithm. The
number of CCS per 1,000 residents were also calculated at county level.
The accumulated COVID-19 confirmed case numbers were extracted at county level. We
used existing hospitalization data to estimate the average length of stay (ALOS) in acute care,
since it is key for estimating the daily hospitalized patients. For a given state, the current hospital-
ized patients should be equal to the accumulation of hospitalized patients minus the accumula-
tion of deaths and discharged patients within the most recent ALOS. Since no patient discharge
data was available, we assumed that the number of discharged patients was zero. Therefore, we
estimated ALOS by matching (1) the accumulation of hospitalized patients minus the accumula-
tion of deaths in most recent days, and (2) the current number of hospitalized patients, and
finally interpolating by two nearest days or accumulation periods. It turns out to be an optimiza-
tion problem to find a parameter (n) to match the two data sources, as shown in Eq (1).
ALOS ¼ arg min
ðNh;n (cid:0) Ndeath;nÞ (cid:0) Nch
n2<þ
ð1Þ
where Nh,n is the accumulated number of hospitalized patients in the past n days, Ndeath,n is the
accumulated number of deaths in the past n days, and Nch is the number of currently hospital-
ized patients.
PLOS ONE | https://doi.org/10.1371/journal.pone.0240348 October 14, 2020
5 / 19
PLOS ONE Spatiotemporal analysis of COVID-19 medical resource deficiencies
Fig 1. Medical feature extraction workflow.
https://doi.org/10.1371/journal.pone.0240348.g001
State hospitalization data were only available recently (starting from March 17, 2020 in NY)
with numerous missing data. By May 1, 2020, among 22 states that have both current and
accumulated numbers of hospitalized patients, eight states (Colorado, Massachusetts, Maine,
Minnesota, Montana, North Dakota, New York, Oklahoma) had complete data for the most
recent 20 days; 12 states (Oklahoma, Wisconsin, Mississippi, Maryland, New Hampshire, New
Mexico, Oregon, South Dakota, Virginia, Wyoming, Rhode Island, Kentucky) only had data
in the most recent 5–15 days; and data from Arkansas, Arizona, and Connecticut were aban-
doned due to poor quality. We calculated the daily ALOS for these 19 states and pooled the
results in Fig 2. The state ALOS ranges from 8.8 (New Mexico)-28.5 (Mississippi) days. The
overall national ALOS weighted by state hospitalized patients is 15.5 days, which is longer than
a previous estimation that the ALOS in acute care were 11 days [18]. It is worth noting that
ALOS is likely to be underestimated since we assumed no discharged patients. Furthermore,
ALOS is subject to change when more hospitalization data become available in the future.
Finally, we define the COVID-19 hospitalized rate as the ratio of the number of current hos-
pitalized patients and the accumulated confirmed case numbers during the most recent ALOS.
If the hospitalized rate remains the same within a state, the daily hospitalized patient number
in a county can be estimated by using the accumulated COVID-19 confirmed case numbers
minus deaths in the most recent ALOS, multiplied by the state average hospitalized rate. If no
PLOS ONE | https://doi.org/10.1371/journal.pone.0240348 October 14, 2020
6 / 19
PLOS ONE Spatiotemporal analysis of COVID-19 medical resource deficiencies
Fig 2. Box-plot (5-number summary) of hospitalized ALOS among 19 states.
https://doi.org/10.1371/journal.pone.0240348.g002
state ALOS is available, we use the overall national average ALOS of 15.5 days. This daily hos-
pitalized patient number can be used to evaluate the daily medical burden at county level.
3.2. Medical resource deficiency indices
The medical resource deficiency indices (MRDI) are defined as an indicator of medical
resource burden at county level. We define two forms of MRDI: general MRDI, and local daily
MRDI (MRDId).
MRDI ¼
Nc (cid:0) Ndeath
Nlicbed � NCCS
ð2Þ
where Nc is the accumulated number of confirmed COVID patients, Ndeath is the accumulated
number of deaths, Nlicbed is the total number of licensed beds, and NCCS is the number of criti-
cal care staff. We assumed that Nlicbed and NCCS were relatively independent at county level,
and the product of them represents the interconnection of these two medical resource features
or factors. Therefore, the MRDI represents the number of accumulated active confirmed cases
normalized by the local maximum potential medical resources (total licensed beds and total
CCS). MRDId is represented as
MRDId ¼
ðNcA (cid:0) NdAÞ � rh
Nicubed � NCCS
ð3Þ
where NcA is the accumulated confirmed case numbers during a most recent ALOS, NdA is the
accumulated death numbers during the same ALOS, rh is the state hospitalized rate derived
PLOS ONE | https://doi.org/10.1371/journal.pone.0240348 October 14, 2020
7 / 19
PLOS ONE Spatiotemporal analysis of COVID-19 medical resource deficiencies
from state hospitalization data, and Nicubed is the number of adult ICU beds. MRDId represents
the local daily medical burden, or the number of hospitalized patients that can be supported
per ICU beds per CCS. MRDId is large (>1) when local medical resources cannot fully support
the hospitalized critically ill patients, or the local medical burden is heavy; and MRDId is small
(<1) when local medical resources are sufficient.
3.3. Visualization analysis using ArcGIS Dashboards
Based on ArcGIS Dashboard, we designed a comprehensive operational dashboard for moni-
toring, analyzing, visualizing, and sharing our medical data and analyzed results. A multi-
stacked map is built at the center of the interface (Fig 3), which represents the spatial distribu-
tions of COVID-related statistics such as MRDI, death rate, and infection rate at county level
over the U.S. In addition to visualizing the macro spatial distribution pattern of those statistics
results, two lists of counties are displayed. Those counties are dynamically filtered by the cur-
rent map extent in map view and are ranked in real-time by hospitalized rate and death rate to
represent the spreading of COVID-19 and the outbreak situation in the selected study area.
Focusing on a specific county, an indicator and two pie charts are applied to display for each
county (Fig 4): 1) the comparison of active COVID—19 cases and the number of overall beds;
2) the percentage of ICU beds in overall beds; and 3) the proportion of each type of CCS. From
the temporal analysis perspective, a time series chart is designed to demonstrate the dynamics
of medical resource deficiencies for each county on a daily basis during the pandemic. In the
following section, we will use the dashboard components to analyze spatiotemporal distribu-
tions of medical resource deficiencies. We will further explore the possible factors relating to
the medical resource deficiencies for specific counties and areas as well as the medical resource
capacity for non-severe COVID-19 patients, the supplies needed for severe cases, and propor-
tion of each type of CCS.
Fig 3. Spatiotemporal visualization interface based on ArcGIS Dashboards.
https://doi.org/10.1371/journal.pone.0240348.g003
PLOS ONE | https://doi.org/10.1371/journal.pone.0240348 October 14, 2020
8 / 19
PLOS ONE Spatiotemporal analysis of COVID-19 medical resource deficiencies
Fig 4. A use case for regional visual analysis of Tennessee.
https://doi.org/10.1371/journal.pone.0240348.g004
4. Results
4.1. Medical resource features
The ICU beds per 1,000 residents (Fig 5A) and CCS per 1,000 residents (Fig 5B) are mapped at
county level. Both maps show that these two medical resources are not homogeneously distrib-
uted across the U.S. Some midwestern states, such as North Dakota, South Dakota, Nebraska,
Kansas, and Montana have more ICU beds, but less CCS. The spatial distribution of CCS
shows a checker board pattern, with many gaps or low numbers across the country. The prod-
uct of ICU beds and CCS per 1,000 residents is shown in Fig 6A. The darkest green zones rep-
resent counties with higher quantities of medical resources including ICU beds and CCS.
A total of 19 major medical centers represent top ranking healthcare facilities in the U.S.
(Table 2) [27]. Medical centers are conglomerations of health care facilities including hospitals
and research facilities that could be affiliated with a medical school. Overlaying the locations
of these 19 medical centers on the map (purple circles on the map), it seems these counties and
medical centers are spatially highly correlated (Fig 6A).
Since senior people (aged 60+) are vulnerable to COVID-19, we also produced a map of the
product of ICU beds and CCS per 1,000 senior residents (Fig 6B). This map represents loca-
tions where the supply of medical resources for seniors is higher.
A regression analysis was conducted to examine the correlation between CCS and adult
ICU beds at county level (Fig 7). If all 3,143 counties are included, the coefficient of determina-
tion (r2) is 0.90. However, this high r2 value is quite misleading, since it is heavily influenced
by several large counties with rich medical resources (blue dots). Removing the top 30 coun-
ties, causes the coefficient of determination (r2) to drop to 0.78, which better represents the
geographic disparity of these two factors in most (3113) of the U.S. counties, as shown in Fig
5A and 5B.
PLOS ONE | https://doi.org/10.1371/journal.pone.0240348 October 14, 2020
9 / 19
PLOS ONE Spatiotemporal analysis of COVID-19 medical resource deficiencies
Fig 5. Geographical distribution of medical resources at county level normalized by local population. (a) ICU beds
per 1,000 residents; (b) CCS per 1,000 residents.
https://doi.org/10.1371/journal.pone.0240348.g005
A total of 671 counties have neither ICU beds nor CCS, and are shown in Fig 8. These coun-
ties are mainly distributed in less-populated rural areas across the U.S., and they are not
included in MRDI or MRDId calculation to avoid a divide-by-zero error. During the COVID-
19 pandemic, individuals requiring a higher level of care in these areas would be sent to
PLOS ONE | https://doi.org/10.1371/journal.pone.0240348 October 14, 2020
10 / 19
PLOS ONE Spatiotemporal analysis of COVID-19 medical resource deficiencies
Fig 6. Overall medical resources at county level normalized by local population (a) The product of ICU beds and CCS per 1,000 residents, and
19 medical centers shown as purple bubbles; (b) The product of ICU beds and CCS per 1,000 for senior residents (aged 60+).
https://doi.org/10.1371/journal.pone.0240348.g006
PLOS ONE | https://doi.org/10.1371/journal.pone.0240348 October 14, 2020
11 / 19
PLOS ONE Spatiotemporal analysis of COVID-19 medical resource deficiencies
Table 2. Major medical centers in the U.S.
#
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
Medical Center
Location
Banner University Medical Center Tucson
Phoenix Healthcare Cluster
Loma Linda University Medical Center
Stanford University Medical Center
UCSF Medical Center
Ronald Reagan UCLA Medical Center
Illinois Medical District
National Institutes of Health
Johns Hopkins
Boston Longwood Area
Washington University Medical Center
Mayo Clinic
NewYork-Presbyterian Hospital
Cleveland Clinic
University of Pennsylvania Health System
University of Pittsburgh Medical Center
Texas Medical Center
South Texas Medical Center
Southwestern Medical District
Tucson, AR
Phoenix, AR
Loma Linda, CA
Stanford, CA
San Francisco, CA
Los Angeles, CA
Chicago, IL
Bethesda, MD
Baltimore, MD
Boston, MA
St. Louis, MO
Rochester, MN & others
New York, NY
Cleveland, Oh
Philadelphia, PA
Pittsburgh, PA
Houston, TX
San Antonio, TX
Dallas, TX
https://doi.org/10.1371/journal.pone.0240348.t002
neighboring counties with sufficient medical resources, and could result in larger MRDId in
the neighboring counties.
4.2. Spatiotemporal trend of MDRI and MDRId
The spatiotemporal dynamics of general MDRI across the U.S. is illustrated at: http://mrd-
dashboard.stcenter.net/. The general MDRI represents the number of accumulated active con-
firmed COVID-19 cases normalized by local maximum potential medical resources, while the
dynamic view provides an insightful alternative visualization of COVID-19 U.S. cases by
county. Six snapshot maps are illustrated in Fig 9A–9F, which demonstrate six time-stamped
frames taken on February 15, March 15, April 15, May 15, June 15, and July 15, 2020. A pro-
portional symbol map is used with semi-transparent red circles to represent the general
MDRI. This visualization technique enhances clustering patterns, and there is a clear trend
where the general medical burden shifted from the east coast of the U.S. to midwestern states.
As of July 2020, it would seem that Louisiana, Mississippi, Georgia, Tennessee, Indiana, and
Iowa are possibly suffering a new wave of medical resource deficiencies due to the rapid
increase of accumulated active confirmed cases in some counties.
Furthermore, the spatiotemporal dynamics of local daily MRDId is also illustrated in the
dashboards. Since hospitalization data has been available only recently, we illustrate two
frames taken on May 1, and August 1, 2020 (Fig 10A and 10B). The red circle symbols are
semi-transparent, and county-level medical resource deficiencies are visually enhanced by
searching the reddest clustering patterns in the map. During this COVID-19 infection period,
it seems that Mississippi, Louisiana, Tennessee, and Indiana were suffering from medical
resource deficiencies, which would have required special attention when relocating medical
resources if necessary. These hotspots have been partially confirmed from local news reports.
For example, there were 5,153 known presumptive cases with the total death toll of 201 in Mis-
sissippi on April 23, 2020 [28]; new cases of COVID-19 rose sharply on May 1 in East Baton
PLOS ONE | https://doi.org/10.1371/journal.pone.0240348 October 14, 2020
12 / 19
PLOS ONE Spatiotemporal analysis of COVID-19 medical resource deficiencies
Fig 7. The correlation between CCS and adult ICU beds at county level. Blue dots represent 30 counties with rich medical resources. Orange dots
represent the other 3113 counties. The blue dashed line is the overall regression line, and the green solid line is the regression line for the 3113 counties only.
https://doi.org/10.1371/journal.pone.0240348.g007
Rouge, Louisiana, as deaths approached 350 in the region [29]; the nation’s highest infection
rate was in a county in Trousdale County, Tennessee, where 1,300 cases of Covid-19 were
reported, and most of them traced back to a state correction center [30]; and Indiana passed
1,000 COVID-19 deaths on April 29, 2020 [31].
4.3. Spatiotemporal visualization and analysis interface
In the center of the dashboard, several map layers could be selected to show the general spatial
distribution of MRDI, death rate, infection rate and active cases over licensed beds per capita.
After interactive map scaling (by zooming in/out) and moving (by dragging) operations, or
using the polygon selection tool, the charts and rank list are linked and self-adapted to the
analysis region of interest to a user. By clicking the polygon of a selected county, attribute
PLOS ONE | https://doi.org/10.1371/journal.pone.0240348 October 14, 2020
13 / 19
PLOS ONE Spatiotemporal analysis of COVID-19 medical resource deficiencies
Fig 8. The 671 counties without licensed beds or CCS.
https://doi.org/10.1371/journal.pone.0240348.g008
information about medical resources and COVID-19 related data would popup and the rele-
vant chart is automatically updated in the dashboard.
Northern Tennessee State is presented as a use case to show the possible interactive analysis
(Fig 4). Since western and east coast regions have more medical resources than central regions
(Fig 6A), and the states along the Mississippi River in the southern U.S. show a high risk (Fig
10), we zoom in on the map and select the nearest region with the largest red bubble in Ten-
nessee (Fig 4). Thirty counties are selected as a result, and relevant numbers are calculated and
presented in dashboard charts. The medical bed pie chart shows ICU beds are 10.89% in over-
all licensed beds, and the medical staff pie chart shows the nurses group is the highest (55.77%)
followed by physicians (44.01%), physician assistants (0.18%) and therapists (0.04%). The line
chart shows a time-series trend for MRDI in the northern Tennessee area, and we find the
index varied greatly between April 30, 2020 to May 1, 2020, which could be explained by the
possible tracing of the virus to a correction center outbreak in Trousdale County [32]. On the
right column of the dashboard, the risk factors of medical resource and infection rate is ranked
by the selected region. Trousdale, Davidson, and Sumner County are the top 3 with highest
infection risks, while Trousdale also shows the highest medical resource risk in this region.
The case study in Fig 4 demonstrates the potential of our developed dashboard for interactive
and visual analysis of specific regions of interest for policy makers, other stakeholders, and the
general public.
PLOS ONE | https://doi.org/10.1371/journal.pone.0240348 October 14, 2020
14 / 19
PLOS ONE Spatiotemporal analysis of COVID-19 medical resource deficiencies
Fig 9. General MRDI trend.
https://doi.org/10.1371/journal.pone.0240348.g009
5. Discussion and conclusions
In this study, a data-driven approach has been used to estimate the medical resource deficien-
cies or medical burden at county level during the COVID-19 pandemic across the U.S. Specifi-
cally, spatiotemporal data analysis methods including feature extraction, database structured
PLOS ONE | https://doi.org/10.1371/journal.pone.0240348 October 14, 2020
15 / 19
PLOS ONE Spatiotemporal analysis of COVID-19 medical resource deficiencies
Fig 10. Daily medical burden MRDId trend on May 1 (a) and August 1, 2020 (b).
https://doi.org/10.1371/journal.pone.0240348.g010
PLOS ONE | https://doi.org/10.1371/journal.pone.0240348 October 14, 2020
16 / 19
PLOS ONE Spatiotemporal analysis of COVID-19 medical resource deficiencies
query (SQL), data fusion or aggregation, linear regression analysis, and spatial statistics were
used to extract medical resource features and patient statistics, such as hospital beds, CCS,
local population, COVID-19 confirmed case numbers, and hospitalization data at county level.
The average length of stay (ALOS) was then estimated from hospitalization data at state level,
and the hospitalized rate were calculated at state and county level. Based on these datasets, we
developed two medical resource deficiency indices MRDI and MRDId that measure the local
medical burden from two different perspectives. The first index represents the number of
accumulated active confirmed cases normalized by local maximum potential medical
resources; and the second one represents the number of hospitalized patients that can be sup-
ported per ICU beds per critical care staff. The related medical resource data, MRDI and
MRDId were visualized and analyzed using a dynamic spatiotemporal platform created
through ArcGIS Pro Dashboards, which is a convenient way to enhance the clustering patterns
and trends.
Our analysis showed that (1) the spatial distribution of medical resources (hospital beds,
ICU beds, and CCS) at county level is highly heterogeneous across the U.S., and ICU beds and
CCS are not spatially highly correlated; (2) MRDI and MRDId can provide new insights into
the U.S. pandemic preparedness and local dynamics relating to medical burdens during a peak
period in the COVID-19 pandemic; and (3) a data-driven dynamic spatiotemporal framework
is a powerful data visualization tool to illustrate the trends of MRDI / MRDId and other medi-
cal-related statistics.
It is worth noting that we have not considered the number of discharged patients due to
lack of data, leading to a possible slight underestimate of ALOS during the COVID-19 rapid
infection period. As a result, MRDId may also be slightly underestimated. We also did not con-
sider the ratio of ICU patients and acute hospitalized patients due to lack of data, and assumed
all hospitalized patients were treated as ICU cases. As a result, MRDId was possibly overesti-
mated, and the values calculated here should be viewed as the upper limit of local medical bur-
dens. Some other uncertainties include (1) the numbers of registered hospital beds and CCS
could be incomplete or not up-to-date, although the most recent Definitive Healthcare and
NPI databases have been used, so the medical resources could be underestimated, (2) critically
ill patients in counties without ICU beds and CCS would be sent to neighboring counties with
sufficient medical resources, (3) some numbers of experienced ICU staff may become ill, (4)
the number of trained professionals may have increased based on emergent recruiting, and (5)
the capacity in ICUs and emergency rooms may have been expanded during the crisis. How-
ever, MRDId can still serve as a useful indicator to measure the county-level medical resource
deficiencies, and this index can be improved once more public health data are available in the
future. Furthermore, it could provide reasonable evidence for policy makers in local and state
governments to assess their medical inventories and staff resources, and provide preparedness
for decision of re-opening the economies and public life.
In the future, our work can be combined with epidemic models to either provide driving
parameters or calibrate the models and predict the local medical burdens. The spatiotemporal
analysis used in this study can be extended to include remote sensing data, social media data,
and mobile traffic flow data to estimate severity of pandemic or predict the outbreak cases in
the U.S. and other counties.
Author Contributions
Conceptualization: Dexuan Sha, Xin Miao, Yuyang Tian, Chaowei Yang.
Data curation: Dexuan Sha, Hai Lan, Shiyang Ruan, Yifei Tian, Chaowei Yang.
PLOS ONE | https://doi.org/10.1371/journal.pone.0240348 October 14, 2020
17 / 19
PLOS ONE Spatiotemporal analysis of COVID-19 medical resource deficiencies
Formal analysis: Dexuan Sha, Xin Miao, Hai Lan, Shiyang Ruan, Yifei Tian.
Funding acquisition: Xin Miao, Chaowei Yang.
Investigation: Yuyang Tian, Chaowei Yang.
Methodology: Dexuan Sha, Xin Miao, Hai Lan, Yifei Tian, Yuyang Tian, Chaowei Yang.
Project administration: Xin Miao, Chaowei Yang.
Resources: Yuyang Tian.
Software: Dexuan Sha, Hai Lan, Yifei Tian.
Supervision: Xin Miao, Chaowei Yang.
Validation: Dexuan Sha, Xin Miao, Hai Lan, Shiyang Ruan, Yuyang Tian, Chaowei Yang.
Visualization: Dexuan Sha, Hai Lan.
Writing – original draft: Dexuan Sha, Xin Miao, Hai Lan, Kathleen Stewart, Yuyang Tian,
Chaowei Yang.
Writing – review & editing: Kathleen Stewart, Chaowei Yang.
References
1.
2019 Global Health Security Index [Internet]. 2019 [cited April 25, 2020]. Available from: https://www.
ghsindex.org/wp-content/uploads/2020/04/2019-Global-Health-Security-Index.pdf.
2. Rhodes A, Ferdinande P, Flaatten H, Guidet B, Metnitz PG, Moreno RP. The variability of critical care
bed numbers in Europe. Intensive Care Medicine. 2012; 38:1647–53. https://doi.org/10.1007/s00134-
012-2627-8 PMID: 22777516
3. Wallace DJ, Angus DC, Seymour CW, Barnato AE, Kahn JM. Critical Care Bed Growth in the United
States. A Comparison of Regional and National Trends. American Journal of Respiratory and Critical
Care Medicine. 2015; 191(4):410–6. https://doi.org/10.1164/rccm.201409-1746OC PMID: 25522054
4. How prepared is the US to respond to COVID-19 relative to other countries? [Internet]. 2020 [cited April
25, 2020]. Available from: https://www.healthsystemtracker.org/chart-collection/how-prepared-is-the-
us-to-respond-to-covid-19-relative-to-other-countries/#item-start.
5. Estimating COVID-19 Prevalence in Symptomatic Americans [Internet]. 2020 [cited April 21]. Available
from: https://news.gallup.com/opinion/gallup/306458/estimating-covid-prevalence-symptomatic-
americans.aspx.
6. U.S. ICU Resource Availability for COVID-19 [Internet]. 2020 [cited April 21]. Available from: https://
sccm.org/Blog/March-2020/United-States-Resource-Availability-for-COVID-19.
7. Guan W-j, Ni Z-y, Hu Y, Liang W-h, Ou C-q, He J-x, et al. Clinical Characteristics of Coronavirus Dis-
ease 2019 in China. The New England Journal of Medicine. 2020. https://doi.org/10.1056/
NEJMoa2002032 PMID: 32109013
8.
Laboratory-Confirmed COVID-19-Associated Hospitalizations [Internet]. 2020 [cited April 25, 2020].
Available from: https://gis.cdc.gov/grasp/covidnet/COVID19_3.html.
9. Dong E, Du H, Gardner L. An interactive web-based dashboard to track COVID-19 in real time. Lancet
Infect Dis. 2020. Epub 2020/02/23. https://doi.org/10.1016/s1473-3099(20)30120-1 PMID: 32087114;
PubMed Central PMCID: PMC7159018.
10. Yang C, Sha D, Liu Q, Li Y, Lan H, Guan WW, et al. Taking the pulse of COVID-19: A spatiotemporal
perspective. International Journal of Digital Earth, https://doi.org/10.1080/17538947.2020.1809723
11. Yang C, Clarke K, Shekhar S, Tao CV. Big Spatiotemporal Data Analytics: a research and innovation
frontier.2020. 34(6): 1075–1088.
12. Rubinson L, Vaughn F, Nelson S, Giordano S, Kallstrom T, Buckley T, et al. Mechanical ventilators in
US acute care hospitals. Disaster medicine and public health preparedness. 2010; 4(3):199–206.
https://doi.org/10.1001/dmp.2010.18 PMID: 21149215
13. Huang H-C, Araz OM, Morton DP, Johnson GP, Damien P, Clements B, et al. Stockpiling Ventilators for
Influenza Pandemics. Emerging Infectious Diseases. 2017; 23(6):914–21. https://doi.org/10.3201/
eid2306.161417 PMID: 28518041
PLOS ONE | https://doi.org/10.1371/journal.pone.0240348 October 14, 2020
18 / 19
PLOS ONE Spatiotemporal analysis of COVID-19 medical resource deficiencies
14. Ventilator Stockpiling and Availability in the US [Internet]. Johns Hopkins Center for Health Security.
2020 [cited April 26, 2020]. Available from: https://www.centerforhealthsecurity.org/resources/COVID-
19/COVID-19-fact-sheets/200214-VentilatorAvailability-factsheet.pdf.
15. Malatino E. Strategic national stockpile: overview and ventilator assets. Respir Care. 2008; 53(1):91–5.
PMID: 18173863
16. Ajao A, Nystrom SV, Koonin LM, Patel A, Howell DR, Baccam P, et al. Assessing the Capacity of the
US Health Care System to Use Additional Mechanical Ventilators During a Large-Scale Public Health
Emergency. Disaster Med Public Health Prep. 2015; 9(6):634–41. https://doi.org/10.1017/dmp.2015.
105 PMID: 26450633
17. Branas CC, Rundle A, Pei S, Yang W, Carr BG, Sims S, et al. Flattening the curve before it flattens us:
hospital critical care capacity limits and mortality from novel coronavirus (SARS-CoV2) cases in US
counties. medRxiv. 2020:2020.04.01.20049759. https://doi.org/10.1101/2020.04.01.20049759
18. Giannakeas V, Bhatia D, Warkentin MT, Bogoch II, Stall NM. Estimating the Maximum Capacity of
COVID-19 Cases Manageable per Day Given a Health Care System’s Constrained Resources. Annals
of Internal Medicine. 2020. https://doi.org/10.7326/m20-1169 PMID: 32298412
19. U.S. Census Bureau. TIGER/Line Shapefiles and TIGER/Line Files. [Internet]. 2020 [cited April 29,
2020]. Available from: https://www2.census.gov/geo/tiger/TIGER2019/
20. Where The ICU Beds Are [Internet]. 2020 [cited April 25, 2020]. Available from: https://khn.org/news/
as-coronavirus-spreads-widely-millions-of-older-americans-live-in-counties-with-no-icu-beds/.
21. Definitive Healthcare: USA Hospital Beds [Internet]. 2020 [cited April 27, 2020]. Available from: https://
www.definitivehc.com/.
22. Hospitals [Internet]. 2019 [cited April 25, 2020]. Available from: https://hifld-geoplatform.opendata.
arcgis.com/datasets/hospitals.
23. NPPES. NDI Registry Database. In: System NPPE, editor. 2020.
24.
Liu Q, Sha D, Liu W, Houser P, Zhang L, Hou R, et al. Spatiotemporal Patterns of COVID-19 Impact on
Human Activities and Environment in Mainland China Using Nighttime Light and Air Quality Data.
Remote Sensing. 2020; 12(10):1576.
25. Coronavirus Locations: COVID-19 Map by County and State [Internet]. 2020 [cited March 1, 2020].
Available from: https://usafacts.org/visualizations/coronavirus-covid-19-spread-map/.
26.
The COVID Tracking Project: Most recent data [Internet]. 2020 [cited April 29, 2020]. Available from:
https://covidtracking.com/data.
27. Medical centers in the United States [Internet]. 2020 [cited May 3, 2020]. Available from: https://en.
wikipedia.org/wiki/Medical_centers_in_the_United_States#Cleveland_Clinic.
28. Mississippi COVID-19 cases now number more than 5,000 [Internet]. 2020 [cited May 10, 2020]. Avail-
able from: https://www.djournal.com/news/mississippi-covid-19-cases-now-number-more-than-5-000/
article_b11840b6-21ed-51f6-8073-8b78d323b6e3.html.
29. Staff. 115 new confirmed coronavirus cases in East Baton Rouge, part of sharp rise; see latest state-
wide data. The Advocate2020.
30. Marema T, Murphy T. Covid-19 Update: April Saw Rural America’s Infection Rate Increase 8-Fold. The
Daily Yonder2020.
31. Colias-Pete M. Indiana passes 1,000 COVID-19 deaths; Lake County reports four more deaths, brings
total to 81. Chicago Tribune2020.
32.
Large spike in coronavirus cases traced to prison in Trousdale County [Internet]. 2020. Available from:
https://communityimpact.com/nashville/southwest-nashville/coronavirus/2020/05/01/large-spike-in-
coronavirus-cases-traced-to-prison-in-trousdale-county/.
PLOS ONE | https://doi.org/10.1371/journal.pone.0240348 October 14, 2020
19 / 19
PLOS ONE |
10.1371_journal.pone.0240029 | RESEARCH ARTICLE
Genetic characterization for lesion mimic and
other traits in relation to spot blotch
resistance in spring wheat
Shweta SinghID
Ram Narayan Ahirwar2, Dwijesh Chandra Mishra3, Sundeep Kumar4, Ramesh Chand5,
Uttam KumarID
6, Suneel Kumar4, Arun Kumar Joshi7,8
2*, Ravindra Nath Kharwar1, Neeraj Budhlakoti3,
1, Vinod Kumar MishraID
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
1 Department of Botany, Institute of Science, Banaras Hindu University, Varanasi, India, 2 Department of
Genetics and Plant Breeding, Institute of Agricultural Sciences, Banaras Hindu University, Varanasi, India,
3 ICAR-Indian Agricultural Statistics Research Institute, New Delhi, India, 4 ICAR-National Bureau of Plant
Genetic Resources, New Delhi, India, 5 Department of Mycology and Plant Pathology, Institute of Agricultural
Sciences, Banaras Hindu University, Varanasi, India, 6 Borlaug Institute for South Asia (BISA), Ladhowal,
Ludhiana, Punjab, India, 7 International Maize and Wheat Improvement Center (CIMMYT), New Delhi, India,
8 Borlaug Institute for South Asia (BISA), New Delhi, India
* vkmbhu@gmail.com
OPEN ACCESS
Abstract
Lesion mimic (Lm) mutants display hypersensitive responses (HR) without any pathogen
attack; their symptoms are similar to those produced by a pathogen and result in cell death.
In wheat, such mutants have been reported to be resistant against leaf rust due to their bio-
trophic nature. However, Lm mutants tend to encourage spot blotch (SB) disease caused by
Bipolarissorokiniana since dead cells facilitate pathogen multiplication. In this study, 289
diverse wheat germplasm lines were phenotyped in three consecutive growing seasons
(2012–2015). Genotype data was generated using the Illumina iSelect beadchip assay plat-
form for wheat germplasm lines. A total of 13,589 single-nucleotide polymorphisms (SNPs)
were selected andused for further association mapping. Lm was positively associated with
Area Under Disease Progress Curve (AUDPC) for SB but negatively with glaucous index
(GI), leaf tip necrosis (Ltn) and latent period (LP). Ltn had a negative association with
AUDPC and Lm but a positive one with LP. In a genome-wide association study (GWAS),
29 markers were significantly associated with these traits and 27 were an notated. Seven
SNP markers associated with Lm were on chromosome 6A; another on 1B was found to be
linked with Ltn. Like wise, seven SNP markers were associated with GI; one on chromo-
some 6A with the others on 6B. Five SNP markers on chromosomes 3B and 3Dwere signifi-
cantly correlated with LP, while nine SNP markers on chromosomes 5A and 5B were
significantly associated with AUDPC for SB. This study is the first to explore the interaction
in wheat between Lm mutants and the hemibiotrophic SB pathogen B.sorokiniana.
Citation: Singh S, Mishra VK, Kharwar RN,
Budhlakoti N, Ahirwar RN, Mishra DC, et al. (2020)
Genetic characterization for lesion mimic and other
traits in relation to spot blotch resistance in spring
wheat. PLoS ONE 15(10): e0240029. https://doi.
org/10.1371/journal.pone.0240029
Editor: Manoj Prasad, National Institute of Plant
Genome Research, INDIA
Received: July 14, 2019
Accepted: September 18, 2020
Published: October 5, 2020
Peer Review History: PLOS recognizes the
benefits of transparency in the peer review
process; therefore, we enable the publication of
all of the content of peer review and author
responses alongside final, published articles. The
editorial history of this article is available here:
https://doi.org/10.1371/journal.pone.0240029
Copyright: © 2020 Singh et al. This is an open
access article distributed under the terms of the
Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: All relevant data are
uploaded to the CIMMYT repository and publicly
accessible via the following URL: http://hdl.handle.
net/11529/10714 (Hdl/11529/10714).
PLOS ONE | https://doi.org/10.1371/journal.pone.0240029 October 5, 2020
1 / 16
PLOS ONE Funding: The Author (Shweta Singh) is grateful to
the University Grant Commission (UGC), New Delhi
for providing JRF, SRF and financial assistance.
The funder had no role in study design, data
collection and analysis, decision to publish, or
preparation of the manuscript.
Competing interests: The authors have declared
that no competing interests exist.
Genetic characterization for lesion mimic and spot blotch resistance in wheat
Introduction
Plant lesion mimic(Lm) mutants exhibitnecrotic symptoms on the leaves that appear in the
absence of any pathogen. These symptoms mimic the HR displayed during plant-pathogen
interactions [1]. First reported in barley [2,3], this type of mimicry has also been witnessed in
Arabidopsis [4–6], maize [7,8], rice [9–11], and wheat [12,13]. Lesion mimics form patches of
dead cells without any natural wound, injury, stress or infection being present in the plants.
HR-associated cell death arrests the growth of biotrophic pathogens by restricting the supply
of essential nutrients from the host. However, it confers no resistance to hemibiotrophs or
necrotrophic pathogens [14] such as Botrytis cinerea [15], where cell death follows infection
[16]. Another trait, leaf tip necrosis (Ltn), provides resistance against various plant pathogens
at the adult plant stage and is linked to genes such as Lr34 [17]. This gene was first reported in
wheat PI58548 [18] and was later described in many wheat varieties [19–21]. It slows develop-
ment of rusts and, under suitable conditions, has the ability in seedlings to provide resistance
to certain rust races including leaf rust (Pucciniatriticina) and stripe rust (P.striiformis) [19].
Wheat genotypes possessing Lr34 also show resistance against SB caused by hemibiotroph
pathogen B.sorokiniana [17]. SB progress may also be inhibited by other components of resis-
tance such as increased latent period (LP) [17] and glaucous index (GI) or waxiness. Disease
potential of the crop may be reduced by utilizing genotypes with a long LP [17] and a high GI.
Wheat SB causes average yield reductions in South Asia and India of 19.6% and 15.5%
respectively [22]. Losses may be20–80% in susceptible genotypes [23] and complete failure can
occur with the most severe infections [24]. B.sorokiniana infects most Poaceae family crops,
but even though it can infect a vast range of cultivars including wild and cultivated varieties,
the chances of the migration of an isolate from one crop to another are remote because the
causal pathogen is mainly seed born [25,26].
Because Lm checks the growth of stem rust pathogen, research using molecular markers
has been undertaken to find robust Quantitative Trait Loci (QTLs) against stem rust [27]. This
is in contrast to the case of Lm and SB where the phenotype-genotype association is not well
understood. To our knowledge associations between Lm, Ltn, GI, LP, and SB have not previ-
ously been studied.
Because lesion mimics restrict the growth of biotrophic pathogens such as rusts, Lm genes
are being introduced into wheat cultivars to achieve a degree of immunity against these patho-
gens. However, the cell death of leaves in response to the Lm genes expression is a major draw-
back, as this provides suitable conditions for the growth of hemi biotroph and necrotrophs
pathogens. Therefore it is important for wheat-cultivating areas globally to establish the effects
of Lm genes on SB, which is caused by a hemibiotrophic fungus.
For complex traits, association mapping can identify significant correlations between phe-
notypes and the corresponding sequence variants within an existing diversity panel [28]. The
present study was undertaken to characterize Lm, Ltn, GI, and LP, and their association with
SB resistance in spring wheat.
Materials and methods
Plant material
The Wheat Association Mapping Initiative (WAMI) panel of 289 diverse wheat germplasm
lines was obtained from the Global Wheat Program (CIMMYT, Mexico). It contains a wide
range of genotypic and phenotypic genotypes which are stable for the traits under examina-
tion. These lines were evaluated for Lm, Ltn, LP, GI and AUDPC for SB. The details of the
germplasm lines used are given in S3 Table.
PLOS ONE | https://doi.org/10.1371/journal.pone.0240029 October 5, 2020
2 / 16
PLOS ONE Genetic characterization for lesion mimic and spot blotch resistance in wheat
Sowing and maintenance of crop under experimental field
The research trials took place at the Agricultural Research Farm, Institute of Agricultural Sci-
ences, Banaras Hindu University, Varanasi, India (25o15’ N, 25o15’, 83o03’E; 70 m above sea
level) during three consecutive crop seasons, 2013–14, 2014–15 and 2015–16. Planting was
done between26th November and 5th December in each crop season to ensure that grain filling
coincided with local high temperatures and relative high humidity. Wheat genotypes were
sown in two replicates each year in an alpha lattice design. Each genotype was sown in two
1-metre rows, with a row-to-row distance of 25 cm and a plant-to-plant distance of 5 cm.
Agronomic practices recommended for normal fertility conditions for irrigated wheat were
followed for all three crops; 120 kg N, 60 kg P2O5 and 40 kg K2O ha-1.
Inoculation of the pathogen
A pure culture of B.sorokiniana (HD 3069/MCC 1572) for artificial inoculation was obtained
from the Department of Mycology and Plant Pathology, Institute of Agricultural Sciences,
Banaras Hindu University, Varanasi [29]. Following multiplication of the isolate on sorghum
grain, a suspension in water of 104 spores/ml was applied uniformly at the heading stage [30],
with the spraying being done in the evening [31]. The field was irrigated the next morning to
provide a favourable environment for disease development.
Scoring for lesion mimic, leaf tip necrosis, SB, glaucous index and latent period
The plants were observed in the experimental plot for the traits Lm, Ltn, SBGI and LP. Five
randomly tagged plants of each genotype were evaluated for the expression of Lm symptoms.
Lm was scored on flag leaf at growth stages (GS) 63, 69 and 77. Leaves showing typical Lm
expression were scored with modification of the 1–9 rating scale [13]. Within the scale, the %
area denotes the leaf area necrosis, where 1 = no visible specks, 2 = 1–10%, 3 = 21–30%,
4 = 31–40%, 5 = 41–50%, 6 = 51–60%, 7 = 61–70%, 8 = 71–80% and 9 = more than 80%. Flag
leaves of the tagged plants of each genotype were evaluated for Ltn at GS69. For Ltn, leaves
were scored in two ways—first, as a presence or absence of Ltn; second, when present, scored
as the level of its expression on a scale of 1–5, where 1 = no Ltn, 2 = 25%, 3 = 50%, 4 = 75% and
5 = necrosis of more than 75% of the flag leaf. Glaucous or waxiness on the 5 tagged plants was
recorded visually at the time of flowering on the peduncle and flag leaf sheath on a scale of
1–5. Here, 1 denotes a very low or minimum appearance of waxiness, 2 denotes low waxiness
appearance, 3 denotes a comparatively moderate level of waxiness, 4 denotes a high level of
waxiness appearance while 5 indicates a maximum level of waxiness.
LP is the period in days between inoculation and spore production and was observed and
recorded using the process described by Parlevliet [32]. Five randomly selected flag leaves
from each tagged plant were examined with the aid of a 20 × magnifying lens to establish when
50% of the primary lesions were sporulated.
Disease assessment
Ten randomly tagged plants of each genotype were evaluated for SB severity at three different
growth stages, GS63 (beginning to half-completion of anthesis), GS69 (anthesis complete) and
GS77 (late milking) using a double-digit scale (DD, 00–99) according to Saari and Prescott
[33]. For each score, the disease severity percentage was calculated using the formula:
%severity ¼ ðD1=9ÞðD2=9Þ100
PLOS ONE | https://doi.org/10.1371/journal.pone.0240029 October 5, 2020
3 / 16
PLOS ONE Genetic characterization for lesion mimic and spot blotch resistance in wheat
Where,
D1 ¼ vertical disease progress on the plant
D2 ¼ the disease severity score on the affected leaves
AUDPC was based on disease severity at GS63, GS69 and GS77 using the percent severity
estimates as outlined in [34], given as:
AUDPC ¼
Xn(cid:0) 1
i¼0
½fðYi þ Yiþ1Þ=2g � ðtiþ1(cid:0) tiÞ�
Where,
Yi ¼ disease severity at time ti
ðtiþ1(cid:0) tiÞ ¼ time interval ðdaysÞ between two disease scores
n ¼ thenumber of dates at which SB was recorded
DNA extraction and SNP genotyping
DNA was extracted from 20-day fresh leaves of each line following the CTAB procedure [35]
and genotyped at CIMMYT, Mexico using the Illumina iSelect beadchip assay [36] for wheat.
To avoid low polymorphic and low-quality SNPs, markers were filtered on the parameter of
minor allele frequency < 0.10. Thus 13,589 out of a total of 15,737 highly polymorphic SNPs
were selected and used for association mapping.
Phenotypic and population structure analysis
Analysis of Variance (ANOVA) was carried out to determine genotype, year, and
genotype × year variances among the traits measured. Correlation analysis was performed to
better understand the relationship among the traits. All these analyses were done using SAS
9.3. The population structure (Q) for the WAMI marker panel was determined using the pro-
gram STRUCTURE v2.3.4 [37]. The number of clusters (K) was predefined as1–10 with a
burn-in of 10,000 iterations followed by 10,000 Markov Chain Monte Carlo (MCMC) repli-
cates, passed as initial parameters for running STRUCTURE. The number of subgroups of the
population was estimated using ‘Structure Harvester’ [38], a web-based utility that provides
maximum likelihood estimates of the proportion of each sample derived from each of the K
populations. The population Q-matrixwas also obtained for further analysis.
Genome-wide association analysis
TASSEL 5.0 [39] was used for the identification of significant marker-trait associations, based
on the Mixed Linear Model (MLM). MLM takes into account both the population structure
(Q-matrix generated through STRUCTURE) as well as the ancestral relatedness i.e. kinship
matrix (K). TASSEL 5.0 was used to calculate the population kinship matrix by applying a
scaled Identity By State (IBS) method. The general mathematical formulation of this mixed lin-
ear model can be written in the following form:
y ¼ Xa þ Qb þ Ku þ e
PLOS ONE | https://doi.org/10.1371/journal.pone.0240029 October 5, 2020
4 / 16
PLOS ONE Genetic characterization for lesion mimic and spot blotch resistance in wheat
Where,
y ¼ the vector of phenotypes
a and b ¼ vectors of fixed effects
u ¼ the vector of random effects ðKinship matrixÞ
e ¼ the vector of random residuals
X ¼ the genotypes of marker
Q ¼ the population structure
K ¼ kinship matrix
Since the Q-matrix is used as a covariate in the model, it controls the structure and also avoids
false positives. MLM is used preferentially because of its efficiency in terms of reducing time com-
plexity [40]; its parameters were left at the default settings when running TASSEL. AP-value
�0.001 was taken as denoting a significant marker-trait association (MTA) and the R2 value was
used to evaluate the magnitude of the QTL effects. For better visualization of results, Manhattan
plots were also generated. Linkage distribution among the markers was also calculated.
Results
Phenotypic analysis
The results of the ANOVA for the five measured traits of the WAMI panel over three consecu-
tive growing seasons are presented in Table 1. It can be observed that the genotypes exhibit dif-
ferences at a 1% level of significance for all the traits. The year was also found to be significant
for all traits except Ltn. Moreover, significant differences were observed for genotype × year
(P<0.01) for all the traits. Partitioning of the total sum of squares indicates that the year
accounted for more variation than genotype for all the traits except Ltn. In addition, AUDPC
for SB was negatively and significantly correlated with Ltn (-0.302) and LP (-0.529), while posi-
tively and significantly with Lm (0.493);Ltn is positively correlated with LP (0.313) but nega-
tively with Lm (-0.426) (Table 2).
Population structure and linkage disequilibrium analysis
From model-based analysis using STRUCTURE (Figs 1 & 2) the optimal K was determined to
be 6. Subpopulation I contained 58 (20.1%) genotypes; II, 56 (19.3%); III, 74 (25.6%); IV, 35
(12.1%); V, 48 (16.7%); and VI, 18 (6.2%). Individuals of each population were categorized as
pure or admixture types. Genotypes with�0.8 of member proportions were considered as
pure were others were labeled admixtures. Considering this criterion, the composition of the
six subpopulations was as follows; I, 2.8% pure and 17.3% admixture; II, 4.8% pure and 14.5%
Table 1. Analysis of variance for the traits lesion mimic, leaf tip necrosis, latent period, glaucous index and area under disease progress curve.
Source
Df
Lm
Mean square
2199.76�
1890.15�
5.20
146.95�
Genotype
Year
Replication
288
2
1
Genotype × year
576
�Significant at P<0.01.
Ltn
F Value Mean square
61.54
52.88
0.15
4.11
648.95�
117.87
36.04
70.65�
https://doi.org/10.1371/journal.pone.0240029.t001
LP
GI
AUDPC
F Value Mean square
F Value Mean square
19.30
3.51
1.07
2.10
14.87�
206.35�
0.27
2.02�
14.99
207.86
4.94
2.06
2.37�
3.63�
1.80
0.75�
F Value Mean square
123267.82�
3170105.74�
6.49
9.92
4.94
2.06
2902.96
7135.74�
F Value
99.50
2558.98
2.34
5.76
PLOS ONE | https://doi.org/10.1371/journal.pone.0240029 October 5, 2020
5 / 16
PLOS ONE Genetic characterization for lesion mimic and spot blotch resistance in wheat
Table 2. Pearson correlation coefficients analysis for the five traits studied.
Traits
AUDPC
GI
Ltn
Lm
GI
Ltn
Lm
LP
-0.052 (0.02)
-0.302 (< .001)
0.493 (< .001)
-0.529 (< .001)
1
0.009 (0.0681)
-0.092 (0.0001)
0.063 (0.0072)
1
-0.426 (< .0001)
0.313 (< .0001)
1
-0.465 (< .0001)
AUDPC = Area under disease progress curve; GI = Glaucousness index; Ltn = Leaf tip necrosis; Lm = Lesion mimic; LP = Latent period. P-values are given in
parentheses.
https://doi.org/10.1371/journal.pone.0240029.t002
admixture; III, 3.1% pure and 22.5% admixture; IV, 1.4% pure and 10.7% admixture; V, 4.2%
pure and 12.5% admixture; and VI, 1.4% pure and 4.8% admixture. Fig 2 shows the population
structure. To evaluate the population composition, Q-matrix (K = 6) and Kinship matrix were
further used as covariates for a GWAS. A total of 13,589 SNPs markers that passed quality fil-
tering were used for mapping. Of these SNPs, 4,967 had loci mapped on A genome, 7,236 on B
genome and 1,386 on D genome (S2 Table). A Linkage Disequilibrium (LD) plot based on the
association among the markers was generated (Fig 3). LD was estimated from all pairs of SNPs
along each chromosome. The average LD R2 was 0.35 for the A sub-genome, 0.37 for B and
0.36 for D. In Fig 3 most of the markers are tightly linked and observed below the diagonalas
large areas of red. This denotes that there was restricted space for recombination between the
markers, which facilitates association mapping of the five traits, and suggests that a minimum
number of markers is required effectively to cover the entire genome. A more detailed distri-
bution of SNPs over chromosomes is presented in S1 Table.
Genome-wide marker-trait association with SNP-markers
A total of 29 SNPs exhibited significant marker-trait associations with Lm, GI, Ltn, LP and SB
AUDPC at P<0.001. They occur across seven different genomic regions (1B, 3B, 3D, 5B, 5A,
Fig 1. Population structure showing genetic relationships of 289 wheat lines. ΔK plot, with K ranging from 1–10.
https://doi.org/10.1371/journal.pone.0240029.g001
PLOS ONE | https://doi.org/10.1371/journal.pone.0240029 October 5, 2020
6 / 16
PLOS ONE Genetic characterization for lesion mimic and spot blotch resistance in wheat
Fig 2. STRUCTURE analysis used to define genetic relationships among 289 wheat lines. The existence of six
subpopulations was inferred. Plot was generated using the mean of the variation posterior distribution over inferred
admixture proportions. The X-axis shows the membership coefficients and Y-axis shows the different genotype entries.
A visual vertical separation represents different subpopulations.
https://doi.org/10.1371/journal.pone.0240029.g002
6A and 6B). Seven SNPs each were found for Lm and GI, one SNP for Ltn, five SNPs for LP
and nine SNPs for SB AUDPC. The P- and R2-values, and other details of these marker-trait
associations are presented in Table 3. The Manhattan plots for each trait are shown in Fig 4.
Individually the SNPs explain5-8% of the total phenotypic variation.
For lesion mimic a total of seven significant marker trait associations (MTAs) were mapped
on chromosome 6A with four situated at 90 cM and the remaining three at 91cM. Each of
these markers explained 6–7% of total phenotypic variation. A single SNP marker was signifi-
cantly associated with Ltn and mapped on chromosome 1B at a distance of 115 cM. This SNP
explained 6% of total phenotypic variation. Seven SNPs were found to be in significant
marker-trait association with GI. One marker was on chromosome 6A at 21 cM. All others
Fig 3. Linkage disequilibrium plot of significant marker associations. R2 and P-values of pair-wise analyses are
indicated by colour in the right-side bars.
https://doi.org/10.1371/journal.pone.0240029.g003
PLOS ONE | https://doi.org/10.1371/journal.pone.0240029 October 5, 2020
7 / 16
PLOS ONE Table 3. List of significant SNPs associated with five different traits—Lm, Ltn, GI, LP and AUDPC—detected in the 289 WAMI spring wheat panel.
Genetic characterization for lesion mimic and spot blotch resistance in wheat
Marker
Chr.
Trait
Lm
Ltn
GI
wsnp_CAP11_c1178_684471
wsnp_Ra_c12086_19452422
Tdurum_contig69065_319
Tdurum_contig55363_297
wsnp_Ku_rep_c102901_89769309
wsnp_RFL_Contig3136_3092151
Tdurum_contig29974_90
Ex_c25733_348
BobWhite_c3714_659
CAP7_c524_326
Kukri_rep_c79491_139
TA001682-1583
RAC875_c17011_373
RAC875_c21938_1408
TA002907-0816
LP
RAC875_c4389_1344
RAC875_c4389_1412
tplb0043c20_1046
tplb0043c20_1046–1
GENE-1851_76
AUDPC
wsnp_Ku_c40334_48581010
BobWhite_c48435_165
Tdurum_contig12066_126
Tdurum_contig12066_247
Tdurum_contig12066_126–1
Tdurum_contig12066_247–1
tplb0027f13_1493
tplb0027f13_1346
tplb0027f13_1346–1
https://doi.org/10.1371/journal.pone.0240029.t003
6A
6A
6A
6A
6A
6A
6A
1B
6A
6B
6B
6B
6B
6B
6B
3B
3B
3B
3D
3B
5B
5B
5A
5A
5B
5B
5B
5A
5B
Pos
90
91
91
90
91
90
90
115
21
118
118
119
122
120
122
11
32
26
18
26
90
90
83
83
90
90
90
83
90
P value
2.39E-05
2.88E-05
3.16E-05
3.36E-05
4.1E-05
4.37E-05
5.67E-05
7.69E-04
4.55E-06
5.85E-06
1.14E-05
3.8E-05
4.17E-05
4.87E-05
8.65E-05
8.83E-06
1.02E-05
3.38E-05
3.38E-05
8.36E-05
3.9E-05
6.53E-05
7.01E-05
7.01E-05
7.01E-05
7.01E-05
7.85E-05
8.01E-05
8.01E-05
Marker R2
0.0694
0.0641
0.0665
0.0635
0.06181
0.06497
0.06444
0.05931
0.07696
0.07547
0.07075
0.06106
0.06214
0.06042
0.05608
0.07531
0.07691
0.06421
0.06421
0.05819
0.06109
0.05752
0.05716
0.05716
0.05716
0.05716
0.05702
0.05616
0.05616
were on chromosome 6B. Two of these were each at 118 cM, another two were at 119 and 120
cM, while the remaining two were mapped at a distance of 122 cM. The phenotypic variations
explained by these loci were in the range 5–8%. For LP, five significant SNPs were identified
on chromosome 3B and 3D. Two mapped together at 26cM on chromosome 3B which sug-
gests that a single QTL for LP may link to these SNPs. Another two occurred on chromosome
3B at 11 cM and 32 cM respectively, while the remaining marker was on chromosome 3D at
18 cM. Each explained 6–8% of total genetic variation. A total of nine SNPs on chromosomes
5A and 5B were found to be in significant marker-trait associations for AUDPC (Table 3). Six
of these SNPs were on chromosome 5B, each at 90 cM; the other three clustered on 5A at 83
cM. It is possible that two QTLs for SB resistance may link to these SNPs. The phenotypic vari-
ation explained by each individual locus was around 6%.
Identification of putative candidate genes and their annotation
To analyze and further annotate significant MTAs, we located them on a reference wheat
genome (RefSeq v1.0). Due to the large size of the wheat genome, extended 250 kb genomic
regions either side of significantly associated SNPs were analyzed to identify putative genes. A
PLOS ONE | https://doi.org/10.1371/journal.pone.0240029 October 5, 2020
8 / 16
PLOS ONE Genetic characterization for lesion mimic and spot blotch resistance in wheat
Fig 4. Manhattan plots for different traits under study: a) GI b) LM c) LP d) Ltn e) AUDPC. The threshold line at P = 0.001 has
been drawn to highlight significant markers.
https://doi.org/10.1371/journal.pone.0240029.g004
total of 27 candidate genes were annotated by function (Table 4). The MTAs that could not be
annotated were Kukri_rep_c79491_139, associated with GI, and Tdurum_contig_12066_126
associated with AUDPC.
For trait Lm, markers mapped within a very narrow interval of 0.09 Mbpon chromosome
6A, in the region from 56.34 Mbp (wsnp_Ku_rep_c102901_89769309 and
wsnp_CAP11_c1178_684471) to 56.43 Mbp (Tdurum_contig55363_297). This interval con-
tains seven genes, of which five have annotations with high confidence (Table 4). Few of the
candidates genes having more than one SNPs at same chromosome position like TraesC-
S6A02G33100 (wsnp_Ra_c12086_19452422, wsnp_Ku_rep_c102901_89769309) and TraesC-
S6A02G331000(wsnp_CAP11_c1178_684471, wsnp_RFL_Contig3136_3092151) at 91 and 90
cM respectively. These genes encode for various classes of proteins and enzymes including a
WD40-repeat-containing domain superfamily member (which regulates a plant-specific devel-
opmental event to control cell cycle) and MAG2-interacting protein (which acts as a precursor
for the accumulation in dry seeds of the two major storage proteins albumin 2S and globulin
12S. The single marker identified for Ltn was located on chromosome 1B and belongs to the
protein kinase-like superfamily, which is common to both serine/threonine and tyrosine pro-
tein kinases and has a catalytic domain that contains anucleotide-binding site (NBS), playing a
critical role in disease resistance.
Similarly, for trait LP, most markers mapped on to chromosome 3B in a 0.21 Mbp interval
at 0.90–1.11 Mbp. Likewise for lesion mimic here alsowe found two SNPs falling under same
candidate gene viz. TraesCS3B02G025200 (tplb0043c20_1046, GENE-1851_76). Their under-
lying identified genes were found to relate to the Cytochrome P450 superfamily and Fructose-
bisphosphate aldolase, which are involved in specific mechanisms like stress and defense
response, energy and metabolism.
For GI, six of the seven MTAs mapped on to chromosome 6B. The identified genes related
to Ubiquitin-conjugating enzyme/RWD-like proteins (with functions involved with plant
innate immunity), Synaptotagmin-like mitochondrial-lipid-binding protein (that acts as
PLOS ONE | https://doi.org/10.1371/journal.pone.0240029 October 5, 2020
9 / 16
PLOS ONE Table 4. Detailed annotation (i.e. underlying genes, their functions and GO terms) of identified markers for different traits related to spot blotch.
Genetic characterization for lesion mimic and spot blotch resistance in wheat
Markers
Chromosome Gene accession
Gene Descriptor
TraesCS6A02G331000 Mitochondrial carrier domain
superfamily
GO Terms
GO:0016021
wsnp_CAP11_c1178_684471
wsnp_Ra_c12086_19452422
Tdurum_contig69065_319
Tdurum_contig55363_297
6A
6A
6A
6A
TraesCS6A02G333100 WD40-repeat-containing domain
superfamily, MAG2-interacting protein
GO:0005737, GO:0006888, GO:0006890,
GO:0032527
TraesCS6A02G333400 Glycoside hydrolase, family 5
TraesCS6A02G332000 Mitochondrial carrier domain
GO:0016021
superfamily
wsnp_Ku_rep_c102901_89769309 6A
TraesCS6A02G333100 WD40-repeat-containing domain
superfamily, MAG2-interacting protein
wsnp_RFL_Contig3136_3092151
6A
TraesCS6A02G331000 Mitochondrial carrier domain
superfamily
GO:0005737, GO:0006888, GO:0006890,
GO:0032527
GO:0005618, GO:0005774, GO:0005794,
GO:0031305, GO:0005315, GO:0009651,
GO:0035435
Tdurum_contig29974_90
Ex_c25733_348
BobWhite_c3714_659
CAP7_c524_326
Kukri_rep_c79491_139
TA001682-1583
RAC875_c17011_373
RAC875_c21938_1408
TA002907-0816
RAC875_c4389_1344
RAC875_c4389_1412
tplb0043c20_1046
tplb0043c20_1046–1
GENE-1851_76
wsnp_Ku_c40334_48581010
BobWhite_c48435_165
Tdurum_contig12066_126
Tdurum_contig12066_126–1
Tdurum_contig12066_247
Tdurum_contig12066_247–1
tplb0027f13_1493
tplb0027f13_1346
tplb0027f13_1346–1
6A
1B
6A
6B
6B
6B
6B
6B
6B
3B
3B
3B
3D
3B
5B
5B
5A
5B
5A
5B
5B
5A
5B
https://doi.org/10.1371/journal.pone.0240029.t004
TraesCS6A02G331500 Cyclin-like superfamily
TraesCS1B02G419400 Protein kinase-like domain superfamily
TraesCS6A02G016200 Ubiquitin-conjugating enzyme/RWD-
like
TraesCS6B02G463000
Synaptotagmin-like mitochondrial-lipid-
binding domain
GO:0016021, GO:0008289, GO:0006869
NA
TraesCS6B02G462300 Haem peroxidase superfamily
TraesCS6B02G472900 Lunapark family
GO:0016021, GO:0071786
TraesCS6B02G465300 AP-5 complex subunit beta-1
GO:0005623, GO:0016021, GO:0016197
TraesCS6B02G471900 Domain unknown function DUF295
TraesCS3B02G025600 Cytochrome P450 superfamily
TraesCS3B02G025600 Cytochrome P450 superfamily
GO:0004497, GO:0005506, GO:0016705,
GO:0020037, GO:0055114
GO:0004497, GO:0005506, GO:0016705,
GO:0020037, GO:0055114
TraesCS3B02G025200 Fructose-bisphosphate aldolase, class-I
GO:0004332, GO:0006096, EC:4.1.2.13
TraesCS3D02G026400 Fructose-bisphosphate aldolase
GO:0004332, GO:0006096, EC:4.1.2.13
TraesCS3B02G025200 Aldolase-type TIM barrel
TraesCS5B02G368600
S-acyltransferase
GO:0005829, GO:0016021, GO:0004332,
GO:0006096, GO:0030388, EC:4.1.2.13
GO:0016021, GO:0019706, EC:2.3.1.225
TraesCS5B02G368500 Potassium transporter
GO:0016021, GO:0015079, GO:0071805
TraesCS5A02G366100 Potassium transporter
GO:0016021, GO:0015079, GO:0071805
NA
TraesCS5A02G366100 Potassium transporter
TraesCS5B02G368500 Potassium transporter
TraesCS5B02G368500 Potassium transporter
GO:0016021, GO:0015079, GO:0071805
TraesCS5A02G366100 Potassium transporter
GO:0016021, GO:0015079, GO:0071805
TraesCS5B02G368500 Potassium transporter
GO:0016021, GO:0015079, GO:0071805
molecular hubs for the exchange of small molecules such as lipids, and of signals, such as cal-
cium ions), haem peroxidase superfamily members (which act as an electron acceptor to cata-
lyze several oxidative reactions), the Lunapark family, and AP-5 complex subunit beta-1 (i.e.
floral organ development and plant reproduction).
A total of 9 SNPs were found to be in significant association to AUDPC of which 6
SNP mapped on chromosome 5B, four SNPs (BobWhite_c48435_165, tplb0027f13_1493,
Tdurum_contig12066_247–1 and tplb0027f13_1346–1) represents same candidate gene i.e.
TraesCS5B02G368500, whereas SNP wsnp_Ku_c40334_48581010 represent candidates
gene involved in S-acyltransferase (i.e. TraesCS5B02G368600) and remaining one
PLOS ONE | https://doi.org/10.1371/journal.pone.0240029 October 5, 2020
10 / 16
PLOS ONE Genetic characterization for lesion mimic and spot blotch resistance in wheat
Fig 5. Graphical representation of categories of Gene Ontology (GO) terms of identified MTA related to Spot blotch:
a) Cellular Component b) Biological Process c) Molecular Function.
https://doi.org/10.1371/journal.pone.0240029.g005
(Tdurum_contig12066_126–1) could not be annotated (Table 4). Remaining 3 SNPs for AUDPC
mapped on chromosome 5A (at 83 cM i.e. Tdurum_contig12066_126, Tdurum_contig12066_247
and tplb0027f13_1346) represent the same candidate gene i.e. TraesCS5A02G366100. Seven of
the nine MTAs identified were found to be involved in potassium transport which has a crucial
role in plant responses to, and tolerance of, abiotic stresses. Detailed annotations (i.e. cellular com-
ponent, biological process and molecular function) are shown in Fig 5.
Discussion
Lm in wheat is governed by four recessive genes—lm, lm1, lm2 [13] and lm3 [41]. Since they
are known to demonstrate a protective effect against biotrophic pathogens [12,27,41], Lm
genes are being introduced in wheat genotypes to provide resistance against these pathogens.
However, these genes express their symptoms as cell death in the leaf tissues of the host geno-
types. If this expression is severe, there may be a negative effect due to a reduced photosyn-
thetic area. In addition, the dead tissue might be a nutrient source for various hemibiotrophs
and necrotrophs. Therefore it is necessary to establish the effects on Lm plants of diseases like
SB, which is one of the major concerns for the Eastern Gangetic Plains (EGP) region of South
Asia covering the >10 Mha wheat belt of India, Nepal and Bangladesh [42–44].
Analysis of variance indicated the presence in the WAMI spring wheat panel of significant vari-
ability for Lm and the other traits investigated. Lm was positively correlated with AUDPC which
indicates that it promotes SB severity, confirming previous reports [45]. Lm appears to enhance
the impact of SB due to the hemibiotrophic nature of B.sorokiniana, which germinates on living
cells but multiplies on dead cells. Since Lm contributes to SB progress, reduced Lm expression can
control the necrotrophic action of the pathogen. The results of this study demonstrate that higher
LP is associated with lower levels of SB and Lm. Higher levels of GI can also restrict the progress of
SB. Therefore, for the development of SB-resistant genotypes, a combination of Ltn and GI with
higher LP may be utilized, which can reduce Lm gene expression as well as SB disease.
An inhibitory effect of Ltn on Lm was observed in this study. The association of LP with
Ltn was positive, which indicates that resistance against SB is enhanced by the presence of Ltn
[17]. Ltn can thus be recommended as a phenotypic marker for the selection of SB resistant
PLOS ONE | https://doi.org/10.1371/journal.pone.0240029 October 5, 2020
11 / 16
PLOS ONE Genetic characterization for lesion mimic and spot blotch resistance in wheat
genotypes. Recently, Singh et al. [46] proposed that stacking Lr34(Sb1), Lr46, and vrn-A1
(which are prevalent in the CIMMYT gene pool) with additional SB resistance QTL can lead
to a high level of SB resistance. Ltn in wheat is associated with Lr34 [47] and it can provide
resistance against rust as well as SB [17,47]. Glaucous index also showed a positive association
with Ltn indicating the cumulative positive effect of Lr34and increased waxiness on SB resis-
tance. For this reason, genotypes with well-expressed GI or waxiness also appear to inhibit the
expression of Lm. Similarly, LP displayed a negative association with Lm, and longer LP con-
tributed to enhanced levels of resistance against SB.
In this study, GWAS was performed using 90 k Illumina SNPs chip markers to establish the
genotypic relationships of five traits—Lm, Ltn, LP, GI, and AUDPC for SB. To validate the
broader applicability of SNPs and GWAS, we also verified the resistances that were previously
detected using DArT markers [48–50]. Nine QTLs relating to AUDPC were identified on
chromosomes 5A and 5B for improved resistance against SB.
In the association study of Adhikari et al. [49], genomic regions associated with SB resis-
tance led to the identification of nine SNPs on chromosomes 1B, 5A, 5B, 6B, and 7B. The
study used 528diverse spring wheat genotypes that were phenotyped for SB and genotyped uti-
lizing a 9K SNP wheat chip [51]. Ahirwar et al. [52] reported 14 SNPs on chromosomes 1B,
5B, 6A and 6B. We detected MTAs that corresponded to nine previously-reported loci on two
chromosomes [47,53,54] from biparental QTL analyses. The present study identified QTLs for
SB resistance that were reported previously at similar positions which confirms the robust
inheritance of QTLs associated with the SNPs mapped here.
This work is the first to study Lm and its association with LP, Ltn, GI, and AUDPC for SB.
Novel SNPs for these traits were firmly identified. The genotypic relationships among these
traits need to be further studied for improvement of the wheat research program. The reported
QTLs from this study should provide a foundation for further research in this area.
Conclusions
Twenty nine significant marker-trait associations were identified in the present investigation.
We found seven markers closely associated with Lm, all on chromosome 6A, one for Ltn on
1B, and seven for GI across 6A and 6B. Five markers for LP were mapped on chromosomes 3B
and 3D while nine SNPs on 5A and 5B were associated with AUDPC for SB.
A notable phenotypic and molecular variation was observed in the WAMI panel, which
confirms the diverse genetic background of the WAMI germplasm. The genotype analysis
showed significant positive correlations between Lm and AUDPC, Ltn and LP, and between
LP and GI, whereas Lm was significantly negatively associated with Ltn, LP, and GI.
This study established for the first time an association of markers for Lm, Ltn, LP and GI,
and QTLs mapped through GWAS. Our data revealed that most of the SNPs were present on
the A and B genomes of wheat. These identified SNP markers linked to different QTLs will be
useful in breeding for Lm and for SB resistance in wheat. The study also establishes a clear
association between Lm and Ltn with AUDPC for SB, GI and LP. Based on a positive associa-
tion of Lm and AUDPC, the pattern of appearance of HR, and necrosis and lesion formation
by Lm, it is evident that the structure, expression, function and pathways of Lm genes can pro-
vide useful information. Utilizing this information to better understand the nature of SB path-
ogens is will be critical in the development and selection of resistance cultivars.
Supporting information
S1 Table. Marker distribution among the population studied.
(DOCX)
PLOS ONE | https://doi.org/10.1371/journal.pone.0240029 October 5, 2020
12 / 16
PLOS ONE Genetic characterization for lesion mimic and spot blotch resistance in wheat
S2 Table. Distribution of 13589 highly polymorphic SNPS throughout the wheat genome.
(DOCX)
S3 Table. List of lines in the spring wheat association mapping (WAMI) evaluated over
three years in BHU, Varanasi.
(DOCX)
Acknowledgments
The authors gratefully acknowledge Matthew P. Reynolds, CIMMYT, Mexico, for providing
WAMI population and the open-access molecular data used in this study.
Author Contributions
Conceptualization: Vinod Kumar Mishra, Ravindra Nath Kharwar, Ramesh Chand, Arun
Kumar Joshi.
Data curation: Vinod Kumar Mishra, Neeraj Budhlakoti, Ram Narayan Ahirwar, Sundeep
Kumar, Ramesh Chand.
Formal analysis: Neeraj Budhlakoti, Ram Narayan Ahirwar, Dwijesh Chandra Mishra, Sun-
deep Kumar.
Investigation: Shweta Singh, Vinod Kumar Mishra, Ravindra Nath Kharwar.
Methodology: Shweta Singh, Vinod Kumar Mishra, Ravindra Nath Kharwar, Ramesh Chand,
Uttam Kumar, Arun Kumar Joshi.
Project administration: Vinod Kumar Mishra.
Resources: Vinod Kumar Mishra, Ramesh Chand, Uttam Kumar, Arun Kumar Joshi.
Software: Neeraj Budhlakoti, Dwijesh Chandra Mishra, Sundeep Kumar.
Supervision: Vinod Kumar Mishra, Arun Kumar Joshi.
Validation: Shweta Singh, Vinod Kumar Mishra, Ram Narayan Ahirwar.
Visualization: Shweta Singh, Ram Narayan Ahirwar, Dwijesh Chandra Mishra, Sundeep
Kumar.
Writing – original draft: Shweta Singh.
Writing – review & editing: Vinod Kumar Mishra, Ravindra Nath Kharwar, Neeraj Budhla-
koti, Ram Narayan Ahirwar, Ramesh Chand, Uttam Kumar, Suneel Kumar, Arun Kumar
Joshi.
References
1. Moeder W, Yoshioka K. Lesion mimic mutants: A classical, yet fundamental approach to study pro-
grammed cell death. Plant Signaling and Behavior. 2008; 3(10): 764–767. https://doi.org/10.4161/psb.
3.10.6545 PMID: 19513227
2. Wolter M, Hollricher K, Salamini F, Schulze-Lefert P. The mlo resistance alleles to powdery mildew
infection in barley trigger a developmentally controlled defence mimic phenotype. Mol. Gen Genet.
1993; 239:122–128. https://doi.org/10.1007/BF00281610 PMID: 8510641
3. Rostoks N, Schmierer D, Kudrna D, Kleinhofs A. Barley putative hypersensitive induced reaction
genes: genetic mapping, sequence analyses and deferential expression in disease lesion mimic
mutants. TheorAppl Genet. 2003; 107:1094–1101.
PLOS ONE | https://doi.org/10.1371/journal.pone.0240029 October 5, 2020
13 / 16
PLOS ONE Genetic characterization for lesion mimic and spot blotch resistance in wheat
4. Pilloff RK, Devadas SK, Enyedi A, Raina R. The Arabidopsis gain-of-function mutant dll1 spontaneously
develops lesions mimicking cell death associated with disease. Plant J. 2002; 30:61–70. https://doi.org/
10.1046/j.1365-313x.2002.01265.x PMID: 11967093
5.
6.
7.
Ishikawa A, Tanaka H, Nakai M, Asahi T. Deletion of a chaperonin 60 gene leads to cell death in the
Arabidopsis lesion initiation 1 mutant. Plant Cell Physiol. 2003; 44:255–261. https://doi.org/10.1093/
pcp/pcg031 PMID: 12668771
Ishikawa A. Tetrapyrrole metabolism is involved in lesion formation, cell death, in the Arabidopsis lesion
initiation 1 mutant. BiosciBiotechnolBiochem. 2005; 69:1929–1934.
Johal GS, Hulbert SH, Briggs SP. Disease lesion mimics of maize—a model for cell-death in plants.
Bioessays. 1995; 17:685–692.
8. Penning BW, Johal GS, McMullen MD. A major suppressor of cell death, slm1, modified the expression
of the maize (Zea mays L.) lesion mimic mutation les23. Genome. 2004; 47:961–969. https://doi.org/
10.1139/g04-046 PMID: 15499410
9.
Takahashi A, Kawasaki T, Henmi K, Shii K, Kodama O, Satoh H. et al. Lesion mimic mutants of rice
with alterations in early signalling events of defence. Plant J. 1999; 17:535–545. https://doi.org/10.
1046/j.1365-313x.1999.00405.x PMID: 10205906
10. Yin Z, Chen J, Zeng L, Goh M, Leung H, Khush GS. et al. Characterizing rice lesion mimic mutants and
identifying a mutant with broad-spectrum resistance to rice blast and bacterial blight. Mol Plant Microbe
Interact. 2000; 13:869–876. https://doi.org/10.1094/MPMI.2000.13.8.869 PMID: 10939258
11. Mori M, Tomita C, Sugimoto K, Hasegawa M, Hayashi N, Dubouzet JG et al. Isolation and molecular
characterization of a spotted leaf 18 mutant by modified activation-tagging in rice. Plant Mol Biol. 2007;
63:847–860. https://doi.org/10.1007/s11103-006-9130-y PMID: 17273822
12. Kamlofski CA, Antonelli E, Bender C, Jaskelio VM, Danna CH, Ugalde R, et al. A lesion-mimic mutant of
wheat with enhanced resistance to leaf rust. Plant Pathol. 2007; 56:46–54.
13. Yao Q, Zhou R, Fu T, Wu W, Zhu Z, Li A. et al. Characterization and mapping of complementary lesion
mimic genes lm1 and lm2 in common wheat. TheorAppl Genet. 2009; 119: 1005–1012. https://doi.org/
10.1007/s00122-009-1104-4 PMID: 19621213
14. Veronese P, Chen X, Bluhm B, Salmeron J, Dietrich R, Mengiste T. The BOS loci of Arabidopsis are
required for resistance to Botrytis cinerea infection. Plant J. 2004; 40:558–74. https://doi.org/10.1111/j.
1365-313X.2004.02232.x PMID: 15500471
15. Govrin EM, Levine A. The hypersensitive response facilitates plant infection by the necrotrophic patho-
gen Botrytis cinerea. Curr Biol. 2000; 10:751–57. https://doi.org/10.1016/s0960-9822(00)00560-1
PMID: 10898976
16. Govrin EM, Rachmilevitch S, Tiwari BS, Solomon M, Levine A. An elicitor from Botrytis cinerea induces
the hypersensitive response in Arabidopsis thaliana and other plants and promotes the gray mold dis-
ease. Phytopathology. 2006; 96:299–307. https://doi.org/10.1094/PHYTO-96-0299 PMID: 18944445
17.
Joshi AK, Chand R, Kumar S, Singh RP. Association of leaf tip necrosis with the SB pathogen in wheat.
Crop Sciences. 2004; 44:792–797.
18. Dyck PL. The association of a gene for leaf rust resistance with the chromosome-7D suppressor of
stem rust resistance in common wheat. Genome. 1982; 29: 467–469.
19. Dyck PL and Sambroski DJ. The inheritance of resistance to Pucciniarecondita in a group of common
wheat cultivars. Canadian J. Gen. Cytol. 1982; 24:273–283.
20. Singh RP. Genetic association of gene Bdv1 for barley yellow dwarf virus with genes Lr34 and Yr18 for
adult plant resistance to rusts in bread wheat. Plant Dis. 1993; 77:1103–1106.
21. Singh RP and Rajaram S. Genetics of adult plant resistance to stripe rust in ten spring bread wheat.
Euphytica 1994; 72: 1–7.
22. Dubin HJ, Van Ginkel. The status of the wheat diseases and disease research in warmer areas. In
Saunders D. A. (Ed) Wheat for the Non-traditional warmer Areas Mexico DF CIMMYT. 1991; pp. 125–
145.
23. Duveiller E, Gilchrist L. Production constraints due to Bipolarissorokiniana in wheat: current situation
and future prospects. In: Saunders D.A and Hettel G.P. (eds.), Wheat in Heat Stressed Environments:
Irrigated, Dry Areas and Rice-Wheat Farming Systems Mexico DE CIMMYT. 1994; pp. 343–352.
24. Mehta YR. Manejo Integrado de Enfermedadas de Trigo-santa Cruz. Bolivia CIAT/IAPAR. 1994;pp.
314.
25. Bakonyi J, Aponyi I, Fischl G. Diseases caused by Bipolarissorokiniana and Drechsleratriticirepentis in
Hungary InDuveiller E, Dubin HJ, Reeves J, McNab A (eds) Helminthosporium Blights of wheat: Spot
Blotch and Tan Spot Mexico DF CIMMYT. 1997; pp. 80–88.
PLOS ONE | https://doi.org/10.1371/journal.pone.0240029 October 5, 2020
14 / 16
PLOS ONE Genetic characterization for lesion mimic and spot blotch resistance in wheat
26. Pandey SP, Kumar S, Kumar U, Chand R, Joshi AK. Sources of inoculum and re-appearance of SB of
wheat in rice-wheat cropping systems in eastern India. European J Plant Pathol. 2005; 111:47–55.
27.
28.
29.
30.
Zhang L, Lavery L, Gill U, Gill K, Steffenson B, Yan GP. et al. A cation/proton-exchanging protein is a
candidate for the barley NecS1 gene controlling necrosis and enhanced defence response to stem rust.
TheorAppl Genet. 2009; 118: 385–397.
Lopes MS, Dreisigacker S, Peña RJ, Sukumaran S, Reynolds MP. Genetic characterization of the
wheat association mapping initiative (WAMI) panel for dissection of complex traits in spring wheat. The-
orAppl Genet 2015; 128:453–464. https://doi.org/10.1007/s00122-014-2444-2 PMID: 25540818
Joshi AK, Chand R. Variation and inheritance of leaf angle and its association with SB (Bipolarissoro-
kiniana) severity in wheat (Triticumaestivum L.). Euphytica. 2002; 24: 283–291.
Zadoks JC, Chang TT, Konjak CF. A decimal code for the growth stages of cereals. Weed Research.
1974; 14: 415–421.
31. Chaurasia S, Joshi AK, Dhari R, Chand R. Resistance to foliar blight of wheat; a research. Genet
Resour Crop Evol. 1999; 46: 469–475.
32. Parlevliet JE. Components of resistance that reduce the rate of epidemic development. Annu Rev Phy-
topathol. 1979; 17: 203–222.
33. Saari EE, Prescott JM. A scale for appraising the foliar intensity of wheat diseases. Plant Disease
Reporter. 1975; 59:377–380.
34. Shaner G, Finney R. The effect of nitrogen fertilization on the expression of slow mildewing in knox
wheat. J Phytopathol. 1977; 67: 1051–1056
35. Saghai-maroof MA, Solima KM, Jorgenson RA, Allard RW. Ribosomal DNA spacer-length polymor-
phisms in barley: Mendelian inheritance, chromosomal location and population dynamics. Proc Natl
AcadSci USA. 1984; 81: 8014–8018.
36. Sivakumar S, Crossa J, Jarquin D, Lopes M, Reynolds MP. Genomic and pedigree prediction with
Genotype × Environment Interaction in spring Wheat Grown in South and Western Asia, North Africa,
and Mexico. CIMMYT Research Data & software Repository Network. 2016; http://hdl.handle.net/
11529/10714. WAMI PHENOTYPIC AND MARKER DATA.zip. Hdl/11529/10714.
37. Pritchard KJ, Matthew Stephens, Peter Donnelly. Inference of population structure using multilocus
genotype data. Genetics. 2000; 155: 945–959. PMID: 10835412
38. Dent Earl A STRUCTURE HARVESTER: a website and program for visualizing STRUCTURE output
and implementing the Evanno method. Conservation Genetics Resources, Volume 4, Issue 2.
2012; pp 359–361.
39. Bradbury PJ, Zhang Z, Kroon DE, Casstevens TM, Ramdoss Y, Buckler ES. TASSEL: software for
association mapping of complex traits in diverse samples. Bioinformatics. 2007; 23: 2633–2635.
https://doi.org/10.1093/bioinformatics/btm308 PMID: 17586829
40. Kang HM, Zaitlen NA, Wade CM, Kirby A, Heckerman D, Daly MJ, et al. Efficient control of population
structure in model organism association mapping. Genetics. 2008; 178: 1709–23. https://doi.org/10.
1534/genetics.107.080101 PMID: 18385116
41. Wang F, Wu W, Wang D, Yang W, Sun J, Liu D. et al. Characterization and genetic analysis of novel
light dependent lesion mimic mutant, lm3, showing adult plant resistance to powdery mildew in common
wheat. PLoS ONE. 2016; 11(5):e0155358, https://doi.org/10.1371/journal.pone.0155358 PMID:
27175509
42.
Joshi AK, Kumari M, Singh V, Reddy C, Kumar S, Rane J, et al. Stay green trait: variation, inheritance
and its association with SB resistance in spring wheat (Triticumaestivum L.). Euphytica. 2007; 153: 59–
71. https://doi.org/10.1007/s10681-006-9235-z
43. Singh PK, Zhang Y, He X., Singh RP, Chand R., Mishra VK, et al. Development and characterization of
the 4th CSISA-SB nursery of bread wheat. Euro. J. Pl. Pathol. 2015; 143: 595–605.
44. Gupta PK, Chand R, Vasistha NK, Pandey SP, Kumar U, Mishra VK, et al. Spot blotch disease of
wheat; The current status of research on genetics and breeding. Plant Pathol. 2018; 67: 508–531.
45. Singh S, Mishra VK, Kharwar RN, Ahirwar RN, Sharma S, Yadav PS. et al. Distribution of leaf tip necro-
sis genes and its association with expression of lesion mimic genes and resistance to SB in spring
wheat. Int. J. Life. Sci. Scienti. Res. 2017; 3(1): 808–816. https://doi.org/10.21276/ijlssr.2017.3.1.10
46. Singh PK, He X., Sansaloni CP, Juliana P, Dreisgacker S, Duveiller E, et al. Resistance to SB in two
mapping population of common wheat is controlled by multiple QTL of minor effect. Inte. J. Mol. Sci.
2018; 19:4054, https://doi.org/10.3390/ijms19124054 PMID: 30558200
47.
Lillemo M, Joshi AK, Prasad R, Chand R, Singh RP. QTL for SB resistance in bread wheat line Saar co-
locate to the biotrophic disease resistance loci Lr34 and Lr46. TheorAppl Genet. 2013; 126(3):711–
719. https://doi.org/10.1007/s00122-012-2012-6 PMID: 23139144
PLOS ONE | https://doi.org/10.1371/journal.pone.0240029 October 5, 2020
15 / 16
PLOS ONE Genetic characterization for lesion mimic and spot blotch resistance in wheat
48. Adhikari TB, Jackson EW, Gurung S, Hansen JM, Bonman JM. Association mapping of quantitative
resistance to Phaeosphaerianodorum in spring wheat landraces from the USDA National Small Grains
Collection. Phytopathology. 2011; 101: 1301–1210. https://doi.org/10.1094/PHYTO-03-11-0076 PMID:
21692647
49. Adhikari TB, Gurung S, Hansen JM, Jackson EW Bonman JM. Association mapping of quantitative trait
loci in spring wheat landraces conferring resistance to bacterial leaf streak and SB. The Plant Genome.
2012; 5: 1–16.
50. Gurung S, Mamidi S, Bonman J, Jackson E, Del, Rı´o L et al. Identification of novel genomic regions
associated with resistance to Pyrenophoratritici-repentis races 1 and 5 in spring wheat landraces using
association analysis. TheorAppl Genet. 2011; 123: 1029–1041.
51. Gurung S, Mamidi S, Bonman JM, Xiong M, Brown-guedira G. Genome-wide association study reveals
novel quantitative trait loci associated with resistance to multiple leaf spot diseases of spring wheat.
PLoS ONE. 2014; 13(12): e0208196 https://doi.org/10.1371/journal.pone.0108179.
52. Ahirwar RN, Mishra VK, Chand R, Budhlakoti N, Mishra DC, Kumar S, et al. Genome-wide association
mapping of SB resistance in wheat association mapping initiative (WAMI) panel of spring wheat (Triticu-
maestivum L.). PLoSONE. 2018; 13(12):e0208196 https://doi.org/10.1371/journal.pone.0208196.
53. Kumar U, Joshi AK, Kumar S, Chand R, Roder MS. Quantitative trait loci for resistance to SB caused by
Bipolarissorokiniana in wheat (T. aestivum L.) lines “Ning 8201” and “Chirya 3”. Mol. Breed. 2010; 26:
477–491.
54. Singh V, Singh G, Chaudhury A, Ojha A, Tyagi BS, Chowdhury AK, et al. Phenotyping at hot spots and
tagging of QTLs conferring SB resistance in bread wheat. MolBiol Rep. 2016; 43: 1293–1303.
PLOS ONE | https://doi.org/10.1371/journal.pone.0240029 October 5, 2020
16 / 16
PLOS ONE |
10.1371_journal.pone.0239320 | RESEARCH ARTICLE
Prevalence and determinants of anemia
among women of reproductive age in Thatta
Pakistan: Findings from a cross-sectional
study
Sumera Aziz AliID
Michael Hambidge3‡, Nancy F. Krebs3‡, Jamie E. Westcott3‡, Elizabeth M. McClure4‡,
Robert L. Goldenberg5‡, Sarah Saleem2☯
1☯*, Zahid Abbasi2☯, Babar Shahid2‡, Ghazal Moin2‡, K.
1 Department of Epidemiology, Mailman School of Public Health, Columbia University, New York City, New
York, United States of America, 2 Department of Community Health Sciences, Aga Khan University, Karachi,
Pakistan, 3 Department of Pediatrics, Section of Nutrition, University of Colorado Anschutz Medical Campus,
Aurora, Colorado, United States of America, 4 Regional Triangulate Institute International, Research Triangle
Park, North Carolina, United States of America, 5 Department of Obstetrics and Gynecology, Columbia
University, New York City, New York, United States of America
☯ These authors contributed equally to this work.
‡ These authors also contributed equally to this work.
* sa3778@cumc.columbia.edu
Abstract
Background
Anemia is a major public health concern among women of reproductive age leading to high
maternal mortality in low- and middle-income countries. Of the prior studies conducted in
Pakistan, most focused on large urban areas and did not explore the determinants of ane-
mia among women of reproductive age (WRA) across socio-demographic, dietary, repro-
ductive, and biological domains. Thus, we aimed to study the prevalence and determinants
of anemia among WRA in rural Pakistan.
Methods
We conducted a cross-sectional study in the Thatta district of Pakistan from September
2018 to January 2019 and enrolled 150 non-pregnant, married women. Data collectors
administered a structured questionnaire to collect sociodemographic, reproductive and die-
tary data from women, who also provided stool and blood samples. We classified all WRA
as anemic if their hemoglobin was <12.0 g/dl. We performed logistic regression analysis to
calculate adjusted odds ratios (aOR) and their respective 95% CIs to assess the determi-
nants of anemia.
Results
In our study, 61.3% of the enrolled women were anemic. In the multivariable analysis, we
found that factors such as serum iron levels of less than 50 μg/dl (aOR: 7.17; 95% CI (2.94,
17.47)), history of breastfeeding (aOR: 2.43; 95% CI (1.04, 5.72)), living in a katcha house
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: Ali SA, Abbasi Z, Shahid B, Moin G,
Hambidge KM, Krebs NF, et al. (2020) Prevalence
and determinants of anemia among women of
reproductive age in Thatta Pakistan: Findings from
a cross-sectional study. PLoS ONE 15(9):
e0239320. https://doi.org/10.1371/journal.
pone.0239320
Editor: Sabine Rohrmann, University of Zurich,
SWITZERLAND
Received: May 31, 2020
Accepted: September 4, 2020
Published: September 24, 2020
Copyright: © 2020 Ali et al. This is an open access
article distributed under the terms of the Creative
Commons Attribution License, which permits
unrestricted use, distribution, and reproduction in
any medium, provided the original author and
source are credited.
Data Availability Statement: All relevant data are
within the paper and its Supporting Information
files.
Funding: The primary and only source of funding
for this study was the University of Colorado-
Denver USA | Anschutz Medical Campus. Neither
my instituion (Aga Khan University), nor any other
organization provided any funding for this cross-
sectional study The funders had no role in study
design, data collection and analysis, except they
PLOS ONE | https://doi.org/10.1371/journal.pone.0239320 September 24, 2020
1 / 16
PLOS ONE Prevalence and determinants of anemia among women of reproductive age
reviewed and approved the protocol of the study
and provided feedback and comments on the final
manuscript and approved the manuscript for
publication. No author received salary for this
cross-sectional study. The authors “Sumera Aziz
Ali and Zahid Abbasi” received a proportion of their
salary for the main and primary Women First
study, but not for this cross-sectional study. The
primary Women First study was funded by
University of Colorado- Denver USA | Anschutz
Medical Campus. We received funding for this
study from the University of Colorado-Denver USA
| Anschutz Medical Campus.
Competing interests: The authors have declared
that no competing interests exist.
(aOR: 6.61; 95% CI (2.21, 19.87)), no consumption of meat (aOR: 4.18; 95% CI (1.66,
9.96)) were significantly associated with anemia among WRA. A history of more than
one abortion (aOR: 0.06; 95% CI (0.01, 0.33) appeared protective for its association with
anemia.
Conclusion
Our findings demonstrate a high burden of anemia and its complex determinants among
WRA in rural Pakistan. A combination of nutritional and educational strategies should be
designed to encourage rural women to consume iron-rich foods in their diet with an access
to adequate food. Breastfeeding women should be encouraged to consume extra calories
with sufficient intake of the food to continue exclusive breastfeeding and reserve the iron
stores through amenorrhea to prevent themselves from becoming anemic.
Introduction
Anemia is one of the major public health issues among women of reproductive age (WRA), as
it leads to high maternal morbidity and mortality and adverse pregnancy outcomes [1]. Ane-
mia occurs when hemoglobin (Hb) concentration falls below an established cut-off value
(<12.0 g/dl) among WRA, consequently impairing the capacity of the blood to transport oxy-
gen to the body [2–4]. About half a billion of WRA are anemic worldwide, with a higher bur-
den of anemia in low and middle-income countries (LMICs), affecting nearly two-thirds of
WRA [5, 6]. More specifically, 41.9% of WRA are anemic in South-East Asia [6]. This is in
contrast to the burden of anemia in developed countries, such as in Europe where 2–5% of
WRA are classified as anemic [7]. Like other LMICs, there is a high burden of anemia in Paki-
stan [8]. Studies have shown that anemia affects 41.7% to 77.0% of WRA in Pakistan [8, 9].
Anemia is more prevalent in the rural areas of Pakistan, where it is often severe and linked to
adverse health consequences such as postpartum hemorrhage, preterm delivery or stillbirth
and low birth weight babies [10]. Furthermore, a study conducted by Hambidge et al 2019
found that anemia is an important and potentially modifiable risk factor on birth outcomes
among rural Pakistani women [11]. More specifically, this study demonstrated that maternal
anemia at baseline modified the treatment effect for birth outcomes such as birth weight,
weight to length ratio as the improvements in these outcomes were significantly greater for
Arm 1 (commencing nutrition intervention prior to conception) vs. Arm 2 (commencing
same nutrition intervention early in gestation) [11].
Although few studies have investigated the determinants of anemia in Pakistan, certain
gaps need to be addressed [12–14]. First, existing studies mainly focused on large urban areas
of Pakistan rather than rural areas where the burden of anemia is higher. In addition, there are
differences in the socio-demographic, cultural, economic, and dietary factors between rural
and urban areas, which do not accurately characterize anemia in a rural setting [15]. Moreover,
rural areas have limited sources of income with poor access to health facilities and other ame-
nities such as electricity, modes of communication and access to health care facilities [16, 17].
Therefore, studies of urban populations may not be relevant in the rural context of Pakistan.
Second, anemia is a complex and multifactorial phenomenon that should be understood holis-
tically. Current studies in urban settings provide important insights about sociodemographic
and reproductive determinants, but leave key biological and dietary determinants unaddressed
PLOS ONE | https://doi.org/10.1371/journal.pone.0239320 September 24, 2020
2 / 16
PLOS ONE Prevalence and determinants of anemia among women of reproductive age
and therefore, poorly understood [18–20]. Lastly, investigators have not utilized objective mea-
surements of biological factors such as iron, ferritin, B12, and folic acid in the serum, hook-
worm in the stool, and a malarial parasite in the blood. Rather, their results are often based on
self-reported data, which are subjective and may have produced biased estimates of associa-
tions between biological determinants and anemia among WRA.
There is a need to investigate the underlying determinants of anemia comprehensively
across the range of socio-demographic, reproductive, dietary, and biological domains in rural
Pakistan, which have not been previously studied in this context. Understanding these deter-
minants holistically will enable the development of local strategies and targeted interventions
to address the burden of anemia among WRA in rural Pakistan. Hence, the overarching objec-
tives of this study are to measure the prevalence and determinants of anemia among WRA in
rural Pakistan. We hypothesized that poor, less educated, older, and multiparous women who
consume iron-deficient diet with abnormal biomarkers will be more anemic than younger,
nulliparous, well-privileged women who consume iron-rich diet.
Materials and methods
Study population and sample
We conducted a cross-sectional study in the district of Thatta, Pakistan from September 2018
to January 2019. We examined the associations of sociodemographic, reproductive, dietary,
and biological determinants of anemia in 150 WRA participating in the Women First (WF)
preconception maternal nutrition trial [21]. At the time of enrollment in the primary WF trial,
women were eligible if they had 0–5 children and planned to conceive in the future after being
enrolled in the WF primary study. We used a subset of women from the WF database (385
women) who provided biological samples in the primary trial, who had conceived during their
enrollment in the WF primary study, and had delivered their baby [21]. For this cross-sec-
tional study, we randomly selected study participants using the WF database as a sampling
frame. Eligibility was contingent on being a resident of the Thatta district, not being pregnant
at the time of survey, agreeing to participate, and intending to provide biological samples after
enrollment in the cross-sectional study.
Data collection tool (questionnaire)
We used a structured and validated questionnaire to collect data on socio-demographic, repro-
ductive, anthropometric, dietary, and biological determinants from WRA. We identified possi-
ble determinants of anemia based on the literature pertaining to the risk of development of
anemia among women in LMICs [19]. To develop a questionnaire for the proposed cross-sec-
tional study, we chose a series of validated questions from different studies that have been con-
ducted in Thatta Pakistan. For example, the Pakistan Demographic Health Survey 2012–2013
(PDHS-2012-2013) was one of such sources to choose the relevant validated questions [15].
We selected questions related to socio-demographic characteristics from the PDHS-2012-2013
survey and adapted the questions wherever necessary [15]. Secondly, a maternal and newborn
health registry (MNHR) is placed in Thatta district [20], whereby data is being collected on the
socio- demographic, and reproductive characteristics of WRA in a similar setting and same
population, where we undertook the cross-sectional study [18]. The MNHR has validated the
questions in a similar population that also helped me to use several relevant questions for the
current cross-sectional study [18]. Lastly, an individual randomized nutritional trial (Woman
First) was conducted from 2012 to 2018 in Thatta district where MNHR is based [19, 21].
The trial used 20 validated questionnaires to explore socio-demographic, reproductive, dietary
factors, anthropometric measurements, and biological factors among WRA in the same
PLOS ONE | https://doi.org/10.1371/journal.pone.0239320 September 24, 2020
3 / 16
PLOS ONE Prevalence and determinants of anemia among women of reproductive age
population [19]. Since all these questionnaires were relevant to the proposed study; therefore,
we choose the most appropriate questions from these validated questionnaires and adapted
these wherever required to develop and finalize the questionnaire for the cross-sectional study.
After developing and finalizing the questionnaire from three main sources, we translated the
questionnaire into the local language of district Thatta (Sindhi) followed by back-translation.
To check the flow of questions, we pre-tested questionnaire on the WRA other than the study
setting of Thatta but having similar characteristics to the study sample. The questionnaire
consisted of different sections and collected socio-demographic, reproductive, dietary, and
anthropometric data.
Study procedures & data collection
After assessing the eligibility criteria and receiving written informed consent, trained study
staff, who were familiar with the local language (Sindhi), administered a structured and vali-
dated questionnaire to the enrolled women in their households. The study staff asked women
about various socio-demographic, reproductive and dietary measures including maternal age,
parity, education, working status, water supply, sanitation, age at the time of marriage, history
of any previous adverse pregnancy outcome, date of the last delivery, consumption of fruits,
vegetables and meat in their diet, and use of smokeless tobacco. The study staff also measured
weight (in kilograms), height (in meters), waist circumference (in centimeters (cm)), hip cir-
cumference (cm), and mid-upper arm circumference (cm). Using a digital weighing scale, the
study staff measured the weight of WRA while barefoot and wearing light clothes to the nearest
0.1 kg. The same staff measured the height of a woman barefoot to the nearest of 0.1 cm using
a stadiometer. Before the measurement, the staff calibrated scales to zero. The staff measured
mid-upper arm circumference on the upper right arm at the midpoint of the acromion process
and the tip of the olecranon (precision 1mm). Similarly, the research staff measured waist cir-
cumference just above the iliac crest in the horizontal plane, and hip circumference at the
point yielding the maximum circumference over the buttocks, all using a standard measuring
tape to the nearest 1mm. We calculated body mass index (BMI) as a ratio of weight (in kilo-
grams) to the square of height (in meters) and the study staff recorded BMI as a continuous
variable. We categorized BMI as underweight (less than 18.5 kg/m2), normal (18.5–24.9 kg/
m2), overweight (25.0–29.9 kg/m2), and obese (�30.0 kg/m2) [22]. For regression analysis, we
merged overweight and obese categories into a single category as overweight/obese due to low
numbers in the obese category.
A trained and professional phlebotomist collected biological samples including blood and
stool samples from the enrolled women. The phlebotomist collected around 12–15 mL of
venous blood by following the standard procedures [19]. The phlebotomist transferred all bio-
logical samples from Thatta collection points to the main laboratory of the Aga Khan Univer-
sity Hospital in Karachi, Pakistan by maintaining the cold chain. The trained laboratory staff
and microbiologist processed the blood samples to measure complete blood count (CBC),
serum iron and ferritin levels, serum B-12 and folic acid levels. In addition, they used the
blood samples to measure the inflammatory markers including erythrocyte sedimentation
rate (ESR) by the Westergren method, and C-Reactive protein (CRP) using a high sensitivity
ELISA assay [23, 24]. Moreover, they used an immunochromatographic test (ICT) technique
to identify the malarial parasite in the blood and examined blood for hemoglobinopathies
including thalassemia minor and major by hemoglobin (Hb) electrophoresis test [25, 26]. Fur-
thermore, they used stool samples to examine different worms in the stool to primarily assess
hookworm infestation. They identified parasites and eggs in the stool samples, using a micro-
scope with conventional techniques such as direct wet smear stain (Lugol stain) [27].
PLOS ONE | https://doi.org/10.1371/journal.pone.0239320 September 24, 2020
4 / 16
PLOS ONE Prevalence and determinants of anemia among women of reproductive age
Primary outcome
The primary outcome of this study was anemia and we used hemoglobin levels (grams per
deciliter) in the blood to classify a woman as anemic. More specifically, we used the World
Health Organization (WHO) cut-off of a hemoglobin level <12.0 g/dL to classify a non-preg-
nant woman aged �15 years as anemic [28–32]. For the current analysis, we created a binary
variable of anemia and no-anemia.
Potential determinants of anemia
Independent variables in this study consisted of various socio-demographic, reproductive, die-
tary, and biological determinants. The socio-demographic determinants included woman’s
age (in years), literacy status (illiterate / literate), residence (urban/ rural), type of house (pucca
house: made of concrete, plastered walls, and roof / katcha house: made of wood, clay, and
stalks), religion (Hinduism/Islam), current working status (are you working currently to earn
money: Yes/No), having electricity (Yes/No), type of family system (nuclear/joint family).
Reproductive determinants included age at the time of marriage (in years), gravidity (total
number of pregnancies), parity (number of children born after 20 weeks of gestation), history
of adverse pregnancy outcomes including stillbirths and abortions (Yes/No), breastfeeding sta-
tus at the time of the survey (Yes/No), current and ever use of family planning method (Yes/
No), and menstrual cycle (regular/irregular).
Dietary determinants included intake of fruits and vegetables in past one month (Yes / No),
consumption of meat (Yes/ No), and history of chewing smokeless tobacco (Yes/No). In addi-
tion, we measured biological determinants such as serum iron (normal range: 50–170 μg/dl),
serum ferritin (normal range: 10 to 150 ng/ml), serum folic acid (normal range: � 2.5–20 ng/
ml), serum CRP (normal range: 0–0.5mg/dl), ESR (normal range: 0–20 mm/ 1st hr), thalasse-
mia minor or major (yes/no), malarial parasite (positive/negative), and parasites in the stool
(seen/unseen).
Statistical methods
To describe the characteristics of the study population, we reported frequencies and propor-
tions. We used a Chi-squared test or Fisher’s exact test to assess the frequency distribution and
the relationship between determinants and anemia status. Similarly, we performed bivariate
logistic regression analyses to determine the individual effect of each significant determinant
on anemia status before the multivariable logistic regression analysis. We used multicollinear-
ity diagnostics such as variance inflation factor (VIF) and tolerance (reciprocal of VIF) to
assess multicollinearity between independent variables [33]. Any VIFs of more than 4 were
investigated, while VIFs greater than 10 or tolerance less than 0.1 indicated serious multicolli-
nearity requiring correction by dropping one of the variables [33, 34].
Finally, we conducted a multivariable logistic regression analysis to determine the effect of
each determinant on anemia by including all predictors and covariates in a single multivari-
able model using a purposeful selection method [35]. We used a p-value cut-off of <0.05 to
determine if a variable should be included in the final regression model. We presented the
results of regression analyses with crude/unadjusted odds ratios (OR) and adjusted odds ratios
(aOR) with 95% confidence intervals (CIs). We used SPSS 19.0 to analyze the data.
Ethical considerations
Ethical approval for this study was obtained from the Aga Khan University Ethical Review.
Committee (2018–0262–342). All study participants provided written informed consent.
PLOS ONE | https://doi.org/10.1371/journal.pone.0239320 September 24, 2020
5 / 16
PLOS ONE Prevalence and determinants of anemia among women of reproductive age
Voluntary participation and the right to ask any questions and decline participation at any
time were emphasized during data collection.
Results
Characteristics of women of reproductive age
We found that more than a quarter of the women (28.0%) were younger than 25 years old and
40.0% were 25–29 years old as shown in Table 1. Moreover, 82.0% of the women were not able
to read or write, a similar proportion of the women (81.3%) were residents of rural areas, and
44.7% reported to be living in a katcha house. More than half of the study participants (55.3%)
did not report working to earn money, and around three-fourths (73.3%) of them were living
in a joint family system (Table 1). Around three-fourths of the study participants (71.3%)
had given birth to at least three children, and 8.7% had a history of more than one abortion.
Slightly more than a third of women (36.0%) reported not breastfeeding their children at the
time of the study, and 46.0% had a normal BMI of 18.5-25kg/m2. Regarding the women’s diet,
we found that 41.3% of the women had never consumed fruits or vegetables in the past month,
and 68.7% had never eaten meat in the past month (Table 1).
We found that 61.3% of women were anemic in the study, and a similar proportion of the
women (62.0%) had low serum iron levels (Table 2). We found that 42.7% of the women had
lower serum ferritin values; however, only 6.0% of the women had low folic acid levels rela-
tive to the normal range of > 2.6ng/ml (Table 2). All of the women had serum B12 levels
within the normal range (>150 pg/ml). Around 10.0% of the women had higher values of
CRP than the normal range of 0–0.5mg/dl, and around 50.0% of the women had higher val-
ues for ESR than the normal range of 0–20 mm/ 1st hr. We found that only 8.7% of the
women had any kind of underlying hemoglobinopathy, and 16.0% of the women’s bloods
were positive for the malarial parasite and 15.2% of the women had parasites in their stool as
shown in Table 2.
Sociodemographic, reproductive, and dietary characteristics of women of
reproductive age by anemia status
We found no differences between anemic and non-anemic women by age, educational level,
parity, working status, place of residence, and type of family system (Table 1). However, we
found that a significantly higher proportion of anemic women (52.5%) were living in a katcha
house when compared with 32.8% of non-anemic women (p-value: 0.02). Similarly, 68.5% of
anemic and 56.9% of non-anemic women reported breastfeeding their children. A signifi-
cantly higher proportion of anemic women (76.1%) had never consumed meat in the past
month compared with non-anemic women (56.9%; p-value: 0.01). About half of anemic
women (46.7%) were underweight, while 39.7% of the non-anemic women were underweight
(Table 1).
The study also found that three-fourths of the anemic women (75.0%) had lower serum
iron levels compared to 41.4% of the non-anemic women. Similarly, around half of anemic
women (51.1%) had lower serum ferritin levels when compared to 29.3% of non-anemic
women as shown in Table 2. There were no differences between anemic and non-anemic
women by levels of serum folic acid, CRP, ESR, hemoglobinopathies, and malarial parasite in
the blood and stool parasites (Table 2).
PLOS ONE | https://doi.org/10.1371/journal.pone.0239320 September 24, 2020
6 / 16
PLOS ONE Table 1. Socio-demographic, reproductive and dietary determinants of women of reproductive age by anemia status in Thatta Pakistan, crude odds ratios and 95%
CI (n = 150).
Prevalence and determinants of anemia among women of reproductive age
Characteristics
Maternal age (Years)
< 25
25–29
�30
Parity
�2
3–4
>4
Educational Status of women
Literate
Illiterate
Current working status of woman
No
Yes
Place of residence
Urban
Rural
Type of Family System
Nuclear
Joint
Type of House
Pakka
Katcha
Age at the time of marriage
>18
� 18
Number of abortions
� 1
>1
Currently Breastfeeding
No
Yes
Current usage of contraception
Yes
No
Intake of vegetables (past one month)
Yes
No
Intake of fruits (past one month)
Yes
No
Intake of meat (past one month)
Yes
No
Chew smokeless tobacco
No
Total n (%)
Anemia n (%)
No Anemia n (%)
Odds Ratio
95%CI
P-value�
42 (28.0)
60(40.0)
48 (32.0)
43 (28.7)
66 (44.0)
41 (27.3)
27 (18.0)
123 (82.0)
83 (55.3)
67 (44.7)
28 (18.7)
122 (81.3)
40 (26.7)
110 (73.3)
83 (55.3)
67 (44.7)
79 (52.7)
71 (47.3)
137 (91.3)
13 (8.7)
54 (36.0)
96 (64.0)
45 (30.0)
105 (70.0)
88 (58.7)
62 (41.3)
88 (58.7)
62 (41.3)
47 (31.3)
103 (68.7)
24(26.1)
36 (39.1)
32(34.8)
24(26.1)
39 (42.4)
29(31.5)
18(19.5)
74(80.4)
54(58.7)
38(41.3)
18(19.6)
74(80.4)
23(25.0)
69(75.0)
44(47.8)
48(52.2)
42(45.7)
50(54.3)
89(96.7)
3(3.3)
29(31.5)
63(68.5)
29(31.5)
63(68.5)
51(55.4)
41(44.6)
52(56.5)
40(43.5)
22(23.9)
70(76.1)
18(31.0)
24(41.4)
16(27.6)
19(32.8)
27(46.6)
12(20.7)
09(15.5)
49 (84.4)
29(50.0)
29(50.0)
10(17.2)
48(82.8)
17(29.3)
41(70.7)
39(67.2)
19(32.8)
29(50.0)
29(50.0)
48(82.8)
10(17.2)
25(43.1)
33(56.9)
16(27.6)
42(72.4)
37(63.8)
21(36.2)
36(62.1)
22(37.9)
25(43.1)
33(56.9)
48 (32.0)
33(35.9)
15(25.9)
1
1.13
1.50
1
1.14
1.91
1
0.75
1
0.70
1
0.85
1
1.24
1
2.24
1
1.19
1
0.16
1
1.65
1
0.81
1
1.42
1
1.26
1
2.41
1
0.51–2.50
0.64–3.53
0.53–2.48
0.77–4.71
0.63
0.33
0.28–1.76
0.45
0.36–1.36
0.29
0.36–2.01
0.59–2.59
0.72
0.56
1.13–4.44
0.02
0.62–2. 29
0.60
0.04–0.62
0.005
0.83–3.25
0.15
0.40–1.71
0.61
0.72–2.78
0.31
0.64–2.46
0.50
1.18–4.88
0.01
0.20
(Continued )
PLOS ONE | https://doi.org/10.1371/journal.pone.0239320 September 24, 2020
7 / 16
PLOS ONE Prevalence and determinants of anemia among women of reproductive age
Table 1. (Continued)
Characteristics
Yes
BMI (kg / m2)
Overweight & Obese
Underweight
Normal
Total n (%)
Anemia n (%)
No Anemia n (%)
Odds Ratio
95%CI
P-value�
102 (68.0)
59(64.1)
43(74.1)
12 (8.0)
69 (46.0)
69 (46.0)
3 (3.3)
43(46.7)
46(50.0)
9 (15.5)
23(39.7)
26(44.8)
0.62
1
4.96
6.00
0.30–1.28
1.23–20.01
1.48–24.31
0.04
�P-values have been calculated using Chi-squared test.
https://doi.org/10.1371/journal.pone.0239320.t001
Determinants of anemia among women of reproductive age: Findings from
a bivariate analysis
We assessed and found multiple statistically significant associations between covariates and
anemia in bivariate analyses. More specifically, we assessed the association between various
sociodemographic, reproductive, dietary, and biological factors and anemia among WRA. We
found that women with more than four children had a 1.91(OR = 1.91; [95% CI: 0.77, 4.71])
times the odds of being anemic when compared with women having less than 2 children
(Table 1). Meanwhile, women living in a katcha house had a 2.24 (OR = 2.24; [95% CI:1.13,
Table 2. Biological determinants of women of reproductive age by anemia status in Thatta Pakistan, crude odds ratios and 95%CI (n = 150).
Characteristics
Serum Iron (μg/dl)
� 50
< 50
Serum Ferritin (ng/ml)
> 10
� 10
Serum Folic Acid (ng/ml)
> 2.6
� 2.6
Serum C-Reactive Protein (mg/dl)
0–0.5
> 0.5
ESR (mm/ 1st hr)
0–20
> 20
Hemoglobinopathies on Hb electrophoresis
No
Yes
Malarial parasite
Negative
Positive
Parasites in stool
Not seen
seen
Total n (%)
Anemia n (%)
No Anemia n (%)
Odds Ratio
95%CI
P-value�
57 (38.0)
93 (62.0)
86 (57.3)
64 (42.7)
141 (94.0)
9 (6.0)
102 (89.5)
12 (10.5)
60 (50.4)
59 (49.6)
137 (91.3)
13 (8.7)
126 (84.0)
24 (16.0)
125 (83.3)
25 (16.7)
23(25.0)
69(75.0)
45(48.9)
47(51.1)
89 (96.7)
3(3.3)
47(88.7)
6(11.3)
34(47.9)
37(52.1)
83 (90.2)
9 (9.8)
78(84.8)
14(15.2)
78 (84.8)
14(15.2)
34 (58.6)
24(41.4)
41 (70.7)
17(29.3)
52 (89.7)
6(10.3)
32(82.1)
7(17.9)
26(54.2)
22(45.8)
54(93.1)
4(6.9)
48(82.8)
10 (17.2)
47(81.0)
11(19.0)
1
4.25
1
2.52
1
0.29
1
0.58
1
1.29
1
1.46
1
0.86
1
0.76
2.1–8.59
<0.001
1.25–5.06
0.01
0.07–1.22
0.09
0.18–1.89
0.36
0.62–2.68
0.50
0.43–4.99
0.54
0.35–2.09
0.74
0.32–1.83
0.55
�P-values have been calculated using Chi-squared test.
https://doi.org/10.1371/journal.pone.0239320.t002
PLOS ONE | https://doi.org/10.1371/journal.pone.0239320 September 24, 2020
8 / 16
PLOS ONE Prevalence and determinants of anemia among women of reproductive age
4.44]) times the odds of being anemic when compared with their counterparts. While explor-
ing the association between reproductive factors and anemia, we found that women who were
breastfeeding their children at the time of the study had a 1.65 (OR = 1.65; [95% CI: 0.83,
3.25]) times the odds of being anemic when compared with their counterparts; however,
results were not statistically significant. Furthermore, women with low and normal BMI had
4.96 (OR = 4.96; [95% CI: 1.23, 20.01]) and 6.00 (OR = 6.00; [95% CI: 1.48, 24.31]) times the
odds of being anemic, respectively, when compared with overweight and obese women. We
also assessed the association between women’s diet and anemia in the bivariate analysis. We
found that women who never consumed fruits and vegetables in the last month had 1.26
(OR = 1.26; [95% CI: 0.64, 2.46]) and 1.42 (OR = 1.42; [95% CI: 0.72, 2.78]) times the odds of
being anemic when compared to women who consumed fruits and vegetables, respectively;
however, these results were statistically insignificant (Table 1). Women who never consumed
meat in the last month had 2.41 times the odds of being anemic when compared women who
reported consuming meat in the past one month (OR = 2.41; [95% CI:1.18, 4.88]). Regarding
the association between biological factors such as serum iron levels and anemia, we found that
women with low serum iron levels had 4.25 (OR = 4.25; [95% CI:2.1,8.59]) times the odds of
being anemic when compared to women having normal serum iron levels as shown in Table 2.
In contrast to other potential factors, we found that a history of more than one abortion was
found to be protective for its association with anemia (OR = 0.16; [95% CI:0.04, 0.62]).
Determinants of anemia among women of reproductive age: Findings of
multivariable analysis
The results of multivariable analyses demonstrated that low serum iron level, living in katcha
house, no consumption of meat in the past month, and history of breastfeeding increased the
risk of anemia (Fig 1). More specifically, significant and positive associations between low
serum iron levels (aOR = 7.17; 95% CI [2.94, 17.47]), living in a katcha house (aOR = 6.61;
[95% CI: 2.21, 19.87]), no consumption of meat in the past month (aOR = 4.18; [95% CI [1.66,
9.96]) and anemia persisted in the multivariable analysis. Moreover, the association between
Fig 1. Determinants of anemia among women of reproductive age in Thatta Pakistan [adjusted OR and 95% CI].
https://doi.org/10.1371/journal.pone.0239320.g001
PLOS ONE | https://doi.org/10.1371/journal.pone.0239320 September 24, 2020
9 / 16
PLOS ONE Prevalence and determinants of anemia among women of reproductive age
history of breastfeeding and anemia turned out to be statistically significant in the adjusted
model (aOR = 2.43; [95% CI:1.04, 5.72]). A history of more than one abortion (aOR: 0.06; 95%
CI (0.01, 0.33) appeared protective for its association with anemia as shown in Fig 1.
Discussion
This was one of the first studies in Pakistan that measured the burden of anemia and its deter-
minants comprehensively across socio-demographic, reproductive, dietary, and biological
domains among WRA in rural settings. Our study demonstrated a high prevalence of anemia
among WRA in a rural area of Pakistan. Factors such as low serum iron levels, no consump-
tion of meat, living in a katcha house, and history of breastfeeding were found to be risk factors
for anemia among WRA. However, women with more than one abortion were found to be at
lower risk for anemia. We did not find any differences between anemic and non-anemic
women by malarial parasite in the blood and stool parasites. Thus, presence of a malarial para-
site or stool parasite did not turn out to be important factors related to anemia in this study.
The results on the prevalence and determinants of anemia among Pakistani women should be
interpreted in their economic and socio-cultural context.
Our observation regarding the prevalence of anemia is consistent with other studies con-
ducted in different rural areas of Pakistan, where the prevalence of anemia has been found to
be between 41.7% to 77.0% [8, 9, 36, 37]. However, it is important to note that a direct compar-
ison of our findings with earlier studies conducted in Pakistan may not be possible because of
the differences in the number and composition of the study subjects enrolled, study settings
and socio-cultural circumstances.
The association of low serum iron levels with a higher risk of anemia in this study is com-
patible with the other studies conducted in urban areas of Pakistan and other low-income
countries, such as Vietnam and Ethiopia [31, 38–40]. Although studies conducted in urban
areas of Pakistan or Ethiopia did not measure serum iron levels in the blood, their results dem-
onstrate that women not consuming iron-rich foods or iron supplements and those with food
insecurity are at higher risk of anemia than women who consume iron-rich foods or take iron
supplements [31, 38–40]. The consistently positive association between low serum iron levels
and anemia may be due to poor diet quality and poor intake of iron-rich foods in our study
population. This was further evident from our study’s findings where two-fifths of the women
reported no consumption of vegetables and fruits in the prior month and more than two-
thirds of the women reported no consumption of meat in the prior month. Thus, inadequate
intake of iron, predominantly caused by reduced access to heme iron, which is primarily
found in meat and is highly bioavailable, can contribute to iron deficiency [41–43]. Further-
more, Lander et al. [44] reported that four-fifths of rural Pakistani women in the same district
(Thatta) were found to consume a diet with inadequate diversity, which might make WRA
susceptible to anemia in Thatta district [44]. Dietary intake data from a similar study also indi-
cated rural Pakistani women had the lowest mean intake of calories, protein and key micronu-
trients [44]. In addition to inadequate dietary diversity, the use of smokeless tobacco, e.g.
“gutkha” is common among rural Pakistani women and has been shown to reduce appetite
and cause inflammation and thus may have affected the intake of iron-rich foods in our study
population [45].
We found a positive association between poor consumption of meat and anemia in our
study population and this finding is consistent with other studies, which reported that poor
intake of meat is a risk factor of anemia among WRA [46, 47]. In rural Vietnam, the consump-
tion of meat of at least three times a week is more common in non-anemic women than in ane-
mic women [48]. Likewise, anemic women in one of the observational studies reported lower
PLOS ONE | https://doi.org/10.1371/journal.pone.0239320 September 24, 2020
10 / 16
PLOS ONE Prevalence and determinants of anemia among women of reproductive age
intake of red meat than non-anemic women [40]. This finding can be explained by the fact
that poor intake of meat can lead to low levels of iron in the body, thus resulting in anemia.
Overall, two types of dietary iron have been found; one is heme iron that comes mainly from
animal flesh foods and the other is non-heme iron, which is the only type found in plant-based
foods, such as grains and vegetables [49, 50]. In humans, the heme form is the most bioavail-
able, with an estimated absorption rate of 15%–35% when compared to absorption rate of 2%
to 20% for non-heme iron [50, 51].
As expected, women living in a katcha house were more likely to be anemic than their
counterparts. Living in a katcha house is one of the proxy indicators of poor socioeconomic
status, which is further related to the intake of a poor diet [28]. This relationship between poor
socio-economic status and anemia is consistent with other studies conducted in different parts
of Asia, Africa, and urban areas of Pakistan [28, 37, 39, 52]. For instance, a study conducted in
northern Pakistan revealed that women with a poor socioeconomic status were more likely to
be anemic than women with a higher socioeconomic status [37]. Likewise, a study conducted
in Bangladesh found that rich women were less likely to be anemic when compared to poor
women [28]. Collectively, these findings can be explained by the fact that a poor woman’s pur-
chasing power is meager; therefore, she cannot afford iron-rich food, such as meat and certain
iron-rich fruits and vegetables, and is therefore consumes a non-nutritious diet consequently
resulting in anemia [53, 54]. Moreover, due to food insecurity, these poor women also tend to
consume smokeless tobacco and non-nutritious substances such as fuller’s earth, clay, and ice
to reduce their appetite, and consequently, are more susceptible to anemia due to a lack of ade-
quate nutrition [55]. Generally, a woman’s diet also reflects her socioeconomic status. There-
fore, living in katcha homes and consuming low meat may all reflect a more marginal living
circumstances of a woman, including more risk of poor sanitation and poorer health status
(more inflammation as reflected by higher values of inflammatory markers and poorer iron
absorption). Thus, diet may well be part of the problem, but it is complicated and interconnec-
ted, and it is likely more than just not eating an adequate iron rich diet.
A history of breastfeeding was found to be associated with anemia among WRA, and this
finding is consistent with findings from other low-income countries [39, 56]. For example, a
study conducted in Ethiopia found that lactating women were at higher risk of anemia when
compared with non-lactating women [39]. There is evidence that a lactating woman needs to
consume more calories with a more diversified diet than non-lactating women [57]. Studies
from Pakistan have reported that women during pregnancy and lactation tend to avoid some
foods (due to misconceptions and beliefs) such as beef, eggs, brinjal, fish and citrus fruits as
these are considered hot and could have ill effects on their babies [58, 59]. Since the amenor-
rhea associated with full exclusive breastfeeding is generally assumed a protective factor for
iron status; insufficient intake of adequate diet could lead to insufficient milk production and
thus shorter exclusive breastfeeding and resumption of menses, which eventually leads to
worse iron status and anemia [60].
Contrary to findings reported in the literature [61–63], a history of more than one abortion
was found to be protective against anemia in our study; however, this finding should be inter-
preted with caution. There could be different plausible explanations for this inverse finding.
Women with more than one abortion were not able to complete the nine months of pregnancy
duration, thus preventing them from physiological anemia caused by pregnancy [64]. This
means that these women avoided anemia that might have occurred had they completed their
full-term pregnancy, as pregnancies generally result in a reduction in the concentration of red
blood cells [65]. Secondly, our study measure regarding the history of abortion was binary
without any further details about the abortions. For example, we did not ask about the type of
abortion from these women [spontaneous vs induced]. In addition, we did not ask women
PLOS ONE | https://doi.org/10.1371/journal.pone.0239320 September 24, 2020
11 / 16
PLOS ONE Prevalence and determinants of anemia among women of reproductive age
about how long they had experienced bleeding after abortion, nor did we investigate the vol-
ume of blood loss or any clots in the blood during the abortion. This lack of information limits
our ability to interpret these results. Lastly, it could be possible that women may not have cor-
rectly reported their history of abortion due to the sensitive nature of the issue and stigma
associated with abortion, or they might have considered a spontaneous abortion as normal
menstrual bleeding. If this reporting were differential across anemic and non-anemic women,
it could have led to a higher number of women with less than one abortion among anemic
than non-anemic women. Due to these potential limitations for this inverse finding, it is hard
to make any conclusion from this protective association between abortion and anemia, and
this needs to be explored further with sufficient details about the history of abortion.
Strengths and limitations
This is one of the first studies in rural Pakistan that investigated potential determinants of ane-
mia comprehensively across sociodemographic, reproductive, dietary and biological domains.
In addition, the random sampling of women suggests that these data are representative of the
overall population in the district. In addition, unlike other studies conducted in a few urban
areas, our data regarding biological determinants of anemia were based on objective measure-
ments rather than relying on self-reported data from women. However, there are some inher-
ent limitations. First, the cross-sectional nature of the survey does not allow temporal
precedence between various socio-demographic, dietary, reproductive, and biological determi-
nants, and anemia to be unambiguously determined. However, unlike other cross-sectional
studies reported in the literature, our study explored the determinants of anemia holistically
across socio-demographic, reproductive, dietary, and biological domains. Secondly, our die-
tary history questionnaire queried only what was eaten in the prior month and may not have
fully represented the daily or seasonal dietary patterns of the women. Although the measured
dietary history did not represent daily or seasonal dietary patterns, it did indicate monthly die-
tary patterns of WRA among rural women. A third limitation is that due to the sensitive nature
of some questions such as the use of contraception and the history of abortion, women might
not have reported the information accurately. However, we tried to overcome this limitation
by asking the sensitive questions in privacy and assured participants about the confidentiality
of their responses. Lastly, the small sample size of the study may limit our findings to be gener-
alized to the larger group of rural Pakistani women, thus more longitudinal and robust tepide-
miological studies with larger sample size are required in the future.
Conclusion
Our study confirmed that anemia is a major public health problem in women of reproductive
age in rural Pakistan and a large proportion of women were found to have low levels of serum
iron. Lack of dietary iron intake mainly contributes to iron deficiency among poor rural Paki-
stani women. These findings provide a rationale to take necessary steps aimed at alleviating
iron deficiency.
There could be several policy implications for this study in the future. The government
should target disadvantageous rural areas and social groups where the purchasing power,
knowledge, and access to health care facilities are limited. More specifically, government of
Pakistan need to develop some mechanisms to reduce the poverty at the household level by
creating job opportunities for the people living in remote areas of Pakistan. One of the ways
for poverty alleviation is to ensure good education in the schools both for boys and girls mainly
in the rural areas, which are highly reliant on public schools. Secondly, government can
PLOS ONE | https://doi.org/10.1371/journal.pone.0239320 September 24, 2020
12 / 16
PLOS ONE Prevalence and determinants of anemia among women of reproductive age
provide small loans or introduce social safety-net programs to help people living below the
poverty line with proper monitoring of the funds allocated for such poor households.
In addition, there is a need to design combined nutrition and educational strategies to
encourage rural women to consume iron-rich foods in their diet. Proactive efforts can be
made to educate women through effective campaigns for motivating them to take iron tablets
according to the prescribed schedule. Moreover, breastfeeding women should be encouraged
to consume extra calories during lactation to prevent themselves from becoming anemic. Simi-
larly, ensuring poor women’s access to information on risks of eating an iron-poor diet and
having a less adequate diet during lactation is also important to reduce anemia among women
in rural Pakistan.
Supporting information
S1 File.
(PDF)
S2 File.
(SAV)
Acknowledgments
The authors would like to extend their gratitude to all the study participants, for their contri-
bution to this research. More specifically, the authors want to acknowledge all data collectors
mainly four field supervisors, Miss Abida Khowaja, Miss Sumaira Fatima, Miss Aziza Khudai
and Miss Shagufta Begum who contributed to field supervision.
Author Contributions
Conceptualization: Sumera Aziz Ali.
Formal analysis: Sumera Aziz Ali, Babar Shahid, Ghazal Moin.
Funding acquisition: Sumera Aziz Ali.
Investigation: Sumera Aziz Ali, K. Michael Hambidge, Nancy F. Krebs, Jamie E. Westcott,
Elizabeth M. McClure, Robert L. Goldenberg, Sarah Saleem.
Methodology: Sumera Aziz Ali.
Project administration: Sumera Aziz Ali, Zahid Abbasi.
Supervision: Sumera Aziz Ali, Zahid Abbasi, Ghazal Moin, K. Michael Hambidge, Nancy F.
Krebs, Jamie E. Westcott, Elizabeth M. McClure, Robert L. Goldenberg, Sarah Saleem.
Writing – original draft: Sumera Aziz Ali.
Writing – review & editing: Sumera Aziz Ali, Zahid Abbasi, Babar Shahid, Ghazal Moin, K.
Michael Hambidge, Nancy F. Krebs, Jamie E. Westcott, Elizabeth M. McClure, Robert L.
Goldenberg, Sarah Saleem.
References
1. Milman N. Anemia—still a major health problem in many parts of the world! Annals of hematology.
2011; 90(4):369–77. https://doi.org/10.1007/s00277-010-1144-5 PMID: 21221586
2. World Health Organization. Haemoglobin concentrations for the diagnosis of anaemia and assessment
of severity. 2011.
PLOS ONE | https://doi.org/10.1371/journal.pone.0239320 September 24, 2020
13 / 16
PLOS ONE Prevalence and determinants of anemia among women of reproductive age
3. Cappellini MD, Motta I, editors. Anemia in clinical practice—definition and classification: does hemoglo-
bin change with aging? Seminars in hematology; 2015: Elsevier.
4. Beutler E, Waalen JJB. The definition of anemia: what is the lower limit of normal of the blood hemoglo-
bin concentration? 2006; 107(5):1747–50. https://doi.org/10.1182/blood-2005-07-3046 PMID:
16189263
5. Oue´draogo S, Koura GK, Bodeau-Livinec F, Accrombessi MM, Massougbodji A, Cot M, et al. Maternal
anemia in pregnancy: assessing the effect of routine preventive measures in a malaria-endemic area.
The American Journal of Tropical Medicine and Hygiene. 2013; 88(2):292–300. https://doi.org/10.4269/
ajtmh.12-0195 PMID: 23296448
6. World Health Organization.The global prevalence of anaemia in 2011. Geneva: World Health Organi-
zation; 2015. 2017.
7. Milman N, Taylor CL, Merkel J, Brannon PM. Iron status in pregnant women and women of reproductive
age in Europe. The American Journal of Clinical Nutrition. 2017; 106(suppl_6):1655S–62S. https://doi.
org/10.3945/ajcn.117.156000 PMID: 29070543
8. Ministry of National Services Regulation and Coordination (MoNHSR&C) GoP. National Nutrition Sur-
vey. 2018.
9. Baig-Ansari N, Badruddin SH, Karmaliani R, Harris H, Jehan I, Pasha O, et al. Anemia prevalence and
risk factors in pregnant women in an urban area of Pakistan. Food and nutrition bulletin. 2008; 29
(2):132–9. https://doi.org/10.1177/156482650802900207 PMID: 18693477
10. Parks S, Hoffman MK, Goudar SS, Patel A, Saleem S, Ali SA, et al. Maternal anaemia and maternal,
fetal, and neonatal outcomes in a prospective cohort study in India and Pakistan. BJOG: An Interna-
tional Journal of Obstetrics & Gynaecology. 2018. https://doi.org/10.1111/1471-0528.15585 PMID:
30554474
11. Hambidge KM, Bann CM, McClure EM, Westcott JE, Garce´s A, Figueroa L, et al. Maternal Characteris-
tics Affect Fetal Growth Response in the Women First Preconception Nutrition Trial. Nutrients. 2019; 11
(10):2534. https://doi.org/10.3390/nu11102534 PMID: 31640153PMID: 31640153
12. Anjum Anam, Manzoor Maleeha, Manzoor Nadia, Shakir HA. Prevalence of anemia during pregnancy
in district Faisalabad, Pakistan. Punjab University Journal of Zoology. 2015; 30(1):15–20.
13. Habib MA, Raynes-Greenow C, Soofi SB, Ali N, Nausheen S, Ahmed I, et al. Prevalence and determi-
nants of iron deficiency anemia among non-pregnant women of reproductive age in Pakistan. 2018; 27
(1):195. https://doi.org/10.6133/apjcn.042017.14 PMID: 29222899
14. Ullah I, Zahid Muhammad, Ismail Khan Muhammad S M. Prevalence of anemia in pregnant women in
district Karak, Khyber Pakhtunkhwa, Pakistan. International Journal of Biosciences. 2013; 3:77–83.
15. National Institute of Population Studies. Pakistan Demographic and Health Survey 2012–13. Isamabad
2013 2017 http://www.nips.org.pk/abstract_files/Priliminary%20Report%20Final.pdf.
16. Ali SA, Saleem S, Sami N, Shabbar M, Ahmed M, Rozi S, et al. Geographic access to working family
planning centers and unintended pregnancies among married women: a community based nested case
control study. Open Journal of Epidemiology. 2016; 6(1):95.
17. Ahmad M. Health care access and barriers for the physically disabled in rural Punjab, Pakistan. Interna-
tional Journal of Sociology and Social Policy. 2013.
18. Goudar SS, Carlo WA, McClure EM, Pasha O, Patel A, Esamai F, et al. The maternal and newborn
health registry study of the global network for women’s and children’s health research. International
Journal of Gynaecology and Obstretics. 2012; 118(3):190–3. https://doi.org/10.1016/j.ijgo.2012.04.022
PMID: 22738806PMID: 22738806
19. Hambidge KM, Krebs NF, Westcott JE, Garces A, Goudar SS, Kodkany BS, et al. Preconception mater-
nal nutrition: a multi-site randomized controlled trial. BMC pregnancy childbirth. 2014; 14(1):1–16.
https://doi.org/10.1186/1471-2393-14-111 PMID: 24650219PMID: 24650219
20. Bose CL, Bauserman M, Goldenberg RL, Goudar SS, McClure EM, Pasha O, et al. The Global Network
Maternal Newborn Health Registry: a multi-national, community-based registry of pregnancy outcomes.
Reproductive Health. 2015; 12(S2):S1. https://doi.org/10.1186/1742-4755-12-S2-S1 PMID:
26063166PMID: 26063166
21. Hambidge KM, Westcott JE, Garce´s A, Figueroa L, Goudar SS, Dhaded SM, et al. A multicountry ran-
domized controlled trial of comprehensive maternal nutrition supplementation initiated before concep-
tion: the Women First trial. The American journal of clinical nutrition. 2019; 109(2):457–69. https://doi.
org/10.1093/ajcn/nqy228 PMID: 30721941
22. Waninge A, Ligthart K, Kramer J, Hoeve S, van der Schans C, Haisma H. World Health Organization,
BMI classification, 2006. Measuring physical fitness. 31(3):38.
23. Namaste SM, Aaron GJ, Varadhan R, Peerson JM, Suchdev P. Methodologic approach for the bio-
markers reflecting inflammation and nutritional determinants of anemia (BRINDA) project. The
PLOS ONE | https://doi.org/10.1371/journal.pone.0239320 September 24, 2020
14 / 16
PLOS ONE Prevalence and determinants of anemia among women of reproductive age
American journal of clinical nutrition. 2017; 106(suppl_1):333S–47S. 10.3945/ajcn.116.142273 PMID:
28615254
24. Preet K, Anchinmane VT, Sankhe S. Evaluation of micro-ESR method with Westergren method for
determination of erythrocyte sedimentation rate. International Journal of Research in Medical Sciences.
2018; 6(2):628–31.
25. Ayong L, Moukoko CEE, Mbacham WF. Diagnosing Malaria: Methods, Tools, and Field Applicability.
Malaria Control and Elimination: Springer; 2019. p. 73–82.
26. Devanesan S, AlQahtani F, AlSalhi MS, Jeyaprakash K, Masilamani V. Diagnosis of thalassemia using
fluorescence spectroscopy, auto-analyzer, and hemoglobin electrophoresis—A prospective study.
Journal of infection public health. 2019; 12(4):585–90. https://doi.org/10.1016/j.jiph.2019.02.018 PMID:
30808594
27. Amin OM. Evaluation of a new system for the fixation, concentration, and staining of intestinal parasites
in fecal specimens, with critical observations on the trichrome stain. Journal of microbiological methods.
2000; 39(2):127–32. https://doi.org/10.1016/s0167-7012(99)00108-6 PMID: 10576702
28. Kamruzzaman M, Rabbani MG, Saw A, Sayem MA, Hossain MG. Differentials in the prevalence of ane-
mia among non-pregnant, ever-married women in Bangladesh: multilevel logistic regression analysis of
data from the 2011 Bangladesh Demographic and Health Survey. BMC women’s health. 2015; 15
(1):54. https://doi.org/10.1186/s12905-015-0211-4 PMID: 26219633PMID: 26219633
29. Bhandary S, Shrestha A. Prevalence of anemia among non-pregnant and pregnant women of reproduc-
tive age in Nepal. Journal of General Practice and Emergency Medicine of Nepal. 2011; 2(2):21–6.
30. Stevens GA, Finucane MM, De-Regil LM, Paciorek CJ, Flaxman SR, Branca F, et al. Global, regional,
and national trends in haemoglobin concentration and prevalence of total and severe anaemia in chil-
dren and pregnant and non-pregnant women for 1995–2011: a systematic analysis of population-repre-
sentative data. The Lancet Global Health. 2013; 1(1):e16–e25. https://doi.org/10.1016/S2214-109X
(13)70001-9 PMID: 25103581
31. Habib MA, Raynes-Greenow C, Soofi SB, Ali N, Nausheen S, Ahmed I, et al. Prevalence and determi-
nants of iron deficiency anemia among non-pregnant women of reproductive age in Pakistan. Asia
Pacific journal of clinical nutrition. 2018; 27(1):195. https://doi.org/10.6133/apjcn.042017.14 PMID:
29222899
32.
Trinh LTT, Dibley M. Anaemia in pregnant, postpartum and non pregnant women in Lak district, Daklak
province of Vietnam. Asia Pacific journal of clinical nutrition. 2007; 16(2). PMID: 17468088
33. Collinearity Diagnostics, Model Fit & Variable Contribution.
34. Senaviratna N, Cooray T. Diagnosing Multicollinearity of Logistic Regression Model. Asian Journal of
Probability and Statistics. 2019:1–9.
35. Hosmer DW Jr, Lemeshow S, Sturdivant RX. Applied logistic regression: John Wiley & Sons; 2013.
36. Parks S, Hoffman MK, Goudar SS, Patel A, Saleem S, Ali SA, et al. Maternal anaemia and maternal,
fetal, and neonatal outcomes in a prospective cohort study in India and Pakistan. BJOG: An Interna-
tional Journal of Obstetrics & Gynaecology. 2019; 126(6):737–43. https://doi.org/10.1111/1471-0528.
15585 PMID: 30554474
37. Ullah I, Zahid M, Khan MI, Shah M. Prevalence of anemia in pregnant women in district Karak, Khyber
Pakhtunkhwa, Pakistan. International Journal of Biosciences. 2013; 3:77–83.
38. Nguyen PH, Gonzalez-Casanova I, Nguyen H, Pham H, Truong T, Nguyen S, et al. Multicausal etiology
of anemia among women of reproductive age in Vietnam. European journal of clinical nutrition. 2015; 69
(1):107–13. https://doi.org/10.1038/ejcn.2014.181 PMID: 25205323
39. Gebremedhin S, Enquselassie F. Correlates of anemia among women of reproductive age in Ethiopia:
evidence from Ethiopian DHS 2005. Ethiopian Journal of Health Development. 2011; 25(1):22–30.
40.
Thomson CA, Stanaway JD, Neuhouser ML, Snetselaar LG, Stefanick ML, Arendell L, et al. Nutrient
intake and anemia risk in the women’s health initiative observational study. Journal of the American Die-
tetic Association. 2011; 111(4):532–41. https://doi.org/10.1016/j.jada.2011.01.017 PMID: 21443985
41. Bhargava A, Bouis HE, Scrimshaw N. Dietary intakes and socioeconomic factors are associated with
the hemoglobin concentration of Bangladeshi women. The Journal of nutrition. 2001; 131(3):758–64.
https://doi.org/10.1093/jn/131.3.758 PMID: 11238756
42. Monsen ER. Iron nutrition and absorption: dietary factors which impact iron bioavailability. Journal of
the American Dietetic Association. 1988; 88(7):786. PMID: 3290310
43.
Liabsuetrakul T. Is international or Asian criteria-based body mass index associated with maternal
anaemia, low birthweight, and preterm Births among Thai population?—an observational study.
Journal of health, population, nutrition. 2011; 29(3):218. https://doi.org/10.3329/jhpn.v29i3.7869 PMID:
21766557
PLOS ONE | https://doi.org/10.1371/journal.pone.0239320 September 24, 2020
15 / 16
PLOS ONE Prevalence and determinants of anemia among women of reproductive age
44.
Lander RL, Hambidge KM, Westcott JE, Tejeda G, Diba TS, Mastiholi SC, et al. Pregnant Women in
Four Low-Middle Income Countries Have a High Prevalence of Inadequate Dietary Intakes That Are
Improved by Dietary Diversity. Nutrients. 2019; 11(7). https://doi.org/10.3390/nu11071560 PMID:
31295916PMID: 31295916
45. Wabe NT, Mohammed MA. What science says about khat (Catha edulis Forsk)? Overview of chemistry,
toxicology and pharmacology. Journal of Experimental & Integrative Medicine. 2012; 2(1).
46. Al-Sayes F, Gari M, Qusti S, Bagatian N, Abuzenadah A. Prevalence of iron deficiency and iron defi-
ciency anemia among females at university stage. Journal of Medical Laboratory Diagnosis. 2011; 2
(1):5–11.
47. Pasricha S-R, Caruana SR, Phuc TQ, Casey GJ, Jolley D, Kingsland S, et al. Anemia, iron deficiency,
meat consumption, and hookworm infection in women of reproductive age in northwest Vietnam. The
American journal of tropical medicine hygiene. 2008; 78(3):375–81. PMID: 18337329
48. Pasricha S-R, Caruana SR, Phuc TQ, Casey GJ, Jolley D, Kingsland S, et al. Anemia, iron deficiency,
meat consumption, and hookworm infection in women of reproductive age in northwest Vietnam. The
American journal of tropical medicine and hygiene. 2008; 78(3):375–81. PMID: 18337329
49. Heath A-LM, Fairweather-Tait S. Clinical implications of changes in the modern diet: iron intake, absorp-
tion and status. Best practice research Clinical haematology. 2002; 15(2):225–41. PMID: 12401305
50. Martinez-Torres C, Layrisse M. Iron absorption from veal muscle. The American journal of clinical nutri-
tion. 1971; 24(5):531–40. https://doi.org/10.1093/ajcn/24.5.531 PMID: 5578513
51. Harvey LJ, Armah CN, Dainty JR, Foxall RJ, Lewis DJ, Langford NJ, et al. Impact of menstrual blood
loss and diet on iron deficiency among women in the UK. British Journal of Nutrition. 2005; 94(4):557–
64. https://doi.org/10.1079/bjn20051493 PMID: 16197581
52. Bentley M, Griffiths P. The burden of anemia among women in India. European journal of clinical nutri-
tion. 2003; 57(1):52–60. https://doi.org/10.1038/sj.ejcn.1601504 PMID: 12548297
53. Ghose B, Tang S, Yaya S, Feng Z. Association between food insecurity and anemia among women of
reproductive age. PeerJ. 2016; 4:e1945. https://doi.org/10.7717/peerj.1945 PMID: 27168968
54. Harding KL, Aguayo VM, Namirembe G, Webb P. Determinants of anemia among women and children
in Nepal and Pakistan: An analysis of recent national survey data. Maternal & child nutrition. 2018; 14:
e12478. https://doi.org/10.1111/mcn.12478 PMID: 28857410PMID: 28857410
55. Young SL, Khalfan SS, Farag TH, Kavle JA, Ali SM, Hajji H, et al. Association of pica with anemia and
gastrointestinal distress among pregnant women in Zanzibar, Tanzania. The American journal of tropi-
cal medicine and hygiene. 2010; 83(1):144–51. https://doi.org/10.4269/ajtmh.2010.09-0442 PMID:
20595493
56. Mona´rrez-Espino J, Martı´nez H, Greiner T. Iron deficiency anemia in Tarahumara women of reproduc-
tive-age in Northern Mexico. salud pu´blica de me´xico. 2001; 43:392–401. https://doi.org/10.1590/
s0036-36342001000500002 PMID: 11763687
57. Medicine Io, Lactation NRCSoND, Lactation IoMSoNd, Lactation IoMSo, Resources USH, Administra-
tion S, et al. Nutrition during lactation: Henry Holt and Company; 1991.
58. Hassan U. Beliefs of women regarding food during lactation. Pakistan Armed Forces Medical Journal.
2013; 63(2):266–70.
59. Niloufer Sultan Ali SIA, Noor Rahat. Womens’ beliefs and practices regarding food restrictions during
pregnancy and lactation: A Hospital Based Study. Journal of Ayub Medical College Abbottabad. 2004;
16(3).
60. Dewey KG, Cohen RJ, Brown KH, Rivera LL. Effects of exclusive breastfeeding for four versus six
months on maternal nutritional status and infant motor development: results of two randomized trials in
Honduras. The Journal of nutrition. 2001; 131(2):262–7. https://doi.org/10.1093/jn/131.2.262 PMID:
11160544
61. Uche-Nwachi E, Odekunle A, Jacinto S, Burnett M, Clapperton M, David Y, et al. Anaemia in pregnancy:
associations with parity, abortions and child spacing in primary healthcare clinic attendees in Trinidad
and Tobago. African health sciences. 2010; 10(1):66. PMID: 20811527
62. Mardani M, Rezapour S, Ahmadipour S, Mohsenzadeh A, Khalkhali Rad A, Roosta S, et al. Prevalence
of anemia and its risk factors among pregnant women in Khorramabad (Iran) 2010–2014. The Journal
of Maternal-Fetal & Neonatal Medicine. 2017; 30(7):826–9.
63. Bahaj AA. Prevalence of and risk factors of anemia among pregnantwomenMukalla-Hadhramout-
Yemen. Tikret Journal of Pharmaceutical Sciences. 2006; 2(1):71–7.
64. Chowdhury S, Rahman M, Moniruddin A. Anemia in pregnancy. Medicine Today. 2014; 26(1):49–52.
65. Akinlaja OJ. Hematological changes in pregnancy-The preparation for intrapartum blood loss. Interna-
tional Journal of Obstretics and Gynecology 2016; 4(3):00109.
PLOS ONE | https://doi.org/10.1371/journal.pone.0239320 September 24, 2020
16 / 16
PLOS ONE |
10.1371_journal.pone.0231405 | RESEARCH ARTICLE
Design of thermal wind sensor with constant
power control and wind vector measurement
method
Congning Liu1, Yunbo Shi1, Xiaoyu YuID
2*, Tengxi Wang3, Maria D. King4
1 The Higher Educational Key Laboratory for Measuring & Control Technology and Instrumentations of
Heilongjiang Province, School of Measurement-Control Technology & Communications Engineering, Harbin
University of Science and Technology, Harbin, Heilongjiang, China, 2 School of Atmospheric Sciences, Sun
Yat-sen University, Zhuhai, Guangdong, China, 3 Texas A&M Transportation Institute, Bryan, Texas, United
States of America, 4 Department of Biological and Agricultural Engineering, Texas A&M University, Texas,
United States of America
* sysyuxiaoyu@163.com
Abstract
This paper presents a conceptual wind vector detector for measuring the velocity and direc-
tion of wind in enclosed or semi-enclosed large spaces. Firstly, a thermal wind sensor with
constant power control was manufactured and then used as a wind velocity sensing unit.
Secondly, a sensor bracket equipped with three thermal wind sensors was designed, the
fluid dynamic response regularity of the measured wind field to the sensor bracket was ana-
lyzed using ANSYS Fluent CFD software, and then its structural parameters were optimized
to improve measurement accuracy. The sensor bracket was fabricated via 3D printing.
Finally, a unique wind vector measurement method was developed for the wind vector
detector. Experimental results showed that the measured velocity range of the thermal wind
sensor satisfied the requirements of being within 0–15 m/s with an accuracy of ±0.3 m/s,
and the wind direction angle range of the wind vector detector was within 0–360˚ with an
accuracy of ±5˚. By changing the applied power control value of the thermal wind sensor
and structural parameters of the sensor bracket, the measurement range and accuracy of
the wind vector detector can be adjusted to suit different applications.
Introduction
For enclosed or semi-enclosed large spaces such as urban underground pipe networks, mine
laneways and large storage warehouses, additional ventilation is required to avoid the accumu-
lation of toxic gases or dust [1,2]. On the one hand, when the wind velocity is too low, toxic
gases or dust cannot be discharged, which may easily cause poisoning or dust explosion. On
the other hand, relatively higher wind velocities will cause dust dispersion, which will adversely
affect the reliable operation of equipment [3]. In such scenarios, the wind velocities are always
less than 15 m/s [4–6]. Furthermore, the wind direction affects the ventilation performance
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: Liu C, Shi Y, Yu X, Wang T, King MD
(2020) Design of thermal wind sensor with
constant power control and wind vector
measurement method. PLoS ONE 15(4):
e0231405. https://doi.org/10.1371/journal.
pone.0231405
Editor: Wei Yao, Huazhong University of Science
and Technology, CHINA
Received: January 28, 2020
Accepted: March 22, 2020
Published: April 14, 2020
Copyright: © 2020 Liu et al. This is an open access
article distributed under the terms of the Creative
Commons Attribution License, which permits
unrestricted use, distribution, and reproduction in
any medium, provided the original author and
source are credited.
Data Availability Statement: All relevant data are
within the manuscript and its Supporting
Information files.
Funding: This work was supported by the Basic
Scientific Research Program of China grant
JCKY2017412C003 and the Young Scientists Fund
grant 61501149. The funders had no role in study
design, data collection and analysis, decision to
publish, or preparation of the manuscript.
Competing interests: The authors have declared
that no competing interests exist.
PLOS ONE | https://doi.org/10.1371/journal.pone.0231405 April 14, 2020
1 / 22
PLOS ONE Thermal wind sensor and wind vector measurement
[7]. Therefore, wind data form an important basis for ventilation management and must be
measured in real time [8].
Wind is a three-dimensional vector generated by air flows. However, in actual measure-
ments, wind is often considered a two-dimensional vector [9]. It is determined by two parame-
ters, namely the wind velocity (the magnitude of the wind vector) and wind direction (the
angle of the wind vector) [10,11]. Over the years, various types of detectors have been devel-
oped to measure the wind vector [12,13]. Wind vector measurements are mainly performed
via mechanical, Pitot tube, ultrasonic, and thermal methods [14]. The mechanical method is
mainly used for meteorological detection and large space turbulence measurements. Based on
their structural characteristics, there are two types of mechanical models: propeller (or vane)
and cup. The first one rotates itself in a plane perpendicular to the measured wind vector, and
the second one rotates itself in a plane parallel to the measured wind vector [15]. These models
have the following advantages: simple structure and process, low cost, convenient daily main-
tenance, and strong anti-interference ability. However, a slightly high-velocity start-up wind is
required, resulting in poor accuracy and sensitivity for low-velocity wind [16]. The Pitot tube
method has the advantages of no inertia delay and high sensitivity, but it requires a homoge-
nous wind field. Therefore, it is easy for the detector to perform inaccurate measurements,
especially in a wind field with dust [17,18]. The ultrasonic method has the advantages of a
wide measurement range and fast response, while being sturdy and durable [19]. As it calcu-
lates the wind vector using the propagation time of ultrasonic waves in the measured fluid
field, it can be easily affected by temperature changes [20]. The thermal method includes hot
wire and heat sensitive, it is compact and easy to integrate into other devices [21,22]. For
example, MEMS wind sensors based on the heat transfer principle are being extensively
researched [23]. However, it has problems such as heat loss [24]. In recent years, many compa-
nies, including TSI in the United States, TESTO-AG in Germany, KIMO in France, DANTEC
in Denmark, and KANOMAX in Japan, manufactured various types of detectors for measur-
ing the wind vector [25]. The detectors mainly differ in terms of their range, accuracy, operat-
ing environment, and application scenarios. Compared to the variety of wind vector detectors
available, the mechanical type and the thermal type have more applications. Considering the
various advantages and disadvantages of these two types, a novel wind vector detector
equipped with wind velocity sensing units of the thermal type and a specific mechanical part
rotating in a plane parallel to the measured wind vector was designed. A preliminary survey
revealed that there is no reported research on a wind vector detector combining a unique
mechanical structure and several wind velocity sensing units of the thermal type, while consid-
ering the fluid dynamic response regularity of the measured wind field to a specific mechanical
part. This conceptual wind vector detector provides a new method for wind vector
measurement.
Therefore, in this work, a wind vector monitoring system was designed. As shown in Fig 1,
this system comprises a wind vector detector, data transceiver and remote monitoring center.
Fig 1. Schematic of the wind vector monitoring system.
https://doi.org/10.1371/journal.pone.0231405.g001
PLOS ONE | https://doi.org/10.1371/journal.pone.0231405 April 14, 2020
2 / 22
PLOS ONE Thermal wind sensor and wind vector measurement
Additionally, data are transmitted between the detector and the computer via wireless commu-
nication modules (SHUNZHOU SZ05 module); the management level software handles the
storage, display, and inquiry functions of measured data.
In practical applications, if the remote monitoring center is considerably far away from the
wind vector detector, the wireless communication module of the wind vector detector should
be replaced by a 4G network module (HUAWEI ME909s-821 module). Furthermore, the com-
puter of the remote monitoring center should be connected to the Internet, and a cloud server
platform needs to be established. Thus, data interaction between the wind vector detector and
remote monitoring center can be realized even when the distance between the two is signifi-
cantly large.
The contribution and novelty of this work can be summarized as follows:
• In terms of the heating unit design, based on the relationship between the temperature in the
special nickel-chromium wire coil and heating power, as well as its thermal field distribution
characteristics at different wind velocities, the coil is innovatively used as the heating unit of
the thermal wind sensor with constant power control.
• In terms of the structural design, for improving the efficiency of research, a sensor bracket
equipped with three thermal wind sensors was fabricated via 3D printing. The conceptual
wind vector detector mainly comprises a unique mechanical structure and three wind veloc-
ity sensing units; it overcomes the limitation of mechanical wind detectors, which are unable
to measure slightly low-velocity wind.
• In terms of measurement methods, the fluid dynamic response regularity of the measured
wind field to the sensor bracket was analyzed; thereafter, the relationship between the distri-
bution characteristics of the wind velocity contours and the structural parameters of the sen-
sor bracket was established. Using these simulation results, a unique wind vector
measurement method was developed; this developed method can be regarded as a novel
approach for the wind vector measurement.
This remainder of the paper is organized as follows. Firstly, the theory of wind measure-
ment is presented; a thermal wind sensor and a wind vector detector are described, along with
a unique wind vector measurement method. Subsequently, the details of the experimental plat-
forms are provided, including tests performed on the thermal wind sensor and wind vector
detector, and related results and discussions. Finally, the conclusions are presented.
Materials and methods
Thermal wind sensor
A thermal wind sensor was designed to measure wind velocity. Firstly, the measurement prin-
ciples of the sensor were studied. Secondly, the thermal field distribution of the special nickel-
chromium wire coil used in the sensor was analyzed, and a strategy of constant power control
for the heating unit of the thermal wind sensor was subsequently developed. Finally, the circuit
for wind velocity measurement was designed.
Measurement principles.
In a thermal wind sensor, a certain voltage or current is applied
to a heating unit, whose temperature is increased to a fixed value. This temperature is continu-
ously measured by a temperature sensor. When air flows, the temperature of the heating unit
is reduced. This reduction in temperature is proportional to the wind velocity. Hence, the heat
loss can characterize wind velocity. Therefore, some parameters such as the physical character-
istics of the heating unit, temperature difference between the heating unit and measured wind
field, and wind velocity can be used to develop a specific mathematical model.
PLOS ONE | https://doi.org/10.1371/journal.pone.0231405 April 14, 2020
3 / 22
PLOS ONE Thermal wind sensor and wind vector measurement
The heat transfer modes mainly include conduction, convection, and radiation [26]. The
sensor response time can be reduced by increasing the conduction and convection efficiencies.
The range and sensitivity of sensor can be improved by increasing the temperature difference
between the heating unit and measured wind field.
Using the heat transfer principle [27] and King’s law [28], the following relationship may be
obtained:
P ¼ I2R ¼ ðA þ B
p
ffiffiffi
v
ÞðTs (cid:0) TeÞ
ð1Þ
where P is the heating power of the sensor; I is the current flowing through the heating unit; R
is the heating resistance; A and B are constants related to the material and structure, respec-
tively; v is the wind velocity; Ts is the temperature of the heating unit; Te is the temperature of
the measured wind field.
If ΔT is the temperature difference between the heating unit and measured wind field, it
can be expressed as
DT ¼ Ts (cid:0) Te
The wind velocity measurement results can then be directly calculated as follows:
v ¼ ½ðP=DT (cid:0) AÞ=B�2
ð2Þ
ð3Þ
Constant power control strategy. The heating unit was designed using a nickel-chro-
mium wire coil, with a wire diameter of 100 μm, coil diameter of 2 mm, and resistance value of
25 O.
The thermal field distribution of the coil with a built-in platinum resistance thermometer
(PT100) was analyzed using FOTRIC AnalyzIR software [29]. The relationship between the
heating power of the coil and surface temperature of PT100 is shown in Fig 2.
Fig 2. Relationship between the heating power of the coil and surface temperature of PT100.
https://doi.org/10.1371/journal.pone.0231405.g002
PLOS ONE | https://doi.org/10.1371/journal.pone.0231405 April 14, 2020
4 / 22
PLOS ONE Thermal wind sensor and wind vector measurement
According to these temperature characteristics, when the heating power of the coil was
below 810 mW, the temperature change was approximately linear. However, at heating powers
above 810 mW, the temperature tended to be nearly constant. Considering the range and accu-
racy required of the wind sensor, as well as the need to reduce power consumption and heat
loss of the coil, a heating power of 160 mW was applied. In this case, PT100 within the coil sta-
bly absorbed heat, and the surface thermal field distribution of PT100 was approximately uni-
form. When the wind velocity was lower than 15 m/s, the thermal field in the coil changed
conspicuously for the measured fluid fields at different wind velocities. However, at wind
velocities above 15 m/s, the thermal field tended to stabilize. Therefore, the goal of measuring
the velocity range of the thermal wind sensor was successfully achieved with the lowest possi-
ble power consumption. For fluid fields with different wind velocities, the temperature distri-
butions of the coil with a built-in PT100 thermometer were observed using a FOTRIC 220
thermal video camera from the top of the sensor (Fig 3).
The physical characteristics of the nickel-chromium coil caused its resistance value to
change with temperature. To guarantee a constant release of heat in the coil, it is necessary to
design a constant power control circuit, similar to the one in Fig 4.
The bridge comprises the nickel-chromium wire coil (RH) and three precision resistors (Ra,
Rb, and Rc) with ultra-low temperature drift. The temperature coefficient of these precision
resistors is 2 ppm/˚C, with an accuracy of 0.01%. Calculations based on the circuit yielded
DU2 ¼
VCC (cid:0) UC
þ 1
Rb
Rc
(cid:0)
VCC (cid:0) UC
þ 1
RH
Ra
ð4Þ
The resistance of the bridge was based on the change in the temperature of the coil. When
wind velocity increased, the temperature of the coil decreased owing to heat loss, the resistance
Fig 3. Top view of temperature distributions inside the coil at different wind velocities. (A) Wind velocity of 0 m/s. (B) Wind velocity of 5 m/s. (C) Wind velocity of
10 m/s. (D) Wind velocity of 15 m/s. (E) Wind velocity of 16 m/s. (F) Wind velocity of 17 m/s.
https://doi.org/10.1371/journal.pone.0231405.g003
PLOS ONE | https://doi.org/10.1371/journal.pone.0231405 April 14, 2020
5 / 22
PLOS ONE Thermal wind sensor and wind vector measurement
Fig 4. Schematic of constant power control circuit.
https://doi.org/10.1371/journal.pone.0231405.g004
value of RH decreased, and the voltage value of ΔU2 decreased. In contrast, when the wind
velocity decreased, the temperature in the coil increased, the resistance value of RH increased,
and the voltage value of ΔU2 increased.
The incremental digital proportional-integral-derivative (PID) adjustment, shown in Fig 5,
had a lead and lag correction that could improve the transient response speed and stability of
the constant power control circuit [30].
The incremental digital PID control equation is given by
uðkÞ ¼ uðk (cid:0) 1Þ þ KP½eðkÞ (cid:0) eðk (cid:0) 1Þ� þ KI½eðkÞ (cid:0) eðk (cid:0) 1Þ� þ KD½eðkÞ (cid:0) 2eðk (cid:0) 1Þ þ eðk
ð5Þ
(cid:0) 2Þ�
where u(k) is the kth time output of the incremental digital PID controller, and e(k) is the dif-
ference between the kth time actual output value and the target value.
The controlled object of the constant power control circuit (Fig 4) was analyzed. The aim of
its closed-loop control strategy was to implement the feedback and correction of ΔU3/
(VCC−UC). The ADC chip recorded voltage values of ΔU3, VCC, and UC at each system sam-
pling period (T). Thereafter, the microprocessor control unit (MCU) received these recorded
values. Based on the deviation of ΔU3/(VCC−UC) for three consecutive sampling periods, the
output values of the DAC chip were adjusted using the incremental digital PID algorithm;
thus, I0 was adjusted. Therefore, the nickel-chromium wire coil (RH) was controlled so as to
work in a constant power mode.
Fig 5. Block diagram of the incremental digital PID controller.
https://doi.org/10.1371/journal.pone.0231405.g005
PLOS ONE | https://doi.org/10.1371/journal.pone.0231405 April 14, 2020
6 / 22
PLOS ONE Thermal wind sensor and wind vector measurement
Table 1. Parameters description.
System parameters
RH
Ra, Rb, Rc
VCC
Ad
PID control parameters
T
TI
TD
KP
KI
KD
Other parameters
4U2
4U3
Resistance of the nickel-chromium wire coil
Precision resistances of 25 O, 100 kO and 100 kO, respectively
Operating voltage (5V)
Amplification gain of the amplifier
System sampling period
Integral time
Derivative time
Proportional gain of ΔU3/(VCC−UC)
Integral gain of ΔU3/(VCC−UC)
Derivative gain of ΔU3/(VCC−UC)
Output voltage of the bridge
Output voltage of the amplifier
https://doi.org/10.1371/journal.pone.0231405.t001
All system parameters, PID controller parameters, and other parameters [31,32] used in
this study are summarized in Table 1.
To simplify the tuning process for the PID control parameters, a virtual instrument for the
adjustment of PID parameters was designed using LabVIEW software. The data transceiver
module was connected to the virtual instrument computer through a universal asynchronous
receiver/transmitter (UART) port and then the system of PID parameter adjustment was ini-
tialized. As shown in Fig 6, the data transceiver module collected the voltage values of ΔU3,
VCC, and UC, the virtual instrument computer adjusted the PID control parameters using the
received data, and the output values of the PID controller [33] were converted through an
appropriate function. Subsequently, the data transceiver module received data from the virtual
instrument computer and drove the power amplifier transistor to provide I0, which was
required by RH, thereby achieving the goal of constant power control.
Fig 6. Schematic of the PID parameter adjustment system.
https://doi.org/10.1371/journal.pone.0231405.g006
PLOS ONE | https://doi.org/10.1371/journal.pone.0231405 April 14, 2020
7 / 22
PLOS ONE Thermal wind sensor and wind vector measurement
The PID control parameters were tuned via the trial and error method. The tuning process
was performed according to the operation sequence of "proportion first, then integration, and
last differentiation", while observing the response curve. Firstly, the proportional gain (KP) was
tuned, integral time (TI) was set to the maximum value, and derivative time (TD) was set to 0,
to ensure that the PID controller worked under pure proportional action. KP was gradually
increased, and the response characteristics of ΔU3/(VCC−UC) were observed. A response curve
with fast response was obtained, and at this time, the response curve was slightly over-tuned.
Secondly, KP was reduced to 60%, TI was gradually reduced until a satisfactory response curve
was obtained, and KP was fine-tuned accordingly. The system yielded a good dynamic perfor-
mance, and the static error of the system could be reduced. Thirdly, TD was gradually
increased, and KP and TI were fine-tuned accordingly. The over-tuned amount and stability of
the response curve were observed, and satisfactory control results were obtained by repeating
the trial and error method. Finally, the PID control parameters were obtained as KP = 260.5, KI
= KPT/TI = 51.3, and KD = KPTD/T = 761.7 in the virtual instrument computer. Based on Eq
(5), the incremental digital PID algorithm was implemented using the MCU (STMICROE-
LECTRONICS STM32F107 chip).
Circuit solutions. The thermal wind sensor circuit, as shown in Fig 7, mainly includes a
constant-power heating circuit, constant current source circuit, and differential amplifier
circuit.
Here, Rs is the PT100 resistance for detecting the heat loss inside the coil, Re is the PT100
resistance for detecting environmental temperature, and the remaining two resistances are
precision resistances each with a resistance of 200 kO. The manufactured thermal wind sensor
is depicted in Fig 8.
The differential bridge circuit was powered by a constant current source chip (LM134). As
the resistance values of Rs and Re were significantly smaller than the two precision resistances
on the other arm of the bridge, the value of the current at Rs and Re was approximately equal to
the output current of the constant current source circuit. Therefore, ΔU1 was calculated as fol-
lows:
DU1 ¼
1
2
IðRs (cid:0) ReÞ
ð6Þ
Fig 7. Wind velocity measurement circuit.
https://doi.org/10.1371/journal.pone.0231405.g007
PLOS ONE | https://doi.org/10.1371/journal.pone.0231405 April 14, 2020
8 / 22
PLOS ONE Thermal wind sensor and wind vector measurement
Fig 8. Photograph of the thermal wind sensor. (A) Schematic. (B) Front view. (C) Top view.
https://doi.org/10.1371/journal.pone.0231405.g008
The relationship between the resistance of PT100 and temperature is expressed as
Rt ¼ R0ð1 þ aTÞ
where R0 is the resistance value of PT100 at 0˚C; T is the temperature of the measured wind
field; and α is the temperature coefficient of PT100.
From Eqs (2), (6), and (7), ΔU1 can be obtained as
DU1 ¼
1
2
IR0aDT
Thereafter, from Eqs (3) and (8), v can be obtained as
!2
(cid:0) A
aPIR0
2DU1
B
v ¼
ð7Þ
ð8Þ
ð9Þ
As indicated in Eq (9), only v and ΔU1 are variables. Hence, a mathematical relationship
was derived between the output voltage value (ΔU1) in the differential bridge circuit and the
wind velocity value (v) of the measured wind field.
Proposed methods
To further measure the wind direction, a wind vector detector was manufactured. Firstly, a
sensor bracket equipped with three thermal wind sensors was designed. The turbulent k-ε
model was used to simulate and analyze the fluid dynamic response regularity of the measured
wind field to the sensor bracket, and then its structural parameters were set. The sensor bracket
was fabricated via 3D printing. Secondly, based on the simulation results, a unique wind vector
measurement method was developed, and the measurement process was established.
Simulation analysis. The sensor bracket was designed using AutoCAD software, as
shown in Fig 9.
It consists of a channel (BC) and two cantilever beams (BL and BR). BC, BL, and BR were
equipped with thermal wind sensors. The structures of BL and BR are completely symmetrical.
According to the size of the thermal wind sensor that was manufactured (Fig 7), the size of BC
(W1 × H1) was set as 22 mm × 22 mm. Then, the length of BC (L1) was calculated using the
simulation results obtained from ANSYS FLUENT CFD software.
PLOS ONE | https://doi.org/10.1371/journal.pone.0231405 April 14, 2020
9 / 22
PLOS ONE
Thermal wind sensor and wind vector measurement
Fig 9. Sensor bracket structure.
https://doi.org/10.1371/journal.pone.0231405.g009
The sensor bracket was produced using a ZRapid iSLA660 3D printer via stereolithography
(SLA). SLA is an additive manufacturing process that involves focusing an ultraviolet (UV)
laser on a photopolymer resin material. The photopolymer is sensitive to UV light, so the resin
is photochemically solidified and forms a single layer of the desired 3D object. This process is
repeated for obtaining each layer until the 3D object is completed. The resolution of the used
3D printer was 0.05 mm (print layer thickness), and the internal support mesh structure of the
sensor bracket was strictly evaluated for eliminating thin-wall problems (minimum wall thick-
ness was 0.8 mm) [34]. The photograph of the sensor bracket equipped with three thermal
wind sensors is shown in Fig 10.
The model of the sensor bracket is explained as follows. The nickel-chromium wire coil of
the thermal wind sensor (SBC) was placed in the inner central area of the channel (BC). The
internal center point of the coil was the coordinate origin of the model, and the directions of
three coordinate axes are shown in Fig 10. The horizontal and vertical planes passing through
the coordinate origin in BC were defined as the XZ and XY planes, respectively. The velocity
contours of the XZ and XY planes were used to analyze the wind velocity distribution in BC.
For studying the effect of wind velocity on the three thermal wind sensors when the sensor
bracket was rotated horizontally, the fluid dynamic response regularity of the measured wind
field to the sensor bracket was analyzed using ANSYS FLUENT CFD software [35–37].
There were irregular pulsations caused by vortices of different sizes in turbulence; that is,
the spatial distribution of the wind vector changed randomly with time and space. The larger-
scale vortices mainly depended on the boundary conditions of the fluid field, and the smaller-
scale vortices mainly depended on the magnitude of the viscous force. They were the main
PLOS ONE | https://doi.org/10.1371/journal.pone.0231405 April 14, 2020
10 / 22
PLOS ONE Thermal wind sensor and wind vector measurement
Fig 10. Photograph of the sensor bracket equipped with three wind sensors.
https://doi.org/10.1371/journal.pone.0231405.g010
causes of the low-frequency and high-frequency pulsations, respectively. In the fully expanded
turbulence area, the size of the turbulence changed at any time. The larger- scale vortices con-
tinuously transferred energy to the smaller-scale vortices under the mutual force between the
vortices, and then smaller-scale vortices receiving energy continuously disappeared under the
action of the viscous force. Simultaneously, under the joint effect of the boundary conditions,
wind disturbance, and velocity gradient, new vortices were continuously generated, resulting
in turbulent motion. Considering the actual wind field inside the sensor bracket, the turbulent
k-ε model was suitable for analyzing the distribution of the wind vector field [38].
When the wind flowed vertically through BC along the negative z-axis, the wind direction
changed slightly at the entrance of BC. This resulted in the formation of a wind field around
the inner center of BC. The higher the velocity of the measured wind field, the more obvious
was the distribution of wind velocity in the inner central area of BC. Therefore, the upper limit
of the velocity range (15 m/s) was selected as the wind velocity value of the measured field.
When wind flowed vertically through BC along the negative z-axis at 15 m/s for different
lengths of BC, the average wind velocities in its inner central area could be determined via sim-
ulation (Table 2).
It can be concluded that when the length of BC was 110 mm, the wind flowing vertically
along the negative z-axis had the weakest impact on its inner central area. Consequently, the
Table 2. Average wind velocity in the inner central area of BC for its different lengths.
length(mm)
velocity(m/s)
30
9.14
40
7.15
50
6.49
60
2.94
70
1.54
80
1.50
90
1.55
100
1.03
110
0.17
120
0.45
130
0.31
https://doi.org/10.1371/journal.pone.0231405.t002
PLOS ONE | https://doi.org/10.1371/journal.pone.0231405 April 14, 2020
11 / 22
PLOS ONE Thermal wind sensor and wind vector measurement
Fig 11. Cloud chart of the wind field when the length of BC is 100 mm. (A) Velocity cloud of the XZ plane in BC when wind flows vertically through
BC along the negative z-axis at 15 m/s. (B) Velocity cloud of the XY plane in BC when wind flows vertically through BC along the negative z-axis at
15 m/s.
https://doi.org/10.1371/journal.pone.0231405.g011
length of BC was set to 110 mm. At the same time, the two other representative lengths of BC
were selected to analyze the velocity contours of the wind field in BC using ANSYS Fluent
CFD-Post software.
Based on the distributions of the velocity contours for different lengths of BC (Figs 11–13),
the difference in the wind velocity distribution within the entire area of BC was relatively obvi-
ous; moreover, it was evident that the distribution of the wind velocity was the smallest in the
inner central area of BC with a length of 110 mm. This observation is consistent with the simu-
lation results presented in Table 2. Therefore, it was clear that the characteristics of the velocity
contours were irregular, and the distribution of wind velocity was related to the length of the
channel (BC). When the length of BC was set to 110 mm, the average wind velocity in the
inner central area of BC (the area where the nickel-chromium wire coil was located) was less
than 0.25 m/s.
Fig 12. Cloud chart of the wind field when the length of BC is 110 mm. (A) Velocity cloud of the XZ plane in BC when wind flows vertically through
BC along the negative z-axis at 15 m/s. (B) Velocity cloud of the XY plane in BC when wind flows vertically through BC along the negative z-axis at
15 m/s.
https://doi.org/10.1371/journal.pone.0231405.g012
PLOS ONE | https://doi.org/10.1371/journal.pone.0231405 April 14, 2020
12 / 22
PLOS ONE Thermal wind sensor and wind vector measurement
Fig 13. Cloud chart of the wind field when the length of BC is 120 mm. (A) Velocity cloud of the XZ plane in BC when wind flows vertically through
BC along the negative z-axis at 15 m/s. (B) Velocity cloud of the XY plane in BC when wind flows vertically through BC along the negative z-axis at
15 m/s.
https://doi.org/10.1371/journal.pone.0231405.g013
For the simulation results shown in Fig 12(A), the entire wind field where the sensor
bracket was placed was further analyzed. The velocity vector photograph of the measured
wind field to the sensor bracket is presented in Fig 14.
It is clear that the wind velocity around the area of BL (windward direction) is greater than
the wind velocity around the area of BR (leeward direction). Based on this result, the wind
direction can be measured (from the windward direction to the leeward direction).
The sensor bracket was placed in the measured wind field with different wind velocities.
When BC was rotated parallel to the wind direction, the average wind velocity around the
areas where three independent thermal wind sensors (SBL, SBC, and SBR) were placed was
determined via simulation (Table 3).
When wind flowed parallel to BC, the wind direction changed at the inlet of BC. It resulted
in the formation of a boundary layer in the channel, and then, the wind velocity around the
Fig 14. Velocity vector photograph of the measured wind field to the sensor bracket.
https://doi.org/10.1371/journal.pone.0231405.g014
PLOS ONE | https://doi.org/10.1371/journal.pone.0231405 April 14, 2020
13 / 22
PLOS ONE Thermal wind sensor and wind vector measurement
Table 3. Average wind velocity of three independent areas.
Wind velocity (m/s)
0.5
1
1.5
2
3
4
5
6
7
8
9
10
11
12
13
14
15
SBL(m/s)
0.52
SBC(m/s)
0.58
SBR(m/s)
0.52
1.03
1.55
2.05
3.07
4.09
5.10
6.12
7.13
8.15
9.16
10.18
11.19
12.21
13.22
14.24
15.25
1.10
1.45
1.93
2.86
3.83
4.80
5.79
6.76
7.73
8.69
9.69
10.67
11.61
12.60
13.60
14.55
1.03
1.54
2.05
3.07
4.08
5.10
6.12
7.13
8.15
9.16
10.18
11.19
12.21
13.22
14.23
15.25
https://doi.org/10.1371/journal.pone.0231405.t003
area of SBC was affected. At the same time, owing to the effect of the cantilever beam boundary
layer, the wind velocities around the areas of SBL and SBR were also affected. The values of the
wind velocities at SBL and SBR had smaller deviations than the values of the wind velocity at
SBC. Therefore, the actual values of wind velocity were obtained by correcting the average val-
ues of the wind velocities at SBL and SBR. From the data in Table 3, the following fitting equa-
tion was obtained using the least square method:
f ðvaÞ ¼ 1:016va þ 0:02093
ð10Þ
where va is the average value of the wind velocity around the areas of SBL and SBR, and f (va) is
the actual value of the wind velocity in the measured field.
Measurement methods. To verify the effectiveness and practicality of the wind vector
measurement method developed in this study, the wind vector detector was manufactured
according to application requirements. It consists of a circuit board for data acquisition and
control, stepper motor, through-hole conductive slip ring, flange coupling, and sensor bracket
(Figs 15 and 16).
The flow chart for the measurement method performed using the wind vector detector is
shown in Fig 17 and includes the following steps:
• The wind vector detector was started, and the sensor bracket was calibrated to the initial
position by the through-beam infrared photoelectric switch. At this time, the initial angle
was set as 0˚.
• The stepper motor drove the sensor bracket to rotate in a clockwise direction.
• When the output voltage of SBC (UBC) just reached its maximum value (that is, the wind
velocity measured by the thermal wind sensor just reached the minimum value.), the wind
direction was perpendicular to the channel (BC). At this moment, BC turned by an angle α.
• The output voltage of SBL was UBL, and that of SBR was UBR. If UBL > UBR, the predicted
value of the wind direction angle was φ = α + 90˚. Otherwise, it was φ = α + 270˚.
PLOS ONE | https://doi.org/10.1371/journal.pone.0231405 April 14, 2020
14 / 22
PLOS ONE Thermal wind sensor and wind vector measurement
• The stepper motor continued to run. When the rotation angle (β) reached an angle between
85˚ and 95˚, the average voltages UBL and UBR were detected once every 1˚. Comparing the
detected values, the wind velocity value corresponding to the minimum average voltage
value was assigned to va. The actual value of the wind velocity in the measured field was cal-
culated using Eq (10).
• The predicted value of the wind direction angle (φ) was amended by θ degrees. Therefore,
the angle of the wind direction was ψ = φ + θ in the measured field.
• Finally, the wind vector data were collected, and the sensor bracket returned to the initial
position in the clockwise direction.
Experimental results and discussion
Experimental platform of thermal wind sensor test
The experimental platform, shown in Fig 18, mainly comprises a fan, wind velocity adjustment
device, wind rectification section (honeycomb and damping net), wind test section, standard
anemometer, data acquisition unit, and power supply.
The wind velocity adjustment device can produce airflow with different velocities. Zero-
point adjustment and velocity calibration of the sensor output voltage (Uout) were performed
using a standard anemometer.
Fig 19 shows the relationship between the wind velocity (v) and wind sensor output voltage
(Uout). The curve shown in the figure indicates that the output voltage values are very regular
and their values decrease as the wind velocity increases.
Based on the curve presented above, the following fitting equation was obtained using
MATLAB software:
(
Uout ¼ 0:0073v2 (cid:0) 0:1292v þ 0:9842;
Uout ¼ (cid:0) 0:01v þ 0:5;
0m=s � v < 9m=s
9m=s � v � 15m=s
ð11Þ
Fig 15. Schematic of the wind vector detector.
https://doi.org/10.1371/journal.pone.0231405.g015
PLOS ONE | https://doi.org/10.1371/journal.pone.0231405 April 14, 2020
15 / 22
PLOS ONE Thermal wind sensor and wind vector measurement
Fig 16. Photograph of the wind vector detector. (A) Front view. (B) Top view. (C) Circuit board.
https://doi.org/10.1371/journal.pone.0231405.g016
Therefore, the wind velocity was calculated from Eq (11) using the output voltage of the
thermal wind sensor, and the sensor sensitivity reduced gradually as the wind velocity
increased. The wind velocity measurement range satisfied the requirement of being in the
range 0–15 m/s.
Fig 20 shows the wave curves of the output voltage deviation values of the sensor (ΔV) and
corresponding wind velocity deviation values (Δv) calculated using Eq (11) at different wind
velocities. The results indicate that the output voltage deviation values and the corresponding
wind velocity deviation values show opposite trends. When the wind velocity was below 9 m/s,
the absolute value of the output voltage deviation was less than 27 mV; however, when the
velocity was between 9 and 15 m/s, the absolute value of output voltage deviation was less than
3 mV. The absolute value of the wind velocity deviation was less than 0.3 m/s, i.e., the wind
velocity measurement accuracy was approximately ±0.3 m/s.
PLOS ONE | https://doi.org/10.1371/journal.pone.0231405 April 14, 2020
16 / 22
PLOS ONE Thermal wind sensor and wind vector measurement
Fig 17. Flow chart of the measurement method.
https://doi.org/10.1371/journal.pone.0231405.g017
Fig 21 shows the output voltage curves of all three thermal wind sensors (SBC, SBL, and SBR)
at different wind velocities under the same experimental conditions. The results indicate that
the output voltage values of three thermal wind sensors (SBC, SBL, and SBR) are the same at a
certain wind velocity, and the sensors can meet the accuracy requirements of the wind vector
detector.
Fig 18. Experimental platform of thermal wind sensor test.
https://doi.org/10.1371/journal.pone.0231405.g018
PLOS ONE | https://doi.org/10.1371/journal.pone.0231405 April 14, 2020
17 / 22
PLOS ONE Thermal wind sensor and wind vector measurement
Fig 19. Output voltage curve of the thermal wind sensor at different wind velocities.
https://doi.org/10.1371/journal.pone.0231405.g019
Experimental platform of wind vector detector test
The experimental platform (Fig 22) comprises a wind vector test box, fan, wind velocity adjust-
ment device, data transceiver, and user interface.
The positive direction of the wind vector detector was marked with an arrow (Fig 16(A)).
When the direction of the channel (BC) of the sensor bracket was the same as that of the
Fig 20. Measurement accuracy of the sensor at different wind velocities.
https://doi.org/10.1371/journal.pone.0231405.g020
PLOS ONE | https://doi.org/10.1371/journal.pone.0231405 April 14, 2020
18 / 22
PLOS ONE Thermal wind sensor and wind vector measurement
Fig 21. Output voltages of three independent sensors (SBC, SBL, and SBR).
https://doi.org/10.1371/journal.pone.0231405.g021
arrow, the angle was set as 0˚. If the direction of the measured wind field was consistent with
the positive direction of the detector, the angle of the wind direction was identified as 0˚. In
this experimental platform, the arrow of the detector was pointed to different angles of the
scale circle, which was equivalent to realizing that the test platform generates wind at different
angles. Therefore, the direction of the measured wind field was an angle value relative to the
positive direction of the detector (i.e., an angle value between the measured wind direction
and the arrow).
As the output values of the three thermal wind sensors (SBC, SBL, and SBR) were fitted using
Eq (10), the effect of the sensor bracket structure on the measured wind field can be ignored.
Therefore, during wind velocity measurements, the range and accuracy of the wind vector
detector were consistent with those of the thermal wind sensor.
Fig 22. Experimental platform of wind vector detector test.
https://doi.org/10.1371/journal.pone.0231405.g022
PLOS ONE | https://doi.org/10.1371/journal.pone.0231405 April 14, 2020
19 / 22
PLOS ONE Thermal wind sensor and wind vector measurement
Fig 23. Angle error curves of the wind direction.
https://doi.org/10.1371/journal.pone.0231405.g023
Fig 23 shows the angle error curves of the wind direction at different wind velocities in the
wind vector testing box. These results indicate that the absolute values of the wind direction
deviations were less than 5˚, i.e., the wind direction measurement accuracy was approximately
±5˚. The angles of wind direction can be measured in the range of 0˚–360˚.
Conclusions
Based on the heat transfer principle, King’s law, and the fluid dynamic response regularity of
the measured wind field to the sensor bracket, a thermal wind sensor with constant power con-
trol was developed, and a unique wind vector measurement method was proposed. The con-
ceptual wind vector detector was subsequently implemented. The designed wind vector
monitoring system can realize real-time online detection of the wind vector. The results of a
series of experiments indicated that the accuracy of wind velocity measurement was ±0.3 m/s
in the range of 0–15 m/s, and the accuracy of wind direction measurement was approximately
±5˚ in the angle range of 0˚–360˚. This proved that the wind vector detector yielded good
accuracy, thereby meeting the requirements for measuring the velocity and direction of wind
in enclosed or semi-enclosed large spaces.
Although important findings were obtained, there are a few limitations to this study. The
wind velocity range only met the measurement requirements for enclosed or semi-enclosed
large spaces, and the thermal wind sensor could not measure wind velocities exceeding 15 m/s.
Furthermore, our results suggest that the measurement range of wind velocity can be
expanded by increasing the heating power value of the nickel-chromium wire coil, and the
structural parameters of the sensor bracket can be set via simulations. Currently, there is no
function to automatically adjust the heating power value and length of the channel (BC)
according to the velocity range of the measured wind field. This issue should be addressed in a
future study. The findings of this study can serve as a theoretical foundation for developing
further applications employing instruments for the wind vector measurement.
PLOS ONE | https://doi.org/10.1371/journal.pone.0231405 April 14, 2020
20 / 22
PLOS ONE Thermal wind sensor and wind vector measurement
Supporting information
S1 Data. Empirical data.
(ZIP)
Author Contributions
Conceptualization: Congning Liu, Maria D. King.
Data curation: Congning Liu, Tengxi Wang.
Formal analysis: Congning Liu, Xiaoyu Yu, Tengxi Wang.
Funding acquisition: Yunbo Shi, Xiaoyu Yu.
Methodology: Congning Liu, Yunbo Shi, Maria D. King.
Project administration: Xiaoyu Yu.
Resources: Yunbo Shi, Maria D. King.
Supervision: Yunbo Shi.
Writing – original draft: Congning Liu.
Writing – review & editing: Congning Liu, Xiaoyu Yu, Maria D. King.
References
1.
Likar J, Čadezˇ J. Ventilation Design of Enclosed Underground Structures. Tunnelling and Underground
Space Technology. 2000; 15(4): 477–480.
2. Parra MT, Villafruela JM, Castro F, Me´ ndez C. Numerical and experimental analysis of different ventila-
tion systems in deep mines. Building and Environment. 2006; 41(2): 87–93.
3. Su S, Chen H, Teakle P, Xue S. Characteristics of coal mine ventilation air flows. Journal of Environ-
mental Management. 2008; 86(1): 44–62. https://doi.org/10.1016/j.jenvman.2006.11.025 PMID:
17239518
4. Sun K. Design of Hot Film Anemometer Based on Constant Temperature Difference Method. M.En.
Thesis, Harbin University of Science and Technology. 2013.
5.
Zhou J. Research on 3D Spatial Data Models and Refined 3D Automatic Modeling Methods for Under-
ground Pipeline. D.En. Thesis, Wuhan University. 2016.
6. Stefopoulos EK, Damigos DG. Design of emergency ventilation system for an underground storage
facility. Tunnelling and Underground Space Technology. 2007; 22(3): 293–302.
7. Semin MA, Levin LY. Stability of air flows in mine ventilation networks. Process Safety and Environmen-
tal Protection. 2019; 124: 167–171.
8. Martinez LM, Agdas D, Webster GD, Masters FJ. Wind Speed Perception and Risk. PLoS ONE. 2012;
7(11): e49944. https://doi.org/10.1371/journal.pone.0049944 PMID: 23226230
9. Counselman CC, Gourevitch SA, King RW, Pettengill GH, Prinn RG, Shapiro II, et al. Wind Velocities
on Venus: Vector Determination by Radio Interferometry. Science. 1979; 203(4382): 805–806. https://
doi.org/10.1126/science.203.4382.805 PMID: 17833005
10. Schuhen N, Thorarinsdottir TL, Gneiting T. Ensemble Model Output Statistics for Wind Vectors. Monthly
Weather Review. 2012; 140(10): 3204–3219.
11. Dias JM, de Winter W, Donker J, Sterk G, van Beem J, Ruessink G. Regional versus local wind speed
and direction at a narrow beach with a high and steep foredune. Plos One. 2020; 15(1): e0226983.
https://doi.org/10.1371/journal.pone.0226983 PMID: 31895928
12. Suomi I, Vihma T. Wind Gust Measurement Techniques—From Traditional Anemometry to New Possi-
bilities. Sensors. 2018; 18(4): 1300.
13. Magar V, Bruecker CH, Mikulich V. Sensing of minute airflow motions near walls using pappus-type
nature-inspired sensors. Plos One. 2017; 12(6): e0179253. https://doi.org/10.1371/journal.pone.
0179253 PMID: 28658272
PLOS ONE | https://doi.org/10.1371/journal.pone.0231405 April 14, 2020
21 / 22
PLOS ONE Thermal wind sensor and wind vector measurement
14. Wu Y, Qi L, Zhang H, Musiu EM, Yang Z, Wang P. Design of UAV Downwash Airflow Field Detection
System Based on Strain Effect Principle. Sensors. 2019; 19(11): 2630.
15. Pindado S, Sanz A, Wery A. Deviation of Cup and Propeller Anemometer Calibration Results with Air
Density. Energies. 2012; 5(3): 683–701.
16. Pindado S, Cubas J, Sorribes-Palmer F. On the harmonic analysis of cup anemometer rotation speed:
A principle to monitor performance and maintenance status of rotating meteorological sensors. Mea-
surement. 2015; 73: 401–418.
17. Klopfenstein R Jr. Air velocity and flow measurement using a Pitot tube. ISA Transactions. 1998; 37(4):
257–263.
18. Cho A, Kim J, Lee S, Kee C. Wind Estimation and Airspeed Calibration using a UAV with a Single-
Antenna GPS Receiver and Pitot Tube. IEEE Transactions on Aerospace and Electronic Systems.
2011; 47(1): 109–117.
19. Dong H, Jun Y. High Accuracy Time of Flight Measurement for Ultrasonic Anemometer Applications. In:
Proceedings of 3rd International Conference on Instrumentation and Measurement, Computer, Com-
munication and Control; September 21–23, 2013; Shenyang, Liaoning, China: IEEE; 2013. p. 61–64.
20. Han D, Kim S, Park S. Two-dimensional ultrasonic anemometer using the directivity angle of an ultra-
sonic sensor. Microelectronics Journal. 2008; 39(10): 1195–1199.
21. Wang D, Xiong W, Zhou Z, Zhu R, Yang X, Li W, et al. Highly Sensitive Hot-Wire Anemometry Based
on Macro-Sized Double-Walled Carbon Nanotube Strands. Sensors. 2017; 17(8): 1756.
22. Ye Y, Yi Z, Gao S, Qin M, Huang Q-A. Octagon-Shaped 2-D Micromachined Thermal Wind Sensor for
High-Accurate Applications. Journal of Microelectromechanical Systems. 2018; 27(4): 739–747.
23. Ke W, Liu M, Li T, Wang Y. MEMS thermal gas flow sensor with self-test function. Journal of Microme-
chanics and Microengineering. 2019; 29(12):125009.
24.
Leu TS, Yu JM, Miau JJ, Chen SJ. MEMS flexible thermal flow sensors for measurement of unsteady
flow above a pitching wind turbine blade. Experimental Thermal and Fluid Science. 2016; 77: 167–178.
25. Yao W. Measurement System of Low Wind Speed and Direction. M.En. Thesis, Hefei University of
Technology. 2012.
26.
Zhu Y, Chen B, Qin M, Huang Q-A. 2-D Micromachined Thermal Wind Sensors—A Review. IEEE Inter-
net of Things Journal. 2014; 1(3): 216–232.
27. Nguyen N-T. A novel wind sensor concept based on thermal image measurement using a temperature
sensor array. Sensors and Actuators A: Physical. 2004; 110(1–3): 323–327.
28. King LV. On the Convection of Heat from Small Cylinders in a Stream of Fluid: Determination of the Con-
vection Constants of Small Platinum Wires, with Applications to Hot-Wire Anemometry. Proceedings of
the Royal Society A: Mathematical, Physical and Engineering Sciences. 1914; 90(622): 563–570.
29.
FOTRIC AnalyzIR software. Version 4.1.7 [software]. 2019 Feb 12 [cited 2019 Sep 26]. Available from:
http://www.fotric.com/.
30. Hang CC, Sin KK. A comparative performance study of PID auto-tuners. IEEE Control Systems. 1991;
11(5): 41–47.
31. Yang B, Zhang X, Yu T, Shu H, Fang Z. Grouped grey wolf optimizer for maximum power point tracking
of doubly-fed induction generator based wind turbine. Energy Conversion and Management. 2017;
133: 427–443.
32. Yang B, Yu T, Shu H, Zhang X, Qu K, Jiang L. Democratic joint operations algorithm for optimal power
extraction of PMSG based wind energy conversion system. Energy Conversion and Management.
2018; 159: 312–326.
33. Sung SW, Lee I-B. An improved algorithm for automatic tuning of PID controllers. Chemical Engineering
Science. 2000; 55(10): 1883–1891.
34. Monteiro J, Carrilho J, Silva MGd, Miranda A, Silva A. 3D Printed Pressure Anemometers. 3D Printing
and Additive Manufacturing. 2017; 4(3): 172–181.
35. Magar V, Poortinga A, van Minnen J, Keijsers J, Riksen M, Goossens D, et al. Measuring Fast-Tempo-
ral Sediment Fluxes with an Analogue Acoustic Sensor: A Wind Tunnel Study. PLoS ONE. 2013; 8(9):
e74007. https://doi.org/10.1371/journal.pone.0074007 PMID: 24058512
36. Awadalla M, Lu T-F, Tian ZF, Dally B, Liu Z. 3D framework combining CFD and MATLAB techniques for
plume source localization research. Building and Environment. 2013; 70: 10–19.
37. Bartzanas T, Kacira M, Zhu H, Karmakar S, Tamimi E, Katsoulas N, et al. Computational fluid dynamics
applications to improve crop production systems. Computers and Electronics in Agriculture. 2013; 93:
151–167.
38. Kuwata Y, Suga K, Sakurai Y. Development and application of a multi-scale k-ε model for turbulent
porous medium flows. International Journal of Heat and Fluid Flow. 2014; 49: 135–150.
PLOS ONE | https://doi.org/10.1371/journal.pone.0231405 April 14, 2020
22 / 22
PLOS ONE |
10.1371_journal.pone.0239340 | RESEARCH ARTICLE
Electrocatalytic CO2 fixation by regenerating
reduced cofactor NADH during Calvin Cycle
using glassy carbon electrode
Irshad Ali1, Saeid Amiri2, Nehar Ullah1, Mohammad YounasID
1, Mashallah RezakazemiID
3*
1 Department of Chemical Engineering, University of Engineering & Technology, Peshawar, Pakistan,
2 Chemical & Petroleum Engineering Department, Sharif University of Technology, Tehran, Iran, 3 Faculty of
Chemical and Materials Engineering, Shahrood University of Technology, Shahrood, Iran
* m.younas@uetpeshawar.edu.pk, mashalah.rezakazemi@gmail.com
Abstract
In this study, an enzymatic pathway has been developed to replicate the Calvin Cycle by
creating the individual steps of the carbon cycle in a bioreactor. The technology known as
“artificial photosynthesis” converts CO2 emissions into a variety of intermediates that serve
as precursors to high-value products. CO2, light, water, and electricity were used as feed-
stock. An electrochemical reactor was also studied for the regeneration of active NADH
operating at constant electrode potential. Initially, a batch electrochemical reactor containing
80 mL of 0.2 mM NAD+ in Tris-buffer (pH 7.40) was used to evaluate the electrode material
operating at normal temperature and pressure. The results showed that the cathode is
highly electrocatalytically efficient and selective to regenerate 97.45±0.8% of NADH from
NAD+ at electrode potential of -2.3 V vs. mercury standard electrode (MSE). The NADH
regeneration system was then integrated with ATP regeneration system and bioreactor con-
taining Ribulose bisphosphate carboxylase/oxygenase (RuBisCO). NADH was regenerated
successfully during the process electrochemically and then was used by the enzymatic
reaction to produce triose phosphate and 3-Phosphoglycerate (3GPA).
Introduction
Economic growth along with high density infrastructural and transportation development is
mainly responsible for rapid energy demand that consequently resulted in an increased emis-
sion of carbon dioxide (CO2). CO2 is the primary anthropogenic greenhouse gas (GHG)
mainly responsible for global warming due to the burning of fossil fuels (coal, hydrocarbons,
peat) for energy [1–3]. Other significant sources of atmospheric CO2 include byproducts from
the fermentation of sugars, the respiration of all living organisms, volcanoes, hot springs, and
geysers [4–7]. It is desired to develop green and sustainable technology to convert CO2 into
valuable materials for carbon capture and sequestration (CCS) technology and reduce the bur-
den on greenhouse gas emissions. Currently, the developed routes for CO2 reduction are the
reduction at the source (i.e. reduced use of fossil fuels), sequestration and the chemical conver-
sion [7]. Instead CO2 upcycling offers a pathway towards more sustainable processes [5, 8–10].
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: Ali I, Amiri S, Ullah N, Younas M,
Rezakazemi M (2020) Electrocatalytic CO2 fixation
by regenerating reduced cofactor NADH during
Calvin Cycle using glassy carbon electrode. PLoS
ONE 15(9): e0239340. https://doi.org/10.1371/
journal.pone.0239340
Editor: Jonghyun Park, Missouri University of
Science and Technology, UNITED STATES
Received: May 22, 2020
Accepted: September 3, 2020
Published: September 17, 2020
Copyright: © 2020 Ali et al. This is an open access
article distributed under the terms of the Creative
Commons Attribution License, which permits
unrestricted use, distribution, and reproduction in
any medium, provided the original author and
source are credited.
Data Availability Statement: All relevant data are
within the manuscript.
Funding: The authors received no specific funding
for this work.
Competing interests: The authors have declared
that no competing interests exist.
PLOS ONE | https://doi.org/10.1371/journal.pone.0239340 September 17, 2020
1 / 20
PLOS ONE Glassy carbon electrode for CO2 fixation
The conversion of CO2 into fuels or value-added chemicals is currently a field of great research
interest. Non-enzymatic electrochemical and photocatalytic methods in recent years have
been investigated for nicotinamide adenine dinucleotide (NADH) regeneration [11]. How-
ever, those non-enzymatic methods require a rare metal-chelating electron mediator or toxic
methylviologen as prerequisites for the NADH regeneration routes [11]. Thus, in view of
increasing environmental and sustainability considerations, seeking a green, sustainable, eco-
nomic, and efficient regeneration method for NADH is indispensable.
An alternate route to reconstruct cellular processes by imitating natural systems through
photosynthesis is the Calvin cycle [12]. Herein, plants use CO2 and sunlight in a process com-
monly known as “photosynthesis” to produce organic matter. We aim to mimic the same prin-
ciples in a non-living system, which we term “artificial photosynthesis” shown Schematically
in Fig 1A along with Calvin Cycle Fig 1B, the driving engine of photosynthesis and our
process.
It involves several enzymes along with light energy-dependent reactions ATP and NADH
to form value-added products. First, CO2 is fixed from inorganic form to organic 3-PGA
(3-Phosphoglycerate) which in turn forms an intermediate product 1,3-Bisphosphoglycerate in
presence of PGK (Phosphoglycerate Kinase) and ATP. This intermediate along with GAPDH
(Glyceraldehyde-3-Phosphate Dehydrogenase) and NADH forms G3P (Glyceraldehyde 3-Phos-
phate). Finally, G3P and ATP are used to regenerate RuBP (Ribulose-1,5-bisphosphate) and to
fix more CO2 in the system [13–17]. In enzyme-catalyzed reactions, nicotinamide adenine dinu-
cleotide (NAD+/NADH) redox couple, is an important component to synthesize precious
medicinal and other value-added products [18–22]. The main function of cofactor NAD(H) is
to supply electrons and hydrogen (proton) in these enzymatic reactions [11]. The 1,4-NADH is
the only enzymatically active isomers among different reduced NADH. 1,4-NADH is rarely
available isomer due to its high cost (ca. $992,000/kg). Therefore, it is very important to regener-
ate NADH in-situ and cost-effectively in a biochemical process for reuse. This will substantially
help in the final product cost reduction and would, therefore, justify the use of such an expen-
sive coenzyme 1,4-NADH.
Many studies have been reported in the literature on the synthetic Calvin Cycle [12, 23–27]
to produce high-value chemicals also commonly called the Reductive pentose phosphate path-
way [28]. It is an enzymatic cycle that catalyzes the photosynthetic assimilation of CO2 and
produces pentoses. Ribulose bisphosphate carboxylase/oxygenase (RubisCO), the only enzyme
capable of CO2 assimilation [28], is the first enzyme in the cycle and catalyzes the fixation of
Fig 1. (a) A schematic of artificial photosynthesis mirroring plant cell photosynthesis: (1) Biochemical reactor cascade, (2) ATP
regeneration system, (3) NADH regeneration system and (b) Calvin Cycle.
https://doi.org/10.1371/journal.pone.0239340.g001
PLOS ONE | https://doi.org/10.1371/journal.pone.0239340 September 17, 2020
2 / 20
PLOS ONE Glassy carbon electrode for CO2 fixation
atmospheric CO2 to Ribulose 1,5-bisphosphate (RuBP). By CO2 fixation to RuBP two mole-
cules of the triose phosphate, 3-Phosphoglycerate (3PGA), are produced. Through a series of
nine enzymes, five of every six triose phosphate molecules produced by the Calvin Cycle are
used for the regeneration of three molecules of RuBP. One triose phosphate molecule is the net
gain from the fixation of three molecules of CO2, which a cell can use in a variety of biosyn-
thetic processes.
By employing RubisCO’s ability to sequester atmospheric CO2 and the Calvin Cycle’s abil-
ity to regenerate the CO2 acceptor, the bioprocess developed in the current work can convert
CO2 emissions into valuable small molecules without-bio or petro-based feedstocks. For exam-
ple, through the application of various, well-known, enzymatic pathways the net gain of triose
phosphate can be easily converted into hexose sugars such as fructose or glucose [28], glycerol,
and numerous high-value chemicals including 2,3-butanediol, propionate, ethanol and buta-
nol [26]. All of the necessary metabolic pathways share Dihydroxyacetone phosphate (DHAP)
upstream of the final product formation. Nonetheless DHAP is the intermediate directly fol-
lowing 3PGA in the Calvin Cycle. By capitalizing on these metabolic pathways, we can easily
tailor biochemical transformation pathways to convert 3PGA into higher-value chemicals.
The reducing power required by Calvin Cycle is supplied in the form of the enzymatic co-
factor NADH. However, the regeneration system for a stable NADH has not yet been reported
which is one of the key elements in artificial photosynthesis process system in order to convert
CO2 into valuable products which are supported by co-factors ATP and NADH.
Several methods have been employed to reduce NAD+ to NADH [11, 29]. However, elec-
trochemistry-based methods are very promising and have gained much attention for the
regeneration of NADH. As widespread use of these methods is due to the intrinsic nature of
electrochemistry when the reactants are electrons, the progress of the reaction can easily be
monitored and controlled which can lead to relatively easy scale-up of the process [11, 30–33].
NADH regeneration via electrochemical reduction of NAD+ can be represented by the fol-
lowing established pathway [34]:
Step 1 : NADþ þ e(cid:0) (cid:0) !NAD �
Step 2a NAD � þe(cid:0) þ Hþ(cid:0) !NADH
ð1Þ
ð2Þ
Step 2a (Eq 2) is the slow step where as a result enzymatically-inactive dimer NAD2 is pro-
duced due to the fast NAD-radical dimerization, step 2b [31, 35–40]:
Step 2b : NAD � þNAD � (cid:0) !NAD2
ð3Þ
The high concentration of H+ will boost the reaction kinetics of Step 2a (Eq 2), as H+ will
readily react with adsorbed NAD-radical on the electrode surface followed by Eley–Rideal
mechanism, or Langmuir–Hinshelwood mechanism. When it follows the former mechanism,
it will first adsorb on the electrode’s surface as M-Hads and then react with the immediately
available NAD-radical [34, 35, 41, 42]. In case, reaction follows the latter mechanism e.g. Lang-
muir–Hinshelwood, it results in the faster kinetics of Step 2a (Eq 2). In this case, the adsorbed
hydrogen (Hads) on the electrode surface significantly affects recovery of the enzymatically
active 1,4- NADH compared to inactive-dimer NAD2. However, it has been found that the
concentration of adsorbed hydrogen, Hads is strongly dependent on the electrode potential
[34, 35, 41]. Therefore, hydrogen coverage or Hads on the surface is an important parameter in
the reduction of NAD+ to active NADH. Namely, on a bare glassy carbon (GC) electrode’s sur-
face used in the developed system, recovery of enzymatically-active 1,4-NADH reached 98% at
electrode potential of –2.3 VMSE in the previous work, it was stated that highest NADH
PLOS ONE | https://doi.org/10.1371/journal.pone.0239340 September 17, 2020
3 / 20
PLOS ONE Glassy carbon electrode for CO2 fixation
recovery can be obtained at this electrode potential [34, 35, 41]. The highest recovery was due
to higher concentration of adsorbed ‘active’ hydrogen, Hads, at the more cathodic potential on
the glassy carbon electrode surface.
In the current work, NAD+ was electrochemically reduced to enzymatically active
1,4-NADH which was subsequently oxidized to NAD+ in the simulated environment of Calvin
cycle. The best experimental conditions were discussed and selected for electrochemical reduc-
tion of NAD+ to enzymatically active NADH to produce different sugars. The optimized
experimental conditions for the electrochemical regeneration of NADH were presented and
discussed in contest of literature. In the second part of the study, the integration of NADH
regenerative system with ATP in a biochemical reactor containing all the essential ingredients
for Calvin cycle were presented. To the best of authors’ knowledge, no published literature is
available on the regeneration of NADH through electrochemical method during Calvin Cycle
(artificial photosynthesis) and is, therefore, the focus of the current work.
The synthetic Calvin Cycle engineered by Ingenuity Lab (Fig 1A and 1B) has the potential
to produce greater than fifty high-value chemicals. However, it is very important to develop a
simple and stable NADH regeneration system which is one of the key elements in artificial
photosynthesis process system since the energy required for the conversion of C O2 into
organic molecules is supported by co-factor molecules ATP and NADH.
The technology is a multi-enzyme platform that generates valuable, small organic molecules
from CO2 produced by industrial processes, sunlight, water, and electricity. It essentially pro-
vides photosynthesis without the energy requirements for reproduction and growth found in
traditional biological carbon fixation platforms. The process is designed as a cascade of biore-
actors, which allows for optimized reaction conditions at each stage of the process.
Materials and methods
Chemicals and solutions
Enzymatically active NADH regeneration was performed in 50 mM Tris-buffer solution con-
taining 0.2 mM of NAD+ (purity 95%, Sigma N0632) in a conventional three-electrode electro-
chemical batch reactor. The initial volume of the electrolyte in the reactor was 80 mL at a
constant pH of 7.4 and 295 K. Deionized water (resistivity 18.2 MW cm, Milli-Q1 systems)
was used in the preparation of Tris-buffer solution which is used to adjust of pH of Hydrochlo-
ric acid. No further purification of the chemicals was carried out and were used as received.
Electrochemical cell and electrodes
A conventional three-electrode based electrochemical batch reactor/cell was used as shown in
Fig 2.
Glassy carbon (GC) electrode with two different geometric areas (12.5 and 50 cm2) was
used as a working electrode (cathode) in the reactor. Graphite rod from McMaster-Carr
9121K71 was utilized as an anode (counter electrode). Initially, the anode was sonicated for 30
min in ethanol followed by rinsing with deionized water. Since oxygen evolves on the anode
during the NADH regeneration, therefore it was separated from the bulk electrolyte using
glass tubes with Nafion membrane to prevent the diffusion of dissolved oxygen to the working
electrode. As a reference electrode, a Mercury/mercurous sulphate electrode (MSE; +0.642 V
vs. SHE) supplied by Fisher Scientific was used and all the reported potentials in this paper are
with reference to MSE. The experiments were carried out under the optimized conditions, for
example, at electrode potential of -2.30 V vs. MSE, room temperature, and biological pH.
Detailed experimental procedure and essays have been described in S1 File.
PLOS ONE | https://doi.org/10.1371/journal.pone.0239340 September 17, 2020
4 / 20
PLOS ONE Glassy carbon electrode for CO2 fixation
Fig 2. Three-electrode batch electrochemical reactor (cell) used in the research.
https://doi.org/10.1371/journal.pone.0239340.g002
Equipment
Linear polarization, cyclic voltammetry, and controlled-potential electrolysis techniques were
carried out by utilizing Ecochemie Autolab potentiostat/galvanostat PGSTAT30/ controlled
using the NOVA v.12.1.1 software. An in-line OceanView UV-Vis spectrophotometer was uti-
lized to track the reduction kinetics of NAD+ (NADH regeneration). The activity assay was
monitored by a Perkin Elymer Lambda 1050 UV-Vis-Nir Spectrophotometer at a wavelength of
340 nm. The detailed procedure of the assay can be found in the literature [31, 35, 36, 41, 43].
Electrode pretreatment
Before the NADH regeneration experiment, a pretreatment wet-polishing of the GC surface
was done with 1200/4000 grid paper to achieve a mirror-finished surface. It was degreased
with ethanol followed by sonication in ethanol for 30 min to get rid of any residues from the
electrode surface during electrode pretreatment. Since impurities on the electrode surface
greatly affect its performance, GC electrode was electrochemically cleaned in 0.5 M H2SO4
(Fisher Scientific 351293) by cyclic voltammetry between –1.8 and 1.8 V at a scan rate of 100
mV s–1, for 50 cycles using the same electrochemical reactor. A very stable and reproducible
cyclic voltammetry (CV) profile confirmed that the surface of the GC electrode is very clean as
shown later in Fig 3.
PLOS ONE | https://doi.org/10.1371/journal.pone.0239340 September 17, 2020
5 / 20
PLOS ONE Glassy carbon electrode for CO2 fixation
Fig 3. Cyclic voltammogram of GC electrode in 0.5 M H2SO4 solution. Scan rate, sr: 100 mV s–1. Temperature, T = 295 K.
https://doi.org/10.1371/journal.pone.0239340.g003
Electrochemical regeneration of active NADH
Electrochemical regeneration of active NADH was performed in the electrochemical reactor
(Fig 1). As counter electrodes, two graphite rods were used, and GC electrode was used as a
working electrode in the developed batch electrochemical reactor. For a uniform electric field
in the reactor, counter electrodes were arranged and placed opposite to both surfaces of work-
ing electrode. A total electrolytic volume of 80 mL was used with an initial 0.2 mM concentra-
tion of NAD+.
Experimental methodology
All measurements were performed in an oxygen-free electrolyte. To achieve this, pure argon
(99.9%) was purged through the electrolyte for 30 min prior to and during electrochemical
NADH regeneration. This also ensured convective mass transport of electroactive species to/
from the electrode surface since NAD+ reduction reaction is mass-transport controlled
reaction.
To investigate the enzymatic activity of the regenerated NADH, activity tests were made
according to the regular Sigma Quality Control Test Procedure (EC 1.8.1.4) which was further
modified for this purpose using lipoamide dehydrogenase (5.3 U/mg, Calzyme laboratories,
Inc. 153A0025) as an enzyme and DL-6,8-thioctic acid amide (MedChem Express, HY-B1142)
as a substrate [36, 44].
The Calvin Cycle started with sequestering CO2 in the RuBisCO foam reactor. While CO2
sparged through the liquid, it formed the foam architecture due to the presence of ranaspu-
min-2 (RSN-2) as a surfactant. The enzyme solution was recycled from the bulk of the liquid
PLOS ONE | https://doi.org/10.1371/journal.pone.0239340 September 17, 2020
6 / 20
PLOS ONE Glassy carbon electrode for CO2 fixation
in the bottom of the reactor to the top, flowing through the foam structure. This allows for
CO2 to come into contact and react with RuBP in the presence of RuBisCO to produce 3PGA,
an intermediate in the Calvin Cycle and the feed for the next reactor. 3PGA was separated
through an ultrafiltration (UF) system and proceeded to the next reactor. Except for the
RuBisCO reactor, in which the enzyme was in the bulk of the solution, i.e. dissolved free form,
all other reactors used enzymes that are immobilized onto beads and packed into the reactor
space. The reactants passed through a packed bed of immobilized enzyme where the reactions
took place. The reactors were designed such that the reactants were completely converted to
the product as they passed along the packed bed. Immobilization eliminates the need to sepa-
rate the enzyme from the solution before going to the next reactor. The reactors that involve
cofactors (ATP or NADH) have nanofiltration (NF) separation system to separate the cofac-
tors to be regenerated and recycled. When possible, other reaction steps were combined into
multi-bed reactors, where immobilized enzymes were sequentially packed, and reactants while
passing along the beds were converted to the reactants of the subsequent bed. For details, S1
File may be read. Experimental work was repeated at least three times and the mean values was
considered. The standard error among the triplicate readings was withing ±5%.
Results and discussion
Activation of glassy carbon electrode
To minimize the interference of hydrogen evolution reaction (HER) and its effect on NAD+
reduction reaction, Glassy Carbon was used as a working electrode (cathode) due to its high
hydrogen reduction overpotential, in addition to low porosity and comparatively better electri-
cal conductivity [45]. Furthermore, it is cheap, easily available, and stable under extreme
experimental conditions, which make it suitable amongst other materials for industrial appli-
cations [45]. During electrochemical measurements, surface of the electrode material plays an
important role. Cleanliness from impurities and surface chemistry of carbon-oxygen function-
alities of the electrodes greatly affect these measurements [46]. Therefore, GC electrode was
electrochemically cleaned in 0.5 M H2SO4 (Fisher Scientific 351293) by cyclic voltammetry
between –1.8 and +1.8 V at a scan rate of 100 mV s–1, for 50 cycles using the same electro-
chemical reactor. The results are plotted in Fig 3.
Fig 3 demonstrates a very stable and reproducible cyclic voltammogram (CV) which clearly
indicates a clean GC electrode. Namely, it is characterized by a large double layer (DL) region
of glassy carbon electrode negative of +1.0 V and positive of –1.0 V [47, 48]. There is no evi-
dence of any redox reaction occurring on the GC surface since the current (reaction rate) is
zero. The increase in the current positive of +1.0 V is indication of oxygen evolution reaction
(OER) while the increase in the current negative of –1.0 V due to the HER. Thus, the electro-
chemical activation of GC involves both oxidation and reduction of the GC electrode surface.
Linear polarization voltammetry (LV)
Linear voltammetry (LV) was conducted in order to determine the working potential region
of NAD+ reduction to active NADH as shown in Fig 4.
Initially, a linear voltammogram was recorded only in the buffer solution without NAD+
molecule (dashed line) which shows a control curve of the glassy carbon electrode that is atypi-
cal behavior of the pure GC electrodes under the applied experimental conditions [35]. Such
behavior of GC electrode is attributed to a wide double layer region (positive of ca. –1.40 V)
and the beginning of HER (negative of ca. –1.50 V). On the other hand, when the same experi-
ment was carried out in the electrolyte containing 4 mM NAD+ (solid line), one can see a
broad and sharp cathodic current peak at a potential negative of ca. –1.30 V as shown in Fig 4.
PLOS ONE | https://doi.org/10.1371/journal.pone.0239340 September 17, 2020
7 / 20
PLOS ONE Glassy carbon electrode for CO2 fixation
Fig 4. Linear voltammogram of GC electrode in 0.1 M phosphate buffer solution (dashed line) and phosphate buffer
solution containing 4 mM NAD+ solution (solid line) at a scan rate (sr) of 10 mV s–1. Temperature, T = 295 K.
https://doi.org/10.1371/journal.pone.0239340.g004
Two major potential regions can be identified in Fig 4 (solid line), i.e. the region negative of
ca. –1.30 V is related to the NAD+ reduction region with a well-pronounced NAD+ reduction
peak negative of ca. –1.40 V. This cathodic current peak is of NAD+ reduction reaction on dif-
ferent electrode materials as evident from literature [31, 37, 49–53]. Thus, LV technique
resulted in some useful information on NAD+ reduction reaction. However, in the actual bio-
chemical reactor, 1,4-NADH regeneration would not be performed under potentiodynamic
conditions, but rather potentiostatic. For this purpose, controlled-potentiostatic measure-
ments were carried out in order to reduce NAD+ to active NADH.
Electrochemical regeneration of active NADH
Electrochemical NADH regeneration was studied in various buffer solutions at different con-
centrations and temperatures in order to optimize the experimental conditions for faster reac-
tion kinetics and high yield of NADH.
Phosphate buffer solution
Initially, the NAD+ reduction kinetics and recovery of regenerated active 1,4-NADH were
investigated in Phosphate Buffer solution at room temperature (295 K ±2) and pH 7.4 at
selected electrode potential of –2.30 V as shown in Fig 5.
Fig 5 shows the time dependence of absorbance at 340 nm at electrode potential of –2.30 V
using Glassy carbon electrode. In the beginning, the absorbance at 340 nm is zero. However,
when electrode potential of –2.30 V was applied, the absorbance started to increase with time.
After 70 min of electrolysis the absorbance reached a plateau indicating the completion of the
reaction. As explained earlier in the manuscript that at 340 nm both NAD2 and 1,4-NADH
absorb, hence Fig 5 does not give the precise information about active NADH regenerated
PLOS ONE | https://doi.org/10.1371/journal.pone.0239340 September 17, 2020
8 / 20
PLOS ONE Glassy carbon electrode for CO2 fixation
Fig 5. Time evolution of absorbance at 340 nm during electrolysis of 0.2 mM NAD+ in phosphate buffer solution at electrode
potential of –2.30 V using GC electrode (geometric area, A = 50 cm2). Temperature, T = 295±2 K.
https://doi.org/10.1371/journal.pone.0239340.g005
rather it only shows the conversion of NAD+ to both 1,4-NADH and NAD2 the latter being
enzymatically-inactive [31, 35, 36, 41, 43]. Therefore, to distinguish between the two species,
an enzymatic assay was conducted. Briefly, the assay uses DL-lipoamide as a substrate to con-
vert it into dihydrolipoamide in the presence of lipoamide dehydrogenase as enzyme, repre-
sented by Eq (4) [44]
1; 4 (cid:0) NADH þ DL(cid:0)
lipoamide lipoamide dehydrogenase NADþ þ dihydrolipoamide
>
ð4Þ
Oxidation of NADH to NAD+ reduces absorbance at 340 nm, which clearly can be seen
and observed. In order to calculate the percentage of active NADH produced, the absorbance
at 340 nm before and after reaction (4) was utilized. It is very important to mention here again
that both active NADH (and its different isomers) and NAD2 absorb at 340 nm and thus can-
not be used to determine the actual amount of active NADH. For this purpose, an activity
assay was conducted for the samples collected from the reactor. The results of the activity assay
show that 97.45±0.8% of active 1, 4-NADH was recovered under the experimental conditions
employed which is in agreement with our previous studies [34, 35, 41]. Now, taking that the
NAD2 formation requires one electron per one NAD+ molecules, while the formation of
1,4-NADH requires the exchange of two electrons (Eq 2), the 97.45% active NADH indicates
that 1.97 electron went to reduce NAD+ to active NADH. The remaining could be both inac-
tive NADH isomers or/and dimer NAD2. Thus, almost 98% of electrons went to reduce NAD+
to active NADH.
PLOS ONE | https://doi.org/10.1371/journal.pone.0239340 September 17, 2020
9 / 20
PLOS ONE Glassy carbon electrode for CO2 fixation
Fig 6. Time evolution of absorbance at 340 nm during electrolysis of 0.2 mM NAD+ in phosphate buffer solution
during reduction and oxidation cycles at electrode potentials of –2.30 V and +0.40 V, respectively using GC
electrode. Temperature, T = 295±2 K.
https://doi.org/10.1371/journal.pone.0239340.g006
To investigate if regenerated 1,4-NADH is stable for a prolonged time and can be oxidized
back to NAD+ electrochemically, cyclic reduction and oxidation were carried out at two differ-
ent temperatures i.e. at room temperature, T = 295±2 K and low temperature, T = 277±2 K as
shown in Figs 6 and 7, respectively.
The results in Fig 6 demonstrate that NAD+ indeed was reduced to NADH as the absor-
bance at 340 nm increased with time from 0 to 0.55 at –2.30 V using GC electrode and it is
reached to its maximum value in approximately 3 hrs that shows the completion of NAD+
reduction reaction [41]. The GC electrode of small geometric area (12.5 cm2) was used as com-
pared to 50 cm2 (Fig 5) which took only 70 min. Soon after the completion of reduction step,
the oxidation of the regenerated NADH was initiated back to NAD+ at electrode potential of
+0.40 V using the same GC electrode. As it can be seen in Fig 6 that the absorbance at 340 nm
started to decrease showing the oxidation of NADH to NAD+ since NAD+ absorbs at 260 nm.
Fig 7. Time evolution of absorbance at 340 nm during electrolysis of 0.2 mM NAD+ in phosphate buffer solution
during reduction and oxidation cycles at electrode potentials of –2.30 V and +0.40 V, respectively using GC
electrode. Temperature, T = 277±2 K.
https://doi.org/10.1371/journal.pone.0239340.g007
PLOS ONE | https://doi.org/10.1371/journal.pone.0239340 September 17, 2020
10 / 20
PLOS ONE Glassy carbon electrode for CO2 fixation
The cycle was repeated in order to investigate how many cycles the system can regenerate
NADH? It was noticed that with each reduction cycle the absorbance at 340 nm decreased and
after 30 hr, the developed system was not able to reduce or oxidize the NAD(H) molecule.
Thus, it was concluded that at room temperature the NAD+/NADH is stable for a maximum
period of 24 to 30 h (4 cycles).
Further, to investigate the effect of temperature on the oxidation/reduction cycle, the exper-
iments were repeated under the same conditions but at a low temperature of 277 K as shown
in Fig 7.
It was very interesting to see that at low temperatures the system was able to continuously
reduce and oxidize the NAD+/NADH up to 96 hrs (Fig 7) as compared to 30 hrs at room tem-
perature (Fig 6). Thus, it could be concluded that at low temperature the NAD(H) is stable for
96 hrs under the applied experimental conditions. These results show that the regenerated
NADH was active for 96 hrs. Thus, the formed radicals do not affect the chemical structure of
NADH and enzyme under the applied experimental conditions.
Tris-buffer solution
Further to study the effect of buffer solution on the kinetics of NAD+ reduction reaction and
recovery of active 1,4-NADH, experiments were conducted under the same operating condi-
tions (Fig 5) but in Tris-buffer with two different concentrations 100 (dotted line) and 260
mM (solid line) as shown in Fig 8.
Fig 8 shows that the kinetics of NAD+ reduction reaction is a bit slower than in phosphate
buffer (Fig 5) and it takes longer to reach a plateau. However, the conversion is much higher
Fig 8. Time evolution of absorbance at 340 nm during electrolysis of 0.2 mM NAD+ in 100 mM Tris-buffer solution (dotted
line) and 260 mM (solid line) at electrode potential of –2.30 V using GC electrode. Temperature, T = 295±2 K.
https://doi.org/10.1371/journal.pone.0239340.g008
PLOS ONE | https://doi.org/10.1371/journal.pone.0239340 September 17, 2020
11 / 20
PLOS ONE Glassy carbon electrode for CO2 fixation
than the phosphate buffer solution. Namely, based on the standard curve for commercial
NADH (figure not shown here) only 50% of initial NAD+ was converted to the products using
phosphate buffer solution. However, the conversion increased to 70% when Tris-buffer was
used. The possible reason for low conversion using the phosphate buffer solution could be the
adsorption of phosphate groups on the electrode surface which blocks the electrode surface
and thus, minimize the electrochemically active surface area (EASA) of the electrode. In addi-
tion, it is very clear from Fig 8 that the concentration of Tris-buffer has no significant effect on
the regeneration kinetics of NAD+ reduction reaction.
Hence, it was found from these experiments that 100 mM Tris-buffer solution resulted an
improved conversion of NAD+ to the reduction products compared to phosphate buffer solu-
tion. Furthermore, the kinetics of NAD+ reduction is slightly slower than phosphate buffer
solution and the concentration of the Tris-buffer solution has no effect on both reduction
kinetics and recovery of active NADH.
In order to replicate the Calvin Cycle in a real biochemical reactor, other chemicals and
enzymes will also present in the reaction mixture. Thus, we performed the NADH regenera-
tion experiment with a reaction mixture of 10 mM MgCl2, 1mM Dithiothreitol (DTT) and 0.2
mM NAD+ in selected 100 mM Tris-buffer at pH 7.4 and room temperature as shown in Fig 9.
The trend in Fig 9 (dashed line and dotted lines) exhibits similar regeneration kinetics.
However, the conversion is decreased from 70 to 50% which may be due the adsorption of
MgCl2 on the surface of electrode. However, it was interesting to see that conversion increased
from 50 to 60% again with the addition of DTT (dashed lines). Since DTT is reducing agent
Fig 9. Time evolution of absorbance at 340 nm during electrolysis of 100 mM Tris solution containing 0.2 mM NAD+ and 10
mM MgCl2 (dotted line) and 0.2 mM NAD+, 10 mM MgCl2 and1 mM DTT (dashed line) at electrode potential of –2.30 V using
GC electrode and control experiment (solid line), no potential applied. Temperature, T = 295±2 K.
https://doi.org/10.1371/journal.pone.0239340.g009
PLOS ONE | https://doi.org/10.1371/journal.pone.0239340 September 17, 2020
12 / 20
PLOS ONE Glassy carbon electrode for CO2 fixation
Fig 10. Schematic of NADH regeneration module using free enzymes.
https://doi.org/10.1371/journal.pone.0239340.g010
therefore to see the reducing effect of DTT on the NAD+ reduction, we performed a control
experiment without applying potential Fig 9 (solid line). As the absorbance at 340 nm is zero
throughout the time of the experiment, therefore it confirmed that the increase in regeneration
kinetics was not due to the reducing effect of DTT.
The next step in the process was to integrate NADH regeneration system with other essen-
tial modules of the Calvin Cycle. However, before the NADH regeneration integration with
other modules, we tested the system with free enzymes in order to make sure that the system is
working. Therefore, experiments were conducted to study the regeneration of a system that
contained free enzyme as well as NAD+ as shown in Fig 10 and the corresponding results are
shown in Fig 11.
Fig 11 clearly shows that initially the absorbance at 340 nm is zero confirming that there
was no NADH in the system. However, the absorbance at 340 nm increases with time when an
electrode potential of –2.30 V was applied. Here, it is important to mention that both NADH
and NAD2 absorb at 340 nm but as we already showed in the first part of the manuscript that
our developed NADH regeneration system yields 97.45±0.8% of active NADH which is con-
firmed by the decrease in absorbance in Fig 11 since according to Eq 4, only enzymatically-
active NADH is consumed during the reaction and not enzymatically-inactive NAD2. There-
fore, the peak at 340 nm here corresponds to active NADH, not NAD2. Cyclic regeneration
was performed and the NAD+ was reduced 3 times over the 45 hr test period.
System integration
The key modules in artificial photosynthesis are: (i) adapted Calvin Cycle and biochemical
transformation which include enzyme production, separation, purification, and immobiliza-
tion, (ii) light-driven ATP synthesis and (iii) electrochemical regeneration of NADH. Fig 12
shows the schematic of corresponding integration of all modules into one system.
For the integration of all three modules the enzymes of the Calvin Cycle were immobilized
in hollow fiber system. The corresponding enzymes and their amounts are shown in Fig 12.
3 Þ. NAD+ was regenerated using a GC
CO2 was provided in the form of bicarbonate ðHCO(cid:0)
electrode in the same setup (Fig 2) at an applied potential of –2.30 V and ATP was generated
PLOS ONE | https://doi.org/10.1371/journal.pone.0239340 September 17, 2020
13 / 20
PLOS ONE Glassy carbon electrode for CO2 fixation
Fig 11. Time evolution of absorbance at 340 nm during free enzymes cycle containing 0.2 mM NAD+ at an electrode
potential of –2.30 V using GC electrode. Temperature, T = 295 K.
https://doi.org/10.1371/journal.pone.0239340.g011
by vesicles that were trapped and recirculated in the final hollow fiber system. To measure the
NADH concentration based on absorbance at 340 nm, two UV/Vis spectrometers were used.
The full cycle experiment was conducted over multiple days as shown in Fig 13.
As the system has no NADH at the start, therefore the absorbance at 340 nm is zero. How-
ever, when an electrode potential of –2.30 V was applied, one can observe a sharp increase in
the absorption at 340 nm. Over the 3-day period, regeneration was done 3 times (3 cycles)
which can be seen as sharp increases in Fig 13. The corresponding rates for NADH regenera-
tion and disappearance during the 3 cycles are shown in Table 1. The concentration of regen-
erated ATP in the integrated system was around 35 μM.
ATP was generated through coupled activity of proton pumping by bR in presence of LED
light to the vesicles trapped in the last hollow fiber system. ATP produced was measured by
Luciferin-Luciferase assay kit. NAD+ was regenerated to NADH over 2 hr. Then, NAD+ regen-
eration was stopped to allow the detection of its reduction over a one-day span. This would
allow for confirming of successful integration of NADH and ATP regeneration subsystems.
Then, regeneration was done again to replenish NADH concentration. To the best of authors’
knowledge, no such system has been reported in the literature till now.
The different modules involved in this technology including ATP regeneration and NADH
regeneration were individually developed successfully. It was showed that we can integrate the
system and run all the modules. However, the methodology for immobilization of the enzymes
needs to be modified since it was not efficient as expected. Although, we could integrate the
system successfully, the amount of capturing CO2 was really low due to the significant decrease
PLOS ONE | https://doi.org/10.1371/journal.pone.0239340 September 17, 2020
14 / 20
PLOS ONE Glassy carbon electrode for CO2 fixation
Fig 12. Schematic of the full Calvin cycle experiment including all three essential modules.
https://doi.org/10.1371/journal.pone.0239340.g012
in efficiency of the enzymes after immobilization. Therefore, we are currently working on
other immobilization techniques with minimum loss of efficiency.
Please note that the data in Table 2 is only related to NADH regeneration from commer-
cially available NAD+ while the current research is related to NADH regeneration during the
Calvin cycle (artificial photosynthesis).
PLOS ONE | https://doi.org/10.1371/journal.pone.0239340 September 17, 2020
15 / 20
PLOS ONE Glassy carbon electrode for CO2 fixation
Fig 13. Time evolution of absorbance at 340 nm during full cycle experiment containing 0.2 mM NAD+ at an electrode
potential of –2.30 V using GC electrode. Temperature, T = 295±2 K.
https://doi.org/10.1371/journal.pone.0239340.g013
Conclusions
An electrochemical reactor was developed for the reduction of NAD+ to NADH during Calvin
Cycle using a Glassy Carbon cathode operated at constant electrode potential. The results
showed that GC electrode is highly efficient and electrocatalytically active in regenerating
NADH from NAD+. At electrode potential of −2.30 V, a very high recovery of 97.45±0.8% was
achieved. The reason for this high recovery was the formation of active hydrogen, Hads on the
electrode surface which is immediately available for the NAD-radical electrochemical hydro-
genation. It was also shown that Tris-buffer solution resulted an improved conversion of
NAD+ to the reduction products compared to phosphate buffer solution with slightly slower
regeneration kinetics. The Tris-buffer concentration has no effect on the regeneration kinetics
and recovery of NADH. In addition, temperature greatly affects the stability of regenerated
NADH. Namely, at room temperature the NAD+/NADH was stable for a maximum period of
24 to 30 hr (4 cycles) while at low temperature it was stable for 96 hr. The developed system
worked effectively to regenerate NADH in a real biochemical reactor in the presence of
enzymes and ATP regeneration systems during the Calvin Cycle. However, the methodology
Table 1. NADH regeneration and consumption rates.
Cycle (day)
Regeneration (μM min-1)
Disappearance (μM min-1)
1
0.50
0.04
2
0.20
0.02
3
0.19
0.02
https://doi.org/10.1371/journal.pone.0239340.t001
PLOS ONE | https://doi.org/10.1371/journal.pone.0239340 September 17, 2020
16 / 20
PLOS ONE Table 2. Efficiency comparison of various methods/systems in the regeneration of enzymatically-active 1,4-NADH (highest efficiency at optimum conditions).
Glassy carbon electrode for CO2 fixation
Electrode
GC
GC
GC-Ru
Au-Pt
Cu
Au
Au-Hg
Pt (non-modified)
Au-Hg/cholesterol
Pt/anion_charged_memb
Hg
Rh3+_PPY/C
https://doi.org/10.1371/journal.pone.0239340.t002
Recovery of active NADH / %
97.45
References
Current work
98
98
64
54
28
10
50
75
65
50
52
[35]
[31]
[36]
[36]
[36]
[52]
[52]
[52]
[39]
[39]
[54]
for immobilization of the enzymes needs to be modified since it was not efficient as expected.
Namely, the amount of capturing CO2 provided by the bicarbonate was really low due to sig-
nificant decrease in efficiency of the enzymes after immobilization. By using CO2, this technol-
ogy not only reduces greenhouse gas emissions, but it generates revenue by producing
valuable and useful chemicals, including ethylene glycol. Therefore, our research will fit into
advanced technology solutions and/or infrastructure for CO2 conversion into valuable, useful
materials.
Supporting information
S1 File. Procedure for enzymatic assay.
(DOCX)
Author Contributions
Conceptualization: Mohammad Younas.
Data curation: Nehar Ullah.
Investigation: Irshad Ali, Saeid Amiri.
Methodology: Irshad Ali.
Project administration: Mashallah Rezakazemi.
Resources: Saeid Amiri.
Supervision: Mohammad Younas.
Visualization: Nehar Ullah, Mashallah Rezakazemi.
Writing – original draft: Nehar Ullah.
Writing – review & editing: Mohammad Younas, Mashallah Rezakazemi.
References
1. Etheridge DM, Steele LP, Langenfelds RL, Francey RJ, Barnola JM, Morgan VI. Natural and anthropo-
genic changes in atmospheric CO2 over the last 1000 years from air in Antarctic ice and firn. Journal of
Geophysical Research: Atmospheres. 1996; 101(D2):4115–28.
PLOS ONE | https://doi.org/10.1371/journal.pone.0239340 September 17, 2020
17 / 20
PLOS ONE Glassy carbon electrode for CO2 fixation
2. Socolow R, Hotinski R, Greenblatt JB, Pacala S. Solving the climate problem: Technologies available to
curb CO 2 emissions. Environment. 2004; 46(10):8–17.
3. Middleton RS, Eccles JK. The complex future of CO2 capture and storage: Variable electricity genera-
tion and fossil fuel power. Applied Energy. 2013; 108:66–73.
4. Song C. Global challenges and strategies for control, conversion and utilization of CO2 for sustainable
development involving energy, catalysis, adsorption and chemical processing. Catalysis Today. 2006;
115(1–4):2–32.
5. Olah GA, Goeppert A, Prakash GKS. Chemical recycling of carbon dioxide to methanol and dimethyl
ether: From greenhouse gas to renewable, environmentally carbon neutral fuels and synthetic hydro-
carbons. Journal of Organic Chemistry. 2009; 74(2):487–98. https://doi.org/10.1021/jo801260f PMID:
19063591
6. Solomon S, Plattner G-K, Knutti R, Friedlingstein P. Irreversible climate change due to carbon dioxide
emissions. 2009.
7.
Jiang Z, Xiao T, Kuznetsov VL, Edwards PP. Turning carbon dioxide into fuel. Philosophical Transac-
tions of the Royal Society A: Mathematical, Physical and Engineering Sciences. 2010; 368
(1923):3343–64. https://doi.org/10.1098/rsta.2010.0119 PMID: 20566515
8. Michael Schwartz RLC, Victoria M. Kehoe, Robert C. MacDuff, Jay Patel, Anthony F. Sammells. Car-
bon Dioxide Reduction to Alcohols using Perovskite-Type Electrocatalysts. Journal of Electrochemical
Society. 1993; 140(3):614–8.
9. Aydin R, Koleli F. Electrochemical reduction of CO2 on a polyaniline electrode under ambient conditions
and at high pressure in methanol. Journal of Electroanalytical Chemistry. 2002; 535(1–2):107–12.
10. Wu J, Risalvato FG, Ke FS, Pellechia PJ, Zhou XD. Electrochemical reduction of carbon dioxide I.
effects of the electrolyte on the selectivity and activity with Sn electrode. Journal of the Electrochemical
Society. 2012; 159(7):F353–F9.
11. Chenault HK, Whitesides GM. Regeneration of nicotinamide cofactors for use in organic synthesis.
Applied Biochemistry and Biotechnology. 1987; 14(2):147–97. https://doi.org/10.1007/BF02798431
PMID: 3304160
12. Wendell D, Todd J, Montemagno C. Artificial photosynthesis in ranaspumin-2 based foam. Nano Lett.
2010; 10(9):3231–6. Epub 2010/03/09. https://doi.org/10.1021/nl100550k PMID: 20205454.
13.
Zhao X, Li WF, Wang Y, Ma ZH, Yang SJ, Zhou Q, et al. Elevated CO2 concentration promotes photo-
synthesis of grape (Vitis vinifera L. cv. ’Pinot noir’) plantlet in vitro by regulating RbcS and Rca revealed
by proteomic and transcriptomic profiles. BMC Plant Biol. 2019; 19(1):42. Epub 2019/01/31. https://doi.
org/10.1186/s12870-019-1644-y PMID: 30696402; PubMed Central PMCID: PMC6352424.
14. Ebenhoh O, Spelberg S. The importance of the photosynthetic Gibbs effect in the elucidation of the Cal-
vin-Benson-Bassham cycle. Biochem Soc Trans. 2018; 46(1):131–40. Epub 2018/01/07. https://doi.
org/10.1042/BST20170245 PMID: 29305411; PubMed Central PMCID: PMC5818666.
15.
Tomimatsu H, Tang Y. Effects of high CO2 levels on dynamic photosynthesis: carbon gain, mecha-
nisms, and environmental interactions. J Plant Res. 2016; 129(3):365–77. Epub 2016/04/21. https://doi.
org/10.1007/s10265-016-0817-0 PMID: 27094437.
16. Bassham JA. Mapping the carbon reduction cycle: a personal retrospective. Photosynth Res. 2003; 76
(1–3):35–52. Epub 2005/10/18. https://doi.org/10.1023/A:1024929725022 PMID: 16228564.
17.
Tongul B, Kavakcıoglu B, Tarhan L. Partitioning and purification of menadione induced NAD (P) H oxi-
dase from Phanerochaete chrysosporium in aqueous two-phase systems. Separation and Purification
Technology. 2016; 163:275–81.
18. Sauve AA. NAD+ and vitamin B3: From metabolism to therapies. Journal of Pharmacology and Experi-
mental Therapeutics. 2008; 324(3):883–93. https://doi.org/10.1124/jpet.107.120758 PMID: 18165311
19. Pollak N, Dolle C, Ziegler M. The power to reduce: Pyridine nucleotides—Small molecules with a multi-
tude of functions. Biochemical Journal. 2007; 402(2):205–18. https://doi.org/10.1042/BJ20061638
PMID: 17295611
20. Delecouls-Servat K, Bergel A, Basseguy R. Membrane electrochemical reactors (MER) for NADH
regeneration in HLADH-catalysed synthesis: Comparison of effectiveness. Bioprocess and Biosystems
Engineering. 2004; 26(4):205–15. https://doi.org/10.1007/s00449-004-0356-2 PMID: 15057554
21. Karyakin AA, Ivanova YN, Karyakina EE. Equilibrium (NAD+/NADH) potential on poly(Neutral Red)
modified electrode. Electrochemistry Communications. 2003; 5(8):677–80.
22. Devaux-Basseguy R, Bergel A, Comtat M. Potential applications of NAD(P)-dependent oxidoreduc-
tases in synthesis: A survey. Enzyme and Microbial Technology. 1997; 20(4):248–58.
23. Ren X, Ghassemi P, Yuan W, Zhou J, Chong P, Noh M. Cell-free artificial photosynthesis system2017.
1859–62 p.
PLOS ONE | https://doi.org/10.1371/journal.pone.0239340 September 17, 2020
18 / 20
PLOS ONE Glassy carbon electrode for CO2 fixation
24. W Wendell D. Artificial photosynthesis processes as a means of producing biofuels2010. 855–60 p.
25. Heldt H-W, Piechulla B. 6—The Calvin cycle catalyzes photosynthetic CO2 assimilation. In: Heldt H-W,
Piechulla B, editors. Plant Biochemistry ( Fourth Edition). San Diego: Academic Press; 2011. p. 163–
91.
26. Kaur G, Srivastava AK, Chand S. Advances in biotechnological production of 1,3-propanediol. Bio-
chemical Engineering Journal. 2012; 64:106–18. https://doi.org/10.1016/J.Bej.2012.03.002
WOS:000304223100013.
27. Choi H-J, Montemagno CD. Artificial Organelle: ATP Synthesis from Cellular Mimetic Polymersomes.
Nano Letters. 2005; 5(12):2538–42. https://doi.org/10.1021/nl051896e PMID: 16351211
28. Heldt H-W, Heldt F. 6—The Calvin cycle catalyzes photosynthetic CO2 assimilation. 2005.
29. Wang X, Saba T, Yiu HH, Howe RF, Anderson JA, Shi J. Cofactor NAD (P) H regeneration inspired by
heterogeneous pathways. Chem. 2017; 2(5):621–54.
30.
Liu W, Wang P. Cofactor regeneration for sustainable enzymatic biosynthesis. Biotechnology
Advances. 2007; 25(4):369–84. https://doi.org/10.1016/j.biotechadv.2007.03.002 PMID: 17459647
31. Azem A, Man F, Omanovic S. Direct regeneration of NADH on a ruthenium modified glassy carbon elec-
trode. Journal of Molecular Catalysis A: Chemical. 2004; 219(2):283–99.
32. Van Der Donk WA, Zhao H. Recent developments in pyridine nucleotide regeneration. Current Opinion
in Biotechnology. 2003; 14(4):421–6. https://doi.org/10.1016/s0958-1669(03)00094-6 PMID: 12943852
33.
Leonida MD. Redox enzymes used in chiral syntheses coupled to coenzyme regeneration. Current
Medicinal Chemistry. 2001; 8(4):345–69. https://doi.org/10.2174/0929867013373390 PMID: 11172694
34. Ali I, Khan T, Omanovic S. Direct electrochemical regeneration of the cofactor NADH on bare Ti, Ni, Co
and Cd electrodes: The influence of electrode potential and electrode material. Journal of Molecular
Catalysis A: Chemical. 2014; 387:86–91. http://dx.doi.org/10.1016/j.molcata.2014.02.029.
35. Ali I, Soomro B, Omanovic S. Electrochemical regeneration of NADH on a glassy carbon electrode sur-
face: The influence of electrolysis potential. Electrochemistry Communications. 2011; 13(6):562–5.
36. Damian A, Maloo K, Omanovic S. Direct electrochemical regeneration of NADH on Au, Cu and Pt-Au
electrodes. Chemical and Biochemical Engineering Quarterly. 2007; 21(1):21–32.
37. Damian A, Omanovic S. Electrochemical reduction of NAD+ on a polycrystalline gold electrode. Journal
of Molecular Catalysis A: Chemical. 2006; 253(1–2):222–33.
38. Elving PJ, Bresnahan WT, Moiroux J, Samec Z. NAD/NADH as a model redox system: Mechanism,
mediation, modification by the environment. Bioelectrochemistry and Bioenergetics. 1982; 9(3):365–78.
39.
Jaegfeldt H. 420—A study of the products formed in the electrochemical reduction of nicotinamide-ade-
nine-dinucleotide. Bioelectrochemistry and Bioenergetics. 1981; 8(3):355–70.
40. Vuorilehto K, Lutz S, Wandrey C. Indirect electrochemical reduction of nicotinamide coenzymes. Bioe-
lectrochemistry. 2004; 65(1):1–7. Epub 2004/11/04. https://doi.org/10.1016/j.bioelechem.2004.05.006
PMID: 15522685.
41. Ali I, Gill A, Omanovic S. Direct electrochemical regeneration of the enzymatic cofactor 1,4-NADH
employing nano-patterned glassy carbon/Pt and glassy carbon/Ni electrodes. Chemical Engineering
Journal. 2012; 188:173–80. https://doi.org/10.1016/j.cej.2012.02.005. WOS:000303271800019.
42. Atkins JdP P. Elements of Physical Chemistry. New York: W. H. Freeman; 2009.
43. Ali I, Omanovic S. Kinetics of electrochemical reduction of NAD+ on a glassy carbon electrode. Interna-
tional Journal of Electrochemical Science. 2013; 8(3):4283–304.
44. Massey V, Gibson QH, Veeger C. Intermediates in the catalytic action of lipoyl dehydrogenase (diapho-
rase). Biochemical Journal. 1960; 77:341–0. https://doi.org/10.1042/bj0770341 PMID: 13767908
45.
Filipe OMS, Brett CMA. Characterization of carbon film electrodes for electroanalysis by electrochemi-
cal impedance. Electroanalysis. 2004; 16(12):994–1001.
46. Conway K. The Electrochemical Double Layer. New Jersey: The Electrochemical Society, Inc; 1997.
47. Wei L, Fan Y-J, Wang H-H, Tian N, Zhou Z-Y, Sun S-G. Electrochemically shape-controlled synthesis
in deep eutectic solvents of Pt nanoflowers with enhanced activity for ethanol oxidation2012. 468–74 p.
48. Ve´ lez J, Muena J, J Aguirre M, Ramı´rez G, Herrera F. Electrochemical Oxidation of Sulfite in Aqueous
Solution by Glassy Carbon Electrode Modified with Polymeric Co(II) Meso- Tetrakis (2-Thienyl) Porphy-
rin. 2012.
49. Rahman G, Lim JY, Jung KD, Joo OS. NAD+ hydrogenation on Au electrode deposited on modified
glassy carbon. Electrochemistry Communications. 12(10):1371–4.
50. Moiroux J, Elving PJ. Adsorption phenomena in the NAD+/NADH system at glassy carbon electrodes.
Journal of Electroanalytical Chemistry. 1979; 102(1):93–108.
PLOS ONE | https://doi.org/10.1371/journal.pone.0239340 September 17, 2020
19 / 20
PLOS ONE Glassy carbon electrode for CO2 fixation
51. Nakamura Y, Suye SI, Kira JI, Tera H, Tabata I, Senda M. Electron-transfer function of NAD+-immobi-
lized alginic acid. Biochimica et Biophysica Acta—General Subjects. 1996; 1289(2):221–5.
52. Baik SH, Kang C, Jeon IC, Yun SE. Direct electrochemical regeneration of NADH from NAD+ using cho-
lesterol-modified gold amalgam electrode. Biotechnology Techniques. 1999; 13(1):1–5.
53. Oh SJ, Kim N, Lee YTJJoMS. Preparation and characterization of PVDF/TiO2 organic–inorganic com-
posite membranes for fouling resistance improvement. 2009; 345(1–2):13–20.
54. Shimizu AK Y., Ito S. and Sasaki K. Denki Kagaku. 1993; 61.
PLOS ONE | https://doi.org/10.1371/journal.pone.0239340 September 17, 2020
20 / 20
PLOS ONE |
10.1371_journal.pone.0238029 | RESEARCH ARTICLE
The incidence and survival after in-hospital
cardiopulmonary cerebral resuscitation in
end-stage kidney disease patients: A
nationwide population-based study
Chia-Hung Yang1, Jia-Jin Chen2,3, Jih-Kai Yeh4, George Kuo2, Cheng-Chia Lee2, I-
Chang Hsieh4, Ming-Jer Hsieh4, Ya-Chung Tian2, Chih-Hsiang ChangID
2,3*
1 Department of Cardiology, New Taipei Municipal TuCheng Hospital, Chang Gung Memorial Hospital and
Chang Gung University, 2 Kidney Research Center, Department of Nephrology, Chang Gung Memorial
Hospital, Linkou Medical Center, Taoyuan, Taiwan, 3 School of Medicine, College of Medicine, Chang Gung
University, Taoyuan, Taiwan, 4 Department of Cardiology, Chang Gung Memorial Hospital, Linkou Medical
Center, Taoyuan, Taiwan
* franwisandsun@gmail.com
Abstract
Background
This study analyzed the survival and protective predictors of in-hospital cardiopulmonary
cerebral resuscitation (CPCR) to potentially help physicians create effective treatment plans
for End-stage kidney disease (ESKD) patients.
Methods
We extracted the data of 7,116 ESKD patients who received their first in-hospital CPCR
after initial dialysis between 2004 and 2012 from the National Health Insurance Research
Database. The primary outcome was the survival rate during the first in-hospital CPCR. The
secondary outcome was the median post-discharge survival.
Results
From 2004 through 2012, the incidence of in-hospital CPCR decreases from 3.97 to 3.67
events per 1,000 admission days (P for linear trend <0.001). The survival rate for the first in-
hospital CPCR did not change significantly across the 9 years (P for trend = 0.244), whereas
the median survival of post-discharge survival increased significantly from 3.0 months in
2004 to 6.8 months in 2011 (P for linear trend <0.001). In addition, multivariable analysis
identified older age as a risk factor and prior intracardiac defibrillator (ICD) or cardiac resyn-
chronization therapy defibrillator (CRT-D) implantation as a protective factor for in-hospital
death during the first in-hospital CPCR.
Conclusion
The incidence of in-hospital CPCR and the duration post-discharge among ESKD patients
improved despite there being no significant difference in the survival rate of ESKD patients
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: Yang C-H, Chen J-J, Yeh J-K, Kuo G, Lee
C-C, Hsieh I-C, et al. (2020) The incidence and
survival after in-hospital cardiopulmonary cerebral
resuscitation in end-stage kidney disease patients:
A nationwide population-based study. PLoS ONE
15(8): e0238029. https://doi.org/10.1371/journal.
pone.0238029
Editor: Antonio Cannatà, King’s College London,
UNITED KINGDOM
Received: November 20, 2019
Accepted: August 7, 2020
Published: August 28, 2020
Peer Review History: PLOS recognizes the
benefits of transparency in the peer review
process; therefore, we enable the publication of
all of the content of peer review and author
responses alongside final, published articles. The
editorial history of this article is available here:
https://doi.org/10.1371/journal.pone.0238029
Copyright: © 2020 Yang et al. This is an open
access article distributed under the terms of the
Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: All relevant data are
within the manuscript and its Supporting
Information files.
PLOS ONE | https://doi.org/10.1371/journal.pone.0238029 August 28, 2020
1 / 12
PLOS ONE The outcomes after CPCR in ESKD patients
after CPCP. Either ICD or CRT-D implantation may be advisable for ESKD patients with a
high risk of sudden cardiac death.
Funding: C-HC was supported by the Ministry of
Science and Technology (106-2314-B-182A-118-
MY3) and grants from Chang Gung Memorial
Hospital, Taiwan (CMRPG3F1653,
CMRPG5H0162). C-HY was supported by Chang
Gung Memorial Hospital, Taiwan (CORPG3G0761).
Competing interests: The authors have declared
that no competing interests exist.
Background
The incidence of in-hospital cardiac arrest in the general population has grown in the past
decade, in line with aging societies and increasingly complex comorbidities [1]. The survival-
to-discharge rate of cardiopulmonary cerebral resuscitation (CPCR), the possibility of favor-
able neurologic recovery, and functional status warrant the most consideration for patients fol-
lowing cardiac arrest. Although the advanced cardiac life support guidelines already define a
cardiac arrest algorithm for use in resuscitation, the effects of survival on in-hospital cardiac
arrest are rarely reported [2].
Despite the decreased overall death rate in dialysis patients since 2001, as found by a United
States Renal Data System report, sudden cardiac arrest and arrhythmia remain the leading cause of
death in end-stage kidney disease (ESKD) patients [3–5].One of the noteworthy features of ESKD
patients with cardiac arrest is the catastrophic outcome, with 30-day and 1-year survival rates of
32% and 15%, respectively [6], higher than those of the general population [7]. Most related
research has focused on risk factors for sudden cardiac death or cardiac arrest in ESKD patients [8–
13]. Some studies have suggested that beta blockers may have a preventive or prognostic effect for
cardiac arrest in ESKD patients and that the effects of intracardiac defibrillators (ICDs) are inconsis-
tent [14–17].According to Saeed et al. [7], patient age of less than 65 years, receiving cardiac cathe-
terization, and hospital teaching status are predictor factors for discharge to home. Beta blockers
and ACEI/ARB but not age are associated with higher survival in ESKD patients with cardiac arrest
in a cohort study by Pun et al. [15]. Studies examining the prognostic factor for ESKD patients with
cardiac arrest after CPCR appear to present controversial findings and are small in number; in addi-
tion, no large population-based study on the topic has been conducted in an Asian country.
Therefore, our aims in performing this large-scale population-based cohort study were to
1) determine the incidence of in-hospital CPCR among ESKD patients, 2) investigate the
trends in survival after in-hospital CPCR among ESKD patients, and 3) verify the protective
factors related to successful CPCR among ESKD patients.
Method
Data source
This was a retrospective cohort study conducted using data from the National Health Insur-
ance Research Database (NHIRD) of Taiwan, which was described in a previous study [18].
The NHIRD contains all diagnosis, treatment, medication, and other claims data for outpa-
tient and inpatient visits from all medical facilities contracted with the National Health Insur-
ance Administration. The diagnostic codes are those of the International Classification of
Diseases, 9th Revision, Clinical Modification (ICD-9-CM). All identification data are
encrypted before the information is released to researchers. The Institutional Review Board of
Linkou Chang Gung Memorial Hospital approved this study.
Study cohort and design
The study population was patients diagnosed with end-stage kidney disease who received
permanent dialysis between January 1, 2004, and December 31, 2012. Permanent dialysis
PLOS ONE | https://doi.org/10.1371/journal.pone.0238029 August 28, 2020
2 / 12
PLOS ONE The outcomes after CPCR in ESKD patients
was verified by ensuring the patients had received approval for a catastrophic illness certifi-
cate (ICD-9-CM: 585). Because the reasons of patients with out-of-hospital cardiac arrest
sometimes were not clear when coming to emergency department and the medical treat-
ments on the ambulance were not reported in NHIRD [19]. Therefore, the study cohort
was limited to ESKD patients who experienced a first in-hospital CPCR after dialysis
initiation.
Of the 93,887 ESKD patients initially extracted from the NHIRD, 99 were excluded because
of missing information, 415 were excluded because they were aged under 20 years at the time
of dialysis initiation, 8,850 were excluded because their follow-up duration was less than 3
months after dialysis initiation, and 77,407 were excluded because they did not experience in-
hospital CPCR after dialysis initiation. In our country, only patients with Do Not Resuscitate
were not performed CPCR. Furthermore, among the ESKD patients with the first in-hospital
CPCR, 728 were excluded because they had been diagnosed with a malignancy before CPCR
and 31 were excluded because they had received a kidney transplant before CPCR. Some
ESKD patients with malignancy had signed Do Not Resuscitate and also poor clinical out-
comes, which were based on cancer types and stages. Besides, patients with kidney transplant
had the experience to take immunosuppressive drugs, which might affected cardiovascular
risks [20]. Finally, the remaining 6,357 ESKD patients with a first in-hospital CPCR after dialy-
sis initiation were eligible for analysis (Fig 1).
Fig 1. Patient selection criteria.
https://doi.org/10.1371/journal.pone.0238029.g001
PLOS ONE | https://doi.org/10.1371/journal.pone.0238029 August 28, 2020
3 / 12
PLOS ONE The outcomes after CPCR in ESKD patients
Study outcome
The primary outcome was the survival rate during the first in-hospital CPCR. The secondary
outcome was the median post-discharge survival. We identified all hospital admissions and all
in-hospital CPCR events that occurred beyond 3 months after initial dialysis based on the
National Health Insurance (NHI) reimbursement codes. We did not include CPCR events that
occurred in emergency departments. In-hospital death was defined as withdrawal from the
NHI system within 7 days of the discharge date of the index in-hospital CPCR. All-cause mor-
tality was also defined as withdrawal from the NHI program. Patients who survived the index
hospitalization with CPCR were followed from the date of discharge until December 31, 2013,
or the date of death, whichever came first.
Patient characteristics
We extracted the characteristics of the study cohort at the time of first CPCR admission. The
data included sex, age, date of initial dialysis, and comorbidities, namely hypertension (ICD-
9-CM: 401–405), diabetes mellitus (ICD-9-CM: 250), dyslipidemia (ICD-9-CM: 272), atrial
fibrillation (ICD-9-CM: 42731), coronary artery disease (ICD-9-CM: 410–414), peripheral
arterial disease (ICD-9-CM: 4400, 4402, 4403, 4408, 4409, 443, 4440, 44422, 4448, 4478, and
4479), and prior ICD or cardiac resynchronization therapy defibrillator (CRT-D) implanta-
tion. Hospitalization history and medication before the first in-hospital CPCR were also
included. The comorbidities were defined as at least 2 outpatient visits or one inpatient diag-
nosis in the year before the index in-hospital CPCR. Hospitalization history could be tracked
to 1997. Prior ICD or CRT-D implantation was determined using NHI reimbursement codes.
Medication usage was extracted for the 3 months before the index in-hospital CPCR.
Statistical analysis
The baseline characteristics of the patients who died during the in-hospital CPCR (the
deceased group) and those who survived the admission (the survivor group) were compared
using the t test for continuous variables and the chi-square test for categorical variables. The
incidence of CPCR was estimated using the total number of in-hospital CPCR events per 1,000
admission days of all ESKD patients. Multiple CPCR events in the same patients in the differ-
ent days contributed to incidence calculation. The trend of CPCR incidence across the study
years was analyzed using a univariate Poisson regression in which the study year was the con-
tinuous explanatory variable and the logarithm of total admissions days was the offset variable.
The trend of surviving the CPCR admission across the study years was analyzed using a uni-
variate logistic regression. The trend of median post-discharge survival across the study years
was assessed using the Jonckheere–Terpstra test. Finally, factors associated with risk of in-hos-
pital death for CPCR were investigated using a multivariable logistic regression analysis in
which all the baseline characteristics were treated as explanatory variables [21].
A 2-sided P value <0.05 was considered to be statistically significant and no adjustment of
multiple testing (multiplicity) was made. The Jonckheere–Terpstra test was a one-sided statis-
tical test. All statistical analyses were performed using SAS version 9.4 (SAS Institute, Cary,
NC)
Results
Baseline characteristics
Table 1 presents the demographic data, including comorbidities, hospitalization history, and
medications, for the entire cohort and stratified by survivor or deceased during the index
PLOS ONE | https://doi.org/10.1371/journal.pone.0238029 August 28, 2020
4 / 12
PLOS ONE Table 1. Baseline characteristics for entire cohort and stratified by survivor or deceased during the index admission.
The outcomes after CPCR in ESKD patients
Variable
Demographic information
Age (years)
Age � 65 years, n (%)
Male sex, n (%)
Comorbidities in the previous year, n (%)
Hypertension
Diabetes mellitus
Dyslipidemia
Atrial fibrillation
Peripheral arterial disease
Coronary artery disease
ICD or CRT-D implantation
Charlson Comorbidity Index score
Hospitalization history, n (%)
Heart failure
Stroke
Myocardial infarction
Infection-related hospitalization
Medication, n (%)
Aspirin/clopidogrel
ACEI/ARB
β-blocker
Loop diuretics
K-sparing diuretics
Statin
Sodium bicarbonate
Calcium supplementation
Follow-up duration (years)
Entire cohort (n = 6,357)
Survivor (n = 1,198)
Deceased (n = 5,159)
P value
69.5 ± 12.1
4,241 (66.7)
3,244 (51.0)
5,474 (86.1)
4,565 (71.8)
1,140 (17.9)
536 (8.4)
994 (15.6)
3,257 (51.2)
22 (0.3)
5.6 ± 2.0
3,093 (48.7)
2,188 (34.4)
1,138 (17.9)
5,542 (87.2)
3,503 (55.1)
3,054 (48.0)
3,023 (47.6)
2,276 (35.8)
221 (3.5)
1,598 (25.1)
180 (2.8)
1,399 (22.0)
0.3 ± 1.1
67.0 ± 12.2
706 (58.9)
595 (49.7)
1,039 (86.7)
905 (75.5)
245 (20.5)
82 (6.8)
197 (16.4)
634 (52.9)
14 (1.2)
5.6 ± 2.0
569 (47.5)
394 (32.9)
203 (16.9)
1,033 (86.2)
693 (57.8)
604 (50.4)
607 (50.7)
422 (35.2)
38 (3.2)
330 (27.5)
35 (2.9)
273 (22.8)
1.6 ± 2.1
70.1 ± 12.0
3,535 (68.5)
2,649 (51.3)
4,435 (86.0)
3,660 (70.9)
895 (17.3)
454 (8.8)
797 (15.4)
2,623 (50.8)
8 (0.2)
5.6 ± 2.0
2,524 (48.9)
1,794 (34.8)
935 (18.1)
4,509 (87.4)
2,810 (54.5)
2,450 (47.5)
2,416 (46.8)
1,854 (35.9)
183 (3.5)
1,268 (24.6)
145 (2.8)
1,126 (21.8)
NA
<0.001
<0.001
0.294
0.492
0.001
0.012
0.028
0.393
0.195
<0.001
0.294
0.373
0.216
0.338
0.274
0.034
0.068
0.017
0.643
0.523
0.033
0.835
0.469
NA
ACEI, angiotensin converting enzyme inhibitor; ARB, angiotensin receptor blocker; CRT-D, cardiac resynchronization therapy defibrillator; ICD, intracardiac
defibrillator; NA, not applicable.
Data are presented as frequency (percentage) or mean ± standard deviation.
https://doi.org/10.1371/journal.pone.0238029.t001
admission. The survival rate from the first in-hospital CPCR to discharge was 18.8% (1,198 of
6,357 patients). The patients in the survivor group were on average younger than those in the
deceased group. Compared with the deceased group, the survivor group had a significantly
higher prevalence of diabetes mellitus, dyslipidemia, and ICD or CRT-D implantation and a
lower prevalence of atrial fibrillation. Regarding medication use in the previous 3 months, the
patients in the survivor group were more likely to be prescribed aspirin/clopidogrel, beta
blockers, and statins than were those in the deceased group.
Incidence of CPCR and median survival of post-discharge survival
From 2004 through 2012, the incidence of CPCR decreased from 3.97 to 3.67 events per 1,000
admission days, respectively (P for trend <0.001) (Fig 2; detailed data in S1 Table). The sur-
vival rate from the first in-hospital CPCR did not change significantly across the 9 years (P for
trend = 0.244), whereas the median survival of post-discharge survival increased significantly
PLOS ONE | https://doi.org/10.1371/journal.pone.0238029 August 28, 2020
5 / 12
PLOS ONE The outcomes after CPCR in ESKD patients
Fig 2. Incidence of Cardiopulmonary Cerebral Resuscitation (CPCR) across the study years.
https://doi.org/10.1371/journal.pone.0238029.g002
from 3.0 months in 2004 to 6.8 months in 2011 (P for trend <0.001) (Fig 3; detailed data in
S2 Table).
Associated factors of in-hospital death of CPCR
Table 2 shows the findings of multivariable logistic regression analysis, including all baseline
characteristics, for predicting in-hospital CPCR death. The result revealed that an older patient
age (odds ratio 1.21, 95% confidence interval [CI] 1.15–1.28) was a risk factor for in-hospital
death, while the presence of ICD or CRT-D implantation (odds ratio 0.15, 95% CI 0.06–0.37)
was a protective factor for in-hospital death.
Fig 3. CPCR admission survival and median post-discharge survival across the study years. The median survival month after
discharge was not calculated in 2012 due to insufficient potential follow-up.
https://doi.org/10.1371/journal.pone.0238029.g003
PLOS ONE | https://doi.org/10.1371/journal.pone.0238029 August 28, 2020
6 / 12
PLOS ONE Table 2. Factors associated with risks of in-hospital death of CPCR.
The outcomes after CPCR in ESKD patients
Variable
Age (per 10 years)
Male sex
Comorbidities
Hypertension
Diabetes mellitus
Dyslipidemia
Atrial fibrillation
Peripheral arterial disease
Coronary artery disease
ICD or CRT-D implantation
Hospitalization history
Heart failure
Stroke
Myocardial infarction
Infection-related hospitalization
Medication
Aspirin/clopidogrel
ACEI/ARB
β-blocker
Loop diuretics
K-sparing diuretics
Statin
Sodium bicarbonate
Calcium supplementation
Odds ratio
1.21
1.13
0.997
0.86
0.91
1.19
0.99
0.93
0.15
1.05
1.06
1.16
1.07
0.89
0.97
0.96
1.06
1.16
1.02
0.93
0.98
95% CI
1.15–1.28
0.99–1.28
0.82–1.21
0.74–1.01
0.76–1.08
0.93–1.53
0.83–1.18
0.80–1.07
0.06–0.37
0.92–1.20
0.92–1.21
0.97–1.39
0.89–1.30
0.77–1.03
0.84–1.11
0.84–1.10
0.92–1.23
0.81–1.67
0.86–1.20
0.63–1.36
0.83–1.14
P value
<0.001
0.074
0.977
0.060
0.280
0.170
0.905
0.318
<0.001
0.476
0.444
0.108
0.460
0.114
0.654
0.571
0.403
0.423
0.857
0.690
0.748
ACEI, angiotensin converting enzyme inhibitor; ARB, angiotensin receptor blocker; CI, confidence interval; CPCR,
cardiopulmonary cerebral resuscitation; CRT-D, cardiac resynchronization therapy defibrillator; ICD, intracardiac
defibrillator.
https://doi.org/10.1371/journal.pone.0238029.t002
Discussion
This study yielded four major findings. First, from 2004 through 2013, the incidence of in-hos-
pital CPCR for ESKD patients decreased significantly, falling from 3.97 to 3.67 events per
1,000 admission days. Second, although the in-hospital survival rate did not exhibit improve-
ment, the median months of post-discharge survival increased significantly to 6.8 months.
Third, multivariable analysis revealed that only age and ICD or CRT-D implantation were
strong predictors for successful CPCR and that the traditional risk factors for cardiac death
were statistical nonsignificant. Finally, neither beta blockers nor ACEI/ARB was associated
with a survival benefit.
Previous studies
The outcomes of ESKD patients after in-hospital CPCR have rarely been reported in previous
studies. Of the relevant research present in the literature, much of the data are older and/or
based on small-population studies, rendering them inadequate for properly reflecting the cur-
rent CPCR algorithm in wide use. The findings of previous related studies are summarized in
S3 Table. As the data show, the prognosis after CPCR in ESKD patients is poor, with a dis-
charge alive rate ranging from 0% to 26.1% in the studies (excluding that of Lafrance et al.
PLOS ONE | https://doi.org/10.1371/journal.pone.0238029 August 28, 2020
7 / 12
PLOS ONE The outcomes after CPCR in ESKD patients
[22]). Three studies [7, 23, 24] have confirmed the prognosis of cardiac arrest or acute coro-
nary syndrome as being worse in ESKD patients than in the general population. Wong et al.
[21] found that the incidence of in-hospital CPCR increased from 2000 through 2010 but that
the survival rate improved in the later years, similar to the findings of Saeed et al. [7]. Herzog
et al. found that the incidence of cardiac arrest increased along with dialysis vintage (cardiac
arrest rates of 93 and 164 events/1,000 patient years at years 1 and 4, respectively), with a
1-year survival rate of 15%. Diabetes mellitus was found to be a risk factor and poor outcome
predictor for sudden cardiac arrest in their study [6]. As mentioned, the clinic outcomes of
ESKD patients improved in previous study but remained poorer than those of the general pop-
ulation. Therefore, determining possible protective factors for dialysis patients after CPCR
should be a crucial goal of future research.
Current study
Compared with that of the U.S.-based study by Wong et al., the incidence of in-hospital CPCR
for ESKD patients was higher in the current Taiwan-based study, even after the incidence
decreased significantly to 3.67 CPR events per 1,000 admission days in 2012. From 2004 to
2008, there was a small increase in the incidence of in-hospital CPCR in our cohort study;
however, the overall trend remained downward from 2004 to 2013. Furthermore, the in-hospi-
tal survival rate of ESKD patients after CPCR improved slightly but was statistical nonsignifi-
cant, in contrast to the studies of Saeed et al. [23] and Wong et al. [21]. There are 2 possible
reasons for a significant decrease in in-hospital CPR. The first is the improvement of overall
medical care quality in Taiwan throughout the NHI era. The second is the passage of the Natu-
ral Death Act in 2000, which was aimed at avoiding unnecessary CPR in terminal patients. The
increase we observed in median months of post-discharge survival suggest that survival and
discharge to home status were better in 2012 than in previous years. Available data from the
general population also indicate that survival and neurologic status after CPR are improving,
probably as a result of better resuscitation care and CPR techniques [23, 25].
In line with previous research, the present study used multivariable regression analysis and
confirmed Pun’s observation [15], indicating that traditional cardiovascular factors including
diabetes mellitus, hypertension, dyslipidemia, and underlying coronary artery disease might
not be correlated with survival after cardiac arrest. This result may be explained by considering
that the leading cause of death in dialysis is sudden cardiac death and arrhythmia and that the
risk factor/pathogenesis of sudden cardiac death in ESKD patient differ from those of cardio-
vascular death in the general population. These risk factors include cardiac pathophysiology
change (e.g., myocardial interstitial fibrosis, microvascular calcification, QT prolongation, left
ventricular hypertrophy, and ventricular dysfunction), rapid electrolyte shift, chronic inflam-
mation, uremia, mineral and bone disorder in chronic kidney disease, hypervolemia, and the
dialysis procedure itself [8–13, 26, 27].
To our knowledge, this is the first nationwide, population-based study to examine the role
of ICD implantation in ESKD patient. Both ICD and CRT-D implantation seemed to be a pos-
sible cardioprotective factor in ESKD patients after CPCR in our study. Wan et al. reported
that ESKD patients with wearable cardioverter defibrillators had better survival after the event
of sudden cardiac arrest [28]. In addition, Herzog at el. found that only 7.6% of ESKD patients
who survived a cardiac arrest had received an ICD implantation. Their study also revealed that
ICD implantation had a secondary preventive role, namely that it was associated with greater
survival in cardiac arrest survivors [16]. A meta-analysis by Chen et al. [29] found that ICD
implantation improved overall survival in ESKD patients with heart failure. The abovemen-
tioned findings may suggest some beneficial effect of ICD implantation in ESRD patients.
PLOS ONE | https://doi.org/10.1371/journal.pone.0238029 August 28, 2020
8 / 12
PLOS ONE The outcomes after CPCR in ESKD patients
However, data from a prospective, randomized study are lacking and no international consen-
sus exists regarding ICD use in ESKD patients. Two prospective, randomized trials, Implant-
able Cardioverter Defibrillators in Dialysis Patients (ISRCTN20479861) [30] and Wearable
Cardioverter Defibrillator in Hemodialysis Patients (NCT02481206) are currently evaluating
the role of ICD in sudden cardiac death prevention in ESKD patients, but results were still
impending as of March 2019.
The present study indicated that aspirin/clopidogrel, beta blockers, and statins might be
associated with survival after CPCR, but these associations were not statistical significant after
multivariable regression analysis. A previous study suggested that the pathogenesis of cardio-
vascular events in ESKD patients differ from those in the general population [5] and, as men-
tioned above, the pathogenesis of cardiovascular death in dialysis also differs. Nevertheless,
many previous studies have shown the beneficial effects of beta blockers in lowing sudden car-
diac mortality and reducing the incidence of sudden cardiac death specifically in ESKD
patients [15, 31–33]. The underlying reason or reasons for this remain unclear. One possible
explanation is that to avoid intradialytic hypotension, the pre-dialytic dosage of cardiovascular
drugs is sometimes reduced by the patient or physician; therefore, the actual dose may be
lower than that prescribed.
Clinical implications
Sudden cardiac death and ventricular arrhythmias are common outcomes for ESKD patients
who receive CPCR [34, 35]. In the general population, ICD implantation could reduce mortal-
ity in patients with sudden cardiac arrest [36]. The role of ICD implantation in the primary
prevention of sudden cardiac death in ESKD patients is obscure and the data from 2 large pro-
spective randomized studies that may shed further light on the underlying mechanism are still
not yet available. The results of our cohort study lend support to the notion that ICD implanta-
tion has a survival benefit in ESKD patients after CPCR. These results imply that ICD implan-
tation might improve clinical outcomes for ESKD patients who are at high risk of sudden
cardiac death or otherwise clinically indicated.
Study limitations
This study was based on data from a large administrative database and thus its design has sev-
eral limitations. First, no personal data such as family history and lifestyle or laboratory data
for parameters including creatinine level, BP records, or lipid data were available. Therefore,
the etiology of in-hospital CPCR, such as the details of the cardiac arrest events including the
initial cardiac arrest rhythm, witness status, bystander status and treatments used during and
after resuscitation, could not be included. Not like patients with OHCA, witness status and
bystander status were less important for patients with in-hospital cardiac arrest [22, 37, 38].
Second, this cohort study only included ESKD patients with in-hospital CPCR and didn’t
include ESKD patient with CPCR in emergency departments. Therefore, the extrapolation
validity was limited to ESKD patients with in-hospital CPCR and selection bias may exist.
Third, we did not have data regarding the patients’ status after in-hospital CPCR, such as their
levels of disability or nursing home use. Fourth, the clinical indications for patients using beta
blockers were unavailable, as was their actual and prescribe doses. Finally, we were unable to
take repeat CPCR in the same day into account.
Conclusion
The incidence of in-hospital CPCR and the survival duration post-discharge among ESKD
patients improved despite there being no significant difference in the survival rate of ESKD
PLOS ONE | https://doi.org/10.1371/journal.pone.0238029 August 28, 2020
9 / 12
PLOS ONE The outcomes after CPCR in ESKD patients
patients after CPCP. Either ICD or CRT-D implantation may be advisable for selected dialysis
patients with a high risk of sudden cardiac death or with clinical indication.
Supporting information
S1 Table. Admitted CPCR events per 1,000 admission days of admitted ESKD patients.
(DOCX)
S2 Table. Median survival months.
(DOCX)
S3 Table. Summary of studies about ESKD patient post CPCR outcome.
(DOCX)
Author Contributions
Data curation: Jih-Kai Yeh.
Formal analysis: George Kuo.
Investigation: Cheng-Chia Lee.
Methodology: I-Chang Hsieh.
Resources: Ming-Jer Hsieh.
Software: Ya-Chung Tian.
Supervision: Chih-Hsiang Chang.
Writing – original draft: Chia-Hung Yang.
Writing – review & editing: Jia-Jin Chen.
References
1. Merchant RM, Yang L, Becker LB, Berg RA, Nadkarni V, Nichol G, et al. Incidence of treated cardiac
arrest in hospitalized patients in the United States. Crit Care Med. 2011; 39(11):2401–6. https://doi.org/
10.1097/CCM.0b013e3182257459 PMID: 21705896; PubMed Central PMCID: PMC3196742.
2. Ecc Committee S, Task Forces of the American Heart A. 2005 American Heart Association Guidelines
for Cardiopulmonary Resuscitation and Emergency Cardiovascular Care. Circulation. 2005; 112(24
Suppl):IV1–203. https://doi.org/10.1161/CIRCULATIONAHA.105.166550 PMID: 16314375.
3. Saravanan P, Davidson NC. Risk assessment for sudden cardiac death in dialysis patients. Circ
Arrhythm Electrophysiol. 2010; 3(5):553–9. Epub 2010/10/21. https://doi.org/10.1161/CIRCEP.110.
937888 PMID: 20959609.
4. Collins AJ, Foley RN, Gilbertson DT, Chen SC. United States Renal Data System public health surveil-
lance of chronic kidney disease and end-stage renal disease. Kidney Int Suppl (2011). 2015; 5(1):2–7.
https://doi.org/10.1038/kisup.2015.2 PMID: 26097778; PubMed Central PMCID: PMC4455192.
5. Wanner C, Krane V, Marz W, Olschewski M, Mann JF, Ruf G, et al. Atorvastatin in patients with type 2
diabetes mellitus undergoing hemodialysis. N Engl J Med. 2005; 353(3):238–48. https://doi.org/10.
1056/NEJMoa043545 PMID: 16034009.
6. Herzog CA. Cardiac arrest in dialysis patients: approaches to alter an abysmal outcome. Kidney Int
Suppl. 2003;(84):S197–200. Epub 2003/04/16. https://doi.org/10.1046/j.1523-1755.63.s84.17.x PMID:
12694343.
7. Saeed F, Adil MM, Malik AA, Schold JD, Holley JL. Outcomes of In-Hospital Cardiopulmonary Resusci-
tation in Maintenance Dialysis Patients. J Am Soc Nephrol. 2015; 26(12):3093–101. https://doi.org/10.
1681/ASN.2014080766 PMID: 25908784; PubMed Central PMCID: PMC4657831.
8. Kanbay M, Afsar B, Goldsmith D, Covic A. Sudden death in hemodialysis: an update. Blood Purif. 2010;
30(2):135–45. Epub 2010/08/28. https://doi.org/10.1159/000320370 PMID: 20798493.
PLOS ONE | https://doi.org/10.1371/journal.pone.0238029 August 28, 2020
10 / 12
PLOS ONE The outcomes after CPCR in ESKD patients
9. Pun PH, Lehrich RW, Honeycutt EF, Herzog CA, Middleton JP. Modifiable risk factors associated with
sudden cardiac arrest within hemodialysis clinics. Kidney Int. 2011; 79(2):218–27. Epub 2010/09/03.
https://doi.org/10.1038/ki.2010.315 PMID: 20811332.
10. Burton JO, Jefferies HJ, Selby NM, McIntyre CW. Hemodialysis-induced cardiac injury: determinants
and associated outcomes. Clin J Am Soc Nephrol. 2009; 4(5):914–20. Epub 2009/04/10. https://doi.
org/10.2215/CJN.03900808 PMID: 19357245; PubMed Central PMCID: PMC2676185.
11. Karnik JA, Young BS, Lew NL, Herget M, Dubinsky C, Lazarus JM, et al. Cardiac arrest and sudden
death in dialysis units. Kidney Int. 2001; 60(1):350–7. https://doi.org/10.1046/j.1523-1755.2001.00806.
x PMID: 11422771.
12. Herzog CA, Strief JW, Collins AJ, Gilbertson DT. Cause-specific mortality of dialysis patients after coro-
nary revascularization: why don’t dialysis patients have better survival after coronary intervention?
Nephrol Dial Transplant. 2008; 23(8):2629–33. https://doi.org/10.1093/ndt/gfn038 PMID: 18299298;
PubMed Central PMCID: PMC2727291.
13. Amann K, Buzello M, Simonaviciene A, Miltenberger-Miltenyi G, Koch A, Nabokov A, et al. Capillary/
myocyte mismatch in the heart in renal failure—a role for erythropoietin? Nephrol Dial Transplant. 2000;
15(7):964–9. https://doi.org/10.1093/ndt/15.7.964 PMID: 10862632.
14.
Jadoul M, Thumma J, Fuller DS, Tentori F, Li Y, Morgenstern H, et al. Modifiable practices associated
with sudden death among hemodialysis patients in the Dialysis Outcomes and Practice Patterns Study.
Clin J Am Soc Nephrol. 2012; 7(5):765–74. Epub 2012/03/10. https://doi.org/10.2215/CJN.08850811
PMID: 22403271; PubMed Central PMCID: PMC3338277.
15. Pun PH, Lehrich RW, Smith SR, Middleton JP. Predictors of survival after cardiac arrest in outpatient
hemodialysis clinics. Clin J Am Soc Nephrol. 2007; 2(3):491–500. https://doi.org/10.2215/CJN.
02360706 PMID: 17699456.
16. Herzog CA, Li S, Weinhandl ED, Strief JW, Collins AJ, Gilbertson DT. Survival of dialysis patients after
cardiac arrest and the impact of implantable cardioverter defibrillators. Kidney Int. 2005; 68(2):818–25.
https://doi.org/10.1111/j.1523-1755.2005.00462.x PMID: 16014061.
17. Pun PH, Hellkamp AS, Sanders GD, Middleton JP, Hammill SC, Al-Khalidi HR, et al. Primary prevention
implantable cardioverter defibrillators in end-stage kidney disease patients on dialysis: a matched
cohort study. Nephrol Dial Transplant. 2015; 30(5):829–35. Epub 2014/11/19. https://doi.org/10.1093/
ndt/gfu274 PMID: 25404241; PubMed Central PMCID: PMC4425479.
18. Hsing AW, Ioannidis JP. Nationwide Population Science: Lessons From the Taiwan National Health
Insurance Research Database. JAMA Intern Med. 2015; 175(9):1527–9. https://doi.org/10.1001/
jamainternmed.2015.3540 PMID: 26192815.
19. Axelsson C, Claesson A, Engdahl J, Herlitz J, Hollenberg J, Lindqvist J, et al. Outcome after out-of-hos-
pital cardiac arrest witnessed by EMS: changes over time and factors of importance for outcome in
Sweden. Resuscitation. 2012; 83(10):1253–8. Epub 2012/03/06. https://doi.org/10.1016/j.resuscitation.
2012.01.043 PMID: 22388092.
20. Melilli E, Manonelles A, Montero N, Grinyo J, Martinez-Castelao A, Bestard O, et al. Impact of immuno-
suppressive therapy on arterial stiffness in kidney transplantation: are all treatments the same? Clin Kid-
ney J. 2018; 11(3):413–21. Epub 2018/07/11. https://doi.org/10.1093/ckj/sfx120 PMID: 29988241;
PubMed Central PMCID: PMC6007381.
21. Wong SP, Kreuter W, Curtis JR, Hall YN, O’Hare AM. Trends in in-hospital cardiopulmonary resuscita-
tion and survival in adults receiving maintenance dialysis. JAMA Intern Med. 2015; 175(6):1028–35.
Epub 2015/04/29. https://doi.org/10.1001/jamainternmed.2015.0406 PMID: 25915762; PubMed Cen-
tral PMCID: PMC4451394.
22.
Lafrance JP, Nolin L, Senecal L, Leblanc M. Predictors and outcome of cardiopulmonary resuscitation
(CPR) calls in a large haemodialysis unit over a seven-year period. Nephrol Dial Transplant. 2006; 21
(4):1006–12. Epub 2005/12/31. https://doi.org/10.1093/ndt/gfk007 PMID: 16384828.
23. Girotra S, Nallamothu BK, Spertus JA, Li Y, Krumholz HM, Chan PS, et al. Trends in survival after in-
hospital cardiac arrest. N Engl J Med. 2012; 367(20):1912–20. https://doi.org/10.1056/
NEJMoa1109148 PMID: 23150959; PubMed Central PMCID: PMC3517894.
24.
Fu CM, Chang CH, Lee CC, Fan PC, Chen SW, Lee CT, et al. Impact of dialysis dependence on prog-
nosis in patients with myocardial infarction: An 11-year population-based study. Medicine (Baltimore).
2018; 97(6):e9833. Epub 2018/02/09. https://doi.org/10.1097/MD.0000000000009833 PMID:
29419688; PubMed Central PMCID: PMC5944684.
25. Kazaure HS, Roman SA, Rosenthal RA, Sosa JA. Cardiac arrest among surgical patients: an analysis
of incidence, patient characteristics, and outcomes in ACS-NSQIP. JAMA Surg. 2013; 148(1):14–21.
https://doi.org/10.1001/jamasurg.2013.671 PMID: 23324834.
26. Yen CL, Tu KH, Lin MS, Chang SW, Fan PC, Hsiao CC, et al. Does a Supplemental Low-Protein Diet
Decrease Mortality and Adverse Events After Commencing Dialysis? A Nationwide Cohort Study.
PLOS ONE | https://doi.org/10.1371/journal.pone.0238029 August 28, 2020
11 / 12
PLOS ONE The outcomes after CPCR in ESKD patients
Nutrients. 2018; 10(8). Epub 2018/08/12. https://doi.org/10.3390/nu10081035 PMID: 30096785;
PubMed Central PMCID: PMC6115816.
27. Chung CM, Lin MS, Chang CH, Cheng HW, Chang ST, Wang PC, et al. Moderate to high intensity statin
in dialysis patients after acute myocardial infarction: A national cohort study in Asia. Atherosclerosis.
2017; 267:158–66. Epub 2017/10/08. https://doi.org/10.1016/j.atherosclerosis.2017.09.018 PMID:
28985950.
28. Wan C, Herzog CA, Zareba W, Szymkiewicz SJ. Sudden cardiac arrest in hemodialysis patients with
wearable cardioverter defibrillator. Ann Noninvasive Electrocardiol. 2014; 19(3):247–57. https://doi.org/
10.1111/anec.12119 PMID: 24252154; PubMed Central PMCID: PMC4034590.
29. Chen TH, Wo HT, Chang PC, Wang CC, Wen MS, Chou CC. A meta-analysis of mortality in end-stage
renal disease patients receiving implantable cardioverter defibrillators (ICDs). PLoS One. 2014; 9(7):
e99418. Epub 2014/07/19. https://doi.org/10.1371/journal.pone.0099418 PMID: 25036181; PubMed
Central PMCID: PMC4103758.
30.
de Bie MK, Lekkerkerker JC, van Dam B, Gaasbeek A, van Buren M, Putter H, et al. Prevention of sud-
den cardiac death: rationale and design of the Implantable Cardioverter Defibrillators in Dialysis patients
(ICD2) Trial—a prospective pilot study. Curr Med Res Opin. 2008; 24(8):2151–7. Epub 2008/06/20.
https://doi.org/10.1185/03007990802237343 PMID: 18561878.
31. Matsue Y, Suzuki M, Nagahori W, Ohno M, Matsumura A, Hashimoto Y. beta-blocker prevents sudden
cardiac death in patients with hemodialysis. Int J Cardiol. 2013; 165(3):519–22. https://doi.org/10.1016/
j.ijcard.2011.09.037 PMID: 21996409.
32. Cice G, Ferrara L, D’Andrea A, D’Isa S, Di Benedetto A, Cittadini A, et al. Carvedilol increases two-year
survivalin dialysis patients with dilated cardiomyopathy: a prospective, placebo-controlled trial. J Am
Coll Cardiol. 2003; 41(9):1438–44. https://doi.org/10.1016/s0735-1097(03)00241-9 PMID: 12742278.
33.
Tang CH, Wang CC, Chen TH, Hong CY, Sue YM. Prognostic Benefits of Carvedilol, Bisoprolol, and
Metoprolol Controlled Release/Extended Release in Hemodialysis Patients with Heart Failure: A 10-
Year Cohort. J Am Heart Assoc. 2016; 5(1). Epub 2016/01/08. https://doi.org/10.1161/JAHA.115.
002584 PMID: 26738790; PubMed Central PMCID: PMC4859376.
34. Davis TR, Young BA, Eisenberg MS, Rea TD, Copass MK, Cobb LA. Outcome of cardiac arrests
attended by emergency medical services staff at community outpatient dialysis centers. Kidney Int.
2008; 73(8):933–9. https://doi.org/10.1038/sj.ki.5002749 PMID: 18172435.
35. Bleyer AJ, Russell GB, Satko SG. Sudden and cardiac death rates in hemodialysis patients. Kidney Int.
1999; 55(4):1553–9. https://doi.org/10.1046/j.1523-1755.1999.00391.x PMID: 10201022.
36. Adabag S, Patton KK, Buxton AE, Rector TS, Ensrud KE, Vakil K, et al. Association of Implantable Car-
dioverter Defibrillators With Survival in Patients With and Without Improved Ejection Fraction: Second-
ary Analysis of the Sudden Cardiac Death in Heart Failure Trial. JAMA Cardiol. 2017; 2(7):767–74.
https://doi.org/10.1001/jamacardio.2017.1413 PMID: 28724134; PubMed Central PMCID:
PMC5710619.
37. Martinell L, Nielsen N, Herlitz J, Karlsson T, Horn J, Wise MP, et al. Early predictors of poor outcome
after out-of-hospital cardiac arrest. Crit Care. 2017; 21(1):96. Epub 2017/04/16. https://doi.org/10.1186/
s13054-017-1677-2 PMID: 28410590; PubMed Central PMCID: PMC5391587.
38.
Imbeault B, Chan CT. Cardiopulmonary Resuscitation in Outpatient Dialysis Clinics: Perception of Futil-
ity? J Am Soc Nephrol. 2019. Epub 2019/02/09. https://doi.org/10.1681/ASN.2019010046 PMID:
30733236; PubMed Central PMCID: PMC6405143.
PLOS ONE | https://doi.org/10.1371/journal.pone.0238029 August 28, 2020
12 / 12
PLOS ONE |
10.1371_journal.pone.0239135 | RESEARCH ARTICLE
Economic effects of policy options restricting
antimicrobial use for high risk cattle placed in
U.S. feedlots
Guillaume LhermieID
Verteramo ChiuID
Morgan ScottID
1,2*, Pierre Sauvage2, Loren William Tauer3, Leslie
1, Karun Kanyiamattam1, Ahmed Ferchiou2, Didier Raboisson2, Harvey
4, David R. Smith5, Yrjo Tapio Grohn1
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: Lhermie G, Sauvage P, Tauer LW, Chiu
LV, Kanyiamattam K, Ferchiou A, et al. (2020)
Economic effects of policy options restricting
antimicrobial use for high risk cattle placed in U.S.
feedlots. PLoS ONE 15(9): e0239135. https://doi.
org/10.1371/journal.pone.0239135
Editor: Juan J. Loor, University of Illinois, UNITED
STATES
Received: March 16, 2020
Accepted: August 31, 2020
Published: September 15, 2020
Copyright: © 2020 Lhermie et al. This is an open
access article distributed under the terms of the
Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: All relevant data are
within the manuscript and its Supporting
Information files.
Funding: This work was supported by Agriculture
and Food Research Initiative Competitive Grant no.
2016-68003-24607 from the USDA National
Institute of Food and Agriculture. HMS and YTG
were recipient of the award. National Institute of
Food and Agriculture - USDA. nifa.usda.gov The
funders had no role in study design, data collection
1 Department of Population Medicine and Diagnostic Sciences, College of Veterinary Medicine, Cornell
University, Ithaca, NY, United States of America, 2 IHAP, Universite´ de Toulouse, INRA, ENVT, Toulouse,
France, 3 Dyson School of Applied Economics and Management, Cornell SC Johnson College of Business,
Cornell University, Ithaca, NY, United States of America, 4 Department of Veterinary Pathobiology, College
of Veterinary Medicine and Biomedical Sciences, Texas A&M University, College Station, TX, United States
of America, 5 Department of Pathobiology and Population Medicine, College of Veterinary Medicine,
Mississippi State University, Starkville, MS, United States of America
* gl467@cornell.edu
Abstract
The rising public health threat of antimicrobial resistance, the influence of food service com-
panies, as well as the overall lack of positive image of using medical products in intensive
farming are major drivers curbing antimicrobial use. In the future, government policies may
affect practices of antimicrobial use in beef production in feedlots, a prominent current user
of antimicrobials in animal agriculture, but also the agricultural industry generating the high-
est cash receipt in the U.S. Our objective was to estimate the cost effect from the following
policies in feedlots: 1) using antimicrobials for disease prevention, control, and treatment; 2)
using antimicrobials only for treatment of disease; and 3) not using antimicrobials for any
reason. We modelled a typical U.S. feedlot, where high risk cattle may be afflicted by dis-
eases requiring antimicrobial therapy, namely respiratory diseases, liver abscesses and
lameness. We calculated the net revenue loss under each policy of antimicrobial use restric-
tion. With moderate disease incidence, the median net revenue loss was $66 and $96 per
animal entering the feedlot, for not using antimicrobials for disease prevention and control,
or not using any antimicrobials, respectively, compared to using antimicrobials for disease
prevention, control, and treatment. Losses arose mainly from an increase of fatality and
morbidity rates, almost doubling for respiratory diseases in the case of antimicrobial use
restrictions. In the case of antimicrobial use prohibition, decreasing the feeder cattle price by
9%, or alternatively, increasing the slaughter cattle price by 6.3%, would offset the net reve-
nue losses for the feedlot operator. If no alternatives to antimicrobial therapy for prevention,
control and treatment of current infectious diseases are implemented, policies that economi-
cally incentivize adoption of non-antimicrobial prevention and control strategies for infec-
tious diseases would be necessary to maintain animal welfare and the profitability of beef
production while simultaneously curbing antimicrobial use.
PLOS ONE | https://doi.org/10.1371/journal.pone.0239135 September 15, 2020
1 / 19
PLOS ONE and analysis, decision to publish, or preparation of
the manuscript.
Competing interests: The authors have declared
that no competing interests exist.
Economics of antimicrobial use in US feedlots
Introduction
Cattle production is the most important agricultural industry in the U.S., accounting in 2019
for $68 billion of the $374 billion of U.S. agricultural commodity cash receipts [1]. Approxi-
mately 11 million tonnes of beef are harvested every year, comprising culled dairy and beef
cows and bulls, and feeder calves either finished on grass or in feedlots [2]. Feedlots are con-
centrated feeding operations, where weaned steers and heifers are grouped and finished with
energy-dense rations, for periods of 100 to 300 days, depending on placement weight, and
feeding conditions. In the context of a growing global demand for meat, feedlot operators
must maintain or even improve their productivity. Stress generated by weaning, transport
from cow-calf or backgrounding operations to feedlots, as well as commingling, constitute risk
factors for infectious respiratory diseases, occurring mainly in the first weeks after arrival to
the feedlot. Diseases afflicting cattle diminish the efficiency of the production process [3].
First, diseases may decrease the output sold, by increasing animal losses due to involuntary
culling or mortality. Second, diseases may decrease the efficiency of production factors, leading
for example to an increase in the feeding period, or a decrease in feed conversion. Third, dis-
ease generates additional expenditures for treatments.
Beef feedlot operators commonly use antimicrobials (AM) to limit the impacts of infectious
diseases. In the remainder of the manuscript, we use the term antimicrobials interchangeably
with antibiotics i.e., compounds having an antibacterial activity. Bovine Respiratory Disease
(BRD) is by far the most frequently occurring clinical disease, affecting up to 36% of cattle
placed on feed [4]. Antimicrobials are used routinely for BRD prevention and control [5, 6].
One of the current AMU practices applied to pens of cattle considered at high risk for BRD
consists of treating the entire cohort of animals on arrival with a long-acting AM, to reduce
the incidence or prevent the appearance of clinical signs. Current treatments include macro-
lides, phenicols, tetracyclines, sulfonamides and cephalosporins [7, 8].
Liver abscesses (LA) are localized infections caused by anaerobic bacteria. It is generally rec-
ognized that these infections are consequences of rumenitis and ruminal acidosis, generated
by ramping up of high-grain feeding [9]. The prevalence of LA at slaughter averages from 10
to 20% in most feedlots [10]. The control of LA is based on in-feed use of AM that have been
approved for LA prevention. Currently in the U.S., 6 AM are indicated for the prevention of
LA; namely, bacitracin, chlortetracycline, oxytetracycline, neomycin, tylosin and virginiamy-
cin. Tylosin is by far the most commonly used [11].
Infectious causes of lameness in feedlot cattle include foot rot and arthritis, due to several
bacteria species. Lameness has a negative impact on animal welfare, as it causes pain and
reduces social interactions, and may consequently decrease feed consumption [12]. Data from
western U.S. feedlots showed that lameness accounted for 16% of health problems and 5% of
fatalities [13]. Prevention of lameness is essentially based on good hygiene, feeding and hous-
ing, and includes nutritional supplements such as zinc. Curative parenteral treatments include
AMU: e.g., ceftiofur, oxytetracycline, or florfenicol [14].
The most up to date data on antimicrobial use for treatment, control, and prevention are
limited to irregular NAHMS reporting (last report dated 2011) which tends to ask all-or-none
questions regarding use. Complementary to these, updated FDA reports on annual sales of
antibiotic classes requests sponsors to estimate which livestock species their products are des-
tined for. Ceftiofur and enrofloxacin each were introduced in the late 1980s. Subsequent labels
for the ceftiofur molecule into longer duration formulations were undertaken to extend prod-
uct self-life and to expand labels to include control (metaphylaxis) indications where a single
dose was necessitated for treating large numbers of animals. Tilmicosin and oxytetracycline
(also from pre-1990s) were earlier such formulations. Lately, the longer duration macrolide
PLOS ONE | https://doi.org/10.1371/journal.pone.0239135 September 15, 2020
2 / 19
PLOS ONE Economics of antimicrobial use in US feedlots
products such as tulathromycin and gamithromycin have (anecdotally) supplanted Ceftiofur
Crystalline-Free Acid (CCFA–long duration ceftiofur) for BRD control. Ceftiofur Crystalline-
Free Acid has a 13-day slaughter withholding period. On the other hand, tulathromycin has a
much longer withholding period (and cannot be used in adult dairy cattle). While ceftiofur
remains widely used in dairy production (owing to its zero-day milk withhold) the macrolides
(including tilmicosin) are the majority of BRD control metaphylaxis products used at present
in beef [15]
Quantitative data on AMU in feedlots are sparse; however, aggregate data from the U.S.
Food and Drug Administration (FDA) have shown that dairy and beef cattle accounted for
approximately 50% of non-medically and medically important AM (in tons of active ingredi-
ents) in food animal production [16]. It is noteworthy that macrolides and 3rd generation
cephalosporins, which are used in feedlots, are classified as highest-priority critically important
AM for human medicine by the World Health Organization, being both (i) used to treat infec-
tions caused by bacteria possibly transmitted from non-human sources, or with resistance
genes from non-human source, as well as (ii) sole, or one of limited available therapies, to treat
serious bacterial infections in people [17]. As the efficacy of AM at controlling bacterial dis-
eases is high, their use clearly enhances overall animal productivity; as such, they remain a
widely used tool [18]. However, the increasing evidence of the contribution of AMU in animal
agriculture to the public health threat of antimicrobial resistance (AMR) has emphasized the
paramount importance of prudent AMU in food animal production. Indeed, AMU in cattle
production leads to selection of resistant bacteria from the commensal and pathogenic enteric
microflora of animals, potentially transmitted to humans [19]. To our knowledge, the quanti-
tative impact of AMU in cattle production on AMR in humans has not been measured. Yet
consumers, businesses and public advocacy organizations have demanded a reduction of
AMU in animal agriculture. In an international context where initiatives aiming to curb AMR
are flourishing, it is possible that global policies on AMU will affect U.S. beef production
through trade restrictions. Our objective is therefore to evaluate the economic impact of differ-
ent policies for AMU in feedlots in the U.S., using a partial budgeting method. We compared
three alternative scenarios 1) using AM for disease prevention, control, and treatment (PCT);
2) not using AM for prevention or control (No-PC); and 3) not using AM for any reason (No-
PCT).
Material and methods
We used a simulation model built in Microsoft Excel (Microsoft, Redmond, WA, USA), repre-
sentative of a 100 head pen of a large U.S. feedlot feeding high risk cattle to estimate the aver-
age net costs and benefits of AMU restrictions in beef production using a partial-budgeting
approach. Sensitivity analysis was then performed to determine the impact of various biologic
and economic parameters of the model.
Scenario definitions
Prevention, Control, and Treatment uses of AM scenario (PCT).
In the PCT scenario,
AM are administered to all cattle in the pen at or shortly after arrival, primarily as a strategy
for preventing or controlling bacterial pneumonia as well as liver abscesses. In prevention, AM
are used before the onset of disease in the population, but given its likelihood of apparition,
and under veterinary prescription. In control, AM are given to a group of animals to mitigate
the impact of disease already observed in some animals of the group. In treatment, AM are
used only to treat individuals with clinical signs of diseases. Subsequently, observed clinical
cases of illness are treated individually. This scenario typically occurs under the oversight and
PLOS ONE | https://doi.org/10.1371/journal.pone.0239135 September 15, 2020
3 / 19
PLOS ONE Economics of antimicrobial use in US feedlots
direction of veterinarians because the AM most commonly used are either prescription medi-
cations or require an order from a veterinarian (Veterinary Feed Directive) in the case of AM
delivered in feed. We assumed that antimicrobial treatment was implemented for cases of
observed disease. Using an ionophores medication was still possible in this scenario.
Producing without antimicrobials: No-PCT scenario. Under the No-PCT scenario,
recourse to therapeutic AMU was prohibited for prevention, control or treatment of disease.
Using ionophores was still possible.
Limiting the use of antimicrobials to curative treatments: No-PC scenario. Under the
No-PC scenario, AM are only used to treat animals displaying clinical signs of diseases.
Feedlot model parameters
Production process characteristics of healthy animals.
In each scenario, we modeled a
generic pen of 100 steers at high risk to contract infectious diseases, averaging 295 kg at arrival
and slaughtered at 590 kg (Table 1). The pen was considered as a closed system, with animals
not displaying clinical signs at the beginning of the feeding period. Apart from AMU, other
production methods were kept similar between the different scenarios. In the PCT scenario,
the feeding period was assumed to be 197 days, divided into two successive first (P1) and sec-
ond (P2) periods: a P1 of 37 days and a P2 of 160 days. The average daily gain (ADG) was 0.97
kg/d and 1.63 kg/d in P1 and P2, with an ADG of 1.50 kg/d over the feeding period. Because
the steers did not receive prophylactic AM in the No-PC and No-PCT scenarios, we assumed,
based on estimates found in the literature, that the ADG was lower in those scenarios (P1: 0.90
kg/d; P2: 1.52 kg/d; whole feeding period: 1.40 kg/d) and that the feeding period was longer
(P1: 40 d; P2: 171 d; whole feeding period: 211 d) [20–22].
Consistent with Wileman et al. [23], the ADG of healthy animals that received AM as pro-
phylaxis for BRD was 7.3% greater than the ADG of animals that did not receive prophylaxis.
Daily production costs were calculated according to Lawrence and Ellis [24]. They reported
feeding and operating costs (labor, manure handling, equipment, and interest). We calculated
the feeding costs according to the feeding ration described by Lawrence and Ellis [24] and the
3 year (2016–2018) average prices of corn, alfalfa hay and 50% dry-matter distillers’ grains
[25]. We assumed veterinary costs (for implants, pest control and vaccination) of $19.44 per
animal for the entire feeding period [24, 26, 27]. Transportation costs were also included. Pro-
duction costs are reported in Table 2.
Table 1. Production characteristics for a healthy steer used the model parametrization.
Production characteristics
Entering weight
Slaughter weight
Average ADG, PCT
ADG, P1
ADG, P2
P1 duration
P2 duration
Feeding period duration
PCT
295 kg
590 kg
1.5 kg/d
0.97 kg/d
1.63 kg/d
37 d
159 d
197 d
Scenario
No-PC
295 kg
590 kg
1.4 kg/d
0.9 kg/d
1.52 kg/d
40 d
171 d
211 d
No-PCT
295 kg
590 kg
1.4 kg/d
0.9 kg/d
1.52 kg/d
40 d
171 d
211 d
References
20, 21, 22
ADG: Average Daily Gain; PCT: Prevention, Control, and Treatment; No-PCT: no Prevention, no Control, and no Treatments; No-PC: no Prevention, no Control; P1:
Period 1; P2: Period 2.
https://doi.org/10.1371/journal.pone.0239135.t001
PLOS ONE | https://doi.org/10.1371/journal.pone.0239135 September 15, 2020
4 / 19
PLOS ONE Economics of antimicrobial use in US feedlots
Table 2. Production costs for model parametrization. Antimicrobial costs are applicable only in prevention and control.
Parameter name
Feeder cattle price ($/kg)
Daily Feed cost per head ($)
Daily operating costs ($)
Transportation cost per head($)
Implant per head ($)
Vaccination per head ($)
Antimicrobials per head ($)
Slaughter cattle price ($/kg)
https://doi.org/10.1371/journal.pone.0239135.t002
Mean
3.44
1.63
0.38
13.99
12
7.44
20
2.62
SD
0.3
-
-
-
-
-
-
Min
2.69
1.17
-
-
-
-
-
Max
4.04
2.73
-
-
-
-
-
0.20
2.23
3.00
References
30
24, 25
24, 26, 27
28, 29
20, 30
In the PCT scenario, AM were used in feed or parenterally for prevention and control of
bovine respiratory diseases (BRD) and liver abscesses (LA). Costs of antimicrobial prevention
and control were set at $20 per head [28, 29].
The average feeder cattle price was set at $3.44 per kg. (i.e., as average of the 295 kg-non-fed
steer price from January 2016 to December 2018) [30].
Diseases. Three main categories of diseases were included in our model, each of them
divided into subcategories.
The first category consisted of bovine respiratory disease. We assumed that BRD could be
observed (i.e., steers present clinical signs of disease) during P1 and P2. In addition, BRD
affected some steers subclinically. Liver abscesses comprised the second category. Consistent
with the literature, we assumed that three grades of liver abscesses could be observed at slaugh-
ter: LA- (one or two small abscesses or presence of abscess scars; LA (two to four well-orga-
nized abscesses); and LA+ (one or more large abscesses or multiple small active abscesses) [10,
31]. The third category, lameness, was divided into two subcategories: foot rot (Lame-FR) and
infectious arthritis (Lame-IA).
We assumed an average incidence for each disease and each scenario.
In the PCT scenario, the incidence of disease in the pen was broken into low, moderate,
and high, with an average incidence risk estimated in each category of disease (See Table 3) in
accordance with the mean incidence rates reported in several studies [10, 12, 22, 31–37, 38].
Prophylactic AMU decreased the incidence of diseases against which an AM is adminis-
tered, if pathogens were susceptible to the administered AM. The disease incidence rates in
Table 3. Cumulative incidence estimates of diseases for model parametrization.
Disease
Low incidence (%)
Moderate incidence (%)
High incidence (%)
References
BRD-Cl1
BRD-Cl2
BRD-SubCl
LA-
LA
LA+
Lame-FR
Lame-IA
PCT
6.4
1.6
10.0
4.5
2.3
4.5
0.1
0.1
Others
12.3
3.1
19.2
16.7
8.5
16.7
0.1
0.1
PCT
12.0
3.0
20.0
8.5
6.0
10.0
1.5
0.5
Others
23.1
5.8
38.5
31.5
22.2
37.0
1.5
0.5
PCT
18.4
4.6
30.0
12.0
9.0
22.0
13.0
7.0
Others
35.4
8.8
57.7
44.4
33.3
81.5
13.0
7.0
22, 28, 33, 34, 35, 36, 37
10, 31, 32
12, 13, 38
PCT: Prevention, Control and Treatment scenario; BRD-Cl1: Bovine Respiratory Disease with clinical signs occurring in the period 1; BRD-Cl2: Bovine Respiratory
Disease with clinical signs occurring in the period 2; BRD-SubCl: Bovine Respiratory Disease with subclinical signs; LA-: Liver Abscess, mild intensity, LA: Liver
Abscess, moderate intensity; LA+: Liver Abscess, severe intensity; Lame-FR: Footrot; Lame-IA: Infectious Arthritis.
https://doi.org/10.1371/journal.pone.0239135.t003
PLOS ONE | https://doi.org/10.1371/journal.pone.0239135 September 15, 2020
5 / 19
PLOS ONE Economics of antimicrobial use in US feedlots
No-PC and No-PCT were similar, and were calculated with the following formula, where
RRMORB was the relative risk of morbidity when AM are used in prophylaxis:
INCIDENCENo(cid:0) PC ¼ INCIDENCEPCT=RRMORB:
The relative risks of morbidity were equal to 0.52 [39] for respiratory diseases and 0.27 [23]
for LA. We did not find published literature concerning the efficacy of prophylactic therapy
against foot rot and infectious arthritis. Thus, we set these relative risks to 1.
Evaluation of the impacts of infectious diseases
The cumulated impact of each disease was estimated per sick animal, according to the follow-
ing components: realizer rate, fatality rate, number of days on feed until slaughter (DOF), qual-
ity grade (QG) (% of sick animals having a grade loss), treatment costs ($), and labor costs ($).
Realizers were cattle that failed to respond to treatment. The realizer rate is the proportion of
cattle marketed early because of morbid condition. Case fatality rate consists of the proportion
of sick animals that died from the disease. We assumed that diseases were independent from
each other, and that they could occur only once.
For the PCT scenario, estimates of impacts provided by previous research were generally
extracted from studies in which AM were used to manage infectious diseases [22, 23, 31, 32,
34–36, 39–43]. Therefore, we used an average value from the literature for each cost impact
component, assuming that AM were used to achieve these values.
For the No-PC scenario, we calculated the realizer rate and the case fatality rate using the
relative risks (RR) or odds ratios found in the literature [39] according to the following formu-
las:
Realizer rateNo(cid:0) PC ¼ RRrealizer rate � realizer ratePCT
Case fatality rateNo(cid:0) PC ¼ RRfatality rate � case fatality ratePCT
These relative risks were set to one for subclinical diseases and when no data were found in
the literature.
In the No-PCT scenario, for each disease, we set intermediate values for case fatality rate
and realizer rate, between those used in the No-PC scenario and the highest estimates found in
the literature.
We set an average value of ADG in each scenario, and allowed ADG to vary in a range
of ± 15% of the average in the model. We assumed that the impacts of each disease on ADG
were consistent between the PCT and No-PC scenarios. In the No-PCT scenario, for each dis-
ease, the disease impact on ADG was intermediate between the value used in the No-PC sce-
nario and the highest value found in the literature. Consistent with the literature, the loss of
ADG in the case of BRD was applied to the entire duration of the feeding phase concerned [32,
34, 35]. In the case of liver abscesses, the loss of ADG was applied from day 0. In the case of
lameness, the loss of ADG (i.e., foot rot and infectious arthritis) was applied from the onset of
the disease [37]. The average day of onset of disease, day of death and day of realizer slaughter
are reported in Table 4.
For modeling purposes, we held constant for healthy cattle the slaughter weight of 590 kg,
and therefore assumed that diseases would decrease the ADG, and consequently increased the
number of days on feed. Thus, the impact of diseases on ADG was converted and expressed as
additional days on feed (Extra-DOF).
The slaughter cattle selling price depends on Yield and Quality Grades of the carcass. The
base price is generally defined for the Quality Grade “Choice” and the Yield Grade “3”. In our
PLOS ONE | https://doi.org/10.1371/journal.pone.0239135 September 15, 2020
6 / 19
PLOS ONE Economics of antimicrobial use in US feedlots
Table 4. Average day of onset, day of death and day of anticipated slaughter, for each infectious disease.
Disease
BRD-Cl1
BRD-Cl2
Lame-FR
Lame-IA
Onset of disease
(days)
Average day of death after
entering feedlot
Average day of realizer slaughter after
entering feedlot
20
83
85
32
50
113
101
48
50
113
118
65
BRD-Cl1: Bovine Respiratory Disease with clinical signs occurring in Period 1; BRD-Cl2: Bovine Respiratory Disease
with clinical signs occurring in Period 2; Lame-FR: Foot rot; Lame-IA: Infectious Arthritis.
https://doi.org/10.1371/journal.pone.0239135.t004
model, we assumed that healthy animals reached this level. Diseases can affect the grades of a
certain percentage of animals that are slaughtered at a regular slaughter weight. We assumed
that 10% of steers presenting clinical diseases and 2% of steers affected by subclinical diseases
had a loss of quality grade [33, 44]. As several studies have reported that diseases did not
impact the Yield Grade enough to induce a penalty, we did not include the Yield Grade in the
model [31, 33, 44, 45].
Additional treatment costs and additional labor costs per head were set at constant values
for each scenario, and their total costs varied only according to the incidence of disease. We
assumed that all sick animals were treated in the No-PC and PCT scenarios. Only treated ani-
mals required additional labor.
For simplification purposes, each disease was modeled at the individual level as only occur-
ring once and each was considered independent from the other diseases. All parameter esti-
mates extracted from available literature are shown in Table 5.
Costs and revenues calculation
Cost and revenue estimates. The impacts were estimated for a pen of 100 steers with
average incidence rates of disease.
First, an average cost was estimated for each cost component. The calculation method of
each cost component was similar for each disease.
Table 5. Default disease impacts for model parametrization.
Impact of disease
Disease
Extra-DOF (d)
Loss-QG (% of
animals)
Fatality rate (%)
Realizer rate (%)
Cost of treatment / sick
animal ($)
Labor cost / sick
animal ($)
PCT No-PC No-PCT PCT No-PC No-PCT PCT No-PC No-PCT PCT No-PC No-PCT PCT No-PC No-PCT PCT No-PC No-PCT
BRD-Cl1
BRD-Cl2
3
0
BRD-SubCl 10
LA-
LA
LA+
Lame-FR
4
11
19
2
3
0
11
4
12
21
3
6
0
14
7
16
23
13
Lame-IA
165
179
179
10
10
2
2
2
2
10
10
10
10
2
2
2
2
10
10
10
10
2
2
2
2
10
10
4.7
4.7
7.6
7.6
14.7
14.7
0
0
0
0
0
0
0
0
0
0
0
0
5
5
0
0
0
0
5
5
0
0
0
0
5
5
0
0
0
0
7.2
22
7.2
22
12.1
30.8
3.1
4.5
3.1
4.5
4.2
5.8
20
20
0
0
0
0
12
17
20
20
0
0
0
0
12
17
0
0
0
0
0
0
0
0
1.18
1.18
1.18
1.18
0
0
0
0
2
2
0
0
0
0
2
2
0
0
0
0
0
0
0
0
Extra-DOF: Additional Days On Feed; Loss-QG: Loss of Quality Grade; BRD-Cl1: Bovine Respiratory Disease with clinical signs occurring in Period 1; BRD-Cl2:
Bovine Respiratory Disease with clinical signs occurring in Period 2; BRD-SubCl: Bovine Respiratory Disease with subclinical signs; LA-: Liver Abscess, mild intensity,
LA: Liver Abscess, moderate intensity; LA+: Liver Abscess, severe intensity; Lame-FR: Foot rot; Lame-IA: Infectious Arthritis. PCT: Prevention Control and Treatment;
No-PC: no Prevention, no Control; No-PCT: no Prevention, no Control and no Treatment.
https://doi.org/10.1371/journal.pone.0239135.t005
PLOS ONE | https://doi.org/10.1371/journal.pone.0239135 September 15, 2020
7 / 19
PLOS ONE Economics of antimicrobial use in US feedlots
Lost revenues. The lost revenue of a dead animal was set as the gross revenue of a healthy
animal at slaughter. The lost revenue of a realizer was calculated by subtracting the gross reve-
nue of a realizer from the gross revenue of a healthy steer at slaughter.
The gross revenue was calculated as the product of weight (kg) at slaughter times the selling
price/kg. To calculate the gross revenue, we used an average slaughter price of $2.62 /kg, corre-
sponding to a 3-year average of the selling price of live-fed animals [20, 30].The selling price of
live realizer was equal to 53% of the price of animals that reach the optimal slaughter weight
(590 kg) [13, 37].
The loss of Quality Grade, from Choice to Select or Standard for example, induces about a
10% price penalty [46]. We used this value to calculate the impact of loss of Quality Grade due
to disease incidence.
Additional costs and cost adjustments. We calculated the daily production costs (sum of feed
and daily operating costs) for a healthy animal. We calculated the costs of treatment and labor
costs for a sick animal, and multiplied the values by the number of sick animals under each
scenario and incidence (low, moderate, high) rate. For each clinical disease, we set an average
day of death or realizer slaughter according to published data [12, 36], and calculated the num-
ber of days for which a steer should have been fed until slaughter (Table 3). The weight of real-
izer and the number of days on feed were calculated. Additional days on feed only affected sick
animals that recovered and were slaughtered at optimal weight (590 kg). We then multiplied
this number by the daily production costs, to obtain the additional (or lower) costs of finishing
a previously diseased animal, assuming the animal recovered sufficiently from the disease
event in order to gain optimal weight.
For each scenario for the three incidence rates, the net revenue was calculated at the pen
level by subtracting the costs and the lost revenue from the gross revenue.
Scenarios analysis. We calculated the net impact by subtracting the net revenues of the
No-PC and No-PCT scenarios from the net revenue of the PCT scenario.
For the prices of purchase and sale of steers, we calculated the variations in feeder cattle
prices and slaughter prices, all other things being equal, required for the feedlot to be indiffer-
ent between the PCT scenario and the alternative scenarios. To do so, we modelled the net rev-
enue as a function of the feeder price or the slaughter cattle price, for each scenario and each
level of disease incidence. Then, we calculated required feeder and slaughter cattle prices to
equalize the net revenue in each scenario.
Sensitivity analysis. Sensitivity analyses were performed using the @Risk (Palisade Cor-
poration, Ithaca, NY, USA) Excel add-in, to determine the influence of stochastic and deter-
ministic input parameters on outcome values. Input components (feed costs, relative risks,
AM effects and costs, and feeder cattle prices) and output components (slaughter price) were
modeled stochastically. The results are presented in a regression tornado diagram, which
depicts the change in the dependent variable when the independent variable increases by one
SD with all other variables held constant. The @Risk Excel add-in runs a multiple regression
analysis for each iteration with the dependent variable of interest (the net impact) and the sim-
ulated values of each stochastic independent variables [47].
Results
Net impact under No-PCT and No-PC scenarios over PCT scenario
Estimates of net impact under the No-PCT scenario and under the No-PC scenario are pre-
sented in Fig 1. In a situation of moderate incidence of diseases, the median net impact per
steer entering the feedlot over the PCT scenario was $66 (25th percentile: 61, 75th percentile:
PLOS ONE | https://doi.org/10.1371/journal.pone.0239135 September 15, 2020
8 / 19
PLOS ONE Economics of antimicrobial use in US feedlots
Fig 1. Net impact, per steer present in the feedlot at day 0, for the No-PC and No-PCT scenarios, compared with the PCT scenario, as a
function of the level of disease incidence. The grey dots represent the median value, written on the right side, the red diamonds represent the mean,
and the bars represent the 25th and 75th percentiles values. PCT: Prevention Control and Treatment; No-PC: no Prevention, no Control; No-PCT: no
Prevention, no Control and no Treatment.
https://doi.org/10.1371/journal.pone.0239135.g001
73) and $96 (25th percentile: 89, 75th percentile: 104), in the No-PC and No-PCT scenarios,
respectively. The net impact increased with the incidence of diseases.
Sensitivity analysis showed that the net impact was most influenced by feed costs, followed
by the relative risk of morbidity when AM are used in prevention and control (RRMORB),
slaughter cattle price and extra days on feed, when comparing the No-PC over the PCT sce-
nario (Fig 2A). The net impact was most influenced by feed costs, followed by slaughter cattle
price, the relative risk of morbidity when AM are used in prevention and control (RRMORB),
and extra days on feed, when comparing the No-PCT over the PCT scenario (Fig 2B). Regard-
less of the scenario, cattle prices and feed costs were the two most important factors influenc-
ing the net impact that were not under the direct control of feedlot operators.
Additional costs and lost revenue imputable to diseases under each
scenario
Costs and lost revenue under the PCT scenario, the No-PCT scenario and under the No-PC
scenario are presented in Fig 3. In a situation of moderate incidence of diseases, the sum of
additional costs and lost revenue per steer were $42, $102, $137 in the PCT, No-PCT and No-
PLOS ONE | https://doi.org/10.1371/journal.pone.0239135 September 15, 2020
9 / 19
PLOS ONE Economics of antimicrobial use in US feedlots
Fig 2. Tornado diagram for factors influencing the net impact (median value) in moderate incidence. Fig 2A: No-PC scenario compared to PCT scenario; Fig 2B:
No-PCT scenario compared to PCT scenario. Solid (positive) or dashed (negative) bars indicates the sign of coefficient For example, a positive coefficient e.g., Feed costs
with solid bar extending to the right, indicates that this input has a positive impact: increasing this input will increase the net impact. The values (in bold) are median
values from Fig 1. PCT: Prevention Control and Treatment; No-PC: no Prevention, no Control; No-PCT: no Prevention, no Control and no Treatment; BRD / RR
MORB: relative risk of Bovine Respiratory Disease morbidity when antimicrobials are not used for prevention. LA / RR MORB: relative risk of liver abscesses morbidity
when antimicrobials are not used for prevention; BRD / RR FAT: relative risk of Bovine Respiratory Disease fatality when antimicrobials are not used for prevention;
Extra-DOF: Additional Days On Feed.
https://doi.org/10.1371/journal.pone.0239135.g002
PC scenarios, respectively. Breaking down additional costs and lost revenues per parameter
shows that mortality and additional days on feed were the largest cost components (Fig 3).
Indifference between scenarios
To produce equal net revenue under the No-PC scenario or No-PCT scenarios compared to
the PCT scenario, feedlot operators would have to purchase steers between 3.5% to 9.7%, and
4.9% to 15.4% respectively, lower than the average purchase price, depending on the incidence
of diseases. Importantly, these lower prices directly affect primary cow-calf producers
upstream in the production system. Alternatively, if feedlot steers could sell between 2.3% to
7%, and 3.3% to 11.6% higher than the average slaughter price, respectively, they would be
indifferent to a change of scenario. The results are depicted in Table 6.
Discussion
The objective of our study was to evaluate the economic impact of different policies for AMU
in feedlots in the U.S., using a partial budgeting method calibrated with data from U.S. and
Canadian feedlots. We evaluated the impact of restrictions regarding therapeutic uses of anti-
microbials only. We compared each of our alternative scenarios with a baseline scenario
(PCT), mimicking the current practices of antimicrobial therapy. In addition, we incorporated
in our model the major diseases reported in feedlot operations. We evaluated such impact for
cattle placed in large feedlots (>1,000 head capacity) in pens of 100 heads, as they represented
81% of the cattle placed in the U.S. in 2019 [48]. We focused on high risk cattle, as the adoption
of programs aiming to lower diseases’ risk, such as preconditioning programs, is still low in
the U.S. [49, 50].
In the PCT scenario and with moderate disease incidence, the average cost of a clinical case
of BRD was $138 and the net revenue obtained from a sick steer was $-58. Our results are con-
sistent with those of Brooks et al. [51], who found that the loss of net revenue per sick animal
ranged from $60 to $143 and the net revenue provided by a sick animal ranged from $-61 to
$-78. Poulsen Nautrup et al. [52] estimated the per case cost of BRD between $28 and $307
when prophylactic AMU was not administered.
PLOS ONE | https://doi.org/10.1371/journal.pone.0239135 September 15, 2020
10 / 19
PLOS ONE Economics of antimicrobial use in US feedlots
PLOS ONE | https://doi.org/10.1371/journal.pone.0239135 September 15, 2020
11 / 19
PLOS ONE Economics of antimicrobial use in US feedlots
Fig 3. Additional mean costs and lost revenues per steer present in the feedlot at day 0, for the PCT, No-PC and
No-PCT scenarios. Mean estimates were used to parameterize the model for each level of incidence. PCT: Prevention
Control and Treatment; No-PC: no Prevention, no Control; No-PCT: no Prevention, no Control and no Treatment;
Extra-DOF: Additional Days On Feed; Loss-QG: Loss of Quality Grade.
https://doi.org/10.1371/journal.pone.0239135.g003
In our model, the loss of net revenue induced by LA was $10, $25, and $41 per steer enter-
ing the feedlot for the carcass grades LA-, LA and LA+, respectively (see S1 Material). These
consequences of LA are a sales price discount [31]. Data regarding the effects of LA are less fre-
quent than for BRD. Though animals do not show clinical signs of LA unless severely affected,
it has been previously reported that cattle presenting abscessed livers had a reduced ADG (up
to 11%) and a lower feed efficiency [40]. This unavoidably increases the fattening duration
and/or decreases the slaughter weight and devalues the quality grade [31].
To our knowledge, only Davis-Unger et al. [37] have calculated the economic impact of the
different causes of lameness in beef feedlots. They calculated that each foot rot case cost $110
and each infectious arthritis case cost $727. The costs were due to decreased ADG, weight loss,
and additional treatment costs. These results are consistent with our estimates ($152 for foot
rot and $585 for infectious arthritis).
In the PCT scenario, the net revenue per steer entering the feedlot was $42. We observed
that our results were sensitive to feed costs and sales prices. Time-series data show that indeed,
this revenue from January 2016 to October 2018 varied from $-500 to $350 in Kansas and aver-
aged $-40 [53].
While discussions regarding the efficacy and economics of preventing and controlling dis-
eases such as liver abscesses and bovine respiratory disease focus on numeric comparisons, it
is clear that there are stark differences in the perceived moral imperatives to use antibiotics for
each indication [54]. Earlier work has shown a varying sense among feedlot veterinarians of
the moral duty to use antibiotics to treat, control, and prevent disease. These differences grow
when comparing to feedlot operators and to the general public. The immediacy of the need
and the predictable differences in threats to animal welfare and mortality while preventing and
controlling respiratory disease far outweigh perceptions of the importance of using medically
important classes of antibiotic to prevent and control liver abscesses, which rarely exhibit ani-
mal welfare concerns.
Though AMU in animal agriculture is increasingly scrutinized, data reporting the eco-
nomic implications of policies aiming at curbing AMU remain uncommon. In the beef, poul-
try and pig sectors, researchers have investigated the impact of banning growth promotion
uses of medically important antimicrobials and concluded that such policies would have a
minor to moderate impact [55–58]. It is noteworthy that since then, the U.S. government has
Table 6. Variations in feeder cattle prices and slaughter prices required for indifference between PCT, No-PCT and No-PC scenarios.
Incidence level
Low
Mean
No-PC scenario
Moderate
Mean
High
Mean
Low
Mean
No-PCT scenario
Moderate
Mean
High
Mean
[min-max]
[min-max]
[min-max]
[min-max]
[min-max]
[min-max]
Feeder price (Δ% from PCT)
-3.5
-6.1
-9.7
-4.9
-9.0
-15.4
[-17.8 –+23.5]
[-20.0 - +20.1]
[-23.0 - +15.5]
[-18.9 - +21.7]
[-22.4 - +16.5]
[-27.9 - +8.3]
Slaughter cattle price (Δ% from PCT)
2.3
4.2
7.0
3.3
6.3
11.6
[-10.8 - +20.0]
[-9.1 - +22.2]
[-6.7 - +25.5]
[-9.9 - +21.2]
[-7.3 - +24.7]
[-26.4 - +31.0]
PCT: Prevention Control and Treatment; No-PC: no Prevention, no Control; No-PCT: no Prevention, no Control and no Treatment.
https://doi.org/10.1371/journal.pone.0239135.t006
PLOS ONE | https://doi.org/10.1371/journal.pone.0239135 September 15, 2020
12 / 19
PLOS ONE Economics of antimicrobial use in US feedlots
fully implemented Guidance for Industry #209 and #213, banning the use of medically impor-
tant AM as growth promoters in food animal production [59, 60]. In dairy production, Lher-
mie et al. estimated the average cost of entirely removing AM in the U.S. at $61 per cow,
leading to a loss of $152 million for the U.S. dairy sector per annum [61, 62]. In the United
States, there has been little movement towards banning various uses of antibiotic classes
deemed to not be of importance to human medicine (e.g., ionophores, bacitracin, bambermy-
cins). While the list derived from the FDA Guidance for Industry 152 [63] differs in modest
ways from the list published and updated regularly by the WHO [64], the policy differences
between, for example Europe, and the United States remain quite entrenched. Changes that
occurred in the United States due to FDA GFI 209/213 mostly affected use of medically impor-
tant antibiotics for growth promotion. Little activity has occurred since in changing, for exam-
ple, labels that allow for continuous feeding with no defined duration. This seems to be the
most likely next target in the United States and it is likely that products such as tylosin used to
prevent/control liver abscesses would be required to be relabeled to better reflect the actual
periods of risk for the development of abscessation.
In our feedlot model, removing prevention and control treatments induced a loss of $62
per steer entering the feedlot, under the scenario of moderate disease incidence. Our results
are consistent with those from Dennis et al. [42], who report that the use of “upper tier” AM
such as macrolides in prophylaxis allowed farmers to achieve a net revenue from $58 to $119
higher than if no prophylaxis was performed. In a meta-analysis, Abell et al. (2017), classified 8
AM used in prevention and control of BRD in lesser, middle, and upper tiers, in function of
their odd-ratio, with “upper tier” AM being the most effective to decrease morbidity and mor-
tality [7]. This additional net revenue ranged from $14 to $42 when “lower tier” AM such as
sulfonamides or phenicols were used [42]. Prophylaxis was valued at 0.96% and 1.17% of the
industry gross revenue, depending on the dataset. In a meta-analysis comparing conventional
and nonconventional beef production, Wileman et al. [23] also reported that metaphylaxis
enabled higher growth performance (0.11 kg/d of additional ADG) for the cattle treated. It is
noteworthy that the existence of compensatory gain may moderate the ADG losses experi-
enced during the clinical phase of disease, which are unlikely to be sustained through slaugh-
ter. We did not account for potential compensatory gains in our model.
Removing any kind of AMU led to negative revenue with moderate and high disease inci-
dence under the assumptions of our model. When incidence rates were low, the net revenues
were $27 and $12 per animal entering the feedlot, in the No-PC and No-PCT scenarios,
respectively. This indicates that it seems feasible to achieve a positive net revenue even if dras-
tic measures are taken concerning AMU in feedlots, but only if the incidence rates of major
infectious diseases are low. This would require a large-scale adoption of non-medical and med-
ical disease control strategies, such as preconditioning and changes in rations and feeding
practices, as well as structural changes in the beef market [38, 65–67]. The effects of such man-
agement and structural changes were beyond the scope of our study and have not been incor-
porated in our model, but they may be worth investigating at the feedlot and supply chain
levels.
One limitation of our model is that it does not account for animals dying from diseases
without having been diagnosed, which may lead to an underestimation of the net impact. Our
estimates do not include any type of aggregate market adjustment changes due to restrictions
where lower production or higher costs might lead to higher finished beef prices. Our model
was designed to provide information regarding short-run consequences of policies. This
enables us to identify in detail which factors influence the loss of revenue, but does not model
or estimate intermediate or long term impacts of policies, including the impact on other mar-
kets. Johnson et al. [68] estimated that reducing the prevalence of BRD would lead to an
PLOS ONE | https://doi.org/10.1371/journal.pone.0239135 September 15, 2020
13 / 19
PLOS ONE Economics of antimicrobial use in US feedlots
increased beef supply, and consequently losses of $2,904 million over 4 years for the beef
industry [68].
Another limitation of our model is that we assumed that diseases were independent of one
another. This hypothesis simplifies a very complex issue. Importantly, some diseases, such as
BRD, LA and infectious arthritis, may be epidemiologically linked, and treatments may help
control multiple diseases simultaneously [37, 69]. This may be associated with changes in ani-
mals’ immune status, as suggested by Duff and Galyean [70].
To assess how cow-calf producers might be affected in the long run, we calculated the
decrease in purchase price (feeder cattle price) required to counterbalance the loss of net reve-
nue in the No-PC and No-PCT scenarios. This presumes that those prices need to occur to
maintain profitability in the feedlot. A decreased purchase price of 6.1% and 9%, for the No-
PC and the No-PCT scenario respectively, allows the feedlot to maintain the revenue obtained
in the PCT scenario. This would result in lower profits to beef cow producers who may
respond by producing fewer calves resulting in higher feeder cattle prices. Another alternative
for feedlot operators to maintain profitability would be to experience an increase in selling
prices. Under moderate disease incidence rates, an increase in selling price of 4.2% and 6.3%
in the No-PC and No-PCT scenario, respectively, enables the feedlot to maintain nearly the
same expected revenue as in the PCT scenario. These increases in selling price is less than the
average premium paid for organic products [71] and lower than twice the price of conven-
tional beef that American consumers appear ready to spend for beef meat raised without AM
and hormones, under supply and demand conditions existing at the time of the surveys. [72,
73].
Prohibition of AMU raises concerns of animal welfare. It is very likely that prohibition
would worsen animal welfare on farms, as sick animals would remain untreated, culled or
euthanized. Sixty-five percent of the conventional U.S. animal producers consider that raising
animals without AM is a threat to animal welfare [74]. Animal welfare evaluation remains
highly complex, as it takes multiple dimensions, and assessing the impacts of not using AM is
needed prior to implementing restrictions. Antimicrobials have been proven effective to con-
trol infectious diseases, and remain a major tool in the therapeutic arsenal. Currently, few
alternatives have emerged [75], but focusing on infection prevention and control, such as bio-
security measures, may help curb the need for AM.
In conclusion, our results suggest that the current beef feedlot sector would incur revenue
losses, estimated between $43 and $139 per steer, in the case of restrictions on AMU. Policies
should be designed to incentivize farmers to adopt non-AM preventive measures aimed at
decreasing the need for AMU. Research for alternative methods to control bacterial diseases
would also be valuable.
Supporting information
S1 Material.
(DOCX)
Acknowledgments
The authors are grateful to Julia Hertl (Cornell University) for valuable advice on the
manuscript.
Author Contributions
Conceptualization: Guillaume Lhermie.
PLOS ONE | https://doi.org/10.1371/journal.pone.0239135 September 15, 2020
14 / 19
PLOS ONE Economics of antimicrobial use in US feedlots
Data curation: Pierre Sauvage.
Formal analysis: Pierre Sauvage.
Funding acquisition: Harvey Morgan Scott, Yrjo Tapio Grohn.
Investigation: Guillaume Lhermie, Pierre Sauvage, Harvey Morgan Scott, Yrjo Tapio Grohn.
Methodology: Guillaume Lhermie, Pierre Sauvage, Loren William Tauer, Didier Raboisson,
Harvey Morgan Scott, Yrjo Tapio Grohn.
Project administration: Guillaume Lhermie.
Resources: Leslie Verteramo Chiu, Karun Kanyiamattam, Ahmed Ferchiou, Didier Raboisson,
David R. Smith.
Software: Guillaume Lhermie, Ahmed Ferchiou.
Supervision: Loren William Tauer, David R. Smith, Yrjo Tapio Grohn.
Validation: Guillaume Lhermie, Loren William Tauer, Leslie Verteramo Chiu, Karun Kanyia-
mattam, Ahmed Ferchiou, Didier Raboisson, David R. Smith, Yrjo Tapio Grohn.
Writing – original draft: Guillaume Lhermie.
Writing – review & editing: Guillaume Lhermie, Pierre Sauvage, Loren William Tauer, Leslie
Verteramo Chiu, Karun Kanyiamattam, Ahmed Ferchiou, Didier Raboisson, Harvey Mor-
gan Scott, David R. Smith, Yrjo Tapio Grohn.
References
1. USDA ERS. Data Products / Farm Income and Wealth Statistics / Annual cash receipts by commodity
[Internet]. 2020. Available from: https://data.ers.usda.gov/reports.aspx?ID=17832
2. USDA:NASS. Beef, Slaughter—Production. Quick Stats. 2018.
3. Mclnerney J. Old Economics for New Problems -Livestock Disease: Presidential Address. J Agric Econ
[Internet]. 1996; 47(1–4):295–314. Available from: http://doi.wiley.com/10.1111/j.1477-9552.1996.
tb00695.x
4. United States Department of Agriculture. Beef 2007–2008. Part I: Reference of beef cow-calf manage-
ment practices in the United States, 2007–2008. USDA Reports. 2008;(February):1–89.
5. O’Connor AM, Coetzee JF, da Silva N, Wang C. A mixed treatment comparison meta-analysis of antibi-
otic treatments for bovine respiratory disease. Prev Vet Med [Internet]. 2013; 110(2):77–87. Available
from: https://doi.org/10.1016/j.prevetmed.2012.11.025 PMID: 23276402
6. DeDonder KD, Apley MD. A Review of the Expected Effects of Antimicrobials in Bovine Respiratory Dis-
ease Treatment and Control Using Outcomes from Published Randomized Clinical Trials with Negative
Controls. Vet Clin North Am—Food Anim Pract [Internet]. 2015; 31(1):97–111. Available from: https://
doi.org/10.1016/j.cvfa.2014.11.003 PMID: 25578389
7. Abell KM, Theurer ME, Larson RL, White BJ, Apley M. A mixed treatment comparison meta-analysis of
metaphylaxis treatments for bovine respiratory disease in beef cattle 1, 2. 2017;(2015).
8. USDA:APHIS:VS. Antimicrobial Use andStewardship on U.S. Feedlots, 2017 [Internet]. 2019. Available
from: https://www.aphis.usda.gov/animal_health/nahms/downloads/amu-feedlots.pdf
9. USDA. Feedlot 2011 Part I: Management Practices on U.S. Feedlots with a Capacity of 1,000 or More
Head. 2013;
10. Amachawadi RG, Nagaraja TG. Liver abscesses in cattle: A review of incidence in Holsteins and of bac-
teriology and vaccine approaches to control in feedlot cattle. J Anim Sci [Internet]. 2016; 94(4):1620–
32. Available from: http://www.ncbi.nlm.nih.gov/pubmed/27136021 https://doi.org/10.2527/jas.2015-
0261 PMID: 27136021
11. Cazer CL, Eldermire ERB, Lhermie G, Murray SA, Scott HM, Gro¨ hn YT. The effect of tylosin on antimi-
crobial resistance in beef cattle enteric bacteria: A systematic review and meta-analysis. Prev Vet Med.
2020 Mar; 176:104934. https://doi.org/10.1016/j.prevetmed.2020.104934 PMID: 32109782
PLOS ONE | https://doi.org/10.1371/journal.pone.0239135 September 15, 2020
15 / 19
PLOS ONE Economics of antimicrobial use in US feedlots
12.
Terrell SP, Reinhardt CD, Larson CK, Vahl CI, Thomson DU. Incidence of lameness and association of
cause and severity of lameness on the outcome for cattle on six commercial beef feedlots. J Am Vet
Med Assoc. 2017; 250(4):437–45. https://doi.org/10.2460/javma.250.4.437 PMID: 28165312
13. Griffin D, Perino L, Hudson D, Griffin N-L, Perino D;, Hudson L. G93-1159 Feedlot Lameness Feedlot
Lameness. 1993; Available from: http://digitalcommons.unl.edu/extensionhist%5Cnhttp://
digitalcommons.unl.edu/extensionhist/196
14. Stokka GL, Lechtenberg K, Edwards T, MacGregor S, Voss K, Griffin D, et al. Lameness in feedlot cat-
tle. Vet Clin North Am Food Anim Pract [Internet]. 2001; 17(1):189–207. Available from: https://doi.org/
10.1016/s0749-0720(15)30062-1 PMID: 11320695
15. NAHMS. Feedlot 2011 Part IV: Health and health managment on U.S. feedlots with a capacity of 1,000
or more head. Usda, Aphis. 2013;Pg.:27–32.
16. US Food and Drug Administration. 2018 Summary Report on Antimicrobials Sold or Distributed for Use
in Food-Producing Animals [Internet]. 2019. Available from: https://www.fda.gov/animal-veterinary/
cvm-updates/fda-releases-annual-summary-report-antimicrobials-sold-or-distributed-2018-use-food-
producing
17. WHO. List of Critically Important Antimicrobials for Human Medicine. World Health Organization. 2017.
5th revision.
18.
19.
Lhermie G, Gro¨ hn YT, Raboisson D. Addressing Antimicrobial Resistance: An Overview of Priority
Actions to Prevent Suboptimal Antimicrobial Use in Food-Animal Production. Front Microbiol [Internet].
2017; 7(January):1–11. Available from: http://journal.frontiersin.org/article/10.3389/fmicb.2016.02114/
full
Tang KL, Caffrey NP, Nobrega DB, Cork SC, Ronksley PE, Barkema HW, et al. Restricting the use of
antibiotics in food-producing animals and its associations with antibiotic resistance in food-producing
animals and human beings: a systematic review and meta-analysis. Lancet Planet Heal. 2017 Nov; 1
(8):e316–27.
20. USDA:NASS. Cattle on Feed [Internet]. surveys. 2019. Available from: https://www.nass.usda.gov/
Surveys/Guide_to_NASS_Surveys/Cattle_On_Feed/index.php
21. USDA. Overview of the United States Cattle Industry. Economics, Statistics and Market Information
System. 2016.
22.
Thompson PN, Stone A, Schultheiss WA. Use of treatment records and lung lesion scoring to estimate
the effect of respiratory disease on growth during early and late finishing periods in South African feedlot
cattle The online version of this article, along with updated information and servic. J Anim Sci. 2006;
(1999):488–98.
23. Wileman BW, Thomson DU, Reinhardt CD, Renter DG. Analysis of modern technologies commonly
used in beef cattle production: Conventional beef production versus nonconventional production using
meta-analysis. J Anim Sci. 2009; 87(10):3418–26. https://doi.org/10.2527/jas.2009-1778 PMID:
19617517
24.
Lawrence J, Ellis S. Monthly Returns from Cattle Feeding [Internet]. 2008. p. 1–3. Available from:
https://www.extension.iastate.edu/agdm/livestock/html/b1-35.html
25. USDA ERS. Feed Grains: Yearbook Tables [Internet]. Prices and Profitability Models | Agricultural Mar-
keting Resource Center. 2019. Available from: https://www.ers.usda.gov/data-products/feed-grains-
database/feed-grains-yearbook-tables/
26. Wildman BK, Perrett T, Abutarbush SM, Guichon PT, Pittman TJ, Booker CW, et al. A comparison of 2
vaccination programs in feedlot calves at ultra-high risk of developing undifferentiated fever/bovine
respiratory disease. Can Vet J [Internet]. 2008; 49(5):463–72. Available from: http://ovidsp.ovid.com/
ovidweb.cgi?T=JS&PAGE=reference&D=emed10&NEWS=N&AN=351771895 PMID: 18512457
27. Wang M, Schneider LG, Hubbard KJ, Grotelueschen DM, Daly RF, Stokka GS, et al. Beef producer sur-
vey of the cost to prevent and treat bovine respiratory disease in preweaned calves. J Am Vet Med
Assoc. 2018; 253(5):617–23. https://doi.org/10.2460/javma.253.5.617 PMID: 30110213
28. Ball JJ, Kegley EB, Sarchet J, Powell JG. Comparison of treatment protocols for bovine respiratory dis-
ease in high-risk, newly received beef calves. Appl Anim Sci [Internet]. 2019; 35(3):278–83. Available
from: http://dx.doi.org/10.15232/aas.2018-01836
29. Nickell JS, White BJ. Metaphylactic antimicrobial therapy for bovine respiratory disease in stocker and
feedlot cattle. Vet Clin North Am—Food Anim Pract [Internet]. 2010; 26(2):285–301. Available from:
https://doi.org/10.1016/j.cvfa.2010.04.006 PMID: 20619185
30. USDA ERS. Livestock & Meat Domestic Data [Internet]. red meat and poultry production. 2019. Avail-
able from: https://www.ers.usda.gov/data-products/livestock-meat-domestic-data/
31. Brown TR, Lawrence TE. Association of liver abnormalities with carcass grading performance and
value. J Anim Sci. 2010; 88(12):4037–43. https://doi.org/10.2527/jas.2010-3219 PMID: 20817855
PLOS ONE | https://doi.org/10.1371/journal.pone.0239135 September 15, 2020
16 / 19
PLOS ONE Economics of antimicrobial use in US feedlots
32. Babcock AH, White BJ, Dritz SS, Thomson DU, Renter DG. Feedlot health and performance effects
associated with the timing of respiratory disease treatment. J Anim Sci. 2009; 87(1):314–27. https://doi.
org/10.2527/jas.2008-1201 PMID: 18765846
33. Schneider MJ, Tait RG, Busby WD, Reecy JM. An evaluation of bovine respiratory disease complex in
feedlot cattle: Impact on performance and carcass traits using treatment records and lung lesion scores.
J Anim Sci. 2009; 87(5):1821–7. https://doi.org/10.2527/jas.2008-1283 PMID: 19181770
34.
35.
Faber R, Hartwig N, Busby D, BreDahl R. The Costs and Predictive Factors of Bovine Respiratory Dis-
ease in Standardized Steer Tests. Beef Res Rep—Iowa State Univ. 1999;
Tennant TC, Ives SE, Harper LB, Renter DG, Lawrence TE. Comparison of tulathromycin and tilmicosin
on the prevalence and severity of bovine respiratory disease in feedlot cattle in association with feedlot
performance, carcass characteristics, and economic factors. J Anim Sci. 2014; 92(11):5203–13. https://
doi.org/10.2527/jas.2014-7814 PMID: 25349362
36.
Fulton RW. Bovine respiratory disease research (1983–2009). Anim Heal Res Rev. 2009; 10(2):131–9.
37. Davis-Unger J, Pajor EA, Schwartzkopf-Genswein K, Marti T, Dorin C, Spackman E, et al. Economic
impacts of lameness in feedlot cattle. Transl Anim Sci. 2017;(July):467–79. https://doi.org/10.2527/
tas2017.0052 PMID: 32704670
38. Checkley SL, Janzen ED, Campbell JR, McKinnon JJ. Efficacy of vaccination againts Fusobacterium
necrophorum infection for control of liver abscesses and footrot in feedlot cattle in western Canada. Can
Vet J. 2005; 46(11):1002–7. PMID: 16363327
39. Baptiste KE, Kyvsgaard NC. Do antimicrobial mass medications work? A systematic review and meta-
analysis of randomised clinical trials investigating antimicrobial prophylaxis or metaphylaxis against nat-
urally occurring bovine respiratory disease. Pathog Dis. 2017; 75(7):1–12.
40. Nagaraja TG, Lechtenberg KF. Liver Abscesses in Feedlot Cattle. Vet Clin North Am—Food Anim
Pract. 2007; 23(2):351–69. https://doi.org/10.1016/j.cvfa.2007.05.002 PMID: 17606156
41. Cernicchiaro N, White BJ, Renter DG, Babcock AH. Respiratory Disease in Commercial Feeder Cattle.
2013; 74(2).
42. Dennis EJ, Schroeder TC, Renter DG, Pendell DL. Value of arrival metaphylaxis in U.S. cattle industry.
J Agric Resour Econ. 2018; 43(2):233–50.
43. Galyean ML, Gunter SA, Malcolm-Callis KJ. Effects of arrival medication with tilmicosin phosphate on
health and performance of newly received beef cattle. J Anim Sci. 1995 May; 73(5):1219–26. https://
doi.org/10.2527/1995.7351219x PMID: 7665351
44. Garcia MD, Thallman RM, Wheeler TL, Shackelford SD, Casas E. Effect of bovine respiratory disease
and overall pathogenic disease incidence on carcass traits. J Anim Sci. 2010; 88(2):491–6. https://doi.
org/10.2527/jas.2009-1874 PMID: 19897630
45. Gardner BA, Dolezal HG, Bryant LK, Owens FN, Smith RA. Health of finishing steers: effects on perfor-
mance, carcass traits, and meat tenderness. J Anim Sci [Internet]. 1999; 77(12):3168–75. Available
from: http://www.ncbi.nlm.nih.gov/pubmed/10641860 https://doi.org/10.2527/1999.77123168x PMID:
10641860
46. USDA Agricultural Marketing Service. DataMart [Internet]. 2019. Available from: https://mpr.datamart.
ams.usda.gov/subReport.do
47. Bewley JM, Boehlje, AW, Hogeveen H, Kenyon SJ, Eicher SD, et al. Stochastic simulation using @Risk
for dairy business investment decisions. Agric Financ Rev [Internet]. 2010; 70(1):97–125. Available
from: http://www.emeraldinsight.com/doi/10.1108/00021461011042666
48. National Cattlemen’s Beef Association. Beef Industry Overview [Internet]. Industry Statistics. 2020.
Available from: https://www.ncba.org/beefindustrystatistics.aspx
49. USDA:APHIS:VS. Beef 2007–08 Part I: Reference of Beef Cow-calf management Practices in the
United States, 2007–08. United States Dept Agric [Internet]. 2009;(February):N512–1008. Available
from: http://scholar.google.com/scholar?hl=en&btnG=Search&q=intitle:Beef+2007-08+Part+1:
+Reference+of+Beef+Cow-calf+Management+Practices+in+the+United+States,+2007-08#9
50. USDA NAHMS. Beef 2007–08 Part III: Changes in the U.S. Beef Cow-calf Industry, 1993–2008. Man-
agement. 2009.
51. Brooks KR, Raper KC, Ward CE, Holland BP, Krehbiel CR, Step DL. Economic effects of bovine respi-
ratory disease on feedlot cattle during backgrounding and finishing phases1. Prof Anim Sci. 2016; 27
(3):195–203.
52. Poulsen Nautrup B, Van Vlaenderen I, Gasper SM, Holland RE. Estimating the comparative clinical and
economic consequences of tulathromycin for treatment of present or anticipated outbreaks of bovine
respiratory disease in feedlot cattle in the United States. J Anim Sci. 2013; 91(12):5868–77. https://doi.
org/10.2527/jas.2013-6709 PMID: 24126273
PLOS ONE | https://doi.org/10.1371/journal.pone.0239135 September 15, 2020
17 / 19
PLOS ONE Economics of antimicrobial use in US feedlots
53.
Langemeier M. Net Return Prospects for Cattle Finishing in 2019 [Internet]. farmdocdaily. 2019. Avail-
able from: https://farmdocdaily.illinois.edu/2019/02/net-return-prospects-for-cattle-finishing-in-2019.
html
54. McIntosh WM, Schulz S, Dean W, Scott HM, Barling KS, Takei I. Feedlot Veterinarians’ Moral and
Instrumental Beliefs Regarding Antimicrobial Use in Feedlot Cattle. J Community Appl Soc Psychol.
2008; 16(December 2008):51–67.
55. Graham JP, Boland JJ, Silbergeld E. Growth Promoting Antibiotics in Food Animal Production: An Eco-
nomic Analysis. Public Health Rep [Internet]. 2007; 122(1):79–87. Available from: http://journals.
sagepub.com/doi/10.1177/003335490712200111 PMID: 17236612
56. Brorsen BW, Lehenbauer T, Ji D, Connor J. Economic Impacts of Banning Subtherapeutic Use of Anti-
biotics in Swine Production. J Agric arn Appl Econ. 2001; 3(December):489–500.
57. Mathews KH Jr. Economic Effects of a Ban against Antimicrobial Drugs Used in U.S. Beef Production. J
Agric Appl Econ. 2002; 34:513–30.
58. Sneeringer S, MacDonald J, Key N, McBride W, Mathews K. Economics of Antibiotic Use in U.S. Live-
stock Production [Internet]. 2015. Available from: papers3://publication/uuid/C11E23CF-DED8-4131-
8C2E-E5915A5A3F8F
59.
60.
61.
62.
Food US and Administration Drug. Guidance for Industry #213 New Animal Drugs and New Animal
Drug Combination Products Administered in or on Medicated Feed or Drinking Water of Food- Produc-
ing Animals: Recommendations for Drug Sponsors for Voluntarily Aligning Product Use Conditions
with. Fed Regist [Internet]. 2013;(December 2013). Available from: https://www.fda.gov/downloads/
AnimalVeterinary/GuidanceComplianceEnforcement/GuidanceforIndustry/UCM299624.pdf
FDA. Guidance for Industry The Judicious Use of Medically Important Antimicrobial Drugs in Food-Pro-
ducing Animals. 2012;
Lhermie G, Tauer LW, Gro¨hn YT. The farm cost of decreasing antimicrobial use in dairy production.
PLoS One. 2018; 13(3).
Lhermie G, Tauer LW, Gro¨hn YT. An assessment of the economic costs to the U.S. dairy market of anti-
microbial use restrictions. Prev Vet Med. 2018; 160:63–7. https://doi.org/10.1016/j.prevetmed.2018.09.
028 PMID: 30388999
63. Administration F and D. FDA guidance (152):"Guidance for Industry: Evaluating the Safety of Antimicro-
bial New Animal Drugs With Regard to Their Microbiological Effects on Bacteria of Human Health Con-
cern". Fed Regist. 2003; 68(207):61221.
64. Aidara-Kane A, Angulo FJ, Conly JM, Minato Y, Silbergeld EK, McEwen SA, et al. World Health Organi-
zation (WHO) guidelines on use of medically important antimicrobials in food-producing animals. Anti-
microb Resist Infect Control [Internet]. 2018; 7(1). Available from: http://dx.doi.org/10.1186/s13756-
017-0294-9
65.
Lhermie G, Verteramo Chiu L, Kaniyamattam K, Tauer LW, Scott HM, Gro¨ hn YT. Antimicrobial Policies
in United States Beef Production: Choosing the Right Instruments to Reduce Antimicrobial Use and
Resistance Under Structural and Market Constraints. Front Vet Sci. 2019; 6(July).
66. Wilson BK, Richards CJ, Step DL, Krehbiel CR. Beef species symposium: Best management practices
for newly weaned calves for improved health and well-being. J Anim Sci. 2017; 95(5):2170–82. https://
doi.org/10.2527/jas.2016.1006 PMID: 28727007
67.
68.
Thrift FA, Thrift TA. Review: Update on preconditioning beef calves prior to sale by cow-calf producers.
Prof Anim Sci. 2011; 27:73–82.
Johnson KK, Pendell DL. Market Impacts of Reducing the Prevalence of Bovine Respiratory Disease in
United States Beef Cattle Feedlots. Front Vet Sci [Internet]. 2017; 4(November):1–9. Available from:
http://journal.frontiersin.org/article/10.3389/fvets.2017.00189/full
69. Gagea MI, Bateman KG, Shanahan RA, Van Dreumel T, McEwen BJ, Carman S, et al. Naturally occur-
ring Mycoplasma bovis-associated pneumonia and polyarthritis in feedlot beef calves. J Vet Diagnostic
Investig. 2006; 18(1):29–40.
70. Duff GC, Galyean ML. Board-invited review: Recent advances in management of highly stressed,
newly received feedlot cattle. J Anim Sci. 2007; 85(3):823–40. https://doi.org/10.2527/jas.2006-501
PMID: 17085724
71. Roberts C, Spurgeon C, Fowler J. Characteristics of the United States Organic Beef Industry. Range
Improv Task Force [Internet]. 2007;1–12. Available from: http://scholar.google.com/scholar?hl=
en&btnG=Search&q=intitle:Characteristics+of+the+United+States+Organic+Beef+Industry#9
72. Ward CE, Lusk JL, Dutton JM. Implicit value of retail beef product attributes. J Agric Resour Econ.
2008; 33(3):364–81.
73. USDA ERS. Meat Price Spreads [Internet]. 2019. Available from: https://www.ers.usda.gov/data-
products/meat-price-spreads/
PLOS ONE | https://doi.org/10.1371/journal.pone.0239135 September 15, 2020
18 / 19
PLOS ONE Economics of antimicrobial use in US feedlots
74. Singer RS, Porter LJ, Thomson DU, Gage M, Beaudoin A, Wishnie JK. Raising animals without antibiot-
ics: producer and veterinarian experiences and opinions. bioRxiv [Internet]. 2019; Available from:
https://www.biorxiv.org/content/early/2019/08/18/600965
75. Sharma C, Rokana N, Chandra M, Singh BP, Gulhane RD, Gill JPS, et al. Antimicrobial Resistance: Its
Surveillance, Impact, and Alternative Management Strategies in Dairy Animals. Front Vet Sci [Internet].
2018; 4(January):1–27. Available from: http://journal.frontiersin.org/article/10.3389/fvets.2017.00237/
full
PLOS ONE | https://doi.org/10.1371/journal.pone.0239135 September 15, 2020
19 / 19
PLOS ONE |
10.1371_journal.pone.0251004 | RESEARCH ARTICLE
Remembering the romantic past:
Autobiographical memory functions and
romantic relationship quality
Cagla AydinID
1,2*, Asuman Buyukcan-Tetik1
1 Psychology Program, Faculty of Arts and Social Sciences, Sabanci University, Istanbul, Turkey,
2 Department of Psychology, Norwegian University of Science and Technology, Trondheim, Norway
* cagla.aydin@sabanciuniv.edu
Abstract
Do the reasons why we think about our memories and share them with others have implica-
tions for our romantic relationship quality? In the present series of studies (total N = 1,102),
we aimed to answer this question by examining whether the self (e.g., creating a stable self-
image), social (e.g., connecting with others) and directive (e.g., guiding future behavior)
functions of regular memories (Study 1, Study 2) and relationship memories (Study 2, Study
3) were related to intimacy and satisfaction in the current relationship. We further investi-
gated these links when relationship memories were shared with the romantic partner (Study
3). Results showed no association between the self-reported uses of memory for regular
events and relationship quality. In contrast, the social function served by the relationship
events was positively associated, and the directive function was negatively associated with
intimacy and relationship satisfaction. When the memories were to be shared with the part-
ner, only social function was related, positively, to the relationship satisfaction. Findings
were discussed in terms of the importance of considering the self-reported reasons for
recalling an event and understanding of the contextual factors in remembering.
Introduction
We remember our personal past for many reasons. According to an influential approach, rea-
sons for autobiographical remembering are categorized into three broad functional categories
[1]. Self function refers to recalling events to maintain a sense that one is the same person over
time and keep a positive image of self [2, 3]. Memories are also recalled and shared interper-
sonally to form and strengthen social bonds or deepen intimacy with others, which is referred
to as the social function [4]. Finally, the directive function refers to using the personal past as a
prescription to guide future behaviors as well as problem-solving [5].
A major tenet of the functional approach is that past experiences are adaptively (re)con-
structed in order to make them meaningful for responding to ongoing changes in one’s eco-
logical context [6]. One such immediate context is one’s romantic relationships where
remembering may be consequential in terms of the degree of intimacy we feel toward our part-
ner or the satisfaction we get from the relationship. Therefore, in the present research
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: Aydin C, Buyukcan-Tetik A (2021)
Remembering the romantic past: Autobiographical
memory functions and romantic relationship
quality. PLoS ONE 16(5): e0251004. https://doi.
org/10.1371/journal.pone.0251004
Editor: Alexandra Kavushansky, Technion Israel
Institute of Technology, ISRAEL
Received: January 17, 2021
Accepted: April 18, 2021
Published: May 3, 2021
Copyright: © 2021 Aydin, Buyukcan-Tetik. This is
an open access article distributed under the terms
of the Creative Commons Attribution License,
which permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: All data files are
available from the Open Science Framework (OSF)
database. URL: https://osf.io/2zty9/.
Funding: The author(s) received no specific
funding for this work.
Competing interests: The authors have declared
that no competing interests exist.
PLOS ONE | https://doi.org/10.1371/journal.pone.0251004 May 3, 2021
1 / 18
PLOS ONE Memory functions of the romantic personal past
program, we considered the extent to which the self-reported functions of autobiographical
memory relate to two main indicators of one’s romantic relationship quality, namely intimacy
and relationship satisfaction in individuals who are currently involved in a romantic
relationship.
Prior research has indeed linked different memory functions with romantic relationship
quality. Firstly, a strong association is shown between relationship satisfaction and social func-
tion. This was not surprising given that social function is defined as recalling past life to be inti-
mate with others [7, 8] or as coordinating the individual histories of the two partners [9]. For
instance, Alea and Bluck [7] targeted a form of social function, intimacy function, served by
autobiographical memories, and showed that simply the act of recalling (personal) relationship
events, in contrast to (impersonal) fictional vignettes, fostered intimacy in long-term romantic
relationships. In a similar vein, Alea and Vick [10] reported that frequently rehearsed relation-
ship-defining memories; that is, memories that are used for maintaining close bonds, pre-
dicted marital satisfaction.
In terms of self function, the evidence is relatively indirect. It was reported that the use of
autobiographical memories for self-related reasons was positively related to having good rela-
tions with friends and significant others [11]. Furthermore, given the findings that self func-
tion has strong relations with self-esteem [11, 12], and relationship satisfaction and self-esteem
are positively associated [13], it is likely that the use of self function would result in increased
satisfaction in close relationships. Similarly, based on the evidence that sharing self-relevant
information helps develop relationship intimacy [14], a positive association between remem-
bering for self-related reasons and intimacy between partners can possibly be formed.
Finally, a positive association is also reported between the use of directive function and rela-
tionship satisfaction. Philippe, Koestner, and Lekes [15] showed that couple-related memories,
by way of satisfying psychological needs, such as autonomy, relatedness and competence,
actively direct relationship satisfaction. It is worth noting here that Philippe et al. used the
term ‘directive function’ rather loosely- to refer to a directive influence on one’s thoughts and
behaviors rather than a memory to be instantly used to guide through a conflict. Based on
prior arguments that directive function helps individuals navigate difficult emotional situa-
tions [5, 16], an increase in the frequency of the use of memories for directive purposes, for
instance, problem-solving, may indicate a tendency to prevent or resolve relationship conflicts
in a constructive manner. This, in turn, may benefit relationship quality.
All in all, the reported positive associations suggest that using autobiographical memories
functionally benefits the romantic relationship experience. So far, the links from memory
functions to relationship quality have been formed by unpacking singular functions. While it
is worthwhile to identify individual functions for a fine-grained analysis on how memories
function in social context [8], a global analysis of functions is also critical to assess the extent to
which functions served by autobiographical memories -relative to each other- are implicated
in the dynamic interpersonal sphere. It has been hypothesized that individual memories may
serve more than one function depending on the current psychosocial needs of the individual
[5, 6]. Therefore, it is possible that uses of memories other than the one targeted in a particular
study inadvertently influence the relationship outcomes. A global assessment of the three func-
tions in tandem would allow for examining the relative contributions of each function to
romantic relationship quality. It also has ‘heuristic utility’ [17] in broadly thinking about the
functions in the close relationship context. Thus, the main goal of present study is to examine
how different functions of autobiographical memories relate to the quality of romantic
relationships.
Global assessments of the functions of autobiographical remembering have been made pos-
sible by using psychometric scales, such as the TALE (Thinking About Life Experiences
PLOS ONE | https://doi.org/10.1371/journal.pone.0251004 May 3, 2021
2 / 18
PLOS ONE Memory functions of the romantic personal past
Questionnaire; [4, 18]) and the RFS (Reminiscence Functions Scale; [19]). Rather than focus-
ing on specific relationship events, these scales focus on recalling over one’s life. For the pres-
ent purposes, there are several advantages of using a scale of this sort. First, when the focus is
on particular events, it is not clear whether functions will generalize across all types of events.
For instance, it has been reported that different classes of memories, such as, positive and neg-
ative memories, fulfil different functions; contributing to a self-concept and social bonding,
and help avoiding dangers, respectively [20]. Second, in previous work on relational outcomes,
how memories are used by individuals are largely inferred either by asking individuals to
remember intimate relationship events so that the use of the social function is implicated or
content analyzing the conversational narratives to identify functional themes [e.g., 17]. Alter-
natively, the items in the TALE are supposed to tap deliberate uses of the past [8, 18] via self-
reported intentions and goals [21]. By explicitly asking participants to think back, and report
the usefulness of their own memories, we aim to qualify the connection between conscious
recollection of the past and perceived romantic relationship quality.
Overview of the present studies
In the present series of studies, we examined the association between people’s use of autobio-
graphical memory to serve self, social and directive functions and the quality of one’s romantic
relationship. Based on the argument that memories can serve more than just one function [5],
this approach would allow for observing the relative contributions of the three conceptually
distinct functions to relationship quality. To do so, we relied on individuals’ evaluations of the
functions of their own memories by using the TALE scale rather than their responses to spe-
cific recollections. Participants had been romantically involved with someone for a minimum
of three months, and they were over 18 years of age. First, in Study 1, we started by investigat-
ing whether the way individuals use their -everyday- memories are related to the quality of
their romantic relationships. Given the finding that all three functions are related to psycho-
logical well-being [11], we pursued the question whether the well-being of the relationships
depended on the functions one’s personal memories serve. In Study 2, relying on the possibil-
ity that relational events may afford potentially different uses compared to everyday personal
memories [22], our focus was on how relationship-related memories are used functionally and
how those uses were associated with relationship quality. To that effect, we slightly modified
the wording in the TALE items to reflect that participants need to think over their romantic
life with their current partner when responding to the uses of their memories. Finally, drawing
on the possibility that uses of recalling an autobiographical memory may shift when one
reports past experiences to a certain addressee, in particular, to the romantic other, we
explored whether functions of romantic memories changed when reporting to the current
romantic partner and how that related to one’s relationship quality (Study 3).
We operationalize romantic relationship quality as one’s subjective and global evaluation of
the relationship [23]. Previous research emphasized the importance of using different rela-
tional qualities (e.g., satisfaction, intimacy, commitment, conflict, ambivalence, love) to have a
nuanced understanding of how each quality affects the outcomes [e.g., 24–26]. For reasons of
brevity, we did not use the whole range of the possible qualities.
Finally, because, to our knowledge, this is the first systematic study to examine the scope of
all three functions of memory and their relations to relationship quality, we included an
important relational construct and an individual difference variable, attachment style, as a con-
trol variable. Attachment style is known to be a strong predictor of relationship quality [27,
28]. Attachment orientations were also reported to be systematically related to what individu-
als recall about relationship events [29]. Since responses to the items in the TALE can also be
PLOS ONE | https://doi.org/10.1371/journal.pone.0251004 May 3, 2021
3 / 18
PLOS ONE Memory functions of the romantic personal past
thought of “stylistic differences of reminiscence across individuals” [8], controlling for a
potentially confounding individual difference variable was critical for our purposes. Attach-
ment orientations have been shown to be associated with memory usage. Whereas attachment
avoidance was not related to any of the functions, higher attachment anxiety was found to be
related with the social function [30]. How individuals with different attachment styles respond
to a break up was found to be mediated by remembering relationship related events [31].
Age and gender were included as the other control variables. Age was included as a control
variable because, in the functional remembering literature, it has been shown to be an impor-
tant factor [e.g., 30]. Gender was included because previous studies on relationship memories
point to slight gender differences [e.g., 29].
Study 1
In Study 1, we establish how functions of generalized memories of life events are associated
with romantic relationship quality. Previously, it has been shown that individuals who use
their specific memories to serve all three functions reported higher levels of psychological well-
being [11, but 32]. Since having intimate romantic relationships are fundamental to socio-
emotional well-being [33], we expected that reasons to remember personal experiences would
have similar associations with relationship well-being. Even though we did not have predic-
tions as to the relative weights of each function, we predict that social function would be asso-
ciated with the relationship outcomes to a greater level than the other functions. This is due to
the fact that social use of the memory entails maintaining intimacy which should reflect on the
quality of our relationships—regardless of the event’s theme (relationship related or not). In
fact, Waters [11] study showed that for the use of memories for the recurring events was highly
associated with having positive relationships.
Materials and methods
Participants
Participant recruitment was conducted through an online crowdsourcing company, Prolific.
Inclusion criteria for this study were: being involved in a romantic relationship currently at least
for 3 months, being older than 18 and younger than 70, and being a native speaker of English. In
the original data, there were 252 entries. Fifteen participants were excluded due to multiple
entries, not meeting the inclusion criteria (e.g., being age 18 or over, being in an on-going rela-
tionship at least for three months, speaking English as their native language) or not providing the
Prolific identity number. Detailed information about excluded participants is available upon
request. Sample characteristics of the remaining 237 participants are presented in Table 1.
We computed our sample size for multiple regression analyses based on the total number of
predictors with the control variables (i.e., 9–14 variables) across our studies. We had an expec-
tation of a medium effect size [34]. Using a desired statistical power level of .8 and a probability
level of .05, minimum required sample size was between 113 and 135 across our studies [35],
which were all exceeded (nStudy-1 = 237, nStudy-2 = 410, nStudy-3 = 455).
We received ethical approval for this study from Sabanci University Research Ethics Coun-
cil with the protocol number FASS-2019-49. All participants gave their electronic informed
consent for participation before they filled in the survey in return of GBP 1.35.
Procedure and measures
Questionnaires were administered online. After completing the consent form and reading the
instructions, the participants first completed the Thinking about Life Experiences
PLOS ONE | https://doi.org/10.1371/journal.pone.0251004 May 3, 2021
4 / 18
PLOS ONE Table 1. Sample characteristics across studies.
Age
Gender (Female %)
Ethnicity (Caucasian %)
Sexual orientation (Heterosexual %)
Relationship type (Married %)
Relationship duration in years
Parents (%)
Number of children
Education (Bachelor’s %)
Education (High school/GED %)
https://doi.org/10.1371/journal.pone.0251004.t001
Memory functions of the romantic personal past
Study 1 (n = 237)
Study 2 (n = 410)
Study 3 (n = 455)
M or %
SD
M or %
SD
M or %
SD
10.74
10.53
0.83
41.96
78.10
93.20
95.80
97.00
16.27
71.30
1.99
37.10
38.00
10.90
9.31
0.88
35.97
74.70
91.00
96.00
58.50
11.35
56.80
1.90
45.90
36.80
11.48
9.86
0.93
36.53
70.40
90.50
97.40
57.30
12.29
56.30
1.93
33.20
42.00
Questionnaire [TALE; 18], and then the relationship intimacy (Inclusion of the Other in Self;
IOS; [36]) and satisfaction (Relationship Assessment Scale by Hendrick [37]) scales. The ques-
tionnaire ended with the Experiences in Close Relationships-Revised Questionnaire [28] in
order to measure attachment orientations and the demographic questions. The TALE was
administered before the relationship quality measures; the order was not counterbalanced.
Other scales administered but not used for the present work are not reported here (see S1
Appendix).
Functions of autobiographical memory. Thinking About Life Experiences Question-
naire [TALE; 18] was used to measure the three functions of memory: self function, social
function, and directive function. We asked why participants think back or talk about their life.
Sample reasons to assess each function were “when I want to feel that I am the same person
that I was before”, “when I hope to also learn more about another person’s life”, and “when I
believe that thinking about the past can help guide my future”, respectively. Additionally, there
were two items two assess the baseline level (i.e., “In general, how often do you think back over
your life?” and “In general, how often do you talk to others about what’s happened in your
life?”). A 5-point Likert scale (1 = “almost never”, 5 = “very frequently”) was administered.
We first conducted an exploratory factor analysis using varimax rotation and maximum
likelihood estimation to investigate the factorial structure of the TALE in our data. Results
revealed three factors with an eigenvalue higher than 1, which altogether explained 52.12% of
the variance. All items except one were clearly loading to their function in the original scale.
The item “when I want to remember something that someone else said or did that might help
me now” had similar loadings on both social function and directive function (loadings of .38
and .34, respectively). Thus, we conducted all analyses first including, and then excluding this
item. Each time, we got the same results in terms of the memory functions’ associations to the
relationship quality indicators. In the reported analyses, we included this item under the direc-
tive function considering the original scale. Besides, exclusion of this item did not increase the
internal reliability of the subscale for directive function. For this particular study, the five-item
scales for self function, social function, and directive functions had good internal reliability;
Cronbach alpha levels of .82, .81, and .84, respectively.
Relationship quality. To assess relationship quality, we used two different indicators: inti-
macy and relationship satisfaction. Intimacy in the participants’ romantic relationship was
assessed using the Inclusion of Other in the Self Scale [IOS; 36]. This assessment is done via a
7-point Likert type item composed of seven pictures (degrees of interlocking or isolated cir-
cles) representing different levels of closeness. The participants were asked to choose the
PLOS ONE | https://doi.org/10.1371/journal.pone.0251004 May 3, 2021
5 / 18
PLOS ONE Memory functions of the romantic personal past
option that best portrays their relationship. Higher scores (i.e., increasingly overlapping cir-
cles) indicated higher levels of closeness. Comparison of different closeness measures showed
that the IOS scale “is a psychologically meaningful and highly reliable measure of the subjective
closeness of relationships” [38, p. 1]. Relationship satisfaction was assessed with six items of the
7-item Relationship Assessment Scale, which was developed by Hendrick [37]. Sample item
was “In general, how satisfied are you with your relationship?” One item from the original
scale “How much do you love your partner?” was mistakenly omitted in the online version.
We administered a 5-point Likert scale (1 = “very low”, 5 = “very high”). Cronbach alpha level
was .94 for this study.
Attachment. Attachment styles were assessed using the Experiences in Close Relation-
ships-Revised Questionnaire [28]. Each subscale for assessing anxious and avoidant attach-
ment styles had 18 items. Sample items were “I often worry that my partner doesn’t really love
me.” and “I find it difficult to allow myself to depend on romantic partners.” for anxious and
avoidant attachment styles, respectively. For all items, we administered a 5-point Likert scale
(1 = “strongly disagree”, 5 = “strongly agree”). Cronbach alpha level was .94 for anxious attach-
ment and .96 for avoidant attachment.
Results and discussion
Descriptive statistics and correlations
Descriptive statistics of and correlations among study variables are presented in Table 2. Cor-
relations revealed that the three functions of autobiographical memory (i.e., self, social, and
directive functions) had moderate positive associations with each other. Out of the three func-
tions, only the self function was significantly, but negatively, associated with intimacy and rela-
tionship satisfaction. Self function had positive associations with both anxious and avoidant
attachment styles. Anxious attachment was positively related to directive function as well.
Regression results. We regressed intimacy and relationship satisfaction onto self, social,
and directive functions of autobiographical memory. We also controlled for the effects of age,
gender, attachment styles, and baseline levels of thinking and talking about life (see the
Method section for the assessment of baseline levels). Regression results showed that none of
the functions of autobiographical memory had any significant associations with either inti-
macy or relationship satisfaction (see Tables 3 and 4).
In Study 1 we examined the association between the functions of regular autobiographical
memories and relationship quality indicators (i.e., relationship satisfaction and intimacy).
Even though the self function was negatively correlated with the relationship quality indica-
tors, the results of the regression analyses did not support our predictions. The functions of
autobiographical memories were not associated with relational outcomes.
Table 2. Descriptive statistics and correlations among the study variables in Study 1.
Variable
1 Self function
2 Social function
3 Directive function
4 Intimacy
5 Relationship satisfaction
6 Anxious attachment
7 Avoidant attachment
Note. All values in bold had a p-value lower than .05.
https://doi.org/10.1371/journal.pone.0251004.t002
M
2.76
2.91
3.27
5.34
3.98
2.29
2.09
SD
0.83
0.79
0.76
1.59
0.94
0.82
0.78
1
-
.41
.53
-.20
-.21
.25
.21
2
-
.59
.00
-.02
.11
-.02
3
-
-.08
-.06
.16
.01
4
5
6
-
.78
-.52
-.65
-
-.57
-.70
-
.64
PLOS ONE | https://doi.org/10.1371/journal.pone.0251004 May 3, 2021
6 / 18
PLOS ONE Table 3. Regression results for intimacy across studies.
Age
Gender
Anxious attachment
Avoidant attachment
Thinking about life
Talking about life
Thinking about romantic relationship
Talking about romantic relationship with others
Talking about romantic relationship with the partner
TALE Self function
TALE Social function
TALE Directive function
TARE Self function
TARE Social function
TARE Directive function
SHARE Self function
SHARE Social function
SHARE Directive function
Memory functions of the romantic personal past
Study 1
Study 2
β
p
β
.01
-.08
-.15
-.55
-.06
-.05
-
-
-
-.04
.10
-.04
-
-
-
-
-
-
.86
.13
.03
.00
.34
.38
-
-
-
.51
.11
.58
-
-
-
-
-
-
-.09
-.07
-.14
-.43
-.06
.02
-.02
-.06
-
.02
-.12
.13
.04
.16
-.20
-
-
-
p
.05
.10
.00
.00
.26
.73
.65
.23
-
.76
.06
.06
.62
.03
.01
-
-
-
Study 3
β
p
-.01
.02
-.13
-.39
-
-
.01
-.06
.10
-
-
-
.10
-.06
-.13
-.02
.11
.01
.89
.61
.01
.00
-
-
.83
.21
.04
-
-
-
.23
.37
.10
.76
.12
.88
Note. All values in bold had a p-value lower than .05. TALE = Thinking About Life Experiences Questionnaire.
TARE = Thinking About Relationship Experiences Questionnaire (see Study 2). SHARE = Sharing Relationship
Experiences (see Study 3).
https://doi.org/10.1371/journal.pone.0251004.t003
Table 4. Regression results for relationship satisfaction across studies.
Age
Gender
Anxious attachment
Avoidant attachment
Thinking about life
Talking about life
Thinking about romantic relationship
Talking about romantic relationship with others
Talking about romantic relationship with the partner
TALE Self function
TALE Social function
TALE Directive function
TARE Self function
TARE Social function
TARE Directive function
SHARE Self function
SHARE Social function
SHARE Directive function
Study 1
Study 2
β
-.10
-.05
-.19
-.55
-.09
.06
-
-
-
-.03
.03
-.02
-
-
-
-
-
-
p
.03
.32
.00
.00
.10
.27
-
-
-
.62
.65
.77
-
-
-
-
-
-
β
-.12
-.03
-.18
-.47
-.07
.05
.04
.01
-
-.04
-.07
.04
.01
.19
-.19
-
-
-
p
.00
.43
.00
.00
.11
.31
.39
.74
-
.59
.23
.55
.93
.01
.01
-
-
-
Study 3
β
p
-.06
.01
-.19
-.46
-
-
.01
.04
.04
-
-
-
.01
.01
-.14
-.05
.17
-.04
.11
.81
.00
.00
-
-
.75
.44
.34
-
-
-
.85
.88
.04
.48
.00
.56
Note. All values in bold had a p-value lower than .05. TALE = Thinking About Life Experiences Questionnaire.
TARE = Thinking About Relationship Experiences Questionnaire. SHARE = Sharing Relationship Experiences.
https://doi.org/10.1371/journal.pone.0251004.t004
PLOS ONE | https://doi.org/10.1371/journal.pone.0251004 May 3, 2021
7 / 18
PLOS ONE Memory functions of the romantic personal past
Study 2
Whereas Study 1 addressed the association between functions of regular autobiographical
memories and romantic relationship quality, in Study 2 we were motivated by the idea that
remembering the romantic past may have more direct implications for the quality of one’s
relationships. What purposes do the memories about the loved ones serve in the relationship
context? While prior work examining this relationship targeted remembering particular inci-
dents in a romantic relationship; such as the first time someone met their spouse [10], here,
again, we explore the different functions that the generalized relationship memories serve. Fol-
lowing up on the reasoning in Study 1, if these global assessments consist of conscious pro-
cesses; that is, individuals are aware of the purposes their romantic memories serve, there may
be direct consequences for the relationship quality. For parsimony, we again adopted a self-
report methodology, and modified the items in the TALE for assessing the functions of roman-
tic relationship-related memories. Changing the instructions in the original TALE so that the
participants answer the items in reference to different classes of memories has also been sug-
gested by the creators of the scale [18]. A secondary aim of Study 2 was to replicate the findings
in Study 1, particularly that functions of regular memories were not related to any one of the
relationship quality indicators.
Since, in previous work, there is no direct evidence informing us about the link between the self
function and romantic relationship quality, we rely on the reported positive associations between
the self function and self-esteem [11, 12], and self-esteem and relationship satisfaction [13]. We
therefore expect the frequency of self-function to be positively related to relationship quality.
Similarly, we expect the social function to be positively associated with the relationship
quality indicators. For instance, if one remembers romantic memories to “coordinate the indi-
vidual histories of the two partners” [9], intimacy, and relationship satisfaction should
increase. In fact, Alea and Vick [10] reported that memories of relationship events with higher
qualitative richness -vivid and rehearsed- would correspond to higher marital satisfaction. In a
similar vein, warmth and closeness in a relationship increased after recalling a relationship
event [7]. Other work with couples has also found that retrieving autobiographical memories
about instances where the couple laughed together, as opposed to individual laughter-related
events, was related to enhanced marital satisfaction [39].
Finally, we also expect the use of the directive function to be positively associated with rela-
tionship quality indicators. Recently, Philippe et al. [15] broadly defined the directive function
of memories as having a long-term impact on the cognitions and emotions. They showed that
by way of satisfying psychological needs, such as autonomy, relatedness and competence, cou-
ple-related memories directively influence relationship satisfaction. In the present context,
using romantic relationship memories to guide behavior or to solve current problems should
be positively associated with the quality of one’s relationship as it implies actively working on
the issues in the relationship.
Materials and methods
Participants
Participant recruitment was conducted through the same resource; Prolific. All inclusion crite-
ria and consenting procedures followed Study 1’s lead. Participants in Study 1 were not
allowed to participate in Study 2. After excluding 48 participants due to several reasons (e.g.,
indications of not responding in an honest matter, such as, failure to mark the requested
option in the quality check items or not fulfilling the inclusion criteria listed in Study 1), the
sample consisted of 410 participants. Sample characteristics are given in Table 1.
PLOS ONE | https://doi.org/10.1371/journal.pone.0251004 May 3, 2021
8 / 18
PLOS ONE Memory functions of the romantic personal past
Procedure and measures
All measures were the same as in Study 1 except for a modified version of the TALE [18] which is
described below. Participants first received the original TALE then the modified Thinking About
Relationship Experiences (TARE) Questionnaire to investigate the functions of memories specific
to the current relationship. The administration order of these questionnaires was fixed because we
wanted the TALE results to be generalizable to the other studies; that is, responses not to be influ-
enced by prior thoughts about relationships. These are followed by the measures of relationship
quality indicators; namely, intimacy and relationship satisfaction. For full descriptions, please see
Study 1. This time, we used the full scale for relationship satisfaction.
Functions of autobiographical memory. Similar to Study 1, to measure the functions of
autobiographical memory, we used the TALE [18]. We again conducted an exploratory factor
analysis using varimax rotation and maximum likelihood estimation to investigate the factorial
structure of the TALE. Results of the factor analyses, which confirmed the original 3-factor
structure, are given in S1 Appendix. For the present study, the five-item scales for self-func-
tion, social function, and directive function had good internal reliability levels of .83, .80, and
.84, respectively.
Functions of relationship-related memories. To measure the functions of relationship-
related autobiographical memories, we slightly modified the items of the original TALE Ques-
tionnaire to help respondents think back and talk about their relationship. For instance, a
directive function item in the original scale being “I think back and talk about my life or cer-
tain periods of my life when I want to learn from my past mistakes” was reworded as “I think
back and talk about my relationship or certain parts of my relationship when I want to learn
from my past mistakes.” For parsimony, we name this version Thinking About Relationship
Experiences (TARE) Questionnaire. The participants were warned at the beginning that they
were going to answer two similar but slightly different questionnaires; one would be about
their life and the other one being about their current romantic relationship. Comparison of all
items across two scales are presented in the S1 Appendix section. Again, a 5-point Likert scale
(1 = “almost never”, 5 = “very frequently”) was administered in all items. Factor analyses
revealed a 3-factor structure as same as the functions in the original TALE (see S1 Appendix),
which altogether explained 58.72% of the variance. Internal reliability, Cronbach alpha, levels
of five-item scales in the TARE for self-function, social function, and directive function of rela-
tionship-related memories (.88, .85, and .88, respectively) were slightly higher compared to the
ones in the original TALE.
Results and discussion
Descriptive statistics and correlations
Table 5 presents the descriptive statistics of and correlations among study variables in Study 2.
Regarding the TALE scale, correlations were very similar to the ones reported in Study 1 in
terms of both significance and magnitude. Two differences were the non-significant correla-
tion between self-function and intimacy, and the positive association between social function
and anxious attachment.
Correlations between the same functions in the TALE and TARE ranged between .69 and
.72, meaning that they overlap with each other to some extent but are not the same constructs.
Significant correlations between functions in the TARE and relationship quality variables
showed that self-function in the TARE was negatively linked to relationship satisfaction while
directive function in the TARE is negatively associated with both intimacy and relationship
satisfaction.
PLOS ONE | https://doi.org/10.1371/journal.pone.0251004 May 3, 2021
9 / 18
PLOS ONE Memory functions of the romantic personal past
Table 5. Descriptive statistics and correlations among the Study 2 variables.
Variable
1 TALE Self function
2 TALE Social function
3 TALE Directive function
4 TARE Self function
5 TARE Social function
6 TARE Directive function
7 Intimacy
8 Relationship satisfaction
9 Anxious attachment
10 Avoidant attachment
M
2.66
2.90
3.26
2.63
2.97
3.07
5.22
4.06
2.48
2.18
SD
0.81
0.77
0.76
0.87
0.84
0.86
1.39
0.73
0.80
0.70
1
-
.45
.56
.72
.40
.41
-.05
-.14
.27
.13
2
-
.54
.46
.69
.50
-.03
.01
.15
-.05
3
4
5
6
7
8
9
-
.52
.50
.69
-.04
-.09
.20
.06
-
.60
.67
-.08
-.13
.26
.13
-
.67
.04
.08
.14
-.08
-
-.10
-.11
.20
.05
-
.67
-.32
-.50
-
-.38
-.60
-
.42
Note. All values in bold had a p-value lower than .05. TALE = Thinking About Life Experiences Questionnaire. TARE = Thinking About Relationship Experiences
Questionnaire.
https://doi.org/10.1371/journal.pone.0251004.t005
Regression results
Regression results with the control variables revealed that, similar to the Study 1 results, there
was no link between functions of memory in the original TALE scale and neither relationship
satisfaction nor intimacy (see Tables 3 and 4). Both the intimacy and relationship satisfaction
had positive and negative links with social and directive functions in the TARE respectively.
These effects were significant although we controlled for the effects of confounding variables
including the attachment types (see Tables 3 and 4). Effect sizes were small (f2 = .01 for the
effect of social function on intimacy, f2 = .02 for the effect of directive function on intimacy as
well as for the effects of social and directive functions on relationship satisfaction).
In our regression analysis, similar to the findings in Study 1, there was no link between
functions of memory in the original TALE scale and neither relationship satisfaction nor inti-
macy (see Tables 3 and 4). In turn, consistent with our predictions, functions measured by
TARE had different associations with the relationship quality indicators. Even after controlling
for attachment, age and gender, social function was positively associated with intimacy and
relationship satisfaction; whereas directive function was negatively related to them. This find-
ing supports the idea that relationship-related memories are used in individuals’ daily lives
and are related to relational outcomes. This pattern contributes to the literature in that not
only singled out episodes in relationships, such as vacation with the partner [7] or first time
someone met their spouse [10] would function to increase intimacy levels or satisfaction in a
relationship but also generalized evaluations of multiple episodes, such as the items in TARE,
have associations with the quality of one’s relationship. Overall, the present study was a first in
showing that the functions of relationship-related memories, when studied together, are
related to the relationship quality. This association is qualitatively different from the pattern
with regular, non-relationship-themed, memories which is a finding points to the need for dis-
tinguishing different themes/classes of memories when examining their functions.
Study 3
It has been established that memory sharing is one of the primary functions of autobiographi-
cal memory [40]. Sharing autobiographical memories has been shown to lead to relationship
closeness across cultural settings [41]. It is, therefore, a critical omission in the memory func-
tion literature that the role of different interlocutors is rarely considered. Given that the func-
tions and characteristics of the shared vs non-shared memories differ [e.g., 42], events shared
PLOS ONE | https://doi.org/10.1371/journal.pone.0251004 May 3, 2021
10 / 18
PLOS ONE Memory functions of the romantic personal past
with an intimate other are likely to differ in function from events shared with other people
[22]. Therefore, the main aim was to explore whether functions of relationship-related events
were similarly associated with relationship quality indicators when they were shared with the
romantic partner vs any other person. It is, for instance, possible that when shared memories
have a tendency to serve a self function, the quality of the relationship may deteriorate due to
the realization of the partner of not being included in the meaning-making process of the
romantic experience. Thus, associations between memory functions and relationship quality
may be different than when they were not shared with the partner or in some cases, such as
when the social function is involved, may be enhanced.
Materials and methods
Participants
We followed the same procedure explained in the first two studies and recruited participants
through Prolific. Participants in the first two studies were not allowed to participate in this
study. There were 455 participants in the dataset after the exclusion of 63 participants because
of various reasons such as failure in quality check questions. Characteristics of the final sample
are given in Table 1.
Procedure and measures
Functions of relationship-related memories. We used the TARE (Thinking about Rela-
tionship Experiences) again to examine whether we can replicate our findings in Study 2. Fur-
thermore, we adapted the items in the TARE to investigate the functions of relationship
memories when they were shared with the current partner. The modified scale is referred to as
SHARE (Sharing Relationship Experiences) from here on. As an example for the difference
between the TARE and SHARE scales, the TARE item “I think back and talk to other people
about my relationship or certain parts of my relationship when I want to learn from my past
mistakes” was used as “I think back and talk to my partner about my relationship or certain
parts of my relationship when I want to learn from my past mistakes” in the SHARE. Compari-
son of items across scales are presented in the S1 Appendix section.
Exploratory factor analysis revealed a 3-factor structure for the TARE in this study too with
an explained variance of 56.72% in total (see S1 Appendix). Although two items loaded simi-
larly onto two different factors, the results were almost identical when those items were
excluded except that the effect of TARE directive function on relationship satisfaction in
Table 4 became marginal (β = -.12, p = .07). Thus, we continued with the original 3-factor
structure. Five-item scales for self-function, social function, and directive function in the
TARE had good internal reliability levels of .87, .83, and .86, respectively.
The SHARE had a 2-factor structure in the exploratory factor analysis with an explained
variance of 58.26% (see S1 Appendix). The first factor was again representing the self function
with the same 5 items. Social and directive functions however, overlapped and constituted a
separate function together. This indicates that talking about the relationship problems with the
partner to guide future behaviors for example (i.e., directive function) also has a role in bond-
ing the partners with each other (i.e., social function). For the present results to be comparable
with the findings using both the TALE and TARE as well as considering the theoretical differ-
ences between social and directive functions, we still used these two factors of SHARE sepa-
rately in our analysis. The internal reliability levels were also supporting our decision to
use the three functions separately. Five-item scales for self function, social function, and direc-
tive function had good internal reliability levels (Cronbach alphas) of .91, .89, and .85,
respectively.
PLOS ONE | https://doi.org/10.1371/journal.pone.0251004 May 3, 2021
11 / 18
PLOS ONE Memory functions of the romantic personal past
Other variables. For assessing relationship quality and attachment styles, we used the
same measures in previous studies. Internal reliability levels for relationship satisfaction, anx-
ious attachment, and avoidant attachment were .92, .93, and .95 respectively.
Results and discussion
Descriptive statistics and correlations
Descriptive statistics of and correlations among study variables in Study 3 are presented in
Table 6. Correlations between functions in the TARE and relationship quality variables
showed negative association of self function and directive function with relationship satisfac-
tion. All three functions were positively linked to anxious attachment. Avoidant attachment
was positively linked to self function, but negatively linked to social function.
Correlations between the same functions in the TARE and SHARE ranged between .52 and
.73, which showed that they somewhat overlap with each other but tap into different con-
structs. Correlations between functions in the SHARE and relationship quality revealed that
only social function had significant associations with intimacy and relationship satisfaction.
Regression results
Regression results in Table 3 showed that none of the functions either in TARE or SHARE had
significant associations with intimacy. Results about relationship satisfaction however, showed
that directive function in the TARE and social function in the SHARE were negatively and
positively associated with relationship satisfaction, respectively. In the model with the control
variables including attachment (Table 4), effect sizes were relatively small: f2 = .01 for the effect
of TARE directive function on relationship satisfaction, and f2 = .02 for the effect of SHARE
social function on relationship satisfaction.
Furthermore, as described in S1 Appendix, we also conducted the same analysis using the
2-factor structure of the SHARE (i.e., social function and the combined factor of social and
directive functions). The results with the 2-factor structure of the SHARE revealed that the
negative effect of TARE directive function on intimacy was in line with the finding in Study 2
(see Table 3). This effect was not significant in Study 3 when the 3-factor structure was used
(see Table 3). The combination of the social and directive functions in the SHARE had a posi-
tive effect on relationship satisfaction. This effect was in line with the positive effect of social
function in Study 3 when the 3-factor structure was used (see Table 4). SHARE directive
Table 6. Descriptive statistics and correlations among the Study 3 variables.
Variable
1 TARE Self function
2 TARE Social function
3 TARE Directive function
4 SHARE Self function
5 SHARE Social function
6 SHARE Directive function
7 Intimacy
8 Relationship satisfaction
9 Anxious attachment
10 Avoidant attachment
M
2.26
2.75
2.89
2.43
3.40
3.09
5.22
4.03
2.43
2.11
SD
0.86
0.84
0.85
0.93
0.90
0.84
1.47
0.79
0.82
0.72
1
-
0.54
0.66
0.73
0.30
0.40
-0.09
-0.16
0.29
0.11
2
-
0.67
0.40
0.52
0.49
-0.03
0.02
0.17
-0.09
3
-
0.52
0.48
0.66
-0.08
-0.10
0.21
0.00
4
-
0.48
0.63
0.00
-0.07
0.23
-0.02
5
6
7
8
9
-
0.72
0.17
0.22
0.12
-
0.07
0.05
0.16
-0.28
-0.16
-
0.66
-0.32
-0.48
-
-0.41
-0.62
-
0.46
Note. All values in bold had a p-value lower than .05. TARE = Thinking About Relationship Experiences Questionnaire. SHARE = Sharing Relationship Experiences.
https://doi.org/10.1371/journal.pone.0251004.t006
PLOS ONE | https://doi.org/10.1371/journal.pone.0251004 May 3, 2021
12 / 18
PLOS ONE Memory functions of the romantic personal past
function alone was not significant (see Table 4). We also explored whether the associations of
memory functions with relationship quality depended on relationship duration across the
three studies (30 interactions in total). The results showed that only one out of 30 interactions
was noteworthy (see S1 Appendix for details). The association of the SHARE social function
with intimacy depended on relationship duration (b = .35, p < .001). Simple slope analyses
showed that the SHARE social function had a significant positive association with intimacy in
people with longer relationship duration (1 SD above the sample mean; b = .56, p < .001), but
no association in people with shorter relationship duration (1 SD below the sample mean; b =
-.13, p = .30). Hence, our exploratory examinations did not reveal a strong role of relationship
duration in the associations between memory functions and relationship quality.
Thus, when the addressee of the memory sharing activity was defined as the romantic part-
ner, the way functions were linked to the quality of relationships slightly differ from when
sharing with the partner is not specified. The positive association between the social function
and relationship satisfaction (see TARE results in Study 2) is still intact however the negative
link between the use of directive function and satisfaction (see TARE results in Study 2 and
Study 3) is not there anymore. This is the first study that we know of to show that the associa-
tions between the intended reasons to share romantic memories and relationship quality may
slightly change when the romantic partner is the interlocutor. The implications of this finding
are discussed further below.
General discussion
In the present study we aimed to examine the association of functional use of memory and
romantic relationship quality in three studies. We did so by focusing on generalized views on
memories in order to examine the three overarching functions together. We first looked at
whether everyday memories’ functions and relationship quality were linked. In a second study,
we shifted our focus to the functions relationship-related memories serve; and in the third
study, we specified the interlocutor as the romantic partner when considering functions of
romantic memory sharing and relationship quality association.
The predicted positive relations between the three functions and functions of everyday
memories were not confirmed; none of the functions were associated with the quality indica-
tors. We did, however, observe the functions of romantic memories to be associated with
romantic relationships outcomes. How shall these findings inform current theorizing regard-
ing functional remembering in social context?
The lack of association between reasons to remember everyday autobiographical events and
the quality of one’s relationships is in contrast with the previous findings showing that func-
tional remembering -all three functions- is related to having positive relationships [11]. A
closer look, however, reveals that the association in the Waters study was reported for single
events only but not for recurring or general events. A summary of one’s lifetime periods, as
indexed by the TALE, may not have the functional power to have an immediate effect on the
relationship quality but a relationship-specific single event, such as “when we saw that movie
together” or even general events such as “our walks to school together” might have a binding
role for other relationship events; and therefore, its functional relations to the quality of
romantic relationships may be more salient. The present findings therefore suggest that the
general tendency with which individuals remember their past life may not be associated with
the quality of their romantic relationships.
Alternatively, the items in the two studies may have tapped different aspects of the so-called
self function. While Waters [11] used the Centrality of Event Scale which measures whether
the event recalled constitutes a key part of one’s identity; therefore, taps the identity aspect; the
PLOS ONE | https://doi.org/10.1371/journal.pone.0251004 May 3, 2021
13 / 18
PLOS ONE Memory functions of the romantic personal past
TALE scale in the present study examines the continuity of the self in time; therefore, coher-
ence of a self-concept. Pillemer [8] has noted that conceptual categories that the TALE mea-
sures may not encompass the full spectrum of the subfunctions of a particular category.
Therefore, self function as indexed by the TALE items may not be associated with the relation-
ship quality but whether or not a particular event helps defining the self is. Further research
with a focus on the spectrum of the subfunctions is required to support this interpretation.
With regards to the romantic memories (Study 2), results suggest that the social function
was positively related to both relationship satisfaction and intimacy; whereas directive function
was found to be negatively correlated with both of them. The positive association between the
social function and how it is related to relationship satisfaction [10] and intimacy [7] has been
shown previously with specific, one-time relationship memories, such as first-sight or first-
kiss. The novelty of the present findings is that across different settings–as the items in the
scale imply- social function is similarly related to the relationship outcomes.
An unexpected finding was the directive function to be negatively associated with relation-
ship quality when relationship memories are shared with other people. Previous research has
tracked the influence of a single episode; a specific memory’s directive power and found that it
was positively associated with one’s satisfaction in a relationship [8, 15]. A relationship-related
memory; for instance, a prior quarrel, could naturally be used constructively within the rela-
tionship for problem-solving purposes or to fine-tune particulars of future behavior. When
faced with a problem, individuals would bring to mind memories of situations involving a sim-
ilar problem and use that particular memory to work through the challenge [1, 17, 43]. This
guidance would reflect positively on relationship satisfaction. In the present study, however,
we were dealing with a global evaluation of how frequently relationship memories are used for
directive purposes. If one’s perception of the frequency of the use of memories for problem-
solving purposes, is high, it might indicate that the frequency of the problems to be solved is
also high. Following that logic, if the perception of the number of problems (that needs to be
solved) in a relationship is high, it is highly likely that the perceived satisfaction in the relation-
ships would not benefit it.
The negative effect of directive function on relationship quality vanishes when relationship
memories are shared with the partner rather than anonymous others (Study 3). This is further
support for the idea that memories are used based on the changing dynamics of the situation
or context [17, 22]. Why is sharing relationship memories for directive purposes with others
detrimental for relationships? Previous research showed that discussing relationship problems
with friends harms relationship quality, if similar discussions do not take place with the part-
ner [44]. Perhaps the discussion with the partner brings the opportunity to take the perspective
of the partner and smoothly resolve the conflicts, which is not possible when relationship
memories are told to others.
Differential results across social and directive functions may also be due to the valence of
the memories. Previous research showed stronger associations of social and directive functions
with positive and negative memories, respectively [20]. Thus, our findings (in Study 2) reveal-
ing the beneficial effect of social function, but detrimental effect of directive function on rela-
tionship quality may not be surprising. Linking this finding with the relationship research,
perhaps social function is more salient in capitalization attempts (i.e., sharing good news with
others/partner; e.g., [45]), whereas directive function is more salient during discussions about
relationship problems [44]. These questions await future research.
An unexpected finding was the lack of social function’s effect on intimacy for the relation-
ship memories shared with the partner despite its positive effect on relationship satisfaction.
Previous studies showed the bonding roles of disclosure and capitalization in romantic rela-
tionships [45, 46]. One possible explanation of not observing the positive trend here is that we
PLOS ONE | https://doi.org/10.1371/journal.pone.0251004 May 3, 2021
14 / 18
PLOS ONE Memory functions of the romantic personal past
did not consider reaction of the partner. Future research should examine whether social func-
tion builds intimacy when the partner shows constructive and supportive responses [47].
All in all, it was critical to show that global evaluations of memory functions also have direct
associations with the quality of the relationships. Further, as we noted previously, here we
operationalize function as the deliberate use of the memories [1]. Prior work focusing on sin-
gled out relationship episodes, in turn, rely on non-conscious uses of memories. Since it has
been suggested that some functions may be less accessible than others [22], future work is
needed to make this distinction clearer. For instance, the present studies could be replicated
with measures other than self-reported uses. One way is to conduct content analysis of func-
tional use on the memory narratives [e.g., 48].
The present study further shows that memory functions in the modified versions of the
TALE, which we called the TARE (Thinking about relationship experiences) and SHARE
(Sharing relationship experiences with the partner) were differentially linked to relationship
quality. We conclude that TARE and SHARE have utility as separate tools to examine reminis-
cence dynamics in the romantic relationship context, and to inform intervention or counseling
programs. A cautionary note, however, is that the present design employed a fixed ordering of
the questionnaires which may have influenced the participants’ responses. Even though there
was a clear warning in the instructions that there would be two very similar questionnaires to
fill out, there is no way to know whether responding to the TALE (or the TARE, in the second
study) questions had an alerting or inhibiting role on the subsequent scales. Since our main
aim in this research was to examine the roles of memory functions by comparing our results
using the original measure (TALE) and modified measures (TARE and SHARE) with each
other, in none of our studies the TARE or SHARE were used separately. Future studies plan-
ning to use these measures should take this into account and test the replicability of our find-
ings in contexts where the TARE or the SHARE are used in isolation.
The three-function model is suggested to be used as a conceptual model [8] with heuristic
utility [17] for understanding functional uses of the memories broadly. In fact, each function is
regarded as an organizatory unit to include many subpurposes [22]. Targeted experimental
manipulations are needed to fine-tune each function’s association to relational outcomes. With
the present design, it may not be entirely possible to rule out alternative explanations; such as,
individuals that are highly satisfied in their relationships may tend to remember for social rea-
sons or individuals who do not possess feelings of intimacy in the relationship may tend to use
memories mostly for problem-solving purposes. Future research should also identify whether
positive or negative relationship memories function the same way. Some newly identified func-
tions, such as the mood-enhancement function, are very loosely captured by the three categories
[49] but would be very relevant to examine for the romantic relationship context.
Our operationalization of the romantic relationship quality (satisfaction and intimacy)
should be considered as a first step in exploring the wide range of possible qualities [e.g., 26].
Exploring conflict, for instance, would be interesting in terms of how memory is used to deal
with negative relational outcomes. Future research should also consider the moderating roles
of other relationship characteristics (e.g., married vs. cohabitating, same-sex vs. heterosexual,
monogamous vs. non-monogamous relationships).
It should also be noted that psychological well-being has been previously associated with
the memory functions [11]. Therefore, it is possible that our memories’ influence on one’s rela-
tionship quality might be through their effects on general wellbeing and not because of their
direct effects on relationships. It would be worthwhile for the future studies to focus on the
mediating effects of psychological health on this mechanism.
In conclusion, our findings suggest that when remembering has consequences in terms of
the quality of a romantic relationship, how memories are used change depending on their
PLOS ONE | https://doi.org/10.1371/journal.pone.0251004 May 3, 2021
15 / 18
PLOS ONE Memory functions of the romantic personal past
theme (relationship-related or not) and who the social partner is (romantic other or not).
Future studies should consider fine-tuning these broad functional categories in order to fur-
ther understand the causal mechanisms. For instance, whether or not remembering a particu-
lar relationship incident directively might hinder relationship satisfaction. Together, the extant
findings suggest adopting a contextual approach as they revealed not only that functions relate
to different social contexts such as romantic relationships but also that they consider the role
of the social partner.
Supporting information
S1 Appendix.
(XLSX)
Acknowledgments
The authors would like to thank the two research assistants, Irem Duman and Lindon Kras-
niqi, for their help in preparation of the study materials and data collection. We also thank
three anonymous reviewers for their constructive feedback.
Author Contributions
Conceptualization: Cagla Aydin, Asuman Buyukcan-Tetik.
Investigation: Cagla Aydin, Asuman Buyukcan-Tetik.
Methodology: Cagla Aydin, Asuman Buyukcan-Tetik.
Project administration: Cagla Aydin, Asuman Buyukcan-Tetik.
Resources: Cagla Aydin, Asuman Buyukcan-Tetik.
Supervision: Cagla Aydin, Asuman Buyukcan-Tetik.
Validation: Cagla Aydin, Asuman Buyukcan-Tetik.
Visualization: Cagla Aydin, Asuman Buyukcan-Tetik.
Writing – original draft: Cagla Aydin, Asuman Buyukcan-Tetik.
Writing – review & editing: Cagla Aydin, Asuman Buyukcan-Tetik.
References
1. Bluck S, Alea N. Exploring the functions of autobiographical memory: Why do I remember the autumn?
In: Webster JD, & Haight BK, editors. Critical advances in reminiscence work: From theory to applica-
tion. Springer Publishing Company: 2002. pp 61–75.
2. Bluck S, Alea N. Remembering being me: The self continuity function of autobiographical memory in
younger and older adults. In: Sani F, editor. Self continuity: Individual and collective perspectives. Psy-
chology Press: 2008. pp. 55–70.
3. Wilson AE, Ross M. The identity function of autobiographical memory: Time is on or side. Memory.
2003; 11(2): 137–149. https://doi.org/10.1080/741938210 PMID: 12820827
4. Bluck S, Alea N, Habermas T, Rubin DC. A TALE of three functions: the self-reported uses of autobio-
graphical memory. Soc Cogn. 2005; 23(1): 91–117. https://doi.org/10.1521/soco.23.1.91.59198
5. Pillemer DB. Directive functions of autobiographical memory: the guiding power of the specific episode.
Memory. 2003; 11(2): 193–202. https://doi.org/10.1080/741938208 PMID: 12820831
6. Bluck S, Alea N, Demiray B. You get what you need: The psychosocial functions of remembering. In:
Mace JH, editor. New perspectives in cognitive psychology. The act of remembering: Toward an under-
standing of how we recall the past. Wiley-Blackwell; 2010. pp 284–307. https://doi.org/10.1002/
9781444328202.ch12
PLOS ONE | https://doi.org/10.1371/journal.pone.0251004 May 3, 2021
16 / 18
PLOS ONE Memory functions of the romantic personal past
7. Alea N, Bluck S. I’ll keep you in mind: the intimacy function of autobiographical memory. Appl Cogn Psy-
chol. 2007; 21(8):1091–1111. https://doi.org/10.1002/acp.1316
8. Pillemer DB. Twenty years after Baddeley (1988): Is the study of autobiographical memory fully func-
tional? Appl Cogn Psychol. 2009; 23(8): 1193–1208. https://doi.org/10.1002/acp.1619
9. Wegner DM, Erber R, Raymond P. Transactive memory in close relationships. J Pers Soc Psychol.
1991; 61(6): 923–929. https://doi.org/10.1037//0022-3514.61.6.923 PMID: 1774630
10. Alea N, Vick SC. The first sight of love: Relationship-defining memories and marital satisfaction across
adulthood. Memory. 2010; 18(7):730–742. https://doi.org/10.1080/09658211.2010.506443 PMID:
20721804
11. Waters TEA. Relations between the functions of autobiographical memory and psychological wellbeing.
Memory. 2014; 22(3): 265–275. https://doi.org/10.1080/09658211.2013.778293 PMID: 23537126
12. Demiray B, Janssen SMJ. The self-enhancement function of autobiographical memory. Appl Cogn Psy-
chol. 2015; 29:49–60. https://doi.org/10.1002/acp.3074
13. Erol RY, Orth U. Development of self-esteem and relationship satisfaction in couples: Two longitudinal
studies. Dev Psychol. 2014; 50: 2291–2303. https://doi.org/10.1037/a0037370 PMID: 24999764
14.
Laurenceau JP, Feldman Barrett LA, Pietromonaco PR. Intimacy as an interpersonal process: The
importance of self-disclosure and perceived partner responsiveness in interpersonal exchanges. J Pers
Soc Psychol. 1998; 74(5): 1238–1251. https://doi.org/10.1037//0022-3514.74.5.1238 PMID: 9599440
15. Philippe FL, Koestner R, Lekes N. On the directive function of episodic memories in people’s lives: A
look at romantic relationships. J Pers Soc Psychol. 2013; 104(1): 164–179. https://doi.org/10.1037/
a0030384 PMID: 23088231
16. Krans J, de Bree J, Bryant RA. Autobiographical memory bias in social anxiety. Memory. 2014; 22(8):
890–897. https://doi.org/10.1080/09658211.2013.844261 PMID: 24111655
17. Pasupathi M, Lucas S, Coombs A. Conversational functions of autobiographical remembering: long-
married couples talk about conflicts and pleasant topics. Discourse Process. 2002; 34(2): 163–192.
https://doi.org/10.1207/S15326950DP3402
18. Bluck S, Alea N. Crafting the TALE: construction of a measure to assess the functions of autobiographi-
cal remembering. Memory. 2011; 19(5): 470–486. https://doi.org/10.1080/09658211.2011.590500
PMID: 21864212
19. Webster JD. Construction and validation of the reminiscence functions scale. J Gerontol. 1993; 48(5):
256–262. https://doi.org/10.1093/geronj/48.5.p256 PMID: 8366271
20. Rasmussen AS, Berntsen D. Emotional valence and the functions of autobiographical memories: posi-
tive and negative memories serve different functions. Mem Cogn. 2009; 37: 477 4492. https://doi.org/
10.3758/MC.37.4.477 PMID: 19460954
21. McClelland DC, Koestner R, Weinberger J. How do self-attributed and implicit motives differ? Psychol
Rev. 1989; 96(4): 690–702. https://doi.org/10.1037/0033-295X.96.4.690
22. Kulkofsky S, Wang Q, Hou Y. Why I remember that: the influence of contextual factors on beliefs about
everyday memory. Mem Cogn. 2010; 38: 461–473. https://doi.org/10.3758/MC.38.4.461 PMID:
20516226
23.
Fincham FD, Bradbury TN. The assessment of marital quality: A reevaluation. J Marriage Fam. 1987;
49: 797–809.
24. Brock RL, Kochanska G. Decline in the quality of family relationships predicts escalation in children’s
internalizing symptoms from middle to late childhood. J Abnorm Child Psychol. 2015; 43(7): 1295–
1308. https://doi.org/10.1007/s10802-015-0008-9 PMID: 25790794
25.
26.
Fincham FD, Linfield KJ. A new look at marital quality: Can spouses feel positive and negative about
their marriage?. J Fam Psychol. 1997; 11(4):489–502.
Totenhagen CJ, Serido J, Curran MA, Butler EA. Daily hassles and uplifts: A diary study on understand-
ing relationship quality. J Fam Psychol. 2012; 26(5): 719–728. https://doi.org/10.1037/a0029628 PMID:
22906123
27. Hazan C, Shaver P. Romantic love conceptualized as an attachment process. J Pers Soc Psychol.
1987; 52(3): 511–524. https://doi.org/10.1037//0022-3514.52.3.511 PMID: 3572722
28.
Fraley RC, Shaver PR. Adult romantic attachment: Theoretical developments, emerging controversies,
and unanswered questions. Rev Gen Psychol. 2000; 4:132–154. https://doi.org/10.1037/1089-2680.4.
2.132
29. Simpson JA, Rholes WS, Winterheld HA. Attachment working models twist memories of relationship
events. Psychol Sci. 2010; 21(2): 252–259. https://doi.org/10.1177/0956797609357175 PMID:
20424054
PLOS ONE | https://doi.org/10.1371/journal.pone.0251004 May 3, 2021
17 / 18
PLOS ONE Memory functions of the romantic personal past
30. Vranić A, Jelić M, Tonković M. Functions of autobiographical memory in younger and older adults. Front
Psychol. 2018; 9: Article 219. https://doi.org/10.3389/fpsyg.2018.00219 PMID: 29599732
31. Marshall TC, Bejanyan K, Ferenczi N. Attachment styles and personal growth following romantic break-
ups: The mediating roles of distress, rumination, and tendency to rebound. PloS One. 2013; 8(9):
e75161. https://doi.org/10.1371/journal.pone.0075161 PMID: 24066169
32. Grace L, Dewhurst SA, Anderson RJ. A dysphoric’s TALE: The relationship between the self-reported
functions of autobiographical memory and symptoms of depression. Memory. 2016; 24(9): 1173–1181.
https://doi.org/10.1080/09658211.2015.1084009 PMID: 26371517
33. Baumeister RF, Leary MR. The need to belong: Desire for interpersonal attachments as a fundamental
human motivation. Psychol Bull.1995; 117(3), 497–529. https://doi.org/10.1037/0033-2909.117.3.497
PMID: 7777651
34. Alea NL. I’ll keep you in mind: The intimacy function of autobiographical memory in adulthood(disserta-
tion). University of Florida; 2004. https://doi.org/10.1023/b:comh.0000035227.57576.46 PMID:
15453084
35. Soper DS. A-priori sample size calculator for multiple regression [software] [cited 2021 March 12]. Avail-
able from: https://www.danielsoper.com/statcalc/calculator.aspx?id=1
36. Aron A, Aron EN, Smollan D. Inclusion of other in the self scale and the structure of interpersonal close-
ness. J Pers Soc Psychol. 1992; 63(4): 596–612. https://doi.org/10.1037/0022-3514.63.4.596
37. Hendrick SS. A generic measure of relationship satisfaction. J Marriage Fam. 1988; 50(1): 93–98.
https://doi.org/10.2307/352430
38. Ga¨chter S, Starmer C, Tufano F. Measuring the closeness of relationships: a comprehensive evaluation
of the ’Inclusion of the Other in the Self’ scale. PloS One. 2015; 10(6): e0129478. https://doi.org/10.
1371/journal.pone.0129478 PMID: 26068873
39. Bazzini DG, Stack ER, Martincin PDet al. The effect of reminiscing about laughter on relationship satis-
faction. Motiv Emot. 2007; 31: 25–34. https://doi.org/10.1007/s11031-006-9045-6
40. Neisser U. Memory: What are the important questions? In: Gruneberg MM, Morris PE, Sykes RN, edi-
tors. Practical aspects of memory. London: Academic Press; 1978. pp. 3–24.
41. Guan L, Wang Q. Does sharing memories make us feel closer? The roles of memory type and culture.
https://doi.org/10.31234/osf.io/7udy4
42.
Lorenzetti R, Lugli L. Sharing autobiographical memories: Effects of arousal and memory. Stud Com-
mun Sci. 2017; 12(1): 53–57. https://doi.org/10.1016/j.scoms.2012.06.011
43. Cohen G. The effects of aging on autobiographical memory. In: Hompson CP, Herrmann DJ, Bruce D,
Read DJ, Payne DG, Toglia MP, editors. Autobiographical Memory: Theoretical and Applied Perspec-
tives. Hillsdale NJ: Lawrence Erlbaum Associates: 1998. pp 105–123.
44.
Jensen JF, Rauer AJ. Turning inward versus outward: Relationship work in young adults and romantic
functioning. Pers Relatsh. 2014; 21(3): 451–467. https://doi.org/10.1111/pere.12042
45. Gable SL, Reis HT, Impett EA, Asher ER. What do you do when things go right? The intrapersonal and
interpersonal benefits of sharing positive events. J Pers Soc Psychol. 2004; 87(2): 228–245. https://
doi.org/10.1037/0022-3514.87.2.228 PMID: 15301629
46.
Finkenauer C, Buyukcan-Tetik A. To know you is to feel intimate with you: Felt knowledge is rooted in
disclosure, solicitation, and intimacy. Fam Sci. 2015; 6:109–118. https://doi.org/10.1080/19424620.
2015.1082012
47. Pagani AF, Parise M, Donato S, Gable SL, Schoebi D. If you shared my happiness, you are part of me:
Capitalization and the experience of couple identity. Pers Soc Psychol Bull. 2020; 46(2): 258–269.
https://doi.org/10.1177/0146167219854449 PMID: 31179894
48. Wasti AS, Aydin C, Altunsu B, Baykent BTRecalling positive and negative events: a cross-cultural
investigation of the functions of work-related memories. J Appl Res Mem Cogn. 2021 February: Online
first https://doi.org/10.1016/j.jarmac.2020.10.001
49. Wolf T, Demiray B. The mood-enhancement function of autobiographical memories: Comparisons with
other functions in terms of emotional valence. Conscious Cogn. 2019; 70: 88–100. https://doi.org/10.
1016/j.concog.2019.03.002 PMID: 30884386
PLOS ONE | https://doi.org/10.1371/journal.pone.0251004 May 3, 2021
18 / 18
PLOS ONE |
10.1371_journal.pone.0246437 | RESEARCH ARTICLE
The outcomes and decision-making process
for neck lymph nodes with indeterminate
fine-needle aspiration cytology
Ping-Chia ChengID
1, Chih-Ming Chang1,2, Li-Jen Liao1,3,4, Po-Wen Cheng1, Wu-Chia Lo1*
1 Department of Otolaryngology Head and Neck Surgery, Far Eastern Memorial Hospital, New Taipei City,
Taiwan (R.O.C.), 2 Department of Biomedical Engineering, National Yang-Ming University, Taipei, Taiwan
(R.O.C.), 3 Department of Electrical Engineering, Yuan Ze University, Taoyuan, Taiwan (R.O.C.), 4 Medical
Engineering Office, Far Eastern Memorial Hospital, New Taipei City, Taiwan (R.O.C.)
* lowuchia@gmail.com
Abstract
Objectives
This study aims to propose a cytological classification, to evaluate predictive factors of the
final malignancy, and to suggest a proper management strategy for neck lymph nodes
(LNs) with indeterminate cytology.
Methods
Patients who had neck lymphadenopathy with indeterminate cytology between 2007 and
2017 were analyzed retrospectively in a tertiary medical center. Cytological classification
was conducted according to the cytological descriptions. We examined the clinical charac-
teristics according to the final diagnosis of the neck lymphadenopathy.
Results
According to the final diagnoses, there were 142 malignant and 95 benign neck LNs among
237 patients. Multivariate analyses using a stepwise logistic regression model showed that
cytological classification [p < 0.001, OR = 5.67 (3.48–9.23)], prior history of malignancy [p =
0.01, OR = 2.97 (1.26–6.99)], long axis [p = 0.01, OR = 3.06 (1.33–7.06)], short-to-long axis
(S/L) ratio [p = 0.047, OR = 2.15 (1.01–4.57)] and internal echogenicity [p = 0.01, OR = 2.72
(1.26–5.86)] were independent predictors of malignancy.
Conclusions
In patients who have neck LNs with indeterminate cytology, a cytological classification and
four other predictors (prior history of malignancy, long axis � 1.93 cm, S/L ratio � 0.64 and
heterogeneity of internal echogenicity) are statistically associated with the risk of malig-
nancy and helpful in guiding further management.
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: Cheng P-C, Chang C-M, Liao L-J, Cheng
P-W, Lo W-C (2021) The outcomes and decision-
making process for neck lymph nodes with
indeterminate fine-needle aspiration cytology.
PLoS ONE 16(2): e0246437. https://doi.org/
10.1371/journal.pone.0246437
Editor: Francis Moore, Jr., Brigham and Women’s
Hospital, Harvard Medical School, UNITED STATES
Received: October 17, 2020
Accepted: January 19, 2021
Published: February 4, 2021
Copyright: © 2021 Cheng et al. This is an open
access article distributed under the terms of the
Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: All relevant data are
within the manuscript and its Supporting
information files.
Funding: This work was supported by grants from
the Far Eastern Memorial Hospital Research
Program (FEMH-2016-C-026 and FEMH-2017-C-
015). The funders had no role in study design, data
collection and analysis, decision to publish, or
preparation of the manuscript.
Competing interests: The authors have declared
that no competing interests exist.
PLOS ONE | https://doi.org/10.1371/journal.pone.0246437 February 4, 2021
1 / 11
PLOS ONE Neck lymph nodes with indeterminate cytology
Introduction
Neck lymphadenopathy (LAP) is a common problem that causes patients to visit surgeons.
The incidence rate of neck LAP is approximately 0.6% to 0.7% per year [1]. The common
causes include reactive hyperplasia, tuberculous lymphadenitis, other granulomatous
lymphadenitis, metastatic tumor and lymphoma [1–3]. For the evaluation of neck
LAP, ultrasound (US) and US-guided fine-needle aspiration (US-FNA) are widely used
due to their low cost, ease of operation and capability for simultaneous sampling. The
sensitivity, specificity and accuracy rates of FNA for diagnosing malignant neck LAP range
from 78% to 98% [2, 4–8], 67% to 100% [2, 4–8] and 82% to 97% [4, 6, 8], respectively.
Despite the good performance of US-FNA, some cytological reports may not demonstrate a
definite diagnosis even with adequate sampling. A previous study defined this cytological
group as an indeterminate group [9]. They also classified the cytological reports of FNA
into 4 groups: nondiagnostic (inadequate sample), benign, indeterminate, and malignant
[9]. For neck lymph node (LN) aspirations, indeterminate cytology accounted for 8% to
10% of all FNA cytological reports [3, 9]. Bandoh et al. showed that the malignancy rate of
indeterminate cytology was 79% (11 of 14) [9]. Tarantino et al. presented that the malig-
nancy rate of atypical cytology was 77% (10 of 13) [3]. However, their case numbers were
small.
Currently, no clear guidelines, such as the Bethesda system utilized for thyroid nodules,
exist with regard to clinical decision-making when faced with a cervical LN with indeterminate
FNA cytology. In addition, we all know that some characteristics, such as demographic and
sonographic features, have been applied to assess the final malignancy risk of neck LNs [10,
11]. The purposes of this study are thus threefold: 1) to propose a cytological classification sys-
tem for indeterminate cytology and to examine whether the new system is a practical method
for determining the final malignancy, 2) to evaluate if demographic and sonographic features
could help to predict the final malignancy in these patients, and 3) to suggest an effective man-
agement strategy when facing a neck LN with indeterminate FNA cytology.
Materials and methods
Ethical considerations
This study was approved by the institutional ethical review board of Far Eastern Memorial
Hospital [IRB No. 107102-E]. The study did not influence the patients’ treatment or outcome.
All data were retrospectively collected using a de-identified form between January 2020 and
March 2020. We then analyzed the results from this anonymized data set. The final data set of
the current study is within the S1 Table.
Inclusion and exclusion criteria
We performed a retrospective study at a tertiary medical center. We followed the strengthen-
ing the reporting of observational studies in epidemiology (STROBE) statement in this study.
Patients who received US and US-FNA due to neck LAP from October 2007 to September
2017 were reviewed. We divided the cytological reports into four classifications: nondiagnos-
tic, benign, indeterminate, and malignant. The patients who had indeterminate cytological
reports were included and subsequently analyzed. All patients either received core needle
biopsy, excisional biopsy, neck dissection or follow-up for at least 1 year to obtain the final
diagnosis. Patients who did not have histological reports or were lost to follow-up for over one
year after the cytological reports were excluded.
PLOS ONE | https://doi.org/10.1371/journal.pone.0246437 February 4, 2021
2 / 11
PLOS ONE Neck lymph nodes with indeterminate cytology
Clinical characteristics and outcomes assessment
The age, sex, side of LAP, prior history of malignancy and cytological reports were recorded
from the medical charts. We assumed that the level of atypia was related to the malignancy
rate. Thus, we further classified patients with indeterminate cytology into low-, moderate- and
high-risk groups according to the cytological descriptions. The low-risk group included those
with mild atypia, focal atypical cells, or tumor necrosis in the cytological report. The moder-
ate-risk group includes those with atypia or some atypical cells in the description. The high-
risk group included those with a statement of highly atypia, highly atypical cells or suspicion of
malignancy in the report. These cytological descriptions were made by our cytopathologists
after examining the cellularity, nuclear/cytoplasmic ratio, nuclear hyperchromatism, mitotic
features, and nuclear outline [12]. We unbiasedly reviewed the formal reports and divided
these indeterminate statements into 3 groups. The US findings of short and long axis, short-
to-long axis (S/L) ratio, and other sonographic features, including boundary, internal echo-
genicity, echogenicity, calcifications, architecture, hilar echogenicity and vascular pattern of
the neck LNs, were retrieved from a Marosis PACS system (Marotech Inc., Seoul, South
Korea). We analyzed the demographic data and sonographic features according to the final
diagnosis of the neck LAP.
Statistical analysis
Statistical analysis was performed using STATA software, version 12.0 (Stata Corporation,
College Station, TX). Statistical significance was defined as p < 0.05. Categorical variables
were compared using the chi-squared or Fisher’s exact test, while continuous variables were
compared using the two-sample t-test. Multivariate analyses were performed to identify the
risk factors of a final malignancy by using a stepwise logistic regression model adjusted by age
and sex. We defined the significance level as 0.05 for removal from the model. Odds ratios
(ORs) with 95% confidence intervals (95% CIs) were reported.
Results
There were 3393 patients who underwent US-FNA in our department from October 2007 to
September 2017. Among them, 237 patients (7%, 273/3393) who had 237 LNs with indetermi-
nate cytology were analyzed in the study (Table 1). The mean (SD) age of these patients was 50
(16) years, ranging from 12 to 88 years. The mean (SD) short and long axes of LAP were 1.19
(0.66) and 1.93 (1.10) cm, respectively. The mean (SD) S/L ratio was 0.64 (0.17). We further
used the mean values of the above factors to dichotomize these patients.
According to the final diagnosis, there were 142 malignant and 95 benign LNs. The malig-
nancy rate was 60% in these LNs. Among the malignant results, oral cancer, lymphoma and
thyroid cancer were the most common diagnoses (Table 1). Among the indeterminate cytol-
ogy results, the final malignancy rate was highest in the high-risk group [94% (97 of 103)],
followed by the moderate-risk group [44% (27 of 62)] and low-risk group [25% (18 of 72)]
(Fig 1).
The demographic data and sonographic features were compared between benign and
malignant LNs according to the final diagnosis. There were significant differences in age
(p = 0.001), sex (p < 0.001), short axis (p < 0.001), long axis (p < 0.001), S/L ratio (p = 0.002),
prior history of malignancy (p < 0.001), cytological classification (p < 0.001), boundary
(p < 0.001), internal echogenicity (p < 0.001), calcifications (p < 0.001), hilar echogenicity
(p < 0.001) and vascular pattern (p = 0.03) between groups, but no significant differences in
the side of LAP or other sonographic features including echogenicity and architecture were
observed (Table 2).
PLOS ONE | https://doi.org/10.1371/journal.pone.0246437 February 4, 2021
3 / 11
PLOS ONE Neck lymph nodes with indeterminate cytology
Table 1. Demographic data of the patients who had neck LNs with indeterminate cytology.
Demographic data
Age, mean (SD), yrs
Sex, No. (%)
Female
Male
Side, No. (%)
Right
Left
Bilateral
Short axis, mean (SD), cm
Long axis, mean (SD), cm
S/L ratio, mean (SD)
Prior history of malignancy, No. (%)
Cytological classification, No. (%)
Low risk
Moderate risk
High risk
Final diagnoses
Malignant lymph nodes (N = 142, 60%)
Oral cancer
Lymphoma
Thyroid cancer
Nasopharyngeal carcinoma
Unknown primary neck cancer
Hypopharyngeal cancer
Laryngeal cancer
Oropharyngeal cancer
Lung cancer
Breast cancer
Esophageal cancer
Parotid cancer
Conjunctival cancer
Bladder cancer
Cervical cancer
Colon cancer
Benign lymph nodes (N = 95, 40%)
Reactive hyperplasia
Tuberculous lymphadenitis
Kikuchi disease
Toxoplasmic lymphadenitis
Other granulomatous lymphadenitis
Abbreviation: S/L, short-to-long axis; LNs, lymph nodes.
https://doi.org/10.1371/journal.pone.0246437.t001
N = 237
50 (16)
81 (34%)
156 (66%)
108 (46%)
102 (43%)
27 (11%)
1.19 (0.66)
1.93 (1.10)
0.64 (0.17)
92 (39%)
72 (30%)
62 (26%)
103 (44%)
33
27
26
14
9
8
6
5
4
3
2
1
1
1
1
1
83
6
3
1
2
Multivariate analyses performed by using a stepwise logistic regression model adjusted for
age and sex showed that cytological classification had the highest OR (5.67, 95% CI: 3.48–
9.23), with a significant difference between groups (p < 0.001). The other independent risk
factors for a final malignancy included long axis [p = 0.01, OR = 3.06 (1.33–7.06)], S/L ratio
PLOS ONE | https://doi.org/10.1371/journal.pone.0246437 February 4, 2021
4 / 11
PLOS ONE Neck lymph nodes with indeterminate cytology
Fig 1. Final malignancy rates of indeterminate cytology among different cytological classifications.
https://doi.org/10.1371/journal.pone.0246437.g001
[p = 0.047, OR = 2.15 (1.01–4.57)], prior history of malignancy [p = 0.01, OR = 2.97 (1.26–
6.99)] and internal echogenicity [p = 0.01, OR = 2.72 (1.26–5.86)] (Table 3). We further per-
formed a trend test to evaluate the consistency between the two variables (long axis, S/L ratio)
and a final malignancy. The results showed significant positive correlations for these two vari-
ables (p < 0.05).
We then used the cytological classification and other four variables (prior history of malig-
nancy, long axis � 1.93 cm, S/L ratio � 0.64 and heterogeneity of internal echogenicity) in
combination to weigh the malignancy rate. The final malignancy rate was 94% in the high-risk
group, and we did not need to check the other four variables in the group. For the low- and
moderate-risk groups, the malignancy rate increased with the number of positive significant
variables (Table 4).
Discussion
Occasionally, cytopathologists may not be able to report a unique diagnosis despite the sample
being adequate for analysis, and this kind of result was deemed to be indeterminate. Indeter-
minate FNA cytology reportedly accounts for 8–10% of all cervical LN FNAs [3, 9]. In this
series, the rate of indeterminate FNA cytology was 7%, which was similar to that in the previ-
ous reports. Furthermore, knowing the malignancy rate of indeterminate FNA cytology results
can help us perform patient counseling as well as guide further management. Tarantino et al.
[3] and Bandoh et al. [9] showed that the final malignancy rates of LNs with indeterminate
cytology were 77% (10/13) and 79% (11/14), respectively. In our study, we analyzed 237
PLOS ONE | https://doi.org/10.1371/journal.pone.0246437 February 4, 2021
5 / 11
PLOS ONE Table 2. Comparison of demographic and sonographic features between patients with benign and malignant final diagnoses from neck LNs with indeterminate
cytology.
Neck lymph nodes with indeterminate cytology
Variables, No. (%)
Demographic data
Age
<50 yrs
�50 yrs
Sex
Female
Male
Side
Right
Left
Bilateral
Short axis (mean = 1.19 cm)
<1.19 cm
�1.19 cm
Long axis (mean = 1.93 cm)
<1.93 cm
�1.93 cm
S/L ratio (mean = 0.64)
<0.64
�0.64
Prior history of malignancy
Absent
Present
Cytological classifications
Low risk
Moderate risk
High risk
Sonographic feature
Boundary
Clear
Vague
Internal echogenicity
Homogenous
Heterogeneous
Echogenicity
Hyper
Iso
Hypo
Calcification
Absent
Present
Architecture
Cystic
Solid
Benign
N = 95 (40%)
Final Diagnoses
Malignant
N = 142 (60%)
P value
58 (51%)
37 (30%)
46 (57%)
49 (31%)
41 (38%)
43 (42%)
11 (41%)
68 (53%)
27 (25%)
72 (50%)
23 (25%)
57 (50%)
38 (31%)
71 (49%)
24 (26%)
54 (75%)
35 (56%)
6 (6%)
79 (48%)
16 (22%)
74 (54%)
21 (21%)
0 (0%)
3 (30%)
92 (41%)
89 (45%)
6 (15%)
7 (32%)
88 (41%)
55 (49%)
87 (70%)
35 (43%)
107 (69%)
67 (62%)
59 (58%)
16 (59%)
61 (47%)
81 (75%)
73 (50%)
69 (75%)
56 (50%)
86 (69%)
74 (51%)
68 (74%)
18 (25%)
27 (44%)
97 (94%)
86 (52%)
56 (78%)
63 (46%)
79 (79%)
4 (100%)
7 (70%)
131 (59%)
108 (55%)
34 (85%)
15 (68%)
127 (59%)
0.001 a)
< 0.001 a)
0.82
< 0.001 a)
< 0.001 a)
0.002 a)
< 0.001 a)
< 0.001 a)
<0.001 a)
<0.001 a)
0.21 b)
<0.001 a)
0.41
(Continued )
PLOS ONE | https://doi.org/10.1371/journal.pone.0246437 February 4, 2021
6 / 11
PLOS ONE Neck lymph nodes with indeterminate cytology
Table 2. (Continued)
Variables, No. (%)
Hilar echogenicity
Absent
Present
Vascular pattern
Hilar or avascular
Other
Benign
N = 95 (40%)
65 (33%)
30 (73%)
79 (44%)
16 (28%)
Final Diagnoses
Malignant
N = 142 (60%)
131 (67%)
11 (27%)
100 (56%)
42 (72%)
P value
<0.001 a)
0.03 a)
Abbreviation: S/L, short-to-long axis; LNs, lymph nodes.
a)Statistical significance, p < 0.05.
b)Calculated with Fisher’s exact test.
https://doi.org/10.1371/journal.pone.0246437.t002
patients who had indeterminate cytological reports, and the final malignancy rate was 60%,
which is slightly lower than that in the previous studies. The discrepancies among studies
regarding the final malignancy rate in this population might be due to differences in the defini-
tion of indeterminate FNA cytology or interpretation among cytopathologists. Borrowing
from the Bethesda classification for thyroid FNA diagnosis, the indeterminate group included
atypia of undetermined significance, follicular neoplasm, and suspicion of malignancy [13].
Similarly, we proposed a cytological classification system for cervical LNs with indeterminate
FNA cytology. We noted that the final malignancy rate was higher in reports noting a suspi-
cion of malignancy or highly atypical cells and lower in those noting focal atypical cells or
tumor necrosis. To the best of our knowledge, the current study is the first and largest study to
further divide cervical LNs with indeterminate FNA cytology into 3 different risk groups.
Among the demographic data and sonographic features, age, sex, short axis, long axis, S/L
ratio, prior history of malignancy, cytological classification, boundary, internal echogenicity,
calcifications, hilar echogenicity and vascular pattern were related to a final malignant
Table 3. Multivariate analyses performed by using a stepwise logistic regression model to identify the risk factors
for malignant results in patients who had neck LNs with indeterminate cytology.
Variables
Age
Sex
Short axis
Long axis
S/L ratio
Prior history of malignancy
Cytological classification
Boundary
Internal echogenicity
Calcification
Hilar echogenicity
Vascular pattern
OR
1.95
1.69
3.06
2.15
2.97
5.67
2.72
95% CI
0.92–4.11
0.79–3.64
1.33–7.06
1.01–4.57
1.26–6.99
3.48–9.23
1.26–5.86
P value
0.08
0.18
0.01 a)
0.047 a)
0.01 a)
< 0.001 a)
0.01 a)
Abbreviation: S/L, short-to-long axis; LNs, lymph nodes.
a)Statistical significance, p < 0.05.
https://doi.org/10.1371/journal.pone.0246437.t003
PLOS ONE | https://doi.org/10.1371/journal.pone.0246437 February 4, 2021
7 / 11
PLOS ONE Neck lymph nodes with indeterminate cytology
Table 4. The final malignancy rate was determined by using the cytological classification and other four predic-
tors in patients who had neck LNs with indeterminate cytology.
Cytological classification
Four factors (prior history of malignancy, long
axis � 1.93 cm, S/L ratio � 0.64, and heterogeneity of
internal echo)
Malignancy
rate
Low risk (mild atypia, focal atypical
cells, or tumor necrosis)
Moderate risk (atypia or some atypical
cells)
All negative
1 positive
2 positive
3 positive
All positive
All negative
1 positive
2 positive
3 positive
All positive
High risk (highly atypia, highly atypical
cells or suspicion of malignancy)
Abbreviation: S/L, short-to-long axis; LNs, lymph nodes.
https://doi.org/10.1371/journal.pone.0246437.t004
2/18 (11%)
4/26 (15%)
5/16 (31%)
7/12 (58%)
Nil
0/12 (0%)
7/16 (44%)
11/24 (46%)
7/8 (88%)
2/2 (100%)
96/103 (93%)
diagnosis (Table 2). Previous studies have shown that size, shape, margin, hilar echogenicity
and vascular pattern are diagnostic factors for malignant neck LNs [10, 14]. The sonographer
performed US-FNA of the suspicious LN with one or more of the abnormal features men-
tioned above. However, when the cytological reports reveal indeterminate results, no study has
evaluated factors that predict if nodes are malignant or not in this circumstance. In this study,
multivariate analyses with a stepwise logistic regression model adjusted by age and sex showed
that cytological classification had the highest OR with a significant difference. Moreover, the
other four variables (prior history of malignancy, long axis, S/L ratio and internal echogeni-
city) were also independent factors in predicting a final malignancy in patients who initially
had nodal aspirations with indeterminate cytological results (Table 3). The size of the malig-
nant node tends to become large as the tumor rapidly grows [11]. The shape of the malignant
node tends to be round, and an increasing S/L ratio might be noted [11]. The internal echo-
genicity of the malignant node tends to be heterogeneous due to its necrotic and solid compo-
sition [15]. The neck recurrence rate in patients with a prior history of malignancy was not
low [16, 17]; thus, these patients tend to have more malignant results. These four variables can
be quickly evaluated during neck US and are easy to review with the reporting system after
knowing the cytological reports.
In reality, unfortunately, head and neck surgeons are commonly forced to make clinical
and operative decisions based on suboptimal conditions, such as indeterminate FNA results.
According to the 2015 American Thyroid Association (ATA) guidelines for thyroid nodules
with indeterminate cytological results, repeat FNA or molecular testing is recommended for
the low malignancy risk group, diagnostic surgical excision is advised for the moderate risk
group, and surgical management similar to that for malignant cytology results is suggested for
those with a suspicion of malignancy, even though the estimated final malignancy rate is 60–
75% [18]. Based on the ATA guidelines for thyroid nodules, we divided our patients according
to the cytological classification system and the other four significant predictors in Table 4. The
final malignancy rate was 94% in the high-risk group. For the low- and moderate-risk groups,
the malignancy rate increased with more positive significant variables. As a result, for neck
LAP with indeterminate cytological results, we can evaluate the cytological classification first
PLOS ONE | https://doi.org/10.1371/journal.pone.0246437 February 4, 2021
8 / 11
PLOS ONE Neck lymph nodes with indeterminate cytology
and then check if there is a prior history of malignancy, long axis � 1.93 cm, S/L ratio � 0.64
and heterogeneity of internal echogenicity before making further decisions. For the high-risk
group, we recommended directly managing the LAP as a malignancy. If no variables were pos-
itive in the low- and moderate-risk groups, the malignancy rates were 11% and 0%, respec-
tively. Under these conditions, close observation or repeat FNA may be the appropriate
treatment option. If one or two predictors were positive in the low- and moderate-risk groups,
the malignancy rates were higher, and we suggested core needle biopsy or excisional biopsy to
further confirm the diagnoses. If three or four variables were positive in the low- and moder-
ate-risk groups, the malignancy rate was 58–100% in this series. We suggested handling the
LAP as a malignancy in these circumstances.
Limitations
There were several limitations in this study. First, not all our final diagnoses were based on
the histopathologic findings. After a follow-up of at least 12 months, patients with nodes that
diminished or were equal in size were deemed as negative. Inevitably, there may be a small
chance that the node was positive when a slow-growing metastatic lesion, such as metastasis
from papillary thyroid carcinoma, was encountered. Second, there could be bias in the results
from including patients with a prior history of malignancy. We tried to divide the patients into
a treatment-naïve group and a group with a prior history of malignancy. After the univariate
and multivariate analyses, the cytological classification remained an independent factor
(p < 0.001) with a high odds ratio in predicting a final malignancy in both groups (not
reported in the present study). This meant that regardless of whether the patient was treatment
naïve, the cytological classification could be used as the first priority to ascertain the malig-
nancy risk in patients with indeterminate FNA results. In this series, we added one parameter
of a prior history of malignancy to minimize the bias from heterogeneity by including treated
and treatment-naïve patients. Third, the data were only center-based and lacked verification.
There is no standard classification for defining mild atypical, atypical or highly atypical cells,
and discrepancies might exist between the interpretations of cytopathologists. A standardized
classification of the degree of the indeterminate FNA results will be greatly advantageous in the
absence of a definitive diagnosis from FNA. Development of a standard classification system in
the future can help determine patient risk for malignant cervical nodes and further guide clini-
cal decision-making. Further large-scale and prospective studies are necessary in the future.
Conclusion
Indeterminate FNA cytology in the evaluation of cervical LAP should raise the suspicion of
malignancy. Cytological classification and four other predictors (prior history of malignancy,
long axis � 1.93 cm, S/L ratio � 0.64 and heterogeneity of internal echogenicity) are all statisti-
cally associated with the risk of malignancy in this group of patients and are helpful in guiding
further management.
Supporting information
S1 Table. Raw data set.
(XLSX)
Acknowledgments
The authors thank the pathologist, Wen-Chih Huang for the explanation of cytological features.
PLOS ONE | https://doi.org/10.1371/journal.pone.0246437 February 4, 2021
9 / 11
PLOS ONE Neck lymph nodes with indeterminate cytology
Author Contributions
Conceptualization: Ping-Chia Cheng, Li-Jen Liao, Wu-Chia Lo.
Data curation: Ping-Chia Cheng.
Formal analysis: Ping-Chia Cheng, Chih-Ming Chang.
Investigation: Chih-Ming Chang, Li-Jen Liao.
Methodology: Chih-Ming Chang, Li-Jen Liao, Po-Wen Cheng.
Project administration: Po-Wen Cheng, Wu-Chia Lo.
Software: Ping-Chia Cheng.
Supervision: Li-Jen Liao, Wu-Chia Lo.
Validation: Li-Jen Liao, Po-Wen Cheng, Wu-Chia Lo.
Visualization: Po-Wen Cheng, Wu-Chia Lo.
Writing – original draft: Ping-Chia Cheng.
Writing – review & editing: Po-Wen Cheng, Wu-Chia Lo.
References
1. Chau I, Kelleher MT, Cunningham D, Norman AR, Wotherspoon A, Trott P, et al. Rapid access multidis-
ciplinary lymph node diagnostic clinic: analysis of 550 patients. Br J Cancer. 2003; 88(3):354–61. Epub
2003/02/06. https://doi.org/10.1038/sj.bjc.6600738 PMID: 12569376.
2. Hafez NH, Tahoun NS. Reliability of fine needle aspiration cytology (FNAC) as a diagnostic tool in
cases of cervical lymphadenopathy. J Egypt Natl Canc Inst. 2011; 23(3):105–14. Epub 2012/07/11.
https://doi.org/10.1016/j.jnci.2011.09.009 PMID: 22776815.
3.
4.
5.
Tarantino DR, McHenry CR, Strickland T, Khiyami A. The role of fine-needle aspiration biopsy and flow
cytometry in the evaluation of persistent neck adenopathy. American journal of surgery. 1998; 176
(5):413–7. Epub 1999/01/05. https://doi.org/10.1016/s0002-9610(98)00233-5 PMID: 9874424.
Johnson JT, Rosen CA, Bailey BJ. Bailey’s head and neck surgery otolaryngology. 5th ed: Wolters
Kluwer Health/Lippincott Williams & Wilkins, Philadelphia, PA; 2014. 1760–87 p.
Layfield LJ. Fine-needle aspiration in the diagnosis of head and neck lesions: a review and discussion
of problems in differential diagnosis. Diagn Cytopathol. 2007; 35(12):798–805. Epub 2007/11/17.
https://doi.org/10.1002/dc.20769 PMID: 18008348.
6. Borhani AA, Monaco SE. Chapter 7 Image-Guided Fine-Needle Aspiration and Core Needle Biopsy of
Neck Lymph Nodes: Techniques, Pearls, and Pitfalls. Semin Ultrasound CT MR. 2017; 38(5):531–41.
Epub 2017/10/17. https://doi.org/10.1053/j.sult.2017.05.007 PMID: 29031369.
7. Houcine Y, Romdhane E, Blel A, Ksentini M, Aloui R, Lahiani R, et al. Evaluation of fine needle
aspiration cytology in the diagnosis of cervical lymph node lymphomas. J Craniomaxillofac Surg.
2018; 46(7):1117–20. Epub 2018/05/22. https://doi.org/10.1016/j.jcms.2018.04.024 PMID:
29779620.
8. Goret CC, Goret NE, Ozdemir ZT, Ozkan EA, Dogan M, Yanik S, et al. Diagnostic value of fine needle
aspiration biopsy in non-thyroidal head and neck lesions: a retrospective study of 866 aspiration materi-
als. International journal of clinical and experimental pathology. 2015; 8(8):8709–16. Epub 2015/10/16.
PMID: 26464615.
9. Bandoh N, Goto T, Akahane T, Ohnuki N, Yamaguchi T, Kamada H, et al. Diagnostic value of liquid-
based cytology with fine needle aspiration specimens for cervical lymphadenopathy. Diagn Cytopathol.
2016; 44(3):169–76. Epub 2016/01/11. https://doi.org/10.1002/dc.23402 PMID: 26748563.
10.
Liao LJ, Wang CT, Young YH, Cheng PW. Real-time and computerized sonographic scoring system for
predicting malignant cervical lymphadenopathy. Head Neck. 2010; 32(5):594–8. Epub 2009/08/21.
https://doi.org/10.1002/hed.21225 PMID: 19693943.
11. Ahuja AT, Ying M, Ho SY, Antonio G, Lee YP, King AD, et al. Ultrasound of malignant cervical lymph
nodes. Cancer Imaging. 2008; 8:48–56. Epub 2008/04/09. https://doi.org/10.1102/1470-7330.2008.
0006 PMID: 18390388.
PLOS ONE | https://doi.org/10.1371/journal.pone.0246437 February 4, 2021
10 / 11
PLOS ONE Neck lymph nodes with indeterminate cytology
12. Pusztaszeri MP, Faquin WC. Cytologic evaluation of cervical lymph node metastases from cancers of
unknown primary origin. Semin Diagn Pathol. 2015; 32(1):32–41. Epub 2015/01/27. https://doi.org/10.
1053/j.semdp.2014.12.002 PMID: 25618223.
13. Cibas ES, Ali SZ. The 2017 Bethesda System for Reporting Thyroid Cytopathology. Thyroid: official
journal of the American Thyroid Association. 2017; 27(11):1341–6. Epub 2017/11/02. https://doi.org/10.
1089/thy.2017.0500 PMID: 29091573.
14. Rotim T, Kristek B, Turk T, Kretic D, Peric M, Puseljic I, et al. Measurable and Unmeasurable Features
of Ultrasound Lymph Node Images in Detection of Malignant Infiltration. Acta clinica Croatica. 2017; 56
(3):415–24. Epub 2018/02/27. https://doi.org/10.20471/acc.2017.56.03.08 PMID: 29479907.
15. Yoshida H, Yusa H, Ueno E, Tohno E, Tsunoda-Shimizu H. Ultrasonographic evaluation of small cervi-
cal lymph nodes in head and neck cancer. Ultrasound in medicine & biology. 1998; 24(5):621–9. Epub
1998/08/08. https://doi.org/10.1016/s0301-5629(98)00025-8 PMID: 9695264.
16. Chang JH, Wu CC, Yuan KS, Wu ATH, Wu SY. Locoregionally recurrent head and neck squamous cell
carcinoma: incidence, survival, prognostic factors, and treatment outcomes. Oncotarget. 2017; 8
(33):55600–12. Epub 2017/09/15. https://doi.org/10.18632/oncotarget.16340 PMID: 28903447.
17. Amar A, Chedid HM, Rapoport A, Dedivitis RA, Cernea CR, Brandao LG, et al. Update of assessment
of survival in head and neck cancer after regional recurrence. J Oncol. 2012; 2012:154303. Epub 2012/
11/06. https://doi.org/10.1155/2012/154303 PMID: 23125856.
18. Haugen BR, Alexander EK, Bible KC, Doherty GM, Mandel SJ, Nikiforov YE, et al. 2015 American Thy-
roid Association Management Guidelines for Adult Patients with Thyroid Nodules and Differentiated
Thyroid Cancer: The American Thyroid Association Guidelines Task Force on Thyroid Nodules and Dif-
ferentiated Thyroid Cancer. Thyroid: official journal of the American Thyroid Association. 2016; 26(1):1–
133. Epub 2015/10/16. https://doi.org/10.1089/thy.2015.0020 PMID: 26462967.
PLOS ONE | https://doi.org/10.1371/journal.pone.0246437 February 4, 2021
11 / 11
PLOS ONE |
10.1371_journal.pone.0247551 | RESEARCH ARTICLE
Variable host responses mediate host
preference in marine flatworm−snail
symbioses
Juhyung LeeID
¤a*, Timothy M. Davidson¤b, Mark E. Torchin
Smithsonian Tropical Research Institute, Balboa, Ancon, Republic of Panama
¤a Current address: Marine Science Center, Northeastern University, Nahant, Massachusetts, United States
of America
¤b Current address: Department of Biological Sciences, California State University, Sacramento, California,
United States of America
* juh.lee@northeastern.edu
Abstract
Host preference of symbionts evolves from fitness trade-offs. However, it is often unclear
how interspecific variations in host response traits influence this evolutionary process.
Using the association between the polyclad flatworm Paraprostatum echinolittorinae and its
intertidal snail hosts on the Pacific Coast of Panama, we assessed how a symbiont’s host
preference is associated with varying host defenses and post-infestation performances. We
first characterized the prevalence and intensity of worm infestation in five snail hosts (Tegula
pellisserpentis, Nerita scabricosta, N. funiculata, Planaxis planicostatus, and Cerithium ster-
cusmuscarum). We then used manipulative experiments to test flatworm’s host choice,
hosts’ behavioral rejection of flatworms, and hosts’ growth and survival following the infesta-
tion. In the field, flatworms were orders of magnitude more prevalent and dense in T. pellis-
serpentis, N. scabricosta, N. funiculata than P. planicostatus and C. stercusmuscarum,
although the three former hosts were not necessarily more abundant. The results from our
laboratory host selection trials mirrored these patterns; flatworms were 3 to 14 times more
likely to choose T. pellisserpentis, N. scabricosta, N. funiculata over P. planicostatus and C.
stercusmuscarum. The less preferred hosts frequently rejected flatworms via mantle con-
tractions and foot withdrawals, which reduced the infestation rate by 39%−67%. These
behaviors were less frequent or absent in the preferred hosts. Flatworm infestation variably
influenced host performances in the field, negligibly affecting the growth and survival of T.
pellisserpentis and N. funiculata but reducing the growth of P. planicostatus. Flatworms thus
preferred less defended hosts that can also support higher worm densities without being
harmed. Stable isotope analysis further revealed that flatworms are unlikely to feed on snail
tissues and may live as a commensal in their preferred hosts. Our study demonstrates that
host response traits can modulate a symbiont’s host choice and calls for more explicit con-
siderations of host response variability in host preference research.
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: Lee J, Davidson TM, Torchin ME (2021)
Variable host responses mediate host preference in
marine flatworm−snail symbioses. PLoS ONE
16(3): e0247551. https://doi.org/10.1371/journal.
pone.0247551
Editor: Erik V. Thuesen, Evergreen State College,
UNITED STATES
Received: January 14, 2020
Accepted: February 9, 2021
Published: March 2, 2021
Copyright: © 2021 Lee et al. This is an open access
article distributed under the terms of the Creative
Commons Attribution License, which permits
unrestricted use, distribution, and reproduction in
any medium, provided the original author and
source are credited.
Data Availability Statement: All relevant data are
within the manuscript and its Supporting
Information files.
Funding: JL received a Smithsonian Tropical
Research Institute (STRI) short-term fellowship.
Competing interests: The authors have declared
that no competing interests exist.
PLOS ONE | https://doi.org/10.1371/journal.pone.0247551 March 2, 2021
1 / 17
PLOS ONE Host preference of marine symbiotic flatworms
Introduction
In symbiotic interactions [1], the identity of partner species can alter the costs and benefits for
hosts and symbionts, leading to differences in the degree to which the interactions are benefi-
cial, neutral, or antagonistic [2]. Symbionts perform differently and achieve variable fitness
levels in different hosts [3–5]. Also, the performance, survival, and reproductive success of
hosts depend strongly on the species and phenotypes of associated symbionts [6, 7].
Conceptual theories posit that this partner-dependence is central to the evolution of host-
preference in symbioses [8]. Host preference may evolve from host-dependent fitness tradeoffs
that are determined by the relative availability and quality of different host species [8–11]. For
instance, symbionts could gain maximum reproductive success and fitness by associating with
more abundant hosts that they can frequently encounter and search efficiently [10–12]. Alter-
natively, symbionts could choose higher quality hosts in which their adults and offspring can
achieve maximum performance or fitness level [8, 9]. The latter prediction is increasingly sup-
ported by empirical findings that symbionts perform better and achieve higher fitness levels in
their preferred hosts [5, 13, 14].
Several different factors could modulate the success of symbionts in different host species
and contribute to their host preference. These include but are not limited to life-history traits
of symbionts and hosts (e.g., life-stage, reproductive status, and body size) [15, 16], environ-
mental conditions [17], ecological interactions with natural enemies or competing species [16,
18], and host response traits (e.g., behavioral or immunological defense against symbionts)
[11, 19, 20]. Assessing the relative importance of these drivers and their complex interactions
is challenging but vital to understand the evolution and maintenance of host preference.
In this study, we evaluate how host response traits modulate host preference during symbi-
otic interactions. Host responses to symbionts, both adaptive and nonadaptive, could affect
symbionts’ success in different hosts [21]. For example, hosts may variably regulate a symbi-
onts’ performance and abundance with their behavioral or immunological defense, as demon-
strated in several symbiotic relationships [22–24]. Different host performance and survival
upon parasite or pathogen infection may also alter the symbionts’ transmission and fitness
[25]. Such findings lead us to predict that interspecific variations in host response traits would
significantly influence a symbionts’ host choice. However, empirical evaluations of the predic-
tion thus far have been largely restricted to a few study systems (e.g., avian brood parasitism
and parasitoid-insect host interaction) [19, 26, 27].
We examined the host preference of symbiotic marine flatworms (Platyhelminthes: Turbel-
laria: Polycladida) and their varying effects on host organisms. Polyclads are predominantly
free-living [28]. However, several species, mostly belonging to the suborder Acotylea, live as
ecto-symbionts of marine invertebrates, including corals, molluscs, crustaceans, and echino-
derms [28–31]. Symbiotic polyclads are potentially useful models for testing ecological and
evolutionary mechanisms underlying host preference in marine symbionts, owing to their
ability to utilize multiple host species and greater ease of manipulation under a laboratory set-
ting [30, 32].
The flatworm Paraprostatum echinolittorinae is one of several Acotylean polyclads known
to infest marine gastropods (e.g., snails and limpets) [28, 30, 33] and is found along the entire
Pacific Coast of Central America [33]. In coastal Panama, the species is associated with snail
species belonging to four different families (i.e., Tegulidae, Neritidae, Planaxidae, and Cerithii-
dae). The flatworm usually inhabits the host mantle or branchial cavity (Fig 1A). While its eco-
logical effects on snails have not been determined, our initial laboratory and field observations
suggested that snail hosts may respond variably to flatworm infestation. The pilot infestation
trials revealed that not all snails readily accept infesting flatworms, and some species (e.g., P.
PLOS ONE | https://doi.org/10.1371/journal.pone.0247551 March 2, 2021
2 / 17
PLOS ONE Host preference of marine symbiotic flatworms
Fig 1. Symbiotic flatworm Paraprostatum echinolittorinae (Platyhelminthes: Turbellaria: Polycladida) and its
snail hosts. (A) A flatworm inside the mantle cavity of snail host (Tegula pelllisserpentis), as indicated by a white
arrow. Five species of intertidal snails common on the Pacific Coast of Panama including (B) Nerita scabricosta, (C) N.
funiculata, (D) Tegula pelllisserpentis, (E) Cerithium stercusmuscarum, and (F) Planaxis planicostatus.
https://doi.org/10.1371/journal.pone.0247551.g001
planicostatus) will actively reject the worms via mantle contractions and foot withdrawals. We
further observed that, whereas most snails rarely host more than a single or a few worms, some
species will frequently support much higher worm loads (e.g., ~ 39 worms per snail in T.
pellisserpentis).
We hypothesized that snail hosts would differentially respond to flatworm infestation and
that this host response variability would mediate the worm’s host choice. Specifically, we pre-
dicted a) snail hosts would reject and repel infesting flatworms to a varying degree, b) snails
showing a stronger rejection behavior would perform more poorly upon flatworm infestation,
and c) flatworms would prefer hosts that are less likely to reject and be harmed by the infesta-
tion. To test our hypothesis, we first conducted field surveys to quantify flatworm prevalence
and abundance in five common snail hosts (Tegula pellisserpentis, Nerita scabricosta, Nerita
funiculata, Planaxis planicostatus, and Cerithium stercusmuscarum; Fig 1B–1F). We then per-
formed two separate experiments under a laboratory condition, each investigating flatworms’
host choice and snails’ behavioral rejection of infesting worms. Finally, we used a field experi-
ment to assess whether flatworms variably influence the growth and survival of their preferred
and less-preferred snail hosts.
Materials and methods
Flatworm prevalence and infestation density in snail hosts
To examine the prevalence and density of flatworm infestation in different snail species, we
conducted field surveys during January 2012. Snails were collected in the rocky intertidal near
the Smithsonian Tropical Research Institute’s Punta Culebra Nature Center (PCNC; 8˚
54’43.54"N, 79˚31’46.17"W) on Naos Island at the Pacific entrance of the Panama Canal. We
collected T. pellisserpentis (n = 87), N. funiculata (n = 97), and P. planicostatus (n = 99) from
mid-intertidal rocks, and N. scabricosta (n = 101) from high-intertidal rocks. In addition to
these species, we collected 114 individuals of C. stercusmuscarum from high-intertidal tide
pools.
Snails were brought back to the laboratory, where we measured shell length, width, and
height to the nearest 0.01 mm and recorded blotted wet mass (BWM, to 0.0001 g). Live flat-
worms were removed by carefully crushing snail shells and dissecting them from the snail
PLOS ONE | https://doi.org/10.1371/journal.pone.0247551 March 2, 2021
3 / 17
PLOS ONE Host preference of marine symbiotic flatworms
mantle. We counted the number and measured BWM of all flatworms that were found within
each dissected snail. We also approximated the volume of mantle cavity in four snail species
(T. pellisserpentis, N. scabricosta, N. funiculata, and P. planicostatus) by measuring the change
in snail wet mass before and after removing all the water from the mantle (by gently pushing
the snail operculum into the shell). For each species, this procedure was repeated with six dif-
ferent-sized individuals, covering a natural range of body size. Change in water mass (g) was
converted into water volume (cm3; 0.98 cm3 per 1 g seawater) for the approximation of snail
mantle volume.
To quantify the relative abundance of snail species, we conducted additional field surveys
during June 2012 at PCNC. Four transects (40−45 m in length, > 3 m apart) were placed per-
pendicularly to the shoreline, from the high intertidal to the low intertidal zone. Along each of
four transects, we placed two quadrats (1 m2) per each of three intertidal zones (high, mid, and
low; total n = 8 quadrats per zone) and counted all snails found within each quadrat. C. stercus-
muscarum was generally limited to tide pool areas within the high intertidal zone, so we laid
four additional transects (10−12 m in length, > 2 m apart) with two quadrats each across tide
pool areas.
Host selection behaviors in flatworms
We performed a laboratory experiment to test the host selection behaviors of symbiotic flat-
worms. Snails were collected from PCNC between February and May 2012. For larger snail
species (T. pellisserpentis, N. scabricosta), we collected smaller individuals to control for the
potential confounding effect of host body size (Shell length; T. pellisserpentis = 16.7−19.6 mm,
N. scabricosta = 12.6−15.2 mm, N. funiculata = 11.7−14.3 mm, P. planicostatus = 19.3−23.8
mm, C. stercusmuscarum = 18.6−24.1 mm). In the laboratory, we sacrificed multiple individu-
als of each of five snail species to extract live flatworms. Due to the scarcity of flatworm infesta-
tion in P. planicostatus and C. stercusmuscarum, experimental flatworms were all obtained
from T. pellisserpentis, N. scabricosta, and N. funiculata. Flatworms that were collected from
different host species were placed in separate containers until further use. To ensure that all
snails used during the experiment were free of flatworms, we flushed the mantle of all study
snails with Flatworm Exit (Salifert1, diluted to 0.2% with seawater) using a needleless syringe.
This solution is used in commercial aquariums and in experimental studies for removing poly-
clad flatworms and is purportedly non-toxic to other marine invertebrates and fishes [34].
After 30 minutes of the exposure, the snails were transferred to an aerated seawater tank and
held for at least one week prior to experimental trials.
Host-selection trials were conducted in a cylindrical glass bowl (diameter = 25 cm), con-
taining one individual of each of five snail species (T. pellisserpentis, N. funiculata, N. scabri-
costa, P. planicostatus, and C. stercusmuscarum). Prior to each trial, we fixed experimental
snails upside-down to the bottom of the bowl using cyanoacrylate adhesive (Krazy Glue 1) in
order to immobilize them and to better observe whether or not flatworms have entered the
snail cavity. Snails were positioned so that they were equidistant from each other and from the
center of the bowl. Once the glue had dried, we gently flushed the experimental bowl several
times with seawater and waited at least 30 minutes before introducing flatworms. At the onset
of each trial, we introduced 9 flatworms (3 worms collected per snail species; T. pellisserpentis,
N. scabricosta, and N. funiculata) to the center of the bowl and monitored their host choice for
45 minutes. We considered the host was “chosen” when the flatworm, for the first time,
entered the host aperture opening or mantle (irrespective of worms being subsequently
repelled by snail hosts or voluntarily exiting the hosts afterward). We calculated the infestation
probability for each snail species as a proportion of introduced worms that selected the
PLOS ONE | https://doi.org/10.1371/journal.pone.0247551 March 2, 2021
4 / 17
PLOS ONE Host preference of marine symbiotic flatworms
experimental snail within a 45-minute timeframe. This trial was repeated 18 times throughout
the study, with different individuals of snails and flatworms.
Host rejection of flatworm infestation
Our preliminary observations suggested that snails do not actively avoid flatworms (e.g., flee-
ing from approaching worms). However, some snails retracted their foot and contracted their
mantle when flatworms attempted to enter their mantle cavity. This often prevented flatworms
from entering snails’ mantle cavity, suggesting the behavior may be defensive. We experimen-
tally compared the frequency of this rejection behavior among five snail host species. In the
laboratory, we placed an individual snail upside-down (with the aperture upwards) on a glass
petri dish (diameter = 7 cm) and then placed a flatworm directly into the shell aperture. This
approach allowed us to observe host behaviors with greater ease when flatworms entered their
cavity. Flatworm infestation caused a similar behavioral response in snail hosts whether they
be placed in an upside-down or upright position. We recorded the presence or absence of host
rejection behavior over a 3-minute period. We then determined whether experimental worms
have successfully entered the hosts by visually inspecting the shell surface and aperture of snail
subjects. Host and flatworm responses were classified into four categories as follows: a) the
host rejects and repels the worm, b) the host rejects the worm, but the worm successfully
infests the host, c) the host does not reject the worm, and the worm does not infest the host
either, and d) the host does not reject the worm, and the worm successfully infests the host.
For each snail species, we repeated this trial 18 times using different individuals of snails and
flatworms.
Host growth and survival following worm infestations
To examine the effect of flatworm infestation on host growth and survival, we conducted a
field caging experiment with the subset of host species (T. pellisserpentis, N. funiculata, and P.
planicostatus). We chose these snails because they overlap in their natural habitat (i.e., mid-
intertidal zone) and represent host species that flatworms show high (T. pellisserpentis), mod-
erate (N. funiculata), and low (P. planicostatus) preference. Snails were collected at PCNC in
April 2012. We measured snails’ shell dimensions and BWM as above and randomly assigned
them into two groups (n = 18 per treatment group per species): without flatworms and with
flatworms. We also sacrificed multiple individuals of T. pellisserpentis and N. funiculata to
obtain flatworms. To ensure that all snails were free of flatworms before being assigned to dif-
ferent treatment groups, we flushed the mantle of experimental snails with Flatworm Exit solu-
tion as above. The snails that had been assigned to the flatworm treatment group were re-
infested with flatworms by placing haphazardly chosen worms (i.e., irrespective of original
host species) directly on the aperture of each snail. For each snail species, we manipulated flat-
worm density differentially to simulate the natural infestation density (no. worm per snail, T.
pellisserpentis = 8, N. funiculata = 4, and P. planicostatus = 2). The worm densities used here
represented a high density that is less common in the field. Still, because we selectively used
smaller flatworms (0.4−0.8 mg), the combined biomass of flatworms infesting each snail was
well within the range of natural infestation. To track snail shell growths, we marked the margin
of each snail shell aperture with paint covered with a thin layer of superglue. At the beginning
of the experiment, the size of snails did not vary between the two groups for all three species
(SL ± SE [mm]; T. pellisserpentis, without flatworm = 13.99 ± 0.76 mm, with flatworm [8
worms] = 14.02 ± 0.75 mm, t34 = − 0.026, P = 0.980; N. funiculata, without flat-
worm = 11.43 ± 0.13 mm, with flatworm [4 worms] = 11.43 ± 0.14 mm, t34 = − 0.038,
PLOS ONE | https://doi.org/10.1371/journal.pone.0247551 March 2, 2021
5 / 17
PLOS ONE Host preference of marine symbiotic flatworms
P = 0.970; P. planicostatus, without flatworm = 15.70 ± 0.30 mm, with flatworm [2 worms] =
15.69 ± 0.30 mm, t34 = 0.037, P = 0.971).
In the mid-intertidal zone at PCNC, we cleared a 3.5 m × 1.0 m area of rocks to create an
even surface and installed 36 cages, arranging them into a 3 × 12 block formation. We con-
structed each cage (S1 Fig) using a single hexagon-shaped brick (width = 19 cm, height = 13
cm) with a threaded hole (hole diameter = 14.5 cm). We filled the hole with similarly sized
rocks to create interstitial spaces for snails and then covered the top and bottom openings with
a wire mesh (0.75 cm × 0.75 cm). We randomly assigned individual snails of each species irre-
spective of the treatment type (i.e., with or without flatworm) to each of the 36 cages. There-
fore, a single cage held three snail individuals, each of different species. This “cage sharing”
was necessitated to minimize the physical disturbance in our study site. The three species co-
occur naturally, and the density and size ranges of the caged snails were commonly observed
in the field.
We maintained the experiment by cleaning all cages every other day and recorded snail
mortality during each visit. After 52 days, we terminated the experiment and measured shell
growth (i.e., the maximum distance between the newly formed margin of shell aperture and
the outer margin of the painted surface) of all surviving snails. Tissues were extracted from the
shells, and the sex of each individual was determined. Snail tissues were oven-dried at 65˚C for
36 hours to obtain dried tissue mass (DTM) as a more robust measure of growth. We calcu-
lated the growth rate as the proportional increase in DTM. The initial DTM of each experi-
mental snail was estimated using empirically determined relationships between external shell
dimensions and DTM as in Byers [35]. We derived this relationship by measuring the shell
length and weighing the DTM of snails over a range of sizes (n = 30) and regressing the DTM
against the shell length (for all species, r2 > 0.90).
Stable isotope analysis of host and flatworm tissues
To determine whether flatworms directly feed on host tissue, we analyzed the stable isotopic
signatures of snail and flatworm tissues. We sampled tissues from freshly collected T. pellisser-
pentis, N. scabricosta, and N. funiculata and flatworms infesting each snail species (apart from
the studies described above). Due to the difficulty of obtaining a sufficient amount of flatworm
tissues from P. planicostatus and C. stercusmuscarum, we excluded the two snail species from
the study. We sampled the gill and muscle tissues (i.e., mantle and foot) for snails and the
whole body for flatworms. All samples were washed with distilled water, dried at 60˚C for 36
hours, and ground to a fine powder prior to analysis. For each tissue category, we analyzed
four to eight separate samples (ca. 2 mg) using an elemental analyzer (Flash HT) and Delta V
Advantage isotope ratio mass spectrometer (CF-IRMS; Thermo Scientific, Bremen, Germany)
in the Stable Isotope Laboratory of the Smithsonian Tropical Research Institute, Panama. Sta-
ble isotope values are expressed in delta notation as parts per thousands (‰) deviations from
the standard using the following equation:
dΧ ¼ ðRsample=Rstandard (cid:0) 1Þ � 1000
where X is 13C or 15N, and R is the ratio between 13C and 12C or 15N and 14N. For standard ref-
erence materials, we used Pee Dee Belemnite for carbon and atmospheric nitrogen N2 for
nitrogen. Each trophic transfer from prey to consumers is typically represented by the average
increase of δ 15N values by ca. 3.4‰ and δ 13C values by ca. 1‰ [36–38]. In general, δ 15N is
used to determine the difference in trophic level, whereas δ 13C is used to distinguish between
primary carbon sources (e.g., phytoplankton and macroalgae) [36].
PLOS ONE | https://doi.org/10.1371/journal.pone.0247551 March 2, 2021
6 / 17
PLOS ONE Host preference of marine symbiotic flatworms
Statistical analysis
We used a generalized linear model (GLM) with a logit link function (i.e., logistic regression),
to compare a) the prevalence of flatworm infestation (infested vs. non-infested) in different
snail hosts, b) the frequencies of different host response types during the host rejection experi-
ment, and c) the snail mortality (live vs. dead) during the field experiment. We used a GLM
with a quasi-Poisson error distribution for over-dispersed data [39] to compare the density of
flatworm infestation in different snail hosts. We used a linear mixed model (LMM, R package
lme4) to compare the probability of worm infestation for different snail species (square root-
transformed to achieve normality and homogeneity of variance) during the host selection
experiment. We treated snail species as a fixed factor and trial ID as a random factor. The sig-
nificance of snail species as a fixed factor was assessed using Type III ANOVA with Sat-
terthwaite approximations to determine denominator degrees of freedom (R package
lmerTest). Finally, we used a generalized linear model with a Gaussian link function to analyze
snail growth (i.e., shell growth and % increase DTM) during the field experiment. For multiple
comparisons among GLMs and LMM outputs, we used Holms-corrected P values (R package
multcomp). We conducted all analyses using R (version 3.5.1; R Core Team 2018).
Ethics statement
A field collection permit (permit no. DGOMI-PICFC-N˚23) was issued by Panama’s Autori-
dad Nacional del Ambiente, and Autoridad de Recursos Acua´ticos de Panama´. We also fol-
lowed all applicable guidelines for the care and use of animals established by the Smithsonian
Institution during our laboratory and field work.
Results
Flatworm prevalence and infestation density in snail hosts
We dissected 498 snails and recovered a total of 844 individual flatworms. The prevalence of
flatworm infestation in each host species was: T. pellisserpentis = 89%, N. scabricosta = 50%, N.
funiculata = 28%, P. planicostatus = 3%, C. stercusmuscarum = 0.01%; and the mean density of
worm infestation (no. worms per snail ± 1 SE) was: T. pellisserpentis = 5.7 ± 0.6, N. scabri-
costa = 2.6 ± 0.7, N. funiculata = 0.8 ± 0.2, P. planicostatus = 0.03 ± 0.02, C. stercusmus-
carum = 0.01 ± 0.01. Both the prevalence (Logistic regression; χ2 = 354.8, df = 4, P < 0.0001;
Fig 2A) and mean density (quasi-Poisson GLM; F[4,475] = 51.1, P < 0.0001; Fig 2B) of flatworm
infestation varied significantly among snail hosts. Morphometric measurements for each snail
species including shell length, BWM, and approximated cavity volumes are summarized in
Table 1.
Among five snail species, N. funiculata was the most widespread species across the entire
intertidal habitat (found in 66% of quadrats), while T. pellisserpentis and N. scabricosta were
the least frequently observed species (found in 28% of quadrats). When pooling all quadrats,
mean density (no. individual per m2 ± 1 SE) for each snail species was as follows: T. pellisser-
pentis = 0.7 ± 0.2, N. scabricosta = 4.1 ± 1.5, N. funiculata = 23.6 ± 5.1, P. planicosta-
tus = 19.1 ± 4.6, C. stercusmuscarum = 21.0 ± 7.1 (Table 1).
Host selection behaviors in flatworms
There was a significant difference in infestation probability among five snail hosts during host
selection trials (F[4,85] = 16.5, P < 0.0001; Fig 2C). Multiple comparisons of LMM outputs
revealed that flatworms preferred snail species in the order of T. pellisserpentis > N. scabricosta
= N. funiculata > P. planicostatus = C. stercusmuscarum (Fig 2C).
PLOS ONE | https://doi.org/10.1371/journal.pone.0247551 March 2, 2021
7 / 17
PLOS ONE Host preference of marine symbiotic flatworms
Fig 2. Prevalence and density of flatworm infestation in the field and flatworms’ host choice in the laboratory. (A)
Proportion of snail individuals infested by flatworms in the field, (B) box plot of flatworm infestation density (no.
worm per snail) in the field, and (C) box plot of infestation probability (i.e., a proportion of individual worms selecting
a particular host species) measured during the host selection experiment. In box plots, mean values for different snail
groups are represented by red diamond symbols. Different letters indicate a significant difference based on post hoc
multiple comparisons (Holms-corrected P < 0.05).
https://doi.org/10.1371/journal.pone.0247551.g002
PLOS ONE | https://doi.org/10.1371/journal.pone.0247551 March 2, 2021
8 / 17
PLOS ONE Host preference of marine symbiotic flatworms
Table 1. Morphometrics, habitat, and field abundance of snail hosts.
Shell length (mm)
BWM (g)
Approximated mantle volume (cm3)
Snail density (no. individual / m2) Tide pool (n = 8)
High zone (n = 8)
Mid zone (n = 8)
Low zone (n = 8)
Total (n = 32)
T. pellisserpentis
30.35 ± 0.88 mm
14.27 ± 0.82 g
1.03 ± 0.25 cm3
0 ± 0 (0%)
0 ± 0 (0%)
1.1 ± 0.5 (63%)
1.8 ± 0.7 (50%)
0.7 ± 0.2 (28%)
N. funiculata
N. scabricosta
25.87 ± 0.38 mm 11.26 ± 0.25 mm
8.89 ± 0.32 g
0.33 ± 0.02 cm3
0 ± 0 (0%)
11.6 ± 4.5 (75%)
2.6 ± 1.7 (25%)
2.3 ± 2.3 (13%)
4.1 ± 1.5 (28%)
0.72 ± 0.04 g
0.14 ± 0.02 cm3
46.6 ± 11.7 (100%)
8.8 ± 4.6 (50%)
6.8 ± 3.0 (50%)
32.1 ± 11.8 (63%)
23.6 ± 5.1 (66%)
P. planicostatus
18.10 ± 0.38 mm
1.15 ± 0.06 g
0.13 ± 0.02 cm3
0 ± 0 (0%)
23.6 ± 10.2 (50%)
27.1 ± 10.1 (63%)
25.8 ± 9.6 (75%)
19.1 ± 4.6 (47%)
C. stercusmuscarum
20.32 ± 0.40 mm
1.01 ± 0.05 g
-
75.5 ± 16.7 (100%)
6.9 ± 6.9 (13%)
1.5 ± 1.5 (13%)
0 ± 0 (0%)
21.0± 7.1 (31%)
Results summarizing morphometric measurements (i.e., shell length, blotted wet mass, and approximated mantle volume) and field density (no. individual per m2) of
five snail hosts. All surveys were conducted between January and June 2012 at the Smithsonian Tropical Research Institute’s Punta Culebra Nature Center (PCNC),
Panama. For each species, shell length and BWM were measured from 87 to 101 individual snails, and mantle volumes were approximated using 6 different-sized
individuals, covering a natural range of body size. Snail density was measured across four distinct zones within the intertidal habitat (i.e., tide pool, high zone, mid zone,
and low zone). Error values indicate 1 SE, and values inside parentheses indicate percentages of quadrats in which respective snail species were found.
https://doi.org/10.1371/journal.pone.0247551.t001
Host rejection of flatworm infestation
All snail species except N. funiculata showed some degree of behavioral rejection when
exposed to flatworms (Fig 3). Frequencies in which snail hosts rejected and subsequently
repelled flatworms varied among snail species (Logistic regression, χ2 = 74.8, df = 4,
P < 0.0001), being significantly higher in P. planicostatus (50%) and C. stercusmuscarum
(55.6%) in comparisons to T. pellisserpentis (11.1%), N. scabricosta (11.1%), and N. funiculata
(0%) (Holms-corrected P < 0.05 for all comparisons) (Fig 3). Conversely, the probability of
successful infestation, irrespective of host rejection responses, was significantly higher in T.
pellisserpentis (88.9%), N. scabricosta (77.8%) and N. funiculata (94.4%) in comparisons to P.
planicostatus (38.9%) and C. stercusmuscarum (27.8%) (Logistic regression, χ2 = 84.7, df = 4,
P < 0.0001; Holms-corrected P < 0.05 for all comparisons) (Fig 3).
Host growth and survival following worm infestations
Flatworms negatively affected the growth of P. planicostatus, significantly decreasing its DTM
(Linear regression GLM; χ2 = 8567.2, df = 29, P = 0.01, Fig 4A) and marginally decreasing shell
growth (χ2 = 42.2, df = 29, P = 0.08, Fig 4B). Flatworms did not significantly affect the growth
of T. pellisserpentis (Fig 4A and 4B) and marginally increased the DTM of N. funiculata (χ2 =
8753.4, df = 32, P = 0.1, Fig 4A). Flatworm infestation did not affect host mortality in all cases.
In the treatment group, 23%, 17%, and 0% of the surviving T. pellisserpentis (n = 13), N. funicu-
lata (n = 18), and P. planicostatus (n = 15) had at least one worm residing inside mantle cavity,
respectively. No new infestation occurred in the control groups (i.e., snails without flatworm).
Stable isotope analysis of host and flatworm tissues
Flatworm tissues showed limited variation in δ 15N (0.0‰−0.3‰) and δ 13C (0.6‰−1.28‰)
values among different host species categories (Fig 5). Snail gill and muscle tissues had similar
δ 15N values in N. scabricosta and N. funiculata, but these values varied more in T. pellisserpen-
tis (~ 2.5‰). Flatworm δ 15N values were within − 1.6‰ to 1.2‰ of host gill and muscle δ 15N,
whereas flatworm δ 13C values were generally lower (1.0‰−3.6‰) compared to host δ 13C
(Fig 5).
PLOS ONE | https://doi.org/10.1371/journal.pone.0247551 March 2, 2021
9 / 17
PLOS ONE Host preference of marine symbiotic flatworms
Fig 3. Host rejection behaviors and flatworm infestation success. Frequencies of different snail and flatworm responses during experimental infestation.
The responses were classified as: the host rejects and repels the worm (red pink), the host rejects the worm, but the worm successfully infests the host (orange),
the host does not reject the worm, but the worm does not infest the host (olive green), and the host does not reject the worm, and the worm successfully infests
the host (light blue). Snail and flatworm responses were tracked over the allotted time period of 3 minutes using 18 different individuals for each species.
https://doi.org/10.1371/journal.pone.0247551.g003
Discussion
This is the first study to examine the host preference of polyclad flatworms in relation to their
varying effects on host organisms. Flatworms’ host selection mirrored their field prevalence
and infestation density in different snail hosts. However, these patterns were not explained by
the relative availability of different hosts as the preferred species (T. pellisserpentis, N. scabri-
costa, and N. funiculata) were not necessarily more abundant in the field. Therefore, consistent
with previous findings by Fujiwara et al. [30], host preference rather than host availability
appears to drive the differential infestation patterns by symbiotic flatworms. The mechanism
by which flatworms distinguish different hosts is unclear, but it is likely influenced by both
chemical [40] and physical cues [30] from snails.
The preferred and less-preferred snail hosts showed marked differences in their behavioral
defenses and post-infestation performances. P. planicostatus and C. stercusmuscarum fre-
quently rejected and repelled infesting worms, but such responses were reduced or absent in
T. pellisserpentis, N. scabricosta, and N. funiculata. Accordingly, flatworms were 39% to 67%
less successful when they attempted to infest P. planicostatus and C. stercusmuscarum com-
pared to T. pellisserpentis, N. scabricosta, and N. funiculata. As predicted, host species display-
ing stronger rejection behaviors performed more poorly upon infestation. Flatworms
negatively influenced the growth of P. planicostatus in the field but negligibly affected the
growth and survival of T. pellisserpentis and N. funiculata, even though we treated the latter
two hosts with higher worm loads. These findings are concurrent with predictions that the
PLOS ONE | https://doi.org/10.1371/journal.pone.0247551 March 2, 2021
10 / 17
PLOS ONE Host preference of marine symbiotic flatworms
Fig 4. Effect of flatworm infestation on snail growth performances. Box plots of (A) percentage increase in snail dried
tissue mass (DTM) and (B) growth in the shell aperture lip (mm) for Tegula pellisserpentis, Nerita funiculata, and Planaxis
planicostatus with (turquoise box) and without (dark orange box) flatworms during the field experiment. Mean values for
different treatment groups are represented by red diamond symbols. The asterisk (�) indicates a significant difference
(P < 0.05).
https://doi.org/10.1371/journal.pone.0247551.g004
PLOS ONE | https://doi.org/10.1371/journal.pone.0247551 March 2, 2021
11 / 17
PLOS ONE Host preference of marine symbiotic flatworms
Fig 5. Stable isotope analysis of flatworm and snail tissues. Tissue δ 13C and δ 15N of flatworms and their snail host species: Nerita
funiculata (diamond), N. scabricosta (square), and Tegula pellisserpentis (triangle). Different colors represent tissues sampled from
flatworm (emerald green), snail gill (light blue), and snail mantle and foot muscle (purple pink). Sample size was n = 4−8 for each snail
species × tissue category. Error bars indicate 1 SE.
https://doi.org/10.1371/journal.pone.0247551.g005
strength of host counter-defense or resistance may vary depending on the potential costs of
hosting a symbiont [41, 42].
We found no evidence to suggest that flatworms are parasitic. Analysis of host and symbi-
ont δ 15N revealed that flatworms are unlikely to feed on hosts’ soft tissues. Flatworm δ 15N val-
ues were well within > 3.4‰ of host gill and muscle δ 15N (− 1.6‰ to 1.2‰ differences) [36].
Also, consistent with previous observations [33], we did not find visible tissue damage in the
gill and cavity wall of infested snail hosts. These findings suggest flatworms are likely commen-
sal, and their deleterious effects on P. planicostatus may be a byproduct of the infestation.
Because flatworms inhabit the host branchial chamber, they might compete for oxygen with
hosts or physically interfere with the hosts’ oxygen uptake, as hypothesized for other gill-dwell-
ing symbionts [43]. P. planicostatus may have been especially prone to such effects due to its
relatively small body size and mantle cavity volume (Table 1) or other anatomical differences
with T. pellisserpentis and N. funiculata (e.g., gill complex morphology) [44].
Our findings support the hypothesis that flatworm’s host preference could be mediated by
the behavioral defense and post-infestation performance of snail hosts. Selecting more defen-
sive P. planicostatus and C. stercusmuscarum would decrease the probability of successful
infestation when flatworms seek new hosts or leave and re-enter snail hosts during foraging
[45]. Flatworms had no direct impact on the survival of P. planicostatus. However, by nega-
tively affecting its performance, flatworm infestation could render the host more susceptible to
predation or physical stress. Host mortality would be detrimental to flatworms in many cases,
as they may easily succumb to desiccation and predation outside snail hosts [30]. Conversely,
selecting less defensive T. pellisserpentis, N. scabricosta, and N. funiculata would increase the
PLOS ONE | https://doi.org/10.1371/journal.pone.0247551 March 2, 2021
12 / 17
PLOS ONE Host preference of marine symbiotic flatworms
probability of successful worm infestation and re-infestation. These hosts could also provide
more stable habitats for flatworms, as they are less likely to be harmed by the infestation.
Differences in host qualities may have further affected the flatworm’s host choice. Among
the preferred hosts, flatworms showed the strongest preference for T. pellisserpentis, which had
the largest body size and cavity volume (Table 1). Indeed, it is not uncommon for symbionts
to prefer and perform better in larger hosts that offer more generous habitat space and food
resources [46, 47]. Flatworms not only attained higher densities but also 2−5 times larger body
sizes in T. pellisserpentis than in other hosts (S1 Table). It is unclear whether flatworms enjoy
higher reproductive success in the host species. However, a previous study found a strong cor-
relation between the maternal body size of symbiotic polyclads and their reproductive output
[45].
Our results corroborate the growing evidence that interspecific variations in host responses
could modulate host preferences in symbioses. In a cleaning mutualism between branchiob-
dellid worms and freshwater crayfish hosts, the symbionts may prefer host species that are less
likely to behaviorally inhibit their colonization (e.g., via grooming activities) [22]. Similarly, in
parasitic interactions between parasitoids and insect hosts [26] or freshwater trematodes and
amphibians [19], the symbionts may prefer host species with less effective immune defenses.
Moreover, some parasites and pathogens may avoid host species in which the infection kills
the host too quickly and impede cross-host transmission [20]. Symbionts commonly interact
with hosts that are phylogenetically distinct or possess different morphological, behavioral,
and physiological traits [48–50], which may contribute to varying host response patterns.
Therefore, we predict that host response variabilities would play broad and significant roles in
the evolution and maintenance of host preference.
Previous studies considered symbiotic polyclads as commensal [28, 45] or parasites that
prey on host tissues or embryos [51, 52]. However, their effects on host organisms have rarely
been tested using empirical approaches. The results from our field experiment and stable iso-
tope analysis agree with Fujiwara et al. [45] that flatworms live as commensal of intertidal
snails. Flatworms may use snails as mobile homes and feed on similar food sources as the hosts
(e.g., algae and other periphytic materials) [53], while it is unclear whether they obtain food
mainly inside or outside their hosts. Future work should more broadly examine the symbiotic
nature and evolutionary history between these organisms. Indeed, similar species associations
have been reported across different continents (e.g., America, Asia, and Africa) and from vari-
ous families of gastropods and polyclads [28, 33].
Our study is not without caveats. In the host selection experiment, we exclusively used flat-
worms collected from three commonly infested host species, which could have biased our find-
ings. Symbionts can imprint to host chemical or visual cues upon colonization, which may
influence their future host choice [3, 54]. Hence, a proper test of host preference should either
control the source hosts (e.g., introducing the same number of worms collected from each host
species) or use “virgin” symbionts with no prior interactions with hosts. Neither was feasible
in our study due to difficulties in obtaining unsettled flatworm larvae and collecting sufficient
adult worms from all snail species. However, we note that flatworms were 90% to 130% more
likely to choose T. pellisserpentis than N. scabricosta and N. funiculata, although our study
used an equal number of worms collected from these hosts. Thus, our results maintain that a
pre-existing host preference would govern the flatworm’s host choice.
In conclusion, we demonstrate that host response traits could modulate host choices in
symbiotic interactions. As predicted, flatworms’ host preference was positively associated with
the reduced host behavioral defense and the hosts’ ability to support flatworm populations
without being harmed. We encourage future studies to more explicitly consider host response
PLOS ONE | https://doi.org/10.1371/journal.pone.0247551 March 2, 2021
13 / 17
PLOS ONE Host preference of marine symbiotic flatworms
variability when examining multi-species symbioses. These tests will improve our theoretical
and empirical understanding of complex biological processes shaping host preference.
Supporting information
S1 Fig. Experimental cages used in the field experiment. The cages (containing experimental
snails) deployed at the rocky intertidal habitats within Smithsonian Tropical Research Insti-
tute’s Punta Culebra Nature Center (PCNC), Panama.
(TIF)
S1 Table. Flatworm biomass in different snail hosts. Mean blotted wet mass (BWM; mg) of
total and individual flatworms collected from different snail species (2012 field surveys). Error
values indicate 1 SE and values inside parentheses represent sample sizes (i.e., number of snails
dissected).
(DOCX)
S1 Data. Supplementary data (flatworm infestation data, host choice and behavioral
response data, host growth and survival data, and stable isotope data).
(XLSX)
Acknowledgments
We thank Sebastien Romero, Silvia Gu¨tschow, Carmen Schloeder, Sara Lipshutz, and Jara
Javier for helpful assistance with our field/laboratory experiment, and Carlos Bonilla, Jose
Ignacio Sanchez for the maintenance of our experimental tank. We are also very grateful to Jeb
Byers for providing guidance on gastropod morphometric analysis.
Author Contributions
Conceptualization: Juhyung Lee, Mark E. Torchin.
Formal analysis: Juhyung Lee, Timothy M. Davidson, Mark E. Torchin.
Funding acquisition: Juhyung Lee.
Investigation: Juhyung Lee.
Methodology: Juhyung Lee, Timothy M. Davidson, Mark E. Torchin.
Resources: Mark E. Torchin.
Supervision: Timothy M. Davidson, Mark E. Torchin.
Visualization: Juhyung Lee.
Writing – original draft: Juhyung Lee.
Writing – review & editing: Juhyung Lee, Timothy M. Davidson, Mark E. Torchin.
References
1. De Bary A. Die Erscheinung der Symbiose: Verlag von Karl J. Tru¨bner Strassburg; 1879.
2. Bronstein JL. Conditional outcomes in mutualistic interactions. Trends Ecol Evol. 1994; 9(6):214–217.
https://doi.org/10.1016/0169-5347(94)90246-1 PMID: 21236825
3. Ocampo EH, Nuñez JD, Cledo´ n M, Baeza JA. Host-specific reproductive benefits, host selection
behavior and host use pattern of the pinnotherid crab Calyptraeotheres garthi. J Exp Mar Biol Ecol.
2012; 429:36–46.
PLOS ONE | https://doi.org/10.1371/journal.pone.0247551 March 2, 2021
14 / 17
PLOS ONE Host preference of marine symbiotic flatworms
4. Poore AG, Steinberg PD. Preference–performance relationships and effects of host plant choice in an
herbivorous marine amphipod. Ecol Monogr. 1999; 69(4):443–464.
5. Gripenberg S, Mayhew PJ, Parnell M, Roslin T. A meta-analysis of preference–performance relation-
ships in phytophagous insects. Ecol Lett. 2010; 13(3):383–393. https://doi.org/10.1111/j.1461-0248.
2009.01433.x PMID: 20100245
6. Parker BJ, Hrček J, McLean AH, Godfray HCJ. Genotype specificity among hosts, pathogens, and ben-
eficial microbes influences the strength of symbiont-mediated protection. Evolution. 2017; 71(5):1222–
1231. https://doi.org/10.1111/evo.13216 PMID: 28252804
7. Stat M, Morris E, Gates RD. Functional diversity in coral–dinoflagellate symbiosis. Proc Natl Acad Sci
USA. 2008; 105(27):9256–9261. https://doi.org/10.1073/pnas.0801328105 PMID: 18591663
8.
Thompson JN. Evolutionary ecology of the relationship between oviposition preference and perfor-
mance of offspring in phytophagous insects. Entomol Exp Appl. 1988; 47(1):3–14.
9. Roff DA. The evolution of life histories: theory and analysis. New York: Chapman & Hall; 1992.
10. Mason PA, Wilkes SR, Lill JT, Singer MS. Abundance trumps quality: bi-trophic performance and para-
sitism risk fail to explain host use in the fall webworm. Oikos. 2011; 120(10):1509–1518.
11. Reichard M, Polačik M, Tarkan AS, Spence R, Gaygusuz O¨ , Ercan E, et al. The bitterling–mussel
coevolutionary relationship in areas of recent and ancient sympatry. Evolution. 2010; 64(10):3047–
3056. https://doi.org/10.1111/j.1558-5646.2010.01032.x PMID: 20482611
12. Mayhew PJ. Adaptive patterns of host-plant selection by phytophagous insects. Oikos. 1997:417–428.
13. Harvey JA, Gols R, Snaas H, Malcicka M, Visser B. Host preference and offspring performance are
linked in three congeneric hyperparasitoid species. Ecol Entomol. 2015; 40(2):114–122.
14. Nguyen HTT, Tran ANT, Ha LTL, Ngo DN, Dang BT, Geffen AJ. Host choice and fitness of anemonefish
Amphiprion ocellaris (Perciformes: Pomacentridae) living with host anemones (Anthozoa: Actiniaria) in
captive conditions. J Fish Biol. 2019; 94(6):937–947. https://doi.org/10.1111/jfb.13910 PMID:
30676646
15. Burkett-Cadena ND, McClure CJ, Ligon RA, Graham SP, Guyer C, Hill GE, et al. Host reproductive phe-
nology drives seasonal patterns of host use in mosquitoes. PLoS One. 2011; 6(3):e17681. https://doi.
org/10.1371/journal.pone.0017681 PMID: 21408172
16. Huebner LK, Dailey B, Titus BM, Khalaf M, Chadwick NE. Host preference and habitat segregation
among Red Sea anemonefish: effects of sea anemone traits and fish life stages. Mar Ecol Prog Ser.
2012; 464:1–15.
17.
18.
Lower SS, Kirshenbaum S, Orians CM. Preference and performance of a willow-feeding leaf beetle: soil
nutrient and flooding effects on host quality. Oecologia. 2003; 136(3):402–411. https://doi.org/10.1007/
s00442-003-1278-4 PMID: 12756526
Ishihara M, Ohgushi T. Enemy-free space? Host preference and larval performance of a willow leaf bee-
tle. Popul Ecol. 2008; 50(1):35–43.
19. Sears BF, Schlunk AD, Rohr JR. Do parasitic trematode cercariae demonstrate a preference for sus-
ceptible host species? PLoS One. 2012; 7(12):e51012. https://doi.org/10.1371/journal.pone.0051012
PMID: 23272084
20.
Lievens EJ, Perreau J, Agnew P, Michalakis Y, Lenormand T. Decomposing parasite fitness reveals
the basis of specialization in a two-host, two-parasite system. Evol Lett. 2018; 2(4):390–405. https://doi.
org/10.1002/evl3.65 PMID: 30283690
21. Råberg L, Graham AL, Read AF. Decomposing health: tolerance and resistance to parasites in animals.
Philos Trans R Soc B. 2009; 364(1513):37–49. https://doi.org/10.1098/rstb.2008.0184 PMID:
18926971
22.
Farrell KJ, Creed RP, Brown BL. Preventing overexploitation in a mutualism: partner regulation in the
crayfish–branchiobdellid symbiosis. Oecologia. 2014; 174(2):501–510. https://doi.org/10.1007/s00442-
013-2780-y PMID: 24072440
23. Rothstein SI. A model system for coevolution: avian brood parasitism. Annu Rev Ecol Evol Syst. 1990;
21(1):481–508.
24. Wang D, Zhou L, Wang Q, Ding J. Plant Chemistry Determines Host Preference and Performance of an
Invasive Insect. Front Plant Sci. 2020; 11:1725. https://doi.org/10.3389/fpls.2020.594663 PMID:
33240306
25.
Foreman-Wykert AK, Miller JF. Hypervirulence and pathogen fitness. Trends Microbiol. 2003; 11
(3):105–108. https://doi.org/10.1016/s0966-842x(03)00007-6 PMID: 12648937
26. Dubuffet A, A´ lvarez CIR, Drezen JM, Van Alphen JJ, Poirie´ M. Do parasitoid preferences for different
host species match virulence? Physiol Entomol. 2006; 31(2):170–177.
PLOS ONE | https://doi.org/10.1371/journal.pone.0247551 March 2, 2021
15 / 17
PLOS ONE Host preference of marine symbiotic flatworms
27. Grim T, Samasˇ P, Moska´t C, Kleven O, Honza M, Moksnes A, et al. Constraints on host choice: why do
parasitic birds rarely exploit some common potential hosts? J Anim Ecol. 2011; 80(3):508–518. https://
doi.org/10.1111/j.1365-2656.2010.01798.x PMID: 21244420
28. Parasitism Jennings J. and commensalism in the Turbellaria. Adv Parasitol. 1971; 9:1–32.
29. Doignon G, Artois T, Deheyn D. Discoplana malagasensis sp. nov., a new turbellarian (Platyhelminthes:
Polycladida: Leptoplanidae) symbiotic in an ophiuroid (Echinodermata), with a cladistic analysis of the
Discoplana/Euplana species. Zool Sci. 2003; 20(3):357–370.
30.
Fujiwara Y, Urabe J, Takeda S. Host preference of a symbiotic flatworm in relation to the ecology of litto-
ral snails. Mar Biol. 2014; 161(8):1873–1882.
31. Bo M, Bavestrello G, Di Muzio G, Canese S, Betti F. First record of a symbiotic relationship between a
polyclad and a black coral with description of Anthoplana antipathellae gen. et sp. nov. (Acotylea, Noto-
planidae). Mar Biodivers. 2019; 49(6):2549–2570.
32. Hume BC, D’Angelo C, Cunnington A, Smith EG, Wiedenmann J. The corallivorous flatworm Amakusa-
plana acroporae: an invasive species threat to coral reefs? Coral Reefs. 2014; 33(1):267–272.
33.
Faubel A, Sluys R, Reid DG. A new genus and species of polyclad flatworm found in the mantle cavities
of gastropod molluscs in the high-intertidal zone of the Pacific coast of Central America. J Mar Biol
Assoc. 2007; 87(2):429–434.
34. Bolton DK, Clark GF, Johnston EL. Novel in situ predator exclusion method reveals the relative effects
of macro and mesopredators on sessile invertebrates in the field. J Exp Mar Biol. 2019; 513:13–20.
35. Byers JE. Competition between two estuarine snails: implications for invasions of exotic species. Ecol-
ogy. 2000; 81(5):1225–1239.
36. Pinnegar J, Campbell N, Polunin N. Unusual stable isotope fractionation patterns observed for fish host
−parasite trophic relationships. J Fish Biol. 2001; 59(3):494–503.
37. DeNiro MJ, Epstein S. Influence of diet on the distribution of carbon isotopes in animals. Geochim Cos-
mochim Acta. 1978; 42(5):495–506.
38. Vander Zanden MJ, Cabana G, Rasmussen JB. Comparing trophic position of freshwater fish calcu-
lated using stable nitrogen isotope ratios (δ 15N) and literature dietary data. Can J Fish Aquat Sci.
1997; 54(5):1142–1158.
39. Bolker BM, Brooks ME, Clark CJ, Geange SW, Poulsen JR, Stevens MHH, et al. Generalized linear
mixed models: a practical guide for ecology and evolution. Trends Ecol Evol. 2009; 24(3):127–135.
https://doi.org/10.1016/j.tree.2008.10.008 PMID: 19185386
40.
Iwai N, Sugiura S, Chiba S. Predation impacts of the invasive flatworm Platydemus manokwari on eggs
and hatchlings of land snails. J Molluscan Stud. 2010; 76(3):275–278.
41. Von Beeren C, Maruyama M, Hashim R, Witte V. Differential host defense against multiple parasites in
ants. Evol Ecol. 2011; 25(2):259–276.
42. Rouchet R, Smith C, Liu H, Methling C, Douda K, Yu D, et al. Avoidance of host resistance in the ovipo-
sition-site preferences of rose bitterling. Evol Ecol. 2017; 31(5):769–783.
43. Huxham M, Maitland D, Mocogni M. Respiration rates in Littorina littorea infected with three species of
digenean parasite. J Mar Biol Assoc. 2001; 81(2):351–352.
44. Gilinsky NL. Does archaeogastropod respiration fail in turbid water? Paleobiology. 1984:459–468.
45.
Fujiwara Y, Iwata T, Urabe J, Takeda S. Life history traits and ecological conditions influencing the sym-
biotic relationship between the flatworm Stylochoplana pusilla and host snail Monodonta labio. J Mar
Biol Assoc. 2016; 96(3):667–672.
46. Hart BL, Hart LA, Mooring MS. Differential foraging of oxpeckers on impala in comparison with sympat-
ric antelope species. Afr J Ecol. 1990; 28(3):240–249.
47. Kleven O, Moksnes A, Røskaft E, Honza M. Host species affects the growth rate of cuckoo (Cuculus
canorus) chicks. Behav Ecol Sociobiol. 1999; 47(1–2):41–46.
48. Brooks WR, Rittschof D. Chemical detection and host selection by the symbiotic crab Porcellana
sayana. Invertebr Biol. 1995:180–185.
49. Nagel L, Grutter A. Host preference and specialization in Gnathia sp., a common parasitic isopod of
coral reef fishes. J Fish Biol. 2007; 70(2):497–508.
50. Detwiler J, Janovy J Jr. The role of phylogeny and ecology in experimental host specificity: Insights from
a eugregarine–host system. J Parasitol. 2008; 94(1):7–12. https://doi.org/10.1645/GE-1308.1 PMID:
18372615
51.
Jokiel PL, Townsley SJ. Biology of the polyclad Prosthiostomum (Prosthiostomum) sp., a new coral par-
asite from Hawaii. Pac Sci. 1974; 28:361–373.
PLOS ONE | https://doi.org/10.1371/journal.pone.0247551 March 2, 2021
16 / 17
PLOS ONE Host preference of marine symbiotic flatworms
52.
53.
Lytwyn M, McDermott J. Incidence, reproduction and feeding of Stylochus zebra, a polyclad turbellarian
symbiont of hermit crabs. Mar Biol. 1976; 38(4):365–372.
Levings SC, Garrity SD. Diel and tidal movement of two co-occurring neritid snails; differences in graz-
ing patterns on a tropical rocky shore. J Exp Mar Biol. 1983; 67(3):261–278.
54. Hayes JA. Prey preference in a Caribbean corallivore, Coralliophlla abbreviata (Lamarck)(Gastropoda,
Coralliophilidae). Bull Mar Sci. 1990; 47(2):557–560.
PLOS ONE | https://doi.org/10.1371/journal.pone.0247551 March 2, 2021
17 / 17
PLOS ONE |
10.1371_journal.pone.0245923 | RESEARCH ARTICLE
Prevalence of unintended pregnancy and its
associated factors: Evidence from six south
Asian countries
Alamgir Sarder1, Sheikh Mohammed Shariful IslamID
Benojir AhammedID
1*
2, Maniruzzaman1, Ashis Talukder1,
1 Statistics Discipline, Khulna University, Khulna, Bangladesh, 2 Institute for Physical Activity and Nutrition
(IPAN), School of Exercise and Nutrition Sciences, Deakin University, Melbourne, Australia
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
* benojirstat@ku.ac.bd
Abstract
Aim
OPEN ACCESS
Citation: Sarder A, Islam SMS, Maniruzzaman ,
Talukder A, Ahammed B (2021) Prevalence of
unintended pregnancy and its associated factors:
Evidence from six south Asian countries. PLoS
ONE 16(2): e0245923. https://doi.org/10.1371/
journal.pone.0245923
Editor: Eugene Kofuor Maafo Darteh, University of
Cape Coast, GHANA
Received: August 18, 2020
Accepted: January 11, 2021
Published: February 1, 2021
Copyright: © 2021 Sarder et al. This is an open
access article distributed under the terms of the
Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: Data are available
from the online: https://dhsprogram.com/data/
available-datasets.cfm.
Funding: The author(s) received no specific
funding for this work. The funders had no role in
study design, data collection and analysis, decision
to publish, or preparation of the manuscript.
Competing interests: The authors have declared
that no competing interests exist.
Unintended pregnancy is a significant public health concern in South Asian countries
because of its negative association with the socioeconomic and health outcomes for both
children and mothers. The present study aimed to explore the prevalence of unintended
pregnancy and explore its determinants among women of reproductive age in six South
Asian countries.
Methods
Nationwide latest demography and health survey data from six South Asian countries,
including Bangladesh (2014), Pakistan (2017–2018), Nepal (2016), Afghanistan (2015),
Maldives (2016–2017) and India (2015–2016) were pooled for the present study. Multivari-
ate analysis was performed to explore the association between unintended pregnancy and
its associated factors.
Results
Amongst the total women (n = 41,689), overall, 19.1% pregnancies were reported as unin-
tended (ranging from 11.9% in India to 28.4% in Bangladesh). The logistic regression model
showed that younger women (15–19 years) had 1.42 times higher chance of unintended
pregnancies. The odds of unintended pregnancies was 1.24 times higher for poorest
women and 1.19 times higher for poorer women. Further, urban women (aOR = 0.70, 95%
CI = 0.50–0.80), women having no children (aOR = 0.10, 95% CI = 0.09–0.12), smaller (�4)
family (aOR = 0.72, 95% CI = 0.67–0.78), those who intent to use contraceptive (aOR =
0.72, 95% CI = 0.60–0.86), currently living with partner (aOR = 0.90, 95% CI = 0.81–0.99),
first cohabitation in teenage (�19 years) (aOR = 0.85, 95% CI = 0.78–0.92) were less likely
to report unintended pregnancies.
PLOS ONE | https://doi.org/10.1371/journal.pone.0245923 February 1, 2021
1 / 13
PLOS ONE Prevalence of unintended pregnancy and its associated factors
Conclusions
This study has showed that women’s age, wealth index, place of residence, number of chil-
dren, family size, the intention of contraceptive use, living with a partner, and first cohabita-
tion age are essential determinants of unintended pregnancy. These factors should be
considered when trying to reduce unintended pregnancy in six South Asian countries. How-
ever, there is a need to improve health education, counselling, skills-building, sex education,
modern contraceptive use and its access in this region. Intervention programs regarding
reproductive health and policies are warranted to reduce rates of unintended pregnancy in
South Asian countries.
Introduction
Unintended pregnancy (UP) is a global public health problem and can affect any sexually
active women. Unintended pregnancy is defined when women did not desire to become preg-
nant at that time or at any time in the future [1,2]. Unintended pregnancies have a negative
impact on women’s personal life, their families, and society. Globally, 74 million women had
unintended pregnancies in low and middle-income countries, and every year around 25 mil-
lion unsafe abortions and 47 thousand maternal deaths occur. A remarkable extent of unin-
tended pregnancies brings about premature birth and another unfavorable pregnancy result
[3,4]. It is the consequences of a wide range of factor including non-use of contraception, con-
traceptive discontinuation, contraceptive defeat, inconsistency and inaccurate use of contra-
ception, and lack of awareness regarding family planning [5]. Globally, unintended pregnancy
had been decreased in high-income countries compared with low- and middle-income coun-
tries [6]. The unintended pregnancy remained a global public health problem and substantially
higher in low- and middle-income regions. Approximately 53.8 million unintended pregnan-
cies occurred each year in Asia. About 5.4% of women aged 15–44 years were suffered from
unintended pregnancies during the year 2010–2014 in Asia [7]. South Asian countries, which
includes the sub-Himalayan and neighboring countries of Bangladesh, Pakistan, Nepal,
Afghanistan, Maldives, India, Sri Lanka and Bhutan are low- and middle-income countries
[8]. India, the prevalence of unintended pregnancy was not demonstrated a lot of variety or to
a narrow dosage had been stale during last one decade [9]. Around one-fourth of the women
in India was reported that their pregnancy was unintended in each of the three rounds of
National Family Health Surveys (NFHS) [10]. In 2012, a study revealed that about 29% of preg-
nancies were unintended and the frequency of unintended pregnancy was higher among
older, poor, and less educated women of rural area in Bangladesh [11]. However, the unin-
tended pregnancy was decreased 4% from the year 1993 to 2011 in Bangladesh [12]. Among
South Asian countries, Nepal has recorded the highest (50%) unintended pregnancy [13]. A
hospital-basedcross-sectional survey in Pakistan reported that 38.2% of pregnancies were
unintended [5]. Another study in Afghanistan showed that 36.9% of unplanned pregnancies
were common among women [14]. In addition, poor knowledge in contraceptive use, low
socioeconomic status, contraceptive failure, sexual violence, shortage in contraceptive supply,
unmarried status, age, religion, number of children, residence, wealth index, the intension of
contraceptive use and first cohabitation age influences the unintended pregnancies [6,7]. How-
ever, few studies found that inconsistent and incorrect condom use, contraceptive failure, and
lack of knowledge on emergency contraception were treated as the reason for current
PLOS ONE | https://doi.org/10.1371/journal.pone.0245923 February 1, 2021
2 / 13
PLOS ONE Prevalence of unintended pregnancy and its associated factors
unintended pregnancy in some other countries [15–17].The unsafe abortion, maternal death,
malnutrition, mental illness and vertical transmission of HIV to children were the most severe
consequences of unintended pregnancy [18–20]. These had negative impacts on women’s
quality of life, and increase the economic cost of families as well as increases the mental stress
of the women that were the causes of maternal and neonatal morbidity and mortality [5].
Most of the previous studies were conducted only on the prevalence and determinants of
unintended pregnancy using cross-sectional data of individual countries in South Asia. A
country-specific investigation was conducted in Bangladesh [11,12], Pakistan [5,21], Nepal
[13], Maldives [22] and India [23,24] to investigate the predictors of unintended pregnancy.
Previously an experiment was conducted in 2015 to assess the association between intimate
partner violence (IPV) and unintended pregnancy in South Asia [8]. To the best of our knowl-
edge, there are no studies investigating the phenomenon across South Asian countries. There-
fore, the main objectives of this study were to (i) identify the prevalence of unintended
pregnancy among women in six South Asian countries aged 15–49 years; (ii) examine the asso-
ciation between sociodemographic and behavioral characteristics and unintended pregnancy;
and (iii) examine the impact of sociodemographic and behavioral factors on unintended preg-
nancy. This analysis will utilize pooled data drawn from the demographic and health survey
(DHS) to address the objectives of this study. The results of this study will be useful to develop
programs to improve reproductive health program to reduce the unintended pregnancy in
South Asian countries.
Methods
Ethical approval
This study was performed by secondary data analysis of data collected from MEASURE
demography health and survey (DHS). The DHS surveys obtained ethical clearance from the
Ethics Committee of ORC Macro Inc., and the Ethics Boards of Ministry of Health of the con-
sidered six South Asian countries. This survey confirmed international ethical standards and
during each of the surveys, either written or verbal consent, was provided by the women. The
details of ethics approval of six South Asian countries is described elsewhere [25–30].
Data sources
DHS is a large scale nationally representative cross-sectional survey of households that collects
data on population, health, HIV, and nutrition through more than 400 surveys in over 90
countries [31]. DHS in different countries collects data of all individual ever-married women
aged 15 to 49 years in the household using personal interviews by trained interviewers and a
well-designed questionnaire. The survey used a two-stage stratified sampling technique, sam-
pling within administrative areas [32]. The detail of the study design was described elsewhere
[33]. This study was a cross-sectional study, used a pooled dataset from current Demographic
and Health Surveys (DHS). Data were limited to the most recent standard DHS from six coun-
tries in South Asia. The six countries are Bangladesh (2014), Pakistan (2017–2018), Nepal
(2016), Afghanistan (2015), Maldives (2016–2017) and India (2015–2016). The principal rea-
son for choosing these six countries is that data on some of the sociodemographic, and behav-
ioral characteristics of interest were only available in these six countries. These six countries
had current DHS information and all the factors of interest for this study. Some South Asian
nations were excluded from this study because of lacking related data. The analysis in this
study is limited to currently ever-married women aged 15 to 49 years (N = 42,578). Further
889 women were excluded due to missing data and the analysis included 41,689 observations.
All dataset is available to the public online [34].
PLOS ONE | https://doi.org/10.1371/journal.pone.0245923 February 1, 2021
3 / 13
PLOS ONE Prevalence of unintended pregnancy and its associated factors
Dependent variable
The primary outcome variable of the study was pregnancy intentions status. “Pregnancy inten-
tions” was accounted for the dependent variables for this study which emerged from the
inquiry concerning whether women intended their present pregnancy or not. The DHS docu-
ment has a question for women as "current pregnancy wanted" and it has three responses,
namely: ‘then’, ‘later’ and ‘not at all’. For simplicity we have coded these three responses as fol-
lows: ‘then’ for intended (0); ‘later or not at all’ for ‘unintended (1)’ based on the definition of
unintended pregnancy [35].
Explanatory variables
A set of categorical explanatory variables was selected to fit the two individual regression
model. At the model-1, we considered the only country of origin as an independent variable.
Based on the several studies twelve explanatory variables were considered as independent vari-
ables in model-2, namely age (15–19, 20–24, 25–29, 30–34, 35–39, 40–49), residence (urban,
rural), educational level (no institutional education, primary, secondary & higher), wealth
index (poorest, poorer, middle, richer, richest), religion (Islam, Hinduism, others), household
head (male, female), the number of children (no children, 1–2 children, 3 or more), family size
(� 4, > 4), the intention of contraceptive use (intends to use, does not intend to use), current
residence with a partner (living with a partner, staying elsewhere), first cohabitation age (� 19
years, > 19 years) and country of origin (Bangladesh, Pakistan, Nepal, Afghanistan, Maldives,
India). These variables were selected as there was a significant association with pregnancy
intention and had been reported as predictors of unintended pregnancy [6,7,36–38].
Statistical analysis
Simple descriptive analysis and bivariate and multivariate statistical analyses were performed
in this study. Descriptive analysis was commenced to describe the frequency and percentage
distribution. Bivariate analysis was used to examine the association between the unintended
pregnancy and selected independent variables. Multivariable binary logistic regression analy-
ses were conducted to assess the effect of different sociodemographic and behavioral factors
on unintended pregnancy. Logistic regression (in Model 1) examines the effect of the country
on unintended pregnancy. Multivariable logistic regression (in Model 2) examines the effect of
selected sociodemographic and behavioral factors on unintended pregnancy. The results of the
logistic regression analysis were presented using adjusted odds ratios (aORs) along with 95%
confidence intervals. The high-risk factors were identified based on p-value (p<0.05). All the
statistical analyses were performed in IBM SPSS v20 (SPSS Inc, Chicago, IL).
Results
Background characteristics of participants
Fig 1 presented the distribution of unintended pregnancies of six South Asia countries. Over-
all, 19.1% of unintended pregnancies were found from six South Asian countries. The preva-
lence of unintended pregnancies was ranged from 11.9% in India to 28.4% in Bangladesh. The
background characteristics of the study participants and bivariate analysis of sociodemo-
graphic and behavioral variables with unintended pregnancy in six South Asian countries
were presented in Table 1. Considering the women age, most of the respondents were from
20–24 years (39.3%) age group. The study women were predominantly rural, with only one-
fourth of the women residing in an urban area. Whereas, the percentage of the study women
was gradually decreased with increasing the educational level. Nearly half of the women were
PLOS ONE | https://doi.org/10.1371/journal.pone.0245923 February 1, 2021
4 / 13
PLOS ONE Prevalence of unintended pregnancy and its associated factors
Fig 1. Prevalence of unintended pregnancy in South Asian countries.
https://doi.org/10.1371/journal.pone.0245923.g001
from the poor and poorest family whereas one-third of the women were from rich and richest
family. Majority of the study women were intended to use contraceptive (75.9%), and they had
1–2 children (43.7%). It was observed that the proportion of unintended pregnancy increases
with increasing age and number of children of the women. Subsequently, it was also observed
that the rate of unintended pregnancy decreases with increasing the educational level and
wealth index of the women. In this study, the rate of unintended pregnancy also higher among
Hindu women (15.0%), women’s having >4 family members (14.4%), intent to contraceptive
use (14.0%), life partner living/staying elsewhere (13.8%) and first cohabitation age �19 years
(14.2%). Except for household head of the respondents, all the selected explanatory variables
were found significantly (p < 0.05) associated with an unintended pregnancy.
Multivariate logistic regression analysis of unintended pregnancy
Table 2 presented the results of the logistic regression analysis to shows the effect of unin-
tended pregnancy among six South Asian countries women by background characteristics.
Unadjusted effect of the country on unintended pregnancies was measured in Model 1.
Among the six south Asian countries, Bangladesh (OR = 2.94, 95% CI = 2.56–3.37) had signifi-
cant highest odds of unintended pregnancy among women followed by Nepal (OR = 2.70, 95%
CI = 2.22–3.29), Maldives (OR = 1.92, 95% CI = 1.44–2.56) and Pakistan (OR = 1.21, 95% 95%
CI = 1.06–1.40) compared with India as a reference country. In Model 2, after adjusting the
socio-demographic and behavioral characteristics, an almost similar result was found for the
country of the region. The results indicated that Bangladesh (aOR = 2.97, 95% CI = 2.57–3.45),
Nepal (aOR = 2.58, 95% CI = 2.09–3.19), and Maldives (aOR = 2.42, 95% CI = 1.78–3.28) had
significantly higher odds of unintended pregnancy among the women, but Pakistan
(aOR = 0.79, 95% CI = 0.68–0.92) and Afghanistan (aOR = 0.57, 95% CI = 0.51–0.63) had sig-
nificantly lower odds of unintended pregnancy compared to women of India.Women aged
15–19 years had higher odds (aOR = 1.42, 95% CI = 1.13–1.78) and all others age categories
had lower odds of unintended pregnancies compared with women aged 40–49 years. Urban
women were less likely to commit to unintended pregnancy, as compared with rural women
(aOR = 0.70, 95% CI = 0.50–0.80). Highest odds occurred among Poorest (aOR = 1.24, 95%
PLOS ONE | https://doi.org/10.1371/journal.pone.0245923 February 1, 2021
5 / 13
PLOS ONE Table 1. Background characteristics of the study participants and bivariate analysis of sociodemographic and behavioral variables with unintended pregnancy in
six South Asian countries.
Prevalence of unintended pregnancy and its associated factors
Variables
Total
Age (in years)
15–19
20–24
25–29
30–34
35–39
40–49
Residence
Urban
Rural
Educational level
No education
Primary
Secondary & higher
Wealth index
Poorest
Poorer
Middle
Richer
Richest
Religion
Islam
Hinduism
Others
Household head
Male
Female
Number of children
No children
1–2 children
3 or more children
Family size
�4
>4
Intention of contraceptive use
Intends to use
Does not intend to use
Current residence with partner
Living with partner
Staying elsewhere
First cohabitation age
�19 years
>19 years
N
41,689
4861
16375
12356
5112
2228
757
10206
31483
16833
16628
8228
9662
9914
8568
7359
6186
32868
5330
3491
37302
4387
14636
18201
8852
13111
28578
31640
10049
37500
4189
25095
16594
https://doi.org/10.1371/journal.pone.0245923.t001
%
Intended, n (%)
Unintended, n (%)
Pregnancy status
Chi-square value (p-value)
11.7
39.3
29.6
12.3
5.3
1.8
24.5
75.5
40.4
39.9
19.7
23.2
23.8
20.6
17.7
14.8
78.8
12.8
8.4
89.5
10.5
35.1
43.7
21.2
31.4
68.6
75.9
24.1
90.0
10.0
60.2
39.8
4350(89.5)
14515(88.6)
10775(87.2)
4355(85.2)
1817(81.6)
554(73.2)
8964(87.8)
27402(87.0)
14304(85.0)
14604(87.8)
7458(90.6)
8178(84.6)
8556(86.3)
7546(88.1)
6508(88.4)
5578(90.2)
28647(87.2)
4530(85.0)
3189(91.3)
32508(87.1)
3858(87.9)
13836(94.5)
15687(86.2)
6843(77.3)
11904(90.8)
24462(85.6)
27207(86.0)
9159(91.1)
32753(87.3)
3613(86.2)
21529(85.8)
14837(89.4)
511(10.5)
1860(11.4)
1581(12.8)
757(14.8)
411(18.4)
203(26.8)
1242(12.2)
4081(13.0)
2529(15.0)
2024(12.2)
770(9.4)
1484(15.4)
1358(13.7)
1022(11.9)
851(11.6)
608(9.8)
4221(12.8)
800(15.0)
302(8.7)
4794(12.9)
529(12.1)
800(5.5)
2514(13.8)
2009(22.7)
1207(9.2)
4116(14.4)
4433(14.0)
890(8.9)
4747(12.7)
576(13.8)
3566(14.2)
1757(10.6)
269.2 (<0.001)
4.4 (0.019)
168.1 (<0.001)
128.9 (<0.001)
77.3 (<0.001)
2.2 (0.071)
1501.7 (<0.001)
217.9 (<0.001)
181.9 (<0.001)
4.0 (<0.024)
117.6 (<0.001)
PLOS ONE | https://doi.org/10.1371/journal.pone.0245923 February 1, 2021
6 / 13
PLOS ONE Table 2. Multivariate logistic regression analyses showing the effect of unintended pregnancy among six south Asian countries women by background
characteristics.
Prevalence of unintended pregnancy and its associated factors
Variables
Country
Bangladesh
Pakistan
Nepal
Afghanistan
Maldives
India
Age (in years)
15–19
20–24
25–29
30–34
35–39
40–49
Residence
Urban
Rural
Educational level
No institutional education
Primary
Secondary & higher
Wealth index
Poorest
Poorer
Middle
Richer
Richest
Religion
Islam
Hinduism
Others
Household head
Male
Female
Number of children
No children
1–2 children
3 or more children
Family size
�4
>4
Intention of contraceptive use
Intends to use
Does not intend to use
Current residence with partner
Living with partner
Sample size
Model I OR [95% CI]
p-value
Model II aOR [95% CI]
p-value
1076
1729
519
6260
286
31819
2.94[2.56–3.37]
1.21[1.06–1.40]
2.70[2.22–3.29]
1.06[0.97–1.15]
1.92[1.44–2.56]
Ref
<0.001
0.006
<0.001
0.189
<0.001
2.97[2.57–3.45]
0.79[0.68–0.92]
2.58[2.09–3.19]
0.57[0.51–0.63]
2.42[1.78–3.28]
Ref
1.42[1.13–1.78]
0.89[0.74–1.08]
0.63[0.52–0.75]
0.55[0.45–0.66]
0.64[0.52–0.78]
Ref
<0.001
0.002
<0.001
<0.001
<0.001
0.002
0.248
<0.001
<0.001
<0.001
0.70[0.50–0.80]
0.034
Ref
0.93[0.83–1.04]
0.97[0.86–1.06]
Ref
1.24[1.09–1.42]
1.19[1.05–1.35]
1.09[0.98–1.23]
1.10[0.97–1.23]
Ref
1.52[1.34–1.73]
1.80[1.52–2.04]
Ref
0.183
0.473
0.001
0.006
0.154
0.132
<0.001
<0.001
1.05[0.95–1.17]
0.352
Ref
0.10[0.09–0.12]
0.40[0.36–0.44]
Ref
<0.001
<0.001
0.72[0.67–0.78]
<0.001
Ref
0.72[0.60–0.86]
<0.001
Ref
0.90[0.81–0.99]
0.037
(Continued )
PLOS ONE | https://doi.org/10.1371/journal.pone.0245923 February 1, 2021
7 / 13
PLOS ONE Prevalence of unintended pregnancy and its associated factors
Table 2. (Continued)
Variables
Staying elsewhere
First cohabitation age
�19 years
>19 years
Sample size
Model I OR [95% CI]
p-value
Model II aOR [95% CI]
p-value
Ref
0.85[0.78–0.92]
<0.001
Ref
https://doi.org/10.1371/journal.pone.0245923.t002
CI = 1.09–1.42) and poorer (aOR = 1.19, 95% CI = 1.05–1.35) women respectively, compared
to richest.Women having religious faith in Islam (aOR = 1.52, 95% CI = 1.34–1.73) and Hin-
duism (aOR = 1.80, 95% CI = 1.52–2.04) had higher odds of unintended pregnancies compare
with other religious women. The women had no children (aOR = 0.10, 95% CI = 0.09–0.12)
and 1–2 children (aOR = 0.40, 95% CI = 0.36–0.44) reported that they had lower odds of unin-
tended pregnancy, as compared with the women who had at least three children. It was also
interesting to find that small family (�4) were less likely to have unintended pregnancy
(aOR = 0.72, 95% CI = 0.67–0.78) than large family (>4). Compared to the women does not
intend to use a contraceptive; the odds of unintended pregnancies were 0.72 times lower for
the women intend to use contraceptive (aOR = 0.72, 95% CI = 0.60–0.86). The women who
lived with her partner recorded the lower odds of unintended pregnancy compared with wom-
en’s partner staying elsewhere (aOR = 0.90, 95% CI = 0.81–0.99). First cohabitation age �19
years of women had fewer likelihoods of unintended pregnancy (aOR = 0.85, 95% CI = 0.78–
0.92), as compared to their counterpart.
Discussion
This study is a large-scale study specifically designed to estimate the prevalence and associated
factors of unintended pregnancy in six South Asian countries. Since most of the countries in
this sub-continent were fallen under lower- and middle-income countries [8], they contribute
a significant portion in unintended pregnancy and most of which end up in life furtive induced
abortions [4]. The unintended pregnancy ranged between 11.9% in India and 28.4% in Bangla-
desh, whereas the overall prevalence is 19.1% among the six South Asian countries. This indi-
cates a larger number of unintended pregnancy for South Asia than previously reported [4].
But the rate of unintended pregnancy was still low in South Asia compared with developed
regions with some significant variation among the South Asian countries [4]. A study from
Bangladesh using 2011 BDHS data found the prevalence of unintended pregnancy was approx-
imately 28% among the women [12]. This result is supported by our study using BDHS 2014
dataset. Our study showed increased rate of unintended pregnancy among Indian women
compared to a previous study [23].
The multivariate analysis revealed that Bangladeshi women have higher chances of
experiencing unintended pregnancies, compared to women in India followed by Nepal,
Maldives and Pakistan. The possible reason may be most people lives in village territory,
and they are not much aware of family planning. Also,unavailability and less knowledge
about contraceptive use, and inadequate health care service are responsible behind unin-
tended pregnancy [12]. Our study also found that women aged 15–19 years were more
likely to experience unintended pregnancy compared to women in other age categories. A
research conducted in Pakistan found that women aged below 20 years had the highest risk
of unintended pregnancy [5]. In general, younger women have higher fertility, higher fre-
quency of sexual intercourse, feel shy to take advice about family planning from relatives or
family care organizations, have higher contraceptive failure relative to older women and
PLOS ONE | https://doi.org/10.1371/journal.pone.0245923 February 1, 2021
8 / 13
PLOS ONE Prevalence of unintended pregnancy and its associated factors
misunderstanding about contraceptive use, which contributes to unintended pregnancies
[5,39,40]. The findings of this study are opposite to a study conducted in sub-Saharan
Africa [6] and one in the rural area in Bangladesh [41]. The prevalence and odds of unin-
tended pregnancy of urban women were lower than rural among the six South Asian coun-
tries. This evidence coincides with other studies [4,5]. This difference may be due to the
difference in sociodemographic characteristics of the people in these regions.
Mostly urban women are well educated and have better knowledge of emergency con-
traceptive use and family planning. The opportunity of attending family planning-related pro-
grams is higher for urban women. The prevalence of unintended pregnancy is lower for urban
women, as they are likely to bear a small number of children [42]. The cohabitation rate of
urban women is more moderate than rural, as most of the urban men and women are engaged
with outdoor work compared than rural women [43]. However, a previous study in Pakistan
reported no statistically significant association between unintended pregnancy and wealth
index [5]. But several authors also reported that the wealth index of women was related with
unintended pregnancy [41,44,45] and this study likewise found that women in the lowest
wealth quintile experienced higher odds of unintended pregnancy than women with the high-
est wealth quintile [6]. An opposite result was found in sub-Saharan Africa [6]. Rural women
are more like to have lower wealth index [46] and their knowledge about emergency con-
traceptive use is also little [47] which may increase the prevalence rate of unintended preg-
nancy [48]. Notably, this study found most of the women in South Asia were Muslims with
higher unintended pregnancy experience like in Sub-Saharan Africa [6]. Overall, Hindu
women had the highest risk of unintended pregnancy in South Asia than other religions, and a
coincide similar findings were reported in India [49]. In previous studies, unintended preg-
nancy was more common in Muslim women compared with non-Muslims. One of the main
reasons was that Muslim women’s activities were restricted than some other religions [50] and
also they were likely to accept pregnancy as “given by Allah”. This conception has no existence
in this modern era among Muslim women. So, further research is required about the religious
prospect that contributes to this higher prevalence or odds of unintended pregnancy among
women globally, considering a limited investigation.
Similar to previous studies [5,8,49], this study confirmed the likelihood of higher unin-
tended pregnancy among women with more children. Small family size obstacles the rate of
unintended pregnancy. A few studies confirmed that the likelihood of experiencing unin-
tended pregnancy among women who had more children is higher [51,52]. Women who
intended to use contraceptive were less likely to experience an unintended pregnancy. A few
studies conducted in sub-Sharan African and central India supported these findings [6,53].
Intended contraceptive use greatly influence to reduce the rate of unintended pregnancies, as
any contraceptive can prevent the pregnancy.
We noted that the likelihood of unintended pregnancy was lower among the women who
lived with their partner or husband. The relationship between living with partner or husband
and unintended pregnancy of women has not been explained in previous studies. Further
research is warranted to investigate the causes of the relationship between living with partner
or husband and unintended pregnancy studies in sub-Saharan Africa and throughout the
world report almost similar finding to this study [6].
Strengths and limitations
The strength of this study is the use of data from national representative surveys of six South
Asian countries and the pooled-analysis. Nevertheless, this study has some limitations; for
example, the data does not allow the foundation of causality of unintended pregnancy in south
PLOS ONE | https://doi.org/10.1371/journal.pone.0245923 February 1, 2021
9 / 13
PLOS ONE Prevalence of unintended pregnancy and its associated factors
Asia. Other limitations of this study were the very nature of the subject and method it
employed, as observed in all DHS programs. Besides, the variables used in the analysis made it
difficult, linking their effect with the outcome variable.
Conclusions
In South Asia, the prevalence of unintended pregnancy was higher among women having a
larger number of children (3 or more) and family member (more than 4), aged, non-educated,
Hindu, and poorest women. The determinates of unintended pregnancy are age, residence,
wealth index, religion, number of children, family size, intension to contraceptive use, current
resident with a partner and first cohabitation age. The current dynamics of unintended preg-
nancy in South Asia can be reversely changed by national-level family planning, and maternal
well being polices, where particular interventions to poor women, rural women and early mar-
riage couple. There is a need to encourage women to use contraceptive methods and delay
pregnancies. The Government and health authorities of the considered six South Asian coun-
tries should ensure promoting family planning programs and making contraceptives widely
available for women.
Acknowledgments
The authors are grateful to MEASURE DHS (Demography and Health Surveys) for making
data available for this study by the request of the corresponding author.
Author Contributions
Conceptualization: Alamgir Sarder, Benojir Ahammed.
Data curation: Alamgir Sarder.
Formal analysis: Alamgir Sarder.
Investigation: Sheikh Mohammed Shariful Islam.
Methodology: Alamgir Sarder, Maniruzzaman.
Software: Alamgir Sarder, Maniruzzaman.
Supervision: Benojir Ahammed.
Visualization: Ashis Talukder.
Writing – original draft: Alamgir Sarder.
Writing – review & editing: Sheikh Mohammed Shariful Islam, Maniruzzaman, Ashis Taluk-
der, Benojir Ahammed.
References
1. Kassahun EA, Zeleke LB, Dessie AA, Gersa BG, Oumer HI, Derseh HA, et al. Factors associated with
unintended pregnancy among women attending antenatal care in Maichew Town, Northern Ethiopia,
2017. BMC Res Notes. 2019; 12: 1–6. https://doi.org/10.1186/s13104-018-4038-6 PMID: 30602384
2. Reproductive Health. Centers for Disease Control and Prevention. Available: https://www.cdc.gov/
reproductivehealth/contraception/unintendedpregnancy/index.htm. Access at: September 12, 2019.
3. Amin Shokravi F, Howden Chapman P, Peyman N. A comparison study: Risk factors of unplanned
pregnancies in a group of Iranian and New Zealander women. Eur J Sci Res. 2009; 26: 108–121.
4. Bearak J, Popinchalk A, Alkema L, Sedgh G. Global, regional, and subregional trends in unintended
pregnancy and its outcomes from 1990 to 2014: estimates from a Bayesian hierarchical model. Lancet
Glob Heal. 2018; 6: e380–e389. https://doi.org/10.1016/S2214-109X(18)30029-9 PMID: 29519649
PLOS ONE | https://doi.org/10.1371/journal.pone.0245923 February 1, 2021
10 / 13
PLOS ONE Prevalence of unintended pregnancy and its associated factors
5. Habib MA, Raynes-Greenow C, Nausheen S, Soofi SB, Sajid M, Bhutta ZA, et al. Prevalence and deter-
minants of unintended pregnancies amongst women attending antenatal clinics in Pakistan. BMC Preg-
nancy Childbirth. 2017; 17: 1–10. https://doi.org/10.1186/s12884-016-1183-6 PMID: 28049520
6. Ameyaw ID, Edward Kwabena, Eugene Budu, Francis Sambah, Linus Baatiema Appiah F, Seidu ID,
et al. Prevalence and determinants of unintended pregnancy in sub-Saharan Africa: A multi-country
analysis of demographic and health surveys. PLoS One. 2020; 24: 187–205. https://doi.org/10.29063/
ajrh2020/v24i2.18
7. Darroch JE, Sedgh G, Ball H. Contraceptive Technologies: Responding to Women ‘ s Needs. Guttma-
cher Inst. 2011; 1–51.
8. Raj A, Mcdougal L. Associations of intimate partner violence with unintended pregnancy and pre-preg-
nancy contraceptive use in South Asia. Contraception. 2015; 91: 456–463. https://doi.org/10.1016/j.
contraception.2015.03.008 PMID: 25769441
9.
Indian Institute of Population Science. National Family Health Survey(NFHS-II) 1998–99 Report. 2000;
539. Available: http://www.rchiips.org/nfhs/pub_nfhs-2.shtml.
10. Health NF. India. 2005.
11. Noor F, Rahman M, Rob U, Bellows B. Unintended pregnancy among rural women in Bangladesh. Int Q
Community Health Educ. 2011; 32: 101–113. https://doi.org/10.2190/IQ.32.2.b PMID: 23000458
12. Huda A, Sarker K, Azmi AJ, Reichenbach L. Understanding Unintended Pregnancy in Bangladesh:
Country Profile Report. STEP UP Res Rep. 2013; 76. Available: http://www.popcouncil.org/uploads/
pdfs/2014STEPUP_BangladeshCountryProfile_icddrb.pdf.
13. Puri M, Singh S, Sundaram A, Hussain R, Tamang A, Crowell M, et al. Abortion Incidence and Unin-
tended Pregnancy in Nepal. 2017; 42: 197–209. https://doi.org/10.1363/42e2116 PMID: 28825899
14.
Todd CS, Nasir A, Raza Stanekzai M, Scott PT, Strathdee SA, Botros BA, et al. Contraceptive utilization
and pregnancy termination among female sex workers in Afghanistan. J Women’s Heal. 2010; 19:
2057–2062. https://doi.org/10.1089/jwh.2010.1947 PMID: 20879869
15. Ajayi AI, Nwokocha EE, Adeniyi OV, Ter Goon D, Akpan W. Unplanned pregnancy-risks and use of
emergency contraception: a survey of two Nigerian Universities. BMC Health Serv Res. 2017; 17: 1–8.
https://doi.org/10.1186/s12913-016-1943-z PMID: 28049468
16. Chanda MM, Ortblad KF, Mwale M, Chongo S, Kanchele C, Kamungoma N, et al. Contraceptive use
and unplanned pregnancy among female sex workers in Zambia. Contraception. 2017; 96: 196–202.
https://doi.org/10.1016/j.contraception.2017.07.003 PMID: 28711643
17. Christofides NJ, Jewkes RK, Dunkle KL, McCarty F, Shai NJ, Nduna M, et al. Risk factors for unplanned
and unwanted teenage pregnancies occurring over two years of follow-up among a cohort of young
South African women. Glob Health Action. 2015; 8. https://doi.org/10.3402/gha.v7.23719 PMID:
25150027
18. Baschieri A, Machiyama K, Floyd S, Dube A, Molesworth A, Chihana M, et al. Unintended Childbearing
and Child Growth in Northern Malawi. Matern Child Health J. 2017; 21: 467–474. https://doi.org/10.
1007/s10995-016-2124-8 PMID: 27491527
19. Central Statistical Agency, ORC Macro. Ethiopia Demographic and Health Survey 2005. Heal San Fr.
2006; [ 446]. Available: http://www.measuredhs.com/pubs/pdf/FR179/FR179[23June2011].pdf.
20. Claridge AM, Chaviano CL. Consideration of Abortion in Pregnancy: Demographic Characteristics,
Mental Health, and Protective Factors. Women Heal. 2013; 53: 777–794. https://doi.org/10.1080/
03630242.2013.831018 PMID: 24215272
21. Sathar Z, Singh S, Rashida G, Shah Z, Niazi R. Induced Abortions and Unintended Pregnancies in
Pakistan. Stud Fam Plann. 2014; 45: 471–491. https://doi.org/10.1111/j.1728-4465.2014.00004.x
PMID: 25469930
22. Nagase T, Kunii O, Wakai S, Khaleel A. Obstacles to modern contraceptive use among married women
in southern urban Maldives. Contraception. 2003; 68: 125–134. https://doi.org/10.1016/s0010-7824(03)
00113-6 PMID: 12954525
23. Singh S, Shekhar C, Acharya R, Moore AM, Stillman M, Pradhan MR, et al. The incidence of abortion
and unintended pregnancy in India, 2015. Lancet Glob Heal. 2018; 6: e111–e120. https://doi.org/10.
1016/S2214-109X(17)30453-9 PMID: 29241602
24. Singh S, Hussain R, Shekhar C, Acharya R, Moore AM, Stillman M, et al. Abortion & unintended preg-
nancy in six indian States: Findings and Implications for Policies and Progeams. Guttmacher Inst. 2018.
25. National Institute of Population Research and Training (NIPORT), Mitra and Associates, and ICF Inter-
national. 2016. Bangladesh Demographic and Health Survey 2014. Dhaka, Bangladesh, and Rockville,
Maryland, USA: NIPORT, Mitra and Associates, and ICF Inter.
26. Central Statistics Organization (CSO), Ministry of Public Health (MoPH), and ICF. 2017. Afghanistan
Demographic and Health Survey 2015. Kabul, Afghanistan: Central Statistics Organization.
PLOS ONE | https://doi.org/10.1371/journal.pone.0245923 February 1, 2021
11 / 13
PLOS ONE Prevalence of unintended pregnancy and its associated factors
27.
International Institute for Population Sciences (IIPS) and ICF. 2017. National Family Health Survey
(NFHS-4), 2015–16: India. Mumbai: IIPS.
28. National Institute of Population Studies (NIPS) [Pakistan] and ICF. 2019. Pakistan Demographic and
Health Survey 2017–18. Islamabad, Pakistan, and Rockville, Maryland, USA: NIPS and ICF.
29. Ministry of Health (MOH) [Maldives] and ICF. 2018. Maldives Demographic and Health Survey 2016–
17. Male´, Maldives, and Rockville, Maryland, USA: MOH and ICF.
30. Ministry of Health, Nepal; New ERA; and ICF. 2017. Nepal Demographic and Health Survey 2016.
Kathmandu, Nepal: Ministry of Health, Nepal.
31. Croft Trevor N., Marshall Aileen M. J., Allen Courtney K. et al. Guide to DHS Statistics. Rockville, Mary-
land, USA ICF. 2018; 22–51.
32. Aliaga A, Ren R. Optimal Sample Sizes for Two-stage Cluster Sampling in Demographic and Health
Surveys. 2006; 18.
33. Corsi DJ, Neuman M, Finlay JE, Subramanian S V. Demographic and health surveys: A profile. Int J
Epidemiol. 2012; 41: 1602–1613. https://doi.org/10.1093/ije/dys184 PMID: 23148108
34.
The DHS Program. Demographic and Health Surveys. Available: https://dhsprogram.com/what-we-do/
survey-Types/dHs.cfm.
35. Bishwajit G, Tang S, Yaya S, Feng Z. Unmet need for contraception and its association with unintended
pregnancy in Bangladesh. BMC Pregnancy Childbirth. 2017; 17: 1–9. https://doi.org/10.1186/s12884-
016-1183-6 PMID: 28049520
36. Maxwell L, Devries K, Zionts D, Alhusen JL, Campbell J. Estimating the effect of intimate partner vio-
lence on women’s use of contraception: A systematic review and meta-analysis. PLoS One. 2015; 10:
1–25. https://doi.org/10.1371/journal.pone.0118234 PMID: 25693056
37. Miller E, Decker MR, McCauley HL, Tancredi DJ, Levenson RR, Waldman J, et al. Pregnancy coercion,
intimate partner violence and unintended pregnancy. Contraception. 2010; 81: 316–322. https://doi.org/
10.1016/j.contraception.2009.12.004 PMID: 20227548
38. Nyarko SH. Unintended Pregnancy among Pregnant Women in Ghana: Prevalence and Predictors. J
Pregnancy. 2019; 2019. https://doi.org/10.1155/2019/2920491 PMID: 30834145
39.
Irani L, Ilene S. Speizer J-CF. Couple Characteristics and Contraceptive Use among Women and their
Partners in Urban Kenya. Internaational Perspect Sex Reprod Heal. 2015; 40: 1–26. https://doi.org/10.
1363/4001114.Couple
40. Exavery A, Kante´ AM, Njozi M, Tani K, Doctor H V., Hingora A, et al. Predictors of mistimed, and
unwanted pregnancies among women of childbearing age in Rufiji, Kilombero, and Ulanga districts of
Tanzania. Reprod Health. 2014; 11: 1–9. https://doi.org/10.1186/1742-4755-11-1 PMID: 24383405
41. Kamal M, Islam A. Prevalence and socioeconomic correlates of unintented pregnancy among women
in rural Bangladesh. Salud Publica Mex. 2011; 53: 108–115. https://doi.org/10.1590/s0036-
36342011000200003 PMID: 21537801
42. Khademi S, Cooke MS. Comparing the attitudes of urban and rural Iranian women toward menopause.
Maturitas. 2003; 46: 113–121. https://doi.org/10.1016/s0378-5122(03)00182-8 PMID: 14559382
43. Williams L, Kabamalan M, Ogena N. Cohabitation in the Philippines: Attitudes and behaviors among
young women and men. J Marriage Fam. 2007; 69: 1244–1256. https://doi.org/10.1111/j.1741-3737.
2007.00444.x
44. Sedgh G, Bankole A, Oye-Adeniran B, Adewole IF, Singh S, Hussain R. Unwanted pregnancy and
associated factors among Nigerian women. Int Fam Plan Perspect. 2006; 32: 175–184. https://doi.org/
10.1363/3217506 PMID: 17237014
45. Rashid M, Shifa N. Mistimed and Unwanted Pregnancies in Bangladesh: Trends and Determinants.
PAA Annu Meet. 2007. Available: http://paa2007.princeton.edu/papers/70486.
46. Mahmud S, Shah NM, Becker S. Measurement of Women’s Empowerment in Rural Bangladesh. World
Dev. 2012; 40: 610–619. https://doi.org/10.1016/j.worlddev.2011.08.003 PMID: 23637468
47.
Tabassum A, Manj YN, Gunjial TR, Nazir S. Perceptions of rural women about contraceptive usage in
District Khushab, Punjab. J Pak Med Assoc. 2016; 66: 1577–1581. https://doi.org/10.1016/j.
ssresearch.2016.01.005 PMID: 28179693
48. Muanda MF, Ndongo GP, Messina LJ, Bertrand JT. Barriers to modern contraceptive use in rural areas
in DRC. Cult Heal Sex. 2017; 19: 1011–1023. https://doi.org/10.1080/13691058.2017.1286690 PMID:
28276915
49. Begum S, Dwivedi SN, Pandey A, Mittal S. Association between domestic violence and unintended
pregnancies in india: Findings from the National Family Health Survey-2 data. Natl Med J India. 2010;
23: 198–200. PMID: 21192511
PLOS ONE | https://doi.org/10.1371/journal.pone.0245923 February 1, 2021
12 / 13
PLOS ONE Prevalence of unintended pregnancy and its associated factors
50. Kuhn R. Routes to low mortality in poor countries revisited. Popul Dev Rev. 2010; 36: 655–692. https://
doi.org/10.1111/j.1728-4457.2010.00353.x PMID: 21174865
51. Geda NR, Lako TK. A population based study on unintended pregnancy among married women in a dis-
trict in Southern Ethiopia. J Geogr Reg Plan. 2011; 4: 417–427. Available: http://www.
academicjournals.org/JGRP.
52. Stephenson R, Koenig MA, Acharya R, Roy TK. Domestic violence, contraceptive use, and unwanted
pregnancy in Rural India. Stud Fam Plann. 2008; 39: 177–186. https://doi.org/10.1111/j.1728-4465.
2008.165.x PMID: 18853639
53. Roy TK, Ram F, Nangia P, Saha U, Khan N. Can women’s childbearing and contraceptive intentions
predict contraceptive demand? Findings from a longitudinal study in Central India. Int Fam Plan Per-
spect. 2003; 29: 25–31. https://doi.org/10.1363/ifpp.29.025.03 PMID: 12709309
PLOS ONE | https://doi.org/10.1371/journal.pone.0245923 February 1, 2021
13 / 13
PLOS ONE |
10.1371_journal.pone.0250260 | RESEARCH ARTICLE
Poisson noisy image restoration via
overlapping group sparse and nonconvex
second-order total variation priors
Kyongson Jon1,2, Jun Liu1, Xiaoguang Lv3, Wensheng Zhu1*
1 Key Laboratory for Applied Statistics of MOE, School of Mathematics and Statistics, Northeast Normal
University, Changchun, P.R. China, 2 Faculty of Mathematics, Kim Il Sung University, Pyongyang, D.P.R. of
Korea, 3 School of Science, Jiangsu Ocean University, Lianyungang, Jiangsu, P.R. China
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
* wszhu@nenu.edu.cn
Abstract
OPEN ACCESS
Citation: Jon K, Liu J, Lv X, Zhu W (2021) Poisson
noisy image restoration via overlapping group
sparse and nonconvex second-order total variation
priors. PLoS ONE 16(4): e0250260. https://doi.org/
10.1371/journal.pone.0250260
Editor: Li Zeng, Chongqing University, CHINA
Received: November 6, 2020
Accepted: April 1, 2021
Published: April 20, 2021
Copyright: © 2021 Jon et al. This is an open access
article distributed under the terms of the Creative
Commons Attribution License, which permits
unrestricted use, distribution, and reproduction in
any medium, provided the original author and
source are credited.
Data Availability Statement: All relevant data are
within the manuscript and its Supporting
information files.
Funding: This work is supported in part by the
National Natural Science Foundation of China
(11771072, 61806024); the Science and
Technology Development Plan of Jilin Province
(20191008004TC, 20180520026JH); Fundamental
Research Funds for the Central Universities
(2412020FZ023); the Nature Science Foundation of
Jiangsu Province (BK20181483).
Competing interests: The authors have declared
that no competing interests exist.
The restoration of the Poisson noisy images is an essential task in many imaging applica-
tions due to the uncertainty of the number of discrete particles incident on the image sensor.
In this paper, we consider utilizing a hybrid regularizer for Poisson noisy image restoration.
The proposed regularizer, which combines the overlapping group sparse (OGS) total varia-
tion with the high-order nonconvex total variation, can alleviate the staircase artifacts while
preserving the original sharp edges. We use the framework of the alternating direction
method of multipliers to design an efficient minimization algorithm for the proposed model.
Since the objective function is the sum of the non-quadratic log-likelihood and nonconvex
nondifferentiable regularizer, we propose to solve the intractable subproblems by the majori-
zation-minimization (MM) method and the iteratively reweighted least squares (IRLS) algo-
rithm, respectively. Numerical experiments show the efficiency of the proposed method for
Poissonian image restoration including denoising and deblurring.
1 Introduction
In many real applications, during the image recording, the measurement of light inevitably
leads to the uncertainty of striking particles on the image sensor. In other words, the finite
number of electrons or photons carrying energy in an image sensor may cause statistical fluc-
tuations, which are usually modeled as the Poisson distribution [1–3]. As another degradation
factor, image formation may involve undesirable blurrings such as motion or out-of-focus.
By rewriting the concerned images into column-major vectorized form, we can regard the
observed image g 2 Rn�n as a realization of Poisson random vector with expected value Hf + b,
where H is a n2 × n2 convolution matrix corresponding to the point spreading function (PSF)
which models blur effects, f 2 Rn�n is the original image, and b 2 Rn�n is a nonnegative con-
stant background [4–6].
Poissonian image restoration calls for applying an inverse procedure to approximate f from
an observation g degraded by blur and Poisson noise. A general option to tackle this problem
is to use the maximum a posteriori (MAP) estimation from Bayesian perspective. Taking the
PLOS ONE | https://doi.org/10.1371/journal.pone.0250260 April 20, 2021
1 / 20
PLOS ONE Hybrid regularizer for image deblurring under poisson noise
Poisson statistics of noise into account, we can write the conditional distribution of the
observed data g as
pðgjf Þ ¼
Yn
i;j¼1
e(cid:0) ðHf þbÞi;j ðHf þ bÞgi;j
gi;j!
i;j
;
ð1Þ
where (�)i,j indicates a vector element corresponding to position (i, j) hereafter. On the premise
that we adopt a Gibbs prior [7–9] of the form
pðf Þ / e(cid:0) �ðf Þ=l
ð2Þ
for f, the use of Bayes’ rule and Stirling’s approximation leads to the following minimization
problem [10, 11]:
min
f
flDKLðHf jjgÞ þ �ðf Þg;
ð3Þ
where λ > 0 is a regularization parameter, ϕ(f) is a nonnegative given function (often called
the regularizer), and DKL(Hf||g) denotes the generalized Kullback-Leibler (KL) divergence of
Hf from g as shown below:
DKLðHf jjgÞ ≔
"
Xn
i;j¼1
ðHf þ bÞi;j (cid:0) gi;j þ gi;jlog
#
:
gi;j
ðHf þ bÞi;j
ð4Þ
2 (usually, Γ is an identity matrix or difference matrix) is probably one of
Since the efficiency of the restoration model hinges on the choice of image regularizer,
numerous authors have proposed different regularization methods. Tikhonov regularization
of form �ðf Þ ¼ kGf k2
the most classical methods. In the literatures, several methods have been proposed to effi-
ciently solve Eq (3) with the Tikhonov regularizer, including the scaled gradient projection
method [12], the split-gradient method [13], the projected Newton-conjugate gradient
(PNCG) method [2], the quasi-Newton projection method [14], the hybrid gradient projec-
tion-reduced Newton method [15, 16] and the scaled gradient projection-type method [17].
Although the KL divergence and Tikhonov regularizer are well-known to be convex and the
unique solution to Eq (3) is guaranteed with inexpensive computational cost, it tends to over-
smooth the important details in restored images.
Another classical method, the total variation (TV) regularizer ϕ(f) = kfkTV (see Section 2),
has been widely accepted since the noisy signals have larger TV than the original signal [18].
Bardsley et al. [19] proved that minimizing the Poisson likelihood function in conjunction
with TV regularization is well-posed, and they also verified its convergence by a nonnegatively
constrained and projected quasi-Newton minimization algorithm. Landi et al. [20] also
adopted the TV regularizer for denoising the medical images collected by the photon-counting
devices and extended the PNCG method introduced in [2]. Tao et al. [21] appended a non-
negativity constraint and used the half-quadratic splitting method to solve the related minimi-
zation problem. In addition to the methods mentioned above, other authors employed the
effective optimization techniques for solving the TV regularized Poisson deblurring model,
including the alternating extra-gradient method [22], the primal-dual method [23, 24] and the
alternating direction method of multipliers (ADMM) [25, 26]. It is well-known that the TV
regularization methods could preserve fine details such as sharp edges, but they often exhibit
false jump discontinuities causing spurious staircase artifacts in smooth regions of the restored
images. This may be due to the fact that the TV regularization tends to transform the smooth
regions of the solution into piecewise constant ones while solving the minimization problem.
PLOS ONE | https://doi.org/10.1371/journal.pone.0250260 April 20, 2021
2 / 20
PLOS ONE Hybrid regularizer for image deblurring under poisson noise
One possible remedy for this oversharpening behavior is to introduce high-order TV,
which could penalize jumps more. Zhou et al. [27] selected the second-order TV ϕ(f) = kfkHTV
(see Section 2) as a regularizer, and solved by using the alternating minimization scheme.
Their numerical experiments indicated the advantage of the second-order TV in avoiding the
staircase artifacts, compared with the classical TV regularization [28]. But the high-order TV
usually transforms the smooth signal into over-smoothing [29], so the hybrid regularizations
combining the high-order TV with other regularizers were also considered. Among the hybrid
models, Zhang et al. [30] obtained better results than the model with a strongly convex term
kf k2
2 proposed in [31], by replacing the first-order TV with a second-order TV. Besides, Jiang
et al. [11] combined the first-order and second-order TV priors to restore images contami-
nated by Poisson noise and solved the minimization model by the ADMM. To further improve
the restoration result, Liu et al. [5] studied a spatially adapted regularization parameter updat-
ing scheme. As an adaptive balancing scheme between the first and second derivatives, Wang
et al. [32] established the Poisson noise removal framework of iterative reweighted total gener-
alized variation (TGV) model based on the EM algorithm for the denoising case. Zhang et al.
[33] combined the fractional-order TV with non-local TV to alleviate the staircase artifacts for
the cartoon component as well as to preserve the details for the texture component. Ma et al.
[34] proposed a hybrid regularizer containing a patch-based sparsity promoting prior over a
learned dictionary and a pixel-based total variation prior, but it requires additional strategies
to reduce the high computational cost.
Considering that the tight framelet framework can facilitate the sparse representation of
images [35], Shi et al. [36] combined the framelet regularization with non-local TV and non-
negativity constraint in the non-blind stage of blind Poisson deblurring. Zhang et al. [37] pro-
posed the nonconvex and noncontinuous model with ℓ0 norm of the tight wavelet representa-
tion as a regularizer. Fang and Yan [38] combined the ℓ1 norm of framelet coefficients with
TV for Poissonian image deconvolution to reduce artifacts yielded by only using TV regulari-
zation. In addition to working with transform-domain sparsity mentioned above, the overlap-
ping group sparsity, which describes the natural tendency of large values to arise near other
large values rather than in isolation, has been widely concerned in the field of image restora-
tion, due to its remarkable ability to exploit the structural information of the natural image
gradient [4, 39–41]. Especially, for the Poisson noisy restoration, Lv et al. [4] focused on the
regularization by the total variation with overlapping group sparsity (TVOGS) and they
showed that the solution obtained by the ADMM framework is superior to the first-order TV
regularized methods [25, 42] and the high-order regularized methods [27].
To sum up, the high-order TV or the overlapping group sparse prior can be the most feasi-
ble option to alleviate the staircase artifacts. Nonetheless, the high-order TV, due to the over-
penalizing tendency of the ℓ1 norm [43], often causes over-smoothing at the edges while the
overlapping group sparsity tends to smoothen out blockiness in the restored images more
globally. Adam and Paramesran, motivated by the work of Chen et al. [44], proposed a hybrid
regularization model that combined the overlapping group sparse total variation with the
high-order nonconvex TV to denoise images corrupted by Gaussian noise [40]. Recently, they
extended it to non-blind deblurring under Gaussian noise [45]. In this paper, we present an
extension of Adam’s regularizer to the problem of restoring the Poisson noisy images. This
regularizer takes the advantages of both the high-order nonconvex regularizer and overlapping
group sparsity regularizer, i.e., it can simultaneously facilitate the pixel-level and structured
sparsities of the natural image in the gradient domain. Therefore, we expect that it can not
only effectively reduce staircase artifacts but also preserve well sharp edges in the restored
image. However, its optimization is more challenging than Gaussian deblurring due to the ill-
PLOS ONE | https://doi.org/10.1371/journal.pone.0250260 April 20, 2021
3 / 20
PLOS ONE Hybrid regularizer for image deblurring under poisson noise
conditioned non-quadratic data fidelity term. We employ the alternating direction method of
multipliers to solve the derived minimization problem. In particular, during the ADMM itera-
tions, we solve two intractable subproblems: one is from overlapping group sparse prior and
solved by majorization-minimization (MM) method with well-known quadratic majorizer;
another is from the nonconvex ℓp(0 < p < 1) quasi norm, which is solved by the iteratively
reweighted least squares (IRLS) algorithm with the motivation that the IRLS is guaranteed to
converge to its local minima provided better theoretical and practical abilities than the itera-
tively reweighted ℓ1 (IRL1) algorithm [52–55].
The rest of this paper is organized as follows. In Section 2, we introduce some notations
that will be used to formulate our proposed method. We also briefly review the essential con-
cepts and tools including the overlapping group sparsity prior, the ADMM algorithm, the MM
method, and the IRLS algorithm. Section 3 establishes a novel Poissonian image deblurring
model which comprises the generalized Kullback Leibler (KL) divergence as data fidelity term,
and a combined first-order and second-order total variation priors with overlapping group
sparsity as a regularization term. Sequentially, we derive an effective algorithm for minimizing
the non-convex and non-smooth objective function under the ADMM optimization frame-
work. Section 4 demonstrates the superiority of our method via numerical experiments, fol-
lowed by analyzing the parameter setting and convergence behavior. We finally conclude this
paper in Section 5.
2 Preliminaries
2.1 Notations
It is necessary to introduce some notations throughout this paper. Assuming that all images
under consideration are gray-scale images of size n × n, we lexicographically stack the columns
of an image matrix into a vector form. For example, the (i, j)th pixel of image f becomes the
((j − 1)n + i)th entry of the vector, just written as fi,j. Under the periodic boundary conditions
for image f, we introduce the discrete forward difference operator Dþ : Rn�n ! ðRn�n; Rn�nÞ
defined by
ðDþf Þi;j ≔ ððDþ
1 f Þi;j; ðDþ
2 f Þi;jÞ 2 R2;
and similarly for the backward difference operator D(cid:0)
: Rn�n ! ðRn�n; Rn�nÞ,
ðD(cid:0) f Þi;j ≔ ððD(cid:0)
1 f Þi;j; ðD(cid:0)
2 f Þi;jÞ 2 R2;
where the definitions of each forward and backward sub-operators are separately:
ðDþ
1 f Þi;j ≔
8
<
:
fi;jþ1 (cid:0)
fi;j;
if j < n
fi;1 (cid:0)
fi;n;
if j ¼ n
; ðDþ
2 f Þi;j ≔
8
<
:
fiþ1;j (cid:0)
fi;j;
if i < n
f1;j (cid:0)
fn;j;
if i ¼ n
;
ðD(cid:0)
1 f Þi;j ≔
8
<
:
fi;j (cid:0)
fi;j(cid:0) 1;
if j > 1
fi;1 (cid:0)
fi;n;
if j ¼ 1
; and ðD(cid:0)
2 f Þi;j ≔
8
<
:
fi;j (cid:0)
fi(cid:0) 1;j;
if i > 1
f1;j (cid:0)
fn;j;
if i ¼ 1
ð5Þ
ð6Þ
ð7Þ
Second-order difference operators can be recursively defined by using the first-order differ-
ence operators such as
and other second-order difference operators can be similarly defined. Based on the above
PLOS ONE | https://doi.org/10.1371/journal.pone.0250260 April 20, 2021
ðD(cid:0) þ
11 f Þi;j ≔ ðD(cid:0)
1 ðDþ
1 f ÞÞi;j;
ð8Þ
4 / 20
PLOS ONE Hybrid regularizer for image deblurring under poisson noise
definitions, we denote the first and second-order TV of f as
kf kTV ≔
Xn
kðrf Þi;jk2 and kf kHTV ≔
i;j¼1
Xn
kðr2f Þi;jk2;
i;j¼1
ð9Þ
where k�k2 means the Euclidean norm, ðrf Þi;j ¼ ððDþ
ððD(cid:0) þ
12 f Þi;j; ðDþþ
21 f Þi;j; ðD(cid:0) þ
11 f Þi;j; ðDþþ
22 f Þi;jÞ.
1 f Þi;j; ðDþ
2 f Þi;jÞ; and ðr2f Þi;j ¼
2.2 TVOGS and MM algorithm
To describe the structural sparsity of image gradient, we define a pixel-group ~v ði;jÞ;K of size
K × K (K is called group size) centered at every position (i, j) of a two-dimensional array
v = (vi,j)n × n as
~v ði;jÞ;K ¼
2
6
6
6
6
6
6
6
6
4
vi(cid:0) m1;j(cid:0) m1
vi(cid:0) m1;j(cid:0) m1þ1
vi(cid:0) m1þ1;j(cid:0) m1
vi(cid:0) m1þ1;j(cid:0) m1þ1
.
..
.
..
viþm2;j(cid:0) m1
viþm2;j(cid:0) m1þ1
� � �
� � �
.
. .
� � �
3
7
7
7
7
7
7
7
7
5
vi(cid:0) m1;jþm2
vi(cid:0) m1þ1;jþm2
.
..
viþm2;jþm2
2 RK�K;
ð10Þ
2 c; m2 ¼ bK
2c, and b�c denotes the floor function, i.e., it converts any real num-
where m1 ¼ bK(cid:0) 1
ber into a nearest integer less than or equal to it. Let v(i,j),K be the column-major vectorized
form of ~v ði;jÞ;K, i.e., vði;jÞ;K ¼ ~v ði;jÞ;Kð:Þ. Then, as in Liu et al.’s work [46], we can denote the
TVOGS regularizer ϕTO(f) as
�TOðf Þ ≔ �Oðr1f Þ þ �Oðr2f Þ;
ð11Þ
1 f and r2f ≔ Dþ
where r1f ≔ Dþ
tively, and �OðvÞ ¼
simplicity, we denote ϕO(rf) = ϕO(r1 f) + ϕO(r2 f).
Pn
2 f denote the horizontal and vertical gradient of f, respec-
i;j¼1 kvði;jÞ;Kk2 is the K-group OGS function of v 2 Rn2
. For the notational
Next, we review a minimization problem of the form
1
f
min
2
v
kv (cid:0) v0k2
2 þ l�OðvÞg;
ð12Þ
and we denote the objective function of Eq (12) as R(v).
To solve the above sophisticated OGS model, the majorization-minimization approach is
used with cðvjuÞ ¼ 1
2
1
kuði;jÞ;K k2
kvði;jÞ;Kk2
2 þ kuði;jÞ;Kk2
i;j¼1
as a quadratic majorizer of ϕO(v),
h
Pn
i
namely, ψ(v|u) � ϕO(v) for all v and u 6¼ 0 and with equality if v = u. Therefore, instead of the
intractable direct optimization of R(v), we can approximately solve it through iterative mini-
mizing a sequence of surrogate convex problems:
vðkþ1Þ ¼ arg min
QðvjvðkÞÞ; k ¼ 0; 1; � � � ;
v
Q ðvjvðkÞÞ ¼
1
2
PLOS ONE | https://doi.org/10.1371/journal.pone.0250260 April 20, 2021
kv (cid:0) v0k2
2 þ lcðvjvðkÞÞ ¼ kv (cid:0) v0k2
�
�
2 þ l LðvðkÞÞv
�
�2
2
;
ð13Þ
5 / 20
PLOS ONE Hybrid regularizer for image deblurring under poisson noise
where Λ(u) is a diagonal matrix containing the following entries along its diagonal
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
#(cid:0) 1
Xm2
2
Xm2
"
v
u
u
t
jur(cid:0) iþk1;t(cid:0) jþk2
j2
½LðuÞ�l;l ¼
i;j¼(cid:0) m1
k1;k2¼(cid:0) m1
with l = (r − 1)n + t, for r, t = 1, 2, � � �, n. Eq (13) has the closed-form solution
�
vðkþ1Þ ¼ In2 þ lL vðkÞ
(cid:0)
�(cid:0) 1
�
(cid:0)
�TL vðkÞ
v0;
ð14Þ
ð15Þ
which is the input for the next MM iteration. The inversion of the matrix In2 þ lLðvðkÞÞTLðvðkÞÞ
can be efficiently done via simple component-wise calculations. In summary, we obtain the fol-
lowing algorithm for Eq (12).
Algorithm 1 MM algorithm for minimizing Eq (12)
input λ > 0, K, Nin(inner iterations)
initialize v(0) = v0, k = 0.
while k < Nin do
q
jvðkÞ
1: ½LðvðkÞÞ�l;l ¼
2: vðkþ1Þ ¼ ðIn2 þ lLðvðkÞÞTLðvðkÞÞÞ(cid:0) 1 v0;
3: k = k + 1;
k1;k2¼(cid:0) m1
r(cid:0) iþk1;t(cid:0)
i;j¼(cid:0) m1
jþk2
. MM inner loop
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Pm2
j2�(cid:0) 1
2
Pm2
;
½
end while
2.3 Alternating direction method of multipliers
The standard form of the ADMM [47] is designed to tackle the distributed convex optimiza-
tion problem with linear equality constraints, and many variants have been developed [48–50].
Recently, Wang et al. [51] extended it to the problems involving the nonconvex and non-
smooth muti-blocks objective, as follows:
Xs
min
x;y
i¼1
F iðxiÞ þ GðyÞ
Xs
subject to
i¼1
Aixi þ By ¼ 0;
ð16Þ
where x ¼ ½x1; � � � ; xs� 2 RN (here xi 2 Rni ; N ¼
i¼1 ni) and y 2 RL are primal variables with
the corresponding coefficient matrix Ai 2 RM�ni and B 2 RM�L respectively, F i : Rni ! R is a
continuous function, and G : RL ! R is a smooth function.
Ps
In general, F i can be nonsmooth and nonconvex, and G can be nonconvex. By introducing
a Lagrangian multiplier w 2 RM and a penalty parameter δ > 0 for the linear constraint
Ps
i¼1 Aixi þ By ¼ 0, we obtain the augmented Lagrangian in a scaled form,
Ldðx; y; wÞ ¼
Xs
F iðxiÞ þ GðyÞ þ
i¼1
d
2
�
�
�
�
�
Xs
i¼1
Aixi þ By þ
�
2
�
�
�
�
2
:
w
d
ð17Þ
The ADMM algorithm proceeds to alternatively update each variable (x(k+1), y(k+1)) until it
reaches a stationary point of Eq (17) or meets a stopping criterion. Then, we have the following
PLOS ONE | https://doi.org/10.1371/journal.pone.0250260 April 20, 2021
6 / 20
PLOS ONE Hybrid regularizer for image deblurring under poisson noise
iterative scheme:
xðkþ1Þ
1
.
..
xðkþ1Þ
s
2 arg min
F 1ðx1Þ þ
x1
�
�
�
�
d
2
A1x1 þ A2xðkÞ
2 þ � � � þ AsxðkÞ
s þ ByðkÞ þ
�
�
�
�
2
2
;
wðkÞ
d
.
..
.
..
2 arg min
F sðxsÞ þ
xs
yðkþ1Þ 2 arg min
GðyÞ þ
y
d
2
d
2
�
�
�
�
�
�
�
�
A1xðkþ1Þ
1 þ � � � þ As(cid:0) 1xðkþ1Þ
s(cid:0) 1 þ Asxs þ ByðkÞ þ
�
2
�
�
�
2
;
wðkÞ
d
ð18Þ
A1xðkþ1Þ
1 þ � � � þ Asxðkþ1Þ
s
þ By þ
�
2
�
�
�
2
;
wðkÞ
d
wðkþ1Þ ¼ wðkÞ þ d
!
Aixðkþ1Þ
i
þ Byðkþ1Þ
:
Xs
i¼1
The following lemma establishes the convergence result of ADMM with the nonconvex and
nonsmooth objective [51].
Lemma 1. Let F ðxÞ ¼ F 1ðx1Þ þ � � � þ F sðxsÞ and A = [A1� � �As]. Suppose that the following
assumptions S1–S5 hold, then the sequence generated by Eq (18) with any sufficiently large δ and
any initialization has at least one limit point, and each limit point is a stationary point of Eq
(17).
S1 (coercivity) Let D ¼ fðx; yÞ 2 RNþL : Ax þ By ¼ 0g be the nonempty feasible set and
F ðxÞ þ GðyÞ is coercive over D;
S2 (feasibility) Im(A) � Im(B), where Im(�) is defined as the image of a matrix;
S3 (Lipschitz sub-minimization paths)
(a) For any fixed x, there exists a Lipschitz continuous map G : ImðBÞ ! RL obeying GðuÞ ¼
arg minyfF ðxÞ þ GðyÞ : By ¼ ug provided a unique minimizer,
(b) For i = 1, � � �, s and fixed x−i and y, there exists a Lipschitz continuous map Fi : ImðAiÞ !
Rni obeying FiðuÞ ¼ arg minxi
x−i ≔ [x1;� � �;xi−1;xi+1;� � �;xs];
fF ðxÞ þ GðyÞ : Aixi ¼ ug provided a unique minimizer, where
S4 (objective-F regularity) F is lower semi-continuous or F 1 is lower semi-continuous and
F i; i ¼ 2; � � � ; s is restricted prox-regular;
S5 (objective-G regularity) G is Lipschitz differentiable with the constant LrG
> 0.
2.4 Iteratively reweighted least squares algorithm
IRLS algorithm [52, 53] solves the following nonconvex ℓp norm sparsity problem:
�
min
z
1
2
kz (cid:0)
˚zk2
2 þ lkzkp
p
�
;
ð19Þ
where z; ˚z 2 Rn2
, and 0 < p < 1 causes the model to be nonconvex. It is showed that the IRLS
is guaranteed to converge to its local minima with better theoretical and practical abilities than
the iteratively reweighted ℓ1 (IRL1) [52–55]. Then, the problem Eq (19) can be approximated
to
(
min
z
1
2
kz (cid:0)
˚zk2
2 þ
PLOS ONE | https://doi.org/10.1371/journal.pone.0250260 April 20, 2021
Xn2
oiz2
i
i¼1
)
ð20Þ
7 / 20
PLOS ONE
Hybrid regularizer for image deblurring under poisson noise
with the weight oi ¼ lpðz2
sion by zero. Given k-th estimation zðkÞ ¼ ðzðkÞ
oðkÞ ¼ ðoðkÞ
1 ; � � � ; oðkÞ
i þ �Þp=2(cid:0) 1 and � � 1 being a small positive number to avoid divi-
1 ; � � � ; zðkÞ
n2 Þ, IRLS first calculates the weight
n2 Þ, then updates z by solving the following problem:
(
zðkþ1Þ ¼ arg min
z
1
2
kz (cid:0)
˚zk2
2 þ
Xn2
oðkÞ
i z2
i
i¼1
)
:
ð21Þ
We summarize the IRLS method in Algorithm 2.
Algorithm 2 IRLS algorithm for minimizing Eq (19)
; p 2 ð0; 1Þ; l > 0; �⪡1.
input ˚z 2 Rn2
initialize zð0Þ ¼ ˚z; k ¼ 0.
while not converged do
1: update weight: ω(k) = λp((z(k))2 + �)p/2−1;
2: update z: zðkþ1Þ ¼ ðIn2 þ diagðoðkÞÞÞ(cid:0) 1˚z;
3: k = k + 1;
end while
3 Proposed model and optimization method
In this section, we first present the proposed model and then solve it in the framework of non-
convex and nonsmooth ADMM.
3.1 Model formulation
Considering the advantages of the overlapping group sparse total variation and the high-order
nonconvex total variation, we investigate the Poisson noisy image restoration problem with
the following regularizer:
�ðf Þ ¼ �TOðf Þ þ Z�HTVpðf Þ;
ð22Þ
where ϕTO(f) is the TVOGS term, �HTVpðf Þ ≔ kðr2f Þkp
nonconvex second-order TV term, and η > 0 is a regularization parameter. Substituting the
hybrid regularizer Eq (22) into Eq (3) with the consideration to Eq (11), we obtain
i;j¼1 kðr2f Þi;jkp
p (0 < p < 1) is the
p ¼
Pn
flDKLðHf jjgÞ þ �Oðrf Þ þ Zkr2f kp
pg:
min
f
ð23Þ
In the model above, λ and η are used to control the data fidelity term and the nonconvex
second-order regularizer, respectively. The data fidelity term is an ill-conditioned non-qua-
dratic log-likelihood, while the regularizer is nonconvex and nonsmooth due to the presence
of the ℓp(0 < p < 1) quasi norm. This may increase the difficulty in minimizing the model. To
circumvent this difficulty, we propose an efficient algorithm based on the framework of non-
convex and nonsmooth ADMM in the following subsection.
3.2 Optimization
In order to apply the variable splitting scheme, we introduce three auxiliary variables to trans-
form the original complicated minimization problem into the following equivalent linear
PLOS ONE | https://doi.org/10.1371/journal.pone.0250260 April 20, 2021
8 / 20
PLOS ONE Hybrid regularizer for image deblurring under poisson noise
constraint minimization problem:
min
f
lDKLðx1jjgÞ þ �Oðx2Þ þ Zkx3kp
p
subject to x1 ¼ Hf ; x2 ¼ rf ; x3 ¼ r2f :
ð24Þ
We establish a corresponding relation between the variables to conform with the ADMM
framework Eq (16), as shown below:
; x2 ¼ rf 2 R2n2
(1). x1 ¼ Hf 2 Rn2
(2). F 1ðx1Þ ¼ lDKLðx1jjgÞ; F 2ðx2Þ ¼ �Oðx2Þ; F 3ðx3Þ ¼ Zkx3kp
p; GðyÞ ¼ 0;
1
0
; x3 ¼ r2f 2 R4n2
; y ¼ f 2 Rn2
0
1
1
0
0
1
;
(3). A1 ¼
B
@
C
A, A2 ¼
B
@
C
A, A3 ¼
B
@
(cid:0) In2
0T
3
02
03
(cid:0) I2n2
0T
1
0T
2
01
(cid:0) I4n2
H
C
A, B ¼
B
@
C
A,
r
r2
where 01, 02, 03 denote zero matrix of size 2n2 × 4n2, 4n2 × n2, n2 × 2n2, respectively, and In2 is
the identity matrix of order n2. By introducing the Lagrangian multiplier w ¼ ½w1; w2; w3� 2
R7n2
d ¼ d1; d2; d3
ing augmented Lagrangian function:
ÞT 2 R3, we can turn Eq (24) into the unconstrained minimization of the follow-
) and the positive penalty parameter vector
(here w1 2 Rn2
ð
; w3 2 R4n2
; w2 2 R2n2
Ld x1; x2; x3; f ; w
ð
Þ ¼ lDKL x1jjg
ð
�
�
�
�
þ
d1
2
x1 (cid:0) Hf þ
�
2
�
�
�
2
w1
d1
þ
d2
2
�
�
�
x2 (cid:0) rf þ
�
Þ þ �O x2ð
�
�
�
�
þ
w2
d2
d3
2
2
2
pþ
Þ þ Zkx3kp
�
�
�
�
x3 (cid:0) r2f þ
�
2
�
�
�
2
:
w3
d3
ð25Þ
Then, the ADMM for solving Eq (24) starts its iteration to update every variable by mini-
mizing Eq (25) alternatively. This iterative scheme could be split into several subproblems.
3.2.1 x1-subproblem. According to the ADMM scheme, pulling out the terms with x1
from Eq (25) yields
Þ
x kþ1
ð
1
(
¼ arg min
x1
ð
lDKL x1jjg
Þ þ
�
�
�
�
�
d1
2
x1 (cid:0) Hf kð Þ þ
�
�
�
�
�
w kð Þ
1
d1
)
2
:
2
ð26Þ
It is apparent that each component of x1 can be solved separately. Through the basic mathe-
matical manipulations in the case of b = 0, we obtain
ðxðkþ1Þ
1
Þi;j ¼
2
4
1
2
ðHf kð Þ (cid:0)
w kð Þ
1 þ l
d1
Þi;j þ
v
u
u
t
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
w kð Þ
4lgi;j
1 þ l
d1
d1
Hf kð Þ (cid:0)
!2
þ
3
5:
i;j
ð27Þ
3.2.2 x2-subproblem. Minimizing Ld x1; x2; x3; f ; w
ð
Þ with respect to x2 leads to the over-
lapping group sparse problem
Þ
x kþ1
ð
2
(
¼ arg min
x2
�
�
�
x2 (cid:0) rf kð Þ (cid:0)
�
�
d2
2
!
�
2
�
�
�
�
2
w kð Þ
2
d2
)
þ�O x2ð
Þ
:
ð28Þ
It is easy to see that this minimization problem matches the framework of Eq (12), thus x2
can be solved by Algorithm 1.
PLOS ONE | https://doi.org/10.1371/journal.pone.0250260 April 20, 2021
9 / 20
PLOS ONE
Hybrid regularizer for image deblurring under poisson noise
3.2.3 x3-subproblem. By omitting terms irrespective of x3, we have
Þ
x kþ1
ð
3
(
¼ arg min
x3
�
�
�
x3 (cid:0) r2f kð Þ (cid:0)
�
�
d3
2
!
�
2
�
�
�
�
2
w kð Þ
3
d3
)
þZkx3kp
p
:
ð29Þ
Since this problem involves the nonconvex ℓp norm, minimizing x3 is not straightforward.
But, after some simple modifications, we can apply the IRLS algorithm according to Algorithm
2.
3.2.4 f-subproblem. Considering Eq (25) with respect to f, we have
(
ð
f kþ1
Þ ¼ arg min
f
�
�
�
�
�
d1
2
Þ
x kþ1
ð
1
(cid:0) Hf þ
�
2
�
�
�
�
2
w kð Þ
1
d1
þ
þ
�
�
�
�
�
�
�
�
�
�
d2
2
d3
2
Þ
x kþ1
ð
2
(cid:0) rf þ
w kð Þ
2
d2
Þ
x kþ1
ð
3
(cid:0) r2f þ
w kð Þ
3
d3
�
�
�
�
�
2
þ
2
�
2
�
�
�
�
2
ð30Þ
)
;
which can be solved by the following normal equation
(cid:0)
d1HTH þ d2rTr þ d3r2Tr2
�
f ¼ d1HT x kþ1
ð
Þ
1 þ
!
þ
w kð Þ
1
d1
þd2rT
x kþ1
ð
Þ
2 þ
!
w kð Þ
2
d2
þ d3r2T x kþ1
ð
Þ
3 þ
!
:
w kð Þ
3
d3
ð31Þ
We assume that both the image and convolution matrix are periodically extended, thus
well-known fast Fourier transform can be adopted to efficiently solve Eq (31) [4, 56, 57], which
results in an optimal solution
� �
Þ ¼ F (cid:0) 1 D
C
ð
f kþ1
with
C ¼ d1F � Hð
�
Þ � F Hð
�
D ¼ F � Hð
Þ � d1F x kþ1
ð
1
Þ þ d2F � rð
�
�
þ F w kð Þ
Þ
1
Þ � F rð
�
�
Þ þ d3F � r2
�
ð
Þ � F r2
�
ð
�
Þ;
þ F � rð
Þ � d2F x kþ1
ð
2
Þ
�
þ F w kð Þ
2
ð32Þ
�
�
þ
ð33Þ
�
�
þ F � r2
ð
Þ � d3F x kþ1
ð
3
�
Þ
�
þ F w kð Þ
3
�
�
;
where F denotes the FFT operator, � and � stand for complex conjugate and element-wise
multiplication, respectively. The division is computed by component-wise fashion. Note that
many subterms need to be computed only once during the iterative updates.
PLOS ONE | https://doi.org/10.1371/journal.pone.0250260 April 20, 2021
10 / 20
PLOS ONE
Hybrid regularizer for image deblurring under poisson noise
3.2.5 w-subproblem. Finally, the updating scheme of the Lagrangian multipliers can be
written as
Þ
w kþ1
ð
1
Þ
w kþ1
ð
2
Þ
w kþ1
ð
3
¼ w kð Þ
¼ w kð Þ
ð
1
�
1 þ d1 x kþ1
�
2 þ d2 x kþ1
�
3 þ d3 x kþ1
ð
2
ð
3
Þ
Þ
Þ
�
ð
(cid:0) Hf kþ1
Þ
;
�
(cid:0) rf kþ1
ð
Þ
;
�
Þ
(cid:0) r2f kþ1
ð
:
ð34Þ
¼ w kð Þ
In summary, the key steps of the ADMM algorithm for solving the suggested model are
described in Algorithm 3.
Algorithm 3 The ADMM algorithm for solving Eq (23)
input p, λ, η, δ, K, Nin.
initialize f(0), k = 0, w(0) = 0.
while a stopping criteria unsatisfied do
loop
. outer
Þ
Þ
ð
1
ð
2
ð
3
according to Eq (27).
according to Eq (28) using Algorithm 1.
according to Eq (29) using Algorithm 2.
1: Update x kþ1
2: Update x kþ1
3: Update x kþ1
4: Update f(k+1) according to Eq (32).
5: Update w kþ1
6: k = k + 1.
; i ¼ 1; 2; 3 according to Eq (34).
ð
i
Þ
Þ
end while
We discuss the convergence of the proposed algorithm. Inspired by [51], by verifying the
assumptions in Lemma 1, we can obtain the following convergence result for Algorithm 3.
Theorem 1. The sequence of (x1, x2, x3, f) generated by Algorithm 3 with any sufficiently large
δ and any start point will converge to a stationary point of the augmented Lagrangian Ld.
Proof. Since our model fits the framework of Eq (16), it remains only to check S1–S5 in
Lemma 1. It is easy to verify that DKL(x1||g) is coercive [58] as well as ϕO(x2) and kr2x3kp
p.
Thus, S1 holds. Ax + By = 0 implies S2. S3 holds because both A and B are full column rank
matrices. DKL(x1,g) is lower semi-continuous and ℓp-quasi norm is restricted prox-regular
[51]. Furthermore, ϕO(x2) is convex, hence restricted prox-regular. It is clear that S5 holds,
which completes the proof.
The convergence of Algorithm 3 can also be verified experimentally.
4 Numerical experiments
In this section, we present the numerical experiments to illustrate the effectiveness of the pro-
posed method for the Poisson noisy image restoration, compared with the closely related state-
of-the-art methods including TVOGS [4], SAHTV [5], SB-FA [38] and FT-ADMM [37]. The
TVOGS method introduced the overlapping group sparse TV prior combined with the box-
constraint as a regularizer and solved the optimization model under the ADMM framework,
but denoising is out of their consideration. In the SAHTV method, Liu et al. [5] combined the
first and second-order TV priors, and updated their pixel-wise weighting coefficients with the
local information during the consecutive estimation of the latent image. FT-ADMM, whose
regularizer is the ℓ0 norm of the wavelet representation of the image, was also proposed to effi-
ciently eliminate the staircase artifacts. Fang and Yan [38] proposed to combine the ℓ1 norm of
the framelet representation of the latent image with the TV prior, and solved the resultant
model by the split Bregman method. Since they recommended using the SB-FA algorithm
PLOS ONE | https://doi.org/10.1371/journal.pone.0250260 April 20, 2021
11 / 20
PLOS ONE Hybrid regularizer for image deblurring under poisson noise
Fig 1. Test images used in our experiments. (a): “beauty” (256 × 256). This image was republished from https://www.
flickr.com/photos/90471071@N00/3201337190 under a CC BY license, with permission from Irina Gheorghita, original
copyright 2009; (b): “lotus” (500 × 500). This image was republished from https://www.flickr.com/photos/
robertlylebolton/7797693474 under a CC BY license, with permission from Robert Lyle Bolton, original copyright 2012;
(c): “dolphin” (512 × 512). This image was republished from https://www.flickr.com/photos/grovecrest/6981622176
under a CC BY license, with permission from Simon Lewis, original copyright 2011.
https://doi.org/10.1371/journal.pone.0250260.g001
without the TV prior through the numerical experiments, we adopt SB-FA as a competitive
method rather than SB-FATV with the TV prior.
All experiments in this paper are implemented on Windows 10 64-bit and Matlab R2016b
running on a desktop computer with an Intel Core i5-4590 CPU 3.3GHz and 8GB of RAM. We
uploaded the code for our algorithm on https://github.com/KSJhon/PoissonDeblur_hybrid.
To proceed with the simulation experiments, we intentionally degrade the clean image to
construct the corrupted image. More specifically, an original clean image is first scaled to a
preset peak value MAXf, which decides the different levels of Poisson noise. After that, the
scaled image is convolved with a given PSF to simulate the blurring effect, and further contam-
inated by the signal-dependent Poisson noise. In the following experiments, we set MAXf to be
100, 200, 300, and 350, in which the lower MAXf indicates the relatively higher noise level. For
the deblurring simulations, we consider three types of blur kernels: (1) a 9 × 9 Gaussian kernel
with standard deviation 1; (2) a linear motion blur with a motion length 5 and an angle 45˚ in
the counterclockwise direction; (3) a kernel from Levin et al.’s public dataset [59]. All blurring
operations are fulfilled by the Matlab built-in function “fspecial”, and the Poisson noise is
generated by “poissrnd” without any additional parameter. The test images with various
sizes are shown in Fig 1.
We quantitatively evaluate the performances of the proposed method and the competing
methods by means of the peak signal-to-noise ratio (PSNR) and the structural similarity
(SSIM), defined as
PSNR ¼ 20 log
nMAXf
kf (cid:0) ^f k2
;
SSIM ¼
�
�
�
�
�
2mf m^f þ C1
�
�
2sf ;^f þ C2
� ;
f þ m2
m2
^f
þ C1
f þ s2
s2
^f
þ C2
ð35Þ
ð36Þ
where f is the original image, ^f is the restored image, mf ; m^f are the respective averages, s2
are the respective variances, sf ;^f is the covariance of f and ^f , and C1 ¼ MAX2
f =104; C2 ¼
9 � MAX2
f =104 by default. Generally, the larger the PSNR and the closer SSIM to 1, the better
f ; s2
^f
PLOS ONE | https://doi.org/10.1371/journal.pone.0250260 April 20, 2021
12 / 20
PLOS ONE Hybrid regularizer for image deblurring under poisson noise
the quality of the restored image. In the experiments, we terminate the iteration in Algorithm
3 when the relative error (RelErr) between two consecutive estimates falls below the predefined
tolerance level ε, as follows:
Þ (cid:0)
ð
kf kþ1
kf kþ1
ð
f kð Þk2
Þk2
< ε:
ð37Þ
4.1 Selection of parameters
Since the convergence of the nonconvex ADMM and its suboptimization in Algorithm 3
depends on the values of parameters, they require careful tuning.
For the choice of p, as in the experiments reported in [40], we set it to be 0.1 throughout the
experiments, with the consideration of the important aspect of ℓp norm, which says that taking
p sufficiently close to 0 favors to preserve sharp edges in the restored image.
The group size K also plays an important role in balancing the trade-off between the global
noise filtering performance and the computational cost. In order to find the optimal K, we
conduct a denoising experiment by varying K while keeping other parameters remained, as
shown in Fig 2. Obviously, K = 3 is the best choice for all noise levels, so we fix K = 3 in the fol-
lowing experiments. In addition, the number of inner iterations, Nin, also affects the MM algo-
rithm. We straightforwardly fix Nin = 5 as indicated in [4, 29, 39, 40]. To balance the accuracy
and speed, we empirically set Nout = 50, ε = 10−3. The other parameters, including the regulari-
zation parameters (λ, η) and penalty parameter δ, are manually adjusted to achieve the highest
improvement in the PSNR and SSIM values. In the following subsections, we compare the pro-
posed method (named HTVp-OGS) with the state-of-the-art methods for denoising and
deblurring images under the Poisson noise. Unless otherwise indicated, all parameters
involved in the competitive algorithms are carefully selected near the given default values to
give the best performance, respectively.
4.2 Denoising
Here, we show that our model provides better results in Poisson noise removal through the
comparison with the state-of-the-art methods: SB-FA [38], SAHTV [5], and FT-ADMM [37].
To evaluate the denoising performance, we just set H as an identity matrix, and empirically
fix δ = (5 � 10−3, 3 � 10−2, 2 � 10−4)T. We can observe that λ = 3 � MAXf can bring the satisfactory
results of Eq (23). The other regularization parameter η is hand-tuned to get the best denoising
performance according to the noise level. The denoising results, in terms of the PSNR and
SSIM values, are shown in Table 1. From Table 1, it is obvious that the PSNR and SSIM values
of the images denoised by our model are higher than those of the other three methods (SB-FA,
SAHTV, FT-ADMM). We display in Fig 3. the denoised versions and the zoom-in regions
obtained by our method and other competitive methods from the noisy images with the noise
level MAXf = 100.
Fig 2. Evolution of PSNR values of denoised images with MAXf = 350, 300, 200 and 100 along the varying group
sizes. (a): “beauty” image; (b): “lotus” image; (c): “dolphin” image.
https://doi.org/10.1371/journal.pone.0250260.g002
PLOS ONE | https://doi.org/10.1371/journal.pone.0250260 April 20, 2021
13 / 20
PLOS ONE Hybrid regularizer for image deblurring under poisson noise
Table 1. The PSNR and SSIM values for denoised images by different methods.
MAXf
clean(f)
noisy(g)
denoised(^f )
350
300
200
100
beauty
lotus
dolphin
beauty
lotus
dolphin
beauty
lotus
dolphin
beauty
lotus
dolphin
27.97/0.649
28.12/0.530
27.95/0.465
27.31/0.620
27.46/0.498
27.28/0.431
25.55/0.539
25.70/0.412
25.52/0.348
22.53/0.402
22.68/0.283
22.51/0.227
https://doi.org/10.1371/journal.pone.0250260.t001
SB-FA
34.96/0.929
36.34/0.913
38.18/0.926
34.29/0.912
35.63/0.890
37.04/0.898
32.16/0.844
33.23/0.796
33.94/0.787
27.74/0.645
28.35/0.542
28.41/0.500
SAHTV
35.41/0.941
37.42/0.936
38.41/0.944
34.77/0.931
36.73/0.922
37.89/0.933
32.81/0.880
34.08/0.843
35.00/0.855
28.06/0.676
28.45/0.562
28.66/0.539
FT-ADMM
HTVp-OGS
34.17/0.920
34.96/0.898
36.75/0.910
33.73/0.906
34.42/0.876
35.67/0.884
31.85/0.846
32.38/0.786
33.06/0.779
28.18/0.660
28.49/0.542
28.58/0.544
37.08/0.955
38.64/0.957
40.14/0.966
36.82/0.953
38.18/0.955
39.61/0.964
35.73/0.944
37.12/0.945
38.44/0.956
33.01/0.917
35.06/0.925
36.70/0.939
4.3 Deblurring
In our deblurring experiments, we consider three types of blur kernels. As mentioned above,
the first two are artificial to respectively mimic the effect of the out-of-focus blur and motion
blur, and the last one is the blur kernel from Levin et al.’s dataset [59]. For the sake of a fair
comparison with other methods, we equally set the tolerance level ε = 10−3 and iteration
Fig 3. Denoised images and a zoom-in region from the Poisson noisy images with the noise level MAXf = 100. first column: Poisson noisy images;
second column: denoised images by SB-FA [38]; third column: denoised images by SAHTV [5]; fourth column: denoised images by FT-ADMM [37];
fifth column: denoised images by HTVp-OGS.
https://doi.org/10.1371/journal.pone.0250260.g003
PLOS ONE | https://doi.org/10.1371/journal.pone.0250260 April 20, 2021
14 / 20
PLOS ONE Hybrid regularizer for image deblurring under poisson noise
Table 2. The PSNR and SSIM values for the Poisson image restoration by different methods in the case of the Gaussian blur with kernel size 9 × 9 and variation 1.
MAXf
clean(f)
degraded(g)
350
300
200
100
beauty
lotus
dolphin
beauty
lotus
dolphin
beauty
lotus
dolphin
beauty
lotus
dolphin
27.00/0.62
27.68/0.51
27.55/0.45
26.47/0.59
27.08/0.47
26.93/0.41
24.95/0.51
25.44/0.39
25.27/0.33
22.22/0.37
22.55/0.27
22.39/0.22
https://doi.org/10.1371/journal.pone.0250260.t002
SB-FA
32.81/0.92
35.82/0.92
36.70/0.93
32.58/0.92
35.43/0.91
36.09/0.92
31.51/0.88
33.98/0.87
34.49/0.87
28.50/0.75
29.98/0.70
30.28/0.68
SAHTV
32.45/0.93
36.31/0.94
36.97/0.94
32.19/0.92
36.02/0.93
36.64/0.94
31.41/0.90
34.96/0.91
35.67/0.92
29.78/0.85
32.44/0.84
33.19/0.85
restored(^f )
TVOGS
31.19/0.92
34.64/0.93
36.67/0.95
31.16/0.92
34.61/0.93
36.50/0.95
31.02/0.91
34.36/0.92
35.99/0.94
30.42/0.90
33.28/0.90
34.53/0.91
FT-ADMM
HTVp-OGS
32.39/0.93
34.90/0.93
35.87/0.95
32.33/0.93
34.84/0.93
35.83/0.94
31.96/0.91
34.50/0.92
35.66/0.93
30.66/0.87
32.86/0.85
33.78/0.86
32.85/0.94
36.66/0.94
37.69/0.95
32.60/0.93
36.37/0.94
37.37/0.95
31.92/0.92
35.48/0.93
36.58/0.95
30.57/0.90
33.87/0.92
35.29/0.94
number Nout = 50 in all experiments. As in the denoising experiments, we fix λ = 3 � MAXf and
select the penalty parameter δ as (0.01, 0.1, 0.01)T. The values of δ and λ are kept constant, and
the other parameter η is selected by the grid search scheme to obtain high-quality restored
images in the following three cases:
Setting 1. Gaussian blur case: We set η = 18, 14, 6, and 2 according to the noise level MAXf =
350, 300, 200, and 100, respectively. The detailed comparison of five methods is provided in
Table 2.
Setting2. linear motion blur case: We keep δ and λ remained and empirically set η = 8, 6, 4, 1,
and the detailed results are summarized in Table 3.
Setting 3. ground truth blur case: In this case, we set η = 4, 3, 2, 0.2, and the detailed results are
summaried in Table 4.
Table 3. The PSNR and SSIM values for the Poisson image restoration by different methods in the case of the motion blur with length 5 and angle 45˚.
MAXf
clean(f)
degraded(g)
350
300
200
100
beauty
lotus
dolphin
beauty
lotus
dolphin
beauty
lotus
dolphin
beauty
lotus
dolphin
26.64/0.61
27.54/0.50
27.45/0.44
26.13/0.58
26.94/0.47
26.85/0.41
24.74/0.50
25.34/0.39
25.23/0.33
22.10/0.37
22.79/0.34
22.37/0.21
https://doi.org/10.1371/journal.pone.0250260.t003
SB-FA
31.34/0.92
34.22/0.93
36.07/0.94
31.32/0.92
34.21/0.92
36.01/0.94
31.11/0.91
34.08/0.92
35.78/0.94
30.39/0.88
31.48/0.84
34.56/0.91
SAHTV
31.90/0.92
35.62/0.93
37.06/0.94
31.71/0.92
35.33/0.93
36.72/0.94
30.97/0.90
34.57/0.92
35.88/0.93
29.88/0.87
31.51/0.83
33.81/0.88
restored(^f )
TVOGS
29.96/0.90
33.39/0.92
35.63/0.94
29.98/0.90
33.39/0.92
35.54/0.94
29.95/0.90
33.24/0.91
35.19/0.94
29.67/0.89
31.17/0.84
33.92/0.91
FT-ADMM
HTVp-OGS
31.76/0.92
34.56/0.93
35.62/0.95
31.58/0.92
34.27/0.93
35.49/0.93
31.12/0.90
33.79/0.91
35.19/0.92
29.82/0.84
30.77/0.80
33.23/0.84
32.33/0.93
36.31/0.94
37.68/0.95
32.12/0.93
36.04/0.94
37.46/0.95
31.56/0.92
35.32/0.93
36.58/0.95
30.37/0.90
31.79/0.85
35.34/0.94
PLOS ONE | https://doi.org/10.1371/journal.pone.0250260 April 20, 2021
15 / 20
PLOS ONE Hybrid regularizer for image deblurring under poisson noise
Table 4. The PSNR and SSIM values for the Poisson image restoration by different methods in the case of the ground truth blur.
MAXf
clean(f)
degraded(g)
350
300
200
100
beauty
lotus
dolphin
beauty
lotus
dolphin
beauty
lotus
dolphin
beauty
lotus
dolphin
25.44/0.56
26.49/0.46
26.83/0.42
25.06/0.53
26.03/0.43
26.31/0.39
23.93/0.46
24.69/0.36
24.84/0.31
21.62/0.33
22.14/0.24
22.17/0.20
https://doi.org/10.1371/journal.pone.0250260.t004
SB-FA
29.81/0.89
32.69/0.91
34.58/0.94
29.80/0.89
32.66/0.91
34.47/0.94
29.72/0.88
32.67/0.91
34.42/0.93
29.30/0.87
32.19/0.89
33.77/0.92
SAHTV
30.96/0.90
34.78/0.92
35.84/0.94
30.74/0.90
34.45/0.92
35.57/0.93
30.01/0.88
33.49/0.90
34.67/0.91
28.74/0.83
31.43/0.84
32.19/0.86
restored(^f )
TVOGS
29.22/0.87
32.70/0.91
34.85/0.94
29.22/0.87
32.67/0.91
34.79/0.94
29.21/0.87
32.51/0.90
34.54/0.93
28.96/0.87
31.91/0.89
33.57/0.91
FT-ADMM
HTVp-OGS
30.78/0.91
33.27/0.92
34.38/0.94
30.66/0.90
33.24/0.91
34.38/0.94
30.27/0.88
33.01/0.90
34.43/0.93
29.26/0.84
32.04/0.86
33.42/0.88
31.17/0.91
34.95/0.93
36.43/0.95
30.97/0.91
34.62/0.92
36.18/0.94
30.31/0.89
33.90/0.92
35.22/0.94
29.20/0.87
32.55/0.90
34.18/0.93
In Fig 4, we show the degraded “dolphin” images which are blurred by the corresponding
kernels and further corrupted by the Poisson noise of noise level 200, as well as its restored ver-
sions obtained through our method and the competitive methods. Fig 5(a) depicts the RelErr
curves for the sequence of temporary estimates of the “lotus” image obtained during the
HTVp-OGS iterations. Besides, in Fig 5(b), we show that, as the iteration proceeds, the PSNR
and SSIM values moderately increase. From the results of the above three experiments, we can
see that in most cases, our method provides the best results in terms of the PSNR and SSIM
Fig 4. Restored “dolphin” images and zoom-in regions from the degraded images with different blur kernels and the Poisson noise of level MAXf
= 200. first row: Gaussian blur; second row: motion blur; third row: ground truth blur; first column: degraded images (PSF is shown at the bottom-left
corner); second column: restored images by SB-FA [38]; third column: restored images by SAHTV [5]; fourth column: restored images by TVOGS [4];
fifth column: restored images by FT-ADMM [37]; sixth column: restored images by HTVp-OGS.
https://doi.org/10.1371/journal.pone.0250260.g004
PLOS ONE | https://doi.org/10.1371/journal.pone.0250260 April 20, 2021
16 / 20
PLOS ONE Hybrid regularizer for image deblurring under poisson noise
Fig 5. Plot of performance along the HTVp-OGS iterations for restoring “lotus” images blurred by different
kernels and corrupted by the Poisson noise of level MAXf = 200. (a) relative error in log scale; (b) PSNR and SSIM
values.
https://doi.org/10.1371/journal.pone.0250260.g005
values, and in extreme cases, we obtain a slightly lower PSNR value than others, but the SSIM
value still keeps the best. It is worth noting that the transform-based methods (SB-FA and
FT-ADMM) are quite time-consuming because they inherently involve some complicated
nonlinear operations [3], while our method works in the gradient domain, making it take a rel-
atively short running time.
5 Conclusions
In this paper, we propose a new method to restore Poissonian images (including denoising
and deblurring) by using a hybrid regularizer, which combines the overlapping group sparse
total variation with the high-order nonconvex total variation. This regularizer allows us to
exploit their advantages in order to alleviate the staircase artifacts while preserving the original
sharp edges simultaneously. The model derived from the Bayesian perspective is effectively
solved by the nonconvex and nonsmooth ADMM optimization framework. We adopt the
MM and IRLS algorithm for solving the subproblems accompanied by the OGS and noncon-
vex second-order total variation priors, respectively. Numerical experiments demonstrate that
the proposed method outperforms the other related methods in terms of the PSNR and SSIM
values. The study on adaptively determining the optimal parameters according to the image
content and noise level will be the future work.
Supporting information
S1 File.
(STY)
Acknowledgments
The authors would like to thank Dr. Houzhang Fang and Dr. Haimiao Zhang for supplying us
with the codes used in our numerical comparisons.
Author Contributions
Conceptualization: Jun Liu.
Formal analysis: Jun Liu.
Methodology: Wensheng Zhu.
Software: Kyongson Jon.
PLOS ONE | https://doi.org/10.1371/journal.pone.0250260 April 20, 2021
17 / 20
PLOS ONE Hybrid regularizer for image deblurring under poisson noise
Supervision: Wensheng Zhu.
Validation: Xiaoguang Lv.
Visualization: Xiaoguang Lv.
Writing – original draft: Kyongson Jon.
Writing – review & editing: Kyongson Jon.
References
1. Vardi Y, Shepp LA, Kaufman L. A statistical model for positron emission tomography. Journal of the
American Statistical Association. 1985; 80(389):8–20. https://doi.org/10.2307/2288037
2.
3.
4.
5.
Landi G, Piccolomini EL. A projected Newton-CG method for nonnegative astronomical image deblur-
ring. Numerical Algorithms. 2008; 48(4):279–300. https://doi.org/10.1007/s11075-008-9198-3
James AP, Dasarathy BV. Medical image fusion: A survey of the state of the art. Information Fusion.
2014; 19:4–19. https://doi.org/10.1016/j.inffus.2013.12.002
Lv XG, Jiang L, Liu J. Deblurring Poisson noisy images by total variation with overlapping group spar-
sity. Applied Mathematics and Computation. 2016; 289:132–148. https://doi.org/10.1016/j.amc.2016.
03.029
Liu J, Huang TZ, Lv XG, Wang S. High-order total variation-based Poissonian image deconvolution with
spatially adapted regularization parameter. Applied Mathematical Modelling. 2017; 45:516–529. https://
doi.org/10.1016/j.apm.2017.01.009
6. Kumar A, Ahmad MO, Swamy M. A framework for image denoising using first and second order frac-
tional overlapping group sparsity (HF-OLGS) regularizer. IEEE Access. 2019; 7:26200–26217. https://
doi.org/10.1109/ACCESS.2019.2901691
7. Hebert TJ, Leahy R. Statistic-based MAP image-reconstruction from Poisson data using Gibbs priors.
IEEE Transactions on Signal Processing. 1992; 40(9):2290–2303. https://doi.org/10.1109/78.157228
8. Sebastiani G, Godtliebsen F. On the use of Gibbs priors for Bayesian image restoration. Signal Pro-
cessing. 1997; 56(1):111–118. https://doi.org/10.1016/S0165-1684(97)00002-9
9. Benvenuto F, La Camera A, Theys C, Ferrari A, Lante´ri H, Bertero M. The study of an iterative method
for the reconstruction of images corrupted by Poisson and Gaussian noise. Inverse Problems. 2008; 24
(3):035016. https://doi.org/10.1088/0266-5611/24/3/035016
10.
11.
Jiang L, Huang J, Lv XG, Liu J. Restoring Poissonian images by a combined first-order and second-
order variation approach. Journal of Mathematics. 2013; 2013:1–11. https://doi.org/10.1155/2013/
274573
Jiang L, Huang J, Lv XG, Liu J. Alternating direction method for the high-order total variation-based
Poisson noise removal problem. Numerical Algorithms. 2015; 69(3):495–516. https://doi.org/10.1007/
s11075-014-9908-y
12. Bertero M, Lante´ri H, Zanni L, et al. Iterative image reconstruction: A point of view. Mathematical Meth-
ods in Biomedical Imaging and Intensity-Modulated Radiation Therapy (IMRT). 2008; 7:37–63.
13.
14.
Lanteri H, Roche M, Aime C. Penalized maximum likelihood image restoration with positivity con-
straints: multiplicative algorithms. Inverse Problems. 2002; 18(5):1397. https://doi.org/10.1088/0266-
5611/18/5/313
Landi G, Piccolomini EL. An improved Newton projection method for nonnegative deblurring of Pois-
son-corrupted images with Tikhonov regularization. Numerical Algorithms. 2012; 60(1):169–188.
https://doi.org/10.1007/s11075-011-9517-y
15. Bardsley JM, Vogel CR. A nonnegatively constrained convex programming method for image recon-
struction. SIAM Journal on Scientific Computing. 2004; 25(4):1326–1343. https://doi.org/10.1137/
S1064827502410451
16. Bardsley JM, Laobeul N. Tikhonov regularized Poisson likelihood estimation: theoretical justification
and a computational method. Inverse Problems in Science and Engineering. 2008; 16(2):199–215.
https://doi.org/10.1080/17415970701404235
17. Bonettini S, Landi G, Piccolomini EL, Zanni L. Scaling techniques for gradient projection-type methods
in astronomical image deblurring. International Journal of Computer Mathematics. 2013; 90(1):9–29.
https://doi.org/10.1080/00207160.2012.716513
18. Chen K. Introduction to variational image-processing models and applications. International Journal of
Computer Mathematics. 2013; 90:1–8. https://doi.org/10.1080/00207160.2012.757073
PLOS ONE | https://doi.org/10.1371/journal.pone.0250260 April 20, 2021
18 / 20
PLOS ONE Hybrid regularizer for image deblurring under poisson noise
19. Bardsley JM, Luttman A. Total variation-penalized Poisson likelihood estimation for ill-posed problems.
Advances in Computational Mathematics. 2009; 31(1-3):35. https://doi.org/10.1007/s10444-008-
9081-8
20.
21.
Landi G, Piccolomini EL. An efficient method for nonnegatively constrained Total Variation-based
denoising of medical images corrupted by Poisson noise. Computerized Medical Imaging and Graphics.
2012; 36(1):38–46. https://doi.org/10.1016/j.compmedimag.2011.07.002 PMID: 21821394
Tao S, Dong W, Xu Z, Tang Z. Fast total variation deconvolution for blurred image contaminated by
Poisson noise. Journal of Visual Communication and Image Representation. 2016; 38:582–594. https://
doi.org/10.1016/j.jvcir.2016.04.005
22. Bonettini S, Ruggiero V. An alternating extragradient method for total variation-based image restoration
from Poisson data. Inverse Problems. 2011; 27(9):095001.
23. Bonettini S, Ruggiero V. On the convergence of primal–dual hybrid gradient algorithms for total variation
image restoration. Journal of Mathematical Imaging and Vision. 2012; 44(3):236–253. https://doi.org/
10.1007/s10851-011-0324-9
24. Wen Y, Chan RH, Zeng T. Primal-dual algorithms for total variation based image restoration under Pois-
son noise. Science China Mathematics. 2016; 59(1):141–160. https://doi.org/10.1007/s11425-015-
5079-0
25. Setzer S, Steidl G, Teuber T. Deblurring Poissonian images by split Bregman techniques. Journal of
Visual Communication and Image Representation. 2010; 21(3):193–199. https://doi.org/10.1016/j.jvcir.
2009.10.006
26. Chang H, Lou Y, Duan Y, Marchesini S. Total variation–based phase retrieval for Poisson noise
removal. SIAM Journal on Imaging Sciences. 2018; 11(1):24–55. https://doi.org/10.1137/16M1103270
27.
28.
29.
30.
31.
Zhou W, Li Q. Poisson noise removal scheme based on fourth-order PDE by alternating minimization
algorithm. Abstract and Applied Analysis. 2012; 2012, Special Issue:1–14.
Le T, Chartrand R, Asaki TJ. A variational approach to reconstructing images corrupted by Poisson
noise. Journal of Mathematical Imaging and Vision. 2007; 27(3):257–263. https://doi.org/10.1007/
s10851-007-0652-y
Liu G, Huang TZ, Liu J, Lv XG. Total variation with overlapping group sparsity for image deblurring
under impulse noise. PloS one. 2015; 10(4):e0122562. https://doi.org/10.1371/journal.pone.0122562
PMID: 25874860
Zhang J, Ma M, Wu Z, Deng C. High-order total bounded variation model and its fast algorithm for Pois-
sonian image restoration. Mathematical Problems in Engineering. 2019; 2019:1–11.
Liu X, Huang L. Total bounded variation-based Poissonian images recovery by split Bregman iteration.
Mathematical Methods in the Applied Sciences. 2012; 35(5):520–529. https://doi.org/10.1002/mma.
1588
32. Wang XD, Feng Xc, Wang Ww, Zhang WJ. Iterative reweighted total generalized variation based Pois-
son noise removal model. Applied Mathematics and Computation. 2013; 223:264–277. https://doi.org/
10.1016/j.amc.2013.07.090
33.
Zhang Z, Zhang J, Wei Z, Xiao L. Cartoon-texture composite regularization based non-blind deblurring
method for partly-textured blurred images with Poisson noise. Signal Processing. 2015; 116:127–140.
https://doi.org/10.1016/j.sigpro.2015.04.020
34. Ma L, Moisan L, Yu J, Zeng T. A dictionary learning approach for Poisson image deblurring. IEEE
Transactions on Medical Imaging. 2013; 32(7):1277–1289. https://doi.org/10.1109/TMI.2013.2255883
PMID: 23549888
35.
Fang H, Yan L, Liu H, Chang Y. Blind Poissonian images deconvolution with framelet regularization.
Optics Letters. 2013; 38(4):389–391. https://doi.org/10.1364/OL.38.000389 PMID: 23455078
36. Shi Y, Song J, Hua X. Poissonian image deblurring method by non-local total variation and framelet reg-
ularization constraint. Computers & Electrical Engineering. 2017; 62:319–329. https://doi.org/10.1016/j.
compeleceng.2016.09.032
37.
38.
39.
Zhang H, Dong Y, Fan Q. Wavelet frame based Poisson noise removal and image deblurring. Signal
Processing. 2017; 137:363–372. https://doi.org/10.1016/j.sigpro.2017.01.025
Fang H, Yan L. Poissonian image deconvolution with analysis sparsity priors. Journal of Electronic
Imaging. 2013; 22(2):023033. https://doi.org/10.1117/1.JEI.22.2.023033
Liu J, Huang TZ, Liu G, Wang S, Lv XG. Total variation with overlapping group sparsity for speckle
noise reduction. Neurocomputing. 2016; 216:502–513. https://doi.org/10.1016/j.neucom.2016.07.049
40. Adam T, Paramesran R. Image denoising using combined higher order non-convex total variation with
overlapping group sparsity. Multidimensional Systems and Signal Processing. 2019; 30(1):503–527.
https://doi.org/10.1007/s11045-018-0567-3
PLOS ONE | https://doi.org/10.1371/journal.pone.0250260 April 20, 2021
19 / 20
PLOS ONE Hybrid regularizer for image deblurring under poisson noise
41. Ding M, Huang TZ, Wang S, Mei JJ, Zhao XL. Total variation with overlapping group sparsity for deblur-
ring images under Cauchy noise. Applied Mathematics and Computation. 2019; 341:128–147. https://
doi.org/10.1016/j.amc.2018.08.014
42.
Figueiredo MA, Bioucas-Dias JM. Restoration of Poissonian images using alternating direction optimi-
zation. IEEE Transactions on Image Processing. 2010; 19(12):3133–3145. https://doi.org/10.1109/TIP.
2010.2053941 PMID: 20833604
43. Gong P, Zhang C, Lu Z, Huang J, Ye J. A general iterative shrinkage and thresholding algorithm for
non-convex regularized optimization problems. In: International Conference on Machine Learning;
2013. p. 37–45.
44. Chen PY, Selesnick IW. Group-sparse signal denoising: non-convex regularization, convex optimiza-
tion. IEEE Transactions on Signal Processing. 2014; 62(13):3464–3478. https://doi.org/10.1109/TSP.
2014.2329274
45. Adam T, Paramesran R. Hybrid non-convex second-order total variation with applications to non-blind
image deblurring. Signal, Image and Video Processing. 2020; 14(1):115–123. https://doi.org/10.1007/
s11760-019-01531-3
46.
Liu J, Huang TZ, Selesnick IW, Lv XG, Chen PY. Image restoration using total variation with overlapping
group sparsity. Information Sciences. 2015; 295:232–246.
47. Chan TFC, Glowinski R. Finite element approximation and iterative solution of a class of mildly non-lin-
ear elliptic equations. Computer Science Department, Stanford University Stanford; 1978.
48. Xiu X, Liu W, Li L, Kong L. Alternating direction method of multipliers for nonconvex fused regression
problems. Computational Statistics & Data Analysis. 2019; 136:59–71. https://doi.org/10.1016/j.csda.
2019.01.002
49.
Lu C, Feng J, Yan S, Lin Z. A unified alternating direction method of multipliers by majorization minimi-
zation. IEEE Transactions on Pattern Analysis and Machine intelligence. 2018; 40(3):527–541. https://
doi.org/10.1109/TPAMI.2017.2689021 PMID: 28368818
50. Deng W, Yin W. On the global and linear convergence of the generalized alternating direction method of
multipliers. Journal of Scientific Computing. 2016; 66(3):889–916. https://doi.org/10.1007/s10915-015-
0048-x
51. Wang Y, Yin W, Zeng J. Global convergence of ADMM in nonconvex nonsmooth optimization. Journal
of Scientific Computing. 2019; 78(1):29–63. https://doi.org/10.1007/s10915-018-0757-z
52.
53.
54.
Lyu Q, Lin Z, She Y, Zhang C. A comparison of typical ℓp minimization algorithms. Neurocomputing.
2013; 119:413–424. https://doi.org/10.1016/j.neucom.2013.03.017
Lai MJ, Xu Y, Yin W. Improved iteratively reweighted least squares for unconstrained smoothed ℓq mini-
mization. SIAM Journal on Numerical Analysis. 2013; 51(2):927–957. https://doi.org/10.1137/
110840364
Foucart S, Lai MJ. Sparsest solutions of underdetermined linear systems via ℓq–minimization for 0 < q �
1. Applied and Computational Harmonic Analysis. 2009; 26(3):395–407.
55. Chartrand R, Yin W. Iteratively reweighted algorithms for compressive sensing. In: IEEE International
Conference on Acoustics, Speech and Signal Processing. IEEE; 2008. p. 3869–3872.
56.
57.
Lv XG, Li F. An iterative decoupled method with weighted nuclear norm minimization for image restora-
tion. International Journal of Computer Mathematics. 2020; 97(3):602–623. https://doi.org/10.1080/
00207160.2019.1581178
Figueiredo MA, Bioucas-Dias JM, Nowak RD. Majorization–minimization algorithms for wavelet-based
image restoration. IEEE Transactions on Image processing. 2007; 16(12):2980–2991. https://doi.org/
10.1109/TIP.2007.909318 PMID: 18092597
58. Woo H. A characterization of the domain of Beta-Divergence and its connection to Bregman variational
model. Entropy. 2017; 19(9):482. https://doi.org/10.3390/e19090482
59.
Levin A, Weiss Y, Durand F, Freeman WT. Understanding and evaluating blind deconvolution algo-
rithms. In: 2009 IEEE Conference on Computer Vision and Pattern Recognition. IEEE; 2009. p. 1964–
1971.
PLOS ONE | https://doi.org/10.1371/journal.pone.0250260 April 20, 2021
20 / 20
PLOS ONE |
10.1371_journal.pone.0238646 | RESEARCH ARTICLE
Endoscopic soft palate augmentation using
injectable materials in dogs to ameliorate
velopharyngeal insufficiency
Emiko Tanaka IsomuraID*, Makoto Matsukawa☯, Kiyoko Nakagawa☯, Ryo Mitsui☯,
Mikihiko Kogo☯
First Department of Oral and Maxillofacial Surgery, Osaka University, Graduate School of Dentistry, Suita
City, Osaka, Japan
☯ These authors contributed equally to this work.
* tanaemi@dent.osaka-u.ac.jp
Abstract
Background
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: Isomura ET, Matsukawa M, Nakagawa K,
Mitsui R, Kogo M (2020) Endoscopic soft palate
augmentation using injectable materials in dogs to
ameliorate velopharyngeal insufficiency. PLoS ONE
15(9): e0238646. https://doi.org/10.1371/journal.
pone.0238646
Velopharyngeal structure augmentation methods are used as alternatives to pharyngeal
flap operations. Recently, we investigated the sites of velopharyngeal structure augmenta-
tion in dogs and reported that the most effective injection location is the soft palate. How-
ever, there have been no reports regarding the optimal materials for implantation or
injection. In this study, we aimed to investigate the injectable materials used in soft palate
augmentation in dogs to ameliorate velopharyngeal insufficiency (VPI).
Editor: Mrinmoy Sanyal, Stanford University
School of Medicine, UNITED STATES
Methods
Received: February 25, 2020
Accepted: August 20, 2020
Published: September 4, 2020
Copyright: © 2020 Isomura et al. This is an open
access article distributed under the terms of the
Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: All relevant data are
within the paper.
Funding: This study was supported by Grants-in-
Aid for Scientific Research from the Japan Society
for the Promotion of Science (Grant Number
19K19229). Neither of the authors has a financial
interest in any of the products, devices, or drugs
mentioned in this manuscript.
Competing interests: The authors have declared
that no competing interests exist.
Endoscopic soft palate augmentation (ESPA) was performed in dogs using purified sodium
hyaluronate, atelocollagen, or autogenic fat tissue. ESPA is an original technique devel-
oped by our group, and this is the first report of its performance. Moreover, we assessed
the amount of nasal air leakage during inspiration at rest and during expiration under the
rebreathing system at 1, 2, 3, 4, 5, and 6 months after injection of these materials.
Results
The amount of nasal air leakage during expiration under the rebreathing system was signifi-
cantly decreased in all dogs injected with the ESPA materials, but neither apnea nor hypop-
nea was observed.
Conclusions
We investigated the optimal materials for use in ESPA, such as purified sodium hyaluronate,
atelocollagen, or autogenic fat tissue. We found that all of them reduced nasal air leakage
and only autogenic fat tissue showed significant histologic differences in dogs at 6 months.
This technique may also be useful for the treatment of patients with VPI.
PLOS ONE | https://doi.org/10.1371/journal.pone.0238646 September 4, 2020
1 / 11
PLOS ONE Endoscopic soft palate augmentation in dogs
Introduction
When treating patients with cleft palate, velopharyngeal insufficiency (VPI) can sometimes
occur after palatoplasty. VPI is the failure of the nose and mouth to separate during speech
because of an anatomical dysfunction of the soft palate. Many cases of VPI are due to shortfall,
poor movement of the soft palate caused by scarring, or poor reconstruction of the muscles of
the soft palate. Furthermore, patients with 22q11.2 deletion syndrome have VPI due to inade-
quate soft palate muscle formation.
In our hospital, speech therapy is the first step in the treatment of VPI. If VPI cannot be
managed with speech therapy, a speech aid is used for closure of the nasopharynx by lifting the
soft palate or filling the gap. Then, after VPI is shown to be controlled with the speech aid, pha-
ryngeal flap surgery is performed to wean the patient off the speech aid [1–3]. However, it is
difficult to apply this treatment in children because it causes fundamental changes to the velo-
pharyngeal form, which may result in sleep apnea or inability to perform nasal intubation dur-
ing future orthodontic surgeries [4–7].
Several reports have described another method for treating VPI [8–22]. Velopharyngeal
structure augmentation is an alternative to pharyngeal flap surgery that utilizes an injectable
material implanted into the tissue around the velopharynx. However, it is not yet a standard
treatment because it has not been extensively studied. Anatomic sites and injection materials
vary widely, owing to the lack of standardized criteria, and their effects also differ among
institutions.
Recently, we investigated the sites of velopharyngeal structure augmentation in dogs and
reported that the most effective injection location is the soft palate, rather than the posterior
pharyngeal wall or bilateral pharyngeal walls [23]. Dogs’ velopharynx exhibit inherently like
VPI; thus, the rhinopharynx is not completely closed, even when the soft palate is lifted [24].
We injected saline intraorally, in 1-mL increments, into the nasal side of the soft palate, poste-
rior pharyngeal wall, or bilateral pharyngeal walls of each dog. The soft palate that was injected
with saline achieved steady augmentation, and nasal air leakage disappeared following the
5-mL saline injection. Conversely, nasal air leakage persisted in the dogs with saline injected in
the posterior pharyngeal wall or bilateral pharyngeal walls.
There have been no reports about the optimal materials for implantation or injection,
however. There are various artificial and biological materials that may be used in the velophar-
yngeal structure, including silicone, Teflon, porous polyethylene, Gore-Tex1, calcium
hydroxyapatite, auricular or costal cartilage, and autologous fat [8–22]. However, when we
augment the soft palate, the material needs to be injectable, because it cannot be implanted
into the nasal side of the soft palate without damaging the levator veli palatini.
In this study, we aimed to investigate injectable materials for soft palate augmentation in
dogs for treatment of VPI. Furthermore, we sought to introduce the endoscopic soft palate
augmentation (ESPA) technique, because it causes less damage to the levator veli palatini than
other techniques. ESPA is an original technique developed by us, and this is the first report of
its performance. If the materials can keep the volume after injection, this technique may also
be useful for the treatment of patients with VPI.
Materials and methods
ESPA was performed at the Large Animal Laboratory of the Graduate School of Dentistry at
Osaka University using 11 beagles (TOYO beagle; Oriental Yeast Co., Tokyo, Japan), aged 20–
24 months and weighing 9–12 kg. All dogs were housed in separate cages and were provided
solid food (Oriental Yeast Co., Tokyo, Japan) and water ad libitum. All experimental protocols
PLOS ONE | https://doi.org/10.1371/journal.pone.0238646 September 4, 2020
2 / 11
PLOS ONE Endoscopic soft palate augmentation in dogs
were reviewed and approved by the Intramural Animal Care and Use Committee of Osaka
University Graduate School of Dentistry (approval number: 29-004-0).
All procedures were performed under general anesthesia administered via an intramuscular
injection of medetomidine (0.02 mg/kg) and midazolam (0.3 mg/kg), followed by an intraperi-
toneal injection of sodium pentobarbital (25 mg/kg) 15 minutes later. Animals were fixed in
the supine position after the ventilation tube was passed through the mouth, and all efforts
were made to minimize suffering.
Using an electric knife, an approximately 8-mm hole was made in the most anterior part of
the soft palate in each dog. This hole was needed because the endoscope cannot be inserted
nasally in dogs, due to narrowness of the canine nasal cavity. The endoscope (i-Vets 8.0;
SCETI K., Tokyo, Japan) was then inserted into the nasal side of the soft palate, and purified
sodium hyaluronate (Hyaluronate Na1, Sawai Pharmaceutical Co., Ltd. Osaka, Japan, n = 3),
atelocollagen (Koken Atelocollagen implant1, Koken Co., Ltd. Tokyo, Japan, n = 3), or auto-
genic fat tissue (n = 4) was injected into the nasal mucosal side of the anterior two-thirds of the
soft palate using a 23-G needle (Interject™, Boston Scientific, Natick, USA) under endoscopic
guidance to directly confirm entrance into the nasal mucosa (Fig 1).
Autogenic fat tissue was taken from the greater omentum and refined using the Coleman
method [25]. Vascular tissue was removed visually from the extracted greater omentum and
centrifuged (3000 rpm, 3 minutes) to separate the three layers (upper layer: oil from crushed
fat cell, middle layer: fat cells, bottom layer: blood, water, and lidocaine used as local anesthe-
sia). Only the middle layer was used as an injection material. Approximately 2 ml of each
material was injected into each dog until the soft palates slightly touched the post-pharyngeal
walls (Fig 2).
We then assessed the amount of nasal air leakage during inspiration at rest and expiration
under a rebreathing system prior to ESPA (= non-treated) and 1, 2, 3, 4, 5, and 6 months after
injection of the materials, as described previously [23]. The tip of the ventilation tube was with-
drawn from the trachea to the oral cavity to allow expiration through the nasal cavity. After
removing the electrode and ventilation tube to prevent oral air leakage during measurement,
the oral cavity was filled with an alginate impression material. While the dogs were under the
rebreathing system, the amount of air leakage from the nasal cavity was measured by a flow
meter (TSD117; BIOPAC Systems Inc., Japan) using the rubber tubes connected to the flow
meter’s sensor in front of both nasal apertures. The external portion of the rubber tubes was
packed with quick, self-curing acrylic resin (UNIFAST II; GC Co., Tokyo, Japan) to prevent
air leakage. Data from the flow meter was recorded on a personal computer (U24a-px3210r
Windows7; ASUSTek Computer Inc., Japan) using data acquisition and analysis software
(Labchart7; AD Instruments, Japan) through a DC Amplifier (DA100C; BIOPAC Systems
Inc., Japan), an analog output module (HLT100-C; BIOPAC Systems Inc., Japan), and an AD
converter (Power lab; AD Instruments Co., Tokyo, Japan).
Data from the nasal air leakage of one breath was separated into the inspiration phase and
expiration phase, and their integral value was measured. We assessed the amount of nasal air
leakage during inspiration at rest to determine the presence of apnea or hypopnea and nasal
air leakage during expiration under the rebreathing system to evaluate the effect of ESPA.
After euthanasia at 6 months post-injection, histological examinations were performed in all
dogs. The soft palates were dissected and stained with hematoxylin and eosin.
Normality of the data was evaluated and, owing to their nonparametric nature, analyzed
using the Kruskal–Wallis test post-hoc Mann–Whitney’s U test (p-value < 0.05). All statistical
analyses were conducted using R version 2.8.1 (CRAN: https://cran-archive.r-project.org/bin/
windows/base/old/2.8.1/).
PLOS ONE | https://doi.org/10.1371/journal.pone.0238646 September 4, 2020
3 / 11
PLOS ONE Endoscopic soft palate augmentation in dogs
Fig 1. The schema of endoscopic soft palate augmentation (ESPA). An approximately 8-mm hole was created in the most
anterior part of the soft palate using an electric knife to insert the endoscope. This hole needed because the endoscope could not
insert by nasal approach in dog, due to narrowness of the canine nasal cavity. Then, the endoscope was inserted to the nasal side
of the soft palate.
https://doi.org/10.1371/journal.pone.0238646.g001
Results
The changes of nasal air leakage, presented in medians and inter-quartile ranges that occur
over time are shown in Figs 3 and 4. Fig 3 shows nasal air leakage during inspiration at rest
(soft palate was not lifted), and Fig 4 shows nasal air leakage during expiration under
Fig 2. Endoscopic image during ESPA. Approximately 2-ml of the materials was injected to each of the dogs until the soft palate
slightly touched the post-pharyngeal wall.
https://doi.org/10.1371/journal.pone.0238646.g002
PLOS ONE | https://doi.org/10.1371/journal.pone.0238646 September 4, 2020
4 / 11
PLOS ONE Endoscopic soft palate augmentation in dogs
Fig 3. The amount of nasal air leakage during inspiration at rest. The changes of nasal air leakage during inspiration at rest
(soft palate was not lifted). The amount of nasal air leakage during inspiration in each dog decreased slightly, compared to the
pre-ESPA value, but the decrease was not enough to cause apnea or hypopnea. (0 = Immediately after ESPA).
https://doi.org/10.1371/journal.pone.0238646.g003
rebreathing (soft palate was lifted due to levator veli palatini action). The amount of nasal air
leakage during inspiration in each dog decreased slightly, compared to the pre-ESPA value,
but the decrease was not enough to cause apnea or hypopnea (Fig 3). Comparison of data at 6
months post-injection among the three materials is shown in Fig 5, and no significant differ-
ence was found.
Conversely, the amount of nasal air leakage during expiration under the rebreathing system
was significantly decreased in all dogs injected with materials used for ESPA, compared with
pre-ESPA (p<0.05) (Fig 4). The median amount of nasal air leakage during expiration in the
non-treated dogs (pre-ESPA: n = 10) was 0.16 L/sec, whereas at 6 months after ESPA, the
median amount of nasal air leakage during expiration was 0.055, 0.089, and 0.049 L/sec in
dogs injected with purified sodium hyaluronate, atelocollagen, and autogenic fat tissue, respec-
tively (Fig 6).
Histologically, the maximum soft palate thickness between the dogs injected with sodium
hyaluronate or atelocollagen and non-treated dogs was the same (Fig 7; Table 1). Fat tissues
were observed around the soft palate injection site of the dogs injected with purified sodium
hyaluronate, whereas fibrous tissues were observed in those injected with atelocollagen. In
contrast, fibrous tissues and vasculatures, appearing as lymphatic or blood vessels with mini-
mal muscle tissues, were observed around the injection site in the dogs injected with autoge-
nous fat.
PLOS ONE | https://doi.org/10.1371/journal.pone.0238646 September 4, 2020
5 / 11
PLOS ONE Endoscopic soft palate augmentation in dogs
Fig 4. The amount of nasal air leakage during expiration under rebreathing. The changes of nasal air leakage
during expiration under rebreathing (soft palate was lifted due to levator veli palatini action). The amount of nasal air
leakage during expiration under the rebreathing system was significantly decreased in all dogs injected with materials
used for ESPA, compared with pre-ESPA (p<0.05). (0 = Immediately after ESPA).
https://doi.org/10.1371/journal.pone.0238646.g004
Fig 5. The amount of nasal air leakage during inspiration at rest. The amount of nasal air leakage during inspiration at rest
decreased in all dogs compared to the pre-ESPA value, but apnea or hypopnea was not observed. Moreover, no significant
difference in the outcomes was observed among the materials.
https://doi.org/10.1371/journal.pone.0238646.g005
PLOS ONE | https://doi.org/10.1371/journal.pone.0238646 September 4, 2020
6 / 11
PLOS ONE Endoscopic soft palate augmentation in dogs
Fig 6. The amount of nasal air leakage during expiration under the rebreathing system. The amount of nasal air leakage
during expiration under the rebreathing system decreased significantly in all the dogs injected with any of the materials during
ESPA. The median amount of nasal air leakage of the non-treated dogs (pre-ESPA: n = 10) was 0.16 L/sec, whereas at 6 months
after ESPA, the median amount of nasal air leakage during expiration was 0.055, 0.089, and 0.049 L/sec in dogs injected with
purified sodium hyaluronate, atelocollagen, and autogenic fat tissue, respectively.
https://doi.org/10.1371/journal.pone.0238646.g006
Fig 7. Histological results. The maximum value of the soft palate thickness between the dogs injected with sodium hyaluronate,
atelocollagen and non-treated dogs was the same. Fat tissues were observed around the injection site of the soft palate of the dogs
injected with purified sodium hyaluronate, whereas fibrous tissues were observed in those injected with atelocollagen. In contrast,
fibrous tissues and vasculatures, appearing as lymphatic or blood vessels, with minimal muscle tissues (arrow) were observed
around the injection site in the dogs injected with autogenous fat. (Asterisks are levator veli palatini).
https://doi.org/10.1371/journal.pone.0238646.g007
PLOS ONE | https://doi.org/10.1371/journal.pone.0238646 September 4, 2020
7 / 11
PLOS ONE Endoscopic soft palate augmentation in dogs
Table 1. The thickness of the soft palate at 6 months after the endoscopic soft palate augmentation (ESPA) (mm).
Non-treatment
7.03 (6.95–7.63)
Sodium hyaluronate
6.84 (6.55–6.85)
atelocollagen
7.56 (7.45–8.19)
Fat tissue
10.0 (8.58–11.1)
https://doi.org/10.1371/journal.pone.0238646.t001
Discussion
In this study, we investigated various materials that can be used in ESPA. ESPA is especially
useful because it does not cause injury to the musculus levator veli palatine, and injectable
materials can be implanted while simultaneously monitoring the volume of augmentation
through an endoscope. Injection of materials intraorally has been demonstrated previously,
but this method is challenging when trying to inject into the mucosal layer, as the needle may
penetrate the nasal mucosa [22]. Using our technique, we had to create a hole to insert the
endoscope because the device cannot pass through the nasal cavity in dogs. However, in
humans it can pass through the nasal cavity; thus, creating a hole is not necessary.
We selected purified sodium hyaluronate, atelocollagen, and autogenic fat tissue as the
materials to investigate because they can be injected into the soft palate through a needle.
Sodium hyaluronate is a mucopolysaccharide, and its molar weight is 1,000,000 g/mol, while
atelocollagen is a protein and its molar weight is 300,000 g/mol. One gram of sodium hyaluro-
nate can hold 6,000 ml of water; hence, even if sodium hyaluronate itself is absorbed gradually,
the surrounding water remains, and that makes it easy to retain the entire volume for an
extended period. And all the materials reduced the amount of nasal air leakage during expira-
tion under rebreathing. Soft palate movement was not hindered by any of the materials, and
apnea or hypopnea did not occur in any of the dogs.
Yasuda investigated histological changes in the skin after injection with sodium hyaluronate
or atelocollagen [26]. He reported that the bulging area where the sodium hyaluronate was
injected disappeared after 60 days, and only a small amount of sodium hyaluronate remained
in the site. He also reported minimal fibrous tissue was observed histologically at 180 days
after injection. Conversely, the bulging area where atelocollagen was injected remained even at
180 days after injection, and the material converted to collagen. In our study, sodium hyaluro-
nate was not observed histologically at 6 months after injection, whereas atelocollagen still
remained at the 6-month follow-up, and new collagen was observed around the musculus leva-
tor veli palatini after injection, consistent with Yasuda’s findings.
Multiple studies have reported that 30%–70% of the autologous transplant of fat graft was
resorbed within a year [27–30]. Guerrerosantos et al. reported that fat grafts injected intramus-
cularly were successful, owing to the excellent circulation of muscular tissue, compared to
those injected subcutaneously [30]. In our study, we could not confirm whether the autogenic
fat tissue was injected intramuscularly or subcutaneously, but histologically, no damage to the
levator veli palatini by the injection was observed. Only a small amount of fat tissue remained
around the muscle tissue during follow-up, and vascularization was observed histologically in
the mucosa layer. Soft palate thickness was biggest in the dogs injected with autogenic fat tis-
sue. It was unclear, however, if this observation was due to increased muscle tissue in the soft
palate. By the final observation, no fat tissue remained in any of the animals, instead it was
replaced by fibrous tissue. This result may indicate that the thickness of the soft palate may
remain for a long period of time.
Although ESPA using purified sodium hyaluronate, atelocollagen, or autogenic fat tissue
showed good results after 6 months, its long-term effects should still be considered. Several
reports have described that human lipoaspirate contains multipotent cells and may represent an
alternative stem cell source for bone marrow-derived mesenchymal stem cells [31, 32]. Li et al.
PLOS ONE | https://doi.org/10.1371/journal.pone.0238646 September 4, 2020
8 / 11
PLOS ONE Endoscopic soft palate augmentation in dogs
reported that fat grafts consisting of 105/ml adipose-derived stem cells constitute an ideal trans-
plant strategy, which may result in decreased absorption and accelerated fat regeneration [33].
Adipose-derived stem cells are one of the most widely used stem cell types for the treatment of
bone and cartilage disease, Crohn’s disease, heart diseases, kidney disease, neurological disease
and respiratory disease, as well as for cosmetic and plastic surgery [34]. In addition, some reports
describe the effectiveness of injectable adipose tissue-derived stem cells in treating stress urinary
incontinence or vocal fold paralysis [35–37]. In the future, we should also investigate whether
transplantation of adipose tissue- derived stem cells can help in treating VPI. If we can make the
musculus levator veli palatini stronger with this method, it may be the best treatment for VPI.
This study has several limitations. The structure of the velopharynx in dogs and that in
humans are quite different, and the assessment of speech cannot be performed in dogs. In
dogs, the larynx is located directly behind the base of the tongue and soft palate and lies
between the pharynx and trachea. The larynx covers the trachea during swallowing so that
food does not enter the windpipe. However, the soft palate of dogs may lift like that of humans
during respiration. VPI cannot strictly be assessed in dogs; only analogous comparisons can be
made. Therefore, we eventually must conduct further investigations in humans. We plan to
perform ESPA on humans in a clinical setting soon.
In a systematic review, Nigh E et al. described that autologous fat injection has been advo-
cated for correction of mild to moderate VPI, but it was difficult to adapt in severe VPI [22].
At present, no ideal treatment is available for severe VPI. However, we have previously
reported that the optimal site for injection is the nasal side of the soft palate and improvement
of VPI was dependent on the amount of injected autologous fat [23]. ESPA may control the
amount of fat easier than traditional methods; hence, we believe it could be adapted to treat
severe VPI as well.
In conclusion, we investigated the optimal materials for use in ESPA, such as purified
sodium hyaluronate, atelocollagen, or autogenic fat tissue. We found that all of them reduced
nasal air leakage and only autogenic fat tissue showed significant histologic differences at 6
months in dogs. These results suggest that this technique may also be useful for the treatment
of patients with VPI.
Acknowledgments
We are grateful to Y. Sato for taking care of the animals.
Author Contributions
Conceptualization: Emiko Tanaka Isomura, Mikihiko Kogo.
Data curation: Emiko Tanaka Isomura, Makoto Matsukawa, Kiyoko Nakagawa, Ryo Mitsui.
Formal analysis: Emiko Tanaka Isomura, Makoto Matsukawa, Kiyoko Nakagawa, Ryo Mitsui.
Funding acquisition: Kiyoko Nakagawa.
Methodology: Emiko Tanaka Isomura, Makoto Matsukawa, Kiyoko Nakagawa, Ryo Mitsui.
Project administration: Emiko Tanaka Isomura, Makoto Matsukawa, Kiyoko Nakagawa, Ryo
Mitsui, Mikihiko Kogo.
Resources: Emiko Tanaka Isomura, Makoto Matsukawa, Kiyoko Nakagawa, Ryo Mitsui.
Writing – original draft: Emiko Tanaka Isomura.
Writing – review & editing: Emiko Tanaka Isomura.
PLOS ONE | https://doi.org/10.1371/journal.pone.0238646 September 4, 2020
9 / 11
PLOS ONE Endoscopic soft palate augmentation in dogs
References
1. Rogers C, Konofaos P, Wallace RD. Superiorly based pharyngeal flap for the surgical treatment of velo-
pharyngeal insufficiency and speech outcomes. J Craniofac Surg. 2016; 27: 1746–1749. https://doi.org/
10.1097/SCS.0000000000003050 PMID: 27763974
2. Naran S, Ford M, Losee JE. What’s new in cleft palate and velopharyngeal dysfunction management?
Plast Reconstr Surg. 2017; 139: 1343e–1355e. https://doi.org/10.1097/PRS.0000000000003335
PMID: 28538580
3. Kogo M, Sakai T, Harada T, Nohara K, Tanaka E, Seikai T, et al. Our unified pharyngeal flap operation.
J Cleft Lip Palate Craniofac Anomal. 2017; 4, Suppl S1: 189–S191.
4. Orr WC, Levine NS, Buchanan RT. Effect of cleft palate repair and pharyngeal flap surgery on upper air-
way obstruction during sleep. Plast Reconstr Surg. 1987; 80: 226–232. https://doi.org/10.1097/
00006534-198708000-00010 PMID: 3602172
5. Valnicek SM, Zuker RM, Halpern LM, Roy WL. Perioperative complications of superior pharyngeal flap
surgery in children. Plast Reconstr Surg. 1994; 93: 954–958. https://doi.org/10.1097/00006534-
199404001-00009 PMID: 8134488
6. Aboloyoun AI, Ghorab S, Farooq MU. Palatal lifting prosthesis and velopharyngeal insufficiency: Prelimi-
nary report. Acta Med Acad. 2013; 42: 55–60. https://doi.org/10.5644/ama2006-124.71 PMID: 23735067
7. Campos LD, Trindade-Suedam IK, Sampaio-Teixeira ACM, Yamashita RP, Lauris JRP, Lorenzi-Filho
G, et al. Obstructive sleep apnea following pharyngeal flap surgery for velopharyngeal insufficiency: A
prospective polysomnographic and aerodynamic study in middle-aged adults. Cleft Palate Craniofac J.
2016; 53: e53–e59. https://doi.org/10.1597/14-152 PMID: 25794015
8. Brauer RO. Retropharyngeal implantation of silicone gel pillows for velopharyngeal incompetence.
Plast Reconstr Surg. 1973; 51: 254–262. https://doi.org/10.1097/00006534-197303000-00003 PMID:
4266055
9. Remacle M, Bertrand B, Eloy P, Marbaix E. The use of injectable collagen to correct velopharyngeal
insufficiency. Laryngoscope. 1990; 100: 269–274. https://doi.org/10.1288/00005537-199003000-
00011 PMID: 2308450
10. Dejonckere PH, van Wijngaarden HA. Retropharyngeal autologous fat transplantation for congenital
short palate: A nasometric assessment of functional results. Ann Otol Rhinol Laryngol. 2001; 110: 168–
172. https://doi.org/10.1177/000348940111000213 PMID: 11219525
11. Klotz DA, Howard J, Hengerer AS, Slupchynskj O. Lipoinjection augmentation of the soft palate for velo-
pharyngeal stress incompetence. Laryngoscope. 2001; 111: 2157–2161. https://doi.org/10.1097/
00005537-200112000-00015 PMID: 11802016
12.
13.
Leuchter I, Schweizer V, Hohlfeld J, Pasche P. Treatment of velopharyngeal insufficiency by autologous
fat injection. Eur Arch Otorhinolaryngol. 2010; 267: 977–983. https://doi.org/10.1007/s00405-009-1157-
7 PMID: 20033195
Leboulanger N, Blanchard M, Denoyelle F, Glynn F, Charrier J-B, Roger G, et al. Autologous fat transfer
in velopharyngeal insufficiency: Indications and results of a 25 procedures series. Int J Pediatr Otorhino-
laryngol. 2011; 75: 1404–1407. https://doi.org/10.1016/j.ijporl.2011.08.001 PMID: 21872348
14. Cantarella G, Mazzola RF, Mantovani M, Baracca G, Pignataro L. Treatment of velopharyngeal insuffi-
ciency by pharyngeal and velar fat injections. Otolaryngol Head Neck Surg. 2011; 145: 401–403.
https://doi.org/10.1177/0194599811411655 PMID: 21636838
15.
Filip C, Matzen M, Aagenæs I, Aukner R, Kjøll L, Høgevold HE, et al. Speech and magnetic resonance
imaging results following autologous fat transplantation to the velopharynx in patients with velopharyn-
geal insufficiency. Cleft Palate Craniofac J. 2011; 48: 708–716. https://doi.org/10.1597/09-161 PMID:
21463181
16. Cantarella G, Mazzola RF, Mantovani M, Mazzola IC, Baracca G, Pignataro L. Fat injections for the
treatment of velopharyngeal insufficiency. J Craniofac Surg. 2012; 23: 634–637. https://doi.org/10.
1097/SCS.0b013e31824db85b PMID: 22565869
17.
18.
Filip C, Matzen M, Aagenæs I, Aukner R, Kjøll L, Høgevold HE, et al. Autologous fat transplantation to
the velopharynx for treating persistent velopharyngeal insufficiency of mild degree secondary to overt or
submucous cleft palate. J Plast Reconstr Aesthet Surg. 2013; 66: 337–344. https://doi.org/10.1016/j.
bjps.2012.11.006 PMID: 23254179
Lau D, Oppenheimer AJ, Buchman SR, Berger M, Kasten SJ. Posterior pharyngeal fat grafting for velo-
pharyngeal insufficiency. Cleft Palate Craniofac J. 2013; 50: 51–58. https://doi.org/10.1597/11-038
PMID: 22329568
19. Bishop A, Hong P, Bezuhly M. Autologous fat grafting for the treatment of velopharyngeal insufficiency:
State of the art. J Plast Reconstr Aesthet Surg. 2014; 67: 1–8. https://doi.org/10.1016/j.bjps.2013.09.
021 PMID: 24090720
PLOS ONE | https://doi.org/10.1371/journal.pone.0238646 September 4, 2020
10 / 11
PLOS ONE Endoscopic soft palate augmentation in dogs
20.
Filip C. Response re: ‘Autologous fat grafting for the treatment of velopharyngeal insufficiency: State of
the art’. J Plast Reconstr Aesthet Surg. 2014; 67: 1155–1156. https://doi.org/10.1016/j.bjps.2014.02.
001 PMID: 24581953
21. Boneti C, Ray PD, Macklem EB, Kohanzadeh S, de la Torre J, Grant JH. Effectiveness and safety of
autologous fat grafting to the soft palate alone. Ann Plast Surg. 2015; 74, Suppl 4: S190–S192.
22. Nigh E, Rubio GA, Hillam J, Armstrong M, Debs L, Thaller SR. Autologous fat injection for treatment of
velopharyngeal insufficiency. J Craniofac Surg. 2017: 28: 1248–1254. https://doi.org/10.1097/SCS.
0000000000003702 PMID: 28582294
23.
Isomura ET, Nakagawa K, Matsukawa M, Mitsui R, Kogo M. Evaluation of sites of velopharyngeal struc-
ture augmentation in dogs for improvement of velopharyngeal insufficiency. PLoS One. 2019; 14:
e0212752. https://doi.org/10.1371/journal.pone.0212752 PMID: 30802272
24. Adachi T, Kogo M, Iida S, Hamaguchi M, Matsuya T. Measurement of velopharyngeal movements
induced by isolated stimulation of levator veli palatini and pharyngeal constrictor muscles. J Dent Res.
1997; 76: 1745–1750. https://doi.org/10.1177/00220345970760110501 PMID: 9372791
25. Coleman SR. Hand rejuvenation with structural fat grafting. Plast Reconstr Surg. 2002; 110: 1731–
1744 PMID: 12447057
26. Yasuda H. Comparative study of skin tissue reactions after hyaluronic acid or collagen injection. Journal
of Kanazawa Medical University. 2006; 31: 233–241. (in Japanese).
27. Chajchir A, Benzaquen I. Liposuction fat grafts in face wrinkles and hemifacial atrophy. Aesthetic Plast
Surg. 1986; 10: 115–117. https://doi.org/10.1007/BF01575279 PMID: 3739848
28. Niechajev I, Sevćuk O. Long-term results of fat transplantation: Clinical and histologic studies. Plast
Reconstr Surg. 1994; 94: 496–506. https://doi.org/10.1097/00006534-199409000-00012 PMID:
8047602
29.
Leong DTW, Hutmacher DW, Chew FT, Lim T-C. Viability and adipogenic potential of human adipose
tissue processed cell population obtained from pump-assisted and syringe-assisted liposuction. J Der-
matol Sci. 2005; 37: 169–176. https://doi.org/10.1016/j.jdermsci.2004.11.009 PMID: 15734286
30. Guerrerosantos J, Gonzalez-Mendoza A, Masmela Y, Gonzalez MA, Deos M, Diaz P. Long-term sur-
vival of free fat grafts in muscle: An experimental study in rats. Aesthetic Plast Surg. 1996; 20: 403–408.
https://doi.org/10.1007/BF02390315 PMID: 8849432
31.
Zuk PA, Zhu M, Mizuno H, Huang J, Futrell JW, Katz AJ, et al. Multilineage cells from human adipose
tissue: Implications for cell-based therapies. Tissue Eng. 2001; 7: 211–228. https://doi.org/10.1089/
107632701300062859 PMID: 11304456
32. Coleman SR. Structural fat grafting: More than a permanent filler. Plast Reconstr Surg. 2006; 118:
108S–120S. https://doi.org/10.1097/01.prs.0000234610.81672.e7 PMID: 16936550
33.
Li K, Li F, Li J, Wang H, Zheng X, Long J, et al. Increased survival of human free fat grafts with varying
densities of human adipose-derived stem cells and platelet-rich plasma. J Tissue Eng Regen Med.
2017; 11: 209–219. https://doi.org/10.1002/term.1903 PMID: 24978937
34. Sheykhhasan M, Wong JKL, Seifalian AM. Human adipose-derived stem cells with great therapeutic
potential. Curr Stem Cell Res Ther. 2019; 14: 532–548. https://doi.org/10.2174/
1574888X14666190411121528 PMID: 30973112
35.
Lin G, Wang G, Banie L, Ning H, Shindel AW, Fandel TM, et al. Treatment of stress urinary incontinence
with adipose tissue-derived stem cells. Cytotherapy. 2010; 12: 88–95. https://doi.org/10.3109/
14653240903350265 PMID: 19878076
36. Wanatabe T, Maruyama S, Yamamoto T, Kamo I, Yasuda K, Saka Y, et al. Increased urethral resis-
tance by periurethral injection of low serum cultured adipose-derived mesenchymal stromal cells in rats.
Int J Urol. 2011; 18: 659–666. https://doi.org/10.1111/j.1442-2042.2011.02795.x PMID: 21707765
37. Nishio N, Fujimoto Y, Suga K, Iwata Y, Toriyama K, Takanari K, et al. Autologous fat injection therapy
including a high concentration of adipose-derived regenerative cells in a vocal fold paralysis model: Ani-
mal pilot study. J Laryngol Otol. 2016; 130: 914–922. https://doi.org/10.1017/S0022215116008707
PMID: 27604559
PLOS ONE | https://doi.org/10.1371/journal.pone.0238646 September 4, 2020
11 / 11
PLOS ONE |
10.1371_journal.pone.0248369 | RESEARCH ARTICLE
Climbing since the early Miocene: The fossil
record of Paullinieae (Sapindaceae)
Nathan A. JudID
G. Chery5*
1,3*, Sarah E. AllenID
2, Chris W. Nelson3¤, Carolina L. Bastos4, Joyce
1 Department of Biology, William Jewell College, Liberty, MO, United States of America, 2 Department of
Biology, Penn State Altoona, Altoona, PA, United States of America, 3 Florida Museum of Natural History,
University of Florida, Gainesville, FL, United States of America, 4 Laboratory of Plant Anatomy, Department of
Botany, Instituto de Biociências, Universidade de São Paulo, São Paulo, SP, Brazil, 5 School of Integrative
Plant Sciences, Section of Plant Biology and the L.H. Bailey Hortorium, Cornell University, Ithaca, NY, United
States of America
¤ Current address: Gainesville, FL, United States of America
* judn@william.jewell.edu (NAJ); jgc235@cornell.edu (JGC)
Abstract
Paullinieae are a diverse group of tropical and subtropical climbing plants that belong to the
soapberry family (Sapindaceae). The six genera in this tribe make up approximately one-
quarter of the species in the family, but a sparse fossil record limits our understanding of
their diversification. Here, we provide the first description of anatomically preserved fossils
of Paullinieae and we re-evaluate other macrofossils that have been attributed to the tribe.
We identified permineralized fossil roots in collections from the lower Miocene Cucaracha
Formation where it was exposed along the Culebra Cut of the Panama Canal. We prepared
the fossils using the cellulose acetate peel technique and compared the anatomy with that
of extant Paullinieae. The fossil roots preserve a combination of characters found only in
Paullinieae, including peripheral secondary vascular strands, vessel dimorphism, alternate
intervessel pitting with coalescent apertures, heterocellular rays, and axial parenchyma
strands of 2–4 cells, often with prismatic crystals. We also searched the paleontological liter-
ature for other occurrences of the tribe. We re-evaluated leaf fossils from western North
America that have been assigned to extant genera in the tribe by comparing their morphol-
ogy to herbarium specimens and cleared leaves. The fossil leaves that were assigned to
Cardiospermum and Serjania from the Paleogene of western North America are likely
Sapindaceae; however, they lack diagnostic characters necessary for inclusion in Paulli-
nieae and should be excluded from those genera. Therefore, the fossils described here as
Ampelorhiza heteroxylon gen. et sp. nov. are the oldest macrofossil evidence of Paullinieae.
They provide direct evidence of the development of a vascular cambial variant associated
with the climbing habit in Sapindaceae and provide strong evidence of the diversification of
crown-group Paullinieae in the tropics by 18.5–19 million years ago.
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: Jud NA, Allen SE, Nelson CW, Bastos CL,
Chery JG (2021) Climbing since the early Miocene:
The fossil record of Paullinieae (Sapindaceae).
PLoS ONE 16(4): e0248369. https://doi.org/
10.1371/journal.pone.0248369
Editor: William Oki Wong, Indiana University
Bloomington, UNITED STATES
Received: October 16, 2020
Accepted: February 23, 2021
Published: April 7, 2021
Copyright: © 2021 Jud et al. This is an open access
article distributed under the terms of the Creative
Commons Attribution License, which permits
unrestricted use, distribution, and reproduction in
any medium, provided the original author and
source are credited.
Data Availability Statement: All relevant data are
within the paper and its Supporting information
files.
Funding: This study received support from the
National Science Foundation (NSF) Award Number
0966884. The funders had no role in study design,
data collection and analysis, decision to publish, or
preparation of the manuscript. No additional
external funding received for this study.
Competing interests: The authors have declared
that no competing interests exist.
PLOS ONE | https://doi.org/10.1371/journal.pone.0248369 April 7, 2021
1 / 22
PLOS ONE Fossil Paullinieae
Introduction
Paullinieae (Sapindaceae) are tropical and subtropical woody vines (i.e., lianas), herbaceous
climbers (i.e., vines), and seldom shrubs [1]. The six genera of Paullinieae–Paullinia L., Serja-
nia L., Cardiospermum Kunth., Urvillea Kunth., Lophostigma Radlk., and Thinouia Triana &
Planch–form a clade [2–4, 21] defined by their tendrilate climbing habit and presence of stip-
ules [21]. With approximately 475 species [21], they comprise nearly one quarter of all species
in Sapindaceae. The Paullinieae are one of the four successively nested tribes of the Supertribe
Paulliniodae sensu by Acevedo-Rodrı´guez et al. [21], however the other members–Athyaneae,
Bridgesieae, Thouinieae–are all trees and shrubs. Numerous members of Paullinieae undergo
developmental re-patterning during the production of secondary xylem (i.e., wood) and sec-
ondary phloem (i.e., inner bark), resulting in the formation of “vascular cambial variants,”
such as continuous or discontinuous successive cambia, neoformations forming peripheral
secondary vascular strands (i.e., corded [5]), compound stems, fissured xylem, divided xylem,
lobed xylem, and phloem wedges [5–19].
The monophyly of Paullinieae within the subfamily Sapindoideae is supported by morphol-
ogy [20] and molecular sequence data [2–4, 21, 22]. Molecular phylogenetic analyses have
repeatedly yielded a long branch subtending the Paullinieae [2–4], suggesting shifts in nucleo-
tide substitution rates potentially associated with the evolution of the climbing habit. Previous
efforts to calibrate the phylogeny of Sapindaceae have yielded Oligocene or Miocene estimates
for the age of crown-group Paullinieae [23–25]; however, critical evaluation of the fossil record
is necessary to constrain the timing of diversification and the evolution of morphology and
anatomy of Paullinieae.
Although the fossil record of Sapindaceae is rich e.g., [1, 26], macrofossils of Paullinieae
are rare and at least some previous identifications are unreliable. Here, we describe the first
anatomically preserved macrofossils of Paullinieae. The fossils are roots, but nonetheless pro-
vide strong evidence of the climbing habit based on wood anatomy associated with climbing
in Sapindaceae. Next, we evaluate fossil leaves that have been attributed to the tribe. Then, we
summarize the fossil record of the tribe with a focus on macrofossils and identify occurrences
best suited for calibrating time-trees [27]. Finally, we discuss the implications of our findings
for future studies of the evolution of Paullinieae.
Materials and methods
Geologic setting
Two fossil roots were identified in a collection from the Lirio East site in lower part of the
Cucaracha Formation along the Culebra Cut (Gaillard Cut) of the Panama Canal (Fig 1).
These collections were made in in 2007 by F. Herrera and S.R. Manchester. The lower Cucara-
cha Formation consists of deltaic and coastal swamp deposits laid down during the early Mio-
cene when the nearby Pedro-Miguel Volcanic Complex was active [28–31]. At the Lirio East
site, fossil fruits as well as woods with bark are preserved as calcareous permineralizations in a
poorly sorted, carbonate-cemented sandstone [32].
So far, remains of Sacoglottis (Humiriaceae) [33], Oreomunnia (Juglandaceae) [34], Parinari
(Chrysobalanaceae) [35], Mammea (Calophyllaceae) [36], Rourea (Connaraceae) [37], and
Spondias (Anacardiaceae) [38], have been described. Plant macrofossils from elsewhere in the
Cucaracha Formation include palm stem fragments [39], Guazuma-like Malvaceae [40],
legume woods [39, 41], and a Malpighialean wood [42]. Fossil pollen from the Cucaracha For-
mation includes at least 52 pollen types [43]. Together, these records suggest the vegetation
was primarily tropical rainforest near the paleoshoreline of central Panama [43].
PLOS ONE | https://doi.org/10.1371/journal.pone.0248369 April 7, 2021
2 / 22
PLOS ONE Fossil Paullinieae
Fig 1. Native distribution of Paullinieae and fossil occurrences. Modern occurrence data from the BIEN database [45, 46]. Red star indicates the location
of the Lirio East fossil site where the fossil roots were collected. Fossil pollen occurrence codes: 1 = Serjania sp., upper Miocene Paraje Solo Formation [47–
49]; 2 = Serjania sp. and Paullinia sp., lower-middle Miocene Me´ndez Formation [50]; 3 = Serjania sp. and Paullinia sp., upper Miocene Gatun Formation
[49, 51]; 4 = Serjania sp., Paullinia sp., and Cardiospermum sp., upper Eocene Gatuncillo Formation [48, 52] Occurrence data were extracted from BIEN
ver. 4.1 database using the RBIEN package [46], supplemented with C. pechuelii data from GBIF [53]. Cardiospermum spp. distribution data follows native
ranges determined by [54, 55] (excluding controversial range in India).
https://doi.org/10.1371/journal.pone.0248369.g001
Fossil preparation
We cut the fossils in transverse and tangential and radial longitudinal sections using a Micro-
slice 2 annular saw and prepared serial sections using the cellulose acetate peel technique [44].
Peels were mounted on 25 x 75 mm glass slides with Canada Balsam or Eukitt mounting
medium and examined using light microscopy. Images of microscopic features were captured
with a Canon EOS digital camera mounted on a Nikon compound microscope with transmit-
ted light and processed with Adobe Photoshop (San Jose, California, USA). All specimens,
peels, and microscope slides are curated at the Florida Museum of Natural History Paleobotan-
ical Collections, Gainesville, Florida, United States.
Terminology and measurement protocols for the wood anatomy generally follow the
IAWA Hardwood List [56] but we adapted our approach for characters particular to Paulli-
nieae [64]. Summary statistics for anatomical characters were calculated from 25 measure-
ments. The fossil exhibits vessel dimorphism; this term has been used for both highly skewed
distributions and bimodal distributions [57–59], so we measured all vessels in the central
xylem cylinder [14] of a single transverse peel (n = 162) from the holotype (UF 19391-63016)
to generate a histogram of the distribution of vessel diameters. Then, we used the densityM-
clust function in the package mclust [60] in R [61] to identify the modes in the distribution
that correspond to the narrow and wide vessel classes. We report “narrow vessel diameter”
and “wide vessel diameter” as two separate characters. All measurements were made in ImageJ
1.50a [62].
PLOS ONE | https://doi.org/10.1371/journal.pone.0248369 April 7, 2021
3 / 22
PLOS ONE Fossil Paullinieae
Table 1. Summary of pre-Quaternary macrofossils that have been assigned to Paullinieae.
Species
Ampelorhiza heteroxylon
Bohlenia spp.
“Cardiospermum” coloradensis
“Cardiospermum” terminale
“Serjania” rara
Serjania mezzalire
Serjania itaquaquecetubensis
Serjania laceolata
Organ
root
Formation
Cucaracha
leaf
leaf
leaf
leaf
leaf
leaf
leaf
Klondike Mountain
Green River
Florissant; Renova
Aycross; Bridger
Rio Claro
Itaquaquecetuba
Itaquaquecetuba
Age
Mi.
Eo.
Eo.
Eo.
Eo.
Ol.
Mi.
Mi.
Country
Panama
USA
USA
USA
USA
Brazil
Brazil
Brazil
References
This study
[85, 86]
[81]
[77, 78, 95]
[74, 75]
[89]
[87]
[87]
Status
accepted
rejected
rejected
rejected
rejected
uncertain
uncertain
uncertain
Each identification is classified as accepted, rejected, or uncertain (material is consistent with Paullinieae, but alternative interpretations have not been ruled out). Mi.:
Miocene, Ol.: Oligocene, Eo.: Eocene. See text for further justification of status.
https://doi.org/10.1371/journal.pone.0248369.t001
Fossil leaves
We searched the literature for fossils identified as Paullinieae (Table 1). Of the species we
found, we examined specimens and images for those from North America and we re-described
their morphology following the format of the Manual of Leaf Architecture [63]. For putative
occurrences from South America and Europe, we evaluated images and descriptions from the
published literature. We used herbarium collections and online images to survey angiosperm
families for leaves with organization, margin type, and venation patterns similar to the fossil
leaf taxa re-described here (originally assigned to modern genera within Paullinieae). Then, we
compared the morphology of the fossils with leaves from extant genera in Paullinieae and with
leaves of selected genera outside Sapindaceae that exhibit similarities in organization, shape,
margin, and venation patterns. Cuticle was not preserved on any of the fossil leaves we exam-
ined and we did not evaluate cuticle for diagnostic characters. Comparisons are based on dried
specimens in the University (UC) and Jepson (JEPS) Herbaria at the University of California
—Berkeley, the R. L. McGregor Herbarium (KANU) at the University of Kansas, images avail-
able online via JSTOR Global Plants, and cleared and stained leaves in the National Cleared
Leaf Collection (NCLC-H; https://collections.peabody.yale.edu/pb/nclc/).
Phylogenetic analysis
We obtained the concatenated multiple sequence alignments from [21] and [22]. From these
datasets, we exclusively selected species within the supertribe Paullionieae as described by Ace-
vedo-Rodrı´guez et al. [21], which includes Athyaneae, Bridgesieae, Thouinieae, and Paulli-
nieae, totalling 100 ITS and 88 trnL intron sequences from [21], and 115 ITS sequences from
[22]. We then combined the two ITS datasets and realigned them in Geneious Prime 2021.0.3
(https://www.geneious.com) using the MUSCLE v3.8.425 aligner under default settings; the
trnL intron sequences were realigned under the same settings.
We then obtained wood anatomy data for 11 terminals from [13] and 33 terminals
from [64], and one terminal from [20], now available on morphobank (morphobank.org/
permalink/?P3910), and scored the fossil for 22 out of the 27 anatomy characters. Finally, we
added the character “habit” (0 = self-supporter, 1 = climber) and scored it for all extant species.
Although the wood anatomy characters scored for extant species were observed in stems and
the fossils are roots, available evidence indicates that differences in wood anatomy between
stems and roots within individual plants tend to be quantitative rather than qualitative [16, 65,
PLOS ONE | https://doi.org/10.1371/journal.pone.0248369 April 7, 2021
4 / 22
PLOS ONE Fossil Paullinieae
66]. The resulting dataset (S1 Appendix) comprises 216 tips and 1517 characters with three
partitions: anatomy (1-28), ITS (29-882), and trnL intron (883-1517).
We estimated the phylogenetic position of the fossil taxon using a Bayesian analysis with
two runs each of four chains (three hot, one cold, temp = 0.02) in MrBayes 3.2.7 [67]. We
applied the GTR+G model of nucleotide evolution to the ITS and trnL intron partitions. The
Mk model with rates drawn from a lognormal distribution was applied to the anatomy parti-
tion. The analysis ran for 12 million generations, sampling trees every 1000th generation. The
analysis converged with a standard deviation of split frequencies of 0.007428 and the estimated
sample size (ESS) of all parameters exceeded 2108. All trees were generated using the post
burnin (25% of initial trees discarded) from the combined MrBayes runs. The allcompat con-
sensus tree (50% majority rule consensus with compatible groups added) was generated with
the MrBayes command: contype = allcompat and annotated using iToL v4 [68]. The maxi-
mum clade credibility (MCC) tree was generated with Tree Annotator v1.10.4 [69], and the
maximum a posteriori tree (MAP) was generated with RevBays v1.10 [70]. The MrBayes input
nexus file (data matrix), allcompat consensus, MCC, and MAP trees, and full accession list
with associated molecular and anatomical data references are provided in (S1 Appendix).
Nomenclature
ptThe electronic version of this article in Portable Document Format (PDF) in a work with an
ISSN or ISBN will represent a published work according to the International Code of Nomencla-
ture for algae, fungi, and plants, and hence the new names contained in the electronic publica-
tion of a PLOS ONE article are effectively published under that Code from the electronic edition
alone, so there is no longer any need to provide printed copies. The online version of this work is
archived and available from the following digital repositories: PubMed Central and LOCKSS.
Results
Fossil roots
Family. Sapindaceae Jussieu.
Subfamily. Sapindoideae Burnett.
Tribe. Paullinieae (Kunth) DC.
Genus. Ampelorhiza Jud, S.E. Allen, Nelson, Bastos & Chery gen. nov.
Generic diagnosis. Roots woody with neoformations forming peripheral secondary vas-
cular strands; vessels of two distinct size classes, wide vessels solitary and in tangential multi-
ples, narrow vessels in long radial multiples; intervessel pits alternate with slit-like coalescent
apertures on the walls of large vessels; heterocellular rays composed of mixed upright, square,
and procumbent cells; axial parenchyma strands 2–4 or more cells tall, often chambered with
prismatic crystals.
Type species. Ampelorhiza heteroxylon Jud, S.E. Allen, Nelson, Bastos & Chery gen. et sp. nov.
Specific diagnosis. As for genus.
Holotype. UF 19391-63016 (Figs 2 and 3).
Paratype. UF 19391-63026 (S1 Fig).
Repository. Florida Museum of Natural History (FLMNH), Gainesville, Florida, U.S.A.
Type locality. Panama; Culebra Cut, northeast side of the Panama Canal (N 9.051375˚, W
79.649027˚, WGS84).
Stratigraphic position and age. Cucaracha Formation; early Miocene, c. 18.5–19 Ma
[30, 31].
Etymology. The genus comes from the Greek ámpelos, meaning vine, and ríza meaning
root. The specific epithet comes from the Greek héteros meaning different and -xylon meaning
PLOS ONE | https://doi.org/10.1371/journal.pone.0248369 April 7, 2021
5 / 22
PLOS ONE Fossil Paullinieae
Fig 2. Cambial variant and vessel characters in Ampelorhiza heteroxylon. (A) Transverse section of the stem
showing diffuse-porous wood of the central cylinder (cc) and peripheral vascular strands (ps) in the external tissues.
Arrow indicates the position of the smaller of two peripheral vascular cylinders. Specimen UF 19391-63016 XS peel 10.
(B) Close up transverse section of the larger of two peripheral vascular strands. Specimen UF 19391-63016 XS peel 10.
(C) Transverse section of the smaller of two peripheral vascular cylinders. There is no pith. Specimen UF 19391-63016
XS peel 10. (D) Close up of A showing the primary vascular parenchyma. Specimen UF 19391-63016 XS peel 10. (E)
Tangential longitudinal section through the tall cells of the primary vascular parenchyma (center right), ray cells
(center left) and juvenile wood (far left). UF 19391-63016 LS peel 16. (F) Transverse section showing wide solitary
vessels (WV) and narrow vessels in long radial multiples (at arrow). Specimen UF 19391-63026 XS peel 6. (G)
Tangential longitudinal section (LS) showing coalescent pit apertures on the vessel wall. Specimen UF 19391-63016 LS
peel 6. (H) Tangential longitudinal section showing alternate polygonal pits on the vessel wall (at arrow). Specimen UF
19391-63016 LS peel 7. (I) Tangential longitudinal section showing narrow vessels (NV) with oblique end walls, and
wide vessels (WV) with simple perforation plates and end walls perpendicular to lateral walls (right arrow). Specimen
UF 19391-63026 TLS peel 1. Scale bars: A = 3 mm; B = 1 mm; C, F, I = 200 μm; D, E = 500 μm; G, H = 100 μm.
https://doi.org/10.1371/journal.pone.0248369.g002
PLOS ONE | https://doi.org/10.1371/journal.pone.0248369 April 7, 2021
6 / 22
PLOS ONE Fossil Paullinieae
Fig 3. Wood anatomy in Ampelorhiza heteroxylon. (A) Tangential longitudinal section showing uniseriate pitting on the fiber walls. Specimen UF 19391-
63016 LS peel 5. (B) Tangential longitudinal section showing axial elements including narrow vessels and uniseriate rays (at arrow). Specimen UF 19391-
63016 LS peel 1. (C) Radial longitudinal section showing ray cells against a vessel. Note the partially preserved vessel-ray parenchyma pitting similar in size
to the intervessel pitting (at arrow). Specimen UF 19391-63016 LS peel 7. (D) Tangential longitudinal section showing uniseriate and biseriate rays (left
arrow) and axial elements with crystals (right arrow). Specimen UF 19391-63016 LS peel 5. (E) Radial longitudinal section showing upright (at arrow),
square, and procumbent ray cells. Specimen UF 19391-63026 LS peel 2. Scale bars: A = 70 μm; B = 150 μm; C = 40 μm; D, E = 100 μm.
https://doi.org/10.1371/journal.pone.0248369.g003
wood, referring to the different sizes of the peripheral secondary vascular strands found in
Paullinieae.
Description (descriptio generico-specifica). The holotype is an axis 1 cm wide and 3 cm
long; the paratype is an axis 0.5 by 1 cm wide and 2.5 cm long. Each consists of bark with one
or two peripheral secondary vascular strands (Fig 2A–2C), surrounding a central woody cylin-
der with a small core of primary vascular parenchyma (Fig 2A–2C). The peripheral vascular
PLOS ONE | https://doi.org/10.1371/journal.pone.0248369 April 7, 2021
7 / 22
PLOS ONE Fossil Paullinieae
strands consist of secondary xylem and phloem derived from C-shaped cambia that lack pri-
mary vascular parenchyma. In the holotype, the two preserved peripheral strands are of differ-
ent sizes. One is c. 3.3 mm by c. 2.0 mm in transverse section and the other is 0.7 mm by c. 0.4
mm (Fig 2A lower arrow, Fig 2C). Primary vascular parenchyma in the central cylinder of the
holotype is an eccentric collection of parenchyma cells 200 μm tall by 500 μm wide (Fig 2D).
Radial files of cells with dark contents also extend away from the center of the central cylinder
on one side (Fig 2D). The primary vascular parenchyma cells are tall (c. 150–300 μm), and
many have dark contents in the lumen (Fig 2E). Secondary xylem is diffuse porous (Fig 2A &
2F). Growth rings are indistinct (Fig 2A & 2F). Vessels are in two distinct size classes: wide ves-
sels 50-270 μm (mean: 104 μm) in tangential diameter, mostly solitary but also in tangential
multiples of 2–3; narrow vessels are 11–50 μm in tangential diameter and arranged in radial
multiples of 2–9 (Fig 2A & 2F). Vessel elements are 153–280 μm long (mean: 223 μm, n = 14).
Mean vessel frequency is 27 per mm2. Vessel element end walls are without scalariform bars;
perforation plates are simple (Fig 2I). Tyloses and helical thickenings were not observed. Inter-
vessel pits alternate with distinct borders and coalescent apertures on the walls of large vessels
(Fig 2G & 2F). Vessel-ray parenchyma pits were difficult to observe; we did not find large sim-
ple pits different from those on the vessel walls (Fig 3C). Fibers are poorly preserved but appear
non-septate with minutely bordered uniseriate pits on the radial walls (Fig 3A). Axial paren-
chyma is diffuse and scanty paratracheal, with strands at least 2–4 cells tall and often cham-
bered with prismatic crystals (Fig 3D). Rays are 1–2 (rarely three) cells wide, less than 1 mm
tall, and heterocellular with rows of procumbent square and upright cells mixed throughout
(Fig 3E). Secretory structures were not observed.
Remarks. Although cambial variants are often associated with the climbing habit, the
presence of peripheral vascular strands is not sufficient to identify the fossils as stems or roots.
Bastos et al. [16, 66] demonstrated that cambial variants can be found in both organs. In stems
of Paullinieae, the pith is conspicuously angular (e.g., triangular, pentangular) in transverse
section with primary vascular bundles at the corners [19]. By contrast, in roots the primary
vascular parenchyma is diarch and this region (i.e., the “medulla”) is oval and smaller than the
stem pith in transverse section (Fig 4). In Ampelorhiza heteroxylon, there is an eccentric oval-
shaped parenchymatous core c. 200 by 500 μm in diameter (Fig 2D); therefore, our interpreta-
tion is that the specimens are roots.
We initially recognized that these fossils might be lianas based on the diameter of the largest
vessels relative to the width of the axis. To illustrate this approach, we used logistic regression
to classify unknown fossil axes from Lirio East as climbers or self-supporters based on maxi-
mum vessel diameter and diameter of the central woody cylinder (S2 Fig). The model was
trained using a dataset of 71 samples obtained from Ewers et al. [71], and predicted the habit
of 22 fossil axes with woody cylinders greater than 5 mm in diameter from the Lirio East fossil
collections. Although the model did predict that the Ampelorhiza fossils (and the Rourea fossil
described by Jud and Nelson [37]) are climbers, the training dataset is only stem material and
therefore may not be suitable for classifying roots, given the patterns found by Ewers et al. [72]
when comparing stems and roots in lianas and trees. Further work on the relationship between
hydraulic path length, vessel diameter, and root diameter in lianas and self-supporters (as has
been done for stems [73]) would be useful for identifying lianas in the fossil record.
Fossil leaves
We found one fossil species assigned to Serjania and two assigned to Cardiospermum from
North America in the literature (Table 1). All three were described from fossils of isolated leaf-
lets or partially complete compound leaves (Fig 5). MacGinitie [74] described “Serjania” rara
PLOS ONE | https://doi.org/10.1371/journal.pone.0248369 April 7, 2021
8 / 22
PLOS ONE Fossil Paullinieae
Fig 4. Wood anatomy of the roots of extant Paullinieae species. A–B: Neoformations forming peripheral vascular strands in Serjania
caracasana (Jacq.) Willd. in transverse section. (A) Root macromorphology presenting a cambial variant. Arrows point to individual
peripheral vascular strands. (B) Close up of the juncture of the central cylinder (cc) and a peripheral vascular strand (ps) with a c-shaped
“pith” (i.e., primary vascular parenchyma of the root). (C) Secondary xylem of Thinouia scandens Triana & Planch. with vessel
dimorphism in transverse section. Note the wide vessels (WV) are solitary or in tangential (upper arrow) or radial multiples (lower
arrow), while the narrow vessels (NV) are in longer radial chains. (D) Primary vascular parenchyma in the center of the the diarch roots
(arrows towards protoxylem poles) of S. caracasana in transverse section. (E) Alternate intervessel pits (lower arrow) and those with
coalescent apertures (upper arrow) in S. caracasana in tangential longitudinal section. (F) Prismatic crystals in the axial parenchyma (�) of
S. caracasana in macerated material. Scale bars: A = 0.5 cm, B = 1 mm, C = 250 μm, D = 100 μm, E = 70 μm, F = 50 μm. �prismatic
crystals in axial parenchyma.
https://doi.org/10.1371/journal.pone.0248369.g004
based on leaves from the Eocene Aycross Formation in northwestern Wyoming. The same
species also occurs in the Eocene Bridger Formation in southwestern Wyoming [75]. “Cardios-
permum” terminale (Lesquereux) MacGinitie was first described from the Eocene Florissant
Formation in central Colorado by Lesquereux [76] as Lomatia. MacGinitie [77] transferred
these specimens and others to Cardiospermum based on the twice-ternate leaf organization
and architecture of lobes, teeth, and major vein framework of the leaflets. This species was
later reported from the late Eocene to early Oligocene Climbing Arrow Member of the Renova
Formation in southwestern Montana [78, 79] as well. Finally, “Cardiospermum” coloradensis
(Knowlton) MacGinitie was first described from the Eocene Green River Formation as Phyl-
lites by Knowlton [80]; and later transferred to Cardiospermum by MacGinitie [81]. This
PLOS ONE | https://doi.org/10.1371/journal.pone.0248369 April 7, 2021
9 / 22
PLOS ONE Fossil Paullinieae
Fig 5. Leaf fossils previously assigned to Paullinieae. (A) “Serjania” rara MacGinitie from the Bridger Formation,
Blue Rim site, Sweetwater County, Wyoming, UF 15761S-57786. (B) “Serjania” rara MacGinitie from the Bridger
Formation, Blue Rim site, Sweetwater County, Wyoming, UF 15761N-61430. (C) Paratype of “Serjania” rara
MacGinitie from the Aycross Formation, Kisinger Lakes site, northwestern Wyoming (Pl 25, Fig 3 in [74]), UCMP PA
108, 5698. (D) Hypotype of “Cardiospermum” coloradensis (Knowlton) MacGinitie from the Green River Formation,
west of Wardell Ranch site, Colorado (Pl 22, Fig 3 in [81]), UCMP PA 321, 20593. Arrow indicates marginal vein. (E)
“Cardiospermum” coloradensis (Knowlton) MacGinitie from the Green River Formation in Rainbow, UT, UCMP
PB02016, 201265. Arrow indicates marginal vein. (F) “Cardiospermum” terminale (Lesquereux) MacGinitie from the
Florissant Formation in central Colorado, FLFO 10147. Scale bars = 1 cm.
https://doi.org/10.1371/journal.pone.0248369.g005
species has been reported from throughout the Green River Formation [81–84]. Updated
descriptions of these three species are provided in the (S2 Appendix).
The extinct genus Bohlenia Wolfe & Wehr [85] was established for sapindaceous leaves and
fruits from the Eocene Republic flora (Klondike Mountain Formation) in Washington, USA
(Table 1). Wolfe and Wehr [85] suggested that B. americana (Brown) Wolfe & Wehr may
belong to Paullinieae based on the course of the secondary veins and on the assumption that
co-occurring samaras belonged to the same species; however, McClain and Manchester [86]
transferred the samaras to Dipteronia brownii McClain & Manchester and noted that Bohlenia
PLOS ONE | https://doi.org/10.1371/journal.pone.0248369 April 7, 2021
10 / 22
PLOS ONE Fossil Paullinieae
foliage is similar to Koelreuteria elegans (Seem.) A.C. Sm. Both of these fossil species are mem-
bers of Sapindaceae, but neither belong to Paullinieae.
We also found three species assigned to Serjania from the Cenozoic of Brazil in the litera-
ture (Table 1). Fittipaldi et al. [87] described Serjania lanceolata Fittipaldi, Simões Giulietti et
Pirani and Serjania itaquaquecetubensis Fittipaldi, Simões Giulietti et Pirani based on incom-
plete unlobed, toothed leaf blades from the Oligocene upper Itaquaquecetuba Formation. To
our knowledge, the characteristic pollen of Paullinieae has not been recognized in palynologi-
cal studies of this formation [88]. Finally, Serjania mezzalire Duarte et Rezende-Martins was
described from fossil leaves in the Miocene Rio Claro Formation [89, 90].
Edwards and Wannacot [91] compiled list of all fossil species that had been assigned to
Paullinia based on leaf morphology from Europe. They concluded that a close relationship to
extant Paullinieae can be rejected or is doubtful for all of them based on morphology or quality
of preservation. We concur, so we did not consider these further.
There is considerable variation in the blade shape, margin type, tertiary venation, and base
shape among extant Paullinieae (Fig 6). Leaf margins may be unlobed or lobed, toothed or
untoothed. Toothed margins may be serrate, dentate, or crenate. Secondary vein framework
may be craspedodromous, semicraspedodromous, brochidodromous, or eucamptodromous.
Leaf organization is also variable. Leaves may be simple, once or twice imparipinnate, or up to
thrice ternate (most commonly twice ternate). In compound leaves, the rachis may be winged
or unwinged. Axillary tendrils may be absent or present. Many of these characters also vary
across Sapindaceae. Based on our observations, isolated fossil leaves or leaflets of Paullinieae
may be recognizable if they preserve a combination of the following characters: Axial tendrils,
stipules, ternate compound organization, winged rachides, and absence of a marginal vein.
Morphological similarities between “Cardiospermum” coloradensis, “C.” terminale, “Serja-
nia” rara, and the leaves of some extant Paullinieae include 1. ternate-compound organization,
2. decurrent (Figs 5C, 5E and 5F and 6A and 6B) or complex leaflet bases (Fig 5A and 5B), 3.
irregular spacing of secondary veins, 4. secondary veins that terminate beyond the apex of
lobes/teeth, 5. secondary veins that terminate in angular (“V-shaped”) sinuses (Fig 5), and 6.
secondary veins that bifurcate around angular sinuses (Fig 5E). However, some or all of these
characters can be found in the leaves of other Sapindaceae (e.g., Thouinia Poit., Koelreuteria
Laxm., Dipterodendron Radlk., Dilodendron Radlk., and Athyana (Griseb.) Radlk) and in other
families (see Discussion section for further commentary); they are not diagnostic of Cardios-
permum, Serjania, nor Paullinieae. Furthermore, a prominent marginal vein like that present
in at least some specimens of the fossil species (Fig 5D and 5E) is not present in extant Serjania
and Cardiospermum (Fig 6A and 6B). The descriptions and images of “Serjania” lanceolata,
“S.” itaquaquecetubensis, and “S.” messalire show the shape of the blade, the presence of a ser-
rate margin, and craspedodromous secondary vein framework [87, 90]. Although these char-
acters are consistent with Serjania, their combination is not diagnostic of the genus.
Phylogenetic position of Ampelorhiza
We evaluated the placement of Ampelorhiza by observation of the allcompat consensus, MCC,
and MAP trees sampled from the posterior distribution. Ampelorhiza is always nested within
extant Paullinieae, however its relationship with extant genera differs based on the method
used to generate the tree, reflecting the uncertainty typical of taxa with a high proportion of
missing data. In the allcompat consensus tree (Fig 7) Ampelorhiza is nested within a clade with
Cardiospermum, Paullinia, and Serjania. The various positions of Ampelorhiza within this
clade is represented as a polytomy that includes several lineages of Serjania and Cardiosper-
mum. In the maximum a posteriori tree (S1 Appendix), Ampelorhiza is nested within Urvillea,
PLOS ONE | https://doi.org/10.1371/journal.pone.0248369 April 7, 2021
11 / 22
PLOS ONE Fossil Paullinieae
Fig 6. Extant leaves. Modern leaves for comparison with the putative Paullinieae fossils. Cleared leaves from the National Cleared Leaf Collection (NCLC).
(A) Serjania rhombea Radlk. (Coll.: W.H. Lewis, J.D. Dwyer, T.S. Elias, and R. Solı´s #72 (UC 1355158), 1966, Panama]. (B) Cardiospermum halicacabum L.
[Coll.: R.D.A. Baylis #5080 (UC 1409568), 1972, South Africa]. (C) Paullinia pinnata L., NCLC 0012. (D) Quercus nigra L., NCLC 0215. (E) Lycopersicum
esculentum L., NCLC 1640. (F) Beauprea balansae Brongn. & Gris, NCLC 6658. Scale bars = 1 cm.
https://doi.org/10.1371/journal.pone.0248369.g006
PLOS ONE | https://doi.org/10.1371/journal.pone.0248369 April 7, 2021
12 / 22
PLOS ONE Fossil Paullinieae
Fig 7. Phylogeny of supertribe Paulliniodae. (A) Majority rule consensus tree with all compatible groups (“allcompat”) of supertribe Paulliniodae sensu
Acevedo et al. [21], generated in MrBayes 3.2.7 from an anatomical and molecular combined dataset of 216 tips. Branch colors indicate posterior
probabilities. The outermost black line indicates the tribe Paullinieae. Note the position of the fossil taxon Ampelorhiza within Paullinieae indicated by the
arrow and the dagger. (B) Summary tree showing the same topology, but pruned to show genera only, assuming all genera are monophyletic. Numbers
above branches are posterior probabilities, dashes indicate genera represented by a single species in the “allcompat” consensus tree.
https://doi.org/10.1371/journal.pone.0248369.g007
whereas in the maximum clade credibility tree Ampelorhiza is nested within Serjania. These
results further supports our circumscription of Ampelorhiza as a distinct genus from extant
Paullinieae. The placement of Ampelorhiza within Paullinieae is supported by vessel dimor-
phism, heterocellular rays, and axial parenchyma strands typically 2-4 cells long. One synapo-
morphy of Paullinieae that we did not observe in the fossil is wide rays (ray dimorphism);
however, we only examined two root fragments and this character is observed in many, but
not all, samples from modern roots [16].
Discussion
Roots
The combination of peripheral vascular strands (Figs 2A–2C and 4A & 4B), vessel dimorphism
(Figs 2F & 2I and 4B–4D), wide vessels solitary or in tangential multiples of 2–3 (Fig 2F and
4C), narrow vessels in long radial multiples of 2–21 (Figs 2F and 4C & 4D), alternate interves-
sel pits with slit-like coalescent apertures (Figs 2G and 2H and 4E), heterocellular rays, pris-
matic crystals in axial parenchyma (Figs 3D and 4F), and dark content (possibly phenolic
compounds) in primary vascular parenchyma and ray parenchyma (Fig 2D and 2E) support
the inclusion of Ampelorhiza in Paullinieae [13, 16, 18, 64, 66, 92, 93, 94]. Two wood anatomi-
cal characters typical of extant Paullinieae were not observed in the fossils: 1) alternating bands
of thin and thick-walled regions in the wood which may either be axial parenchyma alternating
with ordinary fibers (e.g., Serjania spp.) or parenchyma-like fiber bands alternating with
PLOS ONE | https://doi.org/10.1371/journal.pone.0248369 April 7, 2021
13 / 22
PLOS ONE Fossil Paullinieae
ordinary fibers (e.g., Paullinia spp.) and 2) ray dimorphism. Because the bands are clearest in
sufficiently thin, stained sections or polished blocks, it may be that the thickness of the peels
and the absence of stain obscures this feature.
The cambial configuration in stems and roots is highly variable in Paullinieae. Chery et al.
[19] and Cunha Neto et al. [18] together distinguished six ontogenetic pathways in the stems
of Paullinia alone, and we expect that Serjania has the most variation in the tribe based on pre-
liminary observations of images in the Smithsonian Liana databases (Acevedo & Chery, per-
sonal observation). Furthermore, Bastos [16, 66] showed that roots may or may not also have
cambial variants, and when present they do not necessary mirror the configuration of the
stems. An asymmetrical distribution of peripheral secondary vascular strands of different
sizes, as in Ampelorhiza heteroxylon, occurs in the roots of Serjania caracasana (Fig 4A & 4B)
and the stems of some Paullinia [18]. Given the variation among stems and the paucity of data
on cambial variants in roots, the configuration of secondary growth in the fossils does not jus-
tify assignment to one of the extant genera.
Despite some anatomical differences among the genera of Paullinieae, the fossils of Ampe-
lorhiza do not preserve a combination of wood anatomy characters diagnostic of any extant
genus either, they are most similar to some Serjania. The wood of Serjania stems has banded
axial parenchyma, no septate fibers, and crystals confined to axial elements, whereas Paullinia,
Thinouia, and Cardiospermum have scanty axial parenchyma, abundant septate fibers, and
crystals in ray parenchyma. Thinouia differs from Paullinia and Cardiospermum by the
absence of crystals in axial elements [13], and some Paullinia can be recognized by a herring-
bone pattern in the wide rays when viewed in transverse section [13]. The fossils do not have
banded parenchyma, nor do they have wide rays with a herringbone pattern. They do have
crystals in the axial elements but we did not observe them in the rays, nor did we detect septate
fibers.
Leaves
We reject the generic assignments of Cardiospermum and Serjania species described from fos-
sil leaf material. Our search for leaves with organization, margin features, and venation archi-
tecture similar to “C.” coloradensis, “C.” terminale, and “S.” rara outside of Sapindaceae led to
comparisons with Anacardiaceae (e.g., Rhus L.), Fagaceae (e.g., Quercus L.), Proteaceae (e.g.,
Roupala Aubl., Lomatia R. Br., Beauprea Brongn. & Gris), Ranunculaceae (e.g., Clematis L.),
and Solanaceae (e.g., Hyoscyamus L., Chamaesaracha (A. Gray) Benth. & Hook. f., Physalis L.,
Lycopersicum Hill.). Some Rhus (Anacardiaceae) have similar shapes to the fossil material, but
secondary venation in Rhus varies from craspedodromous to cladodromous. Some Fagaceae
have similar blade shape, secondary veins, and major veins that project beyond the margin of
the blade; however, all Fagaceae have simple leaves and the sinuses are generally rounded
rather than angular as in the fossils. Previous authors (e.g., [76, 83]) have attributed fossils like
these to Proteaceae; however, although secondary veins in the Proteaceae are variable (e.g.,
brochidodromous to semicraspedodromous to festooned brochidodromous to festooned
semicraspedodromous), they are unlike the craspedodromous framework in the fossils and
again the sinuses between teeth are generally rounded in Proteaceae rather than angular. The
compound leaves of some lobed and toothed Clematis (Ranunculaceae) can be distinguished
from the fossils because they usually have festooned secondary venation. Finally, several Sola-
naceae have asymmetric blades and similarly shaped teeth and lobes; however, again the
sinuses tended to be rounded rather than angular as in the fossils.
Leaf architectural characters preserved in “C.” coloradensis, “C.” terminale, and “S.” rara
support inclusion in Sapindaceae, yet we consider a close relationship with Paullinieae unlikely
PLOS ONE | https://doi.org/10.1371/journal.pone.0248369 April 7, 2021
14 / 22
PLOS ONE Fossil Paullinieae
based on the presence of a prominent perimarginal vein in the fossils and the absence of co-
occurring fossil fruits or pollen despite decades of intensive sampling in the Green River For-
mation and the Florissant fossil beds. Similarly, in his update of the fossil flora of Florissant,
Manchester [95] doubted the generic assignment of “C.” terminale based on the rather coria-
ceous texture of the fossils compared to extant Cardiospermum and the absence of associated
fruits. Other extant Sapindaceae with similar leaf organization, margin type, teeth, and vena-
tion include: Thouinia Poit., Koelreuteria Laxm., Dipterodendron Radlk., Dilodendron Radlk.,
and Athyana (Griseb.) Radlk.
Evolution of Paullinieae
To our knowledge, the oldest reliable fossil evidence of Paullinieae is heteropolar hemi-tri-syn-
colpate pollen from the Gatuncillo Formation in Panama [52]. Some fossil species of the gen-
era Syncolporites and Proteacidites (used for dispersed pollen) may belong to Paullinieae (or
Proteaceae or Myrtaceae) [96]; however, a review of those species is beyond the scope of this
work. Heteropolar hemi-tri-syncolpate pollen is a synapomorphy of the clade that includes all
Paullinieae except Thinouia and Lophostigma [21, 97, 98]. Therefore, these fossils can be con-
sidered evidence of crown-group Paullinieae in the fossil record. Unfortunately, constraining
the age of these samples is challenging. Montes et al. [99] reported Late Eocene and Oligocene
foraminifera from the Gatuncillo Formation, consistent with the original age estimate from
Graham [52]. More recently, Ramı´rez et al. [100] obtained detrital zircons from two sites that
constrain the maximum age of deposition of the Gantuncillo Formation to Late Eocene, c. 41
Ma and c. 36 Ma respectively, but we do not know their position relative to Graham’s [52] pol-
len sample. Older putative occurrences of Middle Eocene pollen from the Wagon Bed Forma-
tion in Wyoming [101] and the Claiborne Group in northern Alabama [102] were not
described nor figured, and are not reliable [103]. Pollen from the Kisinger Lakes paleoflora in
Wyoming that MacGinitie compared with Serjania [74] was not described; however, one fig-
ure shows a single grain 24 μm across in polar view with a 3-(parasyncol)porate structure. It is
not possible to determine whether it was heteropolar and pollen grains in Paullinieae are larger
than 30 μm across [98, 104]. Therefore, we do not consider this a reliable fossil occurrence of
Paullinieae based on the available information. Younger occurrences include heteropolar
demisyncolpate pollen from the late Miocene Gatun Formation in Panama [43, 49] and the
Pliocene Paraje Solo Formation, also in Panama [47].
The transition to the liana habit occurred only once in Sapindaceae along the branch lead-
ing to crown-group Paullinieae [21]. Accordingly, all members of the tribe share anatomy
associated with the climbing habit such as abrupt changes in vessel diameter, vessel dimor-
phism, and numerous members have cambial variants [19, 105]. The combination of wood
anatomical characters and the presence of the peripheral vascular strands preserved in the fos-
sils provides strong evidence of the climbing habit in Paullinieae by the early Miocene.
Paleoecology
Lianas are a conspicuous element of tropical forests and their fossils contribute to reconstruc-
tions of paleoenvironments and paleocommunities. The Lirio East fossil assemblage includes
at least 32 plant morphotypes have been distinguished and assigned to family based on fossil
fruits, seeds, and woods [32–34, 36–38]. The discovery of Ampelorhiza brings the number of
liana species to a minimum of 8, or 25% of the local assemblage. This value is typical of lowland
tropical forests [106]. Three other potential liana axes were identified using logistic regression
(S2 Fig), but remain to be described (F. Herrera, pers. comm.). At least 31 additional fruit and
seed morphotypes have been distinguished but not yet identified to family [32]. In modern
PLOS ONE | https://doi.org/10.1371/journal.pone.0248369 April 7, 2021
15 / 22
PLOS ONE Fossil Paullinieae
tropical forests liana species richness is highest in seasonally dry tropical forests and locally
near forest edges or in treefall gaps [107–109]. Given the rarity of distinct growth rings in the
co-occurring fossil woods and the preference of Sacoglottis and Oreomunnea for humid tropi-
cal forests [33, 34], we hypothesize that the high proportion of lianas in the Lirio East assem-
blage is a signal of riparian zone disturbance and/or edge effects in a humid tropical forest on a
landscape shaped by nearby volcanic activity [31].
Conclusion
The discovery of Ampelorhiza reported here is the oldest reliable macrofossil evidence of Paul-
linieae. Fossil leaves from the Eocene of North America previously attributed to Cardiosper-
mum and Serjania likely belong to Sapindaceae, but are not reliable occurrences of Paullinieae.
Our findings support the conclusion that diversification of the tribe was underway by at least
18.5–19 Ma (early Miocene) and that the climbing habit had evolved by that time.
Supporting information
S1 Appendix. Folder containing the accession list, mrbayes infile.nex, mcc, map, allcom-
pat, and accession list.
(ZIP)
S2 Appendix. Revised descriptions of the leaf architecture. Descriptions of Bohlenia ameri-
cana, Bohlenia insignis, “Cardiospermum” coloradensis, “Cardiospermum” terminale, and “Ser-
jania” rara.
(PDF)
S1 Fig. Transverse section of the paratype, UF 19391-63026.
(TIF)
S2 Fig. Plot of lianas and self-supporting woody dicots. Filled points are fossil axes from the
Lirio East site classified as either climbers or self-supporters using logistic regression. We
applied a conservative decision threshold of 0.95 for classifying lianas.
(TIF)
Acknowledgments
We thank Bruce MacFadden, Jonathan Bloch, Steven Manchester, Carlos Jaramillo, and
Fabiany Herrera for support in the early phases of this project, Veronica Angylossy for super-
vision over Carolina Basto’s thesis work concerning the root anatomy of Paullinieae, and Lil-
lian Pearson for making initial peels of the fossil during her PCP-PIRE internship. Fabiany
Herrera discovered the fossil site at Lirio East. We thank Ricardo Martinez for donating the
vehicles used for fieldwork in Panama, and the Autoridad del Canal de Panama (ACP) for
access to the site where the fossils were collected. We also thank the staff of the herbaria at the
University of California Berkeley, the Florida Museum of Natural History, and the University
of Kansas, and the staff of the paleobotany collections at UC Berkeley, the Florida Museum of
Natural History, and the Smithsonian for assistance. We thank Sarah DeWitt for comments
on the figures. Finally, we thank the reviewers for helpful feedback and suggestions during the
review process. Any opinions, findings, conclusions, or recommendations expressed in this
article are those of the authors and do not necessarily reflect the views of the NSF.
PLOS ONE | https://doi.org/10.1371/journal.pone.0248369 April 7, 2021
16 / 22
PLOS ONE Fossil Paullinieae
Author Contributions
Conceptualization: Nathan A. Jud, Joyce G. Chery.
Data curation: Nathan A. Jud, Joyce G. Chery.
Formal analysis: Nathan A. Jud, Joyce G. Chery.
Investigation: Nathan A. Jud, Sarah E. Allen, Chris W. Nelson, Carolina L. Bastos, Joyce G.
Chery.
Methodology: Nathan A. Jud, Joyce G. Chery.
Resources: Carolina L. Bastos.
Supervision: Nathan A. Jud.
Writing – original draft: Nathan A. Jud, Sarah E. Allen, Chris W. Nelson, Joyce G. Chery.
Writing – review & editing: Nathan A. Jud, Sarah E. Allen, Joyce G. Chery.
References
1. Acevedo-Rodrı´guez P, Van Welzen PC, Adema F, Van Der Ham RWJM. Flowering Plants. Eudicots.
Sapindaceae. Springer p. 357–407. 2011.
2. Harrington MG, Edwards KJ, Johnson SA, Chase MW, Gadek PA. Phylogenetic Inference in Sapinda-
ceae sensu lato Using Plastid matK and rbcL DNA Sequences. Systematic Botany. 2005; 30(2):366–
382. https://doi.org/10.1600/0363644054223549
3. Buerki S, Forest F, Acevedo-Rodrı´guez P, Callmander MW, Nylander JAA, Harrington M, et al. Plastid
and nuclear DNA markers reveal intricate relationships at subfamilial and tribal levels in the soapberry
family (Sapindaceae). Molecular Phylogenetics and Evolution. 2009; 51(2):238–258. https://doi.org/
10.1016/j.ympev.2009.01.012 PMID: 19405193
4. Buerki S, Lowry PP II, Alvarez N, Razafimandimbison SG, Ku¨ pfer P, Callmander MW. Phylogeny and
circumscription of Sapindaceae revisited: molecular sequence data, morphology and biogeography
support recognition of a new family, Xanthoceraceae. Plant Ecology and Evolution. 2010; 143(2):148–
159. https://doi.org/10.5091/plecevo.2010.437
5.
Tamaio N, Somner GV. Development of corded vascular cylinder in Thinouia restingae Ferruci & Som-
ner (Sapindaceae: Paullinieae). The Journal of the Torrey Botanical Society. 2010; 137(4):319–326.
https://doi.org/10.3159/10-RA-047.1
6. Schenck H. Beitra¨ ge zur Biologie und Anatomie der Lianen im besonderen der in Brasilien einhei-
mischen Arten. 4-5. G. Fischer; 1892. f
7. Radlkofer L. Sapindaceae In: Engler A. & Prantl K. Die Natu¨ rlichen Pflazenfamilien. 1895; 3:277–366.
8. Pfeiffer HH. Das abnorme Dickenwachstum. vol. 9. Gebr. Borntraeger; 1926.
9. Radlkofer L, Engler A, Diels L. Sapindaceae. Engelmann 1933.
10.
Johnson MA, Truscott FH. On the Anatomy of Serjania. I. Path of the Bundles. American Journal of
Botany. 1956; 43(7):509–519. https://doi.org/10.1002/j.1537-2197.1956.tb10525.x
11. Obaton M. Les Lianes ligneuses a structure anormale des forêts denses d’Afrique occidentale.
Annales de scinences naturelles (Botanique) Ser. 12. 1960; 1:1–220
12. Van der Walt JJA, Van der Schijff HP, Schweickerdt HG. Anomalous secondary growth in the stem of
lianas Mikania cordata (Burm. F.) Robins (Compositae) and Paullinia pinnata Linn.(Sapindaceae). Kir-
kia. 1973; 9:109–138.
13. Klaassen R. Wood Anatomy of the Sapindaceae. IAWA J. suppl 2. International Association of Wood
Anatomists, Leiden; 1999.
14.
Tamaio N, Neves MF, Brandes AFN, Vieira RC. Quantitative analyses establish the central vascular
cylinder as the standard for wood-anatomy studies in lianas having compound stems (Paullinieae:
Sapindaceae). Flora—Morphology, Distribution, Functional Ecology of Plants. 2011; 206(11):987–
996. https://doi.org/10.1016/j.flora.2011.07.006
15.
Tamaio N. Wood anatomy of lianas of Sapindaceae commercially used in São Paulo-SP. Cerne.
2011; 17(4):533–540.
16. Bastos CL, Tamaio N, Angyalossy V. Unravelling roots of lianas: a case study in Sapindaceae. Annals
of Botany. 2016; 118(4):733–746. https://doi.org/10.1093/aob/mcw091
PLOS ONE | https://doi.org/10.1371/journal.pone.0248369 April 7, 2021
17 / 22
PLOS ONE Fossil Paullinieae
17.
Lopes WAL, De Souza LA, De Almeida OJG. Procambial and cambial variants in Serjania and Urvillea
species (Sapindaceae: Paullinieae). Journal of the Botanical Research Institute of Texas. 2017; 11
(2):421–432.
18. Cunha Neto ILD, Martins FM, Somner GV, Tamaio N. Successive cambia in liana stems of Paullinieae
and their evolutionary significance in Sapindaceae. Botanical Journal of the Linnean Society. 2018;
186(1):66–88. https://doi.org/10.1093/botlinnean/box080
19. Chery JG, Pace MR, Acevedo-Rodrı´guez P, Specht CD, Rothfels CJ. Modifications during early plant
development promote the evolution of nature’s most complex woods. Current Biology. 2020; 30
(2):237–244. https://doi.org/10.1016/j.cub.2019.11.003
20. Acevedo-Rodriguez P. A revision of Lophostigma (Sapindaceae). Systematic Botany. 1993; 18
(3):379–388. https://doi.org/10.2307/2419414
21. Acevedo-Rodrı´guez P, Wurdack KJ, Ferrucci MS, Johnson G, Dias P, Coelho RG, et al. Generic rela-
tionships and classification of tribe Paullinieae (Sapindaceae) with a new concept of supertribe Paulli-
niodae. Systematic Botany. 2017; 42(1):96–114. https://doi.org/10.1600/036364417X694926
22. Chery JG, Acevedo-Rodrı´guez P, Rothfels CJ, Specht CD. Phylogeny of Paullinia L. (Paullinieae:
Sapindaceae), a diverse genus of lianas with dynamic fruit evolution. Molecular Phylogenetics and
Evolution. 2019; 140:106577. https://doi.org/10.1016/j.ympev.2019.106577
23. Buerki S, Forest F, Alvarez N, Nylander JAA, Arrigo N, Sanmartı´n I. An evaluation of new parsimony-
based versus parametric inference methods in biogeography: a case study using the globally distrib-
uted plant family Sapindaceae. Journal of Biogeography. 2011; 38(3):531–550. https://doi.org/10.
1111/j.1365-2699.2010.02432.x
24. Buerki S, Forest F, Stadler T, Alvarez N. The abrupt climate change at the Eocene–Oligocene bound-
ary and the emergence of South-East Asia triggered the spread of sapindaceous lineages. Annals of
Botany. 2013; 112(1):151–160. https://doi.org/10.1093/aob/mct106
25. Muellner-Riehl AN, Weeks A, Clayton JW, Buerki S, Nauheimer L, Chiang YC, et al. Molecular phylo-
genetics and molecular clock dating of Sapindales based on plastid rbcL, atpB and trnL-trnF DNA
sequences. Taxon. 2016; 65(5):1019–1036. https://doi.org/10.12705/655.5
26. Erwin DM, Stockey RA. Sapindaceous flowers from the Middle Eocene Princeton chert (Allenby For-
mation) of British Columbia, Canada. Canadian Journal of Botany. 1990; 68(9):2025–2034. https://doi.
org/10.1139/b90-265
27. Gandolfo MA, Nixon KC, Crepet WL. Selection of fossils for calibration of molecular dating models.
Annals of the Missouri Botanical Garden. 2008; 95(1):34–42. https://doi.org/10.3417/2007064
28. Retallack GJ, Kirby MX. Middle Miocene global change and paleogeography of Panama. Palaios.
2007; 22(6):667–679. https://doi.org/10.2110/palo.2006.p06-130r
29. Kirby MX, Jones DS, MacFadden BJ. Lower Miocene stratigraphy along the Panama Canal and its
bearing on the Central American Peninsula PLoS One. 2008; 3(7):e2791 https://doi.org/10.1371/
journal.pone.0002791 PMID: 18665219
30. MacFadden BJ, Bloch JI, Evans H, Foster DA, Morgan GS, Rincon A, et al. Temporal calibration and
biochronology of the Centenario Fauna, early Miocene of Panama. The Journal of Geology. 2014; 122
(2):113–135. https://doi.org/10.1086/675244
31. Buchs DM, Irving D, Coombs H, Miranda R, Wang J, Coronado M, et al. Volcanic contribution to emer-
gence of Central Panama in the Early Miocene. Scientific reports. 2019; 9(1):1–16. https://doi.org/10.
1038/s41598-018-37790-2 PMID: 30723222
32. Herrera F. Revealing the floristic and biogeographic composition of Paleocene to Miocene Neotropical
forests. [PhD Thesis] University of Florida, Gainesville, FL; 2014.
33. Herrera F, Manchester SR, Jaramillo CA, MacFadden BJ, da Silva-Caminha SA. Phytogeographic his-
tory and phylogeny of the Humiriaceae. International Journal of Plant Sciences. 2010; 171(4):392–
408. https://doi.org/10.1086/651229
34. Herrera F, Manchester SR, Koll R, Jaramillo CA Fruits of Oreomunnea (Juglandaceae) in the early
Miocene of Panama Paleobotany and biogeography: A Festschrift for Alan Graham in his 80th year:
Monographs in systematic botany from the Missouri Botanical Garden. 2014; 128:124–133.
35.
Jud NA, Nelson CW, Herrera F. Fruits and wood of Parinari from the early Miocene of Panama and the
fossil record of Chrysobalanaceae. American Journal of Botany. 2016; 103(2):277–289. https://doi.
org/10.3732/ajb.1500425
36. Nelson CW, Jud NA Biogeographic implications of Mammea paramericana sp. nov. from the lower
Miocene of Panama and the evolution of Calophyllaceae International Journal of Plant Sciences.
2017; 178(3):241–257. https://doi.org/10.1086/689618
37.
Jud NA, Nelson CW. A liana from the lower Miocene of Panama and the fossil record of Connaraceae.
American Journal of Botany. 2017; 104(5):685–693. https://doi.org/10.3732/ajb.1700080
PLOS ONE | https://doi.org/10.1371/journal.pone.0248369 April 7, 2021
18 / 22
PLOS ONE Fossil Paullinieae
38. Herrera F, Carvalho MR, Jaramillo CA, Manchester SR. 19-Million-Year-Old Spondioid Fruits from
Panama Reveal a Dynamic Dispersal History for Anacardiaceae. International Journal of Plant Sci-
ences. 2019; 180(6):479–492. https://doi.org/10.1086/703551
39. Berry EW. Contributions to the geology and paleontology of the Canal Zone, Panama and geologically
related areas in Central America and the West Indies United States National Museum Bulletin. 1918;
103:15–44.
40. Rodrı´guez-Reyes O, Falcon-Lang HJ, Gasson P, Collinson ME, Jaramillo CA. Fossil woods (Malva-
ceae) from the lower Miocene (early to mid-Burdigalian) part of the Cucaracha Formation of Panama
(Central America) and their biogeographic implications. Review of Palaeobotany and Palynology
2014; 209:11–34. https://doi.org/10.1016/j.revpalbo.2014.05.006
41. Rodrı´guez-Reyes O, Gasson P, Falcon-Lang HJ, Collinson ME. Fossil legume woods of the Prioria-
clade (subfamily Detarioideae) from the lower Miocene (early to mid-Burdigalian) part of the Cucara-
cha Formation of Panama (Central America) and their systematic and palaeoecological implications
Review of Palaeobotany and Palynology. 2017; 246:44–61. https://doi.org/10.1016/j.revpalbo.2017.
06.005
42. Rodrı´guez-Reyes O, Gasson P, Thornton C, Falcon-Lang HJ, Jud NA. Panascleroticoxylon crystallosa
gen. et sp. nov.: a new Miocene malpighialean tree from Panama. IAWA Journal. 2017; 38(4):437–
455. https://doi.org/10.1163/22941932-20170178
43.
44.
Jaramillo CA, Moreno E, Ramı´rez V, da Silva-Caminha SAF, de la Barrera A, de la Barrera A, et al.
Palynological record of the last 20 million years in Panama. Paleobotany and biogeography: A fest-
schrift for Alan Graham in his 80th year. 2014; 128:134–251.
Joy KW, Willis AJ, Lacey WS. A rapid cellulose peel technique in palaeobotany. Annals of Botany.
1956; 20(80):635–637.
45. Enquist BJ, Condit R, Peet RK, Schildhauer M, Thiers BM. Cyberinfrastructure for an integrated botan-
ical information network to investigate the ecological impacts of global climate change on plant biodi-
versity. PeerJ Inc.; 2016. e2615v2. Available from: https://peerj.com/preprints/2615.
46. Maitner BS, Boyle B, Casler N, Condit R, Donoghue J, Dura´ n SM, et al. The bien r package: A tool to
access the Botanical Information and Ecology Network (BIEN) database. Methods in Ecology and
Evolution. 2018; 9(2):373–379. https://doi.org/10.1111/2041-210X.12861
47. Graham A. Studies in neotropical paleobotany. II. The Miocene communities of Veracruz, Mexico.
Annals of the Missouri Botanical Garden. 1976; 63(4):787–842. https://doi.org/10.2307/2395250
48. Graham A. Tropical American Tertiary floras and paleoenvironments: Mexico, Costa Rica, and Pan-
ama. American Journal of Botany. 1987; 74(10):1519–1531. https://doi.org/10.1002/j.1537-2197.
1987.tb12143.x
49. Graham A. Studies in Neotropical Paleobotany. X. The Pliocene Communities of Panama–Composi-
tion, Numerical Representations, and Paleocommunity Paleoenvironmental Reconstructions. Annals
of the Missouri Botanical Garden. 1991; 78(2):465–475. https://doi.org/10.2307/2399574
50. Chavez RP, Rzedowski J. Estudio palinolo´gico de las floras fo´ siles del Mioceno inferior y principios del
Mioceno medio de la regio´n de Pichucalco, Chiapas, Me´xico. Acta Bota´ nica Mexicana. 1993; 24:1–96.
51. Hendy AJW. Spatial and stratigraphic variation of marine paleoenvironments in the middle-upper Mio-
cene Gatun Formation, Isthmus of Panama. PALAIOS. 2013; 28(4):210–227. https://doi.org/10.2110/
palo.2012.p12-024r
52. Graham A. Studies in Neotropical Paleobotany. IV. The Eocene communities of Panama. Annals of
the Missouri Botanical Garden. 1985; 72(3):504–534. https://doi.org/10.2307/2399101
53. GBIF.org [11 January 2021] GBIF Occurrence Download https://doi.org/10.15468/dl.7yscgt
54. Gildenhuys E, Ellis AG, Carroll SP, Le Roux JJ. The ecology, biogeography, history and future of two
globally important weeds: Cardiospermum halicacabum Linn. and C. grandiflorum NeoBiota. 2013;
19:45–65. https://doi.org/10.3897/neobiota.19.5279
55. Gildenhuys E, Ellis AG, Carroll SP, Le Roux JJ. Combining natal range distributions and phylogeny to
resolve biogeographic uncertainties in balloon vines (Cardiospermum, Sapindaceae). Diversity and
Distributions. 2015; 21(2):163–174. https://doi.org/10.1111/ddi.12261
56. Wheeler EA, Baas P, Gasson PE, et al. IAWA list of microscopic features for hardwood identification.
IAWA Journal. 1989.
57. Carlquist S. Wood anatomy of Nepenthaceae. Bulletin of the Torrey Botanical Club. 1981; 108
(3):324–330. https://doi.org/10.2307/2484711
58. Carlquist S. Observations on Functional Wood Histology of Vines and Lianas. Aliso: A Journal of Sys-
tematic and Evolutionary Botany. 1985; 11(2):139–157. https://doi.org/10.5642/aliso.19851102.03
59. Carlquist S. Anatomy of vine and liana stems. In: The Biology of Vines. Cambridge University Press;
1991. p. 53–72.
PLOS ONE | https://doi.org/10.1371/journal.pone.0248369 April 7, 2021
19 / 22
PLOS ONE Fossil Paullinieae
60. Scrucca L, Fop M, Murphy TB, Raftery AE. mclust 5: clustering, classification and density estimation
using Gaussian finite mixture models. The R journal. 2016; 8(1):289. https://doi.org/10.32614/RJ-
2016-021
61. R Core Team. R: A Language and Environment for Statistical Computing. Vienna, Austria; 2020.
Available from: https://www.R-project.org/.
62. Schneider CA, Rasband WS, Eliceiri KW. NIH Image to ImageJ: 25 years of image analysis. Nature
methods. 2012; 9(7):671–675. https://doi.org/10.1038/nmeth.2089
63. Ellis B, Daly DC, Hickey LJ, Mitchell JD, Johnson KR, Wilf P, et al. Manual of Leaf Architecture. First
edition ed. Ithaca, NY: Cornell University Press; 2009.
64. Chery JG, da Cunha Neto IL, Pace MR, Acevedo-Rodrı´guez P, Specht CD, Rothfels CJ. Wood anat-
omy of the neotropical liana lineage Paullinia L. (Sapindaceae). IAWA Journal. 2020; 41(3):278–300.
https://doi.org/10.1163/22941932-bja10027
65. Patel RN. A comparison of the anatomy of the secondary xylem in roots and stems. Holzforschung-
International Journal of the Biology, Chemistry, Physics and Technology of Wood. 1965; 19(3):72–79.
66. Bastos CL. Decifrando raı´zes e caules de lianas de Sapindaceae: diversidade macrosco´ pica e con-
exões vasculares [PhD Thesis]. Universidade de São Paulo; 2015.
67. Ronquist F, Teslenko M, Van Der Mark P, Ayres DL, Darling A, Ho¨ hna S, et al. MrBayes 3.2: efficient
Bayesian phylogenetic inference and model choice across a large model space. Systematic biology.
2012; 61(3)539–542. https://doi.org/10.1093/sysbio/sys029 PMID: 22357727
68.
Letunic I, Bork P. Interactive Tree Of Life (iTOL) v4: recent updates and new developments. Nucleic
Acids Research. 2019; 47(W1):W256–W259. https://doi.org/10.1093/nar/gkz239
69. Bouckaert R, Heled J, Ku¨ hnert D, Vaughan T, Wu C-H, Xie D, et al. BEAST 2: a software platform for
Bayesian evolutionary analysis. PLoS Comput Biol. 2014; 10(4):e1003537. https://doi.org/10.1371/
journal.pcbi.1003537 PMID: 24722319
70. Ho¨ hna S, Landis MJ, Heath TA, Boussau B, Lartillot N, Moore BR, et al. RevBayes: Bayesian phyloge-
netic inference using graphical models and an interactive model-specification language. Systematic
biology. 2016; 65(4):726–736. https://doi.org/10.1093/sysbio/syw021 PMID: 27235697
71. Ewers FW, Fisher JB, Chiu S-T. A survey of vessel dimensions in stems of tropical lianas and other
growth forms. Oecologia 1990; 84(4):544–552. https://doi.org/10.1007/BF00328172
72. Ewers FW, Carlton MR, Fisher JB, Kolb KJ, Tyree MT. Vessel diameters in roots versus stems of tropi-
cal lianas and other growth forms. IAWA Journal. 1997; 18(3):261–279. https://doi.org/10.1163/
22941932-90001490
73. Rosell JA, Olson ME. Do lianas really have wide vessels? Vessel diameter–stem length scaling in non-
self-supporting plants. Perspectives in Plant Ecology, Evolution and Systematics. 2014; 16(6):288–
295. https://doi.org/10.1016/j.ppees.2014.08.001
74. MacGinitie HD. An early middle Eocene flora from the Yellowstone-Absaroka volcanic province, north-
western Wind River Basin, Wyoming. First edition ed. Berkeley: University of California Press; 1974.
75. Allen SE. The uppermost Lower Eocene Blue Rim flora from the Bridger Formation of Southwestern
Wyoming: Floristic Composition, Paleoclimate, and Paleoecology [PhD Thesis]. University of Florida.
Gainesville, FL; 2017.
76.
Lesquereux L. Contributions to the flora of the Western Territories III. The Cretaceous and Tertiary flo-
ras. Rept US Geol Surv Terr. 1883; 8:1–283.
77. MacGinitie HD. Fossil plants of the Florissant beds, Colorado. vol. 599 of Carnegie Institution of Wash-
ington Publication. Carnegie Institution of Washington; 1953.
78. Becker HF. Oligocene plants from the upper Ruby River Basin, southwestern Montana. vol. 82. Geo-
logical Society of America; 1961.
79.
Lielke K, Manchester SR, Meyer H, and others. Reconstructing the environment of the northern Rocky
Mountains during the Eocene/Oligocene transition: constraints from the palaeobotany and geology of
south-western Montana, USA. Acta palaeobotanica. 2012; 52(2):317–358.
80. Knowlton FH. Revision of the flora of the Green River formation, with descriptions of new species;
1923. 131–F. Available from: http://pubs.er.usgs.gov/publication/pp131F.
81. MacGinitie HD. The Eocene Green River flora of northwestern Colorado and northeastern Utah. Uni-
versity of California Press; 1969.
82. Brown RW. Additions to the flora of the Green River Formation. US Geological Survey Professional
Paper. 1929;154:279–292.
83. Brown RW. The recognizable species of the Green River flora. United States. Government Printing
Office. 1934.
PLOS ONE | https://doi.org/10.1371/journal.pone.0248369 April 7, 2021
20 / 22
PLOS ONE Fossil Paullinieae
84.
Johnson KR, Plumb C. Common plant fossils from the Green River Formation at Douglas Pass, Colo-
rado, and Bonanza, Utah. In: The Green River Piceance Creek and eastern Unita Basin: Grand Junc-
tion Colorado. Grand Junction Geological Society. 1995; p. 121–130.
85. Wolfe JA, Wehr W. Middle Eocene dicotyledonous plants from Republic, northeastern Washington.
US Geological Survey Bulletin 1597. 1987;p. 1–25.
86. McClain AM, Manchester SR. Dipteronia (Sapindaceae) from the Tertiary of North America and impli-
cations for the phytogeographic history of the Aceroideae. American Journal of Botany. 2001; 88
(7):1316–1325. https://doi.org/10.2307/3558343
87.
Fittipaldi FC, Simões MG, Giulietti AM, Pirani JR. Fossil plants from the Itaquaquecetuba Formation
(Cenozoic of the São Paulo Basin) and their possible paleoclimatic significance. Boletim IG-USP Publi-
cac¸ão Especial. 1989; 7:183–203.
88. Dos Santos D, Garcia M, Saad A, Bistrichi C. Itaquaquecetuba Formation Palynostratigraphy, São
Paulo Basin, Brazil. Revista Brasileira de Paleontologia. 2010; 13:205–220.
89. Duarte L, Pimentel de Rezende-Martins A. Contribuı´c¸ão ao conhecimento da flora Cenozo´ica do Bra-
sil: Jazigo Vargem Grande do Sul, SP: se´ rie taubate´ . I. Anais da Academia Brasileira de Ciências.
1983; 55(1):109–121.
90. Dos Santos MA, Bernardes-de Oliveira MEC. Taxonomia da tafoflora Neogena da Formac¸ão Rio
Claro, Jaguariuna, Estado de Sao Paulo, Brasil. Revista Brasileira de Paleontolgia. 2013; 16(3):465–
486. https://doi.org/10.4072/rbp.2013.3.07
91. Edwards WN, Wonnacott F. Sapindaceae. In: Fossilum Catalogus 2, Plantae. pars. 14. W. Junk; 1928
p. 1–84.
92. Cunha Neto IL, Martins FM, Somner GV, Tamaio N. Secretory structures in stems of five lianas of
Paullinieae (Sapindaceae): morphology and histochemistry Flora. 2017; 235:29–40
93. Pellissari LCO, Barros CF, Medeiros H, Tamaio N. Cambial patterns of Paullinia (Sapindaceae) in
southwestern Amazonia, Brazil. Flora. 2018; 246-247:71–82. https://doi.org/10.1016/j.flora.2018.07.
002
94. Metcalfe CR, Chalk L. Anatomy of the Dicotyledons, Vols. 1 & 2. Clarendon Press; 1950.
95. Manchester SR. Update on the megafossil flora of Florissant, Colorado. Denver Museum of Nature
and Science. 2001; 4:137–161.
96.
Jaramillo C, Rueda MJ. A Morphological Electronic Database of Cretaceous-Tertiary and Extant Pol-
len and Spores from Northern South America, v. 2020. Smithsonian Tropical Research Institute. http://
biogeodb.stri.si.edu/jaramillosdb/web/morphological/ (accessed 18 Jan 2021)
97.
Ferrucci MS, Anzotegui LM. El polen de Paullinieae (Sapindaceae). Bonplandia. 1993;p. 211–243.
98. Van der Ham R, Tomlik A. Serjania pollen and the origin of the tribe Paullinieae (Sapindaceae).
Review of Palaeobotany and Palynology. 1994; 83(1-3):43–53. https://doi.org/10.1016/0034-6667(94)
90056-6
99. Montes C, Cardona A, McFadden R, Moro´ n SE, Silva CA, Restrepo-Moreno S, et al. Evidence for mid-
dle Eocene and younger land emergence in central Panama: Implications for Isthmus closure. Geolog-
ical Society of America Bulletin. 2012; 124(5-6):780–799. https://doi.org/10.1130/B30528.1
100. Ramı´rez DA, Foster DA, Min K, Montes C, Cardona A, Sadove G. Exhumation of the Panama base-
ment complex and basins: Implications for the closure of the Central American seaway. Geochemistry,
Geophysics, Geosystems. 2016; 17(5):1758–1777. https://doi.org/10.1002/2016GC006289
101.
Leopold EB, MacGinitie HD. Development and affinities of Tertiary floras in the Rocky Mountains. In:
Floristics and Paleofloristics of Asia and Eastern North America. Elsevier Publishing Company; 1972.
p. 147–200.
102. Gray J. Temperate pollen genera in the Eocene (Claiborne) flora, Alabama. Science. 1960; 132
(3430):808–810. https://doi.org/10.1126/science.132.3430.808
103. Muller J. Fossil pollen records of extant angiosperms. The Botanical Review. 1981; 47(1):1–142.
https://doi.org/10.1007/BF02860537
104. Bellonzi TK, Dutra FV, de Souza CN, Rezende AA, Gasparino EC. Pollen types of Sapindaceae from
Brazilian forest fragments: apertural variation. Acta Botanica Brasilica. 2020; 34(2):327–341. https://
doi.org/10.1590/0102-33062020abb0022
105. Carlquist S. Comparative Wood Anatomy: Systematic, Ecological, and Evolutionary Aspects of Dicoty-
ledon Wood. Springer Science & Business Media; 2001.
106. Gentry AH. Distribution and evolution of climbing plants. In: The Biology of Vines. Cambridge Univer-
sity Press; 1991. p. 3–49.
107. Schnitzer SA, Bongers F. The ecology of lianas and their role in forests. Trends in Ecology & Evolution.
2002; 17(5):223–230. https://doi.org/10.1016/S0169-5347(02)02491-6
PLOS ONE | https://doi.org/10.1371/journal.pone.0248369 April 7, 2021
21 / 22
PLOS ONE Fossil Paullinieae
108. Schnitzer SA. A mechanistic explanation for global patterns of liana abundance and distribution. The
American Naturalist. 2005; 166(2):262–276. https://doi.org/10.1086/431250
109. Van Der Heijden GMF, Phillips OL. Environmental effects on Neotropical liana species richness. Jour-
nal of Biogeography. 2009; 36(8):1561–1572. https://doi.org/10.1111/j.1365-2699.2009.02099.x
PLOS ONE | https://doi.org/10.1371/journal.pone.0248369 April 7, 2021
22 / 22
PLOS ONE |
10.1371_journal.pone.0244962 | RESEARCH ARTICLE
Feasibility and validation of a web-based
platform for the self-administered patient
collection of demographics, health status,
anxiety, depression, and cognition in
community dwelling elderly
Matthew Calamia1*, Daniel S. Weitzner1, Alyssa N. De Vito1, John P. K. Bernstein1,
Ray AllenID
2, Jeffrey N. Keller2
1 Department of Psychology, Louisiana State University, Baton Rouge, Louisiana, United States of America,
2 Pennington Biomedical Research Center, Baton Rouge, Louisiana, United States of America
* mcalamia@lsu.edu
Abstract
The coronavirus disease pandemic has brought a new urgency for the development and
deployment of web-based applications which complement, and offer alternatives to, tradi-
tional one-on-one consultations and pencil-and-paper (PaP) based assessments that cur-
rently dominate clinical research. We have recently developed a web-based application that
can be used for the self-administered collection of patient demographics, self-rated health,
depression and anxiety, and cognition as part of a single platform. In this study we report the
findings from a study with 155 cognitively healthy older adults who received established
PaP versions, as well as our novel computerized measures of self-rated health, depression
and anxiety, and cognition. Moderate to high correlations were observed between PaP and
web- based measures of self-rated health (r = 0.77), depression and anxiety (r = 0.72), and
preclinical Alzheimer’s disease cognitive composite (PACC) (r = .61). Test-retest correla-
tions were variable with high correlations for a measure of processing speed and a measure
of delayed episodic memory. Taken together, these data support the feasibility and validity
of utilization of this novel web-based platform as a new alternative for collecting patient
demographics and the assessment of self-rated health, depression and anxiety, and cogni-
tion in the elderly.
Introduction
The coronavirus disease 19 (Covid-19) pandemic and the resulting direct and indirect impacts
of social distancing dramatically interrupted or stopped clinical research around the world.
These and other realities in the wake of Covid-19 have created a new urgency for the genera-
tion of web-based research platforms which provide alternatives to face-to-face and pencil-
and-paper (PaP) based assessments, and reduce the dependence on the manual transfer of PaP
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: Calamia M, Weitzner DS, De Vito AN,
Bernstein JPK, Allen R, Keller JN (2021) Feasibility
and validation of a web-based platform for the self-
administered patient collection of demographics,
health status, anxiety, depression, and cognition in
community dwelling elderly. PLoS ONE 16(1):
e0244962. https://doi.org/10.1371/journal.
pone.0244962
Editor: Simone Reppermund, University of New
South Wales, AUSTRALIA
Received: April 17, 2020
Accepted: December 19, 2020
Published: January 19, 2021
Copyright: © 2021 Calamia et al. This is an open
access article distributed under the terms of the
Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: Data cannot be
shared publicly because of this was not permitted
by the consent form signed by participants. Data
are available from the Keller-Lamar Health
Foundation (info@keller-lamar.org) for researchers
who can provide evidence of IRB approval for
access.
Funding: The study was funded by a contract from
the Keller-Lamar Health Foundation (http://www.
PLOS ONE | https://doi.org/10.1371/journal.pone.0244962 January 19, 2021
1 / 15
PLOS ONE keller-lamar.org/) awarded to MC. The funders had
no role in study design, data collection and
analysis, decision to publish, or preparation of the
manuscript.
Competing interests: The authors have declared
that no competing interests exist.
Web-based platform for self-administered assessments in the elderly
data to an electronic database. Several computerized and web-based applications are currently
available for conducting individualized assessments for a variety of clinical endpoints, includ-
ing cognition [1]. However, these tools typically do not readily interact with a centralized
study database and generally lack the ability to collect the supporting clinical data that accom-
pany clinical research studies (e.g., demographics, secondary endpoint data collection). In
order to address these research gaps, we have created a web-based platform which allows for
the self-administered collection of patient demographics, the delivery and automated scoring
of multiple assessments, and the capability to automatically populate all study data into a single
secure and functional electronic database.
The fastest growing segment of the United States population is those 85 years of age and
older, with age related diseases such as Alzheimer’s disease related dementia (ADRD) expected
to increase from 5 million to 15 million in the next three decades [2, 3]. Recent ADRD research
efforts have focused on developing multicomponent assessments of cognitive function, with
an emphasis on developing composite cognitive assessments that are sensitive enough to mea-
sure the earliest changes relevant to the future development of ADRD. The Alzheimer’s Dis-
ease Cooperative Study Preclinical Alzheimer’s Cognitive Composite (ADCS-PACC) is a PaP
based assessment package that has emerged as the leading clinical research tool for aging, mild
cognitive impairment, and pre-ADRD research. The ADCS-PACC focuses on the assessment
of the three cognitive domains which are the most predictive for the development of ADRD
[4]. The ADCS-PACC is the primary endpoint in one of the largest clinical trials for AD pre-
vention [5], and is a major cognitive endpoint for some of the largest longitudinal and cohort
studies around the world [6]. Computer-based assessments have increasingly been used and
valued for clinical care and research including studies of the elderly [7–9], clinical trials
focused on cognition [10], and longitudinal studies with elderly participants [11, 12]. Cur-
rently there are no computerized/web-based options for the ADCS-PACC even though such
an advance would provide a potential option that decreases the need for face-to-face assess-
ments, manual scoring, manual z-score transformation, and manual data transfer to an elec-
tronic database that currently accompanies all ADCS-PACC efforts.
The current study focused on the validity and feasibility of using the computerized PACC
(cPACC), a novel web-based application which employs a self-administered approach for
elderly participants to provide demographic data as well as undergo assessments of self-rated
health, depression, anxiety, and cognition. Analysis of 155 community dwelling elderly sub-
jects demonstrates the feasibility of collecting data for each of these aspects in a self-adminis-
tered manner that resulted in the automated population of a single, secured, cloud-based
database. We report on the validity for each of the web-based measures with traditional PaP
based assessments and report on their reliability as part of a two-week test-retest design in a
subset of participants.
Methods
The demographic, assessment, and database platform
The platform used in this study was created by developers at Pennington Biomedical Research
Center. The platform consisted of a web application written in Angular v.6 communicating
with an API developed in Microsoft ASP.NET Core v2.1. The participants used the web-based
application to answer a series of questions, and complete the different cognitive tasks, in a self-
guided manner. As each question and assessment was completed the resulting data populated
a central database that contained the demographic profile and assessment scores for each par-
ticipant. All data was stored in a Microsoft SQL Server database. The entire system was oper-
ated as a web application in Microsoft Azure. Data was extracted from Azure using Microsoft
PLOS ONE | https://doi.org/10.1371/journal.pone.0244962 January 19, 2021
2 / 15
PLOS ONE Web-based platform for self-administered assessments in the elderly
SQL Server Management Studio. Administrative rights within the platform were used to con-
trol access to functionality and data.
Participants
Individuals were recruited to the study who were 55–95 years old (inclusive), who did not have
motor or sensory deficits that were sufficient to interfere with the ability of the participant to
complete computerized assessments. Fliers, email blasts to a clinical registry of individuals
aged 50 and over, and word of mouth were used to recruit participants in the Baton Rouge
area. A total of 174 participants met study criteria and provided written informed consent.
Most of these participants (n = 155) had Mini-Mental State Examination (MMSE) scores in a
range suggesting intact global cognition (i.e., greater than or equal to 25) and are the focus of
all analyses other than the one analysis also comparing their scores against a small group of
participants with MMSE scores below 25 (n = 19). Demographic information for the study
sample is provided in Tables 1 and 2. See Table 3 for raw performance data for participants.
Missing data ranged from 0–15 participants across measures.
Table 1. Participant demographics.
MMSE � 25
MMSE < 25
Demographic Variables
Age
Female
Non-Hispanic
Race
Caucasian
African American
Bi-racial
Native American
Highest Degree of Education
GED
Some College
Associate’s Degree
Bachelor’s Degree
Master’s Degree
Doctorate Degree
Marital Status
Married
Widowed
Divorced
Never Married
Common-Law Partner
Living Situation
Living Alone
Residence Type
Single Family Home
Apartment
Assisted Living
Mean (SD)
71.64 (8.13)
-
-
-
-
-
-
-
-
-
-
-
n (%)
-
111 (71.6%)
145 (93.5%)
140 (90.3%)
7 (4.5%)
2 (1.3%)
1 (0.01%)
10 (6.5%)
33 (21.3%)
7 (3.9%)
41 (26.5%)
52 (33.5%)
6 (3.9%)
82 (55.0%)
28 (18.8%)
23 (15.4%)
14 (9.4%)
2 (1.3%)
48 (31.0%)
111 (71.6%)
35 (22.6%)
3 (0.6%)
n (%)
-
6 (31.6%)
18 (94.7%)
17 (89.5%)
1 (5.3%)
-
-
1 (5.3%)
1 (5.3%)
1 (5.3%)
7 (36.8%)
4 (21.1%)
2 (10.5%)
9 (47.4%)
6 (31.6%)
2 (10.5%)
1 (5.3%)
-
5 (26.3%)
11 (57.9%)
4 (21.1%)
3 (15.8%)
Mean (SD)
75.94 (11.10)
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
Note: Demographic information for some variables was unavailable and therefore not all variables will sum to a total of 155 and 19 individuals. MMSE = Mini-Mental
State Examination
https://doi.org/10.1371/journal.pone.0244962.t001
PLOS ONE | https://doi.org/10.1371/journal.pone.0244962 January 19, 2021
3 / 15
PLOS ONE Web-based platform for self-administered assessments in the elderly
Table 2. Prevalence of health conditions in entire sample.
Health Condition
Cardiovascular
High Blood Pressure
High Cholesterol
Diabetes
Heart Attack
Atrial Fibrillation
Neurological
Stroke
Parkinson’s Disease
Multiple Sclerosis
Transient Ischemic Attack
Alzheimer’s Disease
Other Dementia
Other Neurological Disease
Concussion/TBI
Psychiatric
Alcohol Abuse
Drug Abuse
Depression
Anxiety
Other
B12 Deficiency
Sleep Apnea
Thyroid Deficiency
Cancer
n (%)
72 (46.5%)
55 (35.5%)
18 (11.6%)
4 (2.6%)
14 (9.0%)
2 (1.3%)
2 (1.3%)
0 (0.0%)
1 (0.6%)
0 (0.0%)
3 (1.9%)
4 (2.6%)
2 (1.3%)
3 (1.9%)
1 (0.6%)
31 (20.0%)
27 (17.4%)
6 (3.9%)
17 (11.0%)
33 (21.3%)
26 (16.8%)
https://doi.org/10.1371/journal.pone.0244962.t002
Procedures
The measures were administered on the same day, with half of the participants completing the
PaP measures first, and the other half completing the web-battery first. Participants completed
the measures in a quiet and private testing room on either a desktop or laptop computer with a
computer mouse. PaP measures were administered by a trained research assistant. Research
assistants remained in the room while participants completed the computerized measures, but
only to address technological issues (e.g., computer froze/internet connection issues) or pro-
vide encouragement to participants.
A subset of the sample with MMSE scores greater than or equal to 25 (n = 55) were ran-
domly selected to complete a second visit approximately two weeks later during which they
repeated the cPACC to assess for test-retest reliability. The first study visit was on June 11,
2018 and the last study visit was on October 9, 2019. All study procedures were approved by
the LSU Institutional Review Board and were conducted according to the principles
expressed in the Declaration of Helsinki. All data collected during the assessment were stored
immediately at the conclusion of each page. Data were written to a Microsoft SQL Server
2014 database and stored as the raw answer provided by the participant. Answers to some
tests such as the participant typing the name and hobby of a person in an image were reported
as the exact text entered by the participant. Other tests using multiple choice answers or clicks
on a grid were scored as number of correct answers and where applicable number of
attempts.
PLOS ONE | https://doi.org/10.1371/journal.pone.0244962 January 19, 2021
4 / 15
PLOS ONE Web-based platform for self-administered assessments in the elderly
Table 3. Means and standard deviations for cognitive measures and questionnaires in participants scoring above and below 25 on the MMSE.
MMSE � 25
MMSE < 25
Variables
FNHR-IFR
FNHR-IR
FNHR-DFR
FNHR-DR
GLIR
GLDR
SL
SM
VP
VAS
LM-DR
DSC
FCSRT
GAI
GDS
n
148
149
148
149
155
151
149
140
142
152
152
151
151
152
152
Mean (SD)
6.54 (2.90)
28.33 (2.99)
10.24 (3.52)
14.65 (2.01)
14.05 (3.74)
7.03 (2.67)
10.50 (6.44)
24.54 (7.85)
19.27 (5.58)
83.91 (13.33)
6.19 (3.02)
51.11 (13.49)
47.66 (1.48)
1.61 (3.06)
5.54 (4.84)
Min
0
17
2
0
5
0
0
0
1
18
1
24
31
0
0
Max
14
32
16
16
23
12
24
42
28
100
18
90
49
16
25
n
17
16
15
15
16
19
18
16
16
18
17
17
17
18
17
Mean (SD)
2.53 (2.24)
21.06 (7.04)
4.27 (4.85)
10.00 (3.70)
8.44 (3.98)
3.53 (2.95)
4.00 (3.93)
11.44 (7.25)
9.63 (7.16)
82.33 (12.98)
2.24 (2.93)
28.06 (13.83)
40.24 (10.83)
3.00 (2.68)
8.18 (5.87)
Min
0
10
0
4
4
0
0
0
0
60
0
3
4
0
1
Max
8
31
14
16
18
10
14
24
19
100
8
64
48
10
22
Note: SD = Standard Deviation; MMSE = Mini-Mental State Examination; FNHR-IR = Face Name Hobby Recall Immediate Free Recall; FNHR-IFR = Face Name
Hobby Recall Immediate Recognition; FNHR-DFR = Face Name Hobby Recall Delayed Free Recall; FNHR-DR = Face Name Hobby Recall Delayed Recognition;
GLIR = Grid Locations Immediate Recall; GLDR = Grid Locations Delayed Recall; SL = Symbol Line; VP = Visual Patterns; SM = Speeded Matching; VAS = EQ-5D
Visual Analog Scale; Logical Memory–Delayed Recall; DSC = Digit Symbol Coding; FCSRT = Free and Cued Selective Reminding Test; GAI = Geriatric Anxiety
Inventory; GDS = Geriatric Depression Scale.
https://doi.org/10.1371/journal.pone.0244962.t003
Paper and Pencil (PaP) measures
Questionnaires. Participants completed the EQ-5D Visual Analog Scale (VAS) to assess
self-rated health [13]. For this measure, participants rate their current health on a 0 to 100
scale from the “worst health” to “best health” they can imagine. The EQ-5D VAS is sensitive to
individual differences such as age [14] and physical activity [15]. The Geriatric Anxiety Inven-
tory (GAI) and Geriatric Depression Scale (GDS) were used to assess anxiety and depression,
respectively. The GAI is a 20-item geriatric-focused self-report measure of anxiety-related
symptoms [16]. The GAI demonstrates excellent internal consistency (α = 0.91) and test-retest
reliability (r = 0.91) [16] as well as good convergent validity with worry and anxiety measures
[17]. The Geriatric Depression Scale (GDS) is a 30-item self-report which measures depressive
symptoms in older adults [18]. The GDS demonstrates excellent internal consistency (α =
0.94), good test-retest reliability (r = 0.84) [19], and at least adequate convergent validity with
other depression measures such as the Beck Depression Inventory-II (r = .78) [20]. However,
despite good convergent validity, the discriminant validity of these measures is weak with one
study finding a correlation as high as r = .86 between the GAI and GDS [21]. A 12-item com-
puter proficiency questionnaire [22] was used in order to assess how easily older adults felt
they could perform tasks on a computer (e.g., “Use a keyboard to type”) in a 5-point likert
scale format. The sample had a self-reported mean computer proficiency rating of 3.17 (SD =
.98), indicating that on average, they could somewhat easily perform computer-based tasks.
Alzheimer’s Disease Cooperative Study Preclinical Alzheimer’s Cognitive Composite
(ADCS-PACC). A review by Alzheimer’s disease cooperative study (ADCS) identified episodic
memory, executive function, and orientation as the 3 key cognitive domains linked to the
development of mild cognitive impairment and ADRD [4]. A total of 4 pencil-and- paper
(PaP) cognitive assessments were selected to capture these domains as part of the ADCS
PLOS ONE | https://doi.org/10.1371/journal.pone.0244962 January 19, 2021
5 / 15
PLOS ONE Web-based platform for self-administered assessments in the elderly
Preclinical Alzheimer’s Disease Composite (PACC). The ADCS-PACC is comprised of imme-
diate recall on the Free and Cued Selective Reminding Test (FCSRT), delayed recall on one
story from the Logical Memory subtest of the Wechsler Memory Scale-Revised battery, the
Digit-Symbol Test from the Wechsler Adult Intelligence Scale-Revised, and the MMSE.
Studies using the ADCS-PACC have been able to identify study subjects who would go on
to develop clinical biomarkers of ADRDs such as pathological beta amyloid deposition as well
as identify which study subjects exhibit the fastest rate of cognitive decline in longitudinal
studies [23, 24]. Due to these successes, the ADCS-PACC has emerged as one of the most com-
monly utilized cognitive batteries in prominent clinical trials and longitudinal research studies
including the A4 trial and Alzheimer’s Disease Neuroimaging Initiative (ADNI), respectively.
Web-battery measures
Questionnaires. The web battery included survey items regarding participant demo-
graphics (i.e., date, gender, zip code, ethnicity, race, marital status, living situation, and highest
level of education attained), and health history (i.e., a list of conditions presented as a check-
list). Additionally, we collected information on family history of dementia, pain severity and
interference on daily functioning, frequency of exercise, number of medications and medica-
tion adherence, concern about driving and accident history, self-rated health, and subjective
memory complaints that will be a part of future research. Responses to the demographic and
health history questions can be found in Tables 1 and 2.
For the purposes of this study, psychometric validation focused on 1) a self-report measure
of health in which participants make one global rating of their health and 2) a new 17-item
measure of depression and anxiety developed based on widely used measures of depression
and anxiety. Participants were asked to rate how much they felt or experienced certain symp-
toms over the past 2 weeks on a 5-point scale (“not at all” to “extremely”). Given that brief
measures of depression and anxiety show poor discriminant validity [15], these symptoms
were assessed jointly rather than with the aim of developing two separate scales.
Computerized Preclinical Alzheimer’s Cognitive Composite (cPACC). The cognitive measure
in this study was a cognitive composite that was validated against the (ADCS-PACC). Like the
ADCS-PACC, the cPACC was designed to assess the domains of orientation and episodic
memory. The cPACC also includes a measure of processing speed designed to be comparable
to the PaP measure of digit symbol coding which the PACC considers a measure of executive
functioning. Additionally, the cPACC includes measures of working memory given working
memory is related to executive functioning [25], a PACC domain, and a working memory
item is included on the MMSE which is used as part of the PACC.
Orientation. For orientation participants are asked orientation questions on the computer
screen (day, year, time of day, etc.) and select the answers from a list of multiple-choice
response options. Participants receive 1 point for each correct answer.
Face Name Hobby Recall (FNHR). This cPACC component is designed to assess episodic
memory which is one component of the ADCS-PACC. It is based on the short version of the
Face-Name Associative Memory Exam [26–28]. For cPACC Faces and Names the participant
first completes a learning trial in which 8 faces with a name and hobby presented underneath.
The names and hobbies chosen are short in word length (e.g., Amy, Hiker). Faces vary in age,
gender, and race. Stimuli are presented twice and are followed by immediate recall trial each
time in which they have to recall the names and hobbies when presented only with the face by
typing their responses into a text box and then clicking “next” to submit their response. Partic-
ipants receive 1 point each for correctly naming the person’s hobby and their name, for a total
of 2 points per stimulus. An immediate recognition trial then follows in which they must select
PLOS ONE | https://doi.org/10.1371/journal.pone.0244962 January 19, 2021
6 / 15
PLOS ONE Web-based platform for self-administered assessments in the elderly
the correct name and hobby from a multiple-choice list by clicking on the correct stimulus.
Participants receive 1 point for correctly selecting each name and hobby for a total of 2 possible
points. After a ~15-minute delay in which they complete other cPACC measures, delayed
recall and recognition trials are completed
Grid locations. The grid location test is a measure of visual episodic memory designed based
on the Visual Spatial Learning Test [29, 30], a measure designed to be a visual equivalent to
verbal list-learning paradigms. Scores on this measure highly correlated with verbal memory
measures [29, 30]. For cPACC grid locations, participants complete two learning and immedi-
ate memory trials in which they see 6 symbols on a 4x4 grid and then have to select the symbols
they saw and put them in the correct location. For each symbol, participants can earn up to 2
points (1 point for selecting the correct symbol and 1 point for placing the symbol in the cor-
rect location), for a possible of 12 points. The same symbols and locations are used for both
learning trials. Participants then complete a delayed memory trial after ~15-minute delay.
Speeded matching. Speeded matching is a measure of processing speed and executive func-
tion that is based on the Wechsler Adult Intelligence Scale—Revised (WAIS-R) Digit Symbol
Coding subtest [31], a measure included in the ADCS-PACC. For the cPACC participants
have 90 seconds to select symbols that correspond to numbers based on a key matching each
unique symbol to a specific number. As participants select symbols, they appear in the blank
boxes above the numbers. As participants complete more matches, additional numbers with
blank boxes above them appear on the screen. Participants receive 1 point for each correctly
selected symbol.
Symbol line. Symbol line is a measure of visual working memory based on Wechsler Mem-
ory Scale—Fourth Edition (WMS-IV) Symbol Span [32]. Participants see a line of symbols and
then have to correctly select which symbols they saw in the correct order (i.e., left to right). Ini-
tially participants are shown only two symbols in a line, but lines of increasing lengths are
added until a participant makes no correct responses or is presented with a trial of 7 symbols.
For each symbol, participants can achieve a total possible of 2 points. If participants recall
incorrect symbols, they receive 0 points. If all of the correct symbols are recalled, but in the
incorrect order, participants receive 1 point. If participants recall the correct symbols in the
correct order, they receive 2 points.
Visual patterns. Visual patterns is a measure of visual working memory based on Wechsler
Memory Scale - 3rd Edition (WMS-III) Spatial Span [33]. Participants see an array of 9 white
boxes and are asked to recall the order in which boxes are turned black. A box that is turned
black returns to white before the next box turns black. Initially participants are shown only
two boxes that are turned black but increasing numbers of boxes are turned black until a par-
ticipant makes no correct responses or is presented with a trial of 7 boxes. Participants receive
1 point for the correct completion for each sequence.
Analyses
Validity. Pearson correlations were used to examine the relationship between scores on
questionnaire measures administered via PaP or the web-battery. For the GAI and GDS, scores
were first converted to z-scores and a composite was created to compare with the web-based
measure of depression and anxiety symptoms. For the web-battery measure of depression and
anxiety, a confirmatory factor analysis (CFA) was first conducted as part of assessing construct
validity to assess whether a one-factor model provided adequate model fit. To compare the
ACDS-PACC and cPACC using a Pearson correlation, individual tests administered were also
first converted to z-scores. Thus, for the PaP, the FCSRT, Logical Memory Delayed Score, and
MMSE total score were individually standardized into Z-scores, and then summed together to
PLOS ONE | https://doi.org/10.1371/journal.pone.0244962 January 19, 2021
7 / 15
PLOS ONE Web-based platform for self-administered assessments in the elderly
create the PaP composite score. To create the cPACC composite score, the number of correct
responses recalled during the Faces and Names immediate and long-delay free recall and multiple
choice, Grid Locations, Symbol Matching, Symbol Line, and Grid Pattern tasks were individually
standardized into Z-scores. These Z-scores were then summed together to create the cPACC com-
posite score. Only two participants in our cognitively intact sample missed an orientation item.
Therefore, orientation was not included when creating a cPACC composite score.
To assess the sensitivity of the cPACC to cognitive impairment, we calculated effect sizes
using Hedges’ g to determine whether the subtests of the cPACC could differentiate between
those with MMSE scores above and below 25. Hedges’ g was used given the large difference in
sample sizes between the cognitively healthy and cognitively impaired group.
Practice effects. Dependent t-tests and Cohen’s d were used to examine practice effects
on the web-battery in the subsample who completed a second visit approximately two weeks
following the initial visit.
Reliability. To assess internal consistency of the measure of depression and anxiety, coef-
ficient alpha was used. To assess the test-retest reliability of the web-battery, Pearson correla-
tions were used to examine the relationship of questionnaire and cognitive test scores
administered within an approximately two-week test-retest interval.
Results
Validity and reliability of the web-based questionnaire
Adequate fit for a one-factor model for the web-battery measure of depression and anxiety
(CFI = 0.91, RMSEA = 0.08) was obtained when allowing for two pairs of correlated residuals
for items with similar content (i.e., “I was easily upset” and “I was easily annoyed”; “I had diffi-
culty stopping myself from worrying” and “I worried a lot.”). Coefficient alpha for this scale
was .91. A high correlation was observed between the web-battery measure of depression and
anxiety and the GDS/GAI composite, r = 0.70. Similarly, a high correlation, r = .77, was
obtained between the web-battery measure of self-rated health and the EQ-5D VAS (see Fig 1).
Fig 1. Relationships between web-based measures and paper and pencil measures. Note: A.) Relationship between
the Computerized Preclinical Alzheimer’s Cognitive Composite (cPACC) and the Alzheimer’s Disease Cooperative
Study Preclinical Alzheimer’s Cognitive Composite (ADCS-PACC). B.) Relationship between the web-battery measure
of depression and anxiety and the GDS/GAI composite score. C.) Relationship between the web-battery measure of
self-rated health and the EQ-5D Visual Analog Scale (VAS).
https://doi.org/10.1371/journal.pone.0244962.g001
PLOS ONE | https://doi.org/10.1371/journal.pone.0244962 January 19, 2021
8 / 15
PLOS ONE Web-based platform for self-administered assessments in the elderly
In the 55 older adults in the sample who completed a retest after approximately 2 weeks, the
web-battery measure of depression and anxiety and self-rated health both had high test-retest
correlations, r = 0.85 and r = .0.83, respectively.
Validity and reliability of the Computerized Preclinical Alzheimer’s
Cognitive Composite (cPACC)
The composite scores derived from the cPACC and the ADCS-PACC were found to be mod-
erately related (r = .61) (Fig 1). As an additional exploratory analysis using stepwise regres-
sion showed that this same correlation could be obtained using only a subset of measures:
Speeded Match, the immediate trials of Face Name Hobby Recall, the immediate trials of
Face Name Hobby Recognition, and the delayed trial of Grid Locations (F(4,132) = 22.08,
R2 = .401). See S1 Table for the full results of the regression analysis. In addition, the Speeded
Match task moderately correlated to the Digit Symbol Coding subtest (r = .56), thus demon-
strating convergent validity between a measure of the cPACC and a PaP measure it was
designed to match (see S2 Table for relationships among all measures on the CPACC and the
PaP measures).
All of the measures of working memory, episodic memory, and processing scored as part of
the cPACC battery significantly differed (Hedges’ g ranged from 1.12 to 2.30) between those
above and below an MMSE score of 25, which is a common cutoff for cognitive impairment.
The differences between those above and below an MMSE score of 25 were larger in the com-
posite PaP score compared to the composite cPACC score (Hedges’ g = 2.98 vs 2.31). However,
when the MMSE was removed from the PaP composite score, the differences between those
above and below an MMSE score of 25 were larger in the composite cPACC score compared
to the composite PaP score (Hedges g = 2.31 to 2.17).
High test-retest reliability was obtained on delayed free-recall and multiple-choice subtests
of the Faces and Names test (r = .70 to r = .74) as well as on a measure of processing speed sim-
ilar to digit symbol coding on the PaP (r = .73) (Table 4). However, tasks of visual working
memory demonstrated weak to moderate test-retest reliability (r = .36 to r = .45).
Both episodic memory tasks demonstrated significant practice effects (p’s < .01) on both
immediate and delayed-recall trials. However, measures of processing speed and visual work-
ing memory tasks did not (p’s > .05; see Table 4).
Table 4. Test-retest correlations and practice effects between baseline and follow-up visits.
Test
Face Name Hobby Recall Immediate Free Recall
Face Name Hobby Recall Immediate Recognition
Face Name Hobby Recall Delayed Free Recall
Face Name Hobby Recall Delayed Recognition
Grid Locations Immediate Recall
Grid Locations Delayed Recall
Symbol Line
Visual Patterns
Speeded Matching
R
.56
.59
.70
.74
.57
.48
.36
.45
.73
t
9.52���
6.34���
5.84���
3.93���
6.25���
3.60��
1.08
1.37
.075
d
1.25
.78
.61
.39
.78
.49
.16
.19
.02
Note: All correlations significant at p < .01
�� indicates significant dependent t-test value at the p < .01 level
��� indicates significant dependent t-test value at the p < .001 level
d = Cohen’s d
https://doi.org/10.1371/journal.pone.0244962.t004
PLOS ONE | https://doi.org/10.1371/journal.pone.0244962 January 19, 2021
9 / 15
PLOS ONE Web-based platform for self-administered assessments in the elderly
Discussion
The current study demonstrates the feasibility of using a novel web-based application for the
collection of study subject demographics, as well as the results from diverse computer-based
assessments, in the elderly. The current feasibility and validity study was conducted under con-
ditions where data was collected both in traditional research settings (i.e., lab space on a uni-
versity campus) as well as in senior living communities. Although a research assistant was
present in case assistance was needed for participants to navigate the web-battery, for nearly
all participants the interaction was primarily limited to providing encouragement during
testing. Encouragement was needed in large part due to the fact the participants completed
extensive PaP as well as web-based battery in the same day. In a minority of participants
assistance with using the computer and/or providing further clarifications to the questions
and tasks that were being asked. In future studies it will be important to further refine the
delivery of the web-based assessments in order to minimize/eliminate the involvement of
research personnel in the evaluation. Exploratory step wise regression analysis identified that
the use of a greatly abbreviated cPACC battery was sufficient to capture the observed validity
between cPACC and ADCS-PACC (Speeded Match, FNHR, Grid Locations). Together these
observations point to the ability to reduce or eliminate participant frustration by using an
abbreviated cPACC and/or minimizing the amount of PaP assessments in future validation
efforts.
Although further validation is needed, one potential use for this platform is to provide an
option for the self-administered collection of assessments and patient demographics in a clini-
cal setting that involves little to no involvement of clinical staff. Additionally, in the current
study use of this web-based platform occurred in some instances in assisted-livings raising the
potential for conducting evaluations outside of traditional clinic setting, including an individu-
al’s home. Both the limited involvement of clinical staff and ability to administer evaluations
outside of the traditional setting are increasingly important aspect of clinical research given
the impacts of Covid-19.
We observed that multiple assessments within the current platform provided valid mea-
sures for diverse aspects of geriatric health. Specifically, we identified the ability of the platform
to capture self-reported patient demographics as well as valid measurements self-rated health,
depression and anxiety symptoms in a sample of community dwelling elderly. The relationship
that was observed between the web-battery measure of depression and anxiety and the GDS/
GAI composite in the current study was similar to correlations found in other studies report-
ing measures of depression and anxiety (e.g., [34–36]). It is important to point out that the
platform therefore not only contains cognitive assessments but also includes other endpoints
that are routinely required as part of cognition focused studies.
There is a widespread and growing use of the ADCS-PACC in clinical trials and longitudi-
nal studies, and therefore there is a need to produce ADCS-PACC assessment options that
don’t require traditional PaP delivery/capture during periods of significant operational and
safety challenges such as Covid-19. We developed the current web-based battery to provide a
mechanism to capture an ADCS-PACC relevant assessment that could be delivered using a
computer-based application in place of a PaP. While our computer-based assessment taps into
cognitive domains relevant to the ADCS-PACC, and significantly correlates with performance
on a PaP version of the ADCS-PACC (moderate significance), we recognize that there are ver-
bal and mechanical limitations in the current computer-based assessment does not allow for a
complete overlap with the individual assessments comprising the ADCS-PACC. Further, the
tests that comprise the computer-based assessment were designed to address similar constructs
to the ADCS-PACC but the format and demands are different even for tests most similar to
PLOS ONE | https://doi.org/10.1371/journal.pone.0244962 January 19, 2021
10 / 15
PLOS ONE Web-based platform for self-administered assessments in the elderly
one another. For example, orientation was asked using multiple choice questions on the
cPACC while the MMSE asks for a verbal response without cues and the digit symbol coding
requires written copy of symbols while the speeded match task involves using a mouse to click
on a response. In this initial validity study, we identified the cPACC to have a statistically sig-
nificant (moderate correlation) with the PaP version of the ADCS-PACC, and to have compa-
rable discrimination to the ADCS-PACC in terms identifying those with and without
cognitive impairment. Interestingly, when the MMSE was removed from the PaP composite
score, the cPACC was better able to distinguish between those with and without cognitive
impairment. To our knowledge, there has only been one previous validation study of com-
puter-based assessments targeting the ADCS-PACC [10]. That study demonstrated that the
computerized batteries had positive correlations with the ADCS-PACC. Therefore, the results
of the current study add to a limited, but growing literature which represents a potentially
important step in moving from a reliance upon PaP versions of the ADCS-PACC for the mea-
surement of an ADRD relevant cognitive composite. In particular, it will be important to
determine in the near future the ability to extend the findings from this initial validation study
to a larger and more diverse study sample that also includes data as to the feasibility of using
the cPACC for measuring the rates of cognitive change over time.
Inherent cPACC features such as the automated assessment delivery and scoring may facili-
tate cognitive composite measures being conducted in a larger number of clinical and research
settings. The cPACC demonstrated good reliability when assessing delayed memory both
through free recall and when given further cuing through multiple choice on the FHNR test.
To our knowledge this is the first study to describe the use of a recall component in a comput-
erized episodic memory test. The FHNR task is based on the Face-Name Association Memory
Test which has been shown to distinguish between cognitive healthy individuals and those
with MCI and correlates with AD biomarkers such as amyloid deposition [26, 28]. Given the
size of practice effects observed for episodic memory measures, a future goal is to develop
alternate forms to reduce practice effects.
In addition to verbal episodic memory, visual episodic memory has shown to decline in a
similar magnitude in individuals at risk for ADRD [37] and visual episodic memory measures
cognitive impairment beyond verbal episodic memory alone [38]. Measures of visual episodic
memory and visual working memory are extremely feasible and conducive for a computer-
based delivery of cognitive assessments and are components of the cPACC [39, 40]. However,
with the exception of a task assessing processing speed, all other tests demonstrated weak to
moderate test-retest correlations in the current study. One possible solution to improve the
test-retest reliability of the cPACC is to add more trials to the visual episodic memory tests.
Despite this, subtests of the cPACC demonstrated strong effect sizes in distinguishing between
those with and without subtle cognitive impairment. Future studies can explore the ability of
the cPACC to identify subtle cognitive impairments in older adults.
Participants in the current study did not demonstrate variability in responses to the orienta-
tion items (only 3 participants in the entire sample did not get both orientation questions cor-
rect). For the PACC, the MMSE is included given it includes items to measure orientation, a
domain identified in the review as important for assessing preclinical AD. However, the
MMSE, is known to have poor psychometric properties (i.e., ceiling effects and low test-retest-
reliability) in healthy, non-demented, older adults [41]. In some circumstances removing the
MMSE has actually been shown to improve the sensitivity of the ADCS-PACC to measure cog-
nitive decline [42]. Taken together, these data highlight the importance of the need to continue
to optimize the psychometric properties of the ADCS-PACC.
A number of studies have identified important roles of working memory in the develop-
ment of MCI and progression to ADRD. Modifications to the ADCS-PACC which add in a
PLOS ONE | https://doi.org/10.1371/journal.pone.0244962 January 19, 2021
11 / 15
PLOS ONE Web-based platform for self-administered assessments in the elderly
measure of verbal fluency, tasks highly linked to working memory, were found to be better
than the original PACC in capturing longitudinal decline [42, 43]. Verbal fluency measures
can be considered as measures of executive functioning, a domain identified as important in
the assessment of preclinical AD [44]. While verbal fluency measures are difficult to incorpo-
rate into a computerized testing setting, visual working memory measures can be readily
implemented in a computer-based assessment and are sensitive to the identification of cogni-
tive decline associated with ADRD [45]. Further validation efforts and implementation of the
cPACC may identify that it has enhanced sensitivity and utility with which to measure and
monitor cognitive change relevant to the development of ADRD.
The focus of the current study was to conduct an initial validation study of the web-battery,
including the cPACC in a non-demented, community dwelling, sample of older study partici-
pants. Of note, only a subset of the much larger web-battery questionnaire was the focus of
psychometric validation and future studies will need to validate the remaining questions. A
limitation of the current study is observed in the study sample being overwhelmingly Cauca-
sian and well-educated which is not representative of the general population raises caution in
extending the findings from this study to a more ethnically and educationally diverse sample.
Validation of the cPACC was based on cross sectional data and caution should be applied in
determining the ability of the cPACC to measure cognitive change in a longitudinal manner
similar to previous studies reported with the ADCS-PACC. Further, in making comparisons
between those with intact global cognition (i.e., MMSE score of 25 or higher) and reduced
global cognition (i.e., MMSE score less than 25), the current study had a small number of par-
ticipants with reduced global cognition. Future studies can continue to examine the utility of
the cPACC to differentiate between those with intact and reduced cognitive performance.
Supporting information
S1 Table. Stepwise regression of cPACC measures and the PaP composite score.
(DOCX)
S2 Table. Correlations among measures on the cPACC and PaP measures.
(DOCX)
Author Contributions
Conceptualization: Matthew Calamia, Alyssa N. De Vito, John P. K. Bernstein, Jeffrey N.
Keller.
Data curation: Matthew Calamia, Alyssa N. De Vito.
Formal analysis: Matthew Calamia, Daniel S. Weitzner, Alyssa N. De Vito, John P. K.
Bernstein.
Investigation: Matthew Calamia, Daniel S. Weitzner, Alyssa N. De Vito, John P. K. Bernstein,
Ray Allen.
Methodology: Matthew Calamia, Alyssa N. De Vito, John P. K. Bernstein, Ray Allen.
Software: Ray Allen.
Supervision: Matthew Calamia, Alyssa N. De Vito.
Validation: Matthew Calamia, Daniel S. Weitzner.
Writing – original draft: Matthew Calamia, Daniel S. Weitzner, Alyssa N. De Vito, John P. K.
Bernstein.
PLOS ONE | https://doi.org/10.1371/journal.pone.0244962 January 19, 2021
12 / 15
PLOS ONE Web-based platform for self-administered assessments in the elderly
Writing – review & editing: Jeffrey N. Keller.
References
1. Wild K, Howieson D, Webbe F, Seelye A, Kaye J. Status of computerized cognitive testing in aging: a
systematic review. Alzheimers Dement. 2008; 4(6):428–37. https://doi.org/10.1016/j.jalz.2008.07.003
PMID: 19012868
2. Katzman R, Saitoh T. Advances in Alzheimer’s disease. The FASEB journal. 1991; 5(3):278–86. PMID:
2001787
3. Salmon DP, Thomas R, Pay M, Booth A, Hofstetter C, Thal L, et al. Alzheimer’s disease can be accu-
rately diagnosed in very mildly impaired individuals. Neurology. 2002; 59(7):1022–8. https://doi.org/10.
1212/wnl.59.7.1022 PMID: 12370456
4. Donohue MC, Sperling RA, Salmon DP, Rentz DM, Raman R, Thomas RG, et al. The preclinical Alzhei-
mer cognitive composite: measuring amyloid-related decline. JAMA neurology. 2014; 71(8):961–70.
https://doi.org/10.1001/jamaneurol.2014.803 PMID: 24886908
5. Sperling RA, Rentz DM, Johnson KA, Karlawish J, Donohue M, Salmon DP, et al. The A4 study: stop-
ping AD before symptoms begin? Science translational medicine. 2014; 6(228):228fs13–fs13. https://
doi.org/10.1126/scitranslmed.3007941 PMID: 24648338
6. Mueller SG, Weiner MW, Thal LJ, Petersen RC, Jack CR, Jagust W, et al. Ways toward an early diag-
nosis in Alzheimer’s disease: the Alzheimer’s Disease Neuroimaging Initiative (ADNI). Alzheimers
Dement. 2005; 1(1):55–66. https://doi.org/10.1016/j.jalz.2005.06.003 PMID: 17476317
7. Gates NJ, Kochan NA. Computerized and on-line neuropsychological testing for late-life cognition and
neurocognitive disorders: are we there yet?. Curr Opin Psychiatry. 2015; 28(2):165–172. https://doi.
org/10.1097/YCO.0000000000000141 PMID: 25602241
8.
Feenstra HEM, Murre JMJ, Vermeulen IE, Kieffer JM, Schagen SB. Reliability and validity of a self-
administered tool for online neuropsychological testing: The Amsterdam Cognition Scan. J Clin Exp
Neuropsychol. 2018; 40(3):253–273. https://doi.org/10.1080/13803395.2017.1339017 PMID:
28671504
9. Hansen TI, Haferstrom EC, Brunner JF, Lehn H, Håberg AK. Initial validation of a web-based self-
administered neuropsychological test battery for older adults and seniors. J Clin Exp Neuropsychol.
2015; 37(6):581–594. https://doi.org/10.1080/13803395.2015.1038220 PMID: 26009791
10. Buckley RF, Sparks KP, Papp KV, et al. Computerized Cognitive Testing for Use in Clinical Trials: A
Comparison of the NIH Toolbox and Cogstate C3 Batteries. J Prev Alzheimers Dis. 2017; 4(1):3–11.
https://doi.org/10.14283/jpad.2017.1 PMID: 29188853
11. Perrin S, Buckley RF, et al. Unsupervised assessment of cognition in the Healthy Brain Project: Implica-
tions for web-based registries of individuals at risk for Alzheimer’s disease. Alzheimers & Dementia:
Translational Research & Clinical Interventions. https://doi.org/10.1002/trc2.12043
12.
Lim YY, Yassi N, Bransby L, Properzi M, Buckley R. The Healthy Brain Project: An Online Platform for
the Recruitment, Assessment, and Monitoring of Middle-Aged Adults at Risk of Developing Alzheimer’s
Disease. J Alzheimer’s Dis. 2019; 68(3):1211–1228.
13.
Janssen B. EQ-5D User Guide 2015 [Available from: https://euroqol.org/publications/user-guides/.
14. Halaweh H, Willen C, Grimby-Ekman A, Svantesson U. Physical activity and health-related quality of
life among community dwelling elderly. Journal of clinical medicine research. 2015; 7(11):845. https://
doi.org/10.14740/jocmr2307w PMID: 26491496
15. Ko¨ nig H-H, Bernert S, Angermeyer MC, Matschinger H, Martinez M, Vilagut G, et al. Comparison of
population health status in six european countries: results of a representative survey using the EQ-5D
questionnaire. Medical care. 2009:255–61. https://doi.org/10.1097/MLR.0b013e318184759e PMID:
19169128
16. Pachana NA, Byrne GJ, Siddle H, Koloski N, Harley E, Arnold E. Development and validation of the
Geriatric Anxiety Inventory. Int Psychogeriatr. 2007; 19(1):103–14. https://doi.org/10.1017/
S1041610206003504 PMID: 16805925
17. Diefenbach GJ, Tolin DF, Meunier SA, Gilliam CM. Assessment of anxiety in older home care recipi-
ents. The Gerontologist. 2009; 49(2):141–53. https://doi.org/10.1093/geront/gnp019 PMID: 19363010
18. Yesavage JA, Brink TL, Rose TL, Lum O, Huang V, Adey M, et al. Development and validation of a geri-
atric depression screening scale: a preliminary report. J Psychiatr Res. 1982; 17(1):37–49. https://doi.
org/10.1016/0022-3956(82)90033-4 PMID: 7183759
19. Peach J, Koob JJ, Kraus MJ. Psychometric Evaluation of the Geriatric Depression Scale (GDS) Sup-
porting Its Use in Health Care Settings. Clinical Gerontologist. 2001; 23(3–4):57–68.
PLOS ONE | https://doi.org/10.1371/journal.pone.0244962 January 19, 2021
13 / 15
PLOS ONE Web-based platform for self-administered assessments in the elderly
20. Kiernan BU, Wilson D, Suter N, Naqvi A. Comparison of the Geriatric Depression Scale and Beck
Depression Inventory in a nursing home setting. Clinical Gerontologist: The Journal of Aging and Mental
Health. 1986.
21. Ribeiro O, Pau´ l C, Simões MR, Firmino H. Portuguese version of the Geriatric Anxiety Inventory: Trans-
cultural adaptation and psychometric validation. Aging & Mental Health. 2011; 15(6):742–8. https://doi.
org/10.1080/13607863.2011.562177 PMID: 21656405
22. Boot WR, Charness N, Czaja SJ, Sharit J, Rogers WA, Fisk AD, et al. Computer proficiency question-
naire: assessing low and high computer proficient seniors. The Gerontologist. 2015; 55(3):404–11.
https://doi.org/10.1093/geront/gnt117 PMID: 24107443
23. Bransby L, Lim YY, Ames D, Fowler C, Roberston J, Harrington K, et al. Sensitivity of a Preclinical Alz-
heimer’s Cognitive Composite (PACC) to amyloid β load in preclinical Alzheimer’s disease. J Clin Exp
Neuropsychol. 2019; 41(6):591–600. https://doi.org/10.1080/13803395.2019.1593949 PMID:
30924399
24. Mormino EC, Papp KV, Rentz DM, Donohue MC, Amariglio R, Quiroz YT, et al. Early and late change
on the preclinical Alzheimer’s cognitive composite in clinically normal older individuals with elevated
amyloid β. Alzheimers Dement. 2017; 13(9):1004–12. https://doi.org/10.1016/j.jalz.2017.01.018 PMID:
28253478
25. McCabe DP, Roediger HL III, McDaniel MA, Balota DA, Hambrick DZ. The relationship between work-
ing memory capacity and executive functioning: evidence for a common executive attention construct.
Neuropsychology. 2010; 24(2):222. https://doi.org/10.1037/a0017619 PMID: 20230116
26. Papp KV, Amariglio RE, Dekhtyar M, Roy K, Wigman S, Bamfo R, et al. Development of a psychometri-
cally equivalent short form of the face–name associative memory exam for use along the early alzhei-
mer’s disease trajectory. Clin Neuropsychol. 2014; 28(5):771–85. https://doi.org/10.1080/13854046.
2014.911351 PMID: 24815535
27. Greenaway MC, Lacritz LH, Binegar D, Weiner MF, Lipton A, Cullum CM. Patterns of verbal memory
performance in mild cognitive impairment, Alzheimer disease, and normal aging. Cogn Behav Neurol.
2006; 19(2):79–84. https://doi.org/10.1097/01.wnn.0000208290.57370.a3 PMID: 16783130
28. Rubiño J, Andre´ s P. The face-name associative memory test as a tool for early diagnosis of alzheimer’s
disease. Frontiers in psychology. 2018; 9:1464. https://doi.org/10.3389/fpsyg.2018.01464 PMID:
30154753
29. Malec JF, Ivnik RJ, Hinkeldey NS. Visual spatial learning test. Psychological Assessment: A Journal of
Consulting and Clinical Psychology. 1991; 3(1):82.
30. Smith GE, Malec JF, Ivnik RJ. Validity of the construct of nonverbal memory: A factor-analytic study in a
normal elderly sample. J Clin Exp Neuropsychol. 1992; 14(2):211–21. https://doi.org/10.1080/
01688639208402824 PMID: 1572945
31. Wechsler D. Welcher adult intelligence scale-revised. New York: Psychological Corporation; 1981.
32. Wechsler D. WMS-IV Administration and Scoring Manual. San Antonio, Texas: Pearson; 2009.
33. Wechsler D. WMS-III: Wechsler Memory Scale Administration and Scoring Manual. San Antonio,
Texas: Psychological Corp.; 1997.
34. Weitzner D, Calamia M, Stasik-O’Brien SM, De Vito A, Pugh E. Psychometric properties of the
expanded version of the inventory of depression and anxiety symptoms (IDAS-II) in a sample of older
adults. Aging Ment Health. 2019:1–7.
35. Segal DL, June A, Payne M, Coolidge FL, Yochim B. Development and initial validation of a self-report
assessment tool for anxiety among older adults: the Geriatric Anxiety Scale. J Anxiety Disord. 2010; 24
(7):709–14. https://doi.org/10.1016/j.janxdis.2010.05.002 PMID: 20558032
36.
37.
Johnco C, Knight A, Tadic D, Wuthrich VM. Psychometric properties of the Geriatric Anxiety Inventory
(GAI) and its short-form (GAI-SF) in a clinical and non-clinical sample of older adults. Int Psychogeriatr.
2015; 27(7):1089–97. https://doi.org/10.1017/S1041610214001586 PMID: 25111285
Lim YY, Maruff P, Pietrzak RH, Ames D, Ellis KA, Harrington K, et al. Effect of amyloid on memory and
non-memory decline from preclinical to clinical Alzheimer’s disease. Brain. 2014; 137(1):221–31.
https://doi.org/10.1093/brain/awt286 PMID: 24176981
38. Alladi S, Arnold R, Mitchell J, Nestor PJ, Hodges JR. Mild cognitive impairment: applicability of research
criteria in a memory clinic and characterization of cognitive profile. Psychological Medicine. 2006; 36
(4):507–15. https://doi.org/10.1017/S0033291705006744 PMID: 16426486
39. Ba¨ ckman L, Jones S, Berger A-K, Laukka EJ, Small BJ. Cognitive impairment in preclinical Alzheimer’s
disease: a meta-analysis. Neuropsychology. 2005; 19(4):520. https://doi.org/10.1037/0894-4105.19.4.
520 PMID: 16060827
PLOS ONE | https://doi.org/10.1371/journal.pone.0244962 January 19, 2021
14 / 15
PLOS ONE Web-based platform for self-administered assessments in the elderly
40. Han SD, Nguyen CP, Stricker NH, Nation DA. Detectable neuropsychological differences in early pre-
clinical Alzheimer’s disease: A meta-analysis. Neuropsychol Rev. 2017; 27(4):305–25. https://doi.org/
10.1007/s11065-017-9345-5 PMID: 28497179
41. Spencer RJ, Wendell CR, Giggey PP, Katzel LI, Lefkowitz DM, Siegel EL, et al. Psychometric limitations
of the mini-mental state examination among nondemented older adults: an evaluation of neurocognitive
and magnetic resonance imaging correlates. Exp Aging Res. 2013; 39(4):382–97. https://doi.org/10.
1080/0361073X.2013.808109 PMID: 23875837
42.
Lim YY, Snyder PJ, Pietrzak RH, Ukiqi A, Villemagne VL, Ames D, et al. Sensitivity of composite scores
to amyloid burden in preclinical Alzheimer’s disease: Introducing the Z-scores of Attention, Verbal flu-
ency, and Episodic memory for Nondemented older adults composite score. Alzheimer’s & Dementia:
Diagnosis, Assessment & Disease Monitoring. 2016; 2:19–26.
43. Papp KV, Rentz DM, Orlovsky I, Sperling RA, Mormino EC. Optimizing the preclinical Alzheimer’s cog-
nitive composite with semantic processing: The PACC5. Alzheimer’s & Dementia: Translational
Research & Clinical Interventions. 2017; 3(4):668–77. https://doi.org/10.1016/j.trci.2017.10.004 PMID:
29264389
44. Henry JD, Crawford JR, Phillips LH. Verbal fluency performance in dementia of the Alzheimer’s type: a
meta-analysis. Neuropsychologia. 2004; 42(9):1212–22. https://doi.org/10.1016/j.neuropsychologia.
2004.02.001 PMID: 15178173
45. Summers MJ, Saunders NL. Neuropsychological measures predict decline to Alzheimer’s dementia
from mild cognitive impairment. Neuropsychology. 2012; 26(4):498. https://doi.org/10.1037/a0028576
PMID: 22612573
PLOS ONE | https://doi.org/10.1371/journal.pone.0244962 January 19, 2021
15 / 15
PLOS ONE |
10.1371_journal.pone.0251177 | RESEARCH ARTICLE
National estimates from the Youth ’19
Rangatahi smart survey: A survey calibration
approach
C. Rivera-RodriguezID
S. Lewycka6,7
1*, T. C. Clark2, T. Fleming3, D. Archer3, S. Crengle4, R. Peiris-John5,
1 Department of Statistics, The University of Auckland, Auckland, New Zealand, 2 School of Nursing,
University of Auckland, Auckland, New Zealand, 3 School of Health, Victoria University of Wellington,
Wellington, New Zealand, 4 Department of Preventive and Social Medicine, University of Otago, Dunedin,
New Zealand, 5 Department of Epidemiology and Biostatistics, The University of Auckland, Auckland, New
Zealand, 6 Nuffield Department of Medicine, University of Oxford, Oxford, United Kingdom, 7 Oxford
University Clinical Research Unit, Ho Chi Minh City, Vietnam
* c.rodriguez@auckland.ac.nz, clriverarodriguez@gmail.com
Abstract
Background
Significant progress has been made addressing adolescent health needs in New Zealand,
but monitoring and gathering high quality estimates of adolescent health and social issues
remains challenging and resource intensive. Previous nationally representative secondary
school surveys were conducted in New Zealand in 2001, 2007 and 2012, as part of the
Youth2000 survey series. This paper focuses on a fourth survey conducted in 2019 (https://
www.youth19.ac.nz/). The 2019 survey had a regional sampling strategy rather than a
national sampling strategy as in previous years. The survey also included kura kaupapa
Māori schools (Māori language immersion schools), as well as mainstream secondary
schools. This paper presents the overall study methodology, and a weighting and calibration
framework in order to provide estimates that reflect the national student population, and
enable comparisons with the previous surveys to monitor trends.
Methods
Youth19 was a cross sectional, self-administered health and wellbeing survey of New Zea-
land high school students. The survey population was secondary school students of New
Zealand aged 12 to 18 years (school years 9–13). The study population was drawn from
three education regions: Auckland, Tai Tokerau (Northland) and Waikato. These are the
most ethnically diverse regions in New Zealand and account for 46% of the adolescent pop-
ulation in New Zealand. The sampling design was two-stage clustered stratified, where
schools were the clusters, and strata were defined by kura schools and educational regions.
There were four strata, formed as follows: kura schools (Tai Tokerau, Auckland and Waikato
regions combined), mainstream-Auckland, mainstream-Tai Tokerau and mainstream-Wai-
kato. From each stratum, 50% of the schools were randomly sampled and then 30% of stu-
dents from the selected schools were invited to participate. All students in the kura kaupapa
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: Rivera-Rodriguez C, Clark TC, Fleming T,
Archer D, Crengle S, Peiris-John R, et al. (2021)
National estimates from the Youth ’19 Rangatahi
smart survey: A survey calibration approach. PLoS
ONE 16(5): e0251177. https://doi.org/10.1371/
journal.pone.0251177
Editor: Andrew R. Dalby, University of
Westminster, UNITED KINGDOM
Received: September 23, 2020
Accepted: April 21, 2021
Published: May 14, 2021
Copyright: © 2021 Rivera-Rodriguez et al. This is
an open access article distributed under the terms
of the Creative Commons Attribution License,
which permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: The data underlying
this study are owned by Youth19 (https://www.
youth19.ac.nz/) and are available upon request.
Interested and qualified researchers may submit
data access requests to https://www.youth19.ac.
nz/contact. The authors are members of the
Youth2019 research group. However, other
researchers would be able to access the same data
on request.
Funding: TC - HRC (Health Research Council New
Zealand) -grant number 17/315 https://www.hrc.
PLOS ONE | https://doi.org/10.1371/journal.pone.0251177 May 14, 2021
1 / 16
PLOS ONE govt.nz/ The funders had no role in study design,
data collection and analysis, decision to publish, or
preparation of the manuscript. TF - HRC (Health
Research Council New Zealand) -grant number 18/
473 https://www.hrc.govt.nz/ The funders had no
role in study design, data collection and analysis,
decision to publish, or preparation of the
manuscript.
Competing interests: The authors have declared
that no competing interests exist.
National estimates from the Youth ’19 Rangatahi smart survey
schools were invited to participate. In order to make more precise estimates and adjust for
differential non-response, as well as to make nationally relevant estimates and allow com-
parisons with the previous national surveys, we calibrated the sampling weights to reflect
the national secondary school student population.
Results
There were 45 mainstream and 4 kura schools included in the final sample, and 7,374 main-
stream and 347 kura students participated in the survey. There were differences between
the sampled population and the national secondary school student population, particularly in
terms of sex and ethnicity, with a higher proportion of females and Asian students in the
study sample than in the national student population. We calculated estimates of the totals
and proportions for key variables that describe risk and protective factors or health and well-
being factors. Rates of risk-taking behaviours were lower in the sampled population than
what would be expected nationally, based on the demographic profile of the national student
population. For the regional estimates, calibrated weights yield standard errors lower than
those obtained with the unadjusted sampling weights. This leads to significantly narrower
confidence intervals for all the variables in the analysis. The calibrated estimates of national
quantities provide similar results. Additionally, the national estimates for 2019 serve as a
tool to compare to previous surveys, where the sampling population was national.
Conclusions
One of the main goals of this paper is to improve the estimates at the regional level using
calibrated weights to adjust for oversampling of some groups, or non-response bias. Addi-
tionally, we also recommend the use of calibrated estimators as they provide nationally
adjusted estimates, which allow inferences about the whole adolescent population of New
Zealand. They also yield confidence intervals that are significantly narrower than those
obtained using the original sampling weights.
1. Background
High quality population-based data that provides estimates of adolescent behaviours are essen-
tial for the planning of services, programmes, policy and for monitoring equitable outcomes.
However undertaking such surveys are expensive, complex and resource-intensive. Significant
progress has been made addressing adolescent health needs in New Zealand and globally since
the turn of the century with reductions in morbidity and mortality [1, 2], and increased data
surveillance monitoring of adolescent health trends. However, some areas, such as mental
health issues remain a concern [3], alongside new important areas have emerged that impact
adolescent wellbeing, such as vaping and social media use [4]. Monitoring and tracking trends
in adolescent health are vital, particularly for Indigenous, ethnic and sexual minority youth,
those with disabilities and from poor neighbourhoods [5].
To investigate the health and wellbeing of young New Zealanders, as part of the Youth2000
survey series, nationally representative secondary school surveys were conducted in New Zea-
land in 2001, 2007 and 2012, and 2019 [1,2]. These surveys provided an opportunity to assess
the situation at each time point, and monitor trends in key indicators of health and wellbeing.
PLOS ONE | https://doi.org/10.1371/journal.pone.0251177 May 14, 2021
2 / 16
PLOS ONE National estimates from the Youth ’19 Rangatahi smart survey
These surveys randomly sampled secondary schools across New Zealand, and from each
school that consented to take part, a random sample of around 8,500 year 9–13 students were
selected to participate. More recent estimates are required in order to monitor progress and
identify areas that need further attention. In 2019 (https://www.youth19.ac.nz/), the schools
sampled only included three regions (Waikato, Auckland and Tai Tokerau/Northland), rather
than from the whole country, due to loss of Government contract. Alternative funding was
sought, and due to logistical and budgetary constraints a pragmatic decision to survey a
smaller proportion of students and regions was made.
In this paper we present the overall study methodology, and how we have utilized a weight-
ing and calibration framework that can provide estimates that reflect the national student pop-
ulation, ensure that ethnic groups, particularly Māori are adequately represented and enable
comparisons with the estimates from previous surveys.
2. Methods
2.1. Study design
Youth19 was a cross-sectional, self-administered health and wellbeing survey of New Zealand
secondary school students. Full details of the methods have been published elsewhere [6]. The
study had the following aims:
1. To collect, analyse and disseminate accurate, comprehensive and timely information on the
health and wellbeing of young people living in Tai Tokerau, Auckland and Waikato
Regions, in order to inform and improve policies and practices;
2. To evaluate how whanaungatanga influences health outcomes for rangatahi Māori;
3. To test the potential benefits of incorporating opt-in access to links for support services
within a survey.
2.2. Target and study populations
New Zealand secondary school students (aged 13–18 years, school years 9–13) were surveyed
across three regions: Auckland, Tai Tokerau/Northland and Waikato. Almost half the New
Zealand youth population resides in these areas (46%), these are the most ethnically diverse
regions in New Zealand and include a range of urban and rural settings as well as a breadth of
socio-economic groupings. These three regions were chosen to represent the diversity of the
New Zealand population, and to ensure that the number of participants from each of the main
ethnic groups provided sufficient statistical power for sub-group analyses. Previous popula-
tion-based studies have used these three regions and found them to be representative of
national statistics [7].
2.3. Sampling design
We used the Education Counts 2017 national list of schools as our sampling frame [8], and
excluded schools from regions other than Auckland, Tai Tokerau and Waikato. We used a
two-stage cluster sampling design. We included single sex, co-education, public, private and
fully integrated schools that had over 50 students in years 9–13. As in the previous three sur-
veys, schools with under 50 students were excluded for logistical reasons, hence the conclu-
sions presented here are only for students attending schools with over 50 students. Special
schools that only included students who had intellectual or physical disabilities which would
have prevented them from being able to participate in the survey where excluded. We stratified
PLOS ONE | https://doi.org/10.1371/journal.pone.0251177 May 14, 2021
3 / 16
PLOS ONE National estimates from the Youth ’19 Rangatahi smart survey
our sample by kura schools and educational regions. There were four strata, formed as follows:
kura schools (Tai Tokerau, Auckland and Waikato regions), mainstream-Auckland, main-
stream-Tai Tokerau and mainstream-Waikato.
There were 161 eligible mainstream schools (100 in Auckland, 23 in Tai Tokerau and 38 in
Waikato). From each stratum, 50% of the schools were randomly sampled using a random
number generator. All selected schools were invited to participate through email and follow up
phone calls. We piloted the survey in two additional schools from the same sampling frame in
Auckland in 2019, these two schools were purposively selected. These were large ethnically
and socio-economically diverse schools. Minimal changes were made to the survey after pilot-
ing, and these schools have been included in the total. We visited the other schools that agreed
to participate between May and September 2019. We randomly sampled 30% of students on
the school roll to be invited to participate in the study. One mainstream school also requested
100% of students be invited and this was done.
There were 8 eligible kura kaupapa Māori schools in the three study regions, and two from
each region (6 in total) were invited to participate. These schools are smaller than mainstream
schools and include immersion in Māori language and culture. Four schools participated and
all students in these kura kaupapa schools were invited to participate.
We calculated sample weights as inverse probability weights using the sampling design
described above. This design is described in detail in Table 1 and Fig 1.
2.4. Data collection
The survey was refined from previous Youth2000 series questions (https://www.fmhs.
auckland.ac.nz/en/faculty/adolescent-health-research-group/publications-and-reports.html),
validated measures and measures used in other surveys, as well as new questions developed
from a rangatahi and Māori whanau photovoice and qualitative research process, and a digi-
tally integrated survey process. Students completed the web-based survey on tablets in English
or Te Reo Māori (the language of New Zealand’s indigenous people). Questions appeared in
text on the screen and were available via voice-over through headphones [6].
2.5. Regional estimates
Most estimates for this study are based on totals, means or proportions. To help simplify the
exposition, we present the methods in the context of estimating totals. This is applicable to
means and proportions since they are functions of totals. Initially, we have a population of size
Table 1.
Not eligible
schools
Eligible
schools
Invited
schools
Participating
schools
Students in eligible
schools
Students in invited
schools
Students in
participating schools
Students
surveyed
Total
By education region
Tai Tokerau
Auckland
Waikato
Other
regions
By school type
Mainstream
Kura
2362
462
126
238
1536
2275
87
169
25
102
42
0
161
8
https://doi.org/10.1371/journal.pone.0251177.t001
86
10
53
23
0
80
6
49
10
27
12
0
45
4
130692
97415
24074
9203
0
129765
927
71105
52013
14693
4399
0
70408
697
PLOS ONE | https://doi.org/10.1371/journal.pone.0251177 May 14, 2021
42298
31028
6871
4399
0
41828
470
7721
5545
1248
928
0
7374
347
4 / 16
PLOS ONE National estimates from the Youth ’19 Rangatahi smart survey
Fig 1. Sample design.
https://doi.org/10.1371/journal.pone.0251177.g001
PLOS ONE | https://doi.org/10.1371/journal.pone.0251177 May 14, 2021
5 / 16
PLOS ONE National estimates from the Youth ’19 Rangatahi smart survey
PN
N and we are interested in estimating the total of a variable of interest, called y, which can be
written as Ty ¼
i¼1 yi. In the absence of complete data from all the population, Ty cannot be
calculated. Consequently, this should be estimated. Since the sampling design is a stratified-
multistage design, the estimator has to account for this design through weights [9–12]. The
weighted estimator of Ty is
X
^T y ¼
wiyi;
sample
where wi = 1/πi, and πi is the sampling probability for individual i. The weight wi can be inter-
preted as the number of people that individual i represents in the population. This type of esti-
mator and its variances are available from the survey package in R.
2.6. Missing observations and extrapolation to the national population
The weighted estimator presented above accounts for the sampling scheme, but it has several
drawbacks. First, it is unbiased, but only when there is not missing information. Second, it is
known to be inefficient because it yields wider confidence intervals than other estimators of
totals [13]. An approach to attaining more efficient estimators is to use auxiliary information
available for the entire population (e.g. information from the sampling frame). For instance,
for the Youth 2019 surveys, an option would be to use information on the ethnic distribution
of students in the population. This information was not used to inform the sampling design,
but we can use it post design to improve the estimators [14, 15].
Calibration is among these methods, it has been used in the literature when sampling
weights are incorrect, to correct for non-response or to extrapolate to wider populations where
there is compelling evidence that the factors contributing to the estimators are very similar in
the target population and in the wider population [10, 16, 17]. The primary idea of calibration
is to adjust the sampling weights wi such that totals of known quantities are exactly estimated.
To see this, let M denote the total number of Māori students in the population of interest.
From the sampling frame, we know that this number is 24983 for the three regions in the
study, and 59040 for the whole country. Although M is known, it is interesting to investigate
what would be the estimator of M using only the survey data. That is ^M ¼
samplewili, where li
is a binary variable denoting if the individual is Māori or not. Since M is actually known, one
could always modify the sampling weights such that ^M ¼ M. The new weights ( ~wi) are found
by minimizing a distance function between the original sampling weights and the modified
weights subject to the constraint ^M ¼ M. The new weights are known as calibrated weights
and the estimator is denoted ~M ¼
sample ~wili. In theory, the variance of ~M will never be larger
than the variance of ^M, which is based on the original weights. This calibration process can be
done using several variables simultaneously. For example, the weights can be calibrated to
demographic factors that are considered important in the analysis, and are available both for
the sampling frame and the study population. Calibration can be implemented via the survey
package in R with the function calibrate() [18, 19].
P
P
2.7. Calibrated estimates: Regional and national
We use calibrated weights at the regional level (Regions: Auckland, Tai Tokerau and Waikato)
in order to improve the efficiency of our estimates, and adjust for differential non-response. In
our case, we calibrate the regional weights to Regional totals of the demographic variables
available from Education Counts: kura kaupapa Māori, School Deciles, Age, Gender and Eth-
nicity. The deciles are a measure of the socio-economic position of a school’s student
PLOS ONE | https://doi.org/10.1371/journal.pone.0251177 May 14, 2021
6 / 16
PLOS ONE National estimates from the Youth ’19 Rangatahi smart survey
community relative to other schools throughout the country. For example, decile 1 schools are
the 10% of schools with the highest proportion of students from low socio-economic commu-
nities, whereas decile 10 schools are the 10% of schools with the lowest proportion of these stu-
dents. A school’s decile does not indicate the overall socio-economic mix of the school or
reflect the quality of education the school provides. Deciles are used to provide funding to
state and state-integrated schools to enable them to overcome the barriers to learning faced by
students from lower socio-economic communities. The lower the school’s decile, the more
funding they receive [20].The majority of our outcome variables show a significant relation-
ship to at least one of these demographic variables, this can be seen in the descriptive plots in
S1 Statistics. Calibration invokes no assumptions apart from the study sample being a sample
selected using a probabilistic design from a population of interest [16]. In our case, this
assumption holds for regional estimates. However, for national estimates, the population of
interest (national) is different to the population from where the sample was selected.
Our main goal is to generate national statistics that enable us to compare the results to pre-
vious national surveys. In order to do this, we have to assume that the regional sample is
selected from the national population. This means that the distributions of factors contributing
to the estimators are very similar in the Regional population (Regions: Auckland, Tai Tokerau
and Waikato) and in the national population. In order to account for the demographic distri-
bution of the national population, we calibrate these weights to the National totals of the same
demographic variables used for the regional weights (kura kaupapa Māori, School Deciles,
Age, Gender and Ethnicity). This calibration was done using the calibrate() function from the
R package survey. The totals used for calibration are education counts available from https://
www.education.govt.nz/our-work/contact-us/. In order to understand how different weights
affect the estimation of outcomes of interest, we compared results for key health and well-
being indicators (Tables 4 and 5).
2.8. Ethics
In each participating mainstream school, the principal or head of the board of trustees pro-
vided consent for the students to be invited to participate. Information for parents in English
and Te Reo Māori was provided to the school (digitally and or printed) and made available to
parents and caregivers who could opt to have their child excluded from the survey. Ethics
approval was granted by the University of Auckland Human Subjects Ethics Committee
(application #022244).
3. Results
3.1. Study participants
There were 2,531 schools nationally, and 624 in the Auckland, Tai Tokerau and Waikato
regions. We excluded 819 schools from these regions because they had less than 50 year 9–13
students, and five because they were partnership schools. A further 2 were excluded mistakenly
due to human error. This left 161 eligible mainstream schools in the three regions. Two large
ethnically diverse schools were purposively selected for piloting, and 78 schools were randomly
selected, making 80 (49.7%) in total. Of these, 45 (56.3%) agreed to participate. There were
41,828, students at participating mainstream schools, and 7,374 (59.7%) participated. The sam-
pling design is shown in Fig 1.
There were 95 kura kaupapa Māori nationally, and 8 in the Auckland, Tai Tokerau and
Waikato regions. Six were invited, and 4 (66.7%) agreed to participate. There were 470 stu-
dents at participating kura, all of whom were invited to participate, and 347 (71%) participated.
The rest of the results are presented for mainstream schools only.
PLOS ONE | https://doi.org/10.1371/journal.pone.0251177 May 14, 2021
7 / 16
PLOS ONE National estimates from the Youth ’19 Rangatahi smart survey
The unweighted characteristics of participating mainstream schools and students are
shown in Table 2, alongside comparable data for the previous national surveys, and all second-
ary school students in New Zealand. The participation rates were much lower than previous
surveys, both for schools (56.3%) and for students (59.7%). In 16 schools, participation was
under 50%, with a measles outbreak, teacher strikes, and high truancy rates indicated by school
staff as likely to have affected response rates in their schools. Apart from this, illness, assess-
ments and field-trips may have resulted in students being unable to participate. The majority
of non-participating students did not arrive at the room in which the survey was taking place,
and only 49 arrived at the room but declined to participate.
There are some important differences in the demographic characteristics of participating
students compared to previous surveys, and to the national secondary school population distri-
bution. There were a lower proportion of high decile schools included in the sample, but
higher participation at high decile schools means that the 2019 sample population matches the
national student population quite well in terms of school decile. There was a lower proportion
of boys (45.1%) compared to the national student population, and to the surveys in 2001 and
2012, and a slightly higher proportion of students aged 17 and above than in previous surveys,
though this is still lower than the national student population. There were ethnic differences
too, with a much higher proportion of Asian students than previous surveys, and than the
national student population, reflecting a higher proportion of the Asian population living in
the Auckland region.
3.2. Estimates
Tables 3 and 4 display the actual regional and national totals and proportions for variables
used to calibrate the sampling weights. This excludes kura kaupapa Māori schools because pre-
vious waves did not include such schools. We can observe in the confidence intervals that the
variance yielded by calibrated weights is zero for these variables. This is due to the fact that we
are calibrating to the actual totals, therefore the calibrated estimates should be exactly the same
as the actual totals and in consequence there is no uncertainty (or variance). Fig 2 shows the
distribution of calibrated and sampling weights. There is a significant shift (right skewed) in
the distribution of calibrated weights. This happens because a large number of individuals are
overrepresented by the original sampling weights. Thus, calibration decreases the magnitude
of the weights of those individuals that are overrepresented, and increases the weights of indi-
viduals that are underrepresented.
Table 5 shows the estimates of the total student numbers for key health and wellbeing indi-
cators, and Table 6 shows the estimated proportions. For the regional estimates, calibrated
weights yield standard errors lower than those obtained with the unadjusted sampling weights.
This leads to significantly narrower confidence intervals for all the variables in the analysis.
We only present calibrated estimates of national quantities because the sampling design was a
regional design and therefore, we do not have national level sampling weights. However,
national calibration provides a tool to compare to previous surveys, where the sampling popu-
lation was national.
4. Discussion
We have conducted a multistage cluster sample survey of New Zealand secondary school stu-
dents from three regions, to build on three previous national surveys. We used calibrated
inverse probability weighting in order to correct for demographic differences between the
regional and national student populations and for non-response, which enables extrapolation
of the results from the Youth 2019 survey to the whole secondary school population of New
PLOS ONE | https://doi.org/10.1371/journal.pone.0251177 May 14, 2021
8 / 16
PLOS ONE Table 2. Unweighted characteristics of participating mainstream schools and students from previous surveys and national school data prior to the studies.
National estimates from the Youth ’19 Rangatahi smart survey
Previous national surveys
2001
n (%)
389
133
2007
n (%)
389
115
2012
n (%)
397
125
Current survey in 3 regions
2019�(Excluding kura kaupapa Māori)
n (%)
161
80
Schools
Eligible1
Invited
Participated
114 (85.7)
96 (83.5)
91 (72.8)
45 (56.3)
Decile
Low
Medium
High
Unknown
Students
Eligible
Participated2
Year
Year 9
Year 10
Year 11
Year 12
Year 13
Sex
Female
Male
Age
�13
14
15
16
�17
Ethnicity3
European
Māori
Pacific
Asian
Other
NZDep4
Low deprivation
Medium
High deprivation
NA
Decile
Low
Medium
High
23 (20.2)
51 (44.7)
40 (35.1)
0 (0)
15 (15.6)
52 (54.2)
29 (30.2)
0 (0)
26 (28.6)
36 (39.6)
29 (31.9)
0 (0)
13 (28.9)
21 (46.7)
10 (22.2)
1 (2.2)
41,828
9,567 (74.0)
9,107 (74.0)
8,500 (68.0)
7,374 (59.7)
2,457 (26.1)
2,233 (23.7)
2,156 (22.9)
1,580 (16.9)
978 (10.4)
4,414 (46.1)
5,152 (53.9)
2,050 (21.5)
2,285 (23.9)
2,178 (22.8)
1,725 (18.1)
1,308 (13.7)
5,406 (57.4)
2,340 (24.8)
768 (8.2)
679 (7.2)
230 (2.4)
-
-
-
2,176 (24.3)
2,090 (23.4)
1,933 (21.6)
1,669 (18.7)
1,077 (12.0)
4,911 (54.0)
4,187 (46.0)
1,860 (20.4)
2,101 (23.1)
1,973 (21.7)
1,743 (19.2)
1,423 (15.6)
4,797 (52.8)
1,702 (18.7)
924 (10.2)
1,126 (12.4)
531 (5.8)
3,218 (36.3)
3,397 (38.3)
2,250 (25.4)
2,061 (24.3)
1,936 (22.8)
1,727 (20.4)
1,534 (18.1)
1,227 (14.5)
3,874 (45.6)
4,623 (54.4)
1,838 (21.7)
1,896 (22.3)
1,755 (20.7)
1,578 (18.6)
1,422 (16.8)
4,024 (47.4)
1,705 (20.1)
1,201 (14.1)
1,051 (12.4)
511 (6.0)
2,718 (32.4)
3,001 (35.8)
2,674 (31.9)
1,732 (18.1)
4,393 (45.9)
3,445 (36.0)
3,218 (35.3)
3,397 (37.3)
2,250 (24.7)
1,793 (21.1)
3,296 (38.8)
3,411 (40.1)
1681 (23.1)
1609 (22.1)
1603 (22.1)
1364 (18.8)
1009 (13.9)
3990 (54.6)
3321(45.4)
1338 (18.1)
1650 (22.4)
1631 (22.1)
1418 (19.2)
1337 (18.1)
3067 (41.7)
1201(16.3)
936 (12.7)
1776 (24.1)
381 (5.2)
2,105 (28.6)
2,783 (37.7)
1,845 (25.0)
641 (8.7)
1,203 (16.3)
3,242 (44.0)
2,887 (39.2)
National data
2018
n (%)
407
-
-
98 (24.1)
177 (43.5)
125 (30.7)
7 (1.7)
280,163
-
57,784 (20.8)
57,302 (20.6)
58,952 (21.2)
56,070 (20.1)
48,161 (17.3)
140,862 (50.3)
139,301 (49.7)
47,191 (17.0)
56,687 (20.5)
56,900 (20.6)
55,627 (20.1)
60,417 (21.8)
145,487 (51.9)
58,119 (20.7)
26,825 (9.6)
32,739 (11.7)
16,993 (6.1)
-
-
-
46,716 (16.7)
121,810 (43.5)
104,987 (37.5)
(Continued )
PLOS ONE | https://doi.org/10.1371/journal.pone.0251177 May 14, 2021
9 / 16
PLOS ONE National estimates from the Youth ’19 Rangatahi smart survey
Table 2. (Continued)
Unknown
Previous national surveys
2001
n (%)
0 (0)
2007
n (%)
242 (2.7)
2012
n (%)
0 (0)
Current survey in 3 regions
2019�(Excluding kura kaupapa Māori)
n (%)
42 (0.6)
National data
2018
n (%)
6,650 (2.4)
1 Number of schools with more than 50 students in years 9–13
2 Totals for each variable (not shown) are different to the overall total number of participating students due to different numbers of missing data for each.
3 Ethnicity was assigned on the basis of prioritised ethnicity, using the NZ Census ethnicity prioritisation method [21]
4 New Zealand Deprivation Index scores based on census areas [22], combined to form 3 categories
https://doi.org/10.1371/journal.pone.0251177.t002
Table 3. National and regional student population estimates- comparison of actual quantities, estimates using sampling weights and estimates using calibrated
weights.
DEMOGRAPHIC
CHARACTERISTICS
DECILE1
DECILE2
DECILE4
DECILE3
DECILE5
DECILE6
DECILE7
DECILE8
DECILE9
AGE13ANDUNDER
AGE14
AGE15
AGE16
GENDER FEMALE
MĀORI
EUROPEAN
PACIFIC
ASIAN
Actual Total
Regional
11839
8471
14212
9175
10379
10145
17084
8604
22554
22241
26685
26527
26181
64737
24393
53889
19913
22358
https://doi.org/10.1371/journal.pone.0251177.t003
Sampling Weights
Regional
Calibrated Weights
Regional
TOTALS
7321
[-45,14688]
6187
[-2501,14876]
13701
[1033,26369]
7920
[-851,16690]
7554
[-2062,17169]
4978
[-811,10768]
27648
[5303,49994]
3000
[-2714,8715]
35989
[3607,68372]
23522
[17565,29479]
29307
[22100,36514]
29099
[22001,36197]
24892
[18471,31313]
72549
[52895,92203]
20384
[15844,24924]
54902
[37045,72760]
16418
[10311,22525]
31858
[19382,44335]
11839
[11839,11839]
8471
[8471,8471]
14212
[14212,14212]
9175
[9175,9175]
10379
[10379,10379]
10145
[10145,10145]
17084
[17084,17084]
8604
[8604,8604]
22554
[22554,22554]
22241
[22241,22241]
26685
[26685,26685]
26527
[26527,26527]
26181
[26181,26181]
64737
[64737,64737]
24393
[24393,24393]
53889
[53889,53889]
19913
[19913,19913]
22358
[22358,22358]
Actual Total
National
14473
12860
25958
20131
21409
38793
35650
33650
38736
47361
56843
57060
55744
139694
58866
145487
26826
32739
Calibrated Weights
National
14473
[14473,14473]
12860
[12860,12860]
25958
[25958,25958]
20131
[20131,20131]
21409
[21409,21409]
38793
[38793,38793]
35650
[35650,35650]
33650
[33650,33650]
38736
[38736,38736]
47361
[47361,47361]
56843
[56843,56843]
57060
[57060,57060]
55744
[55744,55744]
139694
[139694,139694]
58866
[58866,58866]
145487
[145487,145487]
26826
[26826,26826]
32739
[32739,32739]
PLOS ONE | https://doi.org/10.1371/journal.pone.0251177 May 14, 2021
10 / 16
PLOS ONE Table 4. National and regional estimates- comparison of actual quantities, estimates using sampling weights and estimates using calibrated weights.
PROPORTIONS
National estimates from the Youth ’19 Rangatahi smart survey
DECILE1
DECILE2
DECILE4
DECILE3
DECILE5
DECILE6
DECILE7
DECILE8
DECILE9
AGE13 AND UNDER
AGE14
AGE15
AGE16
GENDER FEMALE
MĀORI
EUROPEAN
PACIFIC
ASIAN
Actual Total Regional Sampling Weights Regional Calibrated Weights Regional Actual Total National Calibrated Weights National
0.091
[0.091,0.091]
0.056
[-0.004,0.116]
0.052
[0.052,0.052]
0.051
0.09
0.065
0.108
0.07
0.079
0.077
0.13
0.066
0.172
0.169
0.203
0.202
0.2
0.493
0.186
0.411
0.152
0.17
0.047
[-0.02,0.115]
0.105
[0.007,0.203]
0.061
[-0.009,0.13]
0.058
[-0.016,0.132]
0.038
[-0.007,0.083]
0.212
[0.049,0.374]
0.023
[-0.021,0.067]
0.276
[0.069,0.482]
0.18
[0.164,0.196]
0.224
[0.207,0.242]
0.223
[0.211,0.234]
0.191
[0.174,0.207]
0.555
[0.424,0.687]
0.156
[0.125,0.188]
0.42
[0.348,0.492]
0.126
[0.078,0.173]
0.244
[0.183,0.305]
0.065
[0.065,0.065]
0.109
[0.109,0.109]
0.07
[0.07,0.07]
0.079
[0.079,0.079]
0.078
[0.078,0.078]
0.131
[0.131,0.131]
0.066
[0.066,0.066]
0.173
[0.173,0.173]
0.17
[0.17,0.17]
0.204
[0.204,0.204]
0.203
[0.203,0.203]
0.2
[0.2,0.2]
0.496
[0.496,0.496]
0.187
[0.187,0.187]
0.413
[0.413,0.413]
0.152
[0.152,0.152]
0.171
[0.171,0.171]
0.046
0.092
0.071
0.076
0.138
0.126
0.119
0.137
0.168
0.202
0.202
0.198
0.496
0.209
0.516
0.095
0.116
0.046
[0.046,0.046]
0.092
[0.092,0.092]
0.072
[0.072,0.072]
0.076
[0.076,0.076]
0.138
[0.138,0.138]
0.127
[0.127,0.127]
0.12
[0.12,0.12]
0.138
[0.138,0.138]
0.169
[0.169,0.169]
0.202
[0.202,0.202]
0.203
[0.203,0.203]
0.198
[0.198,0.198]
0.497
[0.497,0.497]
0.21
[0.21,0.21]
0.518
[0.518,0.518]
0.095
[0.095,0.095]
0.117
[0.117,0.117]
https://doi.org/10.1371/journal.pone.0251177.t004
Zealand. The original sampling design was only representative of the three main regions in
New Zealand (Tai Tokerau, Auckland and Waikato). These three regions are believed to repre-
sent the diversity of the New Zealand population [17], however using these data to make infer-
ences about the national situation is imprecise.
One of the main goals of this paper is to improve the estimates at the regional level using
calibrated weights. Calibration aims to account for oversampling or non-response of some
groups of individuals. An interesting example is the proportion of individuals suffering depres-
sive symptoms. Using the original sampling weights, this proportion is estimated to be 0.25
(0.224,0.275), while national calibration yields a lower estimate of 0.227 (0.216,0.239). A reason
for this is that the original sample could have oversampled individuals more prone to suffer
such symptoms, including a higher proportion of girls and a higher proportion of Pacific and
Asian students. Another example is the proportion of individuals who reported binge drinking
in the last 4 weeks. This proportion is estimated as 0.177 (0.155,0.199) using sampling weights,
PLOS ONE | https://doi.org/10.1371/journal.pone.0251177 May 14, 2021
11 / 16
PLOS ONE National estimates from the Youth ’19 Rangatahi smart survey
Table 5. Estimates of total student numbers for health and wellbeing indicators.
TOTALS
Unweighted
Sampling Weights
Regional
Calibrated Weights
Regional
Calibrated Weights
National
MOVED HOME TWO OR MORE TIMES IN LAST YEAR
PARENTS WORRY ABOUT NOT HAVING ENOUGH MONEY
FOR FOOD (OFTEN, ALL THE TIME)
SPENDING ENOUGH TIME WITH AT LEAST ONE PARENT
(MOSTLY)
FAMILY WANTS TO KNOW WHO YOU’RE WITH AND
WHERE YOU ARE (USUALLY OR ALWAYS)
ADULTS AT SCHOOL CARE (A LOT)
526
896
5130
6681
5653
HAVE AN ADULT FEEL OK TALKING TO OUTSIDE FAMILY
3216
FEEL SAFE IN OWN NEIGHBOURHOOD (ALWAYS)†
BULLIED AT SCHOOL (AT LEAST WEEKLY)
WITNESSED ADULTS AT HOME HIT OR PHYSICALLY HURT
EACH OTHER
SEXUAL ABUSE
RATED GENERAL HEALTH AS FAIR OR POOR
CLINICALLY SIGNIFICANT DEPRESSIVE SYMPTOMS (RADS-
SF SCORE �28)
ATTEMPTED SUICIDE IN LAST 12 MONTHS
AT LEAST MONTHLY CIGARETTE USE
AT LEAST ONE EPISODE OF BINGE DRINKING IN LAST 4
WEEKS
AT LEAST MONTHLY MARIJUANA USE
EVER HAD SEXUAL INTERCOURSE
USED A CONDOM AT LAST SEXUAL INTERCOURSE �
EVER BEEN PREGNANT OR GOT SOMEONE PREGNANT
PHYSICAL ACTIVITY FOR AT LEAST 60 MIN EVERY DAY IN
LAST WEEK
ALWAYS WEAR A SEATBELT WHEN DRIVING/BEING
DRIVEN IN A CAR
PASSENGER IN A CAR DRIVEN BY A RISKY DRIVER IN THE
LAST MONTH
DRIVER ENGAGING IN RISKY DRIVING IN THE LAST
MONTH
ACCESSED HEALTH CARE IN LAST 12 MONTHS
WANTED TO SEE A HEALTH PROVIDER BUT WEREN’T
ABLE TO
�Proportion is out of those who have had sex
https://doi.org/10.1371/journal.pone.0251177.t005
4007
383
405
820
4749
1727
439
262
1221
712
1180
695
60
1065
5231
1791
194
5493
1448
9346
[7533,11159]
15976
[12514,19438]
91642
[69496,113788]
119420
[90995,147845]
101015
[77690,124340]
57512
[43342,71681]
71221
[51788,90654]
6781
[4999,8562]
7254
[5597,8911]
14740
[11643,17837]
84841
[63128,106554]
31281
[24447,38114]
7934
[6275,9593]
4621
[3533,5708]
21395
[14913,27877]
12425
[9557,15293]
20877
[15484,26270]
12353
[8685,16021]
1054
[644,1465]
18948
[13766,24131]
93864
[69473,118254]
31409
[22879,39939]
3379
[1890,4869]
98371
[73250,123492]
25807
[20341,31274]
10426
[9451,11402]
17571
[16148,18993]
92278
[91346,93210]
118686
[117509,119863]
101383
[100040,102727]
58427
[56540,60315]
70452
[68657,72248]
7024
[6221,7827]
7669
[7106,8231]
14996
[13976,16015]
85404
[83932,86875]
29596
[28227,30966]
8626
[7856,9397]
5595
[5010,6181]
23678
[21963,25393]
14236
[13248,15225]
24159
[22947,25371]
13803
[12936,14671]
1261
[989,1532]
20275
[19394,21155]
91480
[89825,93134]
32869
[30828,34911]
3705
[2755,4655]
97292
[95461,99122]
26467
[25242,27692]
21632
[19153,24111]
33242
[30122,36362]
199440
[197157,201723]
255696
[252310,259081]
217973
[215324,220622]
130051
[125174,134929]
155142
[150947,159337]
16529
[13977,19081]
15971
[14651,17292]
32977
[30857,35097]
188209
[184259,192158]
61184
[57997,64372]
17105
[15224,18987]
12760
[11175,14344]
57009
[53164,60853]
32905
[30311,35499]
54767
[52081,57453]
31655
[29317,33993]
2564
[1830,3299]
44005
[42054,45955]
202405
[197714,207095]
73403
[68963,77843]
7595
[5545,9645]
212253
[208725,215781]
54682
[51490,57874]
n
7311
7311
7311
7311
7218
7311
7311
7163
6809
6822
7311
7014
7048
6850
6775
7311
6907
6836
1135
6976
6981
6638
1600
7081
7061
PLOS ONE | https://doi.org/10.1371/journal.pone.0251177 May 14, 2021
12 / 16
PLOS ONE National estimates from the Youth ’19 Rangatahi smart survey
Fig 2. Distribution of regional and national sampling weights and calibrated weights.
https://doi.org/10.1371/journal.pone.0251177.g002
while it is estimated to be 0.219 (0.204,0.234) using nationally calibrated weights. The original
sample could have oversampled individuals that were less likely to engage in binge drinking,
with the higher proportion of girls, Pacific and Asian students representing groups who engage
in binge drinking less.
A question that arises is what estimators are more reliable. In such case we recommend the
use of calibrated estimators as they yield confidence intervals that are significantly narrower
than those obtained using the original sampling weights. This is a well-known property of cali-
bration since it reduces the uncertainty in the sample by incorporating information known
prior to the study [19, 23, 24].
An additional goal of this paper was to use the regional sample to make inferences about
the whole adolescent population of New Zealand. This is particularly important because previ-
ous Youth2000 surveys were designed using a national sampling frame instead of a regional
sampling frame. The 2019 survey was designed using a 3-region sampling frame for logistical
and financial reasons. There is ongoing interest in comparing the results and trends with pre-
vious national surveys. To achieve this, we calibrated our regional sampling weights to repre-
sent the national population based on some of the demographic factors presented in Table 2.
The calibrated estimates presented in Table 4 show some differences between the regional and
national proportions.
There are few nationally-representative data available for health and wellbeing indicators
among New Zealand youth, apart from the Youth 2000 surveys. The ASH Year 10 Snapshot
survey reported that 5.9% of Year 10 students are regular smokers. Our data shows that 4.8%
of all secondary school students are regular smokers, but this includes younger students who
are less likely to smoke. The NZ Health Survey estimated that 78.9% of adults over 15 years vis-
ited their GP in the last 12-months, which compares with our estimate of 78.1%. Likewise, the
NZ Health Survey estimated that 20.6% of adults over 15 years had an unmet need for
PLOS ONE | https://doi.org/10.1371/journal.pone.0251177 May 14, 2021
13 / 16
PLOS ONE National estimates from the Youth ’19 Rangatahi smart survey
Table 6. Health and wellbeing indicators.
PROPORTIONS
Unweighted
Sampling Weights
Regional
Calibrated Weights
Regional
Calibrated Weights
National
n
MOVED HOME TWO OR MORE TIMES IN LAST YEAR
PARENTS WORRY ABOUT NOT HAVING ENOUGH MONEY
FOR FOOD (OFTEN, ALL THE TIME)
SPENDING ENOUGH TIME WITH AT LEAST ONE PARENT
(MOSTLY)
FAMILY WANTS TO KNOW WHO YOU’RE WITH AND WHERE
YOU ARE (USUALLY OR ALWAYS)
ADULTS AT SCHOOL CARE (A LOT)
0.072
0.123
0.702
0.914
0.783
HAVE AN ADULT FEEL OK TALKING TO OUTSIDE FAMILY
0.44
FEEL SAFE IN OWN NEIGHBOURHOOD (ALWAYS)†
BULLIED AT SCHOOL (AT LEAST WEEKLY)
WITNESSED ADULTS AT HOME HIT OR PHYSICALLY HURT
EACH OTHER
SEXUAL ABUSE
RATED GENERAL HEALTH AS FAIR OR POOR
CLINICALLY SIGNIFICANT DEPRESSIVE SYMPTOMS (RADS-
SF SCORE �28)
ATTEMPTED SUICIDE IN LAST 12 MONTHS
AT LEAST MONTHLY CIGARETTE USE
AT LEAST ONE EPISODE OF BINGE DRINKING IN LAST 4
WEEKS
AT LEAST MONTHLY MARIJUANA USE
EVER HAD SEXUAL INTERCOURSE
USED A CONDOM AT LAST SEXUAL INTERCOURSE �
EVER BEEN PREGNANT OR GOT SOMEONE PREGNANT
PHYSICAL ACTIVITY FOR AT LEAST 60 MIN EVERY DAY IN
LAST WEEK
ALWAYS WEAR A SEATBELT WHEN DRIVING/BEING DRIVEN
IN A CAR
PASSENGER IN A CAR DRIVEN BY A RISKY DRIVER IN THE
LAST MONTH
0.548
0.053
0.059
0.12
0.65
0.246
0.062
0.038
0.18
0.097
0.171
0.627
0.053
0.153
0.749
0.27
DRIVER ENGAGING IN RISKY DRIVING IN THE LAST MONTH
0.121
ACCESSED HEALTH CARE IN LAST 12 MONTHS��
WANTED TO SEE A HEALTH PROVIDER BUT WEREN’T ABLE
TO���
0.776
0.205
�Taken from those who have ever had sex.
https://doi.org/10.1371/journal.pone.0251177.t006
0.072
[0.059,0.084]
0.122
[0.098,0.147]
0.702
[0.692,0.711]
0.914
[0.905,0.923]
0.783
[0.763,0.804]
0.44
[0.421,0.46]
0.545
[0.51,0.581]
0.053
[0.046,0.06]
0.06
[0.05,0.069]
0.121
[0.104,0.138]
0.65
[0.63,0.669]
0.25
[0.224,0.275]
0.063
[0.049,0.078]
0.038
[0.03,0.046]
0.177
[0.155,0.199]
0.095
[0.086,0.105]
0.169
[0.152,0.186]
0.629
[0.591,0.667]
0.053
[0.038,0.067]
0.152
[0.137,0.167]
0.753
[0.73,0.777]
0.265
[0.244,0.286]
0.118
[0.1,0.135]
0.778
[0.76,0.795]
0.205
[0.189,0.22]
0.08
[0.072,0.087]
0.135
[0.124,0.145]
0.706
[0.699,0.714]
0.909
[0.9,0.918]
0.787
[0.777,0.798]
0.447
[0.433,0.462]
0.539
[0.526,0.553]
0.055
[0.049,0.061]
0.064
[0.059,0.069]
0.125
[0.116,0.133]
0.654
[0.643,0.665]
0.238
[0.227,0.249]
0.069
[0.063,0.075]
0.046
[0.041,0.051]
0.197
[0.183,0.212]
0.109
[0.101,0.117]
0.197
[0.187,0.208]
0.613
[0.589,0.637]
0.054
[0.043,0.066]
0.164
[0.157,0.171]
0.741
[0.729,0.754]
0.282
[0.266,0.297]
0.127
[0.114,0.14]
0.774
[0.762,0.786]
0.211
[0.202,0.22]
0.077
[0.068,0.086]
0.118
[0.107,0.129]
0.71
[0.702,0.718]
0.91
[0.898,0.922]
0.787
[0.777,0.797]
0.463
[0.446,0.48]
0.552
[0.537,0.567]
0.06
[0.051,0.07]
0.061
[0.056,0.066]
0.126
[0.118,0.134]
0.67
[0.656,0.684]
0.227
[0.216,0.239]
0.063
[0.056,0.07]
0.048
[0.043,0.054]
0.219
[0.204,0.234]
0.117
[0.108,0.126]
0.206
[0.196,0.217]
0.617
[0.59,0.644]
0.049
[0.035,0.063]
0.164
[0.157,0.172]
0.756
[0.74,0.772]
0.289
[0.273,0.305]
0.132
[0.116,0.147]
0.781
[0.77,0.792]
0.202
[0.191,0.213]
7311
7311
7311
7311
7218
7311
7311
7163
6809
6822
7311
7014
7048
6850
6775
7311
6907
6836
1135
6976
6981
6638
1600
7081
7061
PLOS ONE | https://doi.org/10.1371/journal.pone.0251177 May 14, 2021
14 / 16
PLOS ONE National estimates from the Youth ’19 Rangatahi smart survey
healthcare, and our data estimates this to be 20.2% [25, 26] These results highlight that calibra-
tion methods can improve the precision of national estimates when compared to similar sur-
veys, however it should be noted that calibration methods cannot account for factors outside
of demographic features (i.e. unique regional differences) and therefore should be utilised with
this limitation in mind.
Future research will involve calibration of the previous surveys using a similar approach to
reduce bias in the estimates, as well as investigating how different designs can improve the
results and methods for combining the periodic complex surveys done in the years 2001, 2007,
2012 and 2019.
Supporting information
S1 File.
(ZIP)
S1 Statistics.
(PDF)
Author Contributions
Conceptualization: T. Fleming, S. Lewycka.
Data curation: C. Rivera-Rodriguez, D. Archer.
Formal analysis: C. Rivera-Rodriguez.
Funding acquisition: T. C. Clark, T. Fleming.
Investigation: T. C. Clark, T. Fleming, S. Crengle, R. Peiris-John.
Methodology: C. Rivera-Rodriguez.
Project administration: T. C. Clark, D. Archer.
Resources: T. Fleming, D. Archer.
Software: C. Rivera-Rodriguez, D. Archer.
Supervision: T. Fleming, S. Lewycka.
Validation: D. Archer, S. Lewycka.
Visualization: S. Lewycka.
Writing – original draft: C. Rivera-Rodriguez, S. Lewycka.
Writing – review & editing: T. C. Clark, T. Fleming, D. Archer, S. Crengle, R. Peiris-John, S.
Lewycka.
References
1. Clark T, Fleming T, Bullen P, Crengle S, Denny S, et al. (2013) Health and well-being of secondary
school students in New Zealand: trends between 2001, 2007 and 2012. J Paediatr Child Health 49:
925–934. https://doi.org/10.1111/jpc.12427 PMID: 24251658
2.
Lewycka S, Clark T, Peiris-John R, Fenaughty J, Bullen P, et al. (2018) Downwards trends in adoles-
cent risk-taking behaviours in New Zealand: Exploring driving forces for change. J Paediatr Child Health
54: 602–608. https://doi.org/10.1111/jpc.13930 PMID: 29779222
3. Bor W, Dean AJ, Najman J, Hayatbakhsh R (2014) Are child and adolescent mental health problems
increasing in the 21st century? A systematic review. Aust N Z J Psychiatry 48: 606–616. https://doi.org/
10.1177/0004867414533834 PMID: 24829198
PLOS ONE | https://doi.org/10.1371/journal.pone.0251177 May 14, 2021
15 / 16
PLOS ONE National estimates from the Youth ’19 Rangatahi smart survey
4. Soneji S, Barrington-Trimis JL, Wills TA, Leventhal AM, Unger JB, et al. (2017) Association Between Ini-
tial Use of e-Cigarettes and Subsequent Cigarette Smoking Among Adolescents and Young Adults: A
Systematic Review and Meta-analysis. JAMA Pediatr 171: 788–797. https://doi.org/10.1001/
jamapediatrics.2017.1488 PMID: 28654986
5. Clark TC, Le Grice J, Moselen E, Fleming T, Crengle S, et al. (2018) Health and wellbeing of Māori sec-
ondary school students in New Zealand: Trends between 2001, 2007 and 2012. Aust N Z J Public
Health 42: 553–561. https://doi.org/10.1111/1753-6405.12839 PMID: 30370961
6.
Fleming T, Peiris-John R, Crengle S, Archer D, Sutcliffe K, et al. (2020) Youth19 Rangatahi Smart Sur-
vey, Initial Findings:Introduction and Methods. The Youth19 Research Group. The University of Auck-
land and Victoria University of Wellington, New Zealand.
7. Morton SMB, Ramke J, Kinloch J, Grant CC, Carr PA, et al. (2015) Growing Up in New Zealand cohort
alignment with all New Zealand births. Aust N Z J Public Health 39: 82–87. https://doi.org/10.1111/
1753-6405.12220 PMID: 25168836
8. Education Counts: School Rolls (n.d.). Available: https://www.educationcounts.govt.nz/statistics/
schooling/student-numbers/6028. Accessed 22 July 2020.
9. Deville J-C, Sa¨rndal C-E (1992) Calibration estimators in survey sampling. J Am Stat Assoc 87: 376–
382. https://doi.org/10.1080/01621459.1992.10475217
10. Sarndal CE, Swensson B, Wretman J (1992) Model Assisted Survey Sampling. New York.
11.
Fuller WA (2009) Sampling Statistics. Hoboken, NJ, USA: John Wiley & Sons, Inc. https://doi.org/10.
1002/9780470523551
12. Sa¨ rndal C-E, Lundstro¨ m S (2005) Introduction to estimation in the presence of nonresponse. Estimation
in Surveys with Nonresponse. Chichester, UK: John Wiley & Sons, Ltd. pp. 43–48. https://doi.org/10.
1002/0470011351.ch5
13. Breslow NE, Holubkov R (1997) Weighted likelihood, pseudo-likelihood and maximum likelihood meth-
ods for logistic regression analysis of two-stage data. Stat Med 16: 103–116. https://doi.org/10.1002/
(sici)1097-0258(19970115)16:1<103::aid-sim474>3.0.co;2-p PMID: 9004386
14. Spiegelman D, Rivera-Rodriguez CL, Haneuse S (2016) Evaluating Public Health Interventions: 3. The
Two-Stage Design for Confounding Bias Reduction-Having Your Cake and Eating It Two. Am J Public
Health 106: 1223–1226. https://doi.org/10.2105/AJPH.2016.303250 PMID: 27285260
15. Breslow NE, Amorim G, Pettinger MB, Rossouw J (2013) Using the Whole Cohort in the Analysis of
Case-Control Data: Application to the Women’s Health Initiative. Stat Biosci 5. https://doi.org/10.1007/
s12561-013-9080-2 PMID: 24363785
16. Deville JC, Sarndal CE (1992) Calibration estimators in survey sampling. J Am Stat Assoc 87: 376–
17.
18.
382.
Fuller W (2009) Sampling Statistics. John Wiley and Sons.
Lumley T, Shaw PA, Dai JY (2011) Connections between survey calibration estimators and semipara-
metric models for incomplete data. Int Stat Rev 79: 200–220. https://doi.org/10.1111/j.1751-5823.
2011.00138.x PMID: 23833390
19. Breslow NE, Lumley T, Ballantyne CM, Chambless LE, Kulich M (2009) Improved Horvitz-Thompson
Estimation of Model Parameters from Two-phase Stratified Samples: Applications in Epidemiology.
Stat Biosci 1: 32. https://doi.org/10.1007/s12561-009-9001-6 PMID: 20174455
20. School Deciles (2021). New Zealand Ministry of Education. Available: https://www.education.govt.nz/
school/funding-and-financials/resourcing/operational-funding/school-decile-ratings/. Accessed 23
March 2021.
21. Statistics New Zealand (2005) Statistical standard for ethnicity.
22. Salmond C, Crampton P, Sutton F, Atkinson J (2006) NZDep2006 Census Area Unit Data. Available:
http://www.wnmeds.ac.nz/academic/dph/research/socialindicators.html.
23. Rivera-Rodriguez C, Toscano C, Resch S (2019) Improved calibration estimators for the total cost of
health programs and application to immunization in Brazil. PLoS One 14: e0212401. https://doi.org/10.
1371/journal.pone.0212401 PMID: 30840645
24. Breslow NE, Lumley T, Ballantyne CM, Chambless LE, Kulich M (2009) Using the whole cohort in the
analysis of case-cohort data. Am J Epidemiol 169: 1398–1405. https://doi.org/10.1093/aje/kwp055
PMID: 19357328
25. NZHS (2018) Regional Data Explorer. New Zealand Health Survey: Regional Data Explorer. Available:
https://minhealthnz.shinyapps.io/nz-health-survey-2014-17-regional-update/. Accessed 25 March
2021.
26. ASH (2019) 2019 ASH Year 10 Snapshot: Topline Results–Smoking. Action on Smoking and Health.
PLOS ONE | https://doi.org/10.1371/journal.pone.0251177 May 14, 2021
16 / 16
PLOS ONE |
10.1371_journal.pone.0248476 | RESEARCH ARTICLE
Characterization of neonatal opioid
withdrawal syndrome in Arizona from 2010-
2017
Emery R. Eaves1,2,3☯, Jarrett Barber3,4, Ryann WhealyID
Jill Hager Cocking4,5, Joseph Spadafino8, Crystal M. HeppID
3,4,5☯*
4,5, Sara A. Clancey6, Rita Wright7,
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
1 Department of Anthropology, Northern Arizona University, Flagstaff, Arizona, United States of America,
2 Center for Health Equity Research, Northern Arizona University, Flagstaff, Arizona, United States of
America, 3 Southwest Health Equity Research Collaborative, Northern Arizona University, Flagstaff, Arizona,
United States of America, 4 School of Informatics, Computing, and Cyber Systems, Northern Arizona
University, Flagstaff, Arizona, United States of America, 5 Pathogen and Microbiome Institute, Northern
Arizona University, Flagstaff, Arizona, United States of America, 6 Institute for Human Development,
Northern Arizona University, Flagstaff, Arizona, United States of America, 7 Department of Social Work,
Northern Arizona University, Flagstaff, Arizona, United States of America, 8 Arizona Department of Health
Services, Phoenix, Arizona, United States of America
☯ These authors contributed equally to this work.
* crystal.hepp@nau.edu
Abstract
In this paper, we describe a population of mothers who are opioid dependent at the time of
giving birth and neonates exposed to opioids in utero who experience withdrawal following
birth. While there have been studies of national trends in this population, there remains a
gap in studies of regional trends. Using data from the Arizona Department of Health Ser-
vices Hospital Discharge Database, this study aimed to characterize the population of neo-
nates with neonatal opioid withdrawal syndrome (NOWS) and mothers who were opioid
dependent at the time of giving birth, in Arizona. We analyzed approximately 1.2 million elec-
tronic medical records from the Arizona Department of Health Services Hospital Discharge
Database to identify patterns and disparities across socioeconomic, ethnic, racial, and/or
geographic groupings. In addition, we identified comorbid conditions that are differentially
associated with NOWS in neonates or opioid dependence in mothers. Our analysis was
designed to assess whether indicators such as race/ethnicity, insurance payer, marital sta-
tus, and comorbidities are related to the use of opioids while pregnant. Our findings suggest
that women and neonates who are non-Hispanic White and economically disadvantaged,
tend be part of our populations of interest more frequently than expected. Additionally,
women who are opioid dependent at the time of giving birth are unmarried more often than
expected, and we suggest that marital status could be a proxy for support. Finally, we identi-
fied comorbidities associated with neonates who have NOWS and mothers who are opioid
dependent not previously reported.
OPEN ACCESS
Citation: Eaves ER, Barber J, Whealy R, Clancey
SA, Wright R, Cocking JH, et al. (2021)
Characterization of neonatal opioid withdrawal
syndrome in Arizona from 2010-2017. PLoS ONE
16(6): e0248476. https://doi.org/10.1371/journal.
pone.0248476
Editor: Barbara Wilson Engelhardt, Monroe Carell
Junior Children’s Hospital at Vanderbilt, UNITED
STATES
Received: June 10, 2020
Accepted: February 28, 2021
Published: June 3, 2021
Peer Review History: PLOS recognizes the
benefits of transparency in the peer review
process; therefore, we enable the publication of
all of the content of peer review and author
responses alongside final, published articles. The
editorial history of this article is available here:
https://doi.org/10.1371/journal.pone.0248476
Copyright: © 2021 Eaves et al. This is an open
access article distributed under the terms of the
Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: Individual level data
cannot be shared publicly because of identifiability
concerns that could occur with individual level
PLOS ONE | https://doi.org/10.1371/journal.pone.0248476 June 3, 2021
1 / 17
PLOS ONE electronic medical record data. We have provided
aggregate data in S1 Table. Hospital discharge data
are housed at the Arizona Department of Health
Services, and may be provided for future analyses
upon approval of a Human Subjects Research
Board protocol at the discretion of the Arizona
Department of Health Services. Researchers
wishing to reproduce or build on this study will
need to submit a data request to the Arizona
Department of Health Services to be approved:
https://www.azdhs.gov/documents/director/
administrative-counsel-rules/HSRB_
NewProductSubmission.pdf.
Funding: This work was supported by a Pilot grant
project to Crystal Hepp, Ph.D. as part of the NIH/
NIMHD RCMI U54MD012388 (Julie Baldwin, Ph.D.
PI). The authors acknowledge the contributions of
Arizona Department of Health Services staff
members Kyle Gardner and Timothy Flood for
assistance defining the data request, the Northern
Arizona University Information Technology
Services for managing data security issues, and the
Southwest Health Equity Research Collaborative
Research Infrastructure Core for analytical
assistance.
Competing interests: The authors have declared
that no competing interests exist.
NAS in Arizona
Introduction
Neonatal Abstinence Syndrome (NOWS), also called Neonatal Opioid Withdrawal Syndrome
(NOWS), is a consequence of abrupt withdrawal from intrauterine opioid exposure after birth
[1–3]. Clinical abstinence symptoms are observed in 60–80% of substance exposed neonates,
and include neurological, gastrointestinal, and autonomic complications [1]. Chronic opioid
exposure in utero occurs in three different contexts: (1) active, untreated addiction to opioids
(heroin or prescription opioids); (2) opioids used for chronic pain management; and (3) Medi-
cation Assisted Treatment (MAT) such as methadone or buprenorphine during pregnancy
[4]. Common diagnostic criteria for NOWS include tremors, seizures, convulsions, feeding
problems, vomiting, diarrhea, respiratory problems, and other neonatal complications [5]. In
this paper, we describe the results of analysis of Arizona Hospital Discharge records from 2010
to 2017 to characterize the population of neonates born with NOWS and their mothers. Our
findings suggest that other conditions may co-occur with NOWS more often than several of
the commonly used diagnostic criteria. Our results suggest a need for better characterization
of comorbid conditions in NOWS neonates and their mothers. Improving understanding of
comorbid conditions and diagnostic criteria has implications for identification and secondary
prevention of NOWS.
Nationwide, NOWS cases have been increasing exponentially and in national studies, the
costs of NOWS births were at least three times greater than normal births [6]. NOWS births
are more common in rural than in urban areas [7–10] and more likely to be among medicaid
covered births [11]. Between 2004 and 2013, there was a 7-fold increase in NOWS in rural
areas alone [12]. In Arizona, where 80% of the population live in mental health professional
shortage areas [13], increasing availability of illicit drugs and steady rates of prescription opi-
oid pain reliever use impact the population, including pregnant women [14]. The incidence of
NOWS in Arizona continues to rise from 1.3 per 1,000 births in 1999 to 3.9 per 1,000 births in
2013, a threefold increase [15]. The number of opioid-related deaths in Arizona has increased
74% from 2012 to 2016, resulting in more than 2 deaths per day in 2016 [16]. Consistent with
findings that socioeconomic status is a factor in NOWS cases [17], Hussaini and Saavedra
reported that nearly 80% of NOWS cases in Arizona were paid for by Medicaid, especially in
the border regions of the state [14].
On June 5, 2017, Arizona Governor Doug Ducey declared a Public Health State of
Emergency due to the opioid epidemic [18]. An Enhanced Surveillance Advisory went
into effect as a first step toward understanding the current opioid situation in Arizona
and to collect data to develop best practices for interventions. As part of this advisory,
any opioid-related event (opioid-related death, naloxone doses administered, NOWS
cases, etc.) must be reported to Arizona Department of Health Services within 24 hours
[18]. To inform the development of best practices in NOWS intervention in the context of
increased attention, we conducted a comprehensive characterization of the population of
neonates with NOWS and mothers who are opioid dependent. This analysis is the first of its
kind in Arizona, and moves toward a better understanding of the population of Arizona
neonates born with NOWS and their mothers. The analysis presented below is focused
on the 5.5 years prior to and 1.5 years following the implementation of Arizona’s opioid
surveillance policy. While there have been studies of national trends in this population
[17], regional trends and issues are less well understood [9, 19]. In this pilot project, we use
data from the Arizona Department of Health Services Hospital Discharge Database to char-
acterize the population of neonates born with NOWS and their mothers in Arizona from
2010 to 2017.
PLOS ONE | https://doi.org/10.1371/journal.pone.0248476 June 3, 2021
2 / 17
PLOS ONE NAS in Arizona
Methods
Data request
Northern Arizona University (NAU) has a data use agreement with the Arizona Department
of Health Services (ADHS), allowing researchers an expedited path to access records in the
ADHS Hospital Discharge Database and other databases. We submitted a data request to the
Human Subjects Research Board (HSRB) at ADHS, to access electronic medical records for all
neonates who were born and all mothers who gave birth in Arizona from 2010–2017. Notably,
Indian Health Services hospitals are not required to report inpatient and emergency depart-
ment visits to ADHS, so birth events at these hospitals are not captured. The request was
approved as public health surveillance, and we additionally submitted a request for determina-
tion of non-human subjects research to the NAU Institutional Review Board. Based on the
ADHS HSRB’s determination of public health surveillance, NAU IRB determined the research
to be non-human subjects research. This final dataset, which was transferred between two
secure servers at ADHS and NAU, included the electronic medical records from 643,370
mothers and 663,353 neonates. All variables and descriptions included in the final dataset used
in this project are included in S1 Table.
Identifying the population of interest
The purpose of this study is to characterize the population of neonates with NOWS and moth-
ers who are opioid dependent at the time of giving birth. We used insurance codes to identify
subpopulations of interest within the larger mother and neonate dataset. The dataset spans
2010–2017, including both the International Classification of Diseases, Ninth Revision, Clini-
cal Modification (ICD-9-CM, referred to as ICD9) and the International Classification of Dis-
eases, Tenth Revision, Clinical Modification (ICD-10-CM, referred to as ICD10), as the
change from ICD9 to ICD10 was required by all healthcare facilities in the United States no
later than October 1, 2015. To identify neonates with NOWS, we used ICD9 and ICD10 codes
779.5 and P96.1, respectively. Similarly, to identify mothers who were opioid dependent at the
time of giving birth, we used ICD9 codes 304.00–304.03 and 304.70–304.73 and ICD10 code
F11, including all subcategories. These data include neonates of all gestational ages, including
those in neonatal intensive care.
Healthcare utilization
To better understand hospital resource utilization and to serve as a proxy for severity of mor-
bidity, we compared length of stay and total charges of the subpopulations to the total popula-
tion of Arizona mothers who have just given birth and neonates. The two variables were
compared to each other using linear regression to better understand how well one explains the
other, and t-tests were used to determine if the populations of interest had means that were sig-
nificantly different.
Demographic disparities
We conducted chi-square tests to determine if selected subpopulations belonged to specific
racial and/or ethnic groups or used particular insurance payers significantly more often than
expected. Similarly, we used a chi-square test to determine if mothers who were dependent on
opioids at the time of giving birth had certain marital statuses more frequently than expected.
Expected proportions were determined from the entire mother or neonate datasets (S1 Table).
PLOS ONE | https://doi.org/10.1371/journal.pone.0248476 June 3, 2021
3 / 17
PLOS ONE NAS in Arizona
Geographic disparities
To identify geographic locations where there were more opioid dependent mothers at the time
of giving birth than expected based on the total number of mothers who gave birth, we con-
ducted a chi-square analysis. This analysis was completed for all non-tribal primary care areas
in Arizona, aggregated from 2010–2017. A primary care area (PCA) is an area in which most
residents seek primary health care from the same place. The Arizona Department of Health
Services states that the PCA is meant to represent residents’ “primary care seeking patterns”
[20]. In addition, PCAs are aggregated to prevent re-identification of a patient in Arizona
while allowing for resolution of population health issues at a scale better than that at which the
geographically large Arizona counties provide.
Associated comorbid conditions
In addition to demographic information, each inpatient and emergency department electronic
medical record includes up to 26 ICD billing codes, including admitting and principal diagno-
sis codes, which could be any of approximately 13,000 ICD-9-CM or 69,000 ICD-10-CM
codes. In the case of the selected subpopulations, these codes may include information regard-
ing comorbidities of NOWS or opioid dependence. To understand comorbidity association
with NOWS and opioid dependence, we selected comorbidities for their importance in classi-
fying NOWS and opioid dependence as measured by their average minimum depth to the
maximal sub-tree in classification random forests [21]. For this, we used the function var.
select, with options method = ‘md’ and conservative = ‘low’, in the R library package random-
ForestSRC [22, 23]. Our data present us with an imbalanced classification problem [24],
wherein positive cases of NOWS or opioid dependence represent a minority of cases, with the
majority of cases being negative. In such situations, overall classification performance—hence
comorbidity selection—is dominated by the majority class, whereas our interest leans, instead,
toward correct classification of the minority class. We use the method of balanced random for-
ests [24], as implemented in the function imbalanced.rfsrc in the R library package random-
ForestSRC [22, 23], to grow balanced classification random forests for NOWS and opioid
dependence before computing the importance of comorbidities.
Results
NOWS is increasing rapidly in Arizona. Opioid overdose in Arizona has been a cause for great
concern, with suspected overdoses (n = 32,900, ~35 per day) and deaths (n = 3,935, ~4 per
day) at epidemic levels from June 15, 2017 through January 16, 2020 [18]. The large number of
neonates with NOWS born during the same period (n = 1,295), a consequence of the rise in
opioid use, also warrants attention. To address this issue, we characterized the population of
neonates with NOWS and mothers who were dependent on opioids at the time of giving birth
within the context of the entire population of neonates born and mothers who gave birth in
Arizona from 2010 through 2017.
Healthcare utilization
To determine the impact of maternal opioid use during pregnancy on both neonatal and
maternal morbidity as well as on healthcare utilization we compared hospitalization rates,
average length of stay, and total charges of the entire populations versus the populations of
interest. During the period of time represented in these data, the rate of newborn neonates
who have NOWS has more than doubled from approximately 34 in 2010 to 88 in 2017, per
every 10,000 births. (Fig 1). Similarly, the rate of mothers who are opioid dependent at the
PLOS ONE | https://doi.org/10.1371/journal.pone.0248476 June 3, 2021
4 / 17
PLOS ONE NAS in Arizona
Fig 1. NOWS hospitalization rates, per 10,000 births, in Arizona from 2010 to 2017. Mothers who are dependent
on opioids at the time of giving birth (dark grey) and newborn infants with NOWS (light grey).
https://doi.org/10.1371/journal.pone.0248476.g001
time of giving birth has increased from 19 to 85 per every 10,000 mothers who have given
birth (Fig 1). While reporting for neonates was substantially higher than reporting for mothers
in 2010, hospitalization rates have evened out over time.
In Arizona, neonates with NOWS have an average length of stay (mean: 19.71 days, median:
16 days) approximately six times longer than that of all neonates (mean: 3.17 days, median: 2
days) (Table 1). Similarly, average total charges are also significantly higher for neonates who
have NOWS (mean: $84,615, median: $49,887) in comparison to all neonates (mean: $10,784,
median: $3,223). While neonates with NOWS only compose 0.5% of the total neonate popula-
tion, the total charges associated with neonates who have NOWS account for 4.5% of all birth
related charges ($323,230,298 of $7,153,221,072) from 2010 through 2017. Total charges and
length of stay were also significantly higher for mothers who were dependent on opioids at
the time of giving birth, however, the differences were modest in comparison to the neonate
population.
Demographic disparities
Previous studies characterizing neonates with NOWS and mothers who are dependent on opi-
oids at the time of giving birth in the United States found that these populations were, more
often than expected, non-Hispanic white and insured by Medicaid [25, 26]. We additionally
examined these demographics to determine if the most heavily impacted populations in Ari-
zona followed national trends. In agreement with previous studies, we found that both neonate
Table 1. Comparison of hospital utilization variables in the target versus non-target populations and results of the t-test analyses.
POPULATION
INFANTS WITHOUT NAS
INFANTS WITH NAS
NON-OPIOID DEPENDENT MOTHERS
OPIOID DEPENDENT MOTHERS
https://doi.org/10.1371/journal.pone.0248476.t001
LENGTH OF STAY (DAYS)
TOTAL CHARGES ($)
Mean
3.07
19.71
2.44
3.22
Median
p-value
2
16
2
3
P<0.0001
P<0.0001
Mean
10356
84615
16801
23674
Median
p-value
3209
49887
14174
18406
P<0.0001
P<0.0001
PLOS ONE | https://doi.org/10.1371/journal.pone.0248476 June 3, 2021
5 / 17
PLOS ONE NAS in Arizona
Fig 2. Comparison of the observed versus expected proportions of NOWS in each racial/ethnic group of infants. Boxes labelled
higher or lower indicate that observed proportions are significantly higher or lower than expected proportions based on a chi-square
analysis after post-hoc comparisons that incorporate a Bonferroni correction with six groups (p<0.001/6). A: Asian, H/L: Hispanic or
Latino, AI/AN: American Indian or Alaskan Native, B: Black, NH/PI: Native Hawaiian or Pacific Islander, NHW: Non-Hispanic
White.
https://doi.org/10.1371/journal.pone.0248476.g002
(Fig 2) and mother (Fig 3) populations of interest were, significantly more often than expected,
non-Hispanic white, and were Asian or Hispanic/Latino significantly less often than expected.
Our target populations were insured by either Medicaid or Medicare significantly more
often and by private or military (TRICARE) insurance less often than expected (Figs 4 and 5).
We additionally considered maternal marital status and found that women dependent on
opioids at the time of giving birth were unmarried significantly more often than expected
based on the total population proportions while unmarried women were dependent on opioids
significantly less often than expected (Fig 6). We suppressed data from categories where there
were less than 10 mothers (i.e. widowed).
Geographic disparities
Arizona is the 6th largest state in the US by area, but is composed of only 15 counties, where six
are among the top 20 geographically largest counties in the United States. The result of a large
state being spread into relatively few counties is that distinct human populations are forced
into county level estimates which are unlikely to provide a relevant picture of population
health. The ADHS has approached this issue by aggregating and reporting population health
results for many conditions at the level of PCA. In an effort to identify if and where maternal
opioid dependence is clustered, we adopted the ADHS strategy and compared counts across
the 126 PCAs that compose Arizona. Within the entire maternal dataset, 25,936 records did
not include a PCA, including 75 mothers who were dependent on opioids at the time of giving
birth, and these records were not included in the geographic analysis. In addition, we sup-
pressed statistically significant results for PCAs where there were fewer than 10 mothers who
were dependent on opioids at the time of giving birth as well as those that are primarily
PLOS ONE | https://doi.org/10.1371/journal.pone.0248476 June 3, 2021
6 / 17
PLOS ONE NAS in Arizona
Fig 3. Comparison of the observed versus expected proportions of opioid dependence at the time of giving birth in each racial/ethnic group of mothers. Boxes
labelled higher or lower mean that the observed proportions are significantly higher or lower than the expected proportion after post-hoc comparisons that incorporate
a Bonferroni correction with six groups (p<0.001/6). A: Asian, H/L: Hispanic or Latino, AI/AN: American Indian or Alaskan Native, B: Black, NH/PI: Native
Hawaiian or Pacific Islander, NHW: Non-Hispanic White.
https://doi.org/10.1371/journal.pone.0248476.g003
composed of tribal nations. Rather, we have reported those PCAs back to the ADHS for use in
their decision processes. An initial chi-square test revealed that opioid dependence among
mothers who had given birth significantly deviated from the expected distribution across
PCAs. Post-hoc comparisons revealed that there were significantly more mothers who were
dependent on opioids residing in the following PCAs than expected: Casas Adobes, Encanto
Village, Flowing Wells, Globe, North Mountain Village, Prescott, Safford, Kingman, Tucson
Central, Tucson East, Tucson Foothills, and Tucson South (Fig 7). The following PCAs had
significantly fewer mothers who were dependent on opioids at the time of giving birth than
expected: Buckeye, Estrella Village and Tolleson, Gilbert Central, Gilbert South, Maryvale Vil-
lage, and Yuma. Future studies may investigate which PCA characteristics may contribute to
or mitigate opioid dependence among pregnant women and women of child-bearing age.
Associated comorbid conditions
As mentioned in the Methods section, we used random forests to select comorbidities associ-
ated with NOWS and opioid dependence. Presence or absence of an ICD9 or 10 code for
NOWS or opioid dependence was used as the labelled target variable. We analyzed four sets of
data for identification of comorbid conditions: Neonates with ICD9 codes, Neonates with
ICD10 codes, Mothers with ICD9 codes, Mothers with ICD10 codes. For neonates with
NOWS (Table 2), we found that, in agreement with previous studies, feeding problems, respi-
ratory distress (transitory tachypnea), and neonatal jaundice commonly co-occurred with
PLOS ONE | https://doi.org/10.1371/journal.pone.0248476 June 3, 2021
7 / 17
PLOS ONE NAS in Arizona
Fig 4. Comparison of the observed versus expected proportions of NOWS at the time birth in each payor group utilized for infants. Boxes
labelled higher or lower mean that the observed proportions are significantly higher or lower than the expected proportion after post-hoc
comparisons that incorporate a Bonferroni correction with six groups (p<0.001/6). PHI: Private Health Insurance, Self: Self Pay, IHS: Indian
Health Services.
https://doi.org/10.1371/journal.pone.0248476.g004
NOWS [27–29]. We also found that neonatal candidiasis infection and diaper or skin rash
(diaper dermatitis) were among top-ranked comorbid conditions.
Analysis of mothers’ discharge records (Table 3) identified several comorbid conditions in
agreement with previous research on women with opioid dependence during pregnancy. Poly-
substance use, notably tobacco, alcohol, and stimulants, is more common among pregnant
women who use opioids [30]. Chronic pain, mental health conditions, unspecified anxiety,
and other viral illnesses were also highly-ranked comorbid conditions.
Discussion
Neonates with NOWS in Arizona and their mothers from 2010–2017 tended to be socioeco-
nomically disadvantaged, non-Hispanic White, and geographically clustered throughout Ari-
zona. In addition, mothers in this group are unpartnered more frequently than expected,
which may indicate a relative lack of social support. Unsurprisingly, characteristics of mothers
with opioid dependence at the time of birth were closely related to those of neonates with
NOWS. There does not appear to be an increase in reported of maternal opioid dependence
from 2016 to 2017. Mothers who were opioid dependent accumulated nearly $8000 more in
total charges and stayed a half day longer than those who were not opioid dependent. There
does not appear to be a significant age difference between mothers with opioid dependence at
the time of birth and those who are not opioid dependent.
There are major inconsistencies in substance use screening and NOWS diagnosis and treat-
ment [31, 32]. The American Academy of Pediatrics has called for more similarity and
PLOS ONE | https://doi.org/10.1371/journal.pone.0248476 June 3, 2021
8 / 17
PLOS ONE NAS in Arizona
Fig 5. Comparison of the observed versus expected proportions of opioid dependence at the time of giving birth in each payor group
utilized for mothers. Boxes labelled higher or lower mean that the observed proportions are significantly higher or lower than the expected
proportion after post-hoc comparisons that incorporate a Bonferroni correction with six groups (p<0.001/6). PHI: Private Health Insurance, Self:
Self Pay, IHS: Indian Health Services.
https://doi.org/10.1371/journal.pone.0248476.g005
standardization of care for neonates with NOWS [33]. Standardized screening and treatment
also has the potential to improve care for neonates with NOWS and their mothers who are opi-
oid dependent at the time of giving birth [31, 34]. In an effort to improve and standardize
NOWS treatment, there is a need for better and more reliable criteria for screening and
early identification of NOWS cases. Analysis of hospital discharge records is one step toward
better characterization of co-morbid conditions that could improve precision in early NOWS
identification.
In a recent study of vaginal flora, Farr et al observed significantly higher rates of candidiasis
in pregnant mothers receiving medication assisted opioid treatment than in control groups
[35]. Diaper dermatitis is a known condition common among neonates with NOWS, however,
it is not considered a reliable diagnostic criterion [36]. Our data suggest that considerably
more attention should be paid to potential links between dermatitis and vaginal candidiasis
with NOWS. To our knowledge there are no studies linking increased rates of neonatal candi-
diasis with vaginal candidiasis increases among mothers using medication assisted (opioid
maintenance) treatment.
Like the neonate analysis, analysis of mothers’ discharge records identified several comor-
bid conditions that previous studies have found to be associated with opioid dependence dur-
ing pregnancy. Polysubstance use is more common among pregnant women who use opioids
[30]. Tobacco use is particularly prevalent among this population, with estimates of tobacco
use as high as 85–90% among pregnant women treated with buprenorphine or methadone
[37–39]. Pregnant women using opioids have been found to be more likely to be diagnosed
with depression, anxiety, post-traumatic stress disorder, and panic disorder [40], which is also
PLOS ONE | https://doi.org/10.1371/journal.pone.0248476 June 3, 2021
9 / 17
PLOS ONE NAS in Arizona
Fig 6. Comparison of the observed versus expected proportions of opioid dependence at the time of giving birth in by
maternal marital status. Boxes labelled higher or lower mean that the observed proportions are significantly higher or lower
than the expected proportion after post-hoc comparisons that incorporate a Bonferroni correction with six groups (p<0.001/5).
M: Married, K: Unknown, D: Divorced, I: Single, S: Separated.
https://doi.org/10.1371/journal.pone.0248476.g006
in agreement with our study, where maternal mental disorders are highly ranked. Interest-
ingly, while we found studies that identified increased opioid prescriptions written to women
with perineal lacerations [41], to our knowledge no studies have reported an association
between perineal lacerations and opioid dependence at the time of giving birth. Epidural anes-
thesia has been associated with increased perineal laceration [42] and mothers who are opioid
dependent at the time of giving birth are not candidates for pain management with opioids
[43, 44]. Future research should consider whether the risk of perineal laceration among opioid
dependent women is elevated due to pain management practices or other factors.
Although additional studies are necessary to better understand the association between
marital status and opioid use disorder, we suspect that marital status is a proxy for social sup-
port. Previous studies reported that married individuals are less likely to use illicit drugs [45]
and those who participate in substance-abuse treatment programs are more likely to experi-
ence positive outcomes [46–49]. Heinz et al. found that that close spousal relationships were a
good predictor of reduced cocaine and heroin use in individuals during and after treatment
[50]. With the results of these previous studies, our results suggest further investigation into
outcomes associated with marital and perhaps other forms of social support when considering
opioid use disorder in pregnant women and women of childbearing age.
Areas with higher than expected rates of NOWS in relation to population estimates warrant
additional research. In our meetings with stakeholders and investigation of local understand-
ings of areas with higher rates of drug use in the state, observations were consistent with our
findings. For example, the Prescott area is known for a proliferation of “sober living houses” in
PLOS ONE | https://doi.org/10.1371/journal.pone.0248476 June 3, 2021
10 / 17
PLOS ONE NAS in Arizona
Fig 7. Comparison of the observed versus expected proportions of opioid dependence at the time of giving birth by maternal residential PCA. Red or blue
indicates that there is a significantly higher or lower number of mothers using opioids than expected. A map with all PCAs labelled can be found on the ADHS
website: https://www.azdhs.gov/documents/prevention/health-systems-development/data-reports-maps/maps/azpca.pdf.
https://doi.org/10.1371/journal.pone.0248476.g007
PLOS ONE | https://doi.org/10.1371/journal.pone.0248476 June 3, 2021
11 / 17
PLOS ONE NAS in Arizona
Table 2. Top ranked ICD9 and ICD10 codes associated with infants who have NOWS.
Rank ICD9 Code
ICD10 Code (and
Rank)
Description
1
2
3
4
5
6
7
8
9
10
11
16
779.5
779.31
P961 (1) Drug withdrawal syndrome in newborn
P92.1–2 and 8–9
(4)
Feeding problems in newborn
760.72
P04.49 (2) Hallucinogenic agents affecting fetus or newborn via placenta or breast
milk
P599 (5) Unspecified fetal and neonatal jaundice
L22 (3) Diaper or napkin rash
P22.1 (6) Transitory tachypnea of newborn
P04.9 (9�) Other noxious influences affecting fetus or newborn via placenta or
breast milk
P37.5 (7) Neonatal Candida infection
P002 (13) Observation for suspected infectious condition
Z23 (19) Need for prophylactic vaccination and inoculation against viral hepatitis
Z38.00 (8) Single liveborn, born in hospital, delivered without mention of cesarean
section (i.e. delivered vaginally)
Q21.1 (10) Atrial septal defect
7746
6910
770.6
760.79
771.7
V290
V05.3
V30.00
745.4 and
745.6
ICD10 codes were used for ranking infants admitted after Oct. 1 2015, as well as for any infants born in health care
facilities that adopted ICD10 codes prior to Oct. 1, 2015. Conditions are in the ICD9 rank order (see first column),
and the corresponding ICD10 and rank of the ICD10 code are listed. The increase from rank 11 to rank 16 was
allowed so that all top 10 ICD10 codes could be shown.
https://doi.org/10.1371/journal.pone.0248476.t002
recent years, with so many recovering substance users coming to the area from outside that
the trend has been reported in state and national news outlets [51, 52]. Low numbers of medi-
cation assisted treatment providers in areas with the highest rates of opioid use and NOWS
cases are also potentially responsible for higher than expected rates in some, particularly rural,
areas of the state [53].
Limitations
This analysis of hospital discharge records is limited by reliance on secondary data for which
reporting may be inconsistent. We expect that heightened surveillance implemented during
the study period may have improved consistency of reporting, however, we can only comment
on what is recorded at discharge. Further, length of stay and total charges can vary based on
treatment approaches, early identification, and hospital policies. It is outside the scope of this
analysis to determine what approaches were used in Arizona NOWS cases or what impact
those practices had on NOWS cases in Arizona.
Tribal PCAs were not included in this analysis due to data restrictions and non-reporting.
The exclusion of Tribal PCAs limits characterization of racial/ethnic demographics and may
not include accurate accounting of a Native American/Alaska Native populations who experi-
ence the highest incidence rates of NOWS nationwide [54].
Conclusions
The comorbidity analysis, using supervised machine learning, revealed that diaper dermatitis
and jaundice were more importantly associated with NOWS than traditional conditions of
respiratory distress and irritability. Additionally, to our knowledge, neonatal candidiasis has
PLOS ONE | https://doi.org/10.1371/journal.pone.0248476 June 3, 2021
12 / 17
PLOS ONE NAS in Arizona
Table 3. Top ranked ICD9 and ICD10 codes associated with mothers who are opioid dependent at the time of giving birth.
Rank
ICD9
Code
ICD10 Code (and Rank) Description
1
2
3
4
5
6
7
8
9
10
11
12
304
F11.20, F11.90, F1110 (1,3,4) Opioid type dependence, unspecified
648.31
648.41
649.01
338.29
V23.7
644.21
648.91
O99.324 (2) Drug dependence of mother, delivered, with or without mention of antepartum condition
O99.344 (8) Mental disorders of mother, delivered, with or without mention of antepartum condition
O99.334 and F17.210 (5,6) Tobacco use disorder complicating pregnancy, childbirth, or the puerperium, delivered, with or without mention
of antepartum condition
G89.29 (9) Other chronic pain
O09.30 (7) Supervision of high-risk pregnancy with insufficient prenatal care
O60.14X0 (57) Early onset of delivery, delivered, with or without mention of antepartum condition
No specific conversion Other current conditions classifiable elsewhere of mother, delivered, with or without mention of antepartum
condition
647.61 O98.42, 098.511–513, O98.52
(31)
Other viral diseases in the mother, delivered, with or without mention of antepartum condition
300
664.11
F41.9 (15) Anxiety state, unspecified
O70.1 (14) Second-degree perineal laceration, delivered, with or without mention of antepartum condition
656.51 O36.5110–130, O36.5910–30
(26)
Poor fetal growth, affecting management of mother, delivered, with or without mention of antepartum condition
62
305.70–72
F15.10 (10) Amphetamine or related acting sympathomimetic abuse (unspecified, continuous, or episodic)
ICD10 codes were used for ranking mothers admitted after Oct. 1 2015, as well as for any infants born in health care facilities that adopted ICD10 codes prior to Oct. 1,
2015. Conditions are in the ICD9 rank order (see first column), and the corresponding ICD10 and rank of the ICD10 code are listed. The increase from rank 12 to rank
62 was allowed so that the top 10 ICD10 codes could be shown.
https://doi.org/10.1371/journal.pone.0248476.t003
not been previously associated with NOWS. However, our results, coupled with the finding
that pregnant women receiving medication assisted opioid treatment are more frequently col-
onized with Candida [35], suggests that it may be important to screen pregnant women receiv-
ing medication assisted opioid treatment for candidiasis to develop maternal and neonatal
treatment strategies. Our analysis of the Arizona Hospital Discharge Database from 2010 to
2017 suggests a need for better characterization of comorbid conditions in NOWS neonates
and their mothers. Variability of comorbid conditions in NOWS neonates suggest that much
more detailed understanding of contributors to and symptoms of NOWS could begin to
address challenges in standardization of care. Efforts to improve treatment as well as primary
and secondary prevention of NOWS may benefit from better characterization of comorbid
and frequently co-occurring conditions. Future research should also consider potential inter-
actions between opioid exposure and comorbid conditions in utero.
Supporting information
S1 Table. Summary of non-geographic data used in this study by figure. The MothersO col-
umn contains data for all mothers who were dependent on opioids at the time of giving birth.
The InfantsNOWS column contains data for all infants who had NOWS. Totals at the end of each
category are not the same for all categories due to missing data or because some patients reported
atypical categories (e.g. payer was workers compensation or individual was a foreign national).
(DOCX)
Acknowledgments
The authors acknowledge the contributions of Arizona Department of Health Services staff
members Kyle Gardner and Timothy Flood for assistance defining the data request, the
PLOS ONE | https://doi.org/10.1371/journal.pone.0248476 June 3, 2021
13 / 17
PLOS ONE NAS in Arizona
Northern Arizona University Information Technology Services for managing data security
issues, and the Southwestern Health Equity Research Collaborative Research Infrastructure
Core for analytical assistance.
Author Contributions
Conceptualization: Emery R. Eaves, Jarrett Barber, Sara A. Clancey, Joseph Spadafino, Crystal
M. Hepp.
Data curation: Ryann Whealy, Joseph Spadafino, Crystal M. Hepp.
Formal analysis: Jarrett Barber, Ryann Whealy, Sara A. Clancey, Jill Hager Cocking, Joseph
Spadafino, Crystal M. Hepp.
Funding acquisition: Emery R. Eaves, Sara A. Clancey, Crystal M. Hepp.
Investigation: Emery R. Eaves, Jarrett Barber, Sara A. Clancey, Rita Wright, Jill Hager Cock-
ing, Crystal M. Hepp.
Methodology: Jarrett Barber, Rita Wright, Crystal M. Hepp.
Project administration: Crystal M. Hepp.
Supervision: Crystal M. Hepp.
Visualization: Crystal M. Hepp.
Writing – original draft: Emery R. Eaves, Jarrett Barber, Sara A. Clancey, Rita Wright, Crystal
M. Hepp.
Writing – review & editing: Emery R. Eaves, Jarrett Barber, Ryann Whealy, Sara A. Clancey,
Rita Wright, Jill Hager Cocking, Joseph Spadafino.
References
1. Gomez-Pomar E, Finnegan LP. The Epidemic of Neonatal Abstinence Syndrome, Historical Refer-
ences of Its’ Origins, Assessment, and Management. Frontiers in Pediatrics. 2018; 6:33. https://doi.org/
10.3389/fped.2018.00033 PMID: 29520355
2. Kocherlakota P. Chapter 19—Pharmacologic Therapy for Neonatal Abstinence Syndrome. In: Benitz
WE, Smith PB, editors. Infectious Disease and Pharmacology. Philadelphia: Content Repository
Only!; 2019. p. 243–59.
3. Coyle MG, Brogly SB, Ahmed MS, Patrick SW, Jones HE. Neonatal abstinence syndrome. Nature
Reviews Disease Primers. 2018; 4(1):47. https://doi.org/10.1038/s41572-018-0045-0 PMID: 30467370
4. McCarthy JJ, Leamon MH, Finnegan LP, Fassbender C. Opioid dependence and pregnancy: minimiz-
ing stress on the fetal brain. American Journal of Obstetrics and Gynecology. 2017; 216(3):226–31.
https://doi.org/10.1016/j.ajog.2016.10.003 PMID: 27729254
5. Bersani I, Corsello M, Mastandrea M, Patacchiola V, Foligno S, Garofalo V, et al. Neonatal abstinence
syndrome. Early Human Development. 2013; 89:S85–S7. https://doi.org/10.1016/S0378-3782(13)
70112-9
6. Corr TE, Hollenbeak CS. The economic burden of neonatal abstinence syndrome in the United States.
Addiction. 2017; 112(9):1590–9. https://doi.org/10.1111/add.13842 PMID: 28612362
7. Villapiano NL, Winkelman TN, Kozhimannil KB, Davis MM, Patrick SW. Rural and urban differences in
neonatal abstinence syndrome and maternal opioid use, 2004 to 2013. JAMA pediatrics. 2017; 171
(2):194–6. https://doi.org/10.1001/jamapediatrics.2016.3750 PMID: 27942711
8. Brown J, Nolte J. Expanded buprenorphine prescribing privileges: Sandbagging in the midst of the
flood? Res Social Adm Pharm. 2018 Jun; 14(6):617–618. https://doi.org/10.1016/j.sapharm.2018.03.
001 PMID: 29525482
9. Brown JD, Goodin AJ, Talbert JC. Rural and Appalachian disparities in neonatal abstinence syndrome
incidence and access to opioid abuse treatment. The Journal of rural health. 2018; 34(1):6–13. https://
doi.org/10.1111/jrh.12251 PMID: 28685864
PLOS ONE | https://doi.org/10.1371/journal.pone.0248476 June 3, 2021
14 / 17
PLOS ONE NAS in Arizona
10. Admon LK, Bart G, Kozhimannil KB, Richardson CR, Dalton VK, Winkelman TN. Amphetamine-and
Opioid-Affected Births: Incidence, Outcomes, and Costs, United States, 2004–2015. Am J Public
Health. 2018 Nov 29:e1–e7. https://doi.org/10.2105/AJPH.2018.304771 PMID: 30496001
11. Winkelman TN, Villapiano N, Kozhimannil KB, Davis MM, Patrick SW. Incidence and costs of neonatal
abstinence syndrome among infants with Medicaid: 2004–2014. Pediatrics. 2018; 141(4):e20173520.
https://doi.org/10.1542/peds.2017-3520 PMID: 29572288
12. Sanlorenzo LA, Stark AR, Patrick SW. Neonatal abstinence syndrome: an update. Current opinion in
pediatrics. 2018; 30(2):182. https://doi.org/10.1097/MOP.0000000000000589 PMID: 29346142
13. Koppell J, Daugherty DB, Garcia J, Whitsett A. Arizona’s Vulnerable Population. A report prepared by
Arizona State University. 2014. https://aztownhall.org/Resources/Documents/AZTwnHll-0414-FNL-LR.
pdf
14. Hussaini KS, Saavedra LFG. Neonatal Abstinence Syndrome (NOWS) in Southwestern Border States:
Examining Trends, Population Correlates, and Implications for Policy. Matern Child Health J. 2018 Sep;
22(9):1352–1359. https://doi.org/10.1007/s10995-018-2517-y PMID: 29572587
15. National Institute on Drug Abuse. Arizona Opioid Summary. 2018. Retrieved from https://www.
drugabuse.gov/drugs-abuse/opioids/opioid-summaries-by-state/arizona-opioid-summary.
16. Arizona Department of Health Services. 2016 Arizona opioid report. Retrieved from https://www.azdhs.
gov/documents/audiences/clinicians/clinical-guidelines-recommendations/prescribing-guidelines/
arizona-opioid-report.pdf.
17. Patrick SW, Schumacher RE, Benneyworth BD, Krans EE, McAllister JM, Davis MM. Neonatal absti-
nence syndrome and associated health care expenditures: United States, 2000–2009. JAMA. 2012 9;
307(18):1934–40. PMID: 22546608
18. Arizona Department of Health Services. Opioid Epidemic. 2019. Retrieved from https://www.azdhs.gov/
prevention/womens-childrens-health/injury-prevention/opioid-prevention/index.php.
19. Stabler ME, Long DL, Chertok IR, Giacobbi PR Jr, Pilkerton C, Lander LR. Neonatal Abstinence Syn-
drome in West Virginia Substate Regions, 2007–2013. J Rural Health. 2017 Jan; 33(1):92–101. https://
doi.org/10.1111/jrh.12174 PMID: 26879950
20. DATA DOCUMENTATION: Sources and Field Descriptions (To accompany Primary Care Area (PCA)
or Special Area Statistical Profiles) Arizona Department of Health Services: Prevention Services, May
24, 2016.
21.
Ishwaran H, Kogalur UB, Gorodeski EZ, Minn AJ, Lauer MS. High-dimensional variable selection for
survival data. Journal of the American Statistical Association. 2010 Mar 1; 105(489):205–17. https://doi.
org/10.1198/jasa.2009.tm08622
22. R Core Team. R: A Language and Environment for Statistical Com-puting. R Foundation for Statistical
Computing, Vienna, Austria,2019.
23.
Ishwaran, H, Kogalur, UB. Fast Unified Random Forests for Survival, Regression, and Classification
(RF-SRC), 2020. R packageversion 2.9.3.
24. Chen, C, Breiman, L. Tech Report: Using random forest to learn imbalanced data. University of Califor-
nia, Berkeley, 2004.
25. Atwell KA, Weiss HB, Gibson C, Miller R, Corden TE. Neonatal Abstinence Syndrome and Maternal
Substance Use in Wisconsin, 2009–2014. WMJ: official publication of the State Medical Society of Wis-
consin. 2016; 115(6):287–94. Epub 2017/11/03. PMID: 29094858.
26.
Filippelli AC, White LA, Spellman LW, Broderick M, Highfield ES, Sommers E, et al. Non-insertive Acu-
puncture and Neonatal Abstinence Syndrome: A Case Series From an Inner-city Safety Net Hospital.
Glob Adv Health Med. 2012; 1(4):48–52. https://doi.org/10.7453/gahmj.2012.1.4.007 PMID: 24078899
27. Maalouf FI, Cooper WO, Stratton SM, Dudley JA, Ko J, Banerji A, et al. Positive predictive value of
administrative data for neonatal abstinence syndrome. Pediatrics. 2019 Jan 1; 143(1):e20174183.
https://doi.org/10.1542/peds.2017-4183 PMID: 30514781
28.
Jansson LM, Velez M, Harrow C. The opioid-exposed newborn: assessment and pharmacologic man-
agement. J Opioid Manag. 2009; 5(1):47–55. PMID: 19344048
29. Kaltenbach K, Finnegan L. Neonatal abstinence syndrome. Opioid-Use Disorders in Pregnancy: Man-
agement Guidelines for Improving Outcomes. Wright TE, Ed. Cambridge University Press. 2018.
30. Azuine RE, Ji Y, Chang H, et al. Prenatal Risk Factors and Perinatal and Postnatal Outcomes Associ-
ated With Maternal Opioid Exposure in an Urban, Low-Income, Multiethnic US Population. JAMA Netw
Open. 2019; 2(6):e196405. https://doi.org/10.1001/jamanetworkopen.2019.6405 PMID: 31251378
31. Bagley SM, Wachman EM, Holland E, Brogly SB. Review of the assessment and management of neo-
natal abstinence syndrome. Addiction science & clinical practice. 2014; 9(1):19. https://doi.org/10.1186/
1940-0640-9-19 PMID: 25199822
PLOS ONE | https://doi.org/10.1371/journal.pone.0248476 June 3, 2021
15 / 17
PLOS ONE NAS in Arizona
32. Kelly LE, Jansson LM, Moulsdale W, Pereira J, Simpson S, Guttman A, et al. A core outcome set for
neonatal abstinence syndrome: study protocol for a systematic review, parent interviews and a Delphi
survey. Trials. 2016; 17(1):536. https://doi.org/10.1186/s13063-016-1666-9 PMID: 27821184
33. Patrick SW, Schumacher RE, Horbar JD, Buus-Frank ME, Edwards EM, Morrow KA, et al. Improving
care for neonatal abstinence syndrome. Pediatrics. 2016; 137(5):e20153835. https://doi.org/10.1542/
peds.2015-3835 PMID: 27244809
34. Burnette T, Chernicky L, Towers CV. The effect of standardizing treatment when managing neonatal
abstinence syndrome. J Matern Fetal Neonatal Med. 2019 Oct; 32(20):3415–3419. https://doi.org/10.
1080/14767058.2018.1465038 PMID: 29757685
35.
Farr A, Kiss H, Hagmann M, et al. Evaluation of the vaginal flora in pregnant women receiving opioid
maintenance therapy: a matched case-control study. BMC Pregnancy Childbirth. 2016; 16(1):206. Pub-
lished 2016 Aug 5. https://doi.org/10.1186/s12884-016-1003-z PMID: 27495167
36. Bagwell GA, Thomas A, Ryshen G. Improving skin integrity in babies diagnosed with neonatal absti-
nence syndrome. Neonatal Network, Academic of Neonatal Nursing. 2016 Sep 1; 35(5):1–7. https://doi.
org/10.1891/0730-0832.35.S1.1 PMID: 29037278
37. Chisolm MS, Fitzsimons H, Leoutsakos JM, Acquavita SP, Heil SH, Wilson-Murphy M, et al. A compari-
son of cigarette smoking profiles in opioid-dependent pregnant patients receiving methadone or bupre-
norphine. Nicotine Tob Res. 2013 Jul; 15(7):1297–304. PMID: 23288871
38.
Jones HE, Heil SH, O’Grady KE, Martin PR, Kaltenbach K, Coyle MG, et al. Smoking in pregnant
women screened for an opioid agonist medication study compared to related pregnant and non-preg-
nant patient samples. Am J Drug Alcohol Abuse. 2009; 35(5):375–80. https://doi.org/10.1080/
00952990903125235 PMID: 20180667
39. Akerman SC, Brunette MF, Green AI, Goodman DJ, Blunt HB, Heil SH. Treating tobacco use disorder
in pregnant women in medication-assisted treatment for an opioid use disorder: a systematic review. J
Subst Abuse Treat. 2015; 52:40–47. https://doi.org/10.1016/j.jsat.2014.12.002 PMID: 25592332
40. Smith MV, Costello D, Yonkers KA. Clinical Correlates of Prescription Opioid Analgesic Use in Preg-
nancy. Matern Child Health J 2015; 19:548–556. https://doi.org/10.1007/s10995-014-1536-6 PMID:
24951127
41. Prabhu M, Garry EM, Hernandez-Diaz S, MacDonald SC, Huybrechts KF, Bateman BT. Frequency of
opioid dispensing after vaginal delivery. Obstetrics and gynecology. 2018 Aug; 132(2):459–65. https://
doi.org/10.1097/AOG.0000000000002741 PMID: 29995732
42. Robinson JN, Norwitz ER, Cohen AP, McElrath TF, Lieberman ES. Epidural analgesia and third-or
fourth-degree lacerations in nulliparas. Obstetrics & Gynecology. 1999 Aug 1; 94(2):259–62.
43. Hoyt MR, Shah U, Cooley J, Temple M. Use of epidural clonidine for the management of analgesia in
the opioid addicted parturient on buprenorphine maintenance therapy: an observational study. Interna-
tional journal of obstetric anesthesia. 2018 May 1; 34:67–72. https://doi.org/10.1016/j.ijoa.2018.01.001
PMID: 29486974
44. Cassidy B, Cyna AM. Challenges that opioid-dependent women present to the obstetric anaesthetist.
Anaesthesia and intensive care. 2004 Aug; 32(4):494–501. https://doi.org/10.1177/
0310057X0403200406 PMID: 15675209
45. Merline AC, O’Malley PM, Schulenberg JE, Bachman JG, Johnston LD. Substance use among adults
35 years of age: prevalence, adulthood predictors, and impact of adolescent substance use. Am J Pub-
lic Health. 2004; 94(1):96–102. https://doi.org/10.2105/ajph.94.1.96 PMID: 14713705
46. Moos RH, Nichol AC, Moos BS. Risk factors for symptom exacerbation among treated patients with
substance use disorders. Addiction. 2002; 97(1):75–85. https://doi.org/10.1046/j.1360-0443.2002.
00063.x PMID: 11895273
47. Beattie MC, Longabaugh R. Interpersonal factors and post-treatment drinking and subjective wellbeing.
Addiction. 1997; 92(11):1507–21. PMID: 9519493
48. Havassy BE, Wasserman DA, Hall SM. Social relationships and abstinence from cocaine in an Ameri-
can treatment sample. Addiction. 1995; 90(5):699–710. PMID: 7795506
49. Dobkin PL, Civita MD, Paraherakis A, Gill K. The role of functional social support in treatment retention
and outcomes among outpatient adult substance abusers. Addiction. 2002; 97(3):347–56. https://doi.
org/10.1046/j.1360-0443.2002.00083.x PMID: 11964111
50. Heinz AJ, Wu J, Witkiewitz K, Epstein DH, Preston KL. Marriage and relationship closeness as predic-
tors of cocaine and heroin use. Addictive Behaviors. 2009; 34(3):258–63. https://doi.org/10.1016/j.
addbeh.2008.10.020 PMID: 19008050
51. Segal, D, City of Addict Entrepreneurs. December, 2017. The New York Times. https://kjzz.org/content/
287185/prescott-grappling-costs-becoming-national-destination-addiction-recovery;
PLOS ONE | https://doi.org/10.1371/journal.pone.0248476 June 3, 2021
16 / 17
PLOS ONE NAS in Arizona
52. Stone, W. “Prescott Grappling With Costs Of Becoming National Destination For Addiction Recovery”
KJZZ, Rio Salado College Public Radio. April, 2016. https://www.nytimes.com/interactive/2017/12/27/
business/new-drug-rehabs.html
53. Cleveland, C. Pregnant and Addicted: Opioid deaths in Arizona nearly doubled over the past ten years.
The number of babies born exposed to opioids grew more than twice as fast. Arizona Center for Investi-
gative Reporting. January, 2018. https://azcir.org/news/2018/01/26/arizona-opioid-epidemic-grew-
twice-as-fast-for-pregnant-mothers-newborns-babies/
54. Strahan AE, Guy GP, Bohm M, Frey M, Ko JY. Neonatal Abstinence Syndrome Incidence and Health
Care Costs in the United States, 2016. JAMA pediatrics. 2020; 174(2):200–2. https://doi.org/10.1001/
jamapediatrics.2019.4791 PMID: 31841581
PLOS ONE | https://doi.org/10.1371/journal.pone.0248476 June 3, 2021
17 / 17
PLOS ONE |
10.1371_journal.pone.0243209 | RESEARCH ARTICLE
Individual differences in emotion regulation
and face recognition
Ahmed M. MegreyaID
1*, Robert D. Latzman2
1 Department of Psychological Sciences, College of Education, Qatar University, Doha, Qatar, 2 Department
of Psychology, Georgia State University, Atlanta, GA, United States of America
* amegreya@qu.edu.qa
Abstract
Face recognition ability is highly variable among neurologically intact populations. Across
three experiments, this study examined for the first time associations between individual dif-
ferences in a range of adaptive versus maladaptive emotion regulation strategies and face
recognition. Using an immediate face-memory paradigm, in which observers had to identify
a self-paced learned unfamiliar face from a 10-face target-present/ target-absent line-up,
Experiment 1 (N = 42) found high levels of expressive suppression (the ongoing efforts to
inhibit emotion-expressive behaviors), but not cognitive reappraisal (the cognitive re-evalua-
tion of emotional events to change their emotional consequences), were associated with a
lower level of overall face-memory accuracy and higher rates of misidentifications and false
positives. Experiment 2 (N = 53) replicated these finding using a range of face-matching
tasks, where observers were asked to match pairs of same-race or different-race face
images taken on the same day or during different times. Once again, high levels of expres-
sive suppression were associated with a lower level of overall face-matching performance
and higher rates of false positives, but cognitive reappraisal did not correlate with any face-
matching measure. Finally, Experiment 3 (N = 52) revealed that the higher use of maladap-
tive cognitive emotion regulation strategies, especially catastrophizing, was associated with
lower levels of overall face-matching performances and higher rates of false positives. All
told, the current research provides new evidence concerning the important associations
between emotion and cognition.
Introduction
Face recognition ability is highly variable among neurologically intact populations [e.g., for
reviews see 1–3]. On the one hand, individuals with developmental prosopagnosia have severe
face recognition deficits in the absence of any brain damages [for reviews, see 4, 5]. On the
other hand, individuals with extra-ordinary face recognition ability, oftentimes called super-
recognizers, are able to perform rather challenging face recognition tasks with extremely high
levels of accuracy [e.g., for reviews see 6, 7]. Between these two extremes, face recognition abil-
ity of the vast majority of neurologically intact individuals is distributed along this spectrum.
Across a set of studies involving a total of 400 participants, Woodhead and Baddeley (1981)
noticed that d’, the sensitivity index in the signal detection theory, ranges from 0.5 to 6.8 using
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: Megreya AM, Latzman RD (2020)
Individual differences in emotion regulation and
face recognition. PLoS ONE 15(12): e0243209.
https://doi.org/10.1371/journal.pone.0243209
Editor: Peter James Hills, Bournemouth University,
UNITED KINGDOM
Received: February 28, 2020
Accepted: November 18, 2020
Published: December 10, 2020
Copyright: © 2020 Megreya, Latzman. This is an
open access article distributed under the terms of
the Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: All relevant data are
within the paper and its Supporting Information
files.
Funding: This report was made possible by an
UREP grant # (UREP26-008-5-003) from the Qatar
national research fund (a member of Qatar
foundation) to Professor Ahmed M. Megreya. The
statements made herein are solely the
responsibility of the authors. The funders had no
role in study design, data collection and analysis,
decision to publish, or preparation of the
manuscript.
PLOS ONE | https://doi.org/10.1371/journal.pone.0243209 December 10, 2020
1 / 26
PLOS ONE Competing interests: The authors have declared
that no competing interests exist.
Emotion regulation and face recognition
a facial recognition memory task (making old/new decisions for a set of previously studied
faces or distractors) [8]. These wide individual differences were also noticed using face percep-
tion tasks, which do not rely on memory. For example, face-matching performance has gener-
ally been found to range between 50% to 96% accuracy [9] using an array task (matching a
target unfamiliar face to a 10-face target-present/ target-absent lineup [10]. Further, using the
Glasgow Face Matching Test (GFMT), a test that asks participants to match the identity of
pairs of unfamiliar faces, individual performance ranges along a broad continuum from just
above chance to perfect [11]. Importantly, not only are these individual differences reliably
found across studies and tasks, they appear to be stable [12], unrelated to general intelligence
[13], and highly heritable [14]. Therefore, understanding the processes associated with the
individual differences in face recognition is a topic of interest [e.g., for reviews see 1–3].
Demographic-related differences in face recognition
The vast majority of studies that have considered contributors to variation in facial recognition
ability have largely focused on demographic characteristics of participants and face stimuli
including race, age, and gender. For example, an own-race advantage has been well-docu-
mented such that individuals recognize faces belonging to their own-race more accurately
than those belonging to other races [15–18]. In addition, developmental studies have reported
that face recognition improves with development before deteriorating with old age [19, 20].
Furthermore, gender differences have been observed in face recognition, with an own-gender
bias reliably found, especially for females [21–24]. Importantly, however, despite these demo-
graphic-related differences, face recognition ability still remarkably varies widely, even within
demographically-homogeneous groups of individuals (i.e., within individuals belonging to the
same race, age, and gender).
Individual differences among face recognition tasks
Even within demographically-homogeneous groups, a great deal of variation in face recogni-
tion abilities exists. With this in mind, a general face recognition factor (termed f) has been
proposed in order to explain this variation. For example, Verhallen et al. (2017) suggested that
individuals who are good in a task measuring specific aspects of face perception are also good
in other tasks measuring different aspects of face memory [25]. Specifically, Verhallen et al.
(2017) reported positive inter-correlations among three standardized tests measuring different
aspects of face recognition [25]. These included the GFMT, the Cambridge Face Memory Test
(CFMT), a measure of face memory that requires participants to recognize sets of previously
learned faces through three-alternative forced choice tasks, and the Mooney Face Test, a mea-
sure of face closure (a perceptual tendency to view incomplete objects as complete) that
requires observers to identify the gender of faces using high-contrast images consisting of
exclusively dark or light regions [25].
Similar findings have also been reported by McCaffery, Robertson, Young, and Burton
(2018) who found positive inter-correlations among the GFMT, the CFMT, and the Before
They Were Famous task (BTWF), a measure of familiar face recognition that requires observ-
ers to recognize a set of celebrities using photos taken before they became famous; when they
were children or adolescents [26]. Robertson, Black, Chamberlain, Megreya and Davis (2020)
also found positive and relatively strong inter-correlations between matching and memory
tasks, which included same-race and other-race faces [27]. Stacchi, Huguenin-Elie, Caldara,
and Ramon (2020) reported similarly strong positive correlations among a variety of face rec-
ognition tasks [28].
PLOS ONE | https://doi.org/10.1371/journal.pone.0243209 December 10, 2020
2 / 26
PLOS ONE Emotion regulation and face recognition
Further, matching upright unfamiliar faces has been found to positively correlate with rec-
ognition memory [9], immediate memory (identifying a learned face through a subsequent
target-present/ target-absent 10 face line-up) [29], eye-witness identification (identifying a cul-
prit eye-witnessed during a staged crime through a target-present/ target-absent line-up) [30],
and matching inverted familiar and unfamiliar faces [9]. Further underscoring the replicabil-
ity, across assessment modalities, of positive associations among distinct face-related tasks,
performance on the CFMT positively has been found to correlate with the fast periodic visual
stimulation (FPVS) paradigm, an objective EEG index of individual face discrimination in the
right occipitotemporal cortex [31].
Individual differences in face recognition and visual processing
In addition to face-related abilities more specifically, individual differences in face recognition
also appear to co-vary with variation in visual processing abilities more generally. For example,
almost forty-years-ago, Woodhead and Baddeley (1981) found that people who were good at
recognizing faces were also good at recognizing other non-face visual objects [8]. More
recently, Megreya and Burton (2006) found that performance on the 1-in-10 face matching
task positively correlated with visual short term memory, perceptual speed (Finding A’s and
Identical Picture Tests), and Matching Familiar Figure Test, a measure of object matching that
requires participants to match a target line drawing of common objects to a line-up of six
minor variants [9]. Similarly, performance on the GFMT positively correlated with the Match-
ing Familiar Figure Test [11, 26] and with the Navon local processing task, a measure of parts
perception that require observers to identify the identity of the small parts of the compound
letters [26] as has performance on CFMT and the Cambridge Car Memory Test have also been
found to be positively correlated with each other [32]. Performance on recognition memory
for faces and visual processing speed in infants [33] and with visual perceptual speed in 11
year-old children [34] have similarly been found to positively associate with each other.
Although studies have consistently found face recognition to be affected by several visual
processing skills, some studies have reported some distinctions between these two domains.
For example, McCaffery et al. (2018) found that performances on the GFMT, CFMT and
BTWF did not correlate with performances on a range of visual perception tasks including
position discrimination (identifying which squire includes a more precisely central dot), posi-
tion of gap (identifying whether the gaps in two circles are in the same or different positions),
letter detection (detecting the letter “f” from a passage), and Navon global processing (identify-
ing the identity of the large compound letters) [26]. Consistently, Wilhelm et al (2010) found
that found that individual differences in three main aspects of face recognition (face percep-
tion, face memory, and speed of face cognition) could not be reduced to individual differences
in immediate and delayed memory, general cognitive ability, mental speed, and object recogni-
tion [13]. Therefore, although face recognition is correlated with some visual processing tasks
[9, 32], a consistent empirical literature provides support for the face-specific domain theory
[35] suggesting that faces are processed through cognitive and neurological mechanisms that
are not involved in object recognition [13].
Individual differences in face recognition and personality and emotions
Intriguingly, although findings have been mixed, individual differences in face recognition
seem to also relate to other “non-visual” processes including personality and emotions. For
example, some studies have found that observers who exhibit high levels of extraversion and
emotional stability are more accurate at face recognition than individuals with lower levels of
these traits [36, 37]. Cheung, Rutherford, Mayes, and McPartland (2010) further found that
PLOS ONE | https://doi.org/10.1371/journal.pone.0243209 December 10, 2020
3 / 26
PLOS ONE Emotion regulation and face recognition
those with higher in extraversion exhibited a discrepant N170 neurophysiological event-
related potential amplitude, a face-specific brain electrophysiology component, associated
with face inversion that was more prominent in the right hemisphere [38]. However, whereas
neuroticism (i.e., low emotional stability) appears to be associated with a reliable negative
impact on face identification [39], results from studies on extraversion and related traits (e.g.,
shyness [reverse-keyed] have been more mixed [40–43]. For example, Megreya and Binde-
mann (2013) examined the relationship between individual differences in performance in the
1-in-10 face matching task and a range of personality factors and found that correct face iden-
tifications related to low anxiety, low tension, and high emotional stability (i.e., low neuroti-
cism) [44]. Lander and Poyarekar (2015) examined the relationship among upright/ inverted
familiar face recognition (naming a set of British and American celebrities), the GFMT, and a
brief measure personality revealed only one significant association: extraversion positively cor-
related with recognition of familiar faces when presented upright but not upside down; no
other associations between face-recognition and personality emerged [45]. Further complicat-
ing the picture, McCaffery et al (2018) more recently found no relationship between the
GFMT, CFMT and BTWF and personality [26].
One potential explanation for the mixed literature concerning associations between indi-
vidual differences traits and face recognition may be variation in not only traits (i.e., general
tendencies to experience various emotions) but also the way in which individuals regulate
those emotions. Indeed, emotions generally, and emotion regulation more specifically, may
play some roles on individual differences in face recognition. In support of this possibility, an
early study reported that participants in a moderate arousal condition had higher face recogni-
tion ability than those in the high arousal condition [46]. More recently, Hills et al (2019)
examined how being observed affects face recognition and found that being observed during
learning, but not during the test phase, impaired recognition accuracy using an old/new recog-
nition memory and eyewitness identification paradigms [47]. Although being observed was
associated with increased physiological arousal as indexed by galvanic skin response and heart
rate, these authors found that this heightened arousal did not explain the detriments in face
recognition [47]. Nevertheless, face recognition has been found to negatively associate with
generalized anxiety [44, 48–51]; but for an inconsistent finding see [52] as well as social anxiety
more specifically [53]. In addition, observers scoring high on socio-emotional empathy also
appear more accurate at face recognition than people who display lower levels of empathy
[54]. Furthermore, the Matching Familiar Figures Test, which was initially developed as a mea-
sure of impulsiveness in children, has been shown to predict accuracy in face matching tasks
[9, 11].
To summarize, there are wide individual differences in face recognition [e.g., for reviews
see 1–3] and these individual differences are stable across different face perception and recog-
nition tasks [25–28]. There have been many factors identified in the literature that may help to
explain this variation including: (i) demographic characteristics including race [15–18], age
[19, 20], and gender [21, 24]; (ii) a range of visual cognitive abilities such as general memory
[8, 13, 32], object perception [11, 26], and perceptual speed [9, 33, 34]; (iii) personality traits
such as extraversion [36–38, 45], neuroticism/low emotional stability [36–39, 44], impulsive-
ness [9, 11, 26], and socio-emotional empathy [54]; and (iv) anxiety-related symptomatology
including generalized anxiety [44, 48–51], social anxiety [53], physiological arousal [47].
Emotions and emotion regulation
It is well-established that emotions can be substantially modified using a range of emotion reg-
ulation (ER) strategies [e.g., for an extensive review see 55] and some scholars have even
PLOS ONE | https://doi.org/10.1371/journal.pone.0243209 December 10, 2020
4 / 26
PLOS ONE Emotion regulation and face recognition
argued that both emotions and ER might be one process [56, 57]. ER has been conceptualized
as “the processes or strategies through which individuals can modulate or manage which emo-
tions they have, when they have them, and how these emotions are experienced and expressed”
[58]. The widely-used process model of ER [58, 59] includes two broader types of ER strategies:
1) antecedent-focused strategies, referring to things people do before emotional response ten-
dencies have completely activated and 2) response-focused strategies, referring to things peo-
ple do once emotions have already been generated. In addition to these two broad types of
strategies, more specific strategies are thought to fall along a temporal sequencing of emotion
regulation processes. In addition, Garnefski, Kraaij, and Spinhoven (2001) provide a second
widely-used model of ER in which a distinction is made between cognitive (e.g., making plans)
and behavioral (e.g., taking immediate action) ER strategies [60].
A meta-analysis found that anxiety, depression, eating, and substance-related disorders
were negatively associated with two adaptive ER strategies (problem-solving and reappraisal)
and positively associated with three less-adaptive ER strategies (rumination, avoidance, and
suppression) [61]. In addition to associating with various psychopathological outcomes, ER
strategies appear to also have cognitive correlates. For example, Richard and Gross (2000)
found that suppression impaired memory, but reappraisal had no effect [62]. Consistently,
subsequent studies reported associations between ER and working memory [63–65]. For
example, Schmeichel et al (2008) found that the successful suppression of facial expressions of
emotions negatively correlated with working memory capacity [65]. In addition, McRae et al
(2012) found a positive correlation between individual differences in reappraisal ability and
working memory capacity [64]. Therefore, some studies have found that training working
memory could improve ER strategies [66–68]. Notably, significant interactions between face
recognition and working memory have been consistently found in behavioral and neuro-phys-
iological experiments [69–71]. For example, activity in fusiform face area was found to modu-
late as a function of working memory load [69].
Current study
Given reliable associations between variation in emotions and face recognition [44, 46, 48–51,
53], the conceptual overlap of emotions and ER [56, 57], the effects of various ER strategies on
memory [62–65], and the associations between memory and face recognition [9, 32, 69–71], it
is reasonable to assume that ER strategies might influence face recognition. Surprisingly, how-
ever, no previous studies have explicitly investigated this assumption (e.g., for reviews see 1–
3]. Therefore, the present study aimed to examine the relationship between individual differ-
ences in ER and face recognition. This is the first study to examine how face memory (Experi-
ment 1) and face perception (Experiments 2 and 3) might be affected by a range of adaptive
and maladaptive ER strategies.
Experiment 1
Method
Participants. Forty-two undergraduate students (21 females and 21 males) from Qatar
University (Qatar) volunteered to participate in this experiment. Participant’s mean age was
20.1 years (SD = 2.1) and all had normal or corrected to normal vision and none had any his-
tory of psychopathology using a self-reported item asking about whether they had any current
or previous mental health problems. Ethical approval for participation in the three experi-
ments in this study was provided by Qatar University’s institutional review board (QU-IRB)
and all methods were administered in accordance with QU-IRB guidelines and regulations.
PLOS ONE | https://doi.org/10.1371/journal.pone.0243209 December 10, 2020
5 / 26
PLOS ONE Emotion regulation and face recognition
Written informed consent was obtained from all participants for being included in the
experiments.
Measures.
(1) Face immediate memory task. This task consists of 20 images of target faces
along with 40 corresponding 10-face target-present/ target-absent lineups. These stimuli were
taken from an Egyptian face database [72]. Target face images were stills captured from a high-
quality digital camcorder, whereas all lineup images were photographs taken from a high-qual-
ity digital camera. The target and line-up images were taken on the same day under the same
lighting conditions, and each image showed a very similar full-face pose of young, clean-
shaven Egyptian young men who were undergraduate students in an Egyptian university.
Notably, only male faces were used in this study as the vast majority of females in Arab coun-
tries are wearing headscarves, which have particular influences on face perception and recog-
nition [16, 73]. All images were shown in grey-scale and the size of each face was
approximately 5–7 cm. Full details about the construction of this task can be found in Megreya
and Burton (2008) [72].
Participants were tested individually in a session of approximately 10 minutes. On each
trial, they were shown (i) a target face; (ii) an intervening interval of 5 seconds; (iii) a 10 face
target-present or target-absent lineup. Fig 1 shows a schematic representation of these proce-
dures. The individuals pictured in Figs 1 and 2 have provided written informed consent (as
outlined in PLOS consent form) to publish their image alongside the manuscript. There was
no time limit for studying the targets. Consistent with previous studies [29, 72], participants
were instructed to study each target until they felt confident that they could recognized him in
a subsequent 10-face lineup test. After a 5-second gap, participants were instructed that the
face they had just seen might or might not be present in the lineup.
Each participant completed 20 trials: 10 target-present and 10 target-absent. The presence
of targets was counter-balanced across the experiment so that each target appeared equally
often in target-present and target-absent lineups. Using an answer sheet, participants were
asked to write down the number of the face in the lineup or mark X if they decided that the tar-
get was not present.
(2) Emotion Regulation Questionnaire (ERQ) [74]. Gross and John (2003) developed the
Emotion Regulation Questionnaire (ERQ), an instrument designed to measure the two most
commonly used ER strategies: cognitive reappraisal (antecedent-focused) and expressive sup-
pression (response-focused) [74]. Cognitive reappraisal is defined as “a form of cognitive
change that involves construing a potentially emotion-eliciting situation in a way that changes
its emotional impact”, whereas expressive suppression is defined as “a form of response modu-
lation that involves inhibiting ongoing emotion-expressive behavior” [74, p. 349]. The ERQ is
a 10-item self-report measure of cognitive reappraisal (6 items) and expressive suppression (4
items), using a 7-point Likert-type scale ranging from 1 (strongly disagree) to 7 (strongly agree).
Gross and John (2003) reported moderate internal reliabilities for cognitive reappraisal (α =
0.79) and expressive suppression (α = 0.73), with no inter-correlation (r = – 0.01). ERQ has
dominated the ER literature and has been translated into many different languages including
Arabic [75]. Across four Arab countries (Egypt, Kuwait, Qatar, and Kingdom of Saudi Arabia),
Cronbach alpha reliability rates for cognitive reappraisal and expressive suppression were
acceptable to good, ranging from 0.75 to 0.85 [75].
Results
In the face immediate memory task, given previous findings that performances on these trials
are dissociable [9, 29, 72], participant’s responses to target-present versus target-absent arrays
were assessed separately. In target-present trials, we measured hits (the correct identification
PLOS ONE | https://doi.org/10.1371/journal.pone.0243209 December 10, 2020
6 / 26
PLOS ONE Emotion regulation and face recognition
Fig 1. A schematic representation of the unfamiliar immediate memory task.
https://doi.org/10.1371/journal.pone.0243209.g001
of the target face), misses (the incorrect decision that the target was absent), and misidentifica-
tion (the identification of a distractor face). In target-absent trials, false positives (the incorrect
decision that the target was present) were calculated. In addition, we report the overall accu-
racy by combining hits and correct rejection (the complement of false positives). For the ERQ,
we report the averages of individuals’ responses on the items measuring cognitive reappraisal
and expressive suppression. Table 1 shows descriptive statistics for these measures.
PLOS ONE | https://doi.org/10.1371/journal.pone.0243209 December 10, 2020
7 / 26
PLOS ONE Emotion regulation and face recognition
Fig 2. Scatter plots for the relationship between expressive suppression and false positives in face immediate memory in Experiment 1. r (40) = .57, p< 0.001,
with 95% confidence internals of .32 to .74.
https://doi.org/10.1371/journal.pone.0243209.g002
Table 2 shows Pearson Correlation coefficients between face immediate memory and ERQ
strategies. Cognitive reappraisal did not correlate with any measures of face memory (mean r
= .06). However, expressive suppression evidenced a moderate to strong negative correlation
with the overall accuracy of face immediate memory (r = -.52, p < 0.001). Specifically, higher
levels of expressive suppression were associated with a lower level of overall face-memory
accuracy and higher rates of misidentifications and false positives. Fig 2 shows scatter plots for
the relationship between expressive suppression and false positives. No associations were
found for hits and misses.
Table 1. Descriptive statistics for participants’ responses on the face immediate memory task (%) and the ERQ.
Measure
Face Immediate memory
Overall accuracy
Hits
Miss
Misidentification
False Positives
Emotion Regulation Questionnaire
cognitive reappraisal
expressive suppression
https://doi.org/10.1371/journal.pone.0243209.t001
M
61.9
66.2
17.4
16.9
42.4
4.9
3.7
SD
16.8
17.2
15.9
13.9
24.5
.9
1.4
Minimum
Maximum
30
30
0
0
0
2.5
1.3
100
100
50
50
70
6.7
6.3
PLOS ONE | https://doi.org/10.1371/journal.pone.0243209 December 10, 2020
8 / 26
PLOS ONE Table 2. Correlations between face immediate face memory and ERQ strategies.
Emotion regulation and face recognition
Cognitive reappraisal
.06
.06
-.13
.10
-.03
Expressive suppression
-.52��
-.20
-.14
.40��
.57��
Overall accuracy
Hits
Miss
Misidentification
False Positives
Note
�� = p< 0.01.
https://doi.org/10.1371/journal.pone.0243209.t002
Discussion
Consistent with previous experiments [72], overall performance on the face immediate mem-
ory task was rather low (62%), but there were wide individual differences ranging from 30% to
100%. Similarly, wide ranges of individual differences in ERQ strategies were evident (see
Table 1), replicating previous studies [74]. The overall accuracy of immediate face memory
negatively correlated with expressive suppression. This correlation was mainly derived from
misidentification and false positive, which are positively correlated with each other [9, 29, 72].
However, there was no correlation between any measure of face memory and cognitive
reappraisal.
The negative correlation found between expressive suppression and face immediate mem-
ory can be explained by integrating two main findings within the existing literature. First,
expressive suppression has been found to associate with negative emotions and stress-related
symptoms [74]. For example, Butler et al. (2003) found that expressive suppression disrupted
social communication, had a negative impact on the regulators’ emotional experience, and
increased stress levels as indicated by increased blood pressure [76]. Second, face recognition
studies have found that negative emotions, especially anxiety, had detrimental effects on per-
formances on a range of face recognition tasks [44, 46, 48–51, 53]. It is therefore possible that
expressive suppression might mediate the negative relationship between emotions and face
recognition. Future studies are needed to more explicitly examine this possibility, though.
The negative correlation between expressive suppression and face memory also converges
with the results of Richard and Gross (2000) that expressive suppression, but not cognitive
reappraisal, is associated with poor memory [62], as previous studies reported positive associa-
tions between memory and face recognition [9, 32, 69–71]. Although some studies have found
that face memory is positively correlated with face perception [26], face recognition theories
suggested a dissociation between face perception and face memory [13]. Therefore, to more
clearly explicate this potential dissociation, Experiment 2 aimed to examine the associations
between ERQ strategies and a face-matching task.
Experiment 2
Experiment 1 found detrimental effects of expressive suppression on some aspects of immedi-
ate face memory. The aim of the present experiment was to replicate this finding using a range
of 1-in-1 face-matching tasks, in which participants were presented with pairs of unfamiliar
faces and they were asked to make same/different decisions. Surprisingly, previous studies
have repeatedly demonstrated that performance on this seemingly easy task is rather poor,
with an error rate of roughly 20 per cent for overall accuracy [9, 72]. In addition, to maximize
individual differences, we involved highly challenging face-matching conditions including
PLOS ONE | https://doi.org/10.1371/journal.pone.0243209 December 10, 2020
9 / 26
PLOS ONE Emotion regulation and face recognition
same- vs. other-race face images taken in the same day or different times. Previous studies
have reported that other-race faces are perceived and recognized less accurately than own-race
faces [15–18] and that matching performance highly degrades when the task involves images
taken months apart [77].
Method
Participants. Fifty-three undergraduate students (28 females and 25 males) from Qatar
University volunteered to participate in this experiment. The participants’ mean age was 19.5
years (SD = 1.6) and all had normal or corrected to normal vision. None had any history of
psychopathology as self-reported and none had participated in Experiment 1.
Measures.
(1) Face-matching tasks. A total of 200 match/mismatch pairs of Egyptian and
UK male unfamiliar faces were used in this experiment. For each face nationality, there were
60 same-day photo pairs (30 matches and 30 mismatches) and 40 different-day photo pairs (20
matches and 20 mismatches). Fig 3 shows examples of these stimuli.
The Egyptian and UK same-day face-matching pairs were taken from Megreya and Burton
(2008 and 2006), respectively [9, 72]. All images showed a full-face view with a neutral facial
Fig 3. Examples of face-matching tasks used in Experiment 2 and 3. Regarding copyright issues, we could not present examples for the UK face images that were
taken in different times.
https://doi.org/10.1371/journal.pone.0243209.g003
PLOS ONE | https://doi.org/10.1371/journal.pone.0243209 December 10, 2020
10 / 26
PLOS ONE Emotion regulation and face recognition
expression. On the one hand, each matching pair consisted of a still photograph of a target face
captured by a high-quality video camera and a photograph depicting the same person that was
taken from a high-quality digital camera in the same day and under the same lighting condi-
tions. On the other hand, each mismatching pair consisted of a still photograph of a target face
and a digital photograph depicting a distractor face.
The Egyptian and UK different-day face-matching pairs were taken from Megreya et al.,
(2013) and White, Kemp, Jenkins, and Burton, (2014), respectively [77, 78]. For the Egyptian
stimuli, each matching pair consisted of a still photograph of a target and a photograph depict-
ing the same face that was taken months apart, with an average of roughly seven months,
whereas mismatch pairs consisted of stills and photographs depicting different faces. Like the
same-day stimuli, all Egyptian different-day images showed a full-face view with a neutral facial
expression. In addition, all stills and photographs were taken from the same camcorder and dig-
ital cameras as used in the same-day condition and under the same lighting conditions. The UK
face pair stimuli were constructed using sets of images depicting Australian and British celebri-
ties taken from the internet. Matching pairs consisted to two images depicting the same face
that were taken during different times and showed different expressions, whereas mismatching
pairs consisted of two images of two different people. Notably, all celebrities were unknown to
the participants in this study as confirmed by all of them after the experiment.
Each participant completed 100 trials (15 Egyptian same-day matches; 15 Egyptian same-
day mismatches; 10 Egyptian different-day matches; 10 Egyptian different-day mismatches; 15
UK same-day matches; 15 UK same-day mismatches; 10 UK different-day matches; 10 UK dif-
ferent-day mismatches). Two versions of stimuli were created to counter-balance match/mis-
match trials so that each target face was equally presented in match and mismatch pairs across
the experiment.
(2) Emotion Regulation Questionnaire (ERQ) [74]. This was the same instrument described
in Experiment 1 above.
Results
Three indices were calculated for the face-matching tasks. These included (i) hits (correct deci-
sion that the two faces in matching pairs depict the same identity), (ii) false positives (FPs; false
decision that the two faces in mismatching depict the same identity), and (iii) overall accuracy
(hits plus correct rejections; the complement of FPs). Consistent with Experiment 1, we report
the averages of individuals’ responses on the ERQ items measuring cognitive reappraisal and
expressive suppression. Table 3 shows descriptive statistics for these measures.
Correlations between face matching and ERQ strategies. Table 4 shows Pearson corre-
lation coefficients between participants’ performances on the face matching tasks and ERQ
strategies. For general performance across the four face-matching tasks, there were no signifi-
cant correlations between all matching measures and cognitive reappraisal (mean r = -.01).
However, expressive suppression correlated negatively with the overall accuracy (r = -.34,
p = 0.01) and positively with FPs (r = .51, p < 0.001). Therefore, higher levels of expressive sup-
pression were associated with a lower level of overall face-matching performance and higher
rates of false positives. Fig 4 shows scatter plots for the relationship between expressive sup-
pression and false positives in the overall face-matching task. Moderate-to-strong positive cor-
relations between expressive suppression and FPs were consistently found across the four face-
matching tasks (mean r = .39). However, no correlation was found between expressive sup-
pression and hits (see Table 4).
Correlations among the four face-matching tasks. Table 5 shows Pearson correlation
coefficients among the four face-matching tasks. There were strong positive correlations
PLOS ONE | https://doi.org/10.1371/journal.pone.0243209 December 10, 2020
11 / 26
PLOS ONE Table 3. Descriptive statistics for participants’ responses on the 1-in-1 face-matching tasks and ERQ strategies in Experiment 2.
Emotion regulation and face recognition
General Performance
Overall Accuracy
Hits
FPs
Performance on Egyptian same-day task
Overall Accuracy
Hits
FPs
Performance on Egyptian different-day task
Overall Accuracy
Hits
FPs
Performance on UK same-day task
Overall Accuracy
Hits
FPs
Performance on UK different-day task
Overall Accuracy
Hits
FPs
ERQ
Reappraisal
Suppression
https://doi.org/10.1371/journal.pone.0243209.t003
M
79.8
81
21.5
85.4
91.1
20.4
81.5
75.6
12.5
81
84.4
22.3
71.1
73.1
31.0
5.3
3.7
SD
8.5
11.4
14.5
10.8
11.9
17.8
10.4
17.8
13.8
11.4
11.2
19.2
10.0
19.6
20.1
0.9
1.5
Minimum
Maximum
51.7
50
1.7
46.7
40
0
55
20
0
50
60
0
50
20
0
3.2
1
92.9
96.7
63.3
100
100
80
100
100
50
100
100
73.3
90
100
80
6.8
6.3
among all tasks with the following exceptions. Hit scores in matching Egyptian faces that were
taken during the same time did not correlate with hits of matching UK faces that were taken
during different times. In addition, no correlation was observed for FPs in matching Egyptian
and UK faces in the different time condition.
Performances across the four face-matching tasks. The comparability of the Egyptian
and UK same day face-matching tasks were higher than that of the different day tasks. On the
one hand, the Egyptian different day task consisted of images of targets that were taken several
months apart using the same cameras and under the same lighting conditions, similar to the
Egyptian same day task. On the other hand, the UK different day task involved ambient images
taken from the internet. With this caution, we examined the differences among participants’
performances on these four face-matching tasks. Matching measures were subjected to a series
of 2 (face nationality: Egyptian vs. UK) x 2 (times: same-day vs. different-day photos) within-
participant Analysis of Variances (ANOVAs). For overall accuracy, there were main effects of
face nationality, F (1,51) = 55.98, p < 0.001, showing that Egyptian faces were matched more
accurately than UK ones (83.5% vs. 76%), and image times, F (1,51) = 29.06, p<0.001, indicat-
ing that same-day faces were matched more accurately than different-day ones (83.2% vs.
76.3%). In addition, there was an interaction between these two factors, F (1,51) = 15.51,
p<0.001. Subsequent Simple Main Effects (SMEs) reported other-race effects using both
same-day and different-day stimuli, Fs (1,51) = 9.66 and 55.84, ps � 0.01, and confirmed the
image time effects using both same-race (Egyptian) and different-race (UK) faces, Fs (1,51) =
4.51 and 30.25, ps � 0.05. Consistently, significant main effects of face nationality were noticed
PLOS ONE | https://doi.org/10.1371/journal.pone.0243209 December 10, 2020
12 / 26
PLOS ONE Emotion regulation and face recognition
Table 4. Correlations between face matching and ERQ strategies in Experiment 2.
Reappraisal
Suppression
General performance
Overall Accuracy
Hits
FPs
Performance on Egyptian same-day task
Overall Accuracy
Hits
FPs
Performance on Egyptian different-day task
Overall Accuracy
Hits
FPs
Performance on UK same-day task
Overall Accuracy
Hits
FPs
Performance on UK different-day task
Overall Accuracy
Hits
FPs
Note
� = p < 0.05
�� = p< 0.01.
https://doi.org/10.1371/journal.pone.0243209.t004
-0.01
0.06
0.06
0.05
0.15
0.04
0.17
0.18
-0.02
-0.04
0.08
0.09
-0.23
-0.16
0.07
-.34��
0.14
0.51��
-0.42��
-0.03
0.48��
-0.11
0.18
0.39�
-0.31�
0.02
0.38�
-0.23
0.19
0.40�
for both hits and false positives, Fs (1,51) = 5.96 and 29.63, ps � 0.05. Image times had signifi-
cant main effect on hits, F (1,51) = 63.16, p < 0.001, but not for false positives, F < 1.
Discussion
This experiment examined the correlation between participant performance on a range of
face-matching tasks, with different levels of difficulty, and ERQ strategies. Consistent with pre-
vious studies [9, 72], the task of matching images depicting same-race unfamiliar faces that
were taken on the same day was relatively error-prone so that participants falsely rejected
roughly 10% of matching pairs and falsely accepted roughly 20% of mismatching pairs. In
addition, consistent with Megreya et al. (2013), a lower level of performance was noticed when
participants had to match same-race face images that were taken months apart as they falsely
rejected roughly 25% in matching pairs [77]. Furthermore, the present results replicated the
well-established other-race effect [18] in that participants matched their own-race faces more
accurately than other-race faces when the images were taken in the same day (85% vs. 81%) or
in different times (81% vs. 71%).
Experiment 1 found that performance on a face immediate memory task did not correlate
with cognitive reappraisal, whereas overall accuracy, misidentifications and false positives
were associated with a high use of expressive suppression. Providing converging evidence in
support of the stability of these findings, using a range of perceptual tasks, the present experi-
ment demonstrated a close to zero correlation between face-matching performances and cog-
nitive reappraisal, while a lower level of overall accuracy and higher rates of false positives
PLOS ONE | https://doi.org/10.1371/journal.pone.0243209 December 10, 2020
13 / 26
PLOS ONE Emotion regulation and face recognition
Fig 4. Scatter plots for the relationship between expressive suppression false positives in the overall face-matching task in Experiment 2. r (50) = .51, p< 0.001,
with 95% confidence internals of .22 to .75.
https://doi.org/10.1371/journal.pone.0243209.g004
were associated with a higher use of expressive suppression. Therefore, maladaptive ER strate-
gies, in general, might have detrimental effects on face recognition. To examine this suggestion
further, Experiment 3 investigated associations between the same four face-matching tasks
and a broader range of adaptive versus maladaptive cognitive ER strategies.
Table 5. Correlations among the four face-matching tasks in Experiment 2.
Egyptian/ Same Time
Egyptian/ Different Times
UK/
Same Time
Overall Accuracy
Egyptian/ Different Times
UK/ Same Time
UK/ Different Times
Hits
Egyptian/ Different Times
UK/ Same Time
UK/ Different Times
False Positives
Egyptian/ Different Times
UK/ Same Time
UK/ Different Times
Note
� = p < 0.05
�� = p< 0.01.
.48��
.79��
.41��
.48��
.48��
.19
.72��
.80��
.42��
https://doi.org/10.1371/journal.pone.0243209.t005
.39��
.45��
.52��
.35�
.57��
.27
.49��
.57��
.51��
PLOS ONE | https://doi.org/10.1371/journal.pone.0243209 December 10, 2020
14 / 26
PLOS ONE Emotion regulation and face recognition
Experiment 3
Experiments 1 and 2 both found that expressive suppression was negatively associated with the
accuracy of face memory and face perception, respectively. This ER strategy has been generally
considered as maladaptive in that it is associated with increased rates of negative emotions and
psychopathology [61]. The ER literature, however, suggests that additional maladaptive ER
strategies exist which may similarly be associated with face-perception abilities. For example,
Garnefski et al., (2001) developed a nine-factor framework for the cognitive emotion regula-
tion strategies thought to encompass the various ways in which people cognitively manage and
control their emotions during or after the experience of a stressful event [60]. These factors
include self-blame, acceptance, rumination, positive refocusing, refocus on planning, positive
reappraisal, putting into perspective, catastrophizing, and other-blame. Garnefski et al. (2001)
found that these nine factors could be classified into two boarder factors: adaptive strategies
(which include Positive refocusing, Positive reappraisal, Putting into perspective, Refocus on
planning and Acceptance) and maladaptive strategies (which include Rumination, Self-blame,
Other-blame and Catastrophizing) [60]. Megreya, Latzman, Al-Attiyah, & Alrashidi, 2016)
replicated the nine-factor structure of the CERQ among four Arab countries (Egypt, Kingdom
of Saudi Arabic, Kuwait, and Qatar) [79]. To further examine how adaptive versus maladaptive
emotion regulation strategies may influence face recognition, Experiment 3 was conducted to
investigate individual differences in across these cognitive ER strategies and face perception.
Method
Participants. Fifty-two graduate and undergraduate students from Qatar University (28
females and 24 males) volunteered to participate in this experiment. Their mean age was 26.4
years (SD = 4.1) and all had normal or corrected to normal vision. None had history of psycho-
pathology as self-reported and none had participated in Experiments 1 and 2.
Measures.
(1) Face-matching tasks. This experiment used the same face-matching tasks
that were used in Experiment 2. These required participants to match sets of match/ mismatch
pairs of unfamiliar faces, which belonged to the same- or different-race and taken on the same
day or during different times (see Fig 3 for examples).
(2) Cognitive Emotion Regulation Questionnaire (CERQ) [60]. The Cognitive Emotion Reg-
ulation Questionnaire (CERQ) assesses a variety of cognitive strategies that people tend to use
following the experience of negative events and situations. The CERQ measures nine cognitive
ER strategies, which were defined as following [80]:
• Self-blame: the thoughts of blaming oneself for what she/he has experienced,
• Acceptance: the thoughts of resigning what has happened,
• Rumination: thinking all the time on the feelings and thoughts associated with negative
events,
• Positive refocusing: thinking of other, pleasant matters instead of the actual event,
• Refocus on planning: thinking on potential steps to deal with negative events,
• Positive reappraisal: thinking of attaching a positive meaning to the event in terms of per-
sonal growth,
• Putting into perspective: the thoughts of playing down the seriousness of a negative event as
compared to other events,
• Catastrophizing: the explicit emphasize of the terror of negative events,
PLOS ONE | https://doi.org/10.1371/journal.pone.0243209 December 10, 2020
15 / 26
PLOS ONE Emotion regulation and face recognition
• Other-blame: the thoughts of putting the blame for what one has experienced on others.
The CERQ is a 36-item self-report measure of these nine cognitive ER strategies that indi-
viduals tend to use after experiencing a stressful life event. It consists of nine subscales, which
measure a range of adaptive (Positive refocusing, Refocus on planning, Positive reappraisal,
and Putting into perspective) and maladaptive (Self-blame, Acceptance, Rumination, Catastro-
phizing, and Other-blame) cognitive ER strategies. Each subscale consists of 4 items, using a
5-point Likert-type scale ranging from 1 (almost never) to 5 (almost always). Therefore, the
scores of each subscale range between 4 and 20, and a high score reflects a greater use of the
CERQ strategy. All of the nine-CERQ factors positively correlated with each other (rs ranged
between .20 and .62), with moderate to high Cronbach’s alpha reliabilities (αs ranged between
0.68 and 0.83) and high test-retest stability (rs ranged between 0.41 and 0.59). The Arabic ver-
sion of the CERQ has similar adequate psychometric properties in a range of Arabic countries,
including Qatar [79, 80]. Specifically, Megreya et al (2016) replicated the nine-factor structure
(CERQ strategies) and the higher two-factor solution (adaptive versus maladaptive strategies)
of the Arabic version of the CERQ, with acceptable to good Cronbach reliability rates for the
nine CERQ subscales (which ranged from 0.67 to 0.86) [79].
Results
Correlations between face matching and CERQ strategies. Table 6 shows descriptive
statistics for participants’ responses on the face-matching tasks and the CERQ. Table 7 shows
Pearson correlation coefficients between participants’ performances on the face matching
tasks and CERQ strategies. For participants’ general performance in the four face-matching
tasks, overall face-matching accuracy correlated negatively with self-blame, rumination, and
catastrophizing. Hits correlated negatively with self-blame (r = -.32, p = 0.02 and rumination
(r = -.28, p = 0.04) whereas false positives correlated positively with catastrophizing (r = .51,
p < 0.001). Fig 5 shows scatter plots for the relationship between catastrophizing and false pos-
itives in the overall face-matching task. Across all of the four face-matching tasks, strong posi-
tive correlations between false positives and catastrophizing were consistently reported (rs
ranged from .35 to .45, ps < 0.01, mean r = .40). However, no correlation was found between
the face matching measures and the other eight CERQ factors, especially for matching own-
race (Egyptian) faces. When broad dimensions were considered, no correlation was found
between adaptive strategies and any of face-matching measures across all tasks. Importantly
however, maladaptive strategies correlated negatively with overall accuracies in all face-match-
ing tasks (rs ranged from -.28 to -.45, ps < 0.05, mean r = -.35) but not in the task of matching
Egyptian different-day faces (r = -.15, p = .27). That is, maladaptive cognitive emotion regula-
tion strategies, especially catastrophizing, were associated with lower levels of overall face-
matching performances and higher rates of false positives.
Correlations among the four face-matching tasks. Table 8 shows Pearson correlation
coefficients among the four face-matching tasks. There were strong positive correlations
among all tasks (rs ranged from .31 to .67, mean r = .49).
Performances across the four face-matching tasks. A series of 2 (face nationalities) x 2
(image times) within-participants ANOVAs were conducted using the three face-matching
measures. Face nationality and image time factors yielded significant main effects on overall
accuracy, Fs (1,51) = 43.14 and 47.49, ps < 0.001, and hits, Fs (1,51) = 13.83 and 71.24,
ps < 0.001, showing same-race and same-day advantages; respectively. For false positives, the
same-race advantage was also noticed, F (1,51) = 14.68, p < 0.001, but there was no main effect
of image time, F (1,51) < 1. Face nationality and image time yielded interactions for overall
accuracy, F (1,51) = 6.43, p = 0.01, and false positives, F (1,51) = 12.56, p < 0.001, but not for
PLOS ONE | https://doi.org/10.1371/journal.pone.0243209 December 10, 2020
16 / 26
PLOS ONE Table 6. Descriptive statistics for participants’ responses on the face matching tasks and CERQ in Experiment 3.
Face matching tasks
General Performance
Overall Accuracy
Hits
FPs
Performance on Egyptian same-day task
Overall Accuracy
Hits
FPs
Performance on Egyptian different-day task
Overall Accuracy
Hits
FPs
Performance on UK same-day task
Overall Accuracy
Hits
FPs
Performance on UK different-day task
Overall Accuracy
Hits
FPs
CERQ
Self-Blame
Acceptance
Rumination
Positive Refocusing
Refocus on Planning
Positive Reappraisal
Putting into Perspective
Catastrophizing
Other-blame
Adaptive strategies
Maladaptive strategies
https://doi.org/10.1371/journal.pone.0243209.t006
M
80.2
80.2
19.8
85.8
90.5
18.8
81.2
76.2
13.8
82.0
83.7
19.7
71.8
70.6
26.9
12.3
13
14.6
13.3
16.1
15.8
15.0
10.4
10
60.3
60.3
SD
7.8
12.3
13.3
8.5
10.1
14.6
11.4
20.3
13.6
9.8
14.1
18.0
10.8
17.6
21.3
3.5
2.8
3.9
4.0
3.2
3.3
3.5
4.4
3.4
11.7
13.8
Emotion regulation and face recognition
Minimum
Maximum
59.2
50.0
1.7
63.3
60
0
55
20
0
56.7
40
0
45
20
0
5
4
5
5
6
5
4
4
4
20
22
93.8
100
62.5
100
100
60
100
100
60
100
100
86.7
95
100
80
19
20
20
19
20
20
20
20
19
77
91
hits, F(1, 51) <1, p = .71. The results of Simple Main Effects of overall accuracy and false posi-
tives can be found as a supplementary file.
Discussion
Replicating results of Megreya et al (2013) [77] and Experiment 2, hit rates dropped from
90.5% to 76.1% when observers were asked to match same-race face images that were taken on
the same day or months apart. In addition, consistent with Experiment 2, hit rates dropped
from 83.7% to 70.6% when they had to match other-race face images that were taken on the
same day or during different times. Indeed, a large body of experimental studies have provided
good evidence that face matching is rather error-prone [9, 72]. However, the vast majority of
these studies have used face-matching stimuli that were photographed on the same day and
under the same lighting conditions. Consistent with a previous suggestion [77], the present
results suggest that these previous experimental studies likely provide an underestimate of the
PLOS ONE | https://doi.org/10.1371/journal.pone.0243209 December 10, 2020
17 / 26
PLOS ONE Emotion regulation and face recognition
Table 7. Correlations between performances on the face-matching tasks and CERQ strategies in Experiment 3.
Self
Acc
Rum
PRef
Plan
PosR
Pers
Cat
Other
Adaptive
Maladaptive
General Performance
Overall Accuracy
Hits
FPs
Performance on Egyptian same-day task
Overall Accuracy
Hits
FPs
Performance on Egyptian different-day task
Overall Accuracy
Hits
FPs
Performance on UK same-day task
Overall Accuracy
Hits
FPs
Performance on UK different-day task
Overall Accuracy
Hits
FPs
-.36��
-.32�
.12
-.20
-.10
.15
-.32�
-.28�
.11
-.14
-.08
.09
.08
-.15
-.23
-.09
-.14
-.02
-.01
.08
.08
-.42��
.02
.51��
-.12
-.01
.13
-.21
-.27
.06
-.17
-.15
.06
-.32�
-.34��
-.01
-.38��
-.20
.22
-.19
-.12
.14
-.19
-.05
.24
-.26
-.13
.18
-.01
-.02
.01
-.22
-.18
.14
-.17
-.19
.01
-.33�
-.34�
.09
-.25
-.18
.10
-.14
-.16
.06
-.05
-.09
-.05
-.13
-.18
-.01
-.11
.12
.21
.03
-.19
-.17
.03
-.20
-.34
.10
-.12
-.20
.09
.03
-.07
-.09
-.05
.06
-.12
-.23
-.15
-.04
-.17
-.08
.01
.13
.10
-.01
.08
.08
.15
.06
-.16
-.11
-.03
.10
-.01
.20
.18
-.36��
.04
.45��
-.17
.09
.43��
-.51��
-.27
.35�
-.28�
.14
.39��
-.03
.16
.15
.13
.17
.04
-.24
-.18
.12
-.23
-.16
.10
-.05
-.08
-.02
-.10
-.18
-.01
-.01
-.14
-.20
-.06
-.15
-.05
-.01
.15
.13
-.38��
-.17
.29�
-.28�
-.09
.26
-.15
-.03
.21
-.45��
-.36��
.20
-.31�
-.10
.23
Note: Self = Self-blame; Acc = Acceptance; Rum = Rumination; PRef = Positive Refocusing; Plan = Refocus on Planning; PosR = Positive Reappraisal; Pers = Putting
into Perspective; Cat = Catastrophizing; Other = Other-blame; FPs = False positives
� = p<0.05
�� = p<0.01.
https://doi.org/10.1371/journal.pone.0243209.t007
challenges of face identity verification in security settings such as country borders where pass-
port officers have to match face identities of different ethnic groups to their passport photo-
graphs that would not be never taken on the same day.
The results of this experiment also improve our understanding of the relationship between
adaptive versus maladaptive ER strategies and face perception. Adaptive cognitive ER strate-
gies–individually or combined–did not correlate with any face matching measures consistently
in all face-matching tasks. This finding converges with the results of Experiments 1 and 2 that
cognitive reappraisal–as an adaptive ER strategy–does not correlate with face memory and
face perception. Importantly, however, higher use of combined maladaptive cognitive ER
strategies, as well as self-blame, rumination, and catastrophizing more specifically, were associ-
ated with lower levels of overall accuracy and higher rates of false positives in the overall face-
matching task. Within each face-matching task, there were inconsistent correlations with self-
blame and rumination, but catastrophizing was positively correlated with false positives. The
positive correlations between catastrophizing and false positives were strong in magnitude
(with mean r = .49) and robust across all of the four face-matching tasks (see Table 7).
General discussion
Across 3 experiments, the current research examined for the first time associations between
individual differences in a range of adaptive versus maladaptive ER strategies and face recogni-
tion. Using an immediate face-memory paradigm, in which observers had to identify a self-
PLOS ONE | https://doi.org/10.1371/journal.pone.0243209 December 10, 2020
18 / 26
PLOS ONE Emotion regulation and face recognition
Fig 5. Scatter plots for the relationship between catastrophizing and false positives in the overall face-matching task in Experiment 3. r (50) = .51,
p< 0.001, with 95% confidence internals of .29 to .68.
https://doi.org/10.1371/journal.pone.0243209.g005
paced learned unfamiliar face from a 10-face target-present/ target-absent line-up, Experiment
1 reported that higher levels of expressive suppression, but not cognitive reappraisal, were asso-
ciated with a lower level of overall face-memory accuracy and higher rates of misidentifications
Table 8. Correlations among the four face-matching tasks in Experiment 3.
Egyptian/ Same Time
Egyptian/ Different Times
UK/Same Time
Overall Accuracy
Egyptian/ Different Times
UK/ Same Time
UK/ Different Times
Hits
Egyptian/ Different Times
UK/ Same Time
UK/ Different Times
False Positives
Egyptian/ Different Times
UK/ Same Time
UK/ Different Times
Note
� = p < 0.05
�� = p< 0.01.
.50��
.67��
.51��
.61��
.66��
.32�
.53��
.60��
.50��
https://doi.org/10.1371/journal.pone.0243209.t008
.43��
.31�
.61��
.37�
.48��
.36��
.36��
.44��
.49��
PLOS ONE | https://doi.org/10.1371/journal.pone.0243209 December 10, 2020
19 / 26
PLOS ONE Emotion regulation and face recognition
and false positives. Experiment 2 replicated these findings using a range of face-matching tasks,
where observers were asked to match pairs of same-race or different-race face images taken on
the same day or during different times. Higher levels of expressive suppression were associated
with a lower level of overall face-matching performance and higher rates of false positives, but
cognitive reappraisal did not correlate with any face-matching measure. Experiment 3 revealed
that the higher use of maladaptive cognitive ER strategies, especially catastrophizing, was associ-
ated with lower levels of overall face-matching performances and higher rates of false positives.
Therefore, the higher use of maladaptive ER strategies in general, and expressive suppression
and catastrophizing more specifically, has detrimental effects on face memory and face
perception.
Expressive suppression refers to is a form of response modulation that involves inhibiting
ongoing emotion-expressive behavior [74], whereas Catastrophizing refers to recurring
thoughts about how terrible the event has been and about what one has gone through being
the worst thing to happen to a person [80]. Expression suppression is associated with a range
of negative social (e.g., disrupted communication) [76] and cognitive (worsened memory) [62]
consequences. In addition, evidence indicates social anxiety disorder is associated with an
overreliance on expressive suppression [81]. Importantly, results of the current research dem-
onstrated that relatively higher levels of expression suppression have a detrimental effect on
face recognition, a social cognitive ability. Catastrophizing is one of most related ER strategies
to anxiety disorders generally [82] and social anxiety disorder specifically as socially anxious
adults tend to interpret ambiguous events in a negative way and appraise mildly negative
events catastrophically [83, 84]. Taken together, given that trait anxiety [48–51] and social anx-
iety [53] are associated negatively with face recognition, expressive suppression and catastro-
phizing serve as mediators in the negative relationship between anxiety disorders and face
recognition.
As this study is exploratory, the mechanisms by which expressive suppression and catastro-
phizing affect face recognition are not clear. Notably, these two maladaptive ER strategies were
associated with higher false positives but they did not correlate with hits. Intuitively, these find-
ings suggest that the influences of expressive suppression and catastrophizing might be related
to processing new faces, rather than old faces. In line with this suggestion, Experiment 1 found
that higher rates of misidentifications in an immediate face-memory paradigm were also asso-
ciated with higher expressive suppression (see Table 2). Nevertheless, future studies are
encouraged to explore these possibilities.
It is well-known that emotions interact with many aspects of cognitive processing [85]. For
example, Blair et al (2007) found that emotional distractors disrupted goal-directed processing
and goal-directed processing disrupted the neurophysiological responses to emotional photo-
graphs [86]. Previous studies reported that higher levels of expressive suppression, but not cog-
nitive reappraisal, are associated with lower levels of general memory performance [62]. These
findings resulted in a conclusion that “keeping a still face and stiff upper lip decreases one’s
memory for the details of the unfolding emotion-eliciting situation, whereas cognitively trans-
forming the situation by changing one’s thinking does not appear to exact such a cognitive
cost” [62, p. 423]. Our results are consistent with these conclusions, in that we found expressive
suppression and catastrophizing to be associated with mistaken face identification, especially
in target-absent trials. These results have potentially important applications. Indeed, whereas
most previous studies that have aimed to improve face recognition have focused on the cogni-
tive representations of faces [17, 78, 87, 88], our findings suggest that improving face recogni-
tion ability likely requires a more thorough training program that likely includes a focus on
cognitive processing of faces as well as strategies for coping with emotions.
PLOS ONE | https://doi.org/10.1371/journal.pone.0243209 December 10, 2020
20 / 26
PLOS ONE Emotion regulation and face recognition
A large body of laboratory studies reported that matching face identities using photographs
is highly error-prone, suggesting challenges to real-life security settings (such as airports) in
where persons’ identities are verified by matching their faces to photo-IDs [9, 72]. Importantly,
however, there is a critical mismatch between face matching in those laboratory studies and
realistic security settings. Specifically, the vast majority of studies have used face-matching
images that were taken on the same day, while the general appearance of faces does change,
even day to day, in realistic settings. Further, people can use their passports as long as they are
valid, often for many years. Therefore, in one notable study, Megreya et al (2013) simulated
face matching procedures in realistic settings by asking participants to match images of unfa-
miliar faces that were taken on the same day or several months apart [77]. When face images
were taken on the same day, Megreya et al. (2013) reported hit rates of 79% and 90% using the
1-in-10 and 1-in-1 face-matching tasks (respectively) [77]. However, hit rates dropped to 58%
and 70% on these tasks when face images were taken several months apart. Consistently, the
results of Experiments 2 and 3 showed that hit rates were dropped from 91.2% to only 76.7%
when face images were taken on the same day or several months apart.
Nevertheless, these face-matching conditions, matching two images taken in different times
that depict a face belonging to the same race, still do not typically match realistic security set-
tings, which involve both own-race and different-race faces. It has been known for many years
that matching and recognizing other-race faces are more difficult than matching and recogniz-
ing own-race faces [15–18]. Therefore, there are three main challenges for matching faces to
photo-IDs in realistic settings. These are (i) matching faces using photographs that is a highly-
error prone task [9, 72, 77]; (ii) matching face images that were taken on different times
[Experiments, 2 & 3, 77]; and (iii) matching faces that belong to other races [5–18]. The pres-
ent study compared for the first time between the accuracy of matching other-race faces that
were taken on the same day or during different times. Specifically, the results of Experiment 2
and 3 reported that hit rates were dropped from 85.6% to only 73.1% when the face images of
other-race faces were taken on the same day or during different times. Therefore, along with
our previous study [77], the present results suggests that previous laboratory studies on face
matching underestimate its difficulty in real-world situations. Photographs of unfamiliar faces
seem to be unreliable proofs of identity, especially if the ID documents do not use very recent
images of the holders and if the faces to be matched belong to a different race.
Limitations and conclusions
The current research is not without limitations. For example, the results showed that a behav-
ioral (expressive suppression) and a cognitive (catastrophizing) emotion regulation strategy–
as measured by the ERQ and CERQ (respectively)—correlated negatively with the accuracy of
unfamiliar face recognition. Although the ERQ and CERQ are widely-used measures of emo-
tion regulation, these findings need replication using, for example, a behavior measure of
expressive suppression [62] and other report-based scales that focus explicitly on behavioral
strategies [89]. In addition, as both expressive suppression and catastrophizing are considered
maladaptive emotion regulation strategies, it is also important to investigate emotion regula-
tion difficulties [90] influence face recognition. An additional limitation is related to the dis-
crepancy with the way in which the face-matching tasks were set up in this study (where 50%
of trails were mismatches) and the identity verification in real-life security settings (where mis-
matches are far less frequent). In addition, only male faces were used. Finally, it is important to
note that the UK different-day face-matching task included ambient images depicting Western
celebrities who were definitely unknown to the participants. These images, therefore, likely
have different characteristics as compared to the images used in the other face-matching tasks.
PLOS ONE | https://doi.org/10.1371/journal.pone.0243209 December 10, 2020
21 / 26
PLOS ONE Emotion regulation and face recognition
However, our main interest in this study was not to examine this peculiar effect on the other-
race effect.
Limitations notwithstanding, the current research reports for the first time that certain
emotion regulation strategies (expressive suppression and catastrophizing) negatively affect
the accuracy of face recognition. Although additional research with larger samples is needed to
confirm the replicability of these findings, the current research provides new evidence for the
association between emotion and cognition [85]. In addition, this study provides a more realis-
tic face-matching procedure suggesting that previous laboratory studies may have underesti-
mated the difficulty of unfamiliar face matching in real-world situations.
Supporting information
S1 Data. The experiments reported in this study.
(XLSX)
S1 File. Supplementary results for experiments 2 and 3 in this study.
(DOCX)
Author Contributions
Conceptualization: Ahmed M. Megreya.
Data curation: Ahmed M. Megreya.
Formal analysis: Ahmed M. Megreya.
Investigation: Ahmed M. Megreya.
Methodology: Ahmed M. Megreya.
Writing – original draft: Ahmed M. Megreya.
Writing – review & editing: Ahmed M. Megreya, Robert D. Latzman.
References
1.
Lander K, Bruce V & Bindemann M (2018). Use-inspired basic research on individual differences in face
identification: Implications for criminal investigation and security. Cogn Research: 3, 26. https://doi.org/
10.1186/s41235-018-0115-6 PMID: 29984301
2. Wilmer JB (2017). Individual differences in face recognition: a decade of discovery. Curr Dir Psychol
Sci: 26, 225–230. https://doi.org/10.1177/0963721417710693
3. Yovel G, Wilmer J B & Duchaine B (2014). What can individual differences reveal about face process-
ing? Front Hum Neurosci: 8, 562. https://doi.org/10.3389/fnhum.2014.00562 PMID: 25191241
4. Kress T & Daum I (2003). Developmental prosopagnosia: a review. Behav Neurol: 14, 109–21. https://
doi.org/10.1155/2003/520476 PMID: 14757987
5. Susilo T., & Duchaine B. (2013). Advances in developmental prosopagnosia research. Curr Opin Neu-
robiol: 23, 423–9. https://doi.org/10.1016/j.conb.2012.12.011 PMID: 23391526
6. Noyes E, Phillips PJ & O’Toole AJ (2017). What is a super-recogniser? In Bindemann M & Megreya AM
(Eds.), Face processing: Systems, disorders and cultural differences (pp. 173–201). NewYork: Nova
Science Publishers.
7. Ramon M, Bobak AK, & White D (2019). Super-recognizers: From the lab to the world and back again.
Br J Psychol: 110, 461–479. https://doi.org/10.1111/bjop.12368 PMID: 30893478
8. Woodhead MM & Baddeley AD (1981). Individual differences and memory for faces, pictures, and
words. Mem Cognit: 9, 368–370. https://doi.org/10.3758/bf03197561 PMID: 7278625
9. Megreya AM & Burton AM (2006). Unfamiliar faces are not faces: Evidence from a matching task. Mem
& Cognit: 34, 865–876. https://doi.org/10.3758/BF03193433
PLOS ONE | https://doi.org/10.1371/journal.pone.0243209 December 10, 2020
22 / 26
PLOS ONE Emotion regulation and face recognition
10. Bruce V, Henderson Z, Greenwood K, Hancock PJB, Burton AM et al. (1999). Verification of face identi-
ties from images captured on video. J Exp Psychol Appl: 5, 339–360. https://doi.org/10.1037//1076-
898X.5.4.339
11. Burton AM, White D & McNeill A (2010). The Glasgow Face Matching Test. Behav Res Methods: 42,
286–291. https://doi.org/10.3758/BRM.42.1.286 PMID: 20160307
12. Bindemann M, Avetisyan M & Rakow T (2012). Who can recognize unfamiliar faces? Individual differ-
ences and observer consistency in person identification. J Exp Psychol Appl: 18, 277–91. https://doi.
org/10.1037/a0029635 PMID: 22905851
13. Wilhelm O, Herzmann G, Kunina O, Danthiir V, Schacht A & Sommer W (2010). Individual differences
in perceiving and recognizing faces—One element of social cognition. J Pers Soc Psychol: 99, 530–48.
https://doi.org/10.1037/a0019972 PMID: 20677889
14. Wilmer JB, Germine L, Chabris CF, Chatterjee G, Williams M, Loken E, et al. (2010). Human face rec-
ognition ability is specific and highly heritable. Proc Natl Acad Sci U S A: 16, 107, 5238–41. https://doi.
org/10.1073/pnas.0913053107 PMID: 20176944
15. Kokje E, Bindemann M, & Megreya AM (2018). Cross-race correlations in the abilities to match unfamil-
iar faces. Acta Psychol: 185,13–21. https://doi.org/10.1016/j.actpsy.2018.01.006 PMID: 29407241
16. Megreya AM & Bindemann M (2009). Revisiting the processing of internal and external features of unfa-
miliar faces: The headscarf effect. Perception: 38, 1831–1848. https://doi.org/10.1068/p6385 PMID:
20192132
17. Megreya AM & Bindemann M (2018). Feature instructions improve face-matching accuracy. PLoS
One: 15, 13(3):e0193455. https://doi.org/10.1371/journal.pone.0193455 PMID: 29543822
18. Megreya AM, White D & Burton AM (2011). The other race effect does not rely on memory: Evidence
from a matching task. Q J Exp Psychol: 64, 1473–1483. https://doi.org/10.1080/17470218.2011.
575228
19. Germine L. T., Duchaine B., & Nakayama K. (2011). Where cognitive development and aging meet:
Face learning ability peaks after age 30. Cognition, 118, 201–210 https://doi.org/10.1016/j.cognition.
2010.11.002 PMID: 21130422
20. Megreya AM & Bindemann M (2015). Developmental improvement and age-related decline in unfamil-
iar face matching. Perception: 44, 5–22. https://doi.org/10.1068/p7825 PMID: 26489213
21.
Lewin C & Herlitz A (2002). Sex differences in face recognition—Women’s faces make the difference.
Brain Cogn: 50,121–8. https://doi.org/10.1016/s0278-2626(02)00016-7 PMID: 12372357
22. Megreya AM & Bindemann M (2012). Identification accuracy for single- and double-perpetrator crimes:
Does accomplice gender matter? Br J Psychol: 103, 439–53. https://doi.org/10.1111/j.2044-8295.
2011.02084.x PMID: 23034106
23. Megreya AM, Bindemann M & Havard C (2011). Sex differences in unfamiliar face identification: Evi-
dence from matching tasks. Acta Psychol: 137, 83–9. https://doi.org/10.1016/j.actpsy.2011.03.003
PMID: 21459354
24. Rehnman J & Herlitz A (2006). Higher face recognition ability in girls: Magnified by own-sex and own-
ethnicity bias. Memory, 14, 289−296. https://doi.org/10.1080/09658210500233581 PMID: 16574585
25. Verhallen RJ, Bosten JM, Goodbourn PT, Lawrance-Owen AJ, Bargary G & Mollon JD (2017). General
and specific factors in the processing of faces. Vision Res, 141, 217–227. https://doi.org/10.1016/j.
visres.2016.12.014 PMID: 28077292
26. McCaffery JM, Robertson DJ, Young AW, & Burton AM. (2018). Individual differences in face identity
processing. Cogn Res Princ Implic: 3:21. https://doi.org/10.1186/s41235-018-0112-9 PMID: 30009251
27. Robertson DJ, Black J, Chamberlain B, Megreya AM & Davis JP (2020). Super-recognisers show an
advantage for other race face identification. Appl Cognitive Psych, 34, 205–216. https://doi.org/10.
1002/acp.3608
28. Stacchi L, Huguenin-Elie E, Caldara R, & Ramon M (2020). Normative data for two tests of face match-
ing under ecological conditions. Cogn Res Princ Implic: 5(1):8. https://doi.org/10.1186/s41235-019-
0205-0 PMID: 32076893
29. Megreya AM & Burton AM (2007). Hits and false positives in face matching: A familiarity-based dissoci-
ation. Percept Psychophys: 69, 1175–1184. https://doi.org/10.3758/bf03193954 PMID: 18038955
30. Bindemann M, Brown C, Koyas T & Russ A (2012). Individual differences in face identification postdict
eyewitness accuracy. J. Appl. Res. Mem. Cogn, 1, 96–103. https://doi.org/10.1016/j.jarmac.2012.02.
001
31. Xu B, Liu-Shuang J, Rossion B, & Tanaka J (2017). Individual differences in face identity processing
with fast periodic visual stimulation. J Cogn Neurosci: 29,1368–1377. https://doi.org/10.1162/jocn_a_
01126 PMID: 28358660
PLOS ONE | https://doi.org/10.1371/journal.pone.0243209 December 10, 2020
23 / 26
PLOS ONE Emotion regulation and face recognition
32. Dennett HW, McKone E, Tavashmi R, Hall A, Pidcock M, Edwards M, et al (2012). The Cambridge Car
Memory Test: A task matched in format to the Cambridge Face Memory Test, with norms, reliability,
sex differences, dissociations from face memory, and expertise effects. Behav Res Methods: 44,587–
605. https://doi.org/10.3758/s13428-011-0160-2 PMID: 22012343
33. Rose SA, Feldman JF & Jankowski JJ (2003). Infant visual recognition memory: Independent contribu-
tions of speed and attention. Dev Psychol: 39, 563–71. https://doi.org/10.1037/0012-1649.39.3.563
PMID: 12760523
34. Rose SA & Feldman JF (1995). Prediction of IQ and specific cognitive abilities at 11 years from infancy
measures. Dev. Psychol, 31, 685–696. https://doi.org/10.1037/0012-1649.31.4.685
35. Kanwisher N (2000). Domain specificity in face perception. Nat Neurosci: 3,759–63. https://doi.org/10.
1038/77664 PMID: 10903567
36.
Li J, Tian M, Fang H, Xu M, Li H & Liu J (2010). Extroversion predicts individual differences in face rec-
ognition. Commun Integr Biol: 3, 295–8. https://doi.org/10.4161/cib.3.4.12093 PMID: 20798810
37. Saito T, Nakamura T & Endo T (2005). Big five personality factors related to face recognition. Jap J.
Psychol: 75, 517–522. https://doi.org/10.4992/jjpsy.75.517 PMID: 15782589
38. Cheung CHM, Rutherford HJV, Mayes LC, & McPartland JC. (2010). Neural responses to faces reflect
social personality traits. Soc Neurosci: 5, 351–359. https://doi.org/10.1080/17470911003597377
PMID: 20544557
39. Bothwell RK, Brigham JC & Pigott MA (1987). An exploratory study of personality differences in eyewit-
ness memory J Soc Behav Pers: 2, 335–343.
40. Pozzulo JD, Crescini C, Lemieux JMT, & Tawfik A (2007). The effect of shyness on eyewitness memory
and the susceptibility to misinformation. Pers Individ Dif: 43, 1656–1666. https://doi.org/10.1016/j.paid.
2007.05.001
41.
42.
Thompson WB & Mueller JH (1984). Face Memory and Hemispheric Preference: Emotionality and
Extroversion. Brain & Cogn: 3, 239–248. https://doi.org/10.1016/0278-2626(84)90019-8
Trouve´ RJ & Libkuman TM (1992). Eyewitness performance of personality types as a function of
induced arousal. Am J Psychol: 105, 417–433. https://doi.org/10.2307/1423196
43. Ward RA & Loftus EF (1985). Eyewitness performance in different psychological types. The J Gen Psy-
chol: 112,191–200. https://doi.org/10.1080/00221309.1985.9711003 PMID: 4056764
44. Megreya AM & Bindemann M (2013). Individual differences in personality and face identification. J
Cogn Psychol: 25, 30–37. https://doi.org/10.1080/20445911.2012.739153
45.
Lander K & Poyarekar S (2015). Famous face recognition, face matching, and extraversion. Q J Exp
Psychol: 68,1769–1776. https://doi.org/10.1080/17470218.2014.988737
46. Brigham JC, Maass A, Martinez D & Whittenberger G (1983). The effect of arousal on facial recognition.
Basic Appl Soc Psychol: 4, 279–293. https://doi.org/10.1207/s15324834basp0403_6
47. Hills PJ, Dickinson D, Daniels LM, Boobyer CA & Burton R (2019). Being observed caused physiological
stress leading to poorer face recognition. Acta Psychol: 196, 118–128. https://doi.org/10.1016/j.actpsy.
2019.04.012 PMID: 31054376
48. Deffenbacher KA, Bornstein BH, Penrod SD & McGorty K (2004). A meta-analytic review of the effects
of high stress on eyewitness memory. Law Hum Behav: 28, 687–706. https://doi.org/10.1007/s10979-
004-0565-x PMID: 15732653
49. Mueller JH, Bailis KL & Goldstein AG (1979). Depth of processing and anxiety in facial recognition. Br J
Psychol: 70, 511–515. https://doi.org/10.1111/j.2044-8295.1979.tb01724.x PMID: 509018
50. Nowicki S, Winograd E & Millard BA (1979). Memory for faces: A social learning analysis. J Res Pers:
13, 460–468. https://doi.org/10.1016/0092-6566(79)90008-4
51. Valentine T & Mesout J (2009). Eyewitness identification under stress in the London Dungeon. App
Cogn Psychol, 23, 151–161. https://doi.org/10.1002/acp.1463
52. Bobak AK, Pampoulov P, & Bate S (2016). Detecting superior face recognition skills in a large sample
of young British adults. Front Psychol: 22, 7:1378. https://doi.org/10.3389/fpsyg.2016.01378 PMID:
27713706
53. Davis JM, McKone E, Dennett H, O’Connor KB, O’Kearney R & Palermo R et al. (2011) Individual differ-
ences in the ability to recognise facial identity are associated with social anxiety. PLoS ONE: 6(12):
e28800. https://doi.org/10.1371/journal.pone.0028800 PMID: 22194916
54. Bate S, Parris B, Haslam C & Kay J (2010). Socio-emotional functioning and face recognition ability in
the normal population. Pers Individ Dif: 48, 239–242. https://doi.org/10.1016/j.paid.2009.10.005
55. Gross JJ (2014). Emotion regulation: conceptual and empirical foundations. In Gross JJ (Ed) Handbook
of Emotion Regulation (pp. 3–20), second edition. New York, NY: The Guilford Press.
PLOS ONE | https://doi.org/10.1371/journal.pone.0243209 December 10, 2020
24 / 26
PLOS ONE Emotion regulation and face recognition
56. Kappas A (2011). Emotion and regulation are one! EMR: 3, 17–25. https://doi.org/10.1177/
1754073910380971
57.
Thompson RA (2011). Emotion and Emotion Regulation: Two Sides of the Developing Coin. EMR, 1,
53–61. https://doi.org/10.1177/1754073910380969
58. Gross JJ (1998). Antecedent- and response-focused emotion regulation: Divergent consequences for
experience, expression, and physiology. J Pers Soc Psychol: 74, 224–37. https://doi.org/10.1037//
0022-3514.74.1.224 PMID: 9457784
59. Gross JJ (2001). Emotion regulation in adulthood: Timing is everything. Curr Dir Psychol Sci: 10, 214–
219. https://doi.org/10.1111/1467-8721.00152
60. Garnefski N, Kraaij V & Spinhoven P (2001). Negative life events, cognitive emotion regulation, and
emotional problems. Pers Individ Dif: 30, 1311–1327. https://doi.org/10.1016/S0191-8869(00)00113-6
61. Aldao A, Nolen-Hoeksema S & Schweizer S (2010). Emotion-regulation strategies across psychopa-
thology: a meta-analytic review. Clin Psychol Rev: 30, 217–37. https://doi.org/10.1016/j.cpr.2009.11.
004 PMID: 20015584
62. Richards JM & Gross JJ (2000). Emotion regulation and memory: The cognitive costs of keeping one’s
cool. J Pers Soc Psychol: 79, 410–24. https://doi.org/10.1037//0022-3514.79.3.410 PMID: 10981843
63. Marceau E, Kelly P & Solowij N (2018). The relationship between executive functions and emotion regu-
lation in females attending therapeutic community treatment for substance use disorder. Drug Alcohol
Depend: 182, 58–66. https://doi.org/10.1016/j.drugalcdep.2017.10.008 PMID: 29154148
64. McRae K, Jacobs SE, Ray RD, John OP & Gross JJ (2012). Individual differences in reappraisal ability:
Links to reappraisal frequency, well-being, and cognitive control. J Res Pers: 46, 2–7. https://doi.org/
10.1016/j.jrp.2011.10.003
65. Schmeichel BJ, Volokhov RN & Demaree HA (2008). Working memory capacity and the self-regulation
of emotional expression and experience. J Pers Soc Psychol: 95,1526–1540. https://doi.org/10.1037/
a0013345 PMID: 19025300
66. Schweizer S, Grahn J, Hampshire A, Mobbs D & Dalgleish T (2013). Training the emotional brain:
Improving affective control through emotional working memory training. J Neurosci: 33, 5301–5311.
https://doi.org/10.1523/JNEUROSCI.2593-12.2013 PMID: 23516294
67. Xiu L, Zhou R & Jiang Y (2016).Working memory training improves emotion regulation ability: Evidence
from HRV. Physiol Behav: 155, 25–29. https://doi.org/10.1016/j.physbeh.2015.12.004 PMID:
26679738
68. Xiu L, Wu J, Chang L & Zhou R (2018). Working Memory Training Improves Emotion Regulation Ability.
Sci Rep: 8,15012. https://doi.org/10.1038/s41598-018-31495-2 PMID: 30301906
69. Druzgal TJ & D’esposito M (2001a). Activity in fusiform face area modulated as a function of working
memory load. Brain Res Cogn Brain Res: 10, 355–364. https://doi.org/10.1016/s0926-6410(00)00056-
2
70. Druzgal TJ & D’esposito M (2003). Dissecting contributions of prefrontal cortex and fusiform face area
to face working memory. J Cogn Neurosci: 15, 771–784. https://doi.org/10.1162/
089892903322370708 PMID: 14511531
71.
Jackson MC & Raymond JE (2008).Familiarity enhances visual working memory for faces. J Exp Psy-
chol Hum Percept Perform: 34, 556–568. https://doi.org/10.1037/0096-1523.34.3.556 PMID:
18505323
72. Megreya AM & Burton AM (2008). Matching faces to photographs: Poor performance in eyewitness
memory (without the memory). J Exp Psychol Appl: 14, 364–372. https://doi.org/10.1037/a0013464
PMID: 19102619
73. Megreya AM, Memon A & Havard C (2012). The headscarf effect: Direct evidence from eyewitness
identification paradigm. App Cogn Psychol: 26, 308–315. https://doi.org/10.1002/acp.1826
74. Gross JJ & John OP (2003). Individual differences in two emotion regulation processes: Implications for
affect, relationships, and well-being. J Pers Soc Psychol: 85, 348–62. https://doi.org/10.1037/0022-
3514.85.2.348 PMID: 12916575
75. Megreya AM, Latzman RD, Al-Emadi AA & Al-Attiyah AA (2018). An integrative model of emotion regu-
lation and associations with positive and negative affectivity across four Arabic speaking countries and
the USA. Motiv Emot: 42, 566–575. https://doi.org/10.1007/s11031-018-9682-6
76. Butler EA, Egloff B, Wlhelm FH, Smith NC, Erickson EA & Gross JJ (2003). The social consequences of
expressive suppression. Emotion, 3, 48–67. https://doi.org/10.1037/1528-3542.3.1.48 PMID:
12899316
77. Megreya AM, Sandford A & Burton AM (2013). Matching face images taken on the same day or months
apart: The limitations of photo-ID. Appl Cognitive Psych: 27, 700–706. https://doi.org/10.1002/acp.
2965
PLOS ONE | https://doi.org/10.1371/journal.pone.0243209 December 10, 2020
25 / 26
PLOS ONE Emotion regulation and face recognition
78. White D, Kemp RI, Jenkins R & Burton AM (2014). Feedback training for facial image comparison. Psy-
chon B Rev: 21, 100–106. https://doi.org/10.3758/s13423-013-0475-3 PMID: 23835616
79. Megreya AM, Latzman RD, Al-Attiyah AA & Alrashidi M (2016). The robustness of the Nine-Factor
Structure of the Cognitive Emotion Regulation Questionnaire across Four Arabic Speaking Middle East-
ern Countries. J Cross-Cult Psychol: 47, 875–890. https://doi.org/10.1177/0022022116644785
80. Garnefski N, Kraaij V & Spinhoven P (2002). Manual for the use of the Cognitive Emotion Regulation
Questionnaire: A questionnaire measuring cognitive coping strategies. Leiderdorp, The Netherlands,
DATEC.
81. Kivity Y & Huppert J D (2019). Emotion regulation in social anxiety: a systematic investigation and
meta-analysis using self-report, subjective, and event-related potentials measures. Cogn Emot: 33,
213–230. https://doi.org/10.1080/02699931.2018.1446414 PMID: 29514588
82. Garnefski N & Kraaij V (2007). The cognitive emotion regulation questionnaire. psychometric features
and prospective relationships with depression and anxiety in adults. Eur J Psychol Assess: 23, 141–
149. https://doi.org/10.1027/1015-5759.23.3.141
83.
Leigh E & Clark D M (2018). Understanding social anxiety disorder in adolescents and improving treat-
ment outcomes: Applying the cognitive model of Clark and Wells (1995). Clin Child Fam Psychol Rev:
21, 388–414. https://doi.org/10.1007/s10567-018-0258-5 PMID: 29654442
84. Mobini S, Reynolds S & Mackintosh B (2013). Clinical implications of cognitive bias modification for
interpretative biases in social anxiety: An integrative literature review. Cogn Ther Res: 37, 173–182.
https://doi.org/10.1007/s10608-012-9445-8
85. Pessoa L (2008). On the relationship between emotion and cognition. Nat Rev Neurosci: 9,148–58.
https://doi.org/10.1038/nrn2317 PMID: 18209732
86. Blair KS, Smith BW, Mitchell DGV, Morton J, Vythilingam M, et al., (2007). Modulation of emotion by
cognition and cognition by emotion. Neuroimage, 35, 430–440. https://doi.org/10.1016/j.neuroimage.
2006.11.048 PMID: 17239620
87. Megreya AM (2018). Feature-by-feature comparison and holistic processing in unfamiliar face match-
ing. PeerJ, 6, e4437. https://doi.org/10.7717/peerj.4437 PMID: 29503772
88.
Towler A, White D & Kemp RI (2017). Evaluating the feature comparison strategy for forensic face iden-
tification. J Exp Psychol Appl: 23, 47–58. https://doi.org/10.1037/xap0000108 PMID: 28045276
89. Kraaij V & Garnefski N (2019). The Behavioral Emotion Regulation Questionnaire: Development, psy-
chometric properties and relationships with emotional problems and the Cognitive Emotion Regulation
Questionnaire. Pers Individ Dif: 137, 56–61. https://doi.org/10.1016/j.paid.2018.07.036
90. Gratz K L & Roemer L (2004). Multidimensional assessment of emotion regulation and dysregulation:
Development, factor structure, and initial validation of the difficulties in emotion regulation scale. J Psy-
chopath Behav Assess: 26, 41–54. https://doi.org/10.1023/B:JOBA.0000007455.08539.94
PLOS ONE | https://doi.org/10.1371/journal.pone.0243209 December 10, 2020
26 / 26
PLOS ONE |
10.1371_journal.pone.0250044 | RESEARCH ARTICLE
Etiologies of genital inflammation and
ulceration in symptomatic Rwandan men and
women responding to radio promotions of
free screening and treatment services
1*, Julien Nyombayire2, Rachel Parker1, Rosine Ingabire2,
Kristin M. WallID
Jean Bizimana2, Jeannine Mukamuyango2, Amelia Mazzei2, Matt A. Price3, Marie
Aimee UnyuzimanaID
2, Amanda Tichacek1, Susan Allen1, Etienne Karita2
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
1 Rwanda Zambia HIV Research Group, Department of Pathology & Laboratory Medicine, School of
Medicine and Hubert Department of Global Health and Department of Epidemiology, Rollins School of Public
Health, Laney Graduate School, Emory University, Atlanta, Georgia, United States of America, 2 Project San
Francisco, Rwanda Zambia HIV Research Group, Kigali, Rwanda, 3 IAVI, NY, NY, University of California
San Francisco, San Francisco, CA, United States of America
OPEN ACCESS
Citation: Wall KM, Nyombayire J, Parker R,
Ingabire R, Bizimana J, Mukamuyango J, et al.
(2021) Etiologies of genital inflammation and
ulceration in symptomatic Rwandan men and
women responding to radio promotions of free
screening and treatment services. PLoS ONE
16(4): e0250044. https://doi.org/10.1371/journal.
pone.0250044
Editor: Antonella Marangoni, Universita degli Studi
di Bologna Scuola di Medicina e Chirurgia, ITALY
Received: January 13, 2021
Accepted: March 30, 2021
Published: April 20, 2021
Copyright: © 2021 Wall et al. This is an open
access article distributed under the terms of the
Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: Data are available
here: Wall, Kristin, 2021, "Replication Data for:
"Etiologies of genital inflammation and ulceration in
symptomatic Rwandan men and women
responding to radio promotions of free screening
and treatment services"", https://doi.org/10.7910/
DVN/CFX6UU, Harvard Dataverse.
Funding: This work was funded by the National
Institutes of Health (NIH) (NIAID R01 AI51231), the
* kmwall@emory.edu
Abstract
Introduction
The longstanding inadequacies of syndromic management for genital ulceration and inflam-
mation are well-described. The Rwanda National Guidelines for sexually transmitted infec-
tion (STI) syndromic management are not yet informed by the local prevalence and
correlates of STI etiologies, a component World Health Organization guidelines stress as
critical to optimize locally relevant algorithms.
Methods
Radio announcements and pharmacists recruited symptomatic patients to seek free STI
services in Kigali. Clients who sought services were asked to refer sexual partners and
symptomatic friends. Demographic, behavioral risk factor, medical history, and symptom
data were collected. Genital exams were performed by trained research nurses and physi-
cians. We conducted phlebotomy for rapid HIV and rapid plasma reagin (RPR) serologies
and vaginal pool swab for microscopy of wet preparation to diagnose Trichomonas vaginalis
(TV), bacterial vaginosis (BV), and vaginal Candida albicans (VCA). GeneXpert testing for
Neisseria gonorrhoeae (NG) and Chlamydia trachomatis (CT) were conducted. Here we
assess factors associated with diagnosis of NG and CT in men and women. We also explore
factors associated with TV, BV and VCA in women. Finally, we describe genital ulcer and
RPR results by HIV status, gender, and circumcision in men.
Results
Among 974 men (with 1013 visits), 20% were positive for CT and 74% were positive for NG.
Among 569 women (with 579 visits), 17% were positive for CT and 27% were positive for
NG. In multivariate analyses, factors associated with CT in men included younger age,
PLOS ONE | https://doi.org/10.1371/journal.pone.0250044 April 20, 2021
1 / 21
PLOS ONE NIH AIDS International Training and Research
Program Fogarty International Center (D43
TW001042); and the NIH-funded Emory Center for
AIDS Research (P30 AI050409). This work was
partially funded by IAVI with the generous support
of USAID and other donors; a full list of IAVI
donors is available at https://www.iavi.org. The
contents of this manuscript are the responsibility of
the authors and do not necessarily reflect the views
of USAID or the US Government. The funders had
no role in study design, data collection and
analysis, decision to publish, or preparation of the
manuscript.
Competing interests: No authors have competing
interests.
Etiologies of genital abnormalities in Rwandan men and women
responding to radio advertisements, <17 days since suspected exposure, and not having
dysuria. Factors associated with NG in men included not having higher education or full-
time employment, <17 days since suspected exposure, not reporting a genital ulcer, and
having urethral discharge on physical exam. Factors associated with CT in women included
younger age and < = 10 days with symptoms. Factors associated with NG in women
included younger age, lower education and lack of full-time employment, sometimes using
condoms vs. never, using hormonal vs. non-hormonal contraception, not having genital
ulcer or itching, having symptoms < = 10 days, HIV+ status, having BV, endocervical dis-
charge noted on speculum exam, and negative vaginal wet mount for VCA. In multivariate
analyses, only reporting >1 partner was associated with BV; being single and RPR+ was
associated with TV; and having < = 1 partner in the last month, being pregnant, genital itch-
ing, discharge, and being HIV and RPR negative were associated with VCA. Genital ulcers
and positive RPR were associated with being HIV+ and lack of circumcision among men.
HIV+ women were more likely to be RPR+. In HIV+ men and women, ulcers were more
likely to be herpetic rather than syphilitic compared with their HIV- counterparts.
Conclusions
Syndromic management guidelines in Rwanda can be improved with consideration of the
prevalence of confirmed infections from this study of symptomatic men and women repre-
sentative of those who would seek care at government health centers. Inclusion of demo-
graphic and risk factor measures shown to be predictive of STI and non-STI dysbioses may
also increase diagnostic accuracy.
Introduction
Globally, over 1 million new sexually transmitted infections (STI) occur each day [1]. The
prevalence of STI increased an estimated 59% in sub Saharan Africa between 1999 and 2005
and has continued to rise [2]. The World Health Organization (WHO) 2016–2021 Global
Health Sector Strategy on Sexually Transmitted Infections aims to reduce STI 90% by 2030
using “[epidemiologic] information for focused action” [3].
The association between genital ulceration and inflammation (GUI) due to STI and non-
STI etiologies and heterosexual HIV transmission and acquisition has been extensively studied
in Africa [4–12]. Broadly, in observational studies GUI is associated with both transmitting
and acquiring HIV in both men and women, and with transmission of more than one virion,
an otherwise rare event, in cohabiting heterosexual discordant couples which comprise one of
the largest HIV risk groups [6, 13–17].
Ulcerative STI that may facilitate HIV transmission include syphilis (Treponema pallidum,
TP), Herpes simplex virus (HSV), and chancroid (Haemophilus ducreyi, HD) [18–20]. Inflam-
matory STI that increase HIV transmission include gonorrhea (Neisseria gonorrhoeae, NG),
chlamydia (Chlamydia trachomatis, CT), and Trichomonas vaginalis (TV) [21–24]. Common
non-STI dysbioses associated with genital inflammation include bacterial vaginosis (BV) and
vaginal Candida albicans (VCA) [25–29].
Untreated TP, HD, HSV, NG, CT and TV can cause severe morbidity and, along with BV
and VCA (which are troublesome but non-invasive), can contribute to HIV transmission. In
our studies in African HIV discordant heterosexual couples, GUI contribute a substantial pop-
ulation attributable fraction of HIV transmission in both donor and recipient [15].
PLOS ONE | https://doi.org/10.1371/journal.pone.0250044 April 20, 2021
2 / 21
PLOS ONE Etiologies of genital abnormalities in Rwandan men and women
The longstanding inadequacies of syndromic management for GUI are well-described [30–
37] but this approach remains the default in many resource-limited settings in Africa due to
the high cost of molecular and culture-based diagnostics. The Rwanda National Guidelines for
HIV and STI syndromic management were last updated in 2019 but these guidelines are not
yet informed by the local prevalence and correlates of STI etiologies, a component WHO
guidelines stress as critical to optimize locally relevant algorithms. We have previously pub-
lished results of a survey of GUI among Female Sex Workers (FSW) in Kigali, but that study
lacked molecular diagnostics for NG and CT [38].
Here we contribute to the epidemiologic data needed to inform improved diagnostic and
treatment algorithms in Rwanda by exploring demographic, behavioral, medical history,
symptom, genital exam, and laboratory factors associated with molecular diagnosis of NG and
CT in men and women. We also explore factors associated with vaginal pathogens TV, BV and
VCA in women. Finally, we describe genital ulcer and rapid plasma reagin (RPR) results strati-
fied by gender, HIV status, and among men, by male circumcision status.
Methods
Ethics
This program was approved as non-research by the Rwandan National Ethics Committee.
This program was determined to be non-research by the Emory Institutional Review Board
criteria. Diagnostic and treatment were provided anonymously as free services.
Setting
Kigali, the capital of Rwanda, has a population of over 1 million people and an adult HIV prev-
alence of 4.3% [39]. Between January 2016 and August 2019, The Center for Family Health
Research (CFHR), a research site established in Kigali in 1986 and affiliated with Emory Uni-
versity in Atlanta, GA, USA, implemented a program for diagnosis and treatment of symptom-
atic GUI residents of Kigali. CFHR has worked closely with the Rwanda Ministry of Health
(MoH) on research for improved HIV and reproductive health care in government-run health
centers for many years [25, 40–43].
Patient recruitment
Patients were residents of Kigali, Rwanda and were recruited in three ways: radio announce-
ments, partner/friend referral, and pharmacist referral. Radio announcements were made in
Kinyarwanda, Rwanda’s vernacular, encouraging men and women with symptoms suggestive
of GUI (e.g., genital discharge, discomfort, ulcer) to seek free services at CFHR clinic and were
broadcast throughout Kigali. Clients who sought services were then asked to refer sexual part-
ners and symptomatic friends. Local pharmacists were alerted to the program and asked to
refer individuals seeking treatments for suggestive symptoms. There were no inclusion/exclu-
sion criteria applied to participant recruitment. Participants are representative of residents of
Kigali with genital symptoms who self-selected to receive care.
Data collection and diagnostic procedures
Demographics, behavioral risk factors, medical histories, and symptoms were collected using a
standard instrument (S1 Fig). This information was obtained during interviews conducted by
nurses who recorded data on paper and entered it into MS Access. Similarly, findings from
genital exams performed by trained physicians and nurses were recorded on paper and entered
into MS Access. Samples for laboratory testing were taken from all patients and included
PLOS ONE | https://doi.org/10.1371/journal.pone.0250044 April 20, 2021
3 / 21
PLOS ONE Etiologies of genital abnormalities in Rwandan men and women
phlebotomy for rapid HIV and RPR serologies and vaginal pool swab for microscopy of wet
preparation to diagnose TV, BV and VCA. GeneXpert testing for NG and CT (Cepheid, Sun-
nyvale USA) was conducted for all patients using endocervical swabs obtained from women
and either urethral swabs (when discharge was reported or noted on physical exam) or urine
samples from men. In collaboration with the MoH, CFHR developed a uniform alphanumeric
identifier to allow anonymous data recording.
Data analysis
Analyses were conducted with Statistical Analysis Software (SAS, Cary, NC). Frequencies of single
and multiple infections were stratified by gender and HIV status. Demographic, behavioral, medi-
cal history, physical exam, microscopy and serology results were tabulated by gender and by NG
and CT results. Bivariate and multivariate analyses of factors associated with NG or CT are pre-
sented in tables. Multivariable logistic regression models included variables associated with each
outcome at p<0.05 in bivariate analysis and then backward selection was applied. Prevalence
odds ratios (crude and adjusted, cPOR and aPOR, respectively) and 95% confidence intervals
(CIs) and 2-sided p-values are presented. Variable multi-collinearity was assessed. Repeated visits
by STI clients with new complaints were accounted for using the GENMOD procedure.
Bivariate and multivariate factors associated with vaginal pathogens TV, BV and VCA in
women were analyzed in analogous fashion with results summarized in text. Demographic,
behavioral, medical history, and HIV and RPR serology results were considered for model
inclusion. Finally, genital ulcer and RPR results were described by gender, HIV status, and
among men, by male circumcision status.
Results
Unless specified in text, p-values are <0.05 for comparisons with details presented in Tables.
Summary of GUI diagnosed in men and women (Table 1)
GeneXpert for NG and CT were provided to men during 1013 visits (974 unique men)
between March 2017 and February 2019. Men tested HIV+ during 5% of these visits. Preva-
lence of NG was 74% and prevalence of CT was 20%, with no differences by HIV status. In the
975 visits with RPR results, TP prevalence was significantly higher among HIV+ (13%) com-
pared with HIV- (5%) men. Nineteen percent of visits were negative for all pathogens, and
17% of visits had more than one infection identified.
GeneXpert for NG and CT were provided to women during 579 visits (569 unique women)
between March 2017 and February 2019. Women tested HIV+ during 13% of these visits.
Prevalence of NG was 26% and prevalence of CT was 17%, with higher prevalence of NG
among HIV+ women. The prevalence of TV (overall 13%) was higher in HIV+ women,
whereas the prevalence of VCA (overall 21%) was higher in HIV- women. In the 568 visits
with RPR results, TP prevalence was significantly higher among HIV+ (22%) compared with
HIV- (6%) women and having multiple pathogens identified was more prevalent among HIV
+ (36%) compared with HIV- (24%) women’s visits. Conversely, having no pathogen identi-
fied was more prevalent in HIV- (31%) versus HIV+ (18%) women’s visits.
Demographics and factors associated with CT and NG in men (Tables 2 and 3)
Men averaged 30.8 years of age, 77% were single, 64% had at least a secondary education, 55%
were employed full time, 22% reported more than one partner in the last 30 days and 57%
reported never using condoms in the past three months. The most common symptoms
PLOS ONE | https://doi.org/10.1371/journal.pone.0250044 April 20, 2021
4 / 21
PLOS ONE Etiologies of genital abnormalities in Rwandan men and women
Table 1. Distribution of pathogens identified in symptomatic men and women in Kigali, Rwanda.
Total
HIV+ (N = 54)
HIV- (N = 958)
p-value
Among all men (N = 1013 visits)�
None identified
CT
NG
CT and NG
Among men with RPR results (N = 975 visits)
None identified
TP
Any multiple infection
Among all women (N = 579 visits)
None identified
CT
NG
CT and NG
BV
TV
VCA
Among women with RPR results (N = 568 visits)
None identified
TP
Any multiple infection
N
196
204
751
138
184
52
164
N
176
98
152
45
113
72
118
169
46
146
Col %
19%
20%
74%
14%
19%
5%
17%
N
14
7
40
7
14
7
11
Col %
26%
13%
74%
13%
26%
13%
21%
N
182
196
711
131
170
45
153
Col %
19%
20%
74%
14%
18%
5%
17%
Total
HIV+ (N = 75)
HIV- (N = 504)
Col %
30%
17%
26%
8%
21%
13%
21%
30%
8%
26%
N
13
8
34
5
20
15
6
13
16
26
Col %
17%
11%
45%
7%
28%
20%
8%
18%
22%
36%
N
163
90
118
40
93
57
112
156
30
120
Col %
32%
18%
23%
8%
19%
12%
23%
31%
6%
24%
0.210
0.181
0.981
0.882
0.150
0.019
0.433
0.008
0.121
<0.0001
0.702
0.087
0.039
0.004
0.020
<0.0001
0.031
TP: Treponema pallidum, NG: Neisseria gonorrhoeae, CT: Chlamydia trachomatis, TV: Trichomonas vaginalis, BV: bacterial vaginosis, VCA: vaginal Candida albicans;
RPR: rapid plasma regain
�One man missing HIV status
https://doi.org/10.1371/journal.pone.0250044.t001
reported were urethral discharge (89%) and dysuria (80%). Physical findings included urethral
discharge in 91% and genital ulcer in 5% of men (Table 2).
Multivariate analyses (Table 3) showed younger age, responding to radio advertisements,
<17 days since suspected exposure, and not having dysuria as independent factors associated
with CT.
Multivariate analyses (Table 3) showed not having higher education or full-time employ-
ment, <17 days since suspected exposure, not reporting a genital ulcer, and urethral discharge
on physical exam as independent factors associated with NG.
HIV, RPR serologic results, and circumcision status were not associated with either CT or NG.
Demographics and factors associated with CT and NG in women (Tables 4
and 5)
The mean age women was 28.7, they had 1.3 children and desired 1.4 more on average, 54%
were single, 53% had a secondary education or more, 34% had full-time employment, 83%
reported < = 1 partner in the last 30 days and 63% reported never using condoms in the past
three months. Vaginal discharge was the most common presenting symptom (82%) and endo-
cervical inflammation or discharge was noted on 49% of speculum exams. (Table 4)
Multivariate analyses (Table 5) showed younger age and having symptoms < = 10 days as
independent factors associated with CT.
PLOS ONE | https://doi.org/10.1371/journal.pone.0250044 April 20, 2021
5 / 21
PLOS ONE Etiologies of genital abnormalities in Rwandan men and women
Table 2. Factors associated with CT or NG infection in men in Kigali, Rwanda (N = 1013).
Demographics
Age, continuous (years)
Referrer
Radio Advert
Friends/Walk-in/Pharmacy/Contact Partner/
Internet
Living and Marital Status
Married and Cohabiting
Single or Divorced/Separated/Widow
Education Level
None
Primary
Secondary
Higher
Employment Status
Full-time employment
Part-time/Student/Jobless
Sexual behaviors
Number of partners in last 30 days
None or one partner
More than one partner
Condom use during vaginal sex in the last
three months
No partners or always used condoms
Sometimes
Never
Number of days since sexual contact you
suspect STI was acquired from
< = 8
9–16
> = 17
Self-reported symptoms
Urethral discharge
Yes
No
Dysuria
Yes
No
Genital itching
Yes
No
Genital ulcer
Yes
No
Total
(N = 1013)
CT-infected
(N = 204)
CT-uninfected
(N = 809)
p-
value
NG-infected
(N = 751)
NG-uninfected
(N = 262)
p-value
n
/mean
30.8
Col%
/SD
7.1
n
/mean
29.4
Row%
/SD
5.6
n
/mean
31.1
Row%
/SD
n
/mean
Row%
/SD
7.4
0.001
30.5
7.0
n
/mean
31.6
Row%
/SD
7.3
0.029
688
325
232
781
25
339
454
193
552
459
704
203
27
363
517
331
288
323
895
114
810
199
67
854
41
878
68%
32%
23%
77%
2%
34%
45%
19%
55%
45%
78%
22%
3%
40%
57%
35%
31%
34%
89%
11%
80%
20%
7%
93%
4%
96%
151
53
33
171
1
66
89
47
122
82
138
45
4
69
110
83
58
50
188
16
153
51
14
171
6
183
22%
16%
14%
22%
4%
19%
20%
24%
22%
18%
20%
22%
15%
19%
21%
25%
20%
15%
21%
14%
19%
26%
21%
20%
15%
21%
537
272
199
610
24
273
365
146
430
377
566
158
23
294
407
248
230
273
707
98
657
148
53
683
35
695
0.037
0.011
0.095
0.095
0.422
0.555
0.010
0.081
0.034
0.864
0.336
78%
84%
86%
78%
96%
81%
80%
76%
78%
82%
80%
78%
85%
81%
79%
75%
80%
85%
79%
86%
81%
74%
79%
80%
85%
79%
488
263
156
595
16
267
343
123
392
357
518
163
14
279
387
292
235
177
717
32
599
150
39
649
13
681
71%
81%
67%
76%
64%
79%
76%
64%
71%
78%
74%
80%
52%
77%
75%
88%
82%
55%
80%
28%
74%
75%
58%
76%
32%
78%
200
62
76
186
9
72
111
70
160
102
186
40
13
84
130
39
53
146
178
82
211
49
28
205
28
197
0.001
0.006
0.001
0.015
0.051
0.015
<0.0001
<0.0001
0.680
0.001
<0.0001
29%
19%
33%
24%
36%
21%
24%
36%
29%
22%
26%
20%
48%
23%
25%
12%
18%
45%
20%
72%
26%
25%
42%
24%
68%
22%
Number of days with symptoms
1–5
385
41%
100
26%
285
74%
0.004
332
86%
53
14%
<0.0001
PLOS ONE | https://doi.org/10.1371/journal.pone.0250044 April 20, 2021
(Continued )
6 / 21
PLOS ONE Etiologies of genital abnormalities in Rwandan men and women
Table 2. (Continued)
Total
(N = 1013)
CT-infected
(N = 204)
CT-uninfected
(N = 809)
p-
value
NG-infected
(N = 751)
NG-uninfected
(N = 262)
p-value
Demographics
n
/mean
Col%
/SD
n
/mean
Row%
/SD
n
/mean
Row%
/SD
n
/mean
Row%
/SD
n
/mean
Row%
/SD
6–10
11–21
>21
Laboratory and physical exam
HIV Status
Positive
Negative
RPR Result
Positive
Negative
Urethral discharge
Yes
No
Genital ulcer
Yes
No
Circumcision status
Circumcised
Uncircumcised
254
192
105
54
958
52
923
858
87
46
898
524
259
27%
21%
11%
5%
95%
5%
95%
91%
9%
5%
95%
67%
33%
40
33
17
7
196
13
182
178
13
8
183
122
45
16%
17%
16%
13%
20%
25%
20%
21%
15%
17%
20%
23%
17%
214
159
88
47
762
39
741
680
74
38
715
402
214
84%
83%
84%
87%
80%
75%
80%
79%
85%
83%
80%
77%
83%
0.181
0.354
0.199
0.623
0.058
191
121
56
40
711
43
677
692
15
17
686
416
195
75%
63%
53%
74%
74%
83%
73%
81%
17%
37%
76%
79%
75%
63
71
49
14
247
9
246
166
72
29
212
108
64
25%
37%
47%
26%
26%
17%
27%
19%
83%
63%
24%
21%
25%
0.981
0.136
<0.0001
<0.0001
0.192
Not significant not shown include: Self-reported symptoms dyspareunia, unpleasant odor, abdominal pain, anal discharge, anal ulcer, anal warts, and sore throat; genital
exam results white accumulation, condyloma/warts, inguinal adenopathy >1cm unilateral and bilateral, inflammation, and testicular mass/tenderness
RPR: Rapid plasma reagin; STI: Sexually transmitted disease; NG: Neisseria gonorrhoeae, CT: Chlamydia trachomatis
https://doi.org/10.1371/journal.pone.0250044.t002
Multivariate analyses (Table 5) showed younger age, lower education and lack of full-time
employment, sometimes using condoms vs. never, using hormonal contraception vs. other or
no contraception, not having a genital ulcer or itching, having symptoms for < = 10 days, HIV
+ status, endocervical discharge noted on speculum exam, BV, and negative VCA as indepen-
dent factors associated with NG.
Factors associated with of BV, TV and VCA in women (not tabled)
Only reporting >1 partner remained independently associated with BV in multivariate analy-
ses (POR 2.21, p = 0.003). Factors associated with TV in multivariate analyses were being sin-
gle and RPR+ (aPOR 2.05, p = 0.009 and aPOR 2.37, p = 0.023, respectively). Factors
associated with VCA were having < = 1 partner in the last month (aPOR 4.26, p = 0.005),
being pregnant (aPOR 3.05, p = 0.002), always using condoms or not having sex in the last
three months vs. never using condoms (aPOR 2.42, p = 0.023), genital itching (aPOR 1.69,
p = 0.034), genital discharge (aPOR 2.56, p = 0.011), and being HIV and RPR negative (aPOR
2.93, p = 0.025 and aPOR 4.94, p = 0.031, respectively).
Genital ulcers in men and women (not tabled)
Reported and/or observed genital ulcers were more common among HIV+ (20%) compared
with HIV- (5%) men (p<0.001). Genital ulcers were noted during physical examination in
PLOS ONE | https://doi.org/10.1371/journal.pone.0250044 April 20, 2021
7 / 21
PLOS ONE Etiologies of genital abnormalities in Rwandan men and women
Table 3. Univariate and multivariate analysis of factors associated with CT or NG infection in men in Kigali, Rwanda (N = 1013).
Demographics
cPOR
95% CI
CT infection
aPOR
95% CI
p-
value
p-
value
cPOR
95% CI
p-value
aPOR
95% CI
p-value
NG infection
Age (per year increase)
0.96
0.94 0.99
0.001
0.96
0.94 0.98
0.001
0.98
0.96
1.00
0.029
Referrer
Radio Advert
1.44
1.02 2.04
0.038
1.44
1.01 2.07
0.046
ref
---
---
---
Friends/Walk-in/Pharmacy/Contact
ref
ref
1.76
1.28
2.43
0.001
Partner/Internet
Living and Marital Status
Married and Cohabiting
ref
Single or Divorced/Separated/Widow
1.69
1.13 2.54
0.011
Education Level
ref
---
---
---
1.56
1.14
2.15
0.006
None/Primary/Secondary
ref
1.84
1.32
2.58
0.000
2.39
1.57
3.63 <0.0001
Higher
Employment Status
Full-time employment
Part-time/Student/Jobless
Sexual behaviors
Number of partners in last 30 days
None or one partner
More than one partner
Condom use during vaginal sex in the last
3 months
No partners or always used condoms
Sometimes
Never
Number of days since sexual contact you
suspect STI was acquired from
0–16
> = 17
Self-reported symptoms
Urethral discharge
Yes
No
Dysuria
Yes
No
Genital itching
Yes
No
Genital ulcer
Yes
No
1.37
0.95 1.99
0.092
ref
---
---
---
ref
---
---
---
1.30
0.96 1.78
0.094
ref
---
---
---
ref
---
---
---
ref
ref
1.17
0.80 1.71
0.424
0.22 1.90
0.426
0.62 1.21
0.411
0.64
0.87
ref
1.45
1.09
1.92
0.011
1.51
1.05
2.17
0.028
ref
1.5
---
1.02
---
2.2
---
0.040
0.37
1.13
ref
0.17
0.83
---
0.8
1.54
---
0.012
0.450
---
1.61
1.13 2.30
0.009
1.64
1.15 2.35
0.007
4.68
3.43
3.37 <0.0001
3.29
2.30
4.7 <0.0001
ref
1.63
0.94 2.83
0.084
ref
ref
ref
ref
ref
---
---
---
ref
---
---
---
10.00
6.41
15.61 <0.0001
ref
---
---
---
ref
---
---
---
1.48
1.03 2.13
0.034
1.51
1.03 2.22
0.035
1.05
0.74
1.49
0.792
1.05
0.57 1.93
0.872
ref
ref
ref
---
---
---
2.24
1.35
3.72
0.002
ref
---
---
---
ref
---
---
---
1.53
0.63 3.70
0.345
7.50
3.79
14.85 <0.0001
4.50
2.22
9.13 <0.0001
Number of days with symptoms
1–10
> = 11
Laboratory and physical exam
HIV Status
Positive
1.39
0.97 1.98
0.075
ref
3.06
2.26
4.15 <0.0001
ref
---
---
---
0.58
0.26 1.31
0.189
0.99
0.53
1.86
0.980
PLOS ONE | https://doi.org/10.1371/journal.pone.0250044 April 20, 2021
(Continued )
8 / 21
PLOS ONE Etiologies of genital abnormalities in Rwandan men and women
Table 3. (Continued)
Demographics
cPOR
95% CI
CT infection
aPOR
95% CI
p-
value
p-
value
cPOR
95% CI
p-value
aPOR
95% CI
p-value
NG infection
Negative
RPR Result
Positive
Negative
Urethral discharge
Yes
No
Genital ulcer
Yes
No
ref
ref
---
---
---
1.30
0.68 2.52
0.429
ref
1.65
ref
0.82
---
3.3
---
0.158
---
1.49
0.81 2.75
0.204
19.94
11.12 35.76 <0.0001
16.38
7.28 36.89 <0.0001
ref
ref
---
---
---
ref
---
---
---
0.82
0.38 1.80
0.626
ref
ref
---
---
---
5.52
2.96
10.28 <0.0001
aPOR: Adjusted prevalence odds ratio; cPOR: Crude prevalence odds ratio; RPR: Rapid plasma reagin; CI: Confidence interval; STI: Sexually transmitted disease; NG:
Neisseria gonorrhoeae, CT: Chlamydia trachomatis
Not significant not shown include: Self-reported symptoms dyspareunia, unpleasant odor, abdominal pain, anal discharge, anal ulcer, anal warts, and sore throat; genital
exam results white accumulation, condyloma/warts, inguinal adenopathy >1cm
https://doi.org/10.1371/journal.pone.0250044.t003
19% of RPR+ and 4% of RPR- men and conversely 20% of men with ulcers were RPR+ com-
pared to 4% of men without ulcers (p<0.001). Among HIV+ men, none of the seven who were
RPR+ had reported and/or observed ulcers while 23% of 43 HIV+ RPR- men had ulcers
(p = 0.319). In contrast, among HIV- RPR+ men 21% had reported or observed ulcers com-
pared to only 4% of HIV-RPR- men (p<0.001). This suggests that ulcers among HIV+ men
were more likely herpetic while among HIV- men at least one fifth were syphilitic.
Although HIV- men were more likely to be circumcised than HIV+ men (67% vs. 58%) in
our program, this difference was not significant (p = 0.196). Among circumcised men, those
who were HIV+ were more likely to have ulcers (13% vs. 4%, p = 0.074) and to be RPR+ (20%
vs. 4%, p = 0.003). Among uncircumcised men, those who were HIV+ were also more likely to
have ulcers (27% vs. 7%, p = 0.001) while the difference in RPR+ results was not significant
(12% vs. 6%, p = 0.324).
Among women, the prevalence of reported or observed ulcers was not significantly differ-
ent by HIV serostatus (20% in HIV+ vs.14% p = 0.196). Genital ulcers were noted during phys-
ical examination for 28% of RPR+ women compared with 14% of RPR- women (p<0.001). As
with men, the association between RPR results and reported and/or observed ulcers differed in
HIV+ and HIV- women: 25% of HIV+RPR+ vs. 20% of HIV+RPR- had ulcers, p = 0.729, com-
pared with 37% of HIV-RPR+ vs. 13% of HIV-RPR- women having ulcers (p = 0.001).
Discussion
We found a high prevalence of NG and CT among symptomatic men and women in Kigali.
Among men, urethral discharge was strongly associated with a diagnosis of NG while dysuria
was not associated with either infection. Specific symptoms were less helpful in identifying NG
and CT among women. Physical exam findings, demographic variables and reported risk
behaviors were independently predictive of NG and/or CT in both men and women, as were
vaginal wet mount findings and HIV serologies among women. Among women, TV and BV
were associated with sexual risk behaviors but not with symptoms while VCA was associated
with vaginal itching and discharge and with low-risk profiles. There were complex inter-
PLOS ONE | https://doi.org/10.1371/journal.pone.0250044 April 20, 2021
9 / 21
PLOS ONE Etiologies of genital abnormalities in Rwandan men and women
Table 4. Factors associated with CT or NG infection in women in Kigali, Rwanda (N = 579).
Demographics
n
/mean
Col
%/SD
n
/mean
Row
%/SD
n
/mean
Row%
/SD
n
/mean
Row
%/SD
n
/mean
Row
%/SD
Age, continuous (years)
28.7
7.2
25.6
6.1
29.3
7.2
<0.0001
26.8
6.3
29.4
7.4
<0.0001
Total (N = 579)
CT-infected
(N = 98)
CT-uninfected
(N = 481)
p-value
NG-infected
(N = 152)
NG-uninfected
(N = 427)
p-value
Referrer
Radio Advert
Friends/Walk-in/Pharmacy/Contact Partner/
Internet
Living and Marital Status
Married and Cohabiting
Single or Divorced/Separated/Widow
Education Level
None
Primary
Secondary
Higher
Employment Status
Full-time employment
Part-time/Student/Jobless
Sexual behaviors
Number of partners in last 30 days
None or one partner
More than one partner
Condom use during vaginal sex in the last 3
months
No partners or always used condoms
Sometimes
Never
Number of days since sexual contact you suspect
STI was acquired from
< = 8
9–16
> = 17
Number of children under 18, continuous
Number of additional children desired,
continuous
Pregnant
Yes
No
Want more children in next two years
Yes
No
Family planning method among women not
pregnant and do not want more children in next
two years
284
295
268
311
25
242
246
66
199
379
444
88
35
163
334
46
78
409
1.3
1.4
48
528
125
419
49%
51%
46%
54%
4%
42%
42%
11%
34%
66%
83%
17%
7%
31%
63%
9%
15%
77%
1.3
1.1
8%
92%
23%
77%
37
61
34
64
2
38
46
12
30
68
70
20
6
34
50
8
20
61
1.0
1.6
8
90
20
74
13%
21%
13%
21%
8%
16%
19%
18%
15%
18%
16%
23%
17%
21%
15%
17%
26%
15%
1.1
1.1
17%
17%
16%
18%
247
234
234
247
23
204
200
54
169
311
374
68
29
129
284
38
58
348
1.3
1.4
40
438
105
345
87%
79%
87%
79%
92%
84%
81%
82%
85%
82%
84%
77%
83%
79%
85%
83%
74%
85%
1.3
1.2
83%
83%
84%
82%
Non-Hormonal Method (IUD/Condom/Tubal
268
66%
47
18%
221
82%
0.498
Ligation/Natural Method) or No Method
Hormonal Implant
Injectable
50
48
12%
12%
9
5
18%
10%
41
43
82%
90%
0.014
0.012
0.513
67
85
60
92
9
81
54
8
0.383
39
113
0.112
0.259
0.066
0.026
0.040
0.947
0.666
95
43
4
66
68
15
31
92
1.2
1.3
11
141
33
106
56
24
16
24%
29%
22%
30%
36%
33%
22%
12%
20%
30%
21%
49%
11%
40%
20%
33%
40%
22%
1.1
1.0
23%
27%
27%
26%
217
210
208
219
16
161
192
58
160
266
349
45
31
97
266
31
47
317
1.3
1.4
37
387
88
298
0.153
0.050
0.001
0.008
76%
71%
78%
70%
64%
67%
78%
88%
80%
70%
79% <0.0001
51%
89% <0.0001
60%
80%
67%
60%
78%
1.3
1.2
77%
73%
73%
74%
0.003
0.600
0.279
0.569
0.821
21%
212
79%
0.001
48%
33%
26
32
52%
67%
(Continued )
PLOS ONE | https://doi.org/10.1371/journal.pone.0250044 April 20, 2021
10 / 21
PLOS ONE Etiologies of genital abnormalities in Rwandan men and women
Table 4. (Continued)
Demographics
Pills
Family planning method and pregnancy
composite
Pregnant
Hormonal method (implant, injectable, pills)
Non-Hormonal (IUD/Condom/ Tubal Ligation/
Natural Method) or No Method
Self-reported symptoms
Vaginal discharge
Yes
No
Genital itching
Yes
No
Dysuria
Yes
No
Genital ulcer
Yes
No
Number of days with symptoms
1–5
6–10
11–21
>21
Laboratory and physical exam
HIV Status
Positive
Negative
RPR Result
Positive
Negative
Trichomonas
Positive
Negative
Candida
Positive
Negative
Bacterial vaginosis
Positive
Negative
Vaginal Inflammation or Discharge
Yes
No
Endocervical Inflammation or Discharge
Total (N = 579)
CT-infected
(N = 98)
n
/mean
40
Col
%/SD
10%
n
/mean
9
Row
%/SD
23%
CT-uninfected
(N = 481)
n
/mean
31
Row%
/SD
78%
p-value
NG-infected
(N = 152)
NG-uninfected
(N = 427)
p-value
n
/mean
12
Row
%/SD
30%
n
/mean
28
Row
%/SD
70%
48
139
388
475
101
320
254
266
311
64
508
72
77
131
257
75
504
46
522
72
491
118
437
113
438
469
69
8%
24%
67%
82%
18%
56%
44%
46%
54%
11%
89%
13%
14%
24%
48%
13%
87%
8%
92%
13%
87%
21%
79%
21%
79%
87%
13%
8
23
67
78
20
52
44
44
54
9
89
16
17
18
37
8
90
10
88
18
75
12
80
25
65
75
15
17%
17%
17%
16%
20%
16%
17%
17%
17%
14%
18%
22%
22%
14%
14%
11%
18%
22%
17%
25%
15%
10%
18%
22%
15%
16%
22%
40
116
321
397
81
268
210
222
257
55
419
56
60
113
220
67
414
36
434
54
416
106
357
88
373
394
54
0.979
0.412
0.732
0.793
0.489
0.170
0.121
0.401
0.038
0.035
0.062
0.232
83%
83%
83%
84%
80%
84%
83%
83%
83%
86%
82%
78%
78%
86%
86%
89%
82%
78%
83%
75%
85%
90%
82%
78%
85%
84%
78%
11
52
89
123
28
57
92
75
76
9
140
24
27
40
48
34
118
21
128
18
129
13
132
47
96
116
24
23%
37%
23%
26%
28%
18%
36%
28%
24%
14%
28%
33%
35%
31%
19%
45%
23%
46%
25%
25%
26%
11%
30%
42%
22%
25%
35%
37
87
299
352
73
263
162
191
235
55
368
48
50
91
209
41
386
25
394
54
362
105
305
66
342
353
45
0.003
0.704
77%
63%
77%
74%
72%
82% <0.0001
0.306
0.020
0.003
64%
72%
76%
86%
72%
67%
65%
69%
81%
55% <0.0001
77%
54%
75%
75%
74%
0.002
0.818
89% <0.0001
70%
58% <0.0001
78%
75%
65%
0.076
(Continued )
PLOS ONE | https://doi.org/10.1371/journal.pone.0250044 April 20, 2021
11 / 21
PLOS ONE Etiologies of genital abnormalities in Rwandan men and women
Table 4. (Continued)
Demographics
Yes
No
Genital Ulcer
Yes
No
Total (N = 579)
CT-infected
(N = 98)
n
/mean
Col
%/SD
n
/mean
262
275
48
481
49%
51%
9%
91%
55
35
9
78
Row
%/SD
21%
13%
19%
16%
CT-uninfected
(N = 481)
n
/mean
Row%
/SD
207
240
39
403
79%
87%
81%
84%
p-value
NG-infected
(N = 152)
NG-uninfected
(N = 427)
p-value
0.010
0.652
n
/mean
88
52
12
126
Row
%/SD
34%
19%
25%
26%
n
/mean
174
223
36
355
Row
%/SD
66%
81%
75%
74%
<0.001
0.857
Not significant not shown include: Self-reported symptoms anal discharge, anal ulcer, anal warts, and sore throat; genital exam results non-menstrual bleeding (cervix
and vagina), condyloma/warts (cervix and vagina), inguinal adenopathy >1cm unilateral and bilateral, adnexal tenderness and adnexal mass.
IUD: intrauterine device; RPR: Rapid plasma reagin; CI: Confidence interval; STI: Sexually transmitted disease; NG: Neisseria gonorrhoeae, CT: Chlamydia trachomatis
https://doi.org/10.1371/journal.pone.0250044.t004
relationships between HIV and RPR serologies and genital ulcers, and these were further influ-
enced by circumcision status among men. These findings exemplify the locally relevant data
that can inform approaches to diagnosis and treatment in Rwanda as called for by WHO. Our
models had good discrimination and use of these data may offer improvement over the current
algorithm recommended by the Rwandan National Guidelines.
As in other studies, syndromic management may perform better among men compared to
women due to the ease of detecting abnormalities on external genitalia and the high likelihood
of NG among men reporting urethral discharge [44]. Surprisingly, dysuria was as common as
discharge in men but contrary to conventional wisdom we did not find an association between
dysuria and NG or CT [45].
The most common presenting symptom among women was vaginal discharge which was
only associated with VCA and not with NG, CT, BV or TV. Genital itching was reported by
over half of patients and was also predictive of VCA. Itching was also useful in pointing away
from NG, as was reported ulcer. Gynecologic exam, specifically endocervical discharge, was
helpful in the diagnosis of NG. Interestingly, wet mount results were predictive NG (BV+,
VCA-), suggesting that these inexpensive and simple tests should be included in any workup
of symptomatic women. Despite extensive laboratory testing, we failed to find an etiology for a
substantial proportion of women seeking care. This may reflect poor sensitivity of microscopy
as well as non-infectious causes of symptoms. As has been noted elsewhere, factors associated
with NG were more useful in predicting infections than those for CT [46, 47].
For both men and women, younger age was predictive of both NG and CT and lower edu-
cation level and jobless or part-time employment status were predictive of NG. Interestingly,
number of partners was not independently associated with CT or NG. Most men and women
reported never using condoms and very few reported always using condoms. Women who
sometimes used condoms were at higher risk of NG than those who never used them. This
may be due to increased condom use in women with higher risk partners.
Genital ulcers were not a common presenting symptom and were not associated with RPR
results among HIV+ patients. RPR provided a diagnosis for 20% of ulcers among HIV- men
and 15% among HIV- women. As others in Africa have reported, HSV is the most likely diag-
nosis for RPR- ulcers which was more common among HIV+ patients [48]. Non-circumcision
among men is associated with HIV acquisition and with increased prevalence and incidence of
ulcerative STI [49–52]. We have previously shown a relationship between ulcers, smegma and
HIV acquisition in uncircumcised men [15]. Among HIV- men, those who were uncircum-
cised were not more likely to report ulcers or to be RPR+ than their circumcised counterparts.
PLOS ONE | https://doi.org/10.1371/journal.pone.0250044 April 20, 2021
12 / 21
PLOS ONE Etiologies of genital abnormalities in Rwandan men and women
Table 5. Univariate and multivariate analysis of factors associated with CT or NG infection in women in Kigali, Rwanda (N = 579).
cPOR
95% CI
p-value
aPOR
95% CI
p-value
cPOR
95% CI
p-value
aPOR
95% CI
p-value
CT infection
NG infection
Demographics
Age (per year increase)
0.91
0.88 0.95 <0.0001
0.90
0.86 0.94 <0.0001
0.95
0.92 0.97 < .001
0.93
0.89 0.97 <0.001
Referrer
Radio Advert
ref
ref
Friends/Walk-in/Pharmacy/Contact
1.74
1.11 2.72
0.015
1.31
0.91 1.90
0.150
Partner/Internet
Living and Marital Status
Married and Cohabiting
Other
Education Level
None/Primary
Secondary/Higher
Employment Status
Full-time employment
Part-time/Student/Jobless
Sexual behaviors
Number of partners in last 30 days
None or one partner
More than one partner
Condom use during vaginal sex in the last
3 months
No partners or always used condoms
Sometimes
Never
Number of days since sexual contact you
suspect STI was acquired from
0–8
9–16
> = 17
Number of children under 18 (per child
increase)
Number of additional children desired
(per child increase)
Family planning method and pregnancy
composite
Pregnant
Hormonal method (implant, injectable,
pills)
Non-Hormonal (IUD/Condom/Tubal
Ligation/Natural Method) or No Method
Self-reported symptoms
Vaginal discharge
Yes
No
Genital itching
Yes
No
Dysuria
ref
ref
1.78
1.13 2.80
0.012
1.46
1.00 2.13
0.048
ref
1.30
0.83 2.01
0.248
ref
2.09
1.44 3.04
0.000
2.13
1.30 3.48
0.003
ref
ref
ref
ref
1.23
0.77 1.96
0.383
1.76
1.16 2.66
0.008
1.95
1.12 3.39
0.019
ref
ref
1.56
0.89 2.75
0.119
3.53
2.19 5.69 <0.0001
0.46 2.97
0.92 2.42
0.741
0.107
0.53 2.69
1.1
3.49
0.666
0.022
1.17
1.49
ref
1.20
1.96
ref
0.15
1.5
0.207
1.81 4.18 <0.0001
0.22 2.41
1.07 2.98
0.611
0.025
0.74
1.79
ref
0.87 3.16
1.36 3.87
0.126
0.002
0.48
2.75
ref
1.66
2.29
ref
0.82
0.69 0.99
0.037
0.96
0.84 1.10
0.594
1.22
1.02 1.45
0.027
0.92
0.79 1.07
0.274
0.96
0.95
0.43 2.14
0.56 1.59
0.915
0.837
1.00
2.01
0.49 2.03
1.32 3.05
0.999
0.001
1.30
1.73
0.57 2.99
1.02 2.94
0.532
0.040
ref
ref
ref
ref
1.26
0.73 2.18
0.408
1.10
0.68 1.78
0.692
ref
ref
ref
ref
1.08
0.69 1.68
0.738
2.62
1.79 3.84 <0.0001
2.54
1.55 4.17
0.0002
PLOS ONE | https://doi.org/10.1371/journal.pone.0250044 April 20, 2021
(Continued )
13 / 21
PLOS ONE Etiologies of genital abnormalities in Rwandan men and women
Table 5. (Continued)
Yes
No
Genital ulcer
Yes
No
Number of days with symptoms
1–10
11 or more
Laboratory and physical exam
HIV Status
Positive
Negative
RPR Result
Positive
Negative
Trichomonas
Positive
Negative
Candida
Positive
Negative
Bacterial vaginosis
Positive
Negative
Vaginal Inflammation OR Discharge
Yes
No
Endocervical Inflammation or Discharge
Yes
No
Genital Ulcer
Yes
No
cPOR
95% CI
p-value
aPOR
95% CI
p-value
cPOR
95% CI
p-value
aPOR
95% CI
p-value
CT infection
NG infection
ref
1.06
0.68 1.64
0.796
ref
1.21
0.84 1.75
0.303
ref
ref
ref
1.30
0.62 2.73
0.489
2.33
1.12 4.84
0.024
2.52
1.09 5.80
0.030
1.72
1.06 2.78
0.027
1.76
1.07 2.88
0.026
1.76
1.16 2.68
0.008
1.78
1.05 3.00
0.032
ref
ref
ref
ref
ref
2.73
1.66 4.47 <0.0001
2.05
1.10 3.83
0.024
1.83
0.85 3.96
0.124
ref
ref
1.37
0.66 2.88
0.401
2.58
1.41
4.7
0.002
ref
1.85
1.03 3.32
0.041
ref
ref
ref
ref
1.06
0.60 1.88
0.838
ref
ref
1.98
1.04 3.77
0.038
3.56
1.89 6.69 <0.0001
2.20
1.11 4.36
0.024
1.63
0.98 2.72
0.063
2.63
1.67 4.15 <0.0001
1.89
1.07 3.34
0.028
ref
ref
ref
ref
ref
1.47
0.79 2.75
0.2201
1.67
0.97 2.86
0.063
1.83
1.15 2.91
0.010
2.17
1.46 3.23
0.000
1.80
1.11 2.93
0.018
ref
1.19
0.56 2.56
0.649
ref
ref
ref
ref
1.06
0.53 2.10
0.875
IUD: intrauterine device; aPOR: Adjusted prevalence odds ratio; cPOR: Crude prevalence odds ratio; RPR: Rapid plasma reagin; CI: Confidence interval; STI: Sexually
transmitted disease; NG: Neisseria gonorrhoeae, CT: Chlamydia trachomatis
https://doi.org/10.1371/journal.pone.0250044.t005
In contrast, among HIV+ men, those who were uncircumcised were more likely to have an
ulcer and less likely to be RPR+ than circumcised men. Circumcision is widely promoted in
Rwanda and available at no cost in most government health centers as part of HIV prevention
services. Though the focus is on protecting HIV- men, our results here suggest that circumci-
sion can benefit HIV+ men by reducing ulcer incidence [53].
It is likely that we missed other less common ulcer etiologies including HD, lymphogranu-
loma venereum (LGV), and granuloma inguinale (Klebsiella granulomatis) [54]. Our clinicians
did suspect chancroid in a few cases, but the service program did not record detailed descrip-
tions or photographs of ulcers and we lacked laboratory diagnostics. The most recent
PLOS ONE | https://doi.org/10.1371/journal.pone.0250044 April 20, 2021
14 / 21
PLOS ONE Etiologies of genital abnormalities in Rwandan men and women
publication presenting confirmed chancroid diagnoses in Rwanda was based on data collected
in 1992, which found 27% of ulcers in men and 20% in women had culture-confirmed HD
[55–59]. For many years the prevalence of HD had been decreasing in much of Africa [48, 54],
but recent publications indicate HD may be staging a comeback [21]. More investigations are
needed in Rwanda.
Physical exam findings made important contributions in our program. Examination of
male genitalia does not require specialized equipment, but speculum exam requires a skilled cli-
nician, a gynecologic exam table and light which are in limited supply in low resource settings.
While genital exams would not be feasible for all symptomatic patients, targeted genital exams
in specific circumstances would be feasible and potentially very useful. Distinguishing between
vaginal and endocervical discharges would greatly improve diagnostic accuracy and bi-manual
exam would identify pelvic inflammatory disease. Similarly, in our setting where less than one
in five ulcer patients are RPR+, assessing ulcer characteristics may be worthwhile. Visual exam
has traditionally been viewed as unreliable as many ulcers do not have a paradigmatic presenta-
tion (e.g. painless ‘clean’ TP ulcer, painful ‘dirty’ HD with inguinal adenopathy, multiple
chronic or recurrent shallow vesicular HSV lesions). However, a recent study in Jamaica com-
pared clinical diagnosis with M-PCR and found visual diagnoses of TP, HSV, and HD were
67.7%, 53.8%, and 75% sensitive and 91.2%, 83.6%, and 75.4% specific, respectively [60].
The advent of point-of-care diagnostics for NG and CT has transformed STI diagnosis, but
given relatively expensive equipment and reagents, this remains out of reach in many low
resource settings. We have used pooling to reduce the per-patient cost in Zambia and this
could be explored in other settings [61]. GeneXpert kits are also available for TV and they are
more sensitive than microscopy. The US CDC has in-house multiplex PCR (M-PCR) for ulcer
etiologies including syphilis, HSV and chancroid. A focused study would provide prevalence
information that could inform the next update of national guidelines.
Our program has several limitations. Social desirability bias may have led to under-report-
ing of risky sexual behaviors. We focused on symptomatic men and women and thus missed
the many people who are asymptomatically infected [62, 63]. We did not screen for active viral
hepatitis as recent unpublished surveys have shown a low prevalence of both hepatitis B and C
(4% and 3%, respectively reported nationally, 4% and 5% among female sex workers tested in
our laboratory). We did not have funding or resources to perform any direct method of detec-
tion for TP using ulcer material, and thus may have misclassified some recently infected people
who were negative by RPR test. While we did treat TV in male partners referred by TV
+ women, we did not systematically test for TV in men. Microscopy for TV detection in men
is extremely insensitive, and we did not have resources to conduct GeneXpert testing for TV.
TV could therefore be the reason for a portion of the symptomatic men with unknown etiol-
ogy. We did not include HSV serologies because adult seroprevalence is high [64]. Assessment
of cervical intraepithelial neoplasia requires more resources than would be achievable on a
large scale in Rwandan health centers so we did not address this important problem. Fortu-
nately, 93% of Rwandan girls now receive the human papillomavirus vaccine and future gener-
ations will be protected [65]. Lastly, we and others have published an association between
female genital schistosomiasis and HIV [66, 67], but this is most commonly seen with S.Hae-
matobium while only S.Mansoni is endemic in Rwanda, thus we did not screen for genital
schistosomiasis [68].
Conclusions
Syndromic management guidelines in Rwanda can be improved with consideration of the
prevalence of confirmed infections from this program offering services to symptomatic men
PLOS ONE | https://doi.org/10.1371/journal.pone.0250044 April 20, 2021
15 / 21
PLOS ONE Etiologies of genital abnormalities in Rwandan men and women
and women representative of those who would seek care at government health centers. Our
findings indicate that syndromic management performs better among men but is poor among
women. Inclusion of demographic and risk factor measures shown to be predictive of STI and
non-STI dysbioses may also increase diagnostic accuracy. In symptomatic women, wet mount
results for BV and VCA may help diagnose NG and are inexpensive and could be offered for
management of women. Targeted genital exams for women in specific circumstances (e.g., in
women without genital itching) may also be useful to diagnose NG. More data is needed on
how often local prevalence and epidemiology should be reassessed to maintain improved syn-
dromic management.
Supporting information
S1 Fig. STI baseline clinical form.
(DOCX)
Author Contributions
Conceptualization: Julien Nyombayire, Rosine Ingabire, Susan Allen, Etienne Karita.
Data curation: Kristin M. Wall, Julien Nyombayire, Rachel Parker, Susan Allen.
Formal analysis: Kristin M. Wall, Rachel Parker.
Funding acquisition: Susan Allen.
Investigation: Julien Nyombayire, Rosine Ingabire, Jean Bizimana, Jeannine Mukamuyango,
Amelia Mazzei, Matt A. Price, Marie Aimee Unyuzimana, Susan Allen, Etienne Karita.
Methodology: Kristin M. Wall, Jean Bizimana, Matt A. Price, Marie Aimee Unyuzimana,
Amanda Tichacek, Susan Allen, Etienne Karita.
Project administration: Julien Nyombayire, Rosine Ingabire, Jean Bizimana, Jeannine Muka-
muyango, Amelia Mazzei, Amanda Tichacek, Susan Allen, Etienne Karita.
Resources: Susan Allen, Etienne Karita.
Supervision: Julien Nyombayire, Rosine Ingabire, Jean Bizimana, Jeannine Mukamuyango,
Amelia Mazzei, Marie Aimee Unyuzimana, Amanda Tichacek, Susan Allen, Etienne Karita.
Validation: Julien Nyombayire, Rachel Parker, Rosine Ingabire, Jeannine Mukamuyango,
Amelia Mazzei, Matt A. Price, Marie Aimee Unyuzimana, Amanda Tichacek, Susan Allen.
Writing – original draft: Kristin M. Wall, Susan Allen.
Writing – review & editing: Julien Nyombayire, Rachel Parker, Rosine Ingabire, Jean Bizi-
mana, Jeannine Mukamuyango, Amelia Mazzei, Matt A. Price, Marie Aimee Unyuzimana,
Amanda Tichacek, Susan Allen, Etienne Karita.
References
1. World_Health_Organization. Sexually transmitted infections (STIs) Geneva: WHO; 2019 [updated 14
June 2019]. Available from: https://www.who.int/news-room/fact-sheets/detail/sexually-transmitted-
infections-(stis).
2. World_Health_Organization. Sexually Transmiited Infections Geneva: World_Health_Organization;
2019 [updated July 2019; cited 2020 Jan 30]. Available from: https://www.who.int/news-room/fact-
sheets/detail/sexually-transmitted-infections-(stis).
PLOS ONE | https://doi.org/10.1371/journal.pone.0250044 April 20, 2021
16 / 21
PLOS ONE Etiologies of genital abnormalities in Rwandan men and women
3. World_Health_Organization. Global health sector strategy on Sexually Transmitted Infections, 2016–
2021: WHO; 2016 [updated July 2016]. Available from: http://www.who.int/reproductivehealth/
publications/rtis/ghss-stis/en/.
4. Carlson JM, Schaefer M, Monaco DC, Batorsky R, Claiborne DT, Prince J, et al. HIV transmission.
Selection bias at the heterosexual HIV-1 transmission bottleneck. Science (New York, NY). 2014; 345
(6193):1254031. Epub 2014/07/12. https://doi.org/10.1126/science.1254031 PMID: 25013080;
PubMed Central PMCID: PMC4289910.
5. Mujugira A, Magaret AS, Baeten JM, Celum C, Lingappa J. Risk Factors for HSV-2 Infection among
Sexual Partners of HSV-2/HIV-1 Co-Infected Persons. BMC Res Notes. 2011; 4:64. Epub 2011/03/17.
https://doi.org/10.1186/1756-0500-4-64 PMID: 21406077; PubMed Central PMCID: PMC3064615.
6. Haaland RE, Hawkins PA, Salazar-Gonzalez J, Johnson A, Tichacek A, Karita E, et al. Inflammatory
genital infections mitigate a severe genetic bottleneck in heterosexual transmission of subtype A and C
HIV-1. PLoS pathogens. 2009; 5(1):e1000274. Epub 2009/01/24. https://doi.org/10.1371/journal.ppat.
1000274 PMID: 19165325; PubMed Central PMCID: PMC2621345.
7. Auvert B, Buve A, Ferry B, Carael M, Morison L, Lagarde E, et al. Ecological and individual level analy-
sis of risk factors for HIV infection in four urban populations in sub-Saharan Africa with different levels of
HIV infection. AIDS. 2001; 15 Suppl 4:S15–30. Epub 2001/11/01. https://doi.org/10.1097/00002030-
200108004-00003 PMID: 11686462.
8. Piot P. AIDS: the impact of other sexually transmitted diseases. Netw Res Triangle Park N C. 1988; 9
(2):4. Epub 1988/01/01. PMID: 12280952.
9. Dhana A, Luchters S, Moore L, Lafort Y, Roy A, Scorgie F, et al. Systematic review of facility-based sex-
ual and reproductive health services for female sex workers in Africa. Global Health. 2014; 10:46. Epub
2014/06/12. https://doi.org/10.1186/1744-8603-10-46 PMID: 24916010; PubMed Central PMCID:
PMC4070634.
10. Vandenhoudt HM, Langat L, Menten J, Odongo F, Oswago S, Luttah G, et al. Prevalence of HIV and
other sexually transmitted infections among female sex workers in Kisumu, Western Kenya, 1997 and
2008. PLoS One. 2013; 8(1):e54953. Epub 2013/02/02. https://doi.org/10.1371/journal.pone.0054953
PMID: 23372801; PubMed Central PMCID: PMC3553007.
11. Vickerman P, Ndowa F, O’Farrell N, Steen R, Alary M, Delany-Moretlwe S. Using mathematical model-
ling to estimate the impact of periodic presumptive treatment on the transmission of sexually transmitted
infections and HIV among female sex workers. Sex Transm Infect. 2010; 86(3):163–8. Epub 2009/10/
27. https://doi.org/10.1136/sti.2008.034678 PMID: 19854700.
12. Seck K, Samb N, Tempesta S, Mulanga-Kabeya C, Henzel D, Sow PS, et al. Prevalence and risk fac-
tors of cervicovaginal HIV shedding among HIV-1 and HIV-2 infected women in Dakar, Senegal. Sex
Transm Infect. 2001; 77(3):190–3. Epub 2001/06/13. https://doi.org/10.1136/sti.77.3.190 PMID:
11402227; PubMed Central PMCID: PMC1744303.
13. Celum C, Wald A, Lingappa JR, Magaret AS, Wang RS, Mugo N, et al. Acyclovir and transmission of
HIV-1 from persons infected with HIV-1 and HSV-2. The New England journal of medicine. 2010; 362
(5):427–39. Epub 2010/01/22. https://doi.org/10.1056/NEJMoa0904849 PMID: 20089951; PubMed
Central PMCID: PMC2838503.
14. Daniels B, Wand H, Ramjee G, Team MDP. Prevalence of Herpes Simplex Virus 2 (HSV-2) infection
and associated risk factors in a cohort of HIV negative women in Durban, South Africa. BMC Res
Notes. 2016; 9(1):510. Epub 2016/12/14. https://doi.org/10.1186/s13104-016-2319-5 PMID: 27955706;
PubMed Central PMCID: PMC5154041.
15. Wall KM, Kilembe W, Vwalika B, Haddad LB, Hunter E, Lakhi S, et al. Risk of heterosexual HIV trans-
mission attributable to sexually transmitted infections and non-specific genital inflammation in Zambian
discordant couples, 1994–2012. Int J Epidemiol. 2017; 46(5):1593–606. Epub 2017/04/13. https://doi.
org/10.1093/ije/dyx045 PMID: 28402442; PubMed Central PMCID: PMC5837621.
16.
Joseph Davey DL, Wall KM, Kilembe W, Naw HK, Brill I, Vwalika B, et al. HIV Incidence and Predictors
of HIV Acquisition From an Outside Partner in Serodiscordant Couples in Lusaka, Zambia. J Acquir
Immune Defic Syndr. 2017; 76(2):123–31. Epub 2017/07/25. https://doi.org/10.1097/QAI.
0000000000001494 PMID: 28737591; PubMed Central PMCID: PMC5597474.
17. Boeras DI, Hraber PT, Hurlston M, Evans-Strickfaden T, Bhattacharya T, Giorgi EE, et al. Role of donor
genital tract HIV-1 diversity in the transmission bottleneck. Proceedings of the National Academy of Sci-
ences of the United States of America. 2011; 108(46):E1156–63. Epub 2011/11/09. https://doi.org/10.
1073/pnas.1103764108 PMID: 22065783; PubMed Central PMCID: PMC3219102.
18.
Looker KJ, Welton NJ, Sabin KM, Dalal S, Vickerman P, Turner KME, et al. Global and regional esti-
mates of the contribution of herpes simplex virus type 2 infection to HIV incidence: a population attribut-
able fraction analysis using published epidemiological data. Lancet Infect Dis. 2020; 20(2):240–9. Epub
2019/11/23. https://doi.org/10.1016/S1473-3099(19)30470-0 PMID: 31753763; PubMed Central
PMCID: PMC6990396.
PLOS ONE | https://doi.org/10.1371/journal.pone.0250044 April 20, 2021
17 / 21
PLOS ONE Etiologies of genital abnormalities in Rwandan men and women
19. Rietmeijer CA, Mungati M, Kilmarx PH, Barr BT, Gonese E, Kularatne RS, et al. Serological Markers for
Syphilis Among Persons Presenting With Syndromes Associated With Sexually Transmitted Infections:
Results From the Zimbabwe STI Etiology Study. Sex Transm Dis. 2019; 46(9):579–83. Epub 2019/04/
23. https://doi.org/10.1097/OLQ.0000000000001006 PMID: 31008842; PubMed Central PMCID:
PMC6885999.
20. Hayes RJ, Schulz KF, Plummer FA. The cofactor effect of genital ulcers on the per-exposure risk of HIV
transmission in sub-Saharan Africa. J Trop Med Hyg. 1995; 98(1):1–8. Epub 1995/02/01. PMID:
7861474.
21. Phiri S, Zadrozny S, Weiss HA, Martinson F, Nyirenda N, Chen CY, et al. Etiology of genital ulcer dis-
ease and association with HIV infection in Malawi. Sex Transm Dis. 2013; 40(12):923–8. Epub 2013/
11/14. https://doi.org/10.1097/OLQ.0000000000000051 PMID: 24220352.
22.
Takuva S, Mugurungi O, Mutsvangwa J, Machiha A, Mupambo AC, Maseko V, et al. Etiology and anti-
microbial susceptibility of pathogens responsible for urethral discharge among men in Harare, Zimba-
bwe. Sex Transm Dis. 2014; 41(12):713–7. Epub 2015/01/13. https://doi.org/10.1097/OLQ.
0000000000000204 PMID: 25581806.
23. Sylverken AA, Owusu-Dabo E, Yar DD, Salifu SP, Awua-Boateng NY, Amuasi JH, et al. Bacterial etiol-
ogy of sexually transmitted infections at a STI clinic in Ghana; use of multiplex real time PCR. Ghana
Med J. 2016; 50(3):142–8. Epub 2016/10/19. PMID: 27752188; PubMed Central PMCID:
PMC5044789.
24. Chirenje ZM, Dhibi N, Handsfield HH, Gonese E, Tippett Barr B, Gwanzura L, et al. The Etiology of Vag-
inal Discharge Syndrome in Zimbabwe: Results from the Zimbabwe STI Etiology Study. Sex Transm
Dis. 2018; 45(6):422–8. Epub 2018/02/22. https://doi.org/10.1097/OLQ.0000000000000771 PMID:
29465674; PubMed Central PMCID: PMC6879447.
25. Haddad LB, Wall KM, Kilembe W, Vwalika B, Khu NH, Brill I, et al. Bacterial vaginosis modifies the asso-
ciation between hormonal contraception and HIV acquisition. AIDS. 2018; 32(5):595–604. Epub 2018/
01/16. https://doi.org/10.1097/QAD.0000000000001741 PMID: 29334545; PubMed Central PMCID:
PMC5832628.
26. Barnabas SL, Dabee S, Passmore JS, Jaspan HB, Lewis DA, Jaumdally SZ, et al. Converging epidem-
ics of sexually transmitted infections and bacterial vaginosis in southern African female adolescents at
risk of HIV. Int J STD AIDS. 2018; 29(6):531–9. Epub 2017/12/05. https://doi.org/10.1177/
0956462417740487 PMID: 29198180.
27. Masha SC, Cools P, Descheemaeker P, Reynders M, Sanders EJ, Vaneechoutte M. Urogenital patho-
gens, associated with Trichomonas vaginalis, among pregnant women in Kilifi, Kenya: a nested case-
control study. BMC Infect Dis. 2018; 18(1):549. Epub 2018/11/08. https://doi.org/10.1186/s12879-018-
3455-4 PMID: 30400890; PubMed Central PMCID: PMC6219184.
28. Kerubo E, Laserson KF, Otecko N, Odhiambo C, Mason L, Nyothach E, et al. Prevalence of reproduc-
tive tract infections and the predictive value of girls’ symptom-based reporting: findings from a cross-
sectional survey in rural western Kenya. Sex Transm Infect. 2016; 92(4):251–6. Epub 2016/01/29.
https://doi.org/10.1136/sextrans-2015-052371 PMID: 26819339; PubMed Central PMCID:
PMC4893088.
29.
Lewis DA, Marsh K, Radebe F, Maseko V, Hughes G. Trends and associations of Trichomonas vagina-
lis infection in men and women with genital discharge syndromes in Johannesburg, South Africa. Sex
Transm Infect. 2013; 89(6):523–7. Epub 2013/04/23. https://doi.org/10.1136/sextrans-2013-051049
PMID: 23605850.
30. Morikawa E, Mudau M, Olivier D, de Vos L, Joseph Davey D, Price C, et al. Acceptability and Feasibility
of Integrating Point-of-Care Diagnostic Testing of Sexually Transmitted Infections into a South African
Antenatal Care Program for HIV-Infected Pregnant Women. Infect Dis Obstet Gynecol. 2018;
2018:3946862. Epub 2018/06/05. https://doi.org/10.1155/2018/3946862 PMID: 29861622; PubMed
Central PMCID: PMC5971359.
31. Garrett NJ, Osman F, Maharaj B, Naicker N, Gibbs A, Norman E, et al. Beyond syndromic manage-
ment: Opportunities for diagnosis-based treatment of sexually transmitted infections in low- and middle-
income countries. PLoS One. 2018; 13(4):e0196209. https://doi.org/10.1371/journal.pone.0196209
PMID: 29689080; PubMed Central PMCID: PMC5918163.
32.
Francis SC, Ao TT, Vanobberghen FM, Chilongani J, Hashim R, Andreasen A, et al. Epidemiology of
curable sexually transmitted infections among women at increased risk for HIV in northwestern Tanza-
nia: inadequacy of syndromic management. PLoS One. 2014; 9(7):e101221. Epub 2014/07/16. https://
doi.org/10.1371/journal.pone.0101221 PMID: 25025338; PubMed Central PMCID: PMC4099080.
33. Guimaraes H, Castro R, Tavora Tavira L, da LEF. Assessing therapeutic management of vaginal and
urethral symptoms in an anonymous HIV testing centre in Luanda, Angola. J Infect Dev Ctries. 2013; 7
(10):720–5. Epub 2013/10/17. https://doi.org/10.3855/jidc.2752 PMID: 24129624.
PLOS ONE | https://doi.org/10.1371/journal.pone.0250044 April 20, 2021
18 / 21
PLOS ONE Etiologies of genital abnormalities in Rwandan men and women
34. Marx G, John-Stewart G, Bosire R, Wamalwa D, Otieno P, Farquhar C. Diagnosis of sexually transmit-
ted infections and bacterial vaginosis among HIV-1-infected pregnant women in Nairobi. Int J STD
AIDS. 2010; 21(8):549–52. Epub 2010/10/27. https://doi.org/10.1258/ijsa.2010.010005 PMID:
20975086; PubMed Central PMCID: PMC3050991.
35. Black V, Magooa P, Radebe F, Myers M, Pillay C, Lewis DA. The detection of urethritis pathogens
among patients with the male urethritis syndrome, genital ulcer syndrome and HIV voluntary counselling
and testing clients: should South Africa’s syndromic management approach be revised? Sex Transm
Infect. 2008; 84(4):254–8. Epub 2008/01/15. https://doi.org/10.1136/sti.2007.028464 PMID: 18192290.
36.
37.
38.
Frohlich JA, Abdool Karim Q, Mashego MM, Sturm AW, Abdool Karim SS. Opportunities for treating
sexually transmitted infections and reducing HIV risk in rural South Africa. J Adv Nurs. 2007; 60(4):377–
83. Epub 2007/09/08. https://doi.org/10.1111/j.1365-2648.2007.04405.x PMID: 17822425.
Tann CJ, Mpairwe H, Morison L, Nassimu K, Hughes P, Omara M, et al. Lack of effectiveness of syn-
dromic management in targeting vaginal infections in pregnancy in Entebbe, Uganda. Sex Transm
Infect. 2006; 82(4):285–9. Epub 2006/08/01. https://doi.org/10.1136/sti.2005.014845 PMID: 16877576;
PubMed Central PMCID: PMC2564710.
Ingabire R, Parker R, Nyombayire J, Ko JE, Mukamuyango J, Bizimana J, et al. Female sex workers in
Kigali, Rwanda: a key population at risk of HIV, sexually transmitted infections, and unplanned preg-
nancy. Int J STD AIDS. 2019; 30(6):557–68. Epub 2019/02/08. https://doi.org/10.1177/
0956462418817050 PMID: 30727831; PubMed Central PMCID: PMC6512058.
39. RWANDA POPULATION-BASED HIV IMPACT ASSESSMENT RPHIA 2018–2019 2019 [cited 2020
Nov 20]. Available from: https://phia.icap.columbia.edu/wp-content/uploads/2019/10/RPHIA-Summary-
Sheet_Oct-2019.pdf.
40. Conkling M, Shutes EL, Karita E, Chomba E, Tichacek A, Sinkala M, et al. Couples’ voluntary counsel-
ling and testing and nevirapine use in antenatal clinics in two African capitals: a prospective cohort
study. Journal of the International AIDS Society. 2010; 13:10. Epub 2010/03/17. https://doi.org/10.
1186/1758-2652-13-10 PMID: 20230628; PubMed Central PMCID: PMC2851580.
41. Karita E, Nsanzimana S, Ndagije F, Wall KM, Mukamuyango J, Mugwaneza P, et al. Implementation
and Operational Research: Evolution of Couples’ Voluntary Counseling and Testing for HIV in Rwanda:
From Research to Public Health Practice. J Acquir Immune Defic Syndr. 2016; 73(3):e51–e8. Epub
2016/10/16. https://doi.org/10.1097/QAI.0000000000001138 PMID: 27741033; PubMed Central
PMCID: PMC5367509.
42. Mazzei A, Ingabire R, Mukamuyango J, Nyombayire J, Sinabamenye R, Bayingana R, et al. Community
health worker promotions increase uptake of long-acting reversible contraception in Rwanda. Reprod
Health. 2019; 16(1):75. Epub 2019/06/06. https://doi.org/10.1186/s12978-019-0739-0 PMID:
31164155; PubMed Central PMCID: PMC6549304.
43.
Ingabire R, Nyombayire J, Hoagland A, Da Costa V, Mazzei A, Haddad L, et al. Evaluation of a multi-
level intervention to improve postpartum intrauterine device services in Rwanda. Gates Open Res.
2018; 2(38):38. Epub 2019/09/25. https://doi.org/10.12688/gatesopenres.12854.3 PMID: 30569036;
PubMed Central PMCID: PMC6266741.3.
44. Rietmeijer CA, Mungati M, Machiha A, Mugurungi O, Kupara V, Rodgers L, et al. The Etiology of Male
Urethral Discharge in Zimbabwe: Results from the Zimbabwe STI Etiology Study. Sex Transm Dis.
2018; 45(1):56–60. Epub 2017/12/15. https://doi.org/10.1097/OLQ.0000000000000696 PMID:
29240635.
45. Gonorrhea—CDC Fact Sheet (Detailed Version): U.S. Centers for Disease Control; 2020 [cited 2020 6/
29/2020]. Available from: https://www.cdc.gov/std/gonorrhea/stdfact-gonorrhea-detailed.htm.
46. Dela H, Attram N, Behene E, Kumordjie S, Addo KK, Nyarko EO, et al. Risk factors associated with gon-
orrhea and chlamydia transmission in selected health facilities in Ghana. BMC Infect Dis. 2019; 19
(1):425. Epub 2019/05/18. https://doi.org/10.1186/s12879-019-4035-y PMID: 31096920; PubMed Cen-
tral PMCID: PMC6524331.
47. Connolly S, Wall KM, Parker R, Kilembe W, Inambao M, Visoiu AM, et al. Sociodemographic factors
and STIs associated with Chlamydia trachomatis and Neisseria gonorrhoeae infections in Zambian
female sex workers and single mothers. Int J STD AIDS. 2020; 31(4):364–74. Epub 2020/03/05. https://
doi.org/10.1177/0956462419894453 PMID: 32126947.
48. Mungati M, Machiha A, Mugurungi O, Tshimanga M, Kilmarx PH, Nyakura J, et al. The Etiology of Geni-
tal Ulcer Disease and Coinfections With Chlamydia trachomatis and Neisseria gonorrhoeae in Zimba-
bwe: Results From the Zimbabwe STI Etiology Study. Sex Transm Dis. 2018; 45(1):61–8. Epub 2017/
12/15. https://doi.org/10.1097/OLQ.0000000000000694 PMID: 29240636; PubMed Central PMCID:
PMC5994235.
49. MacNeily AE. Editorial comment. J Urol. 2011; 185(6):2306–7. Epub 2011/04/23. https://doi.org/10.
1016/j.juro.2011.02.2702 PMID: 21511302.
PLOS ONE | https://doi.org/10.1371/journal.pone.0250044 April 20, 2021
19 / 21
PLOS ONE Etiologies of genital abnormalities in Rwandan men and women
50. Weiss HA, Thomas SL, Munabi SK, Hayes RJ. Male circumcision and risk of syphilis, chancroid, and
genital herpes: a systematic review and meta-analysis. Sex Transm Infect. 2006; 82(2):101–9; discus-
sion 10. Epub 2006/04/04. https://doi.org/10.1136/sti.2005.017442 PMID: 16581731; PubMed Central
PMCID: PMC2653870.
51. Morris BJ, Hankins CA. Effect of male circumcision on risk of sexually transmitted infections and cervi-
cal cancer in women. Lancet Glob Health. 2017; 5(11):e1054–e5. Epub 2017/10/14. https://doi.org/10.
1016/S2214-109X(17)30386-8 PMID: 29025620.
52. Davis S, Toledo C, Lewis L, Maughan-Brown B, Ayalew K, Kharsany ABM. Does voluntary medical
male circumcision protect against sexually transmitted infections among men and women in real-world
scale-up settings? Findings of a household survey in KwaZulu-Natal, South Africa. BMJ Glob Health.
2019; 4(3):e001389. Epub 2019/07/03. https://doi.org/10.1136/bmjgh-2019-001389 PMID: 31263584;
PubMed Central PMCID: PMC6570991.
53. Mehta SD, Moses S, Parker CB, Agot K, Maclean I, Bailey RC. Circumcision status and incident herpes
simplex virus type 2 infection, genital ulcer disease, and HIV infection. AIDS. 2012; 26(9):1141–9. Epub
2012/03/03. https://doi.org/10.1097/QAD.0b013e328352d116 PMID: 22382150; PubMed Central
PMCID: PMC3668787.
54. Kularatne RS, Muller EE, Maseko DV, Kufa-Chakezha T, Lewis DA. Trends in the relative prevalence of
genital ulcer disease pathogens and association with HIV infection in Johannesburg, South Africa,
2007–2015. PLoS One. 2018; 13(4):e0194125. Epub 2018/04/05. https://doi.org/10.1371/journal.pone.
0194125 PMID: 29617372; PubMed Central PMCID: PMC5884493.
55. Bogaerts J, Kestens L, van Dyck E, Tello WM, Akingeneye J, Mukantabana V. Genital ulcers in a pri-
mary health clinic in Rwanda: impact of HIV infection on diagnosis and ulcer healing (1986–1992). Int J
STD AIDS. 1998; 9(11):706–10. Epub 1998/12/24. https://doi.org/10.1258/0956462981921242 PMID:
9863586.
56. Bogaerts J, Vuylsteke B, Martinez Tello W, Mukantabana V, Akingeneye J, Laga M, et al. Simple algo-
rithms for the management of genital ulcers: evaluation in a primary health care centre in Kigali,
Rwanda. Bull World Health Organ. 1995; 73(6):761–7. Epub 1995/01/01. PMID: 8907769; PubMed
Central PMCID: PMC2486690.
57. Van Dyck E, Bogaerts J, Smet H, Tello WM, Mukantabana V, Piot P. Emergence of Haemophilus
ducreyi resistance to trimethoprim-sulfamethoxazole in Rwanda. Antimicrob Agents Chemother. 1994;
38(7):1647–8. Epub 1994/07/01. https://doi.org/10.1128/aac.38.7.1647 PMID: 7979300; PubMed Cen-
tral PMCID: PMC284607.
58. Roggen EL, Hoofd G, Van Dyck E, Piot P. Enzyme immunoassays (EIAs) for the detection of anti-Hae-
mophilus ducreyi serum IgA, IgG, and IgM antibodies. Sex Transm Dis. 1994; 21(1):36–42. Epub 1994/
01/01. https://doi.org/10.1097/00007435-199401000-00008 PMID: 8140487.
59. Bogaerts J, Ricart CA, Van Dyck E, Piot P. The etiology of genital ulceration in Rwanda. Sex Transm
Dis. 1989; 16(3):123–6. Epub 1989/07/01. https://doi.org/10.1097/00007435-198907000-00001 PMID:
2510325.
60. Behets FM, Brathwaite AR, Hylton-Kong T, Chen CY, Hoffman I, Weiss JB, et al. Genital ulcers: setiol-
ogy, clinical diagnosis, and associated human immunodeficiency virus infection in Kingston, Jamaica.
Clin Infect Dis. 1999; 28(5):1086–90. Epub 1999/08/19. https://doi.org/10.1086/514751 PMID:
10452639.
61. Connolly S, Kilembe W, Inambao M, Visoiu AM, Sharkey T, Parker R, et al. A population-specific opti-
mized GeneXpert pooling algorithm for Chlamydia trachomatis and Neisseria gonorrhoeae to reduce
cost of molecular STI screening in resource-limited settings. J Clin Microbiol. 2020. Epub 2020/06/12.
https://doi.org/10.1128/JCM.00176-20 PMID: 32522828.
62.
Johnson LF, Dorrington RE, Bradshaw D, Coetzee DJ. The effect of syndromic management interven-
tions on the prevalence of sexually transmitted infections in South Africa. Sex Reprod Healthc. 2011; 2
(1):13–20. Epub 2010/12/15. https://doi.org/10.1016/j.srhc.2010.08.006 PMID: 21147454.
63. Garrett N, Mitchev N, Osman F, Naidoo J, Dorward J, Singh R, et al. Diagnostic accuracy of the Xpert
CT/NG and OSOM Trichomonas Rapid assays for point-of-care STI testing among young women in
South Africa: a cross-sectional study. BMJ Open. 2019; 9(2):e026888. Epub 2019/02/21. https://doi.
org/10.1136/bmjopen-2018-026888 PMID: 30782948; PubMed Central PMCID: PMC6367982.
64.
Lingappa JR, Kahle E, Mugo N, Mujugira A, Magaret A, Baeten J, et al. Characteristics of HIV-1 discor-
dant couples enrolled in a trial of HSV-2 suppression to reduce HIV-1 transmission: the partners study.
PLoS One. 2009; 4(4):e5272. Epub 2009/05/01. https://doi.org/10.1371/journal.pone.0005272 PMID:
19404392; PubMed Central PMCID: PMC2671170.
65. Cousins S. How Rwanda could be the first country to wipe out cervical cancer: Mosaicscience.com;
2019 [updated 7May2019]. Available from: https://mosaicscience.com/story/rwanda-cervical-cancer-
hpv-vaccine-gardasil-cervarix/.
PLOS ONE | https://doi.org/10.1371/journal.pone.0250044 April 20, 2021
20 / 21
PLOS ONE Etiologies of genital abnormalities in Rwandan men and women
66. Wall KM, Kilembe W, Vwalika B, Dinh C, Livingston P, Lee YM, et al. Schistosomiasis is associated with
incident HIV transmission and death in Zambia. PLoS Negl Trop Dis. 2018; 12(12):e0006902. Epub
2018/12/14. https://doi.org/10.1371/journal.pntd.0006902 PMID: 30543654; PubMed Central PMCID:
PMC6292564.
67. Secor WE. The effects of schistosomiasis on HIV/AIDS infection, progression and transmission. Cur-
rent opinion in HIV and AIDS. 2012; 7(3):254–9. Epub 2012/02/14. https://doi.org/10.1097/COH.
0b013e328351b9e3 PMID: 22327410.
68. Rujeni N, Morona D, Ruberanziza E, Mazigo HD. Schistosomiasis and soil-transmitted helminthiasis in
Rwanda: an update on their epidemiology and control. Infect Dis Poverty. 2017; 6(1):8. Epub 2017/03/
02. https://doi.org/10.1186/s40249-016-0212-z PMID: 28245883; PubMed Central PMCID:
PMC5331630.
PLOS ONE | https://doi.org/10.1371/journal.pone.0250044 April 20, 2021
21 / 21
PLOS ONE |
10.1371_journal.pone.0244447 | RESEARCH ARTICLE
Behavioral response of insecticide-resistant
mosquitoes against spatial repellent: A
modified self-propelled particle model
simulation
Guofa ZhouID, Leonard Yu, Xiaoming WangID, Daibin Zhong, Ming-chieh Lee,
Solomon KibretID, Guiyun Yan*
Program in Public Health, University of California, Irvine, California, United States of America
* guiyuny@uci.edu
Abstract
Rapidly increasing pyrethroid insecticide resistance and changes in vector biting and resting
behavior pose serious challenges in malaria control. Mosquito repellents, especially spatial
repellents, have received much attention from industry. We attempted to simulate interac-
tions between mosquitoes and repellents using a machine learning method, the Self-Pro-
pelled Particle (SPP) model, which we modified to include attractiveness/repellency effects.
We simulated a random walk scenario and scenarios with insecticide susceptible/resistant
mosquitoes against repellent alone and against repellent plus attractant (to mimic a human
host). Simulation results indicated that without attractant/repellent, mosquitoes would fly
anywhere in the cage at random. With attractant, all mosquitoes were attracted to the source
of the odor by the end. With repellent, all insecticide-susceptible mosquitoes eventually
moved to the corner of the cage farthest from the repellent release point, whereas, a high
proportion of highly resistant mosquitoes might reach the attractant release point (the
human) earlier in the simulation. At fixed concentration, a high proportion of mosquitoes
could be able to reach the host when the relative repellency efficacy (compare to attractant
efficacy) was <1, whereas, no mosquitoes reached the host when the relative repellency
efficacy was > 1. This result implies that repellent may not be sufficient against highly physi-
ologically insecticide resistant mosquitoes, since very high concentrations of repellent are
neither practically feasible nor cost-effective.
Introduction
Malaria remains to be the world’s most common mosquito-borne disease, with an estimated
228 million cases worldwide in 2018 [1]. Control efforts mainly rely on vector control using a
single class of insecticides, the pyrethroids, which is the only class approved for use on Long-
Lasting Insecticidal Nets (LLINs) [2]. Pyrethroids, along with other pesticides, are also widely
used to control agricultural pests on livestock and field crops worldwide [3]. The past decade
has seen a dramatic increase in reports of physiological pyrethroid resistance in malaria vectors
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: Zhou G, Yu L, Wang X, Zhong D, Lee M-c,
Kibret S, et al. (2020) Behavioral response of
insecticide-resistant mosquitoes against spatial
repellent: A modified self-propelled particle model
simulation. PLoS ONE 15(12): e0244447. https://
doi.org/10.1371/journal.pone.0244447
Editor: Jiang-Shiou Hwang, National Taiwan Ocean
University, TAIWAN
Received: September 28, 2020
Accepted: December 9, 2020
Published: December 29, 2020
Copyright: © 2020 Zhou et al. This is an open
access article distributed under the terms of the
Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: All relevant data are
within the manuscript and its Supporting
Information files.
Funding: GZ, XW, DZ, MCL, SK, and GY were
funded by the National Institutes of Health (R01
A1050243, D43 TW01505, and U19 AI129326).
The funders had no role in study design, data
collection and analysis, decision to publish, or
preparation of the manuscript.
Competing interests: The authors have declared
that no competing interests exist.
PLOS ONE | https://doi.org/10.1371/journal.pone.0244447 December 29, 2020
1 / 18
PLOS ONE Behavioral response of insecticide-resistant mosquitoes against spatial repellent
[4–15]. The rapid increase in pyrethroid resistance necessitates an immediate proactive resis-
tance management response to avoid compromising existing effective interventions. In addi-
tion to the increase and spread of insecticide resistance, the biting and resting behaviors of
malaria vectors have evolved. A number of recent studies have documented changes in the bit-
ing behavior of the major African malaria vectors, Anopheles gambiae and Anopheles funestus,
from midnight biting to biting in the early evening and morning hours, and to biting outdoors,
where people are not protected by Indoor Residual Spraying (IRS) or LLINs [16–18]. This
early outdoor biting behavior is likely due to insecticide-induced behavioral changes in malaria
vectors, i.e., avoiding contact with insecticide-treated bed nets and walls [19,20]. Residual
malaria transmission has become a very important challenge in malaria control [21].
Various alternative vector control tools against outdoor transmission have been the subject
of research studies in recent years [22–25]. Among these tools, repellents have received the
most attention from industry. The global mosquito repellent market was valued at approxi-
mately $3.2 billion in 2016 and is expected to reach ~$5 billion in 2022, growth driven chiefly
by recent outbreaks of mosquito-borne diseases [26]. Repellents come in two types: topical
and spatial. Topical repellents such as DEET require constant application and reapplication on
skin, clothing, or other surfaces to discourage mosquitoes (and arthropods in general) from
landing or climbing on the surface [27–34]. Spatial repellents release into the air volatile active
ingredients that interfere with mosquitoes’ ability to find a host, thus preventing mosquitoes
from contacting the host and taking a blood meal, thus preventing disease transmission [35–
40]. Spatial repellents confer protection against mosquito bites through the action of emanated
vapor or airborne chemical particles in a large space, such as a room or a yard surrounding a
household [35–42]. Due to their relatively low cost and ease of deployment, spatial repellents
are popular in developing regions [43]. Spatial repellents may significantly aid in preventing
mosquito-borne diseases if properly incorporated into integrated vector management
approaches [40]. The best known spatial repellent is the burning coil, which releases the chem-
ical into a space such as a room, preventing mosquitoes from entering the entire space [43].
However, spatial repellents, such as some coil products, use pyrethroids and other synthetic
chemicals and plant products as major volatile ingredients to repel mosquitoes [23,25,44–46].
These volatile chemicals prevent mosquitoes from feeding on humans through several mecha-
nisms, including 1) knockdown and mortality; 2) deterrence (mosquitoes are prevented from
entering human dwellings); 3) irritancy (mosquitoes enter houses but leave early); 4) excito-
repellency (mosquitoes exit the house to avoid contacting airborne insecticides); and 5) feed-
ing inhibition (mosquitoes are prevented from biting and getting blood meals)
[23,25,35,37,40,44–50]. Since malaria vectors have developed widespread resistance to insecti-
cides, the efficacy of spatial repellents against these resistant mosquitoes is unknown. Some
recent studies have found that insecticide-resistant mosquitoes behave differently than suscep-
tible mosquitoes when they encounter repellents. For example, Deletre et al. found that DEET
had a reduced repellency effect against resistant strains of An. gambiae but maintained irri-
tancy for the susceptible strain [51]. Similarly, Agossa et al. found that as a result of resistance,
pyrethroid-based malaria control tools have decreased toxicity and repellent effects [52].
Kawada et al. found that the frequency of takeoffs from a pyrethroid-treated surface and flying
times without contacting the surface increased significantly in pyrethroid-susceptible An. gam-
biae s.s., while the An. gambiae s.s. wild colony (i.e., the resistant strain) exhibited no such
behavior [53]. Studies of Anopheles and Aedes mosquitoes show similar results, i.e., physiologi-
cal insecticide resistance modifies vector contact avoidance behavior, and that, in general, a
higher concentration of pyrethroids is needed to deter blood-feeding by resistant vectors [54–
58]. While the evidence from these studies points to that repellents have a reduced repellency
effect against physiologically insecticide resistant mosquitoes, it is not conclusive. In addition,
PLOS ONE | https://doi.org/10.1371/journal.pone.0244447 December 29, 2020
2 / 18
PLOS ONE Behavioral response of insecticide-resistant mosquitoes against spatial repellent
data are limited on the response to spatial repellents by mosquito populations that have
evolved outdoor biting behavior.
Machine learning methods such as self-propelled particle (SPP) models have offered inno-
vative ways to study the collective behaviors of living organisms and their interactions with
their environments, including colonies of wasps, schools of fish, flocks of birds, and others
[59–63]. Many animals can be treated as SPPs; they find energy in their food and exhibit vari-
ous locomotion strategies, from flying to crawling. These biological systems can propel them-
selves based on the presence of chemoattractants [63]. A number of SPP models have been
proposed, ranging from the simplest Active Brownian Particle model to highly elaborate and
specialized models aimed at describing specific systems and situations [64,65]. For example,
an SPP model simulation found that as locust population density increased, locusts changed
their behavior from relatively disordered, independent movement of individuals within the
group to the group moving as a highly aligned whole [65]. This modeling result was supported
in the field: when locust density reached a critical value of 74 locusts/m2, the locusts ceased
making rapid and spontaneous changes of direction and instead marched steadily in the same
direction for the full 8 hours of the experiment [66]. Anopheles mosquitoes, which find their
hosts via human odors, can be treated as SPPs because they propel themselves based on che-
moattractants [67,68]. Mosquito host-seeking involves the movements of individual females as
well as the collective behaviors of groups that result from individuals’ local interactions with
each other and with their environment [67,68]. To model mosquito reactions to spatial repel-
lents, however, other parameters must be considered in the presence of repellents, because it is
not clear how insecticide-resistant mosquitoes respond to these repellents. These additional
parameters may be incorporated by modifying existing models, since SPP models allow for
particle assemblies to be temporally and spatially reversible in complex media, e.g., media that
includes barriers resembling a bed net or repellent situation faced by mosquitoes [69–76].
The aim of this study is to demonstrate mosquitoes’ host-seeking behavior in the presence
of spatial repellent with/without resistance using machine learning simulations, and to exam-
ine how physiological insecticide resistance impacts this behavior at the individual and popula-
tion level. The specific question asked is: Can physiologically insecticide-resistant mosquitoes
reach the host before being repelled by the spatial repellent? If the answer is yes, what propor-
tion can reach the host before they are repelled? Such knowledge is crucial in developing more
efficient methods which improve the field effectiveness of spatial repellents and other means of
mosquito control.
Materials and methods
Model development
In this study, we considered the effects of two elements: i) attractive odors to mimic human
odor and attract mosquitoes, and ii) spatial repellents to repel mosquitoes and protect humans
from mosquito biting. Mosquito flight behavior was simulated using the constrained SPP
model [70,73,74]. We chose this approach based on the prior success of SPP in modeling insect
social behaviors [69–75]. We modified the SPP model by adding attractant and repellent to
influence the mosquito flight path. We used both insecticide-susceptible and -resistant mos-
quitoes, since insecticide-resistant mosquitoes may behave differently than susceptible mos-
quitoes when they encounter spatial repellent. Mosquitoes were thus able to choose their
respective paths and move non-uniformly into the spatial repellent–affected space. This con-
strained, non-uniform movement helped insecticide-resistant mosquitoes to avoid the
repellent.
PLOS ONE | https://doi.org/10.1371/journal.pone.0244447 December 29, 2020
3 / 18
PLOS ONE Behavioral response of insecticide-resistant mosquitoes against spatial repellent
Our modified SPP model, based on the Vicsek model [77], is shown in Eq (1). In this
model, mosquito flight velocity, and location, is simulated as a function of Brownian motion
and modulated by attractants and repellents [78]. With neither attractant nor repellent, we
assume mosquitoes move randomly. With attractant, mosquitoes are drawn toward the attrac-
tant release point; conversely, mosquitoes are driven away from the repellent release point.
The flying velocity of a mosquito at a given time is modeled as:
uðt þ DtÞ ¼ uðtÞ þ σDtε þ γ1αðGðuðtÞ; AÞ (cid:0) θ1ÞDt þ γ2βðFðuðtÞ; RÞ (cid:0) θ2ÞDt
ð1Þ
where u(t) is a 3-dimensional vector and parameters are as shown in Table 1. To simplify the
simulation process, we assume mosquitoes are confined within insect-proof cages. Mosquitoes
cannot escape, they have no food source other than the attractant odor, and the repellent is
pure, with no bed net or IRS use. These conditions rule out any potential outdoor resting or
other avoidance behavior.
The mosquitoes’ response to attractant/repellent depends on the concentration of attrac-
tant/repellent and the mosquitoes’ response strength; i.e., insecticide-resistant and -sensitive
mosquitoes respond differently to repellent. The mosquito response to attractant/repellent is
modeled as an inverse-exponential-decay function:
Kðu; CÞ ¼ ruð1 (cid:0) e(cid:0) aCÞ
ð2Þ
where C is the attractant/repellent concentration and a is the response parameter. C is a func-
tion of space and time and is modeled as a convection-diffusion model with a point source
[79]:
@C
@t
¼ Dr2C (cid:0) r vCð
Þ þ mC0 tð Þ
ð3Þ
where C0 is the attractant/repellent release rate, v is the velocity field that the quantity is mov-
ing with (e.g., wind or temperature), D is the diffusion coefficient, μ is the scale parameter
which measures the attractant/repellent releasing strength, and r represents the concentration
gradient.
An attractant/repellent’s concentration is affected by its release rate. For simplicity, we
assume it is not contingent on temperature, wind speed (i.e., v = 0), or delivery mechanism. In
an individual household setting in rural Africa, a repellent such as burning coils usually can be
Table 1. Descriptions of parameters used in the models.
Parameter
Description
Initial value
Δt
σ
ε
γ1, γ2
α, β
θ1, θ2
ρ
a
F, G
A, R
C0
D
μ
Time and time interval for each step of simulation
Random walk rate in the form of Brownian motion
Gaussian random variable with mean of 0 and variance of 1
Mosquito response rates
Attractance/repellency efficiency
Attractant/repellent release locations
Response scale parameter
Response decay/increase rate parameter
Attractant/repellent response functions in the form of K with different a and ρ
Attractant/repellent concentrations in the form of C with different D, μ and C0
Attractant/repellent release rate
Diffusion coefficient
Scale parameter
0.03 s
1.0
1.0, 1.0
0.5, 0.5
1.0
5.0, 5.0
1.0
1.0
https://doi.org/10.1371/journal.pone.0244447.t001
PLOS ONE | https://doi.org/10.1371/journal.pone.0244447 December 29, 2020
4 / 18
PLOS ONE Behavioral response of insecticide-resistant mosquitoes against spatial repellent
seen as a point source with a constant rate of release, whereas a real person releasing human
odor is not really a point source. However, for simplicity of simulation, we assume both attrac-
tant and repellent are released from a point source with constant speed of release.
Mosquito movement rules
In our simulations, an individual mosquito was treated as point particle; body size was omit-
ted. Mosquitoes had no effect on each other during movement, either attractive or repulsive.
Mosquito motion was restricted to the ‘sealed’ house. Both the repellent volatile chemicals and
the attractant odors had threshold concentrations, below which the mosquitoes navigated ran-
domly; i.e., attractant/repellent did not affect mosquito movement when odor/volatile chemi-
cal concentration was low. For concentrations above the threshold, mosquitoes responded by
changing their moving behavior, moving away from higher levels of repellent or toward the
odor release point. High concentration levels biased mosquitoes’ flight direction, resulting in
flights that were on average moving away from the repellent or toward the attractant. However,
we assumed that repellent/attractant concentration did not affect mosquito flight speed; i.e.,
average speed and deviation did not change.
Step length for mosquito movement was drawn separately for X/Y/Z directions from
Brownian motion, with a mean of 1 cm and variance of 1 cm2. Based on video tracking of fly-
ing speed of Anopheles arabiensis and Anopheles gambiae sensu stricto against untreated nets
and LLINs [68,80], we assumed mosquitoes move 1000 steps per 30 s, which was equivalent to
average of 10 m per 30 s, same as observed in the video tracking experiments. Direction of
movement was determined by (X-1.0, Y-1.0, Z-1.0). However, directions were adjusted either
toward the attractant release point or against the repellent release point, depending on the
resistance level and media concentration.
Host odor and repellent
Host odor and repellent were released independently at different emission and diffusion rates,
both following the random diffusion principle as described in Eq (3). For simplicity we
assumed no other media affected the diffusion of attractant and repellent. However, we did
vary the attractiveness and repellency rates. There were no other constraints, so both attractant
and repellent could potentially fill the entire experimental space, but we did set a maximum
concentration level within the space. (In the real world, houses are not sealed and odor and
repellent chemicals can diffuse out of the house, which means indoor concentration remains
relatively stable rather than increasing indefinitely.) As stated above, we assigned a threshold
concentration for both repellent and attractant. Below the threshold, mosquitoes navigated
randomly. Above the threshold, mosquitoes moved away from the repellent move toward the
host odor.
Mosquito resistance
Because resistant mosquitoes tolerate higher insecticide concentrations, we assumed they also
tolerate a higher concentration of the volatile chemicals released into the air by the repellent.
In other words, host-searching movements of insecticide-resistant mosquitoes may not be
inhibited by repellent if the concentration of volatile chemicals is low. We made this assump-
tion based on observations that spatial repellent had a delayed impact or decreased toxicity
and repellent effects against field-caught insecticide-resistant mosquitoes [52,55–58,81]. For
this study, we assumed resistant mosquitoes tolerated a 2-fold higher concentration of repel-
lent compared to non-resistant/susceptible mosquitoes.
PLOS ONE | https://doi.org/10.1371/journal.pone.0244447 December 29, 2020
5 / 18
PLOS ONE Behavioral response of insecticide-resistant mosquitoes against spatial repellent
Additional rules
We assumed mosquitoes did not rest on the cage walls, as if the mosquitoes were starving
when released into a room with a sleeping host and would not stop flying until they reached
the host. Mosquitoes could not escape from the cages. For simplicity of simulation we also
assumed no knockdown or mortality effects; this is probably not accurate, since many spatial
repellents consist of pyrethroid insecticides (although it is accurate if the concentration of vol-
atile chemicals is low).
Outcome measurements
We measured the following parameters: average landing time, landing rate, average repel-
ling time, repelling rate, distribution of landing rate, and distribution of repelling rate.
Landing on a human or host was defined as a mosquito moving to <1 cm (or 1 step) from
the attractant odor release point. A mosquito was repelled if it was pushed out of the
enclosed space, i.e., moved to <1 cm from the corner of the cage farthest from the repellent
release point.
In real-world hut experiments, three indicators are used to measure repellent efficacy:
deterrence, excito-repellency, and toxicity [82,83]. Deterrence is determined by comparing the
total number of mosquitoes in cages with spatial repellent to the number in control cages with-
out spatial repellent. For the simulation study, we removed a mosquito once it reached the
attractant (found the blood source) or was expelled to the far end of the cage opposite the
repellent release location (exited the house). Excito-repellence is measured as the proportion
of mosquitoes found at the far end of the cage opposite the repellent release location, in the
spatial repellent treatment relative to the control. Toxicity is determined by comparing the
mortality rate in spatial repellent treatment houses to that of the control houses. In this study,
we assumed repellent only repels but does not kill mosquitoes.
Simulation of mosquito response to repellent with/without attractant
Simulations were conducted under four scenarios: a) random walk, i.e., no attractant and
no repellent; b) random walk modified by adding attractant to mimic human host; c) ran-
dom walk modified by adding repellent to simulate prevention efficacy; and d) random
walk modified by adding both attractant and repellent. We also simulated different mos-
quito resistance levels: sensitive (no resistance at all) and resistant (insensitive to a 2-fold
higher level of repellent) [51,52]. Resistance levels were simulated by reducing the repel-
lency rate [51–53].
Our simulations assumed mosquitoes were released in the middle of an enclosed space with
a size of 5m x 5m x 3m, which similar to the size of typical African house [68]. Both attractant
and repellent were released on the ground in one corner of the space, and both were treated as
point objects. In each simulation, 10 mosquitoes were released for up to a maximum of 30
minutes. The simulation was repeated 100 times for each scenario with both susceptible and
resistant mosquitoes.
Simulations were realized using R 3.5.2.
Results and discussion
Flight path tracking
Fig 1 illustrates the potential mosquito flight paths (n = 10 mosquitoes) simulated by the
model under different settings: random walk without attractant or repellent, with attractant,
and with repellent (see S1 Fig for single mosquito flying path). By 30 seconds, 3 of the 10
PLOS ONE | https://doi.org/10.1371/journal.pone.0244447 December 29, 2020
6 / 18
PLOS ONE Behavioral response of insecticide-resistant mosquitoes against spatial repellent
Fig 1. Illustrations of simulated mosquito flying paths. A) 3D display of random walk at different time points; B) 2D projected flying paths at different time points for
random walk (left panel), with attractant (middle panel), and with repellent and no resistance (right panel).
https://doi.org/10.1371/journal.pone.0244447.g001
mosquitoes have been attracted to the “human,” i.e., the attractant release point (Fig 1B, mid-
dle). Fig 2 shows the 2D projected attractant/repellent concentration and the locations of all
mosquitoes at different points in time. Here, simulations indicated that all non-resistant mos-
quitoes were repelled to the corner farthest from the repellent release point within 1 minute
(Fig 2C), whereas most of the resistant mosquitoes were able to stay in areas with low repellent
concentration (Fig 2D). The presence of repellent did make mosquitoes more difficult to find
the host even if when repellent concentration was low (S1B & S1D Fig).
Attractant effect
Fig 3 shows how attractant concentration affected mosquito host-searching behavior. In low-
concentration settings (C0 = 5 unit), the first mosquito had reached the host by 38 s; about
50% of the mosquitoes had reached the host by about 45 s; and all mosquitoes had reached the
host by 117 s (Fig 3). The average time to reach the host was 83.1 (SD 1.5) s.
If the attractant release rate was doubled (C0 = 10 unit), the first mosquito reached the host
by 28 s and all mosquitoes reached the host by 96 s (Fig 3). The average time to reach the host
was 65.7 ± 1.5 s.
PLOS ONE | https://doi.org/10.1371/journal.pone.0244447 December 29, 2020
7 / 18
PLOS ONE Behavioral response of insecticide-resistant mosquitoes against spatial repellent
Fig 2. Current 2D projected locations of mosquitoes at different time points in different settings. A) random walk; B) with attractant; C) with repellent and no
resistance; and D) with repellent and resistant mosquitoes. Colors in the heat maps show different concentrations of attractant odors or repellent volatile particles.
https://doi.org/10.1371/journal.pone.0244447.g002
Repellent effect
Fig 4 shows the distribution of times at which mosquitoes were repelled; no attractant was
released in these simulations. For the non-resistant mosquitoes, some mosquitoes were
repelled as early as 170 s after the repellent was first released, and all were repelled by 198 s
(Fig 4). The average repelling time was 194.5 ± 27.3 s. The resistant mosquitoes started being
repelled at 258 s, and all were repelled by 279 s (Fig 4). The average repelling time was
275.5 ± 42.6 s.
Attractant plus repellent
Here we assumed that the attractant odor and the repellent were released starting at the same
time. For the resistant mosquitoes, relative repellency efficiency (β/α) played an important role
when maximum repellent/attractant concentrations were fixed (Fig 5). If repellency efficiency
PLOS ONE | https://doi.org/10.1371/journal.pone.0244447 December 29, 2020
8 / 18
PLOS ONE Behavioral response of insecticide-resistant mosquitoes against spatial repellent
Fig 3. Frequency distribution of time when mosquitoes reached the host at different attractant release rates. C0 is the attractant
concentration.
https://doi.org/10.1371/journal.pone.0244447.g003
was low (e.g., β/α = 0.8), the simulation results indicated that a surprisingly very high propor-
tion, up to 97%, were able to reach the host before they were repelled (Fig 5A). On average
these mosquitoes reached the host within 40.7 ± 4.2 s (range 32–48 s), after which they were
pushed away from the host by the increased repellency. However, they were not pushed out of
the simulation space until 255 s (Fig 5A). Eventually, all mosquitoes were repelled, with an
average repel time of 278.1 ± 37.7 s, and all mosquitoes were repelled within about 30 s (Fig
5A).
When β/α = 1, i.e., repellent and attractant were equally efficient, about half of the mosqui-
toes were able to reach the host before they were repelled (Fig 5B). On average these mosqui-
toes reached the host within 40.7 ± 3.7 s (range 32–48 s), and they were repelled within
275.9 ± 48.0 s. When β/α = 1.2, i.e., repellent was more efficient than attractant, none of the
mosquitoes were able to reach the host and the average repel time was 277.6 ± 42.0 s (Fig 5C).
For the non-resistant mosquitoes, none reached the host before being repelled (Fig 5D).
The distribution pattern of times when mosquitoes were repelled had a long, left-tailed repel-
ling pattern (Fig 5D), which was rather different from resistant mosquitoes (Figs 4 & 5A–5C).
The first mosquito was repelled at 207 s, about 50 s faster than the resistant mosquitoes, and
the proportion of repelled mosquitoes increased slowly from 207 s to 290 s (Fig 5D). The aver-
age repel time was 273.2 ± 16.1 s, very similar to the resistant mosquitoes.
The scale-up of malaria vector control using LLINs and IRS has led to widespread insecti-
cide resistance as well as behavioral changes in Anopheles vectors. Vector resistance to pyre-
throid insecticides and changes in biting and resting behavior pose serious challenges in
PLOS ONE | https://doi.org/10.1371/journal.pone.0244447 December 29, 2020
9 / 18
PLOS ONE Behavioral response of insecticide-resistant mosquitoes against spatial repellent
Fig 4. Frequency distribution of time when mosquitoes were expelled, for non-resistant (blue bar) and resistant (red bar) mosquitoes.
https://doi.org/10.1371/journal.pone.0244447.g004
malaria control. Spatial repellents are considered to be promising alternatives to the contact
repellents currently in use, as they may delay or diminish the development of insecticide resis-
tance by minimizing the intensity of selection pressure induced by contact-mediated toxicity
mechanisms [37,40]. However, since many spatial repellents contain the same or similar chem-
icals as those used in insecticides for current malaria control, their effectiveness can be com-
promised if insecticide-resistant mosquitoes behave differently when they encounter
repellents, as has been demonstrated by recent field observations [51–58,81,84]. Our model
simulation indicated that, in general, it took more time or a higher repellent concentration to
expel mosquitoes with increased insecticide resistance. In addition, depending on resistance
level and repellency strength, a proportion of mosquitoes continued to locate hosts even in the
presence of a repellent. These results are similar to some field observations [51–58].
Resistant mosquitoes’ ability to reach the host even in the presence of repellent was as
expected. This result is similar to findings from semi-field experiments for Aedes aegypti [84].
In their study on Ae. aegypti, Buhagiar and colleagues used a real house setting and they found
that previously exposed Ae. aegypti were more likely to reach the host in a repellent-treated
room [84]. In addition, if a mosquito in a repellent-treated room had not reached the host
within 30 s, it never would [84]. This is similar to our simulation with non-resistant mosqui-
toes; i.e., none of the non-resistant mosquitoes reached the host in the spatial repellent–pro-
tected space, while resistant mosquitoes were able to reach the host before being expelled. As a
result, higher concentrations of repellent were needed to deter blood-feeding by resistant mos-
quitoes [57]. In other words, physiological insecticide resistance may compromise the efficacy
PLOS ONE | https://doi.org/10.1371/journal.pone.0244447 December 29, 2020
10 / 18
PLOS ONE Behavioral response of insecticide-resistant mosquitoes against spatial repellent
Fig 5. Frequency distribution of time when mosquitoes were attracted to the host or expelled from the simulation space, for non-resistant and resistant mosquitoes.
Red bar: Frequency distribution of resistant mosquitoes attracted to the odor release point at different times. Green bar: Frequency distribution of resistant mosquitoes
repelled from the simulation space at different times. Black curve: Frequency distribution of non-resistant mosquitoes repelled from the simulation space at different
times.
https://doi.org/10.1371/journal.pone.0244447.g005
of spatial repellents. Here, repellency efficacy is important. Previous experiments showed that
repellent may lose its efficacy against pyrethroid resistant mosquitoes [51–58]. Our simulation
indicated that if relative repellency efficacy was lower compare to attractant efficacy, a high
proportion of resistant mosquitoes were able to reach the host before they were repelled. In
Buhagiar and colleagues experiments [85], they found 31% of the mosquitoes reached the host
in the presence of repellent. When relative repellency efficacy was high, no mosquitoes could
reach the host, indicating the importance of interactions between resistance and repellent.
Interestingly, in Buhagiar et al experiments [85], all Aedes mosquitoes in the control group
reached the host at an average time of about 80 s, which was 50 s slower than treatment group.
In Parker and colleagues experiments of insecticide sensitive An. arabiensis in Tanzania [68],
they used human bait with no repellent in a hut and found out that mosquitoes first contacted
the untreated net at a mean of 36 s after release, compared to 46 s in LLINs. In our simulations,
all mosquito reached the host at an average of 83 s without repellent and mosquitoes might
reach the host earlier when attractant concentration increased, more importantly, mosquitoes
could reach the host in a short period of time (around 40 s) in the presence of repellent. These
findings were in agreement with previous experiments [68]. The question is why do mosqui-
toes reach the host earlier in repellent protected space than non-repellent protected space? Pre-
vious studies indicated that insecticides/repellents increase mosquito activity, orthokinesis
[86–88]. Kennedy’s study showed that mosquitoes get excited when they come in to contact
with the repellent due to the poisoning effect [86]. When repellent is released in the space,
mosquitoes sense the urgency to get a blood meal before they are repelled. This behavioral
characteristics is often exhibited in the presence of repellents. An experimental study on cock-
roaches using DEET repellent showed that previously exposed insects exhibited an increased
PLOS ONE | https://doi.org/10.1371/journal.pone.0244447 December 29, 2020
11 / 18
PLOS ONE Behavioral response of insecticide-resistant mosquitoes against spatial repellent
behavioral activity than non-exposed insects [89]. However, this behavioral characteristic
needs to be tested on mosquitoes in carefully designed experiments.
A more interesting finding from this study is the slow repellency effect on non-resistant
mosquitoes in an environment with both attractant and repellent (analogues to a room with a
human host and spatial repellent). If no host were in the room, the same concentration of spa-
tial repellent would repel non-resistant mosquitoes quickly. However, when a host was pres-
ent, the repellency effect was significantly delayed, with a long left tail, as if the mosquitoes
were resistant to repellent. This may be due to the mixture of attractant odors and repellent
volatiles which confuses the mosquitoes as they are attracted by the human odor, and thus
attempt to approach the host, but become disoriented by the repellent volatile chemicals. Simi-
lar finding was documented by Clark and Ray which indicated a prolonged activation of mos-
quito’s receptors in the presence of human odor (e.g. CO2) when in contact with airborne
repellent compounds through mechanisms that are not well understood [90]. These results
however need to be confirmed by semi-field or laboratory experiments.
There are differences between real-world and simulation settings. In typical African set-
tings, houses usually have eaves which allow mosquitoes to exit, so the indoor use of spatial
repellent may actually prevent mosquitoes from entering the house. However, the size of typi-
cal African houses are about 5m x 5m x 2.5 m, which is similar to the hut/house used in previ-
ous studies [68,85]. Our simulation used an ideal enclosed space; mosquitoes were not allowed
to escape. In a real-world setting, the respective diffusions of host odor and repellent volatiles
may differ in many aspects. When burning coils in a room, one can smell the chemicals clearly
at the beginning and see the smoke inside the room. Presumably, the concentration of repel-
lent volatiles is higher than that of human odor, thus preventing mosquitoes from reaching the
host. Studies indicate that spatial repellents compounds such as DEET, linalool, dehydrolina-
lool, catnip oil and citronella interfere with the attraction of mosquitoes to host odors by
blocking natural responses to attractants, hence acting as attraction inhibitors and not repel-
lents [88,91,92]. Results from Lucas et al study suggested that even in the presence of airborne
pyrethroids, mosquitoes were able to detect host odors but were inhibited from feeding [93].
This effect is probably a result of pyrethroid–induced neural hyperexcitation, that can occur at
much lower doses than those required for insect knockdown and mortality. In reality, after
mosquitoes get a full blood meal, they will rest somewhere (such as the wall or ceiling in a typi-
cal African setting) to digest the meal. In the simulation study, we assumed mosquitoes did
not move once they reached the host.
Theoretically, physiological resistance to insecticides may not necessarily affect how mos-
quitoes respond to repellents. However, selection experiments have found that spatial repel-
lent–selected mosquitoes were insensitive to the spatial repellent [83]. Field studies found that
field-collected pyrethroid-resistant Ae. aegypti were resistant to mosquito coils and other
repellents [58,80]. In a malaria vector study in Kenya, Kawada and colleagues observed the
lack of repellency effect of pyrethroids in the wild colony of An. gambiae s.s. with high resis-
tance to pyrethroids, but not in the high-resistance wild colonies of An. arabiensis and An.
funestus [52]. This was likely due to a difference in resistance mechanisms, since An. arabiensis
and An. funestus are not affected by the knockdown effect. In this study, we assumed pyre-
throid-resistant malaria vectors were insensitive to spatial repellent; however, the relationship
between resistance and repellency needs further investigation.
Conclusions
Malaria vector resistance to pyrethroid insecticides poses a serious challenge in malaria con-
trol. Spatial repellents have attracted significant attention in industry. Using SPP machine
PLOS ONE | https://doi.org/10.1371/journal.pone.0244447 December 29, 2020
12 / 18
PLOS ONE Behavioral response of insecticide-resistant mosquitoes against spatial repellent
learning modeling, we simulated the potential impact of insecticide resistance on the response
of malaria vectors to spatial repellents. We found that pyrethroid resistance may compromise
the efficacy of spatial repellents, which warrants intensive investigation.
Supporting information
S1 Fig. Illustrations of simulated 2D-projected mosquito flying paths. Scenarios: A) ran-
dom walk; B) with attractant alone; C) with repellent and no insecticide resistance; D) with
repellent and insecticide resistance; E) with repellent, attractant and resistance assuming β/α =
1.0. Simulation periods: Blue colored section 0–10 s, green colored section 11–30 s and red col-
ored section 31–60 s.
(PPTX)
Acknowledgments
We thank the two reviewers, J Rudi Strickler and the anonymous reviewer, for their construc-
tive comments to make this a better paper and suggestions for future research.
Author Contributions
Conceptualization: Guofa Zhou, Xiaoming Wang, Daibin Zhong, Ming-chieh Lee, Solomon
Kibret, Guiyun Yan.
Formal analysis: Guofa Zhou, Leonard Yu.
Funding acquisition: Guiyun Yan.
Investigation: Daibin Zhong, Solomon Kibret.
Methodology: Guofa Zhou, Ming-chieh Lee, Guiyun Yan.
Project administration: Xiaoming Wang, Guiyun Yan.
Resources: Ming-chieh Lee, Solomon Kibret.
Software: Guofa Zhou, Leonard Yu.
Supervision: Guiyun Yan.
Visualization: Leonard Yu.
Writing – original draft: Guofa Zhou.
Writing – review & editing: Guofa Zhou, Daibin Zhong.
References
1. WHO. World Malaria Report 2019. Geneva: WHO; 2019.
2. WHO. WHO recommended long-lasting insecticidal nets. Geneva: WHO; 2015. http://whoint/whopes/
Long-lasting_insecticidal_nets_November_2015pdf 2015.
3. Wanjala CL, Zhou G, Mbugi J, Simbauni J, Afrane YA, Ototo E, et al. Insecticidal decay effects of long-
lasting insecticide nets and indoor residual spraying on Anopheles gambiae and Anopheles arabiensis
in Western Kenya. Parasit Vectors. 2015; 8:588. PMCID: PMC4644290. https://doi.org/10.1186/
s13071-015-1194-6 PMID: 26567915
4. Coetzee M, Wyk P, Booman M, Koekemoer LL, Hunt RH. Insecticide resistance in malaria vector mos-
quitoes in a gold mining town in Ghana and implications for malaria control. Bull Soc Pathol Exot. 2006;
99(5):400–403. PMID: 17253060.
5. Corbel V1, N’Guessan R, Brengues C, Chandre F, Djogbenou L, Martin T, et al. Multiple insecticide
resistance mechanisms in Anopheles gambiae and Culex quinquefasciatus from Benin, West Africa.
Acta Trop. 2007; 103(3):207–216. https://doi.org/10.1016/j.actatropica.2007.01.005 PMID: 17359927.
PLOS ONE | https://doi.org/10.1371/journal.pone.0244447 December 29, 2020
13 / 18
PLOS ONE Behavioral response of insecticide-resistant mosquitoes against spatial repellent
6. N’Guessan R, Corbel V, Akogbe´to M, Rowland M. Reduced efficacy of insecticide-treated nets and
indoor residual spraying for malaria control in pyrethroid resistance area, Benin. Emerg Infect Dis.
2007; 13(2):199–206. PMCID: PMC2725864. https://doi.org/10.3201/eid1302.060631 PMID:
17479880
7. Ranson H, Abdallah H, Badolo A, Guelbeogo WM, Kerah-Hinzoumbe C, Yangalbe-Kalnone E. Insecti-
cide resistance in Anopheles gambiae: data from the first year of a multi-country study highlight the
extent of the problem. Malar J. 2009; 8:299. PMCID: PMC2804687. https://doi.org/10.1186/1475-
2875-8-299 PMID: 20015411
8. Balkew M, Ibrahim M, Koekemoer LL, Brooke BD, Engers H, Aseffa A. Insecticide resistance in Anoph-
eles arabiensis (Diptera: Culicidae) from villages in central, northern and south west Ethiopia and detec-
tion of kdr mutation. Parasit Vectors. 2010; 3(1):40. PMCID: PMC2868498. https://doi.org/10.1186/
1756-3305-3-40 PMID: 20416109
9. Ranson H, N’Guessan R, Lines J, Moiroux N, Nkuni Z, Corbel V. Pyrethroid resistance in African anoph-
eline mosquitoes: what are the implications for malaria control? Trends Parasitol. 2011; 27(2):91–98.
https://doi.org/10.1016/j.pt.2010.08.004 PMID: 20843745.
10.
Trape J-F, Tall A, Diagne N, Ndiath O, Ly AB, Faye J. Malaria morbidity and pyrethroid resistance after
the introduction of insecticide-treated bednets and artemisinin-based combination therapies: a longitu-
dinal study. Lancet Infect Dis. 2011; 11(12):925–932. https://doi.org/10.1016/S1473-3099(11)70194-3
PMID: 21856232
11. Basilua Kanza JP, El Fahime E, Alaoui S, Essassi el M, Brooke B, Nkebolo Malafu A, et al. Pyrethroid,
DDT and malathion resistance in the malaria vector Anopheles gambiae from the Democratic Republic
of Congo. Trans R Soc Trop Med Hyg. 2012; 107(1):8–14. https://doi.org/10.1093/trstmh/trs002 PMID:
23222943.
12. Knox TB, Juma EO, Ochomo EO, Pates Jamet H, Ndungo L, Chege P, et al. An online tool for mapping
insecticide resistance in major Anopheles vectors of human malaria parasites and review of resistance
status for the Afrotropical region. Parasit Vectors. 2014; 7:76. https://doi.org/10.1186/1756-3305-7-76
PMID: 24559061.
13. Mulamba C, Riveron JM, Ibrahim SS, Irving H, Barnes KG, Mukwaya LG, et al. Widespread pyrethroid
and DDT resistance in the major malaria vector Anopheles funestus in East Africa is driven by metabolic
resistance mechanisms. PLoS One. 2014; 9(10):e110058. https://doi.org/10.1371/journal.pone.
0110058 PMID: 25333491.
14. Mzilahowa T, Chiumia M, Mbewe RB, Uzalili VT, Luka-Banda M, Kutengule A, et al. Increasing insecti-
cide resistance in Anopheles funestus and Anopheles arabiensis in Malawi, 2011–2015. Malar J. 2016;
15(1):563. https://doi.org/10.1186/s12936-016-1610-1 PMID: 27876046.
15. malERA Refresh Consultative Panel on Insecticide and Drug Resistance. malERA: An updated
research agenda for insecticide and drug resistance in malaria elimination and eradication. PLoS Med.
2017; 14:e1002450. https://doi.org/10.1371/journal.pmed.1002450 PMID: 29190671.
16. Moiroux N1, Gomez MB, Pennetier C, Elanga E, Djènontin A, Chandre F, et al. Changes in Anopheles
funestus biting behavior following universal coverage of long-lasting insecticidal nets in Benin. J Infect
Dis. 2012; 206(10):1622–1629. https://doi.org/10.1093/infdis/jis565 PMID: 22966127.
17. Reddy MR, Overgaard HJ, Abaga S, Reddy VP, Caccone A, Kiszewski AE, et al. Outdoor host seeking
behaviour of Anopheles gambiae mosquitoes following initiation of malaria vector control on Bioko
Island, Equatorial Guinea. Malar J. 2011; 10:184. https://doi.org/10.1186/1475-2875-10-184 PMID:
21736750.
18. Russell TL, Govella NJ, Azizi S, Drakeley CJ, Kachur SP, Killeen GF. Increased proportions of outdoor
feeding among residual malaria vector populations following increased use of insecticide-treated nets in
rural Tanzania. Malar J. 2011; 10:80. https://doi.org/10.1186/1475-2875-10-80 PMID: 21477321.
19. Killeen GF, Marshall JM, Kiware SS, South AB, Tusting LS, Chaki PP, et al. Measuring, manipulating
and exploiting behaviours of adult mosquitoes to optimise malaria vector control impact. BMJ Glob
Health. 2017; 2(2):e000212. https://doi.org/10.1136/bmjgh-2016-000212 PMID: 28589023.
20. Killeen GF, Govella NJ, Lwetoijera DW, Okumu FO. Most outdoor malaria transmission by behaviou-
rally-resistant Anopheles arabiensis is mediated by mosquitoes that have previously been inside
houses. Malar J. 2016; 15:225. https://doi.org/10.1186/s12936-016-1280-z PMID: 27093890.
21. World Health Organization. Global strategic Framework for integrated vector management. Genera,
WHO. WHO/CDS/CPE/PVC/2004.102004, 2004.
22. Williams YA, Tusting LS, Hocini S, Graves PM, Killeen GF, Kleinschmidt I, et al. Expanding the vector
control toolbox for malaria elimination: A systematic review of the evidence. Adv Parasitol. 2018;
99:345–379. https://doi.org/10.1016/bs.apar.2018.01.003 PMID: 29530309.
PLOS ONE | https://doi.org/10.1371/journal.pone.0244447 December 29, 2020
14 / 18
PLOS ONE Behavioral response of insecticide-resistant mosquitoes against spatial repellent
23. Sangoro O, Kelly AH, Mtali S, Moore SJ. Feasibility of repellent use in a context of increasing outdoor
transmission: a qualitative study in rural Tanzania. Malar J. 2014; 13:347. https://doi.org/10.1186/1475-
2875-13-347 PMID: 25182272.
24. Sangoro O, Turner E, Simfukwe E, Miller JE, Moore SJ. A cluster-randomized controlled trial to assess
the effectiveness of using 15% DEET topical repellent with long-lasting insecticidal nets (LLINs) com-
pared to a placebo lotion on malaria transmission. Malar J. 2014; 13:324. https://doi.org/10.1186/1475-
2875-13-324 PMID: 25129515.
25. Uzzan B, Konate L, Diop A, Nicolas P, Dia I, Dieng Y, et al. Efficacy of four insect repellents against
mosquito bites: a double-blind randomized placebo-controlled field study in Senegal. Fundam Clin
Pharmacol. 2009; 23:589–594. https://doi.org/10.1111/j.1472-8206.2009.00731.x PMID: 19744033.
26.
Zion Market Research. Global market share for mosquito repellent may hit USD 5.00 Billion by 2022.
http://www.digitaljournal.com/pr/3711867.
27. Schreck CE, Kline DL. Personal protection afforded by controlled-release topical repellents and per-
methrin-treated clothing against natural populations of Aedes taeniorhynchus. J Am Mosq Control
Assoc. 1989; 5(1):77–80. PMID: 2708993.
28. Miot HA, Ferreira DP, Mendes FG, Carrenho FR, de Oliveira Amui I, Carneiro CA, et al. Efficacy of topi-
cal permethrin as repellent against Aedes aegypti’s bites. Dermatol Online J. 2008; 14(7):1. PMID:
18718185.
29.
Tuetun B, Choochote W, Pongpaibul Y, Junkum A, Kanjanapothi D, Chaithong U, et al. Celery-based
topical repellents as a potential natural alternative for personal protection against mosquitoes. Parasitol
Res. 2008; 104:107–115. https://doi.org/10.1007/s00436-008-1167-1 PMID: 18766378.
30. Maia MF, Onyango SP, Thele M, Simfukwe ET, Turner EL, Moore SJ. Do topical repellents divert mos-
quitoes within a community? Health equity implications of topical repellents as a mosquito bite preven-
tion tool. PloS One. 2013; 8(12):e84875. https://doi.org/10.1371/journal.pone.0084875 PMID:
24376852.
31. Wilson AL, Chen-Hussey V, Logan JG, Lindsay SW. Are topical insect repellents effective against
malaria in endemic populations? A systematic review and meta-analysis. Malar J. 2014; 13:446.
https://doi.org/10.1186/1475-2875-13-446 PMID: 25413142.
32. Nguyen NM, Whitehorn JS, Luong Thi Hue T, Nguyen Thanh T, Mai Xuan T, Vo Xuan H, et al. Physi-
cians, primary caregivers and topical repellent: all under-utilised resources in stopping dengue virus
transmission in affected households. PLoS Negl Trop Dis. 2016; 10(5):e0004667. https://doi.org/10.
1371/journal.pntd.0004667 PMID: 27164168.
33. Sluydts V, Durnez L, Heng S, Gryseels C, Canier L, Kim S, et al. Efficacy of topical mosquito repellent
(picaridin) plus long-lasting insecticidal nets versus long-lasting insecticidal nets alone for control of
malaria: a cluster randomised controlled trial. Lancet Infect Dis. 2016; 16(10):1169–1177. https://doi.
org/10.1016/S1473-3099(16)30148-7 PMID: 27371977.
34. Deparis X, Frere B, Lamizana M, N’Guessan R, Leroux F, Lefevre P, et al. Efficacy of permethrin-
treated uniforms in combination with DEET topical repellent for protection of French military troops in
Cote d’Ivoire. J Med Entomol. 2004; 41:914–921. https://doi.org/10.1603/0022-2585-41.5.914 PMID:
15535621.
35. Achee NL, Sardelis MR, Dusfour I, Chauhan KR, Grieco JP. Characterization of spatial repellent, con-
tact irritant, and toxicant chemical actions of standard vector control compounds. J Am Mosq Control
Assoc. 2009; 25(2):156–167. https://doi.org/10.2987/08-5831.1 PMID: 19653497.
36. Said SH, Grieco JP, Achee NL. Evaluation of contact irritant and spatial repellent behavioral responses
of male Aedes aegypti to vector control compounds. J Am Mosq Control Assoc. 2009; 25(4):436–441.
https://doi.org/10.2987/09-5895.1 PMID: 20099590.
37. Achee NL, Bangs MJ, Farlow R, Killeen GF, Lindsay S, Logan JG, et al. Spatial repellents: from discov-
ery and development to evidence-based validation. Malar J. 2012; 11:164. https://doi.org/10.1186/
1475-2875-11-164 PMID: 22583679.
38. Syafruddin D, Bangs MJ, Sidik D, Elyazar I, Asih PB, Chan K, et al. Impact of a spatial repellent on
malaria incidence in two villages in Sumba, Indonesia. Am J Trop Med Hyg. 2014; 91(6):1079–1087.
https://doi.org/10.4269/ajtmh.13-0735 PMID: 25311699.
39.
Tambwe MM, Mbeyela EM, Massinda BM, Moore SJ, Maia MF. Experimental hut evaluation of linalool
spatial repellent agar gel against Anopheles gambiae sensu stricto mosquitoes in a semi-field system in
Bagamoyo, Tanzania. Parasit Vectors. 2014; 7:550. https://doi.org/10.1186/s13071-014-0550-2 PMID:
25477012.
40. Norris JE, Coats RJ. Current and future repellent technologies: the potential of spatial repellents and
their place in mosquito-borne disease control. Int J Environ Res Public Health. 2017; 14(2): E124.
https://doi.org/10.3390/ijerph14020124 PMID: 28146066.
PLOS ONE | https://doi.org/10.1371/journal.pone.0244447 December 29, 2020
15 / 18
PLOS ONE Behavioral response of insecticide-resistant mosquitoes against spatial repellent
41. Obermayr U, Ruther J, Bernier UR, Rose A, Geier M. Evaluation of a push-pull approach for Aedes
aegypti (L.) using a novel dispensing system for spatial repellents in the laboratory and in a semi-field
environment. PLoS One. 2015; 10(6):e0129878. https://doi.org/10.1371/journal.pone.0129878 PMID:
26115365.
42. Charlwood JD, Nenhep S, Protopopoff N, Sovannaroth S, Morgan JC, Hemingway J. Effects of the spa-
tial repellent metofluthrin on landing rates of outdoor biting anophelines in Cambodia, Southeast Asia.
Med Vet Entomol. 2016; 30:229–234. https://doi.org/10.1111/mve.12168 PMID: 26991881.
43. Hogarh JN, Antwi-Agyei P, Obiri-Danso K. Application of mosquito repellent coils and associated self-
reported health issues in Ghana. Malar J. 2016; 15:61. https://doi.org/10.1186/s12936-016-1126-8
PMID: 26847206.
44.
Iwashita H, Dida GO, Sonye GO, Sunahara T, Futami K, Njenga SM, et al. Push by a net, pull by a cow:
can zooprophylaxis enhance the impact of insecticide treated bed nets on malaria control? Parasit Vec-
tors. 2014; 7:52. https://doi.org/10.1186/1756-3305-7-52 PMID: 24472517.
45. Kitau J, Oxborough R, Matowo J, Mosha F, Magesa SM, Rowland M. Indoor residual spraying with
microencapsulated DEET repellent (N, N-diethyl-m-toluamide) for control of Anopheles arabiensis and
Culex quinquefasciatus. Parasit Vectors. 2014; 7:446. https://doi.org/10.1186/1756-3305-7-446 PMID:
25249021.
46. Sangoro O1, Lweitojera D, Simfukwe E, Ngonyani H, Mbeyela E, Lugiko D, et al. Use of a semi-field
system to evaluate the efficacy of topical repellents under user conditions provides a disease exposure
free technique comparable with field data. Malar J. 2014; 13:159. https://doi.org/10.1186/1475-2875-
13-159 PMID: 24767458
47. Revay EE, Kline DL, Xue RD, Qualls WA, Bernier UR, Kravchenko VD, et al. Reduction of mosquito bit-
ing-pressure: spatial repellents or mosquito traps? A field comparison of seven commercially available
products in Israel. Acta Trop. 2013; 127(1):63–68. https://doi.org/10.1016/j.actatropica.2013.03.011
PMID: 23545129.
48. Pennetier C, Corbel V, Hougard JM. Combination of a non-pyrethroid insecticide and a repellent: a new
approach for controlling knockdown-resistant mosquitoes. Am J Trop Med Hyg. 2005; 72:739–744.
PMID: 15964959.
49. Sukumaran D, Sharma AK, Wasu YH, Pandey P, Tyagi V. Knockdown and repellent effect of permeth-
rin-impregnated army uniform cloth against Aedes aegypti after different cycles of washings. Parasitol
Res. 2014; 113:1739–1747. https://doi.org/10.1007/s00436-014-3819-7 PMID: 24595642.
50. Bibbs CS, Fulcher A, Xue RD. Allethrin-based mosquito control device causing knockdown, morbidity,
and mortality in four species of field-caught mosquitoes (Diptera: Culicidae). J Med Entomol. 2015;
52:739–742. https://doi.org/10.1093/jme/tjv065 PMID: 26335485.
51. Deletre E, Martin T, Dume´nil C, Chandre F. Insecticide resistance modifies mosquito response to DEET
and natural repellents. Parasit Vectors. 2019; 12(1): 89. https://doi.org/10.1186/s13071-019-3343-9
PMID: 30867033.
52. Agossa FR, Gnanguenon V, Anagonou R, Azondekon R, Aïzoun N, Sovi A, et al. Impact of insecticide
resistance on the effectiveness of pyrethroid-based malaria vectors control tools in Benin: decreased
toxicity and repellent effect. PLoS One. 2015; 10(12): e0145207. https://doi.org/10.1371/journal.pone.
0145207 PMID: 26674643.
53. Kawada H, Ohashi K, Dida GO, Sonye G, Njenga SM, Mwandawiro C, et al. Insecticidal and repellent
activities of pyrethroids to the three major pyrethroid-resistant malaria vectors in western Kenya. Parasit
Vectors. 2014; 7: 208. https://doi.org/10.1186/1756-3305-7-208 PMID: 24885499.
54. Muenworn V, Akaratanakul P, Bangs MJ, Parbaripai A, Chareonviriyaphap T. Insecticide-induced
behavioral responses in two populations of Anopheles maculatus and Anopheles sawadwongporni,
malaria vectors in Thailand. J Am Mosq Control Assoc. 2006; 22:689–698. https://doi.org/10.2987/
8756-971X(2006)22[689:IBRITP]2.0.CO;2 PMID: 17304938
55. Chareonviriyaphap T, Sungvornyothin S, Ratanatham S, Prabaripai A. Insecticide-induced behavioral
responses of Anopheles minimus, a malaria vector in Thailand. J Am Mosq Control Assoc. 2001;
17:13–22. PMID: 11345412
56. He Z, Zhang J, Shi Z, Liu J, Zhang J, Yan Z, et al. Modification of contact avoidance behaviour associ-
ated with pyrethroid resistance in Anopheles sinensis (Diptera: Culicidae). Malar J. 2019; 18: 131.
https://doi.org/10.1186/s12936-019-2765-3 PMID: 30971253
57. Agramonte NM, Bloomquist JR, Bernier UR. Pyrethroid resistance alters the blood-feeding behavior in
Puerto Rican Aedes aegypti mosquitoes exposed to treated fabric. PLoS Negl Trop Dis. 2017; 11(9):
e0005954. PMCID: PMC5624645. https://doi.org/10.1371/journal.pntd.0005954 PMID: 28931018
58. Wagman JM, Achee NL, Grieco JP. Insensitivity to the spatial repellent action of transfluthrin in Aedes
aegypti: A heritable trait associated with decreased insecticide susceptibility. PLoS Negl Trop Dis.
PLOS ONE | https://doi.org/10.1371/journal.pone.0244447 December 29, 2020
16 / 18
PLOS ONE Behavioral response of insecticide-resistant mosquitoes against spatial repellent
2015; 9(4):e0003726. PMCID: PMC4400042. https://doi.org/10.1371/journal.pntd.0003726 PMID:
25879206
59. Yang X. Metaheuristic Optimization: Nature-Inspired Algorithms and Applications. In: Yang X, ed. Artifi-
cial Intelligence, Evolutionary Computing and Metaheuristics Studies in Computational Intelligence.
Berlin: Springer; 2013.
60. Saffre F, Hildmann H, Deneubourg J-L. Can individual heterogeneity influence self-organised patterns
in the termite nest construction model? Swarm Intell. 2018; 12:101–110. https://doi.org/10.1007/
s11721-017-0143-8.
61.
Jun M. Research on the fish behavior simulation based on Swarm Intelligence. Procedia Eng. 2012;
43:547–551. https://doi.org/10.1016/j.proeng.2012.08.095.
62. Meng X-B, Gao XZ, Lu L, Liu Y, Zhang H. A new bio-inspired optimisation algorithm: Bird Swarm Algo-
rithm. J Exp Theor Artif Intell. 2016; 28:673–687. https://doi.org/10.1080/0952813X.2015.1042530.
63. Miller P. The Smart Swarm: How understanding flocks, schools, and colonies can make us better at
communicating, decision making, and getting things done. New York: Avery; 2010.
64. Schimansky-Geier L, Mieth M, Rose´ H, Malchow H. Structure formation by active Brownian particles.
Phys Lett A. 1995; 207:140–146. https://doi.org/10.1016/0375-9601(95)00700-D.
65. Huepe C, Aldana M. Intermittency and clustering in a system of self-driven particles. Phys Rev Lett.
2004; 92:168701. https://doi.org/10.1103/PhysRevLett.92.168701 PMID: 15169268.
66. Buhl J, Sumpter DJ, Couzin ID, Hale JJ, Despland E, Miller ER, et al. From disorder to order in marching
locusts. Science. 2006; 312:1402–1406. https://doi.org/10.1126/science.1125142 PMID: 16741126.
67. Spitzen J, Koelewijn T, Mukabana WR, Takken W. Visualization of house-entry behaviour of malaria
mosquitoes. Malar J. 2016; 15:233. https://doi.org/10.1186/s12936-016-1293-7 PMID: 27108961.
68. Parker JEA, Angarita Jaimes NC, Gleave K, Mashauri F, Abe M, Martine J, et al. Host-seeking activity
of a Tanzanian population of Anopheles arabiensis at an insecticide treated bed net. Malar J. 2017; 16
(1):270. https://doi.org/10.1186/s12936-017-1909-6 PMID: 28676092.
69. Cagnetta FCF, Gonnella G, Suma A. Large fluctuations and dynamic phase transition in a system of
self-propelled particles. Phys Rev Lett. 2017; 119:158002. https://doi.org/10.1103/PhysRevLett.119.
158002 PMID: 29077467
70. Pan Q, He Y. Recent advances in self-propelled particles. Science China Chem. 2017; 60:1293–1304.
71. Pototsky A, Hahn AM, Stark H. Rectification of self-propelled particles by symmetric barriers. Phys Rev
E Stat Nonlin Soft Matter Phys. 2013; 87(4):042124 https://doi.org/10.1103/PhysRevE.87.042124
PMID: 23679389
72. Aragones JLYS, Alexander-Katz A. Diffusion of self-propelled particles in complex media. arXiv
2018:1801.04741v1.
73. Ahmed I, Ly DQ, Ahmed W. Collective behavior of self-propelled particles in the presence of moving
obstacles. Materials Today: Proceedings. 2017; 4:65–74. https://doi.org/10.1016/j.matpr.2017.01.194.
74. Corberi F, Mazzarisi O, Gambassi A. Dynamics of fluctuations in the Gaussian model with conserved
dynamics. J Stat Mech-Theory E. 2019; 2019(10):104001. http://doi.org/10.1088/1742-5468/ab3bc7.
75. Yates CABR, Erban R, Maini PK. Refining self-propelled particle models for collective behavior. Can
App Math Quarter. 2011; 18(3):299–350.
76.
Takayuki H, Takashi S, Nobuyasu I. Collective motion in repulsive self-propelled particles in confined
geometries. J Physics: Conf Series 2017; 921:012006. https://doi.org/10.1088/1742-6596/921/1/
012006.
77. Chate´ H, Ginelli F, Gre´ goire G, Peruani F, Raynaud F. Modeling collective motion: variations on the Vic-
sek model. Eur Phys J B. 2008; 64:451–456. https://doi.org/10.1140/epjb/e2008-00275-9.
78. Schimansky-Geier L, Mieth M, Rose´ H, Malchow. Structure formation by active Brownian particles.
Physics Let A. 1995; 207:140–146. https://doi.org/10.1016/0375-9601(95)00700-D.
79. Birkhoff G, Gartland EC Jr, Lynch RE. Difference methods for solving convection-diffusion equations.
Compu Math Appl. 1990; 19(11):147–160. https://doi.org/10.1016/0898-1221(90)90158-G.
80. Amelia-Yap ZH, Chen CD, Sofian-Azirun M, Lau KW, Suana IW, Harmonis, et al. Efficacy of mosquito
coils: cross-resistance to pyrethroids in Aedes aegypti (Diptera: Culicidae) from Indonesia. J Econ Ento-
mol. 2018; 111(6):2854–2860. https://doi.org/10.1093/jee/toy296 PMID: 30265353.
81. Parker J, Angarita-Jaimes N, Abe M, Towers CE, Towers D, et al. Infrared video tracking of Anopheles
gambiae at insecticide-treated bed nets reveals rapid decisive impact after brief localised net contact.
Sci Rep. 2015; 5: 13392. https://doi.org/10.1038/srep13392 PMID: 26323965
82. Mukabana WR, Mweresa CK, Otieno B, Omusula P, Smallegange RC, van Loon JJ, et al. A novel syn-
thetic odorant blend for trapping of malaria and other African mosquito species. J Chem Ecol. 2012;
38:235–244. https://doi.org/10.1007/s10886-012-0088-8 PMID: 22426893.
PLOS ONE | https://doi.org/10.1371/journal.pone.0244447 December 29, 2020
17 / 18
PLOS ONE Behavioral response of insecticide-resistant mosquitoes against spatial repellent
83. Spitzen J, Spoor CW, Grieco F, ter Braak C, Beeuwkes J, van Brugge SP, et al. A 3D analysis of flight
behavior of Anopheles gambiae sensu stricto malaria mosquitoes in response to human odor and heat.
PLoS One. 2013; 8(5):e62995. https://doi.org/10.1371/journal.pone.0062995 PMID: 23658792.
84. Wagman JM, Achee NL, Grieco JP. Insensitivity to the spatial repellent action of transfluthrin in Aedes
aegypti: a heritable trait associated with decreased insecticide susceptibility. PLoS Negl Trop Dis. 2015;
9(4):e0003726. https://doi.org/10.1371/journal.pntd.0003726 PMID: 25879206.
85. Buhagiar TS, Devine GJ, Ritchie SA. Metofluthrin: investigations into the use of a volatile spatial pyre-
throid in a global spread of dengue, chikungunya and Zika viruses. Parasit Vectors. 2017; 10(1):270.
https://doi.org/10.1186/s13071-017-2219-0 PMID: 28558804.
86. Kennedy JS. The excitant and repellent effects on mosquitoes of sublethal contacts with DDT. Bull
Entomol Res. 1947; 37:593–607. https://doi.org/10.1017/s0007485300030091 PMID: 20287803
87. Dethier GV, Browne BL, Smith NC. The designation of chemicals in terms of the responses they elicit
from insects. J Econ Entomol. 1960; 53:134–136.
88. Ogoma SB, Ngonyani H, Simfukwe ET, Mseka A, Moore J, Maia MF, et al. The mode of action of spatial
repellents and their impact on vectorial capacity of Anopheles gambiae sensu stricto. PLoS One. 2014;
9(12):e110433. https://doi.org/10.1371/journal.pone.0110433 PMID: 25485850
89. Sfara V, Mougabure-Cueto GA, Gonza´ lez-Audino PA. Modulation of the behavioral and electrical
responses to the repellent DEET elicited by the pre-exposure to the same compound in Blattella germa-
nica. PeerJ. 2016; 4:e2150. https://doi.org/10.7717/peerj.2150 PMID: 27375967
90. Clark JT, Ray A. Olfactory mechanisms for discovery of odorants to reduce insect-host contact. J Chem
Ecol. 2016; 42(9):919–30. https://doi.org/10.1007/s10886-016-0770-3 PMID: 27628342
91. Bohbot JD, Dickens JC. Insect repellents: modulators of mosquito odorant receptor activity. PLoS One.
2010; 5(8):e12138. https://doi.org/10.1371/journal.pone.0012138 PMID: 20725637
92. Dogan EB, Rossignol PA. An olfactometer for discriminating between attraction, inhibition, and repel-
lency in mosquitoes (Diptera: Culicidae). J Med Entomol. 1999; 36(6):788–93. https://doi.org/10.1093/
jmedent/36.6.788 PMID: 10593082
93.
Lucas JR, Shono Y, Iwasaki T, Ishiwatari T, Spero N, Benzon G. U.S. laboratory and field trials of meto-
fluthrin (SumiOne) emanators for reducing mosquito biting outdoors. J Am Mosq Control Assoc. 2007;
23(1):47–54. https://doi.org/10.2987/8756-971X(2007)23[47:ULAFTO]2.0.CO;2 PMID: 17536367
PLOS ONE | https://doi.org/10.1371/journal.pone.0244447 December 29, 2020
18 / 18
PLOS ONE |
10.1371_journal.pone.0244053 | RESEARCH ARTICLE
Gender specific differences in COVID-19
knowledge, behavior and health effects
among adolescents and young adults in Uttar
Pradesh and Bihar, India
Jessie PinchoffID
D. Ngo1
1*, KG Santhya2, Corinne White1, Shilpi Rampal2, Rajib Acharya2, Thoai
1 Population Council, One Dag Hammarskjold Plaza, New York, NY, United States of America, 2 Population
Council, India Habitat Centre, New Delhi, Delhi, India
* jpinchoff@popcouncil.org
Abstract
On March 24, 2020 India implemented a national lockdown to prevent spread of the novel
Coronavirus disease (COVID-19) among its 1.3 billion people. As the pandemic may dispro-
portionately impact women and girls, this study examines gender differences in knowledge
of COVID-19 symptoms and preventive behaviors, as well as the adverse effects of the lock-
down among adolescents and young adults. A mobile phone-based survey was imple-
mented from April 3–22, 2020 in Uttar Pradesh and Bihar among respondents randomly
selected from an existing cohort study. Respondents answered questions related to demo-
graphics, COVID-19 knowledge, attitudes, and preventive behaviors practiced, and impacts
on social, economic and health outcomes. Descriptive analyses and linear probability
regression models were performed for all participants and separately for men and women. A
total of 1,666 adolescents and young adults (18–24 years old) were surveyed; 70% were
women. While most participants had high awareness of disease symptoms and preventive
behaviors, there was variation by gender. Compared to men, women were seven percent-
age points (pp) less likely to know the main symptoms of COVID-19 (coeff = -0.071; 95%
confidence interval: -0.122 - -0.021). Among women, there was variation in knowledge by
education level, urban residence, and household wealth. Women were 22 pp less likely to
practice key preventive behaviors compared to men (coeff = -0.222; 95% CIL -0.263,
-0.181). Women were also more likely to report recent depressive symptoms than men
(coeff = 0.057; 95% CI: 0.004, 0.109). Our findings underscore that COVID-19 is already
disproportionately impacting adolescent girls and young women and that they may require
additional targeted, gender-sensitive messaging to foster behavior change. Gender-sensi-
tive information campaigns and provision of health services must be accessible and provide
women and girls with needed resources and support during the pandemic to ensure gains in
public health and gender equity are not lost.
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: Pinchoff J, Santhya K, White C, Rampal S,
Acharya R, Ngo TD (2020) Gender specific
differences in COVID-19 knowledge, behavior and
health effects among adolescents and young adults
in Uttar Pradesh and Bihar, India. PLoS ONE
15(12): e0244053. https://doi.org/10.1371/journal.
pone.0244053
Editor: Kannan Navaneetham, University of
Botswana, BOTSWANA
Received: August 24, 2020
Accepted: November 25, 2020
Published: December 17, 2020
Peer Review History: PLOS recognizes the
benefits of transparency in the peer review
process; therefore, we enable the publication of
all of the content of peer review and author
responses alongside final, published articles. The
editorial history of this article is available here:
https://doi.org/10.1371/journal.pone.0244053
Copyright: © 2020 Pinchoff et al. This is an open
access article distributed under the terms of the
Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: The data are
accessible via Dataverse (https://doi.org/10.7910/
DVN/8ZVOKW).
PLOS ONE | https://doi.org/10.1371/journal.pone.0244053 December 17, 2020
1 / 13
PLOS ONE Funding: The initial UDAYA cohort was funded by
the Bill and Melinda Gates Foundation and Packard
Foundation. No additional funds were received for
the COVID-19 survey.
Competing interests: The authors have declared
that no competing interests exist.
Gender specific variation in COVID-19 knowledge, behavior and health effects among young adults
Introduction
To control the spread of the novel Coronavirus disease (COVID-19), the Indian government
swiftly instituted a shutdown of international borders and a stay-at-home order on March 24,
2020 [1]. Such ‘lockdown’ policies to prevent the spread of COVID-19 originated in high-
income countries and China; little primary research has explored the potential unintended
consequences in countries including India characterized by densely populated urban slums, a
highly mobile population, high proportions of informal sector workers, and stark variation in
poverty levels [2]. Despite rising cases, the lockdown was lifted on June 8, 2020 to begin a
phased reopening. As of August 2020, India surpassed 2.3 million cases of COVID-19, the
third highest case load after the United States and Brazil [3].
Historically, epidemics and humanitarian crises have disproportionately impacted the most
vulnerable, including women and girls [4]. Entrenched inequalities in access to education, job
opportunities, and healthcare often leave women inadequately equipped to effectively protect
themselves and their families against infection during an outbreak, and they are also more
likely to bear secondary negative effects of prolonged crises, such as economic insecurity or
challenges accessing essential health services [5]. Existing gender disparities in India may be
exacerbated or reinforced by the pandemic and are likely to affect women’s ability to make
informed decisions about adopting behaviors that mitigate risk of COVID-19.
Prevention campaigns and behavior change communication interventions across various
media, including a government-run mobile app (“Aarogya Setu”) that sends automated mes-
sages, are informing the public about COVID-19 symptoms, risk factors, and promoting pre-
ventive behaviors such as handwashing, social distancing, and wearing masks in India. To
date, there is little to no research tracing how COVID-19 messages are reaching men and
women or which sub-groups are adopting these behavioral recommendations. However, a
rapid situational assessment in the South Asia region (not including India) suggests that
women are less likely than men to have received COVID-19 information [6]. Moreover, liter-
acy, internet usage and smartphone ownership is lower among women compared to men in
India [7–9]. Accessing and understanding health promotion messages increases knowledge,
which needs to be accompanied with structural facilitators and access to resources to adopt
promoted preventive behaviors (e.g., making soap and water available for handwashing) [10–
12]. These gender gaps may result in lower adoption of promoted health behaviors and
increased risk of infection for women and girls.
The worsening COVID-19 pandemic in India is causing prolonged social and economic
disruptions that are yielding unintended consequences including economic and food insecu-
rity, and challenges in accessing healthcare. Challenges in accessing essential health services
may lead to increases in other adverse health outcomes, from vaccine preventable diseases to
poor birth outcomes and malnutrition [13,14]. This often disproportionately harms women
who may require healthcare themselves and are also often responsible for taking care of their
family’s health needs. Potential reasons for these challenges may include inability to pay clinic
fees as COVID-19 related economic insecurity persists, mobility challenges, or fear of seeking
care due to stigma or concerns about COVID-19 infection at the facility. Indeed, compared to
March 2019, March 2020 data from the Indian National Health Mission showed marked
reductions in indicators of regular health system usage [2].
In addition to physical health, lockdowns may exacerbate household stress, contributing to
increases in sexual and gender-based violence (SGBV) and poor mental health symptoms
[15,16]. While psychological distress increases generally during crises, experience of depressive
symptoms is more common among women compared to men [17]. In addition to gender, a
recent study also found that adolescents and younger adults (<25 years), those that had lost
PLOS ONE | https://doi.org/10.1371/journal.pone.0244053 December 17, 2020
2 / 13
PLOS ONE Gender specific variation in COVID-19 knowledge, behavior and health effects among young adults
employment, and/or lacked formal education were more likely to experience depressive symp-
toms as a result of the pandemic’s effects [18]. Relatedly, stress and ongoing lockdowns have
been linked with violence against women, as in past humanitarian crises [19]. Some countries
reported increases in SGBV during COVID-19 lockdowns [15,20]. Concerns around these sec-
ondary health and well-being effects are significant.
As India is home to the largest population of adolescents and young adults of any country
worldwide, understanding the impact of the pandemic on this important age cohort will also
be critical. In the age- and gender-stratified settings of India, prevailing gender disparities and
traditional gender norms affect health and well-being of adolescents and young people dispro-
portionately. However, little is known regarding the experience during the COVID-19 pan-
demic of Indian adolescent girls and young women compared to men. A cross-sectional
mobile phone-based survey of households in Uttar Pradesh (UP) and Bihar was carried out
four to six weeks after lockdown was imposed. This analysis highlights the gender specific vari-
ation in COVID-19 knowledge and practice of preventive behaviors, and mental health effects
among a cohort of adolescent and young adults. Findings from this study can inform the
development of social service programs and education campaigns to ensure that adolescent
and young women have access to tailored information and resources during this protracted
crisis to ensure development and equity gains are not lost.
Methods
Sampling strategy
A rapid telephone survey was conducted with a sample of participants drawn from an existing
Population Council cohort study of adolescents and young adults. Understanding the Lives of
Adolescents and Young Adults (UDAYA) is a state-level representative longitudinal study of
adolescent girls and boys (aged 10–19) in rural and urban settings in Bihar (n = 10,433) and
UP (n = 10,161), with baseline conducted in 2015–2016 and endline in 2018–19. The original
UDAYA study objectives were to better understand adolescents’ acquisition of assets and their
transition from adolescence to adulthood [21,22]. UDAYA researchers used the 2011 Indian
Census to create a systematic, multi-stage sampling frame for the selection of 150 primary
sampling units (PSU) in each state, with an equal breakdown between urban and rural areas.
UDAYA was designed to provide estimates for five categories of adolescents, namely unmar-
ried younger boys and girls aged 10–14, unmarried older boys and girls aged 15–19, and mar-
ried older girls aged 15–19 that represent each state [21,22].
UDAYA households eligible for inclusion in the COVID-19 survey were those in which
we interviewed a 15-19-year-old boy or girl in 2015–16. Phone numbers were available for
9,771 of such UDAYA participants– 2,437 boys and men and 7,334 girls and women. We
randomly sampled households for the mobile phone survey from this list of telephone num-
bers, stratified by gender. The enumerators contacted telephone numbers belonging to
5,520 UDAYA participants– 1,512 boys and men and 4,008 girls and women–attempting
each number up to 3 times and completing about 10 interviews per day. Of those attempted,
51% of telephone numbers were no longer functional (of UDAYA participants, 44% of boys
and men and 53% of girls and women). Of numbers we successfully reached, 5% of respon-
dents refused to participate in the study. Overall, participants in the COVID-19 study had
slightly higher educational attainment, were slightly more urban, and had slightly higher
household wealth compared to the source cohort. The characteristics of the UDAYA base-
line cohort compared to those who were enrolled in the COVID-19 mobile-phone survey is
summarized in a S1 Table.
PLOS ONE | https://doi.org/10.1371/journal.pone.0244053 December 17, 2020
3 / 13
PLOS ONE Gender specific variation in COVID-19 knowledge, behavior and health effects among young adults
Mobile phone questionnaire
Participants were contacted via mobile phone to remove the risk to field staff and participants
of COVID-19 infection. After verbal consent for participation, a short questionnaire lasting no
longer than 30 minutes was administered. The questionnaire included questions regarding
basic demographics, awareness of COVID-19 or coronavirus, knowledge of symptoms, risk
groups and transmission, perceived risk, COVID-19 prevention behaviors, and fears or con-
cerns regarding the outbreak. Questions assessing household and individual needs under the
government lockdown were also included. In the survey participants self-reported their sex as
male or female; throughout this paper we will refer to respondents as men and women to illus-
trate that our analysis reports how the pandemic impacts gender (the socially constructed
characteristics of men and women) not biological sex.
Ethical review
We received expedited ethical approval from the Population Council’s Institutional Review
Board (IRB) by meeting criteria for research conducted during COVID-19. The IRB permitted
data collection with participants with previous consent from existing cohort studies, provided
the research is aligned with national mitigation efforts. The UDAYA study protocol originally
received IRB approval in 2015 for longitudinal data collection. Participants were told they
could terminate the study at any time or skip any sections. No incentives were offered for tak-
ing part in the study.
Data management and analysis
The survey responses were entered in mini laptops using instruments developed with CSPro
7.1 and exported to Stata v15 for analysis. Each household had a unique ID number, and all
personally identifiable information was removed to ensure confidentiality.
Two summary outcome variables were created. First, participants who correctly identified
all three COVID-19 symptoms (fever, cough and difficulty in breathing) were considered to
have correct knowledge (dichotomous variable). Participants who reported implementing all
four preventive behaviors (staying home more, wearing a mask, washing hands/using sanitizer,
and staying 2m apart) were categorized as implementing the four main preventive behaviors
(dichotomous variable). Depressive symptoms, as measured by reporting feeling lonely,
depressed or irritable during the lockdown, was collected as a dichotomous variable. To con-
trol for household wealth, we created a proxy variable constructed from the presence of four
basic amenities: safe drinking water, electricity, toilet facility and safe cooking fuel. Educational
attainment was categorized into three levels, with grade 8 indicating completion of primary
education and grade 10 indicating completion of secondary education. Religion was catego-
rized as Hindu or Muslim (dichotomous variable), with 9 indicating ‘other’ and excluded from
models. Lastly, caste was categorized as scheduled caste/tribe (SC/ST), other backward castes
(OBC) and general (neither SC/ST nor OBC); these designations, as provisioned in the Indian
constitution, are used to identify marginalized groups in the population. Only women were
asked if they had experienced any violence in the home in the last 15 days under lockdown.
All survey responses were tabulated by gender and tested for statistical significance
(p<0.05) using chi-square tests. We implemented linear probability regression models based
on three outcomes of interest. First, knowledge of all three key symptoms of COVID-19. Sec-
ond, practicing all four of the key preventive behaviors. The third outcome was self-reported
experience of loneliness, depression, or irritability (dichotomous variable) in the previous
seven days used to define experience of depressive symptoms. Three separate linear probability
PLOS ONE | https://doi.org/10.1371/journal.pone.0244053 December 17, 2020
4 / 13
PLOS ONE Gender specific variation in COVID-19 knowledge, behavior and health effects among young adults
regression models were constructed for each of the three outcome variables, first for the full set
of respondents and then stratified by gender.
Results
A total of 1,666 adolescents and young adults (18–24 years) previously enrolled in the UDAYA
study were surveyed. Of these, 70% were women, over half had completed 10+ years of educa-
tion (72%) and nearly half resided in urban areas (47%) (Table 1). Fewer women (40%) than
Table 1. Demographics and COVID-19 related outcomes of interest tabulated by gender.
Demographic/Household characteristics
Age group
18–19 years
20–24 years
Religion
Hindu
Muslim
Other
Completed years of education
0–7 years
8–9 years
10 and above years
Caste
General caste
Other backward caste (OBC)
Scheduled caste/tribe
Current place of residence
Urban (vs Rural)
Have four key amenities1
Yes (vs No)
COVID-19 Outcomes of Interest
Mental Health: have you felt depressed, lonely or irritable under lockdown?
Never
Sometimes
Most of the time
Knowledge and behaviors
Knows all 3 top symptoms2
Reports practicing all 4 main preventive measures3
Economic and health access effects
Self or household member lost job/income source due to COVID-19
Among women who required each health service but could not access it:
Antenatal care
Family planning
Child immunization
Nutrition
Men
N = 506
Women
N = 1,160
Total
N = 1,666
88 (17%)
418 (83%)
432 (85%)
68 (13%)
6 (1%)
25 (5%)
59 (12%)
422 (83%)
108 (21%)
268 (53%)
130 (26%)
160 (14%)
1,000 (86%)
911 (79%)
246 (21%)
3 (0%)
195 (17%)
191 (16%)
774 (67%)
262 (23%)
615 (53%)
283 (24%)
248 (15%)
1,418 (85%)
1,343 (81%)
314 (18%)
9 (1%)
220 (13%)
250 (15%)
1,196 (72%)
370 (22%)
883 (53%)
413 (25%)
274 (54%)
502 (43%)
776 (47%)
180 (36%)
342 (29%)
522 (31%)
972 (58%)
578 (35%)
116 (7%)
266 (53%)
199 (39%)
321 (63%)
159 (31%)
26 (5%)
463 (40%)
158 (14%)
651 (56%)
419 (36%)
90 (8%)
729 (44%)
357 (21%)
p-value
0.058
<0.001
<0.001
0.785
<0.001
0.014
0.011
<0.001
<0.001
274 (54%)
788 (68%)
1,062 (64%)
<0.001
-
-
-
-
138 (12%)
239 (21%)
433 (37%)
595 (51%)
-
-
-
-
-
-
-
-
Notes
1 Includes source of light i.e. electricity, source of water i.e., improved water, source of clean fuel i.e. LPG/bio-gas and type of toilet facility i.e., own/public flush toilet
2 Three main symptoms are fever, cough, and difficulty breathing
3 Four main behaviors are stay home unless urgent, keep 2m apart from others, wear a mask, and wash hands/use sanitizer
https://doi.org/10.1371/journal.pone.0244053.t001
PLOS ONE | https://doi.org/10.1371/journal.pone.0244053 December 17, 2020
5 / 13
PLOS ONE Gender specific variation in COVID-19 knowledge, behavior and health effects among young adults
Table 2. Linear probability model of factors associated with knowledge of all 3 main COVID-19 symptoms (fever,
cough, difficulty breathing), stratified by gender.
VARIABLES
Model 1: All
Model 2: Men
Model 3: Women
(1)
(2)
(3)
Women (vs men)
Muslim (vs Hindu)
Educational attainment (0–7 years = REF)
8–9 years
10+ years
Caste (OBC = REF)
General Category
Scheduled Caste/ Tribe
-0.069��
(-0.122 - -0.021)
-0.047
NA
-0.067
NA
-0.049
(-0.109–0.015)
(-0.198–0.064)
(-0.120–0.023)
0.059
-0.102
0.088
(-0.028–0.146)
0.242��
(-0.336–0.132)
0.136
(-0.006–0.182)
0.250��
(0.171–0.314)
(-0.070–0.342)
(0.173–0.328)
0.069�
(0.010–0.128)
0.022
0.138�
(0.026–0.251)
0.044
0.040
(-0.029–0.110)
0.014
(-0.035–0.080)
(-0.061–0.149)
(-0.056–0.083)
Age 20–24 (vs 18–19 years)
0.008
0.020
0.001
(-0.056–0.073)
-0.067�
(-0.123 - -0.011)
0.111��
(-0.093–0.133)
-0.045
(-0.146–0.055)
0.104
(-0.078–0.080)
-0.077�
(-0.146 - -0.009)
0.109��
(0.049–0.172)
(-0.004–0.211)
(0.033–0.185)
-0.039
0.017
-0.061
(-0.091–0.013)
(-0.082–0.116)
(-0.123–0.001)
1,666
0.095
506
0.073
1,160
0.092
Rural (vs urban)
Household has all 4 amenities
Bihar (vs UP)
Observations
R-squared
CI in parentheses
�� p<0.01
� p<0.05
https://doi.org/10.1371/journal.pone.0244053.t002
men (53%) knew the main symptoms of COVID-19 and fewer women than men practiced key
preventive behaviors such as staying home unless it is urgent and wearing a mask (Table 1).
Fewer women reported doing all prevention behaviors (14% vs 39% of men). A greater propor-
tion of women respondents reported experience of depressive symptoms.
In the full model, women were less likely than men to know COVID-19 symptoms (coeff =
-0.069; 95% CI: -0.122 - -0.021) (Table 2). The model was then stratified by gender (men- and
women-only models). For the men-only model, there were no key characteristics associated
with more or less knowledge of symptoms, except that those in the general caste category were
14 pp more likely to know the symptoms compared with those in the OBC category
(coeff = 0.138; 95% CI: 0.026–0.251). In the women-only model, several characteristics were
associated with having more knowledge of key symptoms. Women who had completed 10
+ years of education were 25 pp more likely to know the symptoms compared with those only
having zero to seven years of education (coeff = 0.250; 95% CI: 0.173–0.328); relatedly, women
residing in households with key amenities were much more likely to know the symptoms
PLOS ONE | https://doi.org/10.1371/journal.pone.0244053 December 17, 2020
6 / 13
PLOS ONE Gender specific variation in COVID-19 knowledge, behavior and health effects among young adults
Fig 1. Proportion of respondents that practice all four the key preventive behaviors, by gender and educational
attainment.
https://doi.org/10.1371/journal.pone.0244053.g001
(coeff = 0.109; 95% CI: 0.033–0.185). Women living in rural areas had lower knowledge of the
symptoms.
Fig 1 highlights the education and gender differences in reportedly practicing all four main
preventive behaviors; this proportion increases across categories of educational attainment for
both men and women (Fig 1). Findings also show that women respondents with secondary
education (10+ years) were less likely than men respondents with less than primary education
(0–7 years) to report practicing all four prevention measures.
In the full model exploring characteristics associated with doing all four prevention behav-
iors, women were 22 pp less likely than men to report doing all behaviors (coeff = -0.221; 95%
CI: -0.263 - -0.180) (Table 3). The full model was re-run stratified by gender. Among men, sev-
eral characteristics contributed to reportedly practicing all four prevention behaviors. Men
who knew the top three symptoms were more likely to practice the four key preventive behav-
iors (coeff = 0.107; 95% CI: 0.020–0.194). Men in rural areas and in Bihar were much less likely
to carry out the four behaviors. For the women-only model, the only characteristic that was
associated with conducting the four behaviors was knowledge of the three main symptoms
(coeff = 0.160; 95% CI: 0.119, 0.201) (Table 3).
The last model explored characteristics associated with self-reported experience of depres-
sive symptoms. In the full model, women were 5 pp more likely to report that they were
experiencing depressive symptoms compared to men (coeff = 0.052; 95% CI: -0.001, 0.104)
(Table 4). When stratified by gender, among men only, household loss of employment was the
only factor associated with depressive symptoms (coeff = 0.169; 95% CI 0.083, 0.254). Among
women only, household loss of employment, religion, and experience of violence were signifi-
cantly associated with depressive symptoms. Women belonging to the Muslim religion com-
pared to those who identified as Hindu, were more likely to report experience of depressive
symptoms (coeff = 0.084; 95% CI:0.012, 0.156). Women who reported violence in the home in
the last 15 days were 30 pp more likely to report experience of depressive symptoms
(coeff = 0.304; 95% CI: 0.133; 0.475).
Women reported whether they had required health services in the previous week, and if so,
if they were able to access them (this question was not included for men). Most women had
not required health services in the previous week. Of the types of services that were required,
nutrition services and child immunization services were the most reported. Among women
who sought nutrition services, 51% required but could not access them, 1% required and were
able to access them. For child immunization services 37% were unable to access them, none
who needed child immunization services could access them. For family planning, 76% stated
PLOS ONE | https://doi.org/10.1371/journal.pone.0244053 December 17, 2020
7 / 13
PLOS ONE Gender specific variation in COVID-19 knowledge, behavior and health effects among young adults
Table 3. Linear probability model of factors associated with reporting all four main preventive behaviors are
being implemented, by gender.
VARIABLES
Women (vs men)
Knowledge of 3 key COVID symptoms
(1)
Model 1
-0.221��
(-0.263 - -0.180)
0.134��
(2)
(3)
Model 2: Men
Model 3: Women
NA
0.107�
NA
0.160��
Muslim (vs Hindu)
-0.045
-0.090
-0.018
(0.09–0.173)
(0.020–0.194)
(0.119–0.201)
Educational attainment (0–7 years REF)
8–9 years
10+ years
Caste (OBC = REF)
General Category
(-0.096–0.005)
(-0.219–0.039)
(-0.069–0.033)
REF
0.009
REF
0.126
REF
-0.008
(-0.062–0.079)
(-0.105–0.357)
(-0.075–0.059)
0.037
0.187
0.022
(-0.022–0.096)
(-0.015–0.390)
(-0.033–0.077)
REF
-0.022
REF
-0.106
REF
0.020
(-0.070–0.026)
(-0.217–0.004)
(-0.029–0.069)
Scheduled Caste/Tribe
-0.002
0.011
-0.007
(-0.049–0.045)
(-0.092–0.115)
(-0.056–0.042)
-0.004
-0.053
(-0.056–0.049)
-0.019��
(-0.074 - -0.017)
-0.056��
(-0.165–0.059)
-0.147��
(-0.234 - -0.060)
-0.103�
0.018
(-0.039–0.074)
-0.011
(-0.052–0.029)
-0.036
(-0.099 - -0.014)
(-0.201 - -0.006)
(-0.081–0.008)
1,666
0.128
506
0.059
1,160
0.066
Age group
Rural (vs Urban)
Bihar (vs UP)
Observations
R-squared
CI in parentheses
�� p<0.01
� p<0.05
https://doi.org/10.1371/journal.pone.0244053.t003
they did not require this service in the previous week, of those that did, 21% could not access
family planning services (84% of those with a family planning service need) (Fig 2).
Discussion
Conducted early in the pandemic, our study identifies gender disparities in COVID-19 related
knowledge and uptake of promoted preventive behaviors among young people in two states in
India. Overall, women were less likely to be able to identify all three of the main COVID-19
symptoms correctly, potentially due to challenges in accessing information or receiving less
accurate information of COVID-19 symptoms. Women were also less likely to be practicing
the most effective prevention behaviors and they were also more likely to report symptoms of
depression. Access to health services is also reportedly affected by the pandemic, with most
women in need of services unable to access them, including nutrition, child immunization,
family planning and antenatal care services. As of Fall 2020, the pandemic is still not under
control globally, and the threat of continued infections remains; therefore, understanding the
PLOS ONE | https://doi.org/10.1371/journal.pone.0244053 December 17, 2020
8 / 13
PLOS ONE Gender specific variation in COVID-19 knowledge, behavior and health effects among young adults
Table 4. Linear probability model of factors associated with self-reported experience of depressive symptoms dur-
ing lockdown, by gender.
VARIABLES
Women (vs men)
Household lost employment
Educational attainment (0–7 years REF)
8–9 years
10+ years
Age 20–24 (vs 18–19 years)
Rural (vs urban)
Bihar (vs UP)
Muslim (vs Hindu)
Under lockdown, experienced any violence in the home in the last 15
days (women only)
Observations
R-squared
CI in parentheses
�� p<0.01
� p<0.05
(1)
Model 1
0.052�
(0.000–
0.104)
0.133��
(0.083–
0.183)
REF
-0.006
(-0.095–
0.083)
0.018
(-0.055–
0.091)
0.021
(-0.046–
0.088)
-0.013
(-0.061–
0.035)
0.038
(-0.015–
0.092)
0.073��
(0.011–
0.135)
NA
(2)
Model 2:
Men
(3)
Model 3:
Women
NA
NA
0.169��
(0.083–
0.254)
REF
0.019
(-0.211–
0.250)
0.026
(-0.174–
0.226)
-0.001
(-0.113–
0.110)
-0.036
(-0.121–
0.049)
0.059
(-0.038–
0.156)
0.041
(-0.085–
0.168)
NA
0.117��
(0.055–0.179)
REF
-0.022
(-0.121–0.076)
0.013
(-0.067–0.092)
0.033
(-0.050–0.115)
0.002
(-0.057–0.060)
0.021
(-0.044–0.086)
0.084��
(0.012–0.156)
0.304��
(0.133–0.475)
1,658
0.027
501
0.038
1,157
0.028
https://doi.org/10.1371/journal.pone.0244053.t004
needs and experiences of adolescents and young adults is critical to offering resources and
social support, with attention to gender.
Gender differences in accurate knowledge of key COVID-19 symptoms likely reflect young
women’s lower levels of educational attainment and lower media exposure, as well as lower
access to mobile phones [21,22]. Among women, there was significant variation in the charac-
teristics of who had COVID-19 information, such as higher educational attainment, urban res-
idence, and higher economic status. These factors likely reflect higher literacy and access to
information among some young women. Interestingly, no variation was observed within men,
and overall, their knowledge was higher than for women. This finding is supported by available
literature on past pandemics. During an outbreak of influenza A (H1N1) in India, a small
study found that men had more knowledge of H1N1; this was attributed to men having more
PLOS ONE | https://doi.org/10.1371/journal.pone.0244053 December 17, 2020
9 / 13
PLOS ONE Gender specific variation in COVID-19 knowledge, behavior and health effects among young adults
Fig 2. Among women, number requiring health services and of those, number unable to obtain them by type of
service.
https://doi.org/10.1371/journal.pone.0244053.g002
social interactions through employment and having higher literacy rates than women [23].
Higher knowledge among men may be influenced by their greater exposure to risk outside the
home for work and socializing shaped by gendered social norms. A recent study from India
found differential COVID-19 risk and mortality by gender, reporting that most infections are
among men [24]. Our study also suggests that men have higher potential exposure but also
higher knowledge of COVID-19 symptoms and prevention; gender dynamics and social
norms may increase both knowledge and infection risk among men. Among women, lower
adoption of promoted behaviors may also reflect the gender roles and the fact that women
spend more time indoors. If women are not going outside, they may not be wearing masks or
keeping 2m distance from others because they are not interacting outside the household.
Knowledge was the only factor associated adoption of promoted behaviors among women;
potentially there are other unmeasured characteristics that are associated with observed varia-
tion among women. To bridge this knowledge gender gap, additional research on whether and
how the pandemic is reinforcing gender roles may help inform gender sensitive education
campaigns via media that women can access and understand even with limited literacy.
Mental health and healthcare-seeking behavior for young people are also affected. Our find-
ings suggest that loss of employment among household members due to the lockdown was
associated with depressive symptoms among both men and women. Approximately 400 mil-
lion informal sector workers in India have lost their livelihood due to COVID-19 and related
lockdowns [25]; interviews with informal sector workers describe impending poverty, evic-
tions and hunger as incomes and work opportunities are sharply curtailed [26]. Previous
research has also found a link between loss of employment and SGBV, both of which likely
relate to depressive symptoms during lockdown [15,16]. A recent study conducted prior to
COVID-19 of mental health in India found being a woman, younger age, loss of employment,
and other characteristics were associated with symptoms of depression, anxiety and stress [18].
Many women reported that they had forgone necessary medical services, which may lead to
adverse secondary health outcomes and outbreaks of other diseases. Among women surveyed,
most of those who did require a health service could not access them. Public transit commonly
used to visit clinics was closed during lockdown, which may have affected access [2]. Chal-
lenges in accessing health services must be carefully monitored to avoid unintended secondary
health crises, including outbreaks of vaccine preventable disease, stunting/undernutrition, and
unintended pregnancy or poor birth outcomes [27]. While most women reported they did not
require any health services, this study was conducted early in the pandemic. If lockdowns
PLOS ONE | https://doi.org/10.1371/journal.pone.0244053 December 17, 2020
10 / 13
PLOS ONE Gender specific variation in COVID-19 knowledge, behavior and health effects among young adults
resume or access continues to be disrupted, utilization of essential services should be moni-
tored, and steps taken to ensure accessibility.
This study has several limitations. First there are inherent challenges in conducting surveys
that are not face-to-face; mobile phone-based data collection relies on self-reported informa-
tion conveyed by participants who may have challenges understanding questions, and we can-
not guarantee protections for participants who may be vulnerable in their households [28].
Secondly, the representativeness of the sample may be compromised as we could only inter-
view those with working phone numbers from the 2015–16 UDAYA survey. Our survey
respondents had slightly higher educational attainment and household wealth compared to
the full UDAYA cohort, suggesting that the most vulnerable from the original sample were not
reachable. Third, we asked questions regarding knowledge of COVID-19 prevention behav-
iors, then later asked about behaviors respondents were doing. Potentially, question order
nudged recall, which could explain why the proportion aware of certain behaviors was lower
than those who reported implementing them. However, both the knowledge and behavior
questions were based on spontaneous responses, not a list read by the interviewer, so this effect
should be minimal. Lastly, our measure of mental health was very simple and self-reported,
validated depression measures are necessary but challenging to collect via mobile phone
interview.
Our findings suggest that early in the pandemic lockdown, there were significant knowl-
edge gaps and secondary health effects disproportionately impacting adolescent girls and
young women. To increase knowledge of symptoms and preventive behaviors, gender-sensi-
tive behavior change campaigns should be developed, and adapted for cultural context, liter-
acy, and accessibility. Improved access to information may lead to adoption of promoted
behaviors, reducing risk of infection. Relatedly, steps to address mental health and the unin-
tended secondary health impacts of the pandemic are required. To date, the Government of
India has introduced several initiatives to address these issues, for example activating a toll-
free helpline for those requiring psychosocial counseling and issuing guidelines for the sus-
tained provision of essential health services. Government agencies are also launching special
social protection initiatives. It is critical that these measures reach the most vulnerable popula-
tions, including messaging targeted to women. Longer term efforts may also be necessary to
address the prolonged and potentially gendered effects of COVID-19 and ensure that health
and development gains are not lost due to the pandemic, especially as India’s case load has
grown to one of the highest worldwide.
Supporting information
S1 Table. Differences in key background characteristics between respondents aged 15–19
whose number was not available, who were interviewed in the COVID-19 survey and who
were not interviewed in COVID-19 survey.
(TIF)
S1 File. COVID-19 study questionnaire.
(PDF)
Acknowledgments
The authors would like to acknowledge the dedicated team at Population Council Inc. in India
that collected all of these surveys and made this research happen.
Author Contributions
Conceptualization: Jessie Pinchoff, KG Santhya, Rajib Acharya, Thoai D. Ngo.
PLOS ONE | https://doi.org/10.1371/journal.pone.0244053 December 17, 2020
11 / 13
PLOS ONE Gender specific variation in COVID-19 knowledge, behavior and health effects among young adults
Data curation: Shilpi Rampal.
Formal analysis: Jessie Pinchoff, Shilpi Rampal.
Investigation: Rajib Acharya, Thoai D. Ngo.
Methodology: KG Santhya, Corinne White, Shilpi Rampal, Rajib Acharya.
Project administration: KG Santhya, Corinne White, Rajib Acharya.
Resources: KG Santhya.
Supervision: Jessie Pinchoff, KG Santhya, Rajib Acharya, Thoai D. Ngo.
Writing – original draft: Jessie Pinchoff, Corinne White.
Writing – review & editing: KG Santhya, Shilpi Rampal, Rajib Acharya, Thoai D. Ngo.
References
1.
Lancet The. India under COVID-19 lockdown. The Lancet. 2020; 395: 1315. https://doi.org/10.1016/
S0140-6736(20)30938-7 PMID: 32334687
2. Cash R, Patel V. Has COVID-19 subverted global health? The Lancet. 2020; 395: 1687–1688. https://
doi.org/10.1016/S0140-6736(20)31089-8 PMID: 32539939
3. COVID-19 Map. In: Johns Hopkins Coronavirus Resource Center [Internet]. [cited 20 Jul 2020]. Avail-
able: https://coronavirus.jhu.edu/map.html
4. Davies SE, Bennett B. A gendered human rights analysis of Ebola and Zika: locating gender in global
health emergencies. Int Aff. 2016; 92: 1041–1060. https://doi.org/10.1111/1468-2346.12704
5. Kapoor M, Agrawal D, Ravi S, Roy A, Subramanian SV, Guleria R. Missing female patients: an observa-
tional analysis of sex ratio among outpatients in a referral tertiary care public hospital in India. BMJ
Open. 2019; 9: e026850. https://doi.org/10.1136/bmjopen-2018-026850 PMID: 31391189
6. Surveys show that COVID-19 has gendered effects in Asia and the Pacific | UN Women Data Hub.
[cited 12 Jun 2020]. Available: https://data.unwomen.org/resources/surveys-show-covid-19-has-
gendered-effects-asia-and-pacific
7.
India. 27 Nov 2016 [cited 21 Jul 2020]. Available: http://uis.unesco.org/en/country/in
8. UNICEF, editor. Children in a digital world. New York, NY: UNICEF; 2017.
9. GSM Association. Connected Women: The Mobile Gender Gap Report 2020. London, United Kingdom:
GSM Association; 2020. Available: https://www.gsma.com/mobilefordevelopment/wp-content/uploads/
2020/05/GSMA-The-Mobile-Gender-Gap-Report-2020.pdf
10. Prochaska JO, Velicer WF. The transtheoretical model of health behavior change. Am J Health Promot
AJHP. 1997; 12: 38–48. https://doi.org/10.4278/0890-1171-12.1.38 PMID: 10170434
11. Parker W. Rethinking conceptual approaches to behaviour change: The importance of context.: 7.
12. Ajzen I. The theory of planned behavior. Organ Behav Hum Decis Process. 1991; 50: 179–211. https://
doi.org/10.1016/0749-5978(91)90020-T
13. Roberton T, Carter ED, Chou VB, Stegmuller AR, Jackson BD, Tam Y, et al. Early estimates of the indi-
rect effects of the COVID-19 pandemic on maternal and child mortality in low-income and middle-
income countries: a modelling study. Lancet Glob Health. 2020; 8: e901–e908. https://doi.org/10.1016/
S2214-109X(20)30229-1 PMID: 32405459
14. Akseer N, Kandru G, Keats EC, Bhutta ZA. COVID-19 pandemic and mitigation strategies: implications
for maternal and child health and nutrition. Am J Clin Nutr. 2020; 112: 251–256. https://doi.org/10.1093/
ajcn/nqaa171 PMID: 32559276
15. Wanqing Z. Domestic Violence Cases Surge During COVID-19 Epidemic. In: Sixth Tone [Internet]. 2
Mar 2020 [cited 24 Jun 2020]. Available: https://www.sixthtone.com/news/1005253/domestic-violence-
cases-surge-during-covid19-epidemic
16.
Liu N, Zhang F, Wei C, Jia Y, Shang Z, Sun L, et al. Prevalence and predictors of PTSS during COVID-
19 outbreak in China hardest-hit areas: Gender differences matter. Psychiatry Res. 2020; 287: 112921.
https://doi.org/10.1016/j.psychres.2020.112921 PMID: 32240896
17. World Health Organization. Mental health in emergencies. 11 Jun 2019 [cited 20 Jul 2020]. Available:
https://www.who.int/news-room/fact-sheets/detail/mental-health-in-emergencies
PLOS ONE | https://doi.org/10.1371/journal.pone.0244053 December 17, 2020
12 / 13
PLOS ONE Gender specific variation in COVID-19 knowledge, behavior and health effects among young adults
18. Saikarthik J, Saraswathi I, Siva T. Risk factors and protective factors of mental health during COVID-19
outbreak and lockdown in adult Indian population- A cross-sectional study. medRxiv. 2020;
2020.06.13.20130153. https://doi.org/10.1101/2020.06.13.20130153
19.
John N, Casey SE, Carino G, McGovern T. Lessons Never Learned: Crisis and gender-based violence.
Dev World Bioeth. 2020; 20: 65–68. https://doi.org/10.1111/dewb.12261 PMID: 32267607
20. Organization WH. COVID-19 and violence against women: what the health sector/system can do, 7
April 2020. 2020 [cited 21 Oct 2020]. Available: https://apps.who.int/iris/handle/10665/331699
21. Santhya KG, Acharya R, Pandey N, Singh S, Rampal S, Zavier AJ, et al. Understanding the lives of ado-
lescents and young adults (UDAYA) in Bihar, India. Population Council; 2017. https://doi.org/10.1017/
S0021932017000360 PMID: 29160192
22. Santhya KG, Acharya R, Pandey N, Gupta A, Rampal S, Singh S, et al. Understanding the lives of ado-
lescents and young adults (UDAYA) in Uttar Pradesh, India (2015–16). Population Council; 2017.
https://doi.org/10.1017/S0021932017000360 PMID: 29160192
23. Kamate SK, Agrawal A, Chaudhary H, Singh K, Mishra P, Asawa K. Public knowledge, attitude and
behavioural changes in an Indian population during the Influenza A (H1N1) outbreak. J Infect Dev
Ctries. 2010; 4: 007–014. https://doi.org/10.3855/jidc.501 PMID: 20130372
24.
Joe W, Kumar A, Rajpal S, Mishra US, Subramanian SV. Equal risk, unequal burden? Gender differen-
tials in COVID-19 mortality in India. J Glob Health Sci. 2020; 2. https://doi.org/10.35500/jghs.2020.2.
e17
25. Kelley M, Ferrand RA, Muraya K, Chigudu S, Molyneux S, Pai M, et al. An appeal for practical social jus-
tice in the COVID-19 global response in low-income and middle-income countries. Lancet Glob Health.
2020; 8: E888–E889. https://doi.org/10.1016/S2214-109X(20)30249-7 PMID: 32416766
26. Mobarak AM, Barnett-Howell Z. Poor Countries Need to Think Twice About Social Distancing. In: For-
eign Policy [Internet]. [cited 28 May 2020]. Available: https://foreignpolicy.com/2020/04/10/poor-
countries-social-distancing-coronavirus/
27. Ministry of Health and Family Welfare. Guidance Note on Provision of Reproductive, Maternal, New-
born, Child, Adolescent Health Plus Nutrition (RMNCAH+N) services during & post COVID-19 Pan-
demic. Ministry of Health and Family Welfare; 2020 pp. 1–9. Available: https://www.mohfw.gov.in/pdf/
GuidanceNoteonProvisionofessentialRMNCAHNServices24052020.pdf
28. Undie C-C, Mathur S, Haberland N, Vieitez I, Pulerwitz J. Opportunities for SGBV Data Collection in the
Time of COVID-19: The Value of Implementation Science. In: Sexual Violence Research Initiative
[Internet]. 26 Jun 2020. Available: https://svri.org/blog/opportunities-sgbv-data-collection-time-covid-
19-value-implementation-science
PLOS ONE | https://doi.org/10.1371/journal.pone.0244053 December 17, 2020
13 / 13
PLOS ONE |
10.1371_journal.pone.0260553 | RESEARCH ARTICLE
A comparison of three different methods of
eliciting rapid activity-dependent synaptic
plasticity at the Drosophila NMJ
Carolina Maldonado-Dı´azID
1,2, Mariam Vazquez1, Bruno MarieID
1,2*
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: Maldonado-Dı´az C, Vazquez M, Marie B
(2021) A comparison of three different methods of
eliciting rapid activity-dependent synaptic plasticity
at the Drosophila NMJ. PLoS ONE 16(11):
e0260553. https://doi.org/10.1371/journal.
pone.0260553
Editor: Hongkyun Kim, Rosalind Franklin University
of Medicine and Science Chicago Medical School,
UNITED STATES
Received: June 22, 2021
Accepted: November 11, 2021
Published: November 30, 2021
Copyright: © 2021 Maldonado-Dı´az et al. This is an
open access article distributed under the terms of
the Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: All relevant data are
within the paper and its Supporting information
files.
1 Institute of Neurobiology, University of Puerto Rico, Medical Sciences Campus, San Juan, Puerto Rico,
2 Department of Anatomy & Neurobiology, University of Puerto Rico, Medical Sciences Campus, San Juan,
Puerto Rico
* bruno.marie@upr.edu, brunomariemail@gmail.com
Abstract
The Drosophila NMJ is a system of choice for investigating the mechanisms underlying the
structural and functional modifications evoked during activity-dependent synaptic plasticity.
Because fly genetics allows considerable versatility, many strategies can be employed to
elicit this activity. Here, we compare three different stimulation methods for eliciting activity-
dependent changes in structure and function at the Drosophila NMJ. We find that the
method using patterned stimulations driven by a K+-rich solution creates robust structural
modifications but reduces muscle viability, as assessed by resting potential and membrane
resistance. We argue that, using this method, electrophysiological studies that consider the
frequency of events, rather than their amplitude, are the only reliable studies. We contrast
these results with the expression of CsChrimson channels and red-light stimulation at the
NMJ, as well as with the expression of TRPA channels and temperature stimulation. With
both these methods we observed reliable modifications of synaptic structures and consis-
tent changes in electrophysiological properties. Indeed, we observed a rapid appearance of
immature boutons that lack postsynaptic differentiation, and a potentiation of spontaneous
neurotransmission frequency. Surprisingly, a patterned application of temperature changes
alone is sufficient to provoke both structural and functional plasticity. In this context, temper-
ature-dependent TRPA channel activation induces additional structural plasticity but no fur-
ther increase in the frequency of spontaneous neurotransmission, suggesting an
uncoupling of these mechanisms.
Introduction
Funding: This work was supported by the NIH
NINDS-R21NS114774 to BM, the NSF HRD-
1736019 grants to B.M and M.V and the NIH RISE
R25GM061838 grant to CMD. Confocal
microscopy was supported by NIH NIGMS
GM103642 and fly husbandry by NIH-RCMI
U54MD007600. The funders had no role in study
Synaptic plasticity is at the center of cognitive processes such as learning and memory [1, 2].
This plasticity consists of increased or decreased neuronal activity leading to changes at the
synapse that will persist after the activity ceases. This phenomenon, thought to be the cellular
correlate of learning and memory, is referred to as activity-dependent synaptic plasticity [3–6].
While our understanding of the mechanisms underlying and regulating this process has
improved tremendously during the last decades [7–9], a lot is still unknown. To dissect the
PLOS ONE | https://doi.org/10.1371/journal.pone.0260553 November 30, 2021
1 / 22
PLOS ONE design, data collection and analysis, decision to
publish, or preparation of the manuscript.
Competing interests: The authors have declared
that no competing interests exist.
Three methods of eliciting rapid activity-dependent synaptic plasticity at the Drosophila NMJ
molecular mechanisms underlying activity-dependent synaptic plasticity, research has turned
towards studying this phenomenon using a variety of model systems. Indeed, in addition to
the in vivo work carried out in rodents (for review [10]) research also turned to hippocampal
neurons in culture [11–15], Caenorhabditis elegans sensory system and neuromuscular junc-
tion [16–19], and the Drosophila melanogaster glutamatergic neuromuscular junction (NMJ)
[20–23]. Methods to elicit activity-dependent synaptic plasticity at the fruit fly Drosophila mel-
anogaster NMJ have been numerous. Patterned depolarizations of the larval NMJ were first
achieved using a stimulation protocol consisting of 5 cycles of high activity in response to a
90mM KCl saline solution (High K+ protocol), direct nerve stimulation, and optogenetics.
This early work validated acute stimulation results in order to elicit both structural/morpho-
logical and functional/electrophysiological modifications [23]. Since then, the High K+ proto-
col has been extensively used [24–28]. The High K+ protocol was then adapted to a shorter
treatment that was sufficient to induce morphological changes in axotomized preparations
[21, 22, 29]. Another way used to evoke activity-dependent plasticity was direct electrical stim-
ulation. This has been used with different stimulation frequency and duration protocols and
was capable of evoking both electrophysiological and morphological modifications at the NMJ
[23, 29]. Structural plasticity was also reported after a continuous increase in motoneuron
activity induced by expressing TRPA channels in motoneurons and exposing transgenic larva
to TRPA-permissive temperature of 30˚C for 1 hour [30]. These studies presented a variety of
methods to evoke neuronal activity ranging from spaced depolarizations distributed within 28
minutes to a sustained depolarization of 60 minutes. Most studies have employed the High K
+ activity-dependent synaptic plasticity induction protocol that involves the dissection of the
larva and the patterned synchronous stimulation of the pre- and the postsynaptic compart-
ments. Different versions with varying timescales of stimulation and rest periods have been
shown to promote structural plasticity at the NMJ.
However, several questions remain. Is structural plasticity invariably coupled to functional
plasticity? Is characterization of functional modifications hampered by a possible detrimental
effect of the stimulation itself? What is the optimal induction protocol to study activity-depen-
dent synaptic plasticity at the Drosophila NMJ? Furthermore, do all patterns of increased activ-
ity evoke the same physiological response at the NMJ?
Here we investigate and compare three different ways to elicit activity-dependent synaptic
plasticity at the Drosophila NMJ and describe the morphological and electrophysiological
changes after each of these treatments. At every step, we discuss the benefits and disadvantages
of each method. We first evoke activity-dependent synaptic plasticity using a patterned High K
+ stimulation protocol established previously [23], and show that this treatment provokes
robust morphological changes but is detrimental to the physiological state of the muscle cell,
rendering the characterization of physiological modifications difficult. We then use an optoge-
netics method to evoke activity-dependent synaptic plasticity. Although optogenetics has been
used at the NMJ before [23, 31, 32], we provide the first evidence for using CsChrimson chan-
nels [33] to elicit activity-dependent synaptic plasticity at the NMJ using a patterned red light
stimulation. We find this method very efficient, permitting both morphological and
electrophysiological characterization. We finally detail the use of transgenic animals expressing
TRPA1 cationic channels [34–36]. In this case, we use different temperature changes and dif-
ferent patterns to allow activity-dependent synaptic plasticity. Surprisingly, we find that tem-
perature changes alone, in the absence of TRPA1 channel expression, can evoke
morphological and electrophysiological alterations in the synapses. Patterned activation of
temperature-driven TRPA1 can provoke additional morphological changes but no additional
electrophysiological modifications, suggesting a possible uncoupling between morphological
and functional changes.
PLOS ONE | https://doi.org/10.1371/journal.pone.0260553 November 30, 2021
2 / 22
PLOS ONE Three methods of eliciting rapid activity-dependent synaptic plasticity at the Drosophila NMJ
Materials and methods
Fly stocks
The genetic strain w1 was used as wildtype control (Bloomington Drosophila Stock Center
[BDSC] stock #145) for experiments using the High K+ approach. For these experiments we
analyzed both male and female larvae.
We used the Gal4/UAS system [37] for ectopic expression of CsChrimson and TRPA1 con-
structs. Transgene constructs used for optogenetic experiments include: UAS-CsChrimson on
the X chromosome (w1118, P[20XUAS-IVS-CsChrimson.mVenus]attP18; BDSC stock #55134)
and the D42gal4 motoneuron driver (w�; P[GawB]D42, BDSC stock #8816). Genetic controls
were heterozygous for the D42gal4 motoneuron driver insertion lacking the UAS CsChrimson
construct.
For temperature experiments, we used the UAS-TRPA1 insertion on the 2nd chromosome
(w�; P[UASTrpA1(B).K]attP16, BDSC stock #26263). Genetic controls were heterozygous for
the D42gal4 motoneuron driver insertion lacking the UAS TrpA1 construct.
Rearing methods
General rearing. All larvae were reared in standard Drosophila cornmeal media at 25˚C,
except when indicated otherwise. All larvae were reared in Jazz-mix Drosophila food (Fisher
Scientific product number: AS153), prepared as instructed by the manufacturer.
Rearing for optogenetic experiments. Larvae were reared in 400μM all-trans retinal food
at 25˚C, fully protected from the light by covering vials with aluminum foil. All-trans retinal
(Toronto Research Chemicals product number: R240000) was initially diluted to 100mM in
95% ethanol. All-trans retinal was then added to the freshly made Jazz-mix food, only when
the food temperature dropped below 57˚C, for a final concentration of 400μM. The food was
then dispatched in individual vials (protected from the light). Importantly, after the prepara-
tion of the all-trans retinal-containing food, a clean spatula was used to break down the solidi-
fied food within each individual vial to make the food on the bottom accessible for adult flies
to feed on and lay their eggs. We also added 100μL of dH2O (to vials containing around 10mL
of food) for moisture, and we dispersed the water around the inside surfaces of the food vials
(by tapping closed vials against the table). Embryos expressing CsChrimson channels in moto-
neurons (using D42gal4 driver) were not viable when placed in standard cornmeal media
without all-trans retinal, or when all-trans retinal-containing food was not well homogenized
(suggesting that the expression of CsChrimson within motoneurons creates a toxic environ-
ment in the absence of light and exogenous retinal). When transferred for experimental pur-
poses, larvae were always kept in complete darkness since the room lighting was sufficient to
activate channels and produce strong muscle contractions.
Rearing for temperature experiments. Larvae were reared in standard Drosophila corn-
meal media at 20˚C until they reached the wandering third instar stage. They were then trans-
ferred to a thermocycler for exposure to specific temperature shifts (see below). When handled
for experimental purposes, larvae were always kept at room temperature around 21˚C.
Stimulation methods and preparations for Immunohistochemistry and
electrophysiology
High K+ activity-dependent plasticity stimulation protocol. We carried out a protocol
adapted from previously published methods [23, 27]. Five spaced depolarizations were per-
formed on semi-intact third instar larvae by the bath application of a modified haemolymph-
like HL3 saline with high K+ and Ca2+ concentrations (70mM NaCl, 10 mM NaHCO3, 115
PLOS ONE | https://doi.org/10.1371/journal.pone.0260553 November 30, 2021
3 / 22
PLOS ONE Three methods of eliciting rapid activity-dependent synaptic plasticity at the Drosophila NMJ
mM sucrose, 5 mM trehalose, 5 mM HEPES, 10 mM MgCl2, 90 mM KCl and 1.5 mM CaCl2)
for stimulation cycles, while rest periods consisted of application of HL3 saline containing low
K+ and Ca2+ concentrations (70mM NaCl, 10 mM NaHCO3, 115 mM sucrose, 5 mM treha-
lose, 5 mM HEPES, 10 mM MgCl2, 5 mM KCl and 0.1 mM CaCl2). The first three stimulations
are composed of 2-minute pulses followed by 15-minute rest periods. The fourth stimulation
is composed of a 4-minute pulse followed by a 15-minute rest, and a fifth and final stimulation
is composed of a 6-minute stimulation followed by a 15-minute rest. Larval preparations were
then stretched to complete dissection prior to immunohistochemical analysis. For electrophys-
iology, larvae were gently stretched, the central nervous system (CNS) was removed, and the
body was then placed on an electrophysiology rig to acquire intracellular electrophysiological
recordings.
Optogenetic activity-dependent plasticity stimulation protocol. We adapted the opto-
genetics method from [23]. Five spaced depolarizations were performed on intact third instar
larvae by exposing transgenic larvae expressing CsChrimson channels in motoneurons to red
light pulses. Chrimson is a channelrhodopsin that is activated by red light. Upon exposure to
red light these channels allow sudden ion influx to motoneurons [38, 39]. Red light pulses
were delivered by placing larvae in a 617nm LEDs arena (Red-Orange LUXEON Rebel LED–
122 lm; Luxeon Star LEDs, Brantford, Canada). By following a specific light pattern protocol
encoded in MatLab, we achieved patterned depolarizations and elicited activity-dependent
plasticity at the NMJ. All pulses consisted of a 5-minute stimulation followed by 15-minute
rest periods for a total of 100-minutes per protocol. Within each 5 minutes of stimulation, lar-
vae were exposed to 60 rapid pulses of 2 seconds of lights on and 3 seconds of lights off. All lar-
vae were placed in a 4-well clear polystyrene dish plate (Fisher Scientific product #144444),
controls were placed in a separate well from experimental larvae. Each well contained a 1 x 1
inch Kim wipe paper with 30μL of 40% sucrose in dH20. All Chrimson-expressing larvae
showed instantaneous muscle contractions when exposed to the light. We monitored consis-
tent body wall muscle contractions during the “lights on” periods throughout the experimental
procedure. Control larvae carried the same genetic modifications as experimental larvae but
lacked the genetic construct to express Chrimson channels in motoneurons. Control larvae
did not show any behavioral response to the red-light pulses. At the end of the last rest period
larvae were dissected under a dissecting microscope using a blue LED light bulb for illumina-
tion (blue light produced subtle body wall muscle contraction that did not interfere with dis-
section, we avoided white light illumination as it resulted in strong and drastic body wall
muscle contractions). For immunohistochemistry analysis, when dissection was completed,
CNS still in place, all lights were turned off and larval preparations were fixed under minimum
light exposure with 4% paraformaldehyde for 15 minutes at room temperature. For electro-
physiology experiments, larvae were gently stretched, the CNS was removed, and the body was
placed on an electrophysiology rig under a low intensity white light that did not elicit muscle
contraction.
Temperature controlled activity-dependent plasticity stimulation protocol.
Intact
third instar larvae were exposed to temperature shifts controlled by a thermocycler (Eppendorf
Mastercycler personal, model 5332) [40, 41], to activate genetically encoded TRPA1 channels
expressed in motoneurons. Pulses were consistent with the stimulation time used during the
High K+ stimulation paradigm. The first three stimulations were composed of a 2-minute
high-temperature exposure, followed by a 15-minute rest period at a temperature below 24˚C
to avoid the activation of TRPA1 channels. The fourth stimulation was composed of a 4-min-
ute high-temperature exposure, followed by a 15-minute rest, and a fifth and final stimulation
was composed of a 6-minute high-temperature exposure, followed by a 15-minute rest. The
thermocycler settings were established as: 1. T = 29.0˚C or 27.0˚C for 2 mins; 2. T = 21.0˚C or
PLOS ONE | https://doi.org/10.1371/journal.pone.0260553 November 30, 2021
4 / 22
PLOS ONE Three methods of eliciting rapid activity-dependent synaptic plasticity at the Drosophila NMJ
23.0˚C for 15 mins; 3. Go to step 1, repeat 2 times; 4. T = 29.0˚C or 27.0˚C for 4 mins; 5.
T = 21.0˚C or 23.0˚C for 15 mins; 6. T = 29.0˚C or 27.0˚C for 6 mins; 7. T = 21.0˚C or 23.0˚C
for 15 mins; 8. Hold at 23.0˚C. We used the fastest ramp speed in between different tempera-
tures, and the lid temperature was set at 22˚C throughout the protocol. All larvae were individ-
ually placed in small 0.5ml PCR tubes with a 1 x 1 inch Kimwipe paper with 30μL of dH20 to
provide a humid environment. Genetic controls were manipulated alongside experimental lar-
vae in the same PCR machine. At the end of the last rest period, larvae were dissected under a
dissecting microscope at room temperature (around 21˚C). For immunohistochemistry analy-
sis, when dissection was completed and the CNS was still in place, larval preparations were
fixed at room temperature with 4% paraformaldehyde for 15 minutes. For electrophysiology
experiments, larvae were gently stretched, the CNS was removed, and the body was then
placed on an electrophysiology rig for quantal analysis.
Immunohistochemistry
Larval preparations were fixed with 4% paraformaldehyde for 15 minutes and washed in PBT
0.1% for 1 hour. Primary antibody mouse anti-Dlg (1:20; Developmental Studies Hybridoma
Bank, 4F3 anti-discs large) was applied overnight at 4˚C. Larval fillets were then washed in
PBT 0.1% for 1 hour. Anti-Hrp Cy3-conjugated AffiniPure goat anti-horseradish peroxidase
(1:300; Jackson ImmunoResearch product #123-165-021) and secondary antibody goat anti-
mouse Alexa Fluor 488-conjugated AffiniPure goat anti-mouse IgG (1:300; Jackson Immu-
noResearch product #115-545-166) were incubated for 1 hour at room temperature. Then a
final wash in PBT 0.1% for 1 hour was followed by mounting on a glass slide with Vectashield
(Vector Labs).
Quantification of ghost boutons
Identification of new synaptic structures (called ghost boutons, as per the previous literature)
following the activity-dependent plasticity stimulation protocol was achieved by immunologi-
cal staining of the NMJ, using a presynaptic (HRP) and a postsynaptic (Dlg) marker. “Ghost
boutons” are newly formed synaptic boutons that lack postsynaptic differentiation; therefore,
they are identified as being Hrp-positive and Dlg-negative. For each condition, control prepa-
rations were treated together with experimental preparations to account for variations in our
experimental manipulations. Quantifications were performed on NMJs of muscles 6/7 on seg-
ment A3 (right and left side of the larva) and averaged across conditions. We used a Nikon
Eclipse 80i microscope at a magnification of 400X to carry out ghost bouton identification.
Representative images were acquired using a Nikon Eclipse Ti inverted A1R laser scanning
confocal microscope. Images were acquired with oil immersion 40x with a digital zoom of 2X
(only for Fig 1), and oil immersion 60x objective. NIS elements Advance Research 4.5 acquisi-
tion and analysis software was used for image acquisition. Fiji (Image J) image processing soft-
ware was used for conversion of stacks into a single Maximum Intensity Projection, then
converted to RGB color TIFF image file format.
Electrophysiology
Intracellular electrophysiological recordings were used to assess modifications in synaptic
function following the activity-dependent synaptic plasticity protocol. Recordings were per-
formed on muscle 6, segment A3 (right and left side of the larva), using a sharp microelectrode
of borosilicate glass with a resistance of 12–20 MO filled with 3M KCl. All recordings pre-
sented for the quantification of functional plasticity have resting membrane potentials lower
than -60 mV, and muscle input resistance above 5MO. Only Fig 1G and 1H include recordings
PLOS ONE | https://doi.org/10.1371/journal.pone.0260553 November 30, 2021
5 / 22
PLOS ONE Three methods of eliciting rapid activity-dependent synaptic plasticity at the Drosophila NMJ
Fig 1. Potassium-driven stimulation evokes both structural and electrophysiological changes at the NMJ but is detrimental to muscle health. A:
Schematic diagram of the NMJ undergoing patterned stimulation under the Potassium stimulation method and giving rise to de novo synaptic
structures. B: Quantification of the average number of ghost boutons at the muscle 6/7 NMJ (segment A3) with and without stimulation. C: 2
representative NMJs at m6/7 segment A3 with immunofluorescence for a presynaptic membrane marker (red; anti-HRP) and post synaptic Discs-Large
marker (green; anti-DLG). Note that arrows point out at ghost boutons showing presynaptic fluorescence but devoid of postsynaptic immunolabelling.
D: Quantification of the average mEPSP frequency with and without stimulation. E: Quantification of the average mEPSP amplitude with and without
stimulation. F: Representative electrophysiological recordings showing spontaneous mEPSP in control and stimulated preparations. G: Quantification
of the average muscle input resistance with and without stimulation. H: Quantification of the average muscle resting potential with and without
stimulation. All animals are w1. ���� is p < 0.0001; � is p < 0.05. Kruskal-Wallis analysis with Dunn’s post-test was performed in B, G, and H. Unpaired
two-tailed t-test was performed in D and E. All quantifications show SEM. Scale is 10 μm.
https://doi.org/10.1371/journal.pone.0260553.g001
with resting membrane potentials higher (more positive) than -60mV and muscle input resis-
tance below 5 MO, in order to characterize the viability of the muscle after 5 pulses of depolari-
zations with high potassium. Data was quantified with Synaptosoft semi-automated data
analysis software. Frequency and amplitude of spontaneous neurotransmission were estab-
lished by measuring 100 continuous individual mEPSP events per NMJ recorded. The average
mEPSP frequency and amplitude were then averaged per condition. For evoked responses, we
averaged the amplitude of 20 suprathreshold evoked EPSPs for each NMJ, and then averaged
all NMJs analyzed for each condition.
Statistical treatment
We used the GraphPad Prism 6 to analyze the data presented in this manuscript. We first
assessed whether data conformed to a normal distribution by performing a Shapiro-Wilk nor-
mality test. When the Shapiro–Wilk normality test was low (p < 0.05), we ran nonparametric
tests. When comparing more than two different samples, we performed a Kruskal–Wallis test
with a post hoc Dunn’s multiple comparisons test. When comparing two samples, a Mann-
Whitney analysis was performed. When the sample distribution was normal, we ran a
parametric one-way ANOVA when comparing more than two samples. The post hoc Holm-
Sidak’s multiple comparisons test was used for multiple comparisons between data sets. When
only two data sets were compared, we performed an unpaired, two-tailed t test. The results of
PLOS ONE | https://doi.org/10.1371/journal.pone.0260553 November 30, 2021
6 / 22
PLOS ONE Three methods of eliciting rapid activity-dependent synaptic plasticity at the Drosophila NMJ
these statistical treatments are shown in the graphs of the different figures, and the specific test
used is described in the figure legend.
Results
The K+-rich depolarizing method elicits potent synapse remodeling while
reducing muscle health
To evoke activity-dependent synaptic plasticity at the Drosophila NMJ, we first used a method
of patterned depolarization by repeatedly applying a depolarizing solution (rich in calcium
and potassium) followed by a physiological solution (low in calcium and potassium) allowing
the preparation to rest (see Materials and methods section). Using this method, motoneurons
can be stimulated in a way reminiscent of the stimulation received by hippocampal neurons
leading to activity-dependent synaptic plasticity [42–46]. As a result of this stimulation, well-
documented morphological and electrophysiological changes ensue [23–27, 47]. Indeed, de
novo synaptic boutons are formed. They are mature after 24 hours but after 1 to 2 hrs only the
presynaptic side is present, making this stage ideal to identify and quantify them (Fig 1A) [23].
Immunoreactivity revealing the presence of a presynaptic side and the absence of postsynaptic
differentiation allows the identification of these boutons, termed “ghost boutons”, and is used
to quantify the magnitude of activity-dependent synaptic plasticity (Fig 1B and 1C). In our
hands, we see that, while there are few ghost boutons in unstimulated controls (average of
1.5 ± 0.3, Fig 1B and 1C), their numbers increase tremendously after stimulation (average of
7.9 ± 1.2, Fig 1B and 1C; p < 0.0001). Because phenomena of activity-dependent synaptic plas-
ticity are time sensitive, we looked at 2 different times after the start of the repeated stimulation
treatments; a rest of 15 min for a total treatment of 90 min and a rest of 45 min for a total pro-
cedure of 120 min. We did not see any difference between these two conditions (compare
7.9 ± 1.2 for 90 min with 6.4 ± 0.4 for 120 min in Fig 1B, p = 0.51) suggesting that 15 min of
rest after the last pulse of the stimulation is enough to evoke a full activity-dependent synaptic
plasticity response and that this synaptic remodeling persists. In addition to these morphologi-
cal changes, modifications of electrophysiological properties also occur. Because they are mea-
sured 45 min after the last pulse of the stimulation, at the time when ghost boutons are
immature, these changes are thought to be independent of de novo bouton formation and rep-
resent a modification in the basic properties of the original synaptic structures [23, 26]. Indeed,
the frequency of miniature excitatory post synaptic potentials (mEPSPs) is increased (compare
2.3 ± 0.2 Hz at rest with 4.2 ± 0.3 Hz after stimulation; Fig 1D and 1F; p < 0.0001). In addition,
after repeated stimulation, we see a small but statistically significant increase in mEPSP ampli-
tude (0.68 ± 0.045 mV at rest and 0.81mV ± 0.037 mV after stimulation; Fig 1E; p = 0.046) that
might represent an effect also described by others [23]. While the increase in mEPSP frequency
is dependent on transcription, translation [26], and Wingless signaling [23], little is known
about this subtle increase in mEPSP amplitude besides the fact that it is not dependent on
Wingless signaling [23]. It was hypothesized to result from a variety of presynaptic modifica-
tions like the release of multiple vesicles at the same time, or an increase in vesicle size that was
previously reported [48]; or postsynaptic changes like modifications in glutamate receptor
function [23].
During our electrophysiological experiments, we noticed a reduction in stimulated muscles’
input resistance and a depolarization of their resting potential. Because the input resistance
has been characterized as a factor influencing mEJP amplitude [49–51] and because the resting
potential is typically used to assess membrane integrity after electrode penetration, we and oth-
ers have defined criteria allowing recording of mEPSPs and EPSP (see Materials and methods).
Surprisingly, most of the preparations after stimulation failed to pass these criteria (see S1
PLOS ONE | https://doi.org/10.1371/journal.pone.0260553 November 30, 2021
7 / 22
PLOS ONE Three methods of eliciting rapid activity-dependent synaptic plasticity at the Drosophila NMJ
Table). We therefore decided to characterize the effect of K+ stimulation on muscle input
resistance and resting potential which are readily quantifiable criteria for assessing muscle via-
bility [52, 53].
Two factors could be detrimental to muscle health, the time (120 min) left exposed as a
semi-intact preparation in physiological serum [54] (Material and Methods), and the repeated
high-potassium depolarizations. To test the relative importance of these two conditions, we
measured preparations dissected and immediately recorded (0 min), preparations dissected
and left in the physiological serum for 120 min before being recorded, and preparations
repeatedly stimulated for 120 min. We find that muscle health is not greatly affected in semi-
intact preparations spending 120 min in physiological saline; although the resting potential
appears to be slightly depolarized at 120 min there is no statistically significant difference
between median values (compare -65.6 ± 1.2 mV at 0 min with -61.7 ± 1.1 mV at 120 min; Fig
1H; p = 0.17); similarly the change in input resistance is also not statistically significant (com-
pare 15.6 ± 0.9 MΩ at 0 min with 12.3 ± 1.7 MΩ at 120 min; Fig 1G; p = 0.24). In contrast, the
preparations that underwent repeated potassium stimulations show clear signs of muscle dis-
tress because their mean resting potential is significantly reduced by about 9% (-59.5 ± 0.9 mV
at 120 min, Fig 1H; p < 0.0001 compared to controls at 0 min) and their average input resis-
tance is reduced by 56% to 6.8 ± 0.9 MΩ (Fig 1G; p < 0.0001 compared to controls at 0 min).
We also noted that there was no statistically significant difference between controls at 120 min
and stimulated preparations at 120 min for both Input resistance (Fig 1G; p = 0.08) and resting
potential (Fig 1H; p = 0.4). This suggests that it is the combination of both time and K+ stimu-
lation that is responsible for the observed deleterious effects on Input resistance and resting
potential. This does not affect the conclusions/observations we and others made on the fre-
quency of mEPSPs; the increase in frequency after stimulation could be, if anything, underesti-
mated. Indeed, mEPSP decreased amplitude due to the state of the muscle could mean that
some mEPSPs are not counted. In addition, the quantification of mEPSP frequency and ampli-
tude (Fig 1D–1F) were made on the subgroup of synapses presenting an input resistance
greater than 5 MΩ and a resting potential more hyperpolarizing than -60 mV (see Materials
and methods). Nevertheless, this provides a strong argument for the need of a less invasive
form of repeated stimulation to induce activity-dependent synaptic plasticity.
Activation of transgenically-encoded CsChrimson in motoneurons elicits
synapse remodeling and allows electrophysiological recordings
Looking to improve the overall state of our preparations we decided to perform optogenetic
stimulation. We established transgenic animals expressing the red-light gated cation channel
CsChrimson [38] under the control of D42-Gal4 (driving the expression in motor and sensory
neurons; [55]). Like in other optogenetics experiments [33], the opening of the CsChrimson
channel leads to the depolarization of the cell in which it is expressed [38]. Chrimson is a
Channelrhodopsin activated by high wavelengths of light, with the strongest response at
590nm [38], which penetrate the cuticle better than shorter wavelengths [56, 57]. When
expressed in motoneurons it can elicit action potentials leading to muscle depolarization [38].
Nevertheless, to date it has not been used at the Drosophila NMJ to induce synaptic plasticity.
To this effect, we raised these animals on all-trans retinal containing food and submitted them
to a patterned light stimulation (see Materials and methods and Fig 2). We found that this
treatment can provoke the appearance of ghost boutons, the morphological modifications
characteristic of activity-dependent synaptic plasticity. Indeed, our unstimulated control prep-
arations show the appearance of 1.7 ± 0.4 ghost boutons while the preparation stimulated with
our optogenetic strategy showed 4.7 ± 0.5 ghost boutons (Fig 2B and 2C; p = 0.0007). These
PLOS ONE | https://doi.org/10.1371/journal.pone.0260553 November 30, 2021
8 / 22
PLOS ONE Three methods of eliciting rapid activity-dependent synaptic plasticity at the Drosophila NMJ
Fig 2. Chrimson-driven activity-dependent plasticity at the Drosophila NMJ. A: Schematic diagram of the NMJ undergoing optogenetic stimulation
using red-light. B: Quantification of the average number of ghost boutons at the muscle 6/7 NMJ (segment A3) with and without stimulation. C: 2
representative NMJs at m6/7 segment A3 with immunofluorescence for a presynaptic membrane marker (red; anti-HRP) and post synaptic Discs-Large
marker (green; anti-DLG). Arrows indicate ghost boutons showing presynaptic fluorescence but devoid of postsynaptic immunolabelling. D:
Quantification of the average mEPSP frequency with and without stimulation. E: Representative electrophysiological recordings showing spontaneous
mEPSPs in control and stimulated preparations. F: Quantification of the average mEPSP amplitude with and without stimulation. G: Quantification of
the average muscle input resistance with and without stimulation. H: Quantification of the average muscle resting potential with and without
stimulation. I: Quantification of the average evoked EPSP amplitude in unstimulated and stimulated preparations J: Representative traces of evoked
EPSPs in unstimulated and stimulated preparations. K: Quantification of the average quantal content (number of vesicles released by action potential).
Unstimulated animals were D42-Gal4/+ and stimulated animals were UAS-CsChrimson/+ (female) or UAS-CsChrimson/Y (male); D42-Gal4/+; ��� is
p < 0.001; �� is p < 0.01. Kruskal-Wallis analysis with Dunn’s post-test was performed in B. Unpaired two-tailed t-test was performed in D, F, G, I and
K. Mann-Whitney analysis was performed in H. All quantifications show SEM. Scale is 10 μm.
https://doi.org/10.1371/journal.pone.0260553.g002
results are similar to previous observations [23, 32], where the authors used the blue-light-
gated channel Channelrhodopsin-2 under the control of C380-Gal4 and OK6-Gal4 to drive
optogenetic stimulation at the larval NMJ. Like us, they showed that presynaptic optogenetic
stimulation induced fewer structural changes when compared to the K+-driven stimulation.
This is probably because a K+ shock directly depolarizes the postsynaptic muscle. In contrast,
optogenetic stimulation is only driven in a subset of neurons and thus induces activity-depen-
dent synaptic plasticity that is the sole consequence of repeated neuronal synaptic activity. We
also tested whether allowing an additional 30 min of rest would induce the formation of more
ghost boutons. Indeed, in these conditions the unstimulated preparations show the presence
PLOS ONE | https://doi.org/10.1371/journal.pone.0260553 November 30, 2021
9 / 22
PLOS ONE Three methods of eliciting rapid activity-dependent synaptic plasticity at the Drosophila NMJ
of 1.3 ± 0.3 ghost boutons while the stimulated preparations show 4.4 ± 0.7 (Fig 2B;
p = 0.0003). These data are not significantly different from the ones observed after only 15 min
of rest after the last pulse (Fig 2B).
Because the main reason for a shift to optogenetics was to preserve the preparations and
allow for electrophysiological recordings, we tested the input resistance and resting potential
for the preparations subjected to optogenetic stimulation. We found that, in stark contrast to
the K+-driven stimulation, there are no deleterious effects associated with the optogenetic pro-
tocols. Indeed, the mean input resistance of our unstimulated control preparations is
20.2 ± 1.2 MΩ while the preparations stimulated by optogenetics showed a comparable input
resistance of 18.8 ± 0.9 MΩ (Fig 2G; p = 0.3). Importantly, there was no depolarization of mus-
cle resting potential in stimulated preparations; stimulated preparations showed a mean rest-
ing potential of -69.5 ± 2 mV compared to controls (-64.2 ± 2.1 mV), although this difference
was not statistically significant (Fig 2H; p = 0.05). These two electrophysiological characteris-
tics illustrate that optogenetic stimulation does not affect the health of the preparation. Having
achieved this, we then assessed the potentiation of mEPSP release frequency. We found that
there is a 65% increase in mEPSP frequency after stimulation. Control preparations show an
average frequency of 3.1 ± 0.2 Hz while the stimulated preparations show an average of
5.1 ± 0.4 Hz (Fig 2D and 2E; p = 0.0002). The amplitude of the mEPSPs is also increased after
this stimulation paradigm (1 ± 0.05 mV for controls and 1.2 ± 0.04 mV after stimulation, Fig
2E and 2F; p = 0.008). The evoked EPSP amplitudes or the quantal content do not show any
difference compared to controls. After stimulation, the average EPSP amplitude is 27.1 ± 1.7
mV and is not significantly different from control measurements (30.3 ± 2.3 mV; Fig 2I and
2J; p = 0.27). This is reminiscent of previous data showing that EPSP amplitude does not
change upon 5 cycles of spaced depolarizations [23]. The resulting quantal content (number of
vesicles released by action potential) also shows no statistically significant change. Quantal
content is 29.9 ± 3 in controls and 22.6 ± 1.8 in stimulated preparations (Fig 2J and 2K;
p = 0.054). In any case, we show that this optogenetic manipulation is adequate for the assess-
ment of activity-dependent synaptic plasticity both at the morphological and electrophysiolog-
ical levels.
The effect of temperature and of temperature-driven activation of TRPA
on synaptic morphology and function
Another way to manipulate neuronal activity is to use transgenic flies expressing the tempera-
ture sensitive TRPA channel in motoneurons. This warmth-activated channel can elicit depo-
larization in other systems. In adult Drosophila flies, depolarization of photoreceptor cells was
achieved with genetically encoded expression of TRPA1 channels in these cells and exposure
of flies to a continuous stimulus of 29˚C [35]. In larval motoneurons, chronic neuronal overac-
tivation was achieved with cell-specific TRPA1 expression and a continuous exposure to rear-
ing temperatures of 25˚C and 27˚C [58]. The use of temperature as the triggering factor for
motoneuron stimulation is interesting. Indeed, temperature can penetrate tissues more effi-
ciently than light [59] which could constitute an incremental improvement compared to the
optogenetics stimulation. In addition, the experimental setup is simpler and more affordable
(bain-marie or thermocycler; see Materials and methods). Although TRPA-driven stimulation
to promote rapid activity-dependent synaptic plasticity has been used at the Drosophila NMJ,
it was utilized to depolarize motoneurons in one continuous stimulus of permissive tempera-
ture exposure [30].
Because temperature is an important factor influencing an array of behavioral [60] and
physiological characteristics like gene expression [61], RNA editing [62–64], and protein
PLOS ONE | https://doi.org/10.1371/journal.pone.0260553 November 30, 2021
10 / 22
PLOS ONE Three methods of eliciting rapid activity-dependent synaptic plasticity at the Drosophila NMJ
activity including ion channel kinetics [65, 66], we asked whether temperature on its own
could have an effect on NMJ structure. We first subjected preparations lacking TRPA channels
raised at 20˚C to a constant temperature of 29˚C for 90 min, before assessing NMJ structure.
In control preparations (kept at 20˚C), ghost boutons average 0.4 ± 0.2 per synapse while after
90 min at 29˚C they average 1.9 ± 0.5 (Fig 3B and 3C). This is a significant, almost 5-fold
increase (p = 0.0076), strongly suggesting that a constant rise in temperature alone can pro-
voke morphological changes at the synapse. However, our intention in this study was to evoke
patterned depolarizations in motoneurons in an attempt to mimic physiological stimuli [23,
47]. We asked whether the same phenomenon could be observed if we applied a stimulation
protocol consisting of 5 cycles of temperature pulses. We developed a patterned stimulation
protocol using temperature pulses, based on the previously described potassium-based and
optogenetic depolarization protocols (see Materials and methods). We chose 21˚C as the rest-
ing temperature and gave pulses of 29˚C temperature [34, 36]. We also tested whether allowing
for an additional 30 min of rest could have an effect, since structural changes that arise as a
consequence of activity-dependent synaptic activity are expected to promote lasting changes at
the NMJ. In the experiments in which we assessed the appearance of ghost boutons in control
animals 15 min after the last pulse of the stimulation (stimulated 90 min) we observed a mean
of 0.9 ± 0.2 ghost boutons per synapse compared to controls (mean 0.4 ± 0.2 boutons. How-
ever, this change was not statistically significant (Fig 3D; p = 0.85). We then asked if these
ghost boutons could develop after an additional 30 min of rest. To our surprise there was a
large, significant difference, with an average of 3.7 ± 0.7 ghost boutons (Fig 3D and 3E;
p < 0.0001). This suggests that patterned temperature stimulation of control animals is suffi-
cient to provoke morphological changes at the synapse typical of activity-dependent synaptic
plasticity.
We then wondered if this phenomenon of temperature-evoked activity-dependent synaptic
plasticity in control animals would hold for smaller temperature steps. We therefore looked
for the minimal temperature range that would leave the TRPA channel inactive at one extreme
and trigger its activation at the other. Excitatory junction potentials were identified at the larval
NMJ expressing TRPA1 channels with temperatures above 25˚C [34], and tonic spikes were
identified in water baths over 26˚C, while 23–24˚C temperatures did not generate action
potentials [36]. We therefore decided to apply 23˚C as a resting temperature and 27˚C as a
stimulating temperature to control animals. In these conditions, the average number of ghost
boutons was 2.3 ± 0.4 (Stimulated 90 min; Fig 3F and 3G; p = 0.0012) and after allowing for an
additional 30 min of rest 1.7 ± 0.4 (p = 0.06), whereas controls showed only 0.4 ± 0.2 ghost
boutons. This shows that this treatment too can elicit a morphological activity-dependent syn-
aptic plasticity response. Interestingly, because we saw an effect after 90 min with the 23–27˚C
protocol and not with the 21–29˚C protocol, it could mean that 23˚C is enough to provoke a
temperature driven stimulation during rest periods while 21˚C is not. Taken together our data
show that higher temperatures applied continuously or in pulses affect the NMJ and provoke
morphological effects typical of activity-dependent plasticity.
We then asked whether transgenic animals expressing the TRPA construct and submitted
to the same stimuli could show additional changes in synaptic morphology. We first used the
continuous temperature protocol and showed that there is no increased effect due to the pre-
sumed additional TRPA stimulation (Fig 4B and 4C). This interesting result suggests that the
effects observed under these conditions depend on temperature and not TRPA-driven depo-
larization. This might be because TRPA is more sensitive to a change in temperature than to
its absolute value. We therefore asked whether the patterned stimulation protocols could show
increased morphological modifications. Indeed, in all the 4 protocols that we tested (pulses
going from 21˚C to 29˚C and 23˚C to 27˚C; 90 and 120 min after the start of the stimulation;
PLOS ONE | https://doi.org/10.1371/journal.pone.0260553 November 30, 2021
11 / 22
PLOS ONE Three methods of eliciting rapid activity-dependent synaptic plasticity at the Drosophila NMJ
Fig 3. Temperature can drive structural activity-dependent synaptic plasticity. A: Schematic diagram of the NMJ submitted to a temperature
stimulus. B, D, and F: Quantification of the average number of ghost boutons at the muscle 6/7 NMJ (segment A3) with and without stimulation. D and
F: a different rest time was applied after the stimulation protocol and indicated on the graphs (total time of 90 min or 120 min). C, E, and G:
representative NMJs at m6/7 segment A3 with immunofluorescence for a presynaptic membrane marker (red; anti-HRP) and post synaptic Disc-Large
marker (green; anti-DLG). Arrows indicate ghost boutons showing presynaptic fluorescence but devoid of postsynaptic immunolabelling. B and C: A
constant temperature change of 29˚C was applied. D and E: Patterned steps from 21˚C to 29˚C were applied. F and G: Patterned steps from 23˚C to
27˚C were applied. All animals were D42-Gal4/+. ���� is p < 0.0001; �� is p < 0.01. Mann-Whitney analysis was performed in B. Kruskal-Wallis
analysis with Dunn’s post-test was performed in D and F. All quantifications show SEM. Scale is 10 μm.
https://doi.org/10.1371/journal.pone.0260553.g003
Fig 4D–4I) we observed an increase in ghost bouton formation when we compared tempera-
ture stimulation alone to temperature-activated TRPA animals. For stimulation using pulses
from 21˚C to 29˚C, we observed 0.9 ± 0.2 (after 90 min) and 3.7 ± 0.7 (after 120 min) ghost
boutons with temperature alone and 3.7 ± 0.7 (after 90 min; p = 0.001) and 7.6 ± 0.7 (after 120
min; p = 0.007) ghost boutons in animals expressing TRPA. Similarly, using pulses from 23˚C
to 27˚C induced 2.3 ± 0.4 (after 90 min) and 1.7 ± 0.4 (after 120 min) ghost boutons with tem-
perature alone and 6.4 ± 0.5 (after 90 min; p = 0.0012) and 6.5 ± 1.1 (after 120 min; p = 0.025)
PLOS ONE | https://doi.org/10.1371/journal.pone.0260553 November 30, 2021
12 / 22
PLOS ONE Three methods of eliciting rapid activity-dependent synaptic plasticity at the Drosophila NMJ
Fig 4. TRPA-driven activity-dependent structural plasticity at the Drosophila NMJ. A: Schematic diagram of the NMJ submitted to a temperature
stimulus that allows the entry of cations through the TRPA channels. B, D, E, G, and H: Quantification of the average number of ghost boutons at the
muscle 6/7 NMJ (segment A3) with and without stimulation. D and G: Preparations were given 15 min rest time (total procedure time of 90 min). E
and H: a rest time of 45 min was applied after the stimulation protocol (total time of 120 min). C, F, and I: representative stimulated NMJs from animals
expressing the TRPA transgene at muscle 6/7 segment A3 with immunofluorescence for a presynaptic membrane marker (red; anti-HRP) and post
synaptic Disc-Large marker (green; anti-DLG). Arrows indicate ghost boutons showing presynaptic fluorescence but devoid of postsynaptic
immunolabelling. B and C: A constant temperature change of 29˚C was applied. D-F: Patterned steps from 21˚C to 29˚C were applied. G-I: Patterned
steps from 23˚C to 27˚C were applied. B, D and G: data was collected 90min after the start of the protocol. E and H: data was collected 120min after the
start of the protocol. Animals were D42-Gal4/+ (Control and Stimulated temperature) or UAS-TrpA/+; D42-Gal4/+ (Stimulated TrpA). ���� is
p < 0.0001; ��� is p < 0.001; �� is p < 0.01 and � is p < 0.05. Kruskal-Wallis analysis with Dunn’s post-test was performed in all the graphs. All
quantifications show SEM. Scale is 10 μm.
https://doi.org/10.1371/journal.pone.0260553.g004
ghost boutons in animals expressing TRPA. This suggests that TRPA activation can efficiently
depolarize the motoneurons and create morphological modifications as a consequence of
activity-dependent synaptic plasticity. It is worth keeping in mind that temperature affects the
entire organism, including the postsynaptic muscle fiber. Hence the effects observed by depo-
larizing a neuronally-expressed TRPA channel are likely a composite of the effect originating
from the presynaptic TRPA-driven Ca2+ influx and a more general temperature effect.
Intrigued by the ability of temperature and patterned TRPA-driven stimulation to evoke
activity-dependent synaptic plasticity at the morphological level, we asked whether these con-
ditions could drive electrophysiological changes. We focused on the stimuli that consist of pat-
terned pulses of temperature from 23˚C to 27˚C, as being the smallest temperature fluctuation
capable of activating and inactivating TRPA channels. We first wondered about the physiolog-
ical status of the preparations after exposure to pulses of temperatures. We found that the
input resistance of the preparations submitted to 23˚C to 27˚C temperature pulses (average of
PLOS ONE | https://doi.org/10.1371/journal.pone.0260553 November 30, 2021
13 / 22
PLOS ONE Three methods of eliciting rapid activity-dependent synaptic plasticity at the Drosophila NMJ
Fig 5. Temperature and TRPA-driven activity-dependent electrophysiological plasticity at the Drosophila NMJ. A: Quantification of the average
mEPSP frequency with and without stimulation. B: Quantification of the average mEPSP amplitude with and without stimulation. C: Quantification of
the average muscle input resistance with and without stimulation. D: Representative electrophysiological recordings showing spontaneous mEPSP in
control and stimulated preparations. E: Representative electrophysiological recordings of evoked EPSPs in unstimulated and stimulated preparations. F:
Quantification of the average muscle resting potential with and without stimulation. G: Quantification of the average evoked EPSP amplitude in
unstimulated and stimulated preparations H: Quantification of the average quantal content (number of vesicles released by action potential). Animals
were D42-Gal4/+ (Control and Stimulated temperature) and UAS-TrpA/+; D42-Gal4/+ (Stimulated TrpA). ���� is p < 0.0001; �� is p < 0.01; � is
p < 0.05. A one-way ANOVA with Holm-Sidak’s multiple comparisons test was applied in A, B, G, H. A Kruskal-Wallis analysis with Dunn’s post-test
was performed in C and F. There were no significant differences detected by this test in G and H. All quantifications show SEM. Scale is 10 μm.
https://doi.org/10.1371/journal.pone.0260553.g005
15.1 ± 1.1 MΩ; Fig 5C) was no different from controls that were not exposed to pulses of tem-
peratures (18 ± 1 MΩ; Fig 5C; p = 0.17). The stimulated preparations containing the TRPA
transgene showed a slight decrease in input resistance compared to controls (14 ± 0.8 MΩ;
Fig 5C; p = 0.013), but this decrease was much subtler than the one observed with the K+ depo-
larization protocol (Stimulated 120 min showed 6.8 ± 0.9 MΩ; Fig 1G). In addition, an input
resistance of 14 MΩ is still considered to indicate a healthy preparation. When we looked at
the resting potential of these preparations, we found that TRPA-driven stimulated prepara-
tions showed resting potentials more hyperpolarized than control preparations (-69 ± 1.3 mV
compared to -62.8 ± 0.8 mV; Fig 5F; p = 0.007) while temperature alone was as hyperpolarized
as controls (-64.6 ± 1 mV; p = 0.76). Together these results suggest that the preparations are
healthy following such stimulations. This gave us the opportunity to ask whether mEPSP fre-
quency and amplitude as well as EPSP amplitude could be affected by such a treatment. We
PLOS ONE | https://doi.org/10.1371/journal.pone.0260553 November 30, 2021
14 / 22
PLOS ONE Three methods of eliciting rapid activity-dependent synaptic plasticity at the Drosophila NMJ
first asked whether temperature alone or temperature-triggered TRPA opening could provoke
an increase in mEPSP frequency after patterned stimulation. We find that, as with the mor-
phological modifications, patterned temperature pulses alone were sufficient to elicit
electrophysiological changes typical of activity-dependent synaptic plasticity. Indeed, while the
mEPSP frequency of control preparations always kept at 20˚C is 3.6 ± 0.3 Hz on average, it is
6 ± 0.4 Hz for preparations subjected to patterned pulses of temperature from 23˚C to 27˚C
(Fig 5A and 5D; p = 0.0002). Surprisingly, and in contrast to what we observed for morpholog-
ical modifications, there is no added effect of TRPA-driven depolarization on the frequency of
mEPSPs (6.1 ± 0.3 Hz; Fig 5A and 5D; p > 0.99). Regarding the mEPSP amplitudes, we found
that none of the stimuli (stimulated temperature 0.94 ± 0.03 mV and stimulated TrpA
0.83 ± 0.02 mV) can elicit a statistically significant increase in the average of mEPSP ampli-
tudes (control 0.84 ± 0.04 mV; Fig 5B and 5D; p = 0.055 and p = 0.91 respectively) in contrast
to what we observed with optogenetics (Fig 2E and 2F) and the High K+ protocol (Fig 1E and
1F). This suggests that the use of TRPA warmth-gated channels and exposure of transgenic
larva to different temperatures might not evoke the same response than the one we observed
with optogenetics. In these experiments, we did not observe any changes in the evoked
response or the quantal content after temperature- (EPSP amplitude = 39.1 ± 1.6 mV com-
pared to 37.6 ± 1.4 mV in control; p = 0.72; and QC = 43.4 ± 2.1 compared to QC = 45.8 ± 2.1
in control; p = 0.56) or TRPA-driven stimulation (EPSP amplitude = 40.2 ± 1.4 mV; p = 0.49;
and QC = 48.7 ± 2.1; p = 0.56; Fig 5E, 5G and 5H).
Discussion
The Drosophila NMJ is a major model for studying basic phenomena underlying synaptic
growth and function. Because Drosophila research has access to numerous genetics strategies,
many experimental avenues exist for eliciting activity-dependent synaptic plasticity and assess-
ing its mechanisms. Here we focused on three different strategies to depolarize experimental
preparations: the addition of a potassium-rich depolarizing solution, a transgenically-encoded
depolarizing light-sensitive cation channel (CsChrimson), and a transgenically-encoded depo-
larizing warmth-sensitive cation channel (TRPA1). In addition, we contrasted the use of con-
tinuous stimulation with patterned stimulation protocols. Continuous activation of neurons
has been achieved successfully with optogenetics. In the adult fly central nervous system, opto-
genetics using the red-activatable Channelrhodopsin ReaChR showed spiking activity decays
during continuous light stimulation [56]. At the larval NMJ, optogenetics has been used for
acute and chronic activation of the Channelrhodopsin variant ChR2-XXL with blue light
pulses that ranged from 10 seconds to 1 hour of constant light exposure [67]. The same is true
for TRPA [33]. Ectopic expression of TRPA channels in R8 photoreceptor cells of adult flies
allowed for the persistent activation of photoreceptors that extended for days [35]. An earlier
study expressed TRPA channels in circadian neurons of adult flies and achieved continuous
activation of these neurons by exposing flies to 27˚C [68]. At the NMJ, continuous TRPA stim-
ulation for 1 hour can provoke morphological changes at muscle 4 when driving UAS-TRPA1
construct with the VGlut/OK371gal4 [69] motoneuron driver [30]. But the search for a stimu-
lus that better resembles the physiological situation has resulted in labs developing patterned
stimulation protocols, which alternate periods of activity with periods of rest [23]. Indeed, the
patterned stimulation protocols used at the Drosophila NMJ are similar to the protocols of
spaced depolarizations used to promote structural plasticity in dendritic spines of hippocampal
neurons in culture [42]. In the present study we also allowed the preparations different rest
durations; we examined them 15 or 45 minutes after the last stimulation to assure that these
PLOS ONE | https://doi.org/10.1371/journal.pone.0260553 November 30, 2021
15 / 22
PLOS ONE Three methods of eliciting rapid activity-dependent synaptic plasticity at the Drosophila NMJ
modifications were lasting effects, a condition sine qua non of activity-dependent synaptic
plasticity [6].
The stimulation settings used in this work are very diverse and might correspond to differ-
ent experimental needs. The ease of use is an important factor when considering which tech-
nique to select. The potassium rich protocol complies fully with this parameter. Because it
does not use transgenes, it is very accessible and very attractive to laboratories working with
undergraduate student scientists. Nonetheless, in our experience, a successful High K+ stimula-
tion protocol is achieved only after significant training of the researcher. In addition, many
experimental questions will require genetic backgrounds containing specific alleles and/or
expressing transgenes. The addition of more transgenes solely designed to depolarize the prep-
aration can be challenging even to the experienced geneticist. Nevertheless, our present results
indicate that depolarizing the preparations using the potassium-rich protocol limits drastically
any electrophysiological work attempting to measure mEPSP or EPSP amplitudes. Indeed,
after this treatment, the preparations show smaller input resistance and depolarized resting
potential. Because we did not observe similar events with the stimulating protocols using tem-
perature, temperature driven TRPA stimulation or optogenetics, we strongly favor the hypoth-
esis positing that K+ stimulation has a deleterious effect on the muscle. Nevertheless, it is
possible that the changes in membrane properties could also be part of the plasticity and/or a
compensation to this plasticity. Indeed, after heat treatment [70] or during repetitive synaptic
activity (train of electrical stimulation at 20Hz; [71]), a decrease in input resistance has been
observed. In addition, at high temperature rearing, synaptic homeostasis takes place to main-
tain a normal EJP within a terminal that contains increased release sites [72]. This is achieved
by decreasing quantal size through a decrease of muscle input resistance. In any case, the
changes in muscle input resistance and resting potential do not pose a direct limitation for
assessing differences in mEPSP frequency, although it may result in an underestimation caused
by missing smaller events during quantification. Another way to minimize the effect on input
resistance and membrane potential might be the use of direct electrical stimulation [23, 29]. It
is still an invasive preparation, but it utilizes a much more physiological stimulation paradigm.
Another parameter to take into consideration is the potency/scale of the response. Indeed,
not all the methods seem to provoke the same number of morphological changes. A patterned
stimulation using the potassium-rich depolarizing solution appears to provoke the most
important synaptic remodeling (7.9 boutons per synapse) while a more modest response was
observed with optogenetics (4.7 boutons per synapse). The temperature-driven response and
the temperature-triggered TRPA also provoked comparable activity-dependent synaptic plas-
ticity morphological changes (4.4 boutons per synapse and 7.6 boutons per synapse,
respectively).
In addition to the robustness of the response, the specificity of the stimulus should also be
considered. Using optogenetics to produce motoneuron-only patterned stimulation seems to
be the most specific manner to elicit activity-dependent synaptic plasticity. Indeed, our experi-
ments show that control preparations do not show morphological changes induced by the cul-
ture conditions (raised in the dark in presence of all-trans retinal containing food) nor the
stimulation protocol (patterned flashes of red light). We can conclude that the synaptic
changes that we observed are only due to the depolarization of the motoneuron and the resul-
tant presynaptic neurotransmitter release at the NMJ. Surprisingly, the efficacy of the potas-
sium depolarization protocol also shows a requirement for presynaptic stimulation. The
requirement for presynaptic release to elicit morphological changes after repeated stimulation
has been established [21–23, 27]. When presynaptic release is compromised by perturbing
action potential formation (by using a Na+ channel mutant; parats) or presynaptic vesicle
fusion (by using synaptotagmin 1 mutants; a fast mediator of neurotransmitter vesicle release
PLOS ONE | https://doi.org/10.1371/journal.pone.0260553 November 30, 2021
16 / 22
PLOS ONE Three methods of eliciting rapid activity-dependent synaptic plasticity at the Drosophila NMJ
and presynaptic Ca2+ sensor), ghost bouton formation after repeated potassium-driven stimu-
lation is prevented. This suggests that the source of depolarization that elicits these persistent
morphological changes is presynaptic. Nevertheless, it is also known that such potassium
shocks can depolarize the terminal directly both pre-and postsynaptically [73]. Such a general-
ized effect might apply to the temperature-evoked stimulation. Indeed, we showed that tem-
perature alone can elicit morphological and electrophysiological changes. It is, at this stage,
not possible to assess whether this phenomenon is due to an effect of the temperature on neu-
rons or muscles or both. We also showed that driving TRPA opening in neurons can increase
the morphological changes at the synapse when compared to temperature only stimulation,
suggesting that presynaptic only stimulation can provoke an additional effect on morphologi-
cal changes. It is possible that increased locomotory activity within the experimental setup
(thermocycler, see Materials and methods) accounts for some of the effects described here.
Rapid temperature changes increase nociceptive rolling behavior of larva [74] and increased
locomotion can elicit morphological changes at the NMJ [29, 74, 75]. In contrast, we show that
driving TRPA opening in motoneurons does not further increase activity-dependent potentia-
tion of spontaneous neurotransmission when compared to temperature only stimulation, sug-
gesting temperature alone accounts for all electrophysiological changes described.
This demonstration of the effects of temperature on a nerve terminal is quite remarkable
and the first such characterization at the Drosophila NMJ. It is a consistent result that we can
observe with a small temperature increase (4˚C difference, 23˚C to 27˚C applied in a patterned
manner). A number of studies have pointed out the consequences of temperature on the ner-
vous system [76, 77]. For example, numerous studies provide evidence that the properties of
neurotransmission vary depending on the temperature of the synapse. Interestingly, within
the mammalian brain, each brain structure has its own basal temperature, and the subtle dif-
ferences in temperature persist even when the environmental conditions impose drastic
changes in absolute temperatures [76]. In vivo experiments on the mammalian neocortex dem-
onstrate that neuronal activity changes in response to temperature. When brain temperature is
decreased, pyramidal neurons from layer 2/3 of the neocortex are depolarized and their input
resistance increases [78]. This was also found in vitro using acute slices [79]. Using rat hippo-
campal slices, the effect of temperature in evoked neurotransmission was shown to modify the
presynaptic compartment by affecting the amount of vesicles released [80]. At the calyx of
Held, temperature is also capable of modifying the dynamics of exocytosis [81]. Interestingly,
and relevant to our study, acute temperature shifts can modulate short term synaptic plasticity
properties in hippocampal cell cultures. In vitro experiments on rat hippocampal synapses
demonstrated that temperature affects the properties of short-term plasticity [82]. In CA1
pyramidal neurons, constant trains of stimulation at a frequency of 40 Hz at 23˚C vs 33˚C
showed that temperature promotes changes in evoked field EPSPs whereby synapses display
depression at 23˚C and potentiation over 33˚C [82].
In this study we have presented different techniques for eliciting activity-dependent synap-
tic plasticity and described their different characteristics. While a subset allows meaningful
electrophysiological assessment, all of them show morphological modifications. It has been
previously shown that the ghost boutons could be heterogenous [29]. Indeed, intense activity
promotes the rapid appearance of new synaptic boutons, some filled with synaptic vesicles pos-
sibly capable of exocytosis/endocytosis while other ghost boutons lack synaptic vesicles but
contain filamentous matrix and membrane folds. Interestingly, both types of ghost boutons
persist and remain unchanged for at least 60 minutes of rest after the last stimulation [29]. It
was previously reported that some ghost boutons are evident within seconds of the first cycle
of stimulation [21], suggesting some ghost boutons appear from primed synapses ready to
respond to increased activity while others appear later, only after subsequent pulses of
PLOS ONE | https://doi.org/10.1371/journal.pone.0260553 November 30, 2021
17 / 22
PLOS ONE Three methods of eliciting rapid activity-dependent synaptic plasticity at the Drosophila NMJ
patterned depolarizations. Others have described that lasting structural changes appear after
the 4th cycle of repeated stimulation [23]. It remains to be determined whether the different
stimulation techniques described here evoke homogeneous or heterogeneous sets of activity-
dependent changes.
Supporting information
S1 Table. Electrophysiological recordings of the muscle input resistance (IR) and resting
potential (RP) in unstimulated controls and K+ stimulated preparations. The 16 control
preparations have an IR � 5 MΩ and RP � -60 mv. In contrast, 14 stimulated preparations
(marked with grey background) out of 22 (64%) failed to meet these standards.
(TIF)
S1 Data.
(XLSX)
Acknowledgments
We thank Dr. Jonathan Blagburn for his valuable comments on previous versions of this man-
uscript. Thanks to Dr. Alice Robie, Dr. Kristin Branson, Dr. Andrew Seeds and Luis Haddock
for their help with the optogenetics setup. Parts of our schematic diagrams (Figs 1A, 2A, 3A
and 4A) were created with Biorender.com.
Author Contributions
Conceptualization: Carolina Maldonado-Dı´az, Bruno Marie.
Formal analysis: Carolina Maldonado-Dı´az, Bruno Marie.
Funding acquisition: Bruno Marie.
Investigation: Carolina Maldonado-Dı´az, Mariam Vazquez.
Methodology: Carolina Maldonado-Dı´az.
Project administration: Carolina Maldonado-Dı´az, Bruno Marie.
Resources: Carolina Maldonado-Dı´az.
Supervision: Bruno Marie.
Validation: Carolina Maldonado-Dı´az.
Visualization: Carolina Maldonado-Dı´az, Bruno Marie.
Writing – original draft: Bruno Marie.
Writing – review & editing: Carolina Maldonado-Dı´az, Bruno Marie.
References
1.
Takeuchi T, Duszkiewicz AJ, Morris RGM. The synaptic plasticity and memory hypothesis: encoding,
storage and persistence. Philosophical transactions of the Royal Society of London Series B, Biological
sciences. 2014; 369: 20130288. https://doi.org/10.1098/rstb.2013.0288 PMID: 24298167
2. Kukushkin NV, Carew TJ. Memory Takes Time. Neuron. 2017; 95: 259–279. https://doi.org/10.1016/j.
neuron.2017.05.029 PMID: 28728021
3. Humeau Y, Choquet D. The next generation of approaches to investigate the link between synaptic
plasticity and learning. Nature neuroscience. 2019; 22: 1536–1543. https://doi.org/10.1038/s41593-
019-0480-6 PMID: 31477899
PLOS ONE | https://doi.org/10.1371/journal.pone.0260553 November 30, 2021
18 / 22
PLOS ONE Three methods of eliciting rapid activity-dependent synaptic plasticity at the Drosophila NMJ
4. Stuchlik A. Dynamic learning and memory, synaptic plasticity and neurogenesis: an update. Frontiers in
Behavioral Neuroscience. 2014; 8: 106. https://doi.org/10.3389/fnbeh.2014.00106 PMID: 24744707
5. Neves G, Cooke SF, Bliss TVP. Synaptic plasticity, memory and the hippocampus: a neural network
approach to causality. Nature reviews Neuroscience. 2008; 9: 65–75. https://doi.org/10.1038/nrn2303
PMID: 18094707
6. Amtul Z, Atta-Ur-Rahman A-U-R. Neural plasticity and memory: molecular mechanism. Reviews in the
neurosciences. 2015; 26: 253–68. https://doi.org/10.1515/revneuro-2014-0075 PMID: 25995328
7. Ho VM, Lee J-A, Martin KC. The cell biology of synaptic plasticity. Science (New York, NY). 2011; 334:
623–8. https://doi.org/10.1126/science.1209236 PMID: 22053042
8. Pan Y, Monje M. Activity Shapes Neural Circuit Form and Function: A Historical Perspective. The Jour-
nal of neuroscience: the official journal of the Society for Neuroscience. 2020; 40: 944–954. https://doi.
org/10.1523/JNEUROSCI.0740-19.2019 PMID: 31996470
9. Kandel ER, Dudai Y, Mayford MR. The molecular and systems biology of memory. Cell. 2014; 157:
163–86. https://doi.org/10.1016/j.cell.2014.03.001 PMID: 24679534
10. Cheyne JE, Montgomery JM. The cellular and molecular basis of in vivo synaptic plasticity in rodents.
American Journal of Physiology-Cell Physiology. 2020; 318: C1264–C1283. https://doi.org/10.1152/
ajpcell.00416.2019 PMID: 32320288
11. Wayman GA, Impey S, Marks D, Saneyoshi T, Grant WF, Derkach V, et al. Activity-dependent dendritic
arborization mediated by CaM-kinase I activation and enhanced CREB-dependent transcription of Wnt-
2. Neuron. 2006; 50: 897–909. https://doi.org/10.1016/j.neuron.2006.05.008 PMID: 16772171
12. Avila ME, Sepu´lveda FJ, Burgos CF, Moraga-Cid G, Parodi J, Moon RT, et al. Canonical Wnt3a modu-
lates intracellular calcium and enhances excitatory neurotransmission in hippocampal neurons. The
Journal of biological chemistry. 2010; 285: 18939–47. https://doi.org/10.1074/jbc.M110.103028 PMID:
20404321
13. Padamsey Z, McGuinness L, Bardo SJ, Reinhart M, Tong R, Hedegaard A, et al. Activity-Dependent
Exocytosis of Lysosomes Regulates the Structural Plasticity of Dendritic Spines. Neuron. 2017; 93:
132–146. https://doi.org/10.1016/j.neuron.2016.11.013 PMID: 27989455
14. Nakahata Y, Yasuda R. Plasticity of Spine Structure: Local Signaling, Translation and Cytoskeletal
Reorganization. Frontiers in Synaptic Neuroscience. 2018; 10: 29. https://doi.org/10.3389/fnsyn.2018.
00029 PMID: 30210329
15. McLeod F, Bossio A, Marzo A, Ciani L, Sibilla S, Hannan S, et al. Wnt Signaling Mediates LTP-Depen-
dent Spine Plasticity and AMPAR Localization through Frizzled-7 Receptors. Cell Reports. 2018; 23:
1060–1071. https://doi.org/10.1016/j.celrep.2018.03.119 PMID: 29694885
16. Hawk JD, Calvo AC, Liu P, Almoril-Porras A, Aljobeh A, Torruella-Sua´rez ML, et al. Integration of Plas-
ticity Mechanisms within a Single Sensory Neuron of C. elegans Actuates a Memory. Neuron. 2018; 97:
356–367.e4. https://doi.org/10.1016/j.neuron.2017.12.027 PMID: 29307713
17.
18.
Jensen M, Hoerndli FJ, Brockie PJ, Wang R, Johnson E, Maxfield D, et al. Wnt Signaling Regulates
Acetylcholine Receptor Translocation and Synaptic Plasticity in the Adult Nervous System. Cell. 2012;
149: 173–187. https://doi.org/10.1016/j.cell.2011.12.038 PMID: 22464329
Jin Y, Qi YB. Building stereotypic connectivity: mechanistic insights into structural plasticity from C. ele-
gans. Current Opinion in Neurobiology. 2018; 48: 97–105. https://doi.org/10.1016/j.conb.2017.11.005
PMID: 29182952
19. Sasakura H, Mori I. Behavioral plasticity, learning, and memory in C. elegans. Current Opinion in Neuro-
biology. 2013; 23: 92–99. https://doi.org/10.1016/j.conb.2012.09.005 PMID: 23063296
20. Bai Y, Suzuki T. Activity-Dependent Synaptic Plasticity in Drosophila melanogaster. Frontiers in Physi-
ology. 2020; 11: 161. https://doi.org/10.3389/fphys.2020.00161 PMID: 32158405
21. Piccioli ZD, Littleton JT. Retrograde BMP Signaling Modulates Rapid Activity-Dependent Synaptic
Growth via Presynaptic LIM Kinase Regulation of Cofilin. The Journal of Neuroscience. 2014; 34: 4371.
https://doi.org/10.1523/JNEUROSCI.4943-13.2014 PMID: 24647957
22. Vasin A, Zueva L, Torrez C, Volfson D, Littleton JT, Bykhovskaia M. Synapsin regulates activity-depen-
dent outgrowth of synaptic boutons at the Drosophila neuromuscular junction. The Journal of neurosci-
ence: the official journal of the Society for Neuroscience. 2014; 34: 10554–63. https://doi.org/10.1523/
JNEUROSCI.5074-13.2014 PMID: 25100589
23. Ataman B, Ashley J, Gorczyca M, Ramachandran P, Fouquet W, Sigrist SJ, et al. Rapid activity-depen-
dent modifications in synaptic structure and function require bidirectional Wnt signaling. Neuron. 2008;
57: 705–18. https://doi.org/10.1016/j.neuron.2008.01.026 PMID: 18341991
24. Nesler KR, Starke EL, Boin NG, Ritz M, Barbee SA. Presynaptic CamKII regulates activity-dependent
axon terminal growth. Mol Cell Neurosci. 2016; 76: 33–41. https://doi.org/10.1016/j.mcn.2016.08.007
PMID: 27567686
PLOS ONE | https://doi.org/10.1371/journal.pone.0260553 November 30, 2021
19 / 22
PLOS ONE Three methods of eliciting rapid activity-dependent synaptic plasticity at the Drosophila NMJ
25. Nesler KR, Sand RI, Symmes BA, Pradhan SJ, Boin NG, Laun AE, et al. The miRNA pathway controls
rapid changes in activity-dependent synaptic structure at the Drosophila melanogaster neuromuscular
junction. PloS one. 2013; 8: e68385. https://doi.org/10.1371/journal.pone.0068385 PMID: 23844193
26. Kuklin EA, Alkins S, Baskar Bakthavachalu X, Genco MC, Sudhakaran I, Raghavan XKV, et al. The
Long 3UTR mRNA of CaMKII Is Essential for Translation-Dependent Plasticity of Spontaneous Release
in Drosophila melanogaster. 2017. https://doi.org/10.1523/JNEUROSCI.1313-17.2017 PMID:
28954869
27. Alicea D, Perez M, Maldonado C, Dominicci-Cotto C, Marie B. Cortactin Is a Regulator of Activity-
Dependent Synaptic Plasticity Controlled by Wingless. Journal of Neuroscience. 2017; 37. https://doi.
org/10.1523/JNEUROSCI.1375-16.2017 PMID: 28123080
28.
Freeman A, Franciscovich A, Bowers M, Sandstrom DJ, Sanyal S. NFAT regulates pre-synaptic devel-
opment and activity-dependent plasticity in Drosophila. Molecular and cellular neurosciences. 2011; 46:
535. https://doi.org/10.1016/j.mcn.2010.12.010 PMID: 21185939
29. Vasin A, Sabeva N, Torres C, Phan S, Bushong EA, Ellisman MH, et al. Two Pathways for the 772 Activ-
ity-Dependent Growth and Differentiation of Synaptic Boutons in Drosophila. eNeuro. 2019; 773 6(4)
30. Dear ML, Shilts J, Broadie K. Neuronal activity drives FMRP- and HSPG-dependent matrix metallopro-
teinase function required for rapid synaptogenesis. Science Signaling. 2017; 10: 36–39. https://doi.org/
10.1126/scisignal.aan3181 PMID: 29114039
31. Metwally E, Zhao G, Li W, Wang Q, Zhang YQ. Calcium-Activated Calpain Specifically Cleaves Gluta-
mate Receptor IIA But Not IIB at the Drosophila Neuromuscular Junction. The Journal of neuroscience:
the official journal of the Society for Neuroscience. 2019; 39: 2776–2791. https://doi.org/10.1523/
JNEUROSCI.2213-17.2019 PMID: 30705102
32.
Ljaschenko D, Ehmann N, Kittel RJ. Hebbian Plasticity Guides Maturation of Glutamate Receptor Fields
In Vivo. Cell Reports. 2013; 3: 1407–1413. https://doi.org/10.1016/j.celrep.2013.04.003 PMID: 23643532
33. Owald D, Lin S, Waddell S. Light, heat, action: neural control of fruit fly behaviour. Philosophical Trans-
actions of the Royal Society B: Biological Sciences. 2015. https://doi.org/10.1098/rstb.2014.0211
PMID: 26240426
34. Hamada FN, Rosenzweig M, Kang K, Pulver SR, Ghezzi A, Jegla TJ, et al. An internal thermal sensor
controlling temperature preference in Drosophila. Nature. 2008; 454: 217–20. https://doi.org/10.1038/
nature07001 PMID: 18548007
35. Sugie A, Hakeda-Suzuki S, Suzuki E, Silies M, Shimozono M, Mo¨ hl C, et al. Molecular Remodeling of
the Presynaptic Active Zone of Drosophila Photoreceptors via Activity-Dependent Feedback. Neuron.
2015; 86: 711–25. https://doi.org/10.1016/j.neuron.2015.03.046 PMID: 25892303
36. Berni J, Muldal AM, Pulver SR. Using Neurogenetics and the Warmth-Gated Ion Channel TRPA1 to
Study the Neural Basis of Behavior in Drosophila. Journal of undergraduate neuroscience education:
JUNE: a publication of FUN, Faculty for Undergraduate Neuroscience. 2010; 9: A5–A14.
37. Brand AH, Perrimon N. Targeted gene expression as a means of altering cell fates and generating dom-
inant phenotypes. Development (Cambridge, England). 1993; 118: 401–15. PMID: 8223268
38. Klapoetke NC, Murata Y, Kim SS, Pulver SR, Birdsey-Benson A, Cho YK, et al. Independent optical
excitation of distinct neural populations. Nature methods. 2014; 11: 338–46. https://doi.org/10.1038/
nmeth.2836 PMID: 24509633
39. Aso Y, Sitaraman D, Ichinose T, Kaun KR, Vogt K, Belliart-Gue´ rin G, et al. Mushroom body output neu-
rons encode valence and guide memory-based action selection in Drosophila. eLife. 2014; 3: e04580.
https://doi.org/10.7554/eLife.04580 PMID: 25535794
40. Carrillo RA, Olsen DP, Yoon KS, Keshishian H. Presynaptic activity and CaMKII modulate retrograde
semaphorin signaling and synaptic refinement. Neuron. 2010; 68: 32–44. https://doi.org/10.1016/j.
neuron.2010.09.005 PMID: 20920789
41. Vonhoff F, Keshishian H. In Vivo Calcium Signaling during Synaptic Refinement at the Drosophila Neu-
romuscular Junction. The Journal of Neuroscience. 2017; 37: 5511–5526. https://doi.org/10.1523/
JNEUROSCI.2922-16.2017 PMID: 28476946
42. Wu G-Y, Deisseroth K, Tsien RW. Spaced stimuli stabilize MAPK pathway activation and its effects on
dendritic morphology. Nature Neuroscience. 2001; 4: 151–158. https://doi.org/10.1038/83976 PMID:
11175875
43. Yao J, Qi J, Chen G. Actin-Dependent Activation of Presynaptic Silent Synapses Contributes to Long-
Term Synaptic Plasticity in Developing Hippocampal Neurons. The Journal of Neuroscience. 2006.
https://doi.org/10.1523/JNEUROSCI.1183-06.2006 PMID: 16885227
44.
Li Z, Okamoto K-I, Hayashi Y, Sheng M. The Importance of Dendritic Mitochondria in the Morphogene-
sis and Plasticity of Spines and Synapses. Cell. 2004; 119: 873–887. https://doi.org/10.1016/j.cell.
2004.11.003 PMID: 15607982
PLOS ONE | https://doi.org/10.1371/journal.pone.0260553 November 30, 2021
20 / 22
PLOS ONE Three methods of eliciting rapid activity-dependent synaptic plasticity at the Drosophila NMJ
45. Cai Q, Pan P-Y, Sheng Z-H. Syntabulin-Kinesin-1 Family Member 5B-Mediated Axonal Transport Con-
tributes to Activity-Dependent Presynaptic Assembly. 2007. https://doi.org/10.1523/JNEUROSCI.0731-
07.2007 PMID: 17611281
46. Hu X, Viesselmann C, Nam S, Merriam E, Dent EW. Activity-Dependent Dynamic Microtubule
Invasion of Dendritic Spines. 2008. https://doi.org/10.1523/JNEUROSCI.3074-08.2008 PMID:
19052200
47. San Martin A, Rela L, Gelb B, Rafael PM. The Spacing Effect for Structural Synaptic Plasticity Provides
Specificity and Precision in Plastic Changes. The Journal of Neuroscience. 2017. https://doi.org/10.
1523/JNEUROSCI.2607-16.2017 PMID: 28432141
48. Steinert JR, Kuromi H, Hellwig A, Knirr M, Wyatt AW, Kidokoro Y, et al. Experience-dependent forma-
tion and recruitment of large vesicles from reserve pool. Neuron. 2006; 50: 723–33. https://doi.org/10.
1016/j.neuron.2006.04.025 PMID: 16731511
49. Katz B, Thesleff S. KATZ and THESLEFF, 1957. J Physiol. 1957; I37: 267–278.
50.
51.
FALK G, FATT P. LINEAR ELECTRICAL PROPERTIES OF STRIATED MUSCLE FIBRES
OBSERVED WITH. Proceedings of the Royal Society of London Series B, Containing papers. 1964;
160: 69–123. https://doi.org/10.1098/rspb.1964.0030 PMID: 14142170
Lnenicka GA, Mellon D. CHANGES IN ELECTRICAL PROPERTIES AND QUANTAL CURRENT DUR-
ING GROWTH OF IDENTIFIED MUSCLE FIBRES IN THE CRAYFISH. J Physiol. 1983. https://doi.
org/10.1113/jphysiol.1983.sp014977 PMID: 6663501
52. Stewart BA, Atwood HL, Renger JJ, Wang J, Wu CF. Improved stability of Drosophila larval neuromus-
cular preparations in haemolymph-like physiological solutions. Journal of comparative physiology A,
Sensory, neural, and behavioral physiology. 1994; 175: 179–91. https://doi.org/10.1007/BF00215114
PMID: 8071894
53.
Feng Y, UEDA A, WU C-F. A modified minimal hemolymph-like solution, HL3.1, for physiological
recordings at the neuromuscular junctions of normal and mutant Drosophila larvae. Journal of Neuroge-
netics. 2004. https://doi.org/10.1080/01677060490894522 PMID: 15763995
54. Macleod GT, Hegstro¨ m-Wojtowicz M, Charlton MP, Atwood HL. Fast Calcium Signals in Drosophila
Motor Neuron Terminals. Journal of Neurophysiology. 2002; 88: 2659–2663. https://doi.org/10.1152/jn.
00515.2002 PMID: 12424301
55. Sanyal S. Genomic mapping and expression patterns of C380, OK6 and D42 enhancer trap lines in the
larval nervous system of Drosophila. Gene expression patterns: GEP. 2009; 9: 371–80. https://doi.org/
10.1016/j.gep.2009.01.002 PMID: 19602393
56.
Inagaki HK, Jung Y, Hoopfer ED, Wong AM, Mishra N, Lin JY, et al. Optogenetic control of Drosophila
using a red-shifted channelrhodopsin reveals experience-dependent influences on courtship. Nature
methods. 2014; 11: 325–32. https://doi.org/10.1038/nmeth.2765 PMID: 24363022
57. Simpson JH, Looger LL. Functional Imaging and Optogenetics in Drosophila. Genetics. 2018; 208:
1291–1309. https://doi.org/10.1534/genetics.117.300228 PMID: 29618589
58. Oswald M, Brooks PS, Zwart MF, Mukherjee A, West RJ, Ng Giachello C, et al. Reactive oxygen spe-
cies regulate activity-dependent neuronal plasticity in Drosophila. eLife. 2018. https://doi.org/10.7554/
eLife.39393 PMID: 30540251
59. Pulver SR, Pashkovski SL, Hornstein NJ, Garrity PA, Griffith LC. Temporal Dynamics of Neuronal Acti-
vation by Channelrhodopsin-2 and TRPA1 Determine Behavioral Output in Drosophila Larvae. J Neuro-
physiol. 2009; 101: 3075–3088. https://doi.org/10.1152/jn.00071.2009 PMID: 19339465
60. Abram PK, Boivin G, Moiroux J, Brodeur J. Behavioural effects of temperature on ectothermic animals:
unifying thermal physiology and behavioural plasticity. Biological Reviews. 2017; 92: 1859–1876.
https://doi.org/10.1111/brv.12312 PMID: 28980433
61. Boothroyd CE, Wijnen H, Naef F, Saez L, Young MW. Integration of Light and Temperature in the Regu-
lation of Circadian Gene Expression in Drosophila. PLoS Genetics. 2007; 3. https://doi.org/10.1371/
JOURNAL.PGEN.0030054 PMID: 17411344
62. Yablonovitch AL, Fu J, Li K, Mahato S, Kang L, Rashkovetsky E, et al. Regulation of gene expression
and RNA editing in Drosophila adapting to divergent microclimates. Nature Communications. 2017; 8:
1570. https://doi.org/10.1038/s41467-017-01658-2 PMID: 29146998
63. Rieder LE, Savva YA, Reyna MA, Chang Y-J, Dorsky JS, Rezaei A, et al. Dynamic response of RNA
editing to temperature in Drosophila. BMC Biology. 2015; 13: 1. https://doi.org/10.1186/s12915-014-
0111-3 PMID: 25555396
64. Buchumenski I, Bartok O, Ashwal-Fluss R, Pandey V, Porath HT, Levanon EY, et al. Dynamic hyper-
editing underlies temperature adaptation in Drosophila. PLoS Genetics. 2017; 13. https://doi.org/10.
1371/journal.pgen.1006931 PMID: 28746393
PLOS ONE | https://doi.org/10.1371/journal.pone.0260553 November 30, 2021
21 / 22
PLOS ONE Three methods of eliciting rapid activity-dependent synaptic plasticity at the Drosophila NMJ
65. Peng I-F, Berke BA, Zhu Y, Lee W-H, Chen W, Wu C-F. Temperature-Dependent Developmental Plas-
ticity of Drosophila Neurons: Cell-Autonomous Roles of Membrane Excitability, Ca 2 Influx, and cAMP
Signaling. 2007. https://doi.org/10.1523/JNEUROSCI.2179-07.2007 PMID: 18003840
66.
Frolov R V, Singh S. Temperature and functional plasticity of L-type Ca2+ channels in Drosophila. Cell
calcium. 2013; 54. https://doi.org/10.1016/j.ceca.2013.07.005 PMID: 23993048
67. Higgins J, Hermanns C, Malloy C, Cooper RL. Considerations in repetitive activation of light sensitive
ion channels for long-term studies: Channel rhodopsin in the Drosophila model. Neuroscience
Research. 2017; 125: 1–10. https://doi.org/10.1016/j.neures.2017.07.001 PMID: 28728913
68. Parisky KM, Agosto J, Pulver SR, Shang Y, Kuklin E, Hodge JJL, et al. PDF cells are a GABA-respon-
sive wake-promoting component of the Drosophila sleep circuit. Neuron. 2008; 60: 672–82. https://doi.
org/10.1016/j.neuron.2008.10.042 PMID: 19038223
69. Mahr A, Aberle H. The expression pattern of the Drosophila vesicular glutamate transporter: a marker
protein for motoneurons and glutamatergic centers in the brain. Gene expression patterns: GEP. 2006;
6: 299–309. https://doi.org/10.1016/j.modgep.2005.07.006 PMID: 16378756
70. Barclay JW, Robertson RM. Role for calcium in heat shock-mediated synaptic thermoprotection in Dro-
sophila larvae. Journal of Neurobiology. 2003; 56: 360–371. https://doi.org/10.1002/neu.10247 PMID:
12918020
71. Gertner DM, Desai S, Lnenicka GA. Synaptic excitation is regulated by the postsynaptic dSK channel at
the Drosophila larval NMJ. J Neurophysiol. 2014; 111: 2533–2543. https://doi.org/10.1152/jn.00903.
2013 PMID: 24671529
72. Ueda A, Wu CF. The role of cAMP in synaptic homeostasis in response to environmental temperature
challenges and hyperexcitability mutations. Frontiers in Cellular Neuroscience. 2015; 9. https://doi.org/
10.3389/fncel.2015.00010 PMID: 25698925
73. Kopke DL, Broadie K. FM Dye Cycling at the Synapse: Comparing High Potassium Depolarization,
Electrical and Channelrhodopsin Stimulation. Journal of visualized experiments: JoVE. 2018. https://
doi.org/10.3791/57765 PMID: 29889207
74.
Luo J, Shen WL, Montell C. TRPA1 mediates sensation of the rate of temperature change in Drosophila
larvae. Nature neuroscience. 2017; 20: 34–41. https://doi.org/10.1038/nn.4416 PMID: 27749829
75. Sigrist SJ, Reiff DF, Thiel PR, Steinert JR, Schuster CM. Experience-dependent strengthening of Dro-
sophila neuromuscular junctions. The Journal of neuroscience: the official journal of the Society for Neu-
roscience. 2003; 23: 6546–56. https://doi.org/10.1523/JNEUROSCI.23-16-06546.2003 PMID:
12878696
76. Kiyatkin EA. Brain temperature homeostasis: physiological fluctuations and pathological shifts. Fron-
tiers in bioscience (Landmark edition). 2010; 15: 73–92. https://doi.org/10.2741/3608 PMID: 20036808
77. Wang H, Wang B, Normoyle KP, Jackson K, Spitler K, Sharrock MF, et al. Brain temperature and its
fundamental properties: a review for clinical neuroscientists. Frontiers in Neuroscience. 2014; 8: 307.
https://doi.org/10.3389/fnins.2014.00307 PMID: 25339859
78. Kalmbach AS, Waters J. Brain surface temperature under a craniotomy. Journal of Neurophysiology.
2012; 108: 3138–3146. https://doi.org/10.1152/jn.00557.2012 PMID: 22972953
79. Hedrick T, Waters J. Spiking Patterns of Neocortical L5 Pyramidal Neurons in Vitro Change with Tem-
perature. Frontiers in Cellular Neuroscience. 2011; 5: 1. https://doi.org/10.3389/fncel.2011.00001
PMID: 21286222
80. Yang X-F, Ouyang Y, Kennedy BR, Rothman SM. Cooling blocks rat hippocampal neurotransmission
by a presynaptic mechanism: observations using 2-photon microscopy. The Journal of Physiology.
2005; 567: 215–224. https://doi.org/10.1113/jphysiol.2005.088948 PMID: 15961429
81. Kushmerick C, Renden R, von Gersdorff H. Physiological temperatures reduce the rate of vesicle pool
depletion and short-term depression via an acceleration of vesicle recruitment. The Journal of neurosci-
ence: the official journal of the Society for Neuroscience. 2006; 26: 1366–77. https://doi.org/10.1523/
JNEUROSCI.3889-05.2006 PMID: 16452660
82. Klyachko VA, Stevens CF. Temperature-dependent shift of balance among the components of short-
term plasticity in hippocampal synapses. The Journal of neuroscience: the official journal of the Society
for Neuroscience. 2006; 26: 6945–57. https://doi.org/10.1523/JNEUROSCI.1382-06.2006 PMID:
16807324
PLOS ONE | https://doi.org/10.1371/journal.pone.0260553 November 30, 2021
22 / 22
PLOS ONE |
10.1371_journal.pone.0237574 | RESEARCH ARTICLE
Performance of diagnostic and predictive host
blood transcriptomic signatures for
Tuberculosis disease: A systematic review and
meta-analysis
Humphrey MulengaID
Kimbung Mbandi1, Simon C. MendelsohnID
Thomas J. Scriba1☯, Mark Hatherill1☯*
1, Chambrez-Zita Zauchenberger1, Erick W. Bunyasi1, Stanley
1, Benjamin Kagina2, Adam Penn-Nicholson1,
1 South African Tuberculosis Vaccine Initiative (SATVI), Institute of Infectious Disease & Molecular Medicine
and Division of Immunology, Department of Pathology, University of Cape Town, Cape Town, South Africa,
2 Vaccines for Africa Initiative (VACFA), School of Public Health & Family Medicine, University of Cape
Town, Cape Town, South Africa
☯ These authors contributed equally to this work.
* Mark.Hatherill@uct.ac.za
Abstract
Introduction
Host blood transcriptomic biomarkers have potential as rapid point-of-care triage, diagnos-
tic, and predictive tests for Tuberculosis disease. We aimed to summarise the performance
of host blood transcriptomic signatures for diagnosis of and prediction of progression to
Tuberculosis disease; and compare their performance to the recommended World Health
Organisation target product profile.
Methods
A systematic review and meta-analysis of the performance of host blood mRNA signatures
for diagnosing and predicting progression to Tuberculosis disease in HIV-negative adults
and adolescents, in studies with an independent validation cohort. Medline, Scopus, Web of
Science, and EBSCO libraries were searched for articles published between January 2005
and May 2019, complemented by a search of bibliographies. Study selection, data extrac-
tion and quality assessment were done independently by two reviewers. Meta-analysis was
performed for signatures that were validated in �3 comparable cohorts, using a bivariate
random effects model.
Results
Twenty studies evaluating 25 signatures for diagnosis of or prediction of progression to TB
disease in a total of 68 cohorts were included. Eighteen studies evaluated 24 signatures for
TB diagnosis and 17 signatures met at least one TPP minimum performance criterion.
Three diagnostic signatures were validated in clinically relevant cohorts to differentiate TB
from other diseases, with pooled sensitivity 84%, 87% and 90% and pooled specificity 79%,
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: Mulenga H, Zauchenberger C-Z, Bunyasi
EW, Mbandi SK, Mendelsohn SC, Kagina B, et al.
(2020) Performance of diagnostic and predictive
host blood transcriptomic signatures for
Tuberculosis disease: A systematic review and
meta-analysis. PLoS ONE 15(8): e0237574. https://
doi.org/10.1371/journal.pone.0237574
Editor: Sanjai Kumar, Food and Drug
Administration, UNITED STATES
Received: December 11, 2019
Accepted: July 30, 2020
Published: August 21, 2020
Copyright: © 2020 Mulenga et al. This is an open
access article distributed under the terms of the
Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: All relevant data are
within the manuscript and its Supporting
Information files.
Funding: The author(s) received no specific
funding for this work.
Competing interests: I have read the journal’s
policy and the authors of this manuscript have the
following competing interests: Thomas J. Scriba
and Adam Penn-Nicholson are inventors of blood
transcriptomic signatures of risk of TB. This does
PLOS ONE | https://doi.org/10.1371/journal.pone.0237574 August 21, 2020
1 / 17
PLOS ONE Performance of diagnostic and predictive host blood transcriptomic signatures for Tuberculosis disease
not alter our adherence to PLOS ONE policies on
sharing data and materials.
88% and 74%, respectively. Four studies evaluated signatures for progression to TB dis-
ease and performance of one signature, assessed within six months of TB diagnosis, met
the minimal TPP for a predictive test for progression to TB disease.
Conclusion
Host blood mRNA signatures hold promise as triage tests for TB. Further optimisation is
needed if mRNA signatures are to be used as standalone diagnostic or predictive tests for
therapeutic decision-making.
Introduction
The World Health Organisation (WHO) has targeted 2035 to end tuberculosis (TB) and aims
for 90% reduction in new TB cases and 95% reduction in TB deaths compared to 2015 levels
[1]. Non-sputum triage, diagnostic, and predictive tests for TB may play a role in advancing
TB control efforts. The WHO, in conjunction with the Foundation for Innovative New Diag-
nostics and the New Diagnostics Working Group of the Stop TB Partnership, has published
Target Product Profiles (TPPs) for non-sputum biomarker triage, diagnostic, and predictive
tests for progression from latent TB infection (LTBI) to TB disease [2–4]. The TPPs require
minimum 90% sensitivity and 70% specificity for a triage test; 65% sensitivity and 98% speci-
ficity for a diagnostic test [2], and 75% sensitivity and 75% specificity for a test to predict pro-
gression from LTBI to active TB disease within two years [3, 4]. A new predictive test should
also achieve a positive predictive value (PPV) of 5.8% given a 2% pre-test probability [3].
Current commercially available TB diagnostic tests are not optimal. Mycobacterium tuber-
culosis (MTB) culture, considered the gold standard, requires days to weeks to obtain a result
from a reference laboratory, and is thus not ideal for rapid patient management [5, 6].Sputum
smear microscopy has low sensitivity [7] ranging from 32% to 89% [8], resulting in a consider-
able proportion of active pulmonary TB patients being missed [9]. Sputum Xpert MTB/RIF
and Xpert MTB/RIF Ultra have considerably better diagnostic performance [10, 11] than
smear microscopy, but are similarly dependant on obtaining an adequate sputum sample; and
need specialised laboratory equipment and a reliable power supply, which impedes routine
screening in TB-endemic resource-limited settings [12].
Individuals with LTBI, defined by a positive tuberculin skin test (TST) or interferon-
gamma release assay (IGRA), have a higher risk of progression to TB disease than MTB-unin-
fected people [13, 14]. However, only about 10–15% of people who test IGRA or TST positive
will go on to develop TB disease [15, 16]. Predictive specificity of IGRA and TST for incident
TB disease is poor (49.3% and 45% respectively) [14, 17]. While prevention of TB disease aris-
ing from LTBI is key to achieving WHO elimination targets [18], mass preventive therapy
based on IGRA or TST screening in TB-endemic countries would need to treat a significant
proportion of the population [19], most of them unnecessarily, which would be unaffordable
and potentially ineffective, because re-infection would likely occur before programmatic cov-
erage was complete.
In recent years, host blood transcriptomic signatures have offered a promising alternative
as tests for both diagnosis of and prediction of progression to TB disease. These signatures
have also improved our understanding of inflammatory processes associated with progression
[20, 21] in individuals with MTB infection [22] and those with TB disease [23]. Host blood
transcriptomic signatures have been shown to discriminate prevalent TB disease cases from
PLOS ONE | https://doi.org/10.1371/journal.pone.0237574 August 21, 2020
2 / 17
PLOS ONE Performance of diagnostic and predictive host blood transcriptomic signatures for Tuberculosis disease
MTB-uninfected and latently MTB-infected individuals, and individuals with other respiratory
ailments [24]; and predict progression to TB disease in individuals with LTBI [25, 26].
Two systematic reviews have evaluated biomarkers for diagnosis of TB disease in children
[27] and all age groups [28]. These systematic reviews did not include biomarkers for progres-
sion to TB disease; and did not include a meta-analysis, owing to heterogenous study designs,
patient selection, and biomarker composition. Two additional studies performed re-analysis of
patient-level data but did not perform a systematic review of signature performance as
reported by the original studies. Warsinske et al [29] compared 16 signatures for TB diagnosis
by recreating the original model of each signature and evaluated each signature across the
datasets they had identified, and Gupta et al [30] conducted a meta-analysis of patient-level
pooled data for signatures of incipient TB. We present a systematic review and meta-analysis
of transcriptomic signatures which have been evaluated in independent validation cohorts. We
aimed to summarise the performance of host blood transcriptomic signatures for diagnosis of
and prediction of progression to TB disease in HIV-negative adults and adolescents; and to
compare individual signature performance to the WHO TPP. This review is registered with
the International Prospective Register of Systematic Reviews (PROSPERO), registration num-
ber CRD42017073817.
Materials and methods
We conducted a systematic review according to standard guidelines [31, 32] (S1 Table). and
designed the protocol prior to conducting the review [33] (http://dx.doi.org/10.1136/
bmjopen-2018-026612).
Study inclusion/exclusion criteria
We included studies evaluating host blood mRNA signatures for diagnosis of or prediction of
progression to TB disease in HIV-negative adults and/or adolescents (� 12 years old) and pub-
lished in English. Studies were restricted to those published between January 2005 and May
2019 to concentrate on recent evidence because of the fast pace at which the field of transcrip-
tomics is changing and advancing. Only studies comparing TB disease cases versus MTB-
uninfected controls, individuals with other diseases (ODs), or with LTBI; and using a microbi-
ological reference standard of either sputum MTB culture, Xpert MTB/RIF, or smear micros-
copy for TB disease diagnosis, were eligible for inclusion. Studies of prediction of incident TB
disease were required to be prospective, with a follow-up period of at least six months; and
enrolling either TB contacts, latently MTB-infected, or uninfected individuals. Studies were
excluded if: conducted in animals; were in children younger than 12 years, did not report sen-
sitivity and specificity, did not allow recreation of a 2 x 2 contingency table for calculation of
test performance, and where authors did not respond to enquiries for data within four weeks
of inquiry. Unpublished reports and conference proceedings were excluded due to absence of
peer review and difficulty in obtaining data.
Literature search
Medline via PubMed, Scopus, Web of Science, and EBSCO databases were searched for rele-
vant studies. The search strategy developed on PubMed was adapted to other databases and
was as follows:
((((((Tuberculosis [MeSH] OR Mycobacterium tuberculosis [MeSH] OR (Tuberculosis OR
TB OR Mycobacterium tuberculosis OR MTB))) AND ((Diagnosis [MeSH] OR Diagnosis
[subheading] OR Prognosis [MeSH] OR (Diagnosis OR diagnostic OR detect� OR predict�
OR prognosis OR prognostic OR screen�)))) AND ((Biomarkers/Blood [MeSH] OR RNA/
PLOS ONE | https://doi.org/10.1371/journal.pone.0237574 August 21, 2020
3 / 17
PLOS ONE Performance of diagnostic and predictive host blood transcriptomic signatures for Tuberculosis disease
Blood [MeSH] OR Transcription, Genetic [MeSH] /etiology/genetics/immunology OR (Blood
Biomarker OR blood biomarkers OR bio-signature OR gene expression OR genetic transcrip-
tion OR host blood OR immune marker OR immunologic marker OR Ribonucleic Acid OR
RNA OR signature OR surrogate endpoint OR surrogate marker OR transcriptome OR tran-
scriptomic)))) AND ((Area under Curve [MeSH] OR Sensitivity and Specificity [MeSH] OR
(Area under curve OR area under curves OR AUC OR receiver operating characteristic OR
ROC OR Accuracy OR Performance OR sensitivity OR specificity)))) AND (Humans[Mesh]))
AND (("2005/01/01"[Date—Publication]: "2019/05/31"[Date—Publication])). Additionally,
bibliographies of included papers were scrutinised for potential papers that were missed by the
search terms.
Study selection
Two reviewers (HM and CZZ) independently screened search outputs for eligible studies.
Publications were first screened by title and abstract, and thereafter by full text. Articles were
independently categorised as either (i) selected, (ii) not selected, or (iii) pending. The two
reviewers conferred to resolve any disagreements about pending publications, and if a consen-
sus could not be reached, discrepancies were adjudicated by a third reviewer (BK). Fig 1
depicts the study selection process.
Data extraction and management
Data metrics were extracted separately from the relevant articles and double-entered into a
Microsoft SQL Server 2012 database to facilitate electronic comparison of data between
reviewers. For each biomarker test, we extracted the reported performance data; sensitivity,
specificity, true positives (TP) false positives (FP), true negatives (TN), and false negatives
Fig 1. Flow of studies in the review of transcriptomic signatures for diagnosing and predicting progression to TB
disease. DTA; diagnostic test accuracy, LTBI; Latent TB infection, OD; Other diseases.
https://doi.org/10.1371/journal.pone.0237574.g001
PLOS ONE | https://doi.org/10.1371/journal.pone.0237574 August 21, 2020
4 / 17
PLOS ONE Performance of diagnostic and predictive host blood transcriptomic signatures for Tuberculosis disease
(FN); as well as details of the study design, population, and index and reference test character-
istics using a customised data extraction form (S1 File). We requested missing data from
authors by email.
Appraisal of methodological quality of studies
HM and CZZ independently assessed the quality of included studies using a customised form
based on the Quality Assessment of Diagnostic Accuracy Studies-2 (QUADAS-2) [34] assess-
ment tool (S2 File, Methods in S3 File). BMK provided adjudication where a disagreement
occurred. For each of the four domains in QUADAS-2, namely “patient selection”, “index
test”, “reference standard” and “flow and timing”; risk of bias was scored as “low” if all
responses in that domain were answered as “yes”, “high” if any of the responses were answered
as “no” or “unclear” and “unclear” if it was unclear for all the responses. We judged “low appli-
cability concerns” for the patient selection domain if a clinically relevant control population
was used; OD for diagnostic studies and LTBI or household TB contacts for predictive studies
and “high applicability concerns” if other populations were used. For other domains, we
judged “low applicability concerns” if all signalling questions in that domain were answered as
“yes”, “high applicability concerns” if any question was answered as “no” and “unclear applica-
bility concerns” if answered as such.
GRADE quality of evidence
We used the “Grading of Recommendations Assessment, Development and Evaluation”
(GRADE) approach to judge the quality of evidence. Classification of the quality of evidence
was based on study design in conjunction with the five factors that affect study quality; study
limitations, indirectness, inconsistency, imprecision, and publication bias [35].
Data synthesis and analysis
We extracted data to construct 2 x 2 contingency tables of reference test versus index test
results. TB positive and TB negative were defined as participants with and without TB disease
respectively, based on the reference standard. Forest plots of sensitivity and specificity with
95% confidence intervals for each signature were created using RevMan 5.3 [36]. Each entry in
the forest plot represents a signature that was evaluated in a distinct cohort. Several signatures
were tested in multiple cohorts and reported in multiple studies. In naming individual studies,
we used the first author name, year of manuscript publication, and a sequential letter repre-
senting a specific cohort. Similarly, the signature naming convention was first author name,
number of genes, and year of publication.
We reported the index test results as TP, FP, TN and FN. If not explicitly reported, TP, FP,
TN, and FN were estimated from the reported sensitivity and specificity and total number of
TB positive individuals and controls. Similarly, sensitivity and specificity were reported, or cal-
culated if the data were available or obtained from the authors. We also calculated the PPV
and negative predictive values (NPV) at 2% pre-test probability for predictive signatures.
Signatures used for both diagnosis and prediction of TB disease are presented in separate
forest plots. Each evaluation of a signature in a different population, or of different signatures
in the same population, is shown as a separate entry (or entries) in the forest plot. If the same
signature was reported using different models, the best performing model was included in the
analysis. Similarly, if a study reported several signatures with the same number of genes in the
same population, only the best performing signature was chosen.
Considerable clinical and methodological heterogeneity was anticipated due to reporting of
multiple signatures in different populations. Therefore, we did not perform meta-analysis on
PLOS ONE | https://doi.org/10.1371/journal.pone.0237574 August 21, 2020
5 / 17
PLOS ONE Performance of diagnostic and predictive host blood transcriptomic signatures for Tuberculosis disease
all signatures in all studies. In order to address this heterogeneity and compare relative perfor-
mance of signatures, we performed meta-analysis only for signatures that were evaluated in at
least three comparable cohorts with the same control population. Meta-analysis was conducted
in STATA 11, using hierarchical logistic regression. We used the bivariate random effects
model to calculate summary sensitivity and specificity with the corresponding 95% CI [37],
and to create summary receiver operating characteristic curves for each signature. Heterogene-
ity in the diagnostic or predictive performance of the signatures was assessed by visual inspec-
tion of the forest plots and the I2 statistic.
Results
Search results
Our search term returned 2,313 reports of which 27 [23–26, 38–62] satisfied all the pre-
specified inclusion criteria (Fig 1). Twenty-three of the 27 studies reported exclusively on diag-
nostic performance of mRNA signatures for discriminating TB disease cases from controls
with or without ODs, and controls with or without LTBI. Two studies reported exclusively on
predictive performance for progression to TB disease [26, 55]; and two studies reported both
diagnostic and predictive performance [25, 60]. A total of 35 transcriptomic signatures incor-
porating 1,027 genes were identified (S2 Table). Forty-two of the 1,027 genes were employed
in at least three or more transcriptomic signatures and Fc gamma receptor 1A (FCGR1A) was
the most frequently utilised gene (Fig 1 in S3 File)
Quality of diagnostic studies
Four studies [42, 44, 58, 60] employed a cohort design, four [23, 38, 51, 52] a cross sectional
design, and the remainder employed a case-control design. All studies had a reference stan-
dard of either smear microscopy [50], Xpert/MTB RIF [41] or MTB culture; Table 1 in S3 File.
One study [46] did not specify the type of reference test used. Eight [23–25, 38, 44, 45, 49, 56,
59, 60] of the 18 studies evaluated signatures in populations with ODs. Only one study [43]
made reference to blinding of the index test readers. Bias in patient selection arose in most
studies due to case-control design and non-reporting of sampling method, resulting in a non-
representative spectrum of patients. Bias in the index test resulted from the lack of reported
blinding in the index test interpretation. Bias in the reference standard was minimal since 97%
(68/70) of the entries used MTB culture as a reference standard which can be considered objec-
tive (Fig 2; Figs 2 and 3 in S3 File).
Quality of studies for prediction of progression to TB disease
All studies [25, 26, 55, 60] validated signatures in participants from prospective independent
cohorts, although two studies [25, 26] were case-control studies nested in prospective cohorts
Fig 2. Summary of results of QUADAS-2 assessment of diagnostic studies in clinically relevant independent
validation cohorts with other diseases.
https://doi.org/10.1371/journal.pone.0237574.g002
PLOS ONE | https://doi.org/10.1371/journal.pone.0237574 August 21, 2020
6 / 17
PLOS ONE Performance of diagnostic and predictive host blood transcriptomic signatures for Tuberculosis disease
Fig 3. Summary of results of QUADAS-2 assessment in studies of prediction to TB disease in independent
validation cohorts.
https://doi.org/10.1371/journal.pone.0237574.g003
(Table 2 in S3 File). Participants from three studies did not all receive the same reference stan-
dard [26, 55, 60] and blinding was only stated in two studies [25, 26] (Fig 3; Fig 4 in S3 File).
Performance of mRNA signatures for TB Diagnosis
Eighteen studies evaluated 24 transcriptomic signatures for diagnosis of TB disease in 70 dif-
ferent independent validation cohorts (Fig 5 in S3 File). Individual signatures displayed sub-
stantial variation in diagnostic performance with sensitivity ranging from 50%-100% and
specificity ranging from 32%-100%. The observed heterogeneity (I2 = 0.99) in study design
and signature performance precluded pooling of diagnostic accuracy estimates for all signa-
tures. Nine signatures were not evaluated in independent validation cohorts and hence
excluded (Fig 6 in S3 File).
Thirty-three entries (46.5%) representing 17 different signatures in 12 studies met at least
one TPP minimum performance criterion in independent validation sets containing unin-
fected controls, LTBI, ODs, or a combination of these populations (Fig 4). Signature perfor-
mance ranged between 69%-100% for sensitivity, and between 70%-100% for specificity.
Ten studies evaluated 12 signatures for diagnosis of TB disease in clinically relevant popula-
tions with ODs (Fig 5). Signature performance ranged between 50%-100% for sensitivity, and
between 47%-96% for specificity. Seven of these signatures; Berry86_2010, daCosta2_2015,
daCosta3_2015, Francisco2_2017, Kaforou44_2015, Walter47_2016, and Zak16_2016 met the
WHO-recommended minimal TPP for a triage test. None of these signatures met the WHO
minimal TPP for a diagnostic test in the OD population.
Fig 4. Forest plots of sensitivity and specificity of transcriptomic signatures for diagnosis of TB disease that met
at least one minimum TPP performance criterion in independent validation cohorts. HC; Healthy controls, LTBI;
Latent TB infection, OD; Other diseases. Vertical dashed lines correspond to 90% sensitivity and 70% specificity.
https://doi.org/10.1371/journal.pone.0237574.g004
PLOS ONE | https://doi.org/10.1371/journal.pone.0237574 August 21, 2020
7 / 17
PLOS ONE Performance of diagnostic and predictive host blood transcriptomic signatures for Tuberculosis disease
Fig 5. Forest plots of sensitivity and specificity of transcriptomic signatures for diagnosis of TB disease in
independent validation cohorts of clinically relevant populations with other diseases. Vertical dashed lines
correspond to 90% sensitivity and 70% specificity.
https://doi.org/10.1371/journal.pone.0237574.g005
Meta-analysis of performance of mRNA signatures for diagnosis of TB
disease
Ten studies included seven signatures that were validated for diagnosis of TB in at least three
comparable cohorts including either LTBI or OD populations. Signature diagnostic perfor-
mance results are shown in Table 1. We excluded signatures validated exclusively in popula-
tions with uninfected controls, but included signatures evaluated in LTBI or OD populations
that also included uninfected controls. Four signatures were validated in exclusively LTBI pop-
ulations and six signatures were validated exclusively in OD populations; and two signatures
in both LTBI and OD populations. Among the six signatures validated in LTBI cohorts, three
signatures with similar diagnostic accuracy (Berry393_2010, Kaforou27_2013, Zak16_2016)
showed pooled sensitivity and specificity that met the minimal WHO TPP for a non-sputum
biomarker triage test in this population. Summary receiver operating characteristic curves for
the signatures in OD populations are shown as Figs 7a, 7b and 7c in S3 File. Heterogeneity in
the signature performance was not explained by comparison group (I2 = 0.98 for all OD, I2 =
0.95 for all LTBI, I2 = 0.90 for all healthy controls,). Zak16_2016 which used the TPP bench-
mark of 90% sensitivity had a lower I2 of 0.56 compared to Sweeny3_2016’s I2 of 0.93. Some of
the heterogeneity was explained by whether the evaluation used the TPP benchmarks and by
composition of the control population.
Performance of mRNA signatures for prediction of progression to TB
disease
Four studies evaluated five signatures for prediction of progression to TB disease in indepen-
dent validation cohorts with LTBI and uninfected controls. Two of these studies evaluated
LTBI populations only; one study evaluated TB contacts only; and one study evaluated both
TB contacts and LTBI populations (Fig 6). The time window between signature measurement
and TB diagnosis reported in these studies was not consistent and ranged from 6 months to 24
months before TB diagnosis. Since differential gene expression becomes more pronounced as
individuals approach TB diagnosis [21, 25, 26], this variable precluded direct comparison of
signature performance for prediction of progression to TB disease. In LTBI populations, signa-
ture performance ranged between 76%-86% for sensitivity; and between 55%-84% for specific-
ity. Signatures evaluated in TB contacts showed between 53%-67% sensitivity and 83%-99%
specificity. The Sweeney3_2016 signature met the TPP performance criterion (PPV � 5.8%
and 75% sensitivity and 75% specificity) for a test to predict progression when measured
PLOS ONE | https://doi.org/10.1371/journal.pone.0237574 August 21, 2020
8 / 17
PLOS ONE Performance of diagnostic and predictive host blood transcriptomic signatures for Tuberculosis disease
Table 1. Meta-analysis of performance of transcriptomic signatures for diagnosis of TB disease in independent validation cohorts that included LTBI or other dis-
ease populations.
Signature
Berry393_2010
Dawany251_2015
Kaforou27_2013
Samabarey10_2017
Sweeney3_2016
Zak16_2016
Berry86_2010
Sweeney3_2016
Zak16_2016
Study Entry ID
TB/ Control
Control Group
Sensitivity % (95% CI)
Specificity % (95% CI)
Berry2010a
Kaforou2013c
Leong2018a
Walter2016d
Dawany2015b
Dawany2015c
Dawany2015d
Dawany2015e
Kaforou2013a
Leong2018d
Walter2016g
Leong2018e
Sambarey2017a
Sambarey2017c
Leong2018f
20/31
97/83
24/16
35/35
29/38
13/29
21/33
20/31
20/31
24/16
35/35
24/16
51/36
47/47
24/16
LTBI
LTBI
LTBI
LTBI
LTBI
LTBI�
LTBI�
LTBI
LTBI
LTBI
LTBI
LTBI
LTBI
LTBI
LTBI
Warsinske2018a
176/187
LTBI+
Sweeney2016d
Leong2018g
Zak2016a
Zak2016b
Zak2016d
Zak2016e
Zak2016l
Berry2010b
Kaforou2013d
Walter2016e
Francisco2017b
Francisco2017d
Sweeney2016e
Warsinske2018a
Zak2016g
Zak2016h
Zak2016i
Zak2016j
Zak2016k
46/25
24/16
21/21
20/30
51/35
46/48
29/38
20/96
97/83
35/39
275/290
144/209
8/18
176/187
35/16
35/14
35/61
51/34
46/49
LTBI
LTBI
LTBI
LTBI
LTBI
LTBI
LTBI
OD
OD
OD
OD
OD
OD
OD+
OD
OD
OD
OD
OD
90 (83, 94)
92 (82,96)
82 (70, 90)
95 (88, 98)
95 (87, 98)
93 (85, 97)
83 (75, 88)
92 (85, 96)
89 (84, 92)
72 (66, 78)
91 (86, 94)
90 (72, 97)
84 (68, 93)
79 (73, 84)
74 (57, 86)
71 (49, 86)
90 (85, 93)
74 (56, 86)
LTBI; Latent TB infection, OD; Other diseases, HC; Health control
�Includes some HCs, +Consists of HCs, ODs and LTBIs
https://doi.org/10.1371/journal.pone.0237574.t001
Fig 6. Forest plots of sensitivity and specificity of transcriptomic signatures for prediction of progression to TB
disease in independent validation cohorts. Vertical dashed lines correspond to 75% sensitivity and 75% specificity.
Prediction time is the time to TB disease used in each study.
https://doi.org/10.1371/journal.pone.0237574.g006
PLOS ONE | https://doi.org/10.1371/journal.pone.0237574 August 21, 2020
9 / 17
PLOS ONE Performance of diagnostic and predictive host blood transcriptomic signatures for Tuberculosis disease
Fig 7. GRADE evidence profile: Transcriptomic signatures for the diagnosis of TB in clinically relevant
populations with other diseases only.
https://doi.org/10.1371/journal.pone.0237574.g007
within 6 months of TB diagnosis. Performance of the other signatures was not reported for
this 6-month predictive interval (time to TB disease) and none of the other signatures met the
TPP performance criterion over longer predictive intervals. Regardless, two other signatures,
assessed within 12 months of TB diagnosis, also achieved a PPV � 5.8% (Table 3 in S3 File).
GRADE evidence profile
We followed a previously published GRADE guideline [35] to assess the quality of the body of
evidence and produced a GRADE evidence profile (Fig 7 and Fig 8 in S3 File). The overall
quality of evidence supporting the estimates of sensitivity and specificity of mRNA signatures
for the diagnosis of TB disease was rated as “very low”. Consequently, very low confidence is
placed in the estimates obtained from pooling studies in meta-analysis. Similarly, the quality of
evidence for studies of progression to TB disease was also very low.
Discussion
The New Diagnostics Working Group Strategic Framework 2016–2020 aims to “achieve early
and universal diagnosis of all patients with all forms of TB to foster progress towards TB elimi-
nation, by making appropriate and affordable diagnostic solutions available at the right set-
ting” [63]. Comparison of signature performance in similar populations and under similar
conditions is critical to down-select candidate biomarkers for development as rapid point-of-
care (POC) tests. Future implementation decisions hinge on test performance in clinically rele-
vant populations under field conditions, in addition to feasibility and cost considerations that
contribute to the assessment of impact and public health value.
We show that of the 17 mRNA signatures that met at least one minimal TPP performance
criterion, 11 were validated in populations including either uninfected controls, LTBI, or both,
and four were validated exclusively in populations with ODs. Four signatures validated in the
uninfected control category (Dewany251_2015, Sweeney3_2016, Zak16_2016, and Lee2_2016)
have potential as diagnostic tests while the remainder show potential as triage tests. The signa-
tures in the LTBI/uninfected control category that are yet to be validated in populations with
ODs such as DeAraujo1_2016, Lee2_2016, Lee3_2016, and Sambarey10_2017 should also be
validated in such populations to confirm robustness of diagnostic accuracy and allow compari-
son with the signatures above.
Although TB triage tests might be used in community mass screening campaigns or contact
investigations that include uninfected individuals or asymptomatic individuals with LTBI; the
greatest need is for TB diagnostic tests that discriminate TB from ODs among symptomatic
PLOS ONE | https://doi.org/10.1371/journal.pone.0237574 August 21, 2020
10 / 17
PLOS ONE Performance of diagnostic and predictive host blood transcriptomic signatures for Tuberculosis disease
individuals seeking health care. We also recognise that evaluation of novel biomarkers in
cohorts of carefully selected TB cases and uninfected controls tends to over-estimate perfor-
mance. In this regard, diagnostic evaluation of signatures in populations that exclusively
included LTBI or uninfected controls would be considered inferior to validation of perfor-
mance in clinically relevant cohorts that included individuals with other respiratory diseases.
The OD category in several of these studies did not include symptomatic individuals present-
ing to healthcare facilities with suspected TB, but rather individuals with systemic lupus ery-
thematosus, sarcoidosis, or other infrequently encountered conditions. None of the signatures
validated in cohorts with ODs met the minimum WHO TPP for sensitivity and specificity of a
diagnostic test. However, seven signatures; Berry86_2010, daCosta2_2015, daCosta3_2015,
Francisco2_2017, Kaforou44_2015, Walter47_2016, and Zak16_2016 met the minimum
WHO TPP for a triage test. The findings suggest that these seven signatures could be further
evaluated as TB triage tests under field conditions; and that feasibility and unit cost-effective-
ness as potential rapid POC tests would be a consideration for further clinical development.
Other signatures that approached but did not meet the TPP may also warrant further valida-
tion in side-by-side comparison studies. Signatures with smaller number of genes which may
be more adaptable to a POC device could be given preferential consideration for clinical vali-
dation and development. Our findings also suggest that further improvement in performance
of existing signatures, or even further discovery of new signatures with improved performance,
would be necessary for clinical development of a non-sputum TB diagnostic test. For instance,
incorporating covariates such as age and sex in the models.
Validation of diagnostic tests in several geographically different populations is important to
confirm robustness. Relatively few signatures were validated in at least three comparable
cohorts of a similar population and were eligible for meta-analysis. Only one of the signatures
(Zak16_2016), validated in multiple cohorts that included ODs such as pneumonia, lung can-
cer, sarcoidosis, or systemic lupus erythematosus, met the minimum WHO TPP performance
criteria for a triage test in meta-analysis and none met the minimal TPP for a diagnostic test in
the meta-analyses. The finding suggests that this signature has potential as a rapid POC test
and should be considered for clinical development upon validation under field conditions.
Similarly, other signatures approaching the minimum TPP target in meta-analysis should be
considered for field validation.
Tests that will accurately predict which individuals with LTBI will develop TB disease are
needed to ensure that preventive treatment can be targeted for those individuals at increased
risk of incident TB disease, while saving those individuals at lowest risk from the cost, burden,
and side-effects of unnecessary intervention. Current tests for MTB infection, including IGRA
and TSTs, are poor predictors because of their low specificity for incident TB disease. It is not
cost-effective to treat the estimated two billion individuals latently infected with MTB world-
wide and therefore preventive therapy targeted with a more specific biomarker may be a more
feasible alternative [64]. Predictive signatures might be used in community-level TB mass-
screening campaigns or contact investigations, but might also be useful in symptomatic indi-
viduals who have been investigated and found not to have active TB disease at the time of test-
ing. Only one of the four signatures (Sweeney3_2016) met the minimum TPP criteria for both
PPV � 5.8% and 75% sensitivity and 75% specificity for a 6-month period prior to TB diagno-
sis, although this was the only signature for which performance within six-months of TB diag-
nosis was reported. The other studies reported signature performance one or two years prior
to TB diagnosis. It is thus not clear how these signatures would perform during the six-month
period before TB diagnosis. These results highlight that more studies of predictive perfor-
mance are necessary. It is also evident that a two-year predictive horizon may be overly opti-
mistic for prediction of progression to TB disease as progression is very variable in occurrence.
PLOS ONE | https://doi.org/10.1371/journal.pone.0237574 August 21, 2020
11 / 17
PLOS ONE Performance of diagnostic and predictive host blood transcriptomic signatures for Tuberculosis disease
One of the challenges of this systematic review and meta-analysis was that most studies did
not specify whether the signature being tested was intended for triage, diagnostic, or predictive
use, neither did they benchmark performance of signatures against the WHO TPPs. For exam-
ple, if the goal is to discover and validate a TB triage test, a study should report specificity at
90% sensitivity or higher, to allow comparison with other novel biomarkers against this stan-
dard. Similarly, if the goal is to discover and validate a TB diagnostic test, the study should ide-
ally report sensitivity at 98% specificity or higher; and studies aiming to discover and validate a
test to predict progression to TB disease should report sensitivity at 75% specificity or higher,
or PPV and NPV along with the prediction time horizon. However, it must be noted that the
differences in the benchmarks are partially because some of the studies were completed before
publication of the TPPs. Secondly, we found that many studies were sub-optimally designed
and used control populations without clinical relevance. As observed in a previous systematic
review [28], we found that several of transcriptomic signatures for TB diagnosis were discov-
ered but not validated in independent representative cohorts. We also found that a number of
diagnostic accuracy studies did not conform to the reporting guidelines for diagnostic test
accuracy (DTA) studies stipulated in the “Standards for the Reporting of Diagnostic Accuracy
Studies” (STARD) [65]. In several studies, cardinal data on study design, patient selection,
numbers of participants in each group, and diagnostic performance data such as sensitivity
and specificity with their corresponding confidence intervals (CIs) that would enable repro-
duction of the study were not reported. This is a major drawback in synthesising the body of
evidence on DTA studies and thus compliance to STARD in designing DTA studies and
reporting their findings cannot be over-emphasised.
Strengths and limitations of the study
We used an inclusive time frame of January 2005 to May 2019 to include the period in which
we believe all transcriptomic TB biomarker studies were published. We also developed a proto-
col prior to performing the systematic review that explicitly stated a rigorous search strategy
and clear inclusion/exclusion criteria. Unlike previous systematic reviews, our review includes
evaluation of signatures for predicting progression to TB disease and a meta-analysis.
Some signatures were designed to optimise sensitivity while others were designed to opti-
mise specificity. This may have introduced bias in the pooled estimates of sensitivity and speci-
ficity in the meta-analysis, and difficult to compare signature performance. Restricting
included studies to those conducted in HIV-negative adults and adolescents may have
excluded signatures with superior diagnostic performance in studies conducted in children or
in HIV-positive individuals. Additionally, language selection bias cannot be ruled out since we
only included studies reported in English. We did not formally assess publication bias as cur-
rent methods are not suitable for DTA studies [66].
Heterogeneity of study design and reliance on reported data makes it impossible to fairly
compare signature performance. A major finding of this study and limitation is the very low
quality of evidence: preponderance of case control studies, spectrum bias and narrow geogra-
phy. This highlights the need for high quality, prospective studies, with relevant populations of
symptomatic clinic attendees, mass screening endemic community population or high-risk
populations such as household TB contacts which minimise spectrum bias, and from multiple
geographies.
Conclusion
Host blood mRNA signatures show considerable promise as triage tests for TB. Signatures
designed for TB diagnosis meeting at least one TPP minimum performance criterion in
PLOS ONE | https://doi.org/10.1371/journal.pone.0237574 August 21, 2020
12 / 17
PLOS ONE Performance of diagnostic and predictive host blood transcriptomic signatures for Tuberculosis disease
independent validation sets containing healthy controls or LTBI populations should be further
optimised in populations with ODs. Similarly signatures for TB diagnosis validated in popula-
tions with ODs and signatures for prediction of progression from LTBI to TB disease meeting
the minimum TPP should be further optimised and validated under field conditions to con-
firm their accuracy for use as standalone diagnostic or predictive tests for therapeutic deci-
sion-making. There is also need for signature discovery in large “real-world” clinically
appropriate populations, without spectrum bias, need for head-to-head comparison of signa-
tures and adaptation and implementation towards a POC test.
Supporting information
S1 Table. PRISMA checklist. Checklist according to the PRISMA reporting guidelines.
(DOCX)
S2 Table. Gene matrix. A matrix of all identified signatures with their corresponding gene
composition.
(XLSX)
S1 File. Data extraction form. Form used for data extraction.
(PDF)
S2 File. QUADAS-2 form. Modified individual study quality assessment form.
(PDF)
S3 File. Methods and results. Manuscript-specific supplementary methods, figures and tables.
(PDF)
Acknowledgments
The authors acknowledge the contribution of Mrs. Mary Shelton, the Health Sciences’ refer-
ence librarian, at the University of Cape Town, for her assistance in developing the search
strategies.
Author Contributions
Conceptualization: Thomas J. Scriba, Mark Hatherill.
Data curation: Humphrey Mulenga.
Formal analysis: Humphrey Mulenga, Chambrez-Zita Zauchenberger.
Investigation: Humphrey Mulenga, Chambrez-Zita Zauchenberger, Erick W. Bunyasi.
Methodology: Humphrey Mulenga, Erick W. Bunyasi, Thomas J. Scriba, Mark Hatherill.
Supervision: Thomas J. Scriba, Mark Hatherill.
Validation: Benjamin Kagina.
Writing – original draft: Humphrey Mulenga.
Writing – review & editing: Humphrey Mulenga, Chambrez-Zita Zauchenberger, Erick W.
Bunyasi, Stanley Kimbung Mbandi, Simon C. Mendelsohn, Benjamin Kagina, Adam Penn-
Nicholson, Thomas J. Scriba, Mark Hatherill.
References
1. WHO. The End TB Strategy. Geneva, Switzerland: World Health Organization; 2015.
PLOS ONE | https://doi.org/10.1371/journal.pone.0237574 August 21, 2020
13 / 17
PLOS ONE Performance of diagnostic and predictive host blood transcriptomic signatures for Tuberculosis disease
2. WHO. Meeting Report: High-priority target product profiles for new tuberculosis diagnostics: report of a
consensus meeting. Geneva, Switzerland: World Health Organization; 2014. Contract No.: WHO/
HTM/TB/2014.18.
3. Seshadri P, Denkinger C. Target Product Profile: Test for Incipient Tuberculosis. Geneva, Switzerland:
Foundation for Innovative New Diagnostics; 2016. Contract No.: Vs8.2; 20 October 2016.
4. WHO. Consensus Meeting Report: Development of a Target Product Profile (TPP) and a framework for
evaluation for a test for predicting progression from tuberculosis infection to active disease. Report.
Geneva: World Health Organization; 2017. Contract No.: WHO/HTM/TB/2017.18.
5. Ghodbane R, Raoult D, Drancourt M. Dramatic reduction of culture time of Mycobacterium tuberculosis.
Scientific reports. 2014; 4:4236. https://doi.org/10.1038/srep04236 PMID: 24577292
6. Norbis L, Miotto P, Alagna R, Cirillo DM. Tuberculosis: lights and shadows in the current diagnostic
landscape. New Microbiol. 2013; 36(2):111–20. PMID: 23686117
7. Davies PD, Pai M. The diagnosis and misdiagnosis of tuberculosis. The international journal of tubercu-
losis and lung disease: the official journal of the International Union against Tuberculosis and Lung Dis-
ease. 2008; 12(11):1226–34.
8. Steingart KR, Henry M, Ng V, Hopewell PC, Ramsay A, Cunningham J, et al. Fluorescence versus con-
ventional sputum smear microscopy for tuberculosis: a systematic review. The Lancet Infectious dis-
eases. 2006; 6(9):570–81. https://doi.org/10.1016/S1473-3099(06)70578-3 PMID: 16931408
9. Harries AD, Maher D, Nunn P. An approach to the problems of diagnosing and treating adult smear-
negative pulmonary tuberculosis in high-HIV-prevalence settings in sub-Saharan Africa. Bull World
Health Organ. 1998; 76(6):651–62. PMID: 10191561
10.
Li S, Liu B, Peng M, Chen M, Yin W, Tang H, et al. Diagnostic accuracy of Xpert MTB/RIF for tuberculo-
sis detection in different regions with different endemic burden: A systematic review and meta-analysis.
PloS one. 2017; 12(7):e0180725. https://doi.org/10.1371/journal.pone.0180725 PMID: 28708844
11. Dorman SE, Schumacher SG, Alland D, Nabeta P, Armstrong DT, King B, et al. Xpert MTB/RIF Ultra
for detection of Mycobacterium tuberculosis and rifampicin resistance: a prospective multicentre diag-
nostic accuracy study. The Lancet Infectious diseases. 2018; 18(1):76–84. https://doi.org/10.1016/
S1473-3099(17)30691-6 PMID: 29198911
12. Maertzdorf J, Kaufmann SH, Weiner J 3rd. Toward a unified biosignature for tuberculosis. Cold Spring
Harbor perspectives in medicine. 2014; 5(1):a018531. https://doi.org/10.1101/cshperspect.a018531
PMID: 25342061
13. Andrews JR, Hatherill M, Mohamed H, Hanekom WA, Campo M, Hawn TR, et al. The Dynamics of
QuantiFERON-TB Gold In-Tube Conversion and Reversion in a Cohort of South African Adolescents.
American journal of respiratory and critical care medicine. 2015; 191(5):584–91. https://doi.org/10.
1164/rccm.201409-1704OC PMID: 25562578
14. Mahomed H, Hawkridge T, Verver S, Abrahams D, Geiter L, Hatherill M, et al. The tuberculin skin test
versus QuantiFERON TB Gold(R) in predicting tuberculosis disease in an adolescent cohort study in
South Africa. PloS one. 2011; 6(3):e17984. https://doi.org/10.1371/journal.pone.0017984 PMID:
21479236
15. Andrews JR, Noubary F, Walensky RP, Cerda R, Losina E, Horsburgh CR. Risk of progression to active
tuberculosis following reinfection with Mycobacterium tuberculosis. Clinical infectious diseases: an offi-
cial publication of the Infectious Diseases Society of America. 2012; 54(6):784–91.
16. Barry CE, 3rd, Boshoff HI, Dartois V, Dick T, Ehrt S, Flynn J, et al. The spectrum of latent tuberculosis:
rethinking the biology and intervention strategies. Nature reviews Microbiology. 2009; 7(12):845–55.
https://doi.org/10.1038/nrmicro2236 PMID: 19855401
17. Diel R, Loddenkemper R, Nienhaus A. Predictive value of interferon-gamma release assays and tuber-
culin skin testing for progression from latent TB infection to disease state: a meta-analysis. Chest.
2012; 142(1):63–75. https://doi.org/10.1378/chest.11-3157 PMID: 22490872
18. Dye C, Glaziou P, Floyd K, Raviglione M. Prospects for tuberculosis elimination. Annu Rev Public
Health. 2013; 34:271–86. https://doi.org/10.1146/annurev-publhealth-031912-114431 PMID:
23244049
19. Sumner T, Scriba TJ, Penn-Nicholson A, Hatherill M, White RG. Potential population level impact on
tuberculosis incidence of using an mRNA expression signature correlate-of-risk test to target tuberculo-
sis preventive therapy. Scientific reports. 2019;9. https://doi.org/10.1038/s41598-018-36956-2
20. Burel JG, Peters B. Discovering transcriptional signatures of disease for diagnosis versus mechanism.
Nature reviews Immunology. 2018; 18(5):289–90. https://doi.org/10.1038/nri.2018.26 PMID: 29629714
21. Scriba TJ, Penn-Nicholson A, Shankar S, Hraha T, Thompson EG, Sterling D, et al. Sequential inflam-
matory processes define human progression from M. tuberculosis infection to tuberculosis disease.
PLOS ONE | https://doi.org/10.1371/journal.pone.0237574 August 21, 2020
14 / 17
PLOS ONE Performance of diagnostic and predictive host blood transcriptomic signatures for Tuberculosis disease
PLoS pathogens. 2017; 13(11):e1006687. https://doi.org/10.1371/journal.ppat.1006687 PMID:
29145483
22. Blankley S, Berry MPR, Graham CM, Bloom CI, Lipman M, O’Garra A. The application of transcriptional
blood signatures to enhance our understanding of the host response to infection: the example of tuber-
culosis. Philos T R Soc B. 2014; 369(1645).
23. Berry MPR, Graham CM, McNab FW, Xu ZH, Bloch SAA, Oni T, et al. An interferon-inducible neutro-
phil-driven blood transcriptional signature in human tuberculosis. Nature. 2010; 466(7309):973–U98.
https://doi.org/10.1038/nature09247 PMID: 20725040
24. Kaforou M, Wright VJ, Oni T, French N, Anderson ST, Bangani N, et al. Detection of tuberculosis in
HIV-infected and -uninfected African adults using whole blood RNA expression signatures: a case-con-
trol study. PLoS medicine. 2013; 10(10):e1001538. https://doi.org/10.1371/journal.pmed.1001538
PMID: 24167453
25.
Zak DE, Penn-Nicholson A, Scriba TJ, Thompson E, Suliman S, Amon LM, et al. A blood RNA signature
for tuberculosis disease risk: a prospective cohort study. Lancet (London, England). 2016; 387
(10035):2312–22.
26. Suliman S, Thompson EG, Sutherland J, Weiner J, Ota MOC, Shankar S, et al. Four-Gene Pan-African
Blood Signature Predicts Progression to Tuberculosis. American journal of respiratory and critical care
medicine. 2018; 197(9):1198–208. https://doi.org/10.1164/rccm.201711-2340OC PMID: 29624071
27.
Togun TO, MacLean E, Kampmann B, Pai M. Biomarkers for diagnosis of childhood tuberculosis: A
systematic review. PloS one. 2018; 13(9):1–19.
28. MacLean E, Broger T, Yerlikaya S, Fernandez-Carballo BL, Pai M, Denkinger CM. A systematic review
of biomarkers to detect active tuberculosis. Nat Microbiol. 2019; 4(5):748–58. https://doi.org/10.1038/
s41564-019-0380-2 PMID: 30804546
29. Warsinske H, Vashisht R, Khatri P. Host-response-based gene signatures for tuberculosis diagnosis: A
systematic comparison of 16 signatures. PLoS medicine. 2019; 16(4):e1002786. https://doi.org/10.
1371/journal.pmed.1002786 PMID: 31013272
30. Gupta RK, Turner CT, Venturini C, Esmail H, Rangaka MX, Copas A, et al. Concise whole blood tran-
scriptional signatures for incipient tuberculosis: a systematic review and patient-level pooled meta-anal-
ysis. The Lancet Respiratory medicine. 2020; 8(4):395–406. https://doi.org/10.1016/S2213-2600(19)
30282-6 PMID: 31958400
31. Cochrane_Collaboration. Handbook for DTA Reviews: Cochrane; 2017 [Cochrane DTA Working
Group]. Available from: http://methods.cochrane.org/sdt/handbook-dta-reviews.
32. Moher D, Shamseer L, Clarke M, Ghersi D, Liberati A, Petticrew M, et al. Preferred reporting items for
systematic review and meta-analysis protocols (PRISMA-P) 2015 statement. Syst Rev. 2015; 4:1.
https://doi.org/10.1186/2046-4053-4-1 PMID: 25554246
33. Mulenga H, Bunyasi EW, Mbandi SK, Mendelsohn SC, Kagina B, Penn-Nicholson A, et al. Performance
of host blood transcriptomic signatures for diagnosing and predicting progression to tuberculosis dis-
ease in HIV-negative adults and adolescents: a systematic review protocol. BMJ Open. 2019; 9(5):
e026612. https://doi.org/10.1136/bmjopen-2018-026612 PMID: 31122978
34. Whiting PF, Rutjes AWS, Westwood ME, Mallett S, Deeks JJ, Reitsma JB, et al. QUADAS-2: A Revised
Tool for the Quality Assessment of Diagnostic Accuracy Studies. Annals of internal medicine. 2011; 155
(8):529–U104. https://doi.org/10.7326/0003-4819-155-8-201110180-00009 PMID: 22007046
35. Horvath AR. Grading quality of evidence and strength of recommendations for diagnostic tests and
strategies. Clinical chemistry. 2009; 55(5):853–5. https://doi.org/10.1373/clinchem.2008.117614 PMID:
19299540
36. Review Manager (RevMan) [Computer program]. Version 5.3. Copenhagen: The Nordic Cochrane
Centre, The Cochrane Collaboration, 2014.
37. Reitsma JB, Glas AS, Rutjes AW, Scholten RJ, Bossuyt PM, Zwinderman AH. Bivariate analysis of sen-
sitivity and specificity produces informative summary measures in diagnostic reviews. J Clin Epidemiol.
2005; 58(10):982–90. https://doi.org/10.1016/j.jclinepi.2005.02.022 PMID: 16168343
38. Bloom CI, Graham CM, Berry MP, Rozakeas F, Redford PS, Wang Y, et al. Transcriptional blood signa-
tures distinguish pulmonary tuberculosis, pulmonary sarcoidosis, pneumonias and lung cancers. PloS
one. 2013; 8(8):e70630. https://doi.org/10.1371/journal.pone.0070630 PMID: 23940611
39. Satproedprai N, Wichukchinda N, Suphankong S, Inunchot W, Kuntima T, Kumpeerasart S, et al. Diag-
nostic value of blood gene expression signatures in active tuberculosis in Thais: a pilot study. Genes
and immunity. 2015; 16(4):253–60. https://doi.org/10.1038/gene.2015.4 PMID: 25764116
40. Cai Y, Yang Q, Tang Y, Zhang M, Liu H, Zhang G, et al. Increased complement C1q level marks active
disease in human tuberculosis. PloS one. 2014; 9(3):e92340. https://doi.org/10.1371/journal.pone.
0092340 PMID: 24647646
PLOS ONE | https://doi.org/10.1371/journal.pone.0237574 August 21, 2020
15 / 17
PLOS ONE Performance of diagnostic and predictive host blood transcriptomic signatures for Tuberculosis disease
41. Darboe F, Mbandi SK, Thompson EG, Fisher M, Rodo M, van Rooyen M, et al. Diagnostic performance
of an optimized transcriptomic signature of risk of tuberculosis in cryopreserved peripheral blood mono-
nuclear cells. Tuberculosis. 2018; 108:124–6. https://doi.org/10.1016/j.tube.2017.11.001 PMID:
29523312
42. Dawany N, Showe LC, Kossenkov AV, Chang C, Ive P, Conradie F, et al. Identification of a 251 gene
expression signature that can accurately detect M. tuberculosis in patients with and without HIV co-
infection. PloS one. 2014; 9(2):e89925. https://doi.org/10.1371/journal.pone.0089925 PMID: 24587128
43.
44.
de Araujo LS, Vaas LA, Ribeiro-Alves M, Geffers R, Mello FC, de Almeida AS, et al. Transcriptomic Bio-
markers for Tuberculosis: Evaluation of DOCK9. EPHA4, and NPC2 mRNA Expression in Peripheral
Blood. Frontiers in microbiology. 2016; 7:1586. https://doi.org/10.3389/fmicb.2016.01586 PMID:
27826286
Francisco NM, Fang YM, Ding L, Feng S, Yang Y, Wu M, et al. Diagnostic accuracy of a selected signa-
ture gene set that discriminates active pulmonary tuberculosis and other pulmonary diseases. The Jour-
nal of infection. 2017; 75(6):499–510. https://doi.org/10.1016/j.jinf.2017.09.012 PMID: 28941629
45. Hai-Hui H, Xiao-Ying L, Yong L, Hua C, Liang-Yong X. Identification of 13 blood-based gene expression
signatures to accurately distinguish tuberculosis from other pulmonary diseases and healthy controls.
Bio-Medical Materials & Engineering. 2015; 26:S1837–S43.
46. Huang ZK, Yao FY, Xu JQ, Deng Z, Su RG, Peng YP, et al. Microarray Expression Profile of Circular
RNAs in Peripheral Blood Mononuclear Cells from Active Tuberculosis Patients. Cellular physiology
and biochemistry: international journal of experimental cellular physiology, biochemistry, and pharma-
cology. 2018; 45(3):1230–40.
47.
48.
49.
50.
51.
52.
Jacobsen M, Repsilber D, Gutschmidt A, Neher A, Feldmann K, Mollenkopf HJ, et al. Candidate bio-
markers for discrimination between infection and disease caused by Mycobacterium tuberculosis. Jour-
nal of Molecular Medicine-Jmm. 2007; 85(6):613–21.
Joosten SA, Goeman JJ, Sutherland JS, Opmeer L, de Boer KG, Jacobsen M, et al. Identification of bio-
markers for tuberculosis disease using a novel dual-color RT-MLPA assay. Genes and immunity. 2012;
13(1):71–82. https://doi.org/10.1038/gene.2011.64 PMID: 21956656
Laux da Costa L, Delcroix M, Dalla Costa ER, Prestes IV, Milano M, Francis SS, et al. A real-time PCR
signature to discriminate between tuberculosis and other pulmonary diseases. Tuberculosis (Edin-
burgh, Scotland). 2015; 95(4):421–5.
Lee SW, Wu LS, Huang GM, Huang KY, Lee TY, Weng JT. Gene expression profiling identifies candi-
date biomarkers for active and latent tuberculosis. BMC bioinformatics. 2016; 17 Suppl 1:3.
Leong S, Zhao Y, Joseph NM, Hochberg NS, Sarkar S, Pleskunas J, et al. Existing blood transcriptional
classifiers accurately discriminate active tuberculosis from latent infection in individuals from south
India. Tuberculosis. 2018; 109:41–51. https://doi.org/10.1016/j.tube.2018.01.002 PMID: 29559120
Lu C, Wu J, Wang H, Wang S, Diao N, Wang F, et al. Novel biomarkers distinguishing active tuberculo-
sis from latent infection identified by gene expression profile of peripheral blood mononuclear cells.
PloS one. 2011; 6(8):e24290. https://doi.org/10.1371/journal.pone.0024290 PMID: 21904626
53. Maertzdorf J, Repsilber D, Parida SK, Stanley K, Roberts T, Black G, et al. Human gene expression pro-
files of susceptibility and resistance in tuberculosis. Genes and immunity. 2011; 12(1):15–22. https://
doi.org/10.1038/gene.2010.51 PMID: 20861863
54. Pan L, Wei N, Jia H, Gao M, Chen X, Wei R, et al. Genome-wide transcriptional profiling identifies
potential signatures in discriminating active tuberculosis from latent infection. Oncotarget. 2017; 8
(68):112907–16. https://doi.org/10.18632/oncotarget.22889 PMID: 29348876
55. Roe J, Venturini C, Gupta RK, Gurry C, Chain BM, Sun Y, et al. Blood transcriptomic stratification of
short-term risk in contacts of tuberculosis. Clinical infectious diseases: an official publication of the
Infectious Diseases Society of America. 2019.
56. Sambarey A, Devaprasad A, Mohan A, Ahmed A, Nayak S, Swaminathan S, et al. Unbiased Identifica-
tion of Blood-based Biomarkers for Pulmonary Tuberculosis by Modeling and Mining Molecular Interac-
tion Networks. EBioMedicine. 2017; 15:112–26. https://doi.org/10.1016/j.ebiom.2016.12.009 PMID:
28065665
57. Serrano CJ, Cuevas-Cordoba B, Macias-Segura N, Gonzalez-Curiel RA, Martinez-Balderas VY,
Enciso-Moreno L, et al. Transcriptional profiles discriminate patients with pulmonary tuberculosis from
non-tuberculous individuals depending on the presence of non-insulin diabetes mellitus. Clinical immu-
nology (Orlando, Fla). 2016; 162:107–17.
58. Sweeney TE, Braviak L, Tato CM, Khatri P. Genome-wide expression for diagnosis of pulmonary tuber-
culosis: a multicohort analysis. The Lancet Respiratory medicine. 2016; 4(3):213–24. https://doi.org/10.
1016/S2213-2600(16)00048-5 PMID: 26907218
59. Walter ND, Miller MA, Vasquez J, Weiner M, Chapman A, Engle M, et al. Blood Transcriptional Bio-
markers for Active Tuberculosis among Patients in the United States: a Case-Control Study with
PLOS ONE | https://doi.org/10.1371/journal.pone.0237574 August 21, 2020
16 / 17
PLOS ONE Performance of diagnostic and predictive host blood transcriptomic signatures for Tuberculosis disease
Systematic Cross-Classifier Evaluation. Journal of clinical microbiology. 2016; 54(2):274–82. https://
doi.org/10.1128/JCM.01990-15 PMID: 26582831
60. Warsinske HC, Rao AM, Moreira FMF, Santos PCP, Liu AB, Scott M, et al. Assessment of Validity of a
Blood-Based 3-Gene Signature Score for Progression and Diagnosis of Tuberculosis, Disease Sever-
ity, and Treatment Response. JAMA Netw Open. 2018; 1(6):e183779. https://doi.org/10.1001/
jamanetworkopen.2018.3779 PMID: 30646264
61. Wu B, Huang C, Kato-Maeda M, Hopewell PC, Daley CL, Krensky AM, et al. Messenger RNA expres-
sion of IL-8, FOXP3, and IL-12beta differentiates latent tuberculosis infection from disease. Journal of
immunology (Baltimore, Md: 1950). 2007; 178(6):3688–94.
62.
Lesho E, Forestiero FJ, Hirata MH, Hirata RD, Cecon L, Melo FF, et al. Transcriptional responses of
host peripheral blood cells to tuberculosis infection. Tuberculosis (Edinburgh, Scotland). 2011; 91
(5):390–9.
63. New Diagnostics Working Group (NDWG). The Roadmap to new TB diagnostics: Stop TB Partnership;
[Available from: http://www.stoptb.org/wg/new_diagnostics/objectives.asp.
64. Sumner T, Bozzani F, Mudzengi D, Hippner P, Houben RM, Cardenas V, et al. Estimating the Impact of
Tuberculosis Case Detection in Constrained Health Systems: An Example of Case-Finding in South
Africa. Am J Epidemiol. 2019; 188(6):1155–64. https://doi.org/10.1093/aje/kwz038 PMID: 30824911
65. Bossuyt PM, Cohen JF, Gatsonis CA, Korevaar DA, Grp S. STARD 2015: updated reporting guidelines
for all diagnostic accuracy studies. Ann Transl Med. 2016; 4(4).
66. Macaskill P, Gatsonis C, Deeks J, Harbord R, Takwoingi Y. Chapter 10 Analysing and Presenting
Results-Cochrane Handbook for Systematic Reviews of Diagnostic Test Accuracy Version 1.0.: The
Cochrane Collaboration; 2010 [Available from: http://srdta.cochrane.org/.
PLOS ONE | https://doi.org/10.1371/journal.pone.0237574 August 21, 2020
17 / 17
PLOS ONE |
10.1371_journal.pone.0247356 | RESEARCH ARTICLE
Development and validation of an LC-MS/MS
method for determination of
hydroxychloroquine, its two metabolites, and
azithromycin in EDTA-treated human plasma
Vong Sok, Florence Marzan, David Gingrich, Francesca Aweeka, Liusheng HuangID*
Drug Research Unit, Department of Clinical Pharmacy, School of Pharmacy, University of California at San
Francisco, San Francisco, California, United States of America
* Liusheng.huang@ucsf.edu
Abstract
Background
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: Sok V, Marzan F, Gingrich D, Aweeka F,
Huang L (2021) Development and validation of an
LC-MS/MS method for determination of
hydroxychloroquine, its two metabolites, and
azithromycin in EDTA-treated human plasma. PLoS
ONE 16(3): e0247356. https://doi.org/10.1371/
journal.pone.0247356
Editor: Pasquale Avino, Universita degli Studi del
Molise, ITALY
Received: August 27, 2020
Accepted: February 2, 2021
Published: March 5, 2021
Peer Review History: PLOS recognizes the
benefits of transparency in the peer review
process; therefore, we enable the publication of
all of the content of peer review and author
responses alongside final, published articles. The
editorial history of this article is available here:
https://doi.org/10.1371/journal.pone.0247356
Copyright: © 2021 Sok et al. This is an open access
article distributed under the terms of the Creative
Commons Attribution License, which permits
unrestricted use, distribution, and reproduction in
any medium, provided the original author and
source are credited.
Data Availability Statement: All relevant data are
within the manuscript and its Supporting
Information files.
Hydroxychloroquine (HCQ) and azithromycin (AZM) are antimalarial drugs recently reported
to be active against severe acute respiratory syndrome coronavirus- 2 (SARS-CoV-2),
which is causing the global COVID-19 pandemic. In an emergency response to the pan-
demic, we aimed to develop a quantitation method for HCQ, its metabolites desethylhydrox-
ychloroquine (DHCQ) and bisdesethylchloroquine (BDCQ), and AZM in human plasma.
Methods
Liquid chromatography tandem mass spectrometry was used to develop the method. Sam-
ples (20 μL) are extracted by solid-phase extraction and injected onto the LC-MS/MS sys-
tem equipped with a PFP column (2.0 × 50 mm, 3 μm). ESI+ and MRM are used for
detection. Ion pairs m/z 336.1!247.1 for HCQ, 308.1!179.1 for DHCQ, 264.1!179.1 for
BDCQ, and 749.6!591.6 for AZM are selected for quantification. The ion pairs m/z
342.1!253.1, 314.1!181.1, 270.1!181.1, and 754.6!596.6 are selected for the corre-
sponding deuterated internal standards (IS) HCQ-d4, DHCQ-d4, BDCQ-d4, and AZM-d5.
The less abundant IS ions from 37Cl were used to overcome the interference from the
analytes.
Results
Under optimized conditions, retention times are 0.78 min for BDCQ, 0.79 min for DHCQ,
0.92 min for HCQ and 1.87 min for AZM. Total run time is 3.5 min per sample. The calibra-
tion ranges are 2–1000 ng/mL for HCQ and AZM, 1–500 ng/mL for DHCQ and 0.5–250 ng/
mL for BDCQ; samples above the range are validated for up to 10-fold dilution. Recoveries
of the method ranged from 88.9–94.4% for HCQ, 88.6–92.9% for DHCQ, 88.7–90.9% for
BDCQ, and 98.6%-102% for AZM. The IS normalized matrix effect were within (100±10) %
for all 4 analytes. Blood samples are stable for at least 6 hr at room temperature. Plasma
PLOS ONE | https://doi.org/10.1371/journal.pone.0247356 March 5, 2021
1 / 17
PLOS ONE Determination of hydroxychloroquine, its metabolites, and azithromycin
Funding: The APC was funded by UCSF library. No
additional external funding was received for this
study.
samples are stable for at least 66 hr at room temperature, 38 days at -70˚C, and 4 freeze-
thaw cycles.
Competing interests: The authors have declared
that no competing interests exist.
Conclusions
An LC-MS/MS method for simultaneous quantitation of HCQ, DHCQ, BDCQ, and AZM in
human plasma was developed and validated for clinical studies requiring fast turnaround
time and small samples volume.
1. Introduction
The new coronavirus disease 2019 (COVID-19), caused by severe acute respiratory syndrome
coronavirus 2 (SARS-CoV-2), has evolved into a world pandemic since the first 4 cases were
reported in December 29, 2019, in Wuhan, China [1]. As of August 25, 2020, there are 23.5
million COVID-19 cases worldwide with 810,492 deaths [2], among which 5.77 million cases
with178,129 deaths are in USA [3]. There is an urgent need for effective drugs to treat
COVID-19. Initial studies found hydroxychloroquine (HCQ) active against SARS-CoV-2 [4]
and potentially useful for the treatment of COVID-19 illness clinically [5] and azithromycin
(AZM) was evaluated in combination with HCQ to treat COVID-19 [5]. However recent clini-
cal trials, lacking critical pharmacology evaluations to inform optimal dosing and requiring
drug quantitation methods, have reported substantial toxicities and contradicted results. Some
studies reported benefits [5–7] while others reported no benefits [8–10].
HCQ, primarily used previously as an antimalarial drug, has also been used for autoim-
mune diseases such as rheumatoid arthritis for several decades [11, 12]. HCQ is 50% bound to
plasma proteins, absorbed completely and rapidly(70–80% in the gastrointestinal tract) [13]
and is characterized by a long half-life(up to 40 days). Its peak concentration(Cmax) in the con-
text of multiple dosing may reach up to 1000 ng/mL [14, 15]. Hepatically, HCQ is metabolized
by cytochrome p450 (CYP) 2D6 to desethyl-chloroquine (DCQ) and desethyl-hydroxychloro-
quine (DHCQ)—both of which exhibiti activity for rheumatoid arthritis [16]; as well as bis-
desethyl-hydroxychloroquine (BDCQ), a metabolite implicated in HCQ toxicity [17, 18]. At
steady state, DHCQ in blood reaches approximately the same concentration as HCQ while
BDCQ exhibits ~1/10 of HCQ concentrations [18].
AZM is a 2nd generation macrolide antibacterial that inhibits bacterial protein synthesis
[19]. It also exhibits moderate activity against malaria and is used in combination with chloro-
quine for malaria chemoprevention [20] and treatment [21]. AZM Cmax has been reported to
be ~400 ng/mL following a 500 mg single dose [22], with higher Cmax expected following mul-
tiple doses due to its long half-life of ~70 hrs. AZM is characterized by both low oral bioavail-
ability(17–37%) and low plasma protein binding (~30%) [19]. It accumulates in tissues and
blood leukocytes.
As our group is a leading pharmacology laboratory for HIV and malaria and as part of the
wide-spread international emergency response to the outbreak of COVID-19, our laboratory
rapidly developed, validated and received approval from a NIH Division of AIDS (DAIDS)
supported quality assurance program, for a liquid chromatography tandem mass spectrometry
(LC-MS/MS) method. This method was developed to support clinical trials and to assess the
pharmacokinetics (PK) and pharmacodynamics (PD) of HCQ, DHCQ, BDCQ and AZM.
LC-MS/MS is the preferred technique for drug analysis due to its high sensitivity and selectiv-
ity. While LC-MS/MS methods to quantitate HCQ [13, 23, 24] and its metabolites [17, 18, 25]
PLOS ONE | https://doi.org/10.1371/journal.pone.0247356 March 5, 2021
2 / 17
PLOS ONE Determination of hydroxychloroquine, its metabolites, and azithromycin
have been reported, they are mainly for analyzing whole blood samples. A number of LC-MS/
MS methods have also been published to measure AZM in human plasma [26, 27]. To the best
of our knowledge, this is the first method for the simultaneous quantitation of HCQ, its metab-
olites and AZM in human plasma. Previous studies reported plasma/serum HCQ ranges from
1.0 to 2440 ng/mL with the majority of samples being between 50.0–1700 ng/mL [24] while
AZM Cmax in plasma ranges from 200 ng/mL to 1500 ng/mL depending on the dosage [19].
Therefore, the assay calibration curve ranges were tailored for 2–1000 ng/mL for both AZM
and HCQ, 1–500 ng/mL for DHCQ, and 0.5–250 ng/mL for BDCQ. This assay requires only
20 μL plasma sample volume.
2. Materials and methods
2.1. Materials
Azithromycin and hydroxychloroquine (Fig 1) were USP reference standards purchased from
Sigma-Aldrich. Desethyl-hydroxychloroquine, bisdesethyl-chloroquine, the internal standards
azithromycin-d5, hydroxychloroquine-d4, Desethyl-hydroxychloroquine-d4, and bisdesethyl-
chloroquine-d4 were purchased from Toronto Research Chemicals. Trifluoroacetic acid, ace-
tonitrile, methanol and water were purchased from Thermo-fisher (Optima LC/MS grade).
Blank human plasma and blood (K2 or K3 EDTA added as anticoagulants) was obtained from
Biological Specialty Co (Comar, PA, USA).
Fig 1. Chemical structures of azithromycin, hydroxychloroquine, desethylhydroxychoroquine,
bisdesethylchloroquine and the internal standards.
https://doi.org/10.1371/journal.pone.0247356.g001
PLOS ONE | https://doi.org/10.1371/journal.pone.0247356 March 5, 2021
3 / 17
PLOS ONE Determination of hydroxychloroquine, its metabolites, and azithromycin
2.2. Instrumentation
Sciex API5000 tandem mass spectrometer was coupled with a Shimadzu Prominence
20ADXR LC pumps and SIL-20ACXR autosampler. The LC column was Pursuit pentafluoro-
phenyl (PFP) (50×2.0 mm, 3μm) fitted with a guard column (10×2.0 mm, 3μm) (Agilent
Tech., Santa Clara, CA, USA) and eluted with water (A) and acetonitrile (B) both containing
0.05% trifluoroacetic acid (TFA) at a flow rate of 0.5 mL/min in a gradient mode: 20% solvent
B (0–0.2 min), 20 to 50% B (0.2–1.5 min), 50–90% B (1.5–1.6 min), 90 to 100% B (1.6–2.0
min), 100% B (2.0–2.5 min), 100 to 20% B (2.5–2.6 min), and 20% B (2.6–3.5 min). Electro
Spray ionization in positive mode (ESI+) was used as the ion source with multiple reaction
monitoring (MRM) of m/z 749.6!591.6 for AZM, m/z 336.1!247.1 for HCQ, m/z
308.1!179.1 for DHCQ, and m/z 264.1!179.1 for BDCQ. The ion pairs for their correspond-
ing internal standards (IS) were m/z 754.6!596.6 for AZM-d5,342.1!253.1 for HCQ-d4,
314.1!181.1 for DHCQ-d4, and 270.1!181.1 for BDCQ-d4. Samples were diverted into MS
source between 0.5–2.3 min.
2.3. Preparation of stock, calibration standards, and quality control
samples
HCQ stock solution was prepared in water. DHCQ and BDCQ stock solutions were prepared
in methanol. AZM stock solution and all working solutions were prepared in methanol-water
(1:1, v/v). Nine combined calibration standard samples at concentrations of 2, 5, 10, 20, 50,
100, 200, 500 and 1000 ng/mL for AZM and HCQ; 1, 2.5, 5, 10, 25, 50, 100, 250, and 500 ng/
mL for DHCQ; 0.5, 1.25, 2.5, 5, 12.5, 25, 50, 125, 250 ng/mL for BDCQ were prepared in blank
EDTA-treated human plasma by serial dilution from a combined working solution of AZM/
HCQ/DHCQ/BDCQ (40/40/20/10 μg/mL). Quality control (QC) samples QC-L (6.00/6.00/
3.00/1.50 ng/mL for AZM/HCQ/DHCQ/BDCQ), QC-M(60.0/60.0/30.0/15.0 ng/mL for AZM/
HCQ/DHCQ/BDCQ), and QC-H (800/800/400/200 ng/mL for AZM/HCQ/DHCQ/BDCQ)
were prepared in blank plasma from different stock solutions or the same verified stock solu-
tions as those used for calibrators. All solutions and plasma samples were stored at -70˚C
before use.
2.4. Sample preparation
Hydrophilic lipophilic balance (HLB) solid phase extraction micro-elution 96-well plates were
preconditioned with 200μL MeOH and 200μL water sequentially. Plasma samples (20 μL)
were added in the wells containing 40 μL 0.1N NaOH, Then 20 μL combined IS solution(100
ng/mL AZM-d5, 40 ng/mL HCQ-d4, 20 ng/mL DHCQ-d4, and 20 ng/mL BDCQ-d4 in 50%
MeOH) was added and mixed briefly. The wells were washed with 200 μL water, followed by
200 μL 10% MeOH under vacuum, and eluted with 25 μL MeOH containing 0.5%FA under
gradually increased vacuum. Elution was repeated for the second time with another 25 μL elu-
ent. To the collection plate, 150 μL water was added with a 12-channel pipette to constitute a
final volume of 200 μL and mixed by pipetting up and down three times. Sample was then
injected into the LC-MS/MS at 1 μL.
2.5. Validation
The method was validated in accordance with guidelines outlined by both the NIH-sponsored
Clinical Pharmacology Quality Assurance Program (CPQA) [28] and the FDA [29]. A full vali-
dation includes precision and accuracy, dilution integrity, selectivity, matrix effect and recov-
ery, and stability. Dilution integrity was evaluated by diluting the extra-high QC sample (4000/
PLOS ONE | https://doi.org/10.1371/journal.pone.0247356 March 5, 2021
4 / 17
PLOS ONE Determination of hydroxychloroquine, its metabolites, and azithromycin
4000/2000/1000 ng/mL AZM/HCQ/DHCQ/BDCQ) by 10-fold with blank plasma. Stability in
plasma was evaluated at -70˚C, room temperature and after 4 freeze-thaw cycles by comparing
the treated samples with untreated samples in plasticmicrocentrifuge tubes. To evaluate auto-
sampler stability, the processed low and high QC samples were first tested on the same day of
processing (as control) and 3 days after having been in the autosampler. Solution stability was
evaluated by diluting the solutions to within the calibration range with methanol-water (1:1, v/
v). To test stability in blood, blank blood was spiked with analytes at high QC levels and mixed
gently on a rotator for 5 min before centrifuging at 2000 g for 10 min to obtain plasma, which
was analyzed along with freshly spiked calibrators and QCs. The remaining blood was rotated
briefly and left on benchtop. A series of plasma samples was then prepared from this blood at
1hr, 2hr, 4hr, and 6hr. Each timed sample was processed and analyzed immediately following
preparation. All measurements were performed in triplicates at the minimum. Selectivity was
evaluated with 6 different lots of human plasma with K3EDTA as the anticoagulant.
Matrix effect (ME), recovery (RE) and process efficiency (PE) were evaluated with three
sets of samples: Set 1 samples were prepared by spiking 20 μL analytes in 50% MeOH solution
at QC-L, QC-M, QC-H concentrations and 20 μL IS solution (100/40/20/20ng/mL AZM-d5,
HCQ-d4, DHCQ-d4, and BDCQ-d4) into 160 μL 50% MeOH and analyzed in triplicates. Set 2
samples were spiked at the same concentration as Set 1 in extracted solutions from blank
plasma in triplicate. Set 3 samples were prepared by spiking analytes in blank plasma with final
concentrations of 6/6/3/1.5 ng/mL (QC-L), 60/60/30/15 ng/mL (QC-M), and 800/800/400/200
ng/mL (QC-H) for AZM/HCQ/DHCQ/BDCQ. These plasma samples were then processed in
triplicate using protocols as described above.
To test impact of hemolysis on plasma sample analysis, a 1 mL aliquot of whole blood
underwent 3 freeze-thaw cycles to lyse the blood cells. Fifty microliter of the treated blood was
spiked into 950 μL plasma in triplicates to give 2–3% hemolyzed plasma, which were spiked
with AZM/HCQ/DHCQ/BDCQ at QC-L and QC-H concentrations. The prepared QC-L and
QC-H were processed and analyzed along with freshly spiked calibrators and QCs.
Clinical samples are likely to be collected in tubes with K2EDTA instead of K3EDTA as the
anticoagulant. To test the impact on quantification, QC-L and QC-H were prepared in two
lots of K2EDTA plasma and one lot of K3EDTA plasma as the control. Triplicates of these sam-
ples were processed and analyzed along with a set of calibrators.
3. Results and discussion
3.1. Method development
3.1.1. LC-MS/MS optimization. The LC-MS/MS system was optimized in both APCI+
and ESI+ modes. Initially APCI+ was chosen for its less matrix effect and lower baseline signal.
However, APCI+ limited linearity to a range narrower than that desired (S1 Fig). ESI+ ion
source was finally chosen for this assay. The optimized MS/MS parameters are shown in
Table 1. LC separation parameters were adopted from a previous assay we had developed for
piperaquine [30]. It was further discovered that mobile phases 0.05% TFA in water and 0.05%
TFA in acetonitrile gave similar peak shapes and retention to those obtained using 20 mM
NH4FA 0.14% TFA in water and 0.1% TFA in acetonitrile. Therefore, the former solvent com-
bination was used due to its simplicity. The retention times are 0.78 min for BDCQ, 0.79 min
for DHCQ, 0.92 min for HCQ and 1.87 min for AZM. Total run time is 3.5 min per sample.
3.1.2. IS selection. The ideal IS for LC-MS/MS methods are stable isotopically-labelled
analytes. In this method, we were able to obtain deuterated IS. To avoid cross talks from ana-
lytes, the less abundant ion pairs from the 37Cl isotope (i.e. the most abundant ion plus 2) were
PLOS ONE | https://doi.org/10.1371/journal.pone.0247356 March 5, 2021
5 / 17
PLOS ONE Determination of hydroxychloroquine, its metabolites, and azithromycin
Table 1. Optimized MS/MS parameters.
Source parameters
Compound parameters
749.6/591.6 (AZM)
754.6/596.6 (AZM-d5)
336.1/247.1 (HCQ)
342.1/253.1 (HCQ-d4)
308.1/179.1 (DHCQ)
314.1/181.1 (DHCQ-d4)
264.1/179.1 (BDCQ)
270.1/181.1 (BDCQ-d4)
TEM,
˚C
500
DP,
v
50
50
50
50
50
50
50
50
IS,
v
1250
EP,
v
10
10
10
10
10
10
10
10
CAD,
psi
12
CE,
v
40
40
29
29
31
31
30
30
CUR,
psi
25
CXP,
V
39
39
16
16
16
16
16
16
Gas1,
psi
50
Dwell time,
Gas2,
psi
40
ms
50
50
50
50
50
50
50
50
TEM, source temperature; IS, ionspray voltage; CUR, curtain gas, Gas1, nebulizer gas; gas2, auxiliary gas; CAD, collision-activated dissociation; DP, declustering
potential; EP, entrance potential; CE, collision energy; CXP, collision cell exit potential.
https://doi.org/10.1371/journal.pone.0247356.t001
selected for HCQ-d4, DHCQ-d4 and BDCQ-d4. For AZM-d5, a higher concentration was used
to avoid cross talk from AZM especially at ULOQ level.
Selection of the appropriate concentrations for stable isotopically labelled internal standards
is based on both the number of stable isotopic atoms in the IS and the mass abundance of ana-
lytes at the IS mass levels. The IS signal originated from analyte at ULOQ should not be more
than 5%. The naturally occurring isotopic masses of the 4 analytes are calculated using an
online calculator and shown in Table 2 [31]. The percentage is based on the most abundant
mass(EM).
To avoid interference of cross talk signal from the analyte, the formula to calculate the min-
imum concentration of stable-isotope labelled internal standard is as follows [32]:
CIS;min ¼ m% �
ULOQ
5%
ð1Þ
Where m is the cross-signal percentage from analyte to IS.
To avoid interference of cross talk signal from the IS to the analyte, the formula to calculate
the maximum IS concentration is as follows:
CIS;max ¼ 20% �
LLOQ
n%
ð2Þ
Where n is the cross-signal percentage from IS to analyte.
Table 2. Natural abundances of isotopic mass of analytes and minimal IS concentration.
Exact Mass (EM)
EM+4
EM+5
EM+6
CIS,min, ng/mL
�CIS,min, ng/mL
AZM
748.51
HCQ
335.18
DHCQ
307.15
BDCQ
263.12
0.740%
0.601%
0.408%
0.052%
10.4
0.003%
148
0.6
0.002%
120
0.4
0.001%
81.6
0.2
�, 37Cl isotope used for HCQ-d4, DHCQ-d4, and BDCQ-d4.
https://doi.org/10.1371/journal.pone.0247356.t002
PLOS ONE | https://doi.org/10.1371/journal.pone.0247356 March 5, 2021
6 / 17
PLOS ONE Determination of hydroxychloroquine, its metabolites, and azithromycin
According to the equations, IS concentrations higher than 100 ng/mL are needed for HCQ,
DHCQ and BDCQ if the most abundant IS ions (EM+4) are selected. However, in order to pre-
vent the spiking of signals at LLOQ levels by highly concentrated IS, we chose the less abundant
37Cl isotope signal (EM+6) for HCQ-d4, DHCQ-d4, and BDCQ-d4. The final IS concentrations
were 100ng/mL AZM-d5, 40 ng/mL HCQ-d4, 20 ng/mL DHCQ-d4, and 20 ng/mL BDCQ-d4.
3.1.3. Sample preparation. Solid phase extraction (SPE) was used in this assay as it
yielded a cleaner extract than protein precipitation method did. HLB microelution plate was
selected in consideration ofthe small sample volumes. While the PRiME HLB plate is more
user-friendly for its omission of well preconditioning step, samples loaded into its wells would
gradually drain away, leaving little time for the user to homogenize sample with IS and
reagents. As a result, we chose the traditional HLB plate as it allows plasma samples to be
homogenized with IS and reagents in the wells prior to passing the mixture onto the stationary
phase. Methods in literature utilized liquid-liquid extraction with alkalized organic solvents
for AZM extraction [26, 33] and protein precipitation with acidified organic solvents for
extraction of HCQ and its metabolites [25]. Considering these analytes are weak bases, we
alkalized plasma samples with NaOH to help retain analytes on SPE absorbent and acidified
elution solvent to help elute analytes. The combined effect improved recovery greatly.
3.2. Method validation
3.2.1. Calibration curves. Nine combined calibration standards, prepared in K3EDTA
human plasma, consisting of 2, 5, 10, 20, 50, 100, 200, 500 and 1000 ng/mL for AZM and
HCQ; 1, 2.5, 5, 10, 25, 50, 100, 250, and 500 ng/mL for DHCQ; 0.5, 1.25, 2.5, 5, 12.5, 25, 50,
125, 250 ng/mL for BDCQ were used to establish the calibration curves. At the lower limit of
quantitation (LLOQ) (2/2/1/0.5 ng/mL AZM/HCQ/DHCQ/BDCQ) the S/N ratios were 47 for
AZM, 75 for HCQ, 31 for DHCQ, and 18 for BDCQ. The calibration curves were constructed
using concentration vs. peak area ratio fitted with least square linear regression weighted by 1/
x for HCQ, DHCQ, and BDCQ while quadratic fitting weighted by 1/x2 was needed for AZM
for better accuracy at lower concentrations. The relative error(%RE) sum, defined as the sum
of absolute %RE values, was used to evaluate the goodness of fit when using different weighting
factors for calibration curve [34]. The %RE sum of calibrators for the 4 intra/inter-day A&P
runs were the lowest for quadratic 1/x2 weighted curve compared to those of linear regression.
To compare regression models with different parameters, an effective way is to use Akaikes
information criterion (AIC) [35]. This criterion not only takes into account the sum of squares
of relative errors (SSR), it also includes a term proportional to the number of parameters used.
AIC is calculated via the formula:
AIC ¼ n � lnðSSRÞ þ 2M
ð3Þ
Where n is the number of calibrators and M is the number of parameters. The model pro-
ducing the smallest AIC is preferred. Quadratic regression weighted by 1/x2 gave the least AIC
values (Table 3).
The correlation coefficient(r) was typically >0.995. Representative chromatograms for
blank plasma, the lower limit of quantification (LLOQ) and its IS, and double blank plasma
injected after ULOQ and its IS are shown in Fig 2.
Fig 2 (red dash line) demonstrates that there are no significant analyte signal and cross-talk
from the IS (<20% of LLOQ) for all analytes. Compared to calibrator#1 [LLOQ, where peak
area = 2530 (AZM), 29900 (HCQ), 6240 (DHCQ), and 8490 (BDCQ)], the peak areas in blank
samples are minimal: 94.3 (3.7%) for AZM, 839 (2.8%) for HCQ, 364 (5.8%) for DHCQ, and
775 (9.1%) for BDCQ.
PLOS ONE | https://doi.org/10.1371/journal.pone.0247356 March 5, 2021
7 / 17
PLOS ONE Determination of hydroxychloroquine, its metabolites, and azithromycin
Table 3. Comparison of regression models.
Run ID
Linear, 1/x weighted
Linear, 1/x2 weighted
Quadratic, 1/x weighted
Quadratic, 1/x2 weighted
https://doi.org/10.1371/journal.pone.0247356.t003
%RE sum
AIC
2
46
32
22
17
8
96
78
82
60
12
82
41
36
32
17
31
25
20
22
2
60
49
48
43
8
72
64
69
43
12
67
56
54
50
17
51
45
45
45
Fig 2 (grey line) displays each analyte’s and IS’s carryover signals relative to their signal
intensities in LLOQ. The peak areas at the retention times of analytes in double blank samples
injected after ULOQ for AZM, HCQ, DHCQ, and BDCQ are 217 (8.6%), 3000 (10%), 1050
(16.8%), and 1170 (13.9%), respectively, all within 20% of LLOQ signal (Left panel), and no
peaks were found at the retention time of ISs in the chromatograms of double blank samples
after ULOQ for AZM-d5, DHCQ-d4, BDCQ-d4 (right panel). The peak for HCQ-d4 in the
double blank following ULOQ was not significant: peak area = 1250, representing 1.5% IS sig-
nal (83800). The results suggest carryover for all analytes and ISs are not significant. Further-
more, the IS signals from ULOQ of all analytes are no more than the IS signals from the
LLOQ, suggesting the cross talk from analytes are negligible (S2 Fig).
3.2.2. Intra-/inter-day precision and accuracy. Precision is the degree of reproducibility;
it characterizes the degree of agreement among a series of individual measurements. Precision
is calculated as the coefficient of variation (%CV). Accuracy is the degree of correctness and is
expressed as the percent deviation from the true concentration value. Precision and accuracy
(P&A) of method should be no more than 15% except for the lower limit of quantitation
(LLOQ), where �20% is acceptable. For inter-assay precision and accuracy, at least 3 runs
with at least 5 replicates of LLOQ, low, medium, and high concentration validation samples in
each run should be performed. These samples are designated as LLOQ, QC-L, QC-M, and
QC-H with AZM/HCQ/DHCQ/BDCQ concentrations of 2/2/1/0.5 ng/mL, 6/6/3/1.5 ng/mL,
60/60/30/15 ng/mL, and 800/800/400/200 ng/mL, respectively.
During the validation, we found the IS solutions were not stable in glass vial especially for
hydroxychloroquine and its metabolites. This resulted in unacceptable data in two inter-day
P&A runs. Further stability test between storage in glass and in plastic Eppendorf tubes
revealed all ISs lost signals in a few hours in glass vial but remained stable in Eppendorf tube
for at least overnight. AZM-d5 was reduced by ~20% depending on the container size. HCQ-
d4, DHCQ-d4 and BDCQ-d4 in 50% MeOH were reduced much more significantly (over 50%)
due to adsorption on glass surface (S1 Table).
Among the 4 runs of intra-/inter-day P&A in this report, the 1st run was performed with
freshly prepared calibrators and QCs from separately weighed stocks. In summary, intra- day
P&A meet the criteria except for one of four runs at the LLOQ for AZM (dev +25%) using 1/x
weighted linear regression calibration curve. When we reanalyzed the intra-/inter-day P&A
data for AZM using quadratic regression of 1/x2 weighted calibration curve, all 4 runs passed
the acceptance criteria.
The inter-assay precisions(%CV) of the method at low, medium, and high concentrations
are 7.5%, 6.9%, 4.9% for AZM; 7.3%, 5.3%, 6.2% for HCQ; 7.5%, 6.3%, 5.1% for DHCQ; 13%,
7.4%, 8.0% for BDCQ. The overall accuracy (%dev) from nominal low, medium and high con-
centrations are 8.0%, 5.3%, 2.2% for AZM; 0.0%, -0.4%, -1.3% for HCQ; -6.3%, -0.4%, -2.1%
for DHCQ; and -4.9%, —5.3%, -5.6% for BDCQ (Table 4).
Intra-assay precision and accuracy were calculated from 6 replicate samples of low,
medium, and high concentrations analyzed on the same day on 4 unique days. The intra-day
PLOS ONE | https://doi.org/10.1371/journal.pone.0247356 March 5, 2021
8 / 17
PLOS ONE Determination of hydroxychloroquine, its metabolites, and azithromycin
Fig 2. Representative chromatograms of blank (red dash line), the LLOQ (blue solid line) and double blank
samples following ULOQ (grey solid line). Blank sample was processed with IS, double blank sample was processed
without IS.
https://doi.org/10.1371/journal.pone.0247356.g002
precision (%CV) of this method at low, medium, and high concentrations ranges from 5.5–
11%, 4.1–7.2%, 2.0–4.2% for AZM; 3.6–8.9%, 3.1–6.6%, 3.5–6.7% for HCQ; 4.7–6.9%, 3.8–
8.5%, 3.9–6.7% for DHCQ; and 7.6–13%, 4.7–6.0%, 4.1–12% for BDCQ. Accuracy (dev%) for
low, medium and high levels ranges from 6.0–9.3%, 0.6–13%, -3.1–8.0% for AZM; -4.0–8.2%,
-2.5–4.1%, -7.5–2.7% for HCQ; -13-(-0.3)%, -5.9–2.3%, -4.5–0.5% for DHCQ; and -13-8.1%,
-11-2.4%, -7.8-(-3.3)% for BDCQ (Table 4).
LLOQ. Six replicates of validation samples at the lowest calibration concentration (2/2/1/0.5
ng/mL for AZM/HCQ/DHCQ/BDCQ) were analyzed on 4 different days to determine the
inter- and intra- assay precision and accuracy of the lowest point on the calibration curve. The
PLOS ONE | https://doi.org/10.1371/journal.pone.0247356 March 5, 2021
9 / 17
PLOS ONE Determination of hydroxychloroquine, its metabolites, and azithromycin
Table 4. Precision and accuracy.
AZM
Nominal, ng/mL
%RSD
%dev
N
2.00
7.3–12%
-19-6.1%
6
HCQ
Nominal, ng/mL
2.00
%RSD
%dev
N
DHCQ
Nominal, ng/mL
%RSD
%dev
N
BDCQ
Nominal, ng/mL
%RSD
%dev
N
5.7–12%
-3.8–19%
6
1.00
9.2–13%
-12-13%
6
0.500
9.0–14%
-11-12%
6
6.00
5.6–11%
6.0–9.3%
6
6.00
3.6–8.9%
-4.0–8.2%
6
3.00
4.7–6.9%
-13-(0.3%)
6
1.50
7.6–13%
-13-8.1%
6
Intra-day
60.0
4.1–7.2%
0.6–13%
6
Intra-day
60.0
800
1.9–3.7%
-3.1–8.0%
6
800
3.1–6.6%
-2.5–4.1%
3.5–6.7%
-7.5–2.7%
6
Intra-day
30.0
3.8–8.5%
-5.9–2.3
6
Intra-day
15.0
4.7–6.0%
-112.4%
6
6
400
3.9–6.7%
-4.5–0.5%
6
200
4.1–12%
-7.8-(3.3)%
6
2.00
13
-6.2
24
2.00
12
6.4
24
1.00
14
0.8
24
0.500
15
0.7
24
Inter-day
60.0
6.5
2.5
24
Inter-day
60.0
5.3
-0.4
24
Inter-day
30.0
6.3
-0.4
24
Inter-day
15.0
7.4
-5.3
24
6.00
7.4
5.8
24
6.00
7.3
0.0
24
3.00
7.5
-6.3
24
1.50
13
-4.9
24
800
5.4
8.9
24
800
6.2
-1.3
24
400
5.1
-2.1
24
200
8.0
-5.6
24
https://doi.org/10.1371/journal.pone.0247356.t004
inter-assay precision (%CV) is 14% for AZM, 12% for HCQ, 14% for DHCQ, and 15% for
BDCQ. The inter- assay percent deviation is -7.9% for AZM, 6.4% for HCQ, 0.8% for DHCQ,
and 0.7% for BDCQ. The intra-assay %CV for the mean of these 4 replicate days ranges from
7.3–12% for AZM, 5.7–12% for HCQ, 9.2–13% for DHCQ, and 9.0–14% for BDCQ. The mean
accuracy (%dev) ranges from -19-6.1% for AZM, -3.8–19% for HCQ, -12-13% for DHCQ, and
-11-12% for BDCQ (Table 4).
3.2.3. Dilution integrity. An extra-high QC plasma samples at a nominal concentration
of 4000 ng/mL for AZM and HCQ, 2000ng/mL for DHCQ, and 1000ng/mL for BDCQ (4
times the ULOQ concentration) were diluted with blank plasma by 10-fold. Five replicates of
the diluted samples processed and analyzed. The mean values (n = 5) were within 15% of the
nominal concentration for all analytes. The %CV was 5.3%, 9.4%, 5.5% and 3.9%; and %dev
was 14%, -7.0%, -11% and -10% for AZM, HCQ, DHCQ, and BDCQ, respectively, suggesting
the samples can be diluted by up to 10-fold without compromising sample integrity.
3.2.4. Stability. Freeze/thaw stability. QC-L and QC-H samples undergone 4 freeze-thaw
cycles were processed and analyzed along with freshly spiked calibrators and QCs. The percent
differences from freshly made controls at QC-L and QC-H concentrations are -3.7% and -10%
for AZM, -8.8% and -1.1% for HCQ, -12% and -5.3% for DHCQ, -7.1% and -4.2% for BDCQ
(S2 Table). The percent remaining analytes compared to nominal concentrations are all within
100 (±15) % (Table 5). The results reveal that AZM, HCQ, DHCQ and BDCQ in plasma are
stable after 4 freeze-thaw cycles.
Room temperature stability in plasma. After standing on the bench for 66 hours, QC-L and
QC-H samples were processed and analyzed along with freshly prepared QC-L and QC-H
samples and the freshly prepared calibrators. The % change from controls at QC-L and QC-H
concentrations was 4.3% and 6.1% for AZM, -2.3% and 0.2% for HCQ, -7.6 and -6.8% for
DHCQ, -12% and -6.3% for BDCQ, all of which were within ±15% (S2 Table). When com-
pared to spiked nominal concentrations, the remaining concentration are all within 100 (±15)
PLOS ONE | https://doi.org/10.1371/journal.pone.0247356 March 5, 2021
10 / 17
PLOS ONE Determination of hydroxychloroquine, its metabolites, and azithromycin
Table 5. Stability of AZM, HCQ, DHCQ, and BDCQ (n = 3).
Treated conditions
conc., ng/mL
% remaining
conc., ng/mL
% remaining
conc., ng/mL
% remaining
conc., ng/mL
% remaining
AZM
HCQ
DHCQ
BDCQ
38 days, -70˚C
Low
High
6.13±0.26
859±33
66hr, 22±3˚C (RT)
Low
high
5.98±0.13
801±33
Reinjection, 3 days
Low
high
5.67±0.35
845±31
4-freeze-thaw cycles
Blood, 6hr
low
high
5.14±0.36
765±23
high
878±34
Combined working solution, 50%MeOH
RT, 8 days
102
107
99.7
100
94.4
106
85.7
95.6
104
96.8
5.20±0.04
790±32
5.46±0.32
786±36
5.81±0.13
718±64
5.78±0.02
784±17
687±21
86.7
98.8
91.0
98.3
96.9
89.8
96.3
98.0
112
96.6
2.58±0.03
367±25
2.74±0.12
363±15
3.16±0.08
410±2
2.92±0.12
373±11
303±10
1.32±0.13
178±16
1.30±0.07
177±7
1.50±0.14
187±12
1.49±0.12
189±2
135±11
86.0
91.8
91.3
90.8
105
102
97.3
93.3
106
97.1
88.0
89.0
86.7
88.5
100
93.5
99.3
94.5
103
94.5
�% remaining was calculated by comparing to the nominal values (100%) for plasma and reinjection stability, comparing to the spiked blood samples at 22min for blood
stability, and comparing to the same solution frozen at -70˚C for solution room temperature (RT) stability.
https://doi.org/10.1371/journal.pone.0247356.t005
%. The results suggest that plasma samples are stable for at least 66 hours at room temperature
(Table 5).
Reinjection reproducibility/autosampler stability. To test autosampler stability, the analyzed
samples were left in the autosampler and re-injected 3 days after (71hr). The percent remain-
ing drug concentrations from nominal values at QC-L and QC-H were 94.4 and 106% for
AZM, 96.9% and 89.8% for HCQ, 105% and 102% for DHCQ, 100% and 93.5% for BDCQ,
suggesting the processed samples are stable in autosampler for at least 3 days (Table 5).
Long-term stability at -70°C of plasma samples. To test long-term stability at -70˚C, tripli-
cates of the QC-L and QC-H plasma samples stored at -70˚C for 38 days were analyzed along
with freshly spiked calibrators and QC samples as controls in triplicate. The treated samples
were all within ±15% difference from controls. The percent differences from control QC-L
and QC-H samples are 3.7% and -3.0% for AZM, 1.3% and 2.7% for HCQ, -1.9% and -0.9%
for DHCQ, and -2.0% and -5.7% for BDCQ (S2 Table). When compared to nominal values,
the treated samples were also within 100 (±15) % (Table 5). The data demonstrate plasma sam-
ples are stable at -70˚C for at least 38 days. Previous study reported AZM in plasma is stable at
-70˚ C for at least 92 days [26].Stability of blood samples at room temperature. When compared
to plasma samples separated from blood at 22 min, less than 15% difference over 6 hr was
found for all drugs, suggesting AZM, HCQ, DHCQ and BDCQ are stable in blood for at least 6
hr at room temperature (Table 5). When compared to nominal concentration, less than 15%
change was found for AZM over 6 hr at room temperature, suggesting AZM is equally distrib-
uted in blood cells and plasma. Whereas, for HCQ and its metabolites, over 20% (-20% for
HCQ, -30% for DHCQ, -35% for BDCQ) lower than nominal concentration was found at 22
min after the drugs had been spiked into blood (S2 Table), suggesting these analytes concen-
trated in blood cells. Previous studies reported that AZM concentrated in blood leukocytes
and inflamed tissues and its concentration in blood was double of that in plasma on day 3 and
PLOS ONE | https://doi.org/10.1371/journal.pone.0247356 March 5, 2021
11 / 17
PLOS ONE Determination of hydroxychloroquine, its metabolites, and azithromycin
4-fold higher after day 30 [19]–Likely because AZM is slowly distributed into blood cells.
HCQ level is also higher in blood than in plasma [36], consistent with our results.
Solution stability. AZM stock solution (2 mg/mL in 50% MeOH) was stable for at least 23
hr at room temperature and for at least 45 days at -70˚C, the % difference from the untreated
fresh controls are -1.3% and -0.1% respectively. HCQ stock (2 mg/mL in water) was stable at
-70˚C for at least 63 days and at room temperature (19–22˚C) for at least 5 days, with the %dif-
ference from fresh controls at 6.6% and 1.5%, respectively (S2 Table). The combined working
solution (40/40/20/10 μg/mL AZM/HCQ/DHCQ/BDCQ in 50% MeOH) was stable at room
temperature for at least 8 days (Table 5). Previous studies reported that AZM stock in metha-
nol is stable for 92 days at -10˚C [26]. HCQ and DHCQ stocks (0.2 mg/mL) in water are stable
at -80˚C for 12 months [18].
The working solutions for all deuterated ISs were stable at room temperature for at least 23
hr in plastic tube (S1 Table). However, due to adsorption on glass surface, the IS solution
should be prepared in plastic container. Similarly, the stock and working solutions of analytes
—especially those of HCQ, DHCQ and BDCQ should be prepared and stored in plastic con-
tainers—even though the impact of adsorption on glass surface may diminish at higher drug
concentration. For example, AZM stock solution at 0.5 mg/mL in glass vial is comparable to
those in plastic vial (<5% difference) (S2 Table).
3.2.5. Matrix effect, recovery, and process efficiency. Three sets of samples each at three
concentration levels (low, medium, and high validation concentration levels) were prepared
and analyzed to determine ME, RE, and PE. The mean peak area and peak area ratio (analyte/
IS) were calculated for each level in each set of samples and comparisons are presented in
Table 6.
ME ¼
100 � peak area of post extraction spiked sample ðset2Þ
peak area of clean sample ðset1Þ
RE ¼
100 � peak area of pre extraction spiked sample ðset3Þ
peak area of post extraction spiked sample ðset2Þ
PE ¼
100 � peak area of pre extraction spiked sample ðset3Þ
peak area of clean sample ðset1Þ
ð4Þ
ð5Þ
ð6Þ
Table 6. Matrix effect, recovery and process efficiency.
Analytes
AZM
HCQ
DHCQ
BDCQ
Conc.
(ng/ml)
Low (6)
Med (60)
High (800)
Low (6)
Med (60)
High (800)
Low (3)
Med (30)
High (400)
Low (1.5)
Med (15)
High (200)
Matrix Effect
analyte
100
98.8
97.9
110
98.0
103
102
101
106
97.1
103
101
IS
102
96.4
95.7
106
101
100
103
98.9
101
104
110
101
Recovery
analyte
102
99.4
98.6
90.5
94.4
88.9
88.6
90.3
92.9
90.9
88.7
89.0
IS
101
100
102
96.1
97.2
94.1
99.8
99.3
96.7
91.8
94.0
87.7
PE
analyte
102
98.3
96.5
99.8
92.4
91.9
90.4
91.4
98.0
88.3
91.6
90.2
IS
103
96.8
97.8
101
97.7
94.1
103
98.2
97.8
95.6
104
88.4
https://doi.org/10.1371/journal.pone.0247356.t006
PLOS ONE | https://doi.org/10.1371/journal.pone.0247356 March 5, 2021
12 / 17
PLOS ONE Determination of hydroxychloroquine, its metabolites, and azithromycin
Recovery (RE). The recovery of analytes from plasma following sample preparation was
assessed by comparing the peak areas from set 3 and set 2. The recoveries for AZM were 102,
99.4 and 98.6% at low, medium, and high concentration, respectively, and the recovery for the
IS ranged from 100–102%. The recoveries ranged from 88.9–94.4% for HCQ, 88.6–92.9% for
DHCQ, and 88.7–90.9% for BDCQ. The %CV of peak areas for recovery experiment was all
within 12%. These results suggest the assay is highly reproducible across the concentration
range with consistent and high recovery.
Matrix Effect (ME). The matrix effect of analytes from plasma following sample preparation
was assessed by comparing the peak areas from set 2 and set 1. The ME for AZM were 100,
98.8 and 97.9% at low, medium, and high concentration, respectively. ME for the IS ranged
from 95.7–102%. The ME for HCQ, DHCQ and BDCQ ranged from 98.0–110%, 101–106%,
97.1–103%, respectively. The IS normalized ME were within (100±10) % for all 4 analytes.
Process Efficiency (PE). The PE of analytes from plasma following sample preparation was
assessed by comparing the peak areas from set 3 and set 1. The PE for AZM were 102, 98.3 and
96.5% at low, medium, and high concentration, respectively. ME for the IS ranged from 96.8–
103%. The PE for HCQ, DHCQ and BDCQ ranged from 91.9–99.8%, 90.4–98.0%, 88.3–
91.6%, respectively.
3.2.6. Selectivity. To test selectivity, 6 lots of blank plasma were processed without adding
ISs and analyzed along with a LLOQ sample. The results are shown in Fig 3. significant signals
were found at the retention times of both analyte and IS for each analyte. The data suggest the
method is highly selective.
3.2.7. Impact of hemolysis. Compared to controls, the differences of hemolyzed plasma
samples at QC-L and QC-H concentrations were within 15% for all analytes (10.8 and -1.4%
for AZM, 3.0 and 0.94% for HCQ, 6.3% and -8.1% for DHCQ, -4.3 and -3.0 for BDCQ) (S2
Table). The results suggest hemolysis does not impact quantitation of the analytes in this assay.
However, since AZM and HCQ are known to present higher in blood than plasma, it is recom-
mended to avoid hemolysis when processing clinical samples.
3.2.8. Impact of anticoagulant counter ions. The differences of K2EDTA plasma samples
from K3EDTA plasma samples (controls) were all within ±15% at QC-L and QC-H concentra-
tions. For AZM, the differences from the controls at QC-L and QC-H were -1.6% and-8.0%,
-4.2% and -5.0% in the two lots of K2EDTA plasma. For HCQ, the differences at QC-L and
QC-H were 4.6% and -6.3%, 7.3% and 2.1% in the two lots of K2EDTA plasma. For DHCQ,
they are -3.9% and -13%, 9.1% and 0.1%. For BDCQ, they are-3.1% and -11%, 11% and -0.8%,
respectively (S2 Table). The data suggest different counter ions in anticoagulant won’t interfere
assay performance, and K2EDTA collection tubes could be used as an alternative for sample
collection.
4. Conclusions
A high-throughput method for simultaneous quantitation of HCQ, DHCQ, BDCQ and AZM
in plasma was developed and validated based on guidelines from FDA and NIH-sponsored
CPQA, and suitable for clinical studies of those drugs. The method required only 20 μL plasma
sample and 3.5 min run time and carryover is negligible. Although clinical data for HCQ and
AZM use in COVID-19 have resulted in poor outcomes, understanding the pharmacological
basis for toxicity is still of interest to inform any potential future use of these compounds.
Therefore, methods to quantitate these drugs reliably still remain relevant. Furthermore, these
compounds continue to be used or evaluated for other illnesses besides COVID-19 such as
malaria, inflammatory diseases, and other viral and bacterial infections [37].Such evaluations
will require state of the art analytical methods such as one described here. For it requires only
PLOS ONE | https://doi.org/10.1371/journal.pone.0247356 March 5, 2021
13 / 17
PLOS ONE Determination of hydroxychloroquine, its metabolites, and azithromycin
Fig 3. Chromatograms of six lots of blank plasma and LLOQ sample.
https://doi.org/10.1371/journal.pone.0247356.g003
a small sample volume, our method can be used for pediatric studies where sample volume is
limited. and it can be coupled with capillary tube sampling by a finger prick or more advanced
microsampling techniques such as Seventh Sense Tap™ to facilitate clinical studies. With the
highly sensitive LC-MS/MS system, our method may also be modified for dried blood spot
samples.
Supporting information
S1 Fig. Linearity of AZM calibration curve in ESI versus APCI.
(PDF)
PLOS ONE | https://doi.org/10.1371/journal.pone.0247356 March 5, 2021
14 / 17
PLOS ONE Determination of hydroxychloroquine, its metabolites, and azithromycin
S2 Fig. LLOQ versus ULOQ chromatograms.
(PDF)
S1 Table. Adsorption of internal standards on container surface.
(DOCX)
S2 Table. Stability data.
(PDF)
S1 File. Assay standard operating procedure.
(PDF)
Acknowledgments
We wish to thank supporting staff at University of California San Francisco (UCSF). We also
want to thank Difrancesco Robin and Andrew Ocque from University of New York at Buffalo,
and Lane Bushman from University of Corolado for their constructive discussion on the
method validation and thank supporting staff and reviewers for CPQA.
Author Contributions
Conceptualization: Francesca Aweeka.
Data curation: Vong Sok, Florence Marzan, David Gingrich, Liusheng Huang.
Formal analysis: Liusheng Huang.
Funding acquisition: Francesca Aweeka.
Methodology: Liusheng Huang.
Project administration: Florence Marzan.
Resources: Florence Marzan, David Gingrich.
Supervision: Francesca Aweeka.
Validation: Vong Sok, Liusheng Huang.
Writing – original draft: Vong Sok, David Gingrich, Liusheng Huang.
Writing – review & editing: Vong Sok, Florence Marzan, David Gingrich, Francesca Aweeka,
Liusheng Huang.
References
1.
Li Q, Guan X, Wu P, Wang X, Zhou L, Tong Y, et al. Early Transmission Dynamics in Wuhan, China, of
Novel Coronavirus-Infected Pneumonia. N Engl J Med. 2020; 382(13):1199–207. https://doi.org/10.
1056/NEJMoa2001316 PMID: 31995857; PubMed Central PMCID: PMC7121484.
2. WHO. Coronavirus disease (COVID-19): Situation Report–WHO COVID-19 Dashboard https://www.
who.int/emergencies/diseases/novel-coronavirus-2019/situation-reports: World Health Organization;
2020 [cited 2020 August 25].
3.
4.
Johns Hopkins CSSE. Tracking COVID-19: COVID-19 Dashboard by the Center for Systems Science
and Engineering (CSSE) at Johns Hopkins University (JHU) https://systems.jhu.edu/: John Hopkins
University Center for Systems Science and Engineering; 2020 [cited 2020 August 25]. Available from:
https://www.arcgis.com/apps/opsdashboard/index.html#/bda7594740fd40299423467b48e9ecf6.
Liu J, Cao R, Xu M, Wang X, Zhang H, Hu H, et al. Hydroxychloroquine, a less toxic derivative of chloro-
quine, is effective in inhibiting SARS-CoV-2 infection in vitro. Cell Discov. 2020; 6:16. https://doi.org/10.
1038/s41421-020-0156-0 PubMed Central PMCID: PMC7078228. PMID: 32194981
PLOS ONE | https://doi.org/10.1371/journal.pone.0247356 March 5, 2021
15 / 17
PLOS ONE Determination of hydroxychloroquine, its metabolites, and azithromycin
5. Gautret P, Lagier JC, Parola P, Hoang VT, Meddeb L, Mailhe M, et al. Hydroxychloroquine and azithro-
mycin as a treatment of COVID-19: results of an open-label non-randomized clinical trial. Int J Antimi-
crob Agents. 2020:105949. https://doi.org/10.1016/j.ijantimicag.2020.105949 PMID: 32205204;
PubMed Central PMCID: PMC7102549.
6. Arshad S, Kilgore P, Chaudhry ZS, Jacobsen G, Wang DD, Huitsing K, et al. Treatment with hydroxy-
chloroquine, azithromycin, and combination in patients hospitalized with COVID-19. Int J Infect Dis.
2020; 97:396–403. https://doi.org/10.1016/j.ijid.2020.06.099 PMID: 32623082; PubMed Central
PMCID: PMC7330574.
7. Mikami T, Miyashita H, Yamada T, Harrington M, Steinberg D, Dunn A, et al. Risk Factors for Mortality
in Patients with COVID-19 in New York City. J Gen Intern Med. 2020. https://doi.org/10.1007/s11606-
020-05983-z PMID: 32607928; PubMed Central PMCID: PMC7325642.
8. Geleris J, Sun Y, Platt J, Zucker J, Baldwin M, Hripcsak G, et al. Observational Study of Hydroxychloro-
quine in Hospitalized Patients with Covid-19. N Engl J Med. 2020; 382(25):2411–8. https://doi.org/10.
1056/NEJMoa2012410 PMID: 32379955; PubMed Central PMCID: PMC7224609.
9. Boulware DR, Pullen MF, Bangdiwala AS, Pastick KA, Lofgren SM, Okafor EC, et al. A Randomized
Trial of Hydroxychloroquine as Postexposure Prophylaxis for Covid-19. N Engl J Med. 2020; 383
(6):517–25. https://doi.org/10.1056/NEJMoa2016638 PMID: 32492293; PubMed Central PMCID:
PMC7289276.
10. Cavalcanti AB, Zampieri FG, Rosa RG, Azevedo LCP, Veiga VC, Avezum A, et al. Hydroxychloroquine
with or without Azithromycin in Mild-to-Moderate Covid-19. N Engl J Med. 2020. https://doi.org/10.1056/
NEJMoa2019014 PMID: 32706953; PubMed Central PMCID: PMC7397242.
11. Sperber K, Hom C, Chao CP, Shapiro D, Ash J. Systematic review of hydroxychloroquine use in preg-
nant patients with autoimmune diseases. Pediatr Rheumatol Online J. 2009; 7:9. https://doi.org/10.
1186/1546-0096-7-9 PMID: 19439078; PubMed Central PMCID: PMC2690583.
12. Rempenault C, Combe B, Barnetche T, Gaujoux-Viala C, Lukas C, Morel J, et al. Clinical and Structural
Efficacy of Hydroxychloroquine in Rheumatoid Arthritis: A Systematic Review. Arthritis Care Res (Hobo-
ken). 2020; 72(1):36–40. https://doi.org/10.1002/acr.23826 PMID: 30629341.
13.
Fan HW, Ma ZX, Chen J, Yang XY, Cheng JL, Li YB. Pharmacokinetics and Bioequivalence Study of
Hydroxychloroquine Sulfate Tablets in Chinese Healthy Volunteers by LC-MS/MS. Rheumatol Ther.
2015; 2(2):183–95. https://doi.org/10.1007/s40744-015-0012-0 PMID: 27747530; PubMed Central
PMCID: PMC4883261.
14. Colson P, Rolain JM, Lagier JC, Brouqui P, Raoult D. Chloroquine and hydroxychloroquine as available
weapons to fight COVID-19. Int J Antimicrob Agents. 2020:105932. https://doi.org/10.1016/j.
ijantimicag.2020.105932 PMID: 32145363.
15. Mehnert JM, Kaveney AD, Malhotra J, Spencer K, Portal D, Goodin S, et al. A phase I trial of MK-2206
and hydroxychloroquine in patients with advanced solid tumors. Cancer Chemother Pharmacol. 2019;
84(4):899–907. https://doi.org/10.1007/s00280-019-03919-x PMID: 31463691.
16. Munster T, Gibbs JP, Shen D, Baethge BA, Botstein GR, Caldwell J, et al. Hydroxychloroquine concen-
tration-response relationships in patients with rheumatoid arthritis. Arthritis Rheum-Us. 2002; 46
(6):1460–9. https://doi.org/10.1002/art.10307 WOS:000176199200006. PMID: 12115175
17. Soichot M, Megarbane B, Houze P, Chevillard L, Fonsart J, Baud FJ, et al. Development, validation and
clinical application of a LC-MS/MS method for the simultaneous quantification of hydroxychloroquine
and its active metabolites in human whole blood. J Pharm Biomed Anal. 2014; 100:131–7. https://doi.
org/10.1016/j.jpba.2014.07.009 PMID: 25165008.
18. Qu Y, Noe G, Breaud AR, Vidal M, Clarke WA, Zahr N, et al. Development and validation of a clinical
HPLC method for the quantification of hydroxychloroquine and its metabolites in whole blood. Future
Sci OA. 2015; 1(3):FSO26. https://doi.org/10.4155/fso.15.24 PMID: 28031899; PubMed Central
PMCID: PMC5137978.
19. Parnham MJ, Erakovic Haber V, Giamarellos-Bourboulis EJ, Perletti G, Verleden GM, Vos R. Azithro-
mycin: mechanisms of action and their relevance for clinical applications. Pharmacol Ther. 2014; 143
(2):225–45. https://doi.org/10.1016/j.pharmthera.2014.03.003 PMID: 24631273.
20. Chico RM, Chandramohan D. Azithromycin plus chloroquine: combination therapy for protection
against malaria and sexually transmitted infections in pregnancy. Expert Opin Drug Metab Toxicol.
2011; 7(9):1153–67. https://doi.org/10.1517/17425255.2011.598506 PMID: 21736423; PubMed Cen-
tral PMCID: PMC3170143.
21. Kshirsagar NA, Gogtay NJ, Moran D, Utz G, Sethia A, Sarkar S, et al. Treatment of adults with acute
uncomplicated malaria with azithromycin and chloroquine in India, Colombia, and Suriname. Res Rep
Trop Med. 2017; 8:85–104. https://doi.org/10.2147/RRTM.S129741 PMID: 30050349; PubMed Central
PMCID: PMC6038897.
PLOS ONE | https://doi.org/10.1371/journal.pone.0247356 March 5, 2021
16 / 17
PLOS ONE Determination of hydroxychloroquine, its metabolites, and azithromycin
22. Supattanapong S, Konsil J. Solid phase extraction and high performance liquid chromatography for the
determination of azithromycin in human plasma. Southeast Asian J Trop Med Public Health. 2008; 39
(6):978–87. PMID: 19062684.
23. Wang LZ, Ong RYL, Chin TM, Thuya WL, Wan SC, Wong ALA, et al. Method development and valida-
tion for rapid quantification of hydroxychloroquine in human blood using liquid chromatography-tandem
mass Spectrometry. J Pharmaceut Biomed. 2012; 61:86–92. https://doi.org/10.1016/j.jpba.2011.11.
034 WOS:000299861700013. PMID: 22197155
24.
Fuzery AK, Breaud AR, Emezienna N, Schools S, Clarke WA. A rapid and reliable method for the quan-
titation of hydroxychloroquine in serum using turbulent flow liquid chromatography-tandem mass spec-
trometry. Clin Chim Acta. 2013; 421:79–84. https://doi.org/10.1016/j.cca.2013.02.018 PMID:
23485647.
25. Chhonker YS, Sleightholm RL, Li J, Oupicky D, Murry DJ. Simultaneous quantitation of hydroxychloro-
quine and its metabolites in mouse blood and tissues using LC-ESI-MS/MS: An application for pharma-
cokinetic studies. J Chromatogr B. 2018; 1072:320–7. https://doi.org/10.1016/j.jchromb.2017.11.026
WOS:000423637600041. PMID: 29207305
26.
Filist M, Bus-Kwasnik K, Ksycinska H, Rudzki PJ. Simplified LC-MS/MS method enabling the determi-
nation of azithromycin in human plasma after a low 100mg dose administration. J Pharm Biomed Anal.
2014; 100:184–9. https://doi.org/10.1016/j.jpba.2014.07.015 PMID: 25168217.
27. Nirogi RVS, Kandikere VN, Shukla M, Mudigonda K, Maurya S, Boosi R, et al. Sensitive and selective
liquid chromatography-tandem mass spectrometry method for the quantification of azithromycin in
human plasma. Anal Chim Acta. 2005; 553(1–2):1–8. https://doi.org/10.1016/j.aca.2005.08.007
WOS:000233077800001.
28. CPQA. Clinical Pharmacology Quality Assurance (CPQA) Guidelines for Bioanalytical Chro-
matographic Method, Development, Validation, and Application, v1.0. Effective April 1, 2017. 2017.
Epub April 1, 2017.
29.
FDA. US Food and Drug Administration (FDA) Guidance for Industry–Bioanalytical Method Validation
(BMV), May 2018. 2018.
30. Kjellin LL, Dorsey G, Rosenthal PJ, Aweeka F, Huang L. Determination of the antimalarial drug pipera-
quine in small volume pediatric plasma samples by LC-MS/MS. Bioanalysis. 2014; 6(23):3081–9.
https://doi.org/10.4155/bio.14.254 PMID: 25529877; PubMed Central PMCID: PMC4321809.
31. Christoph Gohlke. https://www.lfd.uci.edu/~gohlke/molmass/?q=C14H18ClN3. Available from: https://
www.lfd.uci.edu/~gohlke/molmass/?q=C14H18ClN3.
32.
Tan AM, Awaiye K. Use of Internal Standards in Lc-Ms Bioanalysis. Handbook of Lc-Ms Bioanalysis:
Best Practices, Experimental Protocols, and Regulations. 2013:217–27. Book_Doi https://doi.org/10.
1002/9781118671276 WOS:000332641500018.
33. Ben-Eltriki M, Somayaji V, Padwal RS, Brocks DR. A liquid chromatography-mass spectrometric
method for the quantification of azithromycin in human plasma. Biomed Chromatogr. 2013; 27
(8):1012–7. https://doi.org/10.1002/bmc.2896 PMID: 23494651.
34. Almeida AM, Castel-Branco MM, Falcao AC. Linear regression for calibration lines revisited: weighting
schemes for bioanalytical methods. J Chromatogr B. 2002; 774(2):215–22. Pii S1570-0232(02)00244-1
https://doi.org/10.1016/s1570-0232(02)00244-1 WOS:000176996300010. PMID: 12076691
35. Kirkup L, Mulholland M. Comparison of linear and non-linear equations for univariate calibration. J Chro-
matogr A. 2004; 1029(1–2):1–11. https://doi.org/10.1016/j.chroma.2003.12.013 PMID: 15032343.
36. Brocks DR, Skeith KJ, Johnston C, Emamibafrani J, Davis P, Russell AS, et al. Hematologic disposition
of hydroxychloroquine enantiomers. J Clin Pharmacol. 1994; 34(11):1088–97. https://doi.org/10.1002/j.
1552-4604.1994.tb01986.x PMID: 7876401.
37. Sinha N, Balayla G. Hydroxychloroquine and COVID-19. Postgrad Med J. 2020; 96(1139):550–5.
https://doi.org/10.1136/postgradmedj-2020-137785 PMID: 32295814.
PLOS ONE | https://doi.org/10.1371/journal.pone.0247356 March 5, 2021
17 / 17
PLOS ONE |
10.1371_journal.pone.0243922 | RESEARCH ARTICLE
The prevalence of esophageal cancer after
caustic and pesticide ingestion: A nationwide
cohort study
Han-Wei MuID
Dong-Zong HungID
1*
1,2, Chun-Hung Chen1,2, Kai-Wei Yang1,2, Chi-Syuan Pan1,2, Cheng-Li Lin3,
1 Division of Toxicology, China Medical University Hospital, Taichung, Taiwan, 2 Department of Emergency
Medicine, China Medical University Hospital, Taichung, Taiwan, 3 Management Office of Health Data, China
Medical University Hospital, Taichung, Taiwan
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
* dzhung0224@gmail.com
Abstract
OPEN ACCESS
Citation: Mu H-W, Chen C-H, Yang K-W, Pan C-S,
Lin C-L, Hung D-Z (2020) The prevalence of
esophageal cancer after caustic and pesticide
ingestion: A nationwide cohort study. PLoS ONE
15(12): e0243922. https://doi.org/10.1371/journal.
pone.0243922
Editor: Gianluigi Forloni, Istituto Di Ricerche
Farmacologiche Mario Negri, ITALY
Received: July 5, 2020
Accepted: November 30, 2020
Published: December 29, 2020
Peer Review History: PLOS recognizes the
benefits of transparency in the peer review
process; therefore, we enable the publication of
all of the content of peer review and author
responses alongside final, published articles. The
editorial history of this article is available here:
https://doi.org/10.1371/journal.pone.0243922
Copyright: © 2020 Mu et al. This is an open access
article distributed under the terms of the Creative
Commons Attribution License, which permits
unrestricted use, distribution, and reproduction in
any medium, provided the original author and
source are credited.
Data Availability Statement: All relevant data are
within the paper and its Supporting Information
files.
Habits such as smoking and alcohol drinking and existing esophageal malfunction are con-
sidered the main risk factors for esophageal carcinogenesis. Caustic ingestion of acidic or
alkaline agents or strong irritants can induce severe esophageal corrosive injury and
increase esophageal cancer risk. We studied the relationship between esophageal carci-
noma and acute detergent or pesticide poisoning by using nationwide health insurance
data. Methodology/Principle findings: We compared a pesticide/detergent intoxication
cohort (N = 21,840) and an age- and gender-matched control cohort (N = 21,840) identified
from the National Health Insurance Research Database between 2000 and 2011. We used
the multivariable Cox proportional model to determine esophageal carcinoma risk. The over-
all incidence density of esophageal cancer was 1.66 per 10,000 person-years in the compar-
ison cohort and 4.36 per 10,000 person-years in the pesticide/detergent intoxication cohort.
The corresponding adjusted hazard ratio (HR) for esophageal cancer was 2.33 (95% confi-
dence interval [CI] = 1.41–3.86) in the pesticide/detergent intoxication cohort compared with
the control cohort. Patients with corrosive and detergent intoxication did not have a higher
risk of esophageal cancer (adjusted HR = 0.98, 95% CI = 0.29–3.33) than those without pes-
ticide/detergent intoxication. However, patients with pesticide intoxication had a significantly
higher risk of esophageal cancer (adjusted HR = 2.52, 95% CI = 1.52–4.18) than those with-
out pesticide/detergent intoxication. Conclusion: In the present study, after adjusting for
conventional risk factors, we observed that pesticide intoxication could exert substantial
effects through increased esophageal cancer risk. However, patients with detergent intoxi-
cation may not have an increased risk of esophageal cancer.
Introduction
Essentially, self-ingestion of caustic agents, detergents, and pesticides is a serious public health
problem in Taiwan. According to the Taiwan health statistics, 600 people ingested liquid tox-
ins, including caustic agents and pesticides, for suicidal attempt in 1 year, and this is the third
PLOS ONE | https://doi.org/10.1371/journal.pone.0243922 December 29, 2020
1 / 11
PLOS ONE Funding: This work was supported by grants from
the Ministry of Health and Welfare, Taiwan
(MOHW108-TDU-B-212-133004), China Medical
University Hospital (DMR-107-192, CMU107-
ASIA-19), Academia Sinica Stroke Biosignature
Project (BM10701010021), MOST Clinical Trial
Consortium for Stroke (MOST 107-2321-B-039
-004-), Tseng-Lien Lin Foundation, Taichung,
Taiwan, and Katsuzo and Kiyo Aoshima Memorial
Funds, Japan. The funders had no role in the study
design, data collection and analysis, decision to
publish, or preparation of the manuscript. No
additional external funding was received for this
study.
Competing interests: The authors have declared
that no competing interests exist.
Esophageal cancer after pesticide ingestion
common method for committing suicide. Caustic substance ingestion is most frequently
encountered in children as a result of accidental swallowing or in adults as a result of self-
harm. It often extensively injures the upper gastrointestinal tract and may lead to extensive
necrosis, perforation, and death. Among the agents for pesticide poisoning in Taiwan, organo-
phosphorus, herbicides, and other pesticides account for 45%, 23%, and 23%, respectively,
according to the admission 2009 data from Taiwan National Health Insurance Database.
Esophageal cancer accounts for >500,000 cancer deaths annually, and the incidence is rap-
idly increasing worldwide [1]. In Taiwan, 2,630 new cases of esophageal cancer and 1,792
deaths caused by esophageal cancer occurred in 2013. The mean age at occurrence was 57 in
men and 62 in women. Most esophageal cancer cases in Taiwan are of squamous cell carci-
noma (93%), and the incidence is still increasing. The risk factors for esophageal cancer are
smoking, alcohol consumption, dietary factors such as betel quid chewing and high tempera-
ture beverage consumption, gastroesophageal reflux disease, and underlying esophageal dis-
eases such as achalasia, and they are substantially different in various parts of the world [1–5].
Some studies have shown that caustic ingestion that induced severe esophageal corrosive
injury might increase esophageal cancer risk [6–12]. Some studies with limited data even esti-
mated a 1,000-fold higher risk [6]. However, these studies were based on a small number of
case control studies; hence, the evidence is not strong.
Pesticides protect plants from weeds, fungi, or insects. Pest control agents are usually
applied through chemical dispersal in a hydrocarbon solvent-surfactant system to provide a
homogeneous preparation. In addition to pesticides, these solvent-surfactants, such as the sur-
factant of glyphosate, produce significant mucosal irritation effects. Some epidemiological
studies have demonstrated high risks of certain cancers from exposure to some solvents [13].
Some pesticides are classified as carcinogenic or potentially carcinogenic to humans, such as
captafol, diazinon, malathion, and glyphosate. Here, our study investigated the relationship
between esophageal cancer and esophageal injuries after caustics ingestion and pesticide
poisoning.
Methods
Data source
This study used data from the National Health Insurance Research Database (NHIRD). The
NHIRD was launched in Taiwan in 1995 and covers nearly 99% of the total population of Tai-
wan with comprehensive healthcare benefits. For this study, we used the deidentified data of
the residents to link two data files (subsets of the NHIRD), namely inpatient claims data and
Registry of Beneficiaries. International Classification of Diseases-9-Clinical Modification
(ICD-9-CM) codes were used to define diseases in the NHIRD. This study was approved by
the Ethics Review Board of China Medical University (CMUH-104-REC2-115).
Study population
Patients with pesticide/detergent intoxication were identified from the NHIRD from January
1, 2000, to December 31, 2005, according to ICD-9-CM codes 983, 989.3–989.4, and 989.6.
Patients diagnosed with cancer (ICD-9-CM codes 140–208) before pesticide/detergent intoxi-
cation or those who lacked continuous health insurance coverage preceding cohort entry were
excluded. Furthermore, all patients aged <20 years were excluded. Moreover, the comparison
cohort of individuals without any history of pesticide/detergent intoxication was identified
from the NHIRD. The comparison cohort also excluded those with cancer history, without
health insurance before entering the study, or aged <20 years. In the final cohort, the pesti-
cide/detergent intoxication cohort was matched to the comparison cohort at a 1:1 ratio by
PLOS ONE | https://doi.org/10.1371/journal.pone.0243922 December 29, 2020
2 / 11
PLOS ONE Esophageal cancer after pesticide ingestion
gender, age, and the year of study entry. We designated 50 and 65 years as the age threshold. A
consensus is lacking regarding the age at which an individual can be considered elderly, but
the World Health Organization defines individuals >65 years as elderly in most developed
countries. In less developed countries, for example in parts of Africa, >50 years old is consid-
ered elderly. Thus, we classified participants into the age groups of <49, 50–64, and >65 to
determine the difference between each group.
The index date was defined as the date of first diagnosis of pesticide/detergent intoxication
in the database. All participants were observed until they were diagnosed with esophageal can-
cer (ICD-9-CM code 150), death, or the end of the study period (December 31, 2011).
Outcome, comorbidity, and medication
The primary clinical outcome was esophageal cancer (ICD-9-CM code 150). Furthermore,
participants in the pesticide/detergent intoxication and control cohorts were compared for
common comorbidities, including hypertension (ICD-9-CM codes 401–405), diabetes mellitus
(ICD-9-CM code 250), chronic obstructive pulmonary disease (ICD-9-CM codes 491, 492,
and 496), obesity (ICD-9-CM code 278), alcohol-related illness (ICD-9-CM codes 291, 303,
305, 571.0, 571.1, 571.2, 571.3, 790.3, A215, and V11.3), ischemic heart disease (ICD-9-CM
codes 410–414), cerebrovascular disease (ICD-9-CM codes 430–438), and gastric disease
(ICD-9-CM codes 530–534). Common comorbidities were identified according to the diagno-
sis records in the inpatient file before the index date.
Statistical analysis
We used descriptive statistics to summarize the characteristics of the pesticide/detergent intox-
ication cohort and matched comparison cohort. A continuous variable, such as age, was used
in an independent t test to examine the mean ages between the two cohorts. Categorical vari-
ables are presented as the number and percentage and included sex and common comorbidity
assessed using the chi-square test. Univariable and multivariable Cox proportional hazard
regression analyses were used to determine esophageal cancer risk, and the results are pre-
sented as hazard ratios (HRs) with 95% confidence intervals (CIs). The differences in the
cumulative incidence of esophageal cancer between the pesticide/detergent intoxication and
control cohorts were estimated using the Kaplan–Meier method with the log-rank test. A two-
tailed p value of <0.05 was considered statistically significant. We used SAS software (version
9.4 for windows; SAS Institute, Cary, NC, USA) for all statistical analyses and Kaplan–Meier
survival curve plots.
Results
This study included 21,840 patients with pesticide/detergent intoxication and 21,840 control
patients. The basic characteristics of the two cohorts are shown in Table 1. The mean ages of
the pesticide/detergent intoxication cohort and comparison cohort were 52.1 ± 17.4 and
51.6 ± 17.6, respectively. No significant difference was noted in sex and age. The majority of
pesticide/detergent intoxication patients were men (62.1%) and <49 years old (48.1%). In gen-
eral, a high proportion of pesticide/detergent intoxication patients had hypertension, diabetes
mellitus, gastric disease, ischemic heart disease, cerebrovascular disease, chronic obstructive
pulmonary disease, alcohol-related illness, and obesity (all p < 0.001). The average follow-up
duration was 5.25 ± 3.86 years for the pesticide/detergent cohort and 6.63 ± 3.29 years for the
comparison cohort. The Kaplan–Meier curve showed that the cumulative incidence of esoph-
ageal cancer was higher in the pesticide/detergent cohort than in the comparison cohort
throughout the 12-year follow-up period (Fig 1). The cumulative incidence of esophagus
PLOS ONE | https://doi.org/10.1371/journal.pone.0243922 December 29, 2020
3 / 11
PLOS ONE Table 1. Characteristics of patients with and without pesticide/detergent intoxication.
Pesticide/Detergent intoxication
Yes
(N = 21840)
No
(N = 21840)
Esophageal cancer after pesticide ingestion
Age, year
�49
50–64
� 65
Mean (SD) #
Gender
Female
Male
Comorbidity
Hypertension
Diabetes mellitus
Gastric disease
Ischemic heart disease
Cerebrovascular disease
Chronic obstructive pulmonary disease
Alcohol-related illness
Obesity
Chi-square test.
#t test.
https://doi.org/10.1371/journal.pone.0243922.t001
n
10496
5386
5958
52.1
8269
13571
3454
2136
3027
1783
1875
1044
1071
16
%
48.1
24.7
27.3
17.4
37.9
62.1
15.8
9.78
13.9
8.16
8.59
4.78
4.90
0.07
n
10496
5386
5958
51.6
8269
13571
1870
957
1057
862
934
424
112
6
%
p-value
48.1
24.7
27.3
17.6
37.9
62.1
8.56
4.38
4.84
3.95
3.82
1.94
0.51
0.03
0.99
0.004
0.99
<0.001
<0.001
<0.001
<0.001
<0.001
<0.001
<0.001
<0.001
cancer was significantly different between the pesticide/detergent and comparison cohorts
(log-rank test; p < 0.001).
The overall incidence densities of esophageal cancer were 1.66 and 4.36 per 10,000 person-
years in the comparison and pesticide/detergent cohorts, respectively (Table 2). The corre-
sponding adjusted HR for esophageal cancer was 2.33 (95% CI = 1.41–3.86) compared with
controls after adjusting for age, sex, gastric disease, and alcohol-related illness. Compared with
patients aged <49 years, those aged 50–64 and >65 years had 2.67-fold (95% CI = 1.54–4.64)
and 3.18-fold (95% CI = 1.74–5.80) significantly higher risks of esophageal cancer, respectively.
Compared with women, men had an adjusted HR of 19.8 (95% CI = 4.85–80.8) for esophagus
cancer. Among various comorbidity types, significantly increased risk was observed in those
with alcohol-related illness (adjusted HR = 7.14, 95% CI = 3.63–14.1).
Table 3 presents the incidence and HR of esophageal cancer between patients with and
without pesticide/detergent intoxication. Compared with patients without pesticide/detergent
intoxication, men, patients aged <49 years, and those aged >65 years with pesticide/detergent
intoxication had 2.22-fold (95% CI = 1.34–3.69), 2.84-fold (95% CI = 1.08–7.47), and 2.94-fold
(95% CI = 1.19–7.26) increased esophagus cancer risks, respectively. For patients without
comorbidity, those with pesticide/detergent intoxication had a significantly higher esophageal
cancer risk than those without pesticide/detergent intoxication (adjusted HR = 2.32, 95%
CI = 1.32–4.10). Among patients with non–alcohol-related illness, those with pesticide/deter-
gent intoxication had a higher risk of esophageal cancer than controls (adjusted HR = 2.47,
95% CI = 1.46–4.16).
Table 4 presents the incidence and adjusted HR of esophageal cancer between different
groups of patients with pesticide/detergent intoxication. Patients with corrosive and detergent
PLOS ONE | https://doi.org/10.1371/journal.pone.0243922 December 29, 2020
4 / 11
PLOS ONE Esophageal cancer after pesticide ingestion
Fig 1. Cummulative incidence comparison of esophagus cancer in patients with (dashed line) and without (solid line)
pesticide/detergent intoxication.
https://doi.org/10.1371/journal.pone.0243922.g001
intoxication (ICD-9-CM codes 983 and 989.6) did not have a higher risk of esophageal cancer
(adjusted HR = 0.98, 95% CI = 0.29–3.33) than those without pesticide/detergent intoxication.
Furthermore, patients with only pesticide intoxication (ICD-9-CM codes 989.3 and 989.4) had
a significantly higher risk of esophageal cancer (adjusted HR = 2.52, 95% CI = 1.52–4.18) than
those without pesticide/detergent intoxication.
Discussion
Several factors, including living habits and hobbies, contribute to esophageal cancer develop-
ment. Esophageal cancer has two major subtypes, namely squamous cell carcinoma and ade-
nocarcinoma, which have some same and different risk factors. Several genetic and epigenetic
alterations are implicated in both the development and progression of esophageal cancer.
Mucosal break, inflammation, and toxic injuries caused by excessive alcohol drinking and
heavy smoking, two of the most important and common risk factors, are causes of esophageal
carcinoma. Although the relationship between caustic ingestion and esophageal cancer and
the mechanism of esophageal cancer development are unclear, lye-based cleaner burn has
been found to complicate esophageal strictures and thus increase the risk of esophageal squa-
mous cell carcinoma [6,7,11]. Fewer case series studies have shown that the esophageal cancer
PLOS ONE | https://doi.org/10.1371/journal.pone.0243922 December 29, 2020
5 / 11
PLOS ONE Esophageal cancer after pesticide ingestion
Table 2. Incidence per 10,000 person-years of and risk factors for esophagus cancer.
Variable
Pesticide/Detergent intoxication
Event
PY
Rate#
Crude HR (95% CI)
Adjusted HR& (95% CI)
No
Yes
Age, year
�49
50–64
� 65
Gender
Female
Male
Comorbidity
Hypertension
No
Yes
Diabetes mellitus
No
Yes
Gastric disease
No
Yes
Ischemic heart disease
No
Yes
Cerebrovascular disease
No
Yes
Chronic obstructive pulmonary disease
No
Yes
Alcohol-related illness
No
Yes
Obesity
No
Yes
24
50
22
30
22
2
72
66
8
68
6
64
10
71
3
71
3
71
3
61
13
74
0
144761
114723
137357
66141
55986
100714
158769
238183
21300
247603
11880
242170
17314
248574
10910
248837
10646
254217
5266
254356
5128
259400
83
1.66
4.36
1.60
4.54
3.93
0.20
4.53
2.77
3.76
2.75
5.05
2.64
5.78
2.86
2.75
2.85
2.82
2.79
5.70
2.40
25.4
2.85
0.00
1.00
2.64(1.63, 4.30)���
1.00
2.33(1.41, 3.86)��
1.00
2.85(1.65, 4.95)���
2.55(1.41, 4.61)��
1.00
2.67(1.54, 4.64)���
3.18(1.74, 5.80)���
1.00
22.9(5.63, 93.5)���
1.00
19.8(4.85, 80.8)���
1.00
1.41(0.68, 2.95)
1.00
1.92(0.83, 4.42)
1.00
2.25(1.15, 4.38)�
1.00
0.99(0.31, 3.15)
1.00
1.02(0.32, 3.25)
1.00
2.13(0.67, 6.77)
1.00
1.00
1.00
0.83(0.40, 1.72)
1.00
1.00
1.00
1.00
10.9(5.96, 19.8)���
1.00
7.14(3.63, 14.1)���
1.00
-
1.00
Rate#: incidence rate per 10,000 person-years.
Crude HR, relative hazard ratio.
Adjusted HR&: Multivariable analysis including age, sex, gastric disease, and alcohol-related illness.
�p < 0.05
��p < 0.01
���p < 0.001.
https://doi.org/10.1371/journal.pone.0243922.t002
incidence caused by caustic ingestion is 1.4%–2.6% [6,7,14]. Although the incidence might be
overestimated, most experts agree that corrosive injury might be a risk factor for esophageal
carcinoma and have even alleged that the risk is 1,000 times that in the general population
[15]. However, the results of this study are very different from those in the literature. This
research is a nationwide, population-based cohort study designed to identify whether a
PLOS ONE | https://doi.org/10.1371/journal.pone.0243922 December 29, 2020
6 / 11
PLOS ONE Esophageal cancer after pesticide ingestion
Table 3. Incidence and hazard ratio of esophageal cancer between patients with and without pesticide/detergent intoxication.
Variables
Event
Gender
Female
Male
Age, year
�49
50–64
� 65
Comorbidity
No
Yes
Alcohol-related illness
No
Yes
0
24
6
11
7
20
4
22
2
No
PY
55515
89246
74309
36962
33489
130000
14760
144121
639
Pesticide/Detergent intoxication
Rate#
Event
0.00
2.69
0.81
2.98
2.09
1.54
2.71
1.53
31.3
2
48
16
19
15
30
20
39
11
Yes
PY
45199
69523
63048
29178
22496
90783
23939
110234
4489
Rate#
Crude HR (95% CI)
Adjusted HR&
(95% CI)
0.44
6.90
2.54
6.51
6.67
3.30
8.35
3.54
24.5
-
2.59(1.59, 4.23)���
-
2.22(1.34, 3.69)��
3.16(1.24, 8.08)�
2.23(1.06, 4.68)�
3.19(1.30, 7.83)�
2.84(1.08, 7.47)�
1.63(0.74, 3.57)
2.94(1.19, 7.26)�
2.16(1.22, 3.80)��
3.07(1.05, 8.98)�
2.32(1.32, 4.10)��
2.77(0.92, 8.31)
2.33(1.38, 3.94)���
2.47(1.46, 4.16)���
0.78(0.17, 3.53)
1.26(0.27, 5.94)
PY, person-years.
Rate#: incidence rate per 10,000 person-years.
Crude HR, relative hazard ratio.
Adjusted HR†: Multivariable analysis including age, sex, gastric disease, and alcohol-related illness.
�p < 0.05
��p < 0.01
���p < 0.001.
https://doi.org/10.1371/journal.pone.0243922.t003
significant association exists between caustic ingestion and the risk of subsequent esophageal
cancer. We defined conventional risk factors for esophageal cancer, such as age, sex, smoking,
alcohol abuse, and gastric disease (such as achalasia and GERD), which were already well-
established previously. In this 1-million-people cohort, 4,429 people were included in the
detergent and corrosive intoxication group. The relative risk of esophageal cancer did not
increase in patients with caustic agent and detergent poisoning compared with those without
the poisoning after adjustment for these conventional risk factors. One of the reasons might be
that our study included patients with exposure to detergents with less caustic characteristics.
Detergents with acidic or alkaline characteristics are some of the most used toxic and corrosive
Table 4. Incidence and adjusted hazard ratio of esophageal cancer between different entities of pesticide/detergent intoxication.
Variable
Without Pesticide/Detergent intoxication
With Organophosphate/Carbamate + Pesticide (ICD-9-CM code 989.3, 989.4)
With Detergent (ICD-9-CM code 983, 989.6)
N
21840
17411
4429
No. of Events
24
47
3
Rate#
1.66
5.31
1.14
Adjusted HR†
1.00
2.52
0.98
95% CI
(Reference)
(1.52, 4.18)
(0.29, 3.33)
PY, person-years.
Rate#, incidence rate per 10,000 person-years.
Crude HR, relative hazard ratio.
Adjusted HR†: Multivariable analysis including age, sex, gastric disease, and alcohol-related illness.
�p < 0.05, ��p < 0.01, ���p < 0.001.
https://doi.org/10.1371/journal.pone.0243922.t004
PLOS ONE | https://doi.org/10.1371/journal.pone.0243922 December 29, 2020
7 / 11
PLOS ONE Esophageal cancer after pesticide ingestion
chemicals at home. In general, detergents are classified into three categories according to their
surfactant electrical charge: nonionic, anionic, and cationic. Nonionic and anionic detergents
have low toxicity, although they may be mild to moderate irritants. Most serious toxins are cat-
ionic detergents. Most of the detergents used at home are nonionic and anionic. Therefore,
patients with ICD-9-CM codes 983.1 and 983.2 (acidic and alkali corrosive injury), 983.9
(caustic intoxication), or 989.6 (detergent intoxication) were identified, which expanded the
dataset and weakened the results. The grade of esophagus corrosive injury of these cases is not
available in the database. Thus, the true risk of esophageal cancer might be underestimated
because, theoretically, esophageal cancer commonly occurs in patients with high-grade esoph-
ageal corrosive injury. However, based on these data, the results still have considerable
credibility.
Another reason might be that the exposure interval after intoxication is shorter in this
study than in previous studies (only 12 years with an average follow-up duration of only 5
years more). The results might be different if we increased the data and extended the study
period. In previous studies, lye ingestion resulted in squamous cell carcinoma in the esophagus
rather than adenocarcinoma [16]. Despite its uncertain etiology and pathogenesis, the mecha-
nism of esophageal cancer after caustic agent and pesticide ingestion is probably similar to that
of achalasia or esophageal diverticulum. The severe injury of esophagus after caustic ingestion
causes lumen stricture or decreased esophageal motility. Subsequently, esophageal stasis
occurs, which leads to local chronic inflammatory responses in the esophageal mucosa, which
can lead to carcinogenesis. In cases of chronic irritation caused by foods and gastric fluid in
achalasia, reflux esophagitis, or Barrett’s esophagus, the interval between disease diagnosis and
esophageal carcinoma development was approximately 10–15 years [17]. However, the interval
was considered to be shortened to 4 years for patients with aforementioned diseases who were
exposed to airborne toxins that resulted from the terrorist attack of the World Trade Center
[18]. Chemical hazard exposure can accelerate solid tumor development, such as esophageal
carcinoma. In total, 287 chemicals or chemical groups with potential carcinogenic effects were
identified in the field of the World Trade Center, including several organic solvents used in
pesticide synthesis.
In this study, the relative risk of esophageal cancer increased significantly by 2.52× in the
pesticide group, and it was 2.47× even after excluding the comorbidity of alcohol-related ill-
ness. Some pesticides are considered to become carcinogenic over a long time, including their
main ingredients or organic solvents. However, such carcinogenicity was identified for most
of them after chronic exposure in in vitro, in vivo, or epidemiological studies. No study has
examined the relationship of acute large dose exposure with the occurrence of esophageal can-
cer. However, some studies have reported that esophageal cancer is positively associated with
intensive pesticide exposure. Jansson et al. found increased esophageal adenocarcinoma risk
among people with high exposure to pesticides [19]. Meyer et al. showed that esophageal can-
cer is correlated with pesticide exposure because of the high mortality caused by esophageal
cancer in states in Brazil using a high proportion of pesticides [20]. Several pesticides have
been identified as carcinogens, including their main ingredients or solvents. Animal studies
have demonstrated strong genotoxicity for some pesticides, such as diazinon organophos-
phates, malathion, and glyphosate herbicide, due to DNA and chromosomal damage. Further-
more, numerous animal studies have shown strong cellular oxidative stress reactions for them.
Glyphosate herbicide damages the retro-pharynx and esophagus more severely than other pes-
ticides and causes a high rate of morbidity among patients because of its surfactant (poly-
ethoxylated tallowamine) [21,22]. In this cohort study, a high proportion of patients in the
pesticide/detergent intoxication cohort had hypertension, diabetes mellitus, gastric disease,
ischemic heart disease, cerebrovascular disease, chronic obstructive pulmonary disease,
PLOS ONE | https://doi.org/10.1371/journal.pone.0243922 December 29, 2020
8 / 11
PLOS ONE Esophageal cancer after pesticide ingestion
alcohol-relative illness, and obesity (all p < 0.001). Single severe direct esophageal mucosa
damage as well as subsequent inflammation might be one of the causes of carcinogenicity in
these patients with chronic systemic diseases and on long-term medication, with possible mal-
function of the esophagus and stomach. However, in-depth animal experiments and studies
are required to explore the possible mechanisms of the correlation.
Our study has several limitations. The data were collected based on the ICD-9-CM codes in
the database; therefore, some detailed information could not be obtained. First, the grade of
esophageal corrosive injury after caustic ingestion is not provided in the database. This might
underestimate the true risk of esophageal cancer because, theoretically, esophageal cancer
occurs commonly in patients with high-severity esophageal corrosive injury. Second, although
ICD-9-CM codes are used for acidic and alkali corrosive injury (983.1 and 983.2), most doc-
tors in Taiwan refer such patients for the diagnosis of caustic intoxication (ICD-9-CM code
983.9) or detergent intoxication (ICD-9-CM code 989.6). It makes a huge difference in the
case numbers between these diagnoses. Therefore, we cannot evaluate esophageal cancer risk
in patients with acidic and alkaline caustic injury accurately. Third, although caustic, deter-
gent, and pesticide intoxication in Taiwan are mostly through the oral route, using a diagnostic
code to represent all oral-route intoxication could still slightly affect the results. Fourth, due to
the limitation of the ICD-9-CM diagnostic codes, we could categorize the pesticides used for
further detailed analysis. Furthermore, we were unable to extract the exact pathology reports
from the database; thus, we could not further categorize the pathologies into premalignancy
lesions, such as polyp or hyperplasia, or malignancies, such as adenocarcinoma or squamous
cell carcinoma. Fifth, because a health insurance claims database was used, detailed informa-
tion on certain general characteristics, such as obesity, body mass index, smoking, exercise,
and dietary habits, was lacking. To compensate, we tried to use clinical examination-related
morbidities to correct the individual examination index. Lastly, the present research involved
only the Taiwanese general population, which includes 99.5% Han Chinese; thus, differences
may be apparent in a stratified population.
Conclusively, to determine the association between corrosive and detergent intoxication
and esophageal cancer risk, the present study analyzed a population-based cohort from a
nationwide claims database and adjusted for comorbidities to comprehensively assess corro-
sive intoxication-related esophageal cancer risk. We observed that patients with preexisting
corrosive poisoning did not exhibit a higher esophageal cancer risk than the general popula-
tion. However, preexisting pesticide intoxication was associated with a 2.5-fold higher risk of
esophageal cancer compared with the general population. Further investigations are required
to delineate the association between esophageal carcinoma and esophageal corrosive injury or
pesticide poisoning.
Author Contributions
Conceptualization: Han-Wei Mu, Chi-Syuan Pan.
Data curation: Cheng-Li Lin.
Formal analysis: Cheng-Li Lin.
Methodology: Cheng-Li Lin.
Supervision: Chun-Hung Chen, Dong-Zong Hung.
Writing – original draft: Han-Wei Mu.
Writing – review & editing: Kai-Wei Yang, Dong-Zong Hung.
PLOS ONE | https://doi.org/10.1371/journal.pone.0243922 December 29, 2020
9 / 11
PLOS ONE Esophageal cancer after pesticide ingestion
References
1. Uhlenhopp DJ, Then EO, Sunkara T, Gaduputi V. Epidemiology of esophageal cancer: update in global
trends, etiology and risk factors. Clin J Gastroenterol. 2020. Epub 2020/09/24. https://doi.org/10.1007/
s12328-020-01237-x PMID: 32965635.
2. Enzinger PC, Mayer RJ. Esophageal cancer. N Engl J Med. 2003; 349(23):2241–52. Epub 2003/12/06.
https://doi.org/10.1056/NEJMra035010 PMID: 14657432.
3. Huang FL, Yu SJ. Esophageal cancer: Risk factors, genetic association, and treatment. Asian J Surg.
2018; 41(3):210–5. Epub 2016/12/18. https://doi.org/10.1016/j.asjsur.2016.10.005 PMID: 27986415.
4. Shaheen N, Ransohoff DF. Gastroesophageal reflux, barrett esophagus, and esophageal cancer: sci-
entific review. JAMA. 2002; 287(15):1972–81. Epub 2002/04/19. https://doi.org/10.1001/jama.287.15.
1972 PMID: 11960540.
5. Arnold M, Colquhoun A, Cook MB, Ferlay J, Forman D, Soerjomataram I. Obesity and the Incidence of
Upper Gastrointestinal Cancers: An Ecological Approach to Examine Differences across Age and Sex.
Cancer Epidemiol Biomarkers Prev. 2016; 25(1):90–7. Epub 2015/10/24. https://doi.org/10.1158/1055-
9965.EPI-15-0753 PMID: 26494763; PubMed Central PMCID: PMC4713286.
6. Kiviranta UK. Corrosion carcinoma of the esophagus; 381 cases of corrosion and nine cases of corro-
sion carcinoma. Acta Otolaryngol. 1952; 42(1–2):89–95. Epub 1952/01/01. https://doi.org/10.3109/
00016485209120330 PMID: 14932967.
7. Appelqvist P, Salmo M. Lye corrosion carcinoma of the esophagus: a review of 63 cases. Cancer.
1980; 45(10):2655–8. Epub 1980/05/15. https://doi.org/10.1002/1097-0142(19800515)45:10<2655::
aid-cncr2820451028>3.0.co;2-p PMID: 7378999.
8. Hopkins RA, Postlethwait RW. Caustic burns and carcinoma of the esophagus. Ann Surg. 1981; 194
(2):146–8. Epub 1981/08/01. https://doi.org/10.1097/00000658-198108000-00005 PMID: 7259340;
PubMed Central PMCID: PMC1345231.
9.
Ti TK. Oesophageal carcinoma associated with corrosive injury—prevention and treatment by oeso-
phageal resection. Br J Surg. 1983; 70(4):223–5. Epub 1983/04/01. https://doi.org/10.1002/bjs.
1800700413 PMID: 6831176.
10. Kim YT, Sung SW, Kim JH. Is it necessary to resect the diseased esophagus in performing reconstruc-
tion for corrosive esophageal stricture? Eur J Cardiothorac Surg. 2001; 20(1):1–6. Epub 2001/06/26.
https://doi.org/10.1016/s1010-7940(01)00747-3 PMID: 11423265.
11. Csikos M, Horvath O, Petri A, Petri I, Imre J. Late malignant transformation of chronic corrosive oeso-
phageal strictures. Langenbecks Arch Chir. 1985; 365(4):231–8. Epub 1985/01/01. https://doi.org/10.
1007/BF01459611 PMID: 4087989.
12.
Imre J, Kopp M. Arguments against long-term conservative treatment of oesophageal strictures due to
corrosive burns. Thorax. 1972; 27(5):594–8. Epub 1972/09/01. https://doi.org/10.1136/thx.27.5.594
PMID: 5083728; PubMed Central PMCID: PMC470561.
13. Matic MG, Coric VM, Savic-Radojevic AR, Bulat PV, Pljesa-Ercegovac MS, Dragicevic DP, et al. Does
occupational exposure to solvents and pesticides in association with glutathione S-transferase A1, M1,
P1, and T1 polymorphisms increase the risk of bladder cancer? The Belgrade case-control study. PLoS
One. 2014; 9(6):e99448. Epub 2014/06/11. https://doi.org/10.1371/journal.pone.0099448 PMID:
24914957; PubMed Central PMCID: PMC4051772.
14. Carver GM Jr., Sealy WC, Dillon ML Jr. Management of alkali burns of the esophagus. J Am Med
Assoc. 1956; 160(17):1447–50. Epub 1956/04/28. https://doi.org/10.1001/jama.1956.02960520009003
PMID: 13306573.
15. Contini S, Scarpignato C. Caustic injury of the upper gastrointestinal tract: a comprehensive review.
World J Gastroenterol. 2013; 19(25):3918–30. Epub 2013/07/11. https://doi.org/10.3748/wjg.v19.i25.
3918 PMID: 23840136; PubMed Central PMCID: PMC3703178.
16. Ruol A, Rampado S, Parenti A, Portale G, Giacomelli L, Battaglia G, et al. Caustic ingestion and oeso-
phageal cancer: intra- and peri-tumoral fibrosis is associated with a better prognosis. Eur J Cardiothorac
Surg. 2010; 38(6):659–64. Epub 2010/05/11. https://doi.org/10.1016/j.ejcts.2010.03.057 PMID:
20452236.
17. Meijssen MA, Tilanus HW, van Blankenstein M, Hop WC, Ong GL. Achalasia complicated by oesopha-
geal squamous cell carcinoma: a prospective study in 195 patients. Gut. 1992; 33(2):155–8. Epub
1992/02/01. https://doi.org/10.1136/gut.33.2.155 PMID: 1541408; PubMed Central PMCID:
PMC1373921.
18. Howard J. Minimum Latency & Types or Categories of Cancer: Centers for disease control and preven-
tion; 2012 [updated Jan 6, 2015; cited 2012 Oct 17]. 5 p. Available from: https://www.cdc.gov/wtc/pdfs/
policies/WTCHP-Minimum-Cancer-Latency-PP-01062015-508.pdf.
PLOS ONE | https://doi.org/10.1371/journal.pone.0243922 December 29, 2020
10 / 11
PLOS ONE Esophageal cancer after pesticide ingestion
19.
Jansson C, Plato N, Johansson AL, Nyren O, Lagergren J. Airborne occupational exposures and risk of
oesophageal and cardia adenocarcinoma. Occup Environ Med. 2006; 63(2):107–12. Epub 2006/01/20.
https://doi.org/10.1136/oem.2005.022467 PMID: 16421388; PubMed Central PMCID: PMC2078064.
20. Meyer A, Alexandre PC, Chrisman Jde R, Markowitz SB, Koifman RJ, Koifman S. Esophageal cancer
among Brazilian agricultural workers: case-control study based on death certificates. Int J Hyg Environ
Health. 2011; 214(2):151–5. Epub 2010/12/17. https://doi.org/10.1016/j.ijheh.2010.11.002 PMID:
21159552.
21. Chang CY, Peng YC, Hung DZ, Hu WH, Yang DY, Lin TJ. Clinical impact of upper gastrointestinal tract
injuries in glyphosate-surfactant oral intoxication. Hum Exp Toxicol. 1999; 18(8):475–8. Epub 1999/08/
26. https://doi.org/10.1191/096032799678847078 PMID: 10462358.
22. Hung DZ, Deng JF, Wu TC. Laryngeal survey in glyphosate intoxication: a pathophysiological investiga-
tion. Hum Exp Toxicol. 1997; 16(10):596–9. Epub 1997/11/18. https://doi.org/10.1177/
096032719701601007 PMID: 9363477.
PLOS ONE | https://doi.org/10.1371/journal.pone.0243922 December 29, 2020
11 / 11
PLOS ONE |
10.1371_journal.pone.0247289 | RESEARCH ARTICLE
Small in size, big on taste: Metabolomics
analysis of flavor compounds from Philippine
garlic
Ralph John Emerson J. MolinoID
A. Junio1*
1☯, Klidel Fae B. RellinID
1☯, Ricky B. Nellas2, Hiyas
1 Secondary Metabolites Profiling Lab, Institute of Chemistry, College of Science, University of the
Philippines, Diliman, Quezon City, Metro Manila, Philippines, 2 Virtual Biochemical Explorations Lab, Institute
of Chemistry, College of Science, University of the Philippines, Diliman, Quezon City, Metro Manila,
Philippines
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
☯ These authors contributed equally to this work.
* hajunio@up.edu.ph
Abstract
OPEN ACCESS
Citation: Molino RJEJ, Rellin KFB, Nellas RB, Junio
HA (2021) Small in size, big on taste:
Metabolomics analysis of flavor compounds from
Philippine garlic. PLoS ONE 16(5): e0247289.
https://doi.org/10.1371/journal.pone.0247289
Editor: Timothy J Garrett, University of Florida,
UNITED STATES
Received: February 3, 2021
Accepted: May 4, 2021
Published: May 20, 2021
Peer Review History: PLOS recognizes the
benefits of transparency in the peer review
process; therefore, we enable the publication of
all of the content of peer review and author
responses alongside final, published articles. The
editorial history of this article is available here:
https://doi.org/10.1371/journal.pone.0247289
Copyright: © 2021 Molino et al. This is an open
access article distributed under the terms of the
Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: All relevant data are
within the paper and its Supporting information
files.
Philippine garlic (Allium sativum L.) is arguably known to pack flavor and aroma in smaller
bulbs compared to imported varieties saturating the local market. In this study, ethanolic
extracts of Philippine garlic cultivars were profiled using ultra-high performance liquid chro-
matography-quadrupole time-of-flight mass spectrometry (UHPLC-QTOF). γ-Glu dipep-
tides, oligosaccharides and lipids were determined in Philippine garlic cultivars through
bioinformatics analysis in GNPS Molecular Networking Platform and fragmentation analy-
sis. Multivariate statistical analysis using XCMS Online showed the abundance of γ-Glu allyl
cysteine in Batanes-sourced garlic while γ-Glu propenyl cysteine, γ-Glu methyl cysteine,
and alliin are enriched in the Ilocos cultivar. Principal component analysis showed that the γ-
Glu dipeptides found in local garlic influenced their distinct separation across PC1 from
imported varieties. This presence of high levels of γ-Glu dipeptides and probiotic oligosac-
charides may potentially contribute to the superior flavor and nutritional benefits of Philippine
garlic.
Introduction
Garlic (Allium sativum L.) is a high-demand crop from the family Amaryllidaceae valued for
its nutritional and therapeutic properties [1]. A culinary staple, this bulbous spice has been
extensively studied [2] for its cardioprotective, anticancer, antidiabetic, and antimicrobial
effects often attributed to the sulfonated compound allicin, which is produced from the crush-
ing of the cloves [2]. In addition, previous reports have attributed garlic’s aroma to allicin and
its volatile degradation products [3,4]. Aside from these compounds, Ueda et al in 1990 iso-
lated and identified γ-L-glutamyl-S-(2-propenyl)-L-cysteine (γ-L-glutamyl-S-allyl-L-cysteine),
γ-L-glutam-yl-S-(2-propenyl)-L-cysteine sulfoxide, and glutathione [5]. Sensory evaluation of
these compounds showed that these peptides exhibited flavor-modifying effects that are less
PLOS ONE | https://doi.org/10.1371/journal.pone.0247289 May 20, 2021
1 / 14
PLOS ONE Funding: The authors received no specific funding
from agencies in the public, commercial, or not-
for-profit sectors for this work.
Competing interests: The authors have declared
that no competing interests exist.
Metabolomics analysis of Philippine garlic
pronounced but comparable to glutathione [6]. Furthermore, these peptides, owing to their
sulfur-containing structure, have intrinsic flavor that may be responsible for their flavor-modi-
fying effect [6].
The Philippines is known to produce several garlic cultivars that are smaller in size but
more pungent and aromatic than imports from China and neighboring Southeast Asian coun-
tries [7]. These cultivars are specifically found in the dry regions of the country such as in
Batanes Province and Ilocos Region in the north, and the Mindoro Island in the south [8].
Native garlic cultivation is the main source of livelihood for farmers in this region [9]. Interest-
ingly, a study in 2018 reports that locally produced garlic performs poorly compared to
imported white garlic varieties more commonly found in the market [10]. This setback was
attributed to the heavier and cheaper garlic produced from neighboring Asian countries that
puts local garlic farmers at a disadvantage [11]. In fact, in 2020, garlic importation dominated
the local production by 60,000 metric tons [12].
A 2019 study [13] differentiated four garlic cultivars grown in Mindoro according to their
morphological attributes. Using size classification, Batanes white cultivar was noted to be the
largest followed by Ilocos White, Lubang White, and Mindoro White. Also, the maturation of
local garlic is faster; making its harvest 30 days earlier than its imported counterparts which
are usually cultivated for at least 90 days. [14]. Moreover, there is currently limited knowledge
on the chemical constituents of locally grown garlic across the Philippines.
Liquid chromatography-mass spectrometry (LC-MS)-based metabolomics presents a sensi-
tive and high-throughput platform to study Philippine garlic cultivars [15]. Annotation of
metabolites can be performed in the level of crude extract through Data-Dependent Acquisi-
tion (DDA) [16]. Spectral analysis can be carried out in bioinformatics platforms such as
Global Natural Products Social Molecular Networking (GNPS) [17]. Furthermore, multivari-
ate statistical analysis in the cloud-based XCMS Online [18] can be employed to explore simi-
larities and differences between the native and the imported garlic cultivars in terms of their
metabolite profiles. A detailed understanding of the phytochemistry of Philippine garlic could
give insights on its sensory attributes and nutritional benefits. Furthermore, definitive compar-
ison of native and imported garlic cultivars could provide the basis in promoting cultivation
and trade of Philippine garlic.
Materials and methods
Sample preparation
Authenticated local cultivars were obtained from a farm in Ilocos Norte (ILAU samples;
18.1960˚ N, 120.5927˚ E), and from the Department of Agriculture Regional Office 2—Batanes
Experimental Station in Basco, Batanes (BAU; 20.4634˚ N, 122.0042˚ E). These samples were
authenticated based on seed or clove morphology. Native garlic was sourced from several loca-
tions: in Laoag City, Ilocos Norte (LA), and from the municipalities of San Jose (OMSJ;
12.3474˚ N, 121.0659˚ E) and Sablayan, Occidental Mindoro (OMSB; refer to S1 Figure 1 in S1
Fig). Meanwhile, garlic from Sablayan (12.8564˚ N, 120.9101˚ E), Occidental Mindoro, and in
Zamboanga City (6.9214˚ N, 122.0790˚ E) in Mindanao were ascertained by retailers to be
imported (IMPSB and ZAM respectively). Garlic samples of unverified origin were also bought
in Sablayan, Occidental Mindoro (UNKSB), and in NEPA Q-Market, Quezon City (UNKQC;
14.6178˚ N, 121.0572˚ E; refer to S1 Figure 1 in S1 Fig) and included in this study. Samples
were transported to the laboratory under ambient conditions.
Morphometric measurements involved the selection of five garlic bulbs per sampling loca-
tion as biological replicates. The length, width, and thickness of each clove as well as the num-
ber of cloves per bulb were determined (S1 Table 1 in S1 Table). The authenticated Batanes
PLOS ONE | https://doi.org/10.1371/journal.pone.0247289 May 20, 2021
2 / 14
PLOS ONE Metabolomics analysis of Philippine garlic
native cultivar has 12 or more smaller cloves, a trait shared with the native Ilocos variety. Most
imported cultivars that have been naturalized in the Philippines have bigger but fewer cloves
per bulb.
For each garlic cultivar from one location, five different bulbs were sampled which corre-
sponds to the biological replicates. Fresh garlic cloves were crushed in absolute ethanol (USP
grade, Merck1 Germany, and Scharlab SI, Spain) at a ratio of 3:10 (g/mL). Setups were left to
soak for three days at room temperature (28˚C). Further profiling experiments revealed that
prolonged soaking does not result in the degradation of metabolites (S2 Figure 1 in S2 Fig).
Extracts were dried in vacuo and resuspended in LC-MS grade methanol (Merck1 LiChro-
solv1, Germany) and the same sample mass was resuspended to a final concentration of 1.0
mg/mL for analysis.
LC-MS and MS2 analysis
Positive ion mode LC-MS analysis of garlic extracts were analyzed using a Waters1 Acquity
UPLC1 H-Class system hyphenated to a Xevo1 G2-XS Quadrupole Time-of-Flight (qTOF)
mass spectrometer with an electrospray ionization (ESI) source. Calibration of the qTOF was
done prior to runs using mass calibrants and reference standards suitable for the LockSpray™
of Waters1 MS.
Chromatography was performed through an Acquity UPLC1 CSH Fluoro Phenyl column
(1.7 μm, 50 mm long, 2.1 mm I.D.) maintained at 30˚C. Acetonitrile (Merck1 LiChrosolv1
Hypergrade, Germany) and water (Merck1 LiChrosolv1, Germany) infused with 0.1% formic
acid (Pierce1, Invitrogen, USA) were used as binary mobile phase. The flowrate was kept at a
constant rate of 0.350 mL/min with the following H2O:CH3CN elution gradient: 95:5 at 0.75
min, 75:25 at 1.0 minute, 50:50 at 2.0 minutes, 20:80 at 2.25 mins, 0:100 at 4.50 mins, and then
re-equilibrating back to 95:5 at 5.0 to 5.50 minutes. The injection volume for all samples is
1.0 μL. Each biological replicate is analyzed five times to provide technical replicate profiles for
statistical analysis.
ESI was carried out in the positive ionization mode using the following settings: capillary
voltage at 3.0 V, cone voltage at 42 V, and source offset of 80 V. Source temperature was main-
tained at 150.0˚C, and desolvation gas temperature at 500˚C. Full Scan (MS1) analyses were
done within the mass range of m/z 50.0 to 1500.0 and with a scan time of 0.50 second. These
parameters were fine-tuned using an external standard within the mass range to improve sen-
sitivity and ensure the accuracy and precision of the qTOF in mass detection.
MS2 spectra were collected with the fast DDA mode of the instrument, acquiring MS1 and
MS2 spectra at the range of m/z 50–1500, and a scan time of 0.50 second. MS2 acquisition is
initiated if ion intensity in the full scan exceeds the 3.0 x 105 threshold. Return to MS1 scanning
is triggered by the same total ion intensity of product ions, or after 0.25 seconds of MS2 acqui-
sition. A maximum of eight ions per scan were selected for MS2 analyses. The precursor ions
were subjected to collision-induced dissociation using argon curtain under fixed collision
energies of 6.0 eV, 10.0 eV, and 15.0 eV as well as collision energy ramps of 15.0–30.0 eV,
30.0–45.0 eV, and 45.0–60.0 eV. Additional monitoring involving profiling of QC samples
within two days of initial data acquisition was also performed.
Multivariate statistical analysis
Multivariate analysis of MS1 centroid data was done using XCMS Online [18]. Raw data was
converted to 64-bit open-source mzxml via ProteoWizard MSConvert [19]. GUI Sets of multi-
group analyses were performed for local authentic, local, and imported market samples as well
as with garlic of unverified geographic origins.
PLOS ONE | https://doi.org/10.1371/journal.pone.0247289 May 20, 2021
3 / 14
PLOS ONE Metabolomics analysis of Philippine garlic
Feature detection parameters include 5.0 ppm maximal tolerated m/z deviation, minimum
and maximum peak width of 2.0 and 25.0 s minimum m/z difference of 0.01, signal-to-noise
threshold of 10.0, prefilter intensity set to 1.0×105 and noise filter of 100.0. These parameters
were based on the XCMS-suggested values for Waters high-resolution data [18] and were opti-
mized for the samples in this study. The median-fold-change normalization is built in on the
XCMS workflow for peak integration.
Kruskal-Wallis non-parametric analysis was selected as the statistical test with a p-value
threshold for highly significant and significant features set to 0.01 and 0.02 respectively, with
fold change for highly significant features greater than 1.5. Non-parametric analysis was per-
formed as it does not rely on the distribution of metabolite features across the samples [20].
This is suitable for untargeted metabolomics data since the variances within the population are
heteroscedastic [20].
Annotation parameters include 5.0 ppm error and 0.005 m/z absolute error. A width of
100.0 s is considered for extracted ion chromatograms and mass calibration gaps were also
corrected. Non-metric multidimensional scaling (NMDS) and Principal Component Analysis
(PCA) of MS1 data sets were performed by XCMS Online as part of the multigroup analysis.
NMDS and PCA scores and loadings plots were replotted using Xmgrace. Parameters for the
statistical analysis were optimized according to the samples based on the initial suggested val-
ues of XCMS.
Putative compound identification and molecular networking
MS2 data were subjected to library matching in the cloud-based bioinformatics platform
Global Natural Products Social Molecular Networking (GNPS) [17]. Criteria for library
matching were set as follows: precursor ion mass tolerance of 0.45 Da, fragment ion mass tol-
erance of 0.50 Da, a minimum cosine (similarity) score of 0.70, and minimum matched peaks
of 6. Same platform was used to create molecular networks that aided in the annotation of the
metabolites. Parameters used for the generation of molecular networks include a minimum
similarity (cosine) of 0.70 calculated from the fragmentation pattern of two precursor ions, six
matched peaks, and a maximum number (top K) of 7 neighbors. Values for these analyses
were optimized according to the type of instrument and data acquired for this study. Network
visualization was customized using Cytoscape 3.7.1 [21].
Results and discussions
Metabolite profiling, putative compound identification and molecular
networking
Untargeted metabolomics of authenticated native Philippine garlic (BAU and ILAU) reveals
the presence of a broad range of compounds (Fig 1). Polar components (tR = 0.30–0.60 min)
that were found to be present were amino acids and oligosaccharides (e.g. sucrose, 1-kestose,
and stachyose). Polar dipeptides (tR = 1.30–2.00 min) were putatively characterized based on
their fragmentation pattern, and comparison with the mass spectra of annotated γ-Glu dipep-
tides and available literature information [22,23]. Non-polar region (tR = 3.00–4.00 min) rep-
resents the lipid peaks, consisting of fatty acids, phosphoethanolamine-, and phosphocholine-
type lipids. A list of all metabolites putatively annotated by GNPS is presented in Table 1.
Majority of the compounds identified matched with gold spectra in the GNPS database.
Library spectra categorized as ‘gold’ were derived from purified metabolites with nuclear mag-
netic resonance and crystallographic data [17]. Fragmentation patterns of GNPS-annotated
metabolites can be referenced by precursor ions with no library database hit. Through
PLOS ONE | https://doi.org/10.1371/journal.pone.0247289 May 20, 2021
4 / 14
PLOS ONE Metabolomics analysis of Philippine garlic
Fig 1. Base peak chromatograms of authenticated garlic samples and resulting molecular networks with putatively annotated oligosaccharides
and dipeptides. Full-scan profiles of authenticated Batanes (BAU) and Ilocos (ILAU) garlic were shown in (A). GNPS generated networks with
putatively annotated oligosaccharides and γ-Glu dipeptides are illustrated in (B) and (C), respectively.
https://doi.org/10.1371/journal.pone.0247289.g001
fragment matching and molecular networking, structural relationships of two precursor ions
can be scored for similarity with a cosine score of 0.00 to 0.99 representing no to very high sim-
ilarity, respectively [17]. However, since the DDA mode only samples highly abundant metab-
olites, molecular networking such as shown in Fig 1B and 1C should be used with caution in
deciphering the absence or presence of a metabolite in a group of samples. Nevertheless, these
generated networks are useful in providing insights pertaining to the structural relationships
PLOS ONE | https://doi.org/10.1371/journal.pone.0247289 May 20, 2021
5 / 14
PLOS ONE Metabolomics analysis of Philippine garlic
Table 1. Secondary metabolites putatively annotated from authenticated BAU and ILAU garlic through GNPS analysis.
Compound Name
Arginine
1-Kestose
Stachyose
Annotated pentaoligosaccharide
Annotated hexaoligosaccharide
Sn-glycero-3-phosphocholine
Melibiose
Alliin
γ-Glu methyl cysteine
Indole-3-lactic acid
γ-Glu allyl cysteine
γ-Glu propenyl cysteine
γ-Glu phenylalanine
13-Docosenamide, (Z)-
PE(16:0/18:2)
Beta-Sitosterol
PC(18:2/18:2)
Arachidonoyl-Thio-PC
PC(16:0/18:2)
PC(16:0/18:1)
PC(16:0/20:4)
tR (mins)
0.33
0.47
0.48
0.48
0.48
0.5
0.52
0.60
0.92
1.07
1.33
1.59
1.74
3.84
3.93
4.03
4.38
4.58
4.62
4.74
7.36
Major Ion
[M+H]+
[M+Na]+
[M+Na]+
[M+Na]+
[M+Na]+
[M+H]+
[M+Na]+
[M+H]+
[M+H]+
[M+H-H2O]+
[M+H]+
[M+H]+
[M+H]+
[M+H]+
[M+H]+
[M+H-H2O]+
[M+H]+
[M+H]+
[M+H]+
[M+H]+
[M+Na]+
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
Experimental Mass
Monoisotopic Mass
175.1205
527.1575
689.2119
851.2628
1013.3135
258.1113
365.1064
178.0540
265.0849
188.0710
291.1029
291.1029
295.1310
338.3420
716.5248
397.3829
782.5667
784.5710
758.5681
760.5866
804.5518
175.1195
527.1588
689.2116
851.2645
1013.3173
258.1106
365.1060
178.0538
265.0858
188.0711
291.1015
291.1015
295.1294
338.3423
716.5235
397.3843
782.5700
784.5679
758.5700
760.5856
804.5519
aAcceptable range for UPLC-QTOF is 3.0–5.00 ppm. Value > 5.00 ppm requires further verification.
bMN/FA: molecular network/ fragmentation analysis.
Small molecules with no cosine score were putatively characterized through molecular networking and fragmentation analysis.
https://doi.org/10.1371/journal.pone.0247289.t001
ppm Errora
5.70
Cosine Scoreb
0.89
2.48
0.39
1.93
3.72
2.52
1.14
1.12
0.91
0.81
4.92
4.92
5.43
0.86
1.84
1.34
4.19
3.99
2.48
1.28
0.16
0.92
0.83
MN/FA
MN/FA
0.85
0.84
0.74
MN/FA
0.88
MN/FA
MN/FA
0.88
0.8
0.75
0.81
0.94
0.97
0.91
0.98
0.75
among metabolites and annotation of precursor ions with no library hits. Table 1 includes
additional oligosaccharides and dipeptides present in native garlic, annotated using molecular
networking and fragmentation analysis (S3 in S3 Fig and S4 in S4 Fig).
Precursor ions m/z 851.26 and 1013.23, with no library hits in GNPS, were connected in a
network with putatively annotated oligosaccharides: melibiose, 1-kestose, and stachyose (Fig
1B). These unidentified metabolites were detected in the form of sodiated adducts and showed
losses of 162 Da for dehydrated hexose unit, and 180 Da for intact hexose unit (S3 in S3 Fig
and S4 in S4 Fig). Therefore, m/z 851.26 and 1013.23 were assigned as sodiated ions of penta-
and hexa-oligosaccharides, respectively. These oligosaccharides have been shown to have
nutritional benefits and their characterization in Philippine garlic provides information about
its dietary contribution.
Simple fructooligosaccharide 1-kestose (glu(α1!2)fru(β1!6)fru) is reported to exhibit
bifidogenic activity [24]. It also has a demonstrated positive effect on Clostridium clusters IV
and XIVa bacteria, which produces butyrate and other short-chain fatty acids essential in the
maintenance of gut homeostasis [25]. Stachyose (gal(α1!6)gal(α1!6)glc(α1$2β)fru), a pro-
biotic, has been recognized to relieve colorectal and hepatic inflammation [26].
Three γ-Glu dipeptides: γ-Glu allyl-cysteine, γ-Glu propenyl-cysteine, and γ-Glu methyl-
cysteine, were annotated based on the presence of product ions linked to the cleavage of the
amide bond [19]. Common to putatively characterized γ-Glu dipeptides is the m/z 130.05 ion
associated with glutamate. γ-Glu propenyl Cys ([M+H]+ = 291.1029, tR = 1.59 min) was specif-
ically distinguished from its positional isomer, γ-Glu allyl Cys ([M+H]+ = 291.1029, tR = 1.33
PLOS ONE | https://doi.org/10.1371/journal.pone.0247289 May 20, 2021
6 / 14
PLOS ONE Metabolomics analysis of Philippine garlic
Fig 2. MS2 spectra of isobaric features with m/z 291.10 annotated as γ-glutamyl allyl cysteine and γ-glutamyl propenyl cysteine. Spectrum A was
characterized as γ-glutamyl allyl cysteine while spectrum B is associated with its positional isomer, γ-glutamyl propenyl cysteine (tR = 1.59 mins).
https://doi.org/10.1371/journal.pone.0247289.g002
min) based on the higher intensity of the immonium ion m/z 116.05, and the presence of a
unique product ion with m/z 170.08 (Fig 2). A similar observation was described by Nakabaya-
shi et al [23] on a Fourier transform-ion cylcotron resonance analysis of γ-Glu allyl Cys and γ-
Glu propenyl Cys. In a separate study [6], fast atom bombardment-MS showed the formation
of the pyroglutamate ion at m/z 130 and its complementary ion m/z 162. γ-Glu propenyl Cys
also produces a weak dehydration ion [M+H–H2O]+ at m/z 273, together with the parent pro-
tonated molecule at m/z 291 [6], which is consistent with the results shown in Fig 2. The pro-
posed fragmentation scheme for distinguishing product ions of the isomeric dipeptides can be
found in S3 Fig 2 in S3 Fig.
These γ-Glutamyl dipeptides were initially identified, isolated, and characterized by Ueda
et al from water extracts of garlic [5,6] and were the first to propose that these compounds are
linked to the kokumi flavor [5,6], described by Yang, et. al (2019) as the continuity, thickness,
PLOS ONE | https://doi.org/10.1371/journal.pone.0247289 May 20, 2021
7 / 14
PLOS ONE Metabolomics analysis of Philippine garlic
and mouthful taste of food [27]. These compounds are known to activate oral Ca2+ sensing
receptors (CaSRs) [28]. γ-Glu methyl cysteine and γ-Glu allyl cysteine putatively characterized
in the authenticated samples have been specifically mentioned in previous studies to be flavor-
modifying compounds in garlic [29]. Additionally, γ-Glu dipeptides are antioxidants [27] and
serve as precursor metabolites to volatile S-alkenyl cysteine sulfoxides such as methiin, propiin,
and alliin which contributes to the aroma of garlic [27].
The absence of the bioactive allicin in the samples can be attributed to its high volatility and
poor ionization efficiency in both positive and negative ionization modes [29]. Mass spectrom-
etry-based detection of allicin is usually aided by complexation with transition metal ions,
such as Ag+ [29] which was not employed in the study. Moreover, allicin has been reported to
have a half life of 24 hours in 37˚C [30] implying the possibility of degradation occurring prior
to LC-MS analysis as well as probable deactivation of alliinase, the enzyme that catalyzes the
oxidation of alliin to allicin. Nevertheless, the allicin precursor, alliin, was putatively annotated
in all the samples. Alliin is also known to have antihyperlipidemic and anti-hypertensive bene-
fits [31].
Multivariate statistical analysis
Some key metabolite features annotated here through molecular networking appeared to be
dependent on sampling location (Fig 1B and 1C). Moreover, the relative abundances of metab-
olites associated with flavor and aroma could account for the difference in taste and smell
between samples [32]. To provide an unbiased (i.e. intensity thresholds needed for tandem MS
becomes negligible [17] global survey of the garlic, MS1 profiles of all samples were compared
using unsupervised multivariate statistics via XCMS Online [18]. Principal component analy-
sis (PCA) represents data in directions of maximal variance [20,33]. This means that the first
principal component (PC1; x-axis) has the highest variability followed by the second (PC2; y-
axis) and so on [20,33]. Hence, the first and second PCs are usually sufficient in describing the
data if these two PCs account for the majority of variance. To correct for the heteroskedastic
noise found in MS-generated data, statistical tests used were non-parametric and scores plots
from pairwise and multigroup setups were centered and log-transformed [34].
Shang et al., have previously described amino acid, oligosaccharides, dipeptides, and lipids
in garlic [31] but this is the first time that these compounds were putatively identified in Phil-
ippine cultivars. As selected characteristic features identified in literature, these specific metab-
olite groups will serve as basis in comparing different cultivars. Moreover, some of these small
molecules, such as sulfur-containing dipeptides, have been described in literature for their fla-
vor-modifying effects [3–6].
Authenticated garlic samples BAU and ILAU were compared pairwise to allow the explicit
comparisons of their metabolite features, and the results which were used as basis for all suc-
cessive analyses. ILAU and BAU exhibit a bimodal distribution (Fig 3A), with the scores plot
splitting into distinct regions: IB-1 (PC1: -1.0 to -2.0, PC2: 0.0 to -2.5), IB-2 (PC1: -3.5 to -4.5,
PC2: -10.0 to -12.5), and IB-3 (PC1: -7.0 to -7.5, PC2: 10.0 to -12.5). The first two regions, IB-1
and IB-2 show clusters with overlaps between ILAU and BAU, implying shared features.
Meanwhile, IB-3 is a homogeneous ILAU cluster, indicating that there are features characteris-
tic only to ILAU. In Fig 3B, the biggest contributors to the dispersion of data across PC1 and
PC2 are from the three data points that correspond to features m/z 756.6, m/z 780.6, and m/z
784.6. These features were associated with phospholipids. Hence, the difference in levels of lip-
ids within samples from the same location could influence the behavior of the group overall.
Lipids provide essential fatty acids and contribute to the food mouthfeel [35]. These com-
pounds also serve as precursors in the Maillard reaction that amplifies flavor and aroma [35].
PLOS ONE | https://doi.org/10.1371/journal.pone.0247289 May 20, 2021
8 / 14
PLOS ONE Metabolomics analysis of Philippine garlic
Fig 3. Comparison between authenticated garlic BAU and ILAU. Log-transformed PCA scores plot (A) shows that authenticated garlic BAU and
ILAU clusters based on shared features.
https://doi.org/10.1371/journal.pone.0247289.g003
Consequently, these samples are expected to vary in mouthfeel, aroma, and flavor. Features
that have masses smaller than m/z 750 contributed less to the PCs and have clustered at the
(0,0) region of the loadings plot and are attributed to dipeptides, amino acids, and
oligosaccharides.
Non-Metric Multidimensional Scaling (NMDS) biplot (Fig 3C) provides a simpler visuali-
zation of authenticated samples by representing dissimilarities as a function of distance [36].
PLOS ONE | https://doi.org/10.1371/journal.pone.0247289 May 20, 2021
9 / 14
PLOS ONE Metabolomics analysis of Philippine garlic
Fig 4. Multigroup comparison of known native garlic. Log-transformed PCA scores plot (A) shows that authentic
garlic clusters distinctly from other local samples. Mindoro garlic OMSB and OMSJ share features that separate them
from the rest of the samples. The variance observed from the samples can be attributed to primarily to metabolites such
as oligosaccharide stachyose (m/z 689.2), and other unidentified features (red) as indicated by the loadings plot (B).
https://doi.org/10.1371/journal.pone.0247289.g004
The longer distances between ILAU (black) and BAU (red) emphasize their dissimilarities, as
expected of samples from different origins. Interestingly, NMDS also shows that dissimilarities
also exist between replicates of ILAU as well as in BAU.
By explicitly zooming in at the specific metabolites, it was determined that the clustering of
samples is influenced by the levels of sulfur-containing dipeptides. An example of this dipep-
tide γ-Glu allyl cys (m/z 291.10; p-value: 5.0 ×10−15; fold change >1.5), which is present in all
local samples, but is upregulated in garlic from Ilocos (ILAU and LA; S5 Figure 6 in S5 Fig &
S6 Figure 6 in S6 Fig). Dipeptide γ-Glu methyl cys (m/z 265.10, p-value: 1.99 x 10−12, fold
change > 3.27) and the metabolite alliin (m/z 178.1, 2.66 x 10−15, fold change > 1.68) were
also upregulated in ILAU. On the other hand, γ-Glu propenyl cys (p-value: 9.18 x 10−16, fold
change > 2.24; S5 Figure 5 in S5 Fig) is highly abundant in BAU compared to ILAU. These
dipeptides contribute to flavors derived enzymatically in Allium species [37]. The presence of
these compounds at relatively higher levels in authenticated samples could potentially explain
the stronger sensory impression when compared to other cultivars [3,32].
Authenticated cultivars versus local market (Fig 4A) show four distinct regions: R1 (PC1:
1.0 to 4.5, PC2: 2.5 to 5.0), R2 (PC1: -1.0 to -5.0, PC2: 2.0 to -2.0), R3 (PC1: -7.0 to -8.5, PC2:
-7.5 to -12.5), and R4 (PC1: -8.5 to -10.0, PC2: -15.0 to -18.0). The ILAU and BAU cluster is
represented by R1 while LA, OMSB, and OMSJ gather in R2. R3 and R4 solely describe sam-
ples sourced from Mindoro (OMSB and OMSJ). R1 denotes that market samples (R2, R3, and
R4) do not share the same characteristics as the authenticated garlic. One of the biggest con-
tributors to the spread of the data is the oligosaccharide stachyose (m/z 689; Fig 4B). Stachyose
and 1-kestose were specifically upregulated in LA but downregulated in ILAU and BAU (S5
Figures 2 and 3 in S5 Fig). Oligosaccharides also impart the caramelized notes in food through
the Maillard reaction [38]. Furthermore, galacto- and fructooligosaccharides have shown to
have a probiotic effect to the human gut microbiota [39] and their characterization in Philip-
pine garlic could provide additional nutritional benefits.
Based on the insights obtained from exploring authenticated and local market samples, a
comparison between native and imported garlic (Fig 5A) showed a clear separation across PC1
and PC2. Clustering among native samples is primarily due to higher levels of γ-Glu dipeptides
and amino acids, which may be responsible for the enhanced flavor of Philippine garlic [3,32].
Interestingly, some local biological replicates from San Jose (OMSJ) clustered with imported
ZAM and IMPSB, which may imply that OMSB outliers have features in common with ZAM
PLOS ONE | https://doi.org/10.1371/journal.pone.0247289 May 20, 2021
10 / 14
PLOS ONE Metabolomics analysis of Philippine garlic
Fig 5. Scores plot (log, centered) comparing local and imported samples. Sugars and dipeptides, which influence
clustering, are distributed in both local and imported samples.
https://doi.org/10.1371/journal.pone.0247289.g005
and IMSB that have a greater loadings weight. Highly abundant metabolites could also swamp
out and mask the significant yet less abundant features [34].
From the previous analyses in Figs 3 and 4, identifying the geographical origin of unknown
samples becomes possible as native garlic, whether authenticated or market, have distinguish-
ing hallmarks. In Fig 5B, UNKSB and UNKQC were compared to samples of known sources.
UNKSB clustered with the imported varieties while UNKQC shared metabolite features with
that of the local cultivars. These results signify unsupervised PCA of MS data could be used to
determine the origin of samples by comparing their profiles to that of authenticated local and
imported garlic.
Through multivariate analysis, it was inferred that the perceived stronger taste of Philippine
garlic cultivars, particularly those sourced from Ilocos and Batanes, is potentially due to the
higher levels of flavor compounds such as sulfur-containing dipeptides present in native sam-
ples compared to the imports. The secondary metabolite profiles of Philippine garlic estab-
lished in this study provide a basis for its culinary and nutritional value and its molecular
signatures can be used to identify geographical origin of garlic.
Conclusions
Metabolomics and chemometric studies via LC-MS enabled the putative identification of
inherent compounds from different Philippine native garlic cultivars as well as in the imported
varieties studied here. These are sulfur-containing dipeptides (γ-Glu-Phe, γ-Glu-methyl-cys
derivatives), functional oligosaccharides (1-kestose, stachyose, melibiose), lipids and plant hor-
mones (β-sitosterol, indole-3-lactic acid), and volatile molecules (alliin)—all of which contrib-
ute to the complex aroma and flavor profile of garlic and have been validated in literature
[1,24]. Some of these compounds, particularly amino acids, and sugars, were detected at higher
intensities from native garlic cultivars and can be correlated to the enhanced aroma and flavor
profile of native garlic. PCA and NMDS highlighted similarities arising from the contributions
of shared dipeptides and oligosaccharides between garlic grown in the Philippines. These
metabolites influenced the clustering of local garlic and separated them from imported sam-
ples. This study is a pioneering work on comparative LC-MS metabolomic and chemometric
analyses for a high-value crop cultivated in and imported into the Philippines.
PLOS ONE | https://doi.org/10.1371/journal.pone.0247289 May 20, 2021
11 / 14
PLOS ONE Supporting information
Metabolomics analysis of Philippine garlic
S1 Fig.
(PDF)
S2 Fig.
(PDF)
S3 Fig.
(PDF)
S4 Fig.
(PDF)
S5 Fig.
(PDF)
S6 Fig.
(PDF)
S7 Fig.
(PDF)
S8 Fig.
(PDF)
S9 Fig.
(PDF)
S1 Table. Morphometric measurements of native and imported Allium sativum cultivars.
(PDF)
Acknowledgments
The authors would like to acknowledge the Department of Agriculture Regional Office 02
(DA-RO2)—Batanes Experimental Station through the manager, Mr. Celso Batallones, Mrs.
Edelina Rellin (DA-RO2 Philippine Carabao Center) and Mrs. Shirley Abucay for the garlic
samples. The authors would also like to thank the Department of Science and Technology-
funded Discovery and Development of Health Products Program for the LC-MS facility of the
Institute of Chemistry, University of the Philippines, Diliman, and to Mr. Jokent Gaza for his
assistance in the computational aspect of this study.
Author Contributions
Conceptualization: Hiyas A. Junio.
Data curation: Ralph John Emerson J. Molino, Klidel Fae B. Rellin, Hiyas A. Junio.
Formal analysis: Ralph John Emerson J. Molino, Klidel Fae B. Rellin, Ricky B. Nellas, Hiyas A.
Junio.
Supervision: Ricky B. Nellas, Hiyas A. Junio.
Validation: Ralph John Emerson J. Molino, Klidel Fae B. Rellin, Ricky B. Nellas, Hiyas A.
Junio.
Visualization: Klidel Fae B. Rellin.
PLOS ONE | https://doi.org/10.1371/journal.pone.0247289 May 20, 2021
12 / 14
PLOS ONE Metabolomics analysis of Philippine garlic
Writing – original draft: Ralph John Emerson J. Molino, Klidel Fae B. Rellin.
Writing – review & editing: Ralph John Emerson J. Molino, Klidel Fae B. Rellin, Ricky B. Nel-
las, Hiyas A. Junio.
References
1. Batiha GE, Beshbishy AM, Wasef LG, Elewa YH, Al-Sagan A, Abd El-Hack ME, et al. Chemical Constit-
uents and Pharmacological Activities of Garlic (Allium sativum L.): A Review. Nutrients. 2020; 12
(3):872.
2. Salehi B, Zucca P, Orhan IE, Azzini E, Adetunji CO, Mohammed SA, et al. Allicin and health: A compre-
hensive review. Trends in Food Science & Technology. 2019; 86:502–16.
3. Abe K, Hori Y, Myoda T. Volatile compounds of fresh and processed garlic (Review). Experimental and
Therapeutic Medicine. 2019.
4. Keleş D, Taşkın H, Baktemur G, Kafkas E, Bu¨ yu¨ kalaca S. Comparitive Study on Volatile Aroma Com-
pounds of Two Different Garlic Types (Kastamonu And Chinese) Using Gas Chromatography Mass
Spectrometry (Hs-Gc/Ms) Technique. African Journal of Traditional, Complementary and Alternative
Medicines. 2014; 11(3):217.
5. Ueda Y, Sakaguchi M, Hirayama K, Miyajima R, Kimizuka A. Characteristic Flavor Constituents in
Water Extract of Garlic. Agricultural and Biological Chemistry. 1990; 54(1):163–9.
6. Amino Y, Wakabayashi H, Akashi S, Ishiwatari Y. Structural analysis and taste evaluation of γ-glutamyl
peptides comprising sulfur-containing amino acids. Bioscience, Biotechnology, and Biochemistry.
2018; 82(3):383–94. https://doi.org/10.1080/09168451.2018.1436433 PMID: 29488453
7. Micua L. Garlic, ’white gold’ of Ilocos Region [Internet]. Philippine News Agency RSS. Philippine News
Agency; 2017 [cited 2020 July 6]. https://www.pna.gov.ph/articles/1006060.
8. Andres R. DA, state university to put up onion, garlic research hub [Internet]. Philippine News Agency
RSS. Philippine News Agency; 2020 [cited 2020 July 6]. https://www.pna.gov.ph/articles/1094773.
9. Bureau of Agricultural Statistics, Philippine Statistics Authority. 2013 Costs and Returns of Garlic Pro-
duction. [cited 2020 Jul 6] https://psa.gov.ph/content/costs-and-returns-garlic-production.
10. Philippine Statistics Authority. Agricultural Exports and Imports. Philippine Statistics Authority. [cited
2020 Jul 6]. https://psa.gov.ph/sites/default/files/ais_Agri_exports%20%26%20imports%202018.pdf.
11. Audio T. DA seeks market for Mindoro garlic [Internet]. INQUIRER.net. Philippine Daily Inquirer; 2019
[cited 2020 Jul 6]. https://newsinfo.inquirer.net/1132588/da-seeks-market-for-mindoro-garlic.
12. Simeon LM. Government tightens monitoring of garlic, onion [Internet]. Philstar.com. Philstar.com;
2020 [cited 2021Apr28]. https://www.philstar.com/business/agriculture/2020/01/19/1985830/
government-tightens-monitoring-garlic-onion.
13. Ragas R, Padron FK, Ruedas MY. Analysis of the morpho-anatomical traits of four major garlic (Allium
sativum L.) cultivars in the Philippines. Applied Ecology and Environmental Research. 2019; 17
(1):1143–57.
14. Bureau of Plant Industry, Department of Agriculture. Garlic Production Guide. Department of Agricul-
ture; 2011. [cited 2020 Jul 6] Accessed from: http://bpi.da.gov.ph/bpi/images/Production_guide/pdf/
Garlic.pdf.
15.
Zhou B, Xiao JF, Tuli L, Ressom HW. LC-MS-based metabolomics. Molecular BioSystems. The Royal
Society of Chemistry. 2012; 8(2):470–81. https://doi.org/10.1039/c1mb05350g PMID: 22041788
16. Vidova V, Spacil Z. A review on mass spectrometry-based quantitative proteomics: Targeted and data
independent acquisition. Analytica Chimica Acta. 2017; 964:7–23. https://doi.org/10.1016/j.aca.2017.
01.059 PMID: 28351641
17. Wang M, Carver JJ, Phelan VV, Sanchez LM, Garg N, Peng Y, et al. Sharing and community curation
of mass spectrometry data with Global Natural Products Social Molecular Networking. Nature Biotech-
nology. 2016; 34(8):828–37. https://doi.org/10.1038/nbt.3597 PMID: 27504778
18.
Tautenhahn R, Patti GJ, Rinehart D, Siuzdak G. XCMS Online: A Web-Based Platform to Process
Untargeted Metabolomic Data. Analytical Chemistry. 2012; 84(11):5035–9. https://doi.org/10.1021/
ac300698c PMID: 22533540
19. Chambers MC, Maclean B, Burke R, Amodei D, Ruderman DL, Neumann S, et al. A cross-platform
toolkit for mass spectrometry and proteomics. Nature Biotechnology. 2012; 30(10):918–20. https://doi.
org/10.1038/nbt.2377 PMID: 23051804
20. Gauthier TD, Hawley ME. Statistical Methods. Introduction to Environmental Forensics. 2015;:99–148.
PLOS ONE | https://doi.org/10.1371/journal.pone.0247289 May 20, 2021
13 / 14
PLOS ONE Metabolomics analysis of Philippine garlic
21. Shannon P. Cytoscape: A Software Environment for Integrated Models of Biomolecular Interaction Net-
works. Genome Research. 2003; 13(11):2498–504. https://doi.org/10.1101/gr.1239303 PMID:
14597658
22. Paizs B, Suhai S. Fragmentation pathways of protonated peptides. Mass Spectrometry Reviews. 2005;
24(4):508–48. https://doi.org/10.1002/mas.20024 PMID: 15389847
23. Nakabayashi R, Sawada Y, Aoyagi M, Yamada Y, Hirai MY, Sakurai T, et al. Chemical Assignment of
Structural Isomers of Sulfur-Containing Metabolites in Garlic by Liquid Chromatography–Fourier Trans-
form Ion Cyclotron Resonance–Mass Spectrometry. The Journal of Nutrition. 2016; 146(2). https://doi.
org/10.3945/jn.114.202317 PMID: 26764333
24.
25.
26.
Tochio T, Kadota Y, Tanaka T, Koga Y. 1-Kestose, the Smallest Fructooligosaccharide Component,
Which Efficiently Stimulates Faecalibacterium prausnitzii as Well as Bifidobacteria in Humans. Foods.
2018; 7(9):140. https://doi.org/10.3390/foods7090140 PMID: 30200390
Tochio T, Kitaura Y, Nakamura S, Sugawa C, Takahashi M, Endo A, et al. An Alteration in the Cecal
Microbiota Composition by Feeding of 1-Kestose Results in a Marked Increase in the Cecal Butyrate
Content in Rats. PLOS ONE. 2016; 11(11).
Liu Y, Li T, Alim A, Ren D, Zhao Y, Yang X. Regulatory Effects of Stachyose on Colonic and Hepatic
Inflammation, Gut Microbiota Dysbiosis, and Peripheral CD4+ T Cell Distribution Abnormality in High-
Fat Diet-Fed Mice. Journal of Agricultural and Food Chemistry. 2019; 67(42):11665–74. https://doi.org/
10.1021/acs.jafc.9b04731 PMID: 31588753
27. Yang J, Bai W, Zeng X, Cui C. Gamma glutamyl peptides: The food source, enzymatic synthesis,
kokumi-active and the potential functional properties—A review. Trends in Food Science & Technology.
2019; 91:339–46.
28. Broadhead GK, Mun H-chang, Avlani VA, Jourdon O, Church WB, Christopoulos A, et al. Allosteric
Modulation of the Calcium-sensing Receptor by γ-Glutamyl Peptides. Journal of Biological Chemistry.
2011; 286(11):8786–97. https://doi.org/10.1074/jbc.M110.149724 PMID: 21187282
29. Vlase L, Paˆ rvu M, Toiu A, Paˆ rvu EA, Cobzac SC, Puscas M. Rapid and Simple Analysis of Allicin in
Allium Species via LC-CIS-MS/MS. Studia Universitatis Babes-Bolyai Chemia. 2010; 55(4): 297–304.
30. Prati P, Henrique CM, Souza AS, Silva VS, Pacheco MT. Evaluation of allicin stability in processed gar-
lic of different cultivars. Food Science and Technology (Campinas). 2014; 34(3):623–8.
31. Shang A, Cao S-Y, Xu X-Y, Gan R-Y, Tang G-Y, Corke H, et al. Bioactive Compounds and Biological
Functions of Garlic (Allium sativum L.). Foods. 2019; 8(7):246.
32. Pavagadhi S, Swarup S. Metabolomics for Evaluating Flavor-Associated Metabolites in Plant-Based
Products. Metabolites. 2020; 10(5):197. https://doi.org/10.3390/metabo10050197 PMID: 32429044
33. Ringne´r M. What is principal component analysis? Nature Biotechnology. 2008; 26(3):303–4. https://
doi.org/10.1038/nbt0308-303 PMID: 18327243
34.
van den Berg RA, Hoefsloot HCJ, Westerhuis JA, Smilde AK, van der Werf MJ. Centering, scaling, and
transformations: improving the biological information content of metabolomics data. BMC Genomics.
2006; 7(1).
35. Shahidi F. Lipids in Flavor Formation. Flavor Chemistry. 2000;:24–43.
36. Agarwal S, Lanckriet G, Wills J, Kriegman D, Cayton L, Belongie S. Generalized non-metric multidimen-
sional scaling. In: Proceedings of the Eleventh International Conference on Artificial Intelligence and
Statistics. San Juan; 2007.
37. McGorrin RJ. The Significance of Volatile Sulfur Compounds in Food Flavors. ACS Symposium Series.
2011;:3–31.
38. Cardelle-Cobas A, Costo R, Corzo N, Villamiel M. Fructo-oligosaccharide changes during the storage
of dehydrated commercial garlic and onion samples. International Journal of Food Science & Technol-
ogy. 2009; 44(5):947–52.
39. Sims IM, Tannock GW. Galacto- and Fructo-oligosaccharides Utilized for Growth by Cocultures of Bifi-
dobacterial Species Characteristic of the Infant Gut. Applied and Environmental Microbiology. 2020; 86
(11). https://doi.org/10.1128/AEM.00214-20 PMID: 32220841
PLOS ONE | https://doi.org/10.1371/journal.pone.0247289 May 20, 2021
14 / 14
PLOS ONE |
10.1371_journal.pone.0243633 | RESEARCH ARTICLE
The sexual behaviours of adolescents aged
between 14 and 17 years involved with the
juvenile justice system in Australia: A
community-based survey
1, Jocelyn Jones2, Basil DonovanID
Lorraine YapID
Sophie Davison2,5, Ed Heffernan6,7,8, Alun Richards9, Carla Meurk6,7, Megan Steele7,
Christopher FisherID
10, Bianca Ton3, Tony Butler3*
1, Sally Nathan3, Elizabeth Sullivan4,
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
1 The Kirby Institute, The University of New South Wales, Sydney, Australia, 2 National Drug Research
Institute, Perth, Western Australia, Australia, 3 School of Population Health, The University of New South
Wales, Sydney, Australia, 4 The University of Newcastle, Callaghan, New South Wales, Australia, 5 State
Forensic Mental Health Service, North Metropolitan Health Service, & Office of the Chief Psychiatrist,
Nedlands, Western Australia, Australia, 6 Queensland Forensic Mental Health Service, Brisbane, Australia,
7 Queensland Centre for Mental Health Research, Brisbane, Australia, 8 The University of Queensland,
Brisbane, Australia, 9 Immunisation Program and BBV/STI Unit, Communicable Diseases Branch
Queensland Health, Brisbane, Queensland, Australia, 10 Australian Research Centre in Sex, Health and
Society, La Trobe University, Melbourne, Australia
OPEN ACCESS
* tbutler@unsw.edu.au
Citation: Yap L, Jones J, Donovan B, Nathan S,
Sullivan E, Davison S, et al. (2020) The sexual
behaviours of adolescents aged between 14 and 17
years involved with the juvenile justice system in
Australia: A community-based survey. PLoS ONE
15(12): e0243633. https://doi.org/10.1371/journal.
pone.0243633
Editor: Andrea Knittel, University of North Carolina
at Chapel Hill, UNITED STATES
Received: September 17, 2019
Accepted: November 24, 2020
Published: December 28, 2020
Peer Review History: PLOS recognizes the
benefits of transparency in the peer review
process; therefore, we enable the publication of
all of the content of peer review and author
responses alongside final, published articles. The
editorial history of this article is available here:
https://doi.org/10.1371/journal.pone.0243633
Copyright: © 2020 Yap et al. This is an open access
article distributed under the terms of the Creative
Commons Attribution License, which permits
unrestricted use, distribution, and reproduction in
any medium, provided the original author and
source are credited.
Data Availability Statement: All relevant data are
within the manuscript and its Supporting
Information files.
Abstract
Objectives
To overcome key knowledge gaps in relation to justice involved and vulnerable young peo-
ple and their sexual health and to compare this group with their peers from other youth
health surveys in Australia to determine the extent of the issues.
Methods
Young people, aged between 14 and 17 years, who had ever been or were currently
involved with the criminal justice system were purposively sampled. The survey was anony-
mous and delivered using Computer Assisted Telephone Interview (CATI).
Results
A total of 465 justice involved MeH-JOSH young people, aged between 14 and 17 years,
participated in the study: 44% Aboriginal and/or Torres Strait Islander (Indigenous) and 37%
not attending school. Of the total valid responses, 76% (n = 348) reported having ever had
sex, with sexual initiation at a median age of 14 years. We compared these data with their
peers in other Australian surveys and found that young people in our study had a higher
engagement in sex and start having sex at a younger age, reporting more sexual partners at
all ages.
PLOS ONE | https://doi.org/10.1371/journal.pone.0243633 December 28, 2020
1 / 19
PLOS ONE Funding: The authors disclosed that they received
the following support for their research and/or
authorship of this article: MeH-JOSH study was
funded by the Australian National Health and
Medical Research Council Project Grant No.
1043693 (awarded to TB, LY, BD, ES, SN). Tony
Butler is supported by an NHMRC Fellowship and
Basil Donovan is supported by a NHMRC
Practitioner Fellowship: https://www.nhmrc.gov.au.
Any material published or made publicly available
by the researchers cannot be considered as either
endorsed or an expression of the policies or view
by the Western Australia Department of Justice
and Western Australia Department of Health. Any
errors of omission or commission are the
responsibility of the researchers. The funders had
no role in study design, data collection and
analysis, decision to publish, or preparation of the
manuscript.
Competing interests: The authors have declared
that no competing interests exist.
The sexual behaviours of justice-involved adolescents in Australia
Conclusions
The sexual behaviours of young people involved in the justice system in this study suggest
they may be at a greater risk for sexually transmissible infections than their age-matched
peers in the general population. Policymakers should elevate them to a priority population
for targeting sexual health services and health promotion.
Introduction
Young people involved with the justice system aged between 14–17 years are under-researched
in relation to their sexual health in Australia and internationally. They are also neglected by
sexual health programmes even though they are potentially one of the most vulnerable groups
in society. The lack of health information on young people under youth justice supervision
was recently acknowledged by the Council of Australian Governments, the Australian Human
Rights Commission, and the Royal Australasian College of Physicians [1]. Limited social, edu-
cational and economic opportunities place them at a further disadvantage in reducing the risk
of harm to their health [2, 3]. Itinerancy and disengagement from traditional schooling com-
pounds access to this group from a research perspective [4].
Studies have suggested that drug addiction and mental health disorders that put youth at
risk for offending may also drive their tendency to engage in riskier sexual behaviours [5–8].
One study reported increased sexual risk behaviours among juvenile detainees compared to
the general population [9]. In Australia, a survey was conducted among juvenile justice youth
detention in New South Wales. Of 19 young females surveyed, 31.6% had ever been pregnant,
first pregnancy was at 14.2 years. No questions were asked on abortions, miscarriages, still-
births or drugs/smoking/alcohol while pregnant [3]. Formal comparisons of young offenders
with community populations and their sexual health and behaviours are extremely limited
[10].
In Australia, 83% or 4,568 young people were under juvenile justice community supervision
and 18% or 974 young people were in detention on an average day [11]. Community based
supervision in Australia includes unsentenced orders (supervised or conditional bail) and sen-
tenced orders (probation and similar order, suspended detention, parole or supervised release)
[12].
Young people who do not fall under the formal supervision of the youth justice system but
nonetheless who have had contact with the justice system have been neglected from a research
perspective due to itinerancy and disengagement from education and other mainstream ser-
vices (including government sponsored youth justice services), precluding their involvement
in mainstream health surveys. These young people may include those who have had police
warnings, cautions, fines, good behaviour bonds or who have been diverted to other types of
detention such as, a police watch-house or adult prison, or have been diverted to other pro-
grams (e.g. flexi-learning centres and colleges, drug and alcohol treatment centres). We know
even less about this population.
The Mental Health, Sexual Health and Reproductive Health of Young People in Contact
with the Criminal Justice System (MeH-JOSH) study aimed to describe the sexual and mental
health and risk behaviours of young people (14–17 year olds) in the community who had ever
been in contact with the criminal justice system (e.g. police, youth justice, court system, tribu-
nals). The main aim of the survey was to overcome key knowledge gaps in relation to this pop-
ulation and compare this group with their peers from other health surveys of youth in
Australia.
PLOS ONE | https://doi.org/10.1371/journal.pone.0243633 December 28, 2020
2 / 19
PLOS ONE The sexual behaviours of justice-involved adolescents in Australia
Methods
Sampling and sample size
A purposive sampling design based on strict selection criteria was used to recruit young people
in the community; participants must be aged between 14 and 17 years and had to have had
contact with the criminal justice system in the past or present. This approach was necessary as
permission was not granted to conduct the survey with young people in detention and those
serving community orders.
A sample of young people aged 14–17 years old was selected to reflect the age of the youth
justice population in Australia. Those over 17 (that is, 18 or older) are considered adults. Ethi-
cal issues prevented interviewing those below 14 years, such as obtaining parental/guardian
consent, and also having to ensure that very young respondents could truly understand what
they would be consenting to.
Quota sample sizes were calculated based on known demographic characteristics of the
Australian juvenile offender population (age and gender) [11]. Aboriginal and/or Torres Strait
Islander (Indigenous) young people in this study sample were over represented (44%) com-
pared to 5% in the general Australian population and this reflects the greater involvement of
this population in the justice system [13]. In Australia, the majority of young people under
supervision on an average day in 2017–18 were male (81%) [11]. Females were deliberately
oversampled in this study to enable more advanced statistical analysis.
Recruitment
Recruitment took place between June 2016 and August 2018 using four different recruitment strat-
egies and to minimise selection bias. Firstly, young people who met the selection criteria were
recruited through referrals by programme coordinators from community-based organisations and
youth drop-in centres: 70% of the total participants were recruited using this approach. The second
strategy involved recruiting young people waiting inside or outside the magistrate courts on days
the Children’s Court was in session. A third strategy was to recruit justice involved young people
from youth mental health service centres in Western Australia. Programme coordinators referred
young people who satisfied the selection criteria and agreed to participate in the study. A fourth
strategy was to recruit justice involved young people attending flexi-learning schools or colleges.
These institutions aim to re-engage young people who had been out of the formal school system.
Consent
Due to the nature of this population and from our experiences of conducting surveys among
young offenders in Australia [4], the study expected that a high proportion of young people
would not have a good relationship with their parent(s) or with the adults responsible for
them. Thus human research ethics approval was sought and given for young people recruited
in the survey to be treated as mature minors [14]. Nevertheless, all young people approached
outside the courts were still asked by recruiters if we could contact their parents or guardians
for permission to allow them to participate in the survey. Almost all respondents refused us
contact or to give any contact details of their parents or guardians preferring to give consent
themselves. Recruiters were required to administer a Gillick Competency checklist to ensure
that all respondents met the criteria of a mature minor.
Survey
Having obtained written consent from the young person, the survey was delivered using a
Computer Assisted Telephone Interview (CATI). The average time to complete the survey was
PLOS ONE | https://doi.org/10.1371/journal.pone.0243633 December 28, 2020
3 / 19
PLOS ONE The sexual behaviours of justice-involved adolescents in Australia
approximately 40 minutes. No identifying information was recorded by the interviewer to
ensure anonymity.
Data were collected on the following: socio-demographics; history of justice system involve-
ment; sexual identity, sexual attraction and sexual history; sexual health behaviours and knowl-
edge; human papillomavirus (HPV) vaccination; and sexually transmissible infection (STIs)
history. Survey questions were designed to be consistent with other Australian surveys of
young people to enable comparisons.
Original datasets of the MeH-JOSH survey were age matched with the Young Minds Mat-
ter: 2nd Survey of the Mental Health of Australian Children Survey (YMM), a probability sam-
ple of young people aged between 4 and 17 years old from 5,500 randomly sampled families in
Australia [15]. Separate age matched data analysis was provided by C. Fisher for the 6th
National Survey of Secondary Students and Sexual Health (SSASH), a convenience sample of
adolescents in the community that had agreed to take part in an online survey in Australia
[16]. MeH-JOSH sexual health and behavioural questions for this paper were selected based
on the equivalence or sameness of questions from the YMM and SSASH survey (for example,
MEH-JOSH–“Can you tell me if you have ever had sex?” YMM and SSASH–“Have you ever
had sexual intercourse?”).
After completion of the survey, participants were given either $50 cash or a gift card to
reimburse them for their time and travel. No incentives were given to participants recruited
inside the Children’s Court in Western Australia as it was deemed inappropriate. Conse-
quently no participants from this location agreed to take part in the survey.
Post-survey interview. All participants were given post-survey exit interviews and, if nec-
essary, provided with referrals to relevant health agencies to ensure they were not adversely
impacted by the survey questions. We found that HPV vaccinations were the most sought
after referral request.
Participants
A total of 465 justice involved young people (63% male and 37% female), aged between 14 and
17 years, consented to participate in the MeH-JOSH survey. There was a higher proportion of
males than females in the 16–17 age group (56% versus 44%, χ2(1) = 5.683, p = .017). Of the
total surveyed, 85% were born in Australia, 44% identified as Aboriginal and/or Torres Strait
Islanders (Indigenous) and 37% were not attending secondary school or a flexi-school or col-
lege (Table 1). A higher proportion of males were on a current sentence or order compared to
females (40% versus 30%, χ2(1) = 4.567, p = .033). Site of recruitment was not a confounding
factor for age and gender although it was significant for Aboriginality (Indigenous status)
(χ2(4) = 12.026, p = .017).
MEH-JOSH survey data was compared to two community surveys of: (i) 2004 age-matched
young people from the YMM survey (51% male and 49% female), 89% of whom were attend-
ing secondary school or a flexi-school or college, and (ii) 7170 age-matched young people
(43% male, 55% female and 2% trans or gender diverse) from the SSASH survey, 5% of whom
identified as Aboriginal and/or Torres Strait Islanders (Indigenous), 91% born in Australia
and 94% attending secondary school or a flexi-school or college (Table 1).
Data analysis
Sexual health and behaviours
Descriptive statistics were generated using Statistics SPSS 25. The self-reported sexual health
and behaviours of participants in the MEH-JOSH survey (including sexual identity and attrac-
tion, relationship status, ever having sex (any, vaginal, anal, oral), sexual initiation, lifetime
PLOS ONE | https://doi.org/10.1371/journal.pone.0243633 December 28, 2020
4 / 19
PLOS ONE Table 1. Sociodemographic profiles of community-based justice involved young people (MeH-JOSH) and young people from selected community-based population
surveys in Australia, aged between 14 and 17 years.
DESCRIPTION
Justice-involved population
General population
The sexual behaviours of justice-involved adolescents in Australia
Gender
Age Group
Aboriginal and/or Torres Strait Islander
(Indigenous)
Place of Birth
Male
Female
Trans and Gender Diverse
Total
14–15 years
16–17 years
Total
Yes
No
Prefer not to say/Missing�
Total
Australia
New Zealand
Other
Prefer not to say/Missing�
Total
Education
Goes to school
-
-
-
-
-
-
-
-
Community based
MeH-JOSH
Male
n (%)
Female
n (%)
Total
n (%)
465
293 (63.0)
-
2014
YMM1
2018
SSASH2
p value Total
n (%)
2004
1029
(51.3)
Total
n (%)
7170
3067 (42.8)
172 (37.0)
975 (48.7)
3930 (54.8)
0 (0.0)
465
(100.0)
0 (0.0)
2004
(100.0)
173 (2.4)
7170
(100.0)
130 (44.4) 96 (55.8)
226 (48.6) 0.017
652 (32.5)
2046 (28.5)
163 (55.6) 76 (44.2)
239 (51.4)
293
(100.0)
172
(100.0)
465
(100.0)
124 (42.5) 78 (45.6)
202 (43.6) 0.510
168 (57.5) 93 (54.4)
261 (56.4)
1 (-)
292
(100.0)
1 (-)
171
(100.0)
2 (-)
463
(100.0)
243 (82.9) 154 (89.5) 397 (85.4) 0.147
35 (11.9)
12 (7.0)
47 (10.1)
15 (5.1)
6 (3.5)
21 (4.5)
0 (-)
0 (-)
0 (-)
293
172
(100.0)
(100.0)
184 (62.8) 110 (64.0) 294#
(63.2)
465
(100.0)
1352
(67.5)
2004
(100.0)
-
-
-
-
-
-
-
-
-
0.803
1780
(88.8)
5124 (71.5)
7170
(100.0)
298 (4.6)
6697 (95.7)
175 (-)
7170
(100.0)
6481 (90.8)
89 (1.2)
565 (8.0)
35 (-)
7170
(100.0)
6679 (94.2)
Most serious sentence/order
Not at school
Prefer not to say/Missing�
Total
109 (37.2) 62 (36.0)
171 (36.8)
224 (11.2)
412 (5.8)
0 (-)
293
(100.0)
0 (-)
172
(100.0)
0 (-)
465
(100.0)
0 (-)
2004
(100.0)
79 (-)
7170
(100.0)
Community Low–Fines, Bond, Caution,
Warning
Community High–Probation and Community
Order
Detention–Juvenile Detention, Police
Watchhouse, Prison
130 (44.4) 85 (49.4)
215 (46.2) 0.341
46 (15.7)
25 (14.5)
71 (15.3)
104 (35.5) 50 (29.1)
154 (33.1)
Other Contact–Diversion, Drug court
13 (4.4)
12 (7.0)
25 (5.4)
Total
293
(100.0)
172
(100.0)
465
(100.0)
-
-
-
-
-
-
-
-
-
-
(Continued )
PLOS ONE | https://doi.org/10.1371/journal.pone.0243633 December 28, 2020
5 / 19
PLOS ONE Table 1. (Continued)
DESCRIPTION
Justice-involved population
General population
The sexual behaviours of justice-involved adolescents in Australia
Community based
MeH-JOSH
2014
YMM1
2018
SSASH2
Male
n (%)
Female
n (%)
Total
n (%)
465
p value Total
n (%)
2004
Total
n (%)
7170
Offence categories (May have more than
one answer)
Theft and related offences
Unlawful entry with intent/burglary, break and
enter
109 (37.2) 77 (44.8)
186 (40.0) 0.298
54 (18.4)
13 (7.6)
67 (14.4)
0.344
Robbery, extortion and related offences
46 (15.7)
9 (5.2)
55 (11.8)
0.406
Acts intended to cause injury
63 (21.5)
53 (30.8)
116 (24.9) 0.254
Public order offences
41 (14.0)
22 (12.8)
63 (13.5)
Property damage and environmental pollution
50 (17.1)
5 (2.9)
55 (11.8)
Illicit drug offences
Offences against government procedures,
security and operations
Traffic and vehicle regulatory offences
Abduction, harassment and other offences
against the person
Fraud, deception and related offences
Sexual assault and related offences
Other
Yes
No
Prefer not to say/Missing�
Total
On a current order
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
0.895
0.407
0.699
0.941
0.888
0.980
0.922
0.959
0.906
32 (10.9)
12 (7.0)
19 (6.5)
10 (5.8)
44 (9.5)
29 (6.2)
16 (5.5)
8 (2.7)
6 (2.0)
2 (0.7)
9 (3.1)
7 (4.1)
5 (2.9)
1 (0.6)
2 (1.2)
3 (1.8)
23 (4.9)
13 (2.8)
7 (1.5)
4 (0.9)
12 (2.6)
116 (40.1) 51 (30.2)
167 (35.5) 0.033
173 (59.9) 118 (69.8) 291 (63.5)
4 (-)
289
(100.0)
3 (-)
169
(100.0)
7 (-)
458
(100.0)
� Prefer not to say/Missing not included in the total n (%) and p-value calculations
1 2014 YMM– 2014 Young Minds Matter: 2nd Survey of the Mental Health of Australian Children Survey
2 2018 SSASH– 6th National Survey of Secondary Students and Sexual Health
#Includes secondary high schools and flexi- (flexible) schools and colleges
https://doi.org/10.1371/journal.pone.0243633.t001
sexual encounters, last sexual encounter, STI testing and diagnosis, HPV vaccination, and will-
ingness to have HPV vaccine) were compared with their age-matched peers in the YMM [15]
and SSASH [16] surveys where applicable. Chi-squared tests were used to analyse gender dif-
ferences in these variables in the MEH-JOSH survey. A test of two proportions was used to
analyse gender differences in types of offences committed.
Ethics
Approval for the study was obtained from the University of New South Wales Sydney Human
Research Ethics Committee, (HC13308), the Western Australia Aboriginal Health Ethics
Committee (WAAHEC 625), and Curtin University (HRE0133). Permission was also granted
in Western Australia by the North Metropolitan Health Service Mental Health Research Ethics
Committee (22_2016), the North Metropolitan and East Metropolitan Health Services’
Research Governance, and the Department of the Justice Research Application Advisory Com-
mittee (ref 2016/02161) to recruit from their respective premises.
PLOS ONE | https://doi.org/10.1371/journal.pone.0243633 December 28, 2020
6 / 19
PLOS ONE The sexual behaviours of justice-involved adolescents in Australia
Results
Sexual health and behaviours
Of the MeH-JOSH participants, 91% identified as heterosexual, 6% bisexual, and 1% homosex-
ual/lesbian. Young females in the MeH-JOSH study were more likely to identify as bisexual
than young males, (15% versus 1%, χ2(3) = 34.295, p < .001), and this is also reflected in their
sexual attraction (14% versus 3%, χ2(3) = 20.031, p < .001). This is similar to the SSASH study
whereby young females in the general population were also more likely to identify as bisexual
(C Fisher, personal communication September 2020). Thirty-one per cent of respondents
reported they were in a current relationship, either dating or living with their partner. Among
14–15 year old and 16–17 year old age groups, 27% and 34% respectively, were in a relation-
ship at the time of the interview.
Of the total MeH-JOSH sample, 76% (n = 348) reported having ever had oral, vaginal and/
or anal sex, which was higher than among school aged young people in the SSASH (57%) and
YMM (24%) surveys (Fig 1A and Table 2), but lower than the 97% who reported having had
oral, vaginal and/or anal sex in the NSW Young People in Custody Health Survey (YPiCHS)
(Fig 1A) [3]. In the MeH-JOSH survey, 74% (n = 338) had had vaginal and/or anal sex com-
pared to 45% in the SSASH survey (Fig 1B and Table 2). Condom use at the last sexual encoun-
ter was 55% for both MeH-JOSH and SSASH respondents compared to 64% among YMM
respondents (Table 3).
Of the 348 who reported oral, vaginal and/or anal sex, 76% reported having had sex for the
first time aged 14 years or less (median age 14 years) and mostly (94%) with a person <18
years of age (median age 15) (Table 2). Justice-involved young females were more likely to
have had sex with a person 18 years or older for their first sexual partner (13% versus 3%, χ2(1)
= 12.056, p = .001), and their last sexual partner (32% versus 12%, χ2(1) = 19.812, p < .001),
compared to males. Of those in the MeH-JOSH survey who had engaged in sex, 60% reported
having had oral sex (median age 14 years), 96% vaginal sex (median age 14 years), and 20%
anal sex (median age 15 years). Of those who had had sex, reports of oral sexual experience
among justice-involved young people (MeH-JOSH) were below the proportions reported by
age-matched secondary school students (SSASH) (60% versus 97% respectively), but much
higher for vaginal sexual experience (96% versus 76% respectively) (Table 3).
Peak age for vaginal sexual initiation occurred two years earlier in the MeH-JOSH sample
compared to the school sample (SSASH) (Fig 2A). Anal sex in those who were sexually active
in the justice-involved sample was similar to a national survey on secondary school students
(SSASH) (20% versus 22% respectively) although peak age of anal sex initiation occurred
about one year earlier among MeH-JOSH participants (Fig 2B).
Of those who have had sex, 41% of justice-involved MeH-JOSH participants reported hav-
ing six or more sexual partners in their lifetime compared to 16% among school aged young
people in the YMM survey, with reported higher numbers of sexual partners at all ages among
MeH-JOSH participants (Fig 3A and 3B). Justice-involved young males were more likely to
report numbers of 6 or more sexual partners than females (45% versus 34%, χ2(1) = 4.091, p =
.043).
For their last sexual encounter, justice-involved young females were more likely to have sex
with their current girlfriend/boyfriend (54% versus 40%), while males were more likely to have
sex with a stranger (16% versus 4%, χ2(3) = 13.629, p = .003). Young males were more likely to
use a condom during their last sexual encounter than females (59% versus 48%, χ2(1) = 4.094,
p = .043).
Of the 348 sexually experienced participants, 37% (n = 129) reported prior testing for sexu-
ally transmissible infections (STIs) with young females more likely to have been tested recently
PLOS ONE | https://doi.org/10.1371/journal.pone.0243633 December 28, 2020
7 / 19
PLOS ONE The sexual behaviours of justice-involved adolescents in Australia
Fig 1. a. Proportion of justice-involved and community (SSASH and YMM) young people who had ever had oral,
anal and/or vaginal sex. b. Proportion of justice-involved and community (SSASH) young people who had ever had
anal and/or vaginal sex.
https://doi.org/10.1371/journal.pone.0243633.g001
in the last year compared to males (44% versus 24%, χ2(2) = 9.392, p = .009). Young females
were more likely to be tested by a general practitioner (57% versus 40%), while males were
more likely to be tested in a juvenile detention centre or prison (20% versus 3%, χ2(5) =
19.377, p = .002). Of those who have ever been tested, 18% were diagnosed with an STI: chla-
mydia was the most common STI (12.4% positive); none reported being diagnosed with
gonorrhoea or syphilis (Table 4).
PLOS ONE | https://doi.org/10.1371/journal.pone.0243633 December 28, 2020
8 / 19
PLOS ONE Table 2. A comparison of the sexual identity, sexual attraction, sexual behaviours and condom use of community-based justice involved young people (MeH-JOSH)
and young people from selected community-based population surveys in Australia, aged between 14 and 17 years.
DESCRIPTION
Justice-involved population
General population
The sexual behaviours of justice-involved adolescents in Australia
Community based
MeH-JOSH
2014 YMM1
2018
SSASH2
Male
n (%)
Female
n (%)
Total
n (%)
465
p value Total
n (%)
2004
Total
n (%)
7170
Sexual identity
Heterosexual or straight
279 (96.2)
137 (82.5)
416 (91.2) <0.001 -
Sexual
attraction
Bisexual
Homosexual/Queer
Unsure/Undecided
Prefer not to say/Missing�
Total
4 (1.4)
3 (1.0)
4 (1.4)
3 (-)
290
(100.0)
25 (15.1)
29 (6.4)
3 (1.8)
1 (0.6)
6 (-)
166
(100.0)
6 (1.3)
5 (1.1)
9 (-)
456
(100.0)
-
-
-
-
-
Only to people of the opposite sex
263 (90.4)
130 (77.8)
393 (85.8) <0.001 -
To both sexes
9 (3.1)
23 (13.8)
32 (7.0)
Only to people of own sex
Not sure
Prefer not to say/Missing�
Total
12 (4.1)
10 (6.0)
7 (2.4)
2 (-)
291
(100.0)
4 (2.4)
5 (-)
167
(100.0)
22 (4.8)
11 (2.4)
7 (-)
458
(100.0)
Relationships
Single/Never married
211 (72.3)
109 (63.4)
320 (69.0)
0.083
Living with partner in same residence
14 (4.8)
15 (8.7)
29 (6.3)
In a relationship but not living with
partner
Prefer not to say/Missing�
Total
67 (22.9)
48 (27.9)
115 (24.8)
1 (-)
292
(100.0)
0 (-)
172
(100.0)
1 (-)
464
(100.0)
Ever had sex
Oral, Vaginal, and/or Anal
Sex3
Never had sex
Yes
Prefer not to say/Missing�
Total
63 (21.6)
46 (27.7)
109 (23.9)
0.144
1411 (76.1)
3034 (43.3)
228 (78.4)
120 (72.3)
348 (76.1)
442 (23.9)
3979 (56.7)
2 (-)
291
(100.0)
6 (-)
166
(100.0)
8 (-)
457
(100.0)
Vaginal and/or Anal Sex
Never had sex
63 (21.6)
46 (27.7)
109 (23.9)
0.215
No
Yes
Prefer not to say/Missing�
Total
8 (2.8)
2 (1.2)
10 (2.2)
220 (75.6)
118 (71.1)
338 (73.9)
2 (-)
291
(100.0)
6 (-)
166
(100.0)
8 (-)
457
(100.0)
1 2014 YMM– 2014 Young Minds Matter: 2nd Survey of the Mental Health of Australian Children Survey
2 2018 SSASH– 6th National Survey of Secondary Students and Sexual Health
3 MEH-JOSH–“Can you tell me if you have ever had sex?”; YMM and SSASH–“Have you ever had sexual intercourse?”
� Prefer not to say/Missing not included in the total n (%) and p-value calculations
��SSASH–Mostly same, mostly opposite, and equally both sexes combined for “Both Sexes”
https://doi.org/10.1371/journal.pone.0243633.t002
Of the total sample, 30% of young males and females recalled being vaccinated against the
human papilloma virus (HPV), 44% were not, and 25% had never heard of it or could not
recall if they had ever been vaccinated. Young females were more likely to have been HPV
PLOS ONE | https://doi.org/10.1371/journal.pone.0243633 December 28, 2020
9 / 19
5177 (73.6)
1147 (16.3)
335 (4.8)
377 (5.4)
134 (-)
7170 (100)
4244 (62.7)
2290��
(33.8)
235 (3.5)
0 (0.0)
401 (-)
7170 (100.0)
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
151 (-)
1853
(100.0)
-
-
-
-
-
157 (-)
7170 (100.0)
-
3831 (54.6)
3182 (45.4)
157 (-)
7170 (100.0)
PLOS ONE Table 3. A comparison of sexual initiation, lifetime sexual encounters and sexual behaviours of community-based justice involved young people (MeH-JOSH) and
young people from selected community-based population surveys in Australia, aged between 14 and 17 years.
EVER HAD SEX (ORAL, VAGINAL, ANAL SEX)
Justice-involved population
General population
The sexual behaviours of justice-involved adolescents in Australia
Sexual initiation
Age at sexual initiation
Total
Age of first sexual
partner
Lifetime sexual
encounters
Number of sexual
partners3
How often was a
condom used?
Range
Mean
Median
<18 years
18 years and over
Prefer not to say/Missing�
Total
Range
Mean
Median
1 to 5 people
6 or more people
Prefer not to say/Missing�
Total
All the time
Most/Half /Some
86 (38.2)
49 (41.5)
Never
Prefer not to say/Missing�
Total
Last sexual
encounter
Was the last person you
had sex with . . .?
•Your current girlfriend/boyfriend
• Someone you had known for a while, but had
not had sex with before
• Someone known for a while, had sex with
before, but not your current girlfriend or
boyfriend
41 (18.2)
30 (25.4)
71 (20.7)
3 (-)
2 (-)
5 (-)
225
(100.0)
118
(100.0)
91 (40.1)
65 (54.2)
343
(100.0)
156
(45.0)
54 (23.8)
29 (24.2)
83 (23.9)
44 (19.4)
21 (17.5)
65 (18.7)
• Someone you had just met for the first time
38 (16.7)
5 (4.2)
43 (12.4)
• Prefer not to say/Missing
Total
1 (-)
0 (-)
1 (-)
227
(100.0)
120
(100.0)
347
(100.0)
Community based
MeH-JOSH
Male
n (%)
Female
n (%)
226
6–17
13.51
14
218
(96.9)
120
5–16
13.54
14
102
(87.2)
Total
n (%)
348
346
5–17
13.52
14
320
(93.6)
7 (3.1)
15 (12.8)
22 (6.4)
3 (-)
3 (-)
6 (-)
225
(100.0)
5–48
14.72
15
124
(55.1)
101
(44.9)
3 (-)
225
(100.0)
117
(100.0)
11–55
16.19
16
79 (66.4)
40 (33.6)
1 (-)
119
(100.0)
98 (43.6)
39 (33.1)
342
(100.0)
5–55
15.22
15
203
(59.0)
141
(41.0)
4 (-)
344
(100.0)
137
(39.9)
135
(39.4)
2014
YMM1
p value Total
n (%)
442
2018
SSASH2
Total
n (%)
3979
-
0.001
0.043
0.117
0.003
-
-
-
-
-
-
-
-
-
-
-
371
(83.9)
71 (16.1)
0 (-)
442
(100.0)
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
2079
(64.3)
574 (17.7)
370 (11.4)
213 (6.6)
743 (-)
3979
(100.0)
(Continued )
PLOS ONE | https://doi.org/10.1371/journal.pone.0243633 December 28, 2020
10 / 19
PLOS ONE Table 3. (Continued)
EVER HAD SEX (ORAL, VAGINAL, ANAL SEX)
Justice-involved population
General population
The sexual behaviours of justice-involved adolescents in Australia
Age of last sexual
partner
<18 years
18 years and over
Prefer not to say/Missing�
Total
Range
Mean
Median
Did you use a condom? No
Community based
MeH-JOSH
Male
n (%)
Female
n (%)
196
(87.9)
81 (68.1)
Total
n (%)
348
277
(81.0)
27 (12.1)
38 (31.9)
65 (19.0)
5 (-)
1 (-)
6 (-)
223
(100.0)
11–48
16.29
119
(100.0)
12–35
17.35
16
92 (40.7)
17
62 (52.1)
342
(100.0)
11–48
16.66
16
154
(44.6)
191
(55.4)
2014
YMM1
p value Total
n (%)
442
<0.001 -
-
-
-
-
-
-
0.043
154
(34.8)
284
(64.3)
2018
SSASH2
Total
n (%)
3979
2449
(76.6)
749 (23.4)
141 (-)
3198
(100.0)
-
-
-
1398
(43.0)
1807
(55.6)
4 (0.9)
47 (1.4)
Yes
Not sure
Prefer not to say/Missing�
134
(59.3)
0 (0.0)
2 (-)
57 (47.9)
0 (0.0)
0 (0.0)
1 (-)
3 (-)
Oral sex
Ever had oral sex
Age at sexual initiation
(oral sex)
Vaginal sex
Ever had vaginal sex
Age at sexual initiation
(vaginal sex)
Total
No
Yes
Prefer not to say/Missing�
Total
Total
Range
Mean
Median
No
Yes
Prefer not to say/Missing�
Total
Total
Range
Mean
Median
226
(100.0)
119
(100.0)
345
(100.0)
84 (38.4)
53 (44.5)
135
(61.6)
9 (-)
219
(100.0)
135
9–17
14.06
14
10 (4.4)
215
(95.6)
3 (-)
66 (55.5)
1 (-)
119
(100.0)
65
5–17
13.97
14
3 (2.5)
117
(97.5)
0 (-)
137
(40.5)
201
(59.5)
10 (-)
338
(100.0)
200
5–17
14.03
14
13 (3.8)
332
(96.2)
3 (-)
0.269
-
0.366
225
(100.0)
120
(100.0)
345
(100.0)
212
6–17
13.61
14
117
5–16
13.71
14
329
5–17
13.64
14
-
0 (-)
442
(100.0)
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
87 (-)
3252
(100.0)
119 (3.0)
3860
(97.0)
0 (-)
3979
(100.0)
-
-
-
-
807 (21.1)
3022
(78.9)
150 (-)
3979
(100.0)
-
-
-
-
(Continued )
PLOS ONE | https://doi.org/10.1371/journal.pone.0243633 December 28, 2020
11 / 19
PLOS ONE Table 3. (Continued)
EVER HAD SEX (ORAL, VAGINAL, ANAL SEX)
Justice-involved population
General population
The sexual behaviours of justice-involved adolescents in Australia
How often was a
condom used?
All the time
Most/Half/Some
Never
Prefer not to say/Missing�
Total
Anal sex
Ever had anal sex
No
Yes
Prefer not to say/Missing�
Total
Age at sexual initiation
(anal sex)
Total
Range
Mean
How often was a
condom used?
Median
All the time
Most/Half/Some
Never
Prefer not to say/Missing�
Total
Community based
MeH-JOSH
Male
n (%)
Female
n (%)
95 (44.4)
40 (34.5)
85 (39.7)
50 (43.1)
Total
n (%)
348
135
(40.9)
135
(40.9)
34 (15.9)
26 (22.4)
60 (18.2)
1 (-)
1 (-)
2 (-)
214
(100.0)
180
(79.3)
116
(100.0)
99 (82.5)
330
(100.0)
279
(80.4)
47 (20.7)
21 (17.5)
68 (19.6)
1 (-)
0 (-)
1 (-)
227
(100.0)
120
(100.0)
347
(100.0)
0.151
0.474
47
9–17
14.57
21
12–16
14.81
68
9–17
14.65
-
15
25 (53.2)
15
6 (30.0)
15
31 (46.3)
0.218
6 (12.8)
4 (20.0)
10 (14.9)
16 (34.0)
10 (50.0)
26 (38.8)
0 (-)
1 (-)
1 (-)
47
(100.0)
20
(100.0)
67
(100.0)
2014
YMM1
p value Total
n (%)
442
2018
SSASH2
Total
n (%)
3979
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
3114
(78.3)
865 (21.7)
0 (-)
3979
(100.0)
-
-
-
-
-
-
-
-
-
1 2014 YMM– 2014 Young Minds Matter: 2nd Survey of the Mental Health of Australian Children Survey
2 2018 SSASH– 6th National Survey of Secondary Students and Sexual Health
3 MEH-JOSH–“During your life, with how many people have you had sex?”; YMM–“During your life, with how many people have you had sexual intercourse?”
� Prefer not to say/Missing not included in the total n (%) and p-value calculations
https://doi.org/10.1371/journal.pone.0243633.t003
vaccinated than males (38% versus 26%, χ2(2) = 7.361, p = .025). Of those who reported being
vaccinated (n = 139), Aboriginal and/or Torres Strait Islanders had lower vaccination rates
compared to non-Indigenous (37% versus 63%, χ2(2) = 6.596, p = .037). Among those who
were never vaccinated or were unsure or had never heard of HPV, 58% indicated they were
willing to have the HPV vaccine (Table 5). A higher proportion of young females than males
indicated a willingness to be HPV vaccinated (68% versus 53%, χ2(2) = 6.988, p = .030).
Discussion
We have observed that young people involved in the justice system have a higher engagement
in sex compared with their peers in school surveys (1.3 to 3.2 times higher); (ii) are often start-
ing to have sex at a younger age (median age 14 years compared to 17 years in the general Aus-
tralian population and 15 years in the general Indigenous population) [17, 18]; (iii) are
PLOS ONE | https://doi.org/10.1371/journal.pone.0243633 December 28, 2020
12 / 19
PLOS ONE The sexual behaviours of justice-involved adolescents in Australia
Fig 2. a. Age of sexual initiation (vaginal) among justice-involved and community (SSASH) young people. b. Age of
sexual initiation (anal) among justice-involved and community (SSASH) young people.
https://doi.org/10.1371/journal.pone.0243633.g002
significantly more likely to report large numbers of sexual partners (6+) at a young age. While
similar rates of young people used a condom during their last sexual encounter across MeH-
JOSH, SSASH and YMM surveys, engagement in more frequent and riskier sexual behaviours
suggests that this may pose more of a STI risk to justice-involved young people. Most had
never been or could not recall having been vaccinated for HPV, even though all Australian
young people are offered the vaccine in school at age 12–13 years and first-dose coverage in
young females is 86% and 78% in young males, and increasing [19]. These findings highlight
the critical need for sexual health promotion and vaccination programs to reach this
PLOS ONE | https://doi.org/10.1371/journal.pone.0243633 December 28, 2020
13 / 19
PLOS ONE The sexual behaviours of justice-involved adolescents in Australia
Fig 3. a. Proportion of sexual partners by age in justice-involved young people. MeH-JOSH–Of those who had ever had
oral, vaginal and/or anal sex (n = 344). b. Proportion of sexual partners by age in community (YMM) young people.
YMM–Of those who had ever had sexual intercourse (n = 442).
https://doi.org/10.1371/journal.pone.0243633.g003
population, particularly those of Aboriginal and/or Torres Strait Islander descent. It also repre-
sents an opportunity; 58% of participants in this survey (who had never been or were unsure if
they had been vaccinated) indicated their willingness to be vaccinated for HPV.
Young people who identified as Aboriginal and/or Torres Strait Islander (Indigenous)
descent were overrepresented in the survey sample compared with the general population
(44% versus 5%), reflecting the greater involvement of this population in the justice system. In
2016–2017, on an average day, there were 4,286 young people between the age of 14 and 17
PLOS ONE | https://doi.org/10.1371/journal.pone.0243633 December 28, 2020
14 / 19
PLOS ONE Table 4. STI testing and diagnosis as reported by community-based justice involved young people (MeH-JOSH) in Australia, aged between 14 and 17 years.
The sexual behaviours of justice-involved adolescents in Australia
EVER HAD SEX (ORAL, VAGINAL, ANAL SEX)
Ever tested for an STI/HIV/AIDS
Justice-involved population
Community based
MeH-JOSH
Male
n (%)
228
No
Female
n (%)
120
Total
n (%)
348
p value
157 (69.2)
58 (49.2)
215 (62.3) 0.009
Yes, in the last year
55 (24.2)
52 (44.1)
107 (31.0)
Yes, more than a year ago
Prefer not to say/Missing�
Total
15 (6.6)
8 (6.8)
1 (-)
227
(100.0)
2 (-)
118
(100.0)
23 (6.7)
3 (-)
345
(100.0)
If YES, where did you get your last STI and/or HIV test? Local doctor at a General Practice
28 (40.0)
34 (56.7)
62 (47.7) 0.002
(If ever tested) Ever diagnosed with an STI/HIV/AIDS
Other
Total
No
Yes, in the last year
Yes, more than a year ago
Prefer not to say/Missing�
Total
If YES, which infections have you been diagnosed with? (May have more than one
answer)
Chlamydia
Trichomoniasis
Urinary Tract Infection
Genital warts
Prefer not to say/Missing�
clinic
Juvenile Justice Detention/Prison
14 (20.0)
Hospital
8 (11.4)
Family planning/Sexual health clinic
2 (2.9)
Aboriginal Medical Service
4 (5.7)
2 (3.3)
4 (6.7)
9 (15.0)
6 (10.0)
16 (12.3)
12 (9.2)
11 (8.5)
10 (7.7)
19 (14.6)
130
(100.0)
14 (20.0)
5 (8.3)
70 (100.0)
60 (100.0)
60 (87.0)
46 (76.7)
106 (82.2) 0.312
6 (8.7)
3 (4.3)
1 (-)
9 (15.0)
5 (8.3)
0 (-)
69 (100.0)
60 (100.0)
15 (11.6)
8 (6.2)
1 (-)
129
(100.0)
7 (87.5)
1 (12.5)
0 (0.0)
0 (0.0)
1 (-)
9 (81.8)
16 (84.2) 0.652
1 (9.1)
1 (9.1)
1 (9.1)
3 (-)
2 (10.5)
1 (5.3)
1 (5.3)
4 (-)
�Prefer not to say/Missing not included in the total n (%) and p-value calculations
https://doi.org/10.1371/journal.pone.0243633.t004
Total
8 (100.0)
12 (100.0)
20 (100.0)
years under youth or juvenile justice supervision in Australia, of whom 2,036 (48%) were
Aboriginal and/or Torres Strait Islander [20].
Of the 129 survey participants who reported ever being tested for an STI, 18% had tested
positive in the past, with most being diagnosed with chlamydia. In Australia, chlamydia infec-
tions are the most frequently notified STI and are increasing with females 2.3 times more likely
than males to be diagnosed each year, while diagnosis in Aboriginal and/or Torres Strait
Islander populations is 2.8 times higher than non-Indigenous populations (1193.9 per 100,000
compared with 427.0 per 100,000) [21]. In 2017, 17% of 100,775 notifications were among
those aged 15–19 years underscoring the need to reach these justice involved young people
who fall within this age group and are at sexual health risk [21].
A limitation of this study includes the use of a non-random sampling strategy (necessary to
overcome practical difficulties of reaching such individuals), resulting in a sample that may
not be statistically representative of the target population (i.e. the full range of justice-involved
PLOS ONE | https://doi.org/10.1371/journal.pone.0243633 December 28, 2020
15 / 19
PLOS ONE Table 5. HPV vaccinations as reported by community-based justice involved young people in Australia, aged between 14 and 17 years.
The sexual behaviours of justice-involved adolescents in Australia
DESCRIPTION
Have you ever been vaccinated for HPV?
If YES, how many doses of HPV vaccine did you receive?
Justice-involved population
Community based
MeH-JOSH
p
value
0.025
Male
n (%)
293
Female
n (%)
172
76 (26.2)
63 (38.4)
137
(47.2)
64 (39.0)
77 (26.6)
37 (22.6)
3 (-)
8 (-)
Total
n (%)
465
139
(30.6)
201
(44.3)
114
(25.1)
11 (-)
290
(100.0)
164
(100.0)
454
(100.0)
11 (14.7)
11 (17.5)
22 (15.9)
0.764
14 (18.7)
13 (20.6)
27 (19.6)
10 (13.3)
5 (7.9)
15 (10.9)
Yes�
No
Never heard of it /Don’t
know/Unsure
Prefer not to say/
Missing��
Total
1 Dose
2 Doses
3 Doses
Don’t know/Unsure
40 (53.3)
34 (54.0)
74 (53.6)
Prefer not to say/
Missing��
1 (-)
0 (-)
1 (-)
If NO/NEVER HEARD OF IT/DON’T KNOW/UNSURE, do you want the HPV vaccine to
protect you against [FEMALE: cervical cancer] [MALE: cancers of the anus, mouth/throat, or
penis]?
Total
No
Yes
75
(100.0)
63
(100.0)
138
(100.0)
83 (38.8)
25 (24.8)
113
(52.8)
69 (68.3)
0.030
108
(34.3)
182
(57.8)
25 (7.9)
315
(100.0)
�Of 139 who reported HPV vaccination, Aboriginal and/or Torres Strait Islander (Indigenous) 37% (n = 51), Non-Indigenous 63% (n = 87)
��Prefer not to say/Missing not included in the total n (%) and p-value calculations
https://doi.org/10.1371/journal.pone.0243633.t005
Don’t know / Unsure
18 (8.4)
7 (6.9)
Total
214
(100.0)
101
(100.0)
young people). The study data were compared with a sample of similarly aged secondary stu-
dents (SSASH) [16] and interpretations of the results should take into account the socio-demo-
graphic differences between the two study populations (e.g. education and Aboriginality).
Sexual behaviours were also contrasted from YMM (2014) [15] and the Young People in Cus-
tody Health Survey in New South Wales (2015) [3] to situate the data with other studies in the
general population and in juvenile detention. Another limitation of this study is that it was not
possible to conduct formal comparisons across these different survey samples.
Findings are based on self-report and may be impacted by recall or social desirability bias.
While memory recall is not altogether an accurate indicator other studies in Australia also reveal
similar results for similar questions on HPV (e.g. SSASH 2018), most do not recall if they had
been vaccinated for HPV which is an interesting finding in itself. Perhaps in future studies we
can ask respondents for their consent to match them up with their vaccination records.
Since this was an exploratory study, we did not hypothesise about differences between the
incarcerated youth justice population and those with less involvement. Differences stratified
by involvement with the justice system can be explored in future.
PLOS ONE | https://doi.org/10.1371/journal.pone.0243633 December 28, 2020
16 / 19
PLOS ONE The sexual behaviours of justice-involved adolescents in Australia
This research adds to the dearth of surveys of justice involved young people in the commu-
nity [4, 22]. The study demonstrates how young people involved in the justice system may be
at high risk of STIs and that it is possible to target this usually hard to access population
through youth oriented community spaces without juvenile justice or corrective services
involvement. Permission for access and research from these government departments is some-
times an arduous or impossible process despite the urgent need to address serious health issues
in this population. Reaching justice involved young people in the community presents another
viable alternative to enhance the delivery of sexual health promotion programmes and services
and creates an opportunity to augment it with existing health services (mental health, alcohol
and substance abuse) currently provided in these youth oriented community and drop-in
centres.
Supporting information
S1 File.
(DOCX)
Acknowledgments
We would like to acknowledge the participation and assistance by the Western Australia
Department of Justice, Western Australia Department of Health, Queensland and Western
Australia community youth based organisations and drop-in centres, flexi-schools and col-
leges, and young people who agreed to be part of the survey as a whole in the conduct of this
research.
Author Contributions
Conceptualization: Lorraine Yap.
Data curation: Lorraine Yap.
Formal analysis: Lorraine Yap, Megan Steele, Christopher Fisher, Bianca Ton.
Funding acquisition: Lorraine Yap, Basil Donovan, Sally Nathan, Elizabeth Sullivan, Tony
Butler.
Investigation: Lorraine Yap, Jocelyn Jones, Basil Donovan, Sally Nathan, Sophie Davison,
Alun Richards, Tony Butler.
Methodology: Lorraine Yap, Basil Donovan.
Project administration: Lorraine Yap, Jocelyn Jones, Ed Heffernan.
Resources: Ed Heffernan.
Supervision: Lorraine Yap.
Visualization: Lorraine Yap, Bianca Ton.
Writing – original draft: Lorraine Yap.
Writing – review & editing: Lorraine Yap, Jocelyn Jones, Basil Donovan, Sally Nathan, Eliza-
beth Sullivan, Sophie Davison, Ed Heffernan, Alun Richards, Carla Meurk, Megan Steele,
Christopher Fisher, Bianca Ton, Tony Butler.
PLOS ONE | https://doi.org/10.1371/journal.pone.0243633 December 28, 2020
17 / 19
PLOS ONE The sexual behaviours of justice-involved adolescents in Australia
References
1. Australian Institute of Health and Welfare (2018) National data on the health of justice-involved young
people: a feasibility study. Cat. no. JUV 125. Canberra: AIHW. 2018.
2. Australian Institute of Health and Welfare, Australia’s welfare 2015. Australia’s welfare series no. 12.
Cat. no. AUS 189. Canberra: AIHW. 2015.
3. Young People in Custody Health Survey: Full Report. Justice Health & Forensic Mental Health Network
and Juvenile Justice NSW. NSW Government: Malabar, Sydney. 2015.
4. Kenny D., et al., Young offenders on community orders: Health, welfare and criminogenic needs. Syd-
ney University Press: Australia, 2008.
5. Robbins R.N. and Bryan A., Relationships Between Future Orientation, Impulsive Sensation Seeking,
and Risk Behavior Among Adjudicated Adolescents. J Adolesc Res, 2004. 19(4): p. 428–445. https://
doi.org/10.1177/0743558403258860 PMID: 16429605
6. Bryan A.D., Schmiege S.J., and Magnan R.E., Marijuana use and risky sexual behavior among high-
risk adolescents: trajectories, risk factors, and event-level relationships. Dev Psychol, 2012. 48(5): p.
1429–42. https://doi.org/10.1037/a0027547 PMID: 22390666
7. Elkington K.S., et al., HIV/sexually transmitted infection risk behaviors in delinquent youth with psychiat-
ric disorders: a longitudinal study. J Am Acad Child Adolesc Psychiatry, 2008. 47(8): p. 901–11. https://
doi.org/10.1097/CHI.0b013e318179962b PMID: 18645421
8.
9.
Teplin L.A., et al., Major mental disorders, substance use disorders, comorbidity, and HIV-AIDS risk
behaviors in juvenile detainees. Psychiatr Serv, 2005. 56(7): p. 823–8. https://doi.org/10.1176/appi.ps.
56.7.823 PMID: 16020814
Teplin L.A., et al., HIV and AIDS risk behaviors in juvenile detainees: implications for public health pol-
icy. Am J Public Health, 2003. 93(6): p. 906–12. https://doi.org/10.2105/ajph.93.6.906 PMID:
12773351
10. DiClemente R.J., et al., Comparison of AIDS knowledge, attitudes, and behaviors among incarcerated
adolescents and a public school sample in San Francisco. Am J Public Health, 1991. 81(5): p. 628–30.
https://doi.org/10.2105/ajph.81.5.628 PMID: 2014866
11. Australian Institute of Health and Welfare 2019. Youth Justice in Australia 2017–18. Cat. no. JUV 129.
Canberra: AIHW.
12. Australian Institute of Health and Welfare, Youth justice fact sheet no. 85. Types of community-based
youth justice supervision: 2015–16. Cat. no. JUV 107. Canberra: AIHW. 2017.
13. Australian Institute of Health and Welfare Youth Justice. Table S31: Australian population aged 10–17
by Indigenous status, states and territories, December 2013 to December 2017. Available at: https://
www.aihw.gov.au/getmedia/993d8e77-b18e-4482-b9b9-12fe838db158/aihw-juv-128-youth-detention-
population-in-Australia-2018-data-tables.xlsx.aspx
14. Bird S., Consent to medical treatment: the mature minor. Australian Family Physician, 2011. 40: p.
159–160. PMID: 21597522
15.
16.
Lawrence D, et al., The Mental Health of Children and Adolescents. Report on the Second Australian
Child and Adolescent Survey of Mental Health and Wellbeing. 2015, Department of Health Canberra.
Fisher CM, et al., 6th National Survey of Australian Secondary Students and Sexual Health 2018,
(ARCSHS Monograph Series No. 113), Bundoora: Australian Research Centre in Sex, Health & Soci-
ety, La Trobe University. 2019.
17. Rissel C., et al., First vaginal intercourse and oral sex among a representative sample of Australian
adults: the Second Australian Study of Health and Relationships. Sex Health, 2014. 11(5): p. 406–15.
https://doi.org/10.1071/SH14113 PMID: 25376994
18. Ward J., et al., The GOANNA Survey: Sexual Health and relationships in young Aboriginal and Torres
Strait Islander people. Results from the first national study assessing knowledge, risk practices and
health service use in relation to sexually transmitted infections and blood borne viruses. 2014 https://
www.baker.edu.au/Assets/. . ./Final%20Goanna%20Report%20July%202014.pdf (Last accessed 1
April 2019).
19. Brotherton J., et al., HPV vaccine coverage is increasing in Australia. Med J Aust, 2017. 206: 262.
https://doi.org/10.5694/mja16.00958 PMID: 28359009
20. Australian Institute of Health and Welfare (AIHW) Youth justice in Australia 2016–17 Data Tables
(Table S5a in aihw-juv-116-characteristics-of-young-people-under-supervision-S1-to-S35-data-table.
xlsx). http://www.aihw.gov.au/reports/youth-justice/youth-justice-in-australia-2016-17/data (Last
accessed 20 November 2018).
21. Kirby Institute. HIV, viral hepatitis and sexually transmissible infections in Australia: annual surveillance
report. Sydney: Kirby Institute, UNSW Sydney. 2018.
PLOS ONE | https://doi.org/10.1371/journal.pone.0243633 December 28, 2020
18 / 19
PLOS ONE The sexual behaviours of justice-involved adolescents in Australia
22. Kinner S.A., et al., Complex health needs in the youth justice system: a survey of community-based and
custodial offenders. J Adolesc Health, 2014. 54(5): p. 521–6. https://doi.org/10.1016/j.jadohealth.2013.
10.003 PMID: 24287014
PLOS ONE | https://doi.org/10.1371/journal.pone.0243633 December 28, 2020
19 / 19
PLOS ONE |
10.1371_journal.pone.0246344 | RESEARCH ARTICLE
Oncofertility care for newly diagnosed girls
with cancer in a national pediatric oncology
setting, the first full year experience from the
Princess Ma´xima Center, the PEARL study
1*, Anne-Lotte L. F. van der Kooi2, Marianne D. van de
M. E. Madeleine van der PerkID
Wetering1, Irene M. IJgosse1, Eline van Dulmen-den Broeder1, Simone L. Broer3, Aart
J. Klijn4, A. Birgitta Versluys1, Brigitte Arends3, Ralph J. A. Oude Ophuis3, Hanneke M. van
Santen1,5, Alida F. W. van der Steeg1, Margreet A. Veening1, Marry M. van den Heuvel-
Eibrink1, Annelies M. E. Bos3
1 Princess Ma´xima Center for Pediatric Oncology, Utrecht, Netherlands, 2 Department of Obstetrics and
Gynecology, Erasmus MC–University Medical Center Rotterdam, Rotterdam, Netherlands, 3 Reproductive
Medicine and Gynecology, University Medical Center Utrecht, Utrecht, Netherlands, 4 Pediatric Urology,
University Medical Center Utrecht—Wilhelmina Children’s Hospital, Utrecht, Netherlands, 5 Pediatric
Endocrinology, University Medical Center Utrecht—Wilhelmina Children’s Hospital, Utrecht, Netherlands
* m.e.m.vanderperk@prinsesmaximacentrum.nl
Abstract
Background
Childhood cancer patients often remain uninformed regarding their potential risk of gonadal
damage. In our hospital we introduced a five step standard oncofertility care plan for all
newly diagnosed female patients aiming to identify, inform and triage 100% of patients and
counsel 100% of patients at high risk (HR) of gonadal damage. This observational retro-
spective study (PEARL study) evaluated the use of this standard oncofertility care plan in
the first full year in a national cohort.
Methods
The steps consist of 1)timely (preferably before start of gonadotoxic treatment) identification
of all new patients, 2)triage of gonadal damage risk using a standardized gonadal damage
risk stratification tool, 3)informing all patients and families, 4)counseling of a selected subset
of girls, and 5) fertility preservation including ovarian tissue cryopreservation (OTC) in HR
patients using amended Edinburgh criteria. A survey of the medical records of all girls newly
diagnosed with cancer the first year (1-1-2019 until 31-12-2019) was conducted.
Results
Of 261 girls, 228 (87.4%) were timely identified and triaged. Triage resulted in 151 (66%)
low(LR), 32 (14%) intermediate(IR) and 45 (20%) high risk(HR) patients. Ninety-nine fami-
lies were documented to be timely informed regarding gonadal damage risk. In total, 35 girls
(5 LR, 5 IR, 25 HR) were counseled by an oncofertility expert. 16/25 HR patients underwent
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: van der Perk MEM, van der Kooi A-LLF,
van de Wetering MD, IJgosse IM, van Dulmen-den
Broeder E, Broer SL, et al. (2021) Oncofertility care
for newly diagnosed girls with cancer in a national
pediatric oncology setting, the first full year
experience from the Princess Ma´xima Center, the
PEARL study. PLoS ONE 16(3): e0246344. https://
doi.org/10.1371/journal.pone.0246344
Editor: Alessio Paffoni, Infertility Unit, ASST
Lariana, ITALY
Received: October 26, 2020
Accepted: January 18, 2021
Published: March 5, 2021
Copyright: © 2021 van der Perk et al. This is an
open access article distributed under the terms of
the Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: The data is restricted
as it concerns sensitive and potentially identifying
patient information in a limited population within a
single institute. Ethical consent was obtained from
the METC Utrecht (Medical Ethical Committee
Utrecht) and the need for informed consent was
waived for this retrospective observational part of
the PEARL study (METC research file
NL72115.041.19 version 2, METC-protocol
number 19/783, Netherlands trial register number
PLOS ONE | https://doi.org/10.1371/journal.pone.0246344 March 5, 2021
1 / 18
PLOS ONE NL8192). Thus patients did not provide informed
consent for data to be made public in this manner.
However, data can be shared with interested
researchers upon request via direct contact with
the principle investigators, or via Harm van
Tinteren (contact via H.
vanTinteren@prinsesmaximacentrum.nl).
Funding: M.E.M. van der Perk is supported by
funding from the Princess Ma´xima Center
Foundation and the Child health boost grant 2017.
The funders had no role in study design, data
collection and analysis, decision to publish, or
preparation of the manuscript.
Competing interests: The authors have declared
that no competing interests exist.
Oncofertility care for girls in the Princess Ma´ xima Center
fertility preservation (1 ovariopexy + OTC, oocyte cryopreservation (1 with and 1 without
OTC) and 13 OTC). Fertility preservation did not lead to complications or delay of cancer
treatment in any patient.
Conclusion
We timely identified and triaged most girls (88%) with cancer with a high risk of gonadal
damage to be counseled for fertility preservation. We aim to optimize the oncofertility care
plan and the standardized gonadal damage risk stratification tool based on this experience
and these may be of value to other pediatric oncology centers.
Introduction
Childhood cancer treatment is accompanied by multiple direct and late toxicities [1]. Prefera-
bly, these toxicities are anticipated and ideally prevented already prior to and during cancer
treatment. Impaired future fertility is a major concern for childhood cancer patients and their
parents. In the past, awareness of infertility as unexpected sequelae was raised only after reach-
ing adulthood [2, 3]. Currently, patients, parents, survivors and healthcare professionals
acknowledge the importance of discussing the risk of premature ovarian insufficiency (POI)
and consequent infertility due to gonadotoxic cancer treatment [4, 5]. This includes the need
for preservation already at an early stage even before the start of cancer therapy [4–8]. It is
challenging however, to timely identify patients at risk, to triage and to inform patients of
gonadal damage risk on an individual basis before cancer treatment, and to offer the possibility
of further counseling for fertility preservation by a fertility expert, without delay of cancer
treatment in full cohorts of pediatric patients. The clinical focus upon presentation with new
oncologic disease is often on the diagnostic process and swift stratification towards the most
effective cancer treatment rather than on preventing potential toxicities. Therefore, our dedi-
cated oncofertility working group created a standardized and easy to apply oncofertility care
plan including a gonadal damage risk stratification tool, as this was deemed indispensable to
ensure adequate and timely oncofertility care. The aim of the care plan was to identify, inform
and triage 100% of all new patients and to timely refer 100% of the girls at high risk (HR) of
gonadal damage for expert fertility counseling. This observational retrospective PEARL (Pre-
sErving ovARian function through cryopreservation and informing girLs with cancer about
infertility due to gonadotoxic treatment) study evaluated the use of this standard oncofertility
care plan in the first year 2019 in a full national cohort after centralization of pediatric oncol-
ogy care.
Methods
Pediatric oncology care in the Netherlands, previously dispersed in 7 expertise centers, was
merged into one national center, the Princess Ma´xima Center for pediatric oncology in
Utrecht in May 2018. This centralization serves the mission to further improve cure rates,
while also decreasing early and late toxicity.
The female oncofertility care plan
Since 2015 a multidisciplinary dedicated team with representatives of all departments (S1
Table in S1 Appendix), prepared an oncofertility care plan for newly diagnosed children
PLOS ONE | https://doi.org/10.1371/journal.pone.0246344 March 5, 2021
2 / 18
PLOS ONE Oncofertility care for girls in the Princess Ma´ xima Center
Fig 1. Oncofertility care plan 2019 Princess Ma´xima Center for newly diagnosed girls with cancer. ALL = Acute Lymphoblastic Leukemia;
NHL = Non-Hodgkin lymphoma; OTC-o = ovarian tissue cryopreservation of complete ovary; OTC-p = ovarian tissue cryopreservation of partial ovary.
https://doi.org/10.1371/journal.pone.0246344.g001
respecting their right to receive personalized oncofertility care [9]. The oncofertility care plan
for newly diagnosed girls (Fig 1) is based on international and national literature and profes-
sional experiences [9–12]. The content of our care plan had been intensively discussed with
and is approved of by the joined ethical committee of the UMC Utrecht and Princess Ma´xima
Center. This plan includes five steps (S2 Table in S1 Appendix).
1) Identification of all new patients.
Identifying newly diagnosed girls with cancer in our
center is pursued on a daily basis by a dedicated oncofertility nurse practitioner (coordinator)
together with the involved pediatric oncologist. The daily refreshed financial administration,
the tumor board lists and clinical ward rounds are used to timely identify new patients.
2) Triage gonadal damage risk. Pursuing standardized triage is done using a gonadal
damage risk stratification tool on risk of future gonadal damage. This clinically applicable and
easy to use gonadal damage risk stratification tool covers the (most commonly) used European
PLOS ONE | https://doi.org/10.1371/journal.pone.0246344 March 5, 2021
3 / 18
PLOS ONE Oncofertility care for girls in the Princess Ma´ xima Center
treatment protocols in the Netherlands. As pediatric oncology is a dynamic field, the risk strat-
ification is subject to updates with every new treatment protocol introduced in our hospital
and enhancements over time are possible with increasing knowledge. In 2019 the cyclophos-
phamide equivalent dose (CED) score, expected abdominal radiotherapy (with dose to the
ovary), hematopoietic stem cell transplantation (HSCT), and (ovarian) surgery served as the
basis premise for this gonadal damage risk stratification tool. In 2019 treatment protocols were
classified as LR, IR or HR of gonadal damage with CED scores of 0-4000mg/m2, 4000-
8000mg/m2 and more than 8000mg/m2 respectively including radiotherapy and surgery [13–
15]. Evidently, other factors such as very young age at diagnosis, prognosis, psychosocial and
ethical issues are taken into account upon counseling of patients. The oncofertility plan was
mirrored during a working visit to the oncofertility team of the Cincinnati Children’s Hospital,
USA, in September 2018 before implementation.
3) Informing patients. Timely (preferably before start of gonadotoxic treatment) all
patients are informed on their specific gonadal damage risk by their pediatric oncologist and/
or the oncofertility nurse practitioner (coordinator).
4) Counseling a subset of girls by a oncofertility specialist. Timely (before therapy or
not leading to delay of therapy) counseling by the oncofertility specialist (gynecologist) is avail-
able. In our institute this is available for patients at low risk (LR) and intermediate risk (IR) on
request, but actively encouraged in the subset of patients who are at high risk (HR) of gonadal
damage.
5) Fertility preservation. Preserving fertility is an option for highly selected, counseled,
eligible (HR) patients after shared decision making. Four different methods are available as
standard care fertility preservation in our hospital. a) Ovariopexy (OP), which is only useful in
girls in whom radiation to the ovary is expected to do significant (and chemotherapy minor or
no) damage. b) Oocyte harvest and cryopreservation (OC), after hormonal ovarian stimula-
tion, which is only feasible for oncology patients who are postmenarcheal and when treatment
can safely be postponed at least 2 weeks. c) Ovarian tissue cryopreservation (OTC) by unilat-
eral ovariectomy (OTC-o) or d) partial ovariectomy (OTC-p). The care plan offers OTC to all
girls newly diagnosed with cancer with high risk of gonadal damage in our center who, in gen-
eral, are aged 0–18 years at presentation. The DCOG amended-Edinburgh criteria (S3 Table in
S1 Appendix) are used to carefully confirm eligibility of girls for OTC in our center [9, 16, 17].
The content of the counseling by the oncofertility gynecologist is described in the S2 Table in
S1 Appendix.
Patients
Only newly diagnosed girls with pediatric cancer between 1 January 2019 and 31 December
2019 were included. The evaluation was conducted in spring 2020.
Evaluation of the full first year of oncofertility care implementation
Application of the standard oncofertility care plan for newly diagnosed girls and the five step
process was evaluated. Baseline data including age at diagnosis, type of malignancy, proposed
treatment(-protocol and arm), triage result (gonadal damage risk estimation) and curative or
palliative intention of treatment were collected. The date of diagnosis was defined as the date
of communicating the cancer diagnosis including the explanation of the intended treatment
by the pediatric oncologist with the family as recorded in the patient files. The date of the start
of chemotherapy was defined as the starting date of the intended treatment protocol. Date of
triage, information provision (pediatric oncologist or oncofertility nurse practitioner (coordi-
nator)) and, if applicable, date and content of the fertility preservation counseling were
PLOS ONE | https://doi.org/10.1371/journal.pone.0246344 March 5, 2021
4 / 18
PLOS ONE Oncofertility care for girls in the Princess Ma´ xima Center
retrieved from the medical records. Timely was defined as before starting cancer treatment,
with the exception of acute lymphoblastic leukemia (ALL) and non-Hodgkin lymphoma
(NHL). For these patients triage and informing of the family is generally postponed to the
moment of reaching complete remission (CR) or treatment arm allocation. In ALL and NHL
patients this moment of complete remission (CR) harbors an added benefit for eventual future
use of preserved tissue, because of decreased risk of harboring minimal residual disease or leu-
kemic infiltration. In addition, in renal tumor patients in our center the timely triage and
information moment is defined as after surgery when definitive treatment stratification,
including optional radiotherapy, is defined.
Statistical analyses
Descriptive statistics are reported including 95% confidence intervals (CI) for the main findings.
Ethical approval
Ethical consent was obtained from the METC Utrecht (Medical Ethical Committee Utrecht)
and the need for informed consent was waived for this retrospective observational part of the
PEARL study (METC research file NL72115.041.19 version 2, METC-protocol number 19/
783, Netherlands trial register number NL8192). Information was retrieved from the medical
records of all newly diagnosed girls in the Princess Ma´xima Center in 2019.
Results
In 2019, 261 girls with a median age of 8.4 years (range: 0.0–18.1) were newly diagnosed with
pediatric cancer in the Netherlands (Table 1, Fig 2A and 2B). Two died within days of presen-
tation prior to treatment allocation. Of the remaining 259, 228 (88.0% (95%CI: 0.835–0.914))
patients were timely identified and triaged (Fig 3). Of the 31 patients who had not been timely
identified and triaged, 28 were retrospectively classified as LR and 3 as HR. Characteristics of
triaged patients seem representative (S4 Table in S1 Appendix). Combining the triaged and
the non-triaged patients the 259 patients were classified as LR, IR and HR of gonadal damage
in 179 (69.1% (95%CI: 0.632–0.744)), 32 (12.4% (95%CI: 0.089–0.169)) and 48 (18.5% (95%CI:
0.143–0.237)) cases respectively.
Risk of expected gonadal damage had been communicated timely with parents and child in
99/228 (43.4% (95%CI: 0.372–0.499)) patients out of the aimed 100% of patients to inform. Of
these 42/151 (27.8% (95%CI: 0.213–0.354)) were LR, 28/32 (87.5% (95%CI: 0.719–0.950)) IR
and 29/45 (64.4% (95% CI: 0.498–0.768)) HR (Fig 3). Reasons for not informing timely triaged
patients were available for 125/129 cases, including 14 HR patients. Poor prognosis (n = 13),
low risk of gonadal damage (ALL n = 19, renal tumors n = 11, resection only n = 28, other
n = 20), psychosocial issues (n = 9), wait-and-scan regimens: no therapy (n = 14) and palliative
setting at diagnosis (n = 3) were recurrent reasons for not informing these timely triaged
patients (Table 2). Two patients in the HR group were initially classified as LR with resection
only. However, after treatment intensification their risk shifted to HR and subsequently these
patients had unfortunately not been informed again. Of 42 LR timely informed patients, 5 fam-
ilies requested and received further counseling by a fertility expert. Fertility preservation was
not advised nor pursued in these 5 patients.
Of 45/228 (20%) triaged HR patients, 29 families had received personalized information.
The reasons why 16 HR patients did not receive information after triage are depicted in
Table 2. In all 29 informed HR cases counseling by a dedicated oncofertility gynecologist was
strongly advised and 25 (86.2%) (of the aimed 100% counseled HR patients) appreciated such
counseling. The 4/45 HR families that did not wish referral for counseling felt that they had
PLOS ONE | https://doi.org/10.1371/journal.pone.0246344 March 5, 2021
5 / 18
PLOS ONE Oncofertility care for girls in the Princess Ma´ xima Center
Table 1. Characteristics of female childhood cancer patients admitted to the Princess Ma´xima Center in 2019.
Female childhood cancer patients diagnosed
n = 261�
Age at diagnosis (years)�� Median (range)
n
8.4 (0.0–18.1)
Diagnosis
Hematologic
• Acute Lymphoblastic Leukemia
• Acute Myeloid Leukemia
• Other leukemia’s
• Bone marrow failure/MDS
• Hodgkin lymphoma
• Non-Hodgkin lymphoma
Solid tumors
• Neuroblastoma
• Renal tumor
• Carcinoma (hepatic, gynecological)
• Osteosarcoma
• Ewing sarcoma
• Soft tissue sarcoma
• Germ cell tumor
• Skin cancer (incl. melanoma)
• Liver tumors
Neuro-oncology
• Brain tumor
Other (incl. LCH, Pheochromocytoma etc.)
100
43
9
3
4
23
8
104
22
17
8
6
4
11
19
3
7
57
57
17
%
-
38.3
16.5
3.4
1.1
1.5
8.8
3.1
39.8
8.4
6.5
3.1
2.3
1.5
4.2
7.3
1.5
2.7
21.8
21.8
6.5
MDS: myelodysplastic syndrome; LCH: Langerhans Cell Histiocytosis.
�including 2 patients who died premature within days of presentation prior to establishment of treatment
�� date at which diagnosis was discussed with patient.
https://doi.org/10.1371/journal.pone.0246344.t001
been sufficiently informed (three were eligible for preservation, but did not wish preservation
or counseling and one girl was very young (<6 months)).
In 16 out of the 25 HR girls counseling led to an attempt to preserve fertility (Table 3). One
patient underwent the combination of OC+OTC-o (n = 1). This patient initially had time to
delay treatment for oocyte harvest. However, due to the limited number of harvested oocytes
and insufficient time to perform another cycle, additionally OTC-o was performed. In five
patients fertility preservation was performed prior to the start of cytotoxic treatment, eleven
preserved gonadal tissue during treatment (S1 Fig in S1 Appendix). No intra-operative or
post-operative complications, such as wound infections, bleeding requiring transfusion, reop-
eration, ICU admittance or mortality were reported. Reasons for not preserving ovarian tissue
in HR patients mainly included uncertainty about the success of future auto-transplantation or
use of the tissue (Table 4). The option of OC in young mothers (<40 years) of pediatric cancer
patients for future oocyte donation was mentioned during counseling (S2 Table in S1 Appen-
dix). In 2019 none of the mothers pursued this option.
Discussion
This retrospective study in a national cohort evaluated our first year of centralized oncofertility
care in girls with cancer. It aims to search further improvement of quality of care. In the first
PLOS ONE | https://doi.org/10.1371/journal.pone.0246344 March 5, 2021
6 / 18
PLOS ONE Oncofertility care for girls in the Princess Ma´ xima Center
Fig 2. Age at diagnosis and highest received oncofertility care per age group of girls with cancer.
https://doi.org/10.1371/journal.pone.0246344.g002
full year this risk stratification seems to have led to timely identification and triage of 87.4% of
all newly diagnosed girls with cancer using a protocol(-arm) based oncofertility gonadal risk
stratification tool (Table 5). Fertility preservation in 16 patients did not cause delay in the
onset of oncological treatment. Of the 31 not timely identified and triaged patients, 28 were LR
PLOS ONE | https://doi.org/10.1371/journal.pone.0246344 March 5, 2021
7 / 18
PLOS ONE Oncofertility care for girls in the Princess Ma´ xima Center
Fig 3. Evaluation of oncofertility care for girls with cancer. OC = oocyte cryopreservation; OP = ovariopexy; OTC-o = ovarian tissue
cryopreservation via ovariectomy; OTC-p = ovarian tissue cryopreservation via strips. �2 early deaths occurred in neuro-oncology patients
and died prior to triage.
https://doi.org/10.1371/journal.pone.0246344.g003
and three were HR patients. Hence, although improvement is necessary, identification of most
HR patients seems feasible.
To ensure timely provision of information on potential gonadal damage to all patients, we
experienced the importance of daily and central coordination of identification and triage. For
that purpose we built a logistic administrative system. We learned that appointing a dedicated
oncofertility nurse practitioner who coordinates the navigation of all newly identified patients
in the hospital is of utmost importance. Subsequently, patients can be navigated through the
oncofertility triage system in due time, mostly before starting cancer treatment and in close
communication with the multidisciplinary team (S2 Table in S1 Appendix). Triage can be a
PLOS ONE | https://doi.org/10.1371/journal.pone.0246344 March 5, 2021
8 / 18
PLOS ONE Oncofertility care for girls in the Princess Ma´ xima Center
Table 2. Reasons patients were not informed after triage.
Low gonadal damage risk
Intermediate gonadal damage risk
High gonadal damage risk
Total
n = 109
n = 4
n = 16
n = 129
Poor prognosis
Low risk
Perspective of the oncologist: too burdensome
Family and psychosocial issues�
Palliative treatment
ALL LR
Renal tumor LR
Resection only LR��
Young age���
Long PICU admission in diagnostic phase
Wait and scan policy: no treatment
Unknown ����
10
20
1
3
3
19
11
28
0
0
14
0
0
0
0
1
0
0
0
1
0
0
0
2
3
0
2
5
0
0
0
2
1
1
0
2
13
20
3
9
3
19
11
31
1
1
14
4
ALL LR: Acute Lymphoblastic Leukemia at low risk of gonadal damage; LR: low risk; PICU: Pediatric Intensive Care Unit.
� In some situations the pediatric oncologist sensed that extra information other than the necessary treatment information would be too burdensome for parents. Some
patients had very complex family situations.
�� Two patients in the high risk group were initially classified as low risk with resection only, but after a change in treatment this risk shifted to high risk and patients
were not subsequently informed.
��� This patient was 6 months old at diagnosis.
����Unknown consist of one patient with a medulloblastoma and high risk of gonadal damage receiving ACNS0332. The oncologist of one patient with a neuroblastoma
stage 4 decided that fertility information at that time was not applicable and later when the treatment protocol changed to a high risk of gonadal damage no fertility
information was provided.
https://doi.org/10.1371/journal.pone.0246344.t002
Table 3. Girls in which active fertility preservation intervention was pursued.
Age at preservation (years) Median (range)
Malignancy type
Hematological �
Solid ��
Neuro-oncology
Type of preservation
OTC-o (complete ovariectomy)
OTC-p (partial ovariectomy)
Oocyte cryopreservation (OC)
OC + OTC (complete ovariectomy) ���
Ovariopexy (OP) + OTC (complete ovariectomy)
Fertility preservation in HR group
n = 16
4.6 (2.5–17.5)
4
12
0
13
0
1
1
1
HR: high risk; OTC = Ovarian Tissue Cryopreservation.
� The hematologic malignancies included 1 patient with Acute Lymphoblastic Leukemia–HR protocol, 1 Acute
Lymphoblastic Leukemia stem cell transplantation (SCT), 1 Acute Myeloid Leukemia SCT and 1 Myelodysplastic
syndrome SCT.
�� The solid tumors included 5 patients with neuroblastoma, 1 ovarian carcinoma, 3 Ewing sarcomas, 2 soft tissue
sarcoma and 1 bilateral metastasized Wilms tumor receiving stem cell transplantation.
��� This high risk patient initially had the time to delay treatment for oocyte harvest. However due to the limited
number of harvested oocytes and no time to perform another cycle, it was decided to additionally perform OTC.
https://doi.org/10.1371/journal.pone.0246344.t003
PLOS ONE | https://doi.org/10.1371/journal.pone.0246344 March 5, 2021
9 / 18
PLOS ONE Oncofertility care for girls in the Princess Ma´ xima Center
Table 4. Reasons why patients did not desire to preserve fertility after counseling.
Low or intermediate gonadal damage risk (LR, IR)
No guarantee of success auto-transplantation (HR)
Too burdensome on the child, preference for oocyte cryopreservation after age 16 (HR)
The risks do not outweigh the benefits and genetic parenthood is not the most important (HR)
Fertility is no issue due to genetic mutation (HR)
Preference for oocyte cryopreservation at 16 years (HR)
Total
https://doi.org/10.1371/journal.pone.0246344.t004
Frequency
10
4
2
1
1
1
19
%
3.9
1.5
0.8
0.4
0.4
0.4
7.3
complicated effort, which lies just outside the main priorities and expertise of pediatric oncolo-
gists. Therefore, we developed a standardized gonadal damage risk stratification tool, based on
international protocol(-arms), used at that time in our country. This was deemed instrumental
(Table 5). This tool improved over time and the most recent, currently used gonadal damage
risk stratification tool is available in the S5 Table in S1 Appendix.
We identified and triaged 82.2% of all girls aged 13 years and older and timely informed
49.3% of them (Fig 2a). Unfortunately, our retrospective study shows that 16 HR patients had
not been informed after triage. Several reasons were reported such as very poor expected out-
come, serious (co-)morbidity, psychosocial challenges and young age (Table 2). The three clin-
ical practice guidelines of the American Society of Clinical Oncology [10–12] contain
evidence-based recommendations for fertility preservation for patients with childhood cancer.
A study of compliance with these guidelines reported that none of the 136 patients above the
age of 13 had been counselled for fertility preservation [18, 19]. This illustrates how difficult
the oncofertility care logistics process can be in real life clinical practice. Evidently, the IR and
HR groups are the most relevant group to inform timely on their gonadal damage risk, as these
patients might be eligible for fertility preservation. Nevertheless, providing information also to
LR patients has been shown to be of value for survivors and family [20]. Previous surveys have
shown that many patients and families worry about future infertility at diagnosis already [20–
23]. Compared with available literature our percentage of informed patients is reasonably
high, even though, we did not reach our aimed 100% informed patients. More effort is needed
in the coming years to ensure that all patients with an acceptable cure rate are at least informed
about their fertility. Preferably, HR patients will also be referred for counseling to a fertility
expert (gynecologist) to explore the opportunities of fertility preservation.
We learned that as treatment is sometimes intensified (n = 2), gonadal damage risk may
increase and patients may need to be re-triaged, re-informed and that counseling may need to
be reconsidered. For that purpose, presence of an oncofertility nurse practitioner (coordina-
tor) at the multidisciplinary tumor board can enhance awareness in the oncology team. Why a
large proportion (mainly LR) was not timely informed was not always documented. Hence, we
learned it is important to facilitate a standard documentation process in the summary part of
medical records stating whether oncofertility information is provided, and if not the reason
why. This is now standard in our current practice. This is consistent with the recently pub-
lished consensus of the international guideline harmonization group (IGHG), which stated
that all childhood cancer patients and their families have the right to be informed regarding
their gonadal damage risk [24–26].
As indicated, adjustments have already been made in our current standard care oncoferti-
lity plan. We integrated recent recommendations from the IGHG guideline that classifies a
CED score of 6000mg/m2 as high risk instead of the 8000mg/m2 which we used in 2019 in our
gonadal damage risk stratification tool (S5 Table in S1 Appendix) [24–26]. Additionally, low
PLOS ONE | https://doi.org/10.1371/journal.pone.0246344 March 5, 2021
10 / 18
PLOS ONE Table 5. Gonadal damage risk stratification tool for European treatment protocols used for oncofertility care for girls in the Princess Ma´xima Center.
Tumor
Protocol
Treatment arm
CED mg/m2
Female gonadal
damage risk
Oncofertility care for girls in the Princess Ma´ xima Center
Hematologic malignancies
Acute Lymphoblastic
Leukemia�
ALL-11
Interfant06
EsPhALL
IntReALL
ALLTogether��
SR, MR
HR 1–3 +SCT
HR 1–6 + II
Germline LR/rearranged MR HR—SCT
rearranged MR HR + SCT
Arm A
Arm B
High risk arm
High risk arm
SR treatment arm A
SR treatment arm A with SCT
SR treatment arm B
SR treatment arm B with SCT
HR
R1 standard, experimental
R2 standard, Exp arm A, Exp arm B
R3 standard, Exp InO: IR-high risk
2000
5600 + SCT
9300
3000
3000+ SCT
9000
3000
3976 + SCT
5976 + SCT
1976
1976 + SCT
3400
3400 + SCT
1976 + SCT
0
3000
2000
ABL HR allo-SCT (� 1–3 NOPHO blocks)
2000 +SCT
ABL IR-high
HR BCP SCT 3 blocks
HR BCP chemo 6 blocks
HR T-cell without Nelarabine + HR blocks
HR T-cell with Nelarabine single
HR T-cell with Nelarabine single + HR blocks
HR T-cell with Nelarabine addition
HR T-cell with Nelarabine addition + HR blocks
DS-SR
DS-IR, DS-HR
LCH IV
stratum1 group 1 (MS-LCH) arm A / B / C / D
LCH
(Langerhans Cell
Histiocytosis)
Hodgkin lymphoma �
EuroNet-PHL-C2
Non-B NHL (Non-Hodgkin
Euro LB-02
Lymphoma)
stratum 1 group 2 (SS-LCH)
stratum 2
stratum V without clinical neurodegeneration
stratum V with clinical neurodegeneration
TL1
TL2
TL2 intensified
TL3
TL3 intensified
T-Cell LL stage I-II
T-Cell LL stage III-IV
non-T-Cell LL stage I-II
non-T-Cell LL stage III-IV
B-NHL/B-ALL �
SKION B-NHL/B-ALL
Group A
2000
4200 +SCT
7400
4200
1000
3200
2000
4200
1000
3000
0
0
0
0
0
1000
2000
2500
4000
5000
2000
3000
2000
3000
3000
Low
High
High
Low
High
High
Low
High
High
Low
High
Low
High
High
Low
Low
Low
High
Low
High
Intermediate
Intermediate
Low
Low
Low
Intermediate
Low
Low
Low
Low
Low
Low
Low
Low
Low
Low
Intermediate
Intermediate
Low
Low
Low
Low
Low
(Continued )
PLOS ONE | https://doi.org/10.1371/journal.pone.0246344 March 5, 2021
11 / 18
PLOS ONE Table 5. (Continued)
Tumor
Protocol
Treatment arm
CED mg/m2
Female gonadal
damage risk
Oncofertility care for girls in the Princess Ma´ xima Center
Hematologic malignancies
(B-cell Non-Hodgkin
Lymphoma/
acute lymphoblastic
leukemia)
2008
Inter-B-NHL ritux
Anaplastic Large Cell
Lymphoma
ALCL
Acute Myeloid Leukemia �
Nopho DBH AML
Acute Promyelocytic
Leukemia
Solid tumors
Neuroblastoma �
2012
ICC APL 01
ICC APL 02
DCOG NBL 2009
Group B
Group C1
Group C2
Group B HR
Group C1
Group C3
PMLBL
LR
SR arm 1 SR arm 3, HR arm 1, HR arm 2, HR
arm 3, HR arm 4
SR MRD- / SR MRD+ / HR
SR, HR
OG without N4
OG with 1x N4
OG with 2x N4
OG with 3x N4
OG with 4x N4
MR without N4
MR with N4
HR without N4
HR with N4
DCOG NBL 2009 <1yr
OG with 1x N4 <1yr
OG with 2x N4 <1yr, 3x N4 <1yr, 4x N4 <1yr
MR without N4 <1yr, with N4 <1yr
HR without N4 <1yr, with N4 <1yr
Ewing�
Ewing 2008
Osteosarcoma�
EURAMOS 1
R1 female
R3
R3 + TreoMel
MAP
MAPIE
Renal tumors�
UMBRELLA 2016/SIOP 2001���
AV + AVD, AV + AV1, AV + AV2
Rhabdoid tumors �
of the kidney
(RTK) or of soft
tissue (MRT)
EpSSG NRSTS 2005
EURHAB <18 mo
EURHAB <18 mo HD
EURHAB >18 mo
EURHAB >18 mo HD
AV + HR
Cyclophosphamide
3x DOX, 3x ICE, 3xVCA
2x DOX, 2x ICE, 2x VCA + CARBO Thiotepa
3x DOX, 3x ICE, 3xVCA + RT
2x DOX, 2x ICE, 2x VCA + CARBO Thiotepa
+ RT
NRSTS�
EpSSG NRSTS 2005
(Non-Rhabdomyosarcoma
Soft
3x ifosfamide
4x ifosfamide
PLOS ONE | https://doi.org/10.1371/journal.pone.0246344 March 5, 2021
3300
6800
6800
3300
5800
5800
4500
3352
6328
without SCT
with SCT
0
0
0
2100
4200
6300
8400
10290
18690
12690
21090
/kg
/kg
/kg
/kg
25176
25176
30776
0
14640
0
8100
17000
8892
50928
8892
50928
6588
8784
Low
Intermediate
Intermediate
Low
Intermediate
Intermediate
Intermediate
Low
Intermediate
Low
High
Low
Low
Low
Low
Intermediate
Intermediate
High
High
High
High
High
Low
Intermediate
High
High
High
High
High
Low
High
Low
High
High
High
High
High
High
Intermediate
High
(Continued )
12 / 18
PLOS ONE Table 5. (Continued)
Tumor
Protocol
Treatment arm
CED mg/m2
Female gonadal
damage risk
Oncofertility care for girls in the Princess Ma´ xima Center
Hematologic malignancies
Tissue Sarcoma)
Soft tissue sarcomas�
EpSSG
RMS2005
5x ifosfamide
6x ifosfamide
LR subgroup A
SR subgroup B
SR subgroup C (9x Ifosfamide)
SR subgroup C (5xIfosfamide)
SR subgroup D (9x Ifosfamide)
HR and group A + group C
HR and group A + group D
HR and group B + group C
HR and group B + group D, VHR
Germ cell tumor
SIOP CNS GCT II
NGGCT
Liver tumors
Hepatocellular
carcinoma
PHITT
group A1 very low risk HB
group A2 very low risk HB
group B1 Low risk HB / B2
group C intermediate risk SIOPEL3HR / C5VD/
CDDP-M
10980
13176
0
5800
13176
7320
13176
13176
17376
13176
17376
7320
0
0 (cisplatin)
0 (cisplatin)
0 (cisplatin)
Brain tumors
Opticus glioma
Intradural-extramedulary
tumor
Medulla blastoma
AT/RT (Atypical
teratoid/rhabdoid
tumors)
SIOP LGG 2004
HIT-MED + SCT
SR ACNS0331
HR ACNS0332
EURHAB <18 mo
EURHAB <18 mo HD
EURHAB >18 mo
EURHAB >18 mo HD
group D1 high risk HB SIOPEL4, D2 high risk
HB CDCE, CDVI
0 (cis/
carboplatin)
group E1 resected HCC
group E2 resected HCC PLADO
group F unresected/metastatic PLADO sorafenib,
GEMOX
0
0 (cisplatin)
0 (cisplatin)
Vincristine, carboplatin, etoposide. (In case of
allergy: cyclo)
0
Cyclophosphamide, lomustine
cyclophosphamide
3x DOX, 3x ICE, 3xVCA
2x DOX, 2x ICE, 2x VCA + CARBO Thiotepa
3x DOX, 3x ICE, 3xVCA + RT
2x DOX, 2x ICE, 2x VCA + CARBO Thiotepa
+ RT
49500
13200
12000
8892
50928
8892
50928
8540
0
17600
13200
0
7200
Dysgerminoma
WHO IV SIOP CNS GCT II HR-non-
germinoma
PEI
High grade glioma, Pons
glioma
ACNS0126
Medulloblastoma
PNET 5
Temozolamide
MB-SR / MB-WNT-HR(>16years)
MB-WNT-HR (<16years)
MB-SHH-TP53: No alkylating agents
HGG (High grade glioma)
Infant HGG 2013/HIT SKK
Elements IIs IIIs/1 IIIs/2 IVs
�Total Body irradiation, full abdominal/pelvic radiation upgrades towards high risk. Expected unilateral removal of an ovary as part of the oncologic treatment in
combination with gonadotoxic chemotherapy is also classified as high risk and OTC must be discussed.
�� The ALLTogether protocol was not used in 2019. But in anticipation of the starting study, this was already included in this overview.
���The UMBRELLA protocol was initiated in 2019 an prior to that the SIOP-2001 protocol was used, so both protocols were included.
https://doi.org/10.1371/journal.pone.0246344.t005
PLOS ONE | https://doi.org/10.1371/journal.pone.0246344 March 5, 2021
13 / 18
High
High
Low
Intermediate
High
Intermediate
High
High
High
High
High
Intermediate
Low
Low
Low
Low
Low
Low
Low
Low
Low
High
High
High
High
High
High
High
High
Unknown
High
High
Low
Intermediate
PLOS ONE Oncofertility care for girls in the Princess Ma´ xima Center
risk does not mean no risk, as patients with a very low risk may experience infertility after can-
cer treatment [27]. We should take into account that individual susceptibility and genetic vari-
ation may also influence individual risk of gonadal damage [27–29]. The optimal moment to
provide information regarding gonadal damage risk has been discussed in our oncofertility
working group. Previous studies showed that for both patients and parents the preferred
moment is at diagnosis, before cytotoxic medication is applied [2, 20–22]. This corresponds to
our aim to inform all girls at diagnosis as intervention is still feasible then. However, we dis-
covered that for female ALL and NHL patients the moment of triage and information provi-
sion can be postponed to the moment of CR or treatment arm allocation. We amended this
early in the implementation phase of the oncofertility care plan. In addition, over time, for
children with renal tumors (with the exception of full blown ruptured patients [30]), we post-
poned the information process until after surgery (4–6 weeks), as the final gonadotoxic treat-
ment stratification takes place based on histological stage and subtype. More recently, we
learned that choosing the moment of ovarian preservation in patients with large abdominal
tumors, such as neuroblastoma, 3–6 weeks into treatment may be beneficial for surgical and
safety reasons, despite the adverse gonadotoxic influence of 1 or 2 courses of chemotherapy.
Even though all patients may request additional counseling by experts, we learned this
opportunity is not always utilized. Of 45 HR patients only 25 were counseled and of these 25
patients, only 16 chose to preserve gonadal tissue. OTC-o was the most common procedure.
Oocyte cryopreservation before cancer treatment was no option for most girls in our cohort
due to young age and/or lack of opportunity to delay oncologic treatment (Table 4 and S1 Fig
in S1 Appendix). As no adverse events occurred, we consider OTC a safe procedure although
our numbers are obviously limited. From previous reports, only limited information is avail-
able on the safety of OTC [31]. In 2019, we chose not to not perform OTC-p to avoid previ-
ously reported bleeding risks and as the majority of our population was very young with small
ovaries [31]. Although age is no absolute contra-indication for OTC, we are hesitant to per-
form OTC in children under the age of 1 year in our center based on the published suggested
potential higher anesthesia risk in infants [32]. So far, we did not perform ovarian tissue cryo-
preservation in infants under the age of 1 year. Nevertheless, evidence for this anesthesia risk
in laparoscopic procedures is not strong [32–35]. The risk of gonadal damage will therefore
always need to be weighed against the risk of direct toxicity for individual patients. Thus to
infants who will, with no doubt, receive high dose HSCT (e.g. Juvenile myelomonocytic leuke-
mia (JMML)) or high dose total-abdominal radiotherapy (e.g. after extensive rupture at pre-
sentation in renal tumor or neuroblastoma patients), counseling will be offered and OTC
seriously considered.
Decisions to perform OTC were always based on shared decision making. One of the main
reasons for deciding against the OTC option was the communicated uncertainty of success of
future auto-transplantation (Table 3). Although studies in adults have shown promising
results, ex-vivo maturation of ovarian material harvested during childhood is still not pursued.
Future auto-transplantation of ovarian tissue from children is still considered experimental, in
contrast to OTC, which is now considered standard care [14, 36]. This is explicitly explained
to patient and parents during fertility counseling in our hospital (S2 Table in S1 Appendix).
Future research on auto-transplantation of ovarian tissue harvested in prepubertal girls will
shed more light on the effectiveness of, and may lead to more patients opting for, OTC in the
future [17]. Alternatively, OC after finalizing cytotoxic treatment is a feasible option for
patients older than 16 years. This can be done starting 1 year after the end of treatment and
with a sufficient ovarian reserve. However, there is substantial evidence that patients with
excessive doses of alkylating agents, local irradiation, or following HSCT already have dimin-
ished ovarian reserve. It is conceivable that they may not benefit from such procedures [37–
PLOS ONE | https://doi.org/10.1371/journal.pone.0246344 March 5, 2021
14 / 18
PLOS ONE Oncofertility care for girls in the Princess Ma´ xima Center
40]. Thus, also for these patients preventive strategies at diagnosis, as used in our oncofertility
plan, have a higher chance of creating fertility options in the future.
Even though most young mothers are informed about the option of OC, none pursued this
option in 2019. This may be influenced by the fact that the costs are not covered by insurance
companies, as they qualify as “social freezing”. However, we did not investigate the reasons in
this retrospective study. When this option is pursued the oocytes are stored under the mother’s
name to prevent children to feel obliged to use oocytes of the mother. When age permits,
mothers are advised to freeze their oocytes after the end of cancer treatment of their daughter
and thus at a less stressful time.
To improve our oncofertility care, the prospective part of the PEARL study currently evalu-
ates the oncofertility care from a patient and parent point of view. It will explore in depth the rea-
sons to preserve or not and the effect of ovarian tissue cryopreservation on the ovarian reserve.
Patients’ and parents’ recall of fertility information in cancer survivors is known to be limited
[41]. We also aim to evaluate whether information provided by the pediatric oncologist or the
dedicated nurse practitioner (coordinator) is deemed sufficient by LR and selected IR patients.
The prospective study will further analyze whether the provided information is consumed, com-
prehended, and still remembered at the moment of discontinuation of therapy. Furthermore, we
will analyze whether an information moment at the end of treatment would be a welcome addi-
tion to the quality of care of individual patients. In the prospective study the effectiveness of a
protocol(-arm) based oncofertility gonadal risk stratification tool will also be evaluated.
Conclusion
Our study suggests that it may be valuable and clinically feasible to timely identify and triage
87% of all newly diagnosed girls using a protocol(-arm) based oncofertility gonadal risk strati-
fication tool. OTC seems a safe procedure and shared decision making led to a highly selected
subgroup of patients for OTC. However, the safety of OTC needs to be confirmed in large pro-
spective studies. In our center implementing oncofertility care did not cause delay in the onset
of cancer treatment. We will continue to use the adjusted oncofertility care plan and evaluate
this in the prospective PEARL study which started in 2020. We hope that our oncofertility care
plan and risk stratification tool may be of use to other pediatric oncology institutes.
Supporting information
S1 Appendix.
(DOCX)
Acknowledgments
We acknowledge J.I. Geller, pediatric oncologist and O. Frias, fertility patient navigator, from
the Cincinnati Children’s Hospital, USA, for their lively discussions and their willingness to
critically review our oncofertility care plan in 2018.
Author Contributions
Conceptualization: M. E. Madeleine van der Perk, Anne-Lotte L. F. van der Kooi, Marianne
D. van de Wetering, Irene M. IJgosse, Eline van Dulmen-den Broeder, Simone L. Broer,
Aart J. Klijn, A. Birgitta Versluys, Brigitte Arends, Ralph J. A. Oude Ophuis, Hanneke M.
van Santen, Alida F. W. van der Steeg, Margreet A. Veening, Marry M. van den Heuvel-
Eibrink, Annelies M. E. Bos.
PLOS ONE | https://doi.org/10.1371/journal.pone.0246344 March 5, 2021
15 / 18
PLOS ONE Oncofertility care for girls in the Princess Ma´ xima Center
Data curation: M. E. Madeleine van der Perk.
Formal analysis: M. E. Madeleine van der Perk, Anne-Lotte L. F. van der Kooi, Marry M. van
den Heuvel-Eibrink, Annelies M. E. Bos.
Investigation: M. E. Madeleine van der Perk.
Methodology: M. E. Madeleine van der Perk, Anne-Lotte L. F. van der Kooi, Marry M. van
den Heuvel-Eibrink, Annelies M. E. Bos.
Project administration: M. E. Madeleine van der Perk, Anne-Lotte L. F. van der Kooi, Marry
M. van den Heuvel-Eibrink, Annelies M. E. Bos.
Supervision: Anne-Lotte L. F. van der Kooi, Marry M. van den Heuvel-Eibrink, Annelies M.
E. Bos.
Validation: M. E. Madeleine van der Perk, Anne-Lotte L. F. van der Kooi, Marry M. van den
Heuvel-Eibrink, Annelies M. E. Bos.
Visualization: M. E. Madeleine van der Perk, Anne-Lotte L. F. van der Kooi, Marry M. van
den Heuvel-Eibrink, Annelies M. E. Bos.
Writing – original draft: M. E. Madeleine van der Perk, Anne-Lotte L. F. van der Kooi, Marry
M. van den Heuvel-Eibrink, Annelies M. E. Bos.
Writing – review & editing: M. E. Madeleine van der Perk, Anne-Lotte L. F. van der Kooi,
Marianne D. van de Wetering, Irene M. IJgosse, Eline van Dulmen-den Broeder, Simone L.
Broer, Aart J. Klijn, A. Birgitta Versluys, Brigitte Arends, Ralph J. A. Oude Ophuis, Han-
neke M. van Santen, Alida F. W. van der Steeg, Margreet A. Veening, Marry M. van den
Heuvel-Eibrink, Annelies M. E. Bos.
References
1. Oeffinger KC, Mertens AC, Sklar CA, Kawashima T, Hudson MM, Meadows AT, et al. Chronic health
conditions in adult survivors of childhood cancer. N Engl J Med. 2006; 355(15):1572–82. https://doi.org/
10.1056/NEJMsa060185 PMID: 17035650
2. Anazodo A, Laws P, Logan S, Saunders C, Travaglia J, Gerstl B, et al. How can we improve oncofertility
care for patients? A systematic scoping review of current international practice and models of care.
Hum Reprod Update. 2019; 25(2):159–79. https://doi.org/10.1093/humupd/dmy038 PMID: 30462263
3. Reh AE, Lu L, Weinerman R, Grifo J, Krey L, Noyes N. Treatment outcomes and quality-of-life assess-
ment in a university-based fertility preservation program: results of a registry of female cancer patients
at 2 years. J Assist Reprod Genet. 2011; 28(7):635–41. https://doi.org/10.1007/s10815-011-9559-z
PMID: 21424818
4. Deshpande NA, Braun IM, Meyer FL. Impact of fertility preservation counseling and treatment on psy-
chological outcomes among women with cancer: A systematic review. Cancer. 2015; 121(22):3938–47.
https://doi.org/10.1002/cncr.29637 PMID: 26264701
5.
Letourneau JM, Ebbel EE, Katz PP, Katz A, Ai WZ, Chien AJ, et al. Pretreatment fertility counseling and
fertility preservation improve quality of life in reproductive age women with cancer. Cancer. 2012; 118
(6):1710–7. https://doi.org/10.1002/cncr.26459 PMID: 21887678
6. Skaczkowski G, White V, Thompson K, Bibby H, Coory M, Orme LM, et al. Factors influencing the provi-
sion of fertility counseling and impact on quality of life in adolescents and young adults with cancer. J
Psychosoc Oncol. 2018; 36(4):484–502. https://doi.org/10.1080/07347332.2018.1443986 PMID:
29764330
7.
Jardim FA, Lopes-Junior LC, Nascimento LC, Neves ET, Lima RAG. Fertility-Related Concerns and
Uncertainties in Adolescent and Young Adult Childhood Cancer Survivors. J Adolesc Young Adult
Oncol. 2020: https://doi.org/10.1089/jayao.2020.0058 Online ahead of print. PMID: 32945713
8. Newton K, Howard AF, Thorne S, Kelly MT, Goddard K. Facing the unknown: uncertain fertility in young
adult survivors of childhood cancer. J Cancer Surviv. 2020: https://doi.org/10.1007/s11764-020-00910-
x Online ahead of print. PMID: 32613442
PLOS ONE | https://doi.org/10.1371/journal.pone.0246344 March 5, 2021
16 / 18
PLOS ONE Oncofertility care for girls in the Princess Ma´ xima Center
9. Veening MA, Bos AME, Versluys AB, van Santen HM, van de Wetering MD, van den Heuvel-Eibrink
MM, et al. SKION consensus fertiliteitspreservatie voor meisjes met kanker, van 0–18 jaar. SKION
[Internet]. 2016 [cited 2020 Sept 18]. Available from: https://www.skion.nl/workspace/uploads/
Consensus-fertiliteitspreservatie-mei-2016.pdf. Dutch.
10.
11.
Lee SJ, Schover LR, Partridge AH, Patrizio P, Wallace WH, Hagerty K, et al. American Society of Clini-
cal Oncology recommendations on fertility preservation in cancer patients. J Clin Oncol. 2006; 24
(18):2917–31. https://doi.org/10.1200/JCO.2006.06.5888 PMID: 16651642
Loren AW, Mangu PB, Beck LN, Brennan L, Magdalinski AJ, Partridge AH, et al. Fertility preservation
for patients with cancer: American Society of Clinical Oncology clinical practice guideline update. J Clin
Oncol. 2013; 31(19):2500–10. https://doi.org/10.1200/JCO.2013.49.2678 PMID: 23715580
12. Oktay K, Harvey BE, Partridge AH, Quinn GP, Reinecke J, Taylor HS, et al. Fertility Preservation in
Patients With Cancer: ASCO Clinical Practice Guideline Update. J Clin Oncol. 2018; 36(19):1994–
2001. https://doi.org/10.1200/JCO.2018.78.1914 PMID: 29620997
13. Green DM, Nolan VG, Goodman PJ, Whitton JA, Srivastava D, Leisenring WM, et al. The cyclophos-
phamide equivalent dose as an approach for quantifying alkylating agent exposure: a report from the
Childhood Cancer Survivor Study. Pediatr Blood Cancer. 2014; 61(1):53–67. https://doi.org/10.1002/
pbc.24679 PMID: 23940101
14. Wallace WH, Anderson RA, Irvine DS. Fertility preservation for young patients with cancer: who is at
risk and what can be offered? Lancet Oncol. 2005; 6(4):209–18. https://doi.org/10.1016/S1470-2045
(05)70092-9 PMID: 15811616
15. Wallace WH, Kelsey TW, Anderson RA. Fertility preservation in pre-pubertal girls with cancer: the role
of ovarian tissue cryopreservation. Fertil Steril. 2016; 105(1):6–12. https://doi.org/10.1016/j.fertnstert.
2015.11.041 PMID: 26674557
16. Wallace WH, Smith AG, Kelsey TW, Edgar AE, Anderson RA. Fertility preservation for girls and young
women with cancer: population-based validation of criteria for ovarian tissue cryopreservation. Lancet
Oncol. 2014; 15(10):1129–36. https://doi.org/10.1016/S1470-2045(14)70334-1 PMID: 25130994
17.
van den Heuvel-Eibrink MM, van der Kooi ALF, Wallace WHB. Fertility Preservation in Women. N Engl
J Med. 2018; 378(4):399–400.
18. Anderson RA, Mitchell RT, Kelsey TW, Spears N, Telfer EE, Wallace WH. Cancer treatment and
gonadal function: experimental and established strategies for fertility preservation in children and young
adults. Lancet Diabetes Endocrinol. 2015; 3(7):556–67. https://doi.org/10.1016/S2213-8587(15)00039-
X PMID: 25873571
19. Salih SM, Elsarrag SZ, Prange E, Contreras K, Osman RG, Eikoff JC, et al. Evidence to incorporate
inclusive reproductive health measures in guidelines for childhood and adolescent cancer survivors. J
Pediatr Adolesc Gynecol. 2015; 28(2):95–101. https://doi.org/10.1016/j.jpag.2014.05.012 PMID:
25850590
20. Crawshaw MA, Glaser AW, Hale JP, Sloper P. Male and female experiences of having fertility matters
raised alongside a cancer diagnosis during the teenage and young adult years. Eur J Cancer Care
(Engl). 2009; 18(4):381–90. https://doi.org/10.1111/j.1365-2354.2008.01003.x PMID: 19594609
21. Ellis SJ, Wakefield CE, McLoone JK, Robertson EG, Cohn RJ. Fertility concerns among child and ado-
lescent cancer survivors and their parents: A qualitative analysis. J Psychosoc Oncol. 2016; 34(5):347–
62. https://doi.org/10.1080/07347332.2016.1196806 PMID: 27269305
22.
Lee S, Heytens E, Moy F, Ozkavukcu S, Oktay K. Determinants of access to fertility preservation in
women with breast cancer. Fertil Steril. 2011; 95(6):1932–6. https://doi.org/10.1016/j.fertnstert.2011.
01.169 PMID: 21371704
23. Stein DM, Victorson DE, Choy JT, Waimey KE, Pearman TP, Smith K, et al. Fertility Preservation Pref-
erences and Perspectives Among Adult Male Survivors of Pediatric Cancer and Their Parents. J Ado-
lesc Young Adult Oncol. 2014; 3(2):75–82. https://doi.org/10.1089/jayao.2014.0007 PMID: 24940531
24. Mulder RL, Font-Gonzalez A, Hudson MM, van Santen HM, Loeffen EAH, Burns KC, et al. Fertility pres-
ervation for female patients with childhood, adolescent, and young adult cancer: recommendations
from the PanCareLIFE consortium and the International Late Effects of Childhood Cancer Guideline
Harmonization Group. Lancet Oncol. 2021; 22: e45–56. https://doi.org/10.1016/S1470-2045(20)
30594-5 PMID: 33539753
25. Mulder RL, Font-Gonzalez A, van Dulmen—den Broeder E, Quinn GP, Ginsberg JP, Loeffen EAH,
et al. Communication and ethical considerations for fertility preservation for patients with childhood,
adolescent, and young adult cancer: recommendations from the PanCareLIFE Consortium and the
International Late Effects of Childhood Cancer Guideline Harmonization Group. Lancet Oncol. 2021;
21: e68–80.
26. Mulder RL, Font-Gonzalez A, Green DM, Loeffen EAH, Hudson MM, Loonen J, et al. Fertility preserva-
tion for male patients with childhood, adolescent, and young adult cancer: recommendations from the
PLOS ONE | https://doi.org/10.1371/journal.pone.0246344 March 5, 2021
17 / 18
PLOS ONE Oncofertility care for girls in the Princess Ma´ xima Center
PanCareLIFE consortium and the International Late Effects of Childhood Cancer Guideline Harmoniza-
tion Group. Lancet Oncol. 2021; 22: e57–67. https://doi.org/10.1016/S1470-2045(20)30582-9 PMID:
33539754
27.
Levine JM, Whitton JA, Ginsberg JP, Green DM, Leisenring WM, Stovall M, et al. Nonsurgical prema-
ture menopause and reproductive implications in survivors of childhood cancer: A report from the Child-
hood Cancer Survivor Study. Cancer. 2018; 124(5):1044–52. https://doi.org/10.1002/cncr.31121 PMID:
29338081
28. Clemens E, van der Kooi ALF, Broer L, van Dulmen-den Broeder E, Visscher H, Kremer L, et al. The
influence of genetic variation on late toxicities in childhood cancer survivors: A review. Crit Rev Oncol
Hematol. 2018; 126:154–67. https://doi.org/10.1016/j.critrevonc.2018.04.001 PMID: 29759558
29.
30.
van Dorp W, van den Heuvel-Eibrink MM, Stolk L, Pieters R, Uitterlinden AG, Visser JA, et al. Genetic
variation may modify ovarian reserve in female childhood cancer survivors. Hum Reprod. 2013; 28
(4):1069–76. https://doi.org/10.1093/humrep/des472 PMID: 23360674
van den Heuvel-Eibrink MM, Hol JA, Pritchard-Jones K, van Tinteren H, Furtwangler R, Verschuur AC,
et al. Position paper: Rationale for the treatment of Wilms tumour in the UMBRELLA SIOP-RTSG 2016
protocol. Nat Rev Urol. 2017; 14(12):743–52. https://doi.org/10.1038/nrurol.2017.163 PMID: 29089605
31. Corkum KS, Rhee DS, Wafford QE, Demeestere I, Dasgupta R, Baertschiger R, et al. Fertility and hor-
mone preservation and restoration for female children and adolescents receiving gonadotoxic cancer
treatments: A systematic review. J Pediatr Surg. 2019; 54(11):2200–9. https://doi.org/10.1016/j.
jpedsurg.2018.12.021 PMID: 30773394
32. Habre W, Disma N, Virag K, Becke K, Hansen TG, Johr M, et al. Incidence of severe critical events in
paediatric anaesthesia (APRICOT): a prospective multicentre observational study in 261 hospitals in
Europe. Lancet Respir Med. 2017; 5(5):412–25. https://doi.org/10.1016/S2213-2600(17)30116-9
PMID: 28363725
33.
34.
Iwanaka T, Uchida H, Kawashima H, Nishi A, Kudou S, Satake R. Complications of laparoscopic sur-
gery in neonates and small infants. J Pediatr Surg. 2004; 39(12):1838–41. https://doi.org/10.1016/j.
jpedsurg.2004.08.011 PMID: 15616945
Flick RP, Sprung J, Harrison TE, Gleich SJ, Schroeder DR, Hanson AC, et al. Perioperative cardiac
arrests in children between 1988 and 2005 at a tertiary referral center: a study of 92,881 patients. Anes-
thesiology. 2007; 106(2):226–37; quiz 413–4. https://doi.org/10.1097/00000542-200702000-00009
PMID: 17264715
35. Parelkar SV, Oak SN, Bachani MK, Sanghvi BV, Gupta R, Prakash A, et al. Minimal access surgery in
newborns and small infants; five years experience. J Minim Access Surg. 2013; 9(1):19–24. https://doi.
org/10.4103/0972-9941.107129 PMID: 23626415
36. Practice Committee of the American Society for Reproductive Medicine. Electronic address aao. Fertil-
ity preservation in patients undergoing gonadotoxic therapy or gonadectomy: a committee opinion. Fer-
til Steril. 2019; 112(6):1022–33. https://doi.org/10.1016/j.fertnstert.2019.09.013 PMID: 31843073
37. Carter A, Robison LL, Francisco L, Smith D, Grant M, Baker KS, et al. Prevalence of conception and
pregnancy outcomes after hematopoietic cell transplantation: report from the Bone Marrow Transplant
Survivor Study. Bone Marrow Transplant. 2006; 37(11):1023–9. https://doi.org/10.1038/sj.bmt.
1705364 PMID: 16604098
38. Mertens AC, Ramsay NK, Kouris S, Neglia JP. Patterns of gonadal dysfunction following bone marrow
transplantation. Bone Marrow Transplant. 1998; 22(4):345–50. https://doi.org/10.1038/sj.bmt.1701342
PMID: 9722069
39.
40.
41.
van Beek RD, van den Heuvel-Eibrink MM, Laven JS, de Jong FH, Themmen AP, Hakvoort-Cammel
FG, et al. Anti-Mullerian hormone is a sensitive serum marker for gonadal function in women treated for
Hodgkin’s lymphoma during childhood. J Clin Endocrinol Metab. 2007; 92(10):3869–74. https://doi.org/
10.1210/jc.2006-2374 PMID: 17726078
Lie Fong S, Laven JS, Hakvoort-Cammel FG, Schipper I, Visser JA, Themmen AP, et al. Assessment
of ovarian reserve in adult childhood cancer survivors using anti-Mullerian hormone. Hum Reprod.
2009; 24(4):982–90. https://doi.org/10.1093/humrep/den487 PMID: 19153092
van den Berg H, Langeveld NE. Parental knowledge of fertility in male childhood cancer survivors. Psy-
chooncology. 2008; 17(3):287–91. https://doi.org/10.1002/pon.1248 PMID: 17647263
PLOS ONE | https://doi.org/10.1371/journal.pone.0246344 March 5, 2021
18 / 18
PLOS ONE |
10.1371_journal.pone.0243709 | RESEARCH ARTICLE
Factors associated with admission to intensive
care units in COVID-19 patients in Lyon-France
1,2*, Marie-Paule Gustin1,2, Christelle EliasID
Philippe VanhemsID
Ce´ dric Dananche´ 1,2, Be´ atrice Grisi1,2, Elodie Marion1, Nagham Khanafer1,2,
Delphine Hilliquin1,2, Sophie Gardes1, Solweig Gerbier-Colomban1, Selilah Amour1,2,
Elisabetta Kuczewski1, Vanessa EscuretID
behalf of COVID-Outcomes-HCL Consortium
3,4, Bruno Lina3,4, Mitra Saadatian-Elahi1,2, On
1,2, Laetitia Henaff1,2,
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: Vanhems P, Gustin M-P, Elias C, Henaff
L, Dananche´ C, Grisi B, et al. (2021) Factors
associated with admission to intensive care units in
COVID-19 patients in Lyon-France. PLoS ONE
16(1): e0243709. https://doi.org/10.1371/journal.
pone.0243709
Editor: Surbhi Leekha, University of Maryland
School of Medicine, UNITED STATES
Received: May 30, 2020
Accepted: November 29, 2020
Published: January 27, 2021
Copyright: © 2021 Vanhems et al. This is an open
access article distributed under the terms of the
Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: All relevant data are
within the manuscript and its Supporting
Information files.
Funding: PV received partial funding by REACTing
(Research and ACTion targeting emerging
infectious diseases)- INSERM, France and a
donation from Fondation AnBer (http://
fondationanber.fr/). The funders had no role in the
study design, data collection and analysis, decision
to publish, or preparation of the manuscript. There
1 Service Hygiène, Epide´ miologie, Infectiovigilance et Pre´ vention, Centre Hospitalier Edouard Herriot,
Hospices Civils de Lyon, Lyon, France, 2 CIRI, Centre International de Recherche en Infectiologie (Equipe
Laboratoire des Pathogènes Emergents), Univ Lyon, Inserm, U1111, Universite´ Claude Bernard Lyon 1,
CNRS, UMR5308, Lyon, France, 3 Laboratoire de Virologie, Institut des Agents Infectieux, Hoˆ pital de la
Croix-Rousse, Hospices Civils de Lyon, Lyon, France, 4 Virpath—Grippe, de l’e´mergence au controˆ le,
Centre International de Recherche en Infectiologie (CIRI), Inserm U111, CNRS 5308, ENS, Faculte´ de
Me´decine RTH Lae¨nnec, Lyon, France
* Philippe.vanhems@chu-lyon.fr
Abstract
Introduction
A new respiratory virus, SARS-CoV-2, has emerged and spread worldwide since late 2019.
This study aims at analysing clinical presentation on admission and the determinants asso-
ciated with admission in intensive care units (ICUs) in hospitalized COVID-19 patients.
Patients and methods
In this prospective hospital-based study, socio-demographic, clinical and biological character-
istics, on admission, of adult COVID-19 hospitalized patients presenting from the community
for their first admission were prospectively collected and analysed. Characteristics of patients
hospitalized in medical ward to those admitted in ICU were compared using Mann-Whitney
and Chi-square or Fisher exact test when appropriate. Univariate logistic regression was first
used to identify variables on admission that were associated with the outcome i.e. admission
to an ICU versus total hospital stay in a medical ward. Forward selection was then applied
beginning with sex, age and temperature in the multivariable logistic regression model.
Results
Of the 412 patients included, 325 were discharged and 87 died in hospital. Multivariable regres-
sion showed increasing odds of ICU hospitalization with temperature (OR, 1.56 [95% CI, 1.06–
2.28] per degree Celsius increase), oxygen saturation <90% (OR, 12.45 [95% CI, 5.27–29.4]),
abnormal lung auscultation on admission (OR, 3.58 [95% CI, 1.58–8.11]), elevated level of
CRP (OR, 2.7 [95% CI, 1.29–5.66for CRP>100mg/L vs CRP<10mg/L). and monocytopenia
(OR, 3.28 [95% CI, 1.4–7.68]) were also associated with increasing odds of ICU hospitalization.
Older patients were less likely to be hospitalized in ICU (OR, 0.17 [95%CI, 0.05–0.51].
PLOS ONE | https://doi.org/10.1371/journal.pone.0243709 January 27, 2021
1 / 12
PLOS ONE was no additional external funding received for this
study.
Competing interests: The authors have read the
journal’s policy and have the following competing
interests: PV received grants and fees from Anios,
Pfizer, Astellas, MSD, Gilead and Sanofi. These
grants and fees were not related to this present
manuscript. This does not alter our adherence to
PLOS ONE policies on sharing data and materials.
There are no patents, products in development or
marketed products associated with this research to
declare. All other authors have no competing
interest to declare.
Characteristics of COVID-19 patients at hospital admission and factors of severity
Conclusions
Age and delay between onset of symptoms and hospital admission were associated with
the risk of hospitalisation in ICU. Age being a fixed variable, interventions that shorten this
delay would improve the prognosis of Covid-19 patients.
Introduction
Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), first detected in December
2019 in the Hubei province of China [1–3], was declared as a pandemic by the World Health
Organization on March 11, 2020. Coronavirus disease 2019 (COVID-19) is the emerging
infectious disease due to SARS-CoV-2, associated with lower or upper respiratory infection
even though less typical clinical features or asymptomatic cases have also been reported [4, 5].
The infection fatality rates of COVID-19 varies considerably, with a median of 0.27% across 32
locations included in a recent analysis of seroprevalence studies in the general population [6].
Crude case fatality rate ranges from 2% to 4% but can reach 12% to 15% in the elderly [7].
The first published cases of SARS-CoV2 infection in Europe were travellers from Wuhan
who were tested positive in France (two in Paris and one in Bordeaux) on January 24, 2020 [8].
As of September 24, 2020, 497,230 confirmed cases have been reported in France, including
31, 511 (20, 940 in hospitals) deaths [9]. The Auvergne-Rhoˆne-Alpes region located in the
southeast of France has a population of more than 6 million inhabitants. By April 24th, 1,287
patients, including 200 patients in ICUs were hospitalized in public and private structures in
Rhoˆne and Nord-Isère.
COVID-19 related complications, patient outcomes and mortality rates reported so far
have varied considerably between countries most probably owing to differences in healthcare
systems and the availability of ICU beds. Moreover, the prevalence of underlying chronic dis-
eases such as obesity and diabetes, known to be important determinants in the clinical course
and outcome of COVID-19 [10] are also different throughout the world. In addition, a large
number of published reports so far have described hospitalized COVID-19 patients with
incomplete data vis-à-vis hospital follow-up because a substantial proportion of patients
remained hospitalized at the time of manuscript submission or publication.
Knowledge of the baseline characteristics and outcomes of hospitalized COVID-19 patients
from different parts of the world is crucial for the decision-making process at national and
international levels in order to properly respond to the pandemic.
The aim of this study was to report the clinical features and outcomes of patients filling the
WHO case definition for confirmed COVID-19 and admitted to Lyon university-affiliated
hospitals with complete documentation of the hospital stay from February 8 to April 24, 2020.
Demographic, clinical and biological characteristics on admission associated with the risk of
ICU admission was assessed.
Methods
Study design and participants
This prospective, observational, hospital-based study (NOSO-COR, ClinicalTrials:
NCT04290780) is an ongoing international multicentre study carried out in France and hospi-
tals affiliated with the GABRIEL network [11]. The latter is a network of hospitals involved in
prospective studies on respiratory infection in emerging countries, leaded by Merieux
PLOS ONE | https://doi.org/10.1371/journal.pone.0243709 January 27, 2021
2 / 12
PLOS ONE Characteristics of COVID-19 patients at hospital admission and factors of severity
Foundation (www.https://www.gabriel-network.org/). However, the present paper was limited
to community-acquired COVID-19 confirmed patients admitted to four university-affiliated
hospitals in Lyon (Hospices Civils de Lyon, 5,300 beds).
Any adult patient who presented from the community with an infectious syndrome based
on the WHO definition of COVID-19 [12], admitted for the first time to one of the four partic-
ipating university-affiliated hospitals in Lyon between February 8 and April 24, 2020, and hos-
pitalized for a period of at least 24 hours, was included.
The study was approved by the clinical research and ethics committee of Ile de France V on
March 8, 2020 (No. 20.02.27.69817 Cat 3).
Data collection
Identification of community-acquired confirmed SARS-CoV-2 patients was based on a
daily extraction of real-time reverse transcriptase-polymerase chain reaction (RT-PCR)
positive patients from the virology laboratory. Electronic medical records were the main
source of data collection. Demographic characteristics, underlying comorbidities, clinical,
and biological parameters and patient outcome data were collected prospectively on an
electronic case-report form designed especially for the purpose of the project. Clinical out-
comes were monitored up to hospital discharge or death. All data were double-checked
after computerization.
Nasopharyngeal swab samples were collected as part of the standard care in patients present-
ing signs and symptoms of SARS-CoV-2 infection. Samples were transferred to the French
national reference centre of respiratory viruses for the detection of SARS-CoV2 by RT-PCR
[13]. Patients with positive RT-PCR results were defined as laboratory-confirmed SARS-CoV-2.
Statistical analysis
Given the descriptive nature of this observational study and the emergency context, no statisti-
cal sample size calculation was performed. Sample size was equal to the number of patients
included during the study period.
Continuous variables were reported as median and interquartile range (IQR). Categorical
variables were described as frequencies (%). We compared characteristics of patients hospital-
ized in medical ward to those admitted in ICU using Mann-Whitney and Chi-square or Fisher
exact test when appropriate.
Univariate logistic regression was first used to identify variables on admission that were
associated with the outcome i.e. admission to an ICU versus total hospital stay in a medical
ward. Forward selection was then applied beginning with three variables in the model: sex,
temperature and age or delay between onset of symptoms and hospital admission. Variables
that were significant at 0.15 levels in univariate analysis were first introduced one by one in
turn in the multivariate regression model. Interaction with covariates were tested and the most
significant variable was added in the model. This treatment was repeated with the remaining
variables until reduction in the deviance between the current and the previous model was still
significant at 0.05 level with no excessively large ORs’ confidence interval. Goodness of fit of
the models was assessed using Hosmer-Lemeshow test (function hoslem.test, R package
ResourceSelection). This stepwise multivariable analysis was applied to 321 patients for whom
complete biological data (white blood cells, neutrophil, lymphocyte, monocyte, creatinine, red
blood cells, haemoglobin, C-reactive protein and oxygen saturation) were available. Statistical
tests were 2-tailed with a level of statistical significance of < .05. Statistical analysis was per-
formed using R language version 3.5.2 (https://cran.r-project.org/).
PLOS ONE | https://doi.org/10.1371/journal.pone.0243709 January 27, 2021
3 / 12
PLOS ONE Characteristics of COVID-19 patients at hospital admission and factors of severity
Results
Patient characteristics on admission
From February 8 to April 24, 2020, a total of 412 SARS-CoV-2 confirmed patients with known
date of hospital discharge or death were included. Overall, 66 patients (16.0%) were admitted
directly to ICUs and 320 (77.7%) were hospitalized in medical wards, of whom 26/320 (8.1%)
required subsequent transfer to ICUs. Median age was 72.0 years [IQR, 57–83] and 56.3%
were men in the overall population. A total of 188 (45.6%) patients were younger than 70
years, 139 (33.7%) were aged between 70 and 84 years old and 82 (20.6%) were older than 85
years. One or more pre-existing comorbidities were present in 286 patients (69.4%): cardiovas-
cular diseases (47.6%), diabetes (19.9%) and chronic lung diseases (15.0%) being the most
common. The most frequently reported signs and symptoms on admission were cough
(73.5%), dyspnoea/tachypnoea (64.3%), general weakness (61.4%) and fever (>37.8˚C, 57.0%).
Abnormal lung auscultation was observed in 229 patients (55.6%). Demographic data, clinical
signs and symptoms on admission according to hospitalization ward are summarized in
Table 1. Males were significantly more prone to hospitalization in ICUs (P = .0003). The pro-
portion of patients with comorbidities was similarly distributed among medical wards and
ICUs except for cardiovascular diseases (P = .04). Diffuse or abdominal pain were reported sig-
nificantly more often in patients hospitalized in medical wards (P = .003 and P = .02 respec-
tively). Patients hospitalized in ICUs presented more often with fever (>37.8˚C, 66.3% and vs
54.4%; P = .04), shortness of breath (77.2% vs 44.1%; P < .001), showed more frequently
abnormal lung auscultation (71.7% and 76.9% vs 50.9%; P = .0005) and suffered from dys-
pnoea/tachypnoea (85.9% vs 58.1%; P < .001). Patients hospitalized in ICU were significantly
more likely to have oxygen saturation less than 90%, to receive specific treatment, and to be
under ventilation. Duration of symptoms were also significantly higher in ICU hospitalized
patients than those in medical ward (P < .001). The time between symptom onset and hospital
admission was significantly lower in patients hospitalized in medical wards (median, 6 [IQR,
2–9] vs 7 [IQR,4–10]; P = .004).
As of April 24, 2020, 87 (21.1%) patients had died. The crude case fatality rate differed
between patients hospitalized in medical wards (14.1%), and those hospitalized in ICU (45.7%,
P < .0001).
Initial biological data on admission are represented in Table 2. The majority of biological
parameters were in the normal range although their values differed between patients admitted
to ICUs and those hospitalized in medical wards. Lymphocytopenia and monocytopenia were
found in 70.7% and 37.0% of ICU hospitalized patients as compared to 51.9% and 14.6% in
medical ward patients respectively (P < .001). Elevated levels of aspartate aminotransferase
(AST> 37 U/L), alanine aminotransferase (ALT> 61 U/L), lactate dehydrogenase (LDH> 241
U/L), C-reactive protein (CRP> 100 mg/L) and urea (>6.6 mmol/L) on admission were also
noted in ICU patients.
Patient determinants associated with ICU hospitalization
The analysis was based on 321 patients with complete biological data. The crude case fatality
rate was not statistically different between included patients and those not included (20.0%
(75/321) vs 31.6%, (12/38); P = .26).
The time between the onset of symptoms and hospital admission was inversely associated
with age in our study with a shorter time for older patients (p < 0.05) (Supplementary mate-
rial). Due to this strong correlation, an independent effect of each of these variables could not
PLOS ONE | https://doi.org/10.1371/journal.pone.0243709 January 27, 2021
4 / 12
PLOS ONE Table 1. Demographic and clinical characteristics on admission of 412 confirmed COVID-19 hospitalized patients at Lyon University Hospitals, France.
Characteristics of COVID-19 patients at hospital admission and factors of severity
Age(years)
Age (�75 years)
<50
50–69
70–84
> = 85
Gender
Male
Comorbidities
aCardiovascular disease
Systolic arterial pressure (mmHg)
Diastolic arterial pressure (mmHg)
PAS> = 140 and/or PAD> = 90
Diabetes
Chronic lung disease
Renal diseases
Malignancy
Chronic neurological diseases
Liver diseases
Immunodeficiency
Signs and symptoms at admission
Temperature (˚C)
Fever (>37.8˚C)
Fever (>39.0˚C)
Historic of fever
Cough
General weakness
Shortness of breath
Diffuse pain
Diarrhoea
Myalgias
Headache
Nausea
Runny nose
Confusion
Abdominal pain
Anosmia
Ageusia
Sore throat
Chest pain
Joints pain
Dyspnoea/tachypnoea
Abnormal lung auscultation
Pharyngeal exudate
Oxygen saturation (%)
Oxygen saturation <90%
Treatment for Covid-19
Medical wards (n = 320)
73 (57–84)�
151 (47.2)
54 (16.9)��
87 (27.2)
102 (31.9)
77 (24.1)
165 (51.6)
161(50.3)
p-value
0.03
0.06
ICU (n = 92)
68.5 (55.8–78)
33 (35.9)
15 (16.3)
32 (34.8)
37 (40.2)
8 (8.7)
67 (72.8)
0.0003
35 (38)
136 (120–153.2) [292]
132 (115–147) [85]
78 (67–86) [292]
142/292 (48.6)
77 (62–84) [85]
36/85 (42.4)
64 (20)
48 (15)
39 (12.2)
42 (13.1)
46 (14.4)
21 (6.6)
20 (6.2)
18 (19.6)
14 (15.2)
17 (18.5)
13 (14.1)
7 (7.6)
6 (6.5)
6 (6.5)
0.04
0.15
0.15
0.33
1
1
0.12
0.86
0.11
1
1
38 (37–38.4)
38.3 (37.5–39)
0.0005
174 (54.4)
22 (6.9)
273 (85.3)
237 (74.1)
199 (62.2)
141 (44.1)
107 (33.4)
89 (27.8)
71 (22.2)
55 (17.2)
43 (13.4)
40 (12.5)
27 (8.4)
31 (9.7)
26/305 (8.5)
25/305 (8.2)
20 (6.2)
12 (3.8)
7 (2.2)
186 (58.1)
163 (50.9)
21 (6.6)
61 (66.3)
15 (16.3)
85 (92.4)
66 (71.7)
54 (58.7)
71 (77.2)
16 (17.4)
24 (26.1)
14 (15.2)
12 (13)
11 (12)
9 (9.8)
8 (8.7)
2 (2.2)
4/92 (4.3)
4/92 (4.3)
4 (4.3)
2 (2.2)
1 (1.1)
79 (85.9)
66 (71.7)
6 (6.5)
95 (92–97) [279]
26/279 (9.3)
119/304 (39.1)
88 (82.5–93) [71]
38/71 (53.5)
70/89 (78.7)
0.04
0.01
0.08
0.69
0.55
<0.001
0.003
0.79
0.19
0.42
0.86
0.58
1
0.02
0.26
0.26
0.62
0.74
0.69
<0.001
0.0005
1
<0.001
<0.001
<0.001
(Continued )
PLOS ONE | https://doi.org/10.1371/journal.pone.0243709 January 27, 2021
5 / 12
PLOS ONE Characteristics of COVID-19 patients at hospital admission and factors of severity
Table 1. (Continued)
Ventilation
Duration of symptoms
LOS (Alive)
LOS (Deceased
Delays (days) between
Onset of symptoms and hospital admission
< 3 days
3–10 days
> 10 days
Deceased
Medical wards (n = 320)
18 (5.6)
15 (12–18.8) [198]
8 (4–12) [276]
8.5 (6–13) [44]
6 (2–9)
84 (26.2)
43 (13.4)
193 (60.3)
45 (14.1)
ICU (n = 92)
67 (72.8)
22.5 (20.2–27) [34]
12.5 (9–17.8) [50]
11 (6–18.2) [42]
7 (4–10)
10 (10.9)
22 (23.9)
60 (65.2)
42 (45.7)
p-value
<0.001
<0.001
<0.001
0.11
0.004
<0.001
� Median (IQR) for continious variables
�� % in paranthese for categorial variables
COVID-19: coronavirus disease 2019; ICU: Intensive care unit; IQR: interquartile range; LOS: length of stay.
P < .05 was considered statistically significant
aCardiovascular disease included hypertension and heart failure
[n] indicates the patients without missing values for continuous variables
https://doi.org/10.1371/journal.pone.0243709.t001
be detected. Demographic, clinical and biological characteristics on admission associated with
admission in ICUs are therefore summarized in Table 3 for each of these variables.
The results of multivariate regression analysis including age (Table 3) showed that older
patients (� 85-year-old) were less admitted in ICUs (OR, 0.17 [95%CI, 0.05–0.51] as compared
to those < 70 years old. Temperature (OR, 1.56 [95%CI, 1.06–2.28] per degree Celsius
increase; P = .02) and abnormal lung auscultation on admission (OR, 3.58 [95%CI, 1.58–8.11];
P = .002) were associated with a higher risk of admission in ICUs. Patients with oxygen
saturation < 90% had higher risk of ICU admission (OR, 12.5 [95%CI, 5.27–29.4] compared
to those with oxygen saturation � 90%, P < 0.001). The odds of ICU admission revealed a sta-
tistically significant increasing trend with an elevated level of CRP (OR, 2.7 [95%CI, 1.29–5.66]
for CRP> 100mg/L vs CRP < 10mg/L; P = .008). Monocytopenia (monocytes < 0.3 G/L) was
associated with increased risk of ICU hospitalization (OR, 3.28 [95%CI, 1.4–7.68; P = 0.006]].
When age was replaced by time from onset to hospital admission in the multivariate regression
model (Table 3), CRP was not anymore significantly associated with the risk of ICU admission
(OR, 2.01 [95%CI, 0.95–4.23] for CRP> 100mg/L vs CRP < 10mg/L; P = .07).
Discussion and conclusions
This report of French hospitalized COVID-19 patients with full follow-up data completes epi-
demiological information already available from other European countries [14–16]. Overall,
16.6% of the patients were directly admitted to ICUs and 6.3% were transferred to ICUs from
medical wards. The study comprised 412 patients with 87 deaths and 325 patients discharged
alive.
The observed overall case fatality rate of 21.1% in this series is higher than those reported in
China [17, 18], but is similar to what has been observed in New York City [19]. The relatively
younger age of patients in the Chinese studies (median ages: 56 and 49 respectively) could lead
to less sever disease that explain the lower mortality rates reported in these studies.
Mortality rates of almost 46% in our patients hospitalized in ICU was higher than that of
26% reported in ICU-hospitalized patients in Lombardy [14]. However, at the time of
PLOS ONE | https://doi.org/10.1371/journal.pone.0243709 January 27, 2021
6 / 12
PLOS ONE Characteristics of COVID-19 patients at hospital admission and factors of severity
Table 2. Laboratory measures on admission of confirmed Covid-19 hospitalized patients at Lyon University Hospitals, France.
Reference range
Medical wards (n = 320)
ICU (n = 92)
p-value
Median (IQR)
Complete blood count
White blood cells (G/L)
<4
>10
Neutrophils (G/L)
>7.5
Lymphocytes (G/L)
<1
Monocytes (G/L)
<0.3
Platelets (G/L)
<150
Red blood cells (globules rouges)
Haemoglobin (g/L)
<120
NLR
> = 3.3
PLR
> = 180
[4–10]
[1.8–7.5]
[1–4]
[0.2–0.9]
[150–400]
[4.0–6.0]
[120–170]
Prothrombin time (%)
[70–150]
� 70
Inflammation
5.9 (4.5–7.9)� [308]
49/320 (15.9)��
259/320 (84.1)
4.1(2.8–5.9) [308]
36/320 (11.7)
1 (0.7–1.4) [308]
160/320 (51.9)
0.5 (0.4–0.7) [308]
45/320 (14.6)
195(155–257) [305]
70/320(23)
4.6 (4.1–5) [308]
7.1 (5.2–9.3) [92]
13/92 (14.1)
79/92 (85.9)
5.5 (3.8–8.2) [92]
27/92 (29.3)
0.7 (0.6–1.1) [92]
65/92 (70.7)
0.4 (0.2–0.5) [92]
34/92 (37)
196(152–268.8) [90]
21/92(23.3)
4.6 (4.1–5.1) [92]
134 (120–146) [308]
137.5 (118.8–147.2) [92]
74/320 (24)
4.2(2.4–7.4) [308]
186/320(60.4)
24/92 (26.1)
6.9(4.2–11.7) [92]
76/92(82.6)
197.1(135.2–289.3) [305]
253.6(184.7–392.5) [90]
173/320(56.7)
82(68–90) [239]
175/320(73.2)
68/92(75.6)
76(66.2–87) [74]
47/92(63.5)
CRP (C Reactive Protein) (mg/L)
<5
49.9(18.3–107) [305]
127.2(67.6–178.8) [73]
>100
>20.3
Biochemical
Creatinine (μmol/L)
>104
Urea (mmol/L)
>6.4
[45–104]
[2.5–9.2]
AST (Transaminase ASAT) (U/L)
[15–37]
>37
ALT (Transaminase ALAT) (U/L)
[13–61]
>61
LDH (U/L)
>241
Sodium (mmol/L)
Potassium (mmol/L)
85/320(27.9)
220/320(72.1)
81(63–101) [305]
69/320(22.6)
6.3(4.5–9) [304]
152/320(50)
39(28–59) [244]
132/320(54.1)
26(17–45.5) [255]
33/320(12.9)
47/92(64.4)
69/92(94.5)
83(71.5–114) [91]
26/92(28.6)
7(4.8–9.7) [91]
57/92(62.6)
62(45–80) [79]
71/92(89.9)
35(23–62.5) [79]
21/92(26.6)
[87–241]
[136–145]
[3.5–5.1]
313(231.5–402) [111]
408(358.5–496) [23]
78/320(70.3)
137(134–139) [305]
4.1(3.8–4.4) [301]
21/92(91.3)
136(134–138) [91]
4(3.6–4.4) [90]
� Median (IQR) for continious variables
�� % in paranthese for categorial variables
COVID-19: coronavirus disease 2019; ICU: intensive care unit; IQR: interquartile range; LOS: length of stay; AST: aspartate aminotransferase, ALT: alanine
aminotransferase; LDH: lactate dehydrogenase; CRP: C-reactive protein; NLR: Neutrophils to Lymphocytes Ratio, PLR: Platelets to Lymphocytes Ratio
P < 0.05 was considered statistically significant
[n] indicates the patients without missing values for continuous variable
https://doi.org/10.1371/journal.pone.0243709.t002
PLOS ONE | https://doi.org/10.1371/journal.pone.0243709 January 27, 2021
7 / 12
0.002
0.75
0.75
<0.001
0.0001
0.0001
0.002
<0.001
<0.001
0.98
1
0.47
0.57
0.68
<0.001
<0.001
<0.001
0.001
0.12
0.11
<0.001
<0.001
<0.001
0.06
0.26
0.1
0.04
<0.001
<0.001
<0.001
0.008
0.001
0.04
0.41
0.3
PLOS ONE Characteristics of COVID-19 patients at hospital admission and factors of severity
Table 3. Determinants associated with admission to intensive care units in 321 Covid-19 patients at Lyon Univer-
sity hospitals, France.
A: With age in the model
Sex
Female
(17/147)
Male
(39/174)
Age (years)
< 70
70–85
(27/144)
(21/100)
85–103
aTemperature Continuous variable
(8/77)
Oxygen saturation (%)
90–100
(26/265)
<90
(30/56)
Abnormal lung auscultation
No
Yes
(13/141)
(43/180)
C Reactive Protein, mg/L (Normal range <5 mg/L)
�100
(21/217)
(35/104)
>100
Monocyte count, x109/L (Normal range 0.2–0.9 x109/L)
� 0.3
< 0.3
(36/263)
(20/58)
B: With delay between onset and hospital admission in the model
Sex
Female
(17/147)
Male
(39/174)
Delay between onset and hospital admission (day)
0–2
3–10
(5/73)
(38/200)
(13/48)
>10
aTemperature Continuous variable
Oxygen saturation (%)
90–100
(26/265)
<90
(30/56)
Abnormal lung auscultation
No
Yes
(13/141)
(43/180)
C Reactive Protein, mg/L (Normal range <5 mg/L)
�100
(21/217)
(35/104)
>100
Monocyte count, x109/L (Normal range 0.2–0.9 x109/L)
� 0.3
< 0.3
(36/263)
(20/58)
Adjusted OR
95%CI
P-value
1
1.66
1
0.75
0.17
1.56
1
12.45
1
3.58
1
2.7
1
3.28
0.76–3.62
0.2
0.33–1.7
0.05–0.51
1.06–2.28
0.49
0.002
0.02
5.27–29.4
<0.0001
1.58–8.11
0.002
1.29–5.66
0.008
1.40–7.68
0.006
Adjusted OR
95%CI
P-value
1
1.88
1
5.96
6.90
1.62
1
11.48
1
3.50
1
2.01
1
2.86
0.87–4.10
0.11
1.65–21.5
1.68–28.4
1.10–2.39
0.006
0.007
0.01
4.86–27.1
<0.0001
1.56–7.85
0.002
0.95–4.23
0.07
1.26–6.50
0.02
OR: odds ratio, CI: confdence interval
In multivariable logistic regression: i) ICU admission was the outcome and ii) age or delay between onset and
hospital admission sex, continuous temperature, oxygen saturation, abnormal lung auscultation, monocyte count
and C-reactive protein were additive covariates, Akaike information criterion = 216.0, test of Hosmer and Lemeshow
goodness of fit with 10 bins: P = 0.59
aodds ratio of ICU admission was multiplied by 1.62 per degree celsius increase
https://doi.org/10.1371/journal.pone.0243709.t003
PLOS ONE | https://doi.org/10.1371/journal.pone.0243709 January 27, 2021
8 / 12
PLOS ONE Characteristics of COVID-19 patients at hospital admission and factors of severity
reporting, 58% of patients included in the latter study were still hospitalized. In addition, the
attributable mortality related to Covid-19 could not be assessed in our patients who had several
underlying comorbidities that could contribute to the observed mortality rate.
The most commonly known manifestations of the disease i.e. cough, weakness and fever on
admission in our patients were in general similar to those reported in other studies [17, 20,
21]. As reported earlier, cardiovascular diseases and diabetes were the most common comor-
bidities [18, 22].
In agreement with the results of a recent single-arm meta-analysis [23], men accounted
for a higher proportion of COVID-19 patients than women in the present study. Similar
findings have been reported for MERS-CoV [24]. Women and men traditionally differ in
their perceptions of risk [25]. In women, better adoption of protective behaviours such as
hand-washing [26], in particular in the context of a pandemic [27], could at least in part
explain the observed results. Differences in underlying comorbidities, in particular lower
cardiovascular diseases in women could also explain higher rate of ICU admission in men.
However, the rates of patients with cardiovascular diseases were not different between the
two genders in our study.
Consistent with respiratory viral infections, our hospitalized patients had lymphopenia and
elevated levels of LDH and CRP. These laboratory abnormalities were found more often in
patients hospitalized in ICU. Lymphopenia and increased levels of LDH and CRP were also
reported in the meta-analysis of 1994 COVID-19 patients [21].
Our results suggest that temperature, abnormal lung auscultation on admission, high
levels of CRP, and monocytopenia could increase the risk of ICU hospitalization. A high
level of CRP has been reported to be an independent risk factor to assess the severity of
COVID-19 [28].
We found that older age was associated with lower risk of ICU admission. By comparison
with the known severity of influenza in elderly and due to the lack of sufficient knowledge
about the consequences of COVID-19 on elderly during the first wave of the pandemic in
France, this population could present to the hospital as soon as the onset of symptoms. Rapid
management and start of appropriate treatment thanks to earlier presentation at hospital
could explain the lower rates of ICU hospitalization observed in older patients. The risk of
reduced survival after ICU-related invasive treatment in this fragile population could also
explain, at least in part, the observed inverse relationship between age and ICU admission
[20]. Finally, in the context of a pandemic, shortage of intensive care resources could impact
the decisions about the most appropriate treatment.
The time between the onset of symptoms and hospital admission was strongly associated
with ICU admission and could be influenced by multiple determinants such as socio-eco-
nomic status, personal risk perception, and access to care. This delay could be considered as a
warning marker to alert medical providers on the possibility of critical illness when patients
present later in their disease course.
The prospective design, inclusion of both severe and non-severe cases and complete follow-
up of the study population are the main strengths of the present study. Multivariable analysis
was based on 374 patients with complete biological data. However, selection bias remains low
since the case fatality rate did not differ between patients who were included and not included
in the model. Only biological measures on admission were analysed because repeated mea-
surements were most likely only performed in more severe cases.
In conclusion, age and delay between onset of symptoms and hospital admission were asso-
ciated with the risk of hospitalisation in ICU. Age being a fixed variable, interventions which
shortened this delay would improve the prognosis of Covid-19 patients.
PLOS ONE | https://doi.org/10.1371/journal.pone.0243709 January 27, 2021
9 / 12
PLOS ONE Characteristics of COVID-19 patients at hospital admission and factors of severity
Supporting information
S1 File.
(XLSX)
S2 File. Distribution of the time between the onset of symptoms and hospital admission by
age category.
(DOCX)
Acknowledgments
The authors express their gratitude to:
COVID-Outcomes-HCL Consortium (affiliation for all: Hospices Civils de Lyon): Laurent
Argaud, Fre´de´ric Aubrun, Marc Bonnefoy, Maude Bouscambert-Duchamp, Roland Chapurlat,
Dominique Chassard, Christian Chidiac, Michel Chuzeville, Cyrille Confavreux, Se´bastien
Couraud, Gilles Devouassoux, Isabelle Durieu, Michel Fessy, Sylvain Gaujard, Alexandre Gay-
mard, Arnaud Hot, Bruno Lina, Ge´raldine Martin Gaujard, Emmanuel Morelon, Vincent Pir-
iou, Ve´ronique Potinet, Jean-Christophe Richard, Thomas Rimmele, Pascal Sève, Alain Sigal,
Karim Tazarourte.
2) the Department of Health Data of Lyon Hospital: A. Duclos, F. Gueyffier, S. Vautier and
M. Herve´ for the creation and management of e-CRF,
3) Clinical research associates for data collection and data entry: V. Artizzu, L. Bissuel, S.
Bennina, L. Dehina-Khenniche, A. Darrin, M. Grange, B. Robin,
4) staff of the virology laboratory of the Lyon hospital: Claire Bandolo, Genevieve Billaud,
Maude Bouscambert-Duchamp, Emilie Frobert, Alexandre Gaymard, Laurence Josset, Chris-
tophe Ramiere, Isabelle Schuffenecker, Solange Telusson, Martine Valette, Florence Morfin
for providing the results of RT-PCR tests.
Author Contributions
Conceptualization: Ce´dric Dananche´.
Formal analysis: Marie-Paule Gustin.
Funding acquisition: Philippe Vanhems.
Investigation: Selilah Amour, Elisabetta Kuczewski.
Methodology: Philippe Vanhems, Christelle Elias, Laetitia Henaff, Ce´dric Dananche´, Mitra
Saadatian-Elahi.
Project administration: Philippe Vanhems, Laetitia Henaff, Mitra Saadatian-Elahi.
Supervision: Philippe Vanhems, Mitra Saadatian-Elahi.
Validation: Philippe Vanhems, Marie-Paule Gustin, Christelle Elias, Laetitia Henaff, Mitra
Saadatian-Elahi.
Writing – original draft: Philippe Vanhems, Marie-Paule Gustin, Christelle Elias, Laetitia
Henaff, Mitra Saadatian-Elahi.
Writing – review & editing: Philippe Vanhems, Marie-Paule Gustin, Christelle Elias, Laetitia
Henaff, Ce´dric Dananche´, Be´atrice Grisi, Elodie Marion, Nagham Khanafer, Delphine Hil-
liquin, Sophie Gardes, Solweig Gerbier-Colomban, Selilah Amour, Elisabetta Kuczewski,
Vanessa Escuret, Bruno Lina, Mitra Saadatian-Elahi.
PLOS ONE | https://doi.org/10.1371/journal.pone.0243709 January 27, 2021
10 / 12
PLOS ONE Characteristics of COVID-19 patients at hospital admission and factors of severity
References
1. Hui DS, I Azhar E, Madani TA, Ntoumi F, Kock R, Dar O, et al. The continuing 2019-nCoV epidemic
threat of novel coronaviruses to global health—The latest 2019 novel coronavirus outbreak in Wuhan,
China. Int J Infect Dis. 2020; 91:264-266. https://doi.org/10.1016/j.ijid.2020.01.009 PMID: 31953166
2.
3.
4.
Zhu N, Zhang D, Wang W, Li X, Yang B, Song J, et al. A Novel Coronavirus from Patients with Pneumo-
nia in China, 2019. N Engl J Med. 2020; 382(8):727-733. https://doi.org/10.1056/NEJMoa2001017
PMID: 31978945
Lu H, Stratton CW, Tang YW. Outbreak of pneumonia of unknown etiology in Wuhan, China: The mys-
tery and the miracle. J Med Virol. 2020; 92(4):401-402. https://doi.org/10.1002/jmv.25678 PMID:
31950516
Lupia T, Scabini S, Mornese Pinna S, Di Perri G, De Rosa FG, Corcione S. 2019 novel coronavirus
(2019-nCoV) outbreak: A new challenge [published online ahead of print, 2020 Mar 7]. J Glob Antimi-
crob Resist. 2020; 21:22-27. https://doi.org/10.1016/j.jgar.2020.02.021 PMID: 32156648
5. Hu Z, Song C, Xu C, Jin G, Chen Y, Xu X, et al. Clinical characteristics of 24 asymptomatic infections
with COVID-19 screened among close contacts in Nanjing, China. Sci China Life Sci. 2020; 63(5):706-
711. https://doi.org/10.1007/s11427-020-1661-4 PMID: 32146694
6.
John P.A. Ioannidis. The infection fatality rate of COVID-19 inferred from seroprevalence data. medRxiv
preprint https://doi.org/10.1101/2020.05.13.20101253. https://www.medrxiv.org/content/10.1101/2020.
05.13.20101253v3.full.pdf (Accessed on August 3, 2020).
7. Perlman S. Another Decade, Another Coronavirus. N Engl J Med. 2020; 382(8):760-762. https://doi.
org/10.1056/NEJMe2001126 PMID: 31978944
8. Bernard Stoecklin S, Rolland P, Silue Y, Mailles A, Campese C, Simondon A, et al. First cases of coro-
navirus disease 2019 (COVID-19) in France: surveillance, investigations and control measures, Janu-
ary 2020. Euro Surveill. 2020; 25(6):2000094. https://doi.org/10.2807/1560-7917.ES.2020.25.6.
2000094 PMID: 32070465
9. Sante´ Publique France. https://www.santepubliquefrance.fr/dossiers/coronavirus-covid-19/corona
virus-chiffres-cles-et-evolution-de-la-covid-19-en-france-et-dans-le-monde#block-266151 (Accessed
on Septembre 24 2020).
10. Simonnet A, Chetboun M, Poissy J, Raverdy V, Noulette J, Duhamel A, et al. High prevalence of obesity
in severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2) requiring invasive mechanical venti-
lation [published online ahead of print, 2020 Apr 9]. Obesity (Silver Spring). 2020;10.1002/oby.22831.
https://doi.org/10.1002/oby.22831 PMID: 32271993
11. Saadatian-Elahi M, Picot-Sanchez V, He´ naff L, Pradel F, Escuret V, Dananche´ C, et al. Protocol for a
prospective, observational, hospital-based multicentre study of nosocomial SARS-CoV-2 transmission:
NOSO-COR Project. BMJ Open. 2020; 10(10):e039088. https://doi.org/10.1136/bmjopen-2020-
039088 PMID: 33093033
12. World Health Organisation. Coronavirus disease 2019 (COVID-19) Situation Report– 70. Page 9.
https://www.who.int/docs/default-source/coronaviruse/situation-reports/20200330-sitrep-70-covid-19.
pdf?sfvrsn=7e0fe3f8_2.
13. Corman VM, Landt O, Kaiser M, Molenkamp R, Meijer A, Chu DK, et al. Detection of 2019 novel corona-
virus (2019-nCoV) by real-time RT-PCR. Euro Surveill. 2020; 25(3):2000045. https://doi.org/10.2807/
1560-7917.ES.2020.25.3.2000045 PMID: 31992387
14. Grasselli G, Zangrillo A, Zanella A, Antonelli M, Cabrini L, Castelli A, et al. Baseline Characteristics and
Outcomes of 1591 Patients Infected With SARS-CoV-2 Admitted to ICUs of the Lombardy Region, Italy
[published online ahead of print, 2020 Apr 6]. JAMA. 2020; 323(16):1574-1581. https://doi.org/10.1001/
jama.2020.5394 PMID: 32250385
15. Spiteri G, Fielding J, Diercke M, Campese C, Enouf V, Gaymard A, et al. First cases of coronavirus dis-
ease 2019 (COVID-19) in the WHO European Region, 24 January to 21 February 2020. Euro Surveill.
2020; 25(9):2000178. https://doi.org/10.2807/1560-7917.ES.2020.25.9.2000178 PMID: 32156327
16. Hodcroft EB. Preliminary case report on the SARS-CoV-2 cluster in the UK, France, and Spain. Swiss
Med Wkly. 2020; 150(9–10):10.4414/smw.2020.20212. Published 2020 Feb 27. https://doi.org/10.
4414/smw.2020.20212 PMID: 32227799
17. Wang D, Hu B, Hu C, Zhu F, Liu X, Zhang J, et al. Clinical Characteristics of 138 Hospitalized Patients
With 2019 Novel Coronavirus-Infected Pneumonia in Wuhan, China [published online ahead of print,
2020 Feb 7]. JAMA. 2020; 323(11):1061-1069. https://doi.org/10.1001/jama.2020.1585 PMID:
32031570
18. Huang C, Wang Y, Li X, Ren L, Zhao J, Hu Y, et al. Clinical features of patients infected with 2019 novel
coronavirus in Wuhan, China [published correction appears in Lancet. 2020 Jan 30]. Lancet. 2020; 395
(10223):497–506. https://doi.org/10.1016/S0140-6736(20)30183-5 PMID: 31986264
PLOS ONE | https://doi.org/10.1371/journal.pone.0243709 January 27, 2021
11 / 12
PLOS ONE Characteristics of COVID-19 patients at hospital admission and factors of severity
19. Richardson S, Hirsch JS, Narasimhan M, Crawford JM, McGinn T, Davidson KW, et al. Presenting
Characteristics, Comorbidities, and Outcomes Among 5700 Patients Hospitalized With COVID-19 in
the New York City Area [published online ahead of print, 2020 Apr 22]. JAMA. 2020; e206775. https://
doi.org/10.1001/jama.2020.6775 PMID: 32320003
20. Haas LEM, de Lange DW, van Dijk D, van Delden JJM. Should we deny ICU admission to the elderly?
Ethical considerations in times of COVID-19. Crit Care. 2020; 24(1):321. https://doi.org/10.1186/
s13054-020-03050-x PMID: 32517776
21. Borges do Nascimento IJ, Cacic N, Abdulazeem HM, von Groote TC, Jayarajah U, Weerasekara I,
et al. Novel Coronavirus Infection (COVID-19) in Humans: A Scoping Review and Meta-Analysis. J Clin
Med. 2020; 9(4): E941. Published 2020 Mar 30. https://doi.org/10.3390/jcm9040941 PMID: 32235486
22. Guan WJ, Liang WH, Zhao Y, Liang HR, Chen ZS, Li YM, et al. Comorbidity and its impact on 1590
patients with Covid-19 in China: A Nationwide Analysis [published online ahead of print, 2020 Mar 26].
Eur Respir J. 2020;2000547. https://doi.org/10.1183/13993003.00547-2020 PMID: 32217650
23.
Li LQ, Huang T, Wang YQ, Wang ZP, Liang Y, Huang TB, et al. COVID-19 patients’ clinical characteris-
tics, discharge rate, and fatality rate of meta-analysis [published online ahead of print, 2020 Mar 12]. J
Med Virol. 2020;10.1002/jmv.25757. https://doi.org/10.1002/jmv.25757 PMID: 32162702
24. Badawi A, Ryoo SG. Prevalence of comorbidities in the Middle East respiratory syndrome coronavirus
(MERS-CoV): a systematic review and meta-analysis. Int J Infect Dis. 2016; 49:129-133. https://doi.
org/10.1016/j.ijid.2016.06.015 PMID: 27352628
25. Gustafson PE. Gender differences in risk perception: theoretical and methodological perspectives. Risk
Anal. 1998; 18(6):805–811. https://doi.org/10.1023/b:rian.0000005926.03250.c0 PMID: 9972583
26. Srivastav A, Santibanez TA, Lu PJ, Stringer MC, Dever JA, Bostwick M, et al. Preventive behaviors
adults report using to avoid catching or spreading influenza, United States, 2015–16 influenza season.
PLoS One. 2018; 13(3): e0195085. Published 2018 Mar 30. https://doi.org/10.1371/journal.pone.
0195085 PMID: 29601610
27.
28.
Fung IC, Cairncross S. How often do you wash your hands? A review of studies of hand-washing prac-
tices in the community during and after the SARS outbreak in 2003. Int J Environ Health Res. 2007; 17
(3):161-183. https://doi.org/10.1080/09603120701254276 PMID: 17479381
Zhu Z, Cai T, Fan L, Lou K, Hua X, Huang Z, et al. Clinical value of immune-inflammatory parameters to
assess the severity of coronavirus disease 2019 [published online ahead of print, 2020 Apr 22]. Int J
Infect Dis. 2020; S1201-9712(20)30257-5. https://doi.org/10.1016/j.ijid.2020.04.041 PMID: 32334118
PLOS ONE | https://doi.org/10.1371/journal.pone.0243709 January 27, 2021
12 / 12
PLOS ONE |
10.1371_journal.pone.0257328 | RESEARCH ARTICLE
Validity of caregivers’ reports on prior use of
antibacterials in children under five years
presenting to health facilities in Gulu,
northern Uganda
Hindum LanyeroID
Katureebe Agaba4, Joan N. Kalyango5,6, Jaran Eriksen3,7, Sarah Nanzigu1*
1, Moses Ocan1, Celestino Obua2, Cecilia Stålsby Lundborg3,
1 Department of Pharmacology and Therapeutics, Makerere University College of Health Sciences,
Kampala, Uganda, 2 Mbarara University of Science and Technology, Mbarara, Uganda, 3 Department of
Global Public Health, Karolinska Institutet, Stockholm, Sweden, 4 Infectious Diseases Research
Collaboration, Kampala, Uganda, 5 Department of Pharmacy, Makerere University College of Health
Sciences, Kampala, Uganda, 6 Clinical Epidemiology Unit, Makerere University College of Health Sciences,
Kampala, Uganda, 7 Department of Infectious Diseases, South General Hospital, Stockholm, Sweden
* snanzigu@yahoo.com
Abstract
Introduction
Given the frequent initiation of antibacterial treatment at home by caregivers of children
under five years in low-income countries, there is a need to find out whether caregivers’
reports of prior antibacterial intake by their children before being brought to the healthcare
facility are accurate. The aim of this study was to describe and validate caregivers’ reported
use of antibacterials by their children prior to seeking care at the healthcare facility.
Methods
A cross sectional study was conducted among children under five years seeking care at
healthcare facilities in Gulu district, northern Uganda. Using a researcher administered
questionnaire, data were obtained from caregivers regarding reported prior antibacterial
intake in their children. These reports were validated by comparing them to common anti-
bacterial agents detected in blood and urine samples from the children using liquid chroma-
tography with tandem mass spectrometry (LC-MS/MS) methods.
Results
A total of 355 study participants had a complete set of data on prior antibacterial use col-
lected using both self-report and LC-MS/MS. Of the caregivers, 14.4% (51/355, CI: 10.9–
18.5%) reported giving children antibacterials prior to visiting the healthcare facility. How-
ever, LC-MS/MS detected antibacterials in blood and urine samples in 63.7% (226/355, CI:
58.4–68.7%) of the children. The most common antibacterials detected from the laboratory
analysis were cotrimoxazole (29%, 103/355), ciprofloxacin (13%, 46/355), and metronida-
zole (9.9%, 35/355). The sensitivity, specificity, positive predictive value (PPV), negative
predictive value and agreement of self-reported antibacterial intake prior to healthcare
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: Lanyero H, Ocan M, Obua C, Stålsby
Lundborg C, Agaba K, Kalyango JN, et al. (2021)
Validity of caregivers’ reports on prior use of
antibacterials in children under five years
presenting to health facilities in Gulu, northern
Uganda. PLoS ONE 16(9): e0257328. https://doi.
org/10.1371/journal.pone.0257328
Editor: Orvalho Augusto, University of Washington,
UNITED STATES
Received: February 12, 2021
Accepted: August 28, 2021
Published: September 16, 2021
Copyright: © 2021 Lanyero et al. This is an open
access article distributed under the terms of the
Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: LANYERO, HINDUM
(2021), Validity of caregivers’ reports on prior use
of antibacterials in children under five years
presenting to health facilities in Gulu, northern
Uganda, Dryad, Dataset, https://doi.org/10.5061/
dryad.sj3tx9642.
Funding: Makerere University -SIDA collaboration
(Sida PI0010) The funders had no role in study
design, data collection and analysis, decision to
publish, or preparation of the manuscript.
PLOS ONE | https://doi.org/10.1371/journal.pone.0257328 September 16, 2021
1 / 14
PLOS ONE Validity of caregivers’ reports on prior use of antibacterials in children under five years
Competing interests: The authors have declared
that no competing interests exist.
facility visit were 17.3% (12.6–22.8), 90.7% (84.3–95.1), 76.5% (62.5–87.2), 38.5% (33.0–
44.2) and 43.9% (k 0.06) respectively.
Conclusion
There is low validity of caregivers’ reports on prior intake of antibacterials by these children.
There is need for further research to understand the factors associated with under reporting
of prior antibacterial use.
Introduction
Antibacterial agents are used to treat a wide range of bacterial infections and are essential life-
saving medicines. They are the most commonly used medicines in Sub-Saharan Africa due to
the high prevalence of infectious diseases [1]. Used correctly, they deliver enormous benefits
to the health of the population worldwide [2].
Antibacterials are, according to the national drug policy of Uganda, prescription only medi-
cines [3]. However, they are readily accessible and affordable to most patients within the com-
munities in Uganda, not only as prescription medicines as they can often also be obtained
over-the-counter especially in private medicine outlets [4]. The relative ease with which com-
munities access these medicines poses several challenges for antibacterial stewardship [4]. The
majority of caregivers in low-income countries initiate treatment of their children at home [5].
The use of antibacterials prior to hospital visits is common, especially in low-income countries,
and may influence patient treatment outcomes. According to a study in Nigeria, 85% of
patients reported to have self-medicated before coming to the health facility and antibacterials
were among the most common medicines used [6]. A study in Uganda reported that 62.2% of
patients had used antibacterial agents prior to coming to health facility [4]. Another study
done in Haiti to assess self-medication among patients presenting at an out-patient depart-
ment found that 45.5% practiced self-medication with antibacterials [7].
Caregivers’ ability to report antibacterial intake prior to coming to a health facility is crucial
for appropriate prescription of medicines at the health facility. Self-reports have been shown to
have low validity as they are prone to recall bias and social desirability bias. Respondents nor-
mally provide information that conforms to their perceived expectations of the health workers
or researchers [5, 8]. A study carried out in Uganda in 2009 reported a limited validity of care-
givers’ reports of use of sulfamethoxazole, chloroquine and sulfadoxine in their children prior
to arrival to the hospital [5]. Similarly, a study from Tanzania reported that 97% of the children
without history of prior chloroquine treatment had detectable levels of chloroquine in blood
[9]. Another study in Ghana reported a high prevalence (64%) of antibacterials detected in
urine samples of patients compared to the self-reported use (13%) [10].
To our knowledge no study has validated caregivers’ reports of intake of antibacterials in
children under five years in rural communities in low resource settings. In this study we
describe and validate caregivers’ reported use of antibacterials by their children under five
years for treatment prior to seeking care at the healthcare facility.
Materials and methods
Ethics statement
The protocol was reviewed and approved by the Makerere University School of Biomedical
Sciences Research and Ethics Committee (reference SBS-570) and the Uganda National
PLOS ONE | https://doi.org/10.1371/journal.pone.0257328 September 16, 2021
2 / 14
PLOS ONE Validity of caregivers’ reports on prior use of antibacterials in children under five years
Council of Science and Technology (reference HS235ES) (S1 Appendix). Administrative clear-
ance was obtained from the healthcare facilities where the study was conducted. Written
informed consent was obtained from caregivers of children under five years prior to data col-
lection (S2 Appendix).
Study design and setting
A cross-sectional study was conducted among children under five years and their caregivers in
healthcare facilities in Gulu district, northern Uganda. Gulu is located about 360 km from the
capital city Kampala. In Uganda, the lowest level of the district-based healthcare system con-
sists of the village health teams/community medicine distributors, which constitute level 1 of
health care. This is operated by members of the community who can read and write at least in
the local language of the community. The next level is health centre II which is operated by a
professionally trained nurse with a diploma and is intended to serve 5000 patients. This is fol-
lowed by health centre level III which is operated by a professionally trained clinical officer
with a diploma in clinical medicine and intended to serve 10,000 patients. Above health centre
level III is health centre level IV and then district hospitals headed by medical officers with a
basic degree in medicine and surgery and intended to serve about 100,000 patients. Regionally
there are regional referral hospitals where patients are referred to from the district hospitals.
The regional referral hospitals are expected to have specialist health professionals covering the
major disciplines such as surgery, internal medicine, and paediatrics. At the top of the health
care system are the national referral hospitals [11]. Gulu district has a total of 19 health centre
level II, 10 health centre level III, one health centre level IV, 31 registered pharmacies and 135
licensed drug shops [12–14]. This study was carried out in three health centre level III and one
health centre level IV. These healthcare centers were purposively selected because they serve
the greatest number of patients in the out-patient departments in Gulu district. The most com-
mon diseases in children under five years seeking care at healthcare facilities in this area
include; malaria, diarrhea, pneumonia, acute childhood malnutrition and HIV/AIDS [15–17].
Study population
Sick children under five years and their caregivers seeking care at the four healthcare facilities
were included in the study after caregivers’ consent. Children who were brought to the health
center by caregivers who did not take care of the children from the onset of the current illness
were excluded from the study. Children who had come for review or continuation of treatment
for current illness were also excluded from the study.
Sample size
The sample size was computed based on formula for estimation of sample size for a single pro-
portion [18]. Assuming that the proportion of children getting antibacterial treatment prior to
health facility visit was 50%, in order to have a 95% confidence interval and a 5% margin of
error, the minimum sample size needed was set to 385. The number of children sampled from
each facility was determined from the volume of patients at the health facility using propor-
tionate sampling.
Sampling procedures
The patients were selected by systematic random sampling. On each of the data collection
dates the first patient to be recruited into the study was randomly selected by having a blind-
folded data collector walk around in the waiting area and point at a random patient among the
PLOS ONE | https://doi.org/10.1371/journal.pone.0257328 September 16, 2021
3 / 14
PLOS ONE Validity of caregivers’ reports on prior use of antibacterials in children under five years
patients waiting in line to be seen by the healthcare worker in the outpatient department.
Thereafter, every fourth patient in line towards the entrance of the healthcare workers room
was selected for recruitment. In the event that the selected patient was above five years of age,
they were skipped and the next patient recruited while maintaining the sampling interval.
Approximately 10 days were spent collecting data in each healthcare facility.
Data collection
An interviewer administered questionnaire was used for data collection. The questionnaire
was pre-tested on caregivers of 30 children in outpatient departments of Gulu regional referral
hospital. This tool was adapted from a tool used to collect data on prevalence and predictors of
prior antibacterial use among patients presenting to hospitals in northern Uganda in a previ-
ous study [4], it was written in English and translated to Acholi (the most common local lan-
guage spoken in the study area).
The data collection team was divided into four groups each comprising of two people, one
pharmacy technician (health professional with diploma in pharmacy) and a laboratory techni-
cian. The pharmacy technician conducted interviews while the laboratory technician collected
the blood and urine samples.
Information on the following variables was collected; sub-county of residence, age of child,
age of care-giver, sex of child, sex of caregiver, whether medication was given to child before
coming to the healthcare facility since the onset of this current illness, the type and source of
the medicine, and the person who recommended the medicine. In case the caregiver did not
know the name of the medicine, the interviewer asked them to describe it or show the packing
material if at all they had come with it to the health center. Each interview lasted about 20 min-
utes per patient.
Sample collection and transportation. Two hundred microlitres (200μL) of blood was
collected from the fingertips of children under five years using a 200μL micro-pipette with
ethylenediamine tetra-acetic acid (EDTA), and spotted on a filter paper and left to dry for 3
hours in room temperature. After the blood had dried on the filter paper, each filter paper was
put in a separate plastic zip bag with a desiccant and transported to the laboratory for analysis.
Urine samples were collected in sterile wide mouth containers. In the very young children
who couldn’t void in the wide mouth containers, urine samples were collected by placing a
thick layer of cotton wool inside the child’s nappy and squeezing the urine in the urine sample
bottles. Two hundred microlitres (200μL) of urine was collected from the wide mouth contain-
ers using a plastic pipette and spotted on a filter paper and left to dry for 3 hours at room tem-
perature. After the urine had dried on the filter paper, each filter paper was put in a separate
plastic zip bag with a desiccant and transported to the laboratory for analysis. The dried blood
spot (DBS) and dried urine spot (DUS) samples obtained from patients were stored at -20˚C
and -80˚C respectively until analysis.
Extraction and analysis of antibacterials in dry blood spot and dry urine spot sam-
ples. The whole diameter disk (containing 200μl of blood or urine) was cut out from each
DBS and DUS. The cut disc was placed in an Eppendorf tube (1.5 mL capacity) and mixed
with 1000 μL of methanol (20%) and acetonitrile (80%). The sample was vortex-mixed twice
for 20 s at 10-min intervals and then centrifuged at 3500 revolutions per minute (RPM) for 5
minutes. After the extraction period, the filter paper was removed, and 500 μL of the extract
was transferred into an auto-sampler vial to be injected onto the LC-MS/MS system for
analysis.
A simple, fast, sensitive and selective qualitative LC-MS/MS method for identification of
fifteen (15) antibacterials in DBS and DUS was used for analysis (S3 Appendix). The limit
PLOS ONE | https://doi.org/10.1371/journal.pone.0257328 September 16, 2021
4 / 14
PLOS ONE Validity of caregivers’ reports on prior use of antibacterials in children under five years
of detection for the different antibacterials were: amoxicillin (1.34 ng/mL), ampicillin
(0.001 ng/mL), penicillin G (0.005 ng/mL), penicillin V (0.03 ng/mL), cloxacillin (0.2 ng/mL),
cephalexin (0.22 ng/mL), sulfamethoxazole (0.95 ng/mL), trimethoprim (0.52 ng/mL), eryth-
romycin (1.1 ng/mL), ciprofloxacin (0.1 ng/mL), tetracycline (0.14 ng/mL), clarithromycin
(1.4 ng/mL), metronidazole (0.0004 ng/mL), chloramphenicol (0.0001ng/mL) and azithromy-
cin (0.22 ng/mL).
Data on key pharmacokinetics properties that may have affected the interpretation of our
results, have been presented in the supporting information section (S1 Table), and these
include: clearance, terminal half-life, percentage of medicine excreted in urine, time to peak
plasma concentrations and volume of distribution.
Data management
Double data entry was done using Epi-Data 3.1 software for both the questionnaire and labora-
tory data. The two datasets were reconciled by comparing them for each field in the question-
naire and laboratory result, in case of any discrepancies, the corresponding questionnaire or
patient laboratory record was checked to establish the correct entry. Data were then imported
into Stata 14/IC (Stata Inc., Texas USA) for analysis.
Statistical analysis
Descriptive statistics were presented using median and interquartile range (IQR) for continu-
ous variables or frequencies and proportions for categorical variables. The dependent vari-
ables, treatment of child with antibacterials prior to healthcare facility visit as reported by their
caregiver and detectable antibacterials in DBS or DUS samples, were summarized as propor-
tions. In order to adjust for potential biases associated with point estimates from the sampling
design, we used svy commands in stata to compute proportions and respective 95% confidence
intervals. Pearson’s chi-square test was used to assess associations for the categorical variables.
In order to validate caregivers’ reported use of antibacterials, sensitivity, specificity, positive
predictive value (PPV), negative predictive value (NPV), prevalence, agreement and kappa
coefficient were calculated. Laboratory results for detection of antibacterials in dry blood spot
or dry urine spot samples were considered as the gold standard and caregivers’ reports of use
of antibacterials prior to health facility visit were considered as the test results.
Results
Socio-demographic characteristics of the caregivers and children under five
years
Of the 385 sampled children, 355 (92.2%) had data on both caregiver’s report on antibacterial
use prior to health facility visit and results from urine and blood analysis and were thus
included in the analysis. The 30 (7.8%) observations were dropped because they were missing
blood analysis data. Over half (53.2%, n = 189) of the children were female. The median age of
the children was 29 (IQR: 16–46) months. The majority (96.1%, n = 341) of the caregivers
were female. The median age of the caregivers was 25 (IQR: 21–31) years. About half (53.2%,
n = 189) of the children attended a healthcare facility located in a rural area. (Table 1).
Prevalence of antibacterial use prior to coming to the health facility as
reported by caregivers of children under five years
Out of the 355 children under five years who were included in the analysis, 51 (14.4%, CI:
10.9–18.5) were reported by the caregivers to have been treated with antibacterials prior to
PLOS ONE | https://doi.org/10.1371/journal.pone.0257328 September 16, 2021
5 / 14
PLOS ONE Validity of caregivers’ reports on prior use of antibacterials in children under five years
Table 1. Socio-demographic characteristics and prevalence of antibacterial use in children under five years prior to health facility visit as reported by caregivers of
children under five years in rural communities of Gulu district, northern Uganda (August, 2019).
Characteristics
Description
Respondent’s
Frequency (%)
Proportion of reported
antibacterial use, n (%)
95% CI
P-value (Pearson’s
chi-square test)
Overall
Sex of child
Location of health facility
Age of child (months)
Age of child caregiver (years)
Sex of child caregiver
Source of antibacterials
355 (100)
166 (46.8)
189 (53.2)
166 (46.8)
189 (53.2)
64 (18.0)
176 (49.6)
115 (32.4)
121 (34.1)
168 (47.3)
51 (14.4)
7 (2.0)
8 (2.2)
14 (3.9)
341 (96.1)
Male
Female
Urban
Rural
1–12
13–36
37–59
13–22
23–32
33–42
43–52
� 53
Male
Female
Home cabinet
Public health facility
Private clinics
Drug shops
Retail shops
Traditional healers
Antibacterials recommended by
Caregiver
Other household member
Friend/neighbor
Doctor/nurse
Drug seller/pharmacist
Traditional healer
Age of child (months), median (IQR)
Age of child caregiver (years), median (IQR)
29 (16.46)
25 (21.31)
n: Sample size; CI: Confidence Interval; %: Percentage; IQR: Interquartile range
https://doi.org/10.1371/journal.pone.0257328.t001
51 (14.4)
22 (13.3)
29 (15.3)
36 (21.7)
15 (7.9)
6 (9.4)
32 (18.2)
13 (11.3)
15 (12.4)
27 (16.1)
7 (13.7)
1 (14.3)
1 (12.5)
1 (7.1)
50 (14.7)
11 (23.9)
9 (37.5)
8 (30.8)
18 (35.3)
4 (30.8)
1 (50)
11 (31.4)
3 (37.5)
1 (25.0)
17 (34.0)
18 (29.0)
1 (33.3)
0.575
<0.001
0.119
0.936
0.432
<0.001
10.9–18.5
8.9–19.4
10.9–21.3
16.0–28.6
4.8–12.8
4.2–19.5
13.1–24.6
6.7–18.6
7.6–19.6
11.2–22.5
6.6–26.3
1.7–62.3
1.5–57.5
0.9–39.0
11.3–18.9
12.5–38.8
18.8–59.4
14.3–51.8
22.4–49.9
9.1–61.4
1.3–98.7
16.9–49.3
0.253
8.5–75.5
0.6–80.6
21.2–48.8
18.2–41.9
0.8–90.6
coming to the healthcare facility. Of these 51 children, the prevalence of antibacterial use was
higher in those from urban areas (21.7%, CI: 16.0–28.6) and in those who got antibacterials
from public health facilities (37.5%, CI: 18.8–59.4) (Table 1).
Prevalence of antibacterials detected in blood and urine samples of
children under five years
Of the 355 children under five years who were included in the analysis, 226 (63.7%, CI: 58.4–
68.7) had detectable levels of antibacterials in urine or blood in the samples taken upon arrival
to the healthcare facility (Table 2).
Most commonly used antibacterials. The most commonly used antibacterials as reported
by the care givers were amoxicillin (6.2%, 22/355), cotrimoxazole (2.8%, 10/355), and metroni-
dazole (2.3%, 8/355). The most common antibacterials detected from the laboratory analysis
were cotrimoxazole (29%, 103/355), ciprofloxacin (13%, 46/355), and metronidazole (9.9%,
35/355) (Fig 1)
PLOS ONE | https://doi.org/10.1371/journal.pone.0257328 September 16, 2021
6 / 14
PLOS ONE Validity of caregivers’ reports on prior use of antibacterials in children under five years
Table 2. Prevalence of antibacterials detected in blood or urine samples of children under five years in rural communities of Gulu district, northern Uganda
(August, 2019).
Characteristics
Description
Respondent’s
Frequency (%)
Proportion of
antibacterial detected, n (%)
Overall
Sex of child
Location of health facility
Age of child (months)
Age of child caregiver (years)
Sex of child caregiver
Male
Female
Urban
Rural
1–12
13–36
37–59
13–22
23–32
33–42
43–52
� 53
Male
355 (100)
166 (46.8)
189 (53.2)
166 (46.8)
189 (53.2)
64 (18.0)
176 (49.6)
115 (32.4)
121 (34.1)
168 (47.3)
51 (14.4)
7 (2.0)
8 (2.2)
14 (3.9)
226 (63.7)
108 (65.1)
118 (62.4)
103 (62.0)
123 (65.1)
45 (70.3)
112 (63.6)
69 (60.0)
83 (68.6)
99 (58.9)
36 (70.6)
5 (71.4)
3 (37.5)
8 (57.1)
Female
341 (96.1)
218 (63.9)
n: Sample size; CI: Confidence Interval; %: Percentage
https://doi.org/10.1371/journal.pone.0257328.t002
95% CI
58.4–68.7
57.5–71.9
55.3–69.1
54.4–69.1
57.9–71.6
57.9–80.3
56.2–70.4
50.7–68.6
59.7–76.3
51.3–66.2
56.6–81.5
29.7–93.7
11.4–73.6
30.7–80.1
58.7–68.9
P-value (Pearson’s
chi-square test)
0.608
0.554
0.389
0.164
0.605
Validity of caregivers’ reports of antibacterial intake in children under five
years
The sensitivity, specificity, PPV, NPV, agreement and kappa coefficient of the caregivers’
reports of use of antibacterials for treatment of children prior to healthcare facilities visit were
17.3% (12.6–22.8), 90.7% (84.3–95.1), 76.5% (62.5–87.2), 38.5% (33.0–44.2), 43.9% (38.7–
49.3%) and 0.06 (0.01–0.12) respectively. The sensitivity, specificity, PPV,NPV, agreement and
kappa coefficient varied between the different antibacterials (see Table 3).
Discussion
In this study we demonstrated that the prevalence of antibacterial use prior to health facility
visit was high and that caregivers under reported the use of antibacterials in the children under
five years prior to coming to the health facility. Antibacterial use prior to healthcare facility
visit is a common practice in many resource limited settings globally. Caregivers’ ability to
report antibacterial use before coming to the health facility is crucial for appropriate prescrip-
tion of antibacterial upon reaching health facilities [5]. Appropriate prescription of antibacter-
ials is important because it reduces the emergence of antibacterial resistance, poor clinical
outcomes, increased mortality and wastage of financial resources [19].
In the current study, almost two thirds (63.7%) of the samples (blood and/or urine) tested
positive for antibacterials. This implies that the prevalence of antibacterial use prior to health
facility visit is much higher than what was self-reported (14.4%). This finding is similar to
those from other low and middle income countries (LMIC) [4, 5, 10], a study carried out in
Ghana reported a prevalence of self-reported antibacterial use prior to health facility visit of
13%, however, analysis of urine samples reported a much higher prevalence of 64% [10]. In
Uganda, self-medication with antibacterials is a common practice [1, 4, 20] which is reflected
in the high prevalence of antibacterials found in the samples (blood and/or urine) in the cur-
rent study [1]. Another reason for the high prevalence of antibacterial use in our study is the
PLOS ONE | https://doi.org/10.1371/journal.pone.0257328 September 16, 2021
7 / 14
PLOS ONE Validity of caregivers’ reports on prior use of antibacterials in children under five years
Fig 1. Commonly used antibacterials according to the laboratory analysis.
https://doi.org/10.1371/journal.pone.0257328.g001
high prevalence of infectious diseases in these communities. In Uganda, 71% of children
under five years attending healthcare facilities do so due to acute respiratory infections [21],
however, in the community in this study, the most common diseases in children under five
years seeking care at healthcare facilities include; malaria, diarrhea, pneumonia, acute child-
hood malnutrition and HIV/AIDS [15–17]. High prevalence of antibacterials found in the
samples in the current study could also be due to exposure to antibacterials through
Table 3. Validity of caregivers’ reports of antibacterial intake in children under five years in rural communities of Gulu district, northern Uganda (August, 2019).
Parameters
Sensitivity (95% CI)
Specificity (95% CI)
PPV (95% CI)
NPV (95% CI)
Prevalence (95% CI)
Agreement (95% CI)
κ (95% CI)
Overall
17.3 (12.6–22.8)
90.7 (84.3–95.1)
76.5 (62.5–87.2)
38.5 (33.0–44.2)
63.7 (58.4–68.7)
43.9 (38.7–49.3)
0.06 (0.01–0.12)
Amoxicillin
5.6 (0.1–27.3)
93.8 (90.6–96.1)
4.5 (0.1–22.8)
94.9 (92.0–97.0)
5.1 (3.0–7.9)
89.3 (85.6–92.3)
Cotrimoxazole
5.8 (2.3–12.2)
98.4 (96.0–99.6)
60.0 (26.3–87.8)
71.9 (66.8–76.6)
29.0 (24.3–34.0)
71.5 (66.5–76.2)
-0.01 (-0.11–0.09)
0.06 (-0.01–0.12)
Metronidazole
2.9 (0.1–14.9)
97.8 (95.5–99.1)
12.5 (0.3–52.7)
90.2 (86.6–93.1)
9.9 (7.0–13.4)
88.5 (84.7–91.6)
0.01 (-0.08–0.1)
Ciprofloxacin
4.3 (0.5–14.8)
99.7 (98.2–100)
66.7 (9.4–99.2)
87.5 (83.6–90.8)
13.0 (9.6–16.9)
87.3 (83.4–90.6)
0.07 (-0.03–0.16)
%: percentage; CI: Confidence interval; κ: Kappa coefficient
https://doi.org/10.1371/journal.pone.0257328.t003
PLOS ONE | https://doi.org/10.1371/journal.pone.0257328 September 16, 2021
8 / 14
PLOS ONE Validity of caregivers’ reports on prior use of antibacterials in children under five years
consumption of water, vegetables and animal products [22, 23]. In the Hong Kong survey to
determine the presence of veterinary antibiotics in food, drinking water, and the urine of pre-
school children, it was found that 13 veterinary antibiotics were detectable in the urine of
77.4% of primary school children with norfloxacin and penicillin having the highest detection
rates. Enrofloxacin, penicillin, and erythromycin were the most detected veterinary antibiotics
in raw and cooked food [21]. Studies in Uganda, report a high prevalence of veterinary use of
antibacterials. The most commonly used antibacterials in veterinary medicine in Uganda
include; procaine penicillin, trimethoprim/sulfadiazine, erythromycin sulphate, tylosin tar-
trate, oxytetracycline hydrochloride [24, 25]. This high prevalence of antibacterial use can lead
to increased risk of resistance within the community [26]. A study was carried out in Uganda
to determine the epidemiology and antibiotic susceptibility of Vibrio cholerae associated with
the 2017 outbreak in Kasese district, and it reported that V. cholerae was highly resistant to the
commonly used antibiotics [27].
Most caregivers reported to have given their children amoxicillin, cotrimoxazole and met-
ronidazole. This is consistent with reports from a study in northern Uganda where metronida-
zole, amoxicillin, ciprofloxacin, doxycycline or cotrimoxazole were reported as the most
commonly used antibacterials by patients prior to hospital visit [4]. Metronidazole is com-
monly used for bacterial gastroenteritis, amoxicillin is used for bacterial chest infections, and
cotrimoxazole is used to treat pneumonia, bronchitis, infections of the urinary tract, ears intes-
tines and as prophylaxis against opportunistic infections in HIV [28, 29]. In our study the
most commonly detected antibacterials in the laboratory analysis results were cotrimoxazole,
ciprofloxacin and metronidazole, similar to findings from a study carried out in Ghana which
reported ciprofloxacin, trimethoprim or metronidazole as the most common antibacterials
detected in urine samples [30]. Ciprofloxacin is commonly used to treat pneumonia, typhoid
fever, infectious diarrhea, skin and bone infections [28, 29]. Amoxicillin was the most com-
monly reported antibacterial used and yet it was not among the most commonly detected anti-
bacterials from the laboratory analysis. This could be explained by the pharmacokinetics of
amoxicillin, which has a very short half-life of about 1 hour and will usually be out of the sys-
tem within 5 hours. Thus, meaning that for it to be detected in the blood or urine samples, it
should have been taken within a few hours before healthcare facility visit [28, 29]. We also
observed that the number of children who had cotrimoxazole in their biological samples was
higher than those who reported the use. It is possible that some of these children may have
tested positive for cotrimoxazole since they could have been receiving it as prophylaxis against
opportunistic infections in HIV [31, 32]. The prevalence of HIV/AIDS in northern Uganda as
of 2019 when data for this study was collected, was 7.2% in adults and 0.5% in children under
five years [33]. Since we were interested in antibacterial use for current illness, for which the
children were brought to the healthcare facility, caregivers might not have found it not neces-
sary to report the use of cotrimoxazole as prophylaxis against opportunistic infections in HIV.
The positive predicative value we found for reported use of antibacterials is not high enough
to allow caregivers reports to guide treatment. The high specificity values indicate under
reporting but the negative predictive value indicate that many children were given drugs that
were not reported by caregivers. This study was carried out in rural communities of Gulu dis-
trict in Uganda where the adult literacy levels are low [1, 20], and the inconsistencies in care-
givers’ response to interview questions and laboratory findings, could have been because of
caregivers inability to identify medicines taken as antibacterials. Another reason for the incon-
sistencies in self-reported antibacterial use and laboratory findings could have been due to
social desirability bias [34]. The caregivers could have been aware that self-medication is not a
good practice, and therefore feared to tell the interviewers the truth. Another reason for the
inconsistencies could have been due to consumption of these medicines from diffuse sources
PLOS ONE | https://doi.org/10.1371/journal.pone.0257328 September 16, 2021
9 / 14
PLOS ONE Validity of caregivers’ reports on prior use of antibacterials in children under five years
such as milk, water or food, studies in Uganda have reported veterinary use of antibacterials
[24, 25]. Another worrying explanation for the inconsistencies could be the quality of antibac-
terial medicines, some of these antibacterials may not contain the actual quantity of the active
medicine the manufacturers claim they contain. Although we did not set out to study the qual-
ity of antibacterials in this study, high prevalence of substandard antibacterial medicines has
been previously reported in developing countries [35]. Furthermore, inaccuracies in self-
reports may lead to duplication of therapy, incorrect management of the ill child, failure to
appreciate non-compliance leading to exacerbation of chronic medical conditions, or inaccu-
rate research conclusions [36].
We observed a strong association between high self-reported prior antibacterial use and the
source of antibacterials being from public health facilities. This could be attributed to the low
financial status of the people in these communities [1] forcing them to seek free healthcare
from public healthcare facilities. The district-based healthcare system in Uganda consists of
level I, II, III, IV and district hospitals [11]. This therefore means that by the time these chil-
dren were brought to health care level III and IV, they could have already sought care from the
lower levels and were referred to these higher levels for further management.
There is need for further research to understand the reasons for caregivers’ poor reports on
their children’s prior intake of antibacterials before coming to the health facility. Improved
validity could be promoted by encouraging health care workers to carefully explain to the care-
givers the medicines they administer to these children when they fall sick. Proper documenta-
tion of the medicines given to these children when they are sick could also improve the validity
of self-reported medicine use. There is need for the healthcare workers to educate the caregiv-
ers about the dangers of using antibacterials without consulting a healthcare worker, and also
further research is required to better understand why caregivers initiate antibacterial use at
home without consulting a healthcare service provider. This all will allow policy makers to be
better informed when planning interventions to reduce the large amount of incorrect antibac-
terial use in the community.
The results of our study should be considered in light of some limitations. This study could
have been affected by recall bias, where antibacterials given may have been forgotten. The
study could have also been affected by social desirability bias since the study was carried out in
a hospital setting and probably caregivers feared telling the truth because they thought it could
affect patient care. Under reporting could have been affected by how the questions were
understood by the caregivers. Failure to detect some of the antibacterials in the samples could
have been due to the pharmacokinetics of the antibacterials. Factors such as education level of
the caregivers could have contributed to the under reporting of antibacterial use prior to
healthcare facility visit, however, we didn’t collect this information. This is because adult liter-
acy levels in this community are low [1, 20] and to our knowledge previous studies have not
reported associations between self-report and education level [10, 37]. However, there is need
for further research to determine if there’s an association between caregivers’ education level
and reporting of prior antibacterial use in this setting. The discrepancy between the reported
use and the detected antibacterials in blood/urine samples could have also been because the
antibacterial could have been given for the management of another condition, such as cotri-
moxazole for prophylaxis against HIV related opportunistic infections, but we did not collect
this information. Our tool was designed to capture only antibacterial use for the illness for
which the children were brought to the healthcare facility. We were unable to report the levels
of antibacterials in relation to how far back the antibacterials were taken, this is because we
used a qualitative LC-MS/MS method which was developed to report the presence or absence
of antibacterials and not to quantify them.
PLOS ONE | https://doi.org/10.1371/journal.pone.0257328 September 16, 2021
10 / 14
PLOS ONE Validity of caregivers’ reports on prior use of antibacterials in children under five years
Conclusion
A high proportion of children under five years take antibacterials prior to visiting a healthcare
facility in northern Uganda. However, there is low validity of caregivers’ reports on prior
intake of antibacterials by these children. There is need for further research to understand the
factors associated with under reporting of prior antibacterial medicine use by caregivers of
children under five years. In addition, we suggest that health care workers should endeavor to
explain the role and names of medicines during dispensing, as well as the importance of
reporting correctly on prior medication intake. There is also need to educate the caregivers
about the dangers of using antibacterials without consulting a healthcare worker, and also fur-
ther research is required to better understand why caregivers initiate antibacterial use at home
without consulting a healthcare service provider. This all will allow policy makers to be better
informed when planning interventions to reduce the large amount of incorrect antibacterial
use in the community.
Supporting information
S1 Table. Summary of key pharmacokinetic properties of some of the antibacterials that
are commonly used among children under five years in rural communities of Gulu district,
northern Uganda (August, 2019).
(DOCX)
S1 Appendix. Ethical approval letters.
(DOCX)
S2 Appendix. Consent form.
(DOCX)
S3 Appendix. LC-MS/MS method.
(DOCX)
S4 Appendix. Questionnaire.
(DOCX)
Acknowledgments
We appreciate the tireless effort of the data collection team; Apio Patricia, Ojok Albert, Kagood
Francis, Hassan Chakaal and the village local chair persons for their guidance.
Author Contributions
Conceptualization: Hindum Lanyero, Moses Ocan, Celestino Obua, Cecilia Stålsby Lund-
borg, Joan N. Kalyango, Jaran Eriksen.
Data curation: Hindum Lanyero, Joan N. Kalyango, Jaran Eriksen.
Formal analysis: Hindum Lanyero, Katureebe Agaba.
Funding acquisition: Celestino Obua, Cecilia Stålsby Lundborg, Joan N. Kalyango, Jaran
Eriksen.
Investigation: Hindum Lanyero, Sarah Nanzigu.
Methodology: Hindum Lanyero, Moses Ocan, Katureebe Agaba, Joan N. Kalyango, Sarah
Nanzigu.
PLOS ONE | https://doi.org/10.1371/journal.pone.0257328 September 16, 2021
11 / 14
PLOS ONE Validity of caregivers’ reports on prior use of antibacterials in children under five years
Project administration: Moses Ocan, Celestino Obua, Cecilia Stålsby Lundborg, Joan N.
Kalyango, Jaran Eriksen.
Resources: Hindum Lanyero, Jaran Eriksen.
Software: Hindum Lanyero.
Supervision: Moses Ocan, Celestino Obua, Cecilia Stålsby Lundborg, Joan N. Kalyango, Jaran
Eriksen, Sarah Nanzigu.
Validation: Hindum Lanyero, Jaran Eriksen, Sarah Nanzigu.
Visualization: Hindum Lanyero.
Writing – original draft: Hindum Lanyero.
Writing – review & editing: Hindum Lanyero, Moses Ocan, Celestino Obua, Cecilia Stålsby
Lundborg, Katureebe Agaba, Joan N. Kalyango, Jaran Eriksen, Sarah Nanzigu.
References
1. Ocan M, Bwanga F, Bbosa GS, Bagenda D, Waako P, Ogwal-Ogeng J, et al. Patterns and predictors of
self-medication in Northern Uganda. PLoS ONE. 2014; 9(3):e92323. https://doi.org/10.1371/journal.
pone.0092323 PMID: 24658124
2. Wellcome Trust, UK Government. Safe, secure and controlled:managing the supply chain of antimicro-
bials. Review on antimicrobial Resistance. 2015.
3. Ministry of Health Uganda. Uganda National Drug Policy. 2002. Available from: http://library.health.go.
ug/publications/policy-documents/uganda-national-drug-policy-2002.
4. Ocan M, Manabe YC, Baluku H, Atukwase E, Ogwal-Okeng J, Obua C. Prevalence and predictors of
prior antibacterial use among patients presenting to hospitals in Northern Uganda. BMC Pharmacology
and Toxicology. 2015; 16(1):26. https://doi.org/10.1186/s40360-015-0027-8 PMID: 26407973
5. Hildenwall H, Lindkvist J, Tumwine JK, Bergqvist Y, Pariyo G, Tomson G, et al. Low validity of caretak-
ers’ reports on use of selected antimalarials and antibiotics in children with severe pneumonia at an
urban hospital in Uganda. Transactions of the Royal Society of Tropical Medicine and Hygiene. 2009;
103(1):95–101. https://doi.org/10.1016/j.trstmh.2008.04.046 PMID: 18678381
6. Omolase C, Adeleke O, Afolabi A, Ofolabi O. Self medication amongst general outpatients in a Nigerian
community hospital. Annals of Ibadan postgraduate medicine. 2007; 5(2):64–7. https://doi.org/10.4314/
aipm.v5i2.64032 PMID: 25161435
7. Moise K, Bernard JJ, Henrys JH. Evaluation of antibiotic self-medication among outpatients of the state
university hospital of Port-Au-Prince, Haiti: a cross-sectional study. Pan African Medical Journal. 2017;
28(1). https://doi.org/10.11604/pamj.2017.28.4.12589 PMID: 29138650
8. Stirratt MJ, Dunbar-Jacob J, Crane HM, Simoni JM, Czajkowski S, Hilliard ME, et al. Self-report mea-
sures of medication adherence behavior: recommendations on optimal use. Transl Behav Med. 2015; 5
(4):480–2. https://doi.org/10.1007/s13142-015-0315-2 PMID: 26622919
9. Nsimba SE, Massele AY, Eriksen J, Gustafsson LL, Tomson G, Warsame M. Case management of
malaria in under-fives at primary health care facilities in a Tanzanian district. Trop Med Int Health. 2002;
7(3):201–9. https://doi.org/10.1046/j.1365-3156.2002.00847.x PMID: 11903982
10.
Lerbech AM, Opintan JA, Bekoe SO, Ahiabu M-A, Tersbøl BP, Hansen M, et al. Antibiotic exposure in a
low-income country: screening urine samples for presence of antibiotics and antibiotic resistance in
coagulase negative staphylococcal contaminants. PLoS One. 2014; 9(12):e113055. https://doi.org/10.
1371/journal.pone.0113055 PMID: 25462162
11. Kamwesiga J. Uganda Health Care System. Kampala, Uganda: Makerere University. 2011. https://doi.
org/10.1186/1745-6215-12-192 PMID: 21838913
12. Ministry of Health Uganda. National Health Facilty Master List. 2017. Available from: http://library.
health.go.ug/sites/default/files/resources/National%20Health%Facility%MasterLlist%202017.pdf.
13. National Drug Authority. Licensed Out-lets. 2021.Available from: https://www.nda.or.ug/licensed-
outlets/.
14. National Drug Authority. Drug Shops Licensed. 2021. Available from: https://www.nda.or.ug/drug-
shops-licensed-in-2020/#1539347133280-41e0655f-3957.
PLOS ONE | https://doi.org/10.1371/journal.pone.0257328 September 16, 2021
12 / 14
PLOS ONE Validity of caregivers’ reports on prior use of antibacterials in children under five years
15. Simple O, Mindra A, Obai G, Ovuga E, Odongo-Aginya EI. Influence of Climatic Factors on Malaria Epi-
demic in Gulu District, Northern Uganda: A 10-Year Retrospective Study. Malar Res Treat. 2018;
2018:5482136. https://doi.org/10.1155/2018/5482136 PMID: 30186590
16. Omona S, Malinga GM, Opoke R, Openy G, Opiro R. Prevalence of diarrhoea and associated risk fac-
tors among children under five years old in Pader District, northern Uganda. BMC Infectious Diseases.
2020; 20(1):1–9. https://doi.org/10.1186/s12879-020-4770-0 PMID: 31931735
17. Namakula J, Ssengooba F, Ssali S. Country situation analysis: northern Uganda. ReBUILD Consotium
Kampala: Makerere University School of Public Health. 2011. Available from: https://www.
rebuildconsortium.com/media/1024/country-situation-analysis-uganda.pdf
18. Kish Leslie. Survey Sampling. New York, United States: John Wiley & Sons Inc; 1995.
19. O’Connor R, O’Doherty J, O’Regan A, Dunne C. Antibiotic use for acute respiratory tract infections
(ARTI) in primary care; what factors affect prescribing and why is it important? A narrative review. Irish
Journal of Medical Science (1971-). 2018; 187(4):969–86.
20.
Lanyero H, Eriksen J, Obua C, Stålsby Lundborg C, Nanzigu S, Katureebe A, et al. Use of antibacterials
in the management of symptoms of acute respiratory tract infections among children under five years in
Gulu, northern Uganda: Prevalence and determinants. PloS one. 2020; 15(6):e0235164. https://doi.
org/10.1371/journal.pone.0235164 PMID: 32574206
21. UNICEF. Uganda (UGA)-Demographics, Health & Infant mortality-UNICEF DATA. 2020. Available
from: https://data.unicef.org/country/uga/.
22. Radyowijati A, Haak H. Improving antibiotic use in low-income countries: an overview of evidence on
determinants. Social science & medicine. 2003; 57(4):733–44. https://doi.org/10.1016/s0277-9536(02)
00422-7 PMID: 12821020
23.
Li N, Ho KW, Ying G-G, Deng W-J. Veterinary antibiotics in food, drinking water, and the urine of pre-
school children in Hong Kong. Environment international. 2017; 108:246–52. https://doi.org/10.1016/j.
envint.2017.08.014 PMID: 28889029
24. Nayiga S, Kayendeke M, Nabirye C, Willis LD, Chandler CI, Staedke SG. Use of antibiotics to treat
humans and animals in Uganda: a cross-sectional survey of households and farmers in rural, urban and
peri-urban settings. JAC-Antimicrobial Resistance. 2020; 2(4):dlaa082. https://doi.org/10.1093/jacamr/
dlaa082 PMID: 34223037
25. Bashahun D, Odoch T. Assessment of antibiotic usage in intensive poultry farms in Wakiso District,
Uganda. Livestock Research for Rural Development. 2015; 27(12).
26. Austin DJ, Kristinsson KG, Anderson RM. The relationship between the volume of antimicrobial con-
sumption in human communities and the frequency of resistance. Proceedings of the National Academy
of Sciences. 1999; 96(3):1152–6. https://doi.org/10.1073/pnas.96.3.1152 PMID: 9927709
27.
Iramiot JS, Rwego IB, Kansiime C, Asiimwe BB. Epidemiology and antibiotic susceptibility of Vibrio cho-
lerae associated with the 2017 outbreak in Kasese district, Uganda. BMC public health. 2019; 19(1):1–
9. https://doi.org/10.1186/s12889-018-6343-3 PMID: 30606151
28. Rang HP, Ritter JM, Flower R, Henderson G. Rang & Dale’s pharmacology E-book. Elsevier Health
Sciences; 2014. Available from: https://www.elsevier.com/books/rang-andampamp-dales-
pharmacology/ritter/978-0-7020-5362-7
29.
30.
Flower RJ, Henderson G, Loke YK, MacEwan D, Rang HP. Rang & Dale’s Pharmacology E-Book:
Elsevier Health Sciences; 2018. Available from: https://www.elsevier.com/books/rang-and-dales-
pharmacology/ritter/978-0-7020-7448-6
Lerbech AM, Opintan JA, Bekoe SO, Ahiabu M-A, Tersbøl BP, Hansen M, et al. Antibiotic exposure in a
low-income country: screening urine samples for presence of antibiotics and antibiotic resistance in
coagulase negative staphylococcal contaminants. PloS one. 2014; 9(12):e113055. https://doi.org/10.
1371/journal.pone.0113055 PMID: 25462162
31. Badri M, Ehrlich R, Wood R, Maartens G. Initiating co-trimoxazole prophylaxis in HIV-infected patients
in Africa: an evaluation of the provisional WHO/UNAIDS recommendations. AIDS (London, England).
2001; 15(9):1143–8. https://doi.org/10.1097/00002030-200106150-00009 PMID: 11416716
32. Mermin J, Lule J, Ekwaru JP, Downing R, Hughes P, Bunnell R, et al. Cotrimoxazole prophylaxis by
HIV-infected persons in Uganda reduces morbidity and mortality among HIV-uninfected family mem-
bers. AIDS (London, England). 2005; 19(10):1035–42. https://doi.org/10.1097/01.aids.0000174449.
32756.c7 PMID: 15958834
33. Ministry of Health Uganda. THE 2019 HIV EPIDEMIOLOGICAL SURVEILLANCE REPORT FOR
UGANDA. 2020. Available from: https://www.health.go.ug/cause/the-2019-hiv-epidemiological-
surveillance-report-for-uganda/.
34.
Johnson LC, Beaton R, Murphy S, Pike K. Sampling bias and other methodological threats to the valid-
ity of health survey research. International Journal of Stress Management. 2000; 7(4):247–67.
PLOS ONE | https://doi.org/10.1371/journal.pone.0257328 September 16, 2021
13 / 14
PLOS ONE Validity of caregivers’ reports on prior use of antibacterials in children under five years
35. Caudron JM, Ford N, Henkens M, Mace C, Kiddle-Monroe R, Pinel J. Substandard medicines in
resource-poor settings: a problem that can no longer be ignored. Tropical Medicine & International
Health. 2008; 13(8):1062–72.
36. Monte AA, Heard KJ, Hoppe JA, Vasiliou V, Gonzalez FJ. The accuracy of self-reported drug ingestion
histories in emergency department patients. The Journal of Clinical Pharmacology. 2015; 55(1):33–8.
https://doi.org/10.1002/jcph.368 PMID: 25052325
37. Hafferty JD, Campbell AI, Navrady LB, Adams MJ, MacIntyre D, Lawrie SM, et al. Self-reported medica-
tion use validated through record linkage to national prescribing data. Journal of clinical epidemiology.
2018; 94:132–42. https://doi.org/10.1016/j.jclinepi.2017.10.013 PMID: 29097340
PLOS ONE | https://doi.org/10.1371/journal.pone.0257328 September 16, 2021
14 / 14
PLOS ONE |
10.1371_journal.pone.0260153 | RESEARCH ARTICLE
Contextual factors influencing a training
intervention aimed at improved maternal and
newborn healthcare in a health zone of the
Democratic Republic of Congo
Malin BogrenID
1*, Sylvie Nabintu Mwambali2, Marie BergID
1,2
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
1 Institute of Health and Care Sciences, Sahlgrenska Academy, University of Gothenburg, Gothenburg,
Sweden, 2 Faculty of Medicine and Community Health, Department of Obstetrics and Gynecology,
Evangelical University of Africa, Bukavu, Democratic Republic of Congo
* malin.bogren@gu.se
Abstract
Background
OPEN ACCESS
Citation: Bogren M, Mwambali SN, Berg M (2021)
Contextual factors influencing a training
intervention aimed at improved maternal and
newborn healthcare in a health zone of the
Democratic Republic of Congo. PLoS ONE 16(11):
e0260153. https://doi.org/10.1371/journal.
pone.0260153
Editor: Ashraful (Neeloy) Alam, The University of
Sydney Faculty of Medicine and Health,
AUSTRALIA
Received: May 7, 2021
Accepted: November 3, 2021
Published: November 29, 2021
Peer Review History: PLOS recognizes the
benefits of transparency in the peer review
process; therefore, we enable the publication of
all of the content of peer review and author
responses alongside final, published articles. The
editorial history of this article is available here:
https://doi.org/10.1371/journal.pone.0260153
Copyright: © 2021 Bogren et al. This is an open
access article distributed under the terms of the
Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: Data cannot be
shared publicly because to do so could potentially
Maternal and neonatal mortality and morbidity in the Democratic Republic of Congo (DRC)
are among the highest worldwide. As part of a quality improvement programme in a health
zone in the DRC aimed at contributing to reduced maternal and neonatal mortality and mor-
bidity, a three-pillar training intervention around childbirth was developed and implemented
in collaboration between Swedish and Congolese researchers and healthcare profession-
als. The aim of this study is to explore contextual factors influencing this intervention.
Methods
A qualitative research design was used, with data collected through focus group discussions
(n = 7) with healthcare professionals involved in the intervention before and at the end (n =
9). Transcribed discussions were inductively analysed using content analysis.
Results
Three generic categories describe the contextual factors influencing the intervention: i)
Incentives motivated participants’ efforts to begin a training programme; ii) Involving the
local health authorities was important; and (iii) Having physical space, electricity, and equip-
ment in place was crucial.
Conclusions
This study and similar ones highlight that incentives of various types are crucial contextual
factors that influence training interventions, and have to be considered already in the plan-
ning of such interventions. One such factor is expectations of monetary incentives. To meet
this in a small research project like ours would require a reduction of the scale and thus limit
the implementation of new evidence-based knowledge into practice aimed at reducing
maternal mortality and morbidity.
PLOS ONE | https://doi.org/10.1371/journal.pone.0260153 November 29, 2021
1 / 14
PLOS ONE identify our study participants and individual
healthcare facilities, and assurances were given to
respondents that any publication would not do so.
Requests for access to the data underlying our
findings will be considered by the National Ethical
Committee of Public Health/CNES South Kivu
province, and should be addressed to Prof Kitoka
Moke at kitoka02@yahoo.fr.
Funding: The study was conducted with financial
assistance from the Laerdal Foundation and
Sahlgrensringen. The funders had no role in study
design, data collection and analysis, decision to
publish, or preparation of the manuscript.
Competing interests: The authors have declared
no competing interests exist.
Abbreviations: HBB, Helping Babies Breathe; HMS,
Helping Mothers Survive; DRC, Democratic
Republic of Congo; FGDs, focus group
discussions; SDG, Sustainable Development Goal.
Contextual factors influencing a training intervention in the Democratic Republic of Congo
Introduction
In the Democratic Republic of Congo (DRC), the maternal and neonatal mortality ratio
(MMR) remains high. According to the latest nationally reported statistics, in 2013–14 the
maternal mortality rate was 846 deaths per 100,000 live births and the neonatal mortality rate
28 deaths per 1,000 live births [1]. This, despite the fact that 80% of births were assisted in
healthcare facilities by skilled healthcare providers, consisting of either midwives, physicians,
or nurses [2].
The provision of high-quality care is central in achieving health-related targets within the
Sustainable Development Goals (SDGs), especially the health of mothers and newborns [3,4].
In the strive for high-quality care around childbirth, health systems need to ensure that all
women and their newborns receive quality care, defined as being scientifically evidence-based,
equitable, respectful, effective, timely, efficient, and person-centred [2,5–8]. Furthermore,
health systems need to be adapted to context, i.e. the environment and surroundings in which
a proposed change is to be implemented [9]. Poor quality of care is a greater barrier than insuf-
ficient access to healthcare in the DRC as well as worldwide [3,10].
The training of healthcare providers is known to be a useful tool for reducing maternal and
newborn mortality; however, no specific training strategy is effective in every context [11],
which means that the same interventions have different effects in different contexts. Context
includes anything internal and external to an intervention that may act as a barrier to or facili-
tator of its implementation or effect. Thus, it is essential to understand the context–including
identifying which contextual factors influence a particular quality improvement intervention
and how they do so [9,12]. As part of an implementation project aimed at contributing to
reduced maternal and neonatal mortality and morbidity in a health zone in the South Kivu
province of the DRC, this study’s objective was to explore contextual factors influencing a
training intervention focusing on healthcare practice during childbirth. The lessons learned
from the results are presumed to also be useful in other similar contexts, in the DRC as well as
low-income countries elsewhere, when designing and implementing similar training
interventions.
Method
Study design
The study was approved by the National Ethical Committee of Public Health: CNES 001/
DPSKI/129PM/2019. A qualitative research design was used [13], and data was collected
through focus-group discussions (FGDs) with healthcare professionals participating in the
training intervention.
Setting
The DRC comprises 26 provinces, with more than 500 health zones which are organised to
deliver healthcare at three levels. The primary level of care is offered at healthcare centres,
some of which also offer perinatal care. The secondary level is offered at district hospitals,
which have the capacity to perform C-sections, and the tertiary level is offered at referral hospi-
tals (one per health zone) [14]. The healthcare facilities are governed by either the governmen-
tal or private sector.
The health zone where this implementation project took place is one of three in the provin-
cial capital in the South Kivu province, situated in the eastern part of the DRC. At the time of
the intervention, this health zone served more than 450,000 inhabitants at 40 healthcare facili-
ties, of which 34 were healthcare centres, five were district hospitals, and one was a referral
PLOS ONE | https://doi.org/10.1371/journal.pone.0260153 November 29, 2021
2 / 14
PLOS ONE Contextual factors influencing a training intervention in the Democratic Republic of Congo
Table 1. Level, governance, and number of births at each healthcare facility.
Healthcare facilities
Level of healthcare
Financial and governance structure
Births in 2018
1
Tertiary
Private
3,229
2
3
Secondary
Secondary
Private
847
Private
1,249
4
Primary
Private
547
5
Secondary
Private
418
6
Primary
Private
921
7
Primary
Private
165
https://doi.org/10.1371/journal.pone.0260153.t001
hospital. This study is part of a maternal and newborn healthcare quality improvement project
being conducted at seven of these 40 healthcare facilities: three healthcare centres, three district
hospitals, and the referral hospital. There were 16,101 registered births in the health zone in
2018, 7,416 of which occurred at these seven healthcare facilities. The facilities’ level of health-
care and governance are shown in Table 1:
Intervention
The training programme was developed based on core principles of conducting person-cen-
tred holistic care [15,16] as well as a woman-centred model of childbirth care [17], and with
the overall aim to promote healthy physiologic, vaginal birth. The programme was divided
into three pillars, with activities to 1) promote normal physiologic birth, 2) prevent and man-
age complications during labour and birth, and 3) strengthen the healthcare professionals’
self-reflection skills and self-confidence.
The first and second pillars consisted of theory and simulation-based training using equip-
ment from the Laerdal foundation [18], and additional tools such as birthing balls and Rebozo
sheets. The first pillar was based on principles included in the midwifery model of woman-cen-
tred care during childbirth [17] and in the Rebozo technique [19]. The second pillar was based
on the programmes Helping Mothers Survive Bleeding after Birth (HMS-BAB) developed by
Jhpiego [20], and Helping Babies Breathe (HBB) developed by the American Academy of Pedi-
atrics [21]. Pillar 3 consisted of reflection in groups based on a process-oriented group reflec-
tion model in which the participants reflect on themselves selected own experienced situations
related to their professional role [22].
The implementation of the training programme was planned and steered by a multiprofes-
sional project committee of healthcare professionals from the DRC (n = 3) and Sweden
(n = 3), after having been developed by the Swedish research group. Details of the implementa-
tion are described in Table 2. The Congolese partners chose seven of the 40 healthcare facili-
ties, representing all three healthcare levels. Each facility selected its own training facilitators.
A 25-day training in the three-pillar programme was first given to four selected master train-
ers, of whom three were nurses working as midwives and one was a physician specialising in
gynaecology. Next, the master trainers gave a six-day training to 13 selected training facilita-
tors–two from each of six healthcare facilities, and one from the smallest one–consisting of ten
nurse/midwives, two gynaecologists, and one paediatrician. Next, the seven healthcare facilities
were equipped with equipment to conduct the training. This included uniquely developed
didactic teaching material for using the programme, a birthing ball, a specially designed sheet
for use of the Rebozo technique [19], and the Laerdal products ‘MamaNatalie Complete’,
‘MamaBirthie’, and ‘NeoNatalie Complete’ [18]. Further, based on the ‘low-dose high-fre-
quency’ training pedagogy [21], a detailed schedule was defined for doing week-to-week short
training activities for a period of six months, including weekly training in Pillars 1 and 2 and
process-oriented reflections in groups once per week. The master trainers mentored the facili-
tators at the healthcare facilities and led the process-oriented reflections with the staff. The
master trainers received a small monetary incentive, while the facilitators did not.
PLOS ONE | https://doi.org/10.1371/journal.pone.0260153 November 29, 2021
3 / 14
PLOS ONE Contextual factors influencing a training intervention in the Democratic Republic of Congo
Table 2. Description of the implementation of the three-pillar training programme.
Activities
Project planning
Time
3 months
Content
Development of didactic teaching materials and
procurement of training equipment
Setup of steering committee
Identification of healthcare facilities (one hospital on
tertiary level, three hospitals on secondary level, and three
healthcare centres)
Introducing the local health zone authorities to the training
programme and its concept, and obtaining approval
Introducing the participating healthcare facilities’
management and staff to the three-pillar training
programme and its concept
Collection of baseline statistics on labour and birth
Equipment assessments conducted at the seven facilities
using a) the Jhpiego checklists, and b) the HMS/HBB
training list
Pre-intervention discussions on contextual barriers and
facilitators with the healthcare professionals at the
participating facilities
Training local master
trainers
Training of four master trainers (three midwives and one
physician)
25 (14+11) days
Training local facilitators
Training of 13 facilitators (two from each of six facilities and
one from the smallest one)
Training healthcare
professionals
Distribution of teaching materials, training equipment, and
training schedule
1 month
Continuously facilitation
Monthly visits to each healthcare facility by master trainers
6 months
Introduction of training programme by local facilitators
Follow-up visits to each
facility
Follow-up meeting with
the master trainers
Weekly practice by local facilitators using the low-dose high-
frequency practice
Dialogue with facilitators and healthcare professionals at
each healthcare facility, providing opportunities to share
experiences of implementing the training programme.
Dialogue with master trainers, providing opportunities to
share experiences of working with implementing the
training programme
Last month of the
training programme
https://doi.org/10.1371/journal.pone.0260153.t002
Data collection
Data was gathered through FGDs in two periods: before the training started (FGDs = 7) and at
the end of the programme, when facilitators and master trainers were also interviewed
(FGDs = 9). The local project leader of the training programme (SNM) contacted the health
zone authorities and the managers at each participating healthcare facility and informed them
about the study. Authorities from the health zone and managers at each healthcare facility
approved it, and provided contact information for available healthcare staff working at the
maternity unit who had taken part in the three-pillar training programme. The project leader
contacted these individuals and invited them to participate in FGDs after giving them verbal
and written information about the study, including the fact that participation was voluntary
and that they had the right to withdraw at any time without explanation. All invited healthcare
professionals (n = 61), being either nurses, midwives, or gynaecologists, agreed to participate
and signed informed consent.
All 16 FGDs were conducted by two of us authors (MBo and MBe). There were three to
seven participants in each group. The discussions were led by MBe in French, based on an
interview guide (see S1 Appendix), and were translated continuously during the FGDs into
PLOS ONE | https://doi.org/10.1371/journal.pone.0260153 November 29, 2021
4 / 14
PLOS ONE Contextual factors influencing a training intervention in the Democratic Republic of Congo
Table 3. Examples of the data analysis process from meaning unit to category.
Meaning Unit
Code
Subcategory
Category
We, as trained [facilitators, don’t get any sugar [compensation] for the
training we hold [at the healthcare facilities], and we didn’t get any
compensation when we took part in the training to become facilitators
Compensation is
expected
Monetary incentives for
participating are expected
Incentives influence participants’
efforts to begin a training
programme
https://doi.org/10.1371/journal.pone.0260153.t003
Swedish to MBo, who made field notes and asked clarifying questions. The FGDs were audio-
recorded and lasted 30 to 60 minutes, with a mean of 45 minutes.
Data analysis
The audio-recorded FGDs were analysed following principles of qualitative inductive con-
ducted analysis [23]. First, all transcripts were read several times. Next, in new readings, mean-
ing units were identified that answered the research question ‘What are the contextual factors
influencing the three-pillar training intervention and how do they influence it?’ The meaning
units were then compared and sorted into codes based on similar content, which were thereaf-
ter compared and clustered into subcategories and categories. The analysis process was com-
pleted by MBo and MBe separately, with repeated discussions until full agreement was
reached. An example of the analysis process is shown in Table 3.
Results
Contextual factors identified as influencing the implementation of the three-pillar training
programme were sorted into three generic categories with respective subcategories; for an
overview, see Table 4. In the presentation of the results the FGDs conducted in the two periods
are labelled FGD 1 and FGD 2, respectively, with the facilities where they were held labelled
1–7 (see Table 1).
Incentives motivate participants’ efforts to begin a training programme
The incentives that influenced participants’ efforts to get the three-pillar training programme
up and running consist of three subcategories, as follows.
Gaining increased knowledge and skills motivates. Motivation was high among the par-
ticipants to take part in the three-pillar training programme as it provided them with updated
knowledge and skills for daily practice, which could contribute to healthy and positive
childbirth.
Table 4. Categories and subcategories describing contextual factors influencing the three-pillar training
intervention.
Generic category
Subcategory
Incentives motivate participants’ efforts to begin a training
programme
The importance of involving the local health authorities
Gaining increased knowledge and skills motivates
Women’s utilisation of the healthcare facilities
motivates
Monetary incentives for participating are expected
Authorities from the health zone need to be
involved
The healthcare facilities’ management needs to be
involved
The need to have physical space, electricity, and equipment in
place
Inadequate physical space and electricity
Lack of equipment to promote physiologic birth
https://doi.org/10.1371/journal.pone.0260153.t004
PLOS ONE | https://doi.org/10.1371/journal.pone.0260153 November 29, 2021
5 / 14
PLOS ONE Contextual factors influencing a training intervention in the Democratic Republic of Congo
Most of the participating healthcare professionals had not received any formal in-service
training since their professional education, and the three-pillar training programme provided
them with new knowledge. At two facilities they typically arranged training themselves in
areas where care was not optimally conducted, for example in how to resuscitate a newborn.
At one facility, the church sometimes organised training, and those who had participated
shared their new knowledge with colleagues:
There is a lack of in-service training. We get nothing. Sometimes we’re invited to seminars
organised by the private health sector, but no such seminars are organised by the government
health zone; they just give some random information. (FGD 1, Healthcare Facility 7)
Most of us have not gotten any formal in-clinic training since we completed our pre-service profes-
sional education. We try to solve this through organising our own in-service trainings using our
doctors, who have been at the university. But they only have their own notes, and would have
needed to have PowerPoints and other educational materials. (FGD 1, Healthcare Facility 4)
All participants, except for at one healthcare facility, expressed an awareness that they did
not have the latest scientific evidence-based knowledge and accordingly were not practising
optimally. They were primarily motivated to learn in several areas such as promoting normal
physiologic birth and correctly managing complications like postpartum haemorrhage. The
materials they were provided–birthing balls and Rebozo sheets for use during labour, and pen-
guins and a ventilation mask for aspirating the newborn in need–added to their motivation
and their possibility to practise the skills they had learned:
We don’t have an ambulance that we can use to transport the women who need a C-section,
and the ambulance from the reference hospital often comes much too late. The training we’ve
gotten helps us support a normal birth and lets us handle acute conditions like bleeding. (FGD
2, Healthcare Facility 6)
There was a positive attitude regarding sharing knowledge between the different healthcare
facilities within the same health zone. The three-pillar training was regarded as such a knowl-
edge exchange programme both within an individual healthcare facility as well as between the
different facilities. There was a desire to develop such knowledge sharing even further, as the
healthcare facilities had the same type of patients with similar backgrounds and health condi-
tions. The master trainers, in turn, acknowledged that the training programme had given them
a mandate to have access to and connect with the other facilities within the zone.
Women’s utilisation of the healthcare facilities motivates. Another motivating factor
for the healthcare professionals to participate in the training programme was when they
noticed that their changed care routines had influenced how women informed their peers
about their positive experiences of being cared for at the healthcare facilities, which in turn
positively influenced other women’s decisions to seek care at the facilities. It had also been
noticed that several women arrived earlier when their labour had started, which in turn influ-
enced the outcome:
The training was fantastic, both for us as staff at the clinic and for the women who come in
and give birth. At one of the clinics, many more women giving birth are coming in. Now,
when a woman in labour comes, the staff are close to the woman and massage her. When the
woman goes home she tells others where she lives about her positive experience. (FGD 2 with
master trainers)
PLOS ONE | https://doi.org/10.1371/journal.pone.0260153 November 29, 2021
6 / 14
PLOS ONE Contextual factors influencing a training intervention in the Democratic Republic of Congo
It was acknowledged that caring for an extremely poor, and often low-educated, population
was challenging. Women commonly feared having a C-section. With the newly trained care
routines that promoted a normal physiologic birth, there was a belief among the healthcare pro-
fessionals that their increased knowledge and skills in turn would increase the prevalence of vag-
inal, non-instrumental births, which in turn would motivate women to give birth at the facility:
The population is very poor and has a low education level; this makes it difficult to motivate
the patients for different decisions about care. (FGD 1, Healthcare Facility 4)
The pregnant women are afraid of having repeated C-sections. There are different reasons;
one can be that the family force her to give birth normally in order to be considered a real
woman. Another reason for rejecting a C-section is the cost; they therefore reject having a C-
section. (FGD 1, Healthcare Facility 5)
Monetary incentives for participating are expected. Another strong, motivating factor
for participating in the training programme was the expectation of monetary incentives. This
expectation was the same for master trainers, facilitators, and the healthcare professionals at
the facilities, and was based on the fact that monetary incentives were commonly provided by
other projects they had participated in. As this training project had no such monitoring incen-
tive system in place for the weekly participation in training, it made participants at the health-
care facilities hesitant to attend the sessions:
You know how it is with Africans: they don’t come if they don’t get any sugar [compensation].
You need something to motivate them; money’s needed as motivation. // We, as trained [facil-
itators], don’t get any sugar [compensation] for the training we hold [at the healthcare facili-
ties], and we didn’t get any compensation when we took part in the training to become
facilitators. (FGD 2 with facilitators)
The healthcare professionals at each facility followed work schedules covering 24 hours,
seven days a week, which made it difficult for them to attend every scheduled training activity.
Activities that were part of the Pillar 1 and 2 trainings were often practised in the morning,
which made it challenging for those who had been working the night shift. The importance of
being reimbursed as motivation to participate in the training was stressed, at least being reim-
bursed for transportation costs if the training activities were undertaken when someone was
off duty:
It can be hard to convince the staff to take part in the different training steps. What’s hard is
motivating the staff to stay and train after the end of their workday, as well as motivating
staff to come in on their day off to train. Sometimes they refuse to come in because they’re not
paid for the transportation. (FGD 2 with facilitators)
The master trainers stressed that the provision of incentives would increase their motiva-
tion to conduct the scheduled training activities. This, as being a master trainer and a facilita-
tor was regarded as having dual work responsibilities–both their ordinary work as well as this
training–which therefore required sufficient monetary incentives. The lack of monetary incen-
tives to facilitators and the insufficient incentives to master trainers contributed to a lack of
motivation:
PLOS ONE | https://doi.org/10.1371/journal.pone.0260153 November 29, 2021
7 / 14
PLOS ONE Contextual factors influencing a training intervention in the Democratic Republic of Congo
We’d like to get paid when we train the others; we only get a transportation allowance. When
we had our master training in April and May, we didn’t get any compensation. (FGD 2 with
master trainers)
Because facilitators don’t get paid, they’re not motivated to train their colleagues. (FGD 2
with master trainers)
The importance of involving the local health authorities
This category describes the importance of involving both the authorities from the health zone
as well as the health management at the participating healthcare facilities in the three-pillar
training programme, in order to make it more successful and sustainable.
Authorities from the health zone need to be included. Getting the local health zone author-
ity involved was critical for successful implementation, and obtaining this authority’s approval
was perceived as a requirement for conducting the three-pillar training programme. Involving the
local health authority in the training could therefore support and encourage the healthcare facili-
ties to include the training activities within their daily routines. The authorities from the health
zone were informed about the training action and had granted permission to conduct it; but to
support the project further, according to the master trainers, the health inspectors would need to
be incentivised, and it was suggested that the project consider budgeting for this:
The health inspectors from the health zone are included to some degree, but they’re not moti-
vated to support the project. They want to be a part of this project. They want to be there
when we master trainers go to the clinics, but they don’t want to be trainers. If they were
included more they’d be able to motivate the staff to participate in the project during their reg-
ular visits to the healthcare facilities. . . . The inspectors have expressed that if they’re paid,
like they are in other projects, they can encourage the healthcare facilities to take part in the
training. (FGD 2 with master trainers)
The healthcare facilities’ management needs to be involved.
Involving the healthcare
management at each healthcare facility was stressed to be of critical importance, as they make
all the decisions concerning care and care routines. And if the training programme was to gain
sustainability and continue beyond its scheduled time, creating ownership among the local
management was acknowledged as crucial:
We have a culture in which the responsible parties are higher in rank than us; we others feel
lower than them. So if this training is to continue, you have to involve more of those at man-
agement level at the healthcare facilities. That would increase the ownership. Then, manage-
ment will take greater responsibility, they’ll increase our motivation, create an ownership.
(FGD 2, Healthcare Facility 1)
Being involved entailed not only being informed about and influencing the training strat-
egy, which was a part of the project; it also included receiving monetary incentives. If not, this
could act as an obstacle to the training programme. This was especially clear at the tertiary-
level facility. These leaders hindered the programme in various ways, for example by not taking
part in the training and even stating that they had never heard about it:
If people with management responsibility aren’t directly involved in the project, they’ll turn
the responsibility over to those who are running the project. You have to bring in the boss,
make him part of the project. When the boss talks everybody listens; I can’t ask the boss to do
PLOS ONE | https://doi.org/10.1371/journal.pone.0260153 November 29, 2021
8 / 14
PLOS ONE Contextual factors influencing a training intervention in the Democratic Republic of Congo
various things, I can’t give the boss orders. So the boss has to participate and have a mandate.
And that means that the boss needs money from the project. It’s not a salary; it’s a motivation.
If the boss is motivated, we can get everybody to do what we want in the project. If he doesn’t
get paid he’s not going to participate himself and he’s going to work against it. The other
healthcare facilities are small; this is a gigantic clinic, and the boss has to be involved. (FGD
2with master trainers)
The need to have physical space, electricity, and equipment in place
This category describes, in two subcategories, the need to have physical space, electricity, and
equipment in place in order to carry out the three-pillar training programme.
Inadequate physical space and electricity. According to the participants and our own
observations all seven facilities, and specifically their maternity units, had inadequate physical
space to meet the needs. Hence, this contradicted the training. One challenge that negatively
influenced the training using mannequins was that the mannequins were packed in bags and
stored, and were only picked up for each training session. According to the participants this
was due to a lack of space, and the fact that there was no separate table for the mannequins.
This resulted in the training not always being conducted as they had learned during their train-
ing. Another reason for storing the training products was a fear that the material would be
stolen:
At the labour ward there’s no place; we’ve notified the staff that they can use the material on
the day. We have no room to store it openly and guard it. We have many interns, the African
culture–the material can be stolen. (FGD 2 with master trainers)
No access to, or insufficient availability of, electricity was another challenge that limited the
possibilities to conduct care during labour and birth based on the staff’s new knowledge.
Often, electricity was only available for six to eight hours or even less, but there could also be
two days with no electricity at all. This led to ambiguity about using electric equipment, as it
was almost impossible to rely on equipment like the blood refrigerator, resuscitation equip-
ment, and heating lamps for newborns, which depended on electricity. Solar panels were com-
monly used, but were often insufficient. At one facility, the blood refrigerator used all the
electric capacity. Thus, it was motivating to take part in this training programme and use the
assigned equipment that did not require electricity:
Electricity comes and goes; we cannot say how often. We have solar cells but they’re not strong
enough to drive medical equipment, as it needs electricity. (FGD 1, Healthcare Facility 6)
Lack of equipment to promote physiologic birth. Unanimously, all participants men-
tioned the lack of childbirth care equipment at the healthcare facilities as very limiting to the
provision of high-quality maternal and newborn healthcare. There was especially a lack of
equipment for promoting normal physiologic birth. Thus, the birthing balls and a specially
designed sheet for the use of Rebozo techniques were specifically valued, as this made it possi-
ble to offer alternative pain relief and positions during labour and birth.
The material provided through the project was used often, and all participants stressed a
need for additional birthing balls and Rebozo sheets. At some facilities, the balls and sheets
had been used so often they were now broken/torn, while other facilities were in need of extra
equipment so that it could be used if several births were taking place at the same time. The
importance of cleaning the equipment between women also created a need for extra balls and
sheets:
PLOS ONE | https://doi.org/10.1371/journal.pone.0260153 November 29, 2021
9 / 14
PLOS ONE Contextual factors influencing a training intervention in the Democratic Republic of Congo
I participated in a one-week course last autumn and learned about alternative pain relief
such as Rebozo and birthing balls, but couldn’t practise it as we didn’t a ball like that. (FGD
1, Healthcare Facility 2)
If we have two women at the same time we can’t offer both women a birthing ball. We want
to have a ball for all women giving birth; we have about three deliveries a day. We also need
more Rebozo sheets. We’d like to have three Rebozos. (FGD 2, Healthcare Facility 4)
Discussion
The study identified three contextual factors which influenced the three-pillar training inter-
vention programme aimed at improving healthcare practice during labour and birth: (i) Incen-
tives motivated participants’ efforts to begin the training programme; (ii) Involving the local
health authorities was important; and (iii) Having physical space, electricity, and equipment in
place was crucial. A central feature in these identified factors is the implication of incentives,
which we will discuss further below.
The participants expected to be monetarily reimbursed as an add-on to their monthly sal-
ary. They stressed that participation in training and projects of different kinds mostly implies
being paid, and thus expected to be incentivised in our training programme as well. Only the
master trainers were reimbursed, and they judged the compensation level to be too low. This
may be explained by the fact, found in another study in the DRC, that healthcare professionals
in the DRC, especially nurses and midwives, often lack regular payment or compensation for
their employment [24].
Barriers related to not receiving monetary incentives are not unique in quality improve-
ment interventions. The issue has been identified in other low- and middle-income country
healthcare projects in which community health workers have been used to increase the possi-
bility to achieve the goals; there, payment positively influenced the health workers’ motivation
to contribute [25,26]. The possibility of monetary incentives to increase professionals’ willing-
ness to participate was also observed in a training intervention as part of a neonatal health
project in Vietnam [27].
A main finding in our study was that the local health authorities also expected to be paid if
they were to facilitate and encourage the training programme, even though they were not
involved as trainers. In a recent Cochrane review, the involvement of local leaders in quality
improvement activities was found to be effective in implementing evidence-based practice.
The report stresses the importance of engaging healthcare leaders in interventions aimed at
improved health outcomes [28]. Similar findings have been reported in a review article about
middle managers’ role in healthcare evidence-based practice implementation [29]. Our train-
ing programme was introduced and accepted by the local authorities both in the health zone
and at the involved healthcare facilities, but without their being paid. At one of the participat-
ing facilities it was clearly observed that this led to a hindrance of the programme’s activities
and effects. As a consequence, this barrier negatively affected the expected improved maternal
and newborn health outcomes. These findings are rather disappointing.
The issue of providing or not providing performance-based financing, and its effect on
ensuring the delivery of high-quality health services, has been studied in the DRC, where
almost 500 health workers representing five of the 26 provinces participated. Workers who
had received monetary incentives which had then been stopped when the project ended scored
significantly lower on most dimensions of motivation than did those who had never received
money. The study highlights the potentially negative effect on health workers’ motivation
when large donor-driven projects provide generous incentives for participation in training,
PLOS ONE | https://doi.org/10.1371/journal.pone.0260153 November 29, 2021
10 / 14
PLOS ONE Contextual factors influencing a training intervention in the Democratic Republic of Congo
meetings, and workshops which are then withdrawn after the project-based activities are final-
ised [30].
That large donors’ provision of performance-based financing to health workers and leaders
also influenced our training intervention was obvious, as this had created ‘norms’ among
healthcare providers and leaders that such interventions should be paid for and, if not paid,
they would not participate or could even create a barrier. Thus academic research-based proj-
ects, which usually have smaller budgets to work with, cannot compete with large donor
organisations.
Another, highly positive, contextual factor and effect of our three-pillar training programme
was that the participants–both the master trainers, facilitators, and healthcare professionals–recog-
nised its value for increasing knowledge. A strong preference for learning evidence-based knowl-
edge and skills was shown among the participants; not only to have refresher trainings, but also to
learn new things–which enabled them to better conduct high-quality care during labour and
birth. That continuous education and training serve as important motivation for healthcare pro-
fessionals has been observed elsewhere [31]. Meanwhile, non-monetary incentives, in terms of
level of decision-making among community workers in low- and middle-income countries, have
been shown to be highly effective in increasing intrinsic motivation [32].
Another positive effect of the training programme was that it was immediately transferred
to care practice. Positive changes in caring, such as being closer and using alternative methods
such as birthing ball, had been experienced by the women who in turn had informed their
peers who then wanted to give birth at the same facility.
This is an example that through low-dose high frequency training it is possible to immedi-
ately implement a more woman-centred respectful intrapartum care [17,33,34], and which is
instantly told to society by women being cared.
When it comes to successfully implementing the use of HBS and HMS training pro-
grammes, which constituted pillar two of our three-pillar training programme, it has been con-
cluded that a successful implementation requires country-led commitment, readiness, and
follow-up to create local accountability and ownership [35]. Unfortunately, our study did not
fully involve the health zone authority, which resulted in a lack of ownership. These findings
in our study have offered novel insight regarding contextual factors of incentivising authorities
if a training intervention aims to have their full involvement. Comparing the findings with
those of other studies on maternal and neonatal health improvement in a low-income setting
confirms the need to account for involving local authorities from various levels of the health
system from the planning phase, through budgeting, and throughout the implementation and
evaluation processes [27,36]. This strategy of involvement may reduce the risk of facing hin-
drance in terms of monetary incentives for evidence-based interventions, which are proven to
have an impact on the health outcomes of mothers and newborns.
Another motivating incentive of the programme was that it provided the healthcare facili-
ties with material–both mannequins for training purposes as well as items to use during labour
and birth, such as birthing balls and a specially designed sheet for the use of Rebozo tech-
niques. The use of these materials in combination with better humanised behaviour towards
the women, which was stressed in Pillar 1, gave the healthcare facilities a more positive reputa-
tion through women’s sharing of positive information with peers, and also seemed to reduce
the negative trend of pregnant women’s delay in going to the facilities.
Methodological considerations
Our study is among the first to show evidence of the influence of contextual factors in a train-
ing intervention aimed at improving intrapartum care in the DRC. However, the study has its
PLOS ONE | https://doi.org/10.1371/journal.pone.0260153 November 29, 2021
11 / 14
PLOS ONE Contextual factors influencing a training intervention in the Democratic Republic of Congo
limitations. It was carried out at only seven of 40 healthcare facilities, and in only one health
zone in the DRC; this was due to limited resources. It cannot be assumed that the contextual
factors influencing such an intervention in the other 33 facilities in the chosen health zone, as
well as elsewhere in the DRC, would be the same. Therefore, the results may not be generalised
to the entire health zone, nor the country. Another limitation is that only private healthcare
facilities, and no governmental ones, were included. This may have caused us to miss other
contextual factors influencing such a training intervention.
A strength of the study is the interdisciplinary mix of the participating researchers. M Bog-
ren and M Berg, from Sweden, both hold the degrees of PhD, RM, and RN, are conducting
research in low-income settings including the DRC, and have extensive experience working in
low-income countries through multilateral organisations (M Bogren) and private health sys-
tems, including in the DRC (M Berg). The third author, S N. Mwambali, is a Congolese gynae-
cologist, and was the local project leader of the training intervention.
Conclusions
To conclude, this study found what also has been found earlier, that aspects of the context
influence the implementation of an intervention and its outcomes, and hence its feasibility
and usefulness [12]. That incentives are a critical element of successful health interventions in
impacting sexual, reproductive, maternal, and newborn healthcare quality in low- and middle-
income countries is confirmed in a recent systematic review [37]. A critical lesson learned
from this study in the DRC is that incentives of various aspects are crucial contextual factors to
consider when planning for a training intervention. In a small research project like ours, fully
meeting the expectations of monetary incentives would require a reduction of the scale and
thus limit the implementation of new evidence-based knowledge into practice.
Supporting information
S1 Appendix.
(DOCX)
Acknowledgments
We would like to express our sincere appreciation to all the healthcare providers who partici-
pated in this study. We also want to thank Dr. Prof. Denis Mukwege, Panzi Hospital, Susheela
M Engelbrecht at Jhpiego, Maria Hogena¨s, Art of Life, and Marthe Byamungu Makundane for
their respective contribution in the training programme.
Author Contributions
Conceptualization: Malin Bogren, Marie Berg.
Data curation: Malin Bogren, Marie Berg.
Formal analysis: Malin Bogren, Marie Berg.
Funding acquisition: Malin Bogren, Marie Berg.
Investigation: Malin Bogren, Marie Berg.
Methodology: Malin Bogren, Marie Berg.
Project administration: Malin Bogren, Sylvie Nabintu Mwambali.
Supervision: Malin Bogren, Marie Berg.
PLOS ONE | https://doi.org/10.1371/journal.pone.0260153 November 29, 2021
12 / 14
PLOS ONE Contextual factors influencing a training intervention in the Democratic Republic of Congo
Validation: Malin Bogren, Marie Berg.
Writing – original draft: Malin Bogren, Marie Berg.
Writing – review & editing: Malin Bogren, Sylvie Nabintu Mwambali, Marie Berg.
References
1. Ministère du Plan et Suivi de la Mise en œuvre de la Re´ volution de la Modernite´ (MPSMRM)
MdlSPMaII. Democratic Republic of Congo Demographic and Health Survey 2013–14: Key Findings.
Rockville, Maryland, USA: MPSMRM, MSP et ICF International; 2014.
2. Mpunga Mukendi D, Chenge F, Mapatano MA, Criel B, Wembodinga G. Distribution and quality of
emergency obstetric care service delivery in the Democratic Republic of the Congo: it is time to improve
regulatory mechanisms. Reproductive Health. 2019; 16(1):102. https://doi.org/10.1186/s12978-019-
0772-z PMID: 31307497
3. Kruk ME, Gage AD, Arsenault C, Jordan K, Leslie HH, Roder-DeWan S, et al. High-quality health sys-
tems in the Sustainable Development Goals era: time for a revolution. The Lancet Global Health. 2018;
6(11):e1196–e252. https://doi.org/10.1016/S2214-109X(18)30386-3 PMID: 30196093
4. United Nations. Transforming our world: The 2030 agenda for sustainable development New York,
USA; 2015.
5. Ntambue AM, Malonga FK, Dramaix-Wilmet M, Ngatu RN, Donnen P. Better than nothing? maternal,
newborn, and child health services and perinatal mortality, Lubumbashi, democratic republic of the
Congo: a cohort study. BMC Pregnancy Childbirth. 2016; 16:89. https://doi.org/10.1186/s12884-016-
0879-y PMID: 27118184
6. Ntambue AM, Malonga FK, Cowgill KD, Dramaix-Wilmet M, Donnen P. Incidence of catastrophic
expenditures linked to obstetric and neonatal care at 92 facilities in Lubumbashi, Democratic Republic
of the Congo, 2015. BMC Public Health. 2019; 19(1):948. https://doi.org/10.1186/s12889-019-7260-9
PMID: 31307419
7. Cowgill KD, Ntambue AM. Hospital detention of mothers and their infants at a large provincial hospital:
a mixed-methods descriptive case study, Lubumbashi, Democratic Republic of the Congo. Reprod
Health. 2019; 16(1):111. https://doi.org/10.1186/s12978-019-0777-7 PMID: 31331396
Tunc¸alp O¨ , Were W, MacLennan C, Oladapo O, Gu¨lmezoglu A, Bahl R, et al. Quality of care for preg-
nant women and newborns—the WHO vision. 2015; 122(8):1045–9.
8.
9. McCormack B, Kitson A, Harvey G, Rycroft-Malone J, Titchen A, Seers K. Getting evidence into prac-
tice: the meaning of ’context’. Journal of advanced nursing. 2002; 38(1):94–104. https://doi.org/10.
1046/j.1365-2648.2002.02150.x PMID: 11895535
10. Miller S, Abalos E, Chamillard M, Ciapponi A, Colaci D, Comande´ D, et al. Beyond too little, too late and
too much, too soon: a pathway towards evidence-based, respectful maternity care worldwide. The Lan-
cet. 2016; 388(10056):2176–92. https://doi.org/10.1016/S0140-6736(16)31472-6 PMID: 27642019
11. Rowe AK, Rowe SY, Peters DH, Holloway KA, Chalker J, Ross-Degnan D. Effectiveness of strategies
to improve health-care provider practices in low-income and middle-income countries: a systematic
review. The Lancet Global health. 2018; 6(11):e1163–e75. https://doi.org/10.1016/S2214-109X(18)
30398-X PMID: 30309799
12. Moore GF, Audrey S, Barker M, Bond L, Bonell C, Hardeman W, et al. Process evaluation of complex
interventions: Medical Research Council guidance. BMJ: British Medical Journal. 2015; 350. https://doi.
org/10.1136/bmj.h1258 PMID: 25791983
13. Polit D, Beck C. Nursing Research- Generating and Assessing Evidence for Nursing Practice. Philadel-
phia: Lippincott Williams & Wilkins; 2012.
14. WHO. Improving health system efficiency: Democratic Republic of Congo: improving aid coordination in
the health sector. Financing HSG; 2015.
15. Sayer A. Christian Smith, What is a Person? Journal of Critical Realism. 2012; 11(1):126–32.
16. Kitson A, Marshall A, Bassett K, Zeitz K. What are the core elements of patient-centred care? A narra-
tive review and synthesis of the literature from health policy, medicine and nursing. Journal of advanced
nursing. 2013; 69(1):4–15. https://doi.org/10.1111/j.1365-2648.2012.06064.x PMID: 22709336
17. Berg M, Asta Olafsdottir O, Lundgren I. A midwifery model of woman-centred childbirth care—in Swed-
ish and Icelandic settings. Sexual & reproductive healthcare: official journal of the Swedish Association
of Midwives. 2012; 3(2):79–87. https://doi.org/10.1016/j.srhc.2012.03.001 PMID: 22578755
18.
Laerdal Global Health. Laerdal Global Health Products Stavanger2021 [Available from: https://
laerdalglobalhealth.com/products/.
PLOS ONE | https://doi.org/10.1371/journal.pone.0260153 November 29, 2021
13 / 14
PLOS ONE Contextual factors influencing a training intervention in the Democratic Republic of Congo
19. Cohen SR, Thomas CR. Rebozo Technique for Fetal Malposition in Labor. 2015; 60(4):445–51.
20. Bogren M, Denovan A, Kent F, Berg M, Linden K. Impact of the Helping Mothers Survive Bleeding After
Birth learning programme on care provider skills and maternal health outcomes in low-income countries
—An integrative review. Women and birth: journal of the Australian College of Midwives. 2020.
21. Evans CL, Bazant E, Atukunda I, Williams E, Niermeyer S, Hiner C, et al. Peer-assisted learning after
onsite, low-dose, high-frequency training and practice on simulators to prevent and treat postpartum
hemorrhage and neonatal asphyxia: A pragmatic trial in 12 districts in Uganda. PLoS One. 2018; 13
(12):e0207909. https://doi.org/10.1371/journal.pone.0207909 PMID: 30557350
22. Severinsson E, Johansson I, Lindquist I. Effects of process-oriented group supervision—a comparison
of three groups of student nurses. Journal of nursing management. 2014; 22(4):443–51. https://doi.org/
10.1111/j.1365-2834.2012.01463.x PMID: 23409832
23. Elo S, Kyngas H. The qualitative content analysis process. Journal of advanced nursing. 2008; 62
(1):107–15. https://doi.org/10.1111/j.1365-2648.2007.04569.x PMID: 18352969
24. Bogren M, Grahn M, Kaboru BB, Berg M. Midwives’ challenges and factors that motivate them to
remain in their workplace in the Democratic Republic of Congo-an interview study. Human resources
for health. 2020; 18(1):65–. https://doi.org/10.1186/s12960-020-00510-x PMID: 32943067
25. Singh D, Negin J, Otim M, Orach CG, Cumming R. The effect of payment and incentives on motivation
and focus of community health workers: five case studies from low- and middle-income countries. Hum
Resour Health. 2015; 13:58. https://doi.org/10.1186/s12960-015-0051-1 PMID: 26169179
26. Ormel H, Kok M, Kane S, Ahmed R, Chikaphupha K, Rashid SF, et al. Salaried and voluntary commu-
nity health workers: exploring how incentives and expectation gaps influence motivation. Human
resources for health. 2019; 17(1):59. https://doi.org/10.1186/s12960-019-0387-z PMID: 31324192
27. Duong DM, Bergstro¨ m A, Wallin L, Bui HT, Eriksson L, Eldh AC. Exploring the influence of context in a
community-based facilitation intervention focusing on neonatal health and survival in Vietnam: a qualita-
tive study. BMC Public Health. 2015; 15:814. https://doi.org/10.1186/s12889-015-2142-2 PMID:
26297314
28.
Flodgren G, ’Brien O MA, Parmelli E, Grimshaw JM. Local opinion leaders: effects on professional prac-
tice and healthcare outcomes. The Cochrane database of systematic reviews. 2019; 6(6):Cd000125.
https://doi.org/10.1002/14651858.CD000125.pub5 PMID: 31232458
29. Birken S, Clary A, Tabriz AA, Turner K, Meza R, Zizzi A, et al. Middle managers’ role in implementing
evidence-based practices in healthcare: a systematic review. Implementation science: IS. 2018; 13
(1):149. https://doi.org/10.1186/s13012-018-0843-5 PMID: 30541596
30. Maini R, Lohmann J, Hotchkiss DR, Mounier-Jack S, Borghi J. What Happens When Donors Pull Out?
Examining Differences in Motivation Between Health Workers Who Recently Had Performance-Based
Financing (PBF) Withdrawn With Workers Who Never Received PBF in the Democratic Republic of
Congo. Int J Health Policy Manag. 2019; 8(11):646–61. https://doi.org/10.15171/ijhpm.2019.55 PMID:
31779290
31.
Li SA, Jeffs L, Barwick M, Stevens B. Organizational contextual features that influence the implementa-
tion of evidence-based practices across healthcare settings: a systematic integrative review. System-
atic reviews. 2018; 7(1):72. https://doi.org/10.1186/s13643-018-0734-5 PMID: 29729669
32. Kok MC, Kane SS, Tulloch O, Ormel H, Theobald S, Dieleman M, et al. How does context influence per-
formance of community health workers in low- and middle-income countries? Evidence from the litera-
ture. Health Research Policy and Systems. 2015; 13(1):13. https://doi.org/10.1186/s12961-015-0001-3
PMID: 25890229
33. WHO. WHO recommendations Intrapartum care for a positive childbirth experience. Geneva; 2018.
34. Bohren MA, Tunc¸alp O¨ , Miller S. Transforming intrapartum care: Respectful maternity care. Best Prac-
tice & Research Clinical Obstetrics & Gynaecology. 2020; 67:113–26. https://doi.org/10.1016/j.
bpobgyn.2020.02.005 PMID: 32245630
35. Ersdal HL, Singhal N, Msemo G, Kc A, Data S, Moyo NT, et al. Successful implementation of Helping
Babies Survive and Helping Mothers Survive programs—An Utstein formula for newborn and maternal
survival. PLoS One. 2017; 12(6):e0178073. https://doi.org/10.1371/journal.pone.0178073 PMID:
28591145
36. Eriksson L, Bergstro¨ m A, Hoa DTP, Nga NT, Eldh AC. Sustainability of knowledge implementation in a
low- and middle- income context: Experiences from a facilitation project in Vietnam targeting maternal
and neonatal health. PLoS One. 2017; 12(8):e0182626. https://doi.org/10.1371/journal.pone.0182626
PMID: 28806744
37. Negero MG, Sibbritt D, Dawson A. How can human resources for health interventions contribute to sex-
ual, reproductive, maternal, and newborn healthcare quality across the continuum in low- and lower-
middle-income countries? A systematic review. Human Resources for Health. 2021; 19(1):54. https://
doi.org/10.1186/s12960-021-00601-3 PMID: 33882968
PLOS ONE | https://doi.org/10.1371/journal.pone.0260153 November 29, 2021
14 / 14
PLOS ONE |
10.1371_journal.pone.0258709 | RESEARCH ARTICLE
Spatiotemporal inflection points in human
running: Effects of training level and athletic
modality
Yuta Goto1, Tetsuya Ogawa2, Gaku Kakehata1, Naoya Sazuka3, Atsushi Okubo4,
Yoshihiro Wakita4, Shigeo Iso5, Kazuyuki Kanosue5*
1 Graduate School of Sport Sciences, Waseda University, Saitama, Japan, 2 Department of Clothing,
Faculty of Human Sciences and Design, Women’s University Tokyo, Japan, 3 Tokyo Laboratory 25, R&D
Center, Sony Group Corporation, Tokyo, Japan, 4 Tokyo Laboratory 07, R&D Center, Sony Group
Corporation, Tokyo, Japan, 5 Faculty of Sport Sciences, Waseda University, Saitama, Japan
* kanosue@waseda.jp
Abstract
The effect of the different training regimes and histories on the spatiotemporal characteris-
tics of human running was evaluated in four groups of subjects who had different histories of
engagement in running-specific training; sprinters, distance runners, active athletes, and
sedentary individuals. Subjects ran at a variety of velocities, ranging from slowest to fastest,
over 30 trials in a random order. Group averages of maximal running velocities, ranked from
fastest to slowest, were: sprinters, distance runners, active athletes, and sedentary individu-
als. The velocity-cadence-step length (V-C-S) relationship, made by plotting step length
against cadence at each velocity tested, was analyzed with the segmented regression
method, utilizing two regression lines. In all subject groups, there was a critical velocity,
defined as the inflection point, in the relationship. In the velocity ranges below and above the
inflection point (slower and faster velocity ranges), velocity was modulated primarily by alter-
ing step length and by altering cadence, respectively. This pattern was commonly observed
in all four groups, not only in sprinters and distance runners, as has already been reported,
but also in active athletes and sedentary individuals. This pattern may reflect an energy sav-
ing strategy. When the data from all groups were combined, there were significant correla-
tions between maximal running velocity and both running velocity and step length at the
inflection point. In spite of the wide variety of athletic experience of the subjects, as well as
their maximum running velocities, the inflection point appeared at a similar cadence (3.0 ±
0.2 steps/s) and at a similar relative velocity (65–70%Vmax). These results imply that the
influence of running-specific training on the inflection point is minimal.
Introduction
Human running has been studied extensively from the viewpoint of how its temporal
(cadence) and spatial (step length) components contribute to velocity [1–10]. Velocity equals
the product of cadence and step length, and the relative contribution of each component to
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: Goto Y, Ogawa T, Kakehata G, Sazuka N,
Okubo A, Wakita Y, et al. (2021) Spatiotemporal
inflection points in human running: Effects of
training level and athletic modality. PLoS ONE
16(10): e0258709. https://doi.org/10.1371/journal.
pone.0258709
Editor: Leonardo A. Peyre´-Tartaruga, Universidade
Federal do Rio Grande do Sul, BRAZIL
Received: February 11, 2021
Accepted: October 4, 2021
Published: October 18, 2021
Peer Review History: PLOS recognizes the
benefits of transparency in the peer review
process; therefore, we enable the publication of
all of the content of peer review and author
responses alongside final, published articles. The
editorial history of this article is available here:
https://doi.org/10.1371/journal.pone.0258709
Copyright: © 2021 Goto et al. This is an open
access article distributed under the terms of the
Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: All relevant data are
within the manuscript and its Supporting
Information files.
PLOS ONE | https://doi.org/10.1371/journal.pone.0258709 October 18, 2021
1 / 12
PLOS ONE Funding: This work was supported by Japan
Society for the Promotion of Science (JSPS),
KAKENHI Grant Number 19K22822 (K.K) and by
Grant-in-Aid for JSPS Fellows Number 20J11122
(Y.G) from Ministry of Education, Culture, Sports,
Science and Technology of Japan. Sony Group
Corporation provided support in the form of
salaries for authors [NS, AO, and YW], but did not
have any additional role in the study design, data
collection and analysis, decision to publish, or
preparation of the manuscript. The specific roles of
these authors are articulated in the ‘author
contributions’ section. Sony Group Corporation has
a patent (US20180039751A1) on apparatuses for
helping runners modify the V-C-S property. This
patent does not interfere with the usage of any data
or knowledge presented in the paper.
Competing interests: Sony Group Corporation
provided support in the form of salaries for authors
[NS, AO, and YW].Sony Group Corporation does
not alter the adherence to PLOS ONE policies on
sharing data and materials presented in this paper.
Spatiotemporal inflection points in human running
changing velocity differs across the velocity range. A previous study reported that, at slower
velocities, speed is modulated primarily by adjusting step length, whereas, at faster velocities,
speed is modulated more by changes in cadence [6]. At velocities close to maximum, step
length shows only a small increase or even a decrease as running velocity approaches the maxi-
mum [8]. These characteristics are considered to indicate the spontaneous recruitment of an
adequate motor pattern which minimizes energy expenditure at a given running velocity [5,
11–13]. Mechanical approaches, such as Fenn’s approach, have been used as useful tools to elu-
cidate these energy cost determinants with many practical applications [14]. Yanai and Hay
[12], utilizing a two-dimensional simulation, evaluated the relative contribution of cadence
and step length in the optimization of power production utilizing both anatomical (range of
motion in the hip joint) and spatiotemporal (duration of the stance phase) determinants.
Indeed, if the cadence is voluntarily modified from that occurring under the natural move-
ment pattern at a given running velocity, metabolic rate is lowest when the cadence is in the
range of ±10% of the preferred cadence [15–18]. In addition, in the slower velocity range,
Cavagna et al. [19] reported that preferred cadences take place in the proximity of 3 Hz.
However, the extent to which the above characteristics occur in different populations and
in persons with different physical backgrounds remains unclear. Most of the above-mentioned
studies focused on well-trained individuals, especially those trained for running [7–9, 12, 20].
Therefore, the purpose of the present study was to investigate how a change in running
velocity altered the spatiotemporal adjustment between step length and cadence in subjects
with different histories of engagement in running training. Namely, we studied:1. sprinters, 2.
distance runners, 3. active athletes who had received no running-specific training, and 4. sed-
entary, untrained subjects. The relationships among running velocity, cadence, and step length
over a wide range of running velocities were compared across these subjects. Among the four
groups, the distance runners would be expected to run as efficiently (either mechanically or
metabolically) as possible. As noted above, in the slower velocity ranges, altering stride length
is a more energy saving strategy for changing velocity than is altering cadence [12]. Therefore,
we hypothesized: 1. the running step length/cadence patterns of individuals would be influ-
enced by their running training experience and overall physical activity levels and 2. distance
runners would exhibit the greatest tendency to change velocity by altering step length in the
slower velocity range.
Methods
Subjects
A total of eighty volunteers (69 males and 11 females) with different backgrounds, in terms of
their running experience, participated in the study. They were assigned into one of four groups
depending on their current/previous running training. We utilized four groups of subjects
with different histories of running training. The first and second groups consisted of twenty
sprinters (all men) and twenty distance runners (all men), respectively. The participants in the
third group were twenty active athletes (16 males and 4 females). Although running is involved
in many of the sports, all subjects informed us that they had received no special training for
improving their running speed. For reference, the sports that the participants in the third
group engaged in were: soccer, basketball, softball, weightlifting, boxing, lacrosse, volleyball,
American football, badminton, handball, rowing, judo, and golf. They had all participated in
their sport for at least 5 years. The fourth group consisted of sedentary individuals without a
history of any regular participation in sports activities (13 males and 7 females). Table 1 lists
the characteristics of participants in each group. All participants were informed of the pur-
poses and procedures, and signed an informed consent form. This study was approved by the
PLOS ONE | https://doi.org/10.1371/journal.pone.0258709 October 18, 2021
2 / 12
PLOS ONE Spatiotemporal inflection points in human running
Table 1. Physical characteristics and sport activity history of each subject group.
Sprinters
Distance runners
Active athletes
Sedentary individuals
N
20
20
20
17
age, years
height, cm
sports activity history, years
22 ± 2
20 ± 1
23 ± 2
22 ± 2
176.2 ± 6.1b, c, d
171.0 ± 4.5
170.1 ± 5.8
166.0 ± 6.2
9.7 ± 3.0
7.4 ± 2.0
10.2 ± 4.4
Values are means ± SD. N, number of subjects. b, c, d: values are significantly different from distance runners, active athletes, and sedentary individuals, respectively
(p < 0.05). The sport activity history of the active athletes indicates the number of years of participation in that sport for each subject.
https://doi.org/10.1371/journal.pone.0258709.t001
Human Research Ethics Committee in Faculty of Sport Sciences, Waseda University. The
experiments were conducted in accordance with the Declaration of Helsinki.
Experimental setup and tasks
Experiments were conducted on a 30 m all-weather straight track (only 20 m for the sedentary
group in consideration of their physical strength and lack of stamina) on which color markers
were placed every 0.5 m for video analysis. A sagittal view of each participant was recorded by
panning with a video camera (HDR-CX630V, SONY) placed approximately 10 m lateral to the
center of the running path. An additional 10–30 meters was provided before and after the filming
zone (of 30m or 20m) so that the subjects could accelerate and decelerate and thus maintain run-
ning velocity as constant as possible throughout the recording area. This acceleration distance
differed between trials and was selected by the subject. The video sampling frequency was 60 Hz.
Participants were asked to run along the path 30 times at a variety of velocities, which varied
from slow to the fastest possible. The order of running with different velocities was random-
ized on a subject-by-subject basis. The subjects were directed to run at a particular percentage
of their maximal effort [21]. This instruction included requesting a subjective effort from 10%
to 100% of maximum, as well as “run faster or slower than the previous trial”. The actual run-
ning speed did not necessarily match the exact percentage of their maximal speed. However,
this method did produce the necessary array of running speeds and the subjects might run
more than once at an intensity. When running at the minimum velocity, subjects followed our
instruction to run as slowly as they could while still maintaining a running gait (as opposed to
walking, jumping, hopping, or bounding). The interval between trials ranged from 30 seconds
to 5 minutes, depending on the speed of the previous trial. A 5-minute rest was taken after 15
trials. The participants used their own running shoes. Spiked shoes were not allowed.
Data analysis
Offline data analysis was performed by using video administration software (PlayMemories,
SONY, Japan). On the basis of the video analysis, the running velocity, cadence, and step length
were calculated on a trial-by-trial basis for each subject. Mean running velocity (m/sec) was calcu-
lated by dividing the length of the path (m) by the time taken (sec) to run over the path. The
instant at which the subject passed the start and the end point were identified from the position
of the chest relative to the color markers. Mean cadence (steps/sec) was calculated by dividing the
number of steps by the time taken to cover that distance. The number of steps was counted from
the first ground contact with the path to the last ground contact before passing the end point. The
duration utilized was defined as the time between the instant of first foot-contact after the start
position and that of the last foot-contact before the end. Mean step length (m) was calculated by
dividing the mean running velocity (m/sec) by the mean cadence (steps/sec). Step length was also
PLOS ONE | https://doi.org/10.1371/journal.pone.0258709 October 18, 2021
3 / 12
PLOS ONE Spatiotemporal inflection points in human running
expressed as the ratio of the step length (m) to the height (m) of each subject in order to examine
the influence of the physical characteristics of the subjects. For the running velocity, the fastest
among the 30 trials by each subject was designated as their maximal running velocity.
In the present study, the principal analyses for the spatiotemporal running characteristics
of each subject were performed with MATLAB version R2018a (The MathWorks, Inc., USA).
For each subject, the data were plotted as shown in Fig 1 in order to examine the relationship
between cadence and step length (horizontal axis: cadence, vertical axis: step length). This cor-
respondence involved the Velocity (m/s, dotted line), Cadence (steps/s, horizontal), and Step
length (m, vertical), and is defined as the V-C-S relationship. To quantitatively analyze the crit-
ical point at which the relative contribution of spatiotemporal adjustment changed (cadence
vs. step length), we utilized the segmented regression method which has previously been used
to detect lactate threshold [22] and ventilation threshold [23] during aerobic exercise. This is a
statistical method for determining the point at which a line suddenly changes slope at some
unknown point. We used a segmented regression procedure [23, 24] in which the N data
points were divided into two segments (the lower x data and the upper N-x data, x = 3, 4, . . .,
or N-2). Each segment was fitted with a regression line using the Deming regression [25, 26].
This regression method was adopted to exclude the effects of measurement errors in cadence
and step length. That is, one regression line was obtained with x data points from the ascend-
ing order starting with the minimum velocity, and the other one with N-x data points from the
descending order starting with the maximum velocity. The critical point (“inflection point”),
then, was the intersection of the two regression lines with an x value that minimized orthogo-
nal distance between measurement data and regression line for two data sets (segments) (Fig
1, cross; X). We assumed that the regression lines below and above the inflection point would
adequately represent the spatiotemporal characteristics of running for each subject and group.
Subjects with inflection points, thus obtained, that differed largely from the measured
points, were excluded from the analysis (#18, #19, and #20, as seen in S4 Fig).
Therefore, the final analysis involved 20 sprinters, 20 distance runners, 20 active athletes,
and 17 sedentary individuals. For these subjects, running velocity, cadence, and step length at
Fig 1. The relationship between cadence (steps/s, horizontal) and step length (m, vertical) relative to running
velocity (pale broken line and the second vertical axis) in a single sprinter. The inflection point (cross) was
computed from two regression lines from different data sets by combining the segmented regression method of
Deming regression. The filled and open circle markers represent the data sets below and above the inflection point at
which the relationship between cadence and step length changed abruptly. Inflection point was obtained as the
intersection point of the two regression lines.
https://doi.org/10.1371/journal.pone.0258709.g001
PLOS ONE | https://doi.org/10.1371/journal.pone.0258709 October 18, 2021
4 / 12
PLOS ONE Spatiotemporal inflection points in human running
the inflection point were calculated. Normalized values were determined for each parameter at
the maximal running velocity.
Statistical analysis
Statistical analysis was performed using SPSS Statistics 23 software (IBM, USA). Maximal run-
ning velocity, height of subjects, and all variables related to inflection point in each group were
tested for a normal distribution using the Shapiro-Wilk test. Maximal running velocity, height,
and normalized cadence at the inflection point were found to have non-normal distributions.
Thus, group mean data for maximal running velocity, height of subjects, and all variables related
to inflection point were analyzed among the four subject groups by using a non-parametric Krus-
kal-Wallis test. Next, post-hoc pairwise comparisons using the Dunn-Bonferroni approach were
made to identify additional differences between the groups. In order to further investigate the
possible mechanisms responsible for the inflection point, correlational analyses were performed.
All variables across all subjects related to the inflection point and maximal running velocity
were tested for a normal distribution using the Shapiro-Wilk test. Maximum running velocity,
and step length at maximal running velocity exhibited normal distributions. Likewise, running
velocity (both unnormalized and normalized), step length (both unnormalized and normal-
ized), and unnormalized cadence at the inflection point exhibited normal distributions. How-
ever, cadence at maximal running velocity and normalized cadence at the inflection point
exhibited non-normal distributions. Pearson’s and Spearman’s correlations were performed to
analyze the relationship between maximal running velocity and other parameters at the inflec-
tion point. Significance was set at p < 0.05. The data are presented as mean and standard devi-
ation (mean ± SD).
Results
Fig 1 shows a typical example of the relationship between running velocity, cadence, and step
length for a single sprinter. Both cadence and step length show specific changes in relation to
changing running velocity. The inflection point (cadence: 2.97 steps/s, step length: 1.78 m) was
computed from two regression lines.
Fig 2A shows an inter-group comparison of the mean values of Vmax. A Kruskal-Wallis
test revealed significant differences between the groups in terms of maximum running velocity
(χ2 (3) = 52.463, p < 0.001). The post-hoc comparisons revealed that the maximal velocity of
the sprinters was faster compared to all the other subject groups (distance runner: p = 0.009,
active athlete: p < 0.001, sedentary: p < 0.001). The distance runner group exhibited signifi-
cantly faster maximal running velocity in comparison with the sedentary individual group. Fig
2B–2D illustrates the correlation between maximal running velocity and cadence, absolute
step length and step length normalized to height at the maximal running velocity. There were
significant positive correlations between Vmax and cadence as well as step length both in the
unnormalized and normalized forms (cadence: r = 0.514, p < 0.001; step length (unnorma-
lized): r = 0.843, p < 0.001; step length (normalized): r = 0.803, p < 0.001).
Fig 3A shows mean values of cadence and step length at maximal running velocity (Vmax),
the inflection point, and minimal running velocity (Vmin) for each subject group. As shown
in Fig 3A, maximal running velocity was different across the groups and was the fastest in the
sprinters (I, around 10 m/s) and slowest in the sedentary individuals (IV, mostly less than 8 m/
s). All groups tended to increase step length predominately at the velocities between Vmin
(velocity: 2.17 ± 0.45 m/s, cadence: 2.62 ± 0.14 steps/s, step length: 0.82 ± 0.17 m) and the
inflection point, and then to increase cadence until they reached Vmax. Fig 3B depicts mean
values of cadence and step length normalized to the values obtained under maximal running
PLOS ONE | https://doi.org/10.1371/journal.pone.0258709 October 18, 2021
5 / 12
PLOS ONE Spatiotemporal inflection points in human running
Fig 2. Inter-group comparison of mean values (diamond) of the maximum running velocity (Vmax) (A), and
correlation between the maximal running velocity and the cadence (B), step length (C), and step length normalized by
height (D) at maximal running velocity. In Fig 2A, open circles indicate each individual subject. Significant difference;
���p < 0.001, ��p < 0.01. In Fig 2B–2D, filled circles, open circles, filled triangles, and open triangles represent the
sprinters, distance runners, active athletes, and sedentary individuals, respectively. There are significant positive
correlations between Vmax and the cadence (B) and between Vmax and step length, both absolute velocity and velocity
normalized to maximal running velocity (r = 0.514, p < 0.001; r = 0.843, p < 0.001; r = 0.803, p < 0.001, respectively).
https://doi.org/10.1371/journal.pone.0258709.g002
velocity. The characteristics of the increase in velocity were similar to those from Fig 3A. Due
to differences in the absolute value (Fig 3A) of maximal running velocity, the normalized
cadence varied considerably across the subject groups, while variability in step length below
the inflection point was less evident.
Table 2 shows inter-group comparison of the mean values of all variables related to the
inflection point. A Kruskal-Wallis test revealed significant difference of running velocity, step
length, normalized cadence (χ2 (3) = 31.215, p < 0.001; χ2 (3) = 42.68, p < 0.001; χ2 (3) =
23.623, p < 0.001, respectively). The post-hoc comparisons revealed significant differences
between the subject groups. In the group of sprinters, the running velocity was significantly
faster as compared to the active athlete, and sedentary subject groups (active athlete: p < 0.01,
sedentary: p < 0.001). For the same parameter, the group of distance runners showed signifi-
cantly faster in comparison to the sedentary group (p < 0.01). The step length was significantly
longer in the sprinter group in comparison to all the other subject groups (distance runner:
p < 0.01, active athletes: p < 0.001, sedentary: p < 0.001). For the same parameter, the group
of distance runners was significantly longer than the sedentary group (p < 0.05). In the group
PLOS ONE | https://doi.org/10.1371/journal.pone.0258709 October 18, 2021
6 / 12
PLOS ONE Spatiotemporal inflection points in human running
Fig 3. Mean values of cadence and step length at the maximal running velocity (Vmax), inflection point (IP), and
minimal running velocity (Vmin) (A), and those with cadence and step length normalized to those under Vmax (B) for
each subject group. The error bars depict the standard deviation. The filled circles, open circles, filled triangles, and
open triangles represent sprinters, distance runners, active athletes and sedentary individuals, respectively. Pale broken
lines represent running velocity (A) and running velocity normalized by maximal running velocity (B). The thick
broken line in B illustrates the limiting situation, in which velocity change is only done with a step length change in the
velocity range below the inflection point, and only with a cadence change above the inflection point.
https://doi.org/10.1371/journal.pone.0258709.g003
of sprinters, the normalized cadence was lower as compared to distance runner and sedentary
subject groups (distance runner: p < 0.01, sedentary: p < 0.001).
Fig 4A–4C depicts correlations between maximal running velocity and running velocity,
cadence, and step length at the inflection point. There were significant positive correlations
between Vmax and both velocity and step length at the inflection point (velocity: Fig 4A,
r = 0.738, p < 0.001; step length: Fig 4C, r = 0.827, p < 0.001). Cadence at the inflection point
had no correlation with Vmax, and was approximately constant at 3.0 ± 0.2 steps/s regardless
of the subject group (Fig 4B). Fig 4D–4F illustrates correlation for the same parameters shown
in Fig 4A–4C, but with values normalized to Vmax. Velocity and cadence show negative corre-
lations (velocity: r = -0.300, p < 0.01; cadence: r = -0.621, p < 0.001), while step length has a
positive correlation with Vmax (r = 0.290, p < 0.05).
Discussion
We investigated the relative contribution of cadence and step length changes as running veloc-
ity was modulated in four groups of subjects with different histories of engagement in
Table 2. Kinematic variables at the inflection point.
Sprinters (N = 20)
Distance runners (N = 20)
Active athletes (N = 20)
Sedentary individuals (N = 17)
velocity, m/s
step length, m
cadence, steps/s
normalized velocity, %
normalized step length, %
normalized cadence, %
5.86 ± 0.59c, d
2.03 ± 0.13 b, c, d
2.88 ± 0.26
64.7 ± 7.1
96.5 ± 7.2
67.0 ± 4.7 b, d
5.36 ± 0.60 d
1.75 ± 0.14 d
3.06 ± 0.17
67.0 ± 7.5
92.2 ± 8.2
72.6 ± 4.2
5.00 ± 0.50
1.69 ± 0.18
2.97 ± 0.15
66.7 ± 4.8
94.1 ± 5.8
71.2 ± 6.6
4.50 ± 0.65
1.52 ± 0.21
2.96 ± 0.18
68.6 ± 7.4
90.3 ± 7.7
76.0 ± 6.0
Values are means ± SD. N, number of subjects. b, c, d: values are significantly larger, from distance runners, active athletes, and sedentary individuals, respectively.
Normalized velocity, step length, and cadence were obtained by normalizing with corresponding values at the maximal running velocity, respectively.
https://doi.org/10.1371/journal.pone.0258709.t002
PLOS ONE | https://doi.org/10.1371/journal.pone.0258709 October 18, 2021
7 / 12
PLOS ONE Spatiotemporal inflection points in human running
Fig 4. Correlation between maximal running velocity (Vmax) and: running velocity (A), cadence (B), and step length (C), as well as the same three parameters
normalized to the Vmax (D–F) at the inflection point. Filled and open circles, and filled and open triangles represent the sprinters, distance runners, active athletes,
and sedentary individuals, respectively. The correlations are all significant except for cadence (B).
https://doi.org/10.1371/journal.pone.0258709.g004
running-specific training, utilizing the segmented regression method with two regression lines
(Fig 1). In spite of a large variation in maximal running velocity, the general characteristics of
the V-C-S relationship were similar across the subject groups (Fig 3) as well as across the data
of individuals (S1–S4 Figs).
Basic characteristics of the V-C-S relationship
As expected, compared to the sprinters, maximal running velocities were progressively slower
in the distance runners, active athletes and sedentary groups. There were significant differ-
ences between the sprinters and the other three groups, as well as between the distance runners
and the sedentary individuals (Fig 2A). Both cadence and step length at Vmax were well corre-
lated with Vmax (Fig 2B and 2C, respectively). Among the subject groups, the sprinters were
the tallest and the sedentary group was the shortest. The strong correlation of step length with
Vmax was well-preserved, however, even when step length was normalized to the subjects’
heights (Fig 2D). Thus, faster maximum running velocities were generally accomplished with
both a higher cadence and longer steps. The minimum running velocity was common to all
subject groups at 2.17 ± 0.45 m/s with a cadence of 2.62 ± 0.14 steps/s and a step length of
PLOS ONE | https://doi.org/10.1371/journal.pone.0258709 October 18, 2021
8 / 12
PLOS ONE Spatiotemporal inflection points in human running
0.82 ± 0.17 m (Fig 3A). It appears that a slower cadence would have required “hopping” rather
than running, and for shorter step lengths it became similar to “jogging in place”.
In all four subject groups, an abrupt change in the V-C-S relationship took place at the
inflection point (Fig 3 and Table 2). Velocity changes below the inflection point occurred
mainly by modulating step length and velocity changes above the inflection point occurred
mainly via cadence modulation. These characteristics were demonstrated in preceding studies
conducted on sprinters and distance runners [7, 9], and are particularly prominent in
sprinters.
Running velocity at the inflection point has a significant positive correlation with Vmax
(Fig 4A). Thus, the faster the Vmax, the faster the velocity at the inflection point. A faster
velocity at the inflection point is mainly attained by longer step length (Fig 4C). However, this
correlation was weak when it is normalized with the step length at the Vmax (Fig 4F).
Overall, regardless of the training history, all groups had a similar relative step length quite
close to the maximum step length (about 90%). Interestingly, the cadence at the inflection
point has no correlation with Vmax and remained constant at about 3 steps/sec (Fig 4B). The
history of the training influenced normalized cadence at the inflection point, that is, sprinters
had a lower normalized cadence at the inflection point than the others, although in absolute
terms cadence was the same. In the normalized plane (Fig 3B) inflection points of the different
groups are lined along the isovelocity curve of 65–70%. Scatter plots of all subjects of all the
groups showed only a weak correlation between the Vmax and the velocity at the inflection
point normalized with Vmax (Fig 4D). In spite of the wide range of sports, and thus athletic
modality of the subjects as well as their maximum running velocity, the inflection point
appeared at a similar cadence (3.0 ± 0.2 steps/s) as well as at similar relative velocity (65–70%
Vmax), across all groups. These results imply that the influence of running-specific training on
the inflection point is minimal.
Functional meaning of the V-C-S relationship
Although the basic characteristics of the V-C-S relationship are common across different sub-
ject groups, the quantitative difference could be related to quality/quantity difference in run-
ning-specific training among groups.
In the present study, four groups of subjects, sprinters, distance runners, active athletes uti-
lizing varying degrees of running but no running training, and sedentary individuals, were
studied. Of course, the above order would also be expected for the maximal velocity from fast-
est to the slowest (Fig 2A). Sprinting and distance training involves running on a daily basis,
and running (generally without specific running instruction) forms one aspect of training for
many of the active athletes as well. It seems reasonable that some portion of the observed maxi-
mal velocities reflect differences in training.
Interestingly, step length at the inflection point also follows the same order as the maximal
velocity (Figs 3A and 4C and 4F). In the velocity range below the inflection point, velocity
change is mainly done with a change in step length; for energy-saving this is a more efficient
strategy than is changing the cadence [12]. It would be beneficial for distance runners to run
within this range as much as possible when their velocity is below the inflection point. Indeed,
it was shown that at 4.4 m/s velocity, in the range below the inflection point, the stride length
was associated with better running economy in distance runners [27]. Therefore, we had
hypothesized that the ability to run below the inflection point would be particularly developed
in distance runners. However, sprinters and not distance runners increased velocity by elon-
gating both absolute step length (Fig 4C) and relative step length (Fig 4F), all the way to the
upper running speed limit. Thus, our working hypothesis was rejected. Sprinters rarely train
PLOS ONE | https://doi.org/10.1371/journal.pone.0258709 October 18, 2021
9 / 12
PLOS ONE Spatiotemporal inflection points in human running
in the velocity range below the inflection point. Obviously, maximal velocity is crucial for
sprinters. A faster velocity cannot be accomplished only with power, especially at the highest
levels. Sprinters need to develop both power and economy to the upper limit, and inevitably
and unintentionally develop mechanically efficient movements.
Future studies
Why and by what means are there differences in the various parameters of the V-C-S relation-
ship? In particular, the neural as well as physiomechanical mechanisms of differences in the
V-C-S relationship should prove very interesting. In the future, motion analysis together with
measurements of muscle activity and ground reaction forces could help to answer our overall
question. Although numerical simulation of running and walking has many limitations [11, 12,
28], the differences in the V-C-S relationship could be analyzed with numerical models in terms
of various energy costs. Furthermore, it is very interesting that even in the sedentary subjects,
the basic pattern of V-C-S relationship, which is considered to reflect efficiency [12, 13], was
seen. Is the V-C-S pattern innate or does it develop along the development? This, and also
fatigue [29], aging [30, 31], and sex differences [32], if any, are topics that merit future analysis.
Conclusions
In the present study we analyzed the V-C-S relationship of running with the segmented regres-
sion method and made a quantitative comparison of the “spatiotemporal running characteris-
tics” in subjects with different histories of running-specific training. The common characteristic
of the V-C-S relationship is, in the slower and faster velocity ranges, that velocity is mainly mod-
ulated by altering step length and cadence, respectively. This was observed not only in the
sprinters and distance runners, as shown in previous studies, but in active (general sport) ath-
letes and sedentary subjects as well. In spite of the wide range of athletic modalities of the sub-
jects, and their maximum running velocity, the inflection point appeared at a similar cadence
(3.0 ± 0.2 steps/s) and at similar a relative velocity (65–70%Vmax), across all groups. These
results imply that the influence of running-specific training on the inflection point is minimal.
Supporting information
S1 Fig. The relationship between cadence and step length for all the sprinters. The two
dashed lines depict the regression lines computed from different data below and above the
inflection point, respectively.
(PDF)
S2 Fig. The relationship between cadence and step length for all the distance runners. The
two dashed lines show the regression lines computed from different data below and above the
inflection point, respectively.
(PDF)
S3 Fig. The relationship between cadence and step length for the active athletes. The two
dashed lines show the regression lines computed from different data below and above the
inflection point, respectively. The title of each figure corresponds to each subject’s sports expe-
rience. Characters in parentheses signify male or female subjects.
(PDF)
S4 Fig. The relationship between cadence and step length for the sedentary individuals.
The two dashed lines show the regression lines computed from different data below and above
the inflection point, respectively. In the sedentary group, three subjects were excluded from
PLOS ONE | https://doi.org/10.1371/journal.pone.0258709 October 18, 2021
10 / 12
PLOS ONE Spatiotemporal inflection points in human running
data analysis: two subjects (No. 18 and No. 19) had estimated inflection point fell outside the
range of the original data, and one subject (No. 20) showed two regression lines with almost
the same slope giving the inflection point completely outside the range of measured data.
Characters in parentheses signify male or female subjects.
(PDF)
Acknowledgments
The authors thank Dr. Larry Crawshaw for English editing of the manuscript.
Author Contributions
Conceptualization: Yuta Goto, Tetsuya Ogawa, Gaku Kakehata, Kazuyuki Kanosue.
Formal analysis: Yuta Goto, Naoya Sazuka, Yoshihiro Wakita.
Funding acquisition: Yuta Goto.
Investigation: Yuta Goto, Gaku Kakehata.
Methodology: Yuta Goto, Tetsuya Ogawa, Naoya Sazuka, Yoshihiro Wakita.
Project administration: Yuta Goto, Atsushi Okubo, Kazuyuki Kanosue.
Software: Naoya Sazuka.
Supervision: Kazuyuki Kanosue.
Visualization: Yuta Goto.
Writing – original draft: Yuta Goto, Naoya Sazuka, Yoshihiro Wakita.
Writing – review & editing: Tetsuya Ogawa, Gaku Kakehata, Atsushi Okubo, Shigeo Iso,
Kazuyuki Kanosue.
References
1. Dillman CJ. Kinematic analyses of running. Exerc Sport Sci Rev. 1975; 3:193–218. PMID: 1175666
2.
Luhtanen P, Komi P. Mechanical factors influencing running speed. In: Asmussen E, Jo¨rgensen K, edi-
tors. Biomechanics VI-B. Baltimore: University Park Press.; 1978. p. 23–9.
3. Cavagna GA, Franzetti P, Heglund NC, Willems P. The determinants of the step frequency in running,
trotting and hopping in man and other vertebrates. J Physiol. 1988; 399:81–92. https://doi.org/10.1113/
jphysiol.1988.sp017069 PMID: 3404473
4. Kaneko M. Mechanics and energetics in running with special reference to efficiency. J Biomech. 1990;
23 Suppl 1:57–63. https://doi.org/10.1016/0021-9290(90)90041-z PMID: 2081745
5. Cavagna GA, Willems PA, Franzetti P, Detrembleur C. The two power limits conditioning step fre-
quency in human running. J Physiol. 1991; 437:95–108. https://doi.org/10.1113/jphysiol.1991.
sp018586 PMID: 1890660
6. Hay JG. Cycle rate, length, and speed of progression in human locomotion. J Appl Biomech. 2002; 18
(3):257–70.
7. Weyand PG, Sternlight DB, Bellizzi MJ, Wright S. Faster top running speeds are achieved with greater
ground forces not more rapid leg movements. J Appl Physiol (1985). 2000; 89(5):1991–9.
8. Hunter JP, Marshall RN, McNair PJ. Interaction of step length and step rate during sprint running. Med
Sci Sports Exerc. 2004; 36(2):261–71. https://doi.org/10.1249/01.MSS.0000113664.15777.53 PMID:
14767249
9. Nummela A, Keranen T, Mikkelsson LO. Factors related to top running speed and economy. Int J
Sports Med. 2007; 28(8):655–61. https://doi.org/10.1055/s-2007-964896 PMID: 17549657
10. Salo AI, Bezodis IN, Batterham AM, Kerwin DG. Elite sprinting: are athletes individually step-frequency
or step-length reliant? Med Sci Sports Exerc. 2011; 43(6):1055–62. https://doi.org/10.1249/MSS.
0b013e318201f6f8 PMID: 20980924
PLOS ONE | https://doi.org/10.1371/journal.pone.0258709 October 18, 2021
11 / 12
PLOS ONE Spatiotemporal inflection points in human running
11. Minetti AE, Alexander RM. A theory of metabolic costs for bipedal gaits. J Theor Biol. 1997; 186
(4):467–76. https://doi.org/10.1006/jtbi.1997.0407 PMID: 9278722
12. Yanai T, Hay JG. Combinations of Cycle Rate and Length for Minimizing the Muscle Power Require-
ment in Human Running. J Appl Biomech. 2004; 20(1):51–70.
13. Peyre-Tartaruga LA, Coertjens M. Locomotion as a Powerful Model to Study Integrative Physiology:
Efficiency, Economy, and Power Relationship. Front Physiol. 2018; 9:1789. https://doi.org/10.3389/
fphys.2018.01789 PMID: 30618802
14. Peyre-Tartaruga LA, Dewolf AH, di Prampero PE, Fabrica G, Malatesta D, Minetti AE, et al. Mechanical
work as a (key) determinant of energy cost in human locomotion: recent findings and future directions.
Exp Physiol. 2021; 106(9):1897–908. https://doi.org/10.1113/EP089313 PMID: 34197674
15. Cavanagh PR, Williams KR. The effect of stride length variation on oxygen uptake during distance run-
ning. Med Sci Sports Exerc. 1982; 14(1):30–5. https://doi.org/10.1249/00005768-198201000-00006
PMID: 7070254
16.
de Ruiter CJ, Verdijk PW, Werker W, Zuidema MJ, de Haan A. Stride frequency in relation to oxygen
consumption in experienced and novice runners. Eur J Sport Sci. 2014; 14(3):251–8. https://doi.org/10.
1080/17461391.2013.783627 PMID: 23581294
17. Connick MJ, Li FX. Changes in timing of muscle contractions and running economy with altered stride
pattern during running. Gait Posture. 2014; 39(1):634–7. https://doi.org/10.1016/j.gaitpost.2013.07.112
PMID: 23948332
18.
van Oeveren BT, de Ruiter CJ, Beek PJ, van Dieen JH. Optimal stride frequencies in running at different
speeds. PLoS One. 2017; 12(10):e0184273. https://doi.org/10.1371/journal.pone.0184273 PMID: 29059198
19. Cavagna GA, Mantovani M, Willems PA, Musch G. The resonant step frequency in human running.
Pflugers Arch. 1997; 434(6):678–84. https://doi.org/10.1007/s004240050451 PMID: 9305998
20. Dorn TW, Schache AG, Pandy MG. Muscular strategy shift in human running: dependence of running
speed on hip and ankle muscle performance. J Exp Biol. 2012; 215(Pt 11):1944–56. https://doi.org/10.
1242/jeb.064527 PMID: 22573774
21. Kakehata G, Kobayashi K, Matsuo A, Kanosue K, Iso S. Relationship between subjective effort and
kinematics/kinetics in the 50 m sprint. Journal of Human Sport and Exercise. 2020; 15(1):52–66.
22.
Ivy JL, Withers RT, Van Handel PJ, Elger DH, Costill DL. Muscle respiratory capacity and fiber type as
determinants of the lactate threshold. J Appl Physiol Respir Environ Exerc Physiol. 1980; 48(3):523–7.
https://doi.org/10.1152/jappl.1980.48.3.523 PMID: 7372524
23. Neder JA, Stein R. A simplified strategy for the estimation of the exercise ventilatory thresholds. Med
Sci Sports Exerc. 2006; 38(5):1007–13. https://doi.org/10.1249/01.mss.0000218141.90442.6c PMID:
16672856
24. Chen CWS, Chan JSK, Gerlach R, Hsieh WYL. A comparison of estimators for regression models with
change points. Stat Comput. 2011; 21(3):395–414.
25. Deming W. Statistical adjustment of data. New York: Wiley (Dover Publications Edition, 1985)1943.
https://doi.org/10.1126/science.97.2514.209 PMID: 17780136
26. Martin RF. General deming regression for estimating systematic bias and its confidence interval in
method-comparison studies. Clin Chem. 2000; 46(1):100–4. PMID: 10620577
27.
Tartaruga MP, Brisswalter J, Peyre-Tartaruga LA, Avila AO, Alberton CL, Coertjens M, et al. The rela-
tionship between running economy and biomechanical variables in distance runners. Res Q Exerc
Sport. 2012; 83(3):367–75. https://doi.org/10.1080/02701367.2012.10599870 PMID: 22978185
28. Minetti AE, Saibene F. Mechanical work rate minimization and freely chosen stride frequency of human
walking: a mathematical model. J Exp Biol. 1992; 170:19–34. PMID: 1402610
29.
Fischer G, Storniolo JLL, Peyre´ -Tartaruga LA. Effects of Fatigue on Running Mechanics: Spring-Mass
Behavior in Recreational Runners After 60 Seconds of Countermovement Jumps. Journal of Applied
Biomechanics. 2015; 31(6):445–51.
30. Pantoja PD, Morin JB, Peyre-Tartaruga LA, Brisswalter J. Running Energy Cost and Spring-Mass
Behavior in Young versus Older Trained Athletes. Med Sci Sports Exerc. 2016; 48(9):1779–86. https://
doi.org/10.1249/MSS.0000000000000959 PMID: 27116643
31. Cavagna GA, Legramandi MA, Peyre-Tartaruga LA. Old men running: mechanical work and elastic
bounce. Proc Biol Sci. 2008; 275(1633):411–8. https://doi.org/10.1098/rspb.2007.1288 PMID: 18077249
32. Stiffler-Joachim MR, Wille C, Kliethermes S, Heiderscheit B. Factors Influencing Base of Gait During
Running: Consideration of Sex, Speed, Kinematics, and Anthropometrics. J Athl Train. 2020; 55
(12):1300–6. https://doi.org/10.4085/1062-6050-565-19 PMID: 33064810
PLOS ONE | https://doi.org/10.1371/journal.pone.0258709 October 18, 2021
12 / 12
PLOS ONE |
10.1371_journal.pone.0244089 | RESEARCH ARTICLE
Marine n-3 fatty acid consumption in a
Norwegian renal transplant cohort:
Comparison of a food frequency
questionnaire with plasma phospholipid
marine n-3 levels
Joe ChanID
1,2*, My Svensson1,2, Trond Jenssen2,3, Erik B. Schmidt4, Ivar A. Eide1,3
1 Department of Renal Medicine, Akershus University Hospital, Lørenskog, Norway, 2 Institute of Clinical
Medicine, Faculty of Medicine, University of Oslo, Oslo, Norway, 3 Department of Transplantation Medicine,
Oslo University Hospital, Rikshospitalet, Oslo, Norway, 4 Department of Clinical Medicine, Aalborg University
Hospital, Aalborg, Denmark
* joe.chan@ahus.no
Abstract
Background
High levels of plasma marine n-3 fatty acids (n-3FAs) are associated with improved patient
and graft survival in renal transplant recipients (RTRs). The aim of this study was to evaluate
the utility of a new food frequency questionnaire (FFQ) to estimate marine n-3FA consump-
tion in future epidemiological research.
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: Chan J, Svensson M, Jenssen T, Schmidt
EB, Eide IA (2020) Marine n-3 fatty acid
consumption in a Norwegian renal transplant
cohort: Comparison of a food frequency
questionnaire with plasma phospholipid marine n-3
levels. PLoS ONE 15(12): e0244089. https://doi.
org/10.1371/journal.pone.0244089
Editor: Stefano Turolo, Ospedale Maggiore
Policlinico, ITALY
Methods
Received: April 29, 2020
Accepted: December 2, 2020
Published: December 17, 2020
Peer Review History: PLOS recognizes the
benefits of transparency in the peer review
process; therefore, we enable the publication of
all of the content of peer review and author
responses alongside final, published articles. The
editorial history of this article is available here:
https://doi.org/10.1371/journal.pone.0244089
Copyright: © 2020 Chan et al. This is an open
access article distributed under the terms of the
Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: All relevant data are
within the paper and its Supporting Information
files.
We developed an FFQ with a simple design of 10 questions to assess intake of marine
sources of n-3FAs. RTRs included in the recent ORENTRA trial (n = 132) completed the
study FFQ at the baseline visit eight weeks after engraftment and at the end of study visit
one year post-transplant. We measured the reference biomarker plasma phospholipid (PL)
marine n-3FA levels by gas chromatography at the same time points to evaluate association
and degree of agreement between FFQ based marine n-3FA consumption estimates and
the biomarker.
Results
The median plasma PL marine n-3FA level was 6.0 weight percentage (wt)% (interquartile
range [IQR] 4.7 to 7.3) at baseline and 6.3 wt% (IQR 4.8 to 7.4) at end of study. Median FFQ
based marine n-3FA consumption estimates were 22.8 g/month (IQR 13.0 to 34.0) at base-
line and 20.3 g/month (IQR 14.5 to 32.3) at end of study. FFQ based marine n-3FA con-
sumption estimates showed a moderate correlation with plasma PL marine n-3FA levels at
baseline (Spearman’s correlation coefficient rs = 0.43, p<0.001) and a stronger correlation
at end of study (rs = 0.62, p<0.001). Bland Altman plots showed a reasonable degree of
agreement between the two methods at both time points.
PLOS ONE | https://doi.org/10.1371/journal.pone.0244089 December 17, 2020
1 / 14
PLOS ONE Marine n-3 fatty acid consumption in a Norwegian renal transplant cohort
Funding: The author(s) received no specific
funding for this work.
Competing interests: The authors have declared
that no competing interests exist.
Conclusions
Marine n-3FA consumption estimates based on the FFQ showed a moderate correlation
with the reference biomarker plasma PL marine n-3FA levels. The FFQ might be useful in
epidemiological studies where resources are limited.
Introduction
Marine n-3 fatty acid (n-3FA) consumption may benefit cardiovascular health and renal func-
tion following renal transplantation [1, 2]. Previous clinical trials in renal transplant recipients
(RTRs) report lower triglyceride levels, higher high-density lipoprotein cholesterol levels and
lower diastolic blood pressure after marine n-3FA supplementation [1]. A large cohort study
in Norwegian RTRs showed that high plasma phospholipid (PL) n-3FA levels were associated
with improved patient and graft survival [3, 4]. Antifibrotic and renoprotective effects of long-
term high-dose marine n-3FA supplementation have also been shown for other cardiovascular
high-risk populations like myocardial infarction survivors [5, 6]. The recent “Omega-3 fatty
acids in Renal Transplantation (ORENTRA)” trial performed in Norwegian RTRs found lower
levels of inflammatory biomarkers, less development of renal graft fibrosis and improvement
of endothelial function, as well as reduced triglyceride levels after 44 weeks of high-dose n-3FA
supplementation [2].
Observational studies and randomized clinical trials (RCTs) studying the influence of
marine n-3FA intake on cardiovascular health report conflicting results [7–14]. But a recent
meta-analysis, which included three recent large RCTs [15–17], concluded that marine n-3FA
supplementation was associated with a lower risk of cardiovascular events and death [18]. In
renal transplantation, further studies are warranted to evaluate to what extent marine n-3FA
consumption may improve patient and graft survival.
The major marine n-3FAs eicosapentaenoic acid (EPA) and docosahexaenoic acid (DHA)
are found in fish and other seafood. Plasma PL levels of EPA and DHA can be measured by
fatty acid analysis and are considered valid and reliable measures of marine n-3FA consump-
tion [19]. However, fatty acid analysis is more expensive and time-consuming to apply than a
food frequency questionnaire (FFQ). Hence, replacing fatty acid analyses with an FFQ focused
on marine n-3FA consumption seems attractive in epidemiological research, provided that the
FFQ values show a high degree of agreement and association with the reference biomarker.
The main objective of this study was therefore to evaluate the utility of a new FFQ focused
on marine n-3FA consumption, using plasma PL marine n-3FA level as the reference
biomarker.
Materials and methods
Study participants and design
The study cohort consisted of 132 adult Norwegian RTRs included in the ORENTRA trial [2],
who were randomized to receive daily supplementation of either 2.6 g of marine n-3FAs (EPA
plus DHA) or 3 g of extra virgin olive oil (control oil) for 44 weeks. All patients gave written
informed consent for participation in the trial, which also comprised the study FFQ and fatty
acid analysis. The study was approved by the Regional Committees for Medical and Health
Research Ethics in Norway and was performed in accordance with the Declaration of Helsinki
(Clinical.Trials.gov identifier NCT01744067). FFQ and fatty acid analysis were performed
PLOS ONE | https://doi.org/10.1371/journal.pone.0244089 December 17, 2020
2 / 14
PLOS ONE Marine n-3 fatty acid consumption in a Norwegian renal transplant cohort
eight weeks post-engraftment (baseline visit) and one year after transplantation (end of study
visit). Patients were treated with standard triple maintenance immunosuppressive regimen
consisting of prednisolone, mycophenolate and tacrolimus. Blood samples were drawn in a
fasting state in the morning at the baseline and end of study visits. Gas chromatography was
used to determine individual fatty acid levels in plasma PLs, quantified as weight percentage
(wt%) of total plasma PL fatty acids. We defined marine n-3FA level as the sum of EPA and
DHA. The study was performed at Oslo University Hospital during 2012–2015. Details regard-
ing recruitment of patients, fatty acid analysis and the study FFQ are provided in the S1 File.
For the ORENTRA trial, we developed a specific FFQ with a simple design of 10 multiple-
choice questions (Fig 1), focusing on food items containing marine sources of n-3FA that are
typically found in a Nordic diet [20].
We used three different approaches to estimate marine n-3FA consumption at baseline and
end of study based on the FFQ recordings:
1. Marine n-3FA consumption estimates, calculated by combining data from the FFQ with
known content of EPA and DHA in fish and other seafoods [21], assuming a standard por-
tion size for a Norwegian population (S1 File, S2 Fig).
2. Marine n-3FA consumption estimates calculated as in approach 1 using only data on fatty
fish intake for lunch and dinner (S3 Fig).
3. Number of fish servings per month (S4 Fig).
Statistical analysis
We used correlation analysis (Spearman’s correlation coefficient [rs]) and multivariate regres-
sion analysis (data presented as standardized regression coefficients [Std. β-coeff.]) to study
associations between FFQ based marine n-3FA consumption estimates and plasma PL marine
n-3FA levels. Data obtained by the reference biomarker and the study FFQ were standardized
using z-statistics to produce data for both methods on the same scale. This allowed for a more
meaningful visual presentation (scatter plots) and made it possible to analyze degree of agree-
ment using Bland Altman plots and one-sample t-test. Since the study drug used in the OREN-
TRA trial was high-dose marine n-3FA supplementation, we excluded patients in the
intervention group when performing statistical analysis of data from the end of study visit.
Two patients belonging to the control group did not meet at the end of study visit (n = 66 at
baseline, n = 64 at end of study visit). Patient characteristics at baseline grouped according to
plasma PL marine n-3FA tertiles were evaluated with analysis of variance for continuous data
and Mantel-Haenszel linear-by-linear-trend for categorical data. A two-sided p-value
of < 0.05 was considered statistically significant. We used SPSS1 version 25.0 (IBM, New
York, NY, US) for statistical analyses.
Results
Patient characteristics for the study cohort have previously been published in detail [2].
Selected variables, grouped according to plasma PL n-3FA tertiles at baseline eight weeks
post-transplant, are presented in Table 1. Patients in the upper tertile were older and less often
current smokers. Supplementation with cod liver oil was used by 28% of patients in the upper
tertile compared with 9% in the lower. Median plasma PL n-3FA levels were 6.0 wt% (inter-
quartile range [IQR] 4.7 to 7.3, n = 132) at the baseline visit and 6.3 wt% (IQR 4.8 to 7.4,
n = 64) at the end of study visit. Median FFQ based marine n-3FA consumption estimates
were 22.8 g/month (IQR 13.0 to 34.0, n = 132) at baseline and 20.3 g/month (IQR 14.5 to 32.3,
PLOS ONE | https://doi.org/10.1371/journal.pone.0244089 December 17, 2020
3 / 14
PLOS ONE Marine n-3 fatty acid consumption in a Norwegian renal transplant cohort
Fig 1. Study food frequency questionnaire focusing on food items containing marine n-3 polyunsaturated fatty acids (English
version). The study subjects responded to the question “During a typical month, how often do you eat these food items?” using one of
six response alternatives for each of the ten food item categories.
https://doi.org/10.1371/journal.pone.0244089.g001
PLOS ONE | https://doi.org/10.1371/journal.pone.0244089 December 17, 2020
4 / 14
PLOS ONE Table 1. Patient characteristics at baseline eight weeks after renal transplantation according to plasma phospholipid marine n-3 fatty acid tertiles.
Marine n-3 fatty acid consumption in a Norwegian renal transplant cohort
Variables
Number of patients
FFQ based marine fatty acid consumption estimate, g/month
Number of servings of fish / month
Marine n-3FA supplements, %
Recipient age, years
Recipient gender (Female), %
Ethnicity, White, %
Body mass index, kg/m2
Educational level, %
>3 years at University
1–3 years at University
Secondary school
Primary school
Physical exercise, %
High intensity � twice per week
High intensity once per week
Low intensity � twice per week
Low intensity once per week
None
Smoking habits, %
Daily smoker
Non-daily smoker
Former heavy smoker
Former light smoker
Life-long non-smoker
All patients
Plasma PL marine n-3FA level, wt%
p (trend)
132
26.0 (16.6)
19.9 (15.6)
14.5
�5.1
44
19.8 (15.0)
13.3 (13.8)
9.1
5.2–6.9
44
25.4 (17.5)
16.2 (13.4)
6.8
�7.0
44
33.1 (14.7)
26.9 (16.8)
27.9
0.001
0.12
0.01
53.4 (13.8)
45.7 (12.6)
55.6 (13.6)
59.1 (12.0)
<0.001
25.8
92.4
27.3
93.2
25.0
86.0
25.6
97.7
26.0 (3.9)
25.2 (4.0)
25.9 (3.8)
26.8 (3.7)
29.8
7.6
35.1
27.5
42.1
9.5
34.9
7.1
6.3
12.7
3.2
7.1
35.7
41.3
27.3
6.8
40.9
25.0
51.2
2.4
36.6
9.8
0.0
19.5
7.3
7.3
34.1
31.7
27.3
6.8
36.4
29.5
36.4
11.4
36.4
9.1
6.8
11.4
0.0
4.5
43.2
40.9
34.9
9.3
27.9
27.9
39.0
14.6
31.7
2.4
12.2
7.3
2.4
9.8
29.3
51.2
0.86
0.39
0.16
0.56
0.29
0.03
Patient characteristics are presented as percentage for categorical data and mean value (standard deviation) for continuous variables. Differences between groups were
analyzed by analysis of variance and Mantel Haenszel linear-by-linear trend as appropriate.
https://doi.org/10.1371/journal.pone.0244089.t001
n = 64) at end of study. Marine n-3FA consumption remained stable during follow-up in the
control group with a median increase of plasma PL marine n-3FA level of 0.1 wt% (IQR -0.8 to
1.0) and change in FFQ based marine n-3FA consumption estimates of -1.0 g/month (IQR
-9.4 to 6.3).
At baseline, moderate correlations were found between FFQ based marine n-3FA con-
sumption estimates and the reference biomarker plasma PL marine n-3FA levels (approach 1,
rs = 0.43, p<0.001, n = 132, Fig 2). A reasonable degree of agreement between the study FFQ
estimates and the reference biomarker was shown in a Bland Altman plot (Fig 3) and signifi-
cant bias was ruled out by a one-sample t-test (t = 0.04, p = 0.96). Two groups of outlier obser-
vations were identified. One group consisted of patients reporting high intake of marine n-
3FAs but had average plasma PL marine n-3 FA levels. Another group with high or very high
plasma PL marine n-3FA levels had average marine n-3FA consumption according to the
study FFQ.
Correlations between FFQ based marine n-3FA consumption estimates and plasma PL
marine n-3FA levels were stronger at the end of study (rs = 0.60, p<0.001, n = 64, Fig 4) than
at baseline. One-sample t-test (t = 0.06, p = 0.95) and a Bland Altman plot confirmed an
acceptable degree of agreement between the two methods at this time point (Fig 5).
PLOS ONE | https://doi.org/10.1371/journal.pone.0244089 December 17, 2020
5 / 14
PLOS ONE Marine n-3 fatty acid consumption in a Norwegian renal transplant cohort
Fig 2. Scatterplots of standardized plasma PL marine n-3FA levels and standardized FFQ based marine n-3FA
consumption estimates with regression line at eight weeks post-transplant (n = 132).
https://doi.org/10.1371/journal.pone.0244089.g002
Baseline correlation analysis was repeated for patients belonging to the ORENTRA trial
control group (rs = 0.45, p<0.001, n = 66, S5 Fig) and we found a high degree of agreement
between the methods (S6 Fig), similar to what was shown for the whole study cohort at
baseline.
We performed a multivariate stepwise forward regression analysis, adjusting for the poten-
tial confounding factors recipient age, gender, height, weight, body mass index, renal function,
physical activity, educational level and smoking habits (p<0.10 for inclusion of variables in the
final regression model) at baseline and end of study. The reference biomarker plasma PL
marine n-3FA level was associated with FFQ based marine n-3FA consumption estimates (Std.
β-coeff. 0.24, p = 0.01), as well as recipient age (Std. β-coeff. 0.25, p = 0.01) and smoking habits
(Std. β-coeff. 0.15, p = 0.06) at baseline (n = 132). Together the three variables included in final
Fig 3. Bland-Altman plot assessing degree of agreement between standardized plasma PL marine n-3FA levels
and standardized FFQ based marine n-3FA consumption estimates at baseline eight weeks post-transplant. We
used standardization of data obtained by the study FFQ and reference biomarker, hence the mean value was set at 0.
The upper and lower limits of agreement were set at 2 standard deviations from the mean. The Bland Altman plot
includes all patients enrolled in the ORENTRA trial (n = 132) at the baseline time-point.
https://doi.org/10.1371/journal.pone.0244089.g003
PLOS ONE | https://doi.org/10.1371/journal.pone.0244089 December 17, 2020
6 / 14
PLOS ONE Marine n-3 fatty acid consumption in a Norwegian renal transplant cohort
Fig 4. Scatterplots of standardized plasma PL marine n-3FA levels and standardized FFQ based marine n-3FA
consumption estimates with regression line at one year post-transplant for patients belonging to the control
group of the ORENTRA trial (n = 64).
https://doi.org/10.1371/journal.pone.0244089.g004
regression model explained 23% of the variance in plasma PL marine n-3FA levels. At the end
of study, only FFQ based marine n-3FA consumption estimates (Std. β-coeff. 0.54, p<0.001)
was included in the final regression model, and it explained 29% of the variance in the refer-
ence biomarker.
Correlations between FFQ based marine n-3FA consumption estimates and the reference
biomarker were slightly weaker for fatty fish intake (approach 2, baseline rs = 0.35 and end of
study rs = 0.46) and number of fish servings per month (approach 3, baseline rs = 0.38 and end
of study rs = 0.43) than for total marine n-3FA consumption estimates (approach 1). Correla-
tions with the reference biomarker for individual food items included in the study FFQ are
shown in S7 Fig. The food item cod liver oil showed a low correlation with the reference
Fig 5. Bland-Altman plot assessing degree of agreement between standardized plasma PL marine n-3FA levels
and standardized FFQ based marine n-3FA consumption estimates at one year post-transplant. We used
standardization of data obtained by the study FFQ and reference biomarker, hence the mean value was set at 0. The
upper and lower limits of agreement were set at 2 standard deviations from the mean. The Bland Altman plot includes
only patients belonging to the control group of the ORENTRA trial (n = 64) at the end of study time-point.
https://doi.org/10.1371/journal.pone.0244089.g005
PLOS ONE | https://doi.org/10.1371/journal.pone.0244089 December 17, 2020
7 / 14
PLOS ONE Marine n-3 fatty acid consumption in a Norwegian renal transplant cohort
biomarker at baseline (rs = 0.21). Marine n-3FA supplementation, including cod liver oil, was
discontinued after enrollment in the ORENTRA trial and consequently intake of cod liver oil
did not influence results at one year post-transplant.
Discussion
The main finding of the present study was that marine n-3FA consumption estimates based on
a new FFQ focused on fish consumption showed a moderate correlation with the reference
biomarker plasma PL marine n-3FA levels at eight weeks post-transplant and a slightly stron-
ger correlation at one year post-transplant. The correlations found in the present study are in
the range of what is regarded as acceptable in FFQ validation studies [22].
Previous FFQs have mainly focused on fatty fish intake, assumed to reflect marine n-3FA
consumption better than total fish intake [23–35]. We hypothesized that a more meticulous
approach using weighted response scales based on EPA and DHA content in fatty and lean
fish, other seafoods and marine n-3FA supplements would provide a more precise estimation
of marine n-3FA consumption. In our cohort, approach 1, which estimated total marine n-
3FA consumption from all the data obtained by the study FFQ, showed a stronger correlation
with the reference biomarker than approach 2 (which only focused on fatty fish intake) and 3
(which used the number of fish servings), suggesting that our hypothesis was correct.
However, the study FFQ only provided slightly stronger correlations than most recent
FFQs focused on fish consumption (Table 2) and the utility of the study FFQ will have to be
confirmed by other studies before it can be used in epidemiological research.
Plasma PL marine n-3FA levels did not differ between baseline and end of study visits for
the majority of patients. This is consistent with previous reports from large Norwegian cohorts
and supports the notion that a single fatty acid measurement may be acceptable for epidemio-
logical studies [3, 19]. However, the association between FFQ based marine n-3FA consump-
tion estimates and the reference biomarker was stronger at end of study than at baseline.
There could be several explanations to this finding. Study participants might have become
more aware of their eating habits due to participation in the ORENTRAL trial and reported
fish consumption more accurately when they completed the FFQ the second time. We found a
Table 2. Summary of selected food frequency questionnaire validation studies published during the last six years, focusing on fish and/or marine fatty acid con-
sumption, using circulating phospholipids or erytrocytes as the reference biomarker.
First author (reference)
Published, year
Giovannelli J [23]
Lassale C [24]
Sluik D [25]
Whitton C [26]
Laursen UB [27]
Shen W [28]
Schumacher TL [29]
Allaire J [30]
Brunvoll SH [31]
Lepsch J [32]
Zhou YB [33]
Kobayashi M [34]
Liu MJ [35]
2014
2016
2016
2017
2018
2019
2016
2015
2018
2014
2017
2017
2016
https://doi.org/10.1371/journal.pone.0244089.t002
n
2630
198
383
161
200
108
39
60
49
248
804
188
408
Study population
Reference marine fatty acid biomarker
Correlation coefficient
General population
General population
General population
General population
General population
General population
Hyperlipidemia
Prostate cancer
Breast cancer
Pregnant women
Pregnant women
Pregnant women
Lactating women
Plasma phospholipid
Plasma phospholipid
Plasma phospholipid
Plasma phospholipid
Plasma phospholipid
Whole blood phospholipid
Erythrocyte
Erythrocyte
Serum phospholipid
Serum phospholipid
Plasma phospholipid
Erythrocyte
Serum phospholipid
Plasma phospholipid
Erythrocyte
r = 0.39–0.43
rs = 0.51–0.54
r = 0.43–0.47
r = 0.36
rs = 0.45
r = 0.67
rs = 0.53–0.62
rs = 0.59
r = 0.36–0.53
rs = 0.21–0.26
rs = 0.35
rs = 0.33
rs = 0.33–0.45
rs = 0.36
rs = 0.24
PLOS ONE | https://doi.org/10.1371/journal.pone.0244089 December 17, 2020
8 / 14
PLOS ONE Marine n-3 fatty acid consumption in a Norwegian renal transplant cohort
lower correlation with the reference biomarker for cod liver oil than for other food items in the
study FFQ at baseline, which likely influenced the results. Some patients with high plasma PL
marine n-3FA levels reported only average marine n-3FA intake according to the study FFQ, all
of whom reported frequent use of cod liver oil. This signals that the study FFQ weighted
response scale for cod liver oil likely underestimated marine n-3FA content, thus the study FFQ
in its current form lacks precision for patients taking daily marine n-3FA supplements. Addi-
tionally, patients with average plasma PL marine n-3 FA levels who reported high levels of
marine n-3FA intake in the FFQ, showed this pattern both at baseline and end of study. This
might be due to social desirability bias and has likely influenced results at both time-points.
Fish intake in Norway is higher than in most other European countries, due to the rich fish-
ing grounds along the Norwegian coastline with easy access to fresh cold-water fish [36].
Plasma PL marine n-3FA levels in the present cohort were relatively high, even for a Norwe-
gian population, signaling a selected population that focuses on healthy eating habits. On the
other hand, plasma PL marine n-3FA levels in the present study were comparable to a previous
large cohort study in Norwegian RTRs, suggesting that the sample was representative of a Nor-
wegian transplant cohort [3]. Confounding factors like socioeconomical class, educational
level, smoking habits and physical activity may influence associations between fish intake and
outcomes in epidemiological research [36]. In this cohort, FFQ based marine n-3FA consump-
tion estimates and plasma PL marine n-3FA level were associated with smoking habits, but not
other life-style factors.
Dietary habits are changing in the Nordic countries, with lower fish consumption in youn-
ger patients, including Norwegian RTRs [3], thus necessitating revision of questions and
response categories for the present study FFQ in future studies. Cod liver oil intake is an old
tradition in Norway [37] and was therefore included as one of the food times in the study FFQ.
This may be omitted in areas where intake of cod liver oil or other marine n-3FA supplements
are uncommon.
Strengths of the present study include a well-described cohort, plasma PL fatty acid analysis
and a study FFQ performed at two time points, which might improve accuracy. The study
FFQ has a simple design, is easy to read and understand and only takes a few minutes to
answer, which is desirable in large epidemiological studies.
There were also several limitations, including limitations by design such as recall bias and
social desirability bias and a relatively small sample size. The study FFQ marine n-3FA con-
sumption estimates were based on the sum of weighted response scales for ten food items, con-
taining questions on how frequent the food items were consumed, but not on portion size.
Thus, the weighted responses used to calculate marine n-3FA intake were based on assump-
tions of standard portion size for each item, constituting a major limitation in the present
study. Moreover, the study FFQ did not contain any questions regarding seasonal variations,
which could be relevant for some of the included food items in a Norwegian cohort. The study
FFQ contains rather detailed questions about fish and seafood intake and response categories
with minor differences (Fig 1). This likely improved precision for patients who are well aware
of their eating habits but could have been challenging for other patients, possibly leading to
random responses. Broader response categories might have produced more reliable data [38].
For patients on marine n-3FA supplements, like cod liver oil, weighted responses for this food
item in the study FFQ likely underestimated supplements as a source of marine n-3FAs.
The questionnaire was designed to estimate marine n-3FA consumption in a Norwegian
transplant cohort. Due to dietary differences between regions and between patient popula-
tions, FFQ validation studies designed for one region or one particular target population may
not apply to other regions or other patient cohorts [22]. In other regions, food items and
weighted response scales should be revised to reflect fish consumption in that region.
PLOS ONE | https://doi.org/10.1371/journal.pone.0244089 December 17, 2020
9 / 14
PLOS ONE Marine n-3 fatty acid consumption in a Norwegian renal transplant cohort
Moreover, adjustment for portion size and seasonal variations can be made to improve FFQ
performance.
In conclusion, marine n-3FA consumption estimates based on our study FFQ showed a
moderate correlation with the reference biomarker plasma PL marine n-3FA levels, with com-
parable performance to previous FFQs. We recommend using fatty acid analysis to ensure
objective measurement of marine n-3FA consumption in clinical trials, but our FFQ might be
useful in epidemiological studies where resources are limited.
Supporting information
S1 Fig. Study food frequency questionnaire focusing on food items containing marine n-3
fatty acids (Norwegian version). The study subjects responded to the question “During a typi-
cal month, how often do you eat these food items?” using one of six response alternatives for
each of the ten food item categories.
(PDF)
S2 Fig. Study food frequency questionnaire focusing on food items containing marine n-3
fatty acids (investigator’s scoring sheet version in English). The study subjects responded to
the question “During a typical month, how often do you eat these food items?” using one of six
response alternatives for each of the ten food item categories. Based on EPA and DHA content
in the meat of various fish and other seafoods presented in the US Department of Agriculture
Food Composition Database and assuming a standard portion size for dinner and bread
spread, every potential response was given a weight (shown inside boxes). Total intake of
marine n-3 fatty acids per month was calculated as the sum of the ten weighted responses in
grams.
(PDF)
S3 Fig. Study food frequency questionnaire focusing on food items containing marine n-3
fatty acids (investigator’s scoring sheet version in English comprising fatty fish items
only). The study subjects responded to the question “During a typical month, how often do
you eat these food items?” using one out of six response alternatives for each food item catego-
ries. Based on EPA and DHA content in the meat of various fish and other seafoods presented
in the US Department of Agriculture Food Composition Database and assuming a standard
portion size for dinner and bread spread, every potential response was given a weight (shown
inside boxes). Total intake of marine n-3 fatty acids per month was calculated as the sum of
the weighted responses in grams, which for fatty fish intake comprised the four items shown.
(PDF)
S4 Fig. Study food frequency questionnaire focusing on food items containing marine n-3
fatty acids (investigator’s scoring sheet version in English comprising fish servings per
month). The study subjects responded to the question “During a typical month, how often do
you eat these food items?” using one out of six response alternatives for each of the ten food
item categories. Servings of fish per month was calculated as the sum of the ten responses,
using the center value for each response category as shown.
(PDF)
S5 Fig. Scatterplots of standardized plasma marine n-3FA levels and standardized FFQ
based marine n-3FA consumption estimates with regression lines at baseline eight weeks
post-transplant for patients belonging to the control group of the ORENTRA trial
(n = 66).
(TIF)
PLOS ONE | https://doi.org/10.1371/journal.pone.0244089 December 17, 2020
10 / 14
PLOS ONE Marine n-3 fatty acid consumption in a Norwegian renal transplant cohort
S6 Fig. Bland-Altman plot assessing degree of agreement between standardized plasma
marine n-3FA levels and standardized FFQ based marine n-3FA consumption estimates at
baseline eight weeks post-transplant for patients belonging to the control group of the
ORENTRA trial (n = 66).
(TIF)
S7 Fig. Correlation matrix presenting Spearman’s correlation coefficients at eight weeks
(baseline visit) after renal transplantation for the whole study population (n = 132) and
one year post-transplant (end of study visit) for patients belonging to the control group of
the ORENTRA trial (n = 64).
(TIF)
S1 File. Supporting information. Includes information regarding “Patient screening and
recruitment in the ORENTRA trial”, “Fatty acid analysis”, “Sample Size and Power Calcula-
tion” and “Development of the study Food Frequency Questionnaire”.
(DOCX)
Acknowledgments
We thank coworkers Rikke Bu¨low Eschen, Annette Andreassen, Birthe H. Thomsen and Inge Aar-
destrup at The Lipid Research Laboratory, Aalborg University Hospital, Denmark for analyzing
plasma phospholipid fatty acids. We thank statistician Owen Thomas and colleague dr. Anupam
Chandra at Akershus University Hospital for their contribution to this manuscript. We thank the
funding sources Gidske and Peter Jacob Sørensen Research Fund and the South-Eastern Norway
Regional Health Authority. Finally, we thank the study participants in the ORENTRA trial.
Author Contributions
Conceptualization: My Svensson, Trond Jenssen, Erik B. Schmidt, Ivar A. Eide.
Data curation: Ivar A. Eide.
Formal analysis: Joe Chan, Ivar A. Eide.
Funding acquisition: My Svensson.
Investigation: Erik B. Schmidt, Ivar A. Eide.
Methodology: Joe Chan, My Svensson, Trond Jenssen, Erik B. Schmidt, Ivar A. Eide.
Project administration: Trond Jenssen, Ivar A. Eide.
Resources: My Svensson, Trond Jenssen, Erik B. Schmidt, Ivar A. Eide.
Supervision: My Svensson, Erik B. Schmidt, Ivar A. Eide.
Validation: Ivar A. Eide.
Writing – original draft: Joe Chan.
Writing – review & editing: Joe Chan, My Svensson, Trond Jenssen, Erik B. Schmidt, Ivar A.
Eide.
References
1.
Lim AK, Manley KJ, Roberts MA, Fraenkel MB. Fish oil for kidney transplant recipients. The Cochrane
database of systematic reviews. 2016;(8):Cd005282. Epub 2016/08/19. https://doi.org/10.1002/
14651858.CD005282.pub3 PMID: 27535773.
PLOS ONE | https://doi.org/10.1371/journal.pone.0244089 December 17, 2020
11 / 14
PLOS ONE Marine n-3 fatty acid consumption in a Norwegian renal transplant cohort
2. Eide IA, Reinholt FP, Jenssen T, Hartmann A, Schmidt EB, Asberg A, et al. Effects of marine n-3 fatty
acid supplementation in renal transplantation: A randomized controlled trial. Am J Transplant. 2019; 19
(3):790–800. Epub 2018/08/21. https://doi.org/10.1111/ajt.15080 PMID: 30125457.
3. Eide IA, Jenssen T, Hartmann A, Diep LM, Dahle DO, Reisaeter AV, et al. The association between
marine n-3 polyunsaturated fatty acid levels and survival after renal transplantation. Clinical journal of
the American Society of Nephrology: CJASN. 2015; 10(7):1246–56. Epub 2015/06/13. https://doi.org/
10.2215/CJN.11931214 PMID: 26063768; PubMed Central PMCID: PMC4491303.
4. Eide IA, Jenssen T, Hartmann A, Diep LM, Dahle DO, Reisaeter AV, et al. Plasma levels of marine n-3
polyunsaturated fatty acids and renal allograft survival. Nephrology, dialysis, transplantation: official
publication of the European Dialysis and Transplant Association—European Renal Association. 2016;
31(1):160–7. https://doi.org/10.1093/ndt/gfv339 PMID: 26410884.
5. Hoogeveen EK, Geleijnse JM, Kromhout D, Stijnen T, Gemen EF, Kusters R, et al. Effect of omega-3
fatty acids on kidney function after myocardial infarction: the Alpha Omega Trial. Clinical journal of the
American Society of Nephrology: CJASN. 2014; 9(10):1676–83. https://doi.org/10.2215/CJN.10441013
PMID: 25104273; PubMed Central PMCID: PMC4186521.
6. Heydari B, Abdullah S, Pottala JV, Shah R, Abbasi S, Mandry D, et al. Effect of Omega-3 Acid Ethyl
Esters on Left Ventricular Remodeling After Acute Myocardial Infarction: The OMEGA-REMODEL Ran-
domized Clinical Trial. Circulation. 2016; 134(5):378–91. Epub 2016/08/03. https://doi.org/10.1161/
CIRCULATIONAHA.115.019949 PMID: 27482002; PubMed Central PMCID: PMC4973577.
7. Burr ML, Fehily AM, Gilbert JF, Rogers S, Holliday RM, Sweetnam PM, et al. Effects of changes in fat,
fish, and fibre intakes on death and myocardial reinfarction: diet and reinfarction trial (DART). Lancet.
1989; 2(8666):757–61. Epub 1989/09/30. https://doi.org/10.1016/s0140-6736(89)90828-3 PMID:
2571009.
8.
Tavazzi L, Maggioni AP, Marchioli R, Barlera S, Franzosi MG, Latini R, et al. Effect of n-3 polyunsatu-
rated fatty acids in patients with chronic heart failure (the GISSI-HF trial): a randomised, double-blind,
placebo-controlled trial. Lancet. 2008; 372(9645):1223–30. Epub 2008/09/02. https://doi.org/10.1016/
S0140-6736(08)61239-8 PMID: 18757090.
9. Kromhout D, Giltay EJ, Geleijnse JM, Alpha Omega Trial G. n-3 fatty acids and cardiovascular events
after myocardial infarction. N Engl J Med. 2010; 363(21):2015–26. Epub 2010/10/12. https://doi.org/10.
1056/NEJMoa1003603 PMID: 20929341.
10. Bosch J, Gerstein HC, Dagenais GR, Diaz R, Dyal L, Jung H, et al. n-3 fatty acids and cardiovascular
outcomes in patients with dysglycemia. N Engl J Med. 2012; 367(4):309–18. Epub 2012/06/13. https://
doi.org/10.1056/NEJMoa1203859 PMID: 22686415.
11. Bork CS, Mortensen LT, Hjelmgaard K, Schmidt EB. Marine n-3 fatty acids and CVD: new insights from
recent follow-up studies and clinical supplementation trials. Proc Nutr Soc. 2020:1–7. Epub 2020/04/
03. https://doi.org/10.1017/S0029665120006886 PMID: 32234084.
12. Yokoyama M, Origasa H, Matsuzaki M, Matsuzawa Y, Saito Y, Ishikawa Y, et al. Effects of eicosapenta-
enoic acid on major coronary events in hypercholesterolaemic patients (JELIS): a randomised open-
label, blinded endpoint analysis. Lancet. 2007; 369(9567):1090–8. https://doi.org/10.1016/S0140-6736
(07)60527-3 PMID: 17398308.
13. Galan P, Kesse-Guyot E, Czernichow S, Briancon S, Blacher J, Hercberg S, et al. Effects of B vitamins
and omega 3 fatty acids on cardiovascular diseases: a randomised placebo controlled trial. BMJ. 2010;
341:c6273. Epub 2010/12/01. https://doi.org/10.1136/bmj.c6273 PMID: 21115589; PubMed Central
PMCID: PMC2993045.
14. Rauch B, Schiele R, Schneider S, Diller F, Victor N, Gohlke H, et al. OMEGA, a randomized, placebo-
controlled trial to test the effect of highly purified omega-3 fatty acids on top of modern guideline-
adjusted therapy after myocardial infarction. Circulation. 2010; 122(21):2152–9. Epub 2010/11/10.
https://doi.org/10.1161/CIRCULATIONAHA.110.948562 PMID: 21060071.
15. Bowman L, Mafham M, Wallendszus K, Stevens W, Buck G, Barton J, et al. Effects of n-3 Fatty Acid
Supplements in Diabetes Mellitus. N Engl J Med. 2018; 379(16):1540–50. Epub 2018/08/28. https://doi.
org/10.1056/NEJMoa1804989 PMID: 30146932.
16. Manson JE, Cook NR, Lee IM, Christen W, Bassuk SS, Mora S, et al. Marine n-3 Fatty Acids and Pre-
vention of Cardiovascular Disease and Cancer. N Engl J Med. 2019; 380(1):23–32. Epub 2018/11/13.
https://doi.org/10.1056/NEJMoa1811403 PMID: 30415637; PubMed Central PMCID: PMC6392053.
17. Bhatt DL, Steg PG, Miller M, Brinton EA, Jacobson TA, Ketchum SB, et al. Cardiovascular Risk Reduc-
tion with Icosapent Ethyl for Hypertriglyceridemia. N Engl J Med. 2019; 380(1):11–22. Epub 2018/11/
13. https://doi.org/10.1056/NEJMoa1812792 PMID: 30415628.
18. Hu Y, Hu FB, Manson JE. Marine Omega-3 Supplementation and Cardiovascular Disease: An Updated
Meta-Analysis of 13 Randomized Controlled Trials Involving 127 477 Participants. J Am Heart Assoc.
PLOS ONE | https://doi.org/10.1371/journal.pone.0244089 December 17, 2020
12 / 14
PLOS ONE Marine n-3 fatty acid consumption in a Norwegian renal transplant cohort
2019; 8(19):e013543. Epub 2019/10/01. https://doi.org/10.1161/JAHA.119.013543 PMID: 31567003;
PubMed Central PMCID: PMC6806028.
19.
Lindberg M, Midthjell K, Bjerve KS. Long-term tracking of plasma phospholipid fatty acid concentrations
and their correlation with the dietary intake of marine foods in newly diagnosed diabetic patients: results
from a follow-up of the HUNT Study, Norway. Br J Nutr. 2013; 109(6):1123–34. Epub 2012/08/01.
https://doi.org/10.1017/S0007114512002759 PMID: 22846205.
20. Engeset D, Alsaker E, Ciampi A, Lund E. Dietary patterns and lifestyle factors in the Norwegian EPIC
cohort: the Norwegian Women and Cancer (NOWAC) study. European journal of clinical nutrition.
2005; 59(5):675–84. Epub 2005/03/24. https://doi.org/10.1038/sj.ejcn.1602129 PMID: 15785773.
21. MacLean CH, Mojica WA, Morton SC, Pencharz J, Hasenfeld Garland R, Tu W, et al. Effects of omega-
3 fatty acids on lipids and glycemic control in type II diabetes and the metabolic syndrome and on inflam-
matory bowel disease, rheumatoid arthritis, renal disease, systemic lupus erythematosus, and osteopo-
rosis. Evidence report/technology assessment (Summary). 2004;(89):1–4. Epub 2004/05/12. PMID:
15133890.
22. Willett W. Nutritional epidemiology. Third edition. ed. Oxford; New York: Oxford University Press;
2013. ix, 529 pages p.
23. Giovannelli J, Dallongeville J, Wagner A, Bongard V, Laillet B, Marecaux N, et al. Validation of a short,
qualitative food frequency questionnaire in French adults participating in the MONA LISA-NUT study
2005–2007. J Acad Nutr Diet. 2014; 114(4):552–61. Epub 2013/10/03. https://doi.org/10.1016/j.jand.
2013.07.002 PMID: 24083967.
24.
Lassale C, Castetbon K, Laporte F, Deschamps V, Vernay M, Camilleri GM, et al. Correlations between
Fruit, Vegetables, Fish, Vitamins, and Fatty Acids Estimated by Web-Based Nonconsecutive Dietary
Records and Respective Biomarkers of Nutritional Status. J Acad Nutr Diet. 2016; 116(3):427–38 e5.
Epub 2015/11/03. https://doi.org/10.1016/j.jand.2015.09.017 PMID: 26522988.
25. Sluik D, Geelen A, de Vries JH, Eussen SJ, Brants HA, Meijboom S, et al. A national FFQ for the Neth-
erlands (the FFQ-NL 1.0): validation of a comprehensive FFQ for adults. Br J Nutr. 2016; 116(5):913–
23. Epub 2016/07/28. https://doi.org/10.1017/S0007114516002749 PMID: 27452894.
26. Whitton C, Ho JCY, Tay Z, Rebello SA, Lu Y, Ong CN, et al. Relative Validity and Reproducibility of a
Food Frequency Questionnaire for Assessing Dietary Intakes in a Multi-Ethnic Asian Population Using
24-h Dietary Recalls and Biomarkers. Nutrients. 2017; 9(10). Epub 2017/09/28. https://doi.org/10.3390/
nu9101059 PMID: 28946670; PubMed Central PMCID: PMC5691676.
27.
Laursen UB, Rosenkilde LB, Haugaard AM, Obel T, Toft U, Larsen ML, et al. Validation of the HeartDiet
questionnaire. Dan Med J. 2018; 65(11). Epub 2018/11/02. PMID: 30382020.
28. Shen W, Weaver AM, Salazar C, Samet JM, Diaz-Sanchez D, Tong H. Validation of a Dietary Question-
naire to Screen Omega-3 Fatty Acids Levels in Healthy Adults. Nutrients. 2019; 11(7). Epub 2019/07/
03. https://doi.org/10.3390/nu11071470 PMID: 31261632; PubMed Central PMCID: PMC6682879.
29. Schumacher TL, Burrows TL, Rollo ME, Wood LG, Callister R, Collins CE. Comparison of fatty acid
intakes assessed by a cardiovascular-specific food frequency questionnaire with red blood cell mem-
brane fatty acids in hyperlipidaemic Australian adults: a validation study. European journal of clinical
nutrition. 2016; 70(12):1433–8. Epub 2016/08/11. https://doi.org/10.1038/ejcn.2016.144 PMID:
27507074.
30. Allaire J, Moreel X, Labonte ME, Leger C, Caron A, Julien P, et al. Validation of the omega-3 fatty acid
intake measured by a web-based food frequency questionnaire against omega-3 fatty acids in red
blood cells in men with prostate cancer. European journal of clinical nutrition. 2015; 69(9):1004–8. Epub
2015/03/12. https://doi.org/10.1038/ejcn.2015.7 PMID: 25758837.
31. Brunvoll SH, Thune I, Frydenberg H, Flote VG, Bertheussen GF, Schlichting E, et al. Validation of
repeated self-reported n-3 PUFA intake using serum phospholipid fatty acids as a biomarker in breast
cancer patients during treatment. Nutr J. 2018; 17(1):94. Epub 2018/10/20. https://doi.org/10.1186/
s12937-018-0402-6 PMID: 30333016; PubMed Central PMCID: PMC6192340.
32.
33.
Lepsch J, Vaz JS, Moreira JD, Pinto TJ, Soares-Mota M, Kac G. Food frequency questionnaire as an
indicator of the serum composition of essential n-3 and n-6 polyunsaturated fatty acids in early preg-
nancy, according to body mass index. J Hum Nutr Diet. 2015; 28(1):85–94. Epub 2014/03/14. https://
doi.org/10.1111/jhn.12225 PMID: 24620790.
Zhou YB, Li HT, Trasande L, Wang LL, Zhang YL, Si KY, et al. A Correlation Study of DHA Intake Esti-
mated by a FFQ and Concentrations in Plasma and Erythrocytes in Mid- and Late Pregnancy. Nutrients.
2017; 9(11). Epub 2017/11/17. https://doi.org/10.3390/nu9111256 PMID: 29144430; PubMed Central
PMCID: PMC5707728.
34. Kobayashi M, Jwa SC, Ogawa K, Morisaki N, Fujiwara T. Validity of a food frequency questionnaire to
estimate long-chain polyunsaturated fatty acid intake among Japanese women in early and late
PLOS ONE | https://doi.org/10.1371/journal.pone.0244089 December 17, 2020
13 / 14
PLOS ONE Marine n-3 fatty acid consumption in a Norwegian renal transplant cohort
pregnancy. J Epidemiol. 2017; 27(1):30–5. Epub 2017/01/31. https://doi.org/10.1016/j.je.2016.07.001
PMID: 28135195; PubMed Central PMCID: PMC5328737.
35.
Liu MJ, Li HT, Yu LX, Xu GS, Ge H, Wang LL, et al. A Correlation Study of DHA Dietary Intake and
Plasma, Erythrocyte and Breast Milk DHA Concentrations in Lactating Women from Coastland, Lake-
land, and Inland Areas of China. Nutrients. 2016; 8(5). Epub 2016/05/24. https://doi.org/10.3390/
nu8050312 PMID: 27213448; PubMed Central PMCID: PMC4882724.
36. Micha R, Khatibzadeh S, Shi P, Fahimi S, Lim S, Andrews KG, et al. Global, regional, and national con-
sumption levels of dietary fats and oils in 1990 and 2010: a systematic analysis including 266 country-
specific nutrition surveys. BMJ. 2014; 348:g2272. Epub 2014/04/17. https://doi.org/10.1136/bmj.g2272
PMID: 24736206; PubMed Central PMCID: PMC3987052.
37. Mai XM, Langhammer A, Chen Y, Camargo CA Jr., Cod liver oil intake and incidence of asthma in Nor-
wegian adults—the HUNT study. Thorax. 2013; 68(1):25–30. Epub 2012/09/15. https://doi.org/10.
1136/thoraxjnl-2012-202061 PMID: 22977130.
38. Cade J, Thompson R, Burley V, Warm D. Development, validation and utilisation of food-frequency
questionnaires—a review. Public Health Nutr. 2002; 5(4):567–87. Epub 2002/08/21. https://doi.org/10.
1079/PHN2001318 PMID: 12186666.
PLOS ONE | https://doi.org/10.1371/journal.pone.0244089 December 17, 2020
14 / 14
PLOS ONE |
10.1371_journal.pone.0244789 | RESEARCH ARTICLE
Evaluation of vessel density in disorganization
of retinal inner layers after resolved diabetic
macular edema using optical coherence
tomography angiography
Gilda CennamoID
Fausto Tranfa2
1*, Daniela Montorio2, Federica Fossataro2, Claudia Fossataro2,
1 Eye Clinic, Public Health Department, University of Naples “Federico II”, Naples, Italy, 2 Department of
Neurosciences, Reproductive Sciences and Dentistry, University of Naples “Federico II”, Naples, Italy
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
* xgilda@hotmail.com
Abstract
Purpose
OPEN ACCESS
Citation: Cennamo G, Montorio D, Fossataro F,
Fossataro C, Tranfa F (2021) Evaluation of vessel
density in disorganization of retinal inner layers
after resolved diabetic macular edema using optical
coherence tomography angiography. PLoS ONE
16(1): e0244789. https://doi.org/10.1371/journal.
pone.0244789
Editor: Shree K. Kurup, University Hospitals
Cleveland, UNITED STATES
Received: February 24, 2020
Accepted: December 16, 2020
Published: January 12, 2021
Copyright: © 2021 Cennamo et al. This is an open
access article distributed under the terms of the
Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: All relevant data are
within the manuscript.
Funding: The author(s) received no specific
funding for this work.
Competing interests: The authors have declared
that no competing interests exist.
To evaluate the retinal vessel density (VD) in the macular region and the foveal avascular
zone (FAZ) area using optical coherence tomography angiography (OCTA) in patients with
and without disorganization of retinal inner layers (DRILs) after resolution of diabetic macu-
lar edema.
Methods
Thirty-seven eyes of 37 DRIL patients (mean age 63 ± 13.97 years), 30 eyes of 30 no DRIL
patients and 35 eyes of 35 controls were enrolled in this study. We evaluated the VD in the mac-
ular region of superficial capillary plexus (SCP), deep capillary plexus (DCP) and FAZ area.
Results
DRIL and no DRIL groups showed decreased VD in SCP and DCP (p<0.05) and a larger FAZ
area (p<0.001) compared to controls. However, DRIL patients revealed a statistically signifi-
cant reduction in VD of SCP (p = 0.041) and a greater FAZ area (p<0.001) with respect to no
DRIL patients. We found a significant negative correlation between the VD of the foveal SCP (r
= -0.414, p = 0.011), foveal DCP (r = -0.358, p = 0.025) and best corrected visual acuity
(BCVA) in DRIL group. Moreover there was a significant positive correlation between the FAZ
area (r = 0.425, p = 0.034) and BCVA. Therefore, in presence of DRILs lower VD values of
SCP and DCP and a larger FAZ area correlated with a worse visual acuity. In no DRILs group,
there was a significant negative correlation between the VD of the foveal DCP and BCVA.
Conclusion
OCTA highlights the role of retinal vascular ischemia in the pathogenesis of DRILs. This
parameter could represent an important functional predictive factor in diabetic patients.
PLOS ONE | https://doi.org/10.1371/journal.pone.0244789 January 12, 2021
1 / 9
PLOS ONE OCTA in DRILs
Introduction
Diabetic macular edema (DME) represents the main cause of visual loss in diabetic retinopa-
thy [1] due to presence of intra-retinal and subretinal fluid in macular region [2–4].
Spectral Domain Optical Coherence Tomography (SD-OCT) allows to evaluate the archi-
tecture of the retinal layers identifying the disorganization of the retinal inner layers (DRILs)
that consists of absence of identifiable boundaries between the ganglion cell-inner plexiform
layer complex, the inner nuclear layer and the outer plexiform layer [5, 6].
Previous studies reported the close correlation in diabetic retinopathy between DRILs and
macular capillary non-perfusion, detected by Fluorescein Angiography (FA). Furthermore, the
presence of DRILs was associated with reduction of visual acuity in patients with DME or with
resolved macular edema [7–11].
The evaluation of macular ischemia by FA, in presence of DRILs, did not allow a detailed
evaluation of the superficial capillary plexus (SCP) and deep capillary plexus (DCP) in macular
region [12–14].
The introduction of the OCT angiography (OCTA), a highly sensitive imaging technique,
provided a quantitative assessment of the retinal microvasculature changes in patients with
diabetic retinopathy [15–18].
The aim of this study was to analyze, using OCTA, the vessel density (VD) of SCP and DCP
in macular region and Foveal Avascular Zone (FAZ) area as well as the correlation between
these OCTA parameters and best corrected visual acuity (BCVA) in patients with and without
DRILs after resolved DME.
Materials and methods
Subjects
In this retrospective study, from 95 patients with type 2 diabetes mellitus and ocular history of
DME, a total of 67 eyes of 67 patients with resolved DME after anti-vascular endothelial
growth factor injections were enrolled from January 2018 to July 2019 at the Eye Clinic of the
University of Naples “Federico II”.
Each patient underwent BCVA evaluation according to the Early Treatment of Diabetic
Retinopathy Study (ETDRS) (the BCVA was converted into LogMAR scale for statistical calcu-
lations), slit-lamp biomicroscopy, fundus examination, SD-OCT (Spectralis + HRA; Heidel-
berg Engineering, Heidelberg, Germany) and OCTA (RTVue XR Avanti, Optovue, Inc.,
Freemont, California, USA).
DRIL was defined as disorganizations of the inner retinal layers and, more precisely,
as the inability to identify by SD-OCT the well-known-delineated boundaries between
the ganglion cell-inner plexiform layer, inner nuclear layer and outer plexiform layer
within the central 1500 μm region [7, 8, 10]. Patients showing resolved DME and DRIL
represented the DRIL group; patients with resolved DME and without alterations in
architecture of the retinal inner layers on SD-OCT B scans represented the no DRIL
group.
After a retrospective review of SD-OCT images, we divided the patients into two groups:
DRIL group that included 37 eyes of 37 patients (19 females, 18 males, mean age 63 ± 13.97
years) and no DRIL group including 30 eyes of 30 no patients (14 females, 16 males, mean age
62 ± 11.04).
In the DRIL group, 27 eyes showed non-proliferative diabetic retinopathy; while 10 eyes
presented proliferative diabetic retinopathy. In no DRIL group, 20 eyes and 10 eyes showed
non-proliferative and proliferative diabetic retinopathy, respectively.
PLOS ONE | https://doi.org/10.1371/journal.pone.0244789 January 12, 2021
2 / 9
PLOS ONE OCTA in DRILs
Thirty-five eyes of 35 healthy subjects (16 females, 19 males, mean age 62 ± 10.72 years)
served as control group, showing a normal ophthalmological evaluation, absence of vitreoret-
inal and vascular retinal diseases.
The evaluation of DRILs was performed by two masked examiners (FF, CF) and a senior
expert (GC).
Exclusion criteria were the following: clinically significant lens opacities, low-quality of
OCT and OCTA images, myopia greater than 6 diopters, presence of macular edema, subret-
inal or intraretinal fluid detected by SD-OCT with subfoveal thickness more than 300 μm [19],
outer retinal layer disruption, history of choroidal neovascularization, uveitis, and intraocular
surgery (including prior cataract surgery), absence of vitreoretinal, vascular retinal diseases
(such as retinal vein or artery occlusion), and congenital eye disorders.
The study was approved by the Institutional Review Board of the University of Naples “Fed-
erico II” and all investigations adhered to the tenets of the Declaration of Helsinki. Written
informed consents were obtained from the patients enrolled in the study.
Optical coherence tomography angiography
The XR Avanti AngioVue OCTA (software ReVue version 2017.1.0.151, Optovue Inc., Fre-
mont, CA, USA), is a device with a high speed of 70 000 axial scans per second that uses a light
source of 840 nm and an axial resolution of 5 μm. This system is based on split-spectrum
amplitude de-correlation algorithm (SSADA) which uses blood flow as intrinsic contrast. The
flow is detected as a variation over time in a speckle pattern formed by the interference of light
scattered by red blood cells and adjacent tissue structures [20].
The macular capillary network was visualized in scans centered on the fovea by per-
forming a 6 x 6 mm scan. For each eye enrolled, the AngioAnalyticTM software automati-
cally calculated vessel density in different vascular networks of the retina SCP and DCP,
analyzing the whole image, fovea and parafovea, according to the ETDRS classification of
diabetic retinopathy [21].
The boundaries of superficial network extended from 3 microns below the internal limiting
membrane to 15 microns below the inner plexiform layer (IPL). The deep capillary network
extended from 15 to 70 microns below the IPL [22].
The software automatically calculated vessel density in whole scan area and in all sections
of applied grid in different vascular networks of the retina. Vessel density (VD) was defined as
the percentage area occupied by vessels in the analyzed region [23].
The projection artefact removal software was used in order to ensure correct visualization
of the SCP and DCP [24].
Angiovue software automatically calculated the FAZ area in square millimetres over the
6 mm x 6 mm macular area in the full retinal plexus. The “non-flow” option was selected
from a drop-down menu and the area was automatically selected when the FAZ area was
clicked [25].
The images that presented a signal strength index less than 40, motion artefacts, incorrect
segmentation, low centration and focus were not considered for the analysis.
Statistical analysis
SPSS version 17.0 (SPSS Inc, Chicago, Ill, USA) was used for statistical analysis.
The Analysis of Variance (ANOVA) and chi-squared test were performed to evaluate the
differences between groups in terms of age, BCVA and sex, respectively.
ANOVA with Bonferroni corrected for multiple comparisons was used to evaluate the dif-
ferences in VD in each retinal vascular plexus, among DRIL, no DRIL groups and controls.
PLOS ONE | https://doi.org/10.1371/journal.pone.0244789 January 12, 2021
3 / 9
PLOS ONE OCTA in DRILs
Pearson’s correlation was assessed between OCT angiography parameters (VD in SCP,
DCP and FAZ area) and BCVA. A p value lower than < 0.05 was considered statistically
significant.
Results
DRIL patients, no DRIL patients and controls were not significantly different in terms of age
(p = 0.922) and sex (p = 0.877). Conversely, BCVA was significantly different among the three
groups (p<0.001). No significant association was found between the presence of DRIL or no
DRIL and the stage of the diabetic retinopathy (p = 0.574). The demographic and clinical char-
acteristics of the controls and patients was reported in Table 1.
Compared to controls, DRIL and no DRIL groups revealed a statistically significant reduc-
tion in VD of the SCP and DCP in the whole image, fovea and parafovea (p<0.001). Moreover,
a statistically significant enlargement of the FAZ area was found in both groups with respect to
controls (p<0.001). Lastly, DRIL patients showed significantly lower values of VD in SCP
whole image (p = 0.041) and a greater FAZ area (p<0.001) compared to no DRILs patients
(Fig 1).
The differences in VD and FAZ area values between the three groups are shown in Table 2.
In the DRIL group we found a significant negative correlation between the VD of the foveal
SCP (r = -0.414, p = 0.011), foveal DCP (r = -0.358, p = 0.025) and BCVA and a significant pos-
itive correlation between the FAZ area (r = 0.425, p = 0.034 showed) and BCVA. Therefore,
lower VD values of SCP and DCP and larger FAZ area correlated with a worse visual acuity. In
no DRILs group there was a significant negative correlation between the VD of the foveal DCP
and BCVA (Table 3).
Discussion
To the best of our acknowledge, this is the first study evaluating the VD of the SCP and the
DCP in macular region and the FAZ area in patients with DRILs and no DRILs after resolved
DME.
Table 1. Demographic and clinical characteristics in controls, DRIL and No DRIL group.
Eyes (n.)
Female/Male
Age (years)
BCVA, LogMar (Snellen)
DM duration (years)
Glycated hemoglobin level (%)
Nonproliferative diabetic retinopathy (eyes n.)
Proliferative diabetic retinopathy (eyes n.)
Treatment
Anti-VEGF injections (n.)
Controls
35
16/19
62 ± 10.72
0.06 ± 0.08 (20/20)
-
-
-
-
-
Dril Group
No Dril Group
AnovaP
37
19/18
63 ± 13.97
0.71 ± 0.29 (20/100)
24 ± 11.13
7.2 ± 1.08
27
10
3.8 ± 1.5
30
14/16
62 ± 11.04
0.47 ± 1.65 (20/50)
23.6 ± 9.35
7.3 ± 1.09
20
10
3.2 ± 1.5
0.877
0.922
<0.001
-
-
-
DM: Diabetes Mellitus; VEGF: Vascular Endothelial Growth Factor; BCVA: Best-Corrected Visual Acuity; LogMAR: logarithm of the minimum angle of resolution
(Snellen equivalent inbrackets).
Data expressed as mean ± SD
Anova: Analysis of Variance with Bonferroni correction.
Bold p-values are significant (p <0.05).
https://doi.org/10.1371/journal.pone.0244789.t001
PLOS ONE | https://doi.org/10.1371/journal.pone.0244789 January 12, 2021
4 / 9
PLOS ONE OCTA in DRILs
Fig 1. Multicolor image, Spectral Domain-Optical Coherence Tomography (SD-OCT), Optical Coherence
Tomography Angiography (OCTA) in superficial capillary plexus (SCP), deep capillary plexus (DCP) and foveal
avascular zone (FAZ) in a healthy control (top row, A-E), DRIL (disorganized inner retinal layer) patient (middle row,
A1-E1) and no DRIL patient (bottom row, A2-E2). SD-OCT shows the area of DRIL (arrow) (B1) and well-delineated
boundaries in inner retinal layers in no DRIL patient (B2) and in control (B). OCTA shows a reduction in vessel
density in SCP, DCP and a greater FAZ area in DRIL (C1, D1, E1) and no DRIL (C2, D2, E2) compared control (C1,
D1, E1).
https://doi.org/10.1371/journal.pone.0244789.g001
Sun et al. supported, for the first time, the mechanical pathogenesis of DRILs due to macu-
lar edema that could cause the stretching of neuronal cells [6].
Whereas, Nicholson et al. proposed an ischemic hypothesis describing a significant associa-
tion between the presence of DRILs and the macular non perfusion, evaluated by FA [7].
DRIL represents an OCT biomarker, able to predict the BCVA in patients with DME [10];
a study conducted by Radwan et al. showed the correlation between the length of DRIL and
visual changes [25].
In our study, through a quantitative and detailed analysis of the retinal vascular networks
by OCTA, we hypothesized that retinal ischemia plays a crucial role in the development of
DRILs. Furthermore, we found a significant correlation between the presence of this OCT
parameter and visual acuity loss.
Table 2. Differences in OCT angiography parameters among controls, Dril and No Dril groups.
Controls
Dril Group
P�
No Dril Group
P†
P‡
Anova P
Superficial Capillary Plexus (%)
Whole image
Fovea
Parafovea
Deep Capillary Plexus (%)
Whole image
Fovea
Parafovea
Foveal Avascular Zone (mm2)
48.51 ± 4.32
26.12 ± 8.68
48.47 ± 7.83
41.15 ± 6.13
16.08 ± 6.93
42.14 ± 6.52
50.58 ± 4.37
37.93 ± 8.11
42.16 ± 6.02
30.67 ± 9.13
51.79 ± 5.49
46.43 ± 5.53
0.18 ± 0.06
0.65 ± 0.09
<0.001
<0.001
<0.001
<0.001
0.002
<0.001
<0.001
44.17 ± 3.69
17.26 ± 8.38
43.48 ± 5.26
43.72 ± 4.43
32.20 ± 9.55
47.94 ± 4.27
0.33 ± 0.14
0.002
0.041
<0.001
0.010
1
1
<0.001
0.034
0.011
<0.001
0.634
1
0.714
<0.001
<0.001
<0.001
<0.001
<0.001
0.002
<0.001
<0.001
Data expressed as mean ± SD.
Anova: Analysis of Variance with Bonferroni correction.
P�: Controls vs Dril Group.
P†: controls vs No Dril Group.
P‡: Dril Group vs No Dril Group.
Bold p-values are significant (p <0.05).
https://doi.org/10.1371/journal.pone.0244789.t002
PLOS ONE | https://doi.org/10.1371/journal.pone.0244789 January 12, 2021
5 / 9
PLOS ONE Table 3. Correlations between BCVA and OCTA parameters in Dril and No Dril Groups.
BCVA
r
DRIL GROUP
Superficial Capillary Plexus
Whole
Fovea
Parafovea
Deep Capillary Plexus
Whole
Fovea
Parafovea
Foveal Avascular Zone
NO DRIL GROUP
Superficial Capillary Plexus
Whole
Fovea
Parafovea
Deep Capillary Plexus
Whole
Fovea
Parafovea
Foveal Avascular Zone
-0.207
-0.414
-0.212
-0.290
-0.358
-0.214
0.425
-0.108
-0.064
-0.017
-0.207
-0.254
-0.094
0.055
OCTA in DRILs
p
0.219
0.011
0.208
0.082
0.025
0.203
0.034
0.572
0.736
0.928
0.273
0.047
0.623
0.772
Pearson’s correlation p<0.05.
Bold p-values are significant (p <0.05).
https://doi.org/10.1371/journal.pone.0244789.t003
Similar to the results reported by Moein et al., we found a significant increase in FAZ area
in the DRIL group with respect to the no DRIL group and controls. Moreover, in patients with
DRIL the larger FAZ area significantly correlated with lower BCVA while in the no DRIL
group this relationship was not statistically significant [26].
In addition, DRIL and no DRIL patients showed a lower VD in both retinal vascular net-
works compared to controls, as also confirmed by Onishi et al. who found perfusion deficits in
both SCP and DCP in DRILs areas showing a multilevel capillary non-perfusion in DRIL path-
ogenesis [27].
When comparing DRIL and no DRIL groups, we noticed that the first group showed a sta-
tistically significant reduction in VD of the SCP with respect to the second one. Thus, an
altered perfusion of SCP, the main vascular network of the inner retinal structures [28], could
contribute to the development of changes in the inner retinal architecture, supporting the vas-
cular pathogenesis hypothesis.
In the correlation study, indeed, we reported a significant association between lower VD of
SCP and DCP and decreased BCVA in the DRIL group, while only lower VD of DCP corre-
lated with reduced BCVA in the no DRIL group, showing the possible influence of the vascular
damage on vision loss in presence of DRILs.
We hypothesized that, in addition to the reduced perfusion of DCP mainly involved in dia-
betic retinopathy [29], lower VD of SCP in DRILs may cause an impairment of the metabolic
activity in retinal inner layers, where the cells involved in synaptic transmission processes are
located, contributing to the functional damage [30].
PLOS ONE | https://doi.org/10.1371/journal.pone.0244789 January 12, 2021
6 / 9
PLOS ONE OCTA in DRILs
Possible limitations of this study are represented by its retrospective nature and by the rela-
tively small number of eyes examined in each group.
In conclusion, our findings confirmed that the retinal non-perfusion in macular region
plays an important role in the pathogenesis of DRILs.
OCTA provides an objective and quantitative analysis of VD, useful to better understand
the changes in retinal vascular network occurring in DRILs and could represent an important
functional predictive tool for patients with resolved DME.
Acknowledgments
We thank Laura Connolly (Union Jack School, Naples, Italy) for language assistance.
Author Contributions
Conceptualization: Gilda Cennamo.
Data curation: Federica Fossataro, Claudia Fossataro.
Formal analysis: Daniela Montorio.
Investigation: Daniela Montorio, Federica Fossataro, Claudia Fossataro.
Methodology: Gilda Cennamo, Daniela Montorio.
Supervision: Gilda Cennamo, Fausto Tranfa.
Validation: Gilda Cennamo.
Writing – original draft: Daniela Montorio.
Writing – review & editing: Gilda Cennamo, Federica Fossataro, Claudia Fossataro.
References
1. Klein R, Klein BE, Moss SE, Davis MD, DeMets DL. The Wisconsin epidemiologic study of diabetic reti-
nopathy. IV. Diabetic macular edema. Ophthalmology 1984; 91:1464–1474. https://doi.org/10.1016/
s0161-6420(84)34102-1 PMID: 6521986
2. Moss SE, Klein R, Klein BE. The 14-year incidence of visual loss in a diabetic population. Ophthalmol-
ogy 1998; 105:998–1003. https://doi.org/10.1016/S0161-6420(98)96025-0 PMID: 9627648
3. Sakata K, Funatsu H, Harino S, Noma H, Hori S. Relationship between macular microcirculation and
progression of diabetic macular edema. Ophthalmology 2006; 113:1385–1391. https://doi.org/10.1016/
j.ophtha.2006.04.023 PMID: 16877077
4. Sim DA, Keane PA, Zarranz-Ventura J, Fung S, Powner MB, Platteau E, et al. The effects of macular
ischemia on visual acuity in diabetic retinopathy. Invest Ophthalmol Vis Sci 2013: 54:2353–2360.
https://doi.org/10.1167/iovs.12-11103 PMID: 23449720
5. Browning DJ, Glassman AR, Aiello LP, et al. Diabetic Retinopathy Clinical Research Network. Relation-
ship between optical coherence tomography-measured central retinal thickness and visual acuity in dia-
betic macular edema. Ophthalmology 2007; 114:525–536. https://doi.org/10.1016/j.ophtha.2006.06.
052 PMID: 17123615
6. Sun JK, Lin MM, Lammer J, Prager S, Sarangi R, Silva PS, et al. Disorganization of the retinal inner lay-
ers as a predictor of visual acuity in eyes with center-involved diabetic macular edema. JAMA Ophthal-
mol 2014; 132:1309–1316. https://doi.org/10.1001/jamaophthalmol.2014.2350 PMID: 25058813
7. Nicholson L, Ramu J, Triantafyllopoulou I, Patrao NV, Comyn O, Hykin P, et al. Diagnostic accuracy of
disorganization of the retinal inner layers in detecting macular capillary non-perfusion in diabetic retinop-
athy. Clin Exp Ophthalmol 2015; 43:735–741. https://doi.org/10.1111/ceo.12557 PMID: 25998983
8. Sun JK, Radwan SH, Soliman AZ, Lammer J, Lin MM, Prager SG, et al. Neural retinal disorganization
as a robust marker of visual acuity in current and resolved diabetic macular edema. Diabetes. 2015;
64:2560–2570. https://doi.org/10.2337/db14-0782 PMID: 25633419
9. Babiuch A, Han M, Conti F, Wai K, Silva FQ, Singh RP. Association of disorganization of retinal inner
layers with visual acuity response to anti-vascular endothelial growth factor therapy for macular edema
PLOS ONE | https://doi.org/10.1371/journal.pone.0244789 January 12, 2021
7 / 9
PLOS ONE OCTA in DRILs
secondary to retinal vein occlusion. JAMA Ophthalmol 2019; 137:38–46. https://doi.org/10.1001/
jamaophthalmol.2018.4484 PMID: 30286219
10. Radwan SH, Soliman AZ, Tokarev J, Zhang L, van Kuijk FJ, Koozekanani DD. Association of disorgani-
zation of retinal inner layers with vision after resolution of center-involved diabetic macular edema.
JAMA Ophthalmol 2015; 133:820–825. https://doi.org/10.1001/jamaophthalmol.2015.0972 PMID:
25950417
11. Das R, Spence G, Hogg RE, Stevenson M, Chakravarthy U. Disorganization of inner retina and outer
retinal morphology in diabetic macular edema. JAMA Ophthalmol 2018; 136:202–208. https://doi.org/
10.1001/jamaophthalmol.2017.6256 PMID: 29327033
12. Spaide RF, Fujimoto JG, Waheed NK, Sadda SR, Staurenghi G. Optical coherence tomography angi-
ography. Prog Retin Eye Res. 2018; 64:1–55. https://doi.org/10.1016/j.preteyeres.2017.11.003 PMID:
29229445
13. Coscas G, Lupidi M, Coscas F. Optical Coherence Tomographic Angiography in Diabetic Macular
Ischemia: A New Step Forward. JAMA Ophthalmol. 2016: 134:373–374. https://doi.org/10.1001/
jamaophthalmol.2015.4821 PMID: 26795200
14. Durbin MK, An L, Shemonski ND, Soares M, Santos T, Lopes M, et al. Quantification of Retinal Micro-
vascular Density in Optical Coherence Tomographic Angiography Images in Diabetic Retinopathy.
JAMA Ophthalmol. 2017; 135: 370–376. https://doi.org/10.1001/jamaophthalmol.2017.0080 PMID:
28301651
15. Chalam KV, Sambhav K. Optical coherence tomography angiography in retinal diseases. J Opthalmic
Vis Res 2016; 11:84–92. https://doi.org/10.4103/2008-322X.180709 PMID: 27195091
16. Agrawal R, Xin W, Keane PA, Chhablani J, Agarwal A. Optical coherence tomography angiography: a
non-invasive tool to image end-arterial system. Expert rev Med Devices. 2016; 13:519–521. https://doi.
org/10.1080/17434440.2016.1186540 PMID: 27176114
17. Cennamo G, Lanni V, Abbate R, Velotti N, Fossataro F, Sparnelli F, et al. The Relationship between
Macular Pigment and Vessel Density in Patients with Type 1 Diabetes Mellitus. Ophthalmic Res. 2019;
61:19–25. https://doi.org/10.1159/000492897 PMID: 30308482
18. Kim AY, Chu Z, Shahidzadeh A, Wang RK, Puliafito CA, Kashani AH. Quantifying microvascular density
and morphology in diabetic retinopathy using spectral-domain optical coherence tomography angiogra-
phy. Invest Ophthalmol Vis Sci. 2016; 57:OCT362–370. https://doi.org/10.1167/iovs.15-18904 PMID:
27409494
19. Radwan SH, Soliman AZ, Tokarev J, Zhang L, van Kuijk FJ, Koozekanani DD. Association of Disorgani-
zation of Retinal Inner Layers With Vision After Resolution of Center-Involved Diabetic Macular Edema.
JAMA Ophthalmol. 2015; 133:820–825. https://doi.org/10.1001/jamaophthalmol.2015.0972 PMID:
25950417
20.
Jia Y, Tan O, Tokayer J, Potsaid B, Wang Y, Liu JJ, et al. Split spectrum amplitude-decorrelation angi-
ography with optical coherence tomography. Opt Express. 2012; 20:4710–4725. https://doi.org/10.
1364/OE.20.004710 PMID: 22418228
21. Cennamo G, Montorio D, Mirra F, Comune C, D’Alessandro A, Tranfa F. Study of vessel density in
adult-onset foveomacular vitelliform dystrophy with Optical Coherence Tomography Angiography.
Photodiagnosis Photodyn Ther. 2020;101702. https://doi.org/10.1016/j.pdpdt.2020.101702 PMID:
32126304
22. Samara WA, Shahlaee A, Sridhar J, Khan MA, Ho AC, Hsu J. Quantitative optical coherence tomogra-
phy angiography features and visual function in eyes with branch retinal vein occlusion. Am J Ophthal-
mol 2016; 166:76–83. https://doi.org/10.1016/j.ajo.2016.03.033 PMID: 27038893
23. Huang D, Jia Y, Gao SS, Lumbroso B, Rispoli M. Optical coherence tomography angiography using the
optovue device. Dev Ophthalmol. 2016; 56:6–12. https://doi.org/10.1159/000442770 PMID: 27022989
24. Mastropasqua L, Di Staso S, D’Aloisio R, Mastropasqua A, Di Antonio L, Senatore A, et al. Anatomical
and functional changes after dexamethasone implant and ranibizumab in diabetic macular edema: a
retrospective cohort study. Int J Ophthalmol. 2019; 12:1589–1597. https://doi.org/10.18240/ijo.2019.
10.11 PMID: 31637195
25. Radwan SH, Soliman AZ, Tokarev J, Zhang L, van Kuijk FJ, Koozekanani DD. Association of Disorgani-
zation of Retinal Inner Layers With Vision After Resolution of Center-Involved Diabetic Macular Edema.
JAMA Ophthalmol. 2015; 133:820–5. https://doi.org/10.1001/jamaophthalmol.2015.0972 PMID:
25950417
26. Moein HR, Novais EA, Rebhun CB, Cole ED, Louzada RN, Witkin AJ, et al. Optical coherence tomogra-
phy angiography to detect macular capillary ischemia in patients with inner retinal changes after
resolved diabetic macular edema. Retina. 2018; 38: 2277–2284. https://doi.org/10.1097/IAE.
0000000000001902 PMID: 29068912
PLOS ONE | https://doi.org/10.1371/journal.pone.0244789 January 12, 2021
8 / 9
PLOS ONE OCTA in DRILs
27. Onishi A, Ashraf M, Soetikno B, et al. Multilevel ischemia in disorganization of the retinal inner layers on
projection-resolved optical coherence tomography angiography. Retina. 2019; 39:1588–1594. https://
doi.org/10.1097/IAE.0000000000002179 PMID: 29642239
28.
Lavia C, Couturier A, Erginay A, Dupas B, Tadayoni R, Gaudric A. Reduced vessel density in the super-
ficial and deep plexuses in diabetic retinopathy is associated with structural changes in corresponding
retinal layers. PLoS One. 2019; 14:e0219164. https://doi.org/10.1371/journal.pone.0219164 PMID:
31318880
29. Nesper PL, Roberts PK, Onishi AC, Chai H, Liu L, Jampol LM, et al. Quantifying Microvascular Abnor-
malities With Increasing Severity of Diabetic Retinopathy Using Optical Coherence Tomography Angi-
ography. Invest Ophthalmol Vis Sci. 2017; 58:BIO307–315. https://doi.org/10.1167/iovs.17-21787
PMID: 29059262
30. Balaratnasingam C, Inoue M, Ahn S, McCann J, Dhrami-Gavazi E, Yannuzzi LA, et al. Visual Acuity Is
Correlated with the Area of the Foveal Avascular Zone in Diabetic Retinopathy and Retinal Vein Occlu-
sion. Ophthalmology. 2016; 123:2352–2367. https://doi.org/10.1016/j.ophtha.2016.07.008 PMID:
27523615
PLOS ONE | https://doi.org/10.1371/journal.pone.0244789 January 12, 2021
9 / 9
PLOS ONE |
10.1371_journal.pone.0256592 | RESEARCH ARTICLE
A comparison of prediction approaches for
identifying prodromal Parkinson disease
Mark N. WardenID
Brad A. RacetteID
1, Susan Searles Nielsen1, Alejandra Camacho-Soto1, Roman Garnett2,
1,3*
1 Department of Neurology, Washington University School of Medicine, Saint Louis, Missouri, United States
of America, 2 Department of Computer Science and Engineering, Washington University in Saint Louis, Saint
Louis, Missouri, United States of America, 3 Faculty of Health Sciences, School of Public Heath, University of
the Witwatersrand, Johannesburg, South Africa
* racetteb@wustl.edu
Abstract
Identifying people with Parkinson disease during the prodromal period, including via algo-
rithms in administrative claims data, is an important research and clinical priority. We sought
to improve upon an existing penalized logistic regression model, based on diagnosis and
procedure codes, by adding prescription medication data or using machine learning. Using
Medicare Part D beneficiaries age 66–90 from a population-based case-control study of inci-
dent Parkinson disease, we fit a penalized logistic regression both with and without Part D
data. We also built a predictive algorithm using a random forest classifier for comparison. In
a combined approach, we introduced the probability of Parkinson disease from the random
forest, as a predictor in the penalized regression model. We calculated the receiver operator
characteristic area under the curve (AUC) for each model. All models performed well, with
AUCs ranging from 0.824 (simplest model) to 0.835 (combined approach). We conclude
that medication data and random forests improve Parkinson disease prediction, but are not
essential.
Introduction
Parkinson disease (PD) is a progressive, neurodegenerative disorder that is diagnosed when
patients experience motor symptoms such as resting tremor, bradykinesia, rigidity, and pos-
tural instability. However, before these motor symptoms fully manifest, patients may experi-
ence a variety of non-motor symptoms, including cognitive and mood dysfunction, sleep
disorders, and varying degrees of autonomic dysfunction [1–5]. This period of disease is
termed the “prodromal period” and may provide a critical window of opportunity during
which providers could identify PD patients. In particular, earlier recognition of PD might both
facilitate the identification of disease-modifying medications, as well as their initiation, when
available. Moreover, even without such treatments yet available, earlier identification of PD is
essential. During the prodromal disease window, many PD patients experience potentially pre-
ventable fall-related morbidity, including substantial excesses of both traumatic brain injuries
[6, 7] and fractures [8, 9] relative to comparable individuals without PD.
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: Warden MN, Searles Nielsen S,
Camacho-Soto A, Garnett R, Racette BA (2021) A
comparison of prediction approaches for
identifying prodromal Parkinson disease. PLoS
ONE 16(8): e0256592. https://doi.org/10.1371/
journal.pone.0256592
Editor: Thippa Reddy Gadekallu, Vellore Institute of
Technology: VIT University, INDIA
Received: January 14, 2021
Accepted: August 10, 2021
Published: August 26, 2021
Copyright: © 2021 Warden et al. This is an open
access article distributed under the terms of the
Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: The data underlying
the results presented in this study are from the
Centers for Medicare and Medicaid Services and
CMS does not permit data sharing as per their
legally binding and standard data use agreements.
The exact data used in this study can be purchased
directly from the Centers for Medicare and
Medicaid Services (https://www.cms.gov/
Research-Statistics-Data-and-Systems/Research-
Statistics-Data-and-Systems).
PLOS ONE | https://doi.org/10.1371/journal.pone.0256592 August 26, 2021
1 / 13
PLOS ONE Funding: BAR: Michael J. Fox Foundation grant
#10289 (https://www.michaeljfox.org/); National
Institute of Environmental Health Sciences
K24ES017765 (https://www.niehs.nih.gov/);
Department of Defense PD190057 (https://cdmrp.
army.mil/default); SSN: National Institute of
Environmental Health Sciences K01ES028295
(https://www.niehs.nih.gov/).The funders had no
role in study design, data collection and analysis,
decision to publish, or preparation of the
manuscript.
Competing interests: I have read the journal’s
policy and the authors of this manuscript have the
following competing interests: Dr. Racette serves
on the National Advisory Environmental Health
Sciences Council for the National Institute for
Environmental Health Sciences (NIEHS) for which
he is reimbursed for his time. The NIEHS had no
input or influence on the content of this
manuscript.
Parkinson disease predictive algorithms
Towards these ends, researchers have begun to move beyond traditional predictive model-
ing approaches by applying machine learning methods to a wide variety of data. Several inves-
tigators have used machine learning methods to distinguish PD patients from controls, using
data obtained from both wearable and non-wearable sensors [10, 11]. While these methods
have primarily been used to distinguish newly diagnosed PD patients from controls, other
studies were able to distinguish people with potential prodromal PD symptoms, such as hypos-
mia, from controls [11, 12]. Although these people do have a greater risk of developing PD,
this group remains heterogeneous, and there is no “ideal” prodromal PD population. In con-
trast, retrospective cohort studies using predictor data from the prodromal PD time window
afford an opportunity to confirm the PD diagnosis, while providing potentially extensive vari-
ables to include in predictive models.
Medicare claims are a rich source of population-based data to predict which patients will be
diagnosed eventually with PD. We previously developed a PD predictive model using Medi-
care claims data, specifically diagnosis and procedure codes, from the five years prior to PD
diagnosis [13]. This model contained 536 diagnoses and medical procedures as predictors and
achieved an AUC of 0.857, much higher than the AUC of 0.670 achieved with known demo-
graphic and medical predictors of PD. At the optimal cut point, sensitivity was 73.5% and
specificity was 83.2%. While this least absolute shrinkage and selection operator (LASSO)
penalized regression model performed well, the addition of Medicare Part D prescription med-
ication data or the use of other analytic methods, such as machine learning methods, may have
the potential to improve model performance. The current study builds upon our previous
work by considering whether the addition of prescription medication data improves discrimi-
nation and whether a random forest classifier could perform better or help improve the origi-
nal penalized regression approach [13]. Attempting to improve the model is the logical next
step, since we recently validated our original predictive model in a population-based sample
followed forward for PD [14]. We hypothesized that inclusion of prescription medication data
would improve model performance for four reasons: 1) these medication data offer an alterna-
tive way to capture information available from diagnosis codes, which could be incomplete; 2)
medication data might provide diagnostic confirmation and evidence of disease severity; 3)
medications might serve as proxies for biologic pathways that might be predictive of PD; and
4) some medications might increase or decrease risk of PD, regardless of the indication for the
medication, and thus could be independently predictive. Random forest classifiers use a
completely different methodology than penalized regression. Therefore, we sought to deter-
mine if this innovative approach could outperform or possibly enhance the previous penalized
regression model by introducing the probability from the random forest as a predictor in the
penalized logistic regression model. We were able to demonstrate modest improvements in
model performance.
Methods
Standard protocol approvals
This study was approved by the Washington University School of Medicine Human Research
Protection Office and the Centers for Medicare and Medicaid Services.
Study participants
This was a population-based case-control study using Medicare administrative claims data.
Briefly, all participants were U.S. residents age 66–90 years old and relying solely on Medicare
in 2009. Medicare is the only nationwide health insurance coverage universally available in the
U.S., specifically among those age 65 and older. In this age group >98% of Americans
PLOS ONE | https://doi.org/10.1371/journal.pone.0256592 August 26, 2021
2 / 13
PLOS ONE Parkinson disease predictive algorithms
participate in Medicare Part A/B, which provides medical coverage. From all of these benefi-
ciaries, we identified those who met all study eligibility criteria (age 66–90, no non-Medicare
insurance coverage, and U.S. residence) for the year 2009 using the Medicare “base file.” We
then included all incident PD cases and a random sample of comparable beneficiaries as con-
trols who also had Medicare Part D (pharmacy) coverage. We determined PD case status from
complete Part A and B Medicare claims data for 2004–2009, with cases identified as having at
least one International Classification of Diseases, Ninth Revision, Clinical Modification
(ICD9) code for PD (332 or 332.0) in 2009 but no prior year, and no code for atypical parkin-
sonism or Lewy body dementia. Controls met these same study eligibility criteria, except that
they had no ICD9 code for PD, and were alive in 2009 prior to their randomly assigned refer-
ence date (comparable to the cases’ diagnosis dates). The original study included 89,790 cases
and 118,095 controls. From this original group of participants, we further restricted to the
48,295 (54%) of cases and 52,324 (44%) of controls who were also enrolled in Medicare Part D
and had at least one medication filled under this coverage in 2008–2009. After review of medi-
cations taken by the PD patients, we excluded 12,354 cases who had filled a prescription for a
medication known to cause secondary parkinsonism (aripiprazole, chlorpromazine, fluphen-
azine, haloperidol, loxapine, metoclopramide, molindone, olanzapine, paliperidone, perphe-
nazine, pimozide, prochlorperazine, promethazine, quetiapine [if > 100 mg], reserpine,
risperidone, tetrabenazine, thioridazine, thiothixene, trifluoperazine, trimethobenzamide and/
or ziprasidone) within the 6 months prior to their PD diagnosis in 2009 [15]. This left a total of
35,941 PD cases and 52,324 controls for the present work. We formally divided these partici-
pants into a 90% training dataset and 10% test dataset by stratified random sampling (by case
status), such that we had 90% cases and 90% controls in our training set for developing the
models, and 10% cases and 10% controls in our test set for assessing model performance.
Calculation of predictor variables
We calculated predictor variables, as previously [13, 16]. In total, during the development of
the original predictive model there were 26,468 valid codes (11,063 diagnoses and 15,405 pro-
cedures, including ICD9 procedure codes and Healthcare Common Procedure Coding System
[HCPCS] codes mainly comprised of Current Procedural Terminology [CPT] codes). CPT
codes are part of a formal coding system for billing that encompasses surgical and more minor
procedures that physicians perform in the office, along with some radiology and laboratory
tests, in contrast to ICD9 procedure codes used by hospitals. HCPCS codes are similar to CPT
codes but are specific to Medicare. For ICD9/procedure codes recorded for > 10 PD cases, the
median time between receiving the code and PD diagnosis was 2.41 years. This period was
nearly identical to the median time for the 536 ICD9/procedure codes selected for our original
predictive model ultimately (2.42 years), However, the median time for diagnosis codes indica-
tive of cardinal signs of PD was shorter: 1.51 years for ICD9 333.1 (tremor), 1.98 years for
ICD9 781.2 (gait abnormality), 1.09 years for ICD9 781.0 (abnormal involuntary movement),
and 1.44 years for ICD9 781.3 (lack of coordination). We calculated age and obtained sex and
race/ethnicity from the 2009 beneficiary annual summary file. Given the importance of smok-
ing on PD risk [17], we derived a probability of ever having regularly smoked for each partici-
pant using a logistic regression model built from nationwide data [13, 16]. We previously also
identified that overall use of medical care is an important predictor of PD and included this
variable in our models [13, 18].
Building upon the above data from the beneficiary annual summary file and Part A and B
claims that were available to us when we developed our original PD predictive model, we
obtained Medicare Part D prescription data from 2008–2009, i.e., in the one to two years prior
PLOS ONE | https://doi.org/10.1371/journal.pone.0256592 August 26, 2021
3 / 13
PLOS ONE Parkinson disease predictive algorithms
to PD diagnosis, for use in our predictive models. We derived prescription data from a shorter
pre-diagnosis period than for our other claims data because Part D coverage first became avail-
able in late 2006. For each medication, we identified all associated active ingredients and cre-
ated a dichotomous variable representing whether a pharmacy filled a prescription claim for a
medication containing the active ingredient at any time during this period prior to the PD
diagnosis/control reference date. There were 880 active ingredients represented in these pre-
scription claims data. We did not include 31 active ingredients that could be used to treat PD
(carbidopa-levodopa, pramipexole, ropinirole, entacapone, tolcapone, selegiline, rasagline, tri-
hexiphenidyl, benztropine) or that could cause secondary parkinsonism (22 listed above).
Model building approach
We built all models within the training set (90% stratified random sample) using R version
3.5.0. For all models, we used a two-step model building approach with the same first step for
all. In this first step, we identified diagnosis/procedure codes and active ingredients associated
with PD using multivariable logistic regression. For each code and active ingredient, we fit a
logistic regression model adjusting a priori for age (modeled as a two-part linear spline with a
knot at age 85), sex, race/ethnicity (7 categories [6 dummy variables]), probability of ever
smoking (continuous), and number of unique diagnosis codes (continuous) [18]. These con-
stitute the 11 forced demographic predictors. We then used the Bonferroni correction for mul-
tiple comparisons to select a subset of all codes and active ingredients still significantly
associated with PD to consider in the second step of the model building. This prescreening
retained 983 codes and active ingredients, after we excluded ten that effectively were sex-spe-
cific, i.e. acting as a proxy for the patient’s sex.
Starting with the preselected set of predictor variables from the first step, i.e. the 983 codes/
active ingredient variables and the 11 forced demographic variables, we proceeded to the sec-
ond step, which differed for each model. We produced three models (fit three predetermined
classifiers): two penalized logistic regression models [13] (one with and one without prescrip-
tion medications) and a random forest that considered the prescription medications.
For the penalized logistic regressions, we built the models using only the LASSO regression
using the R package glmnet [19, 20]. In our previous work, we determined that LASSO alone
(i.e., α = 1) produced the optimal model as part of the elastic net algorithm [13]. This proce-
dure selects variables and regularizes coefficients based on penalties for possible overfitting.
The method is particularly suitable for high dimensional data, using ten-fold cross validation
to determine the shrinkage parameter (λ), and improves external validity. We used the area
under the receiver operator characteristic curve (AUC) as the measure of model quality for
selecting λ.
For the random forest, we used the R packages randomForest [21] and varSelRF [22],
which is a variable selection package designed for random forests. Specifically, we used a previ-
ously developed variable selection procedure [23]. Briefly, one large random forest was trained
on the full 90% training set using all 983 predictors and 11 demographic variables. The predic-
tor importance matrix, which contained the mean, un-scaled decrease in prediction accuracy
after variable permutation, was estimated once. Then, the 20% of predictors with the lowest
importance were dropped, and a new forest was trained on this smaller subset. The process
was repeated iteratively, while always using the original importance matrix, until only two pre-
dictor variables remained, i.e., 96 times in the present work. Each smaller subset is contained
within all larger subsets, and the predictor subset that generated the lowest “out of bag” error
was used to construct the final, predetermined random forest classifier. Random forests have
several strengths compared with support vector machines that are beneficial in this
PLOS ONE | https://doi.org/10.1371/journal.pone.0256592 August 26, 2021
4 / 13
PLOS ONE Parkinson disease predictive algorithms
application, including: 1) a useful, published feature selection method comparable to the
LASSO approach [23]; 2) the ability to handle many categorical and/or irrelevant features; 3)
automatic feature relevance determination; and 4) an exceptional generalization performance
on a wide range of tasks [24]. The first three of these are critical for our data and goals with
this study. Additionally, in other machine learning applications in PD, random forests have
consistently performed well [10, 25].
After we completed both the random forest and penalized logistic regression models, we
also experimented with using both approaches (penalized regression and random forest)
simultaneously to produce a single, combined classifier. For this, we fit a penalized logistic
regression model that also used the probability of PD generated by the final random forest as a
predictor. The random forest’s probability might be able to act like a case preprocessing filter,
allowing the penalized regression to detect more complex relationships akin to the strategy of
convolution neural networks [26] and the strategy used in Amoroso et al. (2018) [27]. We
again started with the preselected set of predictor variables from the first step but included the
prediction probabilities from the final random forest classifier as a variable that could be
selected.
Finally, given how close to PD diagnosis the cardinal signs were first coded, we repeated all
analyses while utilizing predictor variables that we calculated as of the timepoint one year
prior to PD diagnosis/control reference. Specifically, we applied a one-year lag.
Assessment of model performance
We formally assessed the performance of all models in the test set (10% stratified random sam-
ple). We were able to separate the model building step from the model diagnostic step in this
way because of the size of the available data, allowing for a clean and straightforward interpre-
tation of the test set, as if it were an external dataset. We applied each of the above models
(three primary models and one combined model) to this test dataset. Then, with PD case status
in this test set as the gold standard, we used R to calculate three summary measures of model
performance [28]: the sensitivity at the cut point that correctly classified the most beneficiaries
in the test set, the specificity at that cut point, and the AUC. We also repeated these calcula-
tions at Youden’s Index [29], the point at which the sum of sensitivity and specificities is maxi-
mized, which is not data dependent. We estimated 95% confidence intervals (CIs) using
bootstrapping with 2,000 replicates within the R package pROC [30] and validated the results
using the Stata command roctab [31]. We also calculated the percent of records in the test set
classified correctly. As further validation for all models, we calculated Spearman’s rho in the
test set between the predicted probabilities of PD for each patient derived from each model.
This inter-method reliability approach does not require a true gold standard in order to
attempt to validate both methods [32]. We compared the AUCs from the penalized regression
with Part D to the one without Part D, to assess whether the inclusion of prescription medica-
tion data improved discrimination [33]. Using the same method, we also compared the AUCs
from the random forest classifier, as well as the combined model, to the penalized regression
with Part D data, to assess whether the application of machine learning improved model
performance.
Results
Characteristics of cases and controls
We observed all known associations [13] between PD and age, sex, race/ethnicity, and smok-
ing (Table 1). On average, cases were 78.8 years old, and controls were 78.1 years old. Cases
PLOS ONE | https://doi.org/10.1371/journal.pone.0256592 August 26, 2021
5 / 13
PLOS ONE Parkinson disease predictive algorithms
Table 1. Characteristics of Parkinson disease cases and controls with Medicare Part D coverage, U.S. Medicare
2009, %.
Cases
N = 35,941
Controls
N = 52,324
Age, years
Female
Race/ethnicity
66–69
70–74
75–79
80–84
85–90
White
Black
Pacific Islander/other
Asian
Hispanic
Native American
Unknown
8.1
19.5
24.2
27.3
21.0
64.7
86.3
6.0
1.2
2.9
3.1
0.3
0.1
16.7
28.3
22.3
19.2
13.4
54.0
83.7
7.8
1.6
3.4
2.9
0.4
0.1
Smoking index � mediana
Age, years, mean (SD)
Number of unique ICD9 codes, mean (SD)
41.1
78.8 (6.1)
99.7 (52.4)
51.5
78.1 (6.2)
76.3 (46.0)
a Predicted probability of ever smoking divided by the person’s total number of unique diagnosis codes.
Abbreviations: ICD9 = International Classification of Diseases, Ninth Revision, Clinical Modification; SD = standard
deviation.
https://doi.org/10.1371/journal.pone.0256592.t001
had substantially more unique ICD9 codes in the five years prior to PD diagnosis as compared
to controls up to their comparable reference date.
Characteristics of the models
In the present dataset, the initial penalized logistic regression model, without prescription
medications, selected 183 ICD9/procedure codes, in addition to the 11 forced demographic
variables for a total of 194 predictors (S1 Table). The second model, which repeated the penal-
ized logistic regression, while including the prescription medications, contained all but two of
the ICD9/procedure codes from the first model, as well as 50 additional ICD9/procedure
codes and 28 prescription medications for a total of 270 predictors (S1 Table). Insofar as the
predictors were the same in both of the penalized regression models, the respective ORs were
generally similar.
For the random forest classifier model, the optimal subset of predictors contained 272 pre-
dictors: 248 ICD9/procedure codes, 18 active ingredients, and 6 of the 11 basic demographic
variables (the two age spline variables, sex, smoking, total count of ICD9 codes, black race)
(S1 Table).
Although 121 predictors in the random forest classifier model were not selected into either
penalized regression model, there was substantial overlap between the three models in terms
of the selected predictors, with 117 predictors (111 ICD9/procedure codes and the above 6
demographic variables) appearing in all three models (Fig 1 and S1 Table). Notably, when we
reviewed the non-overlapping codes it was clear that the random forest favored common diag-
noses/procedures, including those with modest magnitudes of association with PD, whereas
PLOS ONE | https://doi.org/10.1371/journal.pone.0256592 August 26, 2021
6 / 13
PLOS ONE Parkinson disease predictive algorithms
Fig 1. Comparison of distinct and shared predictors between models for predicting Parkinson disease, U.S.
Medicare 2009.
https://doi.org/10.1371/journal.pone.0256592.g001
the penalized logistic regression favored rare diagnoses/procedures if the magnitude of the
association was relatively large or other uncommon codes. For example, the penalized regres-
sion included gout (specifically ICD9 274.9), but the random forest did not.
When we joined the penalized regression and random forest approaches into a combined
model, 232 predictors were selected (S2 Table). These predictors included 193 ICD9/proce-
dure codes and 27 prescription medications in addition to the 11 demographic variables and
the one variable that captured the predicted probability of PD from the random forest. As
expected, we observed the largest OR for the single predictor that represented the random for-
est PD prediction probability. The combined model included 10 codes not selected by any of
the three primary models (S1 and S2 Tables). However, all these codes had ORs close to one.
Model performance
When we applied each of the three primary models to the test set, the AUC was quite similar
for each of the three models (Table 2). Accordingly, the AUC was not significantly improved
either by the addition of the Part D data to the penalized regression, or by using random forest
methods instead of penalized regression. We achieved a slightly greater AUC with the com-
bined model, in which the penalized regression model with Part D predictors also included the
probability of PD for each participant produced by the random forest as a predictor. However,
the AUC was not significantly better as compared to the similar model without this predictor.
When we applied a one-year lag to the claims data, the lagged penalized logistic regression
with Part D data contained 199 ICD9/procedure codes and no medications, while the random
forest contained 155 ICD9/procedure codes and five medications. The lagged penalized regres-
sion had an AUC of 0.742 (95% CI 0.731–0.753) and the random forest had an AUC of 0.740
(95% CI 0.729–0.751).
The three primary models had similar sensitivity and specificity. At the cut point that maxi-
mized the percent of subjects classified correctly, the combined model had greater sensitivity
but slightly less specificity than the penalized regression models (Table 2). At the cut point that
maximized the sum of sensitivity and specificity (Youden’s index) [29], all models had
PLOS ONE | https://doi.org/10.1371/journal.pone.0256592 August 26, 2021
7 / 13
PLOS ONE Parkinson disease predictive algorithms
Table 2. Performance of models for predicting Parkinson disease in the test dataset.
Cut point that maximizes
percent accurately classifieda
Specificity
Sensitivity
Cut point at Youden’s indexa Overall performance
Relative performanceb
Sensitivity
Specificity
AUC(95% CI)
Penalized regression without Part
D
Penalized regression with Part D
Random forest (with Part D)
Combined model (with Part D)c
(95% CI)
65.5 (63.9–
67.1)
67.2 (65.6–
68.7)
66.3 (64.7–
67.8)
72.9 (71.5–
79.6)
(95% CI)
83.4 (82.4–
84.4)
82.6 (81.6–
83.7)
82.8 (81.8–
83.9)
79.6 (78.4–
80.7)
(95% CI)
78.0 (76.7–
79.3)
78.6 (77.2–
79.9)
76.8 (75.4–
78.1)
76.3 (74.9–
77.6)
(95% CI)
73.2 (71.9–
74.4)
73.3 (72.1–
74.6)
75.0 (73.9–
76.2)
76.3 (75.0–
77.4)
0.824 (0.815–0.832)
Reference
model
0.827 (0.818–0.836)
p = 0.61
0.826 (0.818–0.835)
0.835 (0.826–0.843)
–
–
–
Reference
model
p = 0.90
p = 0.23
a Percent sensitivity or specificity, at selected cut points: The cut point that maximizes the percent accurately classified (data dependent) and the cut point at Youden’s
index [29] (not data dependent).
b The AUC is a measure of overall model performance, and the presented p-value assesses relative performance of the specified model as compared to the stated
reference model using the method of DeLong et al. [33] to obtain the p-value. A p-value < 0.05 indicates that the two AUCs being compared are significantly different.
The first comparison tests whether there is a difference in AUC when including Part D prescription medication data in the penalized regression model. The other
comparisons test whether there is a difference in the AUCs across the different approaches in which Part D data were included.
c Random forest classifier’s case prediction probability included as a predictor in a new penalized regression model with Part D prescription medication data.
Abbreviations: AUC = area under the receiver operator characteristic curve; CI = confidence interval.
https://doi.org/10.1371/journal.pone.0256592.t002
sensitivity and specificity estimates that were fairly similar (73.2–78.6%), with the combined
model maximizing specificity. The number of records correctly classified in the test set was
very similar across all models (76.1% for the penalized regression without medications, 76.4%
for the penalized regression with medications, 76.0% for the random forest, and 76.9% for the
combined model).
Agreement between predicted probabilities
For each Medicare beneficiary in our dataset, the two penalized regressions’ probabilities were
in very close agreement, despite the second model including prescription medication data
(Spearman’s rho = 0.995). When we compared the random forest predicted probabilities to
those generated by the penalized regression methods, agreement was still high (Spearman’s
rho = 0.915 with the model without Part D data and rho = 0.912 with the model with Part D
data used as predictors). The combined model had Spearman’s rho’s of 0.96 with all three
models.
Discussion
Identification of people with PD during the prodromal period represents an urgent research
priority due to the need to implement neuroprotective therapies earlier in the neurodegenera-
tive process and to prevent disease related morbidity associated with treatable motor symp-
toms. Our recent, complementary study [14] validated the previous PD predictive model [13],
providing evidence that the model is effective and a possible strategy to identify those in the
prodromal stage of PD. The current study continues to build upon this work by assessing the
value of adding medication data from Medicare Part D to an ICD9/procedure code-based pre-
dictive model, as well as applying machine learning methods to further validate and enhance
our previous work [13, 14]. The current study suggests prescription medication data would
not improve performance of our original predictions had pharmacy data been available for all
PLOS ONE | https://doi.org/10.1371/journal.pone.0256592 August 26, 2021
8 / 13
PLOS ONE Parkinson disease predictive algorithms
of the beneficiaries in that sample, because the AUCs between the models with and without
pharmacy data were quite similar and not statistically different. However, adding a random
forest classifier might slightly improve our model, which had already performed well. Even
though the combined model did not have a statistically significantly higher AUC, such a small
gain might be difficult to detect even in this large dataset. The latter method, which uses an
independent analytic paradigm, also provided confirmation that our previous modeling
approach was well suited to developing a predictive algorithm of undiagnosed PD. In addition,
the high correlations between model predictions and the consistency of the discriminative
ability to detect PD provide evidence that our previous and current models approach the best
possible classifier given the Medicare data structure used in this study. Taken together, this fur-
ther validates our previous predictive model [13].
Interestingly, the addition of medications to the predictive model did not improve the over-
all model performance consequentially. The addition of medications resulted in a model with
27% more diagnosis/procedure codes. In fact, the addition of prescription medications com-
plicated the model without greatly improving prediction, suggesting that the diagnoses for
which the medications were used sufficiently distinguished PD cases from controls. Moreover,
generating hypotheses about the point estimate associations with PD for the medications
selected by our model may be difficult, since some medications can be used for a variety of
medical conditions which may have directionally opposite associations with PD. Nevertheless,
most medications identified in the models consistently aligned with potential pharmacological
treatment options of medical conditions shared by all models. Our penalized regression model
with Medicare Part D confirmed the recently published “protective” association for albuterol
(salbutamol) [34]. However, this might reflect the strong inverse association between tobacco
smoking and PD [35], given that carvedilol, which has the opposite pharmacologic effect on β2
adreonoreceptors, also was selected as a negative predictor, and both medications are indicated
for smoking-related conditions. The random forest did not select these or similar medications
related to smoking but alternatively selected chronic ischemic heart disease and a history of
myocardial infarction, both strongly associated with smoking. The medications positively
associated with PD that remained in the penalized regression model, beyond what was cap-
tured via the diagnosis and procedure codes, were primarily those used to treat depression
(fluoxetine, duloxetine, mirtazapine, paroxetine, sertraline, and citalopram), reflecting the
importance of the non-motor symptoms during the prodromal PD period.
There were some consistent themes to the predictors selected by the different algorithms.
Both random forest and penalized regression models highlighted the importance of key pre-
dictors of PD, such as age, sex, white vs. black race, smoking, the cardinal motor signs of PD,
and dementia/cognitive impairment. The random forest and the respective penalized logistic
regression models (with medication data) shared approximately 43% of the predictors, and
these models were comprised almost entirely of ICD9/procedure codes. All models identified
diagnosis and procedure codes which were suggestive of both motor and non-motor symp-
toms and medical conditions associated with PD. Motor signs and/or symptoms, such as
“abnormal involuntary movement”, “tremor”, “lack of coordination”, and “abnormality of
gait” were recognized by all models as important predictors of PD, as expected. Procedure
codes shared among all three models included various brain and spine imaging codes, physical
therapy, and a variety of non-specific diagnostic tests. These codes likely reflect a combination
of diagnostic workup for prodromal PD symptoms and an attempt to treat progressive motor
problems with non-pharmacological approaches. The codes indicative of non-motor symp-
toms that appeared to identify patients with a high probability of PD reflected gastrointestinal
dysfunction (constipation), dysautonomia (orthostatic hypotension, dizziness), and cognitive/
psychiatric impairments other than general anxiety (memory loss, altered mental status,
PLOS ONE | https://doi.org/10.1371/journal.pone.0256592 August 26, 2021
9 / 13
PLOS ONE Parkinson disease predictive algorithms
mental disorder, and depression). Overall, the codes that were common between the three
models demonstrate a prodromal disease state characterized by non-motor symptoms, tremor,
gait impairment, and an attempt by health care providers to treat or identify the cause of the
symptoms.
The random forest tended to select more common predictors with lower magnitude associ-
ations. In contrast, the penalized logistic model selected conditions that were uncommon but
with a known association with PD, such as gout. Similarly, in our original predictive model
using the same regression method but larger sample size, this approach also selected condi-
tions that are rare but have large magnitude associations with PD, such as REM sleep behavior
disorder. The random forest model identified a greater number of unique codes than the
penalized regression models, yet the conditions/procedures represented by these codes had
weaker associations with PD. Many variables with the highest rank in the importance matrix
included common medical conditions that may reflect the importance of health care utiliza-
tion in being diagnosed with PD [18]. Categories distinguishing the random forest model
from the penalized regression models included: 1) prescription medications commonly pre-
scribed for bowel and bladder disorders, cognitive impairment/dementia, and psychiatric dis-
orders (e.g., depression and anxiety); 2) codes indicating head and other body trauma,
previously identified comorbidities of PD [8]; and 3) codes indicating health care utilization
prior to PD diagnosis. These codes provide interesting insight into an alternative approach to
predicting PD. The distinct methodologies we used in our study clearly identify marked clini-
cal differences between prodromal PD patients and the general population.
A strength of the study is that there were approximately 133 cases and 194 controls for each
predictor considered during the model fitting process. Theoretically, the large sample size to
predictor ratio in our models caused our predictions to approach the asymptotically minimum
achievable error [36, 37] for classifying PD. For this reason, and because the penalized regres-
sion and random forest machine learning are independent analytic approaches, we also com-
bined these into one model by feeding the PD probability from the random forest into the
penalized regression model. This approach increased the AUC by approximately 1% in abso-
lute terms. Although this difference may appear small, a 1% improvement might have a mean-
ingful impact on the absolute number of individuals further screened for PD, when applying
the predictive algorithm to a large dataset. Additionally, this improvement may be relatively
substantial considering the models may already be close to the asymptotic prediction limit.
Interestingly, the combined model’s incorporation of the random forest predictions resulted
in a discrimination gain by improving its sensitivity, reinforcing the idea that the random for-
est captured slightly different information about the cases than the penalized regressions. That
is, this model gained greater discrimination by improving case identification, and did so only
at a small cost to control identification. This is reasonable because the random forest probabil-
ity acts like a PD case preprocessing filter, improving sensitivity. In practice, all of these models
have the advantage of offering users complete flexibility in their application, such that one can
balance sensitivity and specificity to customize to each situation.
Despite the many study strengths, there are several potential limitations. First, Medicare is
only a population-based health care program for individuals older than 65; therefore, applica-
tion of this predictive model to younger individuals would not be appropriate. Second, Medi-
care data are limited to medical claims data, which are filed upon delivery of medical services
or filling of prescriptions. Other datasets, such as electronic medical record systems, may have
greater data granularity that could be leveraged for even greater model performance. With that
said, electronic medical record systems present substantial data quality challenges, as well [38].
Additionally, we only had pharmacy data for the final two years of the five year period prior to
PD diagnosis, which may have limited the usefulness of these data. However, these later years
PLOS ONE | https://doi.org/10.1371/journal.pone.0256592 August 26, 2021
10 / 13
PLOS ONE Parkinson disease predictive algorithms
are likely to be predictive due to the prodromal period of PD, insofar as patient symptoms lead
to new medications being prescribed or patients discontinuing medications due to side effects.
Non-pharmacy data in these later years were quite important to our predictive model. Notably,
we found that motor signs of PD had large ORs in the penalized regressions and high impor-
tance in the random forest. Because these signs and symptoms tend to occur in the later pro-
dromal period, relatively close to PD diagnosis, application of a one-year lag did materially
reduce the AUCs for all of our models. These reductions were similar across all models, but
discrimination remained quite good. We also note that ICD9 codes in the final three months
before PD diagnosis probably were particularly influential in achieving such high AUCs in the
unlagged model. There is an increase in the number of diagnoses (ICD9 codes) assigned to
patients around the time of PD diagnosis, as patients seek out care for either their symptoms
of PD or other medical conditions. The overall number of unique ICD9 codes is an important
predictor, in part because of this phenomenon. In addition, we and others have observed a
marked spike in traumas, likely due to falls, in the three months prior to PD diagnosis [9], but
that increased risk of fractures is evident for six to seven years prior to PD diagnosis. In addi-
tion, non-motor symptoms of PD frequently precede the motor symptoms [13]. Thus, we
believe that additional lagging would have a diminished influence on AUCs. As such, predic-
tion of PD more than five years prior to diagnosis will be an important goal for future studies.
The present work provides a useful foundation for this future work by demonstrating that
these predictive models should be attempted in larger datasets, as utilized in our original pre-
dictive model of PD, rather than restricted to individuals with pharmacy coverage.
Supporting information
S1 Table. Three primary predictive models, PD predictive model, U.S. Medicare 2009.
�HCPCS codes are similar to CPT codes but are specific to Medicare; Abbreviations:
CPT = Current Procedural Terminology; HCPCS = Healthcare Common Procedure Coding
System�; ICD9 = International Classification of Diseases, Ninth Revision; PD = Parkinson dis-
ease.
(PDF)
S2 Table. Combined model, PD predictive model, U.S. Medicare 2009. �HCPCS codes are
similar to CPT codes but are specific to Medicare; Abbreviations: CPT = Current Procedural
Terminology; HCPCS = Healthcare Common Procedure Coding System�;
ICD9 = International Classification of Diseases, Ninth Revision; PD = Parkinson disease.
(PDF)
Author Contributions
Conceptualization: Susan Searles Nielsen, Roman Garnett, Brad A. Racette.
Formal analysis: Mark N. Warden, Susan Searles Nielsen, Alejandra Camacho-Soto.
Funding acquisition: Brad A. Racette.
Methodology: Roman Garnett.
Supervision: Susan Searles Nielsen, Roman Garnett, Brad A. Racette.
Validation: Mark N. Warden, Alejandra Camacho-Soto.
Writing – original draft: Mark N. Warden.
PLOS ONE | https://doi.org/10.1371/journal.pone.0256592 August 26, 2021
11 / 13
PLOS ONE Parkinson disease predictive algorithms
Writing – review & editing: Susan Searles Nielsen, Alejandra Camacho-Soto, Roman Garnett,
Brad A. Racette.
References
1. Siderowf A, Jennings D, Eberly S, Oakes D, Hawkins KA, Ascherio A, et al. Impaired olfaction and other
prodromal features in the Parkinson At-Risk Syndrome Study. Mov Disord. 2012; 27(3):406–412.
https://doi.org/10.1002/mds.24892 PMID: 22237833
2. Plouvier AO, Hameleers RJ, van den Heuvel EA, Bor HH, Olde Hartman TC, Bloem BR, et al. Prodro-
mal symptoms and early detection of Parkinson’s disease in general practice: a nested case-control
study. Fam Pract. 2014; 31(4):373–378. https://doi.org/10.1093/fampra/cmu025 PMID: 24869632
3. Goldman JG, Postuma R. Premotor and nonmotor features of Parkinson’s disease. Curr Opin Neurol.
2014; 27(4):434–441. https://doi.org/10.1097/WCO.0000000000000112 PMID: 24978368
4. Schrag A, Horsfall L, Walters K, Noyce A, Petersen I. Prediagnostic presentations of Parkinson’s dis-
ease in primary care: a case-control study. Lancet Neurol. 2015; 14(1):57–64. https://doi.org/10.1016/
S1474-4422(14)70287-X PMID: 25435387
5. Gustafsson H, Nordstrom A, Nordstrom P. Depression and subsequent risk of Parkinson disease: A
nationwide cohort study. Neurology. 2015; 84(24):2422–2429. https://doi.org/10.1212/WNL.
0000000000001684 PMID: 25995056
6. Camacho-Soto A, Warden MN, Searles Nielsen S, Salter A, Brody DL, Prather H, et al. Traumatic brain
injury in the prodromal period of Parkinson’s disease: A large epidemiological study using medicare
data. Ann Neurol. 2017; 82(5):744–754. https://doi.org/10.1002/ana.25074 PMID: 29024046
7. Kenborg L, Rugbjerg K, Lee PC, Ravnskjaer L, Christensen J, Ritz B, et al. Head injury and risk for Par-
kinson disease: results from a Danish case-control study. Neurology. 2015; 84(11):1098–1103. https://
doi.org/10.1212/WNL.0000000000001362 PMID: 25681453
8. Nystrom H, Nordstrom A, Nordstrom P. Risk of Injurious Fall and Hip Fracture up to 26 y before the
Diagnosis of Parkinson Disease: Nested Case-Control Studies in a Nationwide Cohort. PLoS Med.
2016; 13(2):e1001954. https://doi.org/10.1371/journal.pmed.1001954 PMID: 26836965
9. Camacho-Soto A, Gross A, Searles Nielsen S, Miller AN, Warden MN, Salter A, et al. Fractures in the
prodromal period of Parkinson disease. Neurology. 2020; 94(23):e2448–e2456. https://doi.org/10.
1212/WNL.0000000000009452 PMID: 32345729
10. Khoury N, Attal F, Amirat Y, Oukhellou L, Mohammed S. Data-Driven Based Approach to Aid Parkin-
son’s Disease Diagnosis. Sensors (Basel). 2019; 19(2). https://doi.org/10.3390/s19020242 PMID:
30634600
11. Cavallo F, Moschetti A, Esposito D, Maremmani C, Rovini E. Upper limb motor pre-clinical assessment
in Parkinson’s disease using machine learning. Parkinsonism Relat Disord. 2019; 63:111–116. https://
doi.org/10.1016/j.parkreldis.2019.02.028 PMID: 30826265
12. Pena-Nogales O, Ellmore TM, de Luis-Garcia R, Suescun J, Schiess MC, Giancardo L. Longitudinal
Connectomes as a Candidate Progression Marker for Prodromal Parkinson’s Disease. Front Neurosci.
2018; 12:967. https://doi.org/10.3389/fnins.2018.00967 PMID: 30686966
13. Searles Nielsen S, Warden MN, Camacho-Soto A, Willis AW, Wright BA, Racette BA. A predictive
model to identify Parkinson disease from administrative claims data. Neurology. 2017; 89(14):1448–
1456. https://doi.org/10.1212/WNL.0000000000004536 PMID: 28864676
14.
15.
Faust IM, Racette BA, Searles Nielsen S. Validation of a Parkinson Disease Predictive Model in a Popu-
lation-Based Study. Parkinsons Dis. 2020; 2020:2857608. https://doi.org/10.1155/2020/2857608
PMID: 32148753
Lopez-Sendon J, Mena MA, de Yebenes JG. Drug-induced parkinsonism. Expert Opin Drug Saf. 2013;
12(4):487–496. https://doi.org/10.1517/14740338.2013.787065 PMID: 23540800
16. Centers for Disease Control and Prevention (CDC). BRFSS 2009 Survey Data and Documentation
Atlanta: CDC; 2009 [updated December 4, 2014; accessed May 17, 2016, 2016]. Available from: http://
www.cdc.gov/brfss/annual_data/annual_2009.htm.
17. Ritz B, Ascherio A, Checkoway H, Marder KS, Nelson LM, Rocca WA, et al. Pooled analysis of tobacco
use and risk of Parkinson disease. Arch Neurol. 2007; 64(7):990–997. https://doi.org/10.1001/archneur.
64.7.990 PMID: 17620489
18. Gross A, Racette BA, Camacho-Soto A, Dube U, Searles Nielsen S. Use of medical care biases associ-
ations between Parkinson disease and other medical conditions. Neurology. 2018; 90(24):e2155–
e2165. https://doi.org/10.1212/WNL.0000000000005678 PMID: 29743207
19.
Zou H, Hastie T. Regularization and variable selection via the elastic net. Journal of the Royal Statistical
Society Series B-Statistical Methodology. 2005; 67:301–320.
PLOS ONE | https://doi.org/10.1371/journal.pone.0256592 August 26, 2021
12 / 13
PLOS ONE Parkinson disease predictive algorithms
20.
Friedman J, Hastie T, Tibshirani R. Regularization Paths for Generalized Linear Models via Coordinate
Descent. J Stat Softw. 2010; 33(1):1–22. PMID: 20808728
21.
Liaw A, Wiener M. Classification and Regression by RandomForest. R News. 2002; 2(3):18–22.
22. Diaz-Uriarte R. GeneSrF and varSelRF: a web-based tool and R package for gene selection and classi-
fication using random forest. BMC Bioinformatics. 2007; 8:328. https://doi.org/10.1186/1471-2105-8-
328 PMID: 17767709
23. Diaz-Uriarte R, Alvarez de Andres S. Gene selection and classification of microarray data using random
forest. BMC Bioinformatics. 2006; 7:3. https://doi.org/10.1186/1471-2105-7-3 PMID: 16398926
24.
Fernandez-Delgado M, Cernadas E, Barro S, Amorim D. Do we Need Hundreds of Classifiers to Solve
Real World Classification Problems? J Machine Learning Research. 2014; 15:3133–3181.
25. Rovini E, Maremmani C, Moschetti A, Esposito D, Cavallo F. Comparative Motor Pre-clinical Assess-
ment in Parkinson’s Disease Using Supervised Machine Learning Approaches. Ann Biomed Eng. 2018;
46(12):2057–2068. https://doi.org/10.1007/s10439-018-2104-9 PMID: 30030773
26.
LeCun Y, Bengio Y, Hinton G. Deep learning. Nature. 2015; 521(7553):436–444. https://doi.org/10.
1038/nature14539 PMID: 26017442
27. Amoroso N, La Rocca M, Monaco A, Bellotti R, Tangaro S. Complex networks reveal early MRI markers
of Parkinson’s disease. Med Image Anal. 2018; 48:12–24. https://doi.org/10.1016/j.media.2018.05.004
PMID: 29807313
28. Steyerberg EW, Vickers AJ, Cook NR, Gerds T, Gonen M, Obuchowski N, et al. Assessing the perfor-
mance of prediction models: a framework for traditional and novel measures. Epidemiology. 2010; 21
(1):128–138. https://doi.org/10.1097/EDE.0b013e3181c30fb2 PMID: 20010215
29. Youden WJ. Index for rating diagnostic tests. Cancer. 1950; 3(1):32–35. https://doi.org/10.1002/1097-
0142(1950)3:1<32::aid-cncr2820030106>3.0.co;2-3 PMID: 15405679
30. Robin X, Turck N, Hainard A, Tiberti N, Lisacek F, Sanchez JC, et al. pROC: an open-source package
for R and S+ to analyze and compare ROC curves. BMC Bioinformatics. 2011; 12:77. https://doi.org/10.
1186/1471-2105-12-77 PMID: 21414208
31. StataCorp. Stata MC 14.2. MC 14.2 ed. College Station, TX: StataCorp LP; 2015.
32. White E, Armstrong BK, Saracci R. Principles of Exposure Measurement in Epidemiology: Collecting,
Evaluating, and Improving Measures of Disease Risk Factors. Oxford: Oxford University Press; 2008.
33. DeLong ER, DeLong DM, Clarke-Pearson DL. Comparing the areas under two or more correlated
receiver operating characteristic curves: a nonparametric approach. Biometrics. 1988; 44(3):837–845.
PMID: 3203132
34. Mittal S, Bjornevik K, Im DS, Flierl A, Dong X, Locascio JJ, et al. beta2-Adrenoreceptor is a regulator of
the alpha-synuclein gene driving risk of Parkinson’s disease. Science. 2017; 357(6354):891–898.
https://doi.org/10.1126/science.aaf3934 PMID: 28860381
35. Searles Nielsen S, Gross A, Camacho-Soto A, Willis AW, Racette BA. beta2-adrenoreceptor medica-
tions and risk of Parkinson disease. Ann Neurol. 2018; 84(5):683–693. https://doi.org/10.1002/ana.
25341 PMID: 30225948
36.
37.
Tarca AL, Draghici S, Romero R. Developing classifiers for the detection of cancer using multi-analytes.
Methods Mol Biol. 2009; 520:259–272. https://doi.org/10.1007/978-1-60327-811-9_19 PMID:
19381961
Figueroa RL, Zeng-Treitler Q, Kandula S, Ngo LH. Predicting sample size required for classification per-
formance. BMC Med Inform Decis Mak. 2012; 12:8. https://doi.org/10.1186/1472-6947-12-8 PMID:
22336388
38. Harrington L. Copy-Forward in Electronic Health Records: Lipstick on a Pig. Jt Comm J Qual Patient
Saf. 2017; 43(8):371–374. https://doi.org/10.1016/j.jcjq.2017.04.007 PMID: 28738981
PLOS ONE | https://doi.org/10.1371/journal.pone.0256592 August 26, 2021
13 / 13
PLOS ONE |
10.1371_journal.pone.0255359 | RESEARCH ARTICLE
“Not just a hobby, but a lifestyle”:
Characteristics, preferences and self-
perception of individuals with different levels
of involvement in birdwatching
Emilia JaneczkoID
5
Krzysztof Janeczko4, Natalia KorczID
1, Adrian Łukowski2, Ernest Bielinis3, Małgorzata WoźnickaID
1*,
1 Department of Forest Utilization, Institute of Forest Sciences, Warsaw University of Life Sciences—SGGW,
Warsaw, Poland, 2 Department of Silviculture, Faculty of Forestry, Poznan University of Life Sciences,
Poznan, Poland, 3 Department of Forestry and Forest Ecology, Faculty of Agriculture and Forestry,
University of Warmia and Mazury, Olsztyn, Poland, 4 Department of Forest Management Planning,
Dendrometry and Forest Economics, Institute of Forest Sciences, Warsaw University of Life Sciences—
SGGW, Warsaw, Poland, 5 Department of Natural Foundations of Forestry, Institute of Soil Science and
Environment Management, University of Life Sciences in Lublin, Lublin, Poland
* malgorzata_woznicka@sggw.edu.pl
Abstract
Birdwatching is one of the most sustainable types of nature-based tourism and, at the same
time, a form of recreation that is developing very dynamically. Birdwatching is attracting
more and more people, not only professionals, but also amateurs from many countries. Bird-
watching research is still relatively embryonic, especially when compared to nature tourism
or wildlife tourism. Our main aim was to determine preferences and opinions of birdwatchers
visiting the largest national park in Poland, in relation to their different levels of involvement.
The data were collected in 2018 from a survey of a sample of 357 Polish and foreign bird-
watchers. Results showed that birdwatcher respondents were predominantly male, middle-
aged, and living in a large city. An important tool described in this article is a new scale that
assesses the level of involvement of individual people engaged in birdwatching activity. This
scale corresponds well with the individual characteristics of birdwatchers. Most birdwatchers
defined their birdwatching activity as a permanent rather than a temporary hobby and there-
fore considered it to be more of a lifestyle than a hobby. Engagement in birdwatching activity
increased with age and frequency of trips. The two most important reasons for birding were
‘to be close to nature’ and ‘fascination with birds’. It has been proven that the development
of birdwatching in the future will require a developed infrastructure enabling interaction with
the objects of observation.
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: Janeczko E, Łukowski A, Bielinis E,
Woźnicka M, Janeczko K, Korcz N (2021) “Not just
a hobby, but a lifestyle”: Characteristics,
preferences and self-perception of individuals with
different levels of involvement in birdwatching.
PLoS ONE 16(7): e0255359. https://doi.org/
10.1371/journal.pone.0255359
Editor: Shah Md Atiqul Haq, Shahjalal University of
Science and Technology, BANGLADESH
Received: February 24, 2021
Accepted: July 6, 2021
Published: July 30, 2021
Copyright: © 2021 Janeczko et al. This is an open
access article distributed under the terms of the
Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: All relevant data are
within the manuscript and its Supporting
Information files.
Funding: The authors received no specific funding
for this work.
Introduction
Competing interests: The authors have declared
that no competing interests exist.
There has recently been a growing interest among tourists in the natural environment, and
demand for ecotourism increases annually [1–4]. Ecotourism is one of the fastest growing
PLOS ONE | https://doi.org/10.1371/journal.pone.0255359 July 30, 2021
1 / 16
PLOS ONE Preferences and self-perception of individuals with different levels of involvement in birdwatching
sectors in the tourism industry [3]. There are many benefits associated with nature tourism
activities. For example, wildlife watching, wildlife photography, birdwatching, and birdfeeding
are popular with many people and provides a significant economic contribution to local econ-
omies. Expenditures on wildlife-watching activities generate employment and income in vari-
ous manufacturing industries and service sectors [5]. Thus in general tourists become more
aware of the value of biodiversity and the conservation of natural resources [6], thereby the
environmental and economic wellbeing of the local community improves [7, 8].
Avitourism, and therefore also birdwatching, is one of the most sustainable types of nature-
based tourism [9, 10]. Birdwatching, defined as a nature tourism activity [11], is perceived as a
form of recreation that provides an opportunity for contact with the natural world and an
escape from modern, consumer-oriented society. However, Kronenberg [12] points out that
wildlife viewing is nevertheless a visual experience and often requires evidence. By their pres-
ence and persistence, birdwatchers affect the attractiveness of birds’ breeding, migration or
roosting sites; frighten birds; and otherwise increase pressure on birds and their habitats (e.g.,
by luring birds out of hiding places and stressing them by reproducing their calls, or by expos-
ing birds and their nests to predators). Birdwatching is a relatively new activity for large groups
of people that emerged at the beginning of the 20th century [12]. Previously, birdwatching was
carried out by specialists, mainly ornithologists. Currently, these activities involve not only
professionals, but also amateurs from many countries; birdwatching has developed most rap-
idly in the United States and western Europe, particularly in the UK, the Netherlands and Ger-
many [13]. The study by Cordell and Herbert [14] shows that 70.4 million people in the
United States are interested in birdwatching, indicating that birdwatching is one of the favour-
ite activities of Americans. In Poland, only 7,000–10,000 people currently claim to engage in
birdwatching, although estimates conducted by the Polish Society for the Protection of Birds
show that we have about 3,000 birdwatchers in the country, which is a rather small number
considering the area of the country and the total population [15].
Steven et al. [16] suggest that birdwatching research is still relatively embryonic and that
research interest in avitourism is still much smaller compared to interest in nature tourism or
wildlife tourism. The dominant issue in birdwatching research to date has been the recognition of
the influence of demographic factors on birding behaviour [e.g. 9, 11, 16–22]. Subsequent studies
have examined the environmental preferences of avitourists [13, 21, 23–25] The economic bene-
fits of birdwatching development are also increasingly the subject of research [26–28].
There have also been several attempts to determine the extent to which people engage in
this recreational activity. For example, Boxall and McFarlane [29] asked birders to rank them-
selves as casual, novice, intermediate or advanced, based on level of activity and intensity of
interest. Scott and Thigpen [21] identified four groups of birders, namely casual, interested,
active and skilled, but another study by Scott et al. [30] distinguished three measures of bird
specialization: committed birders, active birders, casual birders. Vaske et al. [19] considered
birdwatching within the broader category of wildlife viewing and distinguished four types of
participants of recreational movement related to wildlife observation: highly involved tourists
(highly engaged), creative tourists (creative), generalists (generalist) and casual tourists (occa-
sionalist). Anderwald [31] distinguished five stages in the development of a birdwatcher’s
knowledge: the silent keeper, the researcher-observer, the bird lover, the ornithological tourist
and the Icarus man. In turn, Hvenegaard [23] divided birders into three distinct groups:
advanced-experienced, advanced-active and novice. It is therefore clear that birdwatching is a
form of recreation that encompasses various aspects of skills and knowledge, behaviour and
stages of involvement that have a direct bearing on lifestyle. Thus, there is a lack of consistency
in the nominal scales used to assess birdwatchers’ involvement in their hobby. Also it is notice-
able that there are no proper guidelines to compare the results for any aspect of birdwatchers.
PLOS ONE | https://doi.org/10.1371/journal.pone.0255359 July 30, 2021
2 / 16
PLOS ONE Preferences and self-perception of individuals with different levels of involvement in birdwatching
Even the authors themselves change the scales they use, probably in search of better solutions
[21, 30]. There is a need to find some simple measure that would define well the level of
involvement in the activity of birdwatching. However, the best solution would be to use a dif-
ferent type of scale (other than a nominal scale) to harmonise this measure across studies.
Our main aim was to determine preferences and opinions of birdwatchers visiting the larg-
est national park in Poland in relation to their different levels of involvement. An additional
goal was to create a tool that would allow for a very precise expression of the level of involve-
ment in birdwatching (using an interval scale) in order to address the lack of unification of the
results obtained by different authors for groups of different levels of involvement. The signifi-
cant rate of birdwatching development indicates the need for a better understanding of the
demography, preferences and behavioural patterns of birdwatchers [6]. The results of prefer-
ence studies may be useful in ensuring the multifunctional development of natural-valued
areas [32]. Also Reichhart and Arnberger [33] believe that understanding visitor preferences is
valuable in developing effective landscape management strategies. Preferences can help
develop avitourism products that meet the needs of individual birdwatchers and help them
plan more enjoyable experiences at their destinations [25]. In addition, Randler et al. [11]
point out that knowledge of birdwatchers’ behaviour is important because it can have major
implications for data collection and long-term analyses of bird data.
Materials and methods
Study area
This study was performed in Poland. Poland is located in central-eastern Europe and covers
over 300,000 km2. The country’s population is close to 38 million people, giving an average
population density of 120 people/km2. To practise birdwatching in Poland, it is not necessary
to obtain a special license or course regarding knowledge. Birdwatchers are bound by common
rules of behaviour in natural areas, as well as special restrictions in protective areas. It is also
not necessary to be a member of any association. As individuals become involved in advanced
work with birds (catching, ringing, measurements and scientific assistance, etc.), additional
approvals and training courses may be required from them.
The Biebrza Valley is the most important breeding area for many species of wetland birds
in Poland and one of the most important in central and western Europe. This importance
increases as wetlands disappear from the European landscape. The Biebrza Valley belongs to
the largest wildlife refuges in Europe and is of great importance for many species of feeding
and resting birds during annual migrations. There are also boreal breeding species, as well as
species whose geographic centre is in the taiga and tundra zone. The majority of this area is
the Biebrza National Park, covered by the Ramsar Convention to protect the wetlands and
bird breeding grounds. Birdwatching in the area is an opportunity to activate the local com-
munity. This is particularly important due to the fact that the area is located in the periphery,
at a distance from larger cities, which means that living conditions here are not easy. Identifi-
cation of needs, expectations and preferences of birdwatchers is necessary to create new tourist
products and develop tourist services based on the principle of sustainable development.
The questionnaire
Data were collected from a research survey prepared for the purposes of these studies. A sam-
ple of 357 Polish and foreign birdwatchers was taken in 2018. The participants are representa-
tive of the birdwatching population because they were selected from the entire spectrum of
people visiting the Biebrza National Park. The survey was conducted in the field in the vicinity
of infrastructure facilities dedicated to birdwatchers such as hides, observation towers and
PLOS ONE | https://doi.org/10.1371/journal.pone.0255359 July 30, 2021
3 / 16
PLOS ONE Preferences and self-perception of individuals with different levels of involvement in birdwatching
birdwatching terraces. Birders were defined, as in the Hvenegaard [23] study, as those visitors
who visibly participated in birding activities (based on clues such as visiting popular birding
sites and using binoculars, spotting scopes, and bird books). Participants were selected by sci-
entists from Warsaw University of Life Sciences between April and October in 2018 in the
park management building. We asked every fifth birdwatcher who visited Biebrza National
Park in those days to complete the questionnaire by engaging in direct conversation with an
interviewer who recorded the responses in paper form (n = 357). The questionnaire was pre-
pared in several basic languages (Polish, German and English). Respondents were asked to
provide information regarding their gender (male, female), age (18–34 years, 35–54 years, �55
years), level of education (primary, secondary, higher), place of living (village; �100,000 resi-
dents; >100,000 residents) and country of origin (native, abroad). Respondents provided
information regarding their country of permanent residence, and we defined people of Polish
nationality as native, because the research area belongs to the territory of Poland. They were
questioned on many different topics that are presented and discussed in more detail in the
Results and Discussion sections. Our research survey also allowed us to determine how bird-
watchers’ preferences regarding the active realization of their hobbies in the field are formed.
Respondents also had an opportunity to express an opinion on the subject of birdwatching in
Poland. The time that respondents needed to complete the questionnaire ranged from 5 to 7
minutes.
The survey was conducted in full agreement with the national and international regulations
in compliance with the Declaration of Helsinki (2000). The personal information and data of
the participants were anonymous according to the General Data Protection Regulation of the
European Parliament (GDPR 679/2016). The research was voluntary and did not take into
account minors. In Poland, research of this type does not require the approval of the bioethical
commission.
The questionnaire used in this study was fully anonymous. We did not collect any sensitive,
personal information. Participation in the study was completely voluntary and informed.
Respondents were not obligated in any way to participate in the study, they could refuse to par-
ticipate and/or stop the interview at any time. Before the interview began, each respondent ver-
bally expressed their willingness to participate in the study. We surveyed only adults. There is
no question of an ethical violation in this situation. The consent of the university committee
was not required in this regard.
Data analysis
We wanted to find some simple measure that defines well the level of involvement in the activ-
ity of birdwatching. In this study, we developed an ‘involvement score’ based on six statements
defining respondents’ reasons that prompted them to take up this hobby (motives) and on
respondents’ involvement in six important activities related to birdwatching (performed activi-
ties; Table 1). The items that we used were selected on the basis of literature, our experience
and observations of important issues related to birdwatching. We used principal component
analysis to assess the construct validity of items and Cronbach’s α to measure the internal con-
sistency [34]. To allow the use of Likert’s five-point scale, respondents were given the following
five choices when they assessed their motives: definitely yes (+2), rather yes (+1), hard to say
(0), rather no (–1), definitely no (–2). Thus, the score for motives ranged from –12 to +12
accounting for the responses to all six statements. To this, we added the results corresponding
to respondents’ performed activities. It was a multiple answer question, where each selected
option was awarded 1 point. There were 0 points awarded for not choosing a single answer.
The combined, final involvement score for each respondent could range from –12 to +18.
PLOS ONE | https://doi.org/10.1371/journal.pone.0255359 July 30, 2021
4 / 16
PLOS ONE Preferences and self-perception of individuals with different levels of involvement in birdwatching
Table 1. Birdwatchers’ responses to statements about their motives and activities performed for birdwatching.
Items
n1
% yes
n2
% no
What motivated
you to practice
birdwatching?
What activities related
to birdwatching do
you engage in?
1
2
3
4
5
6
1
2
3
4
5
6
I am simply interested in broadly understood nature
I am interested in and impressed by birds
I want to impress other people through my hobby
I consider birdwatching as a fad
I like to actively rest in nature
I want to take unique photos
Items
I feed the birds and observe their behaviour
I hang nesting boxes for birds
I belong to various associations and groups of
people with similar ornithological interests
I improve my photography skills to take better photos
I read the popular articles and scientific articles in
the field of ornithology
I create my own ornithological notes (I publish
some of them)
353
346
44
76
301
187
n1
252
92
126
215
185
157
https://doi.org/10.1371/journal.pone.0255359.t001
99
97
12
21
84
52
2
10
268
248
27
120
0.5
2.5
76
70
8
34
n3
2
1
45
33
29
50
% hard
to say
Factor
loading
PC1
Factor
loading
PC2
0.5
0.5
12
9
8
14
0.15
0.45
-0.45
-0.57
0.27
0.35
0.25
0.24
0.70
0.67
-0.30
-0.21
% of the indicated answers
Factor
loading PC1
Factor
loading PC2
71
26
35
60
52
28
0.12
0.19
0.64
0.47
0.61
0.70
0.21
0.12
0.28
0.25
0.36
0.21
We used non-parametrical Wilcoxon / Kruskal-Wallis statistical tests (due to the lack of
normal distribution–Shapiro-Wilk W Test; W = 0.9542; P < 0.0001) and post hoc nonpara-
metric comparisons for each pair by Wilcoxon method (P < 0.05) to compare differences in
the involvement score among categories of respondents, and we used omega square (ω2) to
investigate effect size, with cut-off levels of ω2 > 0.01 for a small effect, ω2 > 0.06 for a medium
effect and ω2 > 0.14 for a large effect, as is widely accepted [35]. Abbreviation ‘SE’ means stan-
dard error of the mean.
Results
In total, we surveyed 357 respondents, and none were excluded, such that we used the
responses of all respondents for further analyses.
General characteristic of respondents
Among respondents, the majority were men (58%). The age of the respondents was analysed
in groups, and most people were aged 35–54 years (42%). In general, the remaining age groups
were represented similarly (18–34 years = 31% and �55 years = 27%). Approximately one
third of respondents (37%) lived in places with more than 100,000 inhabitants (village = 31%
and places with no more than 100,000 inhabitants = 32%). The majority of birdwatchers came
from Poland (71%). The remaining respondents (29%) came from different countries, such as
UK, Germany, Netherlands, Belgium, Spain, Czech Republic, Austria and Italy; in our analyses
we consider them collectively as foreigners. Nearly all respondents had higher education. Only
14% and 2% had secondary and primary education, respectively.
Involvement score
Nearly all birdwatchers claimed that they were simply interested in broadly understood nature
(Table 1). Nearly all also declared that they were interested in and impressed by birds. The next
two items were constructed in such a way as to make the respondents reflect more deeply on
PLOS ONE | https://doi.org/10.1371/journal.pone.0255359 July 30, 2021
5 / 16
PLOS ONE Preferences and self-perception of individuals with different levels of involvement in birdwatching
birdwatching (items 3 and 4 were reversed in order to eliminate automatism during the exami-
nation). In both cases, the vast majority (about three quarters) did not agree with the statement
that their hobby is a temporary fad or that they wanted to impress others with their hobby. Over
80% of respondents indicated that birdwatching is a convenient activity for them, which helps
them to relax and rest in nature. More than half of the respondents agreed it was important to
take unique photos during active birdwatching, but one third did not support this view.
In the second part, where respondents were asked about which activities related to bird-
watching they engage in, they could select as many answers as they wanted (therefore, for each
question, 100% means the total number of respondents). The activity of feeding birds and
observing their behaviour was characterized by the highest percentage of responses. Hanging
nesting boxes for birds turned out to be the least popular activity among the respondents.
More than half of the respondents read the popular and scientific articles in the field of orni-
thology, but only less than a quarter of them create their own ornithological notes or publish
some of them. A relatively small group of respondents (slightly more than one third) feel the
need to associate with people with similar interests and therefore belong to various associa-
tions and groups of people with similar ornithological interests. Sixty per cent of respondents
indicated that photography and the activities related to it are an important part of
birdwatching.
Table 1 Responses of birdwatchers to six statements defining their motives and six per-
formed activities related to birdwatching. Loadings (from principal component analysis) of
each item on principal component one (PC1) and two (PC2) are also shown. Items in bold spe-
cifically investigated negative motives (points for answers were assigned in the opposite way
than for the rest). In order to simplify the answers from questionnaires, we changed ‘definitely
yes’ and ‘rather yes’ to YES and ‘rather no’ and ‘definitely no’ to NO. Data were collected in
2018 from a sample of 357 birdwatchers visiting Biebrza National Park.
The six statements defining respondents’ motives (Cronbach’s α = 0.52) and six important
performed activities (Cronbach’s α = 0.57) had different factor loadings ranging from –0.57 to
0.69 on principal component 1 (PC1) and from –0.30 to 0.70 on principal component 2 (PC2),
which were the only components with an eigenvalue >1 (PC1 1.84, PC2 1.25). PC1 explained
30.7% and PC2 20.8% of the variance, and thus we judged that it was sufficient to use these
principal components only. The sum of these components was correlated with the involve-
ment score (r = 0.694, P < 0.001), which is easier to interpret than PC1 or PC2. We therefore
concluded that the involvement score is a suitable measure of involvement of birdwatchers in
their activity. Almost all respondents (99.5%) held an involvement score that was at least
slightly positive (i.e., had involvement score �1; Fig 1); mean involvement score was 9.66
(n = 357, SE = 3.1). This result indicates that the group of respondents includes only people
who are actually, to a greater or lesser extent, associated with birdwatching.
Fig 1 Histogram showing the proportion of birdwatchers with each involvement score.
Involvement score was based on the six statements defining respondents’ motives and six impor-
tant performed activities of birdwatching and ranged from –12 to +18 (Materials and Methods).
Data were collected in 2018 from a sample of 357 birdwatchers visiting Biebrza National Park.
Detailed characteristics, preferences and opinions of respondents
Characteristics. When analysing socio-demographic items, it should be stated that the
group of birdwatchers was very homogeneous in terms of their involvement in their activity,
as most of the differentiating features are statistically insignificant (Table 2). Only age turned
out to be a differentiating factor with medium effect size, and the post hoc test showed that the
involvement score increased with age.
PLOS ONE | https://doi.org/10.1371/journal.pone.0255359 July 30, 2021
6 / 16
PLOS ONE Preferences and self-perception of individuals with different levels of involvement in birdwatching
Fig 1. Percentage of birdwatchers with an engagement score.
https://doi.org/10.1371/journal.pone.0255359.g001
Involvement score were positively related with time since their activity began (years of
interest in birdwatching), as personally assessed by each respondent (χ2
Fig 2). People with more seniority (3 years and more) had higher scores than those with less
seniority. It should be emphasized that the study group differed significantly in the case of n:
the highest number of respondents declared that they were interested in birdwatching for
more than 4 years (n = 260, 73%). Respondents who were interested in birdwatching for less
than 2 years constituted only 17%.
3,353 = 9.10, P = 0.0280;
The respondents were asked about their knowledge of specific species in the area where
they intended to engage in birdwatching (Biebrza Valley). Each respondent was asked to pro-
vide four common or Latin names of birds. Involvement score were higher in people who
properly reported the names of birds (9.93 ± 0.16 SE, χ2
2,354 = 15.52, P < 0.0001) than in their
counterparts who did not (7.50 ± 0.57 SE). It should be emphasized that the study group dif-
fered significantly in the case of n: the highest number of respondents properly reported the
names (n = 317, 89%).
Table 2. Average involvement scores of birdwatchers in relation to their socio-demographic characteristics.
Socio-demographic items (options)
Involvement score
P value
F
ω2
Effect size
Gender (male, female)
Age (18–34 years; 35–54 years; �55 years)
Current place of residence (village; �100,000 residents; >100,000 residents)
Education (primary, secondary, higher)
Country of origin (native, abroad)
https://doi.org/10.1371/journal.pone.0255359.t002
First option Second option Third option
9.61 ±0.22
8.64c ±0.30
9.90 ±0.34
10.90 ±0.81
9.65 ±0.29
9.69 ±0,24
9.64b ±0.25
9.96 ±0.24
9.41 ±0.48
9.64 ±0.20
0.8125
0.0564
0.00
-------
10.81a ±0.27 <0.0001
9.16 ±0.27
0.0726
9.64 ±0.18
0.3820
0.9731
13.4301
0.07 medium
2.6422
0,9649
0.0011
0.01
0.00
0.00
small
-------
-------
PLOS ONE | https://doi.org/10.1371/journal.pone.0255359 July 30, 2021
7 / 16
PLOS ONE Preferences and self-perception of individuals with different levels of involvement in birdwatching
Fig 2. Mean involvement scores of birdwatchers in relation to time since their activity began (years of interest in birdwatching). Different
letters indicate a statistically significant difference based on the post hoc nonparametric comparisons for each pair by Wilcoxon method (P < 0.05).
https://doi.org/10.1371/journal.pone.0255359.g002
Involvement score was based on the six statements defining respondents’ motives and six
important performed activities of birdwatching and ranged from –12 to +18 (Materials and
Methods). Different letters indicate a statistically significant difference based on the post hoc
nonparametric comparisons for each pair by Wilcoxon method (P < 0.05). Data were collected
in 2018 from a sample of 357 birdwatchers visiting Biebrza National Park, Poland.
Preferences
Half of the respondents declared that they participated in field trips to observe birds more than
three times a year (n = 180). Slightly less than 30% of the respondents answered that they par-
ticipated two to three times (n = 107), and the rest that they only participated once a year
(n = 70). Involvement score results defining involvement of birdwatchers in their hobby were
positively related with the frequency of field trips related to bird watching (χ2
P < 0.0001). The highest number of points was obtained by the respondents from the groups
who participated regularly (10.16 ± 0.23 SE) or at least two to three times a year (9.84 ± 0.26
SE), and these results were not statistically significantly different. People participating in bird-
watching only once a year had 8.06 (± 0.37 SE) points and differed significantly from both
above-mentioned groups.
2,354 = 26.04,
Using a multiple-choice question, we checked what methods of observation were used by
the respondents. About three-quarters (n = 270) of respondents used a variety of optical
devices for observation. On the other hand, observations with no instruments were made by
PLOS ONE | https://doi.org/10.1371/journal.pone.0255359 July 30, 2021
8 / 16
PLOS ONE Preferences and self-perception of individuals with different levels of involvement in birdwatching
approximately 17% (n = 61). One-third of the respondents also used the method of listening to
bird sounds (n = 119). Moreover, about 17% of the respondents (n = 62) used artificial shelters
in their observations, in the form of shelters, tents or camouflage nets. Among those who had
never used an additional shelter, around 15% declared that they were considering doing so in
the near future.
Mean score results were positively related with duration of the birdwatcher’s field trips (Fig
3). People who took longer trips had much higher involvement scores. It should be emphasized
that this question was multiple choice, so it is hard to specify whether the groups differed sig-
nificantly. The highest number of respondents declared that they usually participated in trips
lasting a few hours (n = 129). For the remaining respondents, we obtained the results n = 98,
n = 110, n = 13, n = 40 for trips lasting 1 day, several days, 1 week and more than 1 week,
respectively.
Respondents were linked to their favourite observing sites, as 91% said that they often went
to the same sites to observe birds. They also declared that during birdwatching, they engaged
in other forms of recreational activity, such as walking, fishing, etc. (n = 311, 87%).
Opinions
As part of our research, we asked respondents a number of detailed questions to define their
opinion on factors determining the development of birdwatching in Poland (Table 3). In gen-
eral, it can be concluded that they think that the greatest influence on the development of bird-
watching was due to factors such as natural values of Poland, social networks and general
Fig 3. Mean involvement scores of birdwatchers in relation to duration of the field trips in which they participated.
https://doi.org/10.1371/journal.pone.0255359.g003
PLOS ONE | https://doi.org/10.1371/journal.pone.0255359 July 30, 2021
9 / 16
PLOS ONE Preferences and self-perception of individuals with different levels of involvement in birdwatching
Table 3. Responses of birdwatchers regarding selected factors determining the development of birdwatching in Poland and requirement of specialized infrastruc-
ture for birdwatching in Poland.
Do these selected factors determine
the development of birdwatching in Poland?
Does the development of birdwatching in Poland
require specialized infrastructure such as:
https://doi.org/10.1371/journal.pone.0255359.t003
Items
Natural values of Poland
n1 % yes
349
98
n2 % no
4
1
n3 % hard to say
4
1
Diverse and available for birdwatching infrastructure
173
Affordable prices of equipment helpful in observation 153
Wide range of offers from travel agencies
Popularization of birdwatching in the media
Wide offer of publications about birds
Development of social networks
enabling the exchange of experiences
General public interest in ecotourism
Observation tower
Ground observation points
Educational paths
Platforms and viewing terraces
104
215
244
290
267
326
331
294
315
48
43
29
60
68
81
77
91
93
82
88
96
95
142
92
36
26
40
21
11
36
19
27
27
40
26
10
7
11
6
3
10
5
88
110
111
50
77
41
40
10
15
27
23
25
31
31
14
22
11
11
3
4
8
6
1
2
3
4
5
6
7
8
1
2
3
4
public interest in ecotourism. Among the factors assessed, they indicated that birdwatching
infrastructure, prices of equipment helpful in observation and unsatisfactory offerings of travel
agencies potentially inhibited the development of birdwatching in Poland.
We asked two additional detailed questions. In the first question, we asked which special-
ized infrastructure was required in Poland to develop birdwatching (Table 3). The respon-
dents indicated that basically each of the proposed options of infrastructure should help in
development of their activity. Also, we asked about how often they used the currently available
offerings of tourist offices. It turned out that most of the respondents (n = 305, 85%) did not
use such offerings, and others used them sporadically (n = 47, 13%) or always (n = 5, 2%).
In order to simplify the answers from questionnaires, we changed in Table 3 ‘definitely yes’
and ‘rather yes’ to YES and changed ‘rather no’ and ‘definitely no’ to NO. Data were collected
in 2018 from a sample of 357 birdwatchers visiting Biebrza National Park.
Discussion
Methodological aspects
In the research we used a survey questionnaire. It is a very popular research tool, which can be
used to investigate the expectations and preferences of tourists and visitors to valuable nature
areas. The questionnaire was used in birdwatching studies conducted by, among others:
Dwyer [17], Williams and La Montagne [18], Hvenegaard [23]. The questionnaire used in our
study, after some modification, can be used to assess the degree of involvement in various
forms of recreational activities based on nature (for example anglers, hunting). Due to the sur-
vey, we can predict how intensively birdwatching is developing, as well as the level of environ-
mental awareness of its participants. The questionnaire we have constructed can also be used
in the future to select people who can help in ornithological research (e.g. verification of
knowledge of species, indication of breeding sites etc.). It can also be helpful for people prepar-
ing an individual offer for a tourist, adjusted to his/her abilities and level of involvement. The
advantage of surveys is that they allow for relatively easy and quick testing, and that they guar-
antee the anonymity of respondents. However, it is not an ideal tool, it is very difficult to
ensure the truthfulness of the data obtained. Therefore, we decided to use the form of inter-
views, in which questions were read out and the interviewer recorded the respondents’
PLOS ONE | https://doi.org/10.1371/journal.pone.0255359 July 30, 2021
10 / 16
PLOS ONE Preferences and self-perception of individuals with different levels of involvement in birdwatching
answers. Direct contact makes it easier for the researcher to determine whether the answers
obtained can be treated as reliable.
General characteristics of respondents
In our study, birdwatchers were predominantly male, similar to those of Frątczak et al. [13]
and Scott and Thigpen [21]. In the study of Ellis and Vogelsong [6], women were slightly
over-represented, but, as the authors explained, this was probably due to the fact that in the
male-female two-person groups encountered in the field, most often only one questionnaire
was returned, and it was completed by a woman. Also, results from Adams et al. [36], Lee
et al. [28] and Conradie [25] indicated that the proportion of women interested in bird-
watching was higher than men. However, in both articles of Lee et al. [28] and Conradie
[25] the respondents were not, as in our study, birdwatchers met in the field, but partici-
pants in festivals, large events organized with avitourism in mind. Certain demographic
characteristics, mainly gender and age, in addition to socio-economic characteristics, are
linked with cultural consumption patterns and participation in cultural events. Women, in
general, are known to be more active consumers of cultural products than men [37]. Also,
mature aged individuals were more represented among tourists particularly at some attrac-
tions such as cultural festivals [37]. A study by Kim et al. [38] found that women were 1.28
times more likely than men to attend local festivals and fairs. This might be the reason for
the difference between the findings of Lee et al. [28] or Conradie [25] and our results in rela-
tion to the gender of birdwatchers. In our study, the proportion of older people, �55 years
old, was approximately one-third of the total. Research by Dwyer [17] showed that partici-
pation in birdwatching increased with the age of observers. Birdwatching is mainly prac-
ticed by people aged 45–64 years. The respondents aged 18–24 years showed the least
interest in birdwatching. This observation is confirmed by the studies of Ellis and Vogel-
song [6] and Williams and La Montagne [18].
In our study, more than 80% of the respondents had a university degree. Also in the studies
of Adams et al. [36] and Lee et al. [28], more than 70% of birders declared that they had a bach-
elor’s degree or higher.
In our study, approximately one-third of respondents each lived in towns, cities and vil-
lages. The results of Skłodowski and Jurkowska [39] indicated that Polish birdwatchers were
predominantly city dwellers, especially in cities with more than 500,000 inhabitants. In turn,
Dwyer’s [17] research indicated that among birdwatching participants, the most numerous
group were residents of cities with populations below 50,000. In general, Dwyer [17] suggested
that individuals living in areas with populations of less than 50,000 were more likely to partici-
pate in activities generally associated with substantial wildland areas than were individuals liv-
ing in areas with larger populations. Activities requiring the development of specialized
facilities or programs are more likely to involve individuals living in large urban areas, where
there are usually substantial recreation facilities and programs. In Poland, due to socio-eco-
nomic conditions, the tourism activity of rural residents is lower than that of urban residents
[40]. Besides, as Seweryn and Niemczyk’s [41] study showed, rural residents preferred destina-
tions with a different character than the environment in which they live every day (i.e., cities).
Among the birdwatchers we surveyed, the largest group were Poles, but it is noteworthy that
almost 30% of the respondents were residents of other European countries, mainly those
(Great Britain, Germany, the Netherlands, Belgium) where birdwatching is developing very
dynamically [13]. It could be stated that our respondent group is very similar to those
described in previous reports from Poland, as well as to groups from other countries with a
similar culture.
PLOS ONE | https://doi.org/10.1371/journal.pone.0255359 July 30, 2021
11 / 16
PLOS ONE Preferences and self-perception of individuals with different levels of involvement in birdwatching
Involvement score
In Adams’ [36] study, the two most important factors for birding were ‘to be close to nature’
and ‘fascination with birds’. These two factors also resonate in our research. Summing up first
two items from Table 1, it can be said that the research group is quite homogeneous and is def-
initely aware of the relationship between birdwatching and nature. On the one hand, it is clear
that birdwatching is a form of outdoor recreation activity, which, like fishing, backpacking,
camping in a tent or hunting, for example, is strongly rooted in nature. On the other hand, it is
also an activity that involves emotional, spiritual, physical and mental stimulation [25]. Fasci-
nation with birds leads to greater engagement, to acquiring more degrees of knowledge and to
becoming a more professional ‘obsessive’ birdwatcher. It is therefore not surprising that the
vast majority of respondents in our study disagreed with the statement that their hobby is a fad
or that they wanted to impress others with their unique hobby. Birdwatching is a way of life
and, at the same time, an activity that guarantees contact with nature and helps people to relax
and unwind. The fascination with birds involves the search for ‘souvenirs’ or ‘trophies’ that
prove that the visitor has actually visited a place or seen a particular species [12]. Hence, more
than half of the respondents to our survey believe that it was important to take unique photos
while actively birdwatching. Taking photographs is an important part of birdwatchers’ activi-
ties. This is also confirmed by the research of Skłodowski and Jurkowska [39]. Birdwatchers
are people with high environmental awareness and nature sensitivity. They include both those
who start their birdwatching by feeding birds (the highest response rate) and those who are
characterized by a desire to obtain nature knowledge and also to share it with other people
(just over one-third of respondents). More than half of the respondents in our study read pop-
ular scientific articles on ornithology, and approximately 25% of the respondents belonged to
various associations and groups of people with similar ornithological interests.
One of the goals of our work was to address the lack of an appropriate tool to assess bird-
watcher involvement in their activity. Researchers most often use various nominal scales or
leave the assessment of involvement to the respondents. Usually, the use of such a scale pre-
vents accurate comparisons of the results and some are indicative, and so we proposed a differ-
ent approach. In this research, we used a different type of scale often used in other studies, an
interval scale, to quantify involvement by assigning a point value to each respondent. It turned
out that the values obtained, expressed by the involvement scores, corresponded well with the
results of measurements of individual traits indicating involvement in birdwatching (e.g.,
increase of involvement scores of birdwatchers in relation to prolongation of duration of the
field trips in which they participate). Therefore, we want to emphasize that the illustrated result
in the form of a distribution close to normal (Fig 1) allows us to conclude that the set of issues
used to express motives (six statements defining respondents’ motives) as well as activities per-
formed (six important performed activities related to birdwatching) may be used in the future,
during other studies. For example, similar tools have been created to unify the results achieved
in the case of research on attitude toward hunting (attitude toward hunting score [42, 43]).
Detailed characteristics, preferences and opinions of the respondents
Birdwatchers had a good knowledge of birds occurring in the area where they intend to practice
their hobby (Biebrza Valley). They knew which bird species occur in this area, and they were
determined to observe specific bird species, mainly Acrocephalus paludicola (Vieillot 1817), Phi-
lomachus pugnax (Linnaeus 1758), Crex crex (Linnaeus 1758), Anser erythropus (Linnaeus
1758) and Clanga clanga (Pallas 1811). Clearly, these were not casual tourists but people
engaged in their hobby, devoting significant amount of time to it. It seems that Biebrza Valley
met their expectations in this respect, as most of the respondents declared that they often went
PLOS ONE | https://doi.org/10.1371/journal.pone.0255359 July 30, 2021
12 / 16
PLOS ONE Preferences and self-perception of individuals with different levels of involvement in birdwatching
to the same places to observe birds. We also observed that involvement in birdwatching
increased with age. This fact is confirmed by Frątczak et al. [13], who found that there were
more older people among professional observers than among novice birdwatchers. This
involvement increased with the length of the birdwatching period. The involvement of bird-
watchers in their hobby was related to the frequency of field trips to observe birds. The majority
of birdwatchers participated in nature observation trips more than three times a year, with trips
most often lasting several hours. Also, research by Skłodowski and Jurkowska [39] indicated a
prevalence of expeditions lasting mostly a few hours. With many hours of interaction with
nature and focus, many experience a heightened sense of awareness and a higher level of inter-
action with the natural world. During expeditions, birdwatchers used, as both our studies have
shown, many types of equipment, mostly optical devices. Research by Skłodowski and Jur-
kowska [39] found that the amount of equipment used increased with the level of professional-
ism of the birding participant. Birding is a broad concept, involving other senses, such as
hearing, in addition to sight [15]. Hence, over 30% of our respondents also used the method of
listening to bird sounds when birdwatching. The increased interest in birdwatching, or wildlife
watching more broadly, leads to the development of a specialized leisure industry. There are sit-
uations when being invisible is the only way to observe rare and valuable species up close.
Hence, the number of birdwatchers interested in using artificial shelters, tents or camouflage
nets in their observations is also increasing, as shown, among others, by our research. The
majority of our respondents participated in other activities when on a birding trip, such as walk-
ing or fishing. Also, a study by Conradie [25] showed that apart from engaging in birdwatching,
birding tourism participants also pursued other activities such as observing wildlife, other ani-
mals, trees, wild flowers and butterflies. Interestingly, water-related recreational activities such
as diving, snorkelling, beachcombing and boating were not as important.
Our research showed that the development of birdwatching corresponded with natural val-
ues. Research by Steven et al. [44] showed that birders were attracted to sites that provide high
levels of biodiversity and presence of endemic species. With birdwatchers in mind, special
facilities are being developed in the field (e.g., paths with points/chats for observing and photo-
graphing from hidden places, viewing platforms). This infrastructure makes it possible to mul-
tiply experiences, aesthetic sensations and provides unforgettable impressions [15]. Our
research showed that respondents considered such facilities to be helpful in the development
of the hobby of birdwatching, pointing above all to viewing towers and ground observation
posts as the most desirable type of infrastructure. Ellis and Vogelsong [6] suggested that recrea-
tional infrastructure such as well-marked, accessible trails; quality signage; observation towers;
and observational blinds could increase birders’ satisfaction. This is supported by Conradie
[25], whose research showed that the most important attributes at birding destinations were
accessible walking trails, information about birds, bird lists and possibility of spending time in
bird hides.
Our research showed that the vast majority of birdwatchers did not use the offerings of
tourist offices. Also Conradie [25] suggested that avitourists prefer to travel in pairs, small
groups or independently. This means that birdwatching fits into the general trends of tourism
development in Poland, set out in the government document “Tourism Development Pro-
gramme to 2020” [45], accompanied by a change in tourists’ expectations. Visitors increasingly
expect services tailored to their individual needs, providing a more authentic experience.
Conclusion
An important tool described in this article is the new scale that allows assessment of the level
of involvement of individual people engaged in birdwatching. This scale corresponded well
PLOS ONE | https://doi.org/10.1371/journal.pone.0255359 July 30, 2021
13 / 16
PLOS ONE Preferences and self-perception of individuals with different levels of involvement in birdwatching
with the individual characteristics of birdwatchers. Its statements were created based on the
experience of the authors and literature. Statistically, the most common birdwatcher was male,
middle-aged, living in a large city. Most birdwatchers defined their birdwatching activity as a
permanent rather than a temporary hobby, and it was therefore considered to be more of a life-
style. An important fact about the birdwatchers’ community is that people’s involvement in
this activity increased with their age. It was also observed that the level of involvement in bird-
watching activity was related to the frequency of trips. Birdwatchers also used equipment
when observing birds. It has also been proven that certain natural values were related to the
development of birdwatching in the opinion of the respondents and that the development of
birdwatching in the future will require a developed infrastructure enabling interaction with
the objects of observation. These insights are important for the future development of bird-
watching tourism and can help guide the development of regional development strategies in
the future.
Supporting information
S1 File. Survey questionnaire.
(DOCX)
Acknowledgments
We would like to thank the Biebrza National Park personnel for their help in creating the sur-
vey questionnaire and collecting the research material. Their knowledge and experience were
very useful to us. We would like to thank the two anonymous reviewers for their comments,
which were helpful in improving the final version of the article.
Author Contributions
Conceptualization: Emilia Janeczko, Małgorzata Woźnicka, Krzysztof Janeczko.
Formal analysis: Adrian Łukowski, Ernest Bielinis, Małgorzata Woźnicka, Krzysztof Janeczko,
Natalia Korcz.
Investigation: Małgorzata Woźnicka, Krzysztof Janeczko, Natalia Korcz.
Methodology: Emilia Janeczko, Adrian Łukowski, Ernest Bielinis, Małgorzata Woźnicka,
Krzysztof Janeczko, Natalia Korcz.
Resources: Emilia Janeczko, Adrian Łukowski, Ernest Bielinis.
Supervision: Emilia Janeczko.
Writing – original draft: Emilia Janeczko.
References
1. Go¨ssling S. 1999. Ecotourism: A means to safeguard biodiversity and ecosystem functions? Ecol.
Econ. 29, pp. 303–320
2. Coccossis H.; Mexa A.; Collovini, A. 2002. Defining, measuring and evaluating carrying capacity in
european tourism destinations, B4-3040/2000/294577/MAR/D2, University of the Aegean, Athens
3. Das M. Chatterjee B. 2015. Ecotourism: A panacea or a predicament? Tour. Manag. Perspect. 2015,
14, 3–16
4. Carvache-Franco M.; Segarra-Oña M.; Carrascosa-Lo´ pez C. 2019. Segmentation by Motivation in Eco-
tourism: Application to Protected Areas in Guayas, Ecuador. Sustainability; 11, pp. 1–19. Available
online: https://doi.org/10.3390/su11010240
PLOS ONE | https://doi.org/10.1371/journal.pone.0255359 July 30, 2021
14 / 16
PLOS ONE Preferences and self-perception of individuals with different levels of involvement in birdwatching
5. Poudel J., Munn I., A., Henderson J. E. 2017. Economic contributions of wildlife watching recreation
expenditures (2006 & 2011) across the U.S. south: An input-output analysis. Journal of Outdoor Recre-
ation and Tourism 17 (2017) 93–99
6. Ellis L., C., Vogelsong F. 2004. Measuring birdwatcher preferences through importance- performance
analysis. Murdy, James, comp., ed. 2004. Proceedings of the 2003 Northeastern Recreation Research
Symposium. Gen. Tech. Rep. NE-317. Newtown Square, PA: U.S. Department of Agriculture, Forest
Service, Northeastern Research Station, 203–210
7. Vincent V. C. and Thompson W. 2002. Assessing community support and sustainability for ecotourism
development. Journal of Travel Research, 41, 153–160
8.
Lee C.K., Mjelde J.W. 2007. Valuation of ecotourism resources using a contingent valuation method:
The case of the Korean DMZ. Ecological Economics, 63(2–3), 511–520.
9. Connell J. 2009. Birdwatching, twitching and tourism: towards an Australian perspective. Australian
Geographer 40 (2), 203–217
10.
Li F., Zhu Q., Yang Z. 2013. Birding tourism development in Sichuan, China. Tourism Economics 19
(2), 257–273
11. Ch Randler. Tryjanowski P., Jokima¨ ki J., Kaisanlahti-Jokima¨ki M. J, Staller N. 2020. SARS-CoV2
(COVID-19) Pandemic Lockdown Influences Nature-Based Recreational Activity: The Case of Birders.
Int. J. Environ. Res. Public Health 2020, 17, 7310
12. Kronenberg J. 2014. Environmental Impacts of the Use of Ecosystem Services: Case Study of Bird-
watching. Environmental Management (2014) 54:617–630 https://doi.org/10.1007/s00267-014-0317-8
PMID: 24993794
13.
Frątczak M, Sparks TH, Randler C, Tryjanowski P. 2020. Circadian preferences of birdwatchers in
Poland: do “owls” prefer watching night birds, and “larks” prefer daytime ones? PeerJ 8:e8673 https://
doi.org/10.7717/peerj.8673 PMID: 32175191
14. Cordell H. Ken; Herbert Nancy G. 2002. The Popularity of Birding is Still Growing. Birding. 34, 54–61
15.
Janeczko E., Anderwald D. 2011. Birdwatching jako nowa forma turystyki na obszarach leśnych i przyr-
odniczo cennych w Polsce. Studia i Materiały Centrum Edukacji Przyrodniczo-Leśnej. Zeszyt 2(27):
301–308
16. Steven R., Morrison C., Castle J.G. 2014. Birdwatching and avitourism: a global review of research in
its participant markets, distribution and impacts, highlighting future research priorities to inform sustain-
able avitourism management. Journal of Sustainable Tourism, https://doi.org/10.1080/09669582.2014.
924955
17. Dwyer J., F. 1993. Customer diversity and the future demand for outdoor recreation. Proceedings of the
1993 Northeastern recreation research symposium, 1993 April 18–20, Saratoga Springs, NY. Gen.
Tech.Rep. NE-185. PA: U.S.Department of Agriculture, Forest Service, Northeastern Forest Experi-
ment Station: 59–63
18. Williams S., La Montagne K. 2001. Birding in the United States: A Demographicand Economic Analysis,
U.S. Fish and Wildlife Service, Washington.
19. Vaske J. J., Wittmann K., Williams T.V., Hardesty K., Sikorowski L. 2001. Wildlife Viewing in Colorado:
A Review and Synthesis of Existing Data. Project Rep. for the Colo. Div. of Wildlife. Human Dimensions
in Nat. Res. Unit Rep. No. 33, Colorado State Univ., Fort Collins. 41
20.
Jones D., N. 2001. Birdwatching tourism in Australia. Series: Wildlife tourism research report, Coopera-
tive Research Centre (CRC) for Sustainable Tourism, no. 10
21. Scott D., Thigpen J. 2003. Understanding the birder as tourist: Segmenting visitors to the Texas Hum-
mer/Bird Celebration. Hum. Dimens. Wildl. 2003, 8, 199–218
22. Moore R.L.; Scott D.; Moore A. 2008. Gender-based diferences in birdwatchers’ participation and com-
mitment. Hum. Dimens. Wildl. 2008, 13, 89–101
23. Hvenegaard G.T. 2002. Birder specialization diferences in conservation involvement, demographics,
and motivations. Hum. Dimens. Wildl. 2002, 7, 21–36
24. Naidoo R. and Adamowicz W. L. 2005. Biodiversity and nature-based tourism at forest reserves in
Uganda. Environment and Development Economics, 10, 159–178
25. Conradie N. 2015. Profiling the international avitourist: preferences of avitourists at the British and
Dutch birdwatching fairs. African Journal of Hospitality, Tourism and Leisure Vol. 4 (1)
26.
27.
Jin J.J., Wang Z.S., Liu X.M. 2008. Valuing black-faced spoonbill conservation in Macao: A policy and
contingent valuation study. Ecological Economics, 68 (1–2), 328–335
Lee C.K., Lee J.H., Mjelde J.W., Scott D., Kim T.K. 2009. Assessing the economic value of public bird-
watching interpretive service using a contingent valuation method. International Journal of Tourism
Research, 11(6), 583–593
PLOS ONE | https://doi.org/10.1371/journal.pone.0255359 July 30, 2021
15 / 16
PLOS ONE Preferences and self-perception of individuals with different levels of involvement in birdwatching
28.
Lee C. K., Lee J. H., Kim T. K., Mjelde J., W. 2010. Preferences and willingness to pay for bird-watching
tour and interpretive services using a choice experiment. Journal of Sustainable Tourism 18: 5, 695–
708, https://doi.org/10.1080/09669581003602333
29. Boxall P. C., McFarlane B. L. 1993. Human dimensions of Christmas Bird Counts: Implications for non-
consumptive wildlife recreation programs. Wildlife Society Bul- letin, 21, 390–396
30. Scott D., Ditton R.B., Stoll J.R., and Eubanks T.L. 2005. Measuring specialization among birders: utility
of a self- classification measure. Human Dimensions of Wildlife: An International Journal, 10(1):53
31. Anderwald D. 2007. Turystyka ornitologiczna–co to takiego? Turystyka w obszarach Natura 2000. W:
Wnuk Z., Ziaja M. (red.) Wyd. Uniwersytet Rzeszowski, Rzeszo´w: 275–288.
32. Gundersen V., Vistad I., O. 2016. Public opinions and Use of Various Types of Recreational Infrastruc-
ture in Boreal Forest Settings. Forests 2016, 7, 113; https://doi.org/10.3390/f7060113
33. Reichhart T.; Arnberger A. 2010. Exploring the influence of speed, social, managerial and physical fac-
tors on shared trail preferences using a 3D computer animated choice experiment. Landsc. Urban Plan.
2010, 96, 1–11.
34. Vaske J. 2008. Survey research and analysis: applications in parks, recreation and human dimensions.
Venture Publishing, State College, Pennsylvania, USA.
35.
Field A. 2013. Discovering Statistics with IBM SPSS Statistics. Newbury Park, CA: Sage
36. Adams C. E., Leifester J. E. And Herron J. S. C. 1997. Understanding wildlife constituents: Birders and
waterfowl hunters. Wildlife Society Bulletin, 25, 653–660
37.
Zeppel H., Hall M. 1991. Selling art and history: Cultural heritage and tourism. Journal of Tourism Stud-
ies, 2 (1) (1991), pp. 29–45
38. Kim H., Cheng C-K., T.O’Leary J. 2007. Understanding participation patterns and trends in tourism cul-
tural attractions Tourism Management. Volume 28, Issue 5, pp. 1366–1371, https://doi.org/10.1016/j.
tourman.2006.09.023
39. Skłodowski J., Jurkowska A. 2015. Charakterystyka sylwetki i zainteresowań uczestniko´ w turystyki
birdwatchingowej w Polsce. Studia i Materiały Centrum Edukacji Przyrodniczo-Leśnej. Zeszyt 17(45):
203–208
40. Niemczyk A. 2015. Czynniki geograficzne jako determinanta wyboru miejsca docelowego podro´ży w
procesie zaspokajania potrzeb turystycznych, „Marketing i Rynek”, nr 8.
41. Seweryn R., Niemczyk A. 2018. Turystyka miejska mieszkańco´w wsi (na przykładzie turysto´ w odwied-
zających Krako´w). Zesz. Nauk. UEK; 4(976): 39–58
42.
Ljung PE, Riley SJ, Heberlein TA, Ericsson G, 2012. Eat prey and love: game-meat consumption and
attitudes toward hunting. Wildlife Soc B 36(4):669–675. https://doi.org/10.1002/wsb.208
43. Krokowska-Paluszak M., Łukowski A., Wierzbicka A., Gruchała A., Sagan J., Skorupski M. 2020. Atti-
tudes towards hunting in Polish society and the related impacts of hunting experience, socialisation and
social networks. European Journal of Wildlife Research, 66(5), 1–8.
44. Steven R., Smart J. C. R., Morrison C., Castley J., G. 2017. Using a choice experiment and birder pref-
erences to guide bird-conservation funding. Conserv Biol. 31(4):818–827. https://doi.org/10.1111/cobi.
12849 PMID: 27696549
45. Program rozwoju turystyki do 2020—dokument przyjęty Uchwałą nr 143/2015 Rady Ministro´ w z dnia 18
sierpnia 2015 r.
PLOS ONE | https://doi.org/10.1371/journal.pone.0255359 July 30, 2021
16 / 16
PLOS ONE |
10.1371_journal.pone.0258085 | RESEARCH ARTICLE
Level of engagement of recreational physical
activity of urban villagers in Luohu, Shenzhen,
China
Lu ShiID
1*, Willie Leung2, Qingming Zheng3, Jie Wu3
1 Public Health, School of Social and Behavioral Health Science, College of Public Health and Human
Sciences, Oregon State University, Corvallis, OR, United States of America, 2 Department of Health
Sciences and Human Performance, College of Natural and Health Sciences, The University of Tampa,
Tampa, FL, United States of America, 3 Shenzhen Luohu Disease Prevention and Control Center,
Shenzhen, Guangdong, China
* shil@oregonstate.edu
Abstract
Physical activity is important for health. However, there is a lack of literature related to the
physical activity levels of adults living in urban villagers, which is a vulnerable population in
China. The aim of this study is to compare the physical activity and sedentary behavior
engagements between urban villagers and non-urban villagers using the 2019 Luohu Shen-
zhen, China Community Diagnosis Questionnaire. A total of 1205 adults living in urban vil-
lages and non-urban villages were included in the analysis. Unadjusted and multiple
multivariate logistic regression were conducted for the dependent variable of engagement in
recreational physical activity, frequency of recreational physical activity per week, and hours
spent in sedentary behaviors per day. Descriptive analysis was conducted to identify the
reasons for not engaging in physical activity among urban villagers and non-urban villagers.
Across the included sample, 29.05% were urban villagers and 70.95% were non-urban vil-
lagers. The results suggested that urban villagers are more likely to engage in physical activ-
ity than non-urban villager (OR = 1.90, 95% CI [1.40, 2.59], p < 0.001). However, it was also
found that urban village status had no significant association for frequency in engaging in
physical activity and average hours spent in sedentary behaviors. Both urban villagers and
non-urban villages indicated that lack of time, lack of safe and appropriate environment, and
working in labor intensive occupations as some of the reasons for not engaging in physical
activity. There is a need for tailed interventions and policies for promoting physical activity
among urban villagers and non-urban villagers. Additional studies are needed to further our
understanding of the physical activity behaviors among urban villagers in China.
Introduction
Benefits of physical activity
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: Shi L, Leung W, Zheng Q, Wu J (2021)
Level of engagement of recreational physical
activity of urban villagers in Luohu, Shenzhen,
China. PLoS ONE 16(10): e0258085. https://doi.
org/10.1371/journal.pone.0258085
Editor: Francisco Javier Huertas-Delgado, La
Inmaculada Teacher Training Centre (University of
Granada), SPAIN
Received: November 21, 2020
Accepted: September 20, 2021
Published: October 28, 2021
Copyright: © 2021 Shi et al. This is an open access
article distributed under the terms of the Creative
Commons Attribution License, which permits
unrestricted use, distribution, and reproduction in
any medium, provided the original author and
source are credited.
Data Availability Statement: Due to legal and
participants confidentiality, data will only be
available upon request. The data underlying the
results presented in the study are available from
Shenzhen Luohu Disease Prevention and Control
Center via contacting Weihong Chen, director of
Shenzhen Luohu Disease Prevention and Control
Center, at 1433529760@qq.com.
Funding: The authors received no specific funding
for this work.
The benefit of engagement of physical activity is well documented [1, 2]. The numerous bene-
fits included weight management, lower blood cholesterol levels and blood pressure,
PLOS ONE | https://doi.org/10.1371/journal.pone.0258085 October 28, 2021
1 / 17
PLOS ONE Competing interests: The authors have declared
that no competing interests exist.
Urban villagers’ physical activity levels
strengthening bones, muscles, and joint, and reducing risk of cardiovascular disease and cer-
tain types of cancers [3]. In addition to physical health-related benefits, engagement in physical
activity could lead to benefits of social and mental benefits. Regular engagement in physical
activity is associated with reduced stress, improved mental health, emotional regulation, low-
ered depression, increased social functioning, and increased sense of community [4]. Further,
engagement of regular physical activity is related to reduce the risk of developing disabilities
and maintenance of functional independences [5, 6].
Currently, physical inactivity is the fourth leading cause of mortality, according to the
World Health Organization (WHO) [7]. WHO’s physical activity guidelines are 150 minutes
of moderate physical activity or 75 minutes of vigorous physical activity per week or an equiva-
lent combination of moderate- and vigorous-intensity activity for adults [7]. Individuals can
perform various activities, such as leisure time physical activity, active transportation, and
occupational activities to accumulate the minutes required to meet the guidelines. These
guidelines apply to all individuals regardless of gender, race, ethnicity, or income levels.
Physical activity levels of Chinese people
Past literature had examined the physical activity levels individuals living in China [8, 9].
Using the data from the 2012 to 2015 China Hypertension Survey (CHS), it was found that
28.1% of Chinese adults were overweight and 5.2% were obese [10]. The results also found that
regionals different of the prevalence of overweight and obesity different between Northern and
Southern China with adults from Northern China more likely to be obese and overweight.
According to a report published in the official Report on Cardiovascular Diseases in China
2017, 290 millions of Chinese adults are suffering from cardiovascular disease [11]. It was also
found that China is facing a fast growing cardiovascular disease epidemic with a widening
rural-urban disparities [12].
Similar physical activity trends found in Western countries were observed among Chinese
adults as well. Trends such as male are more likely to engage in physical activity than female
and older adults are less physical active than younger adults were found among individuals liv-
ing in China [8, 9]. It was found that 66.3% of adults between the ages of 35 to 74 years were
physically active according to the data from the International Collaborative Study of Cardio-
vascular Disease in Asia from 2000–2001 [9]. Using accelerometers to measure physical activ-
ity, it was found that Chinese adults in Shanghai spent 317 minutes per day in physical activity,
while spent 509 minutes per day in sedentary behaviors [13]. It was reported that Chinese
adults are more likely to report engaging in work-related or occupational physical activity
(63.3%) than leisure time physical or recreational physical activity (24.5%) [9]. There were dis-
parities between urban and rural residents with more rural residents (78.1%) spending time in
physical activity than urban residents (21.8%) [9]. In addition to regional different, it was
found that socioeconomic status (SES) impact physical activity levels among Chinese adults
[14]. Using a community-based survey with 3567 adults living in Jiaxing, China, Chen et al.
found that adults with lower SES are more likely to engage in household physical activity,
adults with middle SES engages in higher intensity of occupational physical activity, and adults
with higher SES levels were more likely to exercise but spent longer time in sedentary behav-
iors [14].
The physical activity of subpopulation of Chinese adults had been well examined, especially
for adults with different living area (rural vs. urban) and SES [14, 15]. However, there is a lack
of literature examining the physical activity levels of urban villagers. Urban villagers refer to
the individuals living in urban village. Urban village or chengzhongcun are typically low quality
and high density with many closely packed apartment blocks of between 2 and 8 floors [16].
PLOS ONE | https://doi.org/10.1371/journal.pone.0258085 October 28, 2021
2 / 17
PLOS ONE Urban villagers’ physical activity levels
Urban villages are transitional neighborhoods typically found in urban areas or cities with
rapid economic growth [16, 17]. Urban villages can be described as narrow roads, face-to-face
buildings, a thin strip of sky, and inner streets packed with shops, grocery stores and service
outlets [16]. Many of these urban villages are associated with unsuitable land use, poor housing
construction, severe infrastructure deficiencies, intensified social disorder, and deteriorated
urban environment [18]. In addition, urban villages often have poor sanitary condition, where
pipelines and drainage systems are poorly constructed and water flows over the ground along
with garage [17]. Many urban villagers are individuals with low SES status due to financial situ-
ation. These urban villagers could include rural-to-urban migrants workers with limited skill-
sets and educations or individuals who recently graduated from colleges and universities. They
are attracted to urban villages due to the cheap housing accommodation. Overall, these urban
villagers aggregate in urban village in large cities, such as Guangzhou and Beijing with limited
infrastructure and poor living environments due to affordable living accommodations.
Due to the unique living situations of urban villages and limited healthcare resources [19],
urban villagers’ physical activity need to be better examined [20]. Knowing physical activity-
related information of urban villagers could better design and develop interventions targeting
the needs of urban villagers in the community. Regular engagement in physical activity is asso-
ciated with better health-related outcomes [21], considering urban villagers is more at risk for
poor health outcomes due to poor living situation [22, 23]. Previous studies had examined the
physical activity levels of youths and adolescents living in urban village [24, 25]. Therefore, to
better understand the physical activity levels of adult urban villagers, the purpose of this study
is to compare the physical activity and sedentary behaviors engagements between urban villag-
ers and non-urban villagers using the 2019 Luohu Shenzhen, China Community Diagnosis
Questionnaire.
Materials and methods
Design and sample
This study is secondary data analysis using data from the 2019 Luohu Shenzhen, China Com-
munity Diagnosis Questionnaire. The questionnaire is part of a community health diagnosis
program funded by the Center for Disease Control and Prevention of Shenzhen. Due to the
unique status of Shenzhen as the Special Economic Zones (SEZ), it attracted various Chinese
citizens with different background to settle in the areas. This allows assessments of health-
related behaviors on various groups of Chinese citizens (e.g., household registration status,
migrants status, employments status, income levels, etc.) within the same survey and living
within the same area. The goal of the survey is to grasp the main health problems existing in
the residents of Luohu District, determine the causes of community health problems, and
determine the priority needs of the public health services and factors affecting residents’
health. The survey also served as an evaluation of Shenzhen residents satisfaction on the vari-
ous healthcare institutes available to them, such as community health centers. The survey con-
sisted of seven parts: 1) family demographics, 2) family medical history, 3) adults healthcare
needs and access to healthcare, 4) health and quality of life of adults over the ages of 60 years
old, 5) health, healthcare and reproductive healthcare needs of married women under the ages
of 50 years old, 6) healthcare needs and health of children, and 7) examination of blood pres-
sure, height, weight, hip length, and waist length. Data collection of the survey was approved
by the IRB at Shenzhen Luohu Disease Prevention and Control Center. Analysis of the survey
data was approved by the IRB at Oregon State University.
Participants of the survey were selected by multiple stages of random selection. First, seven
communities were randomly selected in Dongmen community, Luohu district, Shenzhen,
PLOS ONE | https://doi.org/10.1371/journal.pone.0258085 October 28, 2021
3 / 17
PLOS ONE Urban villagers’ physical activity levels
Luohu as seen in Fig 1. Then 116 community grids were randomly selected from the seven
selected communities in Dongmen community, Luohu district. Lastly, family household, serv-
ing as survey unit, were randomly selected for interview based on the size of the community. All
members of the household participated in the survey. Further, only individuals living in Shen-
zhen for at least six months prior to the interview were included in the survey. The number of
household participants in the survey is based on the size of the community. 200 households
were randomly selected if the community sample size have more than two million individuals,
150 households for community sample size between one to two million, 100 households for
community sample size between half of a million to one million, and 50 households for commu-
nity less than half of a million. The random selection of communities was to identify individuals
living in the various type of communities within the Shenzhen area. All data were collected
between January and September of 2019. All data were collected through face-to-face interview.
A total of 2122 participants were interviewed for the survey. However, only 2089 participants
completed the survey with valid data. 1205 adults were included in the analysis.
Across the sample, 54.52% of the participants were female and 45.48% of the participants
were male. The average age of the participants were 38.8 years old. The average BMI were 22.88
Fig 1. Participants recruitment process.
https://doi.org/10.1371/journal.pone.0258085.g001
PLOS ONE | https://doi.org/10.1371/journal.pone.0258085 October 28, 2021
4 / 17
PLOS ONE Urban villagers’ physical activity levels
kg/cm2 with the hip-to-waist ratio of .89. A majority of the participants were employed
(83.24%). Interestingly, 32.20% of the participants have no formal education and completed pri-
mary education, which made up almost of one third of the sample. 28.54% of the participants
completed middle school, 26.97% completed high school, and 12.28% completed professional
school, college, and university. 76.02% of the participants were married or partnered and
23.98% were singled or not partnered. The sample consisted of more participants with non-
Shenzhen hukou (62.41%). Across the sample, there were more participants without diagnosis
of hypertension and diabetes. Only 6.56% and 1.91% of the participants reported having hyper-
tension and diabetes, respectively. 76.43% of the participants reported they did not smoke.
Measures
The independent variable of the analysis is the living status of the participants. The variable is
based on the location of the community grids participants resides in. Shenzhen used the com-
munity grid system to identify local community [26]. Urban villages are typically located
within one grid. Therefore, community grids serve as an indicator for urban villages. Partici-
pants were classified as urban villagers if they live in an urban village and all other participants
were classified as non-urban villagers.
Physical activity and sedentary behavior-related variables from the survey component
about health and quality of life of adults from 18 to 59 years old were selected for this analysis.
A total of four variables were determined to be related to physical activity from the survey. The
variables were engagement in recreational physical activity, frequency of recreational physical
activity per week, hours spend in sedentary behaviors per day, and reasons for not engaging in
physical activity. All variables were categorical variables. Engagement in recreational physical
activity was based on the question of “Within the past six months, what types of recreational
physical activity did you participated in?”. The respond options included: 1) did not participate
in any activities, 2) machine equipment physical activity, 3) aerobic activity or aerobic dances,
4) swimming, 5) ambulatory activity (e.g., brisk walking, jogging, running, hiking), 6) ball-
related sports (e.g. basketball, baseball, soccer, etc.), 7) sports or fitness competition, 8) martial
arts, or 9) other. Participants were considered not to be engaged in physical activity when they
responded with did not participate in any activities, else participants were classified as engaged
in physical activity. Recreational physical activity is defined as physical activity that is done at
leisure time. The variable of frequency of recreational physical activity were based on the ques-
tion of “Within the past six months, how often do you exercise per week?” with the respond
options of 1) 6 or more times per week, 2) 3 to 5 times per week, 3) 1 to 2 times per week, and
4) lesser than 1 time. The variable of hours spend in sedentary behaviors was based on the
respond to the question of “In the past month, what is the average accumulated hours spend in
sedentary activities (e.g., studying, working, watching TV, using computer, etc.)?”. The
respond options included 1) lesser than 2 hours per day, 2) 2 to 4 hours per day, 3) 4 to 8
hours per day, 4) 8 to 12 hours per day, and 5) more than 12 hours per day. Reasons for not
engaging in physical activity were only for participants who responded that they engaged in
physical activity within the past six months. The survey item aims to identify how prevent
them from engaging in physical activity throughout their routine. Participants were asked the
reasons when they were unable to engage in physical activity weekly. Participants were able to
select multiple options of 1) no recreational physical activity is needed due to labor intensive
occupations, 2) no time to engage in physical activity, 3) there were no appropriate places and/
or environments for physical activity, 4) I feel healthy, I do not need physical activity, 5) do
not want to engage in physical activity, 6) feeling ill, unable to participate in physical activity,
and 7) other reasons.
PLOS ONE | https://doi.org/10.1371/journal.pone.0258085 October 28, 2021
5 / 17
PLOS ONE Urban villagers’ physical activity levels
The covariates included in the analysis were gender, age, employment status, education,
marital status, household registration, body mass index (BMI), diagnosis of hypertension,
diagnosis of diabetes, and smoking status. Gender was a binary variable consisted of male and
female. Age was a continuous variable between 18 to 59 years old. Employment status was a
binary variable of being employed or unemployed. Education was a categorical variable
including professional college and university, high school, middle school, and primary school
or no formal education. Marital status was a binary variable of either being married or single.
Household registration or Hukou were based on participants self-reporting their registration
of either Shenzhen Hukou or non-Shenzhen Hukou. BMI is a continuous variable between
15.02 to 36.11 kg/cm2, which was calculated based on the participants’ height and weight by
the survey. Hip-to-waist was calculated based on the hip and waist of the participants by the
survey. Diagnosis of hypertension, diagnosis of diabetes, and smoking status were all binary
variables with yes and no. These covariates were selected due to their relationship with physical
activity engagement.
Data analyses
Descriptive analysis was conducted for the independent variables, dependent variables, and
the covariates. To determine the physical activity engagement between urban villagers and
non-urban villagers, unadjusted and multiple multivariate logistic regression were conducted
for the dependent variable of engagement in recreational physical activity, frequency of recrea-
tional physical activity per week, hours spend in sedentary behaviors per day, and reasons for
not engaging in physical activity. All analyses were conducted using STATA version 16 (Stata-
Corp LLC., College Station, TX, USA). The alpha levels were set at .05. The study protocol was
approved by the Oregon State University (IRB: IRB-2020-0509).
Results
Across the sample, 29.05% (n = 350) of the participants were urban villagers and 70.95%
(n = 855) were non-urban villagers. Pearson’s chi square test found significant different
between education levels, marital status, and household registration status between the urban
villagers and non-urban villagers. There were more non-urban villagers with completed mid-
dle school, high school, and professional school, college, and university (χ2 = 99.46, p < 0.001).
There were more non-urban villagers who were either married or partnered than urban villag-
ers (χ2 = 3.77, p = 0.05). Regrading to household registration or hukou, there were higher pro-
portion of non-urban villagers with Shenzhen hukou and higher proportion of urban villagers
with non-Shenzhen hukou (χ2 = 180.60, p < 0.001). Also, there were significant different in
age found between the two groups with non-urban villagers had a higher average age. Non-sig-
nificant differences were found between urban villagers and non-urban villagers among other
covariates (e.g., gender, employment, diagnosis of hypertension, diagnosis of diabetes, smok-
ing status, BMI, and hip-to-waist ratio).
Engagement in recreational physical activity
From the total sample size (n = 1205), 63.73% (n = 768) of participants reported not engage in
any recreational physical activity while 36.27% (n = 474) reported engaged in recreational
physical activity. A significant difference in proportion of engaging in recreational physical
activity were found between urban and non-urban villagers (χ2 = 60.79, p < 0.001) with higher
proportion of urban villagers (53.14%) reported engaging in recreational physical activity than
non-urban villagers (29.36%) as shown in Table 1. The unadjusted logistic regression found
that urban villagers were 2.73 (95% CI [2.11, 3.53], p < 0.001) times the odds of non-urban
PLOS ONE | https://doi.org/10.1371/journal.pone.0258085 October 28, 2021
6 / 17
PLOS ONE Table 1. Characteristics of urban villagers and non-urban villagers engaging in recreation physical activity.
Urban Villagers
Non-Urban Villagers
Total
n Mean/Proportion
n Mean/Proportion
n
Mean/Proportion
χ2/ t
P
Urban villagers’ physical activity levels
Engagement in recreational physical activity, %
Yes
No
Frequency of recreational physical activity per week, %
> 6 times
3–5 times
1–2 times
< 1 time
Average hours spend in sedentary behaviors per day, %
> 12 hours
9–12 hours
5–8 hours
2–4 hours
< 2 hours
Gender, %
Female
Male
Age, years
Employment status, %
Yes
No
Education levels, %
College & university
High school
Middle school
Primary school & none
Marital Status
Married/partnered
Single
Household registration (hukou), %
Shenzhen hukou
Non-Shenzhen hukou
Body Mass Index, kg/m2
Hip-to-waist ratio, %
Hypertension, %
Yes
No
Diabetes, %
Yes
No
Smoking status, %
Yes
No
186
164
35
57
62
10
15
46
99
86
104
183
167
350
302
48
46
108
122
74
253
97
29
321
350
350
19
331
8
342
94
256
53.14
46.86
21.34
34.76
37.80
6.10
4.29
13.14
28.29
24.57
29.71
52.29
47.71
37.75
86.29
13.71
13.14
30.86
34.86
21.14
72.29
27.71
8.29
91.71
22.99
89
5.43
94.57
2.29
97.71
26.86
73.14
251
604
154
201
216
34
43
113
221
261
217
474
381
855
701
154
342
236
203
74
663
192
424
431
855
855
60
795
15
840
190
665
29.36
70.64
25.45
33.22
35.70
5.62
5.03
13.22
25.85
30.53
25.38
55.44
44.56
39.24
81.99
18.01
40.00
27.60
23.74
8.65
77.54
22.46
49.59
50.41
22.83
89
7.02
92.98
1.75
98.25
22.22
77.78
437
768
189
358
278
44
58
159
320
347
321
657
548
1205
1003
202
148
325
344
388
916
286
453
752
1205
1205
79
1126
23
1182
284
921
36.27
63.73
21.75
41.20
31.99
5.06
4.81
13.20
26.56
28.80
26.64
54.52
45.48
38.8
83.24
16.76
12.28
26.97
28.55
32.20
76.02
23.98
37.59
62.41
22.87
89
6.56
93.44
1.91
98.09
81.66
18.34
60.79
<0.001�
1.19
0.76
5.65
0.23
1.00
0.32
2.21
0.03�
3.29
0.07
99.46
<0.001�
3.77
0.05
180.60
0 < .001�
-0.70
0.41
0.48
0.68
1.02
0.31
0.37
0.54
2.96
0.09
Note. n, sample size; χ2, chi-square statistic comparing between Urban Villagers and non-Urban Villagers for categorical variables; t, t-statistic comparing between
Urban Villagers and non-Urban Villagers for continuous variables, p, p-value associated with the statistic comparison test;
�, p < 0.05.
https://doi.org/10.1371/journal.pone.0258085.t001
PLOS ONE | https://doi.org/10.1371/journal.pone.0258085 October 28, 2021
7 / 17
PLOS ONE Urban villagers’ physical activity levels
villagers in engaging in recreational physical activity as shown in Table 2. The results of the
multivariate logistic regression found that Urban Villagers were 1.90 (95% CI [1.40, 2.57],
p < 0.001) times the odds of non-urban villagers in engaging in recreational physical activity
after controlling for covariates. The analysis also found the education levels, household regis-
tration, and BMI are significant factors contributing to the results of the odds ratios between
urban villagers and non-urban villagers in engaging in recreational physical activity.
Frequency of recreational physical activity per week
21.34% of urban villagers reported engaging in recreational physical activity more than six times
per week, in compared to 25.45% of non-urban villagers reported the same frequency. 34.76% of
urban villagers and 33.72% of non- urban villagers reported engaging in recreational physical
activity 3 to 5 time per week, 37.80% of urban villagers and 35.70% of non- urban villagers
reported engaging recreational physical activity 1 to 2 times per week. And 6.10% of urban vil-
lagers and 5.62% of non-urban villagers reported engaged in lesser than recreational physical
activity per week. No significant different was found between the two groups regarding the fre-
quency of engaging recreational in physical activity per week (χ2 = 1.19, p = 0.76). The odds
ratio of the unadjusted logistic regression for each level of the frequency of engaging in recrea-
tional physical activity per week with references of less than 1 time per week were 0.98 (95% CI
[0.46, 2.09], p = 0.95) for 1 to 2 time per week, 0.96 (95% CI [0.45, 2.07], p = 0.93) for 3 to 5
times per week, and 0.77 (95% CI [0.35, 1.71], p = 0.95) for more than six times per week for
urban villagers in engaging in recreational physical activity compared to non-urban villagers.
The results of the multivariate logistic regress found that urban villagers status is not a significant
factor in estimating the odds ratio of frequency in engaging recreational physical activity per
week with the reference groups of lesser than 1 time per week as shown in Tables 3 and 4.
Average hours spend in sedentary behaviors per day
4.29% of urban villagers and 5.03% non-urban villagers reported spending more than 12 hours
per day in sedentary, which made up the smallest proportion of the participants in their
respective group. 13.14% of urban villagers and 13.22% of non-urban villagers reported
Table 2. Odd ratios of urban villagers and non-urban villagers in engaging in recreational physical activity.
Urban villagers
Non-urban villagers
Engagement in recreational physical activity
Unadjusted Modelb
Adjusted Modelc
OR
2.73�
1 (ref.)
95% CI
2.11, 3.53
OR
1.90�
1 (ref.)
95% CI
1.40, 2.57
Abbreviations: OR, odds ratio; CI, confidence interval.
aBoldfaced numerals indicate p-value <0.05.
bOdd ratio from logistic regression model were computed for the outcome variable of engagement in recreational
physical activity (yes/no) with the exposure variable of living situation (urban village/non-urban village).
cOdd ratio from multivariable logistic regression model were computed for the outcome variable of engagement in
recreational physical activity (yes/no) with the exposure variable of living situation (urban village/non-urban village)
adjusted for gender (male/female), age (continuous), employment status (yes/no), education levels (college &
university, high school, middle school, primary school & none), marital status (married & partnered/single),
household registration (hukou) (Shenzhen/non-Shenzhen), BMI (continuous), hip-to-waist ratio (continuous),
hypertension (yes/no), diabetes (yes/no), and smoking status (yes/no).
d Detail adjusted model outcome were showed in S1 Table.
https://doi.org/10.1371/journal.pone.0258085.t002
PLOS ONE | https://doi.org/10.1371/journal.pone.0258085 October 28, 2021
8 / 17
PLOS ONE Urban villagers’ physical activity levels
Table 3. Odd ratios of frequency of engaging in recreational physical activity per week between urban villagers and non-urban villagers.
1–2 times vs. < 1 time
(ref.)
OR
0.98
95% CI
0.46, 2.09
Urban villagers
Non-urban villagers
1 (ref.)
Unadjusted odd ratiosb
3–5 times vs. < 1 time
(ref.)
> 6 times vs. < 1 time
(ref.)
1–2 times vs. < 1
time (ref.)
Adjusted odd ratiosc
3–5 times vs. < 1
time (ref.)
> 6 times vs. < 1
time (ref.)
OR
0.96
1 (ref.)
95% CI
0.45, 2.07
95% CI
0.35, 1.71
OR
0.77
1 (ref.)
95% CI
.44, 2.64
AOR
1.07
1 (ref.)
AOR
0.98
1 (ref.)
95% CI
.39, 2.43
AOR
0.83
1 (ref.)
95% CI
.32, 2.15
Abbreviations: OR, odds ratio; CI, confidence interval.
aBoldfaced numerals indicate p-value <0.05.
bOdd ratio from logistic regression model were computed for the outcome variable of engagement in recreational physical activity (yes/no) with the exposure variable of
living situation (urban village/non-urban village).
cOdd ratio from multivariable logistic regression model were computed for the outcome variable of engagement in recreational physical activity (yes/no) with the
exposure variable of living situation (urban village/non-urban village) adjusted for gender (male/female), age (continuous), employment status (yes/no), education levels
(college & university, high school, middle school, primary school & none), marital status (married & partnered/single), household registration (hukou) (Shenzhen/non-
Shenzhen), BMI (continuous), hip-to-waist ratio (continuous), hypertension (yes/no), diabetes (yes/no), and smoking status (yes/no).
d Detail adjusted model outcome were showed in S2 Table.
https://doi.org/10.1371/journal.pone.0258085.t003
spending 8 to 12 hours per day in sedentary behaviors. 28.29% of urban villagers and 25.85%
of non-urban villagers reported spending 4 to 8 hours per day on sedentary behaviors, while
24.57% and 30.53% of urban villagers and non-urban villagers spend 2 to 4 hours per day on
sedentary behaviors. For lowest amount of time spend in sedentary behaviors, 29.71% of
urban villagers and 25.36% of non-urban villagers reported spending lesser than 2 hours on it.
Non-significant different was found between the two groups regarded to the self-reported
hours spend in sedentary hours (χ2 = 5.65, p = 0.23). From the unadjusted logistic regression
with the reference group of spending less than 2 hours per day in sedentary behaviors and
urban villagers, the odd ratios were 0.69 (95% CI [0.49, 0.96], p = 0.03) for 2 to 4 hours, 0.93
(95% CI [0.67, 1.30], p = .69) for 4 to 8 hours, and 0.85 (95% CI [0.56, 1.29], p = 0.44) for 8 to
12 hours. The results of the multivariate logistic regression found that urban villagers status is
not a significant factor in estimating the hours spend in sedentary behaviors per day with the
reference groups of lesser than 2 hours per day as shown in Table 4. However, across all levels
Table 4. Odd ratios of average hours spend in sedentary behaviors per day between urban villagers and non-urban villagers.
2–4 hours vs. <
2 hours
5–8 hours vs. <
2 hours
9–12 hours vs. <
2 hours
>12 hours vs. <
2 hours
2–4 hours vs. <
2 hours
5–8 hours vs. <
2 hours
9–12 hours vs. <
2 hours
>12 hours vs.
< 2 hours
OR
0.69�
1 (ref.)
Urban villagers
Non-urban
villagers
95% CI OR
95% CI
.49, .96
0.93
.67, 1.30
OR
0.85
95% CI
.56, 1.29
OR
0.73
95% CI
.39, 1.37
OR
0.85
95% CI
.58, 1.25
OR
1.18
95% CI
.79, 1.75
OR
1.43
95% CI
.86, 2.38
OR
0.06
95% CI
0, 2.07
1 (ref.)
1 (ref.)
1 (ref.)
1 (ref.)
1 (ref.)
1 (ref.)
1 (ref.)
Abbreviations: OR, odds ratio; CI, confidence interval.
aBoldfaced numerals indicate p-value <0.05.
bOdd ratio from logistic regression model were computed for the outcome variable of engagement in recreational physical activity (yes/no) with the exposure variable of
living situation (urban village/non-urban village).
cOdd ratio from multivariable logistic regression model were computed for the outcome variable of engagement in recreational physical activity (yes/no) with the
exposure variable of living situation (urban village/non-urban village) adjusted for gender (male/female), age (continuous), employment status (yes/no), education levels
(college & university, high school, middle school, primary school & none), marital status (married & partnered/single), household registration (hukou) (Shenzhen/non-
Shenzhen), BMI (continuous), hip-to-waist ratio (continuous), hypertension (yes/no), diabetes (yes/no), and smoking status (yes/no).
d Detail adjusted model outcome were showed in S3 Table.
https://doi.org/10.1371/journal.pone.0258085.t004
PLOS ONE | https://doi.org/10.1371/journal.pone.0258085 October 28, 2021
9 / 17
PLOS ONE Urban villagers’ physical activity levels
Fig 2. Reasons for not engaging in physical activity among urban villagers and non-urban villagers who engage in physical activity.
https://doi.org/10.1371/journal.pone.0258085.g002
of hours spend in sedentary behaviors, completing professional school, college, and university
had a higher odd of spending more time in sedentary behaviors.
Reasons for not engagement in recreational physical activity
Among participants who engage in recreational physical activity, many indicated that no time
to exercise as the main reason why they did not engage in physical activity (n = 273) as shown
in Fig 2. The second top reasons participants selected as the reasons for not engaging in recrea-
tion physical activity was no need to exercise due to labor intensive occupation (n = 91), follow
by unwilling to exercise and no place to exercise (n = 69). Some participants also respond that
they did not engage in recreation physical activity due to feeling healthy (n = 14) and no need
to exercise and unable to engage in recreational physical activity due to illness (n = 5).
When stratified by urban village status, lack of time is the most cited reason for not engag-
ing in physical activity for both urban villagers (n = 107) and non-urban villagers (n = 166).
There were more urban villagers (n = 58) compared to non-urban villagers expressed that they
do not need to engage in physical activity due to occupations being labor intensive. There were
more non-urban villagers (n = 51) expressed that they were unwilling to engage in physical
activity than urban villagers (n = 18). Also, higher number of non-urban villagers (n = 28)
reported not having appropriate places and/or environments for physical activity compared to
urban villagers (n = 19).
Discussion
The purpose of this secondary data analysis is to determine and compare the prevalence of
physical activity engagement among the special population of Chinese urban villagers and
non-urban villagers. Both the unadjusted and adjusted logistic regression identified that urban
PLOS ONE | https://doi.org/10.1371/journal.pone.0258085 October 28, 2021
10 / 17
PLOS ONE Urban villagers’ physical activity levels
villagers are more likely to engage in recreational physical activity than their counterpart of
non-urban villagers. No significant relationship was found between the frequency of engage-
ment in recreational physical activity and urban village status. The multinomial logistic regres-
sion also found no significant relationship between hours spend in sedentary behaviors and
urban village status. Descriptive analysis shown that both urban villagers and non-urban vil-
lagers shared reasons for not engaging in recreational physical activity, such as lack of time to
exercise. However, more urban villagers indicated that their labor-intensive occupations are
sufficient enough for physical activity. While more non- urban villagers indicated that they are
more unwilling to exercise and there are no appropriate places and/or environments for recre-
ational physical activity.
While both urban villagers and non-urban villagers live in urban and well-developed area,
the levels of engagement in recreational physical activity were different between the two
groups. The results demonstrated that even within the same city, engagement in recreational
physical activity could be different by social characteristics. Urban villagers, like non-urban vil-
lagers, have access to different public physical activity facilities within the urban area. Physical
activity facilities such as parks, sidewalks, and outside of the urban villages are facilities that
urban villagers have access to. This is supported by the results that less urban villagers indi-
cated that there are a lack of appropriate places and/or environments for recreational physical
activity in compared to non-urban villagers. The ability of utilizing free public physical activity
facilities increase the opportunities for urban villagers to engage in recreational physical activ-
ity. Having these opportunities allow for urban villagers to obtain a healthier lifestyle of regu-
larly engagement in recreational of physical activity. While it has been found that lower-
income neighborhoods, such as urban villages, have less commercial physical activity-related
facilities [27]. The results of this study was different from the study conducted by Ortiz-Her-
na´ndez and Ramos-Iba´ñez [28], where they found that Mexican adults living in urban locali-
ties and cities with low socio-economic status had a lower probability of engaging in physical
activity. However, it is difficult to compare results across different countries as culture and
environments are widely different between the countries. Therefore, it is not appropriate to
compare the results between the studies. Studies conducted in the US [29] and in the Europe
[30] found similar results of adults living in rural areas less likely to engage in physical activity
and other psychosocial factors could influence physical activity behaviors. These highlight that
there is a need of global effort to promote physical activity in various countries. Further, due to
the unique situation of urban village in China, where the housing is surrounded by well-devel-
oped buildings and infrastructures, urban villagers have easy access to these different
infrastructures.
Income status could potentially be one of the factors explaining the different proportion of
urban villagers and non-urban villagers in engagement of recreational physical activity. Indi-
viduals living in urban village are more likely to be individuals with lower economic status.
Many of these individuals chose to reside in urban village due to the cheap accommodation
[31, 32]. Further, many of these individuals might held lower wages and labor-intensive occu-
pations. As evidence by the results of reasons for not engaging recreational physical activity,
more urban villagers reported that their occupations are labor intensive enough that either
they are too tired to engage in additional physical activity or they felt that they do not need to
engage in additional physical activity. This aligned with previous study finding that more rural
adults in China engage in work-related physical activity than urban adults [9]. In comparison
to urban villagers, fewer participants in the non-urban village group reporting their occupa-
tions are too physically demanding that they felt that engagement in recreational physical
activity is not necessary. Non-urban villagers are more likely to held office-related occupations,
therefore, it limits their ability to engage in physical activity. Past studies had demonstrated
PLOS ONE | https://doi.org/10.1371/journal.pone.0258085 October 28, 2021
11 / 17
PLOS ONE Urban villagers’ physical activity levels
that officer workers are more likely to engage in less physical activity and more sedentary
behaviors [33]. Further, non-urban villagers might be more likely to have better technology
access than urban villagers. Technology such as television and media are found to be associated
with lower physical activity levels and high sedentary behavior [34, 35]. This might relate to
the higher number of non-Urban Villagers reporting unwilling to engage in recreational physi-
cal activity. It is also surprising to find that there are higher numbers of non-urban villagers
indicating that the reason for not engaging in recreational physical activity was lack of appro-
priate places and/or environments. Being consistent with previous research by Munter et al.
[9] where Chinese urban adults are less likely to engage in physical activity than Chinese adults
with lower economic status living in rural area.
Based on the results of this study, more tailed intervention is needed for Chinese adults not
living in urban villages. Even though urban villagers are more likely to be in poor health due to
poor housing situation [36, 37], they are more likely to engage in recreational physical activity
than non-urban villagers. While the two groups have large number of participants reporting
lack of time to engage in recreational physical activity, different interventions should be devel-
oped for the two groups. Due to differences in living situations, economic status, and occupa-
tions, different reactions and responses to interventions might be different between urban
villagers and non-urban villagers. When designing physical activity interventions, there is a
need to consider demographic characteristics and socioeconomic factors. For urban villagers,
tailed interventions are needed to target group of individuals that believe that physical activity
performed during their job are sufficient enough for health. Multiple studies had demon-
strated that leisure time physical activity and recreational physical activity are associated with
better health quality of life [38–40]. Occupational-related physical activity is not considered to
be recreation or leisure physical activity. Therefore, specific interventions are needed targeting
urban villagers. Developing interventions in targeting these reasons and solving these barriers
for non-urban villagers will be important step for increase the proportion of non-urban villag-
ers in engaging in recreational physical activity. For example, Gu et al. [41] found change in
physical activity among office workers after the implementation of a worksite intervention
programs at 17 worksites in the urban city of Shanghai with pedometers for 100 days. The goal
of using and developing physical activity interventions are to promote recreational physical
activity levels among both urban villagers and non-urban villagers.
Further research and studies are warrants in determine the physical activity levels
among urban villagers and non-urban villagers. Study had done in the past to examine the
physical activity levels of Chines adults [9, 13, 42], but there is a lack of empirical evidence
on the physical activity levels of urban villagers. Using additional techniques, such as
accelerometers, to collected more detailed data could increase our understanding of physi-
cal activity levels of urban villagers. More detailed data such as minutes spend in each
intensity of physical activity or number of steps taken each day can better represent the
physical activity levels of urban villagers. It has been proposed that an intersectionality
approach should be taken when measuring and discussing physical activity levels [43–45].
The interacting factors could provide more detail information on the physical activity of
special population such as urban villagers. Often, urban villagers might be considered as
individuals living in urban area. However, due to the unique situation of urban village,
they are considered a special population living in the urban area. This study demonstrated
that there is a need to examine the physical activity levels of special populations living in
China. As shown in this study, the proportion of urban villagers and non-urban villagers
engaging in recreational physical activity is different, so more research is needed. This
data could further facilitate the development of physical activity intervention targeting
urban villagers and non-urban villagers. Future researches should also focus on urban
PLOS ONE | https://doi.org/10.1371/journal.pone.0258085 October 28, 2021
12 / 17
PLOS ONE Urban villagers’ physical activity levels
villagers and non-urban villagers in meeting physical activity guidelines by the World
Health Organization [21]. The current physical activity guidelines for adults over the ages
of 18 years old is at least 150 minutes of moderate-to-vigorous physical activity or 75 min-
utes of vigorous physical activity per week. Examining the prevalence of urban villagers
and non-urban villagers in meeting these physical activity guidelines could increase our
understanding of the physical activity behaviors and dose-response relationship between
physical activity and health among these populations. It is important to note that there is a
lack of national and regional physical activity guidelines in China [46]. Developing these
physical activity guidelines could be beneficial for Chinese citizen as there is a guideline
for them to follow.
One interesting find of the analysis was that education might have an influence on physical
activity-related outcomes among urban villagers and non-urban villagers. Based on the
adjusted logistic regression model, in compare to no formal education and only completing
primary education, other education levels (i.e., middle school, high school, professional college
and university) are less likely to engage in physical activity. The analysis also found that higher
education is associated with longer time spent in sedentary behaviors. The results align with
previous study examining the decline of physical activity levels among Chinese adults [14].
The study found that the greater availability of higher educational institutions is strongly asso-
ciated with the declines of physical activities based on data from the 1991–2006 China Health
and Nutrition Surveys [47]. Individuals with higher education are more likely to have office-
related positions. Officer workers are more likely to spend more time in sedentary behaviors
[48]. In addition, it was found that Chinese adults who completed high school education are
less likely to engage in occupational-related physical activity [9]. These results suggested that
physical activity interventions are needed for individuals with higher education. To ensure
that physical activity become a lifelong habit among Chinese adults, there is need to develop
physical activity intervention targeting adults at various educational levels. For example,
requiring physical education or physical activity classes for students in middle schools, high
schools, and colleges and universities. Requirement of physical education in early childhood is
positively associated with physical activity levels in adulthood [49]. Individuals who had taken
a physical activity course while in colleges and universities report higher physical activity levels
in adulthood compared to those that did not take a physical activity course [50]. Continuation
promotion of physical activity through various different educational institutions could poten-
tially increase physical activity levels of adults.
Limitation
To the authors’ knowledge, this is the first of the few studies that examined the physical activity
levels of urban villagers in China. The strength of this study is including the special population
of urban villagers. However, this study is not without its limitation. The data used in the analy-
sis are based on self-reported data. There could be potential recall and social bias. These biases
could lead to misclassification of data and results [51]. In addition to biases, there could be low
generalizability of the results. Due to the data only included participants living in the Luohu,
Shenzhen, China, the results might be only generalized to this particular populations living in
Shenzhen. However, it is assumed that urban villagers across China shared the similar charac-
teristics of lower economic status, migrant workers, labor intensive worker, poor living situa-
tion, and lack of infrastructures. It is important to note that the survey did not utilized the
International Physical Activity Questionary (IPAQ) in the surveillance system. This could lead
to misunderstanding of questions by the participants. To limit misunderstanding, all data col-
lected were in Chinese via face-to-face interview by trained personals.
PLOS ONE | https://doi.org/10.1371/journal.pone.0258085 October 28, 2021
13 / 17
PLOS ONE Urban villagers’ physical activity levels
Conclusion
Overall, the proportion of urban villagers and non-urban villagers in engaging in recreational
physical activity are different with urban villagers more likely to engage in recreational physical
activity. While participants from both groups expressed that lack of time as a barrier in engag-
ing in recreational physical activity, non-urban villagers are more likely to reported that they
are unwilling to participate in recreational physical activity and lack appropriate place and/
environment for recreational physical activity. Urban villagers are more likely to reported that
they do not engage in recreational physical activity due to work-related physical activity. Physi-
cal activity interventions are needed to target these various barriers in preventing urban villag-
ers and non-urban villagers in participating from recreational physical activity. Further
research is warranted in order to better understanding the physical activity levels of the special
population of urban villagers living in China.
Supporting information
S1 Table. Odd ratios of urban villagers and non-urban villagers in engaging in recreational
physical activity: Adjusted model outcomes.
(DOCX)
S2 Table. Odd ratios of frequency of engaging in recreational physical activity per week
between urban villagers and non-urban villagers: Adjusted model outcomes.
(DOCX)
S3 Table. Odd ratios of average hours spend in sedentary behaviors per day between urban
villagers and non-urban villagers.
(DOCX)
S1 File. Questionnaire Chinese.
(DOCX)
S2 File. Questionnaire English.
(DOCX)
Author Contributions
Conceptualization: Lu Shi, Willie Leung, Qingming Zheng, Jie Wu.
Data curation: Lu Shi, Qingming Zheng, Jie Wu.
Formal analysis: Lu Shi, Willie Leung.
Investigation: Jie Wu.
Methodology: Lu Shi, Willie Leung, Qingming Zheng, Jie Wu.
Project administration: Qingming Zheng, Jie Wu.
Resources: Qingming Zheng.
Software: Lu Shi.
Supervision: Qingming Zheng.
Visualization: Lu Shi.
Writing – original draft: Lu Shi, Willie Leung.
Writing – review & editing: Lu Shi, Willie Leung.
PLOS ONE | https://doi.org/10.1371/journal.pone.0258085 October 28, 2021
14 / 17
PLOS ONE Urban villagers’ physical activity levels
References
1. Haskell WL, Blair SN, Hill JO. Physical activity: Health outcomes and importance for public health pol-
icy. Preventive Medicine. 2009 Oct 1; 49(4):280–2. https://doi.org/10.1016/j.ypmed.2009.05.002 PMID:
19463850
2. Steinbeck KS. The importance of physical activity in the prevention of overweight and obesity in child-
hood: a review and an opinion. Obesity Reviews. 2001; 2(2):117–30. https://doi.org/10.1046/j.1467-
789x.2001.00033.x PMID: 12119663
3. Warburton DER, Nicol CW, Bredin SSD. Health benefits of physical activity: the evidence. Canadian Medi-
cal Association Journal. 2006 Mar 14; 174(6):801–9. https://doi.org/10.1503/cmaj.051351 PMID: 16534088
4. Eime RM, Young JA, Harvey JT, Charity MJ, Payne WR. A systematic review of the psychological and
social benefits of participation in sport for adults: informing development of a conceptual model of health
through sport. Int J Behav Nutr Phys Act. 2013 Dec 7; 10:135. https://doi.org/10.1186/1479-5868-10-
135 PMID: 24313992
5. Singh MAF. Exercise to prevent and treat functional disability. Clin Geriatr Med. 2002 Aug; 18(3):431–
62, vi–vii. https://doi.org/10.1016/s0749-0690(02)00016-2 PMID: 12424867
6.
Tak E, Kuiper R, Chorus A, Hopman-Rock M. Prevention of onset and progression of basic ADL disabil-
ity by physical activity in community dwelling older adults: A meta-analysis. Ageing Research Reviews.
2013 Jan 1; 12(1):329–38. https://doi.org/10.1016/j.arr.2012.10.001 PMID: 23063488
7. World Health Organization. WHO | Physical Inactivity: A Global Public Health Problem [Internet]. WHO.
World Health Organization; 2020 [cited 2020 Sep 22]. Available from: https://www.who.int/
dietphysicalactivity/factsheet_inactivity/en/.
8.
Li F. Physical activity and health in the presence of China’s economic growth: Meeting the public health
challenges of the aging population. Journal of Sport and Health Science. 2016 Sep 1; 5(3):258–69.
https://doi.org/10.1016/j.jshs.2016.06.004 PMID: 30356539
9. Muntner P, Gu D, Wildman RP, Chen J, Qan W, Whelton PK, et al. Prevalence of Physical Activity
Among Chinese Adults: Results From the International Collaborative Study of Cardiovascular Disease
in Asia. Am J Public Health. 2005 Sep 1; 95(9):1631–6. https://doi.org/10.2105/AJPH.2004.044743
PMID: 16051938
10.
Zhang L, Wang Z, Wang X, Chen Z, Shao L, Tian Y, et al. Prevalence of overweight and obesity in
China: Results from a cross-sectional study of 441 thousand adults, 2012–2015. Obesity Research &
Clinical Practice. 2020 Mar 1; 14(2):119–26.
11. Natinal Center for Cardiovascular Disease. China. Report on Cardiovascular Diseases in China 2017.
Beijing: Encyclopedia of China Publishing House; 2018.
12.
Liu L. Rural-Urban Disparities in Cardiovascular Disease Mortality Among Middle-Age Men in China.
Asia Pac J Public Health. 2020 Sep 11;1010539520956446. https://doi.org/10.1177/
1010539520956446 PMID: 32917101
13. Peters TM, Moore SC, Xiang YB, Yang G, Shu XO, Ekelund U, et al. Accelerometer-Measured Physical
Activity in Chinese Adults. American Journal of Preventive Medicine. 2010 Jun 1; 38(6):583–91. https://
doi.org/10.1016/j.amepre.2010.02.012 PMID: 20494234
14. Chen M, Wu Y, Narimatsu H, Li X, Wang C, Luo J, et al. Socioeconomic Status and Physical Activity in
Chinese Adults: A Report from a Community-Based Survey in Jiaxing, China. PLOS ONE. 2015 Jul 15;
10(7):e0132918. https://doi.org/10.1371/journal.pone.0132918 PMID: 26177205
15. Shi Z, Lien N, Kumar BN, Holmboe-Ottesen G. Physical activity and associated socio-demographic fac-
tors among school adolescents in Jiangsu Province, China. Preventive Medicine. 2006 Sep 1; 43
(3):218–21. https://doi.org/10.1016/j.ypmed.2006.04.017 PMID: 16762405
16.
17.
18.
Liu Y, He S. Chinese Urban Villages as Marginalized Neighbourhoods under Rapid Urbanization. In:
Wu F, Webster C, editors. Marginalization in Urban China: Comparative Perspectives [Internet]. Lon-
don: Palgrave Macmillan UK; 2010 [cited 2020 Oct 2]. p. 177–200. (International Political Economy
Series). Available from: https://doi.org/10.1057/9780230299122_10
Liu Y, He S, Wu F, Webster C. Urban villages under China’s rapid urbanization: Unregulated assets
and transitional neighbourhoods. Habitat International. 2010 Apr 1; 34(2):135–44.
Zhang L, Zhao SXB, Tian JP. Self-help in housing and chengzhongcun in China’s urbanization. Interna-
tional Journal of Urban and Regional Research. 2003; 27(4):912–37.
19. Shi L, Patil VP, Leung W, Zheng Q. Willingness to use and satisfaction of primary care services among
locals and migrants in Shenzhen, China. Health & Social Care in the Community [Internet]. [cited 2021 Jul
7];n/a(n/a). Available from: https://onlinelibrary.wiley.com/doi/abs/10.1111/hsc.13418 PMID: 33978287
20. Yang X. Research on the Optimization of Social Sports Participation Path in “Urban Villages.” In 2018
[cited 2020 Oct 23]. Available from: https://kns.cnki.net/KCMS/detail/detail.aspx?dbcode=IPFD&db
PLOS ONE | https://doi.org/10.1371/journal.pone.0258085 October 28, 2021
15 / 17
PLOS ONE Urban villagers’ physical activity levels
name=IPFDLAST2018&filename=LRCM201802002061&v=MTc1OTRiS0lGc1hLVC9JWTdHNEg5b
k1yWTlGWnVzSkRSTkt1aGRobmo5OFRuanFxeGRFZU1PVUtyaWZaZVp1RmluZ1Vy.
21. World Health Organization. Global recommendations on physical activity for health. Geneva: World
Health Organization; 2010.
22. Bork T, Kraas F, Xue D, Li Z. Urban environmental health challenges in China’s villages-in-the-city.
Geographische Zeitschrift. 2011; 99(1):16–35.
23.
Jiang J, Wang P. Health status in a transitional society: urban-rural disparities from a dynamic perspec-
tive in China. Population Health Metrics. 2018 Dec 27; 16(1):22. https://doi.org/10.1186/s12963-018-
0179-z PMID: 30591053
24. Duan J, Cao D. Investigation on the Status Quo of Mass Sports Participation in Urban Villages——Tak-
ing Wanbolin District of Taiyuan City as an Example [Internet]. North University of China; 2017 [cited
2020 Oct 23]. Available from: https://kns.cnki.net/KCMS/detail/detail.aspx?dbcode=CJFQ&dbname=
CJFDLAST2017&filename=TYGJ201703008&v=MDMxOTk4ZVgxTHV4WVM3RGgxVDNxVHJXT
TFGckNVUjdxZVorWnJGaUhuVnIvTE1UVE1aTEc0SDliTXJJOUZiSVI=.
25.
Long H, Li Y, Yu H, Yang W, Wu M. Research on Youth Sports Participation in Urban Villages of Kun-
ming City in the Process of Urbanization. Contemporary Sports Technology. 2019; 9(06):167–8.
26. Chen X. The Problems Research of Griddization Management in Longgang District of Shenzhen City
[Internet] [Master]. Central China Normal University; 2015 [cited 2020 Oct 24]. Available from: https://
gb.oversea.cnki.net/kcms/detail/detail.aspx?recid=&FileName=1016038151.nh&DbName=
CMFD201602&DbCode=CMFD.
27. Powell LM, Slater S, Chaloupka FJ, Harper D. Availability of Physical Activity–Related Facilities and
Neighborhood Demographic and Socioeconomic Characteristics: A National Study. Am J Public Health.
2006 Sep; 96(9):1676–80. https://doi.org/10.2105/AJPH.2005.065573 PMID: 16873753
28. Ortiz-Herna´ ndez L, Ramos-Iba´ñez N. Sociodemographic factors associated with physical activity in
Mexican adults. Public Health Nutrition. 2010 Jul; 13(7):1131–8. https://doi.org/10.1017/
S1368980010000261 PMID: 20196912
29. Wilcox S, Castro C, King AC, Housemann R, Brownson RC. Determinants of leisure time physical activ-
ity in rural compared with urban older and ethnically diverse women in the United States. Journal of Epi-
demiology & Community Health. 2000 Sep 1; 54(9):667–72. https://doi.org/10.1136/jech.54.9.667
PMID: 10942445
30. Moreno-Llamas A, Garcı´a-Mayor J, De la Cruz-Sa´nchez E. Urban-rural differences in trajectories of
physical activity in Europe from 2002 to 2017. Health & Place. 2021 May 1; 69:102570. https://doi.org/
10.1016/j.healthplace.2021.102570 PMID: 33873131
31. Keung Wong DF, Li CY, Song HX. Rural migrant workers in urban China: living a marginalised life:
Rural migrant workers in urban China. International Journal of Social Welfare. 2007 Jan; 16(1):32–40.
32.
Lu Z, Song S. Rural–urban migration and wage determination: The case of Tianjin, China. China Eco-
nomic Review. 2006 Jan; 17(3):337–45.
33. Parry S, Straker L. The contribution of office work to sedentary behaviour associated risk. BMC Public
Health. 2013 Apr 4; 13:296. https://doi.org/10.1186/1471-2458-13-296 PMID: 23557495
34. Harris JL, Bargh JA. The Relationship between Television Viewing and Unhealthy Eating: Implications
for Children and Media Interventions. Health Commun. 2009 Oct; 24(7):660–73. https://doi.org/10.
1080/10410230903242267 PMID: 20183373
35. Keadle SK, Arem H, Moore SC, Sampson JN, Matthews CE. Impact of changes in television viewing
time and physical activity on longevity: a prospective cohort study. International Journal of Behavioral
Nutrition and Physical Activity. 2015 Dec 18; 12(1):156. https://doi.org/10.1186/s12966-015-0315-0
PMID: 26678502
36. Gao Y, Shahab S, Ahmadpoor N. Morphology of Urban Villages in China: A Case Study of Dayuan Vil-
lage in Guangzhou. Urban Science. 2020 May 7; 4(2):23.
37. Wu F. Housing in Chinese Urban Villages: The Dwellers, Conditions and Tenancy Informality. Housing
Studies. 2016 Oct 2; 31(7):852–70.
38.
Tessier S, Vuillemin A, Bertrais S, Boini S, Le Bihan E, Oppert J-M, et al. Association between leisure-
time physical activity and health-related quality of life changes over time. Preventive Medicine. 2007
Mar 1; 44(3):202–8. https://doi.org/10.1016/j.ypmed.2006.11.012 PMID: 17208289
39. Vuillemin A, Boini S, Bertrais S, Tessier S, Oppert J-M, Hercberg S, et al. Leisure time physical activity
and health-related quality of life. Preventive Medicine. 2005 Aug 1; 41(2):562–9. https://doi.org/10.
1016/j.ypmed.2005.01.006 PMID: 15917053
40. Wendel-Vos GCW, Schuit AJ, Tijhuis MAR, Kromhout D. Leisure time physical activity and health-
related quality of life: Cross-sectional and longitudinal associations. Qual Life Res. 2004 Apr 1; 13
(3):667–77. https://doi.org/10.1023/B:QURE.0000021313.51397.33 PMID: 15130029
PLOS ONE | https://doi.org/10.1371/journal.pone.0258085 October 28, 2021
16 / 17
PLOS ONE Urban villagers’ physical activity levels
41. Gu M, Wang Y, Shi Y, Yu J, Xu J, Jia Y, et al. Impact of a group-based intervention program on physical
activity and health-related outcomes in worksite settings. BMC Public Health. 2020 Jun 15; 20(1):935.
https://doi.org/10.1186/s12889-020-09036-2 PMID: 32539787
42.
Zhu W, Chi A, Sun Y. Physical activity among older Chinese adults living in urban and rural areas: A
review. Journal of Sport and Health Science. 2016 Sep 1; 5(3):281–6. https://doi.org/10.1016/j.jshs.
2016.07.004 PMID: 30356525
43. Abichahine H, Veenstra G. Inter-categorical intersectionality and leisure-based physical activity in Can-
ada. Health Promot Int. 2017 Aug 1; 32(4):691–701. https://doi.org/10.1093/heapro/daw009 PMID:
26976822
44. Herrick SSC, Duncan LR. A Qualitative Exploration of LGBTQ+ and Intersecting Identities Within Physi-
cal Activity Contexts. Journal of Sport and Exercise Psychology. 2018 Dec 1; 40(6):325–35. https://doi.
org/10.1123/jsep.2018-0090 PMID: 30537884
45. Ray R. An Intersectional Analysis to Explaining a Lack of Physical Activity Among Middle Class Black
Women. Sociology Compass. 2014; 8(6):780–91.
46. Xu J, Gao C. Physical activity guidelines for Chinese children and adolescents: The next essential step.
J Sport Health Sci. 2018 Jan; 7(1):120–2. https://doi.org/10.1016/j.jshs.2017.07.001 PMID: 30356441
47. Ng SW, Norton EC, Popkin BM. Why have physical activity levels declined among Chinese adults?
Findings from the 1991–2006 China health and nutrition surveys. Social Science & Medicine. 2009 Apr
1; 68(7):1305–14. https://doi.org/10.1016/j.socscimed.2009.01.035 PMID: 19232811
48. Clemes SA, O’Connell SE, Edwardson CL. Office Workers’ Objectively Measured Sedentary Behavior
and Physical Activity During and Outside Working Hours. Journal of Occupational and Environmental
Medicine. 2014 Mar; 56(3):298–303. https://doi.org/10.1097/JOM.0000000000000101 PMID:
24603203
49.
Trudeau F, Laurencelle L, Shephard RJ. Tracking of physical activity from childhood to adulthood. Med
Sci Sports Exerc. 2004 Nov; 36(11):1937–43. https://doi.org/10.1249/01.mss.0000145525.29140.3b
PMID: 15514510
50. Sparling PB, Snow TK. Physical activity patterns in recent college alumni. Res Q Exerc Sport. 2002
Jun; 73(2):200–5. https://doi.org/10.1080/02701367.2002.10609009 PMID: 12092895
51. Althubaiti A. Information bias in health research: definition, pitfalls, and adjustment methods. J Multidis-
cip Healthc. 2016 May 4; 9:211–7. https://doi.org/10.2147/JMDH.S104807 PMID: 27217764
PLOS ONE | https://doi.org/10.1371/journal.pone.0258085 October 28, 2021
17 / 17
PLOS ONE |
10.1371_journal.pone.0255335 | RESEARCH ARTICLE
Impaired immune response mediated by
prostaglandin E2 promotes severe COVID-19
disease
1‡*, Elisabeth Stelling1‡, Lisa Lasswitz2, Antonia P. Gunesch2,3,4,
2,5,6, Thomas Pietschmann2,3, Virginie Montiel7, Jean-Luc Balligand7,
Melanie Ricke-HochID
Martina Kasten1, Francisco J. Zapatero-Belincho´ n2,5, Graham Brogden2,
Gisa GeroldID
Federica Facciotti8, Emilio Hirsch9, Thomas GausepohlID
F. Rimmelzwaan10, Anne Ho¨ fer11,12, Mark P. Ku¨ hnel11,12, Danny Jonigk11,12,
Julian Eigendorf13, Uwe Tegtbur13, Lena Mink13, Michaela Scherr14, Thomas Illig15,
Axel Schambach16,17, Tobias J. Pfeffer1, Andres Hilfiker18, Axel Haverich18,
Denise Hilfiker-Kleiner1,19
1, Husni ElbaheshID
10, Guus
1 Department of Cardiology and Angiology, Hannover Medical School, Hanover, Germany, 2 Institute of
Experimental Virology, TWINCORE, Center for Experimental and Clinical Infection Research Hannover,
Hanover, Germany, 3 German Center for Infection Research, Hanover-Braunschweig Site, Braunschweig,
Germany, 4 Department of Gastroenterology, Hepatology and Endocrinology, Hannover Medical School,
Hanover, Germany, 5 Department of Clinical Microbiology, Virology & Wallenberg Centre for Molecular
Medicine (WCMM), Umeå University, Umeå, Sweden, 6 Department of Biochemistry, University of
Veterinary Medicine Hannover, Hanover Germany, 7 Pole of Pharmacology and Therapeutics, Institut de
Recherche Expe´ rimentale et Clinique, and Cliniques Universitaires Saint-Luc, Universite´ catholique de
Louvain (UCLouvain), Brussels, Belgium, 8 Department of Experimental Oncology, European Institute of
Oncology IRCCS, Milan, Italy, 9 Department of Molecular Biotechnology and Health Sciences, Molecular
Biotechnology Center, University of Torino, Torino, Italy, 10 Research Center for Emerging Infections and
Zoonoses (RIZ), University of Veterinary Medicine in Hannover (TiHo), Hannover, Germany, 11 Biomedical
Research in Endstage and Obstructive Lung Disease (BREATH), German Center for Lung Research,
Hanover, Germany, 12 Institute for Pathology, Hannover Medical School, Hanover, Germany, 13 Institute of
Sports Medicine, Hannover Medical School, Hanover, Germany, 14 Department of Hematology, Hemostasis,
Oncology and Stem Cell Transplantation, Hannover Medical School, Hanover, Germany, 15 Hannover
Unified Biobank (HUB), Hannover Medical School, Hanover, Germany, 16 Institute of Experimental
Hematology, Hannover Medical School, Hanover, Germany, 17 Division of Hematology and Oncology,
Boston Children’s Hospital, Harvard Medical School, Boston, MA, United States of America, 18 Department
of Cardiac, Thoracic, Transplantation and Vascular Surgery, Hannover Medical School, Hanover, Germany,
19 Department of Cardiovascular Complications of Oncologic Therapies, Medical Faculty of the Philipps
University Marburg, Marburg, Germany
‡ SIGR and MK contributed equally to this work as first co-authors.
* hoch.melanie@mh-hannover.de
Abstract
The SARS-CoV-2 coronavirus has led to a pandemic with millions of people affected. The
present study finds that risk-factors for severe COVID-19 disease courses, i.e. male sex,
older age and sedentary life style are associated with higher prostaglandin E2 (PGE2)
serum levels in blood samples from unaffected subjects. In COVID-19 patients, PGE2 blood
levels are markedly elevated and correlate positively with disease severity. SARS-CoV-2
induces PGE2 generation and secretion in infected lung epithelial cells by upregulating
cyclo-oxygenase (COX)-2 and reducing the PG-degrading enzyme 15-hydroxyprostaglan-
din-dehydrogenase. Also living human precision cut lung slices (PCLS) infected with SARS-
CoV-2 display upregulated COX-2. Regular exercise in aged individuals lowers PGE2
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: Ricke-Hoch M, Stelling E, Lasswitz L,
Gunesch AP, Kasten M, Zapatero-Belincho´n FJ, et
al. (2021) Impaired immune response mediated by
prostaglandin E2 promotes severe COVID-19
disease. PLoS ONE 16(8): e0255335. https://doi.
org/10.1371/journal.pone.0255335
Editor: Paulo Lee Ho, Instituto Butantan, BRAZIL
Received: March 11, 2021
Accepted: July 14, 2021
Published: August 4, 2021
Copyright: © 2021 Ricke-Hoch et al. This is an
open access article distributed under the terms of
the Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: All relevant data are
within the manuscript and its Supporting
Information files.
Funding: This work was supported by: The
German Research Foundation (DFG, HI 842/3-2 to
D.H.-K.), by the DFG Clinical Research Group (DFG
KFO311, HI 842/10-1, HI 842/10-2 to D.H.-K.; RI
2531/2-1, RI 2531/2-2 to M.R.-H.), by REBIRTH I/
II to D.H.-K., by the Foundation Leducq (Project ID
19CVD02) to D.H.-K. and E.H., DFG as part of the
German Strategy for Excellence (EXC 2155
“RESIST”, Project ID 39087428 to D.J.), The
PLOS ONE | https://doi.org/10.1371/journal.pone.0255335 August 4, 2021
1 / 24
PLOS ONE DEFEAT PANDEMIcs (AP6-9, to D.J. and M.P.K.),
by the (DFG – Projektnummer 158989968 - SFB
900 project C7 and DFG project GE 2145/3-2 to G.
G.), the ‘Niedersa¨chsischen Vorab’ program
(project 76251-99-3/19 to G.G.) through the
Ministry of Lower Saxony (MWK) and the
Volkswagen Foundation (Volkswagen Stiftung), by
the Federal Ministry of education and research
(project COVID-Protect, Project: 01KI20143C to G.
G.), the Knut and Alice Wallenberg Foundation and
the Federal Ministry of Education and Research
together with the the Ministry for Science and
Culture (MWG) through the ‘Professorinnen
Programm III’ to G.G., by the European Research
Council Consolidator Grant (XHale; 771883 to D.
J.), by REBIRTH I/II and REBIRTH Center for
Regenerative Translational Medicine (MWK, project
ZN3440) to A.S., by Cariplo Foundation (Project
#2018-0498 to E.H.), by MWG project 14-76103-
184 CORONA-1/20 to T.I. and by the European
Virus Archive GLOBAL (EVA-GLOBAL) project
funded by the European Union’s Horizon 2020
research and innovation program under grant
agreement No 871029 (to Christian Drosten). This
work was partly supported by the Alexander von
Humboldt Foundation in the framework of the
Alexander von Humboldt Professorship endowed
by the German Federal Ministry of Education and
Research and by funding from the Ministry for
Science and Culture (MWK), Lower Saxony,
Germany (14 - 76103-184 CORONA-15/20 to G.F.
R.). A.P.G. was supported by the Deutsches
Zentrum fu¨r Infektionsforschung (DZIF; German
Center for Infection Research; Grant No. TTU
05.816 00 to T.P.). Work by J.L.B. was supported
by grants from Fonds National de la Recherche
Scientifique (FNRS) and WEBIO. J.L.B is an
established investigator of the WELBIO institute.
The funders had no role in study design, data
collection and analysis, decision to publish, or
preparation of the manuscript.
Competing interests: The authors have declared
that no competing interests exist.
Impaired immune response promotes severe COVID-19 disease
serum levels, which leads to increased Paired-Box-Protein-Pax-5 (PAX5) expression, a
master regulator of B-cell survival, proliferation and differentiation also towards long lived
memory B-cells, in human pre-B-cell lines. Moreover, PGE2 levels in serum of COVID-19
patients lowers the expression of PAX5 in human pre-B-cell lines. The PGE2 inhibitor Taxi-
folin reduces SARS-CoV-2-induced PGE2 production. In conclusion, SARS-CoV-2, male
sex, old age, and sedentary life style increase PGE2 levels, which may reduce the early
anti-viral defense as well as the development of immunity promoting severe disease
courses and multiple infections. Regular exercise and Taxifolin treatment may reduce these
risks and prevent severe disease courses.
Introduction
The 2019 strain of coronavirus (severe acute respiratory syndrome coronavirus-2 SARS-CoV-
2) caused a pandemic with COVID-19 disease affecting millions of people worldwide. Patients
with serious disease courses frequently present with severe acute respiratory syndrome that
can progress to pneumonia and acute respiratory distress syndrome and shock [1–3]. Systemic
inflammation, acute cardiac injury, heart failure, and hypercoagulability are critical complica-
tions in COVID-19 disease [1, 4–9]. Identified cell types infected with SARS-CoV-2 include
pulmonary epithelial cells, renal cells, cardiomyocytes, endothelial cells and pericytes [10–12].
An increased risk for infection and severe disease courses have been found in association
with older age, male sex, cardiovascular comorbidities and air pollution [7, 13–15]. Immuno-
thrombosis integrates innate immunity, activation of platelets, and clotting factors to fight
invading pathogens and concurrently promotes inflammation-related tissue damage; in the
context of COVID-19 disease, this may explain the systemic hypercoagulability frequently
present in COVID-19 patients [8]. Further alterations in the immune system with partially
opposing mechanisms have been reported in acute and chronic COVID-19 disease. On one
hand, COVID-19 infection appears associated with an upregulation and activation of neutro-
phils while at the same time lymphocytes are diminished [16]. Reduced lymphocyte popula-
tions seem to correlate with more severe organ injury and higher mortality in hospitalized
COVID-19 patients [16]. In this regard, T-cell exhaustion [3, 17], reduced circulating and resi-
dent B-cell population and loss of germinal centers associated with viral persistence and severe
disease courses correlate with high mortality in the acute phase [3, 18, 19]. On the other hand,
a growing body of clinical data suggests that a cytokine storm is associated with COVID-19
severity and is also a crucial cause of death from COVID-19 [20–22]. Among potential mecha-
nisms, SARS-CoV-2 induced formation of autoantibodies, tissue and organ injury as well as
secondary infection with bacteria and fungi [23, 24].
Prostaglandin (PG) E2, a metabolite of arachidonic acid, is a well-known modulator of viral
infection [25]. As such, PGE2 suppresses the adaptive and innate immune systems and pro-
motes infection, e.g., by influenza A virus (IAV) [26, 27]. Moreover, increased circulating
PGE2 levels have been associated with reduced immunity in response to IAV vaccination [26,
27]. Interestingly, IAV infection also promotes the production of PGE2 [28]. Cyclooxygenase-
2 (COX-2) is a rate-limiting enzyme for the generation of PGE2 and Hydroxyprostaglandin
Dehydrogenase 15-(NAD) (HPGD) is an enzyme responsible for the degradation of PGE2
[29]. These findings, supported further by a recent literature review [30] naturally suggested a
connection between arachidonic acid metabolism and PGE2 in COVID-19 disease.
We hypothesized that PGE2 modulates the immune response in individuals at risk for
severe COVID-19 disease. To test this, we first measured serum PGE2 levels in COVID-19
PLOS ONE | https://doi.org/10.1371/journal.pone.0255335 August 4, 2021
2 / 24
PLOS ONE Impaired immune response promotes severe COVID-19 disease
patients with different levels of disease severity, as well as in subjects with putative risk factors
(age, sex, physical fitness) for a severe disease course. To analyze the direct effects of SARS-
CoV-2 on PGE2 production, we infected human lung epithelial cells and human precision-
cut-lung-slices (PCLS) with SARS-CoV-2. Additionally, we further dissected the mechanisms
of PGE2 modulation of immune defense, e.g. through B-cell maturation and the formation of
memory cells, and correlated disease severity with lung B-cell content in patient samples. We
further tested strategies to reduce PGE2 production or the effect on the above parameters as
preventive or therapeutic modalities against severe COVID-19.
Materials and methods
Unless otherwise stated, chemicals and reagents were all purchased from Sigma-Aldrich.
Study design
COVID-19 study.
In this study of 89 patients diagnosed with COVID-19, 41 presented
with mild/moderate symptoms and 48 were hospitalized with severe disease. Blood samples
were also obtained from male (n = 18) and female subjects (n = 28) (age 18–50 years) from a
healthy population established by Hannover Unified Biobank (HUB).
At the time of blood sampling, for 29 patients it was known whether they obtained corti-
coids or not. Among those n = 14 obtained no corticoids and n = 15 COVID-19 patients with
mild and severe disease course received corticoids (Dexamethasone n = 11 or Medrol n = 4).
Information on the use of NSAIDs or leukotriene modifiers were not available. None of the
healthy controls were under corticoids or nonsteroidal anti-inflammatory drugs (NSAIDs)
treatment.
The local ethics committees at Hannover Medical School, Comite´ d’Ethique Hospitalo-
Facultaire of UCLouvain, and the Ethical Committee of IEO has been obtained (IEO1271)
approved this study. All patients and healthy control subjects provided written informed con-
sent. The study conforms to the principles outlined in the Declaration of Helsinki.
Physical assessment and exercise program in healthy elderly individuals (rebirth 60plus
cohort, DRKS00013885). All subjects in the Rebirth 60plus cohort (DRKS00013885) were
initially tested for maximum power output on a cycle ergometer with graded exercise test
(GXT). Based on their activities, physical fitness and pathologies, each subject was given an
aerobic exercise training program. Once a month, the subjects were contacted by phone to
assess training progress and adjust the exercise program, if necessary. All subjects of the
Rebirth 60plus study were informed about benefits and risks regarding all study procedures.
Height and weight were measured using a scale (seca gmbh & co. kg, Hamburg, Germany).
Body fat was measured with a medical Body Composition Analyzer mBCA (seca gmbh & co.
kg, Hamburg, Germany). The physical activity was tracked using a GPS watch Forerunner 30
(Garmin Deutschland GmbH, Munich, Germany) and a daily diary where all physical activi-
ties were additionally documented. All study procedures were approved by the local ethics
committee of Hannover Medical School (Vote #7617) and all subjects provided informed writ-
ten consent prior to the commencement of the study procedures.
Blood sampling and blood tests
Blood samples were collected in S-Monovette1 tubes containing ethylenediaminetetraacetic
acid (EDTA, for plasma) or clot activator (for serum) at the time of hospital admission or at
study inclusion (baseline, BL) and at the follow-up (FU) visits after 12 months for the Rebirth
60Plus male and female subjects (age >60 years). Blood samples were also obtained from
young male and female subjects (age 18–50 years) from a healthy population established by
PLOS ONE | https://doi.org/10.1371/journal.pone.0255335 August 4, 2021
3 / 24
PLOS ONE Impaired immune response promotes severe COVID-19 disease
Hannover Unified Biobank (HUB). Plasma or serum was separated by centrifugation at 1500
rpm for 10 min and aliquots were stored at -80˚C. Laboratory workup was performed as part
of routine analysis by hospital laboratories for leukocytes, neutrophils, lymphocytes, platelets,
CRP and LDH. PGE2 serum and plasma levels were measured using the prostaglandin E2
ELISA kit (abcam ab133021) according to the manufacturer’s protocol.
Infection of Calu-3 cells with SARS-CoV-2 and Taxifolin treatment
Calu-3 cells (kindly provided by Prof. Po¨hlmann, German Primate Center, Go¨ttingen; ATCC
Cat# HTB-55; RRID:CVCL_0609) were maintained in Dulbecco’s’ modified Eagle medium
and Vero cells (ATCC-CCL-81; Lot 58484194) in Advanced MEM at 37˚C and 5% CO2. Both
media were supplemented with 10% fetal bovine serum, 2 mM glutamine, 0.1 mM non-essen-
tial amino acids and 1% Penicillin/Streptomycin. Calu-3 cells (4.5x105 cells/well) were seeded
in collagen-coated 24-well plates. For infection, the SARS-CoV-2 (strain SARS-CoV-2/Mu¨n-
chen-1.2/2020/984,p3) [31] kindly provided by Christian Drosten (Charite´, Berlin) through
the European Virus Archive–Global (EVAg) was used. The isolate was propagated and titrated
in Vero cells. Calu-3 cells were pretreated with 100 μM Taxifolin or DMSO (0.15%) for 24 h.
Infection with SARS-CoV-2 isolate was performed at a multiplicity of infection (MOI) of
2.0x10-5 for 4 h at 37˚C in the presence of the compounds. Heat-inactivated virus (15 min,
70˚C) served a negative control. After infection, cells were washed twice with PBS before the
medium containing the respective compound was added. At 48 h post infection, culture super-
natant was collected and heat-inactivated (15 min, 70˚C) prior to the detection of PGE2. RNA
was isolated from cell lysates using a NucleoSpin RNA kit (Macherey-Nagel) according to the
manufacturer’s instructions to analyze virus genome copy numbers, COX-2, HPGD, PTGES2,
PTGES3, TNFa and IFNg expression.
Virus titration in Vero E6 cells for infection of lung slices with SARS-CoV-
2
Vero E6 (ATCC CRL-1586) and Vero cells (ATCC CCL-81) were maintained in Eagle’s Mini-
mum Essential Medium (EMEM) (Lonza) supplemented with 25 mM of HEPES (Gibco),
1 × GlutaMAX (Gibco), 100 U/ml penicillin and 100 μg/ml streptomycin. SARS-CoV2 isolate
(strain SARS-CoV-2/Mu¨nchen-1.2/2020/984,p3) [31] was kindly provided by Christian Dros-
ten. SARS-CoV-2 seed stocks were generated by inoculating Vero E6 (ATCC CRL-1586) at a
MOI of 0.001, collecting and aliqouting the culture supernatant at 72 h post infection (hpi),
then storing at -80˚C in aliquots. SARS-CoV-2 working stocks were generated by an additional
passage on Vero cells (ATCC CCL-81) at a MOI of 0.001. Plaque and median tissue culture
infectious dose (TCID50) assays were performed to titrate the cultured virus after both passages
using Vero cells. This stock was used for the ex vivo infections of human tissues.
Infections of precision-cut human lung slices (PCLS) with SARS-CoV-2
PCLS were maintained in DMEM/F12 medium (Gibco, Thermo Fisher Scientific) supple-
mented with 2 mM of HEPES (Gibco), 1 × GlutaMAX (Gibco), 100 U/ml penicillin and
100 μg/ml streptomycin; this media was also used for virus dilutions and post-infection incu-
bation. On the day of infection, PCLS were rinsed with PBS (without Mg2+ and Ca2+) then
inoculated with 1 × 105 PFU SARS-CoV-2 in 250 μl of media per well in 48-well plates and
incubated at 37˚C. After 2 h, the inoculum was removed and the PCLS were then cultured in
250 μl of DMEM/F12 medium. At 72 and 120 hpi, supernatants were collected and PCLS were
fixed with fixation buffer (4% PFA, 0.1% glutaraldehyde and 200 mM HEPES in ddH2O) for 1
h at room temperature followed by 24 h at 4˚C.
PLOS ONE | https://doi.org/10.1371/journal.pone.0255335 August 4, 2021
4 / 24
PLOS ONE Impaired immune response promotes severe COVID-19 disease
QRT-PCR for NSP7 to confirm SARS-CoV-2 infection
SARS-CoV-2 infections in human Calu-3 cells and human lung slices and tissue were verified
by NSP7 mRNA expression using qRT-PCR (forward primer: GGG CTC AAT GTG TCC
AGT TAC, reverse primer: TTG CCC TGT CCA GCA TT).
Human lung biopsies from acute COVID-19 patients
Patients with acute COVID-19 (AC, n = 6) have been diagnosed with COVID-19 and were
positively tested via PCR as described [4]. All AC patients used in this study showed typical
acute respiratory distress syndrome (ARDS) histopathology typical for COVID-19 disease. In
addition, NSP7 expression was used to detect SARS-CoV-2 virus in biopsies with the limita-
tion that due to heterogeneous distribution of the virus or already cleared acute infection, PCR
is not always positive in every area of the lungs and therefore NSP7 might be not detected.
Multiplex immunohistochemistry of human lung biopsies
The FFPE sections for each group (Control (Ctrl) n = 3, acute COVID-19 (AC) n = 6, trans-
plant rejected (TR) n = 4) were representatively stained with the manual Opal 7-Color IHC Kit
(Akoya Biosciences, Marlborough, MA) as previously described [32]. The primary antibodies
CD4 (Cytomed SP35, 1:50), CD8 (Dako M0755, 1:600), CD68 (Dako PGM1, 1:750) and CD20
(Dako M0755, 1:1000) were combined in sequence with the opal fluorophore CD4-Opal520,
CD8-Opal570, CD20-Opal540 and CD68-Opal650. The sections were scanned with the Vectra
3 System (Akoya Biosciences, Marlborough, MA). The Regions of Interest (ROIs) were
selected representative for small, medium and large vessels for the entire tissue section. The
number of analyzed stamps was 43 for Ctrl, 74 for AC and 56 for TR. For the detection of
CD20+ B cells, the analysis was performed with the inForm Advanced Image Analysis Software
Version 2.3.0 (Akoya Biosciences, Marlborough, MA) and ImageJ 1.53c (Wayne Rasband,
National Institutes of Health, USA). Statistical analysis was performed using the generalized
linear model with Gaussian distribution and weights adjusted according to the number of
ROIs per patient.
Stimulation of human pre-B-cell lines
Human pre-B-cell lines 697 (ACC42 DSMZ collection) and SUP-B15 (ACC389 DSMZ collec-
tion) were cultivated in RPMI (Gibco) supplemented with 10% FBS. 5x105 cells per ml were
pre-incubated with either the EP1/EP2 receptor antagonist AH6809 (10 μM, Tocris) or the
EP4 receptor antagonist GW627368 (10 μM, Tocris) for 2 h. PGE2 (10 μM, Sigma-Aldrich)
was added and cells were harvested after 48 h in TRIzol, or stained with trypan blue (Bio-Rad
laboratories) and counted for measuring live to dead ratio and cell numbers using the TC20
automated cell counter (Bio-Rad laboratories). Control cells were incubated with dissolvents
(DMSO or ethanol (ETHO), 1 μL/ml media). Alternatively, 5x105 per ml 697 and SUP-B15
cells were incubated with 10% human serum from older individuals (>60 y) prior to the com-
mencement of the exercise program at baseline (BL) and after 12M (12M FU) for 48 h and har-
vested in TRIzol. SUP-B15 cells were incubated with 10% human serum from COVID-19
patients and from healthy controls. Cells were harvested after 48 h in TRIzol.
PGE2 and prostaglandin D2 (PGD2) detection in supernatants of Calu-3
PGE2 and PGD2 levels in the supernatants of the cell lines Calu-3 (normalized to total RNA
content) were measured using the prostaglandin E2 ELISA kit (abcam ab133021) or the
PLOS ONE | https://doi.org/10.1371/journal.pone.0255335 August 4, 2021
5 / 24
PLOS ONE Impaired immune response promotes severe COVID-19 disease
prostaglandin D2 ELISA kit (Cayman Chemicals, No. 512031) respectively, according to the
manufacturer’s protocols.
Isolation of RNA and qRT-PCR
Total RNA was isolated with TRIzol (Thermo Fisher Scientific) and cDNA synthesis was per-
formed as described previously [33]. Real-time PCR with the SYBR green dye method (Bril-
liant SYBR Green Mastermix-Kit, Thermo Fisher Scientific) was performed with the AriaMx
Real-Time PCR System (Agilent Technologies) as described [33]. Expression of mRNA levels
was normalized using the 2-ΔΔCT method relative to 18S, beta-2-microglobulin (B2M) and
glyceraldehyde-3-phosphate dehydrogenase (GAPDH). A list of qRT-PCR primers used in
this study is provided in the supplements file S1 Table.
RNA isolation from formalin fixed and paraffin embedded tissue
RNA isolation from formalin-fixed and paraffin embedded tissue was performed using the
Maxwell1 RSC RNA FFPE Purification Kit (Promega Corporation, Madison, WI). RNA con-
tent was measured by using the Qubit RNA IQ Assay (Thermo Fisher Scientific, Waltham,
MA).
Statistical analyses
Statistical analysis was performed using GraphPad Prism version 5.0a, 7.0 and 8.1.2 for Mac
OS X (GraphPad Software, San Diego, CA, USA).
Normal distribution was tested using the D’Agostino normality test or Shapiro-Wilk nor-
mality test if the sample was too small for D’Agostino normality test. Continuous data were
expressed as mean ± SD or median and interquartile range (IQR), according to the normality
of distribution. Comparison between two groups was performed using one sample t-test or
unpaired two-tailed t-test for Gaussian distributed data and the Mann-Whitney-U test where
at least one column was not normally distributed. When comparing more than two groups,
ANOVA and Bonferroni’s post hoc test or Dunnett’s post hoc test were used according to the
normality of distribution. Categorical variables are presented as frequencies (percentages) and
compared using Fisher’s exact test. A two-tailed P value of <0.05 was considered statistically
significant. Correlation for BMI, BW, body fat content and age was analyzed via ozone correla-
tion analysis by using Pearson correlation coefficients for Gaussian distributions or for non-
parametric Spearman correlation coefficients for non-normal distribution.
Results
PGE2 levels in healthy individuals in relation to sex and age
In healthy control individuals aged <50, circulating PGE2 levels were higher (P>0.01) in men
than in women (Fig 1A). Sex-related differences in circulating PGE2 levels were not observed
in older (<60 years) healthy individuals (Fig 1B). Circulating PGE2 levels were markedly
higher in older (>60 years) healthy males and females than in respective sex-matched younger
(<50 years) individuals (Fig 1C and 1D). Both males and females showed a significant positive
correlation of circulating PGE2 levels with age (Fig 1E and 1F), while no correlation with BMI,
body weight (BW) or body fat content was observed (S2 Table, S1 Fig). Controlled physical
exercise for 12 months reduced PGE2 in elderly male and female individuals compared with
their baseline (BL) levels (Fig 1G and 1H and S2 Table). With these indications, we next set to
explore whether PGE2 levels changes in COVID-19 and whether differences in PGE2 levels
could explain severe disease courses after SARS-CoV-2 infection.
PLOS ONE | https://doi.org/10.1371/journal.pone.0255335 August 4, 2021
6 / 24
PLOS ONE Impaired immune response promotes severe COVID-19 disease
Fig 1. (A) The dot plots summarize circulating serum PGE2 levels (pg/ml) of males (n = 18) and females (n = 28)
below the age of 50 years. (B) Dot plots summarize circulating serum PGE2 levels (pg/ml) of males (n = 40) and
females (n = 46) over the age of 60 years. (C) The dot plots summarize circulating serum PGE2 levels (pg/ml) of males
(n = 18) <50y and males (n = 40) >60y. (D) Dot plots summarize circulating serum PGE2 levels (pg/ml) of females
(n = 28) <50y and females (n = 46) >60y. Ozone correlation analysis of serum PGE2 levels with age in (E) males
(n = 66, Spearman r: 0.2564, P-value: 0.0377) and (F) females (n = 76, Spearman r: 0.638, P-value: <0.0001).
Circulating serum PGE2 levels at baseline (BL) and after 12-months follow-up (FU) following controlled physical
training from (G) males (n = 31) and (H) females (n = 37). (A, B, D, G, H) Data are presented as median±IQR,
��P<0.01, ���P<0.001, ����P<0.0001, Mann-Whitney-U test. (C) Data are presented as mean±SD, ��P<0.01,
unpaired two-tailed t-test. (E, F) Ozone correlation, Spearman correlation coefficients, two-tailed P value. Underlying
data can be found in S1 Data.
https://doi.org/10.1371/journal.pone.0255335.g001
PLOS ONE | https://doi.org/10.1371/journal.pone.0255335 August 4, 2021
7 / 24
PLOS ONE Impaired immune response promotes severe COVID-19 disease
Circulating levels of PGE2 in COVID-19 patients and age-matched healthy
controls
We analyzed PGE2 levels in individuals with mild/moderate (n = 41) and severe (n = 48)
COVID-19 disease from hospitals in Hanover (Germany), Milan (Italy) and Brussels (Bel-
gium) and in age-matched healthy controls (n = 31) (Table 1, S3 Table). Clinical data and labo-
ratory characteristics of the COVID-19 patients revealed that the more severely affected
patients were significantly older with a higher proportion of males than females compared
with the mildly/moderately affected group (Table 1). BMI and diabetes rate are increased in
the entire COVID-19 cohort with no significant difference between the mild/moderate and
the severe groups (Table 1). In addition, C reactive protein (CRP) was elevated, while the total
leukocyte- and neutrophil counts were within the normal range, although some patients dis-
played markedly increased levels (Table 1). The mean lymphocyte counts (T- and B-cells) were
reduced in the majority of COVID-19 patients and were specifically low in patients with severe
disease courses (Table 1). Platelets were in the normal range in all COVID-19 patient groups
and lactate dehydrogenase (LDH) was increased and highest in the severely affected patients
(Table 1). Mortality was 15% for the entire cohort with no patient deaths in the mild/moderate
group and 27% of patients dying in the severe disease group who were all of male sex
(Table 1). Circulating PGE2 levels were increased in COVID-19 patients at the time of hospital
admission compared with healthy controls, and PGE2 levels were significantly higher in the
severely affected patients compared with mildly/moderately affected patients (Fig 2A–2F,
Table 1). A direct relationship of PGE2 levels to death events was not observed (Fig 2D–2F).
COVID-19 patients who need hospitalization were defined as severe COVID-19 patients.
Body mass index (BMI), C-reactive protein (CRP), lactate dehydrogenase (LDH), leukocytes
Table 1. Summary of clinical data of the COVID-19 patients.
Parameters
COVID-19 patients total
Mild to moderate COVID-19 disease
Severe COVID-19 disease
(N = 89)
59 (46–68)
30% (27/89)
79.5 (67.75–96.5)
(n = 42)
171.9±9.4
(n = 43)
27.2 (23.7–30.2)
(n = 43)
28% (11/50)
8036±5831
(n = 50)
4999±2697
(n = 35)
1114±564
(n = 36)
107±83
(n = 50)
Age (years, median ± IQR)
Sex female (%)
Body weight (kg, median ± IQR)
Body height (cm, mean ± SD)
BMI (median ± IQR)
Diabetes (%)
Total leucocytes, counts/μl (mean ± SD)
Standard value: 3900–10200 counts/μl
Neutrophils, counts/μl (mean ± SD)
Standard value: 1500–7700 counts/μl
Lymphocytes, counts/μl (mean ± SD)
Standard value: 1100–4500 counts/μl
CRP mg/L (mean ± SD)
Standard value: <5 mg/L
LDH at hospitalization UI/L (median ± IQR)
Standard value: <248 UI/L
Platelets at hospitalization 103/μl (mean ± SD)
Standard value: 160–370 103/μl
Mortality (%)
https://doi.org/10.1371/journal.pone.0255335.t001
(N = 41)
51 (40–67)
44% (18/41)
75 (65.5–86.5)
(n = 25)
170±9.28
(n = 25)
(N = 48)
62 (51–68.75)�
19% (9/48)�
90 (76–100)�
(n = 17)
174.6±9.1
(n = 18)
26.56 (22.96–28.9)
28 (25.5–31.95)
(n = 26)
19% (5/26)
6803±3213
(n = 26)
4664±2614
(n = 21)
1266±608
(n = 21)
69.4±55.95
(n = 26)
(n = 17)
25% (6/24)
9373±7592
(n = 24)
5502±2839
(n = 14)
902±428
(n = 15)
147.1±89���
(n = 24)
459 (348–659)���
(n = 23)
221±83
(n = 24)
27% (13/48)���
363.5 (263.8–518.8)
299 (229–375)
(n = 48)
229±81
(n = 50)
15% (13/89)
(n = 25)
235±80
(n = 26)
0% (0/41)
PLOS ONE | https://doi.org/10.1371/journal.pone.0255335 August 4, 2021
8 / 24
PLOS ONE Impaired immune response promotes severe COVID-19 disease
Fig 2. Circulating PGE2 levels are elevated in COVID-19 patients. The dot plots summarize circulating serum PGE2
levels (pg/ml) of (A) COVID-19 patients (n = 29) and healthy controls (n = 31), and separately (B) for males (COVID-
19 male patients n = 19; healthy male controls n = 14) and (C) for females (COVID-19 female patients n = 10; healthy
female controls n = 17). The dot plots summarize relative circulating plasma PGE2 levels (in %) (D) of patients with
severe (n = 36) and mild (n = 24) disease, and separately (E) for males (severe affected males n = 29; mild affected
males n = 14) and (F) for females (severe affected females n = 7; mild affected females n = 10); the median of patients
with mild disease was set at 100%. Dots representing patients who died with COVID-19 disease are highlighted in red.
(A, B, D-F) Data are presented as median±IQR, �P<0.05, ���P<0.001, ����P<0.0001, Mann-Whitney-U test. (C) Data
are presented as mean±SD, �P<0.05, unpaired two-tailed t-test. Underlying data can be found in S1 Data.
https://doi.org/10.1371/journal.pone.0255335.g002
normal count, neutrophils normal count, and lymphocytes below normal counts, were ana-
lyzed at the time of hospital admission in routine clinical lab tests. Standard values of blood
parameters were indicated in the parameter column. Values outside the normal range were
indicated in bold font. Comparison between the groups of mild and severe COVID-19 was
performed using Student’s t-test for Gaussian distributed data (presented as mean ± SD) and
the Mann-Whitney-U test where at least one column was not normally distributed (presented
as median and interquartile range (IQR)). Categorical variables are presented as frequencies
(percentages) and were compared using Fisher’s exact test. �P<0.05, ��P<0.01, ���P<0.001
severe COVID-19 vs mild to moderate COVID-19 disease. Underlying data can be found in
S1 Data.
Expression of COX-2 and HPGD and secretion of PGE2 in human lung
epithelial cells and precision-cut lung slices infected with SARS-CoV-2
Next, we investigated whether SARS-CoV-2 would enhance PGE2 production in infected host
cells. Human lung epithelial cells (Calu-3 cells) were infected with SARS-CoV-2 (strain SARS-
CoV-2/Mu¨nchen-1.2/2020/984,p3) [31] and infection was confirmed with qRT-PCR for the
SARS-CoV-2 gene encoding nonstructural protein (NSP)7 [34] (Fig 3A). Heat-inactivation of
SARS-CoV-2 infected supernatants of Calu-3 cells was not associated with degradation of
PGE2 (S2A Fig). Infected cells displayed increased secretion of PGE2, which was specifically
prevented by incubation with the PGE2 inhibitor Taxifolin [35, 36] (Fig 3B). The synthesis of
PLOS ONE | https://doi.org/10.1371/journal.pone.0255335 August 4, 2021
9 / 24
PLOS ONE Impaired immune response promotes severe COVID-19 disease
other prostaglandins like PGD2 was not altered by Taxifolin in infected Calu-3 cells (S3A Fig).
Moreover, Taxifolin treatment was not associated with changes in the proliferation capacity of
Calu-3 cells (S3B–S3D Fig).
SARS-CoV-2 infection increased the expression of COX-2 and reduced the expression of
the PGE2 degrading enzyme HPGD but did not alter the expression of the PGE synthase
(PTGES) in Calu-3 cells (Fig 3C–3E). In contrast, the expression of PGE synthase 2 (PTGES2)
and PGE synthase 3 (PTGES3) were significantly reduced by SARS-CoV-2 in Calu-3 lung cells
(Fig 3F and 3G). In line with these results, the production of PGD2 was also increased in
infected Calu-3 cells (S3E Fig). Additionally, SARS-CoV-2 infection markedly induced the
expression of TNFα (644-fold; P<0.05, S2B Fig), which is known to induce COX-2 expression
and with this the PGE2 production in human fibroblasts [37]. The expression of IFNγ could
not be detected in control or in SARS-CoV-2 infected in human Calu-3 lung cells. Also, the ex
vivo infection of living human PCLS with SARS-CoV-2 (viral infection analyzed by NSP7
qRT-PCR, Fig 3H) led to an upregulation of COX-2 expression compared with non-infected
control slices, while HPGD mRNA levels were unchanged and PGE synthase (PTGES) expres-
sion tended to be increased (Fig 3I–3K).
Effect of PGE2 on the expression of pre-B-cell differentiation and survival
factor PAX5 in human pre-B-cells
PGE2 is known to attenuate the proliferation, differentiation and survival of B-cells [38, 39].
Here, we observed that the addition of PGE2 (10 μM, i.e. 3525 pg/ml), in the range measured
in COVID-19 patients’ sera (1300 to >20.000 pg/ml), to two human B-cell precursor lines, 697
and SUP-B15, significantly reduced PAX5 mRNA expression (Fig 4A and 4B). The effect of
PGE2 on PAX5 in 697 and SUP-B15 cells could be blocked by co-treatment with the PGE2
receptor 4 (EP4; PTGER4) antagonist, GW627368 but not with the EP2 receptor antagonist,
AH6809 (Fig 4A). The expression of PTGER4 in 697 and SUP-B15. Cells was confirmed by
qRT-PCR (S4A and S4B Fig). Additionally, PGE2 (10 μM) stimulation was associated with a
reduced 697 cell number (51%) compared to control (100%, p<0.01) treated 697 cells. The
ratio of live to dead pre-B-cells was not altered through PGE2 stimulation indicating that the
decrease in pre-B-cell number is not mediated by enhanced cell death (S4C Fig). However,
PGE2 stimulation was associated with a reduced expression of the proliferationmarkers Ki67,
TOP2A and TPX2 (S4D–S4F Fig) indicating that it reduces the proliferation capacity of pre-B
cells.
Effect of PGE2 on the expression of inflammatory cytokines TNFα and
IFNγ in human pre-B-cells
During SARS-CoV-2 infection upregulation of PANoptosis inducing cytokines, i.e. TNFα and
IFNγ have been reported in immune cells [40]. Here, PGE2 stimulation reduced the expression
of TNFα in both pre-B cell lines 697 and SUP-B15 (S5A and S5B Fig). The expression of IFNγ
was not changed in 697 cells and in SUP-B15 cells, PGE2 reduced its expression (S5C and S5D
Fig).
Effect of serum from elderly individuals before/after physical exercise on
PAX5 expression in human pre-B-cells
PAX5 expression was higher in 697 and SUP-B15 pre-B-cells incubated with serum from
elderly individuals collected after 12 months of controlled physical exercise compared with
their serum before exercise (Fig 4C and 4D and S2 Table). In addition, the EP4 antagonist,
PLOS ONE | https://doi.org/10.1371/journal.pone.0255335 August 4, 2021
10 / 24
PLOS ONE Impaired immune response promotes severe COVID-19 disease
Fig 3. SARS-CoV-2 infection modulates PGE2 secretion and COX-2 and HPGD expression. (A) Representative gel
image of NSP7 mRNA expression of Calu-3 cells infected with SARS-CoV-2 and control cells. (B) The bar graph
summarizes PGE2 content in supernatants of Calu-3 cells infected with SARS-CoV-2 and treated with Taxifolin
(n = 4) compared with untreated mock (n = 6), DMSO control (n = 8) and heat-inactivated (h.i.) SARS-CoV-2 (n = 6)
normalized to total RNA. The bar graphs summarize mRNA expressions of (C) COX-2, (D) HPGD, (E) PTGES, (F)
PTGES2 and (G) PTGES3 of SARS-CoV-2 infected Calu-3 cells (n = 3). (H) Representative gel image of NSP7 and
B2M mRNA expression of SARS-CoV-2 infected lung slices (120 hpi) and control slices. The bar graphs summarize
mRNA expressions of (I) COX-2, (J) HPGD and (K) PTGES of SARS-CoV-2 infected lung slices (120 hpi; n = 3 for
ctrl, n = 4 for SARS-CoV-2 infection). Data are presented as mean±SD, (B) unpaired two-tailed t-test, �P<0.05 vs.
mock, ��P<0.01 vs. mock, ##P<0.01 vs. SARS-CoV-2 + DMSO. (C-G) One sample t-test, �P<0.05, ��P<0.01 vs. ctrl,
(I-K) unpaired two-tailed t-test, �P<0.05 vs. ctrl. Underlying data can be found in S1 Data and uncropped gel images
in S6 Fig.
https://doi.org/10.1371/journal.pone.0255335.g003
PLOS ONE | https://doi.org/10.1371/journal.pone.0255335 August 4, 2021
11 / 24
PLOS ONE Impaired immune response promotes severe COVID-19 disease
Fig 4. PGE2 stimulation of B-cells modulates the immune response. (A) The bar graph summarizes PAX5 mRNA
expression of 697 pre-B-cells treated with AH6809 (10 μM) or GW627368 (10 μM) and PGE2 (10 μM) for 48 h (n = 18
for ctrl and PGE2 treated cells, n = 3 for AH6809 treated cells and n = 6 for GW627368 treated cells). (B) The bar graph
summarizes PAX5 mRNA expression of human pre-B-cell line SUP-B15 with PGE2 (10 μM) for 48 h (n = 9). (C) The
bar graph summarizes PAX5 mRNA expression of 697 pre-B-cells treated with human serum collected at BL and after
12-months FU of controlled physical training (n = 11). (D) The bar graph summarizes PAX5 mRNA expression of
SUP-B15 pre-B-cells treated with human serum collected at BL and after 12-months FU of controlled physical training
(n = 4). The bar graph summarizes PAX5 mRNA expression of (E) 697 (n = 11) and (F) SUP-B15 (n = 2) pre-B-cells
treated with serum from elderly individuals with high PGE2 levels with and without GW627368 (10 μM). Control pre-
B cells were treated with the solvent DMSO. (A) unpaired two-tailed t-test, ��P<0.01 vs. ctrl, #P<0.05 vs. PGE2, (B-F)
One sample t-test, �P<0.05, ��P<0.01 vs. ctrl or BL, the mean of ctrl or BL was set at 100%. Underlying data can be
found in S1 Data.
https://doi.org/10.1371/journal.pone.0255335.g004
GW627368 increased PAX5 in 697 and SUP-B15 pre-B-cells exposed to serum collected before
physical exercise, indicating that the suppressive effect is mediated by PGE2-EP4 (Fig 4E and
4F).
Effect of serum from COVID-19 patients on PAX5 expression in human
pre-B-cells
Serum from COVID-19 patients with elevated PGE2 levels reduced the expression of PAX5 in
SUP-B15 cells compared with serum from healthy controls. Again, this effect was blocked
upon co-treatment with the PGE2 receptor 4 (EP4) antagonist, GW627368 (Fig 5A).
PLOS ONE | https://doi.org/10.1371/journal.pone.0255335 August 4, 2021
12 / 24
PLOS ONE Impaired immune response promotes severe COVID-19 disease
Fig 5. Modulation of the immune response in COVID-19 patients. (A) The bar graph summarizes PAX5 mRNA
expression of SUP-B15 pre-B-cells treated with serum from healthy controls (serum pooled from 9 controls) and from
COVID-19 patients (serum pooled from 9 COVID-19 patients) incubated with and without GW627368 (10 μM).
Control cells were treated with the solvent DMSO (n = 6 wells with control serum and n = 3 wells with serum of
COVID-19 patients with and without GW627368). (B) Representative gel image of NSP7 and B2M mRNA expression
in control lung tissue (ctrl), in lung tissue of patients with severe acute COVID-19 disease (AC) and in lung tissue
obtained after transplant rejection (TR). (C) Immunohistochemical staining for CD68+, CD4+, CD8+ and CD20+
immune cells (scale bar: 100 μm), (D) Dot plot summarizing the immunohistological quantification of CD20 positive
B-cells per area (mm), dot plots summarize mRNA expression of (E) CD20 and (F) of CD138 in control lung tissue
(ctrl), in the lung tissue of patients with severe acute COVID-19 disease (AC) and in lung tissue obtained after
transplant rejection (TR). (A) One sample t-test, ��P<0.01 vs. ctrl, # P<0.05 vs. serum from COVID-19 patients. (D)
Statistical analysis was performed using the generalized linear model with Gaussian distribution and weights adjusted
according to the number of ROIs per patient, ���P<0.001 vs. ctrl, ###P<0.001 vs AC. (E, F) unpaired two-tailed t-test,
��P<0.01 vs. ctrl, �P<0.05 vs. ctrl. Underlying data can be found in S1 Data and uncropped gel images in S7 Fig.
https://doi.org/10.1371/journal.pone.0255335.g005
Analyses of B-cells in lungs from patients who died of severe acute COVID-
19 disease compared with healthy controls and transplant rejection
biopsies
In lung biopsies from patients who died of severe acute COVID-19 disease (AC group, con-
firmed by qRT-PCR for NSP7, Fig 5B), the signals for CD20 pre-B-cells (qRT-PCR and immu-
nohistochemical quantification) and plasma cells (qRT-PCR for CD138) were barely
PLOS ONE | https://doi.org/10.1371/journal.pone.0255335 August 4, 2021
13 / 24
PLOS ONE Impaired immune response promotes severe COVID-19 disease
detectable and lower than in control lung tissue (ctrl) and markedly lower than in lung tissue
obtained after transplant rejection (TR, Fig 5C–5F). Lung tissue immunostaining showed
increased numbers of CD68+ macrophages and CD4+ T-cells in AC and TR compared with
ctrl lung biopsies (Fig 5C).
Discussion
The key finding of this study is that PGE2 is elevated in patients with COVID-19 disease, with
the highest blood levels observed in those severely affected. Furthermore, SARS-CoV-2 itself
upregulates PGE2 in infected host cells and risk factors such as male sex, age and sedentary life
style are also associated with higher PGE2 serum levels. Finally, PGE2 impairs the B-cell medi-
ated immune response at least in part by reducing PAX5 while the PGE2 inhibitor Taxifolin
attenuates SARS-CoV-2 induced PGE2 production. Moreover, regular exercise also reduces
PGE2 levels in elderly subjects, which is associated with increased PAX5 production in B-cells
exposed to these sera. Thus, PGE2 may emerge as a modulating factor for disease severity and
development of immunity and could therefore be a therapeutic target in COVID-19 preven-
tion and treatment.
Since it is known that PGE2 can exert immunosuppressive effects during viral infection
[25–27], its elevation might critically reduce the initial defense against SARS-CoV-2 and may
thereby lead to more severe disease courses. Interestingly, our data show that the SARS-CoV-2
virus, not only hijacks the host cell gene expression machinery in order to replicate, but also
forces infected host cells to produce PGE2 by upregulating the PGE-generating enzyme COX-
2, and at least in part by reducing the expression of the PGE2-degrading enzyme HPGD (Fig
6). In line with the upregulation of COX-2 but without a specific upregulation of PGE2
synthases by SARS-CoV-2 in infected human lung cells, the production of another prostaglan-
din, PGD2, was also increased. However, to study the regulation and role of PGD2 in COVID-
19 disease was beyond the scope of the present study and needs further investigation. In addi-
tion, we provide evidence that reported risk factors for more severe COVID-19 disease
courses, i.e. male sex, age and a sedentary life style [13, 41] are associated with higher PGE2
levels as PGE2 serum levels are higher in men than women, higher in elderly (>60 years) indi-
viduals of both sexes than in younger individuals, and PGE2 levels in elderly could be reduced
by regular exercise (Fig 6). These findings might explain why males or elderly individuals are
more affected than females or younger individuals. Sex-related differences in circulating PGE2
levels appeared to be specific for younger individuals since in the healthy cohort older <60
years no such differences were observed. Whether age-related hormonal changes in older
females contributes to the age effect in women needs to elucidated in future studies.
In addition to already known effects of PGE2 on immune cells, we discovered a novel
mechanism by which PGE2 in serum from COVID-19 patients specifically impacts on pre-B-
cells since PGE2 in the sera of COVID-19 patients reduces the expression of PAX5 in human
pre-B-cells via its EP4 receptor. PAX5 is a master regulator of most aspects of the life cycle of
B-cells as it represses the transcription of genes required for the development of other hemato-
poietic lineages and plasma cells and by controlling numerous genes that are required for early
development, antigen-receptor recombination, signaling and adhesion [42–44]. Moreover,
while high PAX5 expression is necessary for the above described processes, its reduction is
important for the final differentiation of short-lived plasma cells and their antibody (AB) pro-
duction. Thereby, high PGE2 serum levels on one hand reduces the number of pre-B-cells, but
on the other hand boosts the terminal differentiation of B-cells towards short-lived plasma
cells, two features that on the long run would lead to depleting the B-cell reservoir. This feature
may explain why some patients with initially high SARS-CoV-2-directed AB titers but evolving
PLOS ONE | https://doi.org/10.1371/journal.pone.0255335 August 4, 2021
14 / 24
PLOS ONE Impaired immune response promotes severe COVID-19 disease
PLOS ONE | https://doi.org/10.1371/journal.pone.0255335 August 4, 2021
15 / 24
PLOS ONE Impaired immune response promotes severe COVID-19 disease
Fig 6. Schematic representation of pleiotropic influences of SARS-CoV-2 infection, physical activity and age on
PGE2 levels and the ensuing altered immune response. (A) Modulators of PGE2 synthesis and degradation are
SARS-CoV-2 infection, but also physical inactivity, sex and older age, which are all risk factors for more severe
COVID-19 disease courses [7, 13–15]. Additionally, SARS-CoV-2 infection induces TNFα expression that is known to
mediate increased COX-2 expression [37]. These modulators upregulate the expression of the PGE2-generating
enzyme COX-2 and at least in part reduce the expression of the PGE2-degrading enzyme HPGD, which results in
increased generation and secretion of PGE2. PGE2 targets the innate immune system (monocytes/macrophages),
where it lowers its efficacy to remove pathogens in part by reducing the release of cytokines [46, 59]. Additionally,
PGE2 impairs the response of the adaptive immune system against pathogens by lowering proliferation and survival of
T-cells and inducing T-cell [3, 52, 60, 61]. Furthermore, PGE2 is impairing the B-cell response to pathogens in part by
directly suppressing the B-cell specific transcription factor PAX5 [62]. Increased PGE2 secretion can be prevented by
physical exercise and specific PGE2 inhibitors such as Taxifolin. In addition, Taxifolin reduces viral replication. The
low immune response (phase 1) may enable the entry of secondary infections with bacteria and fungi and reinfections
with SARS-CoV-2 associated with tissue and organ injury, formation of autoantibodies potentially leading to a
cytokine storm and an excessive immune response [20–24]. (B) In pre-B-cells, PAX5 is responsible for suppressing
other hematopoietic differentiation programs and promotes proliferation, survival and differentiation of pre-B-cells
[42–44]. PGE2 reduces PAX5 expression via its EP4 receptor, which not only reduces their survival and proliferation
but boosts the differentiation of B-cells towards plasma cells and may even allow transdifferentiation, features that may
lead to the cytokine storm but also the depletion of the B-cell pool (and germinal centers) [19, 42]. In addition, since
PAX5 is important for the formation memory cells, PGE2 is therefore also lowering the formation of immunity [44,
47]. Blocking the EP4 signaling with the EP4 receptor antagonist GW627368 prevents downregulation of PAX5 in pre-
B-cells and may improve viral defense and formation of immunity against SARS-CoV-2.
https://doi.org/10.1371/journal.pone.0255335.g006
towards a severe disease course display a reduction in germinal centers [19] and reduced B-cell
response thereafter. Our findings in postmortem lung tissue of patients who died of COVID-
19 are in line with this interpretation. Indeed, we detected reduced CD20+ B-cells numbers in
COVID-19 lung tissue in comparison with healthy control tissue or with transplant rejection
lung biopsies. Likewise, other reports show no significant lymphocyte invasion in cardiac tis-
sue despite the presence of SARS-CoV-2 particles [10, 11]. Additional studies suggest higher
risks for severe disease courses in COVID-19 patients with dysfunctional B-cells due to com-
mon variable immune deficiencies (CVIDs) [18], while in turn, patients with larger pools of
naïve B-cells seem to build a more effective immune response to SARS-CoV-2 [45].
The observed low B-cell signals in lung biopsies from patients who died during acute
SARS-CoV-2 infection may also point to loss of these immune cells by PANoptosis (inflamma-
tory cell death). In this regard, Karki et al. reported that during SARS-CoV-2 infection a com-
bination of TNFα and IFNγ could induce PANoptosis [40]. However, we observed the
opposite, i.e. PGE2 reduced the expression of TNFα and IFNγ in pre-B-cells, a feature that has
also been reported for monocytes and macrophages [46].
Moreover, we found that PGE2 reduces the proliferation of human pre-B-cells, an observa-
tion that fits well with the PGE2-mediated reduction of PAX5 and may thereby contribute to
rarification of B-cells in infected tissue.
In addition, since we observed that SARS-CoV-2-infected lung cells upregulate TNFα
expression and since TNFα is known to induce COX-2 expression, we found one possible
mechanism how SARS-CoV-2 may upregulated PGE2 production in infected tissues (Fig 6A)
[37].
As reported above, high PGE2 in COVID-19 serum impairs the B-cell mediated immune
response at least in part by reducing PAX5. PAX5 expression is also necessary for the develop-
ment of memory B-cells after follicular B-cells have encountered antigens [44, 47]. In this
regard, elevated PGE2 would also reduce the ability of an organism to develop longstanding
immunity after COVID-19 infection. Indeed, there are reports on reinfection in individuals
with SARS-CoV-2 [48–50] including a recent case report of a patient with a CD20+ B-cell
acute lymphoblastic leukemia who developed high AB titers against COVID-19 after an initial
recovery. However, the patient experienced a viral reactivation after she lost her COVID-19
PLOS ONE | https://doi.org/10.1371/journal.pone.0255335 August 4, 2021
16 / 24
PLOS ONE Impaired immune response promotes severe COVID-19 disease
AB following the administration of rituximab, cytarabine, and dasatinib for her leukemia, and
experienced severe COVID-19 pneumonia with lymphopenia and high inflammatory markers
[51]. PGE2 not only affects B-cells, but also promotes T-cell exhaustion and viral expansion
through EP2 and EP4, as revealed by recent studies [52] and immunosuppression caused by
T-cell depletion and exhaustion have been suggested as contributing to viral persistence and
mortality in COVID-19 patients [3].
Based on the suspected crucial role of PGE2 for COVID-19 disease courses, we tested the
potential of the PGE2 inhibitor Taxifolin, also known as dihydroquercetin, to limit SARS-CoV-
2-induced PGE2 production in human lung cells (Fig 6A). In agreement with our hypothesis
that PGE2 contributes to severe COVID-19 disease, Taxifolin significantly reduced PGE2 pro-
duction in infected lung cells. Additionally, a recent publication on screening for natural inhibi-
tors for SARS-CoV-2 in silico identified Taxifolin as a direct inhibitor of the SARS-CoV-2 main
protease [53]. Taxifolin is a potent flavonoid with anti-inflammatory activity, which is present as
a natural compound in vegetables and fruits and the Siberian larch, Larix sibirica, [35, 36]. It is
readily available in foodstuffs and could be tested directly in COVID-19 patients. PGE2 synthesis
can be inhibited by NSAIDs, which block COX-1 and -2. However, it is known that NSAIDs are
interfering with the RAAS [54] and in this context, controversial data have been reported sug-
gesting that NSAIDs may favor SARS-CoV-2 entry by upregulating ACE2 [55, 56]. Moreover,
NSAIDs by inhibiting COX-1 and -2 may also reduce the generation of additional prostaglan-
dins, which may have beneficial effects. Therefore, and because the safety of using NSAIDs in
the treatment of COVID-19 patients is discussed critical, we decided to use Taxifolin as an alter-
native treatment strategy. Indeed, we could show that Taxifolin blocked only the SARS-CoV-
2-induced PGE2 synthesis but not the synthesis PGD2 in infected lung cells. Inhibition of the
microsomal prostaglandin E synthase-1 (mPGES-1) by sonlicromanol (Khondrion; a drug cur-
rently in phase 2b studies for mitochondrial disease), may also be beneficial in COVID-19
patients (Fig 6A). Moreover, COVID-19 patients could also benefit from COX-inhibitors such
as aspirin and ibuprofen in the early phase of disease as suggested by a recent review [57]. Treat-
ment of mild and severely affected patients with corticoids, like Dexamethasone or Medrol, has
been associated with better outcome. Here, we observed that corticoids seem to have no effect
on circulating PGE2 levels although number of patients in these subgroup analyses was too low
to be conclusive. Finally, we provide evidence that regular physical activity lowers PGE2 in the
serum of elderly individuals without COVID-19 infection and may thereby support their
immune systems in fighting SARS-CoV-2 infection (Fig 6A).
Thus, known risk factors for severe COVID-19 disease such as age, sex and physical inactiv-
ity are associated with elevated PGE2 levels prior infection and may thereby contribute to a
reduced immune response at the time of SARS-CoV-2 infection. In addition, the SARS-CoV-2
infection may further compromise the immune response by further upregulating PGE2 in
those individuals with pre-existing higher PGE2 levels. Furthermore, it is known that also the
exposure to high levels SARS-CoV-2 virus particles contribute to severe COVID-19 disease
also in individuals with otherwise low risk factors (for example severe disease cases in nurses
and physicians) [58]. As we could demonstrate that SARS-CoV-2-infected host cells produce
high levels of PGE2, a massive infection with SARS-CoV-2 virus may lead to high PGE2 secre-
tion and high circulating PGE2 levels, which subsequently reduced the immune response also
in individual with otherwise low risk for severe disease.
Conclusions
In conclusion, our data suggest that PGE2 production, either induced by SARS-CoV-2 infec-
tion or determined by endogenous and exogenous risk factors critically influences COVID-19
PLOS ONE | https://doi.org/10.1371/journal.pone.0255335 August 4, 2021
17 / 24
PLOS ONE Impaired immune response promotes severe COVID-19 disease
disease severity, (Fig 6A). Mechanistically, we show that PGE2 specifically targets B-cells by
reducing PAX5, a key factor for B-cell proliferation and differentiation (Fig 6A and 6B).
Reducing PGE2 levels preventively and/or during COVID-19 disease may therefore provide a
valuable therapeutic strategy to prevent and fight SARS-CoV-2 infection and to enhance and
prolong immunity.
Limitations of the study
Limitations of our study include the limited numbers of blood samples from COVID-19
patients and that clinical data on COVID-19 patients, i.e. as C-reactive protein (CRP), lactate
dehydrogenase (LDH), leukocytes normal count, neutrophils normal count, and lymphocytes
were not available for all patients.
PGE2 synthesis can be blocked by corticosteroids that inhibit the phospholipases or by
NSAIDs that inhibit the cyclooxygenase. In this study, at the time of blood sampling a part of
the COVID-19 patients with mild or severe disease were treated with corticosteroids or
NSAIDs. Information on the use of NSAIDs or leukotriene modifiers were not available. PGE2
levels in those patients might be underestimated, since both medications may reduce PGE2
biosynthesis.
Most individuals in the healthy elderly collective displayed age-related normal BMI and
numbers in subgroup with increased or reduced BMI were too low to perform conclusive cor-
relation analyses with PGE2 levels.
Serum and plasma samples have to be stored at -80˚C immediately to avoid degradation of
PGE2 and to avoid further prostaglandin synthesis by COX-2. For the present study serum
and plasma was immediately being processed, frozen and stored at -80˚C.
Venipuncture and ex vivo platelet activation may alter plasma prostanoid concentrations, a
feature that cannot be completely excluded.
Supporting information
S1 Fig. PGE2 serum levels showed no correlation with BMI, BW or body fat content.
Ozone correlation analysis of serum PGE2 levels with (A-E) BMI ((A) males: n = 40, Spearman
r: -0.1485, P value: 0.3604; (B) males in normal range BMI 25–30: n = 24, Spearman r: -0.1231,
P value: 0.5667; (C) males with a BMI >30: n = 9, Spearman r: 0.3167, P value: 0.4101 (D)
males with a BMI <25: n = 7, Spearman r: -0.2143, P value: 0.6615 (E) females: n = 45, Pearson
r: 0.03956, P value: 0.7964), (F, G) BW (males: n = 40, Spearman r:-0.08246, P value: 0.6130;
females: n = 45, Pearson r: 0.05614, P value: 0.7142) and (H, I) body fat content (males: n = 37,
Pearson r:-0.03295, P value: 0.8465; females: n = 43, Pearson r: 0.1374, P value: 0.3797) in
(A-D, F, H) males and (E, G, I) females. (A-I) Ozone correlation, Spearman or Pearson corre-
lation coefficients, two-tailed P value. Underlying data can be found in S1 Data.
(TIFF)
S2 Fig. SARS-CoV-2 infection in Calu-3 cells. (A) Heat-inactivation (h.i.) of PGE2 for 30
min at 70˚C compared to untreated PGE2 (ctrl) from the same sample (n = 4). Data are pre-
sented as mean±SD, ctrl was set at 100%, one-sample t-test. (B) The bar graph summarizes
TNFa mRNA expression of SARS-CoV-2 infection in Calu-3 cells in cell culture lysates (n = 3
independent cell culture experiments). Data are presented as mean±SD, mock was set at 100%,
�P<0.05 vs mock, #P<0.05 vs h.i., one-way ANOVA, Dunnett post hoc test. Underlying data
can be found in S1 Data.
(TIFF)
PLOS ONE | https://doi.org/10.1371/journal.pone.0255335 August 4, 2021
18 / 24
PLOS ONE Impaired immune response promotes severe COVID-19 disease
S3 Fig. Taxifolin treatment has no effect on the secretion of PGD2 or the proliferation
capacity of Calu-3 cells. The bar graph summarizes PGD2 content in supernatants of Calu-3
cells infected with SARS-CoV-2 and treated for 48 h with Taxifolin (100 μM; n = 12) compared
with DMSO control (n = 10). The bar graphs summarize the mRNA expression of the prolifer-
ation markers (B) Ki67, (C) TOP2A and (D) TPX2 of Calu-3 cells treated with Taxifolin
(100 μM) for 48 h (n = 7 for ctrl and PGE2 treated cells). (E) The bar graph summarizes PGD2
content in supernatants of Calu-3 cells infected with SARS-CoV-2 (n = 6) compared with
untreated mock (n = 6) and heat-inactivated (h.i.) SARS-CoV-2 (n = 6) normalized to total
RNA. (A-E) Data are presented as mean±SD, (A, B, C) unpaired two-tailed t-test, n.s. (D)
Mann-Whitney-U test, n.s. (E) mock was set at 100%, ��P<0.01 vs mock, #P<0.05 vs h.i., one-
way ANOVA, Bonferroni’s post hoc test. Underlying data can be found in S1 Data.
(TIFF)
S4 Fig. PGE2 stimulation of pre-B-cells modulates the cell number due to alterations of
the proliferation capacity. Representative gel images of PTGER4 and B2M mRNA expression
in pre-B-cell lines (A) 697 and (B) SUP-B15. (C) The bar graph summarizes the percentage of
live cells of control treated and PGE2 (10 μM) treated 697 cells after 48h stimulation. Total cell
number was set at 100%. (D-F) The bar graphs summarize the mRNA expression of the prolif-
eration markers (C) Ki67, (D) TOP2A and (E) TPX2 of pre-B-cells 697 treated with PGE2
(10 μM) for 48 h (n = 5 for ctrl and PGE2 treated cells). (C-E) Data are presented as mean±SD,
(C-E) n. s., ��P<0.01 vs ctrl, unpaired two-tailed t-test. Underlying data can be found in S1
Data and uncropped gel images in S8 Fig.
(TIFF)
S5 Fig. PGE2 stimulation of pre-B-cells is not associated with elevated TNFa or IFNg
expression. The bar graph summarizes TNFa mRNA expression of (A) 697 or (B) SUP-B18
pre-B-cells treated with PGE2 (10 μM) after 48 h (n = 5). Control cells were treated with the
solvent ETHO (n = 5). The bar graph summarizes IFNg mRNA expression of (C) 697 or (D)
SUP-B18 pre-B-cells treated with PGE2 (10 μM) after 48 h (n = 5). Control cells were treated
with the solvent ETHO (n = 4). (A-D) Data are presented as mean±SD, (A, C, D) n. s.,
��P<0.01 vs ctrl, unpaired two-tailed t-test and (B) ��P<0.01, Mann-Whitney-U test. Underly-
ing data can be found in S1 Data.
(TIFF)
S6 Fig. The uncropped gel for Fig 3A and 3H.
(TIFF)
S7 Fig. The uncropped gel for Fig 5B.
(TIFF)
S8 Fig. The uncropped gel for S4A and S4B Fig.
(TIFF)
S1 Data. Numerical raw data. All numerical raw data are combined in a single excel file,
“S1_Data.xlsx,” this file consists of several spreadsheets and each contains the data of 1 figure
or table.
(XLSX)
S1 Table. List of human qRT-PCR primers.
(DOCX)
S2 Table. Summary of clinical data from male and female probands baseline (BL) and
after 12 M Follow-Up (FU) controlled exercise (E). Body mass index (BMI) was determined
PLOS ONE | https://doi.org/10.1371/journal.pone.0255335 August 4, 2021
19 / 24
PLOS ONE Impaired immune response promotes severe COVID-19 disease
with BMI = bodyweight (BW) / squared height. Body weight, body height, BMI, body fat and
activity were analyzed at BL and after 12M FU controlled exercise. Comparison between the
groups BL vs 12M FU was performed using Student’s t-test for Gaussian distributed data (pre-
sented as mean ± SD) and the Mann-Whitney-U test where at least one column was not nor-
mally distributed (presented as median and interquartile range (IQR)). ���P<0.001,
����P<0.00001 BL vs 12M FU. Underlying data can be found in S1 Data.
(DOCX)
S3 Table. Summary of clinical data of healthy controls from Fig 1A.
(DOCX)
Acknowledgments
We thank Sergej Erschow, Silvia Gutzke, Brigit Brandt, Angelica Julieth Diaz Basabe, Delphine
De Mulder, Thomas Gerlach and Giulietta Saletti for excellent technical assistance, and Dr.
Helge Stark for the bioinformatic analysis input. The SARS-CoV-2 virus isolate was kindly
provided by Christian Drosten, Charite´, Berlin.
Author Contributions
Conceptualization: Melanie Ricke-Hoch, Elisabeth Stelling, Thomas Pietschmann, Emilio
Hirsch, Danny Jonigk, Uwe Tegtbur, Axel Schambach, Axel Haverich, Denise Hilfiker-
Kleiner.
Data curation: Melanie Ricke-Hoch, Elisabeth Stelling, Martina Kasten, Thomas Gausepohl,
Anne Ho¨fer, Danny Jonigk, Julian Eigendorf, Lena Mink, Michaela Scherr, Tobias J. Pfeffer,
Denise Hilfiker-Kleiner.
Formal analysis: Melanie Ricke-Hoch, Elisabeth Stelling, Graham Brogden, Gisa Gerold,
Anne Ho¨fer, Danny Jonigk, Denise Hilfiker-Kleiner.
Funding acquisition: Melanie Ricke-Hoch, Gisa Gerold, Thomas Pietschmann, Jean-Luc Bal-
ligand, Emilio Hirsch, Guus F. Rimmelzwaan, Mark P. Ku¨hnel, Danny Jonigk, Thomas
Illig, Axel Schambach, Denise Hilfiker-Kleiner.
Investigation: Melanie Ricke-Hoch, Elisabeth Stelling, Lisa Lasswitz, Antonia P. Gunesch,
Francisco J. Zapatero-Belincho´n, Graham Brogden, Gisa Gerold, Federica Facciotti, Husni
Elbahesh, Guus F. Rimmelzwaan, Anne Ho¨fer, Danny Jonigk, Julian Eigendorf, Uwe Tegt-
bur, Lena Mink, Tobias J. Pfeffer, Axel Haverich, Denise Hilfiker-Kleiner.
Methodology: Melanie Ricke-Hoch, Elisabeth Stelling, Lisa Lasswitz, Antonia P. Gunesch,
Francisco J. Zapatero-Belincho´n, Graham Brogden, Gisa Gerold, Thomas Pietschmann,
Husni Elbahesh, Anne Ho¨fer, Danny Jonigk, Thomas Illig, Andres Hilfiker, Axel Haverich,
Denise Hilfiker-Kleiner.
Project administration: Melanie Ricke-Hoch, Gisa Gerold, Danny Jonigk, Julian Eigendorf,
Uwe Tegtbur, Denise Hilfiker-Kleiner.
Resources: Lisa Lasswitz, Antonia P. Gunesch, Francisco J. Zapatero-Belincho´n, Graham
Brogden, Gisa Gerold, Virginie Montiel, Jean-Luc Balligand, Federica Facciotti, Emilio
Hirsch, Anne Ho¨fer, Mark P. Ku¨hnel, Danny Jonigk, Uwe Tegtbur, Lena Mink, Thomas
Illig, Tobias J. Pfeffer, Andres Hilfiker, Axel Haverich.
Supervision: Melanie Ricke-Hoch, Gisa Gerold, Thomas Pietschmann, Mark P. Ku¨hnel,
Danny Jonigk, Denise Hilfiker-Kleiner.
PLOS ONE | https://doi.org/10.1371/journal.pone.0255335 August 4, 2021
20 / 24
PLOS ONE Impaired immune response promotes severe COVID-19 disease
Validation: Melanie Ricke-Hoch, Elisabeth Stelling, Gisa Gerold, Anne Ho¨fer, Danny Jonigk,
Denise Hilfiker-Kleiner.
Visualization: Melanie Ricke-Hoch, Elisabeth Stelling, Graham Brogden, Anne Ho¨fer, Danny
Jonigk, Michaela Scherr, Denise Hilfiker-Kleiner.
Writing – original draft: Melanie Ricke-Hoch, Elisabeth Stelling, Lisa Lasswitz, Antonia P.
Gunesch, Husni Elbahesh, Anne Ho¨fer, Mark P. Ku¨hnel, Danny Jonigk, Denise Hilfiker-
Kleiner.
Writing – review & editing: Melanie Ricke-Hoch, Elisabeth Stelling, Francisco J. Zapatero-
Belincho´n, Graham Brogden, Gisa Gerold, Thomas Pietschmann, Virginie Montiel, Jean-
Luc Balligand, Thomas Gausepohl, Husni Elbahesh, Guus F. Rimmelzwaan, Anne Ho¨fer,
Mark P. Ku¨hnel, Danny Jonigk, Julian Eigendorf, Uwe Tegtbur, Lena Mink, Thomas Illig,
Axel Schambach, Tobias J. Pfeffer, Andres Hilfiker, Axel Haverich, Denise Hilfiker-Kleiner.
References
1.
Fried JA, Ramasubbu K, Bhatt R, Topkara VK, Clerkin KJ, Horn E, et al. The Variety of Cardiovascular
Presentations of COVID-19. Circulation. 2020; 141(23):1930–6. https://doi.org/10.1161/
CIRCULATIONAHA.120.047164 PMID: 32243205
2.
Lau MSY, Grenfell B, Thomas M, Bryan M, Nelson K, Lopman B. Characterizing superspreading events
and age-specific infectiousness of SARS-CoV-2 transmission in Georgia, USA. Proc Natl Acad Sci U S
A. 2020; 117(36):22430–5. https://doi.org/10.1073/pnas.2011802117 PMID: 32820074
3. Vardhana SA, Wolchok JD. The many faces of the anti-COVID immune response. The Journal of exper-
imental medicine. 2020; 217(6). https://doi.org/10.1084/jem.20200678 PMID: 32353870
4. Ackermann M, Verleden SE, Kuehnel M, Haverich A, Welte T, Laenger F, et al. Pulmonary Vascular
Endothelialitis, Thrombosis, and Angiogenesis in Covid-19. N Engl J Med. 2020; 383(2):120–8. https://
doi.org/10.1056/NEJMoa2015432 PMID: 32437596
5. Yin S, Huang M, Li D, Tang N. Difference of coagulation features between severe pneumonia induced
by SARS-CoV2 and non-SARS-CoV2. J Thromb Thrombolysis. 2020.
6.
7.
Zheng YY, Ma YT, Zhang JY, Xie X. COVID-19 and the cardiovascular system. Nature reviews Cardiol-
ogy. 2020. https://doi.org/10.1038/s41569-020-0360-5 PMID: 32139904
Zheng Z, Peng F, Xu B, Zhao J, Liu H, Peng J, et al. Risk factors of critical & mortal COVID-19 cases: A
systematic literature review and meta-analysis. J Infect. 2020.
8. Nicolai L, Leunig A, Brambs S, Kaiser R, Weinberger T, Weigand M, et al. Immunothrombotic Dysregu-
lation in COVID-19 Pneumonia Is Associated With Respiratory Failure and Coagulopathy. Circulation.
2020; 142(12):1176–89. https://doi.org/10.1161/CIRCULATIONAHA.120.048488 PMID: 32755393
9. Szekely Y, Lichter Y, Taieb P, Banai A, Hochstadt A, Merdler I, et al. Spectrum of Cardiac Manifesta-
tions in COVID-19: A Systematic Echocardiographic Study. Circulation. 2020; 142(4):342–53. https://
doi.org/10.1161/CIRCULATIONAHA.120.047971 PMID: 32469253
10. Chen L, Li X, Chen M, Feng Y, Xiong C. The ACE2 expression in human heart indicates new potential
mechanism of heart injury among patients infected with SARS-CoV-2. Cardiovasc Res. 2020; 116
(6):1097–100. https://doi.org/10.1093/cvr/cvaa078 PMID: 32227090
11.
Fox SE, Li G, Akmatbekov A, Harbert JL, Lameira FS, Brown JQ, et al. Unexpected Features of Cardiac
Pathology in COVID-19 Infection. Circulation. 2020; 142(11):1123–5. https://doi.org/10.1161/
CIRCULATIONAHA.120.049465 PMID: 32689809
12. Sharma A, Garcia G Jr., Wang Y, Plummer JT, Morizono K, Arumugaswami V, et al. Human iPSC-
Derived Cardiomyocytes Are Susceptible to SARS-CoV-2 Infection. Cell Rep Med. 2020; 1(4):100052.
https://doi.org/10.1016/j.xcrm.2020.100052 PMID: 32835305
13. Wenham C, Smith J, Morgan R, Gender, Group C-W. COVID-19: the gendered impacts of the outbreak.
Lancet. 2020; 395(10227):846–8. https://doi.org/10.1016/S0140-6736(20)30526-2 PMID: 32151325
14. Brandt EB, Beck AF, Mersha TB. Air pollution, racial disparities, and COVID-19 mortality. J Allergy Clin
Immunol. 2020; 146(1):61–3. https://doi.org/10.1016/j.jaci.2020.04.035 PMID: 32389591
15. Martelletti L, Martelletti P. Air Pollution and the Novel Covid-19 Disease: a Putative Disease Risk Factor.
SN Compr Clin Med. 2020:1–5. https://doi.org/10.1007/s42399-020-00274-4 PMID: 32296757
PLOS ONE | https://doi.org/10.1371/journal.pone.0255335 August 4, 2021
21 / 24
PLOS ONE Impaired immune response promotes severe COVID-19 disease
16.
Li D, Chen Y, Liu H, Jia Y, Li F, Wang W, et al. Immune dysfunction leads to mortality and organ injury in
patients with COVID-19 in China: insights from ERS-COVID-19 study. Signal Transduct Target Ther.
2020; 5(1):62. https://doi.org/10.1038/s41392-020-0163-5 PMID: 32371949
17. Wilk AJ, Rustagi A, Zhao NQ, Roque J, Martinez-Colon GJ, McKechnie JL, et al. A single-cell atlas of
the peripheral immune response in patients with severe COVID-19. Nat Med. 2020; 26(7):1070–6.
https://doi.org/10.1038/s41591-020-0944-y PMID: 32514174
18. Quinti I, Lougaris V, Milito C, Cinetto F, Pecoraro A, Mezzaroma I, et al. A possible role for B cells in
COVID-19? Lesson from patients with agammaglobulinemia. J Allergy Clin Immunol. 2020. https://doi.
org/10.1016/j.jaci.2020.04.013 PMID: 32333914
19. Kaneko N, Kuo HH, Boucau J, Farmer JR, Allard-Chamard H, Mahajan VS, et al. Loss of Bcl-6-
Expressing T Follicular Helper Cells and Germinal Centers in COVID-19. Cell. 2020.
20. Catanzaro M, Fagiani F, Racchi M, Corsini E, Govoni S, Lanni C. Immune response in COVID-19:
addressing a pharmacological challenge by targeting pathways triggered by SARS-CoV-2. Signal
Transduct Target Ther. 2020; 5(1):84. https://doi.org/10.1038/s41392-020-0191-1 PMID: 32467561
21. Hu B, Huang S, Yin L. The cytokine storm and COVID-19. J Med Virol. 2020.
22. Mehta P, McAuley DF, Brown M, Sanchez E, Tattersall RS, Manson JJ, et al. COVID-19: consider cyto-
kine storm syndromes and immunosuppression. Lancet. 2020; 395(10229):1033–4. https://doi.org/10.
1016/S0140-6736(20)30628-0 PMID: 32192578
23. Wang SF, Tseng SP, Yen CH, Yang JY, Tsao CH, Shen CW, et al. Antibody-dependent SARS corona-
virus infection is mediated by antibodies against spike proteins. Biochemical and biophysical research
communications. 2014; 451(2):208–14. https://doi.org/10.1016/j.bbrc.2014.07.090 PMID: 25073113
24.
Zhang Q, Bastard P, Liu Z, Le Pen J, Moncada-Velez M, Chen J, et al. Inborn errors of type I IFN immu-
nity in patients with life-threatening COVID-19. Science. 2020; 370(6515).
25. Sander WJ, O’Neill HG, Pohl CH. Prostaglandin E2 As a Modulator of Viral Infections. Front Physiol.
2017; 8:89. https://doi.org/10.3389/fphys.2017.00089 PMID: 28261111
26. Coulombe F, Jaworska J, Verway M, Tzelepis F, Massoud A, Gillard J, et al. Targeted prostaglandin E2
inhibition enhances antiviral immunity through induction of type I interferon and apoptosis in macro-
phages. Immunity. 2014; 40(4):554–68. https://doi.org/10.1016/j.immuni.2014.02.013 PMID: 24726877
27.
28.
Full F, Gack MU. Prostaglandin E2: the villain in the host response to influenza virus. Immunity. 2014;
40(4):453–4. https://doi.org/10.1016/j.immuni.2014.03.008 PMID: 24745327
Liu L, Cao Z, Chen J, Li R, Cao Y, Zhu C, et al. Influenza A virus induces interleukin-27 through cycloox-
ygenase-2 and protein kinase A signaling. J Biol Chem. 2012; 287(15):11899–910. https://doi.org/10.
1074/jbc.M111.308064 PMID: 22343630
29. Schmidleithner L, Thabet Y, Schonfeld E, Kohne M, Sommer D, Abdullah Z, et al. Enzymatic Activity of
HPGD in Treg Cells Suppresses Tconv Cells to Maintain Adipose Tissue Homeostasis and Prevent
Metabolic Dysfunction. Immunity. 2019; 50(5):1232–48 e14. https://doi.org/10.1016/j.immuni.2019.03.
014 PMID: 31027998
30. Hoxha M. What about COVID-19 and arachidonic acid pathway? Eur J Clin Pharmacol. 2020; 76
(11):1501–4. https://doi.org/10.1007/s00228-020-02941-w PMID: 32583353
31. Wolfel R, Corman VM, Guggemos W, Seilmaier M, Zange S, Muller MA, et al. Virological assessment
of hospitalized patients with COVID-2019. Nature. 2020; 581(7809):465–9. https://doi.org/10.1038/
s41586-020-2196-x PMID: 32235945
32. Mascaux C, Angelova M, Vasaturo A, Beane J, Hijazi K, Anthoine G, et al. Immune evasion before
tumour invasion in early lung squamous carcinogenesis. Nature. 2019; 571(7766):570–5. https://doi.
org/10.1038/s41586-019-1330-0 PMID: 31243362
33. Hoch M, Fischer P, Stapel B, Missol-Kolka E, Sekkali B, Scherr M, et al. Erythropoietin preserves the
endothelial differentiation capacity of cardiac progenitor cells and reduces heart failure during antican-
cer therapies. Cell Stem Cell. 2011; 9(2):131–43. https://doi.org/10.1016/j.stem.2011.07.001 PMID:
21816364
34. Khailany RA, Safdar M, Ozaslan M. Genomic characterization of a novel SARS-CoV-2. Gene Rep.
2020; 19:100682. https://doi.org/10.1016/j.genrep.2020.100682 PMID: 32300673
35. Hamalainen M, Nieminen R, Asmawi MZ, Vuorela P, Vapaatalo H, Moilanen E. Effects of flavonoids on
prostaglandin E2 production and on COX-2 and mPGES-1 expressions in activated macrophages.
Planta Med. 2011; 77(13):1504–11. https://doi.org/10.1055/s-0030-1270762 PMID: 21341175
36. Wang Y, Wang Q, Bao X, Ding Y, Shentu J, Cui W, et al. Taxifolin prevents beta-amyloid-induced
impairments of synaptic formation and deficits of memory via the inhibition of cytosolic phospholipase
A2/prostaglandin E2 content. Metab Brain Dis. 2018; 33(4):1069–79. https://doi.org/10.1007/s11011-
018-0207-5 PMID: 29542038
PLOS ONE | https://doi.org/10.1371/journal.pone.0255335 August 4, 2021
22 / 24
PLOS ONE Impaired immune response promotes severe COVID-19 disease
37. Nakao S, Ogtata Y, Shimizu E, Yamazaki M, Furuyama S, Sugiya H. Tumor necrosis factor alpha
(TNF-alpha)-induced prostaglandin E2 release is mediated by the activation of cyclooxygenase-2
(COX-2) transcription via NFkappaB in human gingival fibroblasts. Mol Cell Biochem. 2002; 238(1–
2):11–8. https://doi.org/10.1023/a:1019927616000 PMID: 12349897
38. Soleymani Fard S, Jeddi Tehrani M, Ardekani AM. Prostaglandin E2 induces growth inhibition, apopto-
sis and differentiation in T and B cell-derived acute lymphoblastic leukemia cell lines (CCRF-CEM and
Nalm-6). Prostaglandins Leukot Essent Fatty Acids. 2012; 87(1):17–24. https://doi.org/10.1016/j.plefa.
2012.04.012 PMID: 22749740
39. Chen R, Cao Y, Tian Y, Gu Y, Lu H, Zhang S, et al. PGE2 ameliorated viral myocarditis development
and promoted IL-10-producing regulatory B cell expansion via MAPKs/AKT-AP1 axis or AhR signaling.
Cell Immunol. 2020; 347:104025. https://doi.org/10.1016/j.cellimm.2019.104025 PMID: 31837749
40. Karki R, Sharma BR, Tuladhar S, Williams EP, Zalduondo L, Samir P, et al. Synergism of TNF-alpha
and IFN-gamma Triggers Inflammatory Cell Death, Tissue Damage, and Mortality in SARS-CoV-2
Infection and Cytokine Shock Syndromes. Cell. 2021; 184(1):149–68 e17. https://doi.org/10.1016/j.cell.
2020.11.025 PMID: 33278357
41. Davies NG, Klepac P, Liu Y, Prem K, Jit M, group CC-w, et al. Age-dependent effects in the transmis-
sion and control of COVID-19 epidemics. Nat Med. 2020; 26(8):1205–11. https://doi.org/10.1038/
s41591-020-0962-9 PMID: 32546824
42. Delogu A, Schebesta A, Sun Q, Aschenbrenner K, Perlot T, Busslinger M. Gene repression by Pax5 in
B cells is essential for blood cell homeostasis and is reversed in plasma cells. Immunity. 2006; 24
(3):269–81. https://doi.org/10.1016/j.immuni.2006.01.012 PMID: 16546096
43. Schebesta A, McManus S, Salvagiotto G, Delogu A, Busslinger GA, Busslinger M. Transcription factor
Pax5 activates the chromatin of key genes involved in B cell signaling, adhesion, migration, and immune
function. Immunity. 2007; 27(1):49–63. https://doi.org/10.1016/j.immuni.2007.05.019 PMID: 17658281
44. Nutt SL, Tarlinton DM. Germinal center B and follicular helper T cells: siblings, cousins or just good
friends? Nat Immunol. 2011; 12(6):472–7. https://doi.org/10.1038/ni.2019 PMID: 21739669
45. Schultheiss C, Paschold L, Simnica D, Mohme M, Willscher E, von Wenserski L, et al. Next-Generation
Sequencing of T and B Cell Receptor Repertoires from COVID-19 Patients Showed Signatures Associ-
ated with Severity of Disease. Immunity. 2020; 53(2):442–55 e4. https://doi.org/10.1016/j.immuni.2020.
06.024 PMID: 32668194
46. Agard M, Asakrah S, Morici LA. PGE(2) suppression of innate immunity during mucosal bacterial infec-
tion. Front Cell Infect Microbiol. 2013; 3:45. https://doi.org/10.3389/fcimb.2013.00045 PMID: 23971009
47.
48.
49.
50.
51.
Johnson K, Shapiro-Shelef M, Tunyaplin C, Calame K. Regulatory events in early and late B-cell differ-
entiation. Mol Immunol. 2005; 42(7):749–61. https://doi.org/10.1016/j.molimm.2004.06.039 PMID:
15829263
Iwasaki A. What reinfections mean for COVID-19. Lancet Infect Dis. 2020. https://doi.org/10.1016/
S1473-3099(20)30783-0 PMID: 33058796
Tillett RL, Sevinsky JR, Hartley PD, Kerwin H, Crawford N, Gorzalski A, et al. Genomic evidence for
reinfection with SARS-CoV-2: a case study. Lancet Infect Dis. 2020. https://doi.org/10.1016/S1473-
3099(20)30764-7 PMID: 33058797
Legros V, Denolly S, Vogrig M, Boson B, Siret E, Rigaill J, et al. A longitudinal study of SARS-CoV-2-
infected patients reveals a high correlation between neutralizing antibodies and COVID-19 severity.
Cell Mol Immunol. 2021; 18(2):318–27. https://doi.org/10.1038/s41423-020-00588-2 PMID: 33408342
Lancman G, Mascarenhas J, Bar-Natan M. Severe COVID-19 virus reactivation following treatment for
B cell acute lymphoblastic leukemia. J Hematol Oncol. 2020; 13(1):131. https://doi.org/10.1186/
s13045-020-00968-1 PMID: 33008453
52. Sajiki Y, Konnai S, Okagawa T, Nishimori A, Maekawa N, Goto S, et al. Prostaglandin E2-Induced
Immune Exhaustion and Enhancement of Antiviral Effects by Anti-PD-L1 Antibody Combined with
COX-2 Inhibitor in Bovine Leukemia Virus Infection. J Immunol. 2019; 203(5):1313–24. https://doi.org/
10.4049/jimmunol.1900342 PMID: 31366713
53. Andre´ F, Manuel S, Santhosh N, Markus A. L., Martin S. Inhibitors for Novel Coronavirus Protease Iden-
tified by Virtual Screening of 687 Million Compounds 2020.
54. Meune C, Mourad JJ, Bergmann JF, Spaulding C. Interaction between cyclooxygenase and the renin-
angiotensin-aldosterone system: rationale and clinical relevance. J Renin Angiotensin Aldosterone
Syst. 2003; 4(3):149–54. https://doi.org/10.3317/jraas.2003.023 PMID: 14608518
55. Smart L, Fawkes N, Goggin P, Pennick G, Rainsford KD, Charlesworth B, et al. A narrative review of
the potential pharmacological influence and safety of ibuprofen on coronavirus disease 19 (COVID-19),
ACE2, and the immune system: a dichotomy of expectation and reality. Inflammopharmacology. 2020;
28(5):1141–52. https://doi.org/10.1007/s10787-020-00745-z PMID: 32797326
PLOS ONE | https://doi.org/10.1371/journal.pone.0255335 August 4, 2021
23 / 24
PLOS ONE Impaired immune response promotes severe COVID-19 disease
56. Valenzuela R, Pedrosa MA, Garrido-Gil P, Labandeira CM, Navarro G, Franco R, et al. Interactions
between ibuprofen, ACE2, renin-angiotensin system, and spike protein in the lung. Implications for
COVID-19. Clin Transl Med. 2021; 11(4):e371. https://doi.org/10.1002/ctm2.371 PMID: 33931961
57. Baghaki S, Yalcin CE, Baghaki HS, Aydin SY, Daghan B, Yavuz E. COX2 inhibition in the treatment of
COVID-19: Review of literature to propose repositioning of celecoxib for randomized controlled studies.
Int J Infect Dis. 2020; 101:29–32. https://doi.org/10.1016/j.ijid.2020.09.1466 PMID: 33007455
58.
59.
Fajnzylber J, Regan J, Coxen K, Corry H, Wong C, Rosenthal A, et al. SARS-CoV-2 viral load is associ-
ated with increased disease severity and mortality. Nature communications. 2020; 11(1):5493. https://
doi.org/10.1038/s41467-020-19057-5 PMID: 33127906
Tang T, Scambler TE, Smallie T, Cunliffe HE, Ross EA, Rosner DR, et al. Macrophage responses to
lipopolysaccharide are modulated by a feedback loop involving prostaglandin E2, dual specificity phos-
phatase 1 and tristetraprolin. Sci Rep. 2017; 7(1):4350. https://doi.org/10.1038/s41598-017-04100-1
PMID: 28659609
60. Boniface K, Bak-Jensen KS, Li Y, Blumenschein WM, McGeachy MJ, McClanahan TK, et al. Prosta-
glandin E2 regulates Th17 cell differentiation and function through cyclic AMP and EP2/EP4 receptor
signaling. The Journal of experimental medicine. 2009; 206(3):535–48. https://doi.org/10.1084/jem.
20082293 PMID: 19273625
61.
Lee J, Aoki T, Thumkeo D, Siriwach R, Yao C, Narumiya S. T cell-intrinsic prostaglandin E2-EP2/EP4
signaling is critical in pathogenic TH17 cell-driven inflammation. J Allergy Clin Immunol. 2019; 143
(2):631–43. https://doi.org/10.1016/j.jaci.2018.05.036 PMID: 29935220
62. Simkin NJ, Jelinek DF, Lipsky PE. Inhibition of human B cell responsiveness by prostaglandin E2. J
Immunol. 1987; 138(4):1074–81. PMID: 3027169
PLOS ONE | https://doi.org/10.1371/journal.pone.0255335 August 4, 2021
24 / 24
PLOS ONE |
10.1371_journal.pone.0248381 | RESEARCH ARTICLE
THz irradiation inhibits cell division by
affecting actin dynamics
Shota YamazakiID
Yuusuke Yamaguchi3, Chiko Otani1, Yuichi Ogawa4, Masahiko Harata2*,
Hiromichi Hoshina1*
1*, Yuya Ueno2, Ryosuke Hosoki2, Takanori Saito2, Toshitaka Idehara3,
1 Terahertz Sensing and Imaging Research Team, RIKEN Center for Advanced Photonics, Sendai, Miyagi,
Japan, 2 Laboratory of Molecular Biology, Graduate School of Agricultural Science, Tohoku University,
Sendai, Miyagi, Japan, 3 Research Center for Development of Far-Infrared Region, University of Fukui (FIR
UF), Bunkyo, Fukui, Japan, 4 Laboratory of Bio-Sensing Engineering, Graduate School of Agriculture, Kyoto
University, Kitashirakawa-Oiwakecho, Sakyo-ku, Kyoto, Japan
* shota.yamazaki.fc@riken.jp (SY); masahiko.harata.b6@tohoku.ac.jp (MH); hoshina@riken.jp (HH)
Abstract
Biological phenomena induced by terahertz (THz) irradiation are described in recent reports,
but underlying mechanisms, structural and dynamical change of specific molecules are still
unclear. In this paper, we performed time-lapse morphological analysis of human cells and
found that THz irradiation halts cell division at cytokinesis. At the end of cytokinesis, the con-
tractile ring, which consists of filamentous actin (F-actin), needs to disappear; however, it
remained for 1 hour under THz irradiation. Induction of the functional structures of F-actin
was also observed in interphase cells. Similar phenomena were also observed under chemi-
cal treatment (jasplakinolide), indicating that THz irradiation assists actin polymerization.
We previously reported that THz irradiation enhances the polymerization of purified actin in
vitro; our current work shows that it increases cytoplasmic F-actin in vivo. Thus, we identified
one of the key biomechanisms affected by THz waves.
Introduction
The recently developed technology of terahertz (THz) light sources indicate the bloom of
applications in a wide range of fields, such as chemical sensing [1, 2], security imaging motion
sensing [3–6], and telecommunications [7–12]. For example, in the wireless technology "6G"
aiming for practical use in the 2030s, the use of sub-THz electromagnetic waves is being stud-
ied. The use of the “sub-THz” is also being considered for the acquisition of high-precision
position information in radars required for autonomous driving and motion sensors. Over the
next decades, THz light sources will become miniaturized, powerful, cheap, and familiar to
everyday life. To facilitate such practical THz applications, the safety of THz radiation for
human health must be guaranteed [13].
The interaction between THz radiation and biological systems has been previously investi-
gated. Two projects, the European THz-BRIDGE and the International EMF project in the
SCENIHR [14], summarize recent studies about the biological effects of THz radiation. For
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: Yamazaki S, Ueno Y, Hosoki R, Saito T,
Idehara T, Yamaguchi Y, et al. (2021) THz
irradiation inhibits cell division by affecting actin
dynamics. PLoS ONE 16(8): e0248381. https://doi.
org/10.1371/journal.pone.0248381
Editor: Yuval Garini, Technion Israel Institute of
Technology, ISRAEL
Received: February 24, 2021
Accepted: July 17, 2021
Published: August 2, 2021
Copyright: © 2021 Yamazaki et al. This is an open
access article distributed under the terms of the
Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: All relevant data are
within the manuscript and its Supporting
information files.
Funding: This work was supported by Japan
Society for the Promotion of Science (JSPS)
KAKENHI Grant Numbers JP19K15812,
JP20H05378, JP21H02151, JP20K21261,
JP21H04746, the JSPS Core-to- Core Program
(Advanced Research Networks) entitled
“Establishment of international agricultural
immunology research-core for a quantum
improvement in food safety, the RIKEN–AIST Joint
PLOS ONE | https://doi.org/10.1371/journal.pone.0248381 August 2, 2021
1 / 15
PLOS ONE Research Fund (Semi-full research), the
Cooperative Research Program of Research Center
for Development of Far-Infrared Region University
of Fukui (R02FIRDM022A, R03FIRDG018A) and
the Japan-Czech Republic Research Cooperative
Program between JSPS and CAS grant number
JPJSBP120202501. The funders had no role in
study design, data collection and analysis, decision
to publish, or preparation of the manuscript.
Competing interests: The authors have declared
that no competing interests exist.
The effects of THz wave on biomolecules
example, THz irradiation was shown to inhibit cell proliferation and to change the adhesive
properties of the nerve cell membrane [15, 16]. Other studies showed THz-induced DNA
destabilization [17–19], which causes chromosomal aberrations in human lymphocytes [20].
The transcriptional activation of wound-responsive genes in mouse skin [21] and the induc-
tion of DNA damage in an artificial human 3D skin tissue model [22] were also reported as
effects of THz irradiation. However, the mechanisms are still unclear because such phenome-
nological studies cannot reveal the underlying molecular origin in the complex biological
systems.
An important point to consider for THz irradiation experiments is the THz radiation
source itself. The THz power density must not be too high to avoid detrimental thermal effects
on the sample. Many studies have shown the effect of heating on cells, such as tissue damage
[23, 24], heat-induced cellular death [25, 26], and DNA damage [27, 28]. Thus, the THz beam
should not be focused tightly to prevent an increase in the temperature on the sample. Two
studies have shown that millimeter-wave radiation induces specific cellular responses that dif-
fer from direct thermal effects [29, 30]; however, the underlying mechanism and exact targets
are poorly defined. In addition to the effect of heating, the generation of the acoustic waves in
aqueous solution must be considered when using the pulsed THz sources. In our previous
works, we observed that THz pulses generate shockwaves at the surface of liquid water [31].
The generated shockwaves propagate to a depth of several millimeters, and disrupt protein
structures in living cells [32]. To avoid such acoustic effects, the peak power of the THz pulses
should be kept at a sufficiently low level.
In this study, we investigated the “non-thermal” and “non-acoustic” effects of THz irradia-
tion on the morphology of living HeLa cells. The energy of THz was 6 mJ/cm2 with a duration
of 10 ms, giving a peak power less than 0.6 W/cm2, which is eight orders of magnitude smaller
than that in our previous studies [32]. The THz fluence was low enough to keep the tempera-
ture rise less than 0.2 ˚C during irradiation. Morphological observation showed that cell divi-
sion in the cell cycle is arrested at mitosis during THz irradiation. Fluorescence microscopy
revealed that this phenomenon is due to the stabilizing of the contractile ring, which is
required to disappear to complete the cytokinesis—the last step of cell division. We found that
the contractile ring was stabilized because of the enhancement of actin polymerization by THz
irradiation. This work is the first to identify the key molecule and mechanism by which THz
waves affect biological systems in a non-thermal and non-acoustic manner.
Materials and methods
THz source
We used a gyrotron (FU CW GVIB [33]) to generate 0.46-THz waves. We designed an appara-
tus that exposed samples to the radiation, which had a peak power density of 0.6 W/cm2. A
schematic representation of the device is shown in Fig 1A. The THz gyrotron produced
10-ms-long pulses with a 1-Hz repetition rate [34]. The time-averaged intensity is 6 mW/cm2.
As a second source of THz irradiation, we used a compact solid-state device based on an
IMPATT-diode (TeraSense Group Inc), which ensured coherent continuous-wave emission of
THz waves with a frequency of 0.28 THz and output power of 20 mW. THz radiation was out-
putted from the horn antenna (4 mm× 4 mm), and emitted from the bottom of the dish with-
out focusing the beam, and with a power density of 125 mW/cm2.
THz irradiation of HeLa cells
HeLa cells were seeded on 0.15 mm-thick cover glass and cultured in Dulbecco’s Modified
Eagle’s Medium (Gibco) supplemented with 10% fetal bovine serum and antibiotics (penicillin
PLOS ONE | https://doi.org/10.1371/journal.pone.0248381 August 2, 2021
2 / 15
PLOS ONE The effects of THz wave on biomolecules
Fig 1. Effects of THz irradiation on cell morphology. (A) Schematic illustration of the experimental setup. THz
waves with a power density of 0.6 W/cm2, frequency of 0.46 THz, pulse duration of 10 ms, and a repetition rate of 1 Hz
were generated by a gyrotron at FIR-UF. The THz beam passed vertically from the bottom of the dish via an aperture
of 4 mm in the heating stage. As a second source of THz irradiation, we used a IMPATT-diode which ensured
coherent continuous-wave emission of THz waves with a frequency of 0.28 THz and output power of 20 mW. THz
radiation was outputted from the horn antenna with a power density of 125 mW/cm2. HeLa cells were seeded on the
film bottom dish and cultured for 24 hours before the experiments. The culture medium was kept at 37 ˚C by the
heating stage during the experiments. (B) Microscopy images of cells at 0, 30, and 60 minutes. Irradiation was started
PLOS ONE | https://doi.org/10.1371/journal.pone.0248381 August 2, 2021
3 / 15
PLOS ONE The effects of THz wave on biomolecules
at 0 minutes and continued for 60 minutes. The bottom panels show the magnified images of the black squares in the
upper panels. The red arrows indicate a pair of cells with a round shape. The scale bar represents 200 μm (upper panel)
and 20 μm (bottom panel).
https://doi.org/10.1371/journal.pone.0248381.g001
and streptomycin) at 37 ˚C in a 5% CO2 humidified atmosphere. Actin filaments were stained
with SiR-actin by adding probes from a 1 mM dimethyl sulfoxide (DMSO) stock solution to
the growth medium (final concentration: 3 μM) and incubating for 1 hour at 37 ˚C in a 5%
CO2 humidified atmosphere. Note that the actin filaments were stained with SiR-actin in
accordance with our earlier study [32], in which SiR-actin does not affect actin dynamics. The
film dish was set on a heating stage (LINKAM: 10002L) to maintain a culture temperature at
37 ± 1 ˚C. The THz beam passed vertically through a 4-mm hole in the heating stage. During
THz irradiation, fluorescence microscopy images were obtained with a UV light source (Thor-
labs, X-Cite 200DC lamp), dichroic mirror (Thorlabs, DMLP650R), two optical filters (excita-
tion band pass: 625 nm/25 nm; emission long pass: 675 nm), objective lens (Olympus,
LUMFLN60XW; Nikon, N10X-PF), and an sCMOS camera (Thorlabs, CS2100M-USB).
Fig 1A shows a schematic diagram of the experimental setup for THz irradiation. Cells treated
with 10 nM jasplakinolide in DMSO were used as a positive control.
For the quantitative analysis of the cells at cytokinesis, cells were synchronized at the
mitotic phase using 25 μg/ml nocodazole. Cells were cultured at 16 hours after the addition of
nocodazole. Before each experiment, nocodazole was washed out by changing the culture
medium, and cells proceeded to mitosis with or without THz irradiation.
Image analysis was performed using Fiji software. To measure the mean signal intensity in
the membrane compartment, the outline of each cell was selected using the area selection tools
in the software. The mean signal intensity of the signal over the area of the cell was recorded.
The number of cells is shown as n. Statistical significance was calculated using F- and T-tests.
Morphological analysis
To measure the cell area and perimeter, the outlines of cells were selected (in the x–y plane)
using the area selection tools in the Fiji software. The form factors of individual cells were cal-
culated as 4πS/L2, where S is the projected cell area and L is the cell perimeter. This index
reflects the irregularity of the cell shape: a perfectly round cell has a value of one, and a stellate
cell has a value lower than one. Data are presented as the mean ± standard deviation. The
number of cells is shown as n. Statistical significance was calculated using F- and T-tests.
Results
THz irradiation halts cell division of cultured cells
To observe the non-thermal and non-acoustic effects of the THz irradiation, we irradiated liv-
ing cells with a THz beam with relatively low peak power. The sample was irradiated with the
output of the gyrotron (0.46 THz), without focusing the beam and with a peak power density
of 0.6 W/cm2. This radiation power is eight orders of magnitude lower than the power in
which acoustic waves were generated in our previous work [32]. The radiation source was
pulsed with a duty ratio of 1% (10-ms duration, 1-Hz repetition rate) to reduce heating of the
sample. HeLa cells were grown on a film-bottom dish, and the culture medium was kept at 37
˚C by a heating stage during the experiment. THz radiation was emitted from the bottom of
the culture dish for 60 minutes (Fig 1A). The high absorbance of water (160 cm−1 at 21 ˚C,
0.46 THz) limits the penetration depth of the THz waves to about 100 μm. Because the
PLOS ONE | https://doi.org/10.1371/journal.pone.0248381 August 2, 2021
4 / 15
PLOS ONE The effects of THz wave on biomolecules
thickness of the cells is less than 30 μm, THz waves reached all regions of the cell. To evaluate
the effect of THz irradiation, we performed time-lapse microscopy imaging of the HeLa cells
(Fig 1B).
Under THz irradiation, the appearance of a characteristic form of cells, which consists of a
pair of round cells, was frequently observed (Fig 1B, red arrow), and the characteristic cells are
maintained during THz irradiation up to 60 min (Fig 1B, bottom panel (zoomed images)).
The round shape of the cells is a typical morphology of mitotic cells, and the pairing of two
round cells is observed at the last step of mitosis, called cytokinesis (Fig 2A). Cytokinesis is
generally completed within 15 minutes [35]. Therefore, the persistence of the paired round
cells indicates that THz irradiation inhibited the progression of cytokinesis.
For the quantitative evaluation of the arrested cells at cytokinesis, cells were synchronized
at the initial phase of mitosis using 25 μg/ml nocodazole, and released into the culture medium
without nocodazole to proceed with the mitosis. Fig 2B shows the percentage of cells arrested
at cytokinesis. Whereas cytokinetic-arrested cells are not observed under the control condi-
tion, THz irradiation induced cytokinetic arrest at 30 minutes after nocodazole release and the
arrest was further continued (Fig 2B, THz). We also analyzed the effect of heat on the progres-
sion of cytokinesis. The culture medium was kept at 42 ˚C by the heating stage during the pro-
gression of mitosis; however, this did not increase the number of cells arrested at cytokinesis
(Fig 2B, 42 ˚C). Since the temperature rise during THz irradiation was less than 0.2 ˚C (S1 Fig
in S1 File), some other reasons than the temperature increase are supposed for the inhibition
of cytokinesis.
Persistence of the contractile ring during THz irradiation
The dominant regulator of cytokinesis is the contractile ring, which consists of actin filaments
(Fig 2A) [36]. At the start of cytokinesis, a G (globular)- to F (filamentous)-actin transition is
induced to make the contractile ring (polymerization reaction). Then, the opposite transition
of F- to G-actin disassembles the contractile ring to complete cell division (depolymerization
reaction). After THz irradiation, the percentage of cells arrested at cytokinesis significantly
increased in comparison with control cells (Fig 2B), suggesting that THz irradiation affects the
disassembly of the contractile ring.
To observe the behavior of the contractile ring under THz irradiation, we stained actin fila-
ments in living cells with SiR-actin [37], and performed time-lapse imaging under a fluores-
cence microscope. The formation of the contractile ring was observed with and without THz
irradiation (Fig 3, Cytokinesis, red arrow). Without THz irradiation, the contractile ring disap-
peared after 30 minutes, and the two daughter cells separated completely (Fig 3, Control, white
arrows). By contrast, under THz irradiation, the contractile ring remained for at least 30 min-
utes (Fig 3, THz, 30 min later). In cells cultured at 42 ˚C, the contractile ring disappeared, and
cell division was completed within 30 minutes (Fig 3, 42 ˚C, 30 min later). This result suggests
that the depolymerization reaction of actin progresses in a non-thermal manner. Cytokinesis
is generally completed within 15 minutes, and the dynamic turnover of actin filaments to G-
actin is required for its completion [38–41]. Importantly, the chemical induction of actin poly-
merization with jasplakinolide inhibits the completion of cytokinesis by stabilizing the con-
tractile ring [42]. Taken together, these results support the notion that THz irradiation inhibits
the completion of cytokinesis by affecting the actin dynamics.
Effects of THz irradiation on actin filaments inside cells
Actin filaments are relevant to various cellular functions, and their dynamics are tightly regu-
lated. For example, cytoplasmic actin polymerization in the plasma membrane is an essential
PLOS ONE | https://doi.org/10.1371/journal.pone.0248381 August 2, 2021
5 / 15
PLOS ONE The effects of THz wave on biomolecules
Fig 2. THz irradiation halts cytokinesis. (A) Schematic representation of mitotic progression. In the process of mitosis, actin polymerization is
induced to make the contractile ring, which is required for starting the division of the mother cell into two daughter cells. Then, the contractile ring is
squeezed and completes cell division. Cytokinesis is generally completed within 15 minutes. (B) Percentage of cells arrested at cytokinesis. The cell cycle
was synchronized to the mitosis phase with 25 μg/ml nocodazole before each experiment. Nocodazole interferes with the polymerization of
microtubules and arrests the initial step of mitosis. Cells were determined to be arrested at cytokinesis when the contractile ring was retained for more
than 30 minutes after the release from nocodazole. The error bars show the standard deviation of three independent experiments. More than 184 cells
were measured in each experiment.
https://doi.org/10.1371/journal.pone.0248381.g002
and versatile process that defines the cellular shape and confers mobility to cells. To evaluate
the effects of THz irradiation on the actin dynamics observed in living cells, we stained actin
filaments in living HeLa cells with SiR-actin [37], and performed time-lapse imaging with fluo-
rescence microscopy. The fluctuation of the cellular actin filaments can be quantitatively esti-
mated by the fluorescence intensity of SiR-actin, which increases up to 100-fold in the actin
PLOS ONE | https://doi.org/10.1371/journal.pone.0248381 August 2, 2021
6 / 15
PLOS ONE The effects of THz wave on biomolecules
Fig 3. Persistence of the contractile ring during THz irradiation. Live-cell imaging of cells with a contractile ring.
The cellular actin filaments were stained with SiR-actin and observed 30 minutes after generation of the contractile
ring. The culture medium was kept at 37 ˚C by the heating stage during the experiments with and without THz
irradiation. To observe the thermal effects on cytokinesis progression, cells were cultured at 42 ˚C and observed
(bottom panel). The red arrows show the contractile ring and the white arrows show the daughter cells. The white bar
shows a scale of 20 μm.
https://doi.org/10.1371/journal.pone.0248381.g003
PLOS ONE | https://doi.org/10.1371/journal.pone.0248381 August 2, 2021
7 / 15
PLOS ONE The effects of THz wave on biomolecules
filaments. Cells treated with 10 nM jasplakinolide, which induce actin polymerization, were
also analyzed as a positive control.
As Fig 4A shows, most cells stayed adherent during the 60-minute observation period, with
a few cells detaching from the bottom of the dish. In addition, the area of the cells remained
constant for 60 minutes during both THz irradiation and jasplakinolide treatment (S2 Fig in
S1 File), suggesting that abnormal shape changes, such as atrophy and hypertrophy, did not
occur. Fig 4B shows the mean fluorescence intensity of SiR-actin in individual cells at 0, 30,
and 60 minutes. The box plot shows the mean fluorescence intensity of SiR-actin in the cells,
and the error bar represents the standard deviation. The fluorescence intensities of SiR-actin
in the cells were kept constant for 60 minutes in the control experiment (Fig 4B, control),
showing that fluorescence bleaching did not occur during the observation period. After 60
minutes of THz irradiation, the fluorescence intensity of SiR-actin increased, indicating that
actin polymerization was accelerated and the number of filaments increased inside the cells
(Fig 4B, THz). A similar effect was observed for the ‘chemical’ induction of actin filamentation
using jasplakinolide (Fig 4B, Jasplakinolide). These results show that THz irradiation acceler-
ates actin filamentation in living HeLa cells.
To confirm the THz irradiation effect by another type of radiation source, same experiment
was performed by a solid-state semiconductor device (TeraSense: IMPATT diode), which out-
puts continuous-wave at 0.28 THz with a power of 20 mW. THz wave was emitted from the
diagonal horn antenna with a size of 4 mm × 4 mm, attached at the bottom of the film-bottom
dish (Fig 1A). The irradiation power density was about 125 mW/cm2. Fig 4C shows the mean
fluorescence intensity of SiR-actin in the individual cells at 30, 60, 90, and 120 minutes. After
90 minutes of irradiation, the fluorescence intensity of SiR-actin was significantly increased
compared with the control cells (Fig 4C, THz).
The fluorescence intensity under irradiation from the IMPATT diode increased more
slowly than under gyrotron irradiation because of the different parameters of the two light
sources. Specifically, the peak power of the IMPATT diode (125 mW/cm2) was about five
times lower than that of the gyrotron (600 mW/cm2). Moreover, the frequency of the IMPATT
diode (0.28 THz) was much lower than that of the gyrotron (0.46 THz). At present, we do not
know which of these two parameters controls the speed of actin filamentation. We note that
the average energy flux of the IMPATT diode (125 mJ/cm2/s) was higher than that of the gyro-
tron (6 mJ/cm2/s). However, the speed of actin filamentation does not depend on the average
energy flux.
Effects of THz irradiation on actin-including structures in interphase cells
In addition to the formation of the contractile ring in cytokinesis, actin polymerization is
required for forming cellular structures in interphase cells, including stress fibers, lamellipodial
meshworks, and transverse arcs (Fig 5A). Stress fibers exist along the cell membrane and form
the cytoskeleton, which maintains the cell shape. Lamellipodial meshworks are observed at the
leading edge of cells and are required for cell migration. Transverse arcs are generated in the
peripheral regions of the cell membrane and move to the center of the cell [43]; this movement
is generally the initial step of cell migration, and actin polymerization is required for move-
ment. To analyze the effect of THz irradiation on actin polymerization, we analyzed actin-
including structures in living cells using fluorescent microscopy. Note that we did not observe
any change of lamellipodial meshworks in this study. It is known that the production of lamel-
lipodial meshworks induces the reorganization of the cell into an asymmetric shape. To con-
firm the cellular shape transition, we analyzed the form factor, which is close to 1 for a round
shape, and close to 0 for an asymmetric shape [44]. The form factor was the same for the
PLOS ONE | https://doi.org/10.1371/journal.pone.0248381 August 2, 2021
8 / 15
PLOS ONE The effects of THz wave on biomolecules
Fig 4. THz waves enhance actin polymerization in cells. (A) Fluorescence microscopy images of cells stained with
SiR-actin at 0, 30, and 60 minutes. THz irradiation was started at 0 minutes and continued for 60 minutes. As a
positive control, cells were treated with 10 nM jasplakinolide at 0 minutes to induce actin polymerization. The white
bar shows a scale of 200 μm. (B) Mean fluorescence intensity of SiR-actin in individual cells measured from the
fluorescence microscopy images. The box plot shows the mean value relative to 0 minutes. The standard deviations of
PLOS ONE | https://doi.org/10.1371/journal.pone.0248381 August 2, 2021
9 / 15
PLOS ONE The effects of THz wave on biomolecules
three independent experiments are shown. More than 77 cells were measured in each experiment. (C) Irradiation with
THz waves generated by the IMPATT-diode source was started at 0 minutes and continued for 120 minutes. The mean
fluorescence intensity of SiR-actin in individual cells was measured from the fluorescence microscopy images. The
box plot shows the mean value relative to that measured at 30 minutes. The standard deviations of three independent
experiments are shown. More than 120 cells were measured in each experiment.
https://doi.org/10.1371/journal.pone.0248381.g004
control, THz irradiation, and jasplakinolide-treated samples for 60 minutes (S3 Fig in S1 File).
Therefore, we concluded that lamellipodial meshworks were not induced by the 60-minute
THz irradiation.
Fig 5B shows time-lapse images of a single cell stained with SiR-actin at 0, 10, 20, and 30
minutes. The white dotted lines show the position of the cell membrane. The fluorescence
intensity of SiR-actin increased near the cell membrane in the control, indicating that stress
fibers were generated during the measurement (Fig 5B, Control, yellow arrows). Under THz
irradiation, in addition to the stress fiber formation, transverse arcs were formed in the periph-
ery, and this structure moves from the cell membrane towards the center of the cell (Fig 5B,
THz, red arrows) (S1 Movie).
Fig 5C shows the number of cells in which transverse arcs were generated during the
30-minute experiment. 27% of cells contained a transverse arc in the control experiment (Fig
5C, Cont). By contrast, over 60% of the cells contained a transverse arc as a result of either
THz irradiation or jasplakinolide treatment (Fig 5C, THz and Jasp). These results suggest that
THz irradiation affects actin polymerization not only in the contractile ring but also in the
cytoplasm of interphase cells.
Discussion
In our previous study, we subjected an aqueous solution of purified actin protein to THz irra-
diation for the purpose of developing a physical technique for macromolecular manipulation
[34]. In that study, we found that actin filaments were generated effectively under THz
Fig 5. Effect of THz irradiation on actin-including structures. (A) Illustration of the functional structures that include actin filaments inside cells. In
the cytoplasm, actin filaments form massive assemblies, which can be categorized as stress fibers, lamellipodial meshworks, and transverse arcs. Stress
fibers are static structures that exist along the cell membrane; the lamellipodial meshwork is observed in the leading edge of the cell; and transverse arcs
are generated in the cell membrane and move to the center of the cell. (B) Live-cell imaging of actin filaments with and without THz irradiation. The
white dotted line marks the cell membrane. The yellow arrow shows stress fibers, which appeared along the cell membrane. The red arrow shows a
transverse arc, which was generated in the cell membrane and moved to the center of the cell for 20–30 minutes. The scale bar represents 10 μm. (C)
Percentage of cells, in which transverse arcs appeared during microscopy observation for 30 minutes. As a positive control, cells were treated with 10 nM
jasplakinolide at 0 minutes to induce actin polymerization. The error bar shows the standard deviation of three independent experiments. More than
184 cells were measured in each experiment.
https://doi.org/10.1371/journal.pone.0248381.g005
PLOS ONE | https://doi.org/10.1371/journal.pone.0248381 August 2, 2021
10 / 15
PLOS ONE The effects of THz wave on biomolecules
irradiation in living cells. Furthermore, THz irradiation caused the generation and retention
of massive assemblies of actin filaments, such as contractile rings and transverse arcs (Figs 3
and 5).
Because the formation of biological molecules is sensitive to temperature, the simplest
explanation for the enhancement of actin polymerization might be a transient increase of tem-
perature owing to the absorption of THz irradiation by water. However, it has been demon-
strated that the effect of a temperature rise on actin polymerization is negligible (Fig 2B) [34,
45]. In addition, we estimated the temperature change during THz irradiation as 0.23 ˚C using
an adiabatic model (S1 Fig in S1 File). Therefore, it is unlikely that a temperature change due
to THz irradiation enhances actin polymerization in living cells, and other mechanisms should
be considered.
Another possible explanation is THz-induced shockwaves. In our previous study, we found
that shockwaves were generated by THz pulses of 80 μJ/cm2 with a duration of 5 ps (peak
power of 16 MW/cm2) [32]. Intense THz pulses are absorbed at the water surface and the
energy concentration results in shockwave generation. The shockwaves propagate for a few
millimeters in the aqueous medium, and disrupt the morphology of actin filaments in living
cells. However, in the present study, the energy of each THz pulse was 6 mJ/cm2 with a dura-
tion of 10 ms, giving a peak power of just 0.6 W/cm2, which is eight orders of magnitude
smaller than that used in Ref. 32, which generated shockwaves. Therefore, we consider that
THz irradiation did not induce shockwaves under the experimental conditions of the present
study.
We attribute the observed phenomena to non-thermal and non-acoustic effects of THz irra-
diation (i.e. the direct interaction between THz photons and the dynamical motion of the
actin proteins). Because the vibration frequencies of the higher-order conformations of pro-
teins and the surrounding water molecules are in the THz band [46–48], THz irradiation per-
turbates the intra- and inter-molecular dynamics of the actin proteins. The actin
polymerization process consists of three phases: nucleation, elongation, and equilibrium. In
our previous study, we found that THz irradiation enhances actin polymerization reaction in
the aqueous solution [34]. We concluded that THz irradiation accelerates the elongation pro-
cess because the actin filaments undergo additional elongation under THz irradiation in the
equilibrium state. Those results showed that THz irradiation affects the dynamics of actin mol-
ecules during the elongation reaction.
Our previous in vitro THz irradiation experiment for the same molecule helps us under-
stand the mechanism of in vivo THz irradiation. The observed phenomena—the inhibition of
cytokinesis and formation of transverse arcs—suggest the enhancement of actin filamentation
in living cells, which we also quantitatively confirmed from the fluorescence intensity of SiR-
actin. In the in vitro experiment, such enhancement of actin filamentation was not due to the
expression of the intra-cellular system, such as activation of cell signaling, changes of tran-
scriptional regulations, and induction of cellular responses, but was due to the direct enhance-
ment of the elongation reaction of the actin filament. Using actin molecules, we succeeded in
elucidating the effects of THz irradiation on molecular reactions and cellular expression.
Actin filament is a major component of the cytoskeleton, and has crucial roles in determin-
ing cell shape, and for cell motility and division [49, 50]. Moreover, the recent development of
fluorescence probes has led to the revelation that nuclear actin filaments are required for tran-
scriptional regulation, DNA repair, and gene reprogramming [51–53]. Therefore, THz irradia-
tion has potential as a novel biological tool. In fact, we discovered that the effect of THz
irradiation is similar to that of jasplakinolide treatment. Jasplakinolide, a naturally occurring
cyclic peptide from the marine sponge Jaspis sp [54], is a membrane-permeable, actin-poly-
merizing, and filament-stabilizing drug [55]. Jasplakinolide has a wide range of known
PLOS ONE | https://doi.org/10.1371/journal.pone.0248381 August 2, 2021
11 / 15
PLOS ONE The effects of THz wave on biomolecules
biological functions, which include antifungal and antitumor activities [56, 57]. Thus, by anal-
ogy with jasplakinolide, we suggest that THz irradiation can be used to manipulate cell func-
tions via actin polymerization. In this study, we also demonstrated that the actin filamentation
is induced by an IMPATT diode source. The IMPATT diode is small, operated at room tem-
perature, and works with lower electrical power. Such solid-state semiconductor THz-sources
are widely available for experiments with biological samples.
Conclusions
We found that THz irradiation enhances the formation and stabilization of actin assemblies in
living cells. Therefore, we propose that THz irradiation can be used for the optical manipula-
tion of cellular functions via the modulation of actin dynamics, leading to a better understand-
ing of the function of actin.
Supporting information
S1 File.
(DOCX)
S1 Movie. Live-cell imaging of actin filaments with THz irradiation. Actin filaments were
stained with SiR-actin and observed by fluorescence microscopy.
(AVI)
Acknowledgments
We thank Adam Brotchie, PhD, from Edanz Group (https://en-author-services.edanz.com/ac)
for editing a draft of this manuscript and helping to draft the abstract.
Author Contributions
Conceptualization: Shota Yamazaki.
Data curation: Shota Yamazaki, Yuya Ueno, Hiromichi Hoshina.
Formal analysis: Shota Yamazaki, Yuya Ueno, Ryosuke Hosoki, Takanori Saito, Masahiko
Harata, Hiromichi Hoshina.
Funding acquisition: Shota Yamazaki.
Investigation: Shota Yamazaki, Yuusuke Yamaguchi.
Methodology: Shota Yamazaki, Yuusuke Yamaguchi, Yuichi Ogawa.
Project administration: Shota Yamazaki.
Resources: Yuusuke Yamaguchi.
Supervision: Toshitaka Idehara, Chiko Otani, Yuichi Ogawa, Masahiko Harata, Hiromichi
Hoshina.
Visualization: Shota Yamazaki.
Writing – original draft: Shota Yamazaki, Masahiko Harata, Hiromichi Hoshina.
Writing – review & editing: Shota Yamazaki, Masahiko Harata, Hiromichi Hoshina.
References
1. Peiponen K-E, Zeitler A, Kuwata-Gonokami M. Terahertz spectroscopy and imaging. Springer; 2012.
PLOS ONE | https://doi.org/10.1371/journal.pone.0248381 August 2, 2021
12 / 15
PLOS ONE The effects of THz wave on biomolecules
2. UENO Y, AJITO K. Analytical Terahertz Spectroscopy. Anal Sci. 2008; 24: 185–192. https://doi.org/10.
2116/analsci.24.185 PMID: 18270407
3. Mittleman DM. Perspective: Terahertz science and technology. J Appl Phys. 2017; 122: 230901.
https://doi.org/10.1063/1.5007683
4. Dean P, Valavanis A, Keeley J, Bertling K, Lim YL, Alhathlool R, et al. Terahertz imaging using quantum
cascade lasers—a review of systems and applications. J Phys D Appl Phys. 2014; 47: 374008. https://
doi.org/10.1088/0022-3727/47/37/374008
5. Guillet JP, Recur B, Frederique L, Bousquet B, Canioni L, Manek-Ho¨ nninger I, et al. Review of terahertz
tomography techniques. J Infrared, Millimeter, Terahertz Waves. 2014; 35: 382–411. https://doi.org/10.
1007/s10762-014-0057-0
6.
Lien J, Gillian N, Karagozler ME, Amihood P, Schwesig C, Olson E, et al. Soli: Ubiquitous gesture sens-
ing with millimeter wave radar. ACM Trans Graph. 2016; 35: 1–19. https://doi.org/10.1145/2897824.
2925953
7. Giordani M, Polese M, Mezzavilla M, Rangan S, Zorzi M. Toward 6G Networks: Use Cases and Tech-
nologies. IEEE Commun Mag. 2020; 58: 55–61. https://doi.org/10.1109/MCOM.001.1900411
8. Mourad A, Yang R, Lehne PH, De La Oliva A. A baseline roadmap for advanced wireless research
beyond 5G. Electron. 2020; 9: 1–14. https://doi.org/10.3390/electronics9020351
9. Saad W, Bennis M, Chen M. A Vision of 6G Wireless Systems: Applications, Trends, Technologies,
and Open Research Problems. IEEE Netw. 2020; 34: 134–142. https://doi.org/10.1109/MNET.001.
1900287
10. Nagatsuma T. Advances in Terahertz Communications Accelerated by Photonics Technologies.
OECC/PSC 2019—24th Optoelectron Commun Conf Conf Photonics Switch Comput 2019. 2019; 1–3.
11. Koenig S, Lopez-Diaz D, Antes J, Boes F, Henneberger R, Leuther A, et al. Wireless sub-THz commu-
nication system with high data rate. Nat Photonics. 2013; 7: 977–981. https://doi.org/10.1038/nphoton.
2013.275
12. O’Hara JF, Ekin S, Choi W, Song I. A Perspective on Terahertz Next-Generation Wireless Communica-
tions. Technologies. 2019; 7: 43. https://doi.org/10.3390/technologies7020043
13.
14.
Jamshed MA, Heliot F, Brown TWC. A Survey on Electromagnetic Risk Assessment and Evaluation
Mechanism for Future Wireless Communication Systems. IEEE J Electromagn RF Microwaves Med
Biol. 2020; 4: 24–36. https://doi.org/10.1109/JERM.2019.2917766
Leitgeb N, Auvinen A, Danker-hopfe H, Mild KH. SCENIHR (Scientific Committee on Emerging and
Newly Identified Health Risks), Potential health effects of exposure to electromagnetic fields (EMF), Sci-
entific Committee on Emerging and Newly Identified Health Risks SCENIHR Opinion on Potential
health. 2016.
15. Munzarova AF, Kozlov AS, Zelentsov EL. Effect of terahertz laser irradiation on red blood cells aggrega-
tion in healthy blood. Vestn NSU Phys Ser. 2013; 8: 117–123.
16. Olshevskaya JS, Kozlov AS, Petrov AK, Zapara TA, Ratushnyak AS. Cell membrane permeability under
the influence of terahertz (submillimeter) laser radiation. Vestn Novosib State Univ. 2010; 5: 177–181.
17. Bock J, Fukuyo Y, Kang S, Lisa Phipps M, Alexandrov LB, Rasmussen KO, et al. Mammalian stem
cells reprogramming in response to terahertz radiation. PLoS One. 2010; 5: 8–13. https://doi.org/10.
1371/journal.pone.0015806 PMID: 21209821
18. Alexandrov BS, Rasmussen KØ, Bishop AR, Usheva A, Alexandrov LB, Chong S, et al. Non-thermal
effects of terahertz radiation on gene expression in mouse stem cells. Biomed Opt Express. 2011; 2:
2679. https://doi.org/10.1364/BOE.2.002679 PMID: 21991556
19. Alexandrov BS, Lisa Phipps M, Alexandrov LB, Booshehri LG, Erat A, Zabolotny J, et al. Specificity and
Heterogeneity of Terahertz Radiation Effect on Gene Expression in Mouse Mesenchymal Stem Cells.
Sci Rep. 2013; 3: 1–8. https://doi.org/10.1038/srep01184 PMID: 23378916
20. Korenstein-Ilan A, Barbul A, Hasin P, Eliran A, Gover A, Korenstein R. Terahertz Radiation Increases
Genomic Instability in Human Lymphocytes. Radiat Res. 2008; 170: 224–234. https://doi.org/10.1667/
RR0944.1 PMID: 18666810
21. Kim KT, Park J, Jo SJ, Jung S, Kwon OS, Gallerano GP, et al. High-power femtosecond-terahertz pulse
induces a wound response in mouse skin. Sci Rep. 2013; 3: 1–7. https://doi.org/10.1038/srep02296
PMID: 23907528
22.
Titova L V., Ayesheshim AK, Golubov A, Fogen D, Rodriguez-Juarez R, Hegmann FA, et al. Intense
THz pulses cause H2AX phosphorylation and activate DNA damage response in human skin tissue.
Biomed Opt Express. 2013; 4: 559. https://doi.org/10.1364/BOE.4.000559 PMID: 23577291
23. Henriques FC, Moritz AR. Studies of Thermal Injury: I. The Conduction of Heat to and through Skin and
the Temperatures Attained Therein. A Theoretical and an Experimental Investigation. Am J Pathol.
1947; 23: 530–49. http://www.ncbi.nlm.nih.gov/pubmed/19970945
PLOS ONE | https://doi.org/10.1371/journal.pone.0248381 August 2, 2021
13 / 15
PLOS ONE The effects of THz wave on biomolecules
24. Pearce JA. Models for Thermal Damage in Tissues: Processes and Applications. Crit Rev Biomed Eng.
2010; 38: 1–20. https://doi.org/10.1615/critrevbiomedeng.v38.i1.20 PMID: 21175400
25. Boreham DR, Mitchel REJM. Heat-induced thermal tolerance and radiation resistance to apoptosis in
human lymphocytes. Biochem Cell Biol. 1997; 75: 393–397. https://doi.org/10.1139/o97-077 PMID:
9493962
26. Wilmink GJ, Opalenik SR, Beckham JT, Abraham AA, Nanney LB, Mahadevan-Jansen A, et al. Molecu-
lar Imaging-Assisted Optimization of Hsp70 Expression during Laser-Induced Thermal Preconditioning
for Wound Repair Enhancement. J Invest Dermatol. 2009; 129: 205–216. https://doi.org/10.1038/jid.
2008.175 PMID: 18580963
27. MORI E, TAKAHASHI A, OHNISHI T. The Biology of Heat-induced DNA Double-Strand Breaks. Therm
Med. 2008; 24: 39–50. https://doi.org/10.3191/thermalmed.24.39
28. Roti Roti JL, Pandita RK, Mueller JD, Novak P, Moros EG, Laszlo A. Severe, short-duration (0–3 min)
heat shocks (50–52˚C) inhibit the repair of DNA damage. Int J Hyperth. 2010; 26: 67–78. https://doi.org/
10.3109/02656730903417947 PMID: 20100054
29. Habauzit D, Le Que´ment C, Zhadobov M, Martin C, Aubry M, Sauleau R, et al. Transcriptome Analysis
Reveals the Contribution of Thermal and the Specific Effects in Cellular Response to Millimeter Wave
Exposure. Scarfi MR, editor. PLoS One. 2014; 9: e109435. https://doi.org/10.1371/journal.pone.
0109435 PMID: 25302706
30. Romanenko S, Harvey AR, Hool L, Fan S, Wallace VP. Millimeter Wave Radiation Activates Leech
Nociceptors via TRPV1-Like Receptor Sensitization. Biophys J. 2019; 116: 2331–2345. https://doi.org/
10.1016/j.bpj.2019.04.021 PMID: 31103236
31.
Tsubouchi M, Hoshina H, Nagai M, Isoyama G. Plane photoacoustic wave generation in liquid water
using irradiation of terahertz pulses. Sci Rep. 2020; 10: 18537. https://doi.org/10.1038/s41598-020-
75337-6 PMID: 33116209
32. Yamazaki S, Harata M, Ueno Y, Tsubouchi M, Konagaya K, Ogawa Y, et al. Propagation of THz irradia-
tion energy through aqueous layers: Demolition of actin filaments in living cells. Sci Rep. 2020; 10: 1–
10.
33.
Idehara T, Tatematsu Y, Yamaguchi Y, Khutoryan EM, Kuleshov AN, Ueda K, et al. The development
of 460 GHz gyrotrons for 700 MHz DNP-NMR spectroscopy. J Infrared, Millimeter, Terahertz Waves.
2015; 36: 613–627.
34. Yamazaki S, Harata M, Idehara T, Konagaya K, Yokoyama G, Hoshina H, et al. Actin polymerization is
activated by terahertz irradiation. Sci Rep. 2018; 8. https://doi.org/10.1038/s41598-018-28245-9 PMID:
29968762
35. Milo R, Phillips R. Cell Biology by the Numbers. Garland Science; 2015. https://doi.org/10.1201/
9780429258770
36. Bezanilla M, Gladfelter AS, Kovar DR, Lee WL. Cytoskeletal dynamics: A view from the membrane. J
Cell Biol. 2015; 209: 329–337. https://doi.org/10.1083/jcb.201502062 PMID: 25963816
37. D’Este E, Hell SW, Waldmann H, Go¨ ttfert F, Gerlich DW, Masharina A, et al. Fluorogenic probes for
live-cell imaging of the cytoskeleton. Nat Methods. 2014; 11: 731–733. https://doi.org/10.1038/nmeth.
2972 PMID: 24859753
38. Pelham RJ, Chang F. Actin dynamics in the contractile ring during cytokinesis in fission yeast. Nature.
2002; 419: 82–86. https://doi.org/10.1038/nature00999 PMID: 12214236
39. Group CB. Single Particle Tracking of Surface Receptor. 1994; 127: 963–971.
40. O’Connell CB, Warner AK, Wang Y li. Distinct roles of the equatorial and polar cortices in the cleavage
of adherent cells. Curr Biol. 2001; 11: 702–707. https://doi.org/10.1016/s0960-9822(01)00181-6 PMID:
11369234
41. Murthy K, Wadsworth P. Myosin-II-Dependent Localization and Dynamics of F-Actin during Cytokine-
sis. 2005; 15: 724–731. https://doi.org/10.1016/j.cub.2005.02.055 PMID: 15854904
42. Mendes Pinto I, Rubinstein B, Kucharavy A, Unruh JR, Li R. Actin Depolymerization Drives Actomyosin
Ring Contraction during Budding Yeast Cytokinesis. Dev Cell. 2012; 22: 1247–1260. https://doi.org/10.
1016/j.devcel.2012.04.015 PMID: 22698284
43. Vallenius T. Actin stress fibre subtypes in mesenchymal-migrating cells. Open Biol. 2013; 3. https://doi.
org/10.1098/rsob.130001 PMID: 23782578
44. Altankov G, Grinneu F. Depletion of Intracellular Potassium Disrupts Coated Pits and Reversibly Inhibits
Cell Polarization During Fibroblast Spreading. 1993; 120. https://doi.org/10.1083/jcb.120.6.1449 PMID:
8449988
45. KAWAMURA M, MARUYAMA K. A Further Study of Electron Microscopic Particle Length of F-Actin
Polymerized in Vitro. J Biochem. 1972; 72: 179–188. https://doi.org/10.1093/oxfordjournals.jbchem.
a129884 PMID: 4560743
PLOS ONE | https://doi.org/10.1371/journal.pone.0248381 August 2, 2021
14 / 15
PLOS ONE The effects of THz wave on biomolecules
46. Yamamoto N, Ohta K, Tamura A, Tominaga K. Broadband Dielectric Spectroscopy on Lysozyme in the
Sub-Gigahertz to Terahertz Frequency Regions: Effects of Hydration and Thermal Excitation. J Phys
Chem B. 2016; 120: 4743–4755. https://doi.org/10.1021/acs.jpcb.6b01491 PMID: 27158918
47. Xu Y, Havenith M. Perspective: Watching low-frequency vibrations of water in biomolecular recognition
by THz spectroscopy. J Chem Phys. 2015; 143: 170901. https://doi.org/10.1063/1.4934504 PMID:
26547148
48. Conti Nibali V, Havenith M. New Insights into the Role of Water in Biological Function: Studying Sol-
vated Biomolecules Using Terahertz Absorption Spectroscopy in Conjunction with Molecular Dynamics
Simulations. J Am Chem Soc. 2014; 136: 12800–12807. https://doi.org/10.1021/ja504441h PMID:
25127002
49. Pollard TD, Cooper JA. Actin, a central player in cell shape and movement. Science. 2009; 326: 1208–
1212. https://doi.org/10.1126/science.1175862 PMID: 19965462
50. Small JV, Rottner K, Kaverina I, Anderson KI. Assembling an actin cytoskeleton for cell attachment and
movement. Biochim Biophys Acta—Mol Cell Res. 1998; 1404: 271–281. https://doi.org/10.1016/s0167-
4889(98)00080-9 PMID: 9739149
51. Yamazaki S, Yamamoto K, de Lanerolle P, Harata M. Nuclear F-actin enhances the transcriptional
activity of β-catenin by increasing its nuclear localization and binding to chromatin. Histochem Cell Biol.
2016; 145: 389–399. https://doi.org/10.1007/s00418-016-1416-9 PMID: 26900020
52. Caridi CP, Plessner M, Grosse R, Chiolo I. Nuclear actin filaments in DNA repair dynamics. Nat Cell
Biol. 2019; 21: 1068–1077. https://doi.org/10.1038/s41556-019-0379-1 PMID: 31481797
53. Miyamoto K, Pasque V, Jullien J, Gurdon JB. Nuclear actin polymerization is required for transcriptional
reprogramming of Oct4 by oocytes. Genes Dev. 2011; 25: 946–958. https://doi.org/10.1101/gad.
615211 PMID: 21536734
54. White KN, Tenney K, Crews P. The Bengamides: A Mini-Review of Natural Sources, Analogues, Bio-
logical Properties, Biosynthetic Origins, and Future Prospects. J Nat Prod. 2017; 80: 740–755. https://
doi.org/10.1021/acs.jnatprod.6b00970 PMID: 28185457
55. Bubb MR, Spector I, Beyer BB, Fosen KM. Effects of jasplakinolide on the kinetics of actin polymeriza-
tion. An explanation for certain in vivo observations. J Biol Chem. 2000; 275: 5163–5170. https://doi.
org/10.1074/jbc.275.7.5163 PMID: 10671562
56. Senderowicz AM, Kaur G, Sainz E, Laing C, Inman WD, Rodrı´guez J, et al. Jasplakinolide’s inhibition of
the growth of prostate carcinoma cells in vitro with disruption of the actin cytoskeleton. J Natl Cancer
Inst. 1995; 87: 46–51. https://doi.org/10.1093/jnci/87.1.46 PMID: 7666463
57. Eisemann T, Costa B, Strelau J, Mittelbronn M, Angel P, Peterziel H. An advanced glioma cell invasion
assay based on organotypic brain slice cultures. BMC Cancer. 2018; 18: 1–10.
PLOS ONE | https://doi.org/10.1371/journal.pone.0248381 August 2, 2021
15 / 15
PLOS ONE |
10.1371_journal.pone.0255193 | RESEARCH ARTICLE
Factors affecting the choice of delivery place
in a rural area in Laos: A qualitative analysis
Yoshiko Kawaguchi1☯, Ahmad M. Sayed2,3☯, Alliya Shafi3,4, Sengchanh Kounnavong5,
Tiengkham Pongvongsa6, Angkhana Lasaphonh5, Khamsamay Xaylovong1, Miho Sato1,
Mitsuaki Matsui1, Atsuko Imoto1, Nguyen Tien HuyID
1,3*, Kazuhiko Moji1*
1 School of Tropical Medicine and Global Health, Nagasaki University, Nagasaki, Japan, 2 Department of
Organic Chemistry, College of Pharmacy, Al-Azhar University, Cairo, Egypt, 3 Online Research Club (https://
www.onlineresearchclub.org/), Nagasaki, Japan, 4 School of Medicine, American University of Caribbean,
Sint Maarten, 5 Laos Tropical and Pubic Health Institute, Vientiane, Lao PDR, 6 Savannakhet Provincial
Health Department, Savannnakhet, Lao PDR
☯ These authors contributed equally to this work.
* tienhuy@nagasaki-u.ac.jp (NTH); moji-k@nagasaki-u.ac.jp (KM)
Abstract
Background
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: Kawaguchi Y, Sayed AM, Shafi A,
Kounnavong S, Pongvongsa T, Lasaphonh A, et al.
(2021) Factors affecting the choice of delivery
place in a rural area in Laos: A qualitative analysis.
PLoS ONE 16(8): e0255193. https://doi.org/
10.1371/journal.pone.0255193
Editor: Shah Md Atiqul Haq, Shahjalal University of
Science and Technology, BANGLADESH
Received: April 7, 2021
Accepted: July 11, 2021
Published: August 2, 2021
Copyright: © 2021 Kawaguchi et al. This is an open
access article distributed under the terms of the
Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: All relevant data are
within the manuscript and its Supporting
Information files.
Funding: The authors received no specific funding
for this work.
Competing interests: The authors have declared
that no competing interests exist.
Home delivery (HD) without skilled birth attendants (SBAs) are considered crucial risk fac-
tors increasing maternal and child mortality rates in Loa PDR. While a few studies in the liter-
ature discuss the choice of delivery in remote areas of minority ethnic groups; our work aims
to identify factors that indicated their delivery place, at home or in the health facilities.
Methods
A community-based qualitative study was conducted between February and March 2020.
Three types of interviews were implemented, In-depth interviews with 16 women of eight
rural villages who delivered in the last 12 months in Xepon District, Savannakhet Province,
Lao PDR. Also, three focus group discussions (FGDs) with nine HCPs and key-informant
interviews of ten VHVs were managed. Factors affecting the choice of the delivery place
were categorized according to the social-ecological model.
Results
Our sample included five Tri women and two Mangkong women in the HD group, while the
FD group included three Tri women, two Mangkong women, one Phoutai woman, two Lao-
lung women and one Vietnamese. Our investigation inside the targeted minority showed
that both positive perceptions of home delivery (HD) and low-risk perception minorities were
the main reasons for the choice of HD, on the individual level. On the other hand, fear of
complication, the experience of stillbirth, and prolonged labour pain during HD were reasons
for facility-based delivery (FD). Notably, the women in our minority reported no link between
their preference and their language, while the HCPs dated the low knowledge to the lan-
guage barrier. On the interpersonal level, the FD women had better communication with
their families, and better preparation for delivery compared to the HD group. The FD family
PLOS ONE | https://doi.org/10.1371/journal.pone.0255193 August 2, 2021
1 / 16
PLOS ONE Choice of delivery place in a rural area in Laos
prepared cash and transportation using their social network. At the community level, the
trend of the delivery place had shifted from HD to FD. Improved accessibility and increased
knowledge through community health education were the factors of the trend. At the societal
(national policy) level, the free delivery policy and limitation of HCPs’ assisted childbirth only
in health facilities were the factors of increasing FD, while the absence of other incentives
like transportation and food allowance was the factor of remaining of HD.
Conclusions
Based on the main findings of this study, we urge the enhancement of family communication
on birth preparedness and birthplace. Furthermore, our findings support the need to educate
mothers, especially those of younger ages, about their best options regarding the place of
delivery. We propose implementing secondary services of HD to minimize the emergency
risks of HD. We encourage local authorities to be aware of the medical needs of the commu-
nity especially those of pregnant females and their right for a free delivery policy.
Background
Despite the progress made over recent decades to upgrade maternal health around the globe;
yet maternal mortalities are still deemed a considerable portion of deaths [1,2]. The majority
of these mortalities occur in developing countries that lack antenatal care systems [3]. While
giving birth in healthcare institutions lowers delivery mortalities, the utilization of these facili-
ties did not supersede. Hence, the reduction in mortalities not only depend on the spread of
delivery institutions but also depends on other factors that lower the usage of those facilities.
Those factors should be unsealed for better maternal healthcare.
Lao People’s Democratic Republic (Lao PDR) is a Southeast Asian country that has 49 offi-
cial ethnic groups [4]. About two-thirds of the population belongs to Lao-Tai ethnolinguistic
family whose first language is the Lao language, and the rest of them are from other families
[5]. The proportion of ethnic minorities is higher in a remote area; for instance, in the Xepon
district of Savannakhet Province where over 83% of people are not from the Lao-Tai family
[6]. Regarding maternal health, Lao PDR still has a high maternal mortality rate (MMR) of
197/100,000 live birth, while the lifetime risk of maternal death is 1 in 150 in 2015 [7]. On the
other hand, the home delivery (HD) rate is still high in Lao PDR; with 34.5% in national-wide,
and 38.4% in Savannakhet Province in 2017 [6].
Individuals residing in remote areas are less likely to deliver at health care facilities. Kasaye
et al. highlight that the absence of a health facility within a 30-minute distance of walking sig-
nificantly increased HD. Additionally, financial status also affects the choice of delivery loca-
tion [8]. Meanwhile, the government of Lao PDR launched free delivery and child care policy
where pregnant women, women giving childbirth, postpartum women (42 days after delivery)
and children under five years old can receive medical and non-medical benefits (transport
cost, transfer cost and food allowance) [9]. In Lao PDR, unlike other regions where recent
studies found a relation between the education level and the HD [10], the majority of the pop-
ulation are low-educated which masked any potential effect [11].
Starting from the childbirth experience, the perception, easiness, convenience, nearness of
health facilities and the financial cost are considered causes of lack of institutional usage. For
instance, in Xepon; the people’s perception of childbirth was negatively skewed to the facility-
PLOS ONE | https://doi.org/10.1371/journal.pone.0255193 August 2, 2021
2 / 16
PLOS ONE Choice of delivery place in a rural area in Laos
based delivery (FD) [12]. Smith et al. have declared that previous experience of complications
and prolonged labour played a vital role in swaying the preference for FD [13]. Fear of compli-
cation and trust in the quality of the delivery process constituted other reasons [14]. Besides,
the experience with the healthcare facilities, lack of privacy, improper behaviours of HCPs
reversed the pointer to the HD [14]. In a study, the researchers reported that Lao PDR has a
poor quality of care, bad attitude of HCPs, and lack of privacy, which affected the preference of
the FD [15]. The ministry of health in Laos plans to increase the number of midwives and
train them to improve the quality of care. Currently, trained midwives increased from 88 to
1784 in six years from 2010 to 2015. Yet, two-thirds of health centres (HCs) did not have a
midwife in 2014 [9].
In the present study, we aimed to illustrate the factors that determined the choice of delivery
place between HD and FD. We investigated factors like women who gave birth, the health care
providers (HCPs), village health volunteers (VHVs and their role on the delivery-place deter-
mination in the targeted community. Our findings hail the limitations of the strategies devel-
oped inside Lao PDR to upgrade the current policies for the improvement of delivery services.
Methodology
Study area
This study was conducted in villages of Xepon district which is in the northeast of Savannakhet
Province, central of Lao PDR. Our inclusion criteria were to choose a village that: (1) their
dwellers still using the HD. (2) to be about 5–20 km away from the closest health facility,
Xepon District Hospital, and the healthcare centres. (3) to be covered in the Xepon HDSS.
According to the pre-mentioned criteria and situation of COVID-19 pandemic; we have
selected Vangmorthoum, Puong, La-or-kao, Keangthong nok, Kalouk kao-mai, Keangthong
nok, Alai noy, Kaengluang nok and Kadpu. The total population of the eight villages is 2,236
in 2019 according to the Xepon district health office (DHO) (Fig 1).
Study design and population
This community-based qualitative study was implemented in Xepon district, Savannakhet
province between February and March 2020. The selection of women to volunteer in this
study was performed with the support of two local research assistants from the Xepon District
Hospital. The inclusion criteria for choosing our participants were to choose only women
Fig 1. Map of the study site (Xepon district) with village names.
https://doi.org/10.1371/journal.pone.0255193.g001
PLOS ONE | https://doi.org/10.1371/journal.pone.0255193 August 2, 2021
3 / 16
PLOS ONE Choice of delivery place in a rural area in Laos
who: (1) had a child within 12 months, (2) a dweller of one of the eight villages, and (3) gave
consent to participate. A total of 16 women who delivered a child within a year in the eight vil-
lages were selected with the support of Xepon District Hospital and the HCs. The interviews
were conducted in each village. From each village, one woman who did HD and one woman
who did FD were selected based on their availability and consensus to participate. Notably,
one village of our selected 8 villages reported no HD within 12 months from the start of our
investigation. In this village, we interviewed two women, who gave birth in the Xepon District
Hospital and the HC, respectively.
Data collection method
We collected data through three types of interviews: In-depth interviews (IDIs), focus group
discussions (FGDs), key informant interviews (KIIs). Due to the low education level of the
women inside our minorities; we could not use the surveys to collect data. However, the IDIs
were conducted by the principal author in the Lao language and recorded by an IC recorder,
with the permission of our participants, to explain their experience and point of view [16].
FGDs and KIIs were conducted to explain the group concerns, experiences and point of views
[17]. The VHVs, HCPs or husbands interpreted from women’s local languages into Lao, when
a woman could not speak Lao. The detailed results of IDIs, FGDs and KIIs are supported in
(Supplementary A, B and C in S1 File).
Data analysis
The interviews were audio-recorded and transcribed into English by YK. Then, the accuracy
of her translation was checked by three native Laos speakers: TP, AL and KX to validate the
translation. Those translations that included characteristics and statements for each partici-
pant are attached (Supplementary D in S1 File). Then, we divided the participants into HD
and FD groups, to analyze their data. The factors that determined the decision-making of the
delivery place were extracted from those interviews. Then, those factors were categorized and
analysed manually according to the socio-ecological model.
Validity and reliability
The validity/reliability of this data collection process is enhanced by the different sources used
to extract data. Our sample size is representative of the rural community we are targeting,
making this result to be generalized to a larger population.
Ethical statement
Ethical approval was pre-obtained from both the National Research Ethics Committee for
Health Research of Ministry of Health, Lao PDR and the Research Ethics Committee of School
of Tropical Medicine and Global Health, Nagasaki University (Supplementary E in S1 File).
Informed consent, in the Lao language explaining the research aims, was obtained from partic-
ipants either by signing or thumbprint. Individuals’ data was anonymized by codes to ensure
their privacy.
Results
A total of sixteen individuals participated in the study; seven and nine participants had HD
and FD, respectively. Our methodology was to choose from each village: an HD case and an
FD case; however, in one village there was no HD case so we had to choose two FD cases.
Among the FDs mothers; four mothers had a birth in Xepon District Hospital (DHP) while
PLOS ONE | https://doi.org/10.1371/journal.pone.0255193 August 2, 2021
4 / 16
PLOS ONE Table 1. Characteristics data of our included participants.
Choice of delivery place in a rural area in Laos
Average (year) (range)
31.4 (17–38)
24.0 (17–36)
Home Delivery (n = 7)
Facility-based Delivery (n = 9)
Age
Ethnicity
Religion
Occupation
Tri
Mangkong
Phoutai
Laolung
Vietnamese
Buddhism
Animism
Christianity
Farmer
Day labourer
Farmer
Day labourer
5
2
0
0
0
1
6
0
6
1
6
3
2
1
2
1
3
4
2
9
0
3
5
2
3
7
2
2
Education period
No education (number of people)
Language (Lao language)
Husband’s age
Husband’s occupation
Average (year) (range)
0.9 (0–6)
Yes
No
3
4
4.1 (0–8)
6
3
Average (year) (range)
33.1 (19–40)
26.6 (19–37)
Husband’s education
No education (number of people)
Average (year) (range)
2.1 (0–7)
6.0 (0–13)
Husband’s language (Lao language)
Number of people (/household)
Property (transportation)
Marriage status
Marriage age �first marriage
Age of first birth
Number of births
https://doi.org/10.1371/journal.pone.0255193.t001
Yes
No
Average (range)
Motorbike (only)
Motorbike and tractor
Nothing
Marriage
Divorce
Remarriage
Average (year) (range)
Average (year) (range)
Average (times) (range)
4
3
7
2
7.1 (6–10)
6.9 (4–9)
3
2
2 2
4
0
3
5
2
2
8
1
0
17.6 (15–23)
18.7 (15–24)
4.8 (1–8)
18.0 (15–27)
18.9 (15–28)
3.1 (1–9)
the rest delivered in the HC. The HD group and FD group had a mean age of 31.4 and 24.0
years with a range of 17–38 and 17–36, respectively. Animism is the principal religion with
(62.5%). Nine participants received no education at all, while the rest received about) 0-9 (year
of education. The majority of our participants (93.75%) were farmers, while only one partici-
pant was a daily labourer.
Regarding ethnicity, the women in the HD group were five out of Tri ethnicity and two
women out of Mangkong ethnicity. However, the FD group included three Tri women, two
Mangkong women, one Phoutai woman, two Laoolung women and one Vietnamese woman.
Regarding the husband’s education, the husband in the HD group and FD group spent about
0.9 and 2.1 years in education, respectively (Tables 1 & 2).
Theoretical foundation and findings
In this study, we followed the social-ecological model of health [18] that categorized the factors
determining the delivery place decision-making into 1) individual level, 2) interpersonal level, 3)
PLOS ONE | https://doi.org/10.1371/journal.pone.0255193 August 2, 2021
5 / 16
PLOS ONE Table 2. Detailed data of the included participants.
Village Age Delivery
Ethnicity
Occupation Education
priod
1
2
3
4
5
6
7
8
place
28 DHP
22 HC
20 DHP
34 Home
28 HC
38 Home
17 HC
36 Home
20 HC
37 Home
36 DHP
33 Home
23 HC
25 Home
22 DHP
17 Home
Laolung
Phoutai
Farmer
Farmer
Vietnamese Farmer
Tri
Farmer
Mangkong
Farmer
Mangkong
Farmer
Mangkong
Farmer
Mangkong
Farmer
Tri
Tri
Tri
Tri
Tri
Tri
Laolung
Tri
Farmer
Day
labourer
Farmer
Farmer
Farmer
Farmer
Farmer
Farmer
https://doi.org/10.1371/journal.pone.0255193.t002
Choice of delivery place in a rural area in Laos
Speaking the Lao
language
Number of
childbirths
ANC visits
(times)
VHVs’ visits
(times)
6 Yes
5 Yes
7 No
0 No
0 Yes
0 Yes
6 Yes
0 No
0 No
0 No
0 No
0 No
8 Yes
0 Yes
5 Yes
6 Yes
1
4
2
5
5
6
1
8
1
6
9
5
3
3
2
1
8
5
1
0
2
2
5
5
0
0
4
0
4
2
5
1
0
2
0
3
0
5
4
5
0
5
5
4
2
2
4
0
community level, and 4) society level (national policy). In Fig 2, we have summarized all the
reported factors under each category.
The individual level
Perception and experience of women. Women who experienced HD showed a positive
perception towards it, through using words like ‘‘Kud sabaii’ (or ‘Sabaii’) [easy delivery]. Some
factors like the short duration of labour, dearth of complication/abnormal symptoms and nor-
mal pain during delivery endorsed this positive perception. For instance, the following is a
quote from a mother:
Fig 2. Summary of the factors affecting the choice of delivery place in a rural area in Laos.
https://doi.org/10.1371/journal.pone.0255193.g002
PLOS ONE | https://doi.org/10.1371/journal.pone.0255193 August 2, 2021
6 / 16
PLOS ONE Choice of delivery place in a rural area in Laos
“Childbirth is ‘Kud sabaii’, so I haven’t ever thought of FD.” (38yrs/HD/multipara).
On the other hand, many FD women clarified similar perception for FD. They said FD was
‘Sabaii’ (convenient). Some factors like the existence of HCPs, the medical care, the hospital
equipment and the available medicines in the health facility bolstered the positive perception.
“There are medical staffs to help me in the HC. There is only my husband to help me at
home. Husband, he doesn’t know his wife [her condition].” (28yrs/FD/multipara).
Some of them reasoned the rush for FD by the fear of difficult delivery or complications
which was the main reason for thrusting them to FD.
“I decided to go to the hospital because we were afraid of some accidents [complications]
happened. So, I went to the hospital.” (20yrs/FD/multipara).
Likely, the FD women without HD experience showed negative thoughts regarding the
HD; as they recognized HD as unsafe, unhealthy and limited towards any abrupt
complications.
“It is not safe to give a childbirth at home.” (28yrs/FD/multipara).
Woman with experience of delivering in the Xepon district hospitals also claimed the pro-
fessionality of the HCPs.
“HCPs took good care and did monitoring well when I felt labour pain” (22yrs/FD/
multipara).
Even women delivered at home showed a positive impact of FD due to awareness cam-
paigns and person-to-person feedback.
“I know friends with FD experience, she said it was good to give a childbirth in the DHP.”
(36yrs/HD/multipara).
The only negative opinion about FD was raised by a shy young mother.
“I didn’t want to go FD because of my shyness. Women who can do FD are not shy.”
(25yrs/HD/multipara).
Unfortunately, five women (four HD women and one FD woman) experienced child deaths
after HD; one woman lost her child two days after HD. However, this history did not affect
their opinion and they insisted on their previous experience regardless of the bad experience.
Forced choice. HD women linked the need to go to the hospital/HC by the existence of a
complication.
“I have to go to the hospital if the bleeding cannot stop. I think like that. There was no case
like bleeding (in the past). I experienced only ‘Kud sabaii’. Previously, we didn’t know,
nobody explains and provided health education. But I understand that now.” (36yrs/HD/
multipara).
For example, two FD women experienced HD in the past and they moved to the hospital/
HC after two- or three-days suffering from labour pain at home without any delivery.
PLOS ONE | https://doi.org/10.1371/journal.pone.0255193 August 2, 2021
7 / 16
PLOS ONE Choice of delivery place in a rural area in Laos
“I intended to give a childbirth at home. But I had felt pain for two days without delivery.
So, I went to the HC.” (38yrs/FD/multipara)
Health education, knowledge and education level. Regarding medical awareness, no
women reported postnatal care (PNC). Concurrently, both numbers of antenatal care (ANC)
visits and VHVs’ visits were similar in 5/8 villages (Table 3). Notably, one FD woman received
neither ANC nor VHVs’ visit. Most women said they received health education from HCPs
and/or VHVs during the latest pregnancies. Two women reported a lack of health awareness
about the delivery place even though with their contact with ANC or VHVs. Some women
could not identify the irregular symptoms, but basically, HD women recorded the same degree
of knowledge as the FD group. The HCPs are linking this low knowledge with the education
level and the language barrier. Nevertheless, the women did not raise the language barrier as a
direct reason for HD. We found that women, who only speak their local language, could use a
familial member for translation into the Lao language,
“One person has a good husband. The husband studied and understood health education.
Another one, the husband says ‘Yes, yes.’ during health education, and they do not come at
the timing of childbirth. He says ‘I am busy to go somewhere.’” (HCP in Manchy HC)
Interpersonal level
Decision-making and communication in a family.
In the FD group, the interfamilial
conversation during pregnancy was more active than in the HD group. The FD-women mainly
concluded FD with their husbands.
Table 3. Data of our participants related to their latest childbirths.
Antenatal care (ANC) visit
village health volunteers (VHV) visit
No
1–3 times
More than 4
No
1–3 times
More than 4
Health education about the delivery place
Yes (received)
No
Health education provider
HCPs/VHVs/Other
HCPs/VHVs
VHVs/Other
Husband/Parents-in-law
Home Delivery (n = 7)
Facility based Delivery (n = 9)
3
3
1
1
2
4
6
1
4
0
1
1
1
2
6
4
2
3
9
0
6
2
1
0
‘Yu-fai’1) duration (days)
‘Yu-fai’ place
‘Yu-fai’ tool for fire
Food restriction after delivery
Average (range)
6.1 (3–11)
7.3 (3–12)
House (including kitchen)
Small hut
Small hut and house
Charcoal
Wood
Yes
No
4
2
1
0
7
1
6
9
0
0
2
7
3
6
1) ‘Yu-fai’ is a traditional postpartum custom in Lao PDR. Women after delivery stay and take a rest near the fire.
https://doi.org/10.1371/journal.pone.0255193.t003
PLOS ONE | https://doi.org/10.1371/journal.pone.0255193 August 2, 2021
8 / 16
PLOS ONE Choice of delivery place in a rural area in Laos
“I and husband talked together that we would go to the hospital or the HC, or at home for
our childbirth. My husband initiated the discussion. (. . .) I replied it was ‘Sabaii’ to give
childbirth in the hospital.” (23yrs/FD/multipara).
“I talked with my family where I would go for my childbirth, and how we would go.”
(22yrs/FD/multipara).
In some cases, parents or parents-in-law participated in the decision-making. Women and
their families seemed to reach an agreement to do FD without dissent in each family.
“No, parents-in-law didn’t say anything about our decision to do FD. They said that it was
‘Sabaii’ to give childbirth in the HC.” (23yrs/FD/multipara)
Some of the HD women were the main decision-makers, as the HD families did not take
the delivery place seriously. Therefore, they had an access to the FD; however, they chose the
HD.
“My husband asked me whether to go to the hospital. I rejected it because of ‘Kud
sabaii’[easy delivery].” (36yrs/HD/multipara).
“My husband said he would take me to the HC for a childbirth. I said I didn’t want to go.”
(25yrs/HD/multipara).
Even one of our cases preferred to go for the FD, however, she went to the HD unintention-
ally. She had labour pain so they rushed to the HD.
“I tried to say to take me to the HC. But my husband didn’t understand. I was painful and
tried to say. My husband said that I ate too much spicy food. (. . .) I felt angry. I felt angry
with my husband. I hoped him to take me to the HC.” (17yrs/HD/primipara).
Only three young mothers depended on their decisions over their families.
“My mother-in-law said to me to go to the HC for a childbirth because of fear of difficult
birth. (. . .) I didn’t feel anything against mother-in-law’s decision, just went.” (20yrs/FD/
primipara).
Transportation.
Inadequate proper means of transportation and the abrupt labour pain
constituted the reasons for the HD preferability.
“There was no vehicle to go.” (33yrs/HD/multipara).
“I felt labour pain in the big house to sleep. I moved to the small hut on foot and gave a
childbirth (she explained a short duration of labour).” (25yrs/HD/multipara).
Our findings showed that there is no significant difference between the two groups regard-
ing the means of transportation. Directing to the health facility was done through owned
motorbikes or hired means of transportation.
“We talked to look for a car, to go by car. I talked together.” (22yrs/FD/multipara).
PLOS ONE | https://doi.org/10.1371/journal.pone.0255193 August 2, 2021
9 / 16
PLOS ONE Choice of delivery place in a rural area in Laos
Financial situation. Both interviewed mothers and the VHVs agreed that the finance may
be considered a reason directing towards a delivery place, but it is not the only one. Notably,
about 42% of our sample dated the preference for HD to poverty. However, one HCP said it
was a good excuse not to do FD. The VHVs reasoned this to the absence of cash or the will to
pay.
“HD woman said she had no money and did not go to the hospital; she would go if she had
money. It is easy to say.” (HCP in DHP).
“They would prefer to use the money for something else because they have no complication
during delivery. It is easy, a baby is born.” (VHV).
While the authorities in Lao PDR implement the free delivery policy, the policy is not noto-
rious for women. In addition to the added cost for FD like gasoline fee, food and unofficial
payment to HCPs as a token of appreciation. Therefore, they tended to take the cost as a
burden.
FD cost the families to pay for gasoline, food, water and baby goods like clothes. Some
women reported an unofficial payment to HCPs as a token of appreciation. On the other
hand, HCPs and VHVs said that women are not obliged to pay for HCPs any additional
payment.
“And ‘Kha mu’ [unofficial payment to HCPs] they want to give 10,000 or 20,000 kip? It
depends on them. It is possible not to pay them if they don’t have money.” (VHV).
Cash preparation during pregnancy was mentioned only in the FD group. One young FD
woman did not know the policy of free delivery or the details of cost because her husband cov-
ered it. Normally, families get money from hunting, gathering, selling livestock, day labour
and saving.
“I prepared the mother and child book, diaper, cloth for holding a baby, Lao skirt [a tradi-
tional skirt] and money.” (22yrs/FD/multipara).
One multipara received finance and transportation support from her family because of a
shortage of money. In her case, she did not arrange this support in advance because she did
not plan FD.
“I borrowed money from my brother-in-law [the husband of her old sister]. I did not have,
and my husband did not have money.” (36yrs/FD/multipara).
Family’s role and interpersonal relationship. Preparation for delivery was common in
the FD group more than in the HD group. In contrast, three HD families constructed a small
hut for HD between six to eight months of pregnancy.
“I told my husband when it was near childbirth. My husband found a small hut as a place to
childbirth.” [Her husband constructed the small hut around six months of pregnancy.]
(38yrs/HD/multipara).
One male VHV mentioned it was one of the husband’s role to stay with their wives in the
last month of pregnancy.
PLOS ONE | https://doi.org/10.1371/journal.pone.0255193 August 2, 2021
10 / 16
PLOS ONE Choice of delivery place in a rural area in Laos
“A wife is pregnant and she will give a childbirth this month, her husband doesn’t go any-
where and have to stay with his wife to take care of his wife, go together anywhere she
wants to go.” (VHV).
Relatives and friends with experience of FD played the role of giving information to HD women.
“I heard. My relatives said the experience of FD before.” (34yrs/HD/multipara).
A male VHV also mentioned villagers shared their experience with others and there was a
possibility of persuasion to change their behaviour.
“Women who gave a childbirth at a health facility before explain to their friend when they
gave a childbirth and felt happy.” (VHV).
Factually, there are cases that villagers positively affect HD by agreeing with HD.
“Relatives said that HD was ‘Kud sabaii’. Not afraid.” (36yrs/HD/multipara).
Notably, the HD husbands supported their wives through the delivery process including
other members of their families. The support extended to include direct care as receiving a
baby, cutting an umbilical cord with a razor or wood, and placenta delivery.
“My husband holds my body. There were me and a friend inside, my husband. And they
received me. I did strain and gave childbirth. Next, cord-cutting, I cut it by myself, and a
placenta came out by myself. Only mother-in-law and my friend wiped a baby and hold a
baby by cloth.” (34yrs/HD/multipara).
On the other hand, the FD husbands’ missions were to prepare the cash and transportation
for their wives to the hospital or HCs. Mother-in-law and female relatives accompanied them.
.“My husband took me to the HC. I went with my husband and old sister, two people.”
(28yrs/FD/multipara)
The role of relatives in health facilities was to psychologically support mothers. They had no
rule regarding health care such as taking care of the new-born. For instance, a husband entered
a delivery room and encouraged his wife.
“My husband and parents and relatives did not help anything. They went and just
observed.” (28yrs/FD/primipara).
Community-level
The trend of delivery place.
In two remote villages with difficult means of transportation,
HD is still common. The VHVs accepted the HD and recommended the shift to FD in case of
delivery linked complications.
“They have to go to the HC if there are symptoms of complications. If there is nothing, not
painful, common pain, they just give a birth at home.” (VHV).
“Strong pain makes a husband take his wife to the hospital. He is afraid that a problem hap-
pens. Because. . .”(VHV).
PLOS ONE | https://doi.org/10.1371/journal.pone.0255193 August 2, 2021
11 / 16
PLOS ONE Choice of delivery place in a rural area in Laos
On the other hand, the shift of delivery place from home to health facilities became a recent
trend in Xepon district including the targeted villages of this study.
“The smaller number of women do HD than before.” (VHV).
Accessibility. One hour and a half to reach the HC, or to use the hard off-road raised the
fear of women giving birth on the road or facing troubles. The distance and inaccessibility of
women to reach the centres skipped them to the HD.
“Uh. The road is difficult. If the road is available, then it’s easier to go FD.” (VHV).
“We want to give a childbirth in the HC, but afraid to give a childbirth on the way.” (VHV).
The other VHVs and HCPs mentioned the construction of the 9E road (main road of
Xepon district which connects the centre of Savannakhet Province to Vietnam) contributed to
the hike in the number of FD.
“In 2010–2012, the road [9E road] was completely opened and it became convenient.”
(HCP in DHP).
On the other hand, some VHVs did not agree that improving accessibility like the new road
or building new HCs directly affect the delivery place. They mentioned other reasons to
change people’s behaviour.
“They can go to the HC, but FD hasn’t been all yet. Around this (latest) three years, they
come to the HC for childbirths. There are still people with HD.” (HCP).
VHV’s service and health education (community). The VHVs’ mission is to raise medi-
cal awareness for pregnant women and their families; three women with no ANC visits
received their visits from the volunteers.
“Because they understand to go to the HC. Delivery in an HC is safe. There is no right thing
at home. If they use the incorrect thing, it is not healthy, pain and fever, mother and child.
Tool for delivery at home is not like ones in an HC, not clean. We are afraid of becoming
diseases. Many people understand that and go to HC.” (VHV).
VHVs reported that awareness campaigns contributed to improving their knowledge about the
benefit of FD, risk of HD and free delivery policy. Therefore, more people choose FD than before.
Societal level
Free delivery policy and implementation. The free delivery policy implemented in the
region granted the women free service; however, they did not receive non-medical support
like transportation fees. Besides, some women reported their ignorance regarding this policy.
Delivery with the skilled birth attendant at home. Limited resources hindered the HCPs
to introduce the medical services at home; so, women need to go to the facilities to receive the
service from the skilled birth attendant.
“They come if they have a car. In the case that a placenta does not come out after HD, a
family comes to an HC. But the staff does a night shift alone (she cannot go to a village), so
we ask them to take a mother and child here.” (HCP).
PLOS ONE | https://doi.org/10.1371/journal.pone.0255193 August 2, 2021
12 / 16
PLOS ONE Choice of delivery place in a rural area in Laos
Discussion
This study is formed as a community-based qualitative study to be able to figure out the
answers to its aim. As we intended to extract data about the social behaviour of the minorities
living in those rural villages in Lao PDR; we had to use qualitative research methods. The rea-
son is that the quantitative methods are known for their limitations in describing the social
experiences [19]. Besides, this research is investigating the behaviour of minorities that have
no many published reports about them; therefore the qualitative approach was selected to give
in-depth information about the problem [20]. From the quantitative methods; the commu-
nity-based was selected to involve the minorities community to draw their social preferences
and reasoning it form their exact point of view [21]. Using the community-based model we
had surveys, interviews, focus groups, observations or the literature. Coming to literature, the
data published are rare. Also, choosing the surveys was not a good choice due to the low educa-
tion level and the misunderstanding that may happen. Therefore, we used all that we can to
extract the responses.
Our findings suggested that family communication influenced the delivery place and the
preparation for childbirth. Unlike some studies [12,22], we found that our participants had
more autonomy among the Lao PDR minorities to choose their delivery place. We can date
this freedom to the surge of public awareness and the implementation of new policies which
have empowered pregnant women. On the other hand, we found that when the communica-
tion inside the family abates, the HD upsurges. This is supported by Kifle et al. who stated that
the couple is likely to choose the FD when they choose together more than if one of them
decides alone [23]. Moreover, Kabakyenga et al found that the discussion with a third party,
like the HCPs, boosted the possibility of FD [24]. Overall, interfamilial communication along-
side medical awareness can scale down HD.
Besides, the pre-preparation for birth was more familiar in the FD group. Our finding is
coming in accordance with a study that stated that women with good preparedness for birth
were skewed to the FD [25]. As well, Stiefane et al. have figured a rope between the institu-
tional delivery and the rehearsal for birth [26], which is backing our finding. In Lao PDR, the
current “mother-and-child book” urges the people to create a plan about the delivery place,
companion(s), transportation and the financial issues for delivery. Accordingly, this step
should boost the direction of new mothers towards the FD. Currently, there is a responsibility
on the HCPs with the VHVs to aware the pregnant women including their families to establish
their plan and follow the guidelines mentioned in the “mother-and-child-book”.
Surprisingly, we reported that the crucial factor supporting HD is the inherited perception
of upcoming mothers from their ancestors. In accordance, previous studies found women,
with previous experience with the HD, continued to implement it [8,12]. Another study found
that the low risk in birth addressed women to acknowledge the HD [27]. Overall, the multi-
paras continued to conduct the HD even after they acknowledged the risks, and it is hard to
convince them of the opposite. On the opposite, women who experienced FD continued this
path. Hence, it is better to focus the effort on primiparas to reduce the total number of HD.
Notably, the awareness should be introduced to primipara with their families, as the young
mothers are likely to depend on their families in their decisions. In literature, young mothers
tended to give birth to institutions [27].
On the other hand, our findings revealed that multiparas intend to repeat the experience
which hails the importance of making it easy for them to get help if any complications hap-
pened during delivery. Issues like illness perception, finance, distance/transportation should
be studied carefully to reduce the number of HDs [28]. Minimizing the risk of HD is not
avoidable to reduce the mortality rates among mothers and babies. For instance, the HCPs and
PLOS ONE | https://doi.org/10.1371/journal.pone.0255193 August 2, 2021
13 / 16
PLOS ONE Choice of delivery place in a rural area in Laos
VHVs do have a responsibility to educate women and their families on the critical timing that
they should switch to FD. Moreover, they should work with the community to ease transporta-
tions in such cases. Furthermore, the availability of skilled birth attendants inside the Lao PDR
communities may solve this dilemma.
Our findings determined the absence of postnatal care for the mothers included in this
study, which resulted in the absence of any assessment or preventive measures shared with the
families. On the other hand, the literature is stating that postnatal care significantly reduced
neonatal death [29]. The value of postnatal care is not only to promote FD but also to increase
the care and reduce morbidities and mortalities. Therefore, the WHO inquire that the first
postnatal care should be done on the first day after delivery [30].
Our findings found that women received the medical benefits of the free-delivery-policy
and no one did receive the non-medical benefits (transport cost, transfer cost and food allow-
ance) that have been included. Therefore, we think that to get the optimum benefit out of the
free delivery policy; a wider and deeper awareness campaign should be performed to cover
most pregnant women. The respondents mentioned the financial difficulty as an obstacle of
FD, even with their knowledge about free delivery. Therefore, non-medical support consti-
tuted a supportive idea to promote FD. In a study in Nepal, incentive and free delivery pro-
grams could enhance the FD [31].
Limitation of the study
The main limitation of this study is the small number of respondents. Therefore, these findings
are not sufficient for generalization to the population. We intended to investigate more vil-
lages; but the pandemic Covid-19 hindered our progress. In the targeted villages, people speak
the Lao language and/or their tongue which shaped a barrier for communication. There was a
possibility that the opinions of women who cannot speak the Lao dialect were modified.
Besides, there was a possibility of social-desirability bias. Besides, we could not stay enough
time to observe the situation of the villages because of COVID-19. However, with the few limi-
tations we mentioned, this paper is providing full detail regarding the remoted-area women to
improve maternal care services.
Conclusion
In this study, we have found that the inherited perception of HD and the low risks of the deliv-
ery process constituted the main factor that propagated the choice of HD. Meanwhile, pro-
longed labour pain with complications played a vital role in the designation of institutional
birthing. We figured, also, that the hike in the interfamilial communication with the good
preparation of the delivery process directed the mothers towards the FD. Advantageously, the
birth trend is switched towards the FD because of the raised knowledge and improved accessi-
bility. Our results suggest increasing the awareness of the non-medical benefits of the Lao PDR
free delivery policy and to boost the number of skilled birth attendants. Finally, the spread of
the FD and childbirth with SBAs are critical in reducing both morbidities and mortalities
among mothers and babies. Besides, we urge the healthcare authorities to launch awareness
campaigns to introduce the community in those rural places to their free delivery policies.
Supporting information
S1 File. Additional files. Supplementary A: Focus group discussions (FGDs) with health care
providers (HCPs). Supplementary B: Key informant interviews (KIIs) to Village health volun-
teers (VHVs). Supplementary C: Questionnaire/interview guides used for the research. Supple-
mentary D: Summary sheets of women’s characteristic and statements. Supplementary G:
PLOS ONE | https://doi.org/10.1371/journal.pone.0255193 August 2, 2021
14 / 16
PLOS ONE Choice of delivery place in a rural area in Laos
Ethical approvals.
(DOCX)
Author Contributions
Conceptualization: Yoshiko Kawaguchi, Alliya Shafi, Nguyen Tien Huy, Kazuhiko Moji.
Data curation: Sengchanh Kounnavong, Tiengkham Pongvongsa, Angkhana Lasaphonh,
Khamsamay Xaylovong, Miho Sato, Mitsuaki Matsui, Atsuko Imoto.
Investigation: Yoshiko Kawaguchi.
Methodology: Nguyen Tien Huy.
Writing – original draft: Ahmad M. Sayed.
References
1. Darmstadt GL, Lee AC, Cousens S, Sibley L, Bhutta ZA, Donnay F, et al: 60 million non-facility births:
who can deliver in community settings to reduce intrapartum-related deaths? International Journal of
Gynecology & Obstetrics 2009, 107:S89–S112. https://doi.org/10.1016/j.ijgo.2009.07.010 PMID:
19815200
2. Nasah BT, Mati JK, Kasonde JM: Contemporary issues in maternal health care in Africa: CRC Press;
1994.
3. Sarker BK, Rahman M, Rahman T, Hossain J, Reichenbach L, Mitra DK: Reasons for preference of
home delivery with traditional birth attendants (TBAs) in rural Bangladesh: a qualitative exploration.
PloS one 2016, 11(1):e0146161. https://doi.org/10.1371/journal.pone.0146161 PMID: 26731276
4. Sychareun V, Phengsavanh A, Hansana V, Somphet V, Menorah S: Cultural beliefs and traditional ritu-
als about child birth practices in Lao PDR. Kuala Lumpur: The Asian-Pacific Resource & Research
Centre for Women (ARROW) 2009.
5. Schlemmer G: Ethnic 10 Belonging in Laos: A Politico-Historical Perspective. Changing Lives in Laos:
Society, Politics and Culture in a Post-Socialist State 2017:251.
6. Bureau LS: Lao social Indicator survey II 2017, survey findings report. Vientiane: Lao Statistics Bureau
and UNICEF 2018.
7. Organization WH, Health WHOR: Health worker role in providing safe abortion care and post abortion
contraception: World Health Organization; 2015.
8. Kasaye HK, Endale ZM, Gudayu TW, Desta MS: Home delivery among antenatal care booked women
in their last pregnancy and associated factors: community-based cross sectional study in Debremarkos
town, North West Ethiopia, January 2016. BMC pregnancy and childbirth 2017, 17(1):225. https://doi.
org/10.1186/s12884-017-1409-2 PMID: 28705188
9.
Lao P: Ministry of Health (Lao PDR MoH). WHO Health service delivery profile Lao PDR 2012.
10. Gitimu A, Herr C, Oruko H, Karijo E, Gichuki R, Ofware P, et al: Determinants of use of skilled birth
attendant at delivery in Makueni, Kenya: a cross sectional study. BMC pregnancy and childbirth 2015,
15(1):9. https://doi.org/10.1186/s12884-015-0442-2 PMID: 25645900
11. Phathammavong O, Ali M, Souksavat S, Chounramany K, Kuroiwa C: Antenatal care among ethnic
populations in Luang Namtha Province, Lao PDR. Southeast Asian journal of tropical medicine and pub-
lic health 2010, 41(3):705.
12. Sato C, Phongluxa K, Toyama N, Gregorio ER, Miyoshi C, Nishimoto F, et al: Factors influencing the
choice of facility-based delivery in the ethnic minority villages of Lao PDR: a qualitative case study.
Tropical medicine and health 2019, 47(1):50. https://doi.org/10.1186/s41182-019-0177-2 PMID:
31516363
13. Sanaphay A, Daenseekaew S, Smith JF, Eckermann E, Scopaz A: Home delivery in southern Lao
PDR: challenges to achieving MDG 4 & 5 targets. Philippine journal of nursing 2014, 84(1):24–31.
14. Belay A, Sendo E: Factors determining choice of delivery place among women of child bearing age in
Dega Damot District, North West of Ethiopia: a community based cross-sectional study. BMC preg-
nancy and childbirth 2016, 16(1):229. https://doi.org/10.1186/s12884-016-1020-y PMID: 27535126
15. Sychareun V, Hansana V, Somphet V, Xayavong S, Phengsavanh A, Popenoe R: Reasons rural Lao-
tians choose home deliveries over delivery at health facilities: a qualitative study. BMC pregnancy and
childbirth 2012, 12(1):86. https://doi.org/10.1186/1471-2393-12-86 PMID: 22925107
PLOS ONE | https://doi.org/10.1371/journal.pone.0255193 August 2, 2021
15 / 16
PLOS ONE Choice of delivery place in a rural area in Laos
16. Dahlgren L, Emmelin M, Winkvist A: Umeå University; 2007. Qualitative methodology for international
public health: 218.
17. Barbour R, Kitzinger J: Developing focus group research: politics, theory and practice: Sage; 1998.
18. Clinical, Awards TS: Community Engagement Key Function Committee. Principles of Community
Engagement Second Edition 2011.
19. Sudeshna D, Datt S: Limitations and weakness of quantitative research methods. In.: Los Angeles:
Scribd Inc; 2016.
20. Queiro´s A, Faria D, Almeida F: Strengths and limitations of qualitative and quantitative research meth-
ods. European Journal of Education Studies 2017.
21. Hills M, Mullett J: Community-based research: Creating evidence-based practice for health and social
change. 2000.
22. Kaiser JL, Fong RM, Hamer DH, Biemba G, Ngoma T, Tusing B, et al: How a woman’s interpersonal
relationships can delay care-seeking and access during the maternity period in rural Zambia: an inter-
section of the social ecological model with the three delays framework. Social Science & Medicine
2019, 220:312–321.
23. Kifle MM, Kesete HF, Gaim HT, Angosom GS, Araya MB: Health facility or home delivery? Factors influ-
encing the choice of delivery place among mothers living in rural communities of Eritrea. Journal of
Health, Population and Nutrition 2018, 37(1):22.
24. Kabakyenga JK, O¨ stergren P-O, Turyakira E, Pettersson KO: Influence of birth preparedness, decision-
making on location of birth and assistance by skilled birth attendants among women in south-western
Uganda. PloS one 2012, 7(4):e35747. https://doi.org/10.1371/journal.pone.0035747 PMID: 22558214
25. Belda SS, Gebremariam MB: Birth preparedness, complication readiness and other determinants of
place of delivery among mothers in Goba District, Bale Zone, South East Ethiopia. BMC pregnancy and
childbirth 2016, 16(1):73. https://doi.org/10.1186/s12884-016-0837-8 PMID: 27053241
26. Sitefane GG, Banerjee J, Mohan D, Lee CS, Ricca J, Betron ML, et al: Do male engagement and cou-
ples’ communication influence maternal health care-seeking? Findings from a household survey in
Mozambique. BMC Pregnancy and Childbirth 2020, 20(1):1–13. https://doi.org/10.1186/s12884-020-
02984-y PMID: 32527233
27. Sialubanje C, Massar K, Hamer DH, Ruiter RA: Reasons for home delivery and use of traditional birth
attendants in rural Zambia: a qualitative study. BMC pregnancy and childbirth 2015, 15(1):216. https://
doi.org/10.1186/s12884-015-0652-7 PMID: 26361976
28.
29.
Thaddeus S, Maine D: Too far to walk: maternal mortality in context. Social science & medicine 1994,
38(8):1091–1110. https://doi.org/10.1016/0277-9536(94)90226-7 PMID: 8042057
Titaley CR, Dibley MJ, Agho K, Roberts CL, Hall J: Determinants of neonatal mortality in Indonesia.
BMC public health 2008, 8(1):232. https://doi.org/10.1186/1471-2458-8-232 PMID: 18613953
30. Organization WH: WHO recommendations on postnatal care of the mother and newborn: World Health
Organization; 2014.
31. Bhatt H, Tiwari S, Ensor T, Ghimire DR, Gavidia T: Contribution of Nepal’s free delivery care policies in
improving utilisation of maternal health services. International journal of health policy and management
2018, 7(7):645. https://doi.org/10.15171/ijhpm.2018.01 PMID: 29996584
PLOS ONE | https://doi.org/10.1371/journal.pone.0255193 August 2, 2021
16 / 16
PLOS ONE |
10.1371_journal.pone.0254520 | RESEARCH ARTICLE
Potency of propofol for inducing loss of
consciousness in end-stage kidney disease
patients
Mi Roung Jun1☯, Mun Gyu Kim2☯, Ki Seob HanID
Young ParkID
2*, Sanghoon Song2, Jae Hwa Yoo2, Ji Won Chung2, Sang Ho Kim2
1, Ji Eun Park1, Ho Bum Cho2, Sun
1 Department of Anesthesiology and Pain Medicine, Samsung Changwon Hospital, Sungkyunkwan
University School of Medicine, Changwon, Republic of Korea, 2 Department of Anaesthesiology and Pain
Medicine, Soonchunhyang University Hospital Seoul, Seoul, Republic of Korea
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
☯ These authors contributed equally to this work.
* sunnypark97@schmc.ac.kr
Abstract
It can be difficult for anesthesiologists to determine the optimal dose of propofol for end-stage
kidney disease (ESKD) patients due to changes in drug disposition. The purpose of this
study was to evaluate the potency of propofol for inducing loss of consciousness in ESKD
patients. Patients with normal kidney function (Control group, n = 15), those with ESKD
(ESKD group, n = 15), and those with ESKD undergoing cervical epidural anesthesia
(ESKD-CEB group, n = 15) were administered propofol by target-controlled infusion (TCI)
using the Schneider model. The effect-site concentration (Ce) of propofol started at 0.5 μg/ml
and increased in increments of 0.5 μg/ml until the patient did not respond to verbal com-
mands. The relationship between the probability (P) of loss of consciousness and the Ce of
propofol was analyzed in each group using logistic regression. The Ce values of propofol at
the time of loss of consciousness were 4.3 ± 0.9, 3.7 ± 0.9, and 3.3 ± 1.0 μg/ml for the Control,
ESKD, and ESKD-CEB* groups, respectively (*significant difference vs. control, P < 0.05).
The estimated Ce50 values for lost ability to respond to verbal command were 4.56, 3.75, and
3.21 μg/ml for the Control, ESKD, and ESKD-CEB groups, respectively. In conclusion, when
inducing anesthesia in ESKD patients, we recommend using an initial dose similar to that of
patients with normal kidney function, or rather starting with a lower dose.
OPEN ACCESS
Citation: Jun MR, Kim MG, Han KS, Park JE, Cho
HB, Park SY, et al. (2021) Potency of propofol for
inducing loss of consciousness in end-stage
kidney disease patients. PLoS ONE 16(8):
e0254520. https://doi.org/10.1371/journal.
pone.0254520
Editor: Pasqual Barretti, Universidade Estadual
Paulista Julio de Mesquita Filho, BRAZIL
Received: February 26, 2021
Accepted: June 2, 2021
Published: August 12, 2021
Copyright: © 2021 Jun et al. This is an open access
article distributed under the terms of the Creative
Commons Attribution License, which permits
unrestricted use, distribution, and reproduction in
any medium, provided the original author and
source are credited.
Data Availability Statement: All relevant data are
within the manuscript and its S1 Appendix files.
Introduction
Funding: This research was supported by the
Soonchunhyang University Research Fund and the
National Research Foundation of Korea (NRF) grant
funded by the Korea government (MSIT)
(2017R1C1B5076787).
Competing interests: The authors have declared
that no competing interests exist.
Propofol is a short-acting, lipophilic intravenous general anesthetic. The hypnotic action of
propofol is probably mediated through γ-aminobutyric acid (GABA) receptor (agonist) and
N-methyl-D-aspartate (NMDA) receptor (antagonist). Propofol has a protein binding of about
98% and is rapidly metabolized to water-soluble inactive metabolites in the liver and excreted
through the kidneys [1, 2].
End-stage kidney disease (ESKD) is defined as irreversible decline in a person’s own kidney
function, which is severe enough to be fatal in the absence of dialysis or transplantation [3].
PLOS ONE | https://doi.org/10.1371/journal.pone.0254520 August 12, 2021
1 / 9
PLOS ONE Potency of propofol for inducing LOC in ESKD patients
Abnormal increases in toxins, inflammatory factors, and parathyroid hormone (PTH) can
lead to uremic symptoms in patients with ESKD, which can in turn affect drug disposition [4,
5]. For example, the volume of distribution (Vd) may increase due to interference with protein
binding [4]. Thus, an appropriate dose of a drug in patients with normal kidney function can
appear inadequate or excessive in ESKD patients.
Propofol is commonly used for sedation and general anesthesia in ESKD patients due to
rapid recovery after continuous infusion. Previous studies have shown that the pharmacoki-
netics of propofol are not significantly different between patients with normal kidney function
and ESKD [6–8]. Although some investigators have suggested that the hyperdynamic circula-
tion caused by anemia increases the dose requirement of propofol for inducing anesthesia in
ESKD patients [9], studies are conflicting and controversial.
It is important to determine the optimal initial dose of propofol, as this drug can cause car-
diovascular instability during the induction of anesthesia. However, information about the
concentration-response relationship of propofol during induction of anesthesia in ESKD
patients is limited.
This study was performed to investigate the potency of propofol for loss of consciousness,
administered by target-controlled infusion (TCI) for inducing general anesthesia in ESKD
patients compared to patients with normal kidney function. In addition, we evaluated the
potency of propofol for inducing loss of consciousness in ESKD patients who received a cervi-
cal epidural block (CEB).
Materials and methods
Patient population
This study was approved by the Institutional Review Board of hospital (Ref. 2016-05-006-003),
and was registered with the International Clinical Trials Registry Platform (http://cris.nih.go.
kr). Written informed consent was obtained from all patients on the day before surgery.
Fifteen patients with normal kidney function scheduled for elective surgery under general
anesthesia were recruited as the control group (Control group). Thirty patients with ESKD
presenting for arteriovenous fistulae surgery were enrolled between October 2016 and April
2018. Fifteen of the thirty patients with ESKD received general anesthesia (ESKD group);
and the other fifteen patients received a CEB and underwent monitored anesthesia care
(ESKD-CEB group). All ESKD patients had been dialyzed the day before surgery. Preoperative
laboratory testing was performed before surgery in all patients, and after dialysis in the ESKD
patients. Patients were excluded for the following reasons: a history of neurological or psycho-
logical disease or the presence of current neurological symptoms; the presence of a liver func-
tion abnormality; body mass index � 30 kg/m2; a history of hearing impairment; a history of
an allergic reaction to soybean; a history of an adverse drug reaction to propofol or local anes-
thetic; or kidney transplantation. Patients who were on medications that affected the central
nervous system (CNS) (e.g., analgesics, anticonvulsants, and hypnotics) were also excluded
from the study.
Study procedure
Electrocardiography, noninvasive arterial blood pressure, and peripheral oxygen saturation
monitoring were initiated when patients arrived in the operating room. An epidural catheter
was inserted at C7-T1, and 15 ml of 0.45% ropivacaine was administered after negative aspira-
tion in the ESKD-CEB group. Adequate sensory block was confirmed 20 min after injecting
the local anesthetic; CNS symptoms (e.g., dizziness, visual and auditory disturbances,
PLOS ONE | https://doi.org/10.1371/journal.pone.0254520 August 12, 2021
2 / 9
PLOS ONE Potency of propofol for inducing LOC in ESKD patients
disorientation, drowsiness, tremors, shivering, muscular twitching, and generalized tonic-
clonic convulsions) caused by the local anesthetics were observed and recorded.
In all patients, disposable bispectral index (BIS) sensors were placed on the forehead, as rec-
ommended by the manufacturer, and connected to a BIS monitor (BIS1 Monitor; Medtronic,
Dublin, Ireland). The effect-site concentration (Ce) of propofol started at 0.5 μg/ml and was
delivered via a TCI pump (Orchestra1 Base Primea; Fresenius Vial, Bre´zins, France) using the
Schnider model [10]. The Ce was increased in increments of 0.5 μg/ml, 3 min after the previ-
ous target Ce was reached, until the patient did not respond to the loud verbal command,
“open your eyes”, which was defined as loss of consciousness (LOC). The BIS values, blood
pressure, heart rate, peripheral oxygen saturation, and total amount of propofol infused were
recorded 3 min after the target Ce was reached, and the investigator then evaluated the level of
sedation. Oxygen was administered at 5 L/min via a face mask during the study period.
Adverse events requiring cardiovascular or respiratory support during propofol-induced seda-
tion were observed and recorded.
Sample size
In the preliminary study, the total amount of propofol at the time of loss of consciousness were
156.4 ± 15.0 mg, 82.6 ± 8.6 mg, and 63.9 ± 15.0 mg for the Control, ESKD, and ESKD-CEB
groups, respectively. Based on this, the sample sizes required to confirm the difference in the
total amount of propofol between the Control and ESKD groups, ESKD and ESKD-CEB
groups, and Control and ESKD-CEB groups was 11 (α = 0.05, β = 0.2). A 30% dropout rate
was applied here, and the final sample size was 15 subjects per group. The power of the sample
size was reconfirmed using R package wmwpow.
Statistics
One-way analysis of variance or the Kruskal-Wallis test, followed by multiple comparison pro-
cedures using the Holm-Sidak method, was performed to compare the three groups. Continu-
ous variables are presented as mean ± standard deviation, and categorical variables as integers.
All statistical analyses were performed using SigmaPlot 13.0 for Windows (Systat Software,
Inc., Chicago, IL, USA). A P-value < 0.05 was considered significant. We used R package
wmwpow for power analysis.
Probability of LOC and pharmacodynamic analysis
The following sigmoid Emax model was used to determine the relationship between the proba-
bility (P) of LOC and the Ce of propofol in each group:
p of LOC ¼
Cg
e
Cg
e50 Group
þ Cg
e
Where Ce50_Group is the Ce associated with a 50% probability of LOC in each study group,
and γ is the slope of the concentration versus the unconscious probability curve. The pharma-
codynamic model parameters were estimated using the “LIKELIHOOD LAPLACE
METHOD = conditional” option in NONMEM1 7 level 3 (ICON Development Solutions,
Dublin, Ireland). The IIV(random inter-individual effect) of the Ce50 was fixed at zero.
Results
The study flow diagram is shown in Fig 1. No signs of CNS toxicity due to the local anesthetic
were detected during the CEB in any patients in the ESKD-CEB group. No significant adverse
PLOS ONE | https://doi.org/10.1371/journal.pone.0254520 August 12, 2021
3 / 9
PLOS ONE Potency of propofol for inducing LOC in ESKD patients
Fig 1. The flow diagram of the study.
https://doi.org/10.1371/journal.pone.0254520.g001
events requiring cardiovascular or respiratory support occurred during propofol-induced
sedation. The patients’ characteristics are summarized in Table 1. There were expected differ-
ences in laboratory measures of renal function, hemoglobin, and platelets. In addition, in the
ESKD patients who had received a CEB were older with lower BMI and weight. Neither lean
body mass or albumin was different.
PLOS ONE | https://doi.org/10.1371/journal.pone.0254520 August 12, 2021
4 / 9
PLOS ONE Potency of propofol for inducing LOC in ESKD patients
Table 1. Patient characteristics and preoperative laboratory values.
Age (yr)
Sex, M/F
Height (cm)
Weight (kg)
Body mass index (kg/m2)
Lean body mass (kg)
Albumin (g/dL)
Hemoglobin (g/dL)
Platelet (ⅹ103/μL)
BUN (mg/dL)
Creatinine (mg/dL)
eGFR (ml/min/1.73m2)
Potassium (mmol/L)
Prothrombin time (INR)
Control
49.7 ± 9.5
8/7
165.7 ± 9.0
68.4 ± 12.6
24.8 ± 3.3
49.1 (34.3–49.3)
4.6 (4.2–4.6)
13.9 ± 1.4
247 ± 82
9.7 (36.2–74.7)
0.8 (0.7–0.8)
101.9 (97.6–111.2)
4.3 ± 0.7
1.1 (1.0–1.1)
ESKD
59.1 ± 13.9
6/9
157.4 ± 7.1�
56.4 ± 9.9�
22.8 ± 3.7
43.0 (37.8–46.0)
4.3 (4.0–4.4)
10.7 ± 2.0�
162 ± 71�
48.9 (36.5–62.7)�
8.8 (5.1–9.8)�
5.3 (4.4–9.2)�
4.9 ± 0.5�
1.0 (0.9–1.1)�
ESKD_CEB
66.1 ± 11.6�
7/8
159.0 ± 8.1
52.4 ± 11.0�
20.5 ± 3.9�
43.4 (37.9–57.5)
4.4 (4.1–4.8)
11.9 ± 1.4�†
177 ± 45�
54.1 (8.0–14.2)�
7.4 (4.9–11.1)�
7.2 (3.7–9.9)�
4.8 ± 0.9
1.0 (1.0–1.0)�
P
0.002
0.023
0.002
0.016
0.132
0.184
<0.001
0.003
<0.001
<0.001
<0.001
0.031
0.015
Values expressed in mean ± SD, median (25th–75th percentiles) or numbers of patients. ESKD, End stage kidney disease; CEB, Cervical epidural block; eGFR, Estimated
glomerular filtration rate; INR, International normalized ratio.
�There is significant difference vs Control (P < 0.05)
†There is significant difference vs ESKD (P < 0.05).
https://doi.org/10.1371/journal.pone.0254520.t001
The Ce values of propofol at the time of loss of consciousness were 4.3 ± 0.9, 3.7 ± 0.9, and
3.3 ± 1.0 μg/ml for the Control, ESKD, and ESKD-CEB groups, respectively. The Ce value at
the time of LOC was lower in ESKD patients; however, a significant difference was only
detected between the Control and ESKD-CEB groups (Table 2). The statistical power to com-
pare the difference was calculated 99.8% and 73.3% between Control group and ESKD group
and between Control group and ESKD-CEB group, respectively. The pharmacodynamic
model parameters and logistic regression curve depicting the relationship between the study
groups and the probability of LOC are shown in Fig 2. The Ce50 values were estimated to be
4.56, 3.75, and 3.21 μg/ml for the Control, ESKD, and ESKD-CEB groups, respectively.
Discussion
In our small study, propofol Ce and Ce50 values were lower after infusion to achieve loss of
consciousness than controls in ESKD patients who had received a CEB. In ESKD patients
without a CEB, lower Ce50 was not statistically significant although likely was of clinical
significance.
The incidence of ESKD requiring dialysis or kidney transplantation has increased with the
increasing prevalence of hypertension and diabetes in the elderly population [11].
Table 2. Effect-site concentration (Ce), total amount of propofol and bispectral index (BIS) at loss of consciousness.
Ce (μg/ml)
Total amount (mg)
BIS
Control
4.0 (3.5–5.0)
124.0 (99.2–152.0)
77.0 (68.0–81.0)
ESKD
4.0 (3.0–4.0)
94.5 (68.5–120.0)
71.0 (65.0–77.0)
ESKD_CEB
3.0 (2.5–3.5)�
81.0 (62.3–95.0)�
73.0 (68.0–84.0)
P
0.009
0.004
0.567
Values expressed in median (25th–75th percentiles). ESKD, end stage kidney disease; CEB, cervical epidural block.
�There is significant difference vs Control (P < 0.05).
https://doi.org/10.1371/journal.pone.0254520.t002
PLOS ONE | https://doi.org/10.1371/journal.pone.0254520 August 12, 2021
5 / 9
PLOS ONE Potency of propofol for inducing LOC in ESKD patients
Fig 2. Relationship between the probability of LOC and the Ce of propofol at each study groups. Ce50, effect-site concentration associated with a 50% probability of
LOC; ESRD, end stage renal disease; LOC, loss of consciousness; CEB, cervical epidural block; γ, steepness of the concentration versus response relationship; RSE,
relative standard error.
https://doi.org/10.1371/journal.pone.0254520.g002
Accordingly, ESKD patients required to undergo surgical procedures are also increasing.
Anesthesiologists should be aware of the anesthetic consideration of these patients. It is gener-
ally accepted that renal clearance decreases due to decreased estimated glomerular filtration
rate in ESKD patients, which may affect the pharmacokinetics of the drugs cleared by the
PLOS ONE | https://doi.org/10.1371/journal.pone.0254520 August 12, 2021
6 / 9
PLOS ONE Potency of propofol for inducing LOC in ESKD patients
kidneys. However, many drugs, metabolized by the liver, are also known to affect drug distri-
bution, metabolism and transport in ESKD patients and these pharmacokinetic alterations are
thought to be due to the effects of uremia [12]. Changes in the biochemical function of a drug
may be induced by uremic toxins and have various clinical manifestations [11, 13]. Dialysis is
initiated in ESKD patients to treat uremic symptoms caused by toxic solutes; however, conven-
tional dialysis does not eliminate the symptoms because it can only remove two-thirds of the
total urea and a limited number of solutes [14].
Propofol, a drug that is not cleared by the kidneys, is primarily metabolized through O-glu-
curonidation in the conjugation pathway, or by oxidation via cytochrome P450 (CYP) in the
liver, and is excreted in the urine [12, 15, 16]. This series of metabolic steps can be altered by
uremic compounds that downregulate CYPs, or uremic toxins that act as competitive inhibi-
tors interfering with CYP enzyme activity [4, 12]. In addition, the PTH level increases, and the
inflammatory response is activated in ESKD patients [17]. Michaud et al. demonstrated that
PTH and inflammatory cytokines are the main factors downregulating CYP in rats [5]. As a
result, the propofol dose requirement in patients with ESKD can be difficult to predict and is
expected to be different from that in patients with normal kidney function.
Among the serum proteins, propofol binds strongly to albumin and hemoglobin [16–18].
The protein binding of drugs changes in ESKD patients due to the effect of uremic substances,
which act as protein binding inhibitors [12]. Moreover, anemia caused by a relative deficiency
of erythropoietin is a common complication in ESKD patients [19]. The dosage of propofol
should theoretically be reduced in patients with uremic symptoms and low levels of albumin
or hemoglobin, as the fraction of free propofol increases.
Gasperi et al. [20] determined that the mean blood propofol concentration in ESKD
patients was significantly lower than that in normal controls when inducing anesthesia with a
bolus of 2 mg/kg propofol. Goyal et al. [9] concluded that the appropriate induction dose of
propofol is higher in ESKD patients due to hyperdynamic circulation caused by anemia.
However, Ickx et al. [6] reported that the mean propofol blood concentration in ESKD
patients did not differ significantly from that in a normal control group during slow induction
by continuous infusion. Our results are similar to those of Ickx et al. In our study, Although
the Ce was not different in ESKD patients, the Ce50 was lower by 0.81 μg/ml, which although
not statistically significant is likely a notable clinical effect. The reason for the lack of difference
in the Ce of ESKD patients may be the slow infusion of propofol, which could have interfered
with the hyperdynamic circulation.
There are several limitations in this study. First, although it has proper power for compari-
son, our study has small sample sizes. Second, weight, height, and BMI in ESKD patients were
lower than those of patients with normal kidney function, and the ESKD-CEB group was older
than the control group. The pharmacokinetics of propofol could be influenced by those. But
we used TCI pump using the Schnider model. Schnider model include these parameters as
covariates and improved the performance [10, 21]. Therefore, we presumed that the influence
of the differences may be accounted.
Neuraxial anesthesia-induced sensory deafferentation reduces the demand for inhalation
and intravenous anesthetics [22, 23]. Several studies have shown that epidural blockade
reduces the propofol induction requirement [24, 25]. In this small study, we identified a
decrease in the Ce and Ce50 values of ESKD patients who received CEB. However, there were
no statistical differences between the two ESKD groups.
In conclusion, in our study, the Ce50 values were 3.75 and 3.21 μg/ml for the ESKD and
ESKD-CEB groups, respectively. These values were lower 0.81 and 1.35 μg/ml compared with
control group (4.56 μg/ml). Therefore, when inducing anesthesia in ESKD patients, we
PLOS ONE | https://doi.org/10.1371/journal.pone.0254520 August 12, 2021
7 / 9
PLOS ONE Potency of propofol for inducing LOC in ESKD patients
recommend using an initial dose similar to that of patients with normal kidney function, or
rather starting with a lower dose.
Supporting information
S1 Appendix. Control file of the loss of consciousness pharmacodynamic model.
(DOCX)
Author Contributions
Conceptualization: Mi Roung Jun, Sang Ho Kim.
Data curation: Ho Bum Cho, Sanghoon Song.
Formal analysis: Ki Seob Han, Ji Eun Park, Sun Young Park.
Funding acquisition: Sun Young Park.
Investigation: Jae Hwa Yoo.
Supervision: Ji Won Chung.
Writing – original draft: Mi Roung Jun, Mun Gyu Kim.
Writing – review & editing: Mun Gyu Kim, Sun Young Park.
References
1. Dinis-Oliveira RJ. Metabolic Profiles of Propofol and Fospropofol: Clinical and Forensic Interpretative
Aspects. BioMed research international. 2018; 2018:6852857. https://doi.org/10.1155/2018/6852857
PMID: 29992157
2. Mastroianni G, Iannone LF, Roberti R, Gasparini S, Ascoli M, Cianci V, et al. Management of status epi-
lepticus in patients with liver or kidney disease: a narrative review. Expert Rev Neurother. 2020:1–14.
https://doi.org/10.1080/14737175.2021.1862649 PMID: 33297776
3. Abbasi MA, Chertow GM, Hall YN. End-stage renal disease. BMJ Clin Evid. 2010;2010.
4. Naud J, Nolin TD, Leblond FA, Pichette V. Current understanding of drug disposition in kidney disease.
J Clin Pharmacol. 2012; 52(1 Suppl):10S–22S. https://doi.org/10.1177/0091270011413588 PMID:
22232747
5. Michaud J, Naud J, Chouinard J, Desy F, Leblond FA, Desbiens K, et al. Role of parathyroid hormone in
the downregulation of liver cytochrome P450 in chronic renal failure. J Am Soc Nephrol. 2006; 17
(11):3041–8. https://doi.org/10.1681/ASN.2006010035 PMID: 17021269
6.
Ickx B, Cockshott ID, Barvais L, Byttebier G, De Pauw L, Vandesteene A, et al. Propofol infusion for
induction and maintenance of anaesthesia in patients with end-stage renal disease. Br J Anaesth.
1998; 81(6):854–60. https://doi.org/10.1093/bja/81.6.854 PMID: 10211008
7. Nathan N, Debord J, Narcisse F, Dupuis JL, Lagarde M, Benevent D, et al. Pharmacokinetics of propo-
fol and its conjugates after continuous infusion in normal and in renal failure patients: a preliminary
study. Acta Anaesthesiol Belg. 1993; 44(3):77–85. PMID: 8310787
8. Kirvela¨ M, Olkkola KT, Rosenberg PH, Yli-Hankala A, Salmela K, Lindgren L. Pharmacokinetics of pro-
pofol and haemodynamic changes during induction of anaesthesia in uraemic patients. Br J Anaesth.
1992; 68(2):178–82. https://doi.org/10.1093/bja/68.2.178 PMID: 1540461
9. Goyal P, Puri GD, Pandey CK, Srivastva S. Evaluation of induction doses of propofol: comparison
between endstage renal disease and normal renal function patients. Anaesth Intensive Care. 2002; 30
(5):584–7. https://doi.org/10.1177/0310057X0203000506 PMID: 12413256
10. Schnider TW, Minto CF, Gambus PL, Andresen C, Goodale DB, Shafer SL, et al. The influence of
method of administration and covariates on the pharmacokinetics of propofol in adult volunteers. Anes-
thesiology. 1998; 88(5):1170–82. https://doi.org/10.1097/00000542-199805000-00006 PMID: 9605675
11.
Levey AS, Coresh J. Chronic kidney disease. Lancet. 2012; 379(9811):165–80. https://doi.org/10.1016/
S0140-6736(11)60178-5 PMID: 21840587
PLOS ONE | https://doi.org/10.1371/journal.pone.0254520 August 12, 2021
8 / 9
PLOS ONE Potency of propofol for inducing LOC in ESKD patients
12. Yeung CK, Shen DD, Thummel KE, Himmelfarb J. Effects of chronic kidney disease and uremia on
hepatic drug metabolism and transport. Kidney Int. 2014; 85(3):522–8. https://doi.org/10.1038/ki.2013.
399 PMID: 24132209
13. Glorieux G, Schepers E, Vanholder RC. Uremic toxins in chronic renal failure. Prilozi. 2007; 28(1):173–
204. PMID: 17932467
14. Meyer TW, Hostetter TH. Uremia. N Engl J Med. 2007; 357(13):1316–25.
15. Oda Y, Hamaoka N, Hiroi T, Imaoka S, Hase I, Tanaka K, et al. Involvement of human liver cytochrome
P4502B6 in the metabolism of propofol. Br J Clin Pharmacol. 2001; 51(3):281–5. https://doi.org/10.
1046/j.1365-2125.2001.00344.x PMID: 11298076
16. Bryson HM, Fulton BR, Faulds D. Propofol. An update of its use in anaesthesia and conscious sedation.
Drugs. 1995; 50(3):513–59. https://doi.org/10.2165/00003495-199550030-00008 PMID: 8521772
17. Bistrian BR. Role of the systemic inflammatory response syndrome in the development of protein-calo-
rie malnutrition in ESRD. Am J Kidney Dis. 1998; 32(6 Suppl 4):S113–7. https://doi.org/10.1016/s0272-
6386(98)70173-8 PMID: 9892377
18. Altmayer P, Bu¨ch U, Bu¨ch HP. Propofol binding to human blood proteins. Arzneimittelforschung. 1995;
45(10):1053–6. PMID: 8595056
19.
20.
Fishbane S, Spinowitz B. Update on Anemia in ESRD and Earlier Stages of CKD: Core Curriculum
2018. Am J Kidney Dis. 2018; 71(3):423–35. https://doi.org/10.1053/j.ajkd.2017.09.026 PMID:
29336855
de Gasperi A, Mazza E, Noè L, Corti A, Cristalli A, Prosperi M, et al. Pharmacokinetic profile of the
induction dose of propofol in chronic renal failure patients undergoing renal transplantation. Minerva
Anestesiol. 1996; 62(1–2):25–31. PMID: 8768021
21. Schnider TW, Minto CF, Shafer SL, Gambus PL, Andresen C, Goodale DB, et al. The influence of age
on propofol pharmacodynamics. Anesthesiology. 1999; 90(6):1502–16. https://doi.org/10.1097/
00000542-199906000-00003 PMID: 10360845
22. Hodgson PS, Liu SS, Gras TW. Does epidural anesthesia have general anesthetic effects? A prospec-
tive, randomized, double-blind, placebo-controlled trial. Anesthesiology. 1999; 91(6):1687–92. https://
doi.org/10.1097/00000542-199912000-00021 PMID: 10598611
23.
Ishiyama T, Kashimoto S, Oguchi T, Yamaguchi T, Okuyama K, Kumazawa T. Epidural ropivacaine
anesthesia decreases the bispectral index during the awake phase and sevoflurane general anesthesia.
Anesth Analg. 2005; 100(3):728–32, table of contents. https://doi.org/10.1213/01.ANE.0000159868.
06847.47 PMID: 15728060
24. Agarwal A, Pandey R, Dhiraaj S, Singh PK, Raza M, Pandey CK, et al. The effect of epidural bupiva-
caine on induction and maintenance doses of propofol (evaluated by bispectral index) and maintenance
doses of fentanyl and vecuronium. Anesth Analg. 2004; 99(6):1684–8, table of contents. https://doi.org/
10.1213/01.ANE.0000136422.70531.5A PMID: 15562054
25. Senturk M, Gucyetmez B, Ozkan-Seyhan T, Karadeniz M, Dincer S, Akpir D, et al. Comparison of the
effects of thoracic and lumbar epidural anaesthesia on induction and maintenance doses of propofol
during total i.v. anaesthesia. Br J Anaesth. 2008; 101(2):255–60. https://doi.org/10.1093/bja/aen160
PMID: 18534970
PLOS ONE | https://doi.org/10.1371/journal.pone.0254520 August 12, 2021
9 / 9
PLOS ONE |
10.1371_journal.pone.0255730 | RESEARCH ARTICLE
Glycemic profile and associated factors in
indigenous Munduruku, Amazonas
Hanna Lorena Moraes GomesID
Oliveira Cordeiro1, Zilmar Augusto de Souza Filho1, Noeli das Neves Toledo1, Evelyne
Marie Therese MainbourgID
2, Anto´ nio Manuel Sousa3, Gilsirene Scantelbury de Almeida1
1*, Neuliane Melo Sombra1, Eliza Dayanne de
1 Manaus School of Nursing, Federal University of Amazonas, Manaus, Brazil, 2 Leoˆ nidas & Maria Deane
Institute / FIOCRUZ Amazoˆ nia, Manaus, Brazil, 3 Amazonas State University, Manaus, Brazil
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
* hannahlorena.mg@gmail.com
Abstract
Objective
OPEN ACCESS
Citation: Gomes HLM, Sombra NM, Cordeiro
EDdO, Filho ZAdS, Toledo NdN, Mainbourg EMT, et
al. (2021) Glycemic profile and associated factors
in indigenous Munduruku, Amazonas. PLoS ONE
16(9): e0255730. https://doi.org/10.1371/journal.
pone.0255730
Editor: Fernando Guerrero-Romero, Mexican
Social Security Institute, MEXICO
Received: January 9, 2021
Accepted: July 22, 2021
Published: September 3, 2021
Peer Review History: PLOS recognizes the
benefits of transparency in the peer review
process; therefore, we enable the publication of
all of the content of peer review and author
responses alongside final, published articles. The
editorial history of this article is available here:
https://doi.org/10.1371/journal.pone.0255730
Copyright: © 2021 Gomes et al. This is an open
access article distributed under the terms of the
Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: All relevant data are
within the manuscript.
Funding: This study received funding approved by
the National Council for Scientific and
To evaluate the glycemic profile and its association with sociodemographic, anthropometric,
clinical and lifestyle factors of Munduruku indigenous people.
Method
Cross-sectional study with a quantitative and analytical approach, a total of 459 indigenous
people (57.1% men, aged 36.3 ± 14.7 years old) belonging to the Munduruku ethnic group
from the Kwata´ -Laranjal Indigenous Land, in Amazonas, Brazil, were selected by probabilis-
tic sampling in all households in the four most populous villages. Sociodemographic and
anthropometric variables, blood pressure levels and lipid profile were evaluated. Fasting
capillary blood glucose was measured with a digital device. The associations were
assessed by multinomial logistic regression, and p-values�0.05 were considered
significant.
Results
For pre-diabetes, prevalence was 74.3% and, for diabetes, 12.2%. The variables associated
with the risk for pre-diabetes were the following: age (OR = 1.03; 95% CI = 1.00 – 1.06) and
obesity (OR = 9.69; 95% CI = 1.28 – 73.58). The positive associations indicating risk for dia-
betes were as follows: age (OR = 1.05; 95% CI = 1.03 – 1.08), overweight (OR = 4.17; 95%
CI = 1.69 – 10.32) and obesity (OR = 35.26; 95% CI = 4.12 – 302.08).
Conclusions
The risks associated with pre-diabetes and diabetes among the Munduruku indigenous peo-
ple revealed a worrying index. It is necessary to consider changes in eating habits and life-
style, as well as possible environmental and social changes that can affect this and other
groups, with emphasis on those who live in vulnerable conditions.
PLOS ONE | https://doi.org/10.1371/journal.pone.0255730 September 3, 2021
1 / 16
PLOS ONE Technological Development (CNPq) (Proc. 424053
/ 2016-0) and with funding from the Scientific
Article Publication Support Program (PAPAC) and
the Post Support Program -Graduation
(PROSGRAD), both of these are programs of the
Amazonas Research Support Foundation -
FAPEAM. Funders had no role in the study design,
data collection, analysis, decision to publish or
preparation of the manuscript.
Competing interests: The authors have declared
that no competing interests exist.
Glycemic profile and associated factors in indigenous Munduruku, Amazonas.
Introduction
The changes in the globalized world, as a result of the urbanization and industrialization pro-
cesses, brought about changes in habits and lifestyles, contributing to the increase of chronic
non-communicable diseases, among which we can highlight cardiovascular diseases. These
same impacts permeate the indigenous populations, through transitions in life, economic and
sociocultural habits, and in their own lifestyle [1, 2].
The destruction of the ecosystems that the Brazilian Indigenous Lands are facing, together
with the acceleration of the urbanization process, sedentary lifestyle, changes in the diet, obe-
sity and easy access to cities, contribute significantly to the transformations of the daily lives of
indigenous populations, leaving them more vulnerable to certain diseases, which contributes
to the increase of Chronic Non-communicable Diseases (CNCDs) [3, 4].
Social indicators of a national scope classify the North Region as belonging to Class “E” of
social vulnerability, as it consists of extensive rural areas, low demographic density, with a very
low human development index, precarious access to treated water, sewage and electricity,
among other negative results. Compared with the South and Southeast regions of the country,
the North has less capacity to respond to health problems, in terms of Health Care Network
structure [5].
Deaths due to non-communicable diseases (NCDs) represented the highest percentage:
73.4% (95% uncertainty interval [UI] = 72.5 – 74.1) in 2017. In relation to 2007, there was a
22.7% (21.5 – 23.9) increase, equivalent to 7.21 million (7.20 – 8.01) of estimated additional
deaths. There was a major increase in years of life lost due to neoplasms and cardiovascular
diseases.
In the general population, cardiovascular diseases (CVDs) are part of the group of main
causes of mortality. In 2016, approximately 17.9 million people died due to CVDs worldwide.
From this perspective, diabetes mellitus (DM) stands out as a highly prevalent health problem
and one of the main risk factors for CVDs [6–8].
DM is configured as a "metabolic disorder" characterized by persistent hyperglycemia,
resulting from a deficit in the production of insulin or in its action, or even in both mecha-
nisms, leading to long-term complications” (SBD, pg. 19). Data from the International Diabe-
tes Federation point out that, in the world, 8% of adults lived with DM in 2017. DM is a
growing and important health problem that affects the population of all countries, being
responsible for 4 million deaths worldwide in a single year [9, 10].
It is believed that changes in the social, economic and political scopes of indigenous Brazi-
lians may have favored changes in their lifestyle and in their epidemiological profile [1]. In the
Brazilian indigenous population, the first cases of DM began to be investigated from the 1970s,
when the prevalence of diabetes was non-existent [1].
In the state of Mato Grosso do Sul, several studies were carried out with the Terena, Gua-
rani and Kaiowa´ indigenous peoples, where it was found that 4.5% had DM in 2007 and 2008
[11]. Another two studies carried out in the same population found a prevalence rate of 5.8%
in the period from 2009 to 2011, and of 4.5% in 2008 and 2009 [12]. In 2013, among 385 Ter-
ena and Guarani women from the same region, 7% presented altered capillary glycaemia sug-
gestive of DM [13]. Among the Guarani and Tupinikin (ES), in 2003 and 2004, the prevalence
of DM was 4.5% [14]. In Khisêdjê in 2010 and 2011, prevalence was 3.8% [15]. The highest
prevalence rate of DM among indigenous people in Brazil was found among the Xavante in
the state of Mato Grosso (n = 948): 25.9% [16].
The data presented show that diabetes has been growing in indigenous populations [17]
and that is worsened by the increased consumption of industrialized food products, social
problems linked to the economy and the increasingly frequent contact with the non-
PLOS ONE | https://doi.org/10.1371/journal.pone.0255730 September 3, 2021
2 / 16
PLOS ONE Glycemic profile and associated factors in indigenous Munduruku, Amazonas.
indigenous population [1, 17]. Considering that most of the studies refer to ethnicities in the
Brazilian Midwest Region, the objective of the study was to assess the glycemic profile and its
association with sociodemographic, anthropometric, clinical and lifestyle factors of Mundur-
uku indigenous people from the state of Amazonas, Brazilian North Region.
Method
Study locus and population
The study was carried out in the Kwata´-Laranjal indigenous land (Fig 1), located in the munic-
ipality of Borba, state of Amazonas, in the Brazilian North Region. The study population con-
sisted of individuals from the Munduruku ethnic group who live in the villages of Laranjal,
Mucaja´, Kwata´ and Fronteira, members of the Kwata´-Laranjal Indigenous Land, aged between
18 and 80 years old, and of both genders. According to population data, released by the Special
Indigenous Sanitary District of Manaus in 2018, the total population over 18 years old of both
genders living in these four villages consisted in 635 inhabitants, divided as follows: 195 in
Mucaja´, 118 in Laranjal, 186 in Kwata´ and 136 in Fronteira.
Fig 1. Geographic location of the Kwata´-Laranjal Indigenous Land.
https://doi.org/10.1371/journal.pone.0255730.g001
PLOS ONE | https://doi.org/10.1371/journal.pone.0255730 September 3, 2021
3 / 16
PLOS ONE Glycemic profile and associated factors in indigenous Munduruku, Amazonas.
Study participants
The following was accepted for sample calculation: 50.0% proportion of the indigenous popu-
lation and the prevalence values of diabetes pointed out by the Guidelines of the Brazilian Dia-
betes Society and by the study by Soares et al. [9, 18]. The error margin adopted was 5%, 95%
confidence interval, and 10% for losses. The sample consisted of 459 individuals belonging to
the Munduruku ethnicity, from the villages of Mucaja´ (n = 129), Laranjal (n = 93), Kwata´
(n = 136) and Fronteira (n = 101).
The four most populous villages in the Kwata´-Laranjal Indigenous Land (Mucaja´, Laranjal,
Kwata´ and Fronteira) were chosen. Probabilistic sampling of individuals per household was
carried out, in which all members had an equal chance of participating in the study. The study
included indigenous people belonging to the Munduruku ethnic group, aged � 18 years old
and living in the selected villages. It is noted that all the Munduruku indigenous people drawn
to participate in this study were able to fluently communicate in the Portuguese language.
Only those who were ill and pregnant were excluded from the sample.
Data collection
Before starting data collection in the Kwata´-Laranjal Indigenous Land, the team of women
researchers visited the four villages included in this study, which allowed for previous contact
with the local indigenous leaders, closer contact with the health professionals who served in
those villages, and holding a meeting with the indigenous people to present the research objec-
tives and method.
For the data collection stage, the team underwent specific training in order to standardize
the procedures for: measuring blood glucose and capillary lipids after fasting for a minimum
of eight hours, measuring blood pressure, taking anthropometric measurements and conduct-
ing the interview.
At the beginning of data collection, the residents were invited again to be informed about
how the participants would be selected and the procedures for data collection. For each house-
hold, the research participants were selected by means of a draw. The indigenous health agent
assisted the team in locating the homes of the selected participants. The guidelines for data col-
lection were given the day before, with reinforcement regarding the location, day, time and,
mainly, the need for at least 8-hour fasting.
The collection of the anthropometric data, blood pressure, blood glucose and lipids was
always performed at dawn. Before starting the collection of blood drops from the digit pulp,
the indigenous people were asked at what time they had their last meal. Those who reported
breaking the fast were rescheduled for the following day and re-oriented.
In relation to the tests of capillary blood glucose and lipid levels, the equipment used were
as follows: Active portable digital device from the Accu-Chek1 manufacturer for the measure-
ment of capillary blood glucose and the Accutrend1 Plus device for the measurement of cho-
lesterol and triglycerides, both manufactured by Roche Diagno´stica, with their respective test
strips (Accutrend1 Cholesterol and Accutrend1 Triglycerides). The cut-off points used to
assess and classify fasting capillary glucose were as follows: normal < 100 mg/dL, pre-
diabetes � 100 mg/dL and < 126 mg/dL and diabetes � 126 mg/dL [9]. For the lipid levels, the
classification was the following: hypercholesterolemia when � 240 mg/dL and hypertriglyceri-
demia when � 175 mg/dL [19].
The following was used for the evaluation of the anthropometric measures: digital bioimpe-
dance scale (OMRON HBF-514C), portable stadiometer (ALTURA EXATA) and inelastic
measuring tape. The neck circumference measurement was taken at the smallest neck circum-
ference, just above the laryngeal prominence. The waist circumference measurement was
PLOS ONE | https://doi.org/10.1371/journal.pone.0255730 September 3, 2021
4 / 16
PLOS ONE Glycemic profile and associated factors in indigenous Munduruku, Amazonas.
taken at the midpoint between the last rib and the lateral iliac crest, around the narrowest part
of the trunk. The taper index was determined, according to its definition, from the measure-
ments of weight, height and waist circumference. Both BMI and Body Fat Percentage were
assessed using the bioimpedance technique.
The cut-off points adopted to classify neck circumference measurements were as follows: �
37 cm for men and � 34 cm for women; and those for waist circumference were: � 102 cm for
men and � 88 cm for women [20]. For the taper index, the adopted values were: � 1.25 for
men and � 1.18 for women. As for the Body Mass Index (BMI), it was classified as: low weight
(< 18.5 kg/m2), normal weight (18.5 kg/m2-24.9 kg/m2), overweight (25.0 kg/m2-29.9 kg/m2)
and obesity (� 30.0 kg/m2) [21]. The classification of body fat percentage considered the fol-
lowing stratification by age group and gender: low (< 8.0%-< 13.0% for men and < 21.0%-<
30.0% for women), normal (13.0%-24.9% for men and 30.0%-� 35.9% for women) and high
(� 25.0% for men and � 36.0% for women).
Blood pressure levels were measured on the left arm, using an automatic professional blood
pressure monitor (OMRON/Model HBP-1100), properly calibrated. The procedures to per-
form the measurement and classification of blood pressure were conducted according to the
Brazilian Hypertension Directive. The following cut-off points were considered: pre-hyperten-
sion when systolic blood pressure levels are between 140 mmHg and 159 mmHg and/or when
the diastolic blood pressure is between 90 mmHg and 99 mmHg; hypertension when the value
is � 180 mmHg in systolic pressure and/or � 110 mmHg in diastolic pressure. Alternatively,
hypertension could be self-reported, if the indigenous participants reported having been diag-
nosed with hypertension by a physician or if they were taking some antihypertensive medica-
tion, regardless of the blood pressure values measured in the interview [22].
For the assessment of lifestyle, the level of physical activity was investigated using the IPAC
(International Physical Activity Questionnaire), in its short version, an instrument validated
with translation into the Portuguese language. The IPAQ allows quantifying the total minutes
spent in weekly physical activities and surveying the distribution of time by intensity of the
physical activity practiced. The level of physical activity was classified according to the instru-
ment. To assess the intake of alcoholic beverages, the Alcohol Use Disorder Identification Test
(AUDIT) questionnaire was used, allowing the identification of risk and harmful consumption
and of probable dependence on alcohol in the past 12 months.
A form consisting of closed questions related to the following variables was applied: gender,
age, marital status, schooling, paid work, social benefit received, monthly family income, self-
reported hypertension and/or consumption of antihypertensive medications, smoking, level of
physical activity, alcohol consumption and family history of cardiovascular diseases.
The participants who presented changes in capillary glycaemia, triglycerides, total choles-
terol or/and blood pressure, as well as those who were obese were referred directly to the care
provided by the health team at the Base Center (reference health unit, belonging to the Indige-
nous Health Sub-System) for evaluation and monitoring. For the changes in the anthropomet-
ric markers, this information was passed on to the health professionals working in the
respective Base Center.
Statistical analysis
The analysis of the data collected was performed by means of the R software, version 3.5.1. The
Kolmogorov-Smirnov test was used to verify normal distribution of the data. In this way, the
continuous variables were presented using means and standard deviations; and the categorical
variables, with absolute and relative frequencies. For the continuous variables, the Kruskal-
Wallis test was used; and for categorical ones, Fisher’s Exact test. The significance level was set
PLOS ONE | https://doi.org/10.1371/journal.pone.0255730 September 3, 2021
5 / 16
PLOS ONE Glycemic profile and associated factors in indigenous Munduruku, Amazonas.
at 5%. The Wald test was used for the multinomial logistic regression analysis. To verify the
association between the dependent variables (pre-diabetes and diabetes) and the independent
variables of the study, Odds Ratios (ORs) were estimated based on the multinomial regression
model and the respective 95% confidence interval (CI). For this being a multifactorial phe-
nomenon, the independent variables were grouped in blocks (sociodemographic, lifestyle and
anthropometric and clinical factors) and analyzed hierarchically.
Ethical aspects
The data were collected from August to September 2018, after the consent of the leaders of the
Kwata´-Laranjal Indigenous Land, approval by the National Research Ethics Commission
(CAAE 74361617.2.0000.5020), and authorization for entry into indigenous lands of the Min-
istry of Justice National Indian Foundation (43/AAEP/PRES/2018). All the indigenous people
who agreed to participate in the study signed the Free and Informed Consent Form.
Results
As shown in Table 1, the profile of the glycemic levels of the 459 indigenous Munduruku indi-
cates that 86.5% had high serum levels of fasting capillary glycaemia, with 74.3% being sugges-
tive of pre-diabetes and 12.2% of diabetes.
As for the sociodemographic factors, it was observed that 57.1% were men, with a mean age
of 36.6 years old, most with a partner, and 9.6% not having any schooling level. A little over
half of them were unemployed and 61.7% received some social benefit from the Brazilian fed-
eral government. In this way, most of the Munduruku indigenous people had a monthly family
income of up to US$ 470.67.
The general anthropometry assessment allowed identifying that the indigenous people had
high mean values of neck circumference, waist circumference and taper index. The mean BMI
indicated excess weight, in addition to the majority presenting high body fat percentages.
In relation to the clinical factors of the Munduruku indigenous people, the mean pressure
levels indicated normality, but 10.2% presented high levels of systolic and diastolic blood pres-
sure, suggestive of hypertension. Regarding the serum triglyceride levels, the indigenous popu-
lation presented a high mean value but, for total cholesterol, the mean remained within
normal limits.
Regarding the indigenous people’s lifestyle, there was a high prevalence of alcohol con-
sumption (71.2%) and smoking (54.2%), as well as a low prevalence of sedentary lifestyle
(7.6%). It is worth mentioning that most of the indigenous people reported having a family his-
tory of hypertension and diabetes.
Table 1 also shows that the group of Munduruku indigenous people with diabetes presented
statistically significant differences when compared to the other groups, in greater proportion
having some paid work and, in a smaller proportion, receiving some social benefit. The group
of diabetics presents higher values regarding age, mean in the anthropometric markers, preva-
lence of obesity and body fat, prevalence of pre-hypertension and hypertension, mean of tri-
glycerides and total cholesterol, as well as family members with diabetes or/and stroke.
Table 2 shows the unadjusted multinomial logistic regression model. The association of
pre-diabetes with age showed that, for every one-year-old increase in the age of the indigenous
Munduruku, their chance of becoming pre-diabetic increases by 4%. It is also worth noting
that the indigenous people without a partner had a lower risk of being pre-diabetic (OR = 0.55
[95% CI = 0.32 – 0.96]).
As for the association of pre-diabetes with the anthropometric factors, it was observed that,
with a one-centimeter increase in waist circumference (OR = 1.07 [95% CI = 1.03 – 1.10]), in
PLOS ONE | https://doi.org/10.1371/journal.pone.0255730 September 3, 2021
6 / 16
PLOS ONE Table 1. Categorization of the glycemic profile of indigenous Munduruku according to the sociodemographic and anthropometric variables, clinical factors, habits
and lifestyle, and family history.
Glycemic profile and associated factors in indigenous Munduruku, Amazonas.
Variables
Sociodemographic Factors
Gender
Female
Male
Age (years old), mean (SD)
Marital Status
Has a partner
No partner
Schooling
Illiterate
Elementary School
High School
Higher Education or Postgraduate
Paid work
Yes
No
Social benefit
Yes
No
Monthly family income (minimum wagea)
Does not have
<1 minimum wage (US$ 235.34)
1 - 2 minimum wages (US$ 235.35 – US$ 470.67)
3 - 4 minimum wages (US$ 706.01 – US$ 941.35)
� 5 minimum wages (US$ 1,176)
Anthropometric Factors
Neck circumference (cm), mean (SD)
Waist circumference (cm), mean (SD)
Taper index, mean (SD)
BMI (kg/m2), mean (SD)
BMI classification
Low weight (< 18.5 kg/m2)
Normal weight (18.5–24.9 kg/m2)
Overweight (25.0–29.9 kg/m2)
Obesity (�30 kg/m2)
Body fat classification
Low (<8.0%-<13.0% men/<21.0%-<30.0% women)
Normal (13.0%-24.9% men/30.0%-�35.9% women)
High (�25.0% men; �36.0% women)
Clinical Factors
Systolic blood pressure, SBP (mmHg), mean (SD)
Diastolic blood pressure, DBP (mmHg), mean (SD)
Blood pressure classification
Normal
N (%)
62 (13.5)
Glycemic Profile
Pre-diabetes
N (%)
337 (74.3)
Diabetes
N (%)
60 (12.2)
Total
N (%)
459 (100)
22 (35.5)
40 (64.5)
147 (43.6)
190 (56.4)
28 (46.7)
32 (53.3)
197 (42.9)
262 (57.1)
30.2 (±11.2)
36.5 (±14.8)
44.1 (±14.0)
36.6 (±14.7)
35 (56.5)
27 (43.5)
3 (4.8)
19 (30.6)
30 (48.4)
10 (16.1)
22 (35.5)
40 (64.5)
43 (69.4)
19 (30.6)
21 (26.2)
30 (37.5)
22 (27.5)
6 (7.5)
1 (1.3)
35.5 (±3.3)
79.5 (±7.9)
1.20 (±0.08)
23.6 (±2.8)
1 (1.6)
42 (67.7)
18(29.0)
1 (1.6)
1 (1.6)
35 (56.5)
26 (41.9)
236 (70.0)
101 (30.0)
30 (8.9)
134 (39.8)
128 (38.0)
45 (13.4)
138 (40.9)
199 (59.1)
211 (62.6)
126 (37.4)
7 (2.2)
134 (41.7)
115 (35.8)
49 (15.3)
16 (5.0)
36.2 (±3.3)
85.1 (±10.1)
1.24 (±0.09)
25.7 (±4.0)
4 (1.2)
158 (46.9)
127 (37.7)
48 (14.2)
7/337 (2.1)
126 (37.4)
204 (60.5)
41 (68.3)
19 (31.7)
11 (18.3)
22 (36.7)
19 (31.7)
8 (13.3)
34 (56.7)
26 (43.3)
29 (48.3)
31 (51.7)
1 (1.7)
21 (36.2)
24 (41.4)
11 (19.0)
1 (1.7)
312 (68.0)
147 (32.0)
44 (9.6)
175 (38.1)
177 (38.6)
63 (13.7)
194 (42.3)
265 (57.7)
283 (61.7)
176 (38.3)
29 (6.3)
185 (40.3)
161 (35.1)
66 (14.4)
18 (3.9)
37.7 (±3.2)
92.2 (±8.7)
1.29 (±0.07)
28.0 (±3.6)
36.3 (±3.3)
85.3 (±10.2)
1.24 (±0.09)
25.8 (±4.0)
0 (0)
12 (20.0)
31 (51.7)
17 (28.3)
0 (0)
7 (11.7)
53 (88.3)
5 (1.1)
212 (46.2)
176 (38.3)
66 (14.4)
8 (1.7)
168 (36.6)
283 (61.7)
p-value
0.404
<0.001
0.109
0.116
0.039
0.045
0.708
0.001
<0.001
<0.001
<0.001
<0.001
< 0.001
<0.001
110.0 (±12.2)
63.6 (±8.2)
113.6 (±15.0)
66.5 (±8.4)
121.2 (±16.9)
70.4 (±8.8)
114.1 (±15.2)
66.6 (±8.6)
<0.001
<0.001
0.001
(Continued )
PLOS ONE | https://doi.org/10.1371/journal.pone.0255730 September 3, 2021
7 / 16
PLOS ONE Table 1. (Continued)
Variables
Normal (SBP of �120–129 mmHg/DBP �80–84 mmHg)
Pre-hypertension (SBP of �130 mmHg-139 mmHg/BPD �80–89 mmHg)
Hypertension (SBP of �140 mmHg/DBP �90 mmHg)
Triglycerides (mg/dL)
Total cholesterol (mg/dL)
Lifestyle
Smoker
Yes
No
Level of physical activity
Sedentary
Irregularly active
Active
Very active
Alcohol Consumption
Low risk consumption
Risk intake, harmful or probable dependence
Family history
Hypertension
Diabetes
Stroke
Glycemic profile and associated factors in indigenous Munduruku, Amazonas.
Normal
N (%)
62 (13.5)
59 (95.2)
1 (1.6)
2 (3.2)
Glycemic Profile
Pre-diabetes
N (%)
337 (74.3)
294 (87.2)
12 (3.6)
31 (9.2)
Diabetes
N (%)
60 (12.2)
40 (66.7)
6 (10.0)
14 (23.3)
Total
N (%)
459 (100)
393 (85.6)
19 (4.1)
47 (10.2)
p-value
131.9 (±65.8)
171.3 (±25.5)
149.3 (±86.3)
176.5 (±32.3)
206.8 (±124.1)
189.7 (±35.4)
154.5 (±92.1)
177.5 (±32.2)
<0.001
0.003
36 (58.1)
26 (41.9)
2 (3.2)
13 (21.0)
21 (33.9)
26 (41.9)
7 (25.9)
20 (74.1)
43 (84.3)
31 (63.3)
20 (45.5)
184 (54.6)
153 (45.4)
26 (7.7)
92 (27.3)
114 (33.8)
105 (31.2)
35 (30.2)
81 (69.8)
211 (77.0)
161 (61.7)
75 (31.6)
29 (48.3)
31 (51.7)
7 (11.7)
18 (30.0)
24 (40.0)
11 (18.3)
4 (23.5)
13 (76.5)
43 (86.0)
38 (82.6)
19 (52.8)
249 (54.2)
210 (45.8)
35 (7.6)
123 (26.8)
159 (34.6)
142 (30.9)
46 (28.8)
114 (71.2)
297 (79.2)
230 (64.6)
114 (36.0)
0.542
0.125
0.800
0.222
0.023
0.018
Kwata´-Laranjal Indigenous Land, Borba, Amazonas, Brazil, 2018.
a Current minimum wage of R$ 954.00, equivalent to approximately US$ 235.34 in August 2018.
https://doi.org/10.1371/journal.pone.0255730.t001
the taper index (OR = 1.06 [95% CI = 1.02 – 1.09]) and in the BMI (OR = 1.20 [95%
CI = 1.10 – 1.32]), the indigenous people have a chances to develop pre-diabetes of 7%, 6%
and 20%, respectively. Excess weight among the indigenous people also presented an associa-
tion with pre-diabetes, since the chance of the indigenous person who presented overweight to
become pre-diabetic is 87%; and, among those who were obese, the chance becomes 12 times
greater (OR = 12.76 [95% CI = 1.71 – 95.26]). For the indigenous people with high body fat,
the risk of becoming pre-diabetics also increases the chance, but two-fold (OR = 2.18 [95%
CI = 1.25 – 3.79]).
The unadjusted analysis also indicated the association of diabetes with age, schooling, paid
work and any social benefits received. All the anthropometric variables were associated with
diabetes among the indigenous people. It is worth noting that, among the Munduruku indige-
nous people who presented overweight (OR = 6.17 [95% CI = 2.60 – 4.64]) and obesity
(OR = 61.03 [95% CI = 7.34 – 507.08]), the chances increased significantly. The clinical factors
were also associated with diabetes, such as: pre-hypertension, hypertension and an increase in
the total serum cholesterol level. On the other hand, the fact of having a Very Active level of
physical activity (OR = 0.12 [95% CI = 0.02 – 0.68]) reduces by 88% the chance of the indige-
nous Munduruku developing diabetes.
Table 3 shows the Odds Ratio adjusted for gender and age of the variables that presented
statistical significance (p�0.05) in the analyses from Table 2, considering the two outcomes
PLOS ONE | https://doi.org/10.1371/journal.pone.0255730 September 3, 2021
8 / 16
PLOS ONE Table 2. Unadjusted odds ratio and Confidence Interval (CI) for sociodemographic and anthropometric variables, clinical factors, lifestyle and family history asso-
ciated with pre-diabetes and diabetes among the Munduruku indigenous people.
Glycemic profile and associated factors in indigenous Munduruku, Amazonas.
Variables
Sociodemographic Factors
Gender (Ref. Female)
Male
Age (years old)
Marital Status (Ref. Has a partner)
Without partner
Schooling (Ref. Illiterate)
Elementary School
High School
Higher Education or Postgraduate
Paid work (Ref. Yes)
No
Social benefits (Ref. Yes)
No
Monthly family income (Ref. Does not have)
<1 minimum wage (US$: 235.34)
1–2 minimum wages (US$: 235.35–470.67)
3–4 minimum wages (US$: 706.01–941.35)
� 5 minimum wages (US$: 1,176)
Anthropometric Factors
Neck circumference (cm)
Waist circumference (cm)
Taper index
BMI (kg/m2)
BMI classification (Ref. Low weight/Normal weight
Overweight
Obesity
Body Fat Classification (Ref. Normal)
Low
High
Clinical Factors
Systolic blood pressure
Diastolic blood pressure
Blood Pressure Classification (Ref. Normal)
Pre-hypertension
Hypertension
Triglycerides
Total cholesterol
Habits and lifestyle
Smoker (Ref. No)
Yes
Physical activity level (Ref. Sedentary)
Irregularly active
Active
Pre-Diabetes vs Normal
Gross OR
(95% CI)
0.71 (0.40–1.25)
1.04 (1.02–1.07)
0.55 (0.32–0.96)
0.71 (0.20–2.54)
0.43 (0.12–1.49)
0.45 (0.11–1.77)
0.79 (0.45–1.39)
1.35 (0.75–2.42)
1.28 (0.25–6.45)
1.49 (0.29–7.67)
2.33 (0.39–13.91)
4.57 (0.35–59.12)
1.06 (0.98–1.15)
1.07 (1.03–1.10)
1.06 (1.02–1.09)
1.20 (1.10–1.32)
1.87 (1.03–3.40)
12.76 (1.71–95.26)
1.96 (0.23–1.65)
2.18 (1.25–3.79)
1.02 (1.00–1.04)
1.04 (1.01–1.08)
2.41 (0.31–18.88)
3.11 (0.72–13.35)
1.00 (1.00–1.01)
1.01 (1.00–1.02)
1.15 (0.67–1.99)
0.54 (0.12–2.57)
0.42 (0.09–1.89)
p-value
0.235
0.002
0.037
0.593
0.182
0.253
0.420
0.311
0.768
0.631
0.353
0.245
0.164
<0.001
0.002
<0.001
0.040
0.013
0.538
0.006
0.065
0.013
0.403
0.127
0.122
0.222
0.616
0.442
0.258
Diabetes vs Normal
Gross OR
(95% CI)
0.63 (0.30–1.30)
1.07 (1.04–1.11)
0.60 (0.29–1.26)
0.32 (0.08–1.30)
0.17 (0.04–0.70)
0.22 (0.04–1.06)
0.42 (0.20–0.87)
2.42 (1.15–5.07)
1.40 (0.12–16.47)
2.18 (0.18–25.78)
3.67 (0.27–49.30)
2.00 (0.05–78.31)
1.22 (1.09–1.37)
1.14 (1.10–1.19)
1.14 (1.09–1.19)
1.38 (1.24–1.53)
6.17 (2.60–14.64)
61.03 (7.34–507.08)
-
10.19 (3.99–26.00)
1.05 (1.02–1.08)
1.10 (1.05–1.15)
8.85 (1.03–76.36)
10.32 (2.22–47.92)
1.01 (1.00–1.01)
1.02 (1.01–1.03)
1.48 (0.72–3.02)
0.40 (0.07–2.22)
0.33 (0.06–1.75)
p-value
0.211
<0.001
0.177
0.111
0.014
0.059
0.020
0.019
0.789
0.536
0.327
0.711
<0.001
<0.001
<0.001
<0.001
<0.001
<0.001
-
<0.001
<0.001
<0.001
0.047
0.003
<0.001
0.004
0.283
0.292
0.191
(Continued )
PLOS ONE | https://doi.org/10.1371/journal.pone.0255730 September 3, 2021
9 / 16
PLOS ONE Glycemic profile and associated factors in indigenous Munduruku, Amazonas.
Table 2. (Continued)
Variables
Very active
Pre-Diabetes vs Normal
Gross OR
(95% CI)
0.31 (0.07–1.39)
Consumption of Alcohol Beverages (Ref. Low risk consumption)
Risk intake, harmful or probable dependence
1.10 (0.46–2.60)
Family History
Hypertension (Ref. No)
Yes
Diabetes (Ref. No)
Yes
Stroke (Ref. No)
Yes
0.62 (0.28–1.39)
0.93 (0.50–1.76)
0.56 (0.29–1.07)
Kwata´-Laranjal Indigenous Land, Borba, Amazonas, Brazil, 2018.
https://doi.org/10.1371/journal.pone.0255730.t002
p-value
0.127
0.831
0.250
0.834
0.078
Diabetes vs Normal
Gross OR
(95% CI)
0.12 (0.02–0.68)
1.78 (0.49–6.43)
1.14 (0.38–3.43)
2.76 (1.06–7.19)
1.34 (0.55–3.24)
p-value
0.016
0.378
0.812
0.038
0.515
(pre-diabetes and diabetes). Thus, it is noteworthy that pre-diabetes was associated with
increasing age, BMI and obesity. And diabetes remained associated with increasing age, BMI,
overweight and obesity.
Discussion
The prevalence of diabetes among the Munduruku indigenous people (12.2%) was higher than
that found in other studies with indigenous populations, such as the Guarani, Kaiowa´ and Ter-
ena, from Dourados (Mato Grosso do Sul) (4.5%), Aymara, in Chile (1.5%) and was lower
when compared to the Xavante indigenous people (25.9%) from Mato Grosso and to the Pima
indigenous people from the state of Arizona (USA) [11, 16, 23, 24].
The largest participation in the study corresponded to the male gender (57.1%), unlike
studies on cardiovascular risk carried out with other indigenous populations, such as: Xavante
(49.2%) [18], Mura (42.2%) [25], Guarani-Kaiowa´ and Terena (44.2%) [11].
The mean age revealed that the Munduruku indigenous people were young adults: 36.6
years old (±14.7). A number of studies indicate that age is an important indicator for cardio-
vascular risk factors, especially for diabetes [18, 26, 27]. This study revealed that age presented
a positive and significant association with the glycemic profile and, under this perspective, a
study carried out with the Terena and Guarani indigenous peoples in 2016 also presented the
same association [13].
Table 3. Odds ratio adjusted for gender and age and confidence interval (CI) for sociodemographic and anthropometric variables, clinical factors, habits and life-
style and family history associated with pre-diabetes and diabetes among the Munduruku indigenous people.
Variables
Age (years old)
BMI (kg/m2)
Overweight
Obesity
Pre-Diabetes vs Normal
Adjusted OR (95% CI)
1.03 (1.00–1.06)
1.16 (1.06–1.28)
1.48 (0.79–2.77)
9.26 (1.22–70.45)
p-value
0.032
0.002
0.226
0.032
Diabetes vs Normal
Adjusted OR (95% CI)
1.05 (1.02–1.08)
1.28 (1.14–1.43)
4.07 (1.65–10.04)
29.14 (3.38–251.04)
p-value
0.004
<0.001
0.002
0.002
Kwata´-Laranjal Indigenous Land, Borba, Amazonas, Brazil, 2018.
https://doi.org/10.1371/journal.pone.0255730.t003
PLOS ONE | https://doi.org/10.1371/journal.pone.0255730 September 3, 2021
10 / 16
PLOS ONE Glycemic profile and associated factors in indigenous Munduruku, Amazonas.
In relation to the socioeconomic conditions, the findings show a high proportion of low-
income individuals: 46.6% with a family income of less than US$ 235.34, while 57.73% of the
participants had no paid work and 61.66% were receiving social benefits from the Brazilian
federal government. A study carried out with Mura de Autazes indigenous people (Amazonas)
also revealed that 60.2% received income from some social benefits program of the Brazilian
federal government and 59.4% had a family income of less than US$ 237.00 [25]. Another
study carried out with the Guarani-Kaiowa´ and Terena indigenous peoples from Dourados
(Mato Grosso do Sul) presented a percentage of 84.2% of families benefited by the Bolsa Famí-
lia program, highlighting the conditions of social vulnerability experienced by the group and
the possibility of social benefits improving the living conditions of the indigenous people [28].
In this context, it is worth noting that the Munduruku indigenous population presented risk
for diabetes associated with low income.
The anthropometric data presented significant differences, revealing higher mean values
among the indigenous people classified as diabetic compared to pre-diabetics and to those
with normal blood glucose.
For the Body Mass Index, the global mean revealed excess weight [25.8 (±4.0) kg/m2]
among the Munduruku indigenous people, 38.3% of them with overweight and 14% with obe-
sity. A study carried out in 2016 with the Mura de Autazes indigenous people (Amazonas),
showed excess weight, with a BMI of 26.6 (±4.7) kg/m2 [25]. Similar results were found among
the indigenous women from the municipality of Dourados (Mato Grosso do Sul), who pre-
sented a mean BMI of 27.8 (±5.0) kg/m2 [13]. When it comes to the Xavante Indigenous
Reserves of São Marcos and Sangradouro, in the municipality of Volta Grande (Mato Grosso),
the mean BMI indicates obesity among these indigenous people [30.3 (±5.1) kg/m2] [18].
Overweight and obesity are worrisome conditions, as they increase the risk of developing car-
diovascular diseases [18].
Among the Munduruku considered diabetic, the percentage of obesity was 28.3%. Flor
et al. showed that, in 2008, the percentage attributable to obesity associated with diabetes melli-
tus was, for men, 37.3% in the Brazilian North Region against 45.4% in the entire country;
and, for women, 55.1% in the North Region against 58.3% throughout Brazil, and the Brazilian
mean was higher than the mean values found in the international literature [29].
When it comes to indigenous peoples, data for comparative analysis between diabetes and
neck circumference are scarce. In our study, the mean neck circumference was 36 cm (±3.3),
slightly below the national mean for the Brazilian male population (39.5±3.6) and slightly
above the national mean for the Brazilian female population (34.0±2.9) [30]. In relation to
other ethnic groups, such as Asian groups living in different cultural contexts, the mean found
was 33 cm (±4.16), indicating that the increase in fat in the neck region had a greater indica-
tion of cardiometabolic disease when compared to the increase in the body and visceral mass
index [31]. Another two studies conducted with the general American population suggest that
increased neck circumference was associated with hypertension, diabetes, metabolic syndrome
and dyslipidemia [32, 33].
In relation to waist circumference, the Munduruku presented a lower mean [85.3 cm
(±10.2)], when compared to the Xavante indigenous people (Mato Grosso) [95.1 (±8.3) [34],
but higher when compared to the Yanomami (Roraima) [76.3 (±46.8)] [35].
With regard to the Taper Index, the mean was 1.24 (± 0.9) among the Munduruku, similar
to the one found among the Mura (municipality of Autazes, Amazonas) [1.27 (±0.08)] [25].
However, these findings are much lower when compared to the mean of the Brazilian popula-
tion that varies between 1.35 (±0.08) and 1.34 (±0.09) [36, 37].
In relation to the blood pressure levels, the results of this study show that the prevalence of
people with blood pressure levels suggestive of arterial hypertension was 10.2%. A systematic
PLOS ONE | https://doi.org/10.1371/journal.pone.0255730 September 3, 2021
11 / 16
PLOS ONE Glycemic profile and associated factors in indigenous Munduruku, Amazonas.
review study with meta-analysis and meta-regression, conducted with indigenous people from
the North Region (Ianomaˆmi, Suruı´, Tembe´, Amondaua, Parkatêjê, Suruı´), from the Midwest
Region (Terena, Zoro´, Suya´, Kalapalo, Kuikuro, Matipus, Nahukwa´, Mehina´ku, Waura´, Yawa-
lapitı´, Guaranı´, Tupinikin, Xavante, Khisêdjê and indigenous people from the Jaguapiru vil-
lage), and from the Southeast and South Regions (Guaranı´-Mbya´, Kaingang), showed a 12%
increase in the chance of hypertension, in any indigenous person in Brazil, for each year stud-
ied [38]. The meta-analysis of this study showed that there was an increase in the prevalence of
arterial hypertension, since in 1970 it was non-existent in the indigenous population, 0.1%
(95% CI = 0.0% – 0.6%), when compared to 2014, when the highest prevalence of arterial
hypertension was identified: 29.7% (95% CI = 26.1% – 44.4%) [38].
A study that investigated cardiovascular risk factors among different ethnic groups, living
in the same urban area of Manaus (Amazonas), identified that, although the prevalence of
SAH among the indigenous people was lower than in white-skinned (62.5%) and brown-/
black-skinned (60.7%) individuals, that for pre-hypertension and hypertension was 28.6%
among the Satere´-Mawe´ and 46.5% among ethnic groups from the upper Rio Negro [39].
During the assessment of the lipid levels, this study presented a mean of triglycerides of
165.5 (±86.5) mg/dL. In turn, 21.1% of the participants had high levels of triglycerides. These
data are similar to those of the Mura de Autazes indigenous people (Amazonas) [163.5
(±104.7) mg/dL] [25] and Xavante of the São Marcos and Sangradouro Indigenous Reserves
(Mato Grosso) [199.1 (±171.2) mg/dL] [18], differing from the mean among the Guarani-
Mbya´ indigenous people (Rio de Janeiro), which was 116.0 (±74.9) mg/dL [3].
Regarding the total cholesterol levels, the mean was 177.5 (± 32.2) mg/dL, considered
within the boundary range and indicating that the Munduruku presented higher levels when
compared to other ethnicities, such as the Sangradouro and the Guarani-Mbya´ indigenous
peoples, whose mean total cholesterol values were 145.8 (±4.7) mg/dL [16] and 143.8 (±28.8)
mg/dL, respectively [3].
In relation to the diabetics indigenous individuals, 82.6% of them reported having a family
history of diabetes and 52.8%, a family history of stroke. Indigenous people under the age of
55, who live in remote areas of Australia, were 14 times more likely to have an ischemic stroke,
when compared to non-indigenous people belonging to the same age group. It is worth men-
tioning that the prevalence of diabetes found was 70.3% among indigenous people versus 34%
among non-indigenous people [40].
With regard to the findings obtained through Odds Ratio adjusted for gender and age, it is
possible to assert an increase in the chance of developing Pre-diabetes and Diabetes in relation
to age in the group under study. Australian indigenous peoples had a 7% chance of developing
diabetes each year of life [41]. A similar percentage was identified among the Munduruku, in
which at each one year of life increase, there is a 3% chance of having pre-diabetes, and 5% for
diabetes.
As for the BMI, for each increase in the unit of this ratio, the chance for the indigenous per-
son becoming a pre-diabetic is 16%; and 28%, for diabetes. Data found in a comparative study
between the population of the Aracruz Indigenous Reserve (Brazil) and the population of Espı´-
rito Santo (Brazil) showed that obese non-indigenous men and women were twice as likely to
have DM but, when it comes to the indigenous people in this study, no significant differences
were found [42].
Among the overweight indigenous people, the chance of having diabetes is four times
higher, respectively. For the obese, on the other hand, the chances substantially increase, both
for pre-diabetes, which increases to nine times, and for diabetes, which can reach 29 times. In
the study with Guarani, Kaiowa´ and Terena, from the Jaguapiru village (Mato Grosso do Sul),
the prevalence of diabetes among women was 9% and among men, 5%. The study also
PLOS ONE | https://doi.org/10.1371/journal.pone.0255730 September 3, 2021
12 / 16
PLOS ONE Glycemic profile and associated factors in indigenous Munduruku, Amazonas.
indicated a positive and significant association between obesity and diabetes (PR = 1.88; 95%
CI = 1.45 – 2.43; p<0.001). A population-based study carried out in different Brazilian regions
showed that obese individuals had 35% [95% CI = 1.35 – 1.86; p<0.001] chances of developing
diabetes [43]. These findings show that the Munduruku, although still distant from the
national mean of the general Brazilian population, are in an unfavorable condition toward the
development of diabetes in relation to other ethnic groups living in a similar cultural context.
Study limitations
In the absence of specific cut-off points for indigenous populations, those used for the general
population were considered, also adopted in other studies on different ethnic groups.
As it was impossible to apply a dietary recall, it was not possible to verify how much the eat-
ing habits are associated with the values found for glucose, cholesterol and triglycerides.
The instruments adopted in the interview were not specific to indigenous peoples. How-
ever, since it is an essential requirement to achieve the proposed objectives, the adequacy of
language to the understanding of the group under study constituted a task that demanded dif-
ferent moments of planning and evaluation by the team.
Conclusion
The 12% prevalence of glycaemia found among the Munduruku indigenous people is sugges-
tive of diabetes mellitus, and that of 74.3%, revealing pre-diabetes, configure themselves as
worrying indexes, as well as the chance of pre-diabetes, which increases by 20% when the BMI
increases by one unit. It is necessary to consider changes in the eating habits and lifestyle, as
well as environmental and social changes that can affect the health of the Munduruku, and
consider the stress levels, with the possibility of each of these elements contributing or not to
the results of this study. Consequently, it becomes indispensable to develop strategies combin-
ing early diagnosis and treatment actions with actions to reduce the risk factors, in order to
meet the needs and singularities of the Munduruku indigenous people. It is also suggested to
develop new research studies on the topic in order to consolidate these findings in other Mun-
duruku indigenous contexts.
Author Contributions
Conceptualization: Zilmar Augusto de Souza Filho, Noeli das Neves Toledo, Anto´nio Manuel
Sousa, Gilsirene Scantelbury de Almeida.
Data curation: Hanna Lorena Moraes Gomes, Neuliane Melo Sombra.
Formal analysis: Anto´nio Manuel Sousa.
Funding acquisition: Noeli das Neves Toledo.
Investigation: Hanna Lorena Moraes Gomes, Neuliane Melo Sombra.
Methodology: Hanna Lorena Moraes Gomes, Neuliane Melo Sombra.
Project administration: Zilmar Augusto de Souza Filho, Noeli das Neves Toledo, Gilsirene
Scantelbury de Almeida.
Supervision: Zilmar Augusto de Souza Filho, Evelyne Marie Therese Mainbourg, Gilsirene
Scantelbury de Almeida.
Visualization: Hanna Lorena Moraes Gomes, Eliza Dayanne de Oliveira Cordeiro, Zilmar
Augusto de Souza Filho, Evelyne Marie Therese Mainbourg, Gilsirene Scantelbury de
Almeida.
PLOS ONE | https://doi.org/10.1371/journal.pone.0255730 September 3, 2021
13 / 16
PLOS ONE Glycemic profile and associated factors in indigenous Munduruku, Amazonas.
Writing – original draft: Hanna Lorena Moraes Gomes, Neuliane Melo Sombra, Eliza Day-
anne de Oliveira Cordeiro.
Writing – review & editing: Hanna Lorena Moraes Gomes, Neuliane Melo Sombra, Eliza
Dayanne de Oliveira Cordeiro, Evelyne Marie Therese Mainbourg, Gilsirene Scantelbury
de Almeida.
References
1. Coimbra CEA Jr., Santos RV, Escobar AL. Epidemiologia e sau´de dos povos indı´genas no Brasil. Fio
Druz. Abrasco, editor. Epidemiologia e sau´de dos povos indı´genas no Brasil. Rio de Janeiro; 2005.
2. Coimbra CEA Jr. Sau´de e povos indı´genas no Brasil: reflexões a partir do I Inque´rito Nacional de
Sau´ de e Nutric¸ão Indı´gena. Cad Saude Publica [Internet]. 2014; 30(4):855–9. Available from: http://
www.ncbi.nlm.nih.gov/pubmed/24896060 https://doi.org/10.1590/0102-311x00031214 PMID:
24896060
3. Cardoso AM, Mattos IE, Koifman RJ. doenc¸as cardiovasculares na populac¸ão Guaranı´-Mbya´ do
Estado do Rio de Janeiro Prevalenc. Cad Sau´de Pu´blica. 2001; 17(2):345–54. https://doi.org/10.1590/
s0102-311x2001000200009 PMID: 11283765
4. Cardoso AM, Horta BL, Coimbra Ju´ nior C, Folle´ r M-L, Souza MC de, Santos RV. Inque´rito Nacional de
Sau´ de e Nutric¸ão dos Povos Indı´genas [Internet]. Abrasco, editor. Rio de Janeiro; 2009. 496 p. Avail-
able from: https://goo.gl/nvzKjZ%5Cnecos-redenutri.bvs.br/tiki-download_file.php?fileId=1284
5. Codec¸o CT, Vilela D, Coelho F, Bastos LS, Carvalho LM, Gomes MFC, et al. Estimativa de risco de
espalhamento da COVID-19 no Brasil e avaliac¸ão da vulnerabilidade socioeconoˆmica nas microrre-
giões brasileiras. FGV Reposito´ rio Digit. 2020; 2:1–17.
6. Cunha E del BB, Fagundes RP, Scalabrin EE, Herai RH. Avaliac¸ão do perfil lipı´dico de adolescentes.
Int J Cardiovasc Sci. 2018; 31(4):367–73.
7. Schmidt MI, Duncan BB, E Silva GA, Menezes AM, Monteiro CA, Barreto SM, et al. Doenc¸as croˆnicas
não transmissı´veis no Brasil: carga e desafios atuais. Lancet. 2011; 377(9781):1949–61. https://doi.
org/10.1016/S0140-6736(11)60135-9 PMID: 21561658
8. Cardoso FN, Domingues TAM, Silva SS, Lopes J de L. Fatores de risco cardiovascular modifica´veis
em pacientes com hipertensão arterial sistêmica. Reme Rev Min Enferm [Internet]. 2020; 24:1–8. Avail-
able from: https://cdn.publisher.gn1.link/reme.org.br/pdf/e1275.pdf
9.
Lyra R, Oliveira M, Lins D, Cavalcanti N, Gross JL, Maia FFR, et al. Diretrizes Sociedade Brasileira de
Diabetes 2019–2020. Vol. 5, Clannad. 2019. 709–717 p.
10.
IDF. IDF DIABETES ATLAS. 2019. 168 p.
11. Oliveira GF De, Oliveira TRR De, Rodrigues FF, Corrêa LF, Ikejiri AT, Casulari LA. Prevalência de dia-
betes melito e toleraˆ ncia à glicose diminuı´da nos indı´genas da Aldeia Jaguapiru, Brasil. Rev Panam
Salud Publica. 2011; 29(5):315–21. https://doi.org/10.1590/s1020-49892011000500003 PMID:
21709935
12. Oliveira GF, Oliveira TRR, Ikejiri AT, Andraus MP, Galvao TF, Silva MT, et al. Prevalence of hyperten-
sion and associated factors in an indigenous community of Central Brazil: A population-based study.
PLoS One. 2014; 9(1):1–6.
13.
Freitas GA de, Souza MCC de, Lima R da C. Prevalência de diabetes mellitus e fatores associados em
mulheres indı´genas do Municı´pio de Dourados, Mato Grosso do Sul, Brasil. Cad Saude Publica. 2016;
32(8):1–12.
14. Meyerfreund D, Gonc¸alves CP, Cunha Roberto S, Pereira AC, Krieger JE, Mill JG. Age-dependent
increase in blood pressure in two different Native American communities in Brazil. J Hypertens. 2009;
27(9):1753–60. https://doi.org/10.1097/hjh.0b013e32832e0b2b PMID: 19701999
15. Santos KM dos, Tsutsui ML da S, Galvão PP de O, Mazzucchetti L, Rodrigues D, A SG, et al. Grau de
atividade fı´sica e sı´ndrome metabo´ lica: um estudo transversal com indı´genas Khisêdjê do Parque Indı´-
gena do Xingu, Brasil. Cad Sau´ de Pu´ blica. 2012; 28(12):2327–38. https://doi.org/10.1590/s0102-
311x2012001400011 PMID: 23288065
16. Dal Fabbro AL, Franco LJ, Silva AS da, Sartorelli DS, Soares LP, Franco LF, et al. High Prevalence of
type 2 Diabetes Mellitus in Xavante indians from Mato Grosso, Brazil. Ethn Dis. 2014; 24:35–40. PMID:
24620446
PLOS ONE | https://doi.org/10.1371/journal.pone.0255730 September 3, 2021
14 / 16
PLOS ONE Glycemic profile and associated factors in indigenous Munduruku, Amazonas.
17.
Freitas GA de, Souza MCC de, Lima R da C. Prevalência de diabetes mellitus e fatores associados em
mulheres indı´genas do Municı´pio de Dourados, Mato Grosso do Sul, Brasil. Cad Saude Publica. 2016;
32(8):e00023915. https://doi.org/10.1590/0102-311X00023915 PMID: 27626648
18. Soares LP, dal Fabbro AL, Silva AS, Sartorelli DS, Franco LF, Kuhn PC, et al. Cardiovascular risk in
Xavante indigenous population. Arq Bras Cardiol. 2018; 110(6):542–50. https://doi.org/10.5935/abc.
20180090 PMID: 30226913
19.
Faludi AA, Maria Cristina de Oliveira Izar, Kerr Saraiva Jose´ Francisco, Ana Paula Marte Chacra HTB,
Neto AA, Bertolami A, et al. Atualizac¸ão da diretriz brasileira de dislipidemias e prevenc¸ão da atero-
sclerose—2017. Arq Bras Cardiol. 2017; 109(1):76.
20. WHO. Obesity: preventing and managing the global epidemic. Who Tech Rep Ser. 2000; 37(10):253.
PMID: 11234459
21. Gallagher D, Heymsfield SB, Heo M, Jebb SA, Murgatroyd PR, Sakamoto Y. Healthy percentage body
fat ranges: An approach for developing guidelines based on body mass index. Am J Clin Nutr [Internet].
2000; 72(3):694–701. Available from: https://pubmed.ncbi.nlm.nih.gov/10966886/ https://doi.org/10.
1093/ajcn/72.3.694 PMID: 10966886
22. Barroso WKS, Cibele Isaac Saad Rodrigues LAB, Mota-Gomes MA, Brandão AA, Feitosa AD de M,
Machado CA, et al. Diretrizes Diretrizes Brasileiras de Hipertensão Arterial– 2020. Vol. 116. 2021.
516–658 p.
23. Santos JL, Pe´ rez-Bravo F, Carrasco E, Calvilla´ n M, Albala C. Low prevalence of type 2 diabetes despite
a high average body mass index in the aymara natives from Chile. Nutrition. 2001; 17(4):305–9. https://
doi.org/10.1016/s0899-9007(00)00551-7 PMID: 11369169
24. Bennett PH, Burch TA, Miller M. Diabetes Mellitus in American (Pima) Indians. Lancet. 1971; 298
(7716):125–8.
25. De Souza Filho ZA, Ferreira AA, Dos Santos J, Meira KC, Pierin AMG. Cardiovascular risk factors with
an emphasis on hypertension in the Mura Indians from Amazonia. BMC Public Health. 2018; 18(1):1–
12. https://doi.org/10.1186/s12889-018-6160-8 PMID: 30424745
26. Brasil MDS. Diretrizes e Recomendac¸ões para o Cuidado Integral de Doenc¸as Croˆnicas Diretrizes e
Recomendac¸ões para o Cuidado Integral de Doenc¸as Croˆ nicas Não-Transmissı´veis [Internet]. Preven-
tion. 2008. Available from: http://bvsms.saude.gov.br/bvs/publicacoes/diretrizes_recomendacoes_
cuidado_doencas_cronicas.pdf
27. WHO. Technical package for cardiovascular disease management in primary health care. Report.
2016;76.
28. Quermes PA de A, Carvalho JA de. Os impactos dos benefı´cios assistenciais para os povos indı´genas.
Servic¸o Soc e Soc [Internet]. 2013; 116:769–91. Available from: http://www.planalto.gov.br/ccivil_03/
leis/
29.
Flor LS, Campos MR, de Oliveira AF, Schramm JM de A. Diabetes burden in Brazil: Fraction attribut-
able to overweight, obesity, and excess weight. Rev Saude Publica. 2015;49. https://doi.org/10.1590/
S0034-8910.2015049005391 PMID: 26270011
30. Silva AAG de O, de Araujo LF, Diniz M de FHS, Lotufo PA, Bensenor IM, Barreto SM, et al. Neck cir-
cumference and 10-year cardiovascular risk at the baseline of the elsa-brasil study: Difference by sex.
Arq Bras Cardiol. 2020; 115(5):840–8. https://doi.org/10.36660/abc.20190289 PMID: 33295446
31.
Fu W, Zou L, Yin X, Wu J, Zhang S, Mao J, et al. Association between neck circumference and cardio-
metabolic disease in Chinese based cross- - adults: a community- - sectional study. BMJ Open.
2019;1–7. https://doi.org/10.1136/bmjopen-2019-030833 PMID: 31273011
32. Preis SR, Massaro JM, Hoffmann U, D’Agostino RB, Levy D, Robins SJ, et al. Neck Circumference as a
Novel Measure of Cardiometabolic Risck: The Framingham Heart Study. J Clin Endocr Metab. 2010; 95
(8):3701–10. https://doi.org/10.1210/jc.2009-1779 PMID: 20484490
33. Preis SR, Pencina MJ, D’Agostino RB, Meigs JB, Vasan RS, Fox CS. Neck Circumference and the
development of Cardiovascular Disease Risck Factors in the Framingham heart Study. Diabetes Care.
2013;36.
34. Welch JR, Ferreira AA, Tavares FG, Lucena JRM, Oliveira MVG de, Santos R V., et al. The Xavante
Longitudinal Health Study in Brazil: Objectives, design, and key results. Am J Hum Biol. 2019;1–12.
35. Bloch K V., Coutinho E da SF, Loˆ bo MS de C, Oliveira JEP de, Milech A. Pressão arterial, glicemia capi-
lar e medidas antropome´ tricas em uma populac¸ão Yanoma´ mi. Cad Saude Publica. 1993; 9(4):428–38.
36. Barroso TA, Marins LB, Alves R, Gonc¸alves ACS, Barroso SG, Rocha G de S. Associac¸ão Entre a Obe-
sidade Central e a Incidência de Doenc¸as e Fatores de Risco Cardiovascular. Int j Cardiovasc sci.
2017; 30(5):416–24.
PLOS ONE | https://doi.org/10.1371/journal.pone.0255730 September 3, 2021
15 / 16
PLOS ONE Glycemic profile and associated factors in indigenous Munduruku, Amazonas.
37.
Fontela PC, Winkelmann ER, C PRNV. Estudo do ı´ndice de conicidade, ı´ndice de massa corporal e cir-
cunferência abdominal como preditores de doenc¸a arterial coronariana. Rev Port Cardiol [Internet].
2017; 36(5):357–64. Available from: https://doi.org/10.1016/j.repc.2016.09.013 PMID: 28449975
38. Souza Filho ZA de, Ferreira AA, dos Santos B, Pierin AMG. Hypertension prevalence among indige-
nous populations in Brazil: A systematic review with meta-analysis. Rev da Esc Enferm [Internet]. 2015;
49(6):1012–22. Available from: https://www.scielo.br/pdf/reeusp/v49n6/0080-6234-reeusp-49-06-
1016.pdf
39.
Toledo N das N, Almeida GS de, Matos MMM, Balieiro AA da S, Martin LC, Franco RJ da S, et al. Car-
diovascular risk factors: differences between ethnic groups. Rev Bras Enferm. 2020; 73(4):e20180918.
https://doi.org/10.1590/0034-7167-2018-0918 PMID: 32578730
40. Balabanski AH, Goldsmith K, Giarola B, Buxton D, Castle S, McBride K, et al. Stroke incidence and sub-
types in Aboriginal people in remote Australia: A healthcare network population-based study. BMJ
Open. 2020; 10(10):1–9. https://doi.org/10.1136/bmjopen-2020-039533 PMID: 33033097
41. Brown A, Carrington MJ, McGrady M, Lee G, Zeitz C, Krum H, et al. Cardiometabolic risk and disease
in Indigenous Australians: The heart of the heart study. Int J Cardiol [Internet]. 2014; 171(3):377–83.
Available from: https://doi.org/10.1016/j.ijcard.2013.12.026 PMID: 24388543
42. Alvim RDO, Mourao CA, De Oliveira CM, Krieger JE, Mill JG, Pereira AC. Body mass index, waist cir-
cumference, body adiposity index, and risk for type 2 diabetes in two populations in Brazil: General and
Amerindian. PLoS One. 2014; 9(6).
43.
Flor LS, Campos MR. The prevalence of diabetes mellitus and its associated factors in the Brazilian
adult population: evidence from a population-based survey. Rev Bras Epidemiol. 2017; 20(1):16–29.
https://doi.org/10.1590/1980-5497201700010002 PMID: 28513791
PLOS ONE | https://doi.org/10.1371/journal.pone.0255730 September 3, 2021
16 / 16
PLOS ONE |
10.1371_journal.pone.0255694 | RESEARCH ARTICLE
Chemical, physical, and functional properties
of Thai indigenous brown rice flours
David Oppong, Worawan PanpipatID, Manat ChaijanID*
Food Technology and Innovation Research Center of Excellence, School of Agricultural Technology and
Food Industry, Walailak University, Thasala, Nakhon Si Thammarat, Thailand
* cmanat@wu.ac.th
Abstract
Thai indigenous brown rice flours from Nakhon Si Thammarat, Thailand, namely Khai Mod
Rin (KMRF) and Noui Khuea (NKRF), were assessed for quality aspects in comparison with
brown Jasmine rice flour (JMRF) and commercial rice flour (CMRF) from Chai Nat 1 variety.
All the rice flours had different chemical composition, physical characteristic, and techno-
functionality. The KMRF, NKRF, and JMRF were classified as a low amylose type (19.56–
21.25% dw). All rice flours had low total extractable phenolic content (0.1–0.3 mg GAE/g
dw) with some DPPH• scavenging activity (38.87–46.77%). The variations in the bulk den-
sity (1.36–1.83 g/cm3), water absorption capacity (0.71–1.17 g/g), solubility (6.93–13.67%),
oil absorption capacity (1.39–2.49 g/g), and swelling power (5.71–6.84 g/g) were noticeable.
The least gelation concentration ranged from 4.0 to 8.0% where KMRF was easier to form
gel than JMRF, and NKRF/CMRF. The foam capacity of the flours was relatively low (1.30–
2.60%). The pasting properties differed among rice flours and the lowest pasting tempera-
ture was observed in CMRF. Overall, the chemical, physical, functional, and pasting quali-
ties of flours were substantially influenced by rice variety. The findings offered fundamental
information on Thai indigenous rice flour that can be used in food preparations for specific
uses.
Introduction
As one of the pivotal cereal grains in the world, rice (Oryza sativa L.) is a good source of essen-
tial nutrients, especially carbohydrate and protein [1]. The global production of rice is esti-
mated to be doubly increased by 2050 due to the increasing consumer demand [2]. Rice is
mostly consumed as intact kernels, but rice flour can be used for several food preparations
such as traditional foods, noodles, baked goods, extruded products, and novel products (e.g.
gluten free based foods, infant’s foods, and snacks) [1, 3]. It has been reported that the physico-
chemical properties of rice flour significantly affected the qualities of those products. Commer-
cial rice flour (CMRF) is primarily produced from Chai Nat 1 white rice to meet appreciations
of consumers [1, 3], but the flour from the polished grain contains lower nutrients and bioac-
tive compounds due to the elimination of the bran upon polishing/milling [4]. In Thailand,
waxy and high amylose rice varieties are regularly used for CMRF production [5]. Owing to
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: Oppong D, Panpipat W, Chaijan M (2021)
Chemical, physical, and functional properties of
Thai indigenous brown rice flours. PLoS ONE
16(8): e0255694. https://doi.org/10.1371/journal.
pone.0255694
Editor: Umakanta Sarker, Bangabandhu Sheikh
Mujibur Rahman Agricultural University,
BANGLADESH
Received: May 28, 2021
Accepted: July 22, 2021
Published: August 3, 2021
Peer Review History: PLOS recognizes the
benefits of transparency in the peer review
process; therefore, we enable the publication of
all of the content of peer review and author
responses alongside final, published articles. The
editorial history of this article is available here:
https://doi.org/10.1371/journal.pone.0255694
Copyright: © 2021 Oppong et al. This is an open
access article distributed under the terms of the
Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: All relevant data are
within the manuscript.
Funding: This research was funded by Walailak
University, Thailand through the Ph.D. Scholarship
PLOS ONE | https://doi.org/10.1371/journal.pone.0255694 August 3, 2021
1 / 17
PLOS ONE for Outstanding International Students, grant
number [MOE 57 19 00/144/2562]. This research
was also financially supported by the new strategic
research project (P2P), Walailak University,
Thailand. The funders had no role in study design,
data collection and analysis, decision to publish, or
preparation of the manuscript.
Competing interests: The authors have declared
that no competing interests exist.
Characterization of Thai rice flour
the consumers’ demands, indigenous rice flour has been paid more attention as an alternative
raw material for several food products.
Compositional variations in terms of physical, chemical, thermal, and pasting properties
were found among rice varieties, depending on the genotype, agronomic and cultivation con-
ditions, environmental factors, storage conditions, and processing parameters [4, 6]. The qual-
ity of rice does not only include the physical appearance, rather it encompasses the chemical,
functional, and thermal properties [6]. In Thailand, different indigenous rice varieties have
been grown. In Southern Thailand, more than 4,000 local rice varieties have been recorded [7,
8]. Especially in Nakhon Si Thammarat, non-glutinous domestic rices such as Khai Mod Rin
and Noui Khuea have been widely cultivated following the production practices for organic
Thai rice. Recently, the health benefits for the consumption of brown rice were intensively
reported [4]. Therefore, in this research, brown Khai Mod Rin rice flour (KMRF) and brown
Noui Khuea rice flour (NKRF), which are mainly cultivated in Nakhon Si Thammarat, Thai-
land, were investigated for their basic chemical composition, physical property, and techno-
functionality in comparison with organic brown Jasmine rice flour (JMRF) and CMRF.
Material and methods
Chemicals
All chemicals used for analyses such as DPPH (2,2-Diphenyl-1-picrylhydrazyl), acetone, meth-
anol, and gallic acid (GA) were obtained from Sigma-Aldrich Corp. (St. Louis, MO, USA).
Raw materials
Two domestic Southern Thai brown rices (Oryza sativa L., varieties Khai Mod Rin and Noui
Khuea) and brown Jasmine rice were obtained from an organic farm in Phra Phrom, Nakhon
Si Thammarat, Thailand (8˚17’23.4"N, 99˚58’48.2"E, altitude of 9 m), in March 2020. To pre-
pare the flour, brown rices were ground for 5 min using a grinder (MK 5087M Panasonic
Food Processor, Selangor Darul Ehsan, Malaysia) and passed through a 100-mesh sieve.
CMRF, from Chai Nat 1 variety, was obtained from Cho Heng Rice Vermicelli Factory Co.,
Ltd., Nakhon Pathom, Thailand (13˚43’38.1"N, 100˚14’42.3"E, altitude of 10 m), in March
2020. The flours were packed in polythene bags and kept at room temperature (28–30˚C) until
further analysis. The storage time was less than a month. Three different lots of flour were pre-
pared to get triplications for all analyses. The contents for all chemical compositions were
reported on a dry weight (dw) basis.
Proximate composition
The standard methods of AOAC [9] were used for proximate composition analysis including
moisture (AOAC method number 950.46), crude protein (AOAC method number 928.08, a
conversion factor = 5.95), ash (AOAC method number 920.153), fiber (AOAC method num-
ber 962.09), fat (AOAC method number 963.15), and carbohydrate (calculated by difference).
Amylose content
Amylose content was determined according to Kraithong et al. [10]. The sample (100 mg) was
mixed with 95% (v/v) ethanol (1 mL) and 2 M NaOH (9 mL). The mixture was brought to 100
mL with distilled water and then 0.2% (w/v) iodine solution (2 mL) was added. Thereafter, the
absorbance was read at 620 nm (Shimadzu UV-2100 spectrophotometer, Shimadzu Scientific
Instruments Inc., Columbia, MD, USA). A calibration curve was created using standard amy-
lose obtained from potato starch to quantify the amylose content.
PLOS ONE | https://doi.org/10.1371/journal.pone.0255694 August 3, 2021
2 / 17
PLOS ONE Characterization of Thai rice flour
Total extractable phenolic content (TPC) and DPPH• scavenging activity
The method of Sungpud et al. [11] was used for TPC extraction. The flour samples (10 g) were
extracted with 80% (v/v) ethanol (100 mL) at 40˚C in the Lib-300M incubator (Labtech,
Korea) for 24 h under continuous magnetic stirring (300 rpm). Thereafter, the mixtures were
centrifuged at 5,000 rpm for 10 min at room temperature (RC-5B plus centrifuge, Sorvall, Nor-
walk, CT, USA) and the supernatants were collected. After filtration (Whatman No.1), the fil-
trates were analyzed for TPC. One hundred μL of ethanolic flour extract were mixed with 2.0
mL Folin-Ciocalteu reagent (previously diluted to 10-fold with deionized water) and well
mixed. After standing for 5 min, 15% sodium carbonate solution (1.0 mL) was added. The cor-
respondence solution was kept in the dark for 60 min. The absorbance was read at 765 nm
using a UV-vis spectrophotometer (Shimadzu, MD, USA). The TPC was expressed as mg gallic
acid equivalent (GAE)/g sample.
The DPPH• scavenging effect was analyzed according to Limsuwanmanee et al. [12]. Etha-
nolic flour extract (1 mL; 0.1 mg/mL TPC) was mixed with 0.2 mM methanolic DPPH• solu-
tion (1 mL). After incubation in the dark at room temperature (30 min), the absorbance was
measured at 517 nm against blank. A control was prepared using methanol instead of the sam-
ple. DPPH• inhibition was acquired by the following formula:
DPPH� inhibition ð%Þ ¼ ½ðA0 (cid:0) A1Þ=A0� � 100
ð1Þ
where A0 = absorbance of the control and A1 = absorbance of the sample.
Fourier transform infrared (FTIR) spectroscopy
The FTIR spectroscopy is a vibrational spectroscopic technique that can be used to character-
ize the substances by identifying their functional groups presented [13]. FTIR spectra (500–
4,000 cm-1 with the resolution of 4 cm-1 at the average of 16 scans) of the rice flours were
obtained using a horizontal Attenuated Total Reflectance (ATR) Trough plate crystal cell (45˚
ZnSe; 80 mm long, 10 mm wide and 4 mm thick) (Pike Technology, Inc., Madison, WI, USA)
equipped with a Bruker Model Vector 33 FTIR spectrometer (Bruker Co., Ettlingen, Germany)
at room temperature. Analysis of spectral data was carried out using the OPUS 3.0 data collec-
tion software program.
Color
The rice flour color was determined using a Hunterlab colorimeter with 10 standard observers
and illuminant D65 (Hunter Assoc. Laboratory; VA, USA). The L�, a�, and b� values were
recorded.
Bulk density
Fifty grams of flour was taken to a measuring cylinder (100 mL) and tapped carefully. After
reading the volume, the bulk density was estimated from the ratio of mass (g) to volume (mL)
[14].
Bulk density g=mL
ð
Þ ¼
Weight of flour ðgÞ
Volume of flour after settled ðmLÞ
ð2Þ
Water absorption capacity (WAC) and solubility
The WAC and solubility were determined using the method of Kraithong et al. [10]. One
gram of flour was suspended in 10 mL of distilled water and mixed with a vortex mixer for 1
min. The suspensions were heated in a water bath at 30˚C for 30 min with gentle stirring and
PLOS ONE | https://doi.org/10.1371/journal.pone.0255694 August 3, 2021
3 / 17
PLOS ONE Characterization of Thai rice flour
then centrifuged at 1,500 ×g for 10 min (RC-5B plus centrifuge). The supernatant was carefully
poured into an aluminum moisture can before being dried at 105˚C overnight. The sediments
were collected and weighed. The WAC and solubility were calculated using the following for-
mulas:
WAC g=g
ð
Þ ¼
Weight of wet sediment ðgÞ
Dry weight of flour ðgÞ
Solubility %ð Þ ¼
Weight of dried supernatant ðgÞ
Dry weight of flour ðgÞ
� 100
ð3Þ
ð4Þ
Oil absorption capacity (OAC)
The OAC of the rice flour was measured according to Kraithong et al. [10]. One gram of flour
was mixed with 10 mL of soybean oil for 1 min. After standing at room temperature (30 min),
the centrifugation was applied at 1,500 ×g for 10 min (RC-5B plus centrifuge). Thereafter, the
surplus oil was decanted while the residue (weight of oil absorbed) was weighed. The calcula-
tion of OAC was as follows:
ð
OAC g=g
Þ ¼
Weight of oil absorbed ðgÞ
Weight of sample ðgÞ
ð5Þ
Swelling power
The swelling power was measured according to Appiah et al. [15]. The flour sample (1 g) was
mixed with distilled water (30 mL). After heating (85˚C/30 min) in a W350 Memmert water
bath (Schwabach, Germany), sample was cooled to room temperature and centrifuged at 1,500
×g for 20 min (RC-5B plus centrifuge). The swelling power was estimated as the weight of the
paste per weight of the dry sample.
Swelling power g=g
ð
Þ ¼
Weight of the paste ðgÞ
Weight of dry sample ðgÞ
ð6Þ
Least gelation concentration (LGC)
The LGC was measured according to Appiah et al. [15]. Aqueous suspensions of 2, 4, 6, 8, up
to 20% (w/v) flour (5 mL) were heated in boiling water for 1 h. After cooling down in ice bath
and standing at 4˚C for 2 h, the test tube was inverted. The LGC was regarded as the concen-
tration at which the inverted sample did not slip down the side of the test tube.
Foaming capacity (FC)
The flour sample (3 g) was mixed with distilled water (100 mL) at room temperature and
homogenized for 5 min at 13,400 rpm using an IKA1 homogenizer (Model T25 digital Ultra-
Turrax1, Staufen, Germany). The increase in volume of the foam at 30 s after whipping
against the original volume was expressed as FC [16].
FC %ð Þ ¼
Volume after whipping ðmLÞ
Original voloume ðmLÞ
� 100
ð7Þ
Pasting properties
Pasting characteristics were analyzed using the Rapid Visco Analyzer (RVA 4500, Perten
Instruments, Sweden). In a canister, the rice flour sample (3 g) was inserted, and then 25 mL
of distilled water was added (14% moisture basis). The RVA profile was recorded under
PLOS ONE | https://doi.org/10.1371/journal.pone.0255694 August 3, 2021
4 / 17
PLOS ONE Characterization of Thai rice flour
specific conditions. The temperature was held at 50˚C for 1 min and then raised up to 95˚C in
3.8 min (held for 2.5 min). Consequently, it was cooled to 50˚C within 3.8 min and held for 1.4
min. The pasting parameters e.g. pasting temperature, peak viscosity, breakdown, final viscos-
ity, and setback were measured according to the method of Kraithong et al. [10].
Statistical analysis
A completely randomized design was used for experimental design with a single factor of four
rice flours and the experiments were performed in triplicate. Data were subjected to one-way
analysis of variance (ANOVA). Duncan’s multiple-range test was used to analyze significant
differences (p<0.05) among samples, using the SPSS program (Version 23.0, SPSS Inc., Chi-
cago, IL, USA).
Results
Proximate composition and amylose content
The proximate compositions of Thai indigenous brown rice flour, namely KMRF and NKRF,
in comparison with JMRF and CMRF are presented in Table 1. Generally, the brown rice
flours had higher contents of protein, ash, fiber, and fat than CMRF, whereas CMRF had a
higher carbohydrate content (p < 0.05). The moisture contents of the rice flours were in the
range of 4.25–5.06% and the lowest content was found in CMRF (p < 0.05). The flours had
protein ranging between 5.01% and 8.14%. The highest protein content was found in NKRF
(p < 0.05), followed by KFRF/JMRF and CMRF. The ash contents of rice flours were in the
range of 0.38–2.22%. No significant difference in ash content was found among the brown rice
varieties (2.0–2.2%) (p > 0.05). The crude fiber content of the flours ranged from 0.49% to
2.85%. JMRF had the highest crude fiber (2.85%), followed by KMRF (2.54%), NKRF (2.06%),
and CMRF (0.49%). The fat content of the rice flours ranged from 0.77% to 2.0%. NKRF had
the highest fat (2.0%), whereas the lowest value was found in CMRF (0.77%). In the present
experiment, carbohydrate was found to be high in all samples (>85%). CMRF had a higher
carbohydrate content (93.3%) than brown rice flours (85.79–87.82%).
According to Table 1, there is no significant variation in amylose content for the three
brown rice flours (p > 0.05). However, the CMRF was markedly different from them
(p < 0.05). The amylose contents in the rice flours were 19.56–21.25% (Table 1).
Table 1. Chemical compositions of Thai indigenous brown rice flour, namely brown Khai Mod Rin rice flour (KMRF) and brown Noui Khuea rice flour (NKRF), in
comparison with brown Jasmine rice flour (JMRF) and commercial rice flour (CMRF).
Chemical composition
Moisture (% ww)
Protein (% dw)
Ash (% dw)
Fiber (% dw)
Fat (% dw)
Carbohydrate (% dw)
Amylose (% dw)
Total extractable phenolic content (mg GAE/g dw)
DPPH radical scavenging activity (%)
KMRF
NKRF
JMRF
CMRF
Rice flour
5.00 ±0.20a
6.35 ±0.15b
2.20 ±0.01a
2.54 ±0.00b
1.08 ±0.15bc
87.82 ±0.05b
19.56 ±0.23b
0.32 ±0.02a
38.87 ±0.29b
5.06 ±0.12a
8.14 ±0.63a
1.99 ±0.05a
2.06 ±0.02c
2.00 ±0.10a
85.79 ±0.74c
19.60 ±0.01b
0.34 ±0.01a
37.40 ±0.97bc
4.93 ±0.11a
7.01 ±0.75b
2.22 ±0.25a
2.85 ±0.02a
1.39 ±0.23b
86.54 ±0.52c
19.84 ±0.08b
0.25 ±0.01b
35.11 ±0.72c
4.25 ±0.49b
5.01 ±0.26c
0.38 ± 0.00b
0.49 ± 0.02d
0.77 ±0.17c
93.32±0.54a
21.25±0.23a
0.12 ±0.01c
46.77±1.69a
Values are means ± standard deviation from triplicate determinations. Different letters are significantly different along the rows (p < 0.05).
GAE = gallic acid equivalent, ww = wet weight, dw = dry weight
https://doi.org/10.1371/journal.pone.0255694.t001
PLOS ONE | https://doi.org/10.1371/journal.pone.0255694 August 3, 2021
5 / 17
PLOS ONE Characterization of Thai rice flour
TPC, DPPH• scavenging activity, and FTIR spectra
The TPC of the rice flours are given in Table 1. The TPC in the brown rice flours was higher
than the CMRF (p < 0.05). The highest TPC in the brown rice flours was owned by NKRF and
KMRF (~0.3 mg GAE/g), followed by JMRF (0.2 mg GAE/g), and CMRF (0.1 mg GAE/g). The
antioxidant activity of ethanolic extracts of rice flours was determined by DPPH• assay
(Table 1). Inactivation of DPPH• was found in all of the rice flour extracts examined, ranging
from 38.87 to 46.77%. The CMRF extract showed higher antioxidant capacity than those of the
brown rice flours. The FTIR spectra for rice flours are depicted in Fig 1. All the samples
showed similar spectra within a region of 500–4000 cm-1 with different peak intensities.
Color and bulk density
The color attributes (L�, a�, and b�) of the rice flours were remarkably different (p < 0.05)
(Table 2). L� values, which express the lightness, were in the range of 75.85–95.20 with the
highest L� value coming from CMRF. Generally, the L� value of CMRF was higher than
NKRF, KMRF, and JMRF, respectively (p < 0.05). Higher values of a� and b� with lower L�
value were found in all brown rice flours (p < 0.05), while a negative a� value, a low b� value
and the highest L� value was noticeable in CMRF (p < 0.05). The rice flours had bulk densities
ranging between 1.36 g/mL to 1.83 g/mL (Table 2). The mean bulk density was ranged in the
order of CMRF = NKRF � JMRF � KMRF.
Fig 1. FTIR spectra of Thai indigenous organic brown rice flour, namely brown Khai Mod Rin rice flour (KMRF) and brown Noui Khuea rice flour (NKRF), in
comparison with brown Jasmine rice flour (JMRF) and commercial rice flour (CMRF).
https://doi.org/10.1371/journal.pone.0255694.g001
PLOS ONE | https://doi.org/10.1371/journal.pone.0255694 August 3, 2021
6 / 17
PLOS ONE Characterization of Thai rice flour
Table 2. Physical and functional properties of Thai indigenous brown rice flour, namely brown Khai Mod Rin rice flour (KMRF) and brown Noui Khuea rice flour
(NKRF), in comparison with brown Jasmine rice flour (JMRF) and commercial rice flour (CMRF).
Parameters
Color
L�
a�
b�
Bulk density (g/mL)
Water absorption capacity (WAC; g/g)
Solubility (%)
Oil absorption capacity (OAC; g/g)
Swelling power (g/g)
Least gelation concentration (LGC; %)
Foam capacity (FC; %)
KMRF
NKRF
JMRF
CMRF
Rice flour
77.96±0.04c
2.24 ±0.02b
13.64 ±0.15c
1.36 ±0.05b
1.16 ±0.33a
13.67±5.09a
2.31±0.41a
6.84±0.45a
4.00 ± 1.00c
2.61±1.13a
79.74 ±0.02b
1.63 ±0.07c
14.35 ±0.05b
1.78 ± 0.00a
1.06 ±0.06ab
10.83±1.04ab
2.37 ±0.65a
6.49 ±0.39ab
8.00 ± 1.00a
1.50±0.10b
75.85 ±0.03d
2.64 ±0.06a
15.73 ±0.18a
1.69 ± 0.28ab
1.17 ±0.16a
13.40 ± 1.96a
1.39 ±0.98a
6.51 ± 0.76ab
6.00±1.00b
2.00±0.00ab
95.20 ±0.06a
-0.11 ±0.00d
3.14 ±0.11d
1.83 ±0.23a
0.71 ±0.06b
6.93 ±2.18b
2.49 ±0.40a
5.71±0.25b
8.00±1.93a
1.30±0.17b
Values are means ± standard deviation from triplicate determinations. Different letters are significantly different along the rows (p < 0.05).
WAC, solubility, OAC, swelling power, LGC, and FC
https://doi.org/10.1371/journal.pone.0255694.t002
In this study, functional properties of rice flour were tested for WAC, solubility, OAC, swell-
ing power, LGC, FC, and pasting characteristics. WAC of the rice flours ranged from 0.71 g/g to
1.17 g/ g as shown in Table 2. The lowest WAC was found in CMRF (0.71 g/g), followed by
NKRF and KMRF/JMRF. The solubility ranged from 6.93% to 13.67% (Table 2). There were no
significant differences in terms of OAC among the flours (p > 0.05). The OAC of the rice flours
ranged from 1.39 g/g to 2.49 g/g (Table 2). In addition, for all flours, the OAC was higher than
the WAC (Table 2). The flours had swelling power values ranging between 5.71 g/g and 6.84 g/g
(Table 2). The LGC values ranged from 4.0% to 8.0% (Table 2). The FC of the flours was rela-
tively low and the values ranged from 1.30% to 2.6% (Table 2). With regards to the FC, the
value of KMRF � JMRF � NKRF = CMRF.
Pasting properties
The pasting properties of the rice flours are given in Table 3. The pasting temperature of the
rice flours varied between 89˚C and 91˚C (Table 3). All brown rice flours had a higher pasting
temperature than CMRF (p < 0.05). The peak viscosity of the rice flours under study ranged
from 847 BU to 2,250 BU, as shown in Table 3. The result indicated that the CMRF had
Table 3. Pasting of Thai indigenous brown rice flour, namely brown Khai Mod Rin rice flour (KMRF) and brown Noui Khuea rice flour (NKRF), in comparison
with brown Jasmine rice flour (JMRF) and commercial rice flour (CMRF).
Parameters
Pasting temperature (˚C)
Peak viscosity (BU)
Trough viscosity (BU)
Breakdown (BU)
Setback (BU)
Final viscosity (BU)
KMRF
91±0a
898±12c
872±16c
19 ±11c
142±1b
1021±1c
NKRF
91±0a
847±12d
589±23d
258±19b
-85±4d
497±12d
Rice flour
JMRF
91±0a
1,230±24b
1,234±23b
-3±27c
312±6a
1545±1b
CMRF
89±1b
2,250±31a
1,927±9a
321±25a
100±18c
2,028±24a
Values are means ± standard deviation from triplicate determinations. Different letters are significantly different along the rows (p < 0.05).
https://doi.org/10.1371/journal.pone.0255694.t003
PLOS ONE | https://doi.org/10.1371/journal.pone.0255694 August 3, 2021
7 / 17
PLOS ONE Characterization of Thai rice flour
significantly higher peak viscosity followed by JMRF, KMRF, and NKRF (Table 3). The trough
viscosity of the flours ranged from 589 BU to 1,927 BU, as illustrated in Table 3. The highest
setback value was found in JMRF (312 BU) and the highest final viscosity was found in CMRF
(2,028 BU) (p < 0.05; Table 3).
Discussion
Because to varietal differences, geographical location, and processing conditions, all of the rice
flours had varied proximate compositions. The moisture contents of the rice flours in this
study were similar to those reported by Kraithong et al. [10] for Thai pigmented and non-pig-
mented rice flour (5.47–9.87%), but they were lower than those reported previously for Thai
jasmine red rice flour (13.3%) [17] and whole flour from Thai purple rice (11.57%) [18]. The
moisture content of the flours in this investigation, however, was less than the critical moisture
content of 13% [19]. The higher the moisture level of flour, the more likely it is to be spoiled by
microorganisms. As a result, the flours in this study should have a good shelf life. The protein
amount found in this study was comparable to that found in Thai purple rice flours (6.6–
13.0%) [18], black glutinous rice flour (8.0%) [20], and black rice variety (8.0%) [21]. Gener-
ally, rice grain storage proteins are composed of albumins, globulins, glutelins and prolamins,
which are soluble in water, salt solution, dilute acid or alkaline solutions, and aqueous alcohol,
respectively [22]. The ash content of rice flours ranged from 0.38% to 2.22%. The ash content
of 1.8% on brown rice flours reported by Islam et al. [21] was comparable to the values (2.0–
2.2%) found in our investigation. The results of this study were within the range of 0.43–2.34%
reported by Devi et al. [19] on 92 rice varieties, including indigenous, improved, and fragrant
types. The ash content reflected the mineral content in the sample. Thomas et al. [20] reported
ash content of 0.90% in black rice and 0.39% in white rice. The CMRF was significantly lower
than the brown rice flour in this investigation, indicating the same tendency. When compared
to the CMRF, this revealed that brown rice flours could be key sources of minerals.
The crude fiber contents of the rice flours in this study (0.49–2.85%) were similar to those
found in earlier studies. The fiber content of 1.23–1.56% was reported in brown rice flour [3,
23] whereas the content of 0.34% was reported in CMRF [24]. According to Oko and Onyek-
were [24], the average content of fiber in well-milled rice flour was around 0.5–1.0%. The
higher fiber content in brown rice flour was due to the presence of bran fraction. Dietary fiber
has a number of health benefits, including lowering blood cholesterol and/or glucose levels,
acting as a laxative, and lowering the risk of colon cancer, heart disease, and obesity [25]. The
principal components of dietary fiber present in rice are arabinoxylans, β-glucans, cellulose,
and hemicellulose [26, 27]. The number and quantity of these non-starch polysaccharides in
rice, on the other hand, are determined by the rice cultivar, milling degree, and water solubil-
ity. Again, the higher fat content in brown rice flour was due to the presence of bran fraction.
Here, the fat contents for the three brown rice varieties were among the values of brown rice
varieties (0.2–3.86%) reported by Devi et al. [19] and Ye et al. [28]. The fat content of CMRF in
this research was lower than Phitsanulok white rice (1.13%) [10]. Fat content had a substantial
positive link with sensory total points of eating quality of rice, according to Ke-xin et al. [29].
In this study, carbohydrate was higher than 85%. The carbohydrate content of 80.35–91.33%
has been reported in CMRF [23] whereas the value of 77.31% was found in brown rice flour
[21]. A lower carbohydrate content in the brown rice varieties was attributed to higher con-
tents of protein, fat, ash, and fiber.
Herein, the amylose contents in the rice flours were around 20–21%. Juliano [30] classified
the rice based on the amylose content, namely waxy (0–5%), very low (5–12%), low (12–20%),
intermediate (20–25%), and high (25–33%). This implied that the brown rice flours in this
PLOS ONE | https://doi.org/10.1371/journal.pone.0255694 August 3, 2021
8 / 17
PLOS ONE Characterization of Thai rice flour
study were classified as a low amylose type whereas the CMRF was an intermediate type. The
results were in agreement with Ye et al. [28] who reported the amylose content of 10.4–26.5%
in India rice flour. According to Falade and Christopher [31], low amylose rice flour provides
moistness, softness, and chewiness to product textures.
Brown rice is high in phenolic compounds, according to several studies [32–35]. The phe-
nolic compounds are collectively composed of phenoic acids, flavonoids, and anthocyanins/
proanthocyanidins [32]. Several phenolic compounds such as gallic acid, protocatechuic acid,
p-hydroxybezoic acid, vanillic acid, syringic acid, chlorogenic acid, caffeic acid, p-coumaric
acid, sinapic acid, ferulic acid, cinnamic acid, ellagic acid, luteolin, apigenin, tricin, quercetin,
kaempferol, isorhamnetin, myricetin, etc have been identified in brown rice [32–35]. The TPC
obtained from this study was within those reported by Ponjanta et al. [5] who found TPC
around 0.3–2.4 mg GAE/g in Thai rice flours. The highest TPC in the brown rice flours was
found in NKRF/KMRF, followed by JMRF, and CMRF. This implied that indigenous rice
flour is a better source of phenolic compounds. The variation of TPC in brown rice varieties
was governed by genotype, cultivation techniques, and environmental factors [36]. Grain phe-
nolic compounds were eliminated during rice milling and flour preparation, as indicated by a
low TPC in CMRF. In comparison to unmilled rice, Sapna et al. [37] found that the TPC of
milled rice retained roughly 73–89% of the overall TPC.
However, the extractability of the phenolic compounds may also influence the TPC of the
flours. In rice, phenolics can be found in three different forms including free, soluble-conju-
gated, and bound forms [33]. Free phenolics are the most readily available for absorption in
the small intestine [33]. The acidified solvent can be used to enhance the extraction of bound
phenolics (e.g. anthocyanins) from pigmented rice [35]. However, in this study, the non-pig-
mented rice were used. So, the extraction with 80% ethanol, one of the common solvents used
for TPC isolation, at 40˚C for 24 h under continuous stirring was reasonable for TPC
recovery.
For the antioxidant activity of ethanolic extracts of rice flours indicated by DPPH• scaveng-
ing activity, the CMRF extract showed higher antioxidant capacity than those of the brown
rice flours. Maisuthisakul and Changchub [38] reported that white rice varieties had relatively
higher antioxidant power than red rice. In their study using ethanol with shaking method, the
DPPH• inhibitions were ranged from 56.21% to 70.21% for white rice, and from 54.19% to
56.14% for red rice. However, Muntana and Srihanam [39] reported lower antioxidant activity
for white rice compared to red and black rice. Although phenolic substances are responsible
for the antioxidant activity of plant materials, antioxidant potency is not solely characterized
by the TPC [40]. Generally, greater antioxidant activity was positively correlated with the TPC
in extracts [41]. However, in this study, the phenolic compounds were found at very low con-
tent, and thus the free radical scavenging capability varied very slightly. In this investigation,
no link was found between TPC inhibition and DPPH• inhibition. Brown rice flour extract
with a higher TPC had a lower DPPH• inhibition. Sompong et al. [42] also reported a negative
correlation between TPC and DPPH• scavenging activity. Phenolics are not the only determi-
nant of the antioxidant power of plant materials but other phytochemicals with antioxidant
activity (e.g. tocopherols, tocotrienals, γ-oryzanol, phytic acid, and carotenoids) can also be
included [32, 43, 44]. Those phytochemicals can be co-extracted with the solvent used for TPC
extraction and can definitely be found in the final extract [32, 33, 44].
For the FTIR spectra, all the samples showed similar spectra with different peak intensities
indicating the same functional groups with different contents were found. Flores-Morales
et al. [45] reported that the bands at 400 cm-1 and 700 cm-1 were associated with the structural
vibration of amylose and amylopectin. In addition, the same functional groups were observed
in rice flours, including–OH group (3,298–3,278 cm-1),–C-H stretch (2,924–2,854 cm-1),–
PLOS ONE | https://doi.org/10.1371/journal.pone.0255694 August 3, 2021
9 / 17
PLOS ONE Characterization of Thai rice flour
C = O group (1,710–1,000 cm-1), mostly aldehyde group, glucose, cyclodextrin, and–C-OH
bending vibrations (856–573 cm-1). The bands at 3,000–3,500 cm-1 corresponded to the O-H
stretching vibration regions [46]. The results were in agreement with the earlier reports of
Falade and Christopher [31] and Anugrahati et al. [47]. The peak at 1,047 cm-1 represented the
ordered structure of starch and the peak at 1,022 cm-1 was referred to the amorphous structure
of starch [48]. Thus, the degree of order in the starch can be estimated from the ratio of peaks
at 1,047 cm-1/1,022 cm-1 [48]. From the calculation, the ratio 1,047 cm-1/1,022 cm-1 of all sam-
ples was fallen in the narrow range (1.24 for CMRF, 1.19 for JMRF, 1.16 for NKRF, and 1.14
for KMRF). Results indicated that the arrangement of starch of all the rice flours in this study
was in the ordered structure. The use of FIIR analysis for the detection of rice globulin second-
ary structure is based on the amide I region composed of C = O stretching vibrations in the
region of 1,611–1,690 cm-1 [13]. From the results, CMRF had the lowest intensity in this
region, followed by KMRF, and JMRF/NKRF, which was in agreement with the protein con-
tents in the flours.
The color of the rice flours were different. The highest L� value, which express the lightness,
was found in CMRF because the CMRF was prepared from the well-milled rice. CMRFs are
generally white in color, hence the L� value of 93.10 reported for CMRF by Rosniyana et al.
[23] was not surprising. The obvious differences in rice flour color were affected by their poly-
phenols which related to the color of the seed [10]. Since the milling and polishing techniques
were used in the CMRF manufacture. As a result, all brown rice flours had greater a� and b�
values with lower L� values, whereas CMRF had a negative a� value, a low b� value, and the
highest L� value, indicating the darker of the former. Higher a� and b� values with lower L�
value in the three brown rice flours from this study were related to their total phenolic contents
(Table 1). The predominance of phenolics in rice hulls and bran layers led to the increased a�
and b� of the brown rice [17, 49]. It has been reported that rice with light brown pericarp color
presented mainly low molecular weight phenolics whereas in those with dark pericarp color
contained the compounds with higher molecular weight [50]. A negative correlation between
TPC and L� value was reported in various non-pigmented and pigmented rice flour samples
[5].
The rice flours had bulk densities ranging between 1.36 g/cm3 to 1.83 g/cm3. The bulk den-
sity range of 0.65 g/cm3 to 0.89 g/cm3 obtained from some rice flours in Nigeria by Falade and
Christopher [31]. The bulk density of flour is generally affected by the composition and parti-
cle size. The variation in starch content may have caused a modest fluctuation in bulk density.
Iwe et al. [51] suggested that starch content increased the bulk density of flours. The higher
carbohydrate content of CMRF may account for its larger bulk density when compared to the
others, as shown in Table 1. According to Appiah et al. [15], the higher the carbohydrate con-
tent the greater the bulk density. Therefore, the highest bulk density of CMRF/NKRF sug-
gested their suitability be used as a thickening agent in food products.
Other than nutritional aspects, functional features of food are those that are essential for
successfully utilizing the food source [52]. The composition and structure of food components,
as well as their interactions, influence functional properties [52]. WAC represents the ability of
a product to interact with water. It is useful to increase yield and consistency and offer body to
the food [53]. WAC of the rice flours ranged from 0.71 to 1.17 g/g. Rosniyana et al. [23]
reported the WAC of 0.88 g/g on CMRF. The highest WAC of KMRF and JMRF could be due
to the presence of a higher amount of fiber and protein content in these flours. The result
implied that KMRF and JMRF would yield thicker pastes when mixed with water. Flours with
high WAC could contain more hydrophilic proteins [15]. It has been reported that WAC is an
important functional attribute considered in the development of cereal based food since high
WAC might improve product cohesiveness [54]. The variation in WAC between flour may be
PLOS ONE | https://doi.org/10.1371/journal.pone.0255694 August 3, 2021
10 / 17
PLOS ONE Characterization of Thai rice flour
attributed to different protein concentrations, their degree of association with water, and their
conformational aspects. Baxter et al. [55] suggested that the presence of albumin in rice starch
facilitated the uptake of water by starch in during cooking due to the formation of protein-
water-starch interactions. The presence of globulin initially enhanced the rate of water absorp-
tion by rice starch during cooking, but the presence of glutelin slowed it down, according to
Baxter et al. [56].
OAC represents the physical entrapment of oil [3]. The OAC of the rice flours in this study
ranged from 1.39 g/g to 2.49 g/g. Comparatively, the OAC was higher than reported on FARO
44 rice (OAC = 0.46 g/g) by Iwe et al. [45]. Kraithong et al. [10] reported an OAC range of
1.11–1.34 g/g for some Thai organic rice flours. OAC is a prime functionality that intensifies
the mouthfeel while maintaining the flavor of food products. High OAC of the flour suggests
its usefulness in lipid-containing foods e.g. bakery products. The higher OAC recorded in the
rice flours may be due to the presence of more hydrophobic proteins with superior lipid bind-
ing efficacy [28]. Albumin (water-soluble), globulin (salt-soluble), glutelin (alkali-soluble), and
prolamin (alcohol-soluble) are the four components of rice protein [57]. The two primary pro-
teins are glutelin (about 80%) and globulin (about 12%), while albumin (about 5%) and prola-
min (about 3%) are minor ones [57]. Both glutelin and prolamin are hydrophobic proteins
that accumulate in small vacuoles or protein bodies [58]. As a result, rice had a hydrophobic
protein level of more than 80%. Due to the similar OAC, any of the rice flours could be chosen
and employed in food formulations requiring OAC, such as soups, cakes, and sausages [3].
The differences in swelling power and solubility may be due to the amount of protein, amy-
lose content, and lipids [59]. The current findings were consistent with previous research,
which reported swelling power values in rice flours of several rice varieties ranging from 4.7 g/
g to 16.23 g/g [18, 60]. More specifically, the swelling power of the flours in this case was simi-
lar to Jamal et al. [6] who reported swelling power of 5.74–7.64 g/g for rice flours. For water
solubility of rice flour, Kraithong et al. [10] reported the ranges of 2.97–7.05% for some Thai
organic rice flours. A lower swelling power and solubility in CMRF can be attributed to a
higher degree of intermolecular association and higher amylose content compared to the other
rice flours. Swelling is facilitated by amylopectin and disturbed by amylose [61]. The swelling
factor of starches is decreased upon the increase of amylose content in starches.
When flour is mixed with water and heated, the LGC represents the lowest amount of flour
required to gel. Thus, LGC is an important index of gel-forming ability. Here, the LGC values
ranged from 4.0–8.0%. The LGC of the flours was comparable to that of rice flour (6%)
reported by Chandra [62]. However, higher LGC values were reported in some rice flours,
ranging from 8.0% to 22.0% [3, 17, 63]. A lower LGC suggests stronger gelling capacity in gen-
eral. The result suggested that KMRF was easier to form gel than JMRF, and NKRF/CMRF,
respectively. The presence of varying levels of protein, carbohydrates, and lipids in the rice
varieties can cause the differences in LGC [63]. The FC of the flours in this study (1.30–2.60%)
was lower than the previous reports. The FC of the rice flour of 3.52% has been reported by
Chandra [62]. However, higher foaming capacity of 10.40% has been reported for polished
FARO 44 rice variety flour [51]. This might be due to its relatively high protein content (16%)
of the flour form polished FARO 44 rice variety. The brown rice varieties with higher protein
contents (Table 1) had slightly higher foam capacity than the CMRF. A lower protein content
in the rice flours from this experiment may have resulted in the general lower FC of the flours.
The FC of flour is dependent on the protein content because proteins are surface-active agents
which can create the film at the interphase to trap the air bubbles [17, 62]. The FC seemed to
be related to the solubility, in which the lowest value was found in CMRF (Table 2). The foam
capacity of a food material is governed by several factors such as the type and concentration of
PLOS ONE | https://doi.org/10.1371/journal.pone.0255694 August 3, 2021
11 / 17
PLOS ONE Characterization of Thai rice flour
protein, solubility, and other parameters [64]. The low FC was desirable for flours intended for
the application of some bakery products such as crackers and biscuits [17, 62].
Flours are commonly mixed with water for proper hydration before being subjected to ther-
mal treatments that can cause a variety of physicochemical changes in the various components
of flours, including starch gelatinization, protein denaturation, enzyme inactivation, and inter-
actions between them [65]. Flours can serve as a thickening agent, gelling agent, binder, and/
or stabilizer in final food items as a result of various physicochemical processes. Some of the
flour pastes can form a gel after cooling and storage, which is a viscoelastic solid-in-liquid col-
loid with a definite structure but no fluidity [66]. Pasting property of flours is one of the impor-
tant functional characteristics that govern how they are used in food products. The pasting
profiles of flour are species-dependent and are influenced by the chemical composition of
flour [67]. The molecular interactions of starch with other components such as proteins, lipids
and non-starch polysaccharides, are some of the factors involved in determining cooking
behavior of rice [22]. Brown rice flours had a higher pasting temperature than CMRF
(p < 0.05). The CMRF was made from white rice, which was easier to absorb water, swell, and
gelatinize [68]. Thus, the temperature needed for the gelatinization of CMRF was lower than
the others. Cooking the brown rice flours would, therefore, require more energy than the
CMRF. These data back with the theory that rice varieties with higher amylose content require
less time to cook. Riceberry, a pigmented rice flour, had significantly higher pasting tempera-
ture (87.8˚C) than Phitsanulok (white rice flour) (86.8˚C) as reported by Kraithong et al. [10].
White rice flours have lower pasting temperatures than brown rice flours, according to
research findings. This could be due to poor interactions between starch granules, which
results in the lowest pasting temperature. When a starch suspension was subjected to heat
above a critical temperature, granules underwent swelling and amylose dripped into the aque-
ous phase, resulting in increased viscosity [69]. The creation of a starch-lipid complex in
brown rice varieties prevents water from reaching the starch granules, necessitating a higher
temperature because the starch granules have stronger bonds.
The peak viscosity of the rice flours in this study was ranged from 847 BU to 2,250 BU.
However, a higher peak viscosity values of 2,376 to 3,988 BU was reported for six Nigerian rice
flours [31]. Peak viscosity is the indicator of starch granule swelling and high-value peak vis-
cosity indicates a high capacity of swelling of starch [69]. Starches with high peak viscosity are
possible to show high breakdown values, leading to weak gels. Such gels are prone to be
destroyed under shear and heat. This means that peak viscosity and break down have a positive
relationship. The higher peak viscosity in CMRF was due to a larger breakdown during heating
due to weaker heat and shear stress resistance [55]. As a result, the fact that CMRF had the
greatest breakdown value of 321 BU was unsurprising. The high peak viscosity in the CMRF
may be because of its initial high amylose (starch content). The low peak viscosity in the
brown rice flours may be due to their initial high fat and protein contents. Swelling power is
retarded by the presence of fat and protein. Flours with a high starch content have a high gela-
tinization and swelling index. According to the findings, the CMRF and JMRF could be useful
thickening agents. As stated by Kraithong et al. [10], higher peak viscosity (4,067 BU) was
observed in brown jasmine rice flour and the lowest value (2,260 BU) was from white rice
(Phitsanulok). Rice flour with high viscosity is commonly used as a thickening agent in high
viscosity food [28]. Trough viscosity is the viscosity reaching the minimum after cooling [51].
CMRF and JMRF with relatively high amylose content had a higher trough viscosity. The same
trend was described by Sompong et al. [42] who reported a positive correlation between trough
viscosity and amylose content.
The setback is the tendency of flours to reassociate and retrograde on cooling. Setback val-
ues were correlated with the gelling ability of starches to form semi-solid pastes. The higher
PLOS ONE | https://doi.org/10.1371/journal.pone.0255694 August 3, 2021
12 / 17
PLOS ONE Characterization of Thai rice flour
setback values for the JMRF suggested that it had a greater chance to retrograde than the other
rice flours. Flours with resistant to retrograde are suitable for soups and sauces because they
can prevent the loss of viscosity and precipitation [69]. A high setback is related with the syner-
esis. Thus, the lower setback values in NKRF and CMRF indicated that it would form a better
flour paste that could find applications in the confectionery industries. Breakdown viscosity is
indicative of paste stability [31]. The higher the breakdown viscosity, the lower the ability of
the starch to resist shear stress and heating [31]. Thus, JMRF and KMRF, with the lowest
breakdown values might be able to withstand more heating and shear stress than others.
The final viscosity is the tendency of flours to form paste or gel after cooling. Several studies
showed an increase in final peak viscosity than their corresponding peak viscosity. The same
trend was observed in this study for JMRF and KMRF but not for NKRF and CMRF. Other
chemical components (e.g. lipid and protein contents) and swelling power may all have a part
in flour gelatinization, explaining the observed trend. In general, peak, trough, and breakdown
viscosity of rice flour were positively correlated with amylose content [28]. Thus, the differ-
ences in final viscosity values among the rice flours may be due to the differences in amylose
content and other chemical constituents, particularly the type and amount of proteins. Baxter
et al. [56] found that the presence of glutelin in rice starch caused an increase in pasting tem-
perature but a decrease in the viscosity parameters of the starch paste. Contrastingly, the pres-
ence of globulin resulted in a decrease in all the pasting parameters. In the case of prolamin,
the presence of prolamin in rice starch facilitated the water absorption during cooking but the
gelatinized starch absorbed less water compared with control samples without prolamin [70].
Conclusion
This study revealed the variations in chemical, physical, functional, and pasting properties that
exist among four rice flours in Thailand. The compositional variations greatly influenced the
techno-functionality and final quality of rice flours. The brown rice flours in this study were
classified as a low amylose type whereas the CMRF was an intermediate type. All rice flours
had low total phenolic content with moderate DPPH• scavenging activity. Functionalities and
pasting properties of the flours varied among the cultivars. Those variations may be due to the
differences in genetic constitution and processing condition. In addition to their nutritional
and technological benefits, local rice flours have been revealed to be a potential source of bio-
active secondary metabolites, which might be used as functional food ingredients in both
domestic and industrial applications.
Acknowledgments
We would like to thank Food Technology and Innovation Center of Excellence, Walailak Uni-
versity for providing the scientific and technological equipment for this research.
Author Contributions
Conceptualization: David Oppong, Worawan Panpipat, Manat Chaijan.
Data curation: David Oppong.
Formal analysis: Worawan Panpipat.
Funding acquisition: Manat Chaijan.
Investigation: David Oppong.
Methodology: David Oppong, Worawan Panpipat, Manat Chaijan.
PLOS ONE | https://doi.org/10.1371/journal.pone.0255694 August 3, 2021
13 / 17
PLOS ONE Characterization of Thai rice flour
Resources: Worawan Panpipat, Manat Chaijan.
Supervision: Worawan Panpipat, Manat Chaijan.
Validation: Manat Chaijan.
Writing – original draft: David Oppong, Worawan Panpipat, Manat Chaijan.
Writing – review & editing: David Oppong, Worawan Panpipat, Manat Chaijan.
References
1. Puncha-arnon S, Uttapap D. Rice starch vs. rice flour: Differences in their properties when modified by
heat-moisture treatment. Carbohyd Polym. 2013; 91:85–91. https://doi.org/10.1016/j.carbpol.2012.08.
006 PMID: 23044108
2. Ofoedu CE, Akosim CQ, Iwouno JO, Obi CD, Shorstkii I, Okpala COR. Characteristic changes in malt,
wort, and beer produced from different Nigerian rice varieties as influenced by varying malting condi-
tions. PeerJ. 2021; 9:e10968. https://doi.org/10.7717/peerj.10968 PMID: 33777523
3. Ajayi FF, Kusimo OO, Henshaw FO, Ogori AF. Proximate composition and functional properties of rice
(Oryza sativa L.) germinated-cowpea (Vigna unguiculata L.) flour blends. Innov Technol Agric. 2017;
1:32–40.
4. Saleh AS, Wang P, Wang N, Yang L, Xiao Z. Brown rice versus white rice: Nutritional quality, potential
health benefits, development of food products, and preservation technologies. Compr Rev Food Sci
Food Saf. 2019; 18:1070–1096. https://doi.org/10.1111/1541-4337.12449 PMID: 33336992
5. Ponjanta J, Chomsri NO, Meechoui S. Correlation of pasting behaviors with total phenolic compounds
and starch digestibility of indigenous pigmented rice grown in upper Northern Thailand. Funct Foods
Health Dis. 2016; 6:133–143.
6.
Jamal S, Qazi IM, Ahmed I. Comparative studies on flour proximate compositions and functional proper-
ties of selected Pakistani rice varieties. Pakistan Acad Sci. 2016; 53:47–56.
7. Banchuen J, Thammarutwasik P, Ooraikul B, Wuttijumnong P, Sirivongpaisal P. Increasing the bio-
active compounds contents by optimizing the germination conditions of Southern Thai brown rice.
Songklanakarin J Sci Technol. 2010; 32:219–230.
8. Chaijan M, Panpipat W. Nutritional composition and bioactivity of germinated Thai indigenous rice
extracts: A feasibility study. PLoS One. 2020; 15:e0237844. https://doi.org/10.1371/journal.pone.
0237844 PMID: 32834013
9. AOAC. Official methods of analysis. 16th ed. Washington, DC: Association of Official Analytical Chem-
ists; 2000.
10. Kraithong S, Lee S, Rawdkuen S. Physicochemical and functional properties of Thai organic rice flour. J
Cereal Sci. 2018; 79:259–266.
11. Sungpud C, Panpipat W, Sae-Yoo A, Chaijan M. Polyphenol extraction from mangosteen (Garcinia
mangostana Linn) pericarp by bio-based solvents. Int Food Res J. 2020; 27:111–120.
12.
Limsuwanmanee J, Chaijan M, Manurakchinakorn S, Panpipat W, Klomklao S, Benjakul S. 2014. Anti-
oxidant activity of Maillard reaction products derived from stingray (Himantura signifier) non-protein
nitrogenous fraction and sugar model systems. LWT-Food Sci Technol. 2014; 57:718–724.
13. Ellepola SW, Choi SM, Ma CY. Conformational study of globulin from rice (Oryza sativa) seeds by Fou-
rier-transform infrared spectroscopy. Int J Biol Macromol. 2005; 37:12–20. https://doi.org/10.1016/j.
ijbiomac.2005.07.008 PMID: 16140371
14. Goula AM, Adamopoulos KG. Effect of maltodextrin addition during spray drying of tomato pulp in dehu-
midified air: II. Powder properties. Dry Technol. 2008; 26:726–737.
15. Appiah F, Oduro I, Ellis WO. Functional properties of Artocarpus altilis pulp flour as affected by fermen-
tation. Agr Biol J N Am. 2011; 2:773–779.
16. Sulieman MA, El Tinay AH, Elkhalifa AEO, Babiker EE, Elkhalil EA. Solubility as influenced by pH and
NaCl concentration and functional properties of lentil proteins isolate. Pakistan J Nutr. 2006; 5:589–
593.
17. Klunklin W, Savage G. Physicochemical, antioxidant properties and in vitro digestibility of wheat–purple
rice flour mixtures. Int J Food Sci Technol. 2018; 53:1962–1971.
18. Navaratne SB. Selection of polymer based packing material in packing of hygroscopic food products for
long period of storage. Eur Int J Sci Technol. 2013; 2:1–6.
PLOS ONE | https://doi.org/10.1371/journal.pone.0255694 August 3, 2021
14 / 17
PLOS ONE Characterization of Thai rice flour
19. Devi GN, Padmavathi G, Babu VR, Waghray K. Proximate nutritional evaluation of rice (Oryza sativa
L.). J Rice Res. 2015; 8:23–32.
20.
21.
Thomas R, Wan-Nadiah WA, Bhat R. Physiochemical properties, proximate composition, and cooking
qualities of locally grown and imported rice varieties marketed in Penang, Malaysia. Int Food Res J.
2013; 20:1345–1351.
Islam MZ, Taneya MLJ, Shams-Ud-Din M, Syduzzaman M, Hoque MM. Physicochemical and func-
tional properties of brown rice (Oryza sativa) and wheat (Triticum aestivum) flour and quality of compos-
ite biscuit made thereof. Agri. 2012; 10:20–28.
22. Balindong JL, Ward RM, Liu L, Rose TJ, Pallas LA, Ovenden BW, et al. Rice grain protein composition
influences instrumental measures of rice cooking and eating quality. J Cereal Sci. 2018; 79:35–42.
23. Rosniyana A, Hazila KK, Abdullah SNS. Characteristics of local rice flour (MR 220) produced by wet
and dry milling methods. J Trop Agric Food Sci. 2016; 44:147–155.
24. Oko AO, Onyekwere SC. Studies on the proximate chemical composition, and mineral element con-
tents of five new lowland rice varieties planed in Ebonyi State. Int J Biotech Bio. 2010; 6:949–956.
25.
Zhao G, Zhang R, Dong L, Huang F, Tang X, Wei Z, et al. Particle size of insoluble dietary fiber from
rice bran affects its phenolic profile, bioaccessibility and functional properties. LWT-Food Sci Technol.
2018; 87:450–456.
26. Rao RSP, Muralikrishna G. Non-starch polysaccharide-phenolic acid complexes from native and germi-
nated cereals and millet. Food Chem. 2004; 84:527–531.
27.
Lai VMF, Lu S, He WH, Chen HH. Non-starch polysaccharide compositions of rice grains with respect
to rice variety and degree of milling. Food Chem. 2007; 101:1205–1210.
28. Ye L, Wang C, Wang S, Zhou S, Liu X. Thermal and rheological properties of brown flour from Indica
rice. J Cereal Sci. 2016; 70:270–274.
29. Ke-xin Q, Lin-lin L, Tian-yi L, Wen-hong C, Yan-guo S. Correlation between physicochemical properties
and eating qualities of rice. J Northeast Agric Univ. 2014; 21:60–67.
30.
31.
Juliano BO. Structure, chemistry, and function of the rice grain and its fractions. Cereal Foods World.
1992; 37:772–779.
Falade KO, Christopher AS. Physical, functional, pasting and thermal properties of flours and starches
of six Nigerian rice cultivars. Food Hydrocolloids. 2015; 44:478–490.
32. Goufo P, Trindade H. Rice antioxidants: phenolic acids, flavonoids, anthocyanins, proanthocyanidins,
tocopherols, tocotrienols, γ-oryzanol, and phytic acid. Food Sci Nutr. 2014; 2:75–104. https://doi.org/
10.1002/fsn3.86 PMID: 24804068
33. Ravichanthiran K, Ma ZF, Zhang H, Cao Y, Wang CW, Muhammad S, et al. Phytochemical profile of
brown rice and its nutrigenomic implications. Antioxidants. 2018; 7:71. https://doi.org/10.3390/
antiox7060071 PMID: 29789516
34. Ye L, Zhou S, Liu L, Liu L, Waters DL, Zhong K, et al. Phenolic compounds and antioxidant capacity of
brown rice in China. Int J Food Eng. 2016; 12:537–546.
35. Ciulu M, Ca´diz-Gurrea MDLL, Segura-Carretero A. Extraction and analysis of phenolic compounds in
rice: a review. Molecules. 2018; 23:2890. https://doi.org/10.3390/molecules23112890 PMID: 30404149
36. Mpofu A, Sapirstein HD, Beta T. Genotype and environmental variation in phenolic content, phenolic
acid composition, and antioxidant activity of hard spring wheat. J Agric Food Chem. 2006; 54:1265–
1270. https://doi.org/10.1021/jf052683d PMID: 16478246
37. Sapna I, Kamaljit M, Priya R, Jayadeep PA. Milling and thermal treatment induced changes on phenolic
components and antioxidant activities of pigmented rice flours. J Food Sci Technol. 2019; 56:273–280.
https://doi.org/10.1007/s13197-018-3487-1 PMID: 30728569
38. Maisuthisakul P, Changchub L. Effect of extraction on phenolic antioxidant of different Thai rice (Oryza
sativa L.) genotypes. Int J Food Prop. 2014; 17:855–865.
39. Muntana N, Srihanam P. Study on total phenolic contents and their antioxidant activities of Thai white,
red and black rice bran extracts. Pakistan J Bio Sci. 2010; 13:170–174.
40. Shahidi F, Janitha PK, Wanasundara PD. Phenolic antioxidants. Crit Rev Food Sci Nutr. 1992; 32:67–
103. https://doi.org/10.1080/10408399209527581 PMID: 1290586
41. Bryngelsson S, Dimberg LH, Kamal-Eldin A. Effects of commercial processing on levels of antioxidants
in oats (Avena sativa L.). J Agric Food Chem. 2002; 50:1890–1896. https://doi.org/10.1021/jf011222z
PMID: 11902929
42. Sompong R, Siebenhandl-Ehn S, Linsberger-Martin G, Berghofer E. Physicochemical and antioxidative
properties of red and black rice varieties from Thailand, China and Sri Lanka. Food Chem. 2011;
124:132–140.
PLOS ONE | https://doi.org/10.1371/journal.pone.0255694 August 3, 2021
15 / 17
PLOS ONE Characterization of Thai rice flour
43. Kaur M, Asthir B, Mahajan G. Variation in antioxidants, bioactive compounds and antioxidant capacity
in germinated and ungerminated grains of ten rice cultivars. Rice Sci. 2017; 24:349–359.
44.
45.
46.
Ti H, Guo J, Zhang R, Wei Z, Liu L, Bai Y, et al. Phenolic profiles and antioxidant activity in four tissue
fractions of whole brown rice. RSC Adv. 2015; 5:101507–101518.
Flores-Morales A, Jime´nez-Estrada M, Mora-Escobedo R. Determination of the structural changes by
FT-IR, Raman, and CP/MAS 13C NMR spectroscopy on retrograded starch of maize tortillas. Carbohy-
drate Polym. 2012; 87:61–68.
Lohumi S, Lee S, Lee WH, Kim MS, Mo C, Bae H, et al. Detection of starch adulteration in onion powder
by FT-NIR and FT-IR spectroscopy. J Agric Food Chem. 2014; 62:9246–9251. https://doi.org/10.1021/
jf500574m PMID: 25188555
47. Anugrahati NA, Pranoto Y, Marsono Y, Marseno DW. Structural changes in cooked rice treated with
cooling-reheating process and coconut milk addition as observed with FT-IR and 13C NMR. Agritech.
2017; 37:78–81.
48.
49.
Zhang Y, Chen C, Chen Y, Chen Y. Effect of rice protein on the water mobility, water migration and
microstructure of rice starch during retrogradation. Food Hydrocolloids. 2019; 91:136–142.
Jang S, Xu Z. Lipophilic and hydrophilic antioxidants and their antioxidant activities in purple rice bran. J
Agric Food Chem. 2009; 57:858–862. https://doi.org/10.1021/jf803113c PMID: 19138081
50. Walter M, Marchesan E. Phenolic compounds and antioxidant activity of rice. Braz Arch Biol Technol.
2011; 54:371–377.
51.
52.
Iwe MO, Onyeukwu U, Agiriga AN. Proximate, functional and pasting properties of FARO 44 rice, Afri-
can yam bean and brown cowpea seeds composite flour. Cogent Food Agric. 2016; 2:1142409.
Joshi AU, Liu C, Sathe SK. Functional properties of select seed flours. LWT-Food Sci Technol. 2015;
60:325–231.
53. Osundahunsi OF, Fagbemi TN, Kesselman E, Shimoni E. Comparison of the physicochemical proper-
ties and pasting characteristics of flour and starch from red and white sweet potato cultivars. J Agric
Food Chem. 2003; 51:2232–2236. https://doi.org/10.1021/jf0260139 PMID: 12670162
54.
Ideani IA, Jideani VA. Developments on the cereal grains Digitaria exilis (acha) and Digitaria iburua
(iburu). J Food Sci Technol. 2011; 48:251–259. https://doi.org/10.1007/s13197-010-0208-9 PMID:
23572743
55. Baxter G, Zhao J, Blanchard C. Albumin significantly affects pasting and textural characteristics of rice
flour. Cereal Chem. 2010; 87:250–255.
56. Baxter G, Blanchard C, Zhao J. Effects of glutelin and globulin on the physicochemical properties of rice
starch and flour. J Cereal Sci. 2014; 60:414–420.
57.
Ju ZY, Hettiarachchy NS, Rath N. Extraction, denaturation and hydrophobic properties of rice flour pro-
teins. J Food Sci. 2001; 66:229–232.
58. Gonza´ lez-Pe´ rez S, Arellano JB. Vegetable protein isolates. In GO Phillips PA Williams, editors. Hand-
book of Hydrocolloids. USA: Woodhead Publishing; 2009. p. 383–419.
59. Yadav BS, Yadav RB, Kumari M, Khatkar BS. Studies on suitability of wheat flour blends with sweet
potato, colocasia and water chestnut flours for noodle making. LWT-Food Sci Technol. 2014; 57:352–
358.
60. Chaiwanichsiri S, Thumrongchote D, Suzuki T, Laohasongkram K. Properties of non-glutinous Thai rice
flour: effect of rice variety. Res J Pharm Biol Chem Sci. 2012; 3:150–164.
61. Ekanayake EWMDS, Navaratne SB, Wickramasinghe I Abeysundara AT. Determination of changes in
amylose and amylopectin percentages of cowpea and green gram during storage. Nutr Food Sci Int J.
2018; 6:555690.
62. Chandra SS. Assessment of functional properties of different flours. Afr J Agric Res. 2013; 8:4849–
4852.
63.
Thomas R, Bhat R, Kuang YT, Abdullah WNW. Functional and pasting properties of locally grown and
imported exotic rice varieties of Malaysia. Food Sci Technol Res. 2014; 20:469–477.
64. Akubor PI, Eze JI. Quality evaluation and cake making potential of sun and oven dried carrot fruit. Int J
Biosci. 2012; 2:19–27.
65. Yuan TZ, Liu S, Reimer M, Isaak C, Ai Y. Evaluation of pasting and gelling properties of commercial
flours under high heating temperatures using Rapid Visco Analyzer 4800. Food Chem. 2021;
344:128616. https://doi.org/10.1016/j.foodchem.2020.128616 PMID: 33243559
66.
Tabilo-Munizaga G, Barbosa-Ca´novas GV. Rheology for the food industry. J Food Eng. 2005; 67:147–
156.
67. Blazek J, Copeland L. Pasting and swelling properties of wheat flour and starch in relation to amylose
content. Carbohydr Polym. 2008; 71:380–387.
PLOS ONE | https://doi.org/10.1371/journal.pone.0255694 August 3, 2021
16 / 17
PLOS ONE Characterization of Thai rice flour
68. Hsu RJC, Lu S, Chang YH, Chiang W. Effects of added water and retrogradation on starch digestibility
of cooked rice flours with different amylose content. J Cereal Sci. 2015; 61:1–7.
69.
Tsakama M, Mwangwela AM, Manani TA, Mahungu NM. Physicochemical and pasting properties of
starch extracted from eleven sweet potato varieties. Afr J Food Sci Technol. 2010; 1:90–98.
70. Baxter G, Blanchard C, Zhao J. Effects of prolamin on the textural and pasting properties of rice flour
and starch. J Cereal Sci. 2004; 40:205–211.
PLOS ONE | https://doi.org/10.1371/journal.pone.0255694 August 3, 2021
17 / 17
PLOS ONE |
10.1371_journal.pone.0261071 | RESEARCH ARTICLE
Effects of weather and moon phases on
emergency medical use after fall injury: A
population-based nationwide study
Min Ah Yuh1, Kisung KimID
Jinwoo Kim5, Sungyoup HongID
1*
2, Seon Hee Woo3, Sikyoung Jeong1, Juseok OhID
4,
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: Yuh MA, Kim K, Woo SH, Jeong S, Oh J,
Kim J, et al. (2021) Effects of weather and moon
phases on emergency medical use after fall injury:
A population-based nationwide study. PLoS ONE
16(12): e0261071. https://doi.org/10.1371/journal.
pone.0261071
Editor: Quan Yuan, Tsinghua University, CHINA
Received: May 8, 2021
Accepted: November 23, 2021
Published: December 31, 2021
Copyright: © 2021 Yuh et al. This is an open
access article distributed under the terms of the
Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
1 Department of Emergency Medicine, Daejeon St Mary’s Hospital, The Catholic University of Korea College
of Medicine, Seoul, Republic of Korea, 2 BioBrain Inc, Daejeon, Republic of Korea, 3 Department of
Emergency Medicine, Incheon St Mary’s Hospital, The Catholic University of Korea College of Medicine,
Seoul, Republic of Korea, 4 Department of Emergency Medicine, Uijeongbu St Mary’s Hospital, The Catholic
University of Korea College of Medicine, Seoul, Republic of Korea, 5 Department of Emergency Medical
Service, Daejeon Health Institute of Technology, Daejeon, Republic of Korea
* emhong@catholic.ac.kr
Abstract
Background
Previous studies reported that changes in weather and phases of moon are associated with
medical emergencies and injuries. However, such studies were limited to hospital or com-
munity level without explaining the combined effects of weather and moon phases. We
investigated whether changes in weather and moon phases affected emergency depart-
ment (ED) visits due to fall injuries (FIs) based on nationwide emergency patient registry
data.
Methods
Nationwide daily data of ED visits after FI were collected from 11 provinces (7 metropolitan
cities and 4 rural provinces) in Korea between January 2014 and December 2018. The daily
number of FIs was standardized into FI per million population (FPP) in each province. A mul-
tivariate regression analysis was conducted to elucidate the relationship between weather
factors and moon phases with respect to daily FPP in each province. The correlation
between weather factors and FI severity was also analyzed.
Data Availability Statement: Data are available
from the Zenodo database (DOI: 10.5281/zenodo.
5774499).
Results
Funding: We declare that this study was supported
by Daejeon St. Mary’s Hospital, Clinical Research
Institute Grant No. CMCDJ-P-2021013. The
funders had no role in study design, data collection
and analysis, decision to publish, or preparation of
the manuscript.
Competing interests: The authors have declared
that no competing interests exist.
The study analyzed 666,912 patients (418,135 in metropolitan and 248,777 in rural areas)
who visited EDs on weekdays. No regional difference was found in age or gender distribu-
tion between the two areas. Precipitation, minimum temperature and wind speed showed a
significant association with FI in metropolitan areas. In addition, sunshine duration was also
substantial risk factors for FI in rural areas. The incidence of FIs was increased on full moon
days than on other days in rural areas. Injury severity was associated with weather factors
such as minimum temperature, wind speed, and cloud cover.
PLOS ONE | https://doi.org/10.1371/journal.pone.0261071 December 31, 2021
1 / 14
PLOS ONE Effects of weather and moon phases on EM use
Conclusion
Weather changes such as precipitation, minimum temperature, and wind speed are associ-
ated with FI in metropolitan and rural areas. In addition, sunshine duration and full moon are
significantly associated with FI incidence only in rural areas. Weather factors are associated
with FI severity.
Introduction
Accurate prediction of the need for emergency medical care is critical to provide appropriate
services for patients with injuries. Therefore, many countries have an emergency medical sys-
tem data registry collected during pre-hospital and in-hospital phases to design emergency
medical service (EMS) and implement public health monitoring and planning.
Fall injury (FI) is the second major cause of accidental or unintended injury-related deaths
worldwide [1]. The mechanism of injury for falls is vertical deceleration due to the force of
gravity high place or loss of balance on a slippery surface. FIs in older or disabled individuals
increase in winter due to low temperatures and long nights [2]. The slippery ground caused by
melting ice, snow-covered ice, and ice is a typical cause of FI in winter [3]. The incidence of
FIs is known to have a seasonal variation depending on geographical location, such as coun-
tries with a cold climate (Russia, Canada, Sweden, Finland, and Norway) [4, 5] and countries
with a warm tropical climate, such as Hong Kong [6].
Weather conditions have been reported to influence the occurrence of trauma and disease.
Poor weather conditions may lead to traumatic events [7, 8]. However, other studies reported
that outdoor activities even in good weather are related to increased incidence of all kinds of
injuries [9]. If we target FI only, snowfall and icy surfaces were associated with FIs in late
autumn and winter [6, 7, 10]. But another study reported the increased frequency of FIs was
found in better weather with medium mean air temperature and atmospheric pressure during
warm season [4].
A full moon has been reportedly associated with potential emergency department (ED) visits
after traffic accidents [11] and mortality after motorcycle crashes and accidents [12]. However,
Stomp et al. reported that phases other than full moon increased ED visits after all kinds of
trauma [9]. Such difference might be attributed to the use of nationwide statistics of road car
accidents in two of the three studies, whereas Stomp et al. [9] used all types of trauma data from
one ED located in a small suburban area of Netherlands. Therefore, findings from these studies
were limited by small sample size in specific regions, restricted data sources or target injury.
In summary, previous studies evaluated the role of weather and lunar phases; however, the
findings were limited to specific region or involved a small sample size. To prevent unwanted
medical errors due to ED crowding and provide prompt and appropriate EMS, it is vital to
foresee the demand for emergency medical use due to FIs. The primary purpose of this study
was to assess the FI prevalence according to regional characteristics, weather changes and
moon phases using the nationwide longitude data. The secondary goal was to determine the
effects of weather factors and moon phases on FI severity.
Materials and methods
Study design and data collection
To analyze the association of FI incidence with weather factors and lunar phases, a nationwide
epidemiological analysis was conducted based on emergency department usage data obtained
PLOS ONE | https://doi.org/10.1371/journal.pone.0261071 December 31, 2021
2 / 14
PLOS ONE Effects of weather and moon phases on EM use
from all emergency centers in Korea. The population of mainland Korea and its affiliated
islands of 99,000 km2 is approximately 50 million. Korea has a total of 420 registered ERs that
are open to all beneficiaries without restriction. The National Emergency Department Infor-
mation System (NEDIS) operated by the National Emergency Medical Center (NEMC) pro-
spectively collected data of patients who visited all Korean EDs since 2005. This study used the
NEDIS data of patients who visited emergency centers after FI, including age, gender, region
of occurrence, onset time, injury mechanism, injury severity with Korean Triage and Acuity
Scale (KTAS), and outcome of emergency care from January 2014 to December 2018. No per-
sonal identifier was included in these data. Data were stored in a secured personal computer.
KTAS score consists of five levels of acuity: level 1 (resuscitation), level 2 (urgent), level 3
(emergent), level 4 (non-urgent), and level 5 (delayed). The KTAS was developed as a single
triage tool for emergency patients in Korea and has since become nationalized [13].
Daily weather data including precipitation, minimum temperature, mean wind speed
(wind speed), sunshine and fog duration, and cloud cover were obtained from the Korea Mete-
orological Agency (KMA). Daily precipitation was calculated as the sum of hourly measure-
ments for 24 hours. Cloud cover was calculated in integers ranging from 0 to 10 tenths based
on visual cloud cover observations from each observation site. A weather station located in the
capital city of each province in Korea was selected to represent the weather data collection
point. Moon phase data were obtained from a website (https://www.timeanddate.com/moon/
phases/south-korea/).
First, pediatric patients under the age of 15 years were excluded from the collected data. FI
patients on weekdays excluding Saturday, Sunday, and public holidays (New Year’s Day,
Lunar New Year’s Day, Children’s Day, Korean Independence Day, Lunar Thanksgiving Day
and Christmas) were also excluded from this study. We analyzed data from a total of 743
nights (182 new moon days, 186 first quarter and full moon days, and 189 3rd quarter nights).
The full moon period was determined for three days starting from one night before to one
night after the peak full moon night. The same rule was applied to other moon phases.
Annual mid-population data of each province were obtained from the central organization
for Statistics of Korea (http://kostat.go.kr/portal/eng/index.action).
Outcome measurement
We counted the daily number of patients who visited EDs after a FI for each province. The
number of daily FIs was standardized into FIs per million population (FPP) by dividing with
an annual mid-population of the province. This study compared FPP and severity of FI
between two regions: 1) metropolitan areas consisting of provinces with a population of more
than one million; and 2) rural areas without metropolitan cities within the perimeter of the
provincial limit (S1 Fig). The primary outcome was the daily number of ED visits due to FI. It
was calculated as the number of patients per million people. The secondary outcome was
injury severity of patients and was determined by the mean KTAS score of all daily FI patients
in the designated area.
Statistical analysis
The chi-square test is extremely sensitive to sample size. If the sample size is too large (> 500),
any small differences appear statistically significant [14]. Hence, we used Cramer’s V statistics
instead of Chi-square test to estimate the association of ordinary factors between two regions.
The means of the continuous variable were compared using Student’s t-test between two
regions. The number of FI events in a fixed time interval was modeled using Poisson distribu-
tion, and thus a generalized linear model (GLM) with a Poisson distribution and log-linear
PLOS ONE | https://doi.org/10.1371/journal.pone.0261071 December 31, 2021
3 / 14
PLOS ONE Effects of weather and moon phases on EM use
function was used to assess the significance of association between dependent variable (FPP
for each day) and independent variables including weather factors and lunar phases. All vari-
ables with a p value < 0.2 in the univariate analysis were entered into multivariate analysis.
The incidence risk ratios (IRRs) and their 95% confidence intervals (CIs) were calculated for
each independent variable. Theoretical FI incidence-factor curve was approximated via non-
linear curve fitting with Boltzmann sigmoidal function and illustrated with scatter plots. The
correlation between weather factors and daily mean KTAS was analyzed using Pearson corre-
lation coefficients. All statistical analyses were analyzed using Origin Pro (OriginLab, North-
ampton, MA) and Rstudio 1.4.1717 (RStudio Inc, Boston, MA). Statistically significant
difference was indicated by a p value 0.050 or less.
Ethical approval
The study protocol was reviewed and approved by the Institutional Review Board of Daejeon
St Mary’s Hospital, The Catholic University of Korea (DC21ZIS10034).
Results
Demographic characteristics of the study subjects
Of 1,476,652 people included in the registry with FI, 1,065,637 were older than 15 years in
Korea between Jan 2014 and Dec 2018 (Fig 1). FPPs were significant higher on weekends than
on weekdays (P < 0.010, S2 Fig). Hence, we excluded FI cases on weekend and holidays to pre-
vent bias. Finally, 666,912 patients (418,135 in metropolitan and 248,777 in rural areas) on
weekdays were analyzed in this study. The mean age of patients finally enrolled was 54.4 ± 20.4
years. There was no significant association with age distribution between the two regions, but
the proportion of male patients was significantly higher in the rural areas (Table 1, P < 0.01).
The distribution of FI between the two regions was balanced with no monthly difference. Of
the total patients, 69, 563, 66, 755, 69, 573, and 69,783 patients suffered FIs on the new moon,
1st quarter, full moon, and 3rd quarter days, respectively. A notably higher number of FIs
occurred on full moon days (Table 1, P < 0.010) and a significantly higher proportion of
patients were brought to ED in an ambulance in the rural areas (Table 1, p = 0.017).
In the rural areas of this study, the proportion of severe patients with KTAS scores of 1 to 2
was significantly lower, and the proportion of patients with KTAS scores from 3 to 5 was
higher (Table 1, Cramer’s V = 0.076). The proportion of mentally alert patients was higher
(V = 0.022) than in the metropolitan areas. The systolic and diastolic blood pressure and pulse
rate per minute of FI patients in the rural areas were significantly higher in rural areas than in
metropolitan areas.
Relationship of FI incidence with weather and moon phase
Pooled associations of the daily FPP with weather and moon phase are presented in Table 2,
Fig 2 (metropolitan area), and Table 3, Fig 3 (rural area). Among weather factors, precipita-
tion, minimum temperature, and wind speed showed a significant association with FI in met-
ropolitan areas (Table 2). FIs occurred frequently on days with lower precipitation, lower
minimum temperature, and low-wind days in metropolitan areas (Fig 2). In rural areas, FIs
have been shown to increase significantly on days with lower precipitation levels, higher mini-
mum temperatures, higher wind speed and longer sunshine duration (Table 3, Fig 3).
The distribution of FI patients was compared according to moon phase. The frequency of
FIs was higher on full moon days than on new moon days in rural areas (Table 1, p < 0.010).
Full moon was a significant predictor of FIs in univariate analysis in rural area (Table 3,
PLOS ONE | https://doi.org/10.1371/journal.pone.0261071 December 31, 2021
4 / 14
PLOS ONE Effects of weather and moon phases on EM use
Fig 1. Schematic diagram showing the selection of study population for this study.
https://doi.org/10.1371/journal.pone.0261071.g001
p = 0.048) but not significant in multivariate analysis. Based on the interaction analysis, the
new moon phase showed a significant interaction with precipitation and wind speed in rural
areas. However, there was no significant difference in the incidence of FI as similar FIs
occurred on all days in metropolitan areas (Tables 1 and 2).
Correlation of weather factors with FI severity
The severity of FI was measured using KTAS assessed upon ED arrival. KTAS 1 refers to a
state warranting emergency resuscitation, and KTAs 5 indicates absence of emergency. Injury
severity (daily mean KTAS for each province) was significantly correlated with minimum tem-
perature and wind speed and thus the injury severity was increased on cold windless days in
both areas (Table 4). Additionally, in rural areas, the daily mean KTAS was significantly corre-
lated with cloud cover. Precipitation, sunshine, and fog duration were not associated with the
PLOS ONE | https://doi.org/10.1371/journal.pone.0261071 December 31, 2021
5 / 14
PLOS ONE Table 1. Demographic features of subjects who visited ED after a fall injury.
Effects of weather and moon phases on EM use
Variable
Age
Sex
Month
Moon phase
Route
15–19
20–24
25–29
30–34
35–39
40–44
45–49
50–54
55–59
60–64
65–69
70–74
75–79
80–84
85–89
90–94
95–99
100–104
105–109
110–120
Male
Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
Oct
Nov
Dec
New moon
1st quarter
Full moon
Last quarter
other
Ambulance
Private car
Ambulation
Metropolitan
(n = 418,135)
N (%)
Rural
Cramer’s V or p value
(n = 248,777)
N (%)
25,423 (5.0)
28,389 (5.6)
29,192 (5.6)
29,566 (5.8)
27,869 (5.5)
30,840 (6.1)
36,172 (7.1)
43,583 (8.6)
48,404 (9.5)
38,986 (7.7)
35,154 (6.9)
37,268 (7.3)
39,169 (7.7)
31,347 (6.2)
17,844 (3.5)
6,660 (1.3)
1,393 (0.3)
183 (0.0)
24 (0.0)
8 (0.0)
7,299 (5.7)
7,864 (5.5)
7,795 (4.7)
7,988 (5.0)
7,780 (5.6)
8,706 (6.5)
10,245 (7.7)
12,272 (9.1)
13,521 (9.1)
10,742 (7.3)
9,593 (6.3)
10,214 (6.8)
10,960 (7.9)
8,844 (6.6)
5,034 (3.7)
1,893 (1.4)
409 (0.3)
59 (0.1)
13 (0.0)
3 (0.0)
258,579 (51.0)
132,101(53.1)
44,515(8.8)
38,357 (7.7)
39,581 (7.8)
40,136 (7.9)
44,224 (8.7)
38,029 (7.5)
39,968 (7.9)
41,564 (8.2)
43,425 (8.6)
46,051 (9.1)
42,183 (8.3)
49,410 (9.7)
44,461 (10.6)
41,531 (9.9)
43,107 (10.3)
44,059 (10.5)
244,713 (58.6)
439,439 (86.6)
65,574 (12.9)
2,227 (0.4)
12,288 (8.7)
10,381 (7.4)
10,448 (7.4)
10,939 (7.8)
12,420 (8.8)
10,928 (7.7)
11,315 (8.0)
12,131(8.6)
12,371 (8.8)
13,404 (9.5)
11,675 (8.3)
12,931 (9.2)
25,102 (10.1)
25,224 (10.1)
26,466 (10.6)
25,679 (10.3)
146,306 (58.8)
124,180 (87.9)
16,504 (11.7)
520 (0.3)
V = 0.001
P< 0.010
V = 0.020
p <0.010
p = 0.017
(Continued )
PLOS ONE | https://doi.org/10.1371/journal.pone.0261071 December 31, 2021
6 / 14
PLOS ONE Table 1. (Continued)
Variable
KTAS
Mental
SBP
DBP
PR
1
2
3
4
5
Alert
Verbal response
Pain response
Unresponsive
Effects of weather and moon phases on EM use
Metropolitan
(n = 418,135)
N (%)
Rural
Cramer’s V or p value
(n = 248,777)
N (%)
920 (0.80)
5334 (4.61)
34519 (29.83)
63769 (55.11)
11155 (9.64)
487775 (96.1)
11928 (2.4)
5038 (1.0)
2544 (0.5)
134.0 ± 26.6
79.8 ± 19.3
88.8 ± 40.4
223 (0.67)
1195 (3.59)
12620 (37.95)
15897(47.81)
3317(9.97)
137035 (97.0)
2358 (1.7)
1089 (0.8)
738 (0.5)
135.7 ± 27.1
80.7 ± 19.5
89.5 ± 19.5
V = 0.076
V = 0.022
p = 0.001
p <0.010
p <0.010
KTAS, Korean triage and acuity scale; SBP, systolic blood pressure; DBP, diastolic blood pressure; PR, pulse rate.
https://doi.org/10.1371/journal.pone.0261071.t001
severity of FI in the rural or metropolitan areas. There was no significant difference in mean
KTAS depending on the lunar phase in the metropolitan or rural areas (p = 0.394, p = 0.457,
respectively).
Discussion
During the study period of five years, we found that the prevalence and severity of FI were
associated with multiple weather factors such as daily precipitation, minimum temperature,
Table 2. Multivariate regression analysis of relationships between weather factors and moon phase with fall injuries in metropolitan areas.
Univariate analysis
Multivariate analysis
IRR
0.99
0.98
0.78
1.07
1.03
1.00
1.10
1.12
1.25
p value
0.043
<0.001
0.003
0.269
0.020
0.677
0.567
0.532
0.724
Precipitation
Minimum temperature
Wind speed
Cloud cover
Sunshine duration
Fog duration
Moon phase (versus full moon)
New moon
1st quarter
3rd quarter
Interaction effects
precipitation:minimum
precipitation:wind
precipitation:sunshine
minimum:wind
minimum:sunshine
wind:sunshine
IRR, incidence risk ratio; SE, standard error; CI, confidence interval.
https://doi.org/10.1371/journal.pone.0261071.t002
IRR
0.93
1.27
0.80
1.01
1.00
1.01
1.01
0.99
1.02
0.99
p value
0.045
0.005
<0.001
0.516
0.001
0.755
0.414
0.001
0.023
0.131
CI
0.93–0.97
0.97–0.98
0.95–1.01
1.01–1.02
1.00–1.01
1.01–1.01
1.01–1.01
0.99–1.00
1.01–1.03
0.99–0.99
PLOS ONE | https://doi.org/10.1371/journal.pone.0261071 December 31, 2021
7 / 14
PLOS ONE Effects of weather and moon phases on EM use
PLOS ONE | https://doi.org/10.1371/journal.pone.0261071 December 31, 2021
8 / 14
PLOS ONE Fig 2. Scatter plots of the number of ED visits per million people after fall injuries versus weather components in metropolitan areas of Korea. A theoretical
Gaussian regression line estimated by nonlinear curve fitting with the Boltzmann sigmoidal function is shown in red.
https://doi.org/10.1371/journal.pone.0261071.g002
Effects of weather and moon phases on EM use
and wind speed in both metropolitan and rural areas. However, we found that the longer the
sunshine duration was linked with the higher FI in rural area. Moon phases were weakly asso-
ciated with FI, especially in rural areas. The FI severity was closely related to weather factors.
This study enrolled the largest dataset ever collected to determine the association between
weather factors and ED visits due to FI in all 11 provinces of Korea over a 5-year period. A pre-
vious study conducted in a small city of 23,000 people in northern Netherlands reported that
better weather conditions were associated with the incidence of all types of trauma [9]. The
study location was similar to rural South Korea, where weather components including maxi-
mum temperature, sunshine duration, humidity, and precipitation were associated with all
kinds of injury. The present study also found that precipitation, minimum temperature, and
wind speed were typically related to FI. Additionally, sunshine duration was a significant pre-
dictor of FI in rural areas with high agricultural activities. These findings suggest that it is
essential to consider a variety of factors such as geographic location, main industry in the
region, and weather changes during the investigation of injury prevalence.
Ramgopal et al. [8] investigated the association of weather factors with all EMS dispatches
using longitudinal data of ambulance transport in western Pennsylvania and reported
increased EMS responses with rising temperature, snowfall, and rain based on a stratified anal-
ysis of seasonal variables and a day-of-the-week effect week. We found that additional factors
such as wind speed, cloud cover, and sunshine duration were associated with emergency
Table 3. Multivariate regression analysis of relationships between weather factors and moon phase with fall injuries in rural areas.
Univariate analysis
Multivariate analysis
IRR
p value
IRR
p value
0.96
1.25
1.24
2.01
1.19
0.03
0.75
0.61
0.68
<0.001
<0.001
<0.001
0.383
<0.001
0.817
0.002
0.948
0.984
Precipitation
Minimum temperature
Wind speed
Cloud cover
Sunshine duration
Fog duration
Moon phase (versus full moon)
new moon
1st quarter
3rd quarter
Interaction effects
precipitation:wind
precipitation:minimum
precipitation:sunshine
minimum:wind
minimum:sunshine
wind:sunshine
new moon:precipitation
new moon:wind
new moon:sunshine
IRR, incidence risk ratio; SE, standard error; CI, confidence interval.
https://doi.org/10.1371/journal.pone.0261071.t003
0.98
1.20
1.18
1.20
0.78
0.60
0.68
1.01
1.00
0.98
1.01
1.00
1.08
1.01
1.02
0.99
<0.001
<0.001
<0.001
0.040
0.043
0.556
0.984
<0.001
<0.001
<0.001
<0.001
0.590
<0.001
0.027
0.028
0.055
CI
0.90–0.97
0.99–1.51
1.23–1.25
0.67–2.20
0.61–0.85
0.95–1.02
0.99–1.03
1.01–1.01
1.00–1.00
0.96–0.99
1.01–1.02
1.00–1.00
1.06–1.11
0.00–14.6
0.79–1.16
0.02–3.13
PLOS ONE | https://doi.org/10.1371/journal.pone.0261071 December 31, 2021
9 / 14
PLOS ONE Effects of weather and moon phases on EM use
PLOS ONE | https://doi.org/10.1371/journal.pone.0261071 December 31, 2021
10 / 14
PLOS ONE Fig 3. Scatter plots of the number of ED visits per million people after fall injuries versus weather components in rural areas of Korea. A theoretical Gaussian
regression line estimated by nonlinear curve fitting with the Boltzmann sigmoidal function is shown in red.
https://doi.org/10.1371/journal.pone.0261071.g003
Effects of weather and moon phases on EM use
resource use after FI. However, seasonal changes were not included as independent variables
in this study because changes in minimum temperature and precipitation implicated seasonal
variations in weather. We also excluded FI on weekends and holidays because of increased
trauma due to enhanced outdoor leisure activity and distant travel on days that might act as a
confounding variable. We expected no major challenges in the analysis of FIs on weekdays
because of a sufficient number of cases using 5-year large-scale longitudinal data for at the
nationwide level.
Stomp et al. [9] reported that better weather conditions in rural areas were associated with
the incidence of all traumas. Our analysis also found that the frequency of FIs in rural areas
was increased under less precipitation, higher minimum temperature, and longer sunshine
duration such as busy farming seasons.
This is the first study to compare FI-related factors between developed metropolitan cities
and rural areas. During the course of our study, another research paper reported the correla-
tion between weather changes and FI in a small Russian city [4]. It was the only longitudinal
study for FIs like this study but was limited by geographic location of the study area or by
small number of subjects. The daily average of outdoor falls in the cold season was 20.2 per
100,000 people and the slippery surfaces covered with wet snow or ice and temperatures
between -7.0˚C and -0.7˚C were risk factors. As mentioned above, our study results showed a
distinct increase in FIs according to regional characteristics, with a lower temperature trigger-
ing falls on slippery surface in metropolitan areas, and a higher temperature during increased
agricultural activity in rural areas associated with increased FIs. They also reported that the FIs
were increased when the 12-hour precipitation was greater than 0.4 mm; however, the present
study showed that the FIs were increased under low precipitation. This difference is probably
explained by the falling of snow leading to slippery surfaces in Russia with a high altitude,
whereas in Korea located in mid-latitude weather, rain accompanied by summer storms with
strong winds reduced the frequency of outdoor activities. Northern Russia is located at the
highest latitude among countries in the world. As the highest temperature in summer was near
zero, the study failed to reflect changing weather patterns in mid-latitude areas with four clear
seasons.
Good weather conditions accompanied by active agricultural activities and increased night
visibility under moonlight on a full moon might be associated with FIs in rural areas. A moon
phase occurs every 29.53 days and 12.37 times in a year. The four principal moon phases
include: new moon, the 1st quarter, full moon, and the last quarter. Moon phases are known to
drive periodic changes in nighttime illumination, geomagnetic fields, gravitational pull, and
other factors associated with major meteorological and biological changes [15]. We found that
Table 4. Results of Pearson’s correlation analysis between injury severity (mean KTAS) with weather factors.
Precipitation
Minimum temperature
Mean wind speed
Cloud cover
Sunshine duration
Fog duration
Metro
CC
Rural
p value
CC
p value
0.006
0.701
-0.008
0.718
CC, correlation coefficient.
0.049
0.003
0.065
0.004
https://doi.org/10.1371/journal.pone.0261071.t004
0.157
<0.001
0.108
<0.001
-0.045
0.700
0.298
0.007
-0.012
0.957
0.005
0.809
<0.001
0.995
0.030
0.175
PLOS ONE | https://doi.org/10.1371/journal.pone.0261071 December 31, 2021
11 / 14
PLOS ONE Effects of weather and moon phases on EM use
FI-related ED visits on full moon days were significantly increased than on new moon days
only in rural areas. Two of the four rural provinces in this study border the sea and another
province is an island with active fishing activities (S1 Fig). The full moon is a time of full tide
and active fishing activities due to vertical migration of fishes [16]. Therefore, active nighttime
activities and fishing activities might increase FIs in rural areas. However, FIs in metropolitan
areas were less affected by lunar phase due to good visibility under night light that offset the
effects of lunar phases.
A previous study from Japan revealed a significant increase in the risk of emergency trans-
port after traffic accidents on full moon days among those aged �40 years [11]. This finding is
consistent with the results of our study showing a significant increase in FIs during full moon
days especially in rural areas where the elderly individuals reside under weak artificial lighting
at night. A population-based double control study conducted in the United States reported
that deaths from motor traffic accidents are more frequent on full moon nights [12]. The
authors postulated that a full moon might be associated with speeding, long distances, and
unknown routes, resulting in more frequent deaths. In our study, FIs were increased on full
moon days only in rural areas with weak artificial lighting, suggesting that increased visibility
and outdoor activity under moonlight on full moon days are associated with increased FIs.
The analysis of interaction between lunar phases and weather factors showed that the new
moon phase interacted with precipitation, wind speed, and cloud cover, which is consistent
with the findings of a previous study showing an increased number of storms during new
moon phases [17]. The finding suggests that the decrease in FIs in rural areas during new
phases may be a result of weather changes. Thus, the effect of the lunar phase is complex with
increased near-field vision due to moonlight mainly in rural areas and secondary weather
changes associated with lunar phases.
We found that weather factors were correlated with FI severity measured by KTAS, a uni-
fied triage scoring system. KTAS is a five-level triage scale developed in Korea based on Cana-
dian Triage and Acuity Scale (CTAS) and the score is a strong predictor of severity of patients
with higher 30-day mortality [18].
The strength of the study is that it is a population-based analysis of longitudinal data involv-
ing FIs in a mid-latitude country with four clearly distinguished seasons. A few unknown envi-
ronmental factors may confound the study results. Future studies should use more complex
modeling methods and evaluate the effects of moon phases and weather changes. Patients sus-
taining FIs may visit the ED the next day or later instead of on the day of injuries. Morency
et al. [19] reported a significant increase in outdoor falls on days 1–3 after falling temperatures
or snowfall. Therefore, it might be a challenge to compare changes in weather phenomena and
patients visiting the hospital on the same day. We believed that the interval between the
weather change and FIs is not a hindrance because of the gradual changes in weather and FI
incidence over a period of several days. We enrolled subjects regardless of indoor or outdoor
injuries because exposure to slippery terrain under snow or rain can still trigger injuries
indoors. Additionally, snowy and rainy days lead to behavioral changes due to thick clothes
and protective gears. Our study was conducted using large-scale nationwide databases without
analyzing clinical data of patients with emotional stress, alcohol use, and violence. Further, the
effects of other natural events such as earthquakes leading to mass casualties were not
considered.
In summary, we found that the incidence of FI is related to weather factors. Emergency
medical personnel should understand that FIs occur frequently during days of low precipita-
tion, high temperature and low winds linked with active outdoor activity in metropolitan
areas. Additional weather factors have been shown to affect FI incidence in rural areas so that
increased FI rates were noticed on days of low precipitation, high temperature, low winds and
PLOS ONE | https://doi.org/10.1371/journal.pone.0261071 December 31, 2021
12 / 14
PLOS ONE Effects of weather and moon phases on EM use
longer sunshine duration in rural areas. Moon phases are weakly linked to FI incidence rates.
FIs increased only in rural areas during the full moon days compared with new moon days. FI
severity is also affected by weather factors. In both urban and rural areas, the severity of FI sig-
nificantly increased on cold and windy days.
Supporting information
S1 Fig. Areas to be studied were selected by dividing them into A) metropolitan areas (red
color) including seven metropolitan cities with a population exceeding one million and B)
rural areas (blue color) consisting of four provinces without containing metropolitan cities
within its perimeter.
(PDF)
S2 Fig. Distribution of daily fall injuries by weekday for metropolitan and rural areas.
(PDF)
Author Contributions
Conceptualization: Kisung Kim, Sungyoup Hong.
Data curation: Min Ah Yuh, Kisung Kim, Juseok Oh, Jinwoo Kim, Sungyoup Hong.
Formal analysis: Kisung Kim, Seon Hee Woo, Jinwoo Kim, Sungyoup Hong.
Funding acquisition: Juseok Oh, Sungyoup Hong.
Methodology: Min Ah Yuh.
Project administration: Seon Hee Woo.
Software: Kisung Kim.
Supervision: Sungyoup Hong.
Validation: Kisung Kim, Juseok Oh.
Visualization: Min Ah Yuh, Sungyoup Hong.
Writing – original draft: Sikyoung Jeong, Sungyoup Hong.
Writing – review & editing: Seon Hee Woo, Sikyoung Jeong.
References
1. Organization WH, Ageing WHO, Unit LC. WHO global report on falls prevention in older age: World
Health Organization; 2008.
2. Smulders E, Enkelaar L, Weerdesteyn V, Geurts A, van Schrojenstein Lantman-de Valk H. Falls in
older persons with intellectual disabilities: fall rate, circumstances and consequences. J Intellect Disabil
Res. 2013; 57(12):1173–82. https://doi.org/10.1111/j.1365-2788.2012.01643.x PMID: 23106830
3.
Lepy E, Rantala S, Huusko A, Nieminen P, Hippi M, Rautio A. Role of Winter Weather Conditions and
Slipperiness on Tourists’ Accidents in Finland. Int J Environ Res Public Health. 2016; 13(8):822. https://
doi.org/10.3390/ijerph13080822 PMID: 27537899
4. Unguryanu TN, Grjibovski AM, Trovik TA, Ytterstad B, Kudryavtsev AV. Weather conditions and out-
door fall injuries in Northwestern Russia. Int J Environ Res Public Health. 2020; 17(17):6096. https://doi.
org/10.3390/ijerph17176096 PMID: 32825697
5. Sundfør HB, Sagberg F, Høye AJAA, Prevention. Inattention and distraction in fatal road crashes–
Results from in-depth crash investigations in Norway. Accid Anal Prev. 2019; 125:152–7. https://doi.
org/10.1016/j.aap.2019.02.004 PMID: 30763812
6. Yeung P-Y, Chau P-H, Woo J, Yim VW-T, Rainer TH. Higher incidence of falls in winter among older
people in Hong Kong. J of Clin Gerontol Geriatr. 2011; 2(1):13–6.
PLOS ONE | https://doi.org/10.1371/journal.pone.0261071 December 31, 2021
13 / 14
PLOS ONE Effects of weather and moon phases on EM use
7.
Lin L-W, Lin H-Y, Hsu C-Y, Rau H-H, Chen P-L. Effect of weather and time on trauma events deter-
mined using emergency medical service registry data. Injury. 2015; 46(9):1814–20. https://doi.org/10.
1016/j.injury.2015.02.026 PMID: 25818056
8. Ramgopal S, Dunnick J, Owusu-Ansah S, Siripong N, Salcido DD, Martin-Gill C. Weather and temporal
factors associated with use of emergency medical services. Prehosp Emerg Care. 2019; 23(6):802–10.
https://doi.org/10.1080/10903127.2019.1593563 PMID: 30874455
9. Stomp W, Fidler V, ten Duis HJ, Nijsten MW. Relation of the weather and the lunar cycle with the inci-
dence of trauma in the Groningen region over a 36-year period. J Trauma. 2009; 67(5):1103–8. https://
doi.org/10.1097/TA.0b013e3181986941 PMID: 19901675
10. Gevitz K, Madera R, Newbern C, Lojo J, Johnson CC. Risk of fall-related injury due to adverse weather
events, Philadelphia, Pennsylvania, 2006–2011. Public Health Rep. 2017; 132(1_suppl):53S–8S.
https://doi.org/10.1177/0033354917706968 PMID: 28692393
11. Onozuka D, Nishimura K, Hagihara AJSotte. Full moon and traffic accident-related emergency ambu-
lance transport: A nationwide case-crossover study. Sci Total Environ. 2018; 644:801–5. https://doi.
org/10.1016/j.scitotenv.2018.07.053 PMID: 29990928
12. Redelmeier DA, Shafir E. The full moon and motorcycle related mortality: population based double con-
trol study. Br Med J. 2017; 359:j5367. https://doi.org/10.1136/bmj.j5367 PMID: 29229755
13. Park J, Lim T. Korean triage and acuity scale (KTAS). J Korean Soc Emerg Med. 2017; 28(6):547–51.
14. McHugh ML. The chi-square test of independence. Biochem Med. 2013; 23(2):143–9. https://doi.org/
10.11613/bm.2013.018 PMID: 23894860
15. Chakraborty U. Effects of different phases of the lunar month on living organisms. Biol Rhythm Res.
2020; 51(2):254–82.
16. Das D, Pal S, Bhaumik U, Paria T, Mazumdar D, Pal SJIJoF, et al. The optimum fishing day is based on
moon. Int J Fish Aquat Stud. 2015; 2(4):304–9.
17. Pickering WHJPA. Relation of the Moon to the Weather. Pop Astron. 1903; 11:327–8.
18.
Lim YD, Lee DH, Lee BK, Cho YS, Choi G. Validity of the Korean Triage and Acuity Scale for predicting
30-day mortality due to severe trauma: a retrospective single-center study. Eur J Trauma Emerg Surg.
2018; 46(4):1–7. https://doi.org/10.1007/s00068-018-1048-y PMID: 30456416
19. Morency P, Voyer C, Burrows S, Goudreau S. Outdoor falls in an urban context: winter weather impacts
and geographical variations. Can J Public Health. 2012; 103(3):218–22. https://doi.org/10.1007/
BF03403816 PMID: 22905642
PLOS ONE | https://doi.org/10.1371/journal.pone.0261071 December 31, 2021
14 / 14
PLOS ONE |
10.1371_journal.pone.0255090 | REGISTERED REPORT PROTOCOL
Prevalence of complications associated with
the use of a peripherally inserted central
catheter in newborns: A systematic review
protocol
Edienne Rosaˆ ngela Sarmento DinizID
Augusto Rosendo da Silva1, Ricardo Ney CobucciID
1*, Kleyton Santos de Medeiros2, Richardson
3, Angelo Giuseppe Roncalli1
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
This is a Registered Report and may have
an associated publication; please check the
article page on the journal site for any
related articles.
OPEN ACCESS
Citation: Sarmento Diniz ER, de Medeiros KS,
Rosendo da Silva RA, Cobucci RN, Roncalli AG
(2021) Prevalence of complications associated
with the use of a peripherally inserted central
catheter in newborns: A systematic review
protocol. PLoS ONE 16(7): e0255090. https://doi.
org/10.1371/journal.pone.0255090
Editor: Raffaele Serra, University Magna Graecia of
Catanzaro, ITALY
Received: May 4, 2021
Accepted: July 6, 2021
Published: July 23, 2021
Copyright: © 2021 Sarmento Diniz et al. This is an
open access article distributed under the terms of
the Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: All relevant data are
within the paper and its Supporting Information
files.
Funding: Coordination for the Improvement of
Higher Education Personnel (Coor-denac¸ão de
Aperfeic¸oamento de Pessoal de Nı´vel Superior -
CAPES), Finance Code 001. The funders had no
1 Postgraduate Program in Public Health, Federal University of Rio Grande do Norte-UFRN, Natal, Brazil,
2 Health Sciences Postgraduate Program, Federal University of Rio Grande do Norte (UFRN), Natal, RN,
Brazil, 3 Postgraduate Program in Sciences Applied to Women’s Health, Maternidade Escola Janua´ rio Cicco
(MEJC / EBSERH), Federal University of Rio Grande do Norte, Natal, Brazil
* enesarmento@hotmail.com
Abstract
Background
The improper handling of a peripherally inserted central catheter (PICC) in newborns (NBs)
may result in mechanical and infectious complications.
Aim
The aim of this systematic review (SR) is to estimate the prevalence of complications asso-
ciated with the use of PICC in NBs.
Methods
We will utilize PubMed, Embase, CENTRAL, Web of Science, Scopus, Cochrane
Library, CINAHL, and Google Scholar for the databases search. There will be no
restrictions on the search for languages, and observational studies will be selected
wherein the prevalence rate of complications associated with the use of PICC in NBs
has been presented or can be calculated. The systematic review will follow the guide-
lines of the Preferred Reporting Items for Systematic Reviews and Meta-Analyses.
Two reviewers will independently select studies and assess their eligibility using prede-
fined criteria. Using standardized forms, two other reviewers will independently extract
data from each included study, and the random-effects pooled prevalence will be calcu-
lated in the meta-analysis with the respective 95% confidence intervals. The methodo-
logical quality of the studies will be assessed using the modified Newcastle-Ottawa
Scale. Review Manager V.5.3.5 will be used for the qualitative and quantitative synthe-
sis. A protocol was developed and published on PROSPERO (Registration number
CRD42020211983).
PLOS ONE | https://doi.org/10.1371/journal.pone.0255090 July 23, 2021
1 / 9
PLOS ONE Prevalence of PICC complications in newborns: A systematic review protocol
role in study design, data collection and analysis,
decision to publish, or preparation of the
manuscript.
Competing interests: The authors have declared
that no competing interests exist.
Expected results
This SR will show the prevalence of complications caused by the inadequate management
of PICC in NBs, which is information considered important for clinical practice improvement.
Introduction
The insertion of peripherally inserted central catheters (PICC) is common in neonatal inten-
sive care units (NICU) in term newborns (NBs) and premature infants receiving venous thera-
pies with vesicant and irritating drugs and parenteral nutrition. Moreover, depending on the
caliber, it is also used for blood tests, transfusion of blood products, and hemodynamic moni-
toring [1, 2]. These reduce the risk of complications and dispense with the need for surgical
intervention, which is advantageous due to the reduced hospitalization time, reduced hospital
costs, and decreased incidence of neonatal mortality, as PICC reduce the incidence of compli-
cations when compared to other central catheters [3].
Premature NBs generally have low birth weight, and many are dependent on hyperosmotic
or irritating drugs, which must be administered through a long-term venous access [2]. Previ-
ous studies have identified risk factors for complications associated with the use of PICC in
NBs, including gestational age (GA), catheter insertion location, catheter tip location, and
catheter dwell time, especially when it is installed for greater than 35 days, a factor also associ-
ated with an increased catheter-related bloodstream infection [1, 4]. Other mechanical compli-
cations appear with a lower incidence such as rupture of the catheter, migration of the catheter
tip, obstruction, and leakage of drugs [5, 6]. These are responsible for several non-elective
removals, and many of these can be avoided as they are related to the improper handling of the
device [3, 7].
The use of PICC in NICU has become essential in neonatal clinical practice, and monitor-
ing of the risk factors associated with complications is part of the neonatal care quality man-
agement in developed and developing countries, in which teams responsible for the quality
management adopt evidence-based strategies to prevent these complications [8, 9]. Such com-
plications can cause an increase in the costs of health services due to the increase in the length
of stay in the NICU, as well as injuries to NBs, such as tissue, cardiological, and infectious inju-
ries. Previous studies on complications related to the use of PICC in NBs have demonstrated
the prevalence of tissue complications (edema, phlebitis, enlargement, and necrosis) and sys-
temic complications (arrhythmias, cardiac tamponade, and sepsis) [3, 10–12].
Information on the complications associated with the use of PICC in NBs is scarce, and few
studies that were conducted were observations done for short periods and focused on the risk
factors associated with infection, but not on the prevalence of complications [8]. Thus, it is
essential to develop studies to assess the prevalence of complications related to the use of PICC
in NBs. The aim of this systematic review is to estimate the prevalence of complications associ-
ated with the use of PICC in NBs.
Research question
What is the prevalence of complications associated with the use of PICC in NBs?
PLOS ONE | https://doi.org/10.1371/journal.pone.0255090 July 23, 2021
2 / 9
PLOS ONE Prevalence of PICC complications in newborns: A systematic review protocol
Methods
This protocol was designed in accordance with the Preferred Reporting Items for Systematic
Reviews and Meta-Analysis guidelines extension for reporting systematic review protocols
(PRISMA-P) [13]. The protocol was registered with the International Prospective Register of
Systematic Reviews (PROSPERO) (number: CRD42020211983). The systematic review will
follow the guidelines of the Preferred Reporting Items for Systematic Reviews and Meta-Anal-
yses (PRISMA) [14, 15].
Ethics
Ethical approval is not required because this review will retrieve publicly available scientific lit-
erature. Traditional dissemination strategies will be used, including open-access peer-reviewed
publications, scientific presentations, and reports.
Inclusion criteria
This systematic review will include observational studies (case-control, cross-sectional, and
cohort) that describe complications associated with the use of PICC in NBs. There will be no
restrictions on the search for languages and the publication period.
Exclusion criteria
Published articles, but not peer-reviewed articles, will not be included in the review. Random-
ized clinical trials, review articles, reports, and case series will be excluded. Studies that assessed
the prevalence of complications in children (older than 28 days of life) will also be excluded.
The PECOT strategy
• Population/participants: NBs
• Exposure: PICC
• Comparator/control: NBs who do not use the PICC
• Outcome: Complications associated with the use of PICC (mechanical, infectious, and
systemic)
• Types of studies: observational studies (sectional, cohort, and case-control)
Types of participants
Study participants will be NBs using PICC or not, neonates (children under 28 days of age),
extremely premature (<28 weeks) NBs, very premature (28 to 31 weeks and 6 days) NBs, mod-
erate or late preterm (32 to 36 weeks and 6 days) NBs, NBs with low birth weight, and healthy
term NBs [16].
Types of exposures
The included studies will be those describing complications in neonates using PICC, a periph-
erally inserted central catheter that is inserted through a peripheral vein and its tip is destined
for the vena cava, used for medium-and long-term intravenous infusion or therapy with irri-
tating and vesicant drugs [17, 18].
Control: NBs who do not use PICC.
PLOS ONE | https://doi.org/10.1371/journal.pone.0255090 July 23, 2021
3 / 9
PLOS ONE Prevalence of PICC complications in newborns: A systematic review protocol
Outcomes
Primary outcome: Neonatal death
Secondary outcomes:
• Pulmonary complications: pleural effusion, pneumothorax, and hydrothorax
• Cardiovascular complications: arrhythmias, myocardial perforation, and cardiac
tamponade
• Tissue complications: hematomas, phlebitis, pain, local hardening, infiltration, leakage,
and necrosis
• Hematological complications: bleeding, embolism, and thrombosis
• Infectious complications
Types of studies
Observational studies: cross-sectional, cohort, and case-control.
Search strategy
The studies will be obtained through PubMed, Embase, CINAHL, LILACS, CENTRAL, Web
of Science, Scopus, Cochrane Library, and Google Scholar databases. There will be no restric-
tions on the search for languages and year of publication. Articles will also be searched from
the references of the selected studies, and the search strategy used in PubMed is shown in
Table 1.
Data collection and analysis
Study selection. After searching the databases, all identified articles will be exported to
Rayyan software and duplicates will be removed. First, the titles and abstracts will be read inde-
pendently by at least two reviewers (ERSD and KSM) based on the inclusion criteria. The full
texts of these potentially eligible studies will be retrieved and taken independently for eligibility
by two members of the review team (ERSD and RNC). Only studies identified by both pairs of
reviewers based on the inclusion criteria will ultimately be included in the systematic review,
and a third reviewer (RAS) will make a final decision for inclusion in case of discrepancy.
We will maintain a record of the reasons for excluding clinical trials at all stages of the
review. The results of the selection or exclusion of the studies will be reported using the
PRISMA flowchart, as shown in Fig 1.
Data extraction. Using standardized forms, two reviewers (ERSD and KSM) indepen-
dently will extract the following data from each included study: first author, year, place of
study, population and sample, type of study, objective, design, variables analyzed (GA of the
NB, NB weight, NB diagnosis, catheter insertion location, use of parenteral nutrition, use of
vasoactive drugs and antibiotics, time of use of PICC), and complications associated with the
use of PICC. The extracted data cover the issue of the review and will be verified again by three
authors (RNC, RAS, and AGR).
Addressing missing data.
In case of a lack of data, the authors of this article will contact
the respective authors or co-authors of the article in question by telephone or e-mail. If infor-
mation will not be received, the data will be excluded from our analysis and will be covered in
the discussion section.
PLOS ONE | https://doi.org/10.1371/journal.pone.0255090 July 23, 2021
4 / 9
PLOS ONE Prevalence of PICC complications in newborns: A systematic review protocol
Table 1. Search strategy for PubMed.
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
MESH Terms
Infant, Newborn
Neonate
Newborn Infants
Newborn disease
Preterm Infant
Extremely Premature Infant
OR / 1–6
peripherally inserted central venous catheter
PICC Placement
Peripherally Inserted Central Catheter Line Insertion
Venous Catheterizations, Peripheral
OR / 8–11
Complications
catheter infection
Catheter-Related Infections
Embolism
Thrombosis
Bleeding
Arrhythmia
Necrosis
OR / 13–20
Observational Study
Cohort Studies
Case-control Studies
OR / 22–24
7 AND 12 AND 21 AND 25
https://doi.org/10.1371/journal.pone.0255090.t001
Quality assessment of the included studies. The methodological quality of the selected
studies will be assessed using the modified Newcastle-Ottawa Scale [19]. The following items
will be analyzed: exposure, comparability, sample representativeness, sample size, response
rate, outcome assessment, and statistics. The classification of the methodological quality of the
studies will be carried out considering the total number of points received: � 4 for good quality
and <4 for low quality. The divergences found will be discussed and resolved by 3 authors of
the review (ERSD, RNC, and LNS) [20].
Assessment of heterogeneity
Measures of treatment effect. The results of the systematic review will be written in a
structured manner with respect to the characteristics of the target population, type of primary
outcomes (neonatal death), and secondary outcomes (systemic and tissue complications, rup-
ture of the catheter and others).
The general prevalence of complications in NBs with PICC will be calculated using the ran-
dom effects model, considering the heterogeneity between the studies included in the review.
Cochran’s Q test will be used to assess heterogeneity and the I2 statistic for quantification. The
result of I2 � 50% will be considered as low heterogeneity and, in this case, the fixed effects
model will be used. For I2> 50%, high heterogeneity, the random effects model will be used to
calculate the prevalence ratio and 95% confidence interval (95% CI). When possible, we will
use Egger’s funnel plot to assess possible publication bias.
PLOS ONE | https://doi.org/10.1371/journal.pone.0255090 July 23, 2021
5 / 9
PLOS ONE Prevalence of PICC complications in newborns: A systematic review protocol
Fig 1. PRISMA flow diagram for systematic review and meta-analysis.
https://doi.org/10.1371/journal.pone.0255090.g001
Data synthesis. A quantitative synthesis (meta-analysis) will be performed in the RevMan
5.3.5 software using the inverse variance method with the random effects model if there is
more than 50% heterogeneity between the studies. In cases where the data will be insufficient
to calculate an effect estimate, a narrative synthesis will be created, describing the direction
and size of the effects, along with any reported accuracy measures.
Subgroup and sensitivity analysis. Sensitivity analysis will be conducted to verify possible
sources of heterogeneity, removing one study at a time, and verifying if there is a considerable
change in the prevalence estimate and 95% CI. Sensitivity analysis will be performed excluding
studies with a high risk of bias. When the effect estimates of the primary and sensitivity analy-
ses are significantly different, an adjusted sensitivity analysis will be performed.
We plan to perform the following subgroup analyses, wherever possible: type of complica-
tion found (local versus systemic), GA, weight, diagnosis, days of life, and indications for the
use of PICC. If we identify significant differences between the subgroups (test for interaction
<0.05), we will report the results for individual subgroups separately. We will also perform a
formal test for subgroup interactions using the RevMan version 5.3.5.
Grading quality of evidence
Assessment of certainty of evidence. The analysis of the evidence for all outcomes will be
assessed using the Grading of Recommendations Assessment, Development and Evaluation
Working Group (GRADE) methodology [21] by classifying the evidence as high, moderate,
low, or very low.
Discussion
Knowledge about the actual prevalence of complications resulting from the use of PICC is
essential for clinical practice in NICU and may allow the adoption of strategies that reduce
those that are more serious, which can lead to the death of NBs. Previous studies have revealed
a strong association between the time of NBs exposure to PICC during administration of par-
enteral nutrition and the use of antimicrobials with the development of complications,
PLOS ONE | https://doi.org/10.1371/journal.pone.0255090 July 23, 2021
6 / 9
PLOS ONE Prevalence of PICC complications in newborns: A systematic review protocol
especially bloodstream infection [22–24]. However, systematic reviews of the prevalence of
these complications are scarce.
A previous meta-analysis revealed that PICC inserted in the lower limbs did not show
worse results compared to the upper limb group in the NICU, apart from thrombosis [25]. A
finding divergent from this was demonstrated by Pet et al., who reported the occurrence of
complications more frequently with PICC inserted in the upper extremities than in the lower
extremities [26]. In another study, it was observed that the insertion of the PICC in the first 48
hours of life did not increase the prevalence of complications [12].
A recent study conducted in Greece revealed that the complication rates were similar when
comparing the types of catheters used in NBs with low birth weight, and the authors recom-
mend that central venous catheters should be removed early in NICU [27]. Studies have
shown that low birth weight is a risk factor for complications associated with PICC in the
NICU, and Wen et al. demonstrated that infectious and non-infectious complications of PICC
are associated with low weight gain in premature babies [27, 28]. However, there is no reliable
data on the actual prevalence of complications of PICC use in premature or low birth weight
infants.
Biofilm is a fundamental component in the pathogenesis of infectious complications of
PICC. Biofilm can be the cause of PICC extraction and can lead to serious haematogenic infec-
tious complications that can increase the morbidity and mortality of affected babies. In order
to help physicians and nurses to better target their preventive and therapeutic measures, will
be important to understand which organism has the greatest impact on the development of
PICC related bloodstream infections and to study the prevalence of the conditions worldwide,
since effective prevention represent a sensitive target to reduce the prevalence of infectious
complications during the use of PICC [29].
The potential limitations of the systematic review will focus on several aspects of the study
design, searches, and quality appraisal of included studies. Limitations are related to include
cross-sectional and case-control studies to hinder the reliable assessment of the causal relation-
ship between PICC and complications in NBs. Furthermore, a small sample size and a limited
number of studies can influence the validity and reliability of the findings.
Therefore, this systematic review will be carried out using a specific approach with a meta-
analysis of the included studies results if possible. It is justified because knowing the prevalence
of complications associated with the use of PICC in NBs in the NICU can positively impact the
practice of care for the NB during infusion therapy and allow the creation of strategies to
reduce serious complications such as sepsis and death. We expect that it will show the preva-
lence of complications caused by the inadequate management of PICC in NBs, which is infor-
mation considered important for clinical practice improvement.
Supporting information
S1 Checklist. Preferred Reporting Items for Systematic review and Meta-Analysis Proto-
cols (PRISMA-P checklist).
(DOCX)
Author Contributions
Conceptualization: Edienne Rosaˆngela Sarmento Diniz, Kleyton Santos de Medeiros, Ricardo
Ney Cobucci.
Data curation: Richardson Augusto Rosendo da Silva.
PLOS ONE | https://doi.org/10.1371/journal.pone.0255090 July 23, 2021
7 / 9
PLOS ONE Prevalence of PICC complications in newborns: A systematic review protocol
Formal analysis: Edienne Rosaˆngela Sarmento Diniz, Kleyton Santos de Medeiros, Ricardo
Ney Cobucci.
Methodology: Edienne Rosaˆngela Sarmento Diniz, Kleyton Santos de Medeiros, Ricardo Ney
Cobucci.
Supervision: Richardson Augusto Rosendo da Silva, Ricardo Ney Cobucci, Angelo Giuseppe
Roncalli.
Validation: Edienne Rosaˆngela Sarmento Diniz, Richardson Augusto Rosendo da Silva,
Ricardo Ney Cobucci, Angelo Giuseppe Roncalli.
Writing – original draft: Edienne Rosaˆngela Sarmento Diniz, Kleyton Santos de Medeiros.
Writing – review & editing: Ricardo Ney Cobucci, Angelo Giuseppe Roncalli.
References
1. Rosado V, Camargos PA, Anchieta LM, Bouzada MC, Oliveira GM, Clemente WT, et al. Risk factors for
central venous catheter-related infections in a neonatal population: systematic review. J Pediatr. 2018;
94:3–14. https://doi.org/10.1016/j.jped.2017.03.012 PMID: 28866323
2.
Li Renfeng, Xia C, Tian S, Lei X. Application of peripherally inserted central catheters in critically ill new-
borns experience from a neonatal intensive care unit. Medicine [Internet]. 2019; 98:32. https://doi.org/
10.1097/MD.0000000000015837 PMID: 31393341
3. Xiaohe Y, Shaojie Y, Mingjie W, Chuanding C, Zhengchang L, Ying D, et al. Risk Factors Related to
Peripherally Inserted Central Venous Catheter Nonselective Removal in Neonates. BioMed Research
International 2018; Screens:1–6. https://doi.org/10.1155/2018/3769376 PMID: 30003096
4. Milstone AM, Reich NG, Advani S, et al. Catheter dwell time and CLABSIs in neonates with PICCs: a
multicenter cohort study. Pediatrics 2013; 132: e1609–15. https://doi.org/10.1542/peds.2013-1645
PMID: 24218474
5. Vashi PG, Virginkar N, Popiel B, Edwin P, Gupta D. Incidence of and factors associated with catheter-
related bloodstream infection in patients with advanced solid tumors on home parenteral nutrition man-
aged using a standardized catheter care protocol. BMC Infectious Diseases. 2017; 17: 372. https://doi.
org/10.1186/s12879-017-2469-7 PMID: 28558699
6. Sukyung K, Soo mi S, Seoul HL, Kim JH, Kim H, Kim JY, et al. Outcomes of bedside peripherally
inserted central catheter placement: a retrospective study at a single institution. Acute and Critical Care.
2020; 35(1): 31–37. https://doi.org/10.4266/acc.2019.00731 PMID: 32131579
7. Santana FG, Dias PLM. Central Catheter of Peripheral Insertion in Pediatric Oncology: a Retrospective
Study. Brazilian Journal of Cancerology. 2018; 64 (3): 341–347. https://doi.org/10.32635/2176-9745.
RBC.2018v64n3.34
8. Hermansen MC, Hermansen MG. Intravascular catheter complications in the neonatal intensive care
unit. Clin Perinatol. 2005; 32: 141–56. https://doi.org/10.1016/j.clp.2004.11.005 PMID: 15777826
9.
Franceschi AT, Cunha ML. Adverse events related to the use of central venous catheters in hospitalized
newborns. Rev Latino-Am Enfermagem. 2010; 18. Screens 57–63. https://doi.org/10.1590/s0104-
11692010000200009 PMID: 20549118
10. Panagiotounakou P, Antonogeorgos G, Gounari E, Papadakis S, Labadaridis J, Gounaris AK. Peripher-
ally inserted central venous catheters: frequency of complications in premature newborn depends on
the insertion site. Journal of Perinatology. 2014; 34: 461–463. https://doi.org/10.1038/jp.2014.36
PMID: 24625519
11. Singh A, Bajpai M, Panda SS, Jana M. Complications of peripherally inserted central venous catheters
in neonates: Lesson leamed over 2 years in tertiary care center in India.Afr J Paediatr Surg. 2014 Jul-
Sep; 11 (3): 242–7. https://doi.org/10.4103/0189-6725.137334 PMID: 25047317
12.
Li R, Cao X, Shi T, Xiong L. Application of peripherally inserted central catheters in critically ill newborns
experience from a neonatal intensive care unit. Medicine (Baltimore). 2019 Aug; 98 (32): e15837.
https://doi.org/10.1097/MD.0000000000015837 PMID: 31393341
13. Shamseer L, Moher D, Clarke M, Ghersi D, Liberati A, Petticrew M, et al., PRISMA-P Group. Preferred
reporting items for systematic review and meta-analysis protocols (PRISMA-P) 2015: elaboration and
explanation. BMJ. 2015; 349: g7647. https://doi.org/10.1136/bmj.i4086 PMID: 25555855
PLOS ONE | https://doi.org/10.1371/journal.pone.0255090 July 23, 2021
8 / 9
PLOS ONE Prevalence of PICC complications in newborns: A systematic review protocol
14.
Liberati A, Altman DG, Tetzlaff J, et al. The PRISMA statement for reporting systematic reviews and
meta-analyzes of studies that evaluate healthcare interventions: explanation and elaboration. BMJ
2009; 339: b2700. https://doi.org/10.1136/bmj.b2700 PMID: 19622552
15. Moher D, Liberati A, Tetzlaff J, et al. Preferred reporting items for systematic reviews and meta-ana-
lyzes: the PRISMA statement. PLoS Med 2009; 6: e1000097. https://doi.org/10.1371/journal.pmed.
1000097 PMID: 19621072
16. Quinn JA, Munoz FM, Gonik B, et al. Preterm birth: Case definition & guidelines for data collection, anal-
ysis, and presentation of immunization safety data. Vaccine. 2016; 34 (49): 6047–6056. https://doi.org/
10.1016/j.vaccine.2016.03.045 PMID: 27743648
17. Westergaard B, Classen V, Walther-Larsen S. Peripherally inserted central catheters in infants and chil-
dren—indications, techniques, complications and clinical recommendations. Acta Anaesthesiol Scand
2013; 57: 278–87. https://doi.org/10.1111/aas.12024 PMID: 23252685
18. Prince A, Groh-Wargo S. Nutrition management for the promotion of growth in very low birth weight pre-
mature infants. Nutr Clin Pract 2013; 28: 659–68. https://doi.org/10.1177/0884533613506752 PMID:
24163320
19. Wells GA, Shea B, O’Connell D, J Peterson, V Welch, M Losos, P Tugwell. The Newcastle-Ottawa
Scale (NOS) for Assessing the Quality of Nonrandomised Studies in Meta-analyses. [Internet] Ottawa:
Ottawa Hospital Research Institute. Available from: http://www.ohri.ca/programs/clinical_epidemiology/
oxford.asp
20. Stang A, Andreas S. Critical evaluation of the Newcastle-Ottawa scale for the assessment of the quality
of nonrandomized studies in meta-analyzes. Eur J Epidemiol 2010; 25: 603–5. https://doi.org/10.1007/
s10654-010-9491z PMID: 20652370
21. Guyatt G, Oxman AD, Akl EA, Kunz R, Vist G, Brozek J, et al. GRADE guidelines: 1. Introduction-
GRADE evidence profiles and summary of findings tables. J Clin Epidemiol. 2011 Apr; 64(4):383–94.
https://doi.org/10.1016/j.jclinepi.2010.04.026 PMID: 21195583
22. Greenberg RG, Cochran KM, Smith PB, Edson BS, Schulman J, Lee HC, et al. Effect of catheter resi-
dence time on the risk of catheter-associated bloodstream infection in infants. Pediatrics 2015; 136 (6):
1080–6. https://doi.org/10.1542/peds.2015-0573 PMID: 26574587
23. Gordon A, Greenhalgh M, McGuire W. Early planned removal versus expectant management of periph-
erally inserted central catheters to prevent infection in newborn infants. Cochrane Database Syst Rev.
2018 Jun 25; 6 (6): CD012141. https://doi.org/10.1002/14651858 PMID: 29940073
24. Hoang V, Sills J, Chandler M, Busalani E, Clifton-Koeppel R, Modanlou H D. Percutaneously Inserted
Central Catheter for Total Parenteral Nutrition in Neonates: Complications Rates Related to Upper Ver-
sus Lower Extremity Insertion. Pediatrics. 2008; 121(5):e1152–e1159; https://doi.org/10.1542/peds.
2007-1962 PMID: 18390957
25. Chen H, Zhang X, Wang H, Hu X. Complications of upper extremity versus lower extremity placed
peripherally inserted central catheters in neonatal intensive care units: A meta-analysis. Intensive Crit
Care Nurs. 2020 Feb; 56: 102753. https://doi.org/10.1016/j.iccn.2019.08.003 PMID: 31445794
26. Pet GC, Eickhoff JC, McNevin KE, Do J, McAdams RM. Risk factors for peripherally inserted central
catheter complications in neonates. J Perinatol. 2020; 40: 581–588. https://doi.org/10.1038/s41372-
019-0575-7 PMID: 31911643
27. Wen J, Yu Q, Chen H, Chen N, Huang S, Cai W. Peripherally inserted central venous catheter-associ-
ated complications exert negative effects on body weight gain in neonatal intensive care units. Asia Pac
J Clin Nutr. 2017; 26 (1): 1–5. https://doi.org/10.6133/apjcn.112015.07 PMID: 28049254
28.
29.
Foo R, Fujii A, Harris JA. et al. Complications in Tunneled CVL Versus PICC Lines in Very Low Birth
Weight Infants. J Perinatol. 2001; 21: 525–530. https://doi.org/10.1038/sj.jp.7210562 PMID: 11774013
Ielapi N, Nicoletti E, Lorè C, Guasticchi G, Avenoso T, Barbetta A, et al. The Role of Biofilm in Central
Venous Catheter Related Bloodstream Infections: Evidence-based Nursing and Review of the Litera-
ture. Rev Recent Clin Trials. 2020; 15(1):22–27. https://doi.org/10.2174/
1574887114666191018144739 PMID: 31656155.
PLOS ONE | https://doi.org/10.1371/journal.pone.0255090 July 23, 2021
9 / 9
PLOS ONE |
10.1371_journal.pone.0252732 | RESEARCH ARTICLE
Accident and pollution risk assessment for
hazardous cargo in a port environment
Rafi Ullah KhanID, Jingbo Yin*, Faluk Shair Mustafa
Department of International Shipping, School of Naval Architecture, Ocean and Civil Engineering, Shanghai
Jiao Tong University, Shanghai, China
* jingboyin@sjtu.edu.cn
Abstract
The catastrophic environmental, life and monetary losses concomitant to the hazardous
cargo accidents have remained a matter of critical concern for the maritime transportation
officials. The factors that instigate these accidents while dealing with hazardous cargo in a
port environment requires rigorous analysis and evaluation, which still remains in its infancy.
In accord to these prevailing issues, this study focusses on the assessment of multifactor
risks associated with the dealing of hazardous cargos inside a port. The methodology
adopted is the amalgamation of expert judgment and literature review for the identification of
factors and developing their causal relationship, while Bayesian Network (BN) for the infer-
ence, which was based on 348 past accident reports from the year 1990 to 2018. The results
indicate that under normal circumstances, the probability of an accident with considerable
consequences is 59.8, where human and management were found to be the highest contrib-
uting factors. Setting evidence at the environment and pollution accident to occur, the inci-
dence probability of the “management” is raised by 7.06%. A sensitivity analysis was
conducted to determine the most critical factors for the hazardous cargo accident. This
study reveals that in order to evade the hazardous cargo accidents and curtail severity of
the consequences, the port authorities, concerned government departments and other
related institutions should pay specific attention to the qualification, training and attitude of
the involved workforce. Moreover, the development and implementation of stringent safety
protocols was also revealed to have critical prominence. This study holds practical vitality
for enhancing safety and mitigating risks associated to hazardous cargo dealing in a port.
1. Introduction
The port industry all over the world has greatly emphasized the establishment of green ports as
a leading trend towards the achievement of pollution free environment at ports. This aim has
attracted greater consideration and is regarded a common goal to achieve by all countries with
maritime transportation [1]. Goods with inflammable, explosive, and toxic properties carry a
danger along from one port to another in their transportation [2]. The port industry around
the globe thus seriously emphasizes the safer transportation of these hazardous goods. The
excessive adaptation of maritime transportation has led to the development of creating greater
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: Khan RU, Yin J, Mustafa FS (2021)
Accident and pollution risk assessment for
hazardous cargo in a port environment. PLoS ONE
16(6): e0252732. https://doi.org/10.1371/journal.
pone.0252732
Editor: Ahmed Mancy Mosa, Al Mansour
University College-Baghdad-Iraq, IRAQ
Received: February 23, 2021
Accepted: May 21, 2021
Published: June 4, 2021
Copyright: © 2021 Khan et al. This is an open
access article distributed under the terms of the
Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: All relevant data are
within the manuscript and its Supporting
Information files.
Funding: The authors received no specific funding
for this work.
Competing interests: The authors have declared
that no competing interests exist.
PLOS ONE | https://doi.org/10.1371/journal.pone.0252732 June 4, 2021
1 / 20
PLOS ONE Hazardous cargo accident and pollution risk
storage capacities; leading to the increase in throughput of hazardous goods at ports which in
turn requires greater consideration towards the safer production and transportation of the
concerned hazardous goods [3]. The ports with dangerous goods may face unforeseen acci-
dents leading to greater economic loss and may consequence human casualties.
This development in maritime transportation is not free of the risk associated with the
transportation of hazardous goods. This risk on many occasions resulted into dangerous acci-
dents and the most notable accident occurred on August 12, 2015 where fire broke out at the
warehouse that was containing dangerous goods at Ruihai International Logistics Co. Ltd. at
Tianjin Port, Tianjin Binhai New Area in Tianjin city. The fire led to the explosions and caused
serious casualties and economic loss [4]. Another prominent accident in this regard was the
Beirut port explosion, where ammonium nitrate was stored and left their for years which even-
tually resulted in an explosion and brought about serious life, property and environmental
losses [5]. Likewise, a serious oil spill of more than 1000 tons from a Japanese ship at Mauritius
even after spending millions of dollars on the cleaning processes did serious damage to the
local marine environment and conservatories [6]. The significant aspect therefore is to identify
key areas which may prevent accidents happenings at ports during the transportation of haz-
ardous material.
About the standard management of ports operations, a study classified certain functional
data and provided suggestions that focused the safe operation of hazardous chemical industry
at both national and international levels [7]. The safe management of chemical materials was
discussed in detail in an information guide provided in this study. This study in detail focused
on the areas of necessary measures being adopted in transportation of chemical stuff through
various transportation sources like rail, road, air and sea. Various studies have been conducted
on maritime and port logistics of chemicals and hazardous cargo which evaluate the risk posed
by dangerous goods from various perspectives [8–10]. These studies indicate that human fac-
tor is one of the leading accident causation factor. Likewise proper documentation, warehouse
management, equipment and technology, natural factors and container handling and packing
are acknowledged as critical factors for hazardous cargo safety.
A study suggested a well-established department which should maturely consider the pro-
cedural application of an efficient method for the assessment of fires’ risk at ports [11]. Various
studies in past have discussed the standard management to minimize the risk of accidents at
ports; but the recent growth in sea port operations and maritime transportation indeed
requires research to be conducted on every level of operations to minimize the risk of accidents
in port operations and transportation. The available studies provide a solid research founda-
tion for researchers to focus the operational classification and facilitate the safety insurance
during the transportation of hazardous goods. The accidents at ports symbolize the gover-
nance of ports at national level and indicate serious threats of loss during hazardous goods
transportation; if not addressed timely [3, 12]. The need of the hour indeed is, to devise an
innovative procedural mechanism which should focus to limit the risks of hazardous transpor-
tation as its primary concern.
The hazardous cargo risk could be evaluated and assessed through different tools. Some
research studies adopted analytical hierarchy process (AHP) theory for the overall analysis [13,
14]. Bayesian networks are acknowledged as a robust tool in the risk assessment domain, and
its use in the maritime transportation sector is considered highly reliable, efficient and power-
ful tool, but considered still in its infancy [15]. Though there are issues associated with BNs as
they require a lot of data to set up the prior probabilities, however, they widely employed in
the maritime transportation risk assessment domain [16]. The former or prior probability is
the central point of dependence in the Bayesian model as it focuses the data obtained from
past accidents, related literature and expert judgment for future prediction and analysis. BNs
PLOS ONE | https://doi.org/10.1371/journal.pone.0252732 June 4, 2021
2 / 20
PLOS ONE Hazardous cargo accident and pollution risk
that are based on actual data from past accidents are highly favored for their practicality and
accurate results. BNs have been extensively employed in the maritime and hazardous cargo
accident risk assessment studies [17–20].
The risk associated with the handling of dangerous goods in a port environment are related
to various aspects of human, the port environment, infrastructure and facilities itself and issues
with the port authorities and governing bodies. The analysis of these factors holds critical
prominence to advocate the right resources towards right issues at a right time. A reliable
quantification of the concomitant risk in various scenarios provides a significant support in
the decision making associated with safety management and risk mitigation. As discussed ear-
lier, recently there have been many catastrophic hazardous cargo accidents at ports and shores
and have been playing havoc to the human life, property and environment. However, there is a
serious lack of studies that focus on the hazardous cargo accident risk in the port environment
that not only focus on the accident risk, but also the risk posed to the port and surroundings
environment.
Therefore, in accord to this serious gap in the literature, this study is focused on the highly
desired multifactor risk assessment of hazardous cargos in a port environment. Employing the
widespread acceptability of data based BNs, this study aims to develop a BN model. The devel-
oped methodology and model provides useful insights into the effective management of haz-
ardous cargo in a port environment in any region and geography in concurrence to the locally
prevailing conditions. This study employing the past accidents data, determines the most
prominent and decisive factors in developing an efficient, reliable and robust approach for the
hazardous cargo handling and developing sustainable safety plans for port authorities, neigh-
boring environments, supply chains and governments. The results of this study holds practical
vitality for the process safety professional, decision making bodies, port authorities and gov-
ernments in enhancing their efficacy in waning the accidents frequency and circumventing
their cataclysmic consequences.
The rest of the paper is arranged as introduction being followed by literature review, section
3 provides the adopted methodology, while, section provides the detailed hazardous cargo
environment and pollution accident risk assessment, section 5 provides the conclusion of this
study and section 6 provides the references incorporated in this study.
2. Literature review
The continuous growth in the traffic of freight cargo has emphasized the need of long term
sustainability of such growth; playing the role of a key element in devising global policy debate
on trade enhancement and environmental protection [21]. Maritime transportation involves
industrial and port-vessel activities. The industrial activities at port are comprised of all those
activities which concern oil terminals and both chemical and petrochemical plants. While on
the other hand, the port-vessel activities are activities which concern the loading and unload-
ing consignment, oil jetties and its searching [22]. Besides, concerning their situation in coastal
areas, ports are undoubtedly the most intricate structural systems from environmental protec-
tion perspective because a range of goods related activities are performed there [23]. This
aspect as a central point of discussion is discussed and recognized accidental spills as the chief
reason behind water pollution [24]. Port accidents are described as the most frequent accidents
which accumulate 51% of the release that pollutes water. Furthermore, the record port acci-
dents ratio of (59%) is caused and associated with oil spills [23]. Consequently, the rapid
spread of pollution due to oil spills is one of the most serious problems faced at ports during
consignment loading, bulk liquid, fuel supply and navigation.
PLOS ONE | https://doi.org/10.1371/journal.pone.0252732 June 4, 2021
3 / 20
PLOS ONE Hazardous cargo accident and pollution risk
Thus, site-specific strategies at regional level are to be defined, so that the port activities
may well be facilitated on emergency level. Particular considerations should be given during
this procedural implementation to all related oceanographic, meteorological and environmen-
tal circumstances to restrict the spread of pollution, including that which is caused by the spills
[25]. Consequently, the analysis on oil spills should consider the domains of (i) an exact spatial
framework for definite danger evaluation (ii) apposite methodical data which may result into
authentic approach towards risk assessment [25–27].
Environmental risk analysis (ERA) gives the estimated calculation as to what level the spills,
fires or explosions may cause pollution under certain specific circumstances and geographical
location [28]. Furthermore, ERA serves as an essential requirement for the overall process of
spatial planning which incorporate the improvement of ‘inventories and maps’ to sustain
emergency preparation along the development of a green supervision design; leading to the
execution of effective risk management [29, 30]. Presently, ERA for its functionality is not
facilitated by any worldly adapted and acknowledged standard method. There is no globally
accepted standard method to be applied in ERA. Nevertheless, greater considerations should
be given to emphasize the role of ERA towards its approach in handling oil spills with specific
illustration to all those tools which define hazardous situations and support the analysis of spa-
tial risk management [31].
Concerning the recognition of hazards, ERA with regard to spills, fires, explosions and leak-
ages their associated scenarios should elaborate its controlling mechanism. The systematic
procedure of ERA addresses the identification of danger at first level to classify as how the
stressors may get exposed [32]. The identification of environmental dangers has been down
through the adaptation of various innovative tools but none exactly defines the nature of acci-
dents to classify its types in the ERA [23]. Data collected through the recorded accidental spills,
fires and explosions are utilized for this procedural functionality. The databases in this regard,
for instance the FACTS database, are refined with the broader scope to establish international
conventions (OPRC) (UN, 1995) however, this primarily reports outsized spills. Though, 95%
of the overall spill amount is comprised of the small and medium spill ranges which normally
occur at ports and oil terminals during the loading and unloading of consignments with the
estimated ratio of 40% to 29% respectively [33, 34]. The overall databases for these small and
medium sized accidents are usually maintained as records by the concerned local port and
maritime authorities. The further analysis based on the database highlights the actual location
and facility which may experience accidents with a detail of the specific causes and conse-
quences [23]. There is this factor of inaccurate database which may not efficiently characterize
the true nature of hazards as the scope of information is usually very limited in them; giving
consideration to discharge and its appearance. The appearance of the discharge and its source
is helpful in identifying pollution.
The proper definition of ERA for the methodical control over oil spills and hazardous cargo
accidents is done through the proper meteorological and oceanographic representation of
ERA scenarios. In ERA with port-specific location, the variability of (met ocean) tools is not
comprised as representative risk component [35]. Various authors in their studies have
focused on the met-ocean conditions to examine the exact estimate of affected offshore and
coastal area [36]. The basic functionality of these methodological approaches is well defined
through its bases in forecasting systems because the nature of their adaptation is purely opera-
tional [37]. Nonetheless, the emergency based planning procedures require concentrating on
the aspect of prevention. The proper and well defined statistical results in contingency plan-
ning are very significant to comprehend the hazardous cargo accident scenarios.
The assessment of risk can never be ignored as the most significant concern in hazardous
industrial zone, specifically in maritime transportation. The key concern is to identify all those
PLOS ONE | https://doi.org/10.1371/journal.pone.0252732 June 4, 2021
4 / 20
PLOS ONE Hazardous cargo accident and pollution risk
indicators which adversely affect the safety and thus to minimize every possible risk [38]. The
happening probability of an accident and its related adverse consequences are defined as risk.
The risk and its consequences in maritime transportation usually adversely affect the economic
stability in terms of monetary losses, human harms in terms of casualties and environmental
pollution. A Formal Safety Assessment (FSA) was introduced by the International Maritime
Organization to manage the functionality of Maritime Transportation System (MTS). The
MTS functionality scope is affected by the lack of available historical data. The efficient assess-
ment of risk is made possible by the prior knowledge from experts who have worked in this
specific domain, and this knowledge is a significant source of information to base risk assess-
ment on it [39]. One the other hand, the challenge for all models based on prior knowledge is
that the data available is reactive instead of proactive facilitation. A research study highlights
that the reactive approaches are usually passive and their limited systematic approach does not
consider the changes, variables or shortcomings as sources affecting the overall forecast [40,
41]. Certain frameworks in that regard are considered available sources to calculate probabili-
ties and consequences caused by accidents.
The complex system of MTS along with other human, environmental and organizational
factors contribute towards the uncertain functionality scope of MTS [42]. Right from its intro-
duction and adaptation, the Quantitative Risk Assessments (QRA) has been attributed great
appreciation as compared to qualitative assessments in maritime transportation. The basic
approach followed in (QRA) blend expert knowledge and data [43, 44]. A number of QRA
models have adopted tools like fault or event-tree analysis (FTA, ETA), evidential reasoning
(ER), Bayesian Belief Networks BBNs [16, 45–49]. Researchers in maritime risk assessment
have worked on literature review from different perspectives and thus provided wider scope
studies to fill the knowledge gape.
Efforts made possible the availability of an outline on QRA models by combining critical
research ideas from 87 related academic papers [50]. The major objective in their study was to
study risk in collision and grounding through the adaptation of ETA and traffic flow theory. A
study discussed the foundation matters; risk analysis in maritime transportation was further
analyzed through defining and scientific outlooks [51]. Amid all adapted approaches in mari-
time risk assessment the BBN approach efficiently classifies the dependencies related to a cer-
tain accident through conditional probability tables (CPTs) [52]. In addition to that, BBN has
been celebrated for certain other advantages like those of inverse inference ability and network
up gradation [16]. Researchers have in detail studied the comparative sketch of advantages and
challenges associated with BBN in maritime transportation [53]. The BBN was authenticated
by the results of various studies as the most suitable methodology for risk assessment in mari-
time transportation and its related impending decisions. The BBN was further valued for its
positive characteristic of auto up-gradation with available new data. The above mentioned rea-
sons favored BBN adaptation in maritime risk assessment and excessively valued its approach
towards maritime safety.
However, the use of the BBN in the port hazardous cargo accident risk is new and still in its
infancy. Therefore, it is of phenomenal prominence to evaluate the various causation factors of
a hazardous cargo accident in a port environment and provide productive results to aid deci-
sion making, devising policies and developing safety management and risk mitigation systems.
3. Methodology
Bayesian networks are an amalgamation of the probability and graph theory, and hence recog-
nized as an effective tool for the analysis and assessment of uncertainties and vagueness associ-
ated with the causation factors incorporated to the BN model as nodes and states. The data is
PLOS ONE | https://doi.org/10.1371/journal.pone.0252732 June 4, 2021
5 / 20
PLOS ONE Hazardous cargo accident and pollution risk
processed through several tools like artificial intelligence, decision evaluation and the probabil-
ity and graph theory [49]. It is regarded as one of the main features of BN that it can success-
fully process the uncertainty associated with all the data incorporated into the developed
model. Subject to its profound features, BN proves to be an effective tool in the risk assessment
domain, specifically in the maritime transport sector [48]. However, a point of protuberant
concern here is that the matter and issue intended to be studied should thoroughly be checked
for its consistency with BNs.
The process of development of a BN model encompasses the recognition of the relevant
and influencing factors, developing a causal relationship between these factors and demon-
strating it through a proper Directed Acyclic Graph (DAG). Another prominent stage in this
regard is the data incorporation and quantification. In this study, the identification and selec-
tion of the variables and the development of causal relationship among the nodes was done in
concurrence to expert judgment and available related literature. While the data for inference
and quantification was derived from past accidents occurred during 1990 and 2018. The calcu-
lation of probabilities and conditional probability tables (CPTs) was done through parameter
estimation. Parameter estimation is an inbuilt feature of the BayesiaLab software package that
could be utilized when the related data set has been associated with the developed model. All
the fundamental work including the development of model, data association, calculation of the
probabilities and CPTs, inference and the sensitivity analysis was conducted in the academic
version of BAyesiaLab 8 software, which is a comprehensive, consistent, efficient and robust
tool in BN domain. The adopted methodology has been elaborated in detail below.
3.1. Development of nodes and causal relationship
To develop a BN model, the initial stage is to identify the select the variables that affects the
scenario under study. After the variables have been selected, these are represented as nodes
and states in the BN environment. The next stage is to identify and develop the causal relation-
ship between these nodes and impart it an appropriate graphical representation, termed as the
direct acyclic graph. The development of DAG holds critical prominence in imparting a justi-
fied development and interpretation of the cause-consequence relationship. However, to con-
struct these relationships through the incorporation of mathematical expressions is renowned
as a task next to impossible. A DAG consist of nodes and arcs. The node from which the arc
arises is called as the parent node and it ends at the child node.
The most important and significant aspect in the BN domain is the availability of data. It
could be achieved from the relevant literature, accident reports and the databases and the con-
cerned authorities. Another prominent aspect in this regard is the expert judgment, which can
verily be employed as data. However, to augment the reliability and pragmatism, and fade the
ambiguity, the expert judgment should be replaced by the real data whenever becomes avail-
able. Almost all the governments and various international organizations keeps a record of the
accidents that have occurred in their jurisdiction fulfill their scrutiny criteria. Hence, whenever
an accident takes places, it is investigated thoroughly to its details to determine the nature of
the accident, identify the factors that caused the accident along its sequence of occurrence and
quantify the severity of consequences. Reports containing such information can serve as a
potentially rich data source for the Bayesian model.
However, such reports cannot be selected and used on random bases. A strict and compre-
hensive selection criterion is developed for it in concurrence to the expert judgment and litera-
ture review that identifies and shortlists the variables and nodes to be considered for the
proposed study. Hence, only those reports are considered for the study which satisfies the
developed scrutiny and imparts required data on all the selected variables. In this study, the
PLOS ONE | https://doi.org/10.1371/journal.pone.0252732 June 4, 2021
6 / 20
PLOS ONE Hazardous cargo accident and pollution risk
selection of nodes, states, their causal relationship and the criterion for reports selection was
done in concurrence to the available literature in this domain. In concurrence to this criterion,
the DAG was developed, data file was developed and arranged in a format that was in compli-
ance with the BayesiaLab conditions. The selection of nodes, development of causal relation-
ship and the model was also discussed with the subject matter experts working both in the
academic and research fields and the maritime and port industry. Hence, the final form of the
model went through a process of deliberations and changes, unless it was agreed upon by all
the participants.
3.2. Calculating probabilities and CPTs
Once the identification and selection of the nodes is completed, their causal relation has been
identified and the model is developed, the next stage is associated with probabilities calcula-
tion. The quantification of probabilities and CPTs could be achieved through expert judgment,
the retrieved accident reports or an amalgamation of these two. The quantification can be
achieved using various tools like Bayes theorem, logistics regression and maximum likelihood
estimation. In situations where the real data is available and can be associated with the BN
model, the probabilities and CPTs can be calculated through the “Parameter estimation”.
Parameter estimation is an inbuilt feature of the BayesiaLab software package and works on
the principle of maximum likelihood estimation. The mathematical expressions and further
details of the maximum likelihood estimation have been provided in an earlier study of the
authors and can be consulted from there [49].
3.3. Sensitivity analysis
Sensitivity analysis is conducting to determine the most critical factors for a specific result of a
specific scenario in a BN model. Sensitivity analysis imparts the magnitude or strength of the
two way association between the parent and child node. One of the prominent aspect to be
considered while conducting sensitivity analysis is the selection of number of parameters. The
concept of conducting the sensitivity analysis is making variations in the parameters and ana-
lyzing its impact on the other nodes or parameters. It could either be simple in which varia-
tions are made in only one parameter, or complex in which multiple parameters in a CPT are
considered. The reliability is believed to increase with increase in number of parameters and
complexity. Though it requires a comprehensive and far-reaching understanding of the joint
probability distribution and network parameters, but the sensitivity analysis involving multiple
parameters from multiple CPTs is believed to be the most authentic, reliable and holistic [48,
49]. BNs are believed to exhibit a robust and practical interaction between the considered vari-
ables through the induced variations in the selected parameters.
In this study, the sensitivity analysis will be conducted through an inbuilt feature of the
BayesiaLab software package called “Tornado Charts”. These tornado charts displays the maxi-
mum and minimum contribution of all the variables in a model towards a specific node and
state which is specified as the target node and state. The values of these charts could also be
converted in to numeric values and presented in tabular form. Apart from the sensitivity anal-
ysis, in order to determine the consistency and level of confidence in the produced results, a
BN model has to verify the following conditions acknowledged as the validation of model [2, 5,
51, 52]. These conditions could be summarized as, a variation brought in the prior probabili-
ties of parent nodes shall produce a relevant variation in the posterior probabilities of the child
node. Similarly, the magnitude of the effect induced by changing the probabilities shall remain
greater for the set evidence in comparison to the other sub factors in the model.
PLOS ONE | https://doi.org/10.1371/journal.pone.0252732 June 4, 2021
7 / 20
PLOS ONE Hazardous cargo accident and pollution risk
4. Hazardous cargo risk assessment in a port environment
Though the world has seen many advances and improvements in the port safety system, how-
ever, the port environment yet sees the accidents involving hazardous cargo. No port in the
world could be considered totally immune to these accidents. These hazardous cargo accidents
could occur in any part and activity level inside a port. The past accident reports reveals that
such accidents have occurred during the loading and unloading of the hazardous cargo and
transporting these cargos in and out of the port. Moreover, such accidents have also been
found to have taken place inside the port temporary storage areas. These accidents could result
in intermittent fires, catastrophic oil spills and chemical discharges, contamination of the air,
earth and water, corrosiveness, explosions and pollution of the surrounding environment. In
addition to these environmental issues, damage of goods, properties, port infrastructure and
portfolio are also of critical concern. Such accidents could incur huge financial losses.
4.1. Specifying nodes and states of BN
Likewise the manifold consequences of the hazardous cargo accidents, the factors which initi-
ate and cause such accidents are also diverse in nature. The nature of these causation factors
could either be human, organizational, management, facilities and natural. The human factor
is further classified into technical qualification, experience and attitude of the employees
involved in the handling of dangerous goods [10]. If the employees lack the necessary qualifi-
cation required to handle such dangerous goods, it could pose a serious risk. Similarly, a prior
experience of handling the dangerous goods also contributes to enhanced safety and any short-
coming or lack of experience on the part of involved employees could initiate a significant
threat [8, 10]. Moreover, the attitude of the involved employees also plays a very prominent
role in the safe operations associated with hazardous cargo. This attitude could be nature and
carelessness of employees themselves, their lack of training, excessive working hours, lack of
interest due to improper facilities and pay grade and understanding of the situation [8, 9].
The other most prominent dimension is the organization itself. It stands for the proper
warehousing of the dangerous goods at the port environment in concurrence to the rules
defined by national and international standard operating procedures [8, 9]. Similarly, the
development of their own site, project and work nature based safety regulations also holds crit-
ical prominence. Also, the development of a specific department, workforce and organization
to deal with the operations, storage and handling of these goods also plays a noteworthy role in
evading accidents [8].
Subject to the 348 past accident reports from 1990 till 2018 considered for this study, litera-
ture review [42, 54–66] and expert judgment, this study has focused at the important causation
factors of the dangerous cargo accidents in a port environment. These factors have been
arranged into a set of variables, and hence nodes in the model developed. All the factors con-
sidered have been depicted in the Table 1.
4.1.1. Human factor. This variable is included to define the role played by human factor
in the accident instigation. As discussed earlier, human factor is a prominent accident instiga-
tor and in this study three aspects of this variable are considered.
4.1.1.1. Experience. This sub-factor represent the experience level of the ship crew. It depicts
as if the crew was experienced enough while conducting their job or was indicated otherwise
in the accident report.
4.1.1.2. Attitude. This aspect of the human factor depicts the involvement and attitude of
the crew or involved personnel towards their job. It encompass the description in accident
reports about the job seriousness, follow of rules and commands, abuse of authority, profes-
sionalism and the use of alcohol or any other drugs.
PLOS ONE | https://doi.org/10.1371/journal.pone.0252732 June 4, 2021
8 / 20
PLOS ONE Hazardous cargo accident and pollution risk
Table 1. Depiction of the variables considered for this study.
Variable
Qualification
Attitude
Experience
Human Factor
Warehousing
Operations
Emergency
State ‘0’
State ‘1’
Yes
Good
Yes
No
Bad
No
No Effect
Influence
Good
Good
Good
Poor
Poor
Poor
Variable
Infrastructure
Equipment
Facilities
Registration
Safety Protocols
Management
Natural Factors
Organizational Factors
No Effect
Influence
Environment and Pollution Risk
State ‘0’
State ‘1’
Good
Good
Good
Good
Good
Normal
Normal
Normal
Poor
Poor
Poor
Poor
Poor
Bad
Bad
Bad
In the above table, all of the variables are defined as binary in nature and two states as “0” and “1” have been specified for it elaborating its own meaning. Majority of the
variables considered are self-explanatory. However, to avoid any ambiguity, the variables under consideration are being clarified.
https://doi.org/10.1371/journal.pone.0252732.t001
4.1.1.3. Qualification. This sub-factor indicates if the involved staff was qualified enough to
do their job.
4.1.2. Organizational factors. Organizational factors are also considered to be prominent
accident instigators. It is also a broader domain and hence its different aspects have been con-
sidered as sub-factors.
Warehousing. The “Warehousing” node stands for the effective storage and management of
dangerous goods in the port warehouses.
4.1.2.1. Operations. The “Operations” node stands for if the port authorities have
defined and developed a specific team and organization for the storage and handling of the
dangerous goods and also that they are performing their duties in concurrence to the defined
regulations.
4.1.2.2. Emergency. The “Emergency” node depicts the availability and capability of a spe-
cific emergency response team that could timely arrive and manage the situation at site in case
of dangerous cargo accident.
4.1.3. Facilities. This factor indicates how the accident reports have mentioned the role of
involved facilities in accident causation. There could be lack of facilities and their malfunction-
ing or poor state.
4.1.3.1. Infrastructure. The “infrastructure” node depicts if the port infrastructure meets the
requirements of the dangerous goods handling. It could be good if it meets the standards and
there is no objection or bad otherwise.
4.1.3.2. Equipment. The “Equipment” node demonstrates if the port authorities have all the
required equipment and machineries in enough numbers, condition and maintenance backup
to ensure the smooth handling and operations of hazardous cargos.
4.1.4. Management. An effective management and conduction of the port activities, stan-
dard operating procedures and assortment have a critical role in safe hazardous cargo port
operations, which has been depicted by this variable. The different aspects of this variable are
given below.
4.1.4.1. Registration. This sub-factor depicts if all the dangerous cargo have been efficiently
registered and the corresponding data have been put on accessible records. This variable indi-
cates if the accident reports have included the entailment of poor or lack of registration of dan-
gerous goods towards accident causation
4.1.4.2. Safety protocols. The node “Safety Protocols” refers to the development and imple-
mentation of effective and up-to-date safety procedures and guidelines. It also stands for the
PLOS ONE | https://doi.org/10.1371/journal.pone.0252732 June 4, 2021
9 / 20
PLOS ONE Hazardous cargo accident and pollution risk
effective supervision and management of the health and safety situation associated with the
hazardous cargo.
4.1.5. Natural factors. This variable indicates the role played by natural phenomenon in
accident causation. It encompass the effects of rains, storms, winds, thunderstorms, lightening
and tsunamis etc.
4.1.6. Environment and pollution risk. The environmental and pollution risk has been
classified into two states. Where, the “Normal” state stands for small leakages, insignificant smaller
accidents and scenarios in which failure in one or more than other contributory factors had
occurred, but the hazardous cargo accident didn’t take place. While, the “Bad” state stands for acci-
dents which had noticeable consequences and pollution from the leakages, fires and explosions.
Once the nodes and states were specified and the data from accident reports was arranged
in concurrence to theses states and nodes, the selected nodes were developed into Bayesian
Model. This Bayesian Model is also termed as the Directed Acyclic Graph (DAG), where all
the connections between the nodes were developed in agreement to the available literature and
expert judgment. The model developed for this study is depicted in Fig 1.
All the reports from the duration under study which were satisfying the basic criteria of ful-
filling the data availability requirement for all the nodes were set into a data file. This data was
then associated with the developed model using the “associate data” feature of the BayesiaLab
software. Once the data file was associated, all the probabilities and conditional probability
tables were calculated through the parameter estimation function of the software.
4.2. Results and discussion
After running the model in the BayesiaLab environment, the inference results drawn as shown
in Fig 2, indicates that accidents with noticeable pollution, property and monetary losses had
Fig 1. DAG of the developed model.
https://doi.org/10.1371/journal.pone.0252732.g001
PLOS ONE | https://doi.org/10.1371/journal.pone.0252732 June 4, 2021
10 / 20
PLOS ONE Hazardous cargo accident and pollution risk
Fig 2. Inference results under normal conditions.
https://doi.org/10.1371/journal.pone.0252732.g002
highest occurrence probability of 59.80, while the accidents with insignificant or very minor
consequences had a 20% lower rate of occurrence.
Looking at the causation factors in Fig 2, it could be seen that human factor had the highest
contribution with a probability of 76.80, which is very prominent and in concurrence to the
human factor involvement in the maritime transportation accidents domain. Among the sub
factors in Fig 2, the attitude of the staff associated with the dealing of hazardous cargo mattered
the most. The human attitude had the highest probability of 70. The attitude of the staff holds
critical significance towards the control of hazardous cargo accidents. The staff needs to be
evaluated for their vigilance, interests, contribution and enthusiasm towards their duty. A
proper policy needs to be devised for the fatigue surveillance and no staff member shall be put
under duress for extra working hours or situations which are considered unfavorable for work-
ing [10, 67]. Also, the pay grade, safety gear and all other facilities of the staff shall be given due
attention as it contributes towards the mind presence of staff in the work and can be very effi-
cient in situations of an undesired event [8]. Moreover, in case of an undesired event or emer-
gency, the situation handling and management is the responsibility of every person within the
port area to contain the damage and pollution. This responsibility doesn’t limit to the port per-
sonnel only, but extends to the people in the vicinity of the port.
PLOS ONE | https://doi.org/10.1371/journal.pone.0252732 June 4, 2021
11 / 20
PLOS ONE Hazardous cargo accident and pollution risk
The qualification and experience of the staff both has a probability of 60 in the results under
normal situations as depicted in Fig 2. These results highlights the need for the risk mitigation
qualification and training on the emergency response measures [8]. The port authorities and
the involved department of the government shall not believe that such qualifications and skill-
set comes on its own, rather they should focus on developing training programs to enhance
the skillset of the personnel who has to deal with the hazardous cargos in the port environ-
ment. The modern day seaports requires to be highly efficient and safe in order to ensure sus-
tainable businesses and operations. Hence, the port authorities shall focus on the proficiencies
of the involved personnel to enable them of effectively tackling various unforeseen situations
and handling the undesired events [10].
At the international level, the workers organizations have experienced the need for
enhanced skillset required in the modern day seaports subject to the persistently changing job
profiles. Therefore, the personnel should be trained and analyzed for their competency on per-
forming their duties related to the management of hazardous cargos. The international mari-
time organization also enforces that the personnel who deals with the hazardous cargo shall
specifically trained for it. This specified training will enhance their risk mitigation capabilities
and application along adaptation of the defined emergency protocols [8]. Though how critical
this is, port authorities around the globe have been found to violate these recommendations
and rules as evident from the accident reports.
In terms of the causation probability, in Fig 2, the second most prominent factor is the port
hazardous cargo management capability. It encompasses the prominence of both proper regis-
tration and documentation of the hazardous goods, and the development along implementa-
tion of the associated safety protocols, rules, measures and standard operating procedures.
Lack of the design and implementation or not in concurrence to the cotemporary standards of
the related safety protocols, is the highest contributing sub factor in this domain. It also
accounts for the lack of or substandard supervision and management of the staff doing their
duties and the lack of required latest technology for the hazardous cargo operations [68].
The second highest contributing causation factor in this domain is the proper registration
and documentation of the hazardous cargos. The proper and correct documentation holds
critical significance for the launch and success of rescue missions in case of emergencies. It
could be well elaborated by example that the fire extinguishers required and the safety proce-
dures for the annihilation of a fire initiated by gasoline are much different from that of a fire
which is started by kerosene oil, even though both of these materials falls under the same cate-
gory “flammable liquids” of the international maritime organization. Therefore, accurate doc-
umentation is considered to enhance the efficacy of the port safety system and transportation
of the hazardous cargos [68].
The third highest contributing causation factor in Fig 2 is the facilities. It encompasses the
prominence of the port hazardous cargo related infrastructure and the equipment. Port infra-
structure holds critical value as it is an indications of its capability to successfully and reliably
accommodate the hazardous cargo operations. Similarly, the availability of the latest technol-
ogy, modern equipment in the required number, and reliable maintenance along backup is
very critical for the port hazardous cargo operations [8].
Now, employing the inverse propagation property of the Bayesian network, evidence is set
at the environment and pollution related hazardous cargo accident to occur. It determines the
contribution of all the considered factors in a scenario as if the accident has taken place. The
highest involvement in this scenario is of the management factor, which experience an increase
of 7.06% in its causation probability. While, the highest contributing sub factor in this regard
is the safety protocols. It determines the prominence of effective and cutting edge safety proto-
cols and emergency procedures to be in place and their management and implementation be
PLOS ONE | https://doi.org/10.1371/journal.pone.0252732 June 4, 2021
12 / 20
PLOS ONE Hazardous cargo accident and pollution risk
Table 2. Probabilities of the causation factors when evidence set at hazardous cargo environment and pollution accident to occur.
Node
State’s Probabilities
Node
State’s Probabilities
Natural Factors
Safety Protocols
Registration
Management
Equipment
Infrastructure
Facilities
Emergency
0
78.44
23.49
58.61
16.94
63.58
61.95
49.01
30.36
1
21.56
76.51
41.39
83.06
36.42
38.05
50.99
69.64
https://doi.org/10.1371/journal.pone.0252732.t002
Operations
Warehousing
Organizational Factors
Experience
Attitude
Qualification
Human Factors
0
40.35
70.21
47.57
39.46
29.73
39.46
22.26
1
59.65
29.79
52.43
60.54
70.27
60.54
77.74
strictly monitored. While, the second most important factor in this scenario is the human
aspect. The qualification, experience and attitude of the staff dealing with hazardous cargo has
a critical role in the occurrence of a hazardous cargo accident in a port environment. The cau-
sation probabilities of all the factors in this scenario are presented in Table 2.
Ensuring the hazardous cargo related safety in a port environment is a multifactor based
scenario, that specifically revolves around the human, management and organization based
factors [10]. The human factor in this regard is considered most critical as it is embedded in
the other systems too and plays its role in accident causation and evading in both the poor and
efficiently designed systems. From the perspective of warehouse management, not every stor-
age is appropriate for hazardous cargo storage as it requires additional safety and storage mea-
sures. Enhanced fire, explosion and leakage control systems are required in such storage units.
Also, hazardous cargo cannot be stored next to flammable goods and materials that may aid
the fire instigation. Therefore, the port authorities shall assign properly featured areas that has
specific capability for the storage of the hazardous cargos eliminating any threat or risk of an
accident and hence an environmental catastrophe. In these specifically allocated areas, a strict
and efficient supervision requires to be instigated. Port authorities shall devise a warehouse
safety management system that is specifically designed for the hazardous cargos and can effi-
ciently handle the storage and transportation for all the categories and stake holders involved
[8, 9]. Moreover, the port hazardous cargo safety management system shall be based on the
accountability. Where, the employees will focus on their responsibility, skillset and cognizance
that would in turn optimize the safety system bearing fruitful results [10].
The port safety system and goods registration protocols shall be aligned with the interna-
tional and government defined rules [68]. Moreover, these hazardous cargo rules & regulations
on management shall be accustomed to the local prevailing conditions in the port environ-
ment. All the specific characteristic of the port features shall be amalgamated into the extensive
and purpose oriented safety systems of goods transport, storage, and handling and emergency
circumstances. Moreover, the registration data and record shall be maintained in compliance
to the international standards in a comprehensive way such that it could be readily available
for the scientific analysis and research [68]. Moreover, the use of information technology shall
be incorporated in the safety management system of the hazardous cargos so that a compre-
hensive risk management system could be developed not only at the government level but at
the international level focusing at the ports with same natural, infrastructure and management
features. These technological system shall be developed on the basis of the real time data incor-
poration so that all the relevant data could be embedded into the system and made readily
available. This will enable the system of all the pertinent risks and initiate early warning sys-
tems in case of any potential risk or operating system discrepancy.
PLOS ONE | https://doi.org/10.1371/journal.pone.0252732 June 4, 2021
13 / 20
PLOS ONE Hazardous cargo accident and pollution risk
Another prominent perspective in the port hazardous cargo system safety is the efficiency
and appropriateness of port’s relevant equipment and infrastructure [8]. From the infrastruc-
ture and equipment perspective, the type, number and maintenance are the key aspects that
the port authorities shall pay specific attention to. Moreover, the infrastructure and equipment
should be specifically designed and purpose built in concurrence to the geographic, climatic
and terminal features. Moreover, in concurrence to the local prevailing conditions, a robust
safety and emergency response system should be embedded into the system that will impart
the capability of timely and effective measures in case of an accident. The most prominent
safety and risk mitigation systems like pollution containment, sophisticated fire alarm and fire-
fighting system should be implanted. Moreover, these systems and equipment must remain
installed in enough quantities that enables the capability to mitigate any undesired even effi-
ciently [8].
Additionally, the other most protuberant aspect in this regard is the maintenance of such
infrastructure, equipment and safety systems. The port authorities shall conduct regular train-
ings and workshops on the proper, efficient and safe use of the infrastructure and equipment
systems. The concerned staff should be trained on the specific systems uses and must be able
to identify the right maintenance and repair at the right time to avoid any unforeseen cata-
strophic event. Also, the port authorities shall pay keen attention to the international certifica-
tions for their equipment and personnel, regular and timely updates of their systems and
arrange consistent and systematic calibration of all the in-use equipment and technology.
Inability on the part of personnel, management and authorities and unavailability of the
proper infrastructure and equipment can result in serious and calamitous hazardous cargo
accidents. These accidents are believed to have multidimensional consequences. It kills, induce
injuries, destroys properties, incur huge monetary losses and most importantly, it contami-
nates and pollutes the natural environment, which in itself is a multidimensional calamity
being detrimental not only to the humans, but all the living creatures.
4.3. Sensitivity analysis
The purpose of conducting sensitivity analysis is to recognize the most perilous and critical fac-
tors or parameters in the developed model and the scenario under analysis. These highly critical
parameters are associated with high probability of incidence and involvement in comparison to
the other factors considered in the model and study. However, there is no thumb rule or a speci-
fied amount of disparity that would turn a variable considerable. Sensitivity analysis is believed
to play a significant role in the determination of the most critical factors and variables that if pri-
oritized can considerably evade the happening of an accident and substantially curtail the sever-
ity of consequences. However, the ranking classification and selection is highly dependent on
the decision maker rather than forecaster or analyst who conducted the analysis.
The environment and pollution risk was set as the target node and state in constructed BN
model to conduct the sensitivity analysis. In the BayesiaLab environment, the sensitivity analy-
sis could be determined through the use of “tornado charts” feature. These tornado charts dis-
plays the maximum and minimum occurrence and contribution probability of each
considered node and variable and hence its effect on the occurrence of the set target node and
state. In this study, the minimum and maximum incidence probability values of each variable
for the set environment and pollution risk node are incorporated in to a table and presented
below in Table 3. The determined critical factors and their difference in probability were also
verified by setting evidence at each variable in concurrence to the available literature [49].
The results reveal that the management as overall and the devising and implementation of
safety protocols in particular are the most critical factors when it comes to the occurrence of a
PLOS ONE | https://doi.org/10.1371/journal.pone.0252732 June 4, 2021
14 / 20
PLOS ONE Hazardous cargo accident and pollution risk
Table 3. Sensitivity analysis for the environment and pollution accident risk.
Node
Probabilities
Node
Probabilities
Management
Safety Protocols
Facilities
Equipment
Min
42.20
46.83
54.45
54.45
Max
65.36
65.36
66.61
63.37
Infrastructure
Natural Factors
Registration
Human Factors
Min
56.88
58.64
58.41
57.38
Max
61.75
64.48
61.89
60.54
https://doi.org/10.1371/journal.pone.0252732.t003
hazardous cargo accident in a port environment. These factors needs to be given specific atten-
tion allocated specific resources for in order to circumvent the occurrence of such accidents.
Apart from that, the port facilities in terms of the infrastructure and equipment are ranked as
the second most critical factors in the incidence of such accidents. Therefore, the port equip-
ment and infrastructure must be paid a specific attention with regard to their availability,
quantity, maintenance, updating and calibration. The third most critical factor in this regard is
affixed to be the human. Which implies that the port authorities shall specifically emphasize
on the qualification, experience, training, awareness and responsibility realization of all the
staff in general and that associated with the dealings of hazardous cargos specifically. Devising
and implementing a port safety management and risk mitigation plan for the hazardous cargos
around these critical factors will certainly evade the occurrence of dangerous goods accidents
and diminish the severity of consequences.
The results of this study holds practical vitality and are in accordance with the trends
reported by other studies observed in the causation factors. The governance and management
of the hazardous cargo holds critical prominence in ensuring safety. A study aimed at the port
hazardous cargo logistics has reported the governance and management as on the key element
of their three-dimensional risk management plan [9]. Likewise, another study aimed at the
hazardous cargo handling in a port have reported a positive association between the proper
management and hazardous cargo accident risk. They have further reported that proper docu-
mentation, its understanding and implementation of a viable management policy can play a
significant role in the hazardous cargo accident risk reduction [8]. The following of rules for
everyone involved in the hazardous cargo operation and the implementation of properly
devised rules and regulations are key factors in hazardous cargo safety at ports.
Similarly, the availability and proper functioning of the related equipment and machinery
is also being reported as one of the dominant accident causation factors [8–10]. Likewise,
another study aimed at the analysis of hazardous cargo accident causation factors have
reported that around 20% of the accidents in their study were associated with equipment, 41%
were found associated with improper or illegal commands and operations, while 30% were
reported to be associated with inadequate safety management [69]. Similarly, human factor
has also been reported as one of the leading accident causation factors. Human error or factor
is considered to be one of the highest accident instigation factors in the maritime transporta-
tion sector, while in the hazardous cargo accidents it has also been found to be one of the most
significant contributing factor [8]. Its role in hazardous cargo accidents was found counting
for around 28% of the accidents under investigation and in another study was being consid-
ered among the three key components of the hazardous cargo risk management plan [9, 10].
5. Conclusion
Safety and environment protection are the most prominent concerns when it comes to the
dealing of hazardous cargo in the maritime transportation. In a port environment, the
PLOS ONE | https://doi.org/10.1371/journal.pone.0252732 June 4, 2021
15 / 20
PLOS ONE Hazardous cargo accident and pollution risk
hazardous cargo accident and pollution risk is instigated by different factors in a vaguely man-
ner. This study is aimed at the environment and pollution accident’s risk assessment in a port
environment. The methodology adopted is the combination of past accidents and expert judg-
ment for the identification of factors and developing the interdependency connections, while,
the model development and inference analysis is done through Bayesian Networks using Baye-
siaLab software. Past accident reports from 1990 till 2018 were incorporated into a set of vari-
ables and the developed model was run in the BayesiaLab environment. The results indicate
that without setting evidence at any variable, the probability of an environment accident with
considerable damages is 59.80. While, the most prominent contributing factors in this scenario
are the human and management with incidence probabilities of 76.8 and 76 respectively. The
attitude of the employees dealing with the hazardous cargo along the development and imple-
mentation of stringent safety protocols are the most prominent sub factors. If evidence is set at
the hazardous cargo environment and pollution accident to occur, the highest change in the
contribution probability occurs for the Management.
For a hazardous cargo accident to occur, the role of emergency handling and planning
holds critical prominence. If the emergency protocols are designed in wake of the past experi-
ences, concurrence to the latest technologies and latest rules and procedures defined and
adopted across the globe, it can significantly reduce the occurrence of such accidents. More-
over, appropriate, timely, well planned and quick emergency response is believed to signifi-
cantly reduce and control the calamitous consequences of such hazardous cargo accidents.
Additionally, the natural factors also an enhanced role in accident causation in this scenario.
Natural factors are multifaceted from the perspective of their impact, they not only act as insti-
gation agents in the form of rains, winds, storms, heavy tides, high water velocity, lightening
and tsunamis, but also have role in consequence severity. The high tides and water velocity can
result in farther and increased spread of oil spills, chemical leakages, and release of other
obnoxious materials. Likewise, high winds and storms can result in spread of aerial releases of
chemical and gaseous discharges.
A sensitivity analysis was conducted to reveal the most critical factors that could be focused
at in circumventing the hazardous cargo accidents and diminishing their catastrophic conse-
quences. The results reveal that the port authorities and concerned government departments
shall pay specific attention to the qualification and training of the employees and focus at
enhancing their work focus. Similarly, the devising the most advanced and site oriented safety
measures and their strict implementation along supervision can significantly evade the acci-
dent occurrence. The appropriate port infrastructure and availability of the modern and state
of the art equipment also holds critical prominence in evasion of the accidents.
This study has various limitations. All the variables taken into account in this study were
considered as binary variables, which can be improved by considering different states account-
ing for various levels of the variables severity or probability of contribution. Similarly, by
enhancing the number of states in each variable, various aspects of a single variable could be
considered. Different states like negligible, low, medium, high and severe can be added to each
node for its role in the accident causation which will further impart pragmatic knowledge and
comprehension of the level of indulgence of each aspect and factor. Moreover, this study
focused at the quantitative assessment of the involved factors, hence, further research can be
done on various qualitative aspects of the accident causation factors and associated conse-
quences. Moreover, the environment and pollution risk was considered as a broad spectrum
variable without dividing it further depending upon the type and severity. Consequence states
like life and property losses and environmental damages can be added. Also, further categori-
zation of each of these consequence states can be done on the basis of range of losses incurred.
The environment and pollution risk could further be categorized on the basis of its nature like
PLOS ONE | https://doi.org/10.1371/journal.pone.0252732 June 4, 2021
16 / 20
PLOS ONE Hazardous cargo accident and pollution risk
fire, explosion and leakages. Additionally, the spread and damages done by each of these cate-
gories could be quantitatively assessed and associated with specified influence of each instigat-
ing factor.
Supporting information
S1 Data. Model data.
(CSV)
Author Contributions
Conceptualization: Rafi Ullah Khan, Jingbo Yin.
Data curation: Rafi Ullah Khan, Faluk Shair Mustafa.
Formal analysis: Rafi Ullah Khan, Jingbo Yin.
Investigation: Rafi Ullah Khan, Jingbo Yin.
Methodology: Rafi Ullah Khan, Jingbo Yin, Faluk Shair Mustafa.
Project administration: Jingbo Yin.
Resources: Rafi Ullah Khan, Faluk Shair Mustafa.
Software: Rafi Ullah Khan.
Supervision: Jingbo Yin.
Validation: Rafi Ullah Khan.
Visualization: Rafi Ullah Khan.
Writing – original draft: Rafi Ullah Khan, Faluk Shair Mustafa.
Writing – review & editing: Rafi Ullah Khan, Jingbo Yin.
References
1. Chen J, Huang T, Xie X, Lee PT, Hua C. Constructing Governance Framework of a Green and Smart
Port. Journal of Marine Science and Engineering. 2019; 7(4).
2.
Tanackov I, Janković Z, Sremac S, Miličić M, Vasiljević M, Mihaljev-Martinov J, et al. Risk distribution of
dangerous goods in logistics subsystems. Journal of Loss Prevention in the Process Industries. 2018;
54:373–83.
3. Baalisampang T, Abbassi R, Garaniya V, Khan F, Dadashzadeh M. Review and analysis of fire and
explosion accidents in maritime transportation. Ocean Engineering. 2018; 158:350–66.
4.
Zhao B. Facts and lessons related to the explosion accident in Tianjin Port, China. Natural Hazards.
2016; 84(1):707–13.
5. Mehan A, Jansen M. Beirut Blast: A port city in crisis. The Port City Futures Blog 2020.
6. Briefings OAJEE. Oil spill disaster will pressure Mauritian government. (oxan-es).
7.
Zhang H-D, Zheng X-P. Characteristics of hazardous chemical accidents in China: A statistical investi-
gation. Journal of Loss Prevention in the Process Industries. 2012; 25(4):686–93.
8. Saruchera FJJoT, Management SC. Determinants of effective high-risk cargo logistics at sea ports: A
case study. 2020; 14:13. https://doi.org/10.1177/1359105318781904 PMID: 29939086
9.
10.
Liu J, Zhou H, Sun H. A three-dimensional risk management model of port logistics for hazardous
goods. Maritime Policy & Management. 2019; 46(6):715–34.
Lecue M, Darbra RM. Accidents in European ports involving chemical substances: Characteristics and
trends. Safety Science. 2019; 115:278–84.
11. Cao X, Lam JSL. A fast reaction-based port vulnerability assessment: Case of Tianjin Port explosion.
Transportation Research Part A: Policy and Practice. 2019; 128:11–33.
PLOS ONE | https://doi.org/10.1371/journal.pone.0252732 June 4, 2021
17 / 20
PLOS ONE Hazardous cargo accident and pollution risk
12.
Tatarinov V, Kirsanov A. Information support for safety insurance of road transport of dangerous goods.
IOP Conference Series: Materials Science and Engineering. 2019; 492:012006.
13. Santarremigia FE, Molero GD, Poveda-Reyes S, Aguilar-Herrando J. Railway safety by designing the
layout of inland terminals with dangerous goods connected with the rail transport system. Safety Sci-
ence. 2018; 110:206–16.
14. Molero GD, Santarremigia FE, Aragone´ s-Beltra´ n P, Pastor-Ferrando J-P. Total safety by design:
Increased safety and operability of supply chain of inland terminals for containers with dangerous
goods. Safety Science. 2017; 100:168–82.
15. Yang Z, Yang Z, Yin J. Realising advanced risk-based port state control inspection using data-driven
Bayesian networks. Transportation Research Part A: Policy and Practice. 2018; 110:38–56.
16.
17.
18.
Li KX, Yin J, Bang HS, Yang Z, Wang J. Bayesian network with quantitative input for maritime risk analy-
sis. Transportmetrica A: Transport Science. 2014; 10(2):89–118.
Li X, Liu T, Liu Y. Cause Analysis of Unsafe Behaviors in Hazardous Chemical Accidents: Combined
with HFACs and Bayesian Network. International Journal of Environmental Research and Public
Health. 2020; 17(1).
Jiang M, Lu J. Maritime accident risk estimation for sea lanes based on a dynamic Bayesian network.
Maritime Policy & Management. 2020; 47(5):649–64.
19. Panahi R, Ng AKY, Afenyo MK, Haeri F. A novel approach in probabilistic quantification of risks within
the context of maritime supply chain: The case of extreme weather events in the Arctic. Accident Analy-
sis & Prevention. 2020; 144:105673. https://doi.org/10.1016/j.aap.2020.105673 PMID: 32683131
20. Wang L, Wang J, Shi M, Fu S, Zhu M. Critical risk factors in ship fire accidents. Maritime Policy & Man-
agement. 2020:1–19.
21. Baalisampang T, Abbassi R, Khan F. Chapter One—Overview of Marine and Offshore Safety. In: Khan
F, Abbassi R, editors. Methods in Chemical Process Safety. 2: Elsevier; 2018. p. 1–97.
22. Ronza A, Fe´ lez S, Darbra RM, Carol S, Vı´lchez JA, Casal J. Predicting the frequency of accidents in
port areas by developing event trees from historical analysis. Journal of Loss Prevention in the Process
Industries. 2003; 16(6):551–60.
23. Darbra R-M, Casal J. Historical analysis of accidents in seaports. Safety Science. 2004; 42(2):85–98.
24. Peris-Mora E, Orejas JMD, Subirats A, Iba´ ñez S, Alvarez P. Development of a system of indicators for
sustainable port management. Marine Pollution Bulletin. 2005; 50(12):1649–60. https://doi.org/10.
1016/j.marpolbul.2005.06.048 PMID: 16095626
25. Dalton T, Jin D. Extent and frequency of vessel oil spills in US marine protected areas. Marine Pollution
Bulletin. 2010; 60(11):1939–45. https://doi.org/10.1016/j.marpolbul.2010.07.036 PMID: 20797735
26.
27.
Zhang B, Matchinski EJ, Chen B, Ye X, Jing L, Lee K. Chapter 21—Marine Oil Spills—Oil Pollution,
Sources and Effects. In: Sheppard C, editor. World Seas: an Environmental Evaluation ( Second Edi-
tion): Academic Press; 2019. p. 391–406.
Frazão Santos C, Michel J, Neves M, Janeiro J, Andrade F, Orbach M. Marine spatial planning and oil
spill risk analysis: Finding common grounds. Marine Pollution Bulletin. 2013; 74(1):73–81. https://doi.
org/10.1016/j.marpolbul.2013.07.029 PMID: 23941806
28. Arzaghi E, Abbassi R, Garaniya V, Binns J, Khan F. An ecological risk assessment model for Arctic oil
spills from a subsea pipeline. Marine Pollution Bulletin. 2018; 135:1117–27. https://doi.org/10.1016/j.
marpolbul.2018.08.030 PMID: 30301010
29. Boehm PD, Page DS. Exposure Elements in Oil Spill Risk and Natural Resource Damage Assess-
ments: A Review. Human and Ecological Risk Assessment: An International Journal. 2007; 13(2):418–
48.
30. Olita A, Cucco A, Simeone S, Ribotti A, Fazioli L, Sorgente B, et al. Oil spill hazard and risk assessment
for the shorelines of a Mediterranean coastal archipelago. Ocean & Coastal Management. 2012;
57:44–52.
31.
32.
33.
Lam JSL, Lassa JA. Risk assessment framework for exposure of cargo and ports to natural hazards
and climate extremes. Maritime Policy & Management. 2017; 44(1):1–15.
Tolun L, Martens D, Okay OS, Schramm KW. Polycyclic aromatic hydrocarbon contamination in coastal
sediments of the Izmit Bay (Marmara Sea): Case studies before and after the Izmit Earthquake. Envi-
ronment International. 2006; 32(6):758–65. https://doi.org/10.1016/j.envint.2006.03.014 PMID:
16678901
Tauseef SM, Abbasi T, Abbasi SA. Development of a new chemical process-industry accident database
to assist in past accident analysis. Journal of Loss Prevention in the Process Industries. 2011; 24
(4):426–31.
PLOS ONE | https://doi.org/10.1371/journal.pone.0252732 June 4, 2021
18 / 20
PLOS ONE Hazardous cargo accident and pollution risk
34. Ko¨ rvers PMW, Sonnemans PJM. Accidents: A discrepancy between indicators and facts! Safety Sci-
ence. 2008; 46(7):1067–77.
35. Ondiviela B, Juanes JA, Go´ mez AG, Sa´ mano ML, Revilla JA. Methodological procedure for water qual-
ity management in port areas at the EU level. Ecological Indicators. 2012; 13(1):117–28.
36. Cucco A, Sinerchia M, Ribotti A, Olita A, Fazioli L, Perilli A, et al. A high-resolution real-time forecasting
system for predicting the fate of oil spills in the Strait of Bonifacio (western Mediterranean Sea). Marine
Pollution Bulletin. 2012; 64(6):1186–200. https://doi.org/10.1016/j.marpolbul.2012.03.019 PMID:
22498317
37. El-Fadel M, Abdallah R, Rachid G. A modeling approach toward oil spill management along the Eastern
Mediterranean. Journal of Environmental Management. 2012; 113:93–102. https://doi.org/10.1016/j.
jenvman.2012.07.035 PMID: 22996006
38. Ellis J. Undeclared dangerous goods—Risk implications for maritime transport. WMU Journal of Mari-
time Affairs. 2010; 9(1):5–27.
39. Berle Ø, Asbjørnslett BE, Rice JB. Formal Vulnerability Assessment of a maritime transportation sys-
tem. Reliability Engineering & System Safety. 2011; 96(6):696–705.
40. Montewka J, Ehlers S, Goerlandt F, Hinz T, Tabri K, Kujala P. A framework for risk assessment for mari-
time transportation systems—A case study for open sea collisions involving RoPax vessels. Reliability
Engineering & System Safety. 2014; 124:142–57.
41. Goerlandt F, Montewka J. A framework for risk analysis of maritime transportation systems: A case
study for oil spill from tankers in a ship–ship collision. Safety Science. 2015; 76:42–66.
42.
Trucco P, Cagno E, Ruggeri F, Grande O. A Bayesian Belief Network modelling of organisational fac-
tors in risk analysis: A case study in maritime transportation. Reliability Engineering & System Safety.
2008; 93(6):845–56.
43. Antão P, Grande O, Trucco P, Soares CG, Martorell S, Barnett JJS, et al. Analysis of maritime accident
data with BBN models. 2009; 2:3265–73.
44. Wang YF, Xie M, Chin K-S, Fu XJ. Accident analysis model based on Bayesian Network and Evidential
Reasoning approach. Journal of Loss Prevention in the Process Industries. 2013; 26(1):10–21.
45. Celik M, Lavasani SM, Wang J. A risk-based modelling approach to enhance shipping accident investi-
gation. Safety Science. 2010; 48(1):18–27.
46. Khakzad N, Khan F, Amyotte P. Safety analysis in process facilities: Comparison of fault tree and
Bayesian network approaches. Reliability Engineering & System Safety. 2011; 96(8):925–32.
47. Yang ZL, Wang J, Bonsall S, Fang QG. Use of Fuzzy Evidential Reasoning in Maritime Security
Assessment. Risk Analysis. 2009; 29(1):95–120. https://doi.org/10.1111/j.1539-6924.2008.01158.x
PMID: 19141152
48.
Li KX, Yin J, Fan L. Ship safety index. Transportation Research Part A: Policy and Practice. 2014;
66:75–87.
49. Khan RU, Yin J, Mustafa FS, Liu H. Risk Assessment and Decision Support for Sustainable Traffic
Safety in Hong Kong Waters. IEEE Access. 2020; 8:72893–909.
50.
Li S, Meng Q, Qu X. An Overview of Maritime Waterway Quantitative Risk Assessment Models. Risk
Analysis. 2012; 32(3):496–512. https://doi.org/10.1111/j.1539-6924.2011.01697.x PMID: 22035095
51. Goerlandt F, Montewka J. Maritime transportation risk analysis: Review and analysis in light of some
foundational issues. Reliability Engineering & System Safety. 2015; 138:115–34.
52. Ha¨nninen M, Kujala P. Influences of variables on ship collision probability in a Bayesian belief network
model. Reliability Engineering & System Safety. 2012; 102:27–40.
53. Ha¨nninen M, Valdez Banda OA, Kujala P. Bayesian network model of maritime safety management.
Expert Systems with Applications. 2014; 41(17):7837–46.
54. Petukhov I, Steshina L. Training Personalization for Operators of Complex Equipment. Procedia—
Social and Behavioral Sciences. 2015; 186:1240–7.
55. Puisa R, Lin L, Bolbot V, Vassalos D. Unravelling causal factors of maritime incidents and accidents.
Safety Science. 2018; 110:124–41.
56.
Zhao L, Wang X, Qian Y. Analysis of factors that influence hazardous material transportation accidents
based on Bayesian networks: A case study in China. Safety Science. 2012; 50(4):1049–55.
57. Nævestad T-O, Phillips RO, Størkersen KV, Laiou A, Yannis G. Safety culture in maritime transport in
Norway and Greece: Exploring national, sectorial and organizational influences on unsafe behaviours
and work accidents. Marine Policy. 2019; 99:1–13.
58.
Johnson CW, Holloway CM, editors. A Longitudinal Analysis of the Causal Factors in Major Maritime
Accidents in the USA and Canada (1996–2006). The Safety of Systems; 2007 2007//; London:
Springer London.
PLOS ONE | https://doi.org/10.1371/journal.pone.0252732 June 4, 2021
19 / 20
PLOS ONE Hazardous cargo accident and pollution risk
59. Wang S, Yin J, Khan RU. The Multi-State Maritime Transportation System Risk Assessment and Safety
Analysis. Sustainability. 2020; 12(14).
60.
Fan S, Yang Z, Blanco-Davis E, Zhang J, Yan X. Analysis of maritime transport accidents using Bayes-
ian networks. Proceedings of the Institution of Mechanical Engineers, Part O: Journal of Risk and Reli-
ability. 2020; 234(3):439–54.
61. Antão P, Guedes Soares C. Fault-tree models of accident scenarios of RoPax vessels. International
Journal of Automation and Computing. 2006; 3(2):107–16.
62. Wang YF, Faghih Roohi S, Hu XM, Xie M. Investigations of Human and Organizational Factors in haz-
ardous vapor accidents. Journal of Hazardous Materials. 2011; 191(1):69–82. https://doi.org/10.1016/j.
jhazmat.2011.04.040 PMID: 21571433
63. Ahmed IU, Gaweesh SM, Ahmed MM. Exploration of Hazardous Material Truck Crashes on Wyoming’s
Interstate Roads using a Novel Hamiltonian Monte Carlo Markov Chain Bayesian Inference. Transpor-
tation Research Record. 2020:0361198120931103.
64. Ma C, Zhou J, Yang D. Causation Analysis of Hazardous Material Road Transportation Accidents
Based on the Ordered Logit Regression Model. International Journal of Environmental Research and
Public Health. 2020; 17(4). https://doi.org/10.3390/ijerph17041259 PMID: 32075317
65. Batarliene N. Essential Safety Factors for the Transport of Dangerous Goods by Road: A Case Study of
Lithuania. Sustainability. 2020; 12(12).
66. Kadir Z A., Mohammad R, Othman N, Amrin A, Muhtazaruddin MN, Abu-Bakar SH, et al. Risk Manage-
ment Framework for Handling and Storage of Cargo at Major Ports in Malaysia towards Port Sustain-
ability. Sustainability. 2020; 12(2).
67. Cho S-J, Kim D-J, Choi K-SJJotKSoME, Safety. Hazardous and noxious substances (HNS) risk
assessment and accident prevention measures on domestic marine transportation. 2013; 19(2):145–
54.
68. Wu C, Huang L. A new accident causation model based on information flow and its application in Tianjin
Port fire and explosion accident. Reliability Engineering & System Safety. 2019; 182:73–85.
69. Wang B, Li D, Wu C. Characteristics of hazardous chemical accidents during hot season in China from
1989 to 2019: A statistical investigation. Safety Science. 2020; 129:104788.
PLOS ONE | https://doi.org/10.1371/journal.pone.0252732 June 4, 2021
20 / 20
PLOS ONE |
10.1371_journal.pone.0251372 | RESEARCH ARTICLE
Infertility management according to the
Endometriosis Fertility Index in patients
operated for endometriosis: What is the
optimal time frame?
Alexandre BailleulID
Arnaud Fauconnier1,4☯
1☯*, Julien Niro2‡, Joseph Du Cheyron3‡, Pierre Panel2‡,
1 Research Unit EA7285, Risk and Safety in Clinical Medicine for Women and Perinatal Health, Versailles St-
Quentin University, Montigny-le-Bretonneux, Versailles, France, 2 Department of Gynecology & Obstetrics,
Centre Hospitalier Andre´ Mignot, Versailles, France, 3 Clinical Research Department, Centre Hospitalier
Intercommunal de Poissy-Saint-Germain-en-Laye, Poissy, France, 4 Department of Gynecology &
Obstetrics, Centre Hospitalier Intercommunal de Poissy—Saint-Germain, Poissy, France
☯ These authors contributed equally to this work.
‡ These authors also contributed equally to this work.
* alexandre.bailleul@gmail.com
Abstract
Introduction
The Endometriosis Fertility Index (EFI) is a validated score for predicting the postoperative
spontaneous pregnancy rate in patients undergoing endometriosis surgery. However, the
practical use of the EFI to advise patients about postoperative fertility management is
unclear.
Materials and methods
All patients participating in the ENDOQUAL study–a prospective observational bi-center
cohort study conducted between 01/2012 and 06/2018–who underwent surgery for infertility
were asked to complete a questionnaire collecting time and mode of conception. Statistical
analysis was performed with the Fine and Gray model of competing risks and analysis of fer-
tility according to the EFI.
Results
Of the 234 patients analyzed, 104 (44.4%) conceived postoperatively including 58 (55.8%)
spontaneous pregnancies. An EFI of 0–4 for spontaneous pregnancies was associated with
a lower cumulative pregnancy incidence compared to an EFI of 5–10 (52 versus 34 preg-
nancies respectively, Subdistribution Hazard Ratio (SHR) = 0.47; 95% CI [0.2; 1.1]; p =
0.08). An EFI of 0–4 was associated with a higher cumulative pregnancy rate for pregnan-
cies obtained by artificial reproduction technology (ART), compared to an EFI of 5–10 (12
versus 6 pregnancies respectively, SHR = 1.9; CI95% [0.96; 3.8]; p = 0.06). Fecundability
decreased from 12 months for EFI 0–4 and from 24 months for EFI 5–10.
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: Bailleul A, Niro J, Du Cheyron J, Panel P,
Fauconnier A (2021) Infertility management
according to the Endometriosis Fertility Index in
patients operated for endometriosis: What is the
optimal time frame? PLoS ONE 16(5): e0251372.
https://doi.org/10.1371/journal.pone.0251372
Editor: Antonio Simone Laganà, University of
Insubria, ITALY
Received: January 17, 2021
Accepted: April 24, 2021
Published: May 12, 2021
Copyright: © 2021 Bailleul et al. This is an open
access article distributed under the terms of the
Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: All relevant data are
within the manuscript and its Supporting
Information files.
Funding: The author(s) received no specific
funding for this work.
Competing interests: The authors have declared
that no competing interests exist.
PLOS ONE | https://doi.org/10.1371/journal.pone.0251372 May 12, 2021
1 / 11
PLOS ONE Optimal infertility management according to the Endometriosis Fertility Index
Conclusion
Our analysis suggests that patients with an unfavorable EFI (�4) have more ART pregnan-
cies than patients with a favorable EFI (�5) and should be referred for ART shortly after sur-
gery. Patients with a favorable EFI may attempt spontaneous pregnancy for 24 months
before referral.
Introduction
Endometriosis is a benign gynecological disease which affects 6 to 10% of women of reproduc-
tive age [1]. This extrinsic localization of endometrial tissue is responsible for chronic inflam-
mation generating in anatomical pelvic modifications [2]. However, the etiopathogenesis of
endometriosis is a multifactorial process resulting in a heterogeneous disease [3]. Patients usu-
ally present chronic pelvic pain, infertility or impaired quality of life; these symptoms can be
associated with one another [4]. Spontaneous pregnancy rates in women with endometriois-
related infertility have been reported at around 10% [5]. While medical treatments can be
effective in the management of endometriosis, surgery has been demonstrated to be a valid
therapeutic tool [6]. Surgical resection of the lesions can double the chances of woman natural
conception [7, 8].
The Endometriosis Fertility Index (EFI) was developed to predict the spontaneous preg-
nancy rate in women 3 years after surgery for endometriosis [9]. This multifactorial score
includes criteria based on the patient’s characteristics (age, duration of infertility, pregnancy
history), intra-operative lesion description (American Society for Reproductive Medicine
(ASRM), American Fertility Society (AFS) Endometriosis Score) and a functional post-opera-
tive score (Least Function (LF) Score). The EFI is the sum of the surgical and historical factors,
and ranges from 0 to 10. The rate of spontaneous pregnancies is greater in women with higher
EFI scores; cumulative non-ART pregnancy at 36 months was found to be 10% (95%CI: 3, 16;
P< 0.001) for women with an EFI of 0–2, and 69% (95%CI: 58, 79; P< 0.001) for women with
an EFI of 9–10 [10]. The EFI was described by the World Endometriosis Society (WES) in
2017 as being a robust and clinically valid score to predict fertility after surgery in patients
with endometriosis [11]. Consequently, the French College of Gynecologists and Obstetricians
(CNGOF) recommends that the EFI be used for guiding the post-operative strategy. However,
the EFI is a predictive tool rather than a therapeutic decision-making tool and clinical inter-
pretation to determine the best approach for post-operative fertility is ambiguous [12]. In the
literature, the decisional threshold indicating a good likelihood of spontaneous pregnancy is
generally taken to be 5 [13–15]. However, the current interest is to determine the role of the
EFI in the post-operative management of patients with endometriois-related infertility: i.e.,
spontaneous conception or assisted reproductive technology (ART) and the optimal time
frame for each.
The objective of the present study was therefore to assess the value of the EFI in deciding
the optimal time to refer patients for ART after surgery for endometriosis.
Materials and methods
Data source
The patients analyzed in the current study were all part of the ENDOQUAL cohort. ENDOQ-
UAL was an observational, prospective, bi-center study (CHI Poissy—St Germain en Laye
PLOS ONE | https://doi.org/10.1371/journal.pone.0251372 May 12, 2021
2 / 11
PLOS ONE Optimal infertility management according to the Endometriosis Fertility Index
(CHIPS) and CH Mignot de Versailles (CHV)) [16]. The aim of ENDOQUAL was to investi-
gate the impact of different therapeutic modalities (medical treatment, surgical treatment,
ART) on quality of life and fertility by collecting clinical information on volunteer endometri-
osis patients. ENDOQUAL began on January 01, 2012 and the EFI was routinely measured
until January 01, 2017. The study was approved by the Southeast IV Ethics Committee (Sud-
Est, n˚18/002) in France and the French National Committee for Information Technology and
Individual Liberties (N˚906 253). Our study was purely observational and involved no inter-
vention. As such, no written informed consent was required under French law (the Huriet-
Serusclat Act of December 20, 1998). Nevertheless, all the patients received information about
the study and were free to participate or not.
Study design
All patients with endometriosis-related infertility (with or without pain component) and
women with endometriosis with an immediate desire to conceive who underwent surgery
between 01/01/2012 and 31/06/2018 were included. In our expert centers and in accordance
with French gynecologist recommendations, women who require immediate in vitro fertiliza-
tion (IVF) (e.g., for major male factor or tubal anomalies) are not eligible for surgery.
Non-inclusion criteria were women older than 45 years, radical surgery (hysterectomy,
bilateral adnexectomy), included in the medical arm of the ENDOQUAL study (hormonal
treatment or ART), or operated on for an indication other than infertility and without an
immediate desire to conceive.
The objective of the surgical procedure was therefore to improve spontaneous fertility. All
the women underwent a laparoscopic fertility-sparing resection of endometriosis to remove all
endometriosis implants and adhesions. Following the surgery, the women attempted to con-
ceive spontaneously. No specific recommendations were given during the study period about
whether to refer a patient for ART according to their EFI score.
Patients were excluded if the histology was negative for endometriosis, if they had a history
of pelvic surgery known to be at high risk of adhesion (laparotomy), or if they had a history of
multiple (> 2) pelvic surgical procedures for endometriosis.
The primary endpoint was the time of occurrence of a pregnancy (whatever its outcome)
according to the mode of conception: spontaneous or by ART (stimulation, insemination, or
IVF). The secondary endpoint was the postoperative fecundability, which is defined as the
probability of achieving a pregnancy within one menstrual cycle.
Follow-up
Fertility outcomes were collected with two questionnaires survey: (i) one sent one year after
the surgery according to the cohort study protocol, (ii) and a second one sent in January 2020.
If no reply was received, contact was made by telephone after checking for any change of
address. When a patient was lost to follow-up, her doctor was contacted. The questionnaires
covered the following information: how long the patient had attempted to achieve pregnancy,
any attempts at ART, and, for each pregnancy obtained after surgery, the date and means of
conception (spontaneous, ART) and its outcome (miscarriage, therapeutic abortion, voluntary
abortion, birth).
EFI score calculation
As the EFI was not used in regular practice between 2012 and 2017, the EFI score was calcu-
lated a posteriori for women who underwent surgery before 2017 (ENDOQUAL-1) by collect-
ing the relevant data from the detailed operative report of the database and the textual
PLOS ONE | https://doi.org/10.1371/journal.pone.0251372 May 12, 2021
3 / 11
PLOS ONE Optimal infertility management according to the Endometriosis Fertility Index
operative reports. After 2017 (ENDOQUAL-2), the EFI was included in the Case Report Form
(CRF) of the ENDOQUAL study. To avoid calculation bias and confirm the reproducibility of
the EFI, we compared the a posteriori calculation of the EFI scores with the EFI scores from
women included after 2017 [17].
The maximum EFI score is ten: five points are based on the patient’s characteristics (such
as age, duration of infertility, and history of pregnancy); two points on rASRM staging; and
the remaining three points on qualitative assessment by the surgeon (adnexal LF score) on
completion of the procedure [9]. For example, a 28-year-old nulliparous patient, who has been
infertile for 1 year, with only superficial endometriosis without adnexal lesions, has an EFI
score of 9. On the contrary, a 38-year-old nulliparous patient who has been infertile for more
than 3 years, with deep endometriosis (such as a complete posterior cul-de-sac obliteration
and bilateral endometrioma), has an EFI score of 3 after the removal of all implants and
adhesion.
Statistical analysis
The postoperative pregnancy rate and time to pregnancy were analyzed according to the con-
ception mode (spontaneous or ART) for two groups of patients: patients with an EFI of 0–4
and those with an EFI of 5–10. The time from the beginning of the observation period was the
date of the surgery.
Patient characteristics were recorded and compared according to time to pregnancy, age,
endometriosis classification (ASRM, AFS Endometriosis Score, LF Score), EFI, tobacco use,
Body Mass Index (BMI), and Anti-Mu¨llerian Hormone (AMH) levels.
As patients may have two competing events (spontaneous conception and ART concep-
tion), we used a competing risk approach to explore the relation between EFI and fertility [18].
Modeling cumulative incidence curves by the semi-parametric Fine and Gray model (1999)
allowed the calculation of the risk function associated with the cumulative incidence function
(Subdistribution Hazard Ratio—SHR) by a competing risks multivariate analysis. [19, 20].
This analysis compared the cumulative pregnancy curves with various co-variables. This
model (Fine and Gray / SHR) was the most appropriate to analyzed and interpreted our data
and responded to our primary endpoint.
Cause-specific hazard ratios (CSHR) is using a Cox proportional hazard model, in which
patients who experience other events are treated as censored for the event of interest. Sub-dis-
tribution hazard ratios (SHR) is using the Fine and Gray model, in which woman who experi-
ence other events are treated as immune (i.e., “cured” and remaining in the risk sets) to the
event of interest.
Finally, the objective was to determine the time lost before conception between the two
groups. Analyses were based on Area Under the Curve (AUC) and fecundability [21–23]. For
both these analyses, we focused only on spontaneous conception and patients were censored
when they were lost to follow-up or became pregnant with ART.
Calculating the AUC in censored data provides the Restricted Mean Survival Time
(RMST). This time is equivalent to the area under the Kaplan-Meier curve from the start of the
study to an interest time point (36 months in our study, i.e., the maximum follow-up of the
patients’ fertility after surgery in accordance with Adamson et al. [9]). To quantify the differ-
ence between the two EFI groups, the analysis was performed on the difference in RMSTs. It is
interpreted as a gain or loss of event-free survival time for a given period. The RMST, in our
analysis, corresponds to a Restricted Mean Conception Time (RMCT). A decomposition of
6-month postoperative intervals aimed to find the time from which patients with an EFI of
0–4 had a loss of chance for spontaneous conception compared to the other group.
PLOS ONE | https://doi.org/10.1371/journal.pone.0251372 May 12, 2021
4 / 11
PLOS ONE Optimal infertility management according to the Endometriosis Fertility Index
Fecundability is the probability of conception per cycle during an interval [24]. The esti-
mate of the monthly fecundability ^f was obtained by dividing the number of conceptions
observed Ci by the sum of the person-months of exposure Ci during an interval:
^f ¼
SCi
STi
All statistical analysis was performed using available software (R—1.2.5019).
Results
During the study period, 986 patients were enrolled in the ENDOQUAL cohort. Among them
275 meet the inclusion criteria and du to exclusion criteria the final analysis included 234
patients with infertility or immediate desire to conceive after the surgery (Fig 1).
The patients’ clinical characteristics according to pregnancy status and conception mode
are shown in Table 1. During the follow-up, 104/234 patients (44.4%) became pregnant,
including 58 (55.8%) spontaneous conceptions and 46 (44.2%) obtained by ART. The mean
follow-up among the patients who did not conceive was 29.7 months (± SD 24.7). For all preg-
nancies, mean conception time was 17 months (± SD 15.6). The time to conceive was lower
for spontaneous conception compared to ART conception (14.3 months versus 20.4 months,
p = 0.047, 95% CI [-12.1; -0.09]).
The mean a posteriori EFI calculated in patients from ENDOQUAL-1 was 6.1 (± SD 1.73)
versus 5.6 (± SD 2.74) for the patients from ENDOQUAL-2, with no significant difference
between the two results (p = 0.19 95% CI [-0.25; 1.22]). The mean total EFI was 5.9 (± SD
2.04).
Fig 1. Flow chart. N, number of patients; CHIPS, Centre Hospitalier Intercommunal de Poissy St Germain; CHV, Centre Hospitalier de Versailles; EFI,
Endometriosis Fertility Index.
https://doi.org/10.1371/journal.pone.0251372.g001
PLOS ONE | https://doi.org/10.1371/journal.pone.0251372 May 12, 2021
5 / 11
PLOS ONE Table 1. Characteristics of study patients according to pregnancy status and conception mode.
ALL PATIENTS
PREGNANT
NON PREGNANCY
p���
Spontaneous
ART
p��
Optimal infertility management according to the Endometriosis Fertility Index
N = 234
n (%)�
Mean Follow up month (SD)
22.9 (21.2)
Mean Conception time months (SD)
17 (15.6)
Age Mean years (SD)
32 (4.7)
< 35 years
36–39 years
> 40 years
Mean Infertility Time months (SD)
32 (27)
< 3 years
> 3 years
Prior Pregnancy
Yes
No
160 (68)
58 (25)
16 (7)
170 (73)
64 (27)
77 (33)
157 (67)
N = 58
n (%)�
14.3 (12.7)
14.3 (12.7)
30.5 (4.8)
47 (81)
9 (16)
2 (3)
N = 46
n (%)�
20.4 (18.2)
20.4 (18.2)
31.8 (4.3)
31 (68)
13 (28)
2 (4)
23.2 (17.8)
30.5 (18.8)
46 (79)
12 (21)
16 (28)
42 (72)
30 (65)
16 (35)
10 (22)
36 (78)
0.047
0.14
0.29
0.12
0.17
0.65
AFS Total Mean (SD)
43 (37.5)
32.3 (29.9)
53 (42)
< 71
> 71
ASRM
I
II
III
IV
179 (76)
55 (24)
28 (12)
47 (20)
64 (27)
95 (41)
49 (84)
9 (16)
7 (12)
18 (31)
17 (29)
16 (28)
32 (70)
14 (30)
2 (4)
10 (22)
10 (22)
24 (52)
AFS Endometriosis Score Mean (SD)
14.2 (12.8)
LF Score Mean (SD)
5.3 (1.8)
11.7 (10.3)
5.8 (1.7)
16.5 (15.3)
5.1 (1.7)
[1–3]
[4–6]
[7–8]
38 (16)
139 (60)
57 (24)
5 (9)
34 (58)
19 (33)
9 (20)
29 (63)
8 (17)
EFI Mean (SD)
6 (2)
6.4 (1.6)
5.9 (1.9)
[0–2]
[3–4]
[5–6]
[7–8]
[9–10]
Tabac
13 (5.5)
28 (12)
93 (40)
76 (32.5)
24 (10)
40 (17)
1 (2)
5 (9)
25 (43)
21 (36)
6 (10)
7 (12)
4 (9)
8 (17)
14 (30)
16 (35)
4 (9)
4 (9)
AMH (ng/ml) Mean (SD)
BMI (kg/m2) Mean (SD)
4.4 (4.5)
23.3 (4.7)
5 (3.8)
23.2 (4)
4.8 (5.3)
23.7 (5.3)
0.07
0.07
0.03
0.10
0.16
0.27
0.37
0.85
0.63
N = 89
n (%)�
29.7 (24.7)
33 (4.7)
38 (34)
55 (62)
26 (29)
8 (9)
63 (71)
26 (29)
32 (36)
57 (64)
62 (70)
27 (30)
10 (11)
12 (13)
22 (25)
45 (51)
15.2 (13.2)
5.1 (1.9)
18 (20)
53 (60)
18 (20)
6 (7)
11 (12)
39 (44)
25 (28)
8 (9)
17 (19)
5.8 (2.1)
3.4 (3.8)
23 (3.7)
< 0.01
0.11
51 (40)
0.01
0.15
0.05
0.27
0.21
< 0.01
0.10
0.038
0.16
0.04
0.16
0.18
0.56
< 0.01
0.27
0.67
� Results presented are numbers of patients (%) unless otherwise stated.
�� Comparison of pregnant women by method of conception.
��� Comparison groups of pregnant women and non-pregnant women. Legend: ART, Assisted Reproductive Technology; SD, Standard Deviation; AFS, American
Fertility Society; ASRM, American Society for Reproductive Medecine; LF, Least Function; EFI, Endometriosis Fertility Index; AMH, Anti-Mu¨llerian hormone; BMI;
Body mass index.
https://doi.org/10.1371/journal.pone.0251372.t001
Cumulative incidence of all pregnancies with competing risks
Cumulative incidence curves in the presence of competing risks revealed a higher rate of preg-
nancies by spontaneous conception in patients with an EFI of 5–10 (38.8%, 95% CI [25.3;
52.98] at 36 months versus 18.6%, 95% CI [0.42; 64.1] for an EFI of 0–4) and a higher rate of
PLOS ONE | https://doi.org/10.1371/journal.pone.0251372 May 12, 2021
6 / 11
PLOS ONE Optimal infertility management according to the Endometriosis Fertility Index
Fig 2. Cumulative incidence function according to conception mode and EFI with CSHR (Cause Specific Hazard Ratio) and SHR (Sub
Distribution Hazard Ratio) analysis. ART, Assisted Reproductive Technology; CSHR, Cause Specific Hazard Ratio; SHR, Sub-distribution Hazard
Ratio; IC, confidence interval; EFI, Endometriosis Fertility Index.
https://doi.org/10.1371/journal.pone.0251372.g002
pregnancies by ART conception in patients with an EFI of 0–4 (33.3%, 95% CI [9.92; 65.11] of
cumulative pregnancies at 36 months versus 22.5%, 95% CI [10.7; 41.17] for patients with an
EFI of 5–10) (Fig 2).
Using the Fine and Gray model, the likelihood of a natural pregnancy tended to be lower in
patients with an EFI of 0–4 compared with an EFI of 5–10 (SHR = 0.47, 95% CI [0.2; 1.1], p = 0.08).
Patients with an EFI of 0–4 had a higher success with an ART pregnancy than patients with
a an EFI of 5–10. (SHR = 1.9, 95% CI [0.96; 3.8] p = 0.06). The results were consistent with
those posted by the CSHR. The overall results were not significant.
RMCT and fecundability
The difference between the two RMCT groups increased progressively over time with the
main increase occurring 18–24 months postoperatively as shown in Table 2. At 36 months,
PLOS ONE | https://doi.org/10.1371/journal.pone.0251372 May 12, 2021
7 / 11
PLOS ONE Optimal infertility management according to the Endometriosis Fertility Index
Table 2. AUC and fecundability (%, 95%CI) according to EFI for spontaneous pregnancy.
Post Operative Interval
N
AUC
Fecundability
RMCT
Time Lost (months)
EFI score 0–4
EFI score 5–10
EFI score 0–4
EFI score 5–10
0–6 months
6–12 months
12–18 months
18–24 months
24–30 months
30–36 months
0–36 months
17
15
9
7
3
1
52
5,90
5,32
5,00
4,48
4,48
4,48
29,65
5,75
5,08
4,72
4,19
3,65
3,38
26,85
0,15
0,24
0,28
0,29
0,83
1,10
2,80
0.6 [-0.5; 1.7]
2.4 [-0.3; 5.1]
1.2 [-2.2; 3.6]
0
0
0
0.9 [0; 1.8]
1.9 [1; 2.8]
1.8 [0.8; 2.8]
1.6 [0.5; 2.7]
2.1 [0.5; 3.7]
1.3 [-0.1; 2.7]
0.6 [-0.5; 1.7]
1.8 [1.3; 2.7]
N, number of spontaneous pregnancies; AUC, Aera Under Curve; RMCT, Restricted Mean Conception Time; CI, Confidence Interval; EFI, Endometriosis Fertility
Index.
https://doi.org/10.1371/journal.pone.0251372.t002
patients with an EFI of 0–4 had lost 2.8 months (HR = 1.78, 95% CI = [0.76–4.16], p = 0.18) to
achieve a pregnancy, without statistically significant difference.
Over 36 months postoperatively, the fecundability in patients with an EFI of 0–4 was 0.9%
versus 1.8% for patients with an EFI of 5–10. The fecundability of patients with an EFI of 0–4
decreased from 12 months postoperatively. For patients with an EFI of 5–10, fecundability was
stable until 24 months postoperatively, followed by a progressive decrease (Table 2). The prob-
ability of spontaneous conception for patients with an EFI of 0–4 was optimal between 6 and
12 months postoperatively.
Discussion
As well as being a robust tool for predicting the pregnancy rate after surgery in women with
endometriosis-related infertility, the EFI is also an important element to be considered in the
therapeutic decision-making process. We observed more ART pregnancies in patients with an
EFI of 0–4. Patients with a favorable EFI (�5) had more natural pregnancies than patients
with an unfavorable EFI (�4). These results suggest that endometriosis patients with an EFI of
0–4 should be referred to an ART unit rapidly after surgery.
We found that fecundability was optimal up to 12 months postoperatively in women with
an EFI �4, after which it decreased. For patients with an EFI �5, fecundability was stable up
to 24 months postoperatively before decreasing. Fecundability indicates the time during which
a surgeon should refer the patient to an ART unit. Therefore, according to our results, women
with an EFI �4 should be referred within 12 months and those with an EFI �5 within 24
months. These results are confirmed by the AUC analysis. The loss of time for a spontaneous
conception gradually increased postoperatively for women with an EFI of 0–4 versus those
with an EFI of 5–10.
Our analysis was based on the ENDOQUAL study, an important bi-center prospective
cohort of endometriosis patients allowing continual evaluation of health practices. As in other
studies, regular reassessment by questionnaires and telephone interviews constitutes the usual
reference techniques to monitor patients’ fertility [14, 15, 25].
One of the strengths of our study lies in the statistical model we used based on competing
risk which is, to the best of our knowledge, innovative in this context. In their princeps article,
Adamson and Pasta (2010) applied the gold standard for analyzing the effectiveness of infertil-
ity treatments, i.e., analysis by censored data and the survival curve method [22, 23]. However,
the 801 patients analyzed in the article were operated on for infertility and not pain, which is
PLOS ONE | https://doi.org/10.1371/journal.pone.0251372 May 12, 2021
8 / 11
PLOS ONE Optimal infertility management according to the Endometriosis Fertility Index
not the case for all studies on the EFI [9, 26]. Recent advances in epidemiologic and biostatisti-
cal methods have made available multiple tools to describe differences in times to outcomes
related to an exposure in this context [18, 21]. The competing risk model in our study gave
similar results to the AUC analysis and fecundability for inter-group differences and intra-
group evaluation, respectively.
One limitation of our study is the small number of patients in each group when we analyze
time to conception. Nevertheless, despite imperfections from our population, our analysis
allowed us to identify therapeutic profiles and determine an optimal care approach. Another
limitation is that we lacked precise information about how long the patients actually attempted
pregnancy taking into account the duration of postoperative hormonal treatment or post-
operative pain, for example, which may have postponed attempts to conceive naturally. Fur-
thermore, although all patients were allowed to attempt natural conception, we do not know
precisely if women became pregnant naturally after ART failure or how long they were
exposed to natural conception before being referred for ART. Nevertheless, the statistical anal-
ysis tool we used (i.e., SHR and Fine and Gray model) minimized the bias related to this issue.
The monthly conception rate (0% to 2.5%) in our population was lower than that of Mar-
coux et al. (1997) [27]. These authors demonstrated that fertility was improved in women who
underwent resection for endometriosis lesions compared to those would did not (4.7% versus
2.4%; RR = 1.9, 95% = [1.2–3.1]). They focused their analysis on patients with mild to moder-
ate stages of endometriosis (Stage I or II) [28]. Conversely, we included all endometriosis
stages and 68% of our patients had stage III or IV. The low fecundability we observed in
patients with an EFI of 0–4 can be explained by hormonal blockage observed postoperatively
in patients with severe endometriosis.
Physiopathologically, our results can be explained by a beneficial effect of the surgery by
restoring the pelvic anatomy. Patients with an EFI �4 may have residual endometriosis
(incomplete surgery, a poor LF Score). Zhang et al. explained that visible endometriotic lesions
are not sufficient to describe disease severity and that surgery cannot correct the associated
molecular and immune phenomena [25]. However it might be understood there is an impact
of endometriosis surgery on obstetrics outcomes, as Baggio et al. (2015) demonstrate [29].
With all these results, gynecologists may find it difficult to persuade women to opt for an
expectant management, but clinical experience shows that patients prefer to attempt a natural
conception. Surgery for endometriosis infertility was always to improve natural conception.
General conclusion
Although the EFI is a valid tool to predict spontaneous pregnancy rates after surgery for endo-
metriosis, using the EFI in routine clinical practice to triage patients remains a challenge. Our
study suggests that the optimal time frame to manage postoperative infertility varies according
to the EFI: patients with a favorable EFI (�5) may be allowed to have 24 months of spontane-
ous attempts to conceive whereas patients with an unfavorable EFI (�4) should be more rap-
idly referred for ART.
Supporting information
S1 Data.
(XLSX)
Acknowledgments
Thanks to Xavier Paoletti, Aurelien Latouche and Cyrille Huchon for their expertise and advice.
PLOS ONE | https://doi.org/10.1371/journal.pone.0251372 May 12, 2021
9 / 11
PLOS ONE Optimal infertility management according to the Endometriosis Fertility Index
Author Contributions
Conceptualization: Arnaud Fauconnier.
Data curation: Joseph Du Cheyron.
Formal analysis: Alexandre Bailleul.
Investigation: Alexandre Bailleul.
Methodology: Arnaud Fauconnier.
Project administration: Julien Niro, Pierre Panel.
Supervision: Arnaud Fauconnier.
Validation: Arnaud Fauconnier.
Writing – original draft: Alexandre Bailleul.
Writing – review & editing: Arnaud Fauconnier.
References
1.
Fauconnier A, Chapron C. Endometriosis and pelvic pain: epidemiological evidence of the relationship
and implications. Human Reproduction Update. 1 nov 2005; 11(6):595–606. https://doi.org/10.1093/
humupd/dmi029 PMID: 16172113
2. Chapron C, Marcellin L, Borghese B, Santulli P. Rethinking mechanisms, diagnosis and management
of endometriosis. Nat Rev Endocrinol. nov 2019; 15(11):666–82. https://doi.org/10.1038/s41574-019-
0245-z PMID: 31488888
3.
Laganà AS, Garzon S, Go¨ tte M, Viganò P, Franchi M, Ghezzi F, et al. The Pathogenesis of Endometri-
osis: Molecular and Cell Biology Insights. Int J Mol Sci. 10 nov 2019; 20(22). https://doi.org/10.3390/
ijms20225615 PMID: 31717614
4. Brown J, Farquhar C. Endometriosis: an overview of Cochrane Reviews. Cochrane Gynaecology and
Fertility Group, e´ diteur. Cochrane Database of Systematic Reviews [Internet]. 10 mars 2014 [cite´ 18
mai 2020]; Disponible sur: http://doi.wiley.com/10.1002/14651858.CD009590.pub2
5. Barri PN, Coroleu B, Tur R, Barri-Soldevila PN, Rodrı´guez I. Endometriosis-associated infertility: sur-
gery and IVF, a comprehensive therapeutic approach. Reproductive BioMedicine Online. aouˆt 2010; 21
(2):179–85. https://doi.org/10.1016/j.rbmo.2010.04.026 PMID: 20541976
6. Garzon S, Laganà AS, Barra F, Casarin J, Cromi A, Raffaelli R, et al. Novel drug delivery methods for
improving efficacy of endometriosis treatments. Expert Opin Drug Deliv. 27 oct 2020;1–13. https://doi.
org/10.1080/17425247.2021.1829589 PMID: 32981374
7. Duffy JM, Arambage K, Correa FJ, Olive D, Farquhar C, Garry R, et al. Laparoscopic surgery for endo-
metriosis. In: The Cochrane Collaboration, e´diteur. Cochrane Database of Systematic Reviews [Inter-
net]. Chichester, UK: John Wiley & Sons, Ltd; 2014 [cite´ 18 mai 2020]. p. CD011031.pub2. Disponible
sur: http://doi.wiley.com/10.1002/14651858.CD011031.pub2
8. Rizk B, Turki R, Lotfy H, Ranganathan S, Zahed H, Freeman AR, et al. Surgery for endometriosis-asso-
ciated infertility: do we -exaggerate the magnitude of effect? 2015; 10.
9. Adamson GD, Pasta DJ. Endometriosis fertility index: the new, validated endometriosis staging system.
Fertility and Sterility. oct 2010; 94(5):1609–15. https://doi.org/10.1016/j.fertnstert.2009.09.035 PMID:
19931076
10. Vesali S, Razavi M, Rezaeinejad M, Maleki-Hajiagha A, Maroufizadeh S, Sepidarkish M. Endometriosis
Fertility Index for Predicting Non-assisted Reproductive Technology Pregnancy after Endometriosis
Surgery: A Systematic Review and Meta-analysis. BJOG: Int J Obstet Gy. 22 janv 2020;1471–
0528.16107. https://doi.org/10.1111/1471-0528.16107 PMID: 31967727
11.
12.
Johnson NP, Hummelshoj L, Adamson GD, Keckstein J, Taylor HS, Abrao MS, et al. World Endometri-
osis Society consensus on the classification of endometriosis. Hum Reprod. fe´vr 2017; 32(2):315–24.
https://doi.org/10.1093/humrep/dew293 PMID: 27920089
Tomassetti C, D’Hooghe T. Endometriosis and infertility: Insights into the causal link and management
strategies. Best Practice & Research Clinical Obstetrics & Gynaecology. aouˆ t 2018; 51:25–33. https://
doi.org/10.1016/j.bpobgyn.2018.06.002 PMID: 30245115
PLOS ONE | https://doi.org/10.1371/journal.pone.0251372 May 12, 2021
10 / 11
PLOS ONE Optimal infertility management according to the Endometriosis Fertility Index
13. Benoit L, Boujenah J, Poncelet C, Grynberg M, Carbillon L, Nyangoh Timoh K, et al. Predicting the likeli-
hood of a live birth for women with endometriosis-related infertility. European Journal of Obstetrics &
Gynecology and Reproductive Biology. nov 2019; 242:56–62. https://doi.org/10.1016/j.ejogrb.2019.09.
011 PMID: 31563819
14.
15.
Li X, Zeng C, Zhou Y-F, Yang H-X, Shang J, Zhu S-N, et al. Endometriosis Fertility Index for Predicting
Pregnancy after Endometriosis Surgery. Chin Med J. 2017; 130(16):1932. https://doi.org/10.4103/
0366-6999.211892 PMID: 28776545
Zhou Y, Lin L, Chen Z, Wang Y, Chen C, Li E, et al. Fertility performance and the predictive value of the
endometriosis fertility index staging system in women with recurrent endometriosis: A retrospective
study. Medicine. sept 2019; 98(39):e16965. https://doi.org/10.1097/MD.0000000000016965 PMID:
31574795
16. Oppenheimer A, Panel P, Rouquette A, du Cheyron J, Deffieux X, Fauconnier A. Validation of the Sex-
ual Activity Questionnaire in women with endometriosis. Human Reproduction. 1 mai 2019; 34(5):824–
33. https://doi.org/10.1093/humrep/dez037 PMID: 30989214
17.
18.
Tomassetti C, Bafort C, Meuleman C, Welkenhuysen M, Fieuws S, D’Hooghe T. Reproducibility of the
Endometriosis Fertility Index: a prospective inter-/intra-rater agreement study. BJOG: Int J Obstet Gy.
janv 2020; 127(1):107–14. https://doi.org/10.1111/1471-0528.15880 PMID: 31319445
Latouche A, Allignol A, Beyersmann J, Labopin M, Fine JP. A competing risks analysis should report
results on all cause-specific hazards and cumulative incidence functions. Journal of Clinical Epidemiol-
ogy. juin 2013; 66(6):648–53. https://doi.org/10.1016/j.jclinepi.2012.09.017 PMID: 23415868
19. Gray RJ. A Class of $K$-Sample Tests for Comparing the Cumulative Incidence of a Competing Risk.
Ann Statist. sept 1988; 16(3):1141–54.
20.
21.
Fine JP, Gray RJ. A Proportional Hazards Model for the Subdistribution of a Competing Risk. Journal of
the American Statistical Association. juin 1999; 94(446):496–509.
Latouche A, Andersen PK, Rey G, Moreno-Betancur M. A Note on the Measurement of Socioeconomic
Inequalities in Life Years Lost by Cause of Death. Epidemiology [Internet]. 11 juin 2019 [cite´ 5 juin
2020];Publish Ahead of Print. Disponible sur: https://journals.lww.com/00001648-900000000-98537
https://doi.org/10.1097/EDE.0000000000001022 PMID: 31162283
22. Cramer DW, Walker AM, Schiff I. Statistical methods in evaluating the outcome of infertility therapy**-
Supported in part by National Cancer Institute, National Institutes of Health Trainee Grant 2T32 CA
09001–04 (Department of Epidemiology, Harvard University School of Public Health). Fertility and Ste-
rility. juill 1979; 32(1):80–6. PMID: 456633
23. Spira A. Epidemiology of human reproduction. 1986; 5. https://doi.org/10.1093/oxfordjournals.humrep.
a136353 PMID: 3549765
24. Weinberg CR, Wilcox AJ, Baird DD. REDUCED FECUNDABILITY IN WOMEN WITH PRENATAL
EXPOSURE TO CIGARETTE SMOKING. 2018; 7.
25.
Zhang X, Liu D, Huang W, Wang Q, Feng X, Tan J. Prediction of Endometriosis Fertility Index in
patients with endometriosis-associated infertility after laparoscopic treatment. Reproductive BioMedi-
cine Online. juill 2018; 37(1):53–9. https://doi.org/10.1016/j.rbmo.2018.03.012 PMID: 29628331
26. Adamson GD. Endometriosis Fertility Index: is it better than the present staging systems? Current Opin-
ion in Obstetrics and Gynecology. juin 2013; 25(3):186–92. https://doi.org/10.1097/GCO.
0b013e32836091da PMID: 23571831
27. Sylvie M, Rodolphe M, Sylvie B. Laparoscopic Surgery in Infertile Women with Minimal or Mild Endome-
triosis. The New England Journal of Medicine. 1997; 6.
28. Be´ rube´ S, Marcoux S, Langevin M, Maheux R. Fecundity of Infertile Women with Minimal or Mild Endo-
metriosis And Women with Unexplained Infertility. Fertility and Sterility. juin 1998; 69(6):1034–41.
https://doi.org/10.1016/s0015-0282(98)00081-8 PMID: 9627289
29. Baggio S, Pomini P, Zecchin A, Garzon S, Bonin C, Santi L, et al. Delivery and pregnancy outcome in
women with bowel resection for deep endometriosis: a retrospective cohort study. Gynecol Surg. nov
2015; 12(4):279–85.
PLOS ONE | https://doi.org/10.1371/journal.pone.0251372 May 12, 2021
11 / 11
PLOS ONE |
10.1371_journal.pone.0252389 | RESEARCH ARTICLE
Time to recovery from COVID-19 and its
predictors among patients admitted to
treatment center of Wollega University
Referral Hospital (WURH), Western Ethiopia:
Survival analysis of retrospective cohort study
Tadesse TolossaID
Atomssa1, Motuma Getachew1, Getahun FetensaID
1*, Bizuneh WakumaID
2, Dejene Seyoum GebreID
1, Emiru Merdassa
1
2, Diriba Ayala3, Ebisa TuriID
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: Tolossa T, Wakuma B, Seyoum Gebre D,
Merdassa Atomssa E, Getachew M, Fetensa G, et
al. (2021) Time to recovery from COVID-19 and its
predictors among patients admitted to treatment
center of Wollega University Referral Hospital
(WURH), Western Ethiopia: Survival analysis of
retrospective cohort study. PLoS ONE 16(6):
e0252389. https://doi.org/10.1371/journal.
pone.0252389
Editor: Chiara Lazzeri, Azienda Ospedaliero
Universitaria Careggi, ITALY
Received: February 11, 2021
Accepted: May 15, 2021
Published: June 10, 2021
Copyright: © 2021 Tolossa et al. This is an open
access article distributed under the terms of the
Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: All relevant data are
within the manuscript and its S1 File.
Funding: The author(s) received no specific
funding for this work.
Competing interests: The authors have declared
that no competing interests exist.
1 Department of Public Health, Institute of Health Science, Wollega University, Nekemte, Ethiopia,
2 Department of Nursing, Institute of Health Science, Wollega University, Nekemte, Ethiopia, 3 Department
of Midwifery, Institute of Health Science, Wollega University, Nekemte, Ethiopia
* yadanotolasa@gmail.com
Abstract
Introduction
Despite its alarming spread throughout the world, no effective drug and vaccine is discov-
ered for COVID-19 so far. According to WHO, the recovery time from COVID-19 was esti-
mated to be 2 weeks for patients with mild infection, and 3 to 6 weeks for those with serious
illnesses. A studies regarding the median recovery time and its predictors are limited glob-
ally and specifically in Ethiopia. Therefore, the aim of this study was to estimate the median
time to recovery from COVID-19 and its predictors among COVID-19 cases admitted to
WURH, Western Ethiopian.
Methods
This was a hospital-based retrospective cohort study conducted among 263 adult patients
admitted with COVID-19 in WURH treatment center from March 29, 2020 through Septem-
ber 30, 2020. Epidata version 3.2 was used for data entry, and STATA version 14 for analy-
sis. A Cox proportional hazard regression model was fitted to determine factors associated
with recovery time. A variable with P-value � 0.25 at bivariable Cox regression analysis
were selected for multivariable Cox proportional model. Multivariable Cox regression model
with 95% CI and Adjusted Hazard Ratio (AHR) was used to identify a significant predictor of
time to recovery from COVID-19 at P-value < 0.05.
Results
The mean age of patient was 36.8 (SD± 10.68) years. At the end of follow up, two hundred
twenty seven observations were developed an event (recovered) with median time to recov-
ery of 18 days with IQR of 10–27 days. The overall incidence rate of recovery was of 4.38
PLOS ONE | https://doi.org/10.1371/journal.pone.0252389 June 10, 2021
1 / 12
PLOS ONE Time to recovery from COVID-19 and its predictors among patients with COVID-19
per 100 (95% CI: 3.84, 4.99) person-days observations. Being older age (AHR = 1.59, 95%
CI: 1.02, 2.49), presence of fever on admission (AHR = 1.78, 95% CI: 1.21, 2.62), and
comorbidity (AHR = 0.56, 95% CI, 0.34, 0.90) were found to have statistically significant
association with recovery time.
Conclusion and recommendations
In general, the median recovery time of patients with COVID-19 cases was long, and factors
such as older age group, presence of fever, and comorbidity was an independent predictors
of delayed recovery from COVID-19. Intervention to further reduce recovery time at treat-
ment center has to focus on patients those shows symptoms and with comorbidities.
Introduction
Coronaviruses are a large group of viruses, some cause illness to human and some occur in
animals. Rarely, animal coronaviruses can evolve and infect people and then may spread
between people. Human coronaviruses cause routine seasonal respiratory virus infections.
Other coronaviruses, like severe acute respiratory syndrome (SARS) and Middle East repara-
tory syndrome (MERS), can cause serious illness [1]. The emerging and rapidly evolving virus,
the Novel Coronavirus Disease-2019 (SARS-Cov-2), was detected in Wuhan China in Decem-
ber 2019, and was initially related to exposure at a seafood and live animal market [2]. It
spreads throughout countries and was categorized as pandemics by World Health Organiza-
tion (WHO) in March 2020. Coronaviruses are respiratory viruses and most commonly spread
through respiratory secretion of an infected person in close proximity [3].
The epidemiological dynamics of COVID-19 has changed dramatically over the courses of
months. At the time of writing this manuscript COVID-19 has infected more than 149,359,118
people globally, and about 3,149,381 people died out, and 127,040,432 cases were recovered
from COVID-19. In many African countries, the number of cases and number of death are
low compared to European and American countries, this could be due to low test capacity,
underreporting, and young population [4].
Knowing the COVID-19 fatality rate help us understand severity of the disease, identify
risk factors and assess the quality of healthcare. There are difference in mortality rate among
different groups due to age difference and comorbidity [5]. And also the COVID-19 fatality
rate vary across different locations may be due to populations age structure and case-mix of
infected and deceased patients [6].
The median time to recovery from COVID-19 varies among patients and settings, in which
the average recovery time from COVID-19 more than 14 days for some countries and less
than 14 days for others [7]. According to WHO, the recovery time is estimated to be 2 weeks
for patients with mild infection and 3 to 6 weeks for those with serious illnesses [4]. On the
other hand, CDC estimated people with mild to moderate spectrum of symptom and maintain
home isolation have a resolution of 3 days after the fever decreased, and there was substantial
improvement in respiratory symptoms, even without use of medication. A study conducted in
treatment center found in Ethiopia showed that, the rate of prognosis from COVID-19 for
asymptomatic cases was higher when compared to symptomatic COVID-19 cases, and this
study reported the average recovery time of 16 days [8]. Another study conducted in Kotebe
treatment center of Ethiopia revealed the media time to recovery from COVID-19 was 19 days
and it ranges from 2–71 days [9].
PLOS ONE | https://doi.org/10.1371/journal.pone.0252389 June 10, 2021
2 / 12
PLOS ONE Time to recovery from COVID-19 and its predictors among patients with COVID-19
A comprehensive literature review showed that age greater than 65 years, being hyperten-
sive (up to 40% of patients), diabetics, obesity, cardiovascular and lung disease are possible risk
factors for delayed recovery from COVID-19 [10]. In addition, presence of clinical manifesta-
tion on admission also associated with delayed recovery from COVID-19 [8].
Some studies showed improved survival of covid-19 patients depends on quality of health
care services like, patient management, timing of admission, understanding of disease progress
and expand use of steroids [11]. Though some scientific researches have been published, many
aspects of covid-19 still need more detailed valid and reliable information. This retrospective
cohort study aimed at determining time to recovery of Covid-19 infected people and its pre-
dictors among Covid-19 patients admitted to treatment center of Wollega University Referral
Hospital, Western Ethiopia.
Methods
Study area and study period
The study was conducted at wollega University referral hospital COVID-19 treatment center
which is found in Western Ethiopia. The study period was from March 29, 2020 (the first day
when the first case was admitted to this treatment center) to September 30, 2020. The last date
of study period (September 30, 2020) was used as an end date because this date was the last
date when only severe cases of COVID-19 cases was started admitted treatment center. The
data were retrieved between October 30, 2020 and November 15, 2020.
Study design. An institution based retrospective cohort study design was employed.
Study populations. Patients who were tested positive for COVID-19 by using rRT-PCR
test and admitted to WURH treatment center from March 29, 2020 to September 30, 2020
with a definite outcome (event or censored) and whose chart is available during the data col-
lection period. Patients with incomplete outcome variable and important baseline information
such as date admission and outcome occurred were excluded from the analysis.
Sample size and sampling techniques
All Covid-19 patients admitted to the treatment centers during study period (March 29, 2020
to September 30, 2020) and fulfill inclusion criteria was included in this study. A total of 298
COVID-19 cases were admitted to WURH treatment center.
Study variables
Dependent variable. The dependent variable of this study was time to recovery from
COVID-19. The time was estimated in days and it is the time when the patient was diagnosed
positive for COVID 19 by using rRT-PCR test to the patient was diagnosed negative for
COVID 19 and discharged from hospital.
Independent variables. Socio demographic variables such as age, sex, marital status, resi-
dence, contact history; Diseases related variables like type of medical illness, severity, time of
diagnosis, co-morbidity, types of comorbidity; Clinical and laboratory variables such as pres-
ence symptoms on admission were used as an independent variables.
Operational definitions. Survival time is the time in days from the patient was diagnosed
positive for COVID-19 by using rRT-PCR test to the occurrence of the outcome (event/cen-
sored). Event was recovery from COVID-19 or when the patient diagnosed negative after
admission to treatment center by rRT-PCR test. Censored was those patients who were not
developed an event or not recovered from COVID-19 (death, referred to other HI, on treat-
ment when the study was completed). Death is death of patients from COVID-19 while they
PLOS ONE | https://doi.org/10.1371/journal.pone.0252389 June 10, 2021
3 / 12
PLOS ONE Time to recovery from COVID-19 and its predictors among patients with COVID-19
were in the treatment center, and death recorded on card was confirmed by physician. Comor-
bidity (Yes/No) was co-existence of one or more diseases with Covid-19 cases “Yes” and, if not
it was considered as “No”.
Data collection tools and procedure
Data were collected from registration logbook, COVID-19 intake forms and medical cards of
patients. The data extraction tool was prepared from COVID-19 patient medical cards and
log-book that is currently used by the COVID-19 treatment center of the hospital. The check-
list consists of socio demographic related variables, diseases and past medical related variables,
clinical and laboratory variables. Trained health professionals who have been working in the
treatment center was extracted the data. During data collection time, the outcome was con-
firmed by reviewing the chart which was recorded by physician.
Data management and analysis
Epidata version 3.2 was used for data entry, and then the data was exported to STATA version
14 for further analysis. Before analysis, data was cleaned, edited by using simple frequencies
and cross tabulation; re-categorization of categorical variables and categorization of continu-
ous variables was done to be suitable for analysis. Descriptive non-parametric survival analysis
such as Kaplan Meier survival curve was used for the estimation survival probability. Days
were used as time scale to calculate median time to recovery. Log rank test was used to test any
difference in survival probability in categorical covariates.
A cox proportional hazards regression model was used to determine factors associated with
recovery time. Factors associated with recovery time at p-value < 0.25 in bivariable cox regres-
sion were selected for multivariable cox regression analysis. Adjusted Hazard Ratios (AHR)
with 95% confidence intervals was computed and statistical significance was declared when it
is significant at 5% level (p value < 0.05). To assess model adequacy for proportional hazard
model, proportional hazard assumption was checked by log-log plot and global test, and over-
all model adequacy of proportional hazard model was assessed by using cox snell residual
graph.
Ethical consideration
Ethical clearance was obtained from Wollega University research review board. Formal letter
of cooperation was written to WURH treatment center and permission was obtained from the
hospital administration. Personal identifiers were not used on data collection checklist.
Results
Description of study participants
From March 29/ 2020 through September 30/ 2020, a total of 298 patients with COVID-19
were admitted to Wollega University referral hospital treatment center. Of the total patient
cards, 35 patient cards were excluded from analysis due to unregistered outcome (event, cen-
sored, date of admission, date discharge and other baseline data incomplete). Finally, 263
patient cards with complete data were included in final analysis.
Socio-demographic characteristics of patients
The mean age of patient was 36.8 (SD± 10.68) years. The majority (46.4%) of patient’s age
were �25 years. More than half, (57%) of patients were resides in Nekemte city, and around
PLOS ONE | https://doi.org/10.1371/journal.pone.0252389 June 10, 2021
4 / 12
PLOS ONE Time to recovery from COVID-19 and its predictors among patients with COVID-19
Table 1. Socio-demographic characteristics of COVID-19 cases admitted to Wollega University Referral Hospital, 2020.
Variables
Category
Survival status
Total
Age
�25
25–40
�41
Residence
Out of Nekemte
Sex
Contact history
Nekemte
Male
Female
Yes
No
Unknown
https://doi.org/10.1371/journal.pone.0252389.t001
Recovered
No
103
96
28
87
140
155
72
55
84
88
Censored
No
19
13
4
26
10
7
29
4
5
27
No (%)
122 (46.4)
109 (41.4)
32 (12.2)
113 (43.0)
150 (57.0)
162 (61.6)
101 (38.4)
59 (22.4)
89 (33.8)
115 (43.7)
62% of patients were male. More than one third of patients (33.8%) had no contact history and
22.4% of patients had known contact history (Table 1).
Baseline clinical characteristic of patients
Of the total participants, 181 (68.8%) of cases had used oxygen supplementation. Two hundred
(76%) of participants had showed high fever on admission to the treatment center. One hun-
dred 178 (77.7%) and one hundred seventy eight (67.7%) of cases had showed dry cough and
throat pain on admission to treatment, respectively. More than half (53.3) had not clinically
presented with severe headache on entry to hospital (Table 2).
Table 2. Baseline clinical features of COVID-19 cases admitted to Wollega University Referral Hospital, 2020.
Variables
Category
Recovered
Oxygen supplemented
High fever
Dry cough
Sneezing
Throat pain
Severe headache
Difficulty in breathing
Diarrhea
Co morbidity
Yes
No
Yes
No
Yes
No
Yes
No
Yes
No
Yes
No
Yes
No
Yes
No
Yes
No
https://doi.org/10.1371/journal.pone.0252389.t002
No
56
171
176
66
151
76
140
87
155
72
85
142
63
164
42
185
21
206
Survival status
Censored
No
26
10
33
3
27
9
22
14
23
13
30
6
10
26
9
27
22
14
Total
No (%)
82 (31.2)
181 (68.8)
200 (76.0)
63 (24.0)
178 (67.7)
85 (32.3)
162 (61.6)
101 (33.4)
178 (67.7)
85 (32.3)
115 (43.7)
148 (53.3)
73 (27.8)
190 (72.2)
43 (19.4)
220 (86.6)
49 (18.6)
214 (81.4)
PLOS ONE | https://doi.org/10.1371/journal.pone.0252389 June 10, 2021
5 / 12
PLOS ONE Time to recovery from COVID-19 and its predictors among patients with COVID-19
Fig 1. Treatment outcome among patients admitted with COVID-19 in WURH treatment.
https://doi.org/10.1371/journal.pone.0252389.g001
Treatment outcome among patients admitted with COVID-19 in WURH
treatment center
Thirty four (12.9%) observation was censored at the end of the follow-up time. Seventeen (6.5%)
were on treatment when the study was completed, 1.5% was referred for further treatment, and
5.7 were died while they were on the treatment. At the end of follow up, while 229 (87.1%) of the
patients in the cohort were recovered from COVID 19 and recorded as an event (Fig 1).
Recovery rate and median recovery time from COVID-19
A total of 263 patients were followed for a median time of 18 days. Two hundred twenty seven
observations were developed an event (recovered) with median time to recovery of 18 days
with IQR of 10–27 days. During follow-up time, a total of 5177 person-day risks were observed
with a minimum and maximum follow-up time of 5 and 50 days, respectively. The overall inci-
dence rate of recovery was of 4.38 per 100 (95% CI: 3.84, 4.99) person-days observations.
A Kaplan-Meier estimation technique was used to see the estimate of survival time. The
overall graph of Kaplan-Meier survivor function depicted that the graphs decrease rapidly dur-
ing the first 30 days showing most patients recovered from COVID-19 during this time (Fig 2).
A separate Kaplan-Meier survivor functions curve was constructed to estimates the survival
time based on different covariates to see the existence of difference in recovery rate between
categories of individual covariates. From Kaplan Meier survival curve of individual covariates,
there were no difference in recovery rate of among male and female, and being resides in
Nekemte city and out of Nekemte city (Fig 3A). However, there was a difference in survival
probability/recovery rate for the covariates comorbidity and presence of fever on admission
(Fig 3B). To show the significance of survival difference, log rank test was computed at 5% sig-
nificance level. Accordingly, there was significance difference in survival status patients in rela-
tion to comorbidity, and presence of fever.
PLOS ONE | https://doi.org/10.1371/journal.pone.0252389 June 10, 2021
6 / 12
PLOS ONE Time to recovery from COVID-19 and its predictors among patients with COVID-19
Fig 2. Overall Kaplan-Meier survival probability curve.
https://doi.org/10.1371/journal.pone.0252389.g002
Predictors of recovery time from COVID-19
Covariates that had P- value � 0.25 in bivaribale cox regression analysis were selected for
multivariable cox regression analysis. Residence, sex, contact history, comorbidity, presence
of fever and presence of severe headache were selected for multivariable cox regression at
P-value � 0.25. Finally, three of the predictors (Age of patients, fever and comorbidity) were
found to have statistically significant association with recovery time during multivariable cox
proportional regression analysis.
Age of the patients was one of the variables that predicts recovery rate. Recovery rate of
younger age groups (� 24) years was 1.59 times higher as compared to patients who were
aged �41years (AHR = 1.59, 95% CI: 1.02, 2.49). Presence clinical symptom such as fever
was one of the predictors that affect the recovery rate of patients from COVID-19. Patients
who were not detected with fever were at higher rate of recovery than patients who were
showed fever on admission (AHR = 1.78, 95% CI: 1.21, 2.62). Finally, presence of any types
of comorbidity was another factor that determines the recovery time of patients with
COVID-19. Presence of comorbidity lower the rate of recovery by 44% as compared to
those patients who had not admitted with comorbidity (AHR = 0.56, 95% CI, 0.34, 0.90)
(Table 3).
Model goodness-of-fit
After fitting multivariable Cox Proportional Hazard Model, adequacy of a fitted model was
assessed by using cox Snell residuals. Finally, the graph of Nelson-Aalen cumulative hazard
function and the cox Snell residuals variable were compared to the hazard function to the diag-
onal line. The hazard function follows the 45-degree line, which approximately, indicated that
the model fitted the data well (Fig 4).
PLOS ONE | https://doi.org/10.1371/journal.pone.0252389 June 10, 2021
7 / 12
PLOS ONE Time to recovery from COVID-19 and its predictors among patients with COVID-19
Fig 3. (a) Kaplan survival curve for residence and sex. (b) Kaplan survival curve for the presence fever and comorbidity.
https://doi.org/10.1371/journal.pone.0252389.g003
PLOS ONE | https://doi.org/10.1371/journal.pone.0252389 June 10, 2021
8 / 12
PLOS ONE Time to recovery from COVID-19 and its predictors among patients with COVID-19
Table 3. Multivariable Cox regression analysis of median recovery time and its predictors among patients admitted with COVID-19 cases in WURH, 2020.
Variables
Residence
Age
Contact history
Oxygen supplemented
Presence of fever
Presence of headache
Presence of comorbidity
Category
Survival status
Recovered
Not recovered
Rural
Urban
�24
25–40
�41
Yes
No
Unknown
Yes
No
Yes
No
Yes
No
Yes
No
87
140
103
96
28
55
84
88
56
171
176
66
85
142
21
206
26
10
19
13
4
4
5
27
26
10
33
3
30
6
22
14
CHR
1
AHR
1
1.22 (0.93, 1.59)
1.45 (0.95, 2.22)
1.08 (0.82, 1.43)
1.16 (0.88 1.53)
1.59 (1.02, 2.49)
1.17 (0.88, 1.55)
1
1
1
1
1.29 (0.91, 1.83)
1.04 (0.74, 1.47)
1
1.10 (0.76, 1.58)
1.07 (0.75, 1.51)
1
P-value
0.283
0.039�
0.276
0.592
0.692
2.1 (0.72, 3.12)
1.70 (0.92, 2.23)
0.076
1
1
1.29 (0.95, 1.73)
1.78 (1.21, 2.62)
0.003 �
1
1.19 (0.90, 1.56)
0.47 (0.30, 0.74)
1
1
1.06 (0.80, 1.40)
0.56 (0.34, 0.90)
1
0.674
0.017�
AHR: Adjusted Hazard Ratio; CHR: Crude Hazard Ratio
�statistically significant at p<0.05.
WURH: Wollega University Referral Hospital.
https://doi.org/10.1371/journal.pone.0252389.t003
Fig 4. Cox Snell residual test for overall adequacy of the model.
https://doi.org/10.1371/journal.pone.0252389.g004
PLOS ONE | https://doi.org/10.1371/journal.pone.0252389 June 10, 2021
9 / 12
PLOS ONE Time to recovery from COVID-19 and its predictors among patients with COVID-19
Discussion
This study was aimed to determine time to recovery from novel coronavirus disease (SARS
COV-2) and its predictors among patients admitted to WURH with COVID-19 cases. This
study pointed out that the median time to recovery from SARS COV-2 was 18 days. This is
similar to the studies done in Eka Kotebe General Hospital, Ethiopia in which viral clearance
lasted for 19 days [9] and 16 days [8]. This might be due to relative similarity in care and treat-
ment given for the patients in both study areas. Moreover, it is also is consistent with the previ-
ous study findings from Israel (20–21) days [12]. However, the median recovery time was
lower in the previous studies done in Guangzhou Eighth People’s Hospital, China (12 days)
[13], University of California San Diego Health (7 days) [14], Zhejiang University and the
Shenzhen Third People’s Hospital, China (15 days) [15], and in Singapore (12 days) [16]. The
possible reason for the observed discrepancy between the studies might be due to variation in
sample size, study setting, socioeconomic characteristics, and the severity of the disease. Evi-
dences are showing the severe the disease condition, the longer the duration of viral RNA
clearance [17].
The present study found older age as independent predictor of delayed recovery time from
coronavirus disease. This is consistent with previous study findings from Guangzhou, China
[13], Korea [18], Wuhan Pulmonary Hospital, China [19], Shenzhen, China [15], three hospi-
tals in Wuhan, China [17], and Qingdao, China [20]. This might be attributed to older age-
related severity progression of COVID-19 cases which in turn leads to either death or delayed
duration of viral clearance in elderly patients [21]. Moreover, it might be due to the fact that
older age is not without comorbid conditions which are among the major risk factor of lower
recovery rate form coronavirus disease and even death related to COVID-19. Besides, older
age is associated with degeneration of pulmonary function and compromised immunity that
contributes for severe COVID-19 cases and poor clinical outcomes.
The current study has also demonstrated that patients with comorbid condition had 44%
lower odds of recovery rate from coronavirus disease compared to their counterparts. Simi-
larly, existing evidences are supporting the present study finding, for instance, the study done
in Italy [22], Fairfield General Hospital, Bury, UK [23], Wuhan Pulmonary Hospital, China
[19], Jin Yin-tan Hospital and Tongji Hospital [24], and Wuhan, China [15] claim comorbid
conditions majorly cardiovascular diseases attributed to the delayed duration of recovery from
SARS COV-2 cases. Furthermore, our study has also identified absence of fever as a good prog-
nostic factor of COVID-19 cases. This is in line with study conducted in Eka Kotebe treatment
center of Ethiopia, in which presence of clinical manifestation on admission prolong the time
of recovery from COVID-19 [8]. This is also supported by the study finding from Changsha,
China [25]. This could be due to the fact that the function of respiratory system is dependent
on body temperature variations [26]. This can be explained that an increment in body temper-
ature results in increment in respiratory rate which increases the pulmonary work load eventu-
ally leading pulmonary insufficiency and lower recovery rate [26].
Limitations
The study employed advanced statistical model for analysis. However, as the unregistered out-
come and incomplete baseline data were excluded from the analysis, the reviewed records
might lack very important variables that could influence recovery rate from coronavirus dis-
ease. Besides, the subjects made self-report of previously diagnosed medical illness which was
considered as comorbidity.
PLOS ONE | https://doi.org/10.1371/journal.pone.0252389 June 10, 2021
10 / 12
PLOS ONE Time to recovery from COVID-19 and its predictors among patients with COVID-19
Conclusion
In general, this study found the prolonged recovery time from coronavirus disease. The study
revealed that older age, fever at admission, and having at least one comorbid condition as a
poor prognostic factors of novel coronavirus disease. Thus, elders and individuals with comor-
bidity has to get due attention to prevent infection by the virus. Moreover, elders and patients
with comorbidity should get priority in management of coronavirus disease in order to
enhance good clinical outcome.
Supporting information
S1 File. Dataset.
(DTA)
Author Contributions
Conceptualization: Tadesse Tolossa, Dejene Seyoum Gebre, Emiru Merdassa Atomssa, Geta-
hun Fetensa.
Formal analysis: Tadesse Tolossa, Emiru Merdassa Atomssa.
Funding acquisition: Bizuneh Wakuma, Dejene Seyoum Gebre.
Investigation: Dejene Seyoum Gebre, Motuma Getachew, Getahun Fetensa, Diriba Ayala.
Methodology: Tadesse Tolossa, Bizuneh Wakuma.
Project administration: Motuma Getachew, Ebisa Turi.
Resources: Tadesse Tolossa, Bizuneh Wakuma, Diriba Ayala.
Software: Tadesse Tolossa, Emiru Merdassa Atomssa, Motuma Getachew.
Supervision: Emiru Merdassa Atomssa, Getahun Fetensa, Ebisa Turi.
Visualization: Bizuneh Wakuma, Dejene Seyoum Gebre, Motuma Getachew, Diriba Ayala,
Ebisa Turi.
Writing – original draft: Tadesse Tolossa, Bizuneh Wakuma.
Writing – review & editing: Tadesse Tolossa, Emiru Merdassa Atomssa.
References
1. Yin Y. and Wunderink R.G., MERS, SARS and other coronaviruses as causes of pneumonia. Respirol-
ogy, 2018. 23(2): p. 130–137. https://doi.org/10.1111/resp.13196 PMID: 29052924
2. Huang C., et al., Clinical features of patients infected with 2019 novel coronavirus in Wuhan, China. The
lancet, 2020. 395(10223): p. 497–506.
3. Pamuk S., O¨ zkan A., and Polat B., Epidemiology, pathogenesis, diagnosis and management of
COVID-19. The Turkish Journal of Ear Nose and Throat, 2020. 30(Supp: 1): p. 1–9.
4.
Jin Y.-H., et al., A rapid advice guideline for the diagnosis and treatment of 2019 novel coronavirus
(2019-nCoV) infected pneumonia (standard version). Military Medical Research, 2020. 7(1): p. 1–23.
https://doi.org/10.1186/s40779-019-0229-2 PMID: 31928528
5. Organization, W.H., World Health Organization coronavirus disease (COVID-19) dashboard. 2020.
6.
Ioannidis J., The infection fatality rate of COVID-19 inferred from seroprevalence data. MedRxiv, 2020.
https://doi.org/10.2471/BLT.20.265892 PMID: 33716331
7. Bhapkar H.R., et al., Revisited COVID-19 Mortality and Recovery Rates: Are we Missing Recovery
Time Period? J Med Syst, 2020. 44(12): p. 202. https://doi.org/10.1007/s10916-020-01668-6 PMID:
33099706
PLOS ONE | https://doi.org/10.1371/journal.pone.0252389 June 10, 2021
11 / 12
PLOS ONE Time to recovery from COVID-19 and its predictors among patients with COVID-19
8.
Leulseged T.W., et al., Determinants of Time to Convalescence among COVID-19 Patients at Millen-
nium COVID-19 Care Center in Ethiopia: A prospective cohort study. medRxiv, 2020.
9. Abrahim S.A., et al., Time to recovery and its predictors among adults hospitalized with COVID-19: A
prospective cohort study in Ethiopia. PloS one, 2020. 15(12): p. e0244269. https://doi.org/10.1371/
journal.pone.0244269 PMID: 33378367
10. Ortiz-Prado E., et al., Clinical, molecular and epidemiological characterization of the SARS-CoV2 virus
and the Coronavirus disease 2019 (COVID-19), a comprehensive literature review. Diagnostic microbi-
ology and infectious disease, 2020: p. 115094. https://doi.org/10.1016/j.diagmicrobio.2020.115094
PMID: 32623267
11. Xie J., et al., Metabolic syndrome and COVID-19 mortality among adult black patients in New Orleans.
Diabetes care, 2021. 44(1): p. 188–193.
12. Voinsky I., Baristaite G., and Gurwitz D., Effects of age and sex on recovery from COVID-19: Analysis
of 5,769 Israeli patients. Journal of Infection, 2020.
13. Chen X., et al., Associations of clinical characteristics and treatment regimens with the duration of viral
RNA shedding in patients with COVID-19. International Journal of Infectious Diseases, 2020. 98: p.
252–260. https://doi.org/10.1016/j.ijid.2020.06.091 PMID: 32619760
14. Daniels L.B., et al., Relation of statin use prior to admission to severity and recovery among COVID-19
inpatients. The American journal of cardiology, 2020. 136: p. 149–155. https://doi.org/10.1016/j.
amjcard.2020.09.012 PMID: 32946859
15. Xu K., et al., Factors associated with prolonged viral RNA shedding in patients with COVID-19. Clinical
Infectious Diseases, 2020.
16. Young B.E., et al., Epidemiologic features and clinical course of patients infected with SARS-CoV-2 in
Singapore. Jama, 2020. 323(15): p. 1488–1494. https://doi.org/10.1001/jama.2020.3204 PMID:
32125362
17. Xu J., et al., Clinical course and predictors of 60-day mortality in 239 critically ill patients with COVID-19:
a multicenter retrospective study from Wuhan, China. Critical Care, 2020. 24(1): p. 1–11. https://doi.
org/10.1186/s13054-019-2683-3 PMID: 31898531
18. Das A. and Gopalan S.S., Epidemiology of CoVID-19 and predictors of recovery in the Republic of
Korea. medRxiv, 2020. https://doi.org/10.1155/2020/7291698 PMID: 32774918
19. Du R.-H., et al., Predictors of mortality for patients with COVID-19 pneumonia caused by SARS-CoV-2:
a prospective cohort study. European Respiratory Journal, 2020. 55(5).
20. Hu X., et al., Factors associated with negative conversion of viral RNA in patients hospitalized with
COVID-19. Science of the Total Environment, 2020: p. 138812. https://doi.org/10.1016/j.scitotenv.
2020.138812 PMID: 32335406
21. Wu J., et al., Early antiviral treatment contributes to alleviate the severity and improve the prognosis of
patients with novel coronavirus disease (COVID-19). Journal of internal medicine, 2020.
22. Boari G.E., et al., Prognostic factors and predictors of outcome in patients with COVID-19 and related
pneumonia: a retrospective cohort study. Bioscience Reports, 2020. 40(12). https://doi.org/10.1042/
BSR20203455 PMID: 33201172
23. Chinnadurai R., et al., Older age and frailty are the chief predictors of mortality in COVID-19 patients
admitted to an acute medical unit in a secondary care setting-a cohort study. BMC geriatrics, 2020. 20
(1): p. 1–11.
24. Ruan Q., et al., Clinical predictors of mortality due to COVID-19 based on an analysis of data of 150
patients from Wuhan, China. Intensive care medicine, 2020. 46(5): p. 846–848. https://doi.org/10.
1007/s00134-020-05991-x PMID: 32125452
25. Qi L., et al., Factors associated with duration of viral shedding in adults with COVID-19 outside of
Wuhan, China: A retrospective cohort study. International Journal of Infectious Diseases, 2020. https://
doi.org/10.1016/j.ijid.2020.05.045 PMID: 32425636
26. Rubini A., The effect of body warming on respiratory mechanics in rats. Respiratory physiology & neuro-
biology, 2011. 175(2): p. 255–260. https://doi.org/10.1016/j.resp.2010.11.012 PMID: 21112417
PLOS ONE | https://doi.org/10.1371/journal.pone.0252389 June 10, 2021
12 / 12
PLOS ONE |
10.1371_journal.pone.0254310 | RESEARCH ARTICLE
Contextual and individual factors associated
with public dental services utilisation in Brazil:
A multilevel analysis
Maria Helena Rodrigues GalvãoID
Giuseppe Roncalli1
1*, Arthur de Almeida MedeirosID
1,2, Angelo
1 Postgraduate Program in Public Health, Federal University of Rio Grande do Norte, Natal, Rio Grande do
Norte, Brazil, 2 Integrated Health Institute, Federal University of Mato Grosso do Sul, Campo Grande, Mato
Grosso do Sul, Brazil
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
* mhrgalvao@gmail.com
Abstract
Background
OPEN ACCESS
Citation: Galvão MHR, Medeiros AdA, Roncalli AG
(2021) Contextual and individual factors associated
with public dental services utilisation in Brazil: A
multilevel analysis. PLoS ONE 16(7): e0254310.
https://doi.org/10.1371/journal.pone.0254310
Editor: Ratilal Lalloo, University of Queensland,
AUSTRALIA
Received: March 29, 2021
Accepted: June 23, 2021
Published: July 9, 2021
Peer Review History: PLOS recognizes the
benefits of transparency in the peer review
process; therefore, we enable the publication of
all of the content of peer review and author
responses alongside final, published articles. The
editorial history of this article is available here:
https://doi.org/10.1371/journal.pone.0254310
Copyright: © 2021 Galvão et al. This is an open
access article distributed under the terms of the
Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: All data are public
and available on the Brazilian Institute of
Geography and Statistics website (www.ibge.gov.
br).
This study verified the association between contextual and individual factors and public den-
tal services utilisation in Brazil.
Methods
The study was conducted based on a cross-sectional population-based household survey
performed in Brazil (National Health Survey– 2019)). Data was collected between August
2019 and March 2020. Total sample included 43,167 individuals aged �15 years who had
at least one dental appointment in the last 12 months before interview. Study outcome was
‘public dental service utilisation’, and Andersen’s behavioral model was adopted for select-
ing independent variables. A multilevel analysis was performed using individual factors as
first level and federation units as second level.
Results
The highest prevalence of public dental service utilisation on an individual level was
observed among unable to read or write people (PR: 3.31; p<0.001), indigenous (PR:
1.40; p<0.001), black or brown (PR: 1.16; p<0.001), with per capita household income
of up to U$124 (PR: 2.40; p<0.001), living in the rural area (PR: 1.28; p<0.001), and
who self-rated oral health as regular (PR: 1.15; p<0.001) or very bad/bad (PR: 1.26;
p<0.001). On the contextual level, highest PR of public dental service utilisation was
observed among those living in federal units with increased oral health coverage in pri-
mary health care.
Conclusions
Public dental service utilisation is associated with individual and contextual factors. These
results can guide decision-making based on evidence from policymakers, demonstrating
PLOS ONE | https://doi.org/10.1371/journal.pone.0254310 July 9, 2021
1 / 14
PLOS ONE Funding: This study was financed in part by the
Coordenac¸ão de Aperfeic¸oamento de Pessoal de
Nı´vel Superior – Brasil (CAPES) – Finance Code
001. The funding consisted of a postgraduate
studies scholarship to MHRG and payment of
publication fees. Furthermore, it did not interfere
with the study’s design and collection, analysis,
and interpretation of data and writing the
manuscript. There was no additional external
funding received for this study.
Competing interests: The authors have declared
that no competing interests exist.
Factors associated with public dental services utilisation in Brazil
the potential for mitigating oral health inequalities and increasing service coverage in a pub-
lic and universal health system.
Introduction
Brazil is a middle-income country with universal healthcare system covering dental assistance
for all citizens. In 2003, the Brazilian oral health care service was transformed by the National
Oral Health Policy implementation, expanding primary care teams and emphasizing the pri-
mary care-based model [1]. The last Brazilian oral health epidemiological survey, in demon-
strated significant oral health needs, especially in adolescents and adults. Mean values of
decayed, missing, and filled teeth (DMFT) index were 4.2 for adolescents, 16.7 for adults, and
27.5 for older adults. However, “decayed teeth” and “missing teeth” components sharply
reduced compared to previous year. In contrast, “filled teeth” component grew in relative
terms, indicating greater access to dental services for dental restorations [2].
Brazil expanded primary care teams in the public oral health sector, increasing population
coverage from 20.5% (2003) to 43.1% (2019). However, this expansion was not regular over
time. In the first period of policy implementation (2003–2011), a significant expansion
occurred in dental teams, from 6,170 (2003) to 23,076 (2011). A reduction occurred between
2015 and 2018, followed by expansion of 28,991 teams in 2019. Such oscillation was due to
political issues [3]. Furthermore, Brazil has a significant number of dentists (337,137 in Febru-
ary 2021) and has shown a considerable increase in the number of undergraduate courses in
Dentistry in recent years [4].
Although the number of dentists and oral health care teams expanded in the SUS, equity
in dental service access was not reached. For example, 21.6 million people have never had a
dental appointment until 2010 [3]. Furthermore, most dental appointments in Brazil are
paid by either out-of-pocket or private dental insurance plans, despite expansion of public
services, favoring inequalities in dental service utilisation [5]. Last year, dentist appoint-
ments were higher among those with more education, income, and private healthcare cov-
erage and living in the country’s wealthiest regions [5]. Other studies in Brazil revealed that
public dental services are more used by black people from low-income families, living in
small towns, with more than four household residents, and having more dental treatment
need [6,7].
Despite advances in oral health policies, literature lacks studies regarding the profile of den-
tal service users, especially assessing the effectiveness of strategies adopted to expand access to
population with great inequalities. Evaluating multiple determinants of dental service utilisa-
tion based on broader theoretical models and national scope is important to understand the
country’s reality. Therefore, understanding the profile of public dental services helps evaluate
public policy performance regarding equity in oral health. Although other studies [8,9] were
conducted with the same topic, this study presents new and recent contextual elements.
Andersen Behavioral Model comprises a conceptual framework for understanding multiple
dimensions of access to medical and health care outcomes and is valid to evaluate health ser-
vice utilization. The model presents individual and contextual determinants for health service
utilisation, evaluating predisposing, enabling, and need factors at each level [10]. Experts com-
monly use Andersen’s behavioral model to explain access to oral health care [11].
Thus, this study aimed to verify contextual and individual factors associated with public
dental service utilisation by Brazilians aged 15 years or older using concepts of the Andersen
behavior model [10].
PLOS ONE | https://doi.org/10.1371/journal.pone.0254310 July 9, 2021
2 / 14
PLOS ONE Factors associated with public dental services utilisation in Brazil
Materials and methods
Participants and database
Data were collected from the National Health Survey—2019 (PNS), a population-based house-
hold survey that assessed Brazilian determinants, conditions, and health needs. PNS provides a
representative database about the country and people living in private households, contribut-
ing to elaborate public health policies in Brazil and allowed territorial coverage using the Mas-
ter Sample of Integrated Household Research System (SIPD) [12,13].
A three-stage cluster sampling method was used: census tracts selection from primary sam-
pling units, household selection in each PSU,) and one resident aged 15 or older from each
household, randomly selected based on the list of residents obtained during the interview. A
total of 108,457 households were selected (100,541 were occupied), resulting in a database of
279,382 responses (94,114 home interviews).PNS 2019 data were collected between August
2019 and March 2020 [12,13].
The questionnaire was divided into three sections and conducted by trained interviewers
using a mobile device. Third section of the questionnaire included oral health with self-
reported information about last dental appointment, number of missing teeth, and oral health
assessment. This study sample included people aged 15 or older who were selected to answer
the survey questionnaire. Answers to the following question were considered: ‘When was the
last time you visited a dentist?’. Information about last dental appointment was obtained only
for the selected resident who had the last dental appointment up to three years before the inter-
view [12]. Thus, sample consisted of 43,167 individuals.
Characterization of variables
Dependent variable. The study outcome was ‘public dental service utilisation’. We con-
sidered only affirmative or negative responses to the question ‘Has dental consultation been
conducted in the Brazilian National Health System (SUS, from the Portuguese acronym)?’.
Independent variables. Andersen’s behavioral model [10] (Fig 1) was adopted to select
independent variables (Box 1).
Individual independent variables. Regarding individual predisposing factors, we consid-
ered sex (male or female), age (stratified into age groups), skin color/race (white, black, indige-
nous, or Asian), educational level (unlettered, incomplete elementary school, complete
elementary school, high school, or higher education), and per capita household income (up to
U$ 124, from U$125 to U$248, or U$249 or more). Individual facilitating factors were
Fig 1. Conceptual framework adapted by Andersen’s behavioral model.
https://doi.org/10.1371/journal.pone.0254310.g001
PLOS ONE | https://doi.org/10.1371/journal.pone.0254310 July 9, 2021
3 / 14
PLOS ONE Box 1. Description of individual and contextual variables of the study and adaptation strategies for the analysis model. Brazil, 2019
Factors associated with public dental services utilisation in Brazil
Variable
Age
Source
Reference Year
National Health Survey (PNS)
2019
Description
Age, in years, at the time of the interview
Sex
National Health Survey (PNS)
2019
Sex
Skin color/Race
National Health Survey (PNS)
2019
Self-reported skin color
Educational level
National Health Survey (PNS)
2019
Highest educational level reached
Per capita household
income
National Health Survey (PNS)
2019
Per capita household income, converted into dollars, (considering the
average values of December/2019)
Household area
National Health Survey (PNS)
2019
Place of residence
Enrolled in Primary
Health Care
National Health Survey (PNS)
2019
Information regarding household enrolled in a primary care facility.
Type of dental
attendance
National Health Survey (PNS)
2019
Reason for the last dental appointment
Self-rated oral health
National Health Survey (PNS)
2019
Self-rated oral health
Number of lost teeth
National Health Survey (PNS)
2019
PLOS ONE | https://doi.org/10.1371/journal.pone.0254310 July 9, 2021
Original Categorization
(Adapted Categorization)
Age categorized into groups.
15 to 19 years
20 to 39 years
40 to 59 years
60 years or older
Male
Female
White (White)
Black (Black or Brown)
Asian (Asian)
Brown (Black or Brown)
Indigenous (Indigenous)
Unable to read or write (Unable
to read or write)
Incomplete primary school
(Incomplete primary school)
Primary school (Primary school)
Incomplete High School (Primary
school)
High School (High School)
Undergraduate (High School)
Graduation (Higher education)
Continuous variable categorized
in:
U$ 249 and over
U$ 125 to U$ 248
Up to U$ 124
Urban
Rural
Yes
No
Do not know
Cleaning, prevention, or overhaul
(Preventive dental attendance)
Dental pain (Tooth extraction or
dental pain)
Tooth extraction (Tooth
extraction or dental pain)
Dental treatment (Dental
treatment)
Gum problem (Dental treatment)
Mouth wound treatment (Dental
treatment)
Dental implant (Dental
treatment)
Placement/maintenance of braces
on teeth (dental treatment)
Prosthesis or denture placement/
maintenance (Dental treatment)
Other treatments (Dental
treatment)
Very good (Very good or good)
Good (Very good or good)
Moderate (Moderate)
Bad (Bad or very bad)
Very bad (Bad or very bad)
�Continuous variable
(Continued )
4 / 14
PLOS ONE Factors associated with public dental services utilisation in Brazil
Box 1. (Continued)
Human Development
Index
(HDI)
Average per capita
income
Gini Index
Brazilian agency of the United
Nations Development Program
(UNDP)
2017
Brazilian agency of the United
Nations Development Program
(UNDP)
2017
Brazilian agency of the United
Nations Development Program
(UNDP)
2017
Oral Health Coverage in
Primary Health Care
Primary Care Management
and Information System
(E-Gestor)
2019
https://doi.org/10.1371/journal.pone.0254310.t001
Human Development Index refers to geometric mean of dimensions:
Income, Education, and Longevity, with equal weights.
�Continuous variable
Sum of income of all household members, divided by the number of
residents.
�Continuous variable
It measures degree of inequality in the distribution of individuals
according to per capita household income. Its value ranges from 0,
when there is no inequality (per capita household income of all
individuals has the same value), to 1, when inequality is maximum
(only one individual holds all income). The universe of individuals is
limited to those living in permanent private households.
Number of oral health teams in primary care services, divided by the
population in the same year.
�Continuous variable
�Continuous variable
household area (urban or rural) and enrolled in primary health care teams (yes, no, or do not
know). Type of dental attendance (preventive care, dental treatment, or tooth extraction/den-
tal pain), self-rated oral health (very good/good, regular, or very bad/bad), and number of lost
teeth were considered individual need factors. Individual independent variables were collected
from PNS questionnaire.
Context-independent variables. Predisposing contextual factors were Human Develop-
ment Index (HDI), Gini index, and average per capita income obtained from the Brazilian
branch of the United Nations Development Program, considering the latest information avail-
able. Enabling contextual factor was oral health coverage in primary health care (December
2019 as reference) obtained from the System of Information and Management of Primary
Care (Brazil, Ministry of Health). Contextual variables were collected considered all Brazilian
Federation Units (FU) (26 states) and the Federal District.
Statistical analysis
Individual and contextual variables were stored in two databases and merged using determin-
istic linkage technique [14], considering FU codification as reference variable.
All variables were analyzed concerning missing data and outliers. Skin color/race and aver-
age per capita household income presented 0.01% (n = 5) and 0.04% (n = 16) of missing data,
respectively. According to Hair et al., missing data of less than 10% can be ignored [15].
Population expansion was performed for descriptive analysis since this study has a complex
sample design. Expansion factors (or sample weights) were defined to analyze PNS data con-
sidering complex sampling design and distinct selection probabilities for selected households
and residents. Final weight applied was a product of the inverse of selection probability expres-
sions of each stage of sampling plan, including correction for non-responses and adjustments
to total populations8. Prevalence was calculated for individual and contextual variables. In
addition, univariate Poisson regression analysis with robust variance was performed to esti-
mate prevalence ratio (PR) and 95% confidence interval (95% CI). Variables presenting
p�0.20 were included in multilevel analysis model.
Multilevel modeling was chosen because contextual characteristics have a significant effect
on people [16]. Therefore, individual factors and FU were considered first and second levels,
respectively.
PLOS ONE | https://doi.org/10.1371/journal.pone.0254310 July 9, 2021
5 / 14
PLOS ONE Factors associated with public dental services utilisation in Brazil
Multilevel Poisson regression initiated with null model analysis to identify random effects.
Subsequently, modeling was performed with individual and contextual variables. To analyze
interaction between levels, an interaction term was created from the individual variable
‘Record in primary health care teams?’ and the contextual variable ‘oral health coverage in pri-
mary health care.’
Ethical statement
PNS 2019 met all requirements in research involving humans and was approved by the
National Research Ethics Committee (protocol n. 3,529,376). PNS data are public and available
on the Brazilian Institute of Geography and Statistics website (www.ibge.gov.br). Information
regarding contextual variables was collected from a secondary public database.
Results
Descriptive analysis
Regarding the study’s outcome, it was observed that only 23.1% (CI95% 22.3%; 23.9%) of peo-
ple used public oral health services. Regarding individual predisposing factors, most partici-
pants aged between 20 and 39 years (41.2%, 95%CI 40.3–42.1%), were women (56.6%, 95%CI
55.7–57.5%), had high school degree (38.0%, 95CI% 37.1–38.9%), were black or brown (50.9%,
95%CI 49.8–51.9%), and had household income per capita of up to $124 (16.8%, 95%CI 16.2–
17.5%).Average income met the criterion established by the federal government to register in
the national income transfer program for poor people. For individual facilitating factors,
59.0% (95%CI 57.6%; 60.3%) were enrolled in primary care teams and 88.8% (95%CI 88.8%;
89.8) resided in urban areas. Concerning need factors, 75.7% (95%CI 74.9%; 76.5%) rated oral
health as very good or good and 47.3% (95%CI 46.3%; 48.2%) performed preventive care.
Average number of missing teeth was 2.615 ± 0.045 teeth (Table 1).
Regarding predisposing contextual characteristics, mean HDI of FU was 0.777 ± 0.001,
Gini index was 0.523 ± 0.001, and average per capita income was U$372.219 ± 1.128. Average
oral health coverage in primary health was 51.540 ± 0.168 (Table 1).
Univariate analysis
Univariate analysis (Table 2) showed decreased public dental service utilisation according to
age and low prevalence among males (PR: 0.89, 95%CI 0.86–0.93). Educational level and aver-
age per capita household income showed a dose-response effect. Black or brown and indige-
nous were more likely to use public dental services (58% and 121%, respectively) than white
people. Lack of registration by primary health care teams reduced public dental service utilisa-
tion, whereas people living in rural areas were one-fold more likely to use dental services. Den-
tal service utilisation was associated with worse self-rated oral health and tooth extraction or
dental pain.
Regarding contextual factors, public dental service utilisation was more prevalent in FUs
with low HDI, low average income per capita, and high oral health coverage in primary health
(Table 2).
Multilevel analysis
In multilevel modeling, initial null model indicated a contextual effect on prevalence of public
oral health service utilisation. Variance analysis supports this situation since it was different
from zero (0.19—CI95% 0.11; 0.34) and likelihood ratio was significant (LR: 1631.00—
p�0.001) (Table 3).
PLOS ONE | https://doi.org/10.1371/journal.pone.0254310 July 9, 2021
6 / 14
PLOS ONE Factors associated with public dental services utilisation in Brazil
Table 1. Descriptive analysis of individual and contextual characteristics with population expansion. Brazil, 2019.
Variables
n
%
95%CI
Public dental service utilisation
Yes
No
Predisposing
Age
15–19 years
20–39 years
40–59 years
60 years or older
Sex
Female
Male
Educational level
Higher education
High School
Primary school
Incomplete primary school
Unable to read or write
Skin color/Race
White
Black or Brown
Asian
Indigenous
Household income per capita
$249 or more
$125 to $248
$124 or less
Enabling
Registered by primary health care teams
Yes
No
Unknown
Household area
Urban
Rural
Perceived need
Self-rated oral health
Very good or good
Moderate
Bad or very bad
Type of dental attendance
Preventive dental attendance
Dental treatment
Tooth extraction or dental pain
Number of lost teeth
Dependent variable
19,264,898
64,103,920
Individual characteristics
8,805,350
34,369,590
28,248,460
11,945,417
47,188,994
36,179,824
18,056,984
31,684,302
14,927,561
16,684,530
2,015,440
39,762,365
42,396,346
854,911
348,118
46,963,490
22,348,415
14,016,143
49,160,708
24,399,200
9,808,909
74,462,063
8,906,755
63,087,330
17,669,925
2,611,562
39,414,057
30,646,905
13,307,855
83,368,818
23.1
76.9
10.6
41.2
33.9
14.3
56.6
43.4
21.7
38.0
17.9
20.0
2.4
47.7
50.9
1.0
0.4
56.4
26.8
16.8
59.0
29.3
11.8
89.3
10.7
75.7
21.2
3.1
47.3
36.8
16.0
(22.3; 23.9)
(76.1; 77.7)
(9.9; 1.3)
(40.3; 42.1)
(33.1; 34.7)
(13.7; 15.0)
(55.7; 57.5)
(42.5; 44.3)
(20.8; 22.6)
(37.1; 38.9)
(17.2; 18.7)
(19.3; 20.7)
(2.2; 2.7)
(46.6; 48.8)
(49.8; 51.9)
(0.8; 1.3)
(0.3; 0.5)
(55,3; 57,4)
(25,9; 27,7)
(16,2; 17,5)
(57,6; 60,3)
(28,1; 30,5)
(11,1; 12,5)
(88.8; 89.8)
(10.2; 11.2)
(74.9; 76.5)
(20.5; 22.0)
(2.9; 3.4)
(46.3; 48.2)
(35.9; 37.7)
(15.3; 16.6)
2.615 ± 0.045 (2.527; 2.703)
(Continued )
PLOS ONE | https://doi.org/10.1371/journal.pone.0254310 July 9, 2021
7 / 14
PLOS ONE Factors associated with public dental services utilisation in Brazil
Table 1. (Continued)
Variables
n
%
95%CI
Contextual characteristics
Predisposing
Human Development Index
Gini Index
Average Per Capita Income
Enabling
Oral Health Coverage in PHC
PHC: Primary Health Care. CI: Confidence interval.
https://doi.org/10.1371/journal.pone.0254310.t002
83,368,818
83,368,818
83,368,818
0.777 ± 0.001 (0.776; 0.778)
0.523 ± 0.001 (0.522; 0.523)
372.219 ± 1.128 (370.006; 374.432)
83,368,818
51.540 ± 0.168 (51.209; 51.871)
In model 1, only individual variables were included, which maintained the significance
level, except for the variable ‘number of lost teeth’. Most significant adjustments observed were
concerning age. Inversion of PR for education level and skin color/race was observed, with
approximately 50% decrease in PR compared to univariate analysis.
In model 2, when contextual variables were included, no changes were observed in the PR
of individual variables. Gini index lost significance, while PR for HDI largely increased com-
pared to univariate analysis (RP: 189.65—CI95% 0.86; 41,383.46).
Final model included variables presenting statistical significance. PR of all variables
included in the model did not change. Although contextual factors influenced public dental
service utilisation (LR: 270.02; p<0.001), they did not mitigate individual effects.
All individual variables—except for ‘number of lost teeth’—and the contextual variable
‘oral health coverage in primary health care’ were included in the final model. Highest preva-
lence of public dental service utilisation was observed among unable to read or write people
(PR: 3.31–95%CI 3.01; 3.78 –p<0.001), indigenous (PR: 1.40–95%CI 1.18; 1.67– p<0.001),
black or brown (PR: 1.16–95%CI 1.10; 1.21– p<0.001), with per capita household income up
to U$124 (PR: 2.40–95%CI 2.27; 2.55– p<0.001), living in rural areas (PR: 1.28–95%CI 1.22;
1.33– p<0.001), who self-rated oral health as regular (PR: 1.15—CI95% 1.10; 1.20– p<0.001)
or very bad/bad (PR: 1.26—CI95% 1.17; 1.37– p<0.001), and living in FU with high oral health
coverage in the primary care.
Variance between initial null and final models decreased 15%, demonstrating the effects of
the Brazilian FU context on public dental service utilisation.
Discussion
This study verified the association between individual and contextual factors and public dental
service utilisation in Brazil, considering the Andersen Behaviour Model. Our results showed
that contextual and individual characteristics influence public dental service utilisation. At an
individual level, after adjustment for age and sex, educational level, skin color or race, and
household income demonstrated an effect on predisposition to public dental services utilisa-
tion. Enabling factors were living in households enrolled in primary care teams or located in
rural areas. Need factors associated with public dental service utilisation were poor self-rated
oral health and absence of restorative treatment in the last dental attendance. Regarding con-
textual factors, public dental service utilisation was associated with percentage of FU popula-
tion covered by oral health teams in primary care.
Public dental service utilisation by vulnerable groups was evident, demonstrating potential
of the national public policy to expand dental health care access. The reduced utilisation of
PLOS ONE | https://doi.org/10.1371/journal.pone.0254310 July 9, 2021
8 / 14
PLOS ONE Factors associated with public dental services utilisation in Brazil
Table 2. Univariate associations between outcome and the independent variables according to the individual and contextual levels. Brazil, 2019.
Variables
Public dental service utilisation
p-value
PR (95%CI)
No % (95%CI)
Yes % (95%CI)
Individual characteristics
Predisposing
Age
15–19 years
20–39 years
40–59 years
60 years or older
Sex
Female
Male
Educational level
Higher education
High School
Primary school
Incomplete primary school
Unable to read or write
Skin color/Race
White
Black or Brown
Asian
Indigenous
Household income per capita
$249 or more
$125 to $248
$124 or less
Enabling
Registered by primary health care teams
Yes
No
Unknown
Household area
Urban
Rural
Perceived need
Self-rated oral health
Very good or good
Moderate
Bad or very bad
Type of dental attendance
Preventive dental attendance
Dental treatment
Tooth extraction or dental pain
73.9 (70.7; 76.8)
76.2 (75.1; 77.4)
76.9 (75.6; 78.2)
80.9 (79.4; 82.4)
75.3 (74.3; 76.4)
78.9 (77.8; 88.0)
93.8 (92.6; 94.8)
80.9 (79.7; 82.0)
71.6 (69.9; 73.5)
59.3 (57.5; 61.0)
48.4 (43.7; 53.2)
83.2 (82.1; 8436)
70.9 (69.8; 72.1)
86.1 (76.3; 92.2)
59.6 (49.4; 69.0)
89.2 (88.4; 90.0)
68.7 (66.8; 70.4)
48.8 (46.8; 50.8)
69.7 (68.6; 70.8)
88.3 (87.1; 89.4)
84.7 (82.8; 86.3)
79.9 (79.0; 80.7)
51.9 (49.6; 54.1)
80.0 (79.1; 80.9)
68.8 (67.1; 70.4)
56.4 (52.0; 60.7)
78.3 (77.1; 79.5)
83.1 (82.0; 84.2)
58.2 (56.3; 60.1)
26.1 (23.2; 29.3)
23.8 (22.6; 24.9)
23.1 (21.8; 24.4)
19.1 (17.6; 20.6)
24.7 (23.6; 25.7)
21.1 (20.0; 22.2)
6.2 (5.2; 7.4)
19.1 (18.0; 20.3)
28.4 (26.5; 30.4)
40.7 (39.0; 42.5)
51.6 (46.8; 56.3)
16.8 (15.7; 17.9)
29.1 (27.9; 30.2)
13.9 (7.8; 23.7)
40.4 (31.0; 50.6)
10.8 (10.0; 11.6)
31.3 (29.6; 33.2)
51.2 (49.2; 53.2)
30.3 (29.2; 31.4)
11.7 (10.6; 12.9)
15.3 (13.7; 17.2)
20.1 (19.3; 21.0)
48.1 (45.9; 50.4)
20.0 (19.1; 20.9)
31.2 (29.6; 32.9)
43.6 (39.3; 48.0)
21.7 (20.5; 22.9)
16.9 (15.8; 18.0)
41.8 (39.9; 43.7)
Number of lost teeth
2.470 ± 0.051 (2.371; 2.570)
3.097 ± 0.085 (2.930; 3.264)
Contextual characteristics
0.003
<0.001
<0.001
1
0.89 (0.82–0.96)
0.86 (0.79–0.92)
0.77 (0.71–0.84)
1
<0.001
0.89 (0.86–0.93)
<0.001
<0.001
<0.001
<0.001
<0.001
0.150
<0.001
<0.001
<0.001
<0.001
<0.001
1
2.97 (2.73–3.23)
4.63 (4.24–5.05)
6.06 (5.58–6.58)
6.56 (5.91–7.28)
1
1.58 (1.51–1.65)
0.80 (0.0–1.08)
2.21 (1.86–2.63)
1
2.73 (2.59–2.87)
4.23 (4.03–4.45)
1
0.44 (0.42–0.46)
0.54 (0.50–0.58)
1
<0.001
2.06 (1.98–2.15)
<0.001
<0.001
<0.001
<0.001
<0.001
1
1.43 (1.38–1.50)
1.91 (1.77–2.07)
1
0.75 (0.73–0.80)
1.70 (1.63–1.78)
1.01 (1.01–1.02)
Predisposing
Human Development Index
0.781 ± 0.001 (0.780; 0.782)
0.763 ± 0.001 (0.761; 0.765)
<0.001
0.01 (0.01; 0.01)
PLOS ONE | https://doi.org/10.1371/journal.pone.0254310 July 9, 2021
(Continued )
9 / 14
PLOS ONE Factors associated with public dental services utilisation in Brazil
Table 2. (Continued)
Variables
Public dental service utilisation
p-value
PR (95%CI)
Gini Index
Average Income Per Capita
Enabling
Oral Health Coverage in PHC
No % (95%CI)
Yes % (95%CI)
0.522 ± 0.001 (0.521; 0.522)
383.026 ± 1.208 (380.657; 385.395)
0.527 ± 0.001 (0.525; 0.528)
336.259 ± 2.588 (331.186; 341.333)
0.085
<0.001
24.85 (0.64; 961.75)
0.99 (0.99; 0.99)
49.920 ± 0.178 (49.570; 50.270)
56.930 ± 0.391 (56.164; 57.697)
<0.001
1.01 (1.01; 1.02)
PHC: Primary Health Care. CI: Confidence interval. PR: prevalence ratio.
https://doi.org/10.1371/journal.pone.0254310.t003
dental service care is associated with males [9,17–20], black and brown skin color/race
[9,20,21], indigenous people [9,20], low educational level [12,20,22], low-income [9,17,19,20],
lack of health insurance [9,17,20], poor perception of oral health [9,18,23], and living in rural
areas [1,18].
The Brazilian population also demonstrated a social gradient in public dental service utilisa-
tion. Considering dental appointments within last 12 months, the lower the income and edu-
cational level, the higher the number of dental consultations in the SUS. These results
demonstrate that universal public dental service coverage can be a strategy for tackling
inequalities in dental care utilisation. Despite this, inequalities in dental service utilisation per-
sist in Brazil after the National Oral Health Policy implementation [8,24] and may be related
to greater private dental service utilisation. Although public dental care supply increased, the
private sector performed the highest proportion (77.4%) of dental care appoitments in the last
12 months.
The Brazilian scenario of public oral health differs from other countries. Dental care is part
of a universal healthcare system, free of charge at the moment of use and financed by the fed-
eral government with resources from taxes. Oral health teams are included in primary health
care and offer preventive and restorative treatments [1], explaining the potential individual
factor of enrolling in primary care teams to enable public dental services utilisation. This
enrollment enables families to access (e.g., promotion, prevention) and several aspects of fam-
ily and community care. Moreover, oral health teams proved useful as facilitators for access to
Brazilian public dental services, even after adjusting for other variables. This corroborates with
another study, which observed that individuals registered in the Brazilian Family Health Strat-
egy were more likely to use dental services than those unregistered, reducing private insurance
use [25].
Reorientation of oral health care, emphasizing the care model based in primary care, was
the main goal of the National Oral Health Policy. Primary health care has an essential role in
assuming responsibility for detecting needs, providing necessary referrals, monitoring evolu-
tion of rehabilitation, and maintaining rehabilitation in the post-treatment period. Thus, it is
essential to expand the offer of primary care services in oral health. For this purpose, the gov-
ernment is committed to expand and qualify primary care through the family health strategy
[1]. Our results suggest that dental service coverage in primary care increases public dental ser-
vice access. The World Health Organization recommends incorporating primary dental ser-
vices into primary health care initiatives to use pre-existing medical infrastructure and reduce
oral disease burden [26].
In addition to the influence of individual level, characteristics of FUs might affect public
dental service utilisation, whereas socioeconomic factors did not contribute to predisposing
dental public service utilisation. However, oral health coverage in primary care proved to be a
contextual characteristic enabling public dental service access.
PLOS ONE | https://doi.org/10.1371/journal.pone.0254310 July 9, 2021
10 / 14
PLOS ONE Factors associated with public dental services utilisation in Brazil
Table 3. Poisson multilevel regression analysis for public dental services utilisation according to individual and contextual levels. Brazil, 2019.
Variables
Null Model (n = 41,596) Model 1 (n = 41,575) p-value Model 2 (n = 41,575)
PR (95%CI)
PR (95%CI)
Individual characteristics
p-value Final Model (n = 41,575) p-value
PR (95%CI)
Predisposing
Age
15–19 years
20–39 years
40–59 years
60 years or older
Sex
Female
Male
Educational level
Higher education
High School
Primary school
Incomplete primary school
Unable to read or write
Skin color/Race
White
Black or Brown
Asian
Indigenous
Household income per capita
$249 or more
$125 to $248
$124 or less
Enabling
Are you registered by primary health care teams?
Yes
No
Unknown
Household area
Urban
Rural
Perceived need
Self-rated oral health
Very good or good
Moderate
Bad or very bad
Type of dental attendance
Preventive dental attendance
Dental treatment
Tooth extraction or dental pain
Number of lost teeth
Predisposing
Human Development Index
1
1.08 (1.00; 1.17)
1.06 (0.98; 1.15)
1.04 (0.94; 1.14)
0.032
0.116
0.385
1
1.08 (1.00; 1.17)
1.06 (0.98; 1.15)
1.04 (0.94; 1.14)
0.033
0.118
0.387
1
1.08 (1.00; 1.17)
1.06 (0.95; 1.15)
1.04 (0.94; 1.14)
0.034
0.119
0.388
1
1
1
0.89 (0.86; 0.93)
<0.001
0.89 (0.86; 0.93)
<0.001
0.89 (0.86; 0.93)
<0.001
1
2.06 (2.89; 2.25)
2.67 (2.43; 2.93)
3.19 (2.91; 3.49)
3.38 (3.02; 3.78)
<0.001
<0.001
<0.001
<0.001
1
2.06 (1.89; 2.25)
2.67 (2.43; 2.93)
3.19 (2.92; 3.49)
3.38 (3.02; 3.78)
1
1
1.16 (1.11; 1.21)
<0.001
1.16 (1.10; 1.21)
0.82 (0.61; 1.10)
0.192
0.82 (0.61; 1.10)
1.40 (1.18; 1.67)
<0.001
1.40 (1.18; 1.67)
<0.001
<0.001
<0.001
<0.001
<0.001
0.188
<0.001
1
2.06 (1.89; 2.25)
2.67 (2.43; 2.93)
3.19 (2.91; 3.49)
3.37 (3.01; 3.78)
1
1.16 (1.10; 1.21)
0.82 (0.61; 1.10)
1.40 (1.18; 1.67)
<0.001
<0.001
<0.001
<0.001
<0.001
0.191
<0.001
1
1
1
1.85 (1.75; 1.95)
2.41 (2.27; 2.55)
<0.001
<0.001
1.84 (1.75; 1.95)
2.40 (2.27; 2.55)
<0.001
<0.001
1.84 (1.75; 1.95)
2.40 (2.27; 2.55)
<0.001
<0.001
1
1
1
0.64 (0.61; 0.68)
0.72 (0.67; 0.78)
<0.001
<0.001
0.64 (0.61; 0.68)
0.73 (0.68; 0.78)
<0.001
<0.001
0.64 (0.61 (0.68)
0.73 (0.68; 0.78)
<0.001
<0.001
1
1
1
1.28 (1.22; 1.34)
<0.001
1.28 (1.22; 1.33)
<0.001
1.28 (1.22; 1.33)
<0.001
1
1
1.15 (1.10; 1.20)
1.26 (1.17; 1.37)
<0.001
<0.001
1.15 (1.10; 1.20)
1.26 (1.17; 1.37)
<0.001
<0.001
1
1.15 (1.10; 1.20)
1.26 (1.17; 1.37)
<0.001
<0.001
1
1
1
0.64 (0.61; 0.67)
<0.001
0.64 (0.61; 0.67)
1.04 (0.99; 1.09)
0.99 (0.99; 1.00)
0.062
0.929
1.04 (0.99; 1.09)
-
<0.001
0.066
-
Contextual characteristics
189.65 (0.86; 41,383.46)
0.056
0.64 (0.61; 0.67)
1.04 (0.99; 1.09)
<0.001
0.065
-
-
-
-
(Continued )
11 / 14
PLOS ONE | https://doi.org/10.1371/journal.pone.0254310 July 9, 2021
PLOS ONE Factors associated with public dental services utilisation in Brazil
Gini Index
Average per capita income
Enabling
Oral Health Coverage in PHC
Fixed Effects
Intercept (95%CI)
Random Effects
Variance (95%CI)
LR test (Chi2, p-value)
Table 3. (Continued)
Variables
Null Model (n = 41,596) Model 1 (n = 41,575) p-value Model 2 (n = 41,575)
PR (95%CI)
PR (95%CI)
1.19 (0.17; 8.33)
0.99 (0.99; 1.00)
p-value Final Model (n = 41,575) p-value
PR (95%CI)
0.854
0.044
-
-
-
-
1.00 (1.00; 1.01)
0.004
1.00 (1.00; 1.01)
<0.001
-1.36 (-1.52; -1.19)
0.07 (0.06; 0.08)
0.01 (0.01; 0.07)
0.04 (0.03; 0.05)
0.19 (0.11; 0.34)
1631.00 (<0.001)
0.06 (0.03; 0.12)
369.79 (<0.001)
0.03 (0.02; 0.07)
238.63 (<0.001)
0.04 (0.02; 0.08)
270.02 (<0.001)
Model 1: Individual variables; Model 2: Individual variables, maintaining significance level in model 1 and contextual variables; Final model: Individual and contextual
variables, maintaining significance level. PHC: Primary Health Care; LR: Likelihood Ratio.
https://doi.org/10.1371/journal.pone.0254310.t004
Although oral health policies were one of Brazilian government priorities in 2003, current
national agenda [3] neglected oral health care (i.e., low political priority) and excluded oral
health teams from primary care services since 2017. The limited government budget for oral
health care suggests that dental care is unnecessary and should not be provided by the SUS,
unlike other medical services [26]. As shown in this study, reduced policy expansion in Brazil
may threaten equity in dental service utilisation since public services may mitigate inequalities.
Also, public services, part of a universal system, effectively reduced inequalities in dental ser-
vice utilisation, offering an alternative to adopt private dental insurance.
The present study has some strengths and limitations. We used data from a population-
based survey performed with people living in private households. Interviewers were trained in
two stages, and data was collected using digital mobile devices. Urban and rural areas were
estimated for major national regions, FU, capitals, and metropolitan regions [12]. Nonetheless,
this study presents classic limitations of studies with a cross-sectional design. Data were subject
to information and memory bias since the primary outcome was self-reported. However, bias
is expected to be random and small due to sample size.
Despite limitations, this study provides a valuable analysis regarding the profile of dental
service users in Brazil and demonstrates that individual and contextual factors are associated
with public dental service utilisation. At the individual level, sex, educational level, skin color/
race, and household income are predisposing factors for public dental service utilisation,
whereas enrolling in primary care teams and living in rural areas were enabling factors. At the
contextual level, a high percentage of the population covered by oral health in primary care
was an enabling factor for public dental service utilisation.
According to Andersen and Newman [27], the intervention variable must be mutable to
promote equity of access. Changes in health policies may change health service utilisation.
Demographic and social structure variables associated with dental service utilisation have a
low potential for mutability. Alternatively, enabling variables, such as expanding public service
coverage and enrolling families in primary care, have a high potential for mutability through
government actions. Thus, our study revealed that government action is fundamental for
reducing inequalities, observing a mitigating effect of public policies on inequalities associated
with dental services utilisation. This result may guide evidence-based decision-making for pol-
icymakers. Nevertheless, expansion of government actions are needed because coverage is still
low and inequalities are persistent.
PLOS ONE | https://doi.org/10.1371/journal.pone.0254310 July 9, 2021
12 / 14
PLOS ONE Factors associated with public dental services utilisation in Brazil
Acknowledgments
The authors thank Probatus Academic Services for providing scientific language revision and
editing.
Author Contributions
Conceptualization: Maria Helena Rodrigues Galvão, Arthur de Almeida Medeiros, Angelo
Giuseppe Roncalli.
Formal analysis: Maria Helena Rodrigues Galvão, Arthur de Almeida Medeiros.
Methodology: Maria Helena Rodrigues Galvão, Arthur de Almeida Medeiros.
Supervision: Angelo Giuseppe Roncalli.
Writing – original draft: Maria Helena Rodrigues Galvão, Arthur de Almeida Medeiros.
Writing – review & editing: Maria Helena Rodrigues Galvão, Arthur de Almeida Medeiros,
Angelo Giuseppe Roncalli.
References
1. Ministe´ rio da Sau´de (Brazil). Diretrizes da Polı´tica Nacional de Sau´ de bucal.[Internet]. [cited 2021 Jan
27]. Available from: http://189.28.128.100/dab/docs/publicacoes/geral/diretrizes_da_politica_nacional_
de_saude_bucal.pdf.
2. Roncalli AG, Sheiham A, Tsakos G, Watt RG. Socially unequal improvements in dental caries levels in
Brazilian adolescents between 2003 and 2010. Community dentistry and oral epidemiology.2015; 43
(4): 317–324. https://doi.org/10.1111/cdoe.12156 PMID: 25660728
3. Chaves SCL, Aranha-Rossi TR, Lima AMFS. Dental service coverage and oral health promotion com-
munity actions in primary care in Brazil between 2003 and 2019. Heal Policy OPEN. 2020; 1;1:100022.
4. Conselho Federal de Odontologia (Brazil). Dados Estatı´sticos de Profissionais e Entidades Ativas Por
Ano—CFO [Internet]. [cited 2020 Dec 12]. Available from: https://website.cfo.org.br/dados-estatisticos-
de-profissionais-e-entidades-ativas-por-ano/.
5. Mullachery P, Silver D, Macinko J. Changes in health care inequity in Brazil between 2008 and 2013. Int
J Equity Health [Internet]. 2016 Nov 17 [cited 2021 Jan 19]; 15(1):1–12. Available from: https://link.
springer.com/articles/10.1186/s12939-016-0431-8. https://doi.org/10.1186/s12939-016-0431-8 PMID:
27852309
6. Pinto RS, de Abreu MHNG, Vargas AMD. Comparing adult users of public and private dental services
in the state of Minas Gerais, Brazil. BMC Oral Health [Internet]. 2014 Aug 6 [cited 2021 Feb 1]; 14(1).
Available from: https://pubmed.ncbi.nlm.nih.gov/25099268/. https://doi.org/10.1186/1472-6831-14-100
PMID: 25099268
7. Constante HM. Racial inequalities in public dental service utilization: Exploring individual and contextual
determinants among middle-aged Brazilian adults. Community Dent Oral Epidemiol [Internet]. 2020
Aug 1 [cited 2020 Dec 23]; 48(4):302–8. Available from: https://onlinelibrary.wiley.com/doi/10.1111/
cdoe.12533. PMID: 32237080
8. Rebelo Vieira JM, Rebelo MAB, Martins NM de O, Gomes JFF, Vettore MV. Contextual and individual
determinants of non-utilization of dental services among Brazilian adults. J Public Health Dent [Internet].
2019 Dec 1 [cited 2020 Jul 1]; 79(1):60–70. Available from: https://pubmed.ncbi.nlm.nih.gov/30468261/
. https://doi.org/10.1111/jphd.12295 PMID: 30468261
9. Herkrath FJ, Vettore MV, Werneck GL. Contextual and individual factors associated with dental ser-
vices utilisation by Brazilian adults: A multilevel analysis. PLoS One. 2018 February 1;13(2). https://doi.
org/10.1371/journal.pone.0192771 PMID: 29420660
10. Andersen RM, Davidson PL, Baumeister SE. Improving access to care. Changing the US health sys-
tem: Key issues in health services policy and management. John Wiley Sons. 2013;33–69.
11. Shahid M, Shum J, Tadakamadla SK, Kroon J, Peres MA. (2021). Theoretical evidence explaining the
relationship between socio-demographic and psychosocial barriers on access to oral health care
among adults: A scoping review. Journal of Dentistry, 103606. https://doi.org/10.1016/j.jdent.2021.
103606 PMID: 33582113
PLOS ONE | https://doi.org/10.1371/journal.pone.0254310 July 9, 2021
13 / 14
PLOS ONE Factors associated with public dental services utilisation in Brazil
12. Stopa SR, Szwarcwald CL, Oliveira MM de, Gouvea E de CDP, Vieira MLFP, Freitas MPS de, et al.
Pesquisa Nacional de Sau´de 2019: histo´rico, me´ todos e perspectivas. Epidemiol e Serv saude Rev do
Sist Unico Saude do Bras [Internet]. 2020 [cited 2021 Feb 1]; 29(5):e2020315. Available from: http://
scielo.iec.gov.br/scielo.php?script=sci_arttext&pid=S1679-49742020000500035&lng=pt&nrm=
iso&tlng=pt.
13.
14.
Instituto Brasileiro de Geografia e Estatı´stica (Brasil). Pesquisa Nacional por Amostra de Domicı´lios-
Acesso e utilizac¸ão de servic¸os de sau´ de: 2019. Rio de Janeiro (RJ); 2020.
Zhu Y, Matsuyama Y, Ohashi Y, Setoguchi S. When to conduct probabilistic linkage vs. deterministic
linkage? A simulation study. J Biomed Inform. 2015 Aug; 56:80–6. https://doi.org/10.1016/j.jbi.2015.05.
012 Epub 2015 May 22. PMID: 26004791.
15. Hair JF, Black WC, Babin BJ, Anderson RETR. Analise multivariada de dados. 6th ed. Porto Alegre:
Bookman; 2009. 49–99 p.
16. Merlo J, Ohlsson H, Lynch KF, Chaix B, Subramanian SV. Individual and collective bodies: Using mea-
sures of variance and association in contextual epidemiology. J Epidemiol Community Health. 2009; 63
(12):1043–1048. https://doi.org/10.1136/jech.2009.088310 PMID: 19666637
17. Chen M, Wright CD, Tokede O, Yansane A, Montasem A, Kalenderian E, et al. Predictors of dental care
utilization in north-central Appalachia in the USA. Community Dent Oral Epidemiol. 2019 August 1; 47
(4):283–90. https://doi.org/10.1111/cdoe.12453 PMID: 30993747
18. Herkrath FJ, Vettore MV, Werneck GL. Utilisation of dental services by Brazilian adults in rural and
urban areas: a multi-group structural equation analysis using the Andersen behavioural model. BMC
Public Health. 2020 June 17; 20(1):953. https://doi.org/10.1186/s12889-020-09100-x PMID: 32552777
19. Maffioletti F, Vettore MV, Rebelo M, Herkrath F, Queiroz A, Herkrath AP, et al. Predisposing, enabling,
and need characteristics of dental services utilization among socially deprived schoolchildren. J Public
Health Dent [Internet]. 2020 June 1 [cited 2020 July 1]; 80(2):97–106. Available from: https://pubmed.
ncbi.nlm.nih.gov/31788798/. https://doi.org/10.1111/jphd.12349 PMID: 31788798
20. Reda SFSM, Reda SFSM, Murray Thomson W, Schwendicke F. Inequality in Utilization of Dental Ser-
vices: A Systematic Review and Meta-analysis [Internet]. Vol. 108, American Journal of Public Health.
American Public Health Association Inc.; 2018 [cited 2020 July 1]. p. e1–7. Available from: https://
pubmed.ncbi.nlm.nih.gov/29267052/.
21. Chertok IRA, Chertok N, Haile ZT, Chavan B. Association of youth characteristics and recent utilization
of dental services in the United States. Front Pediatr [Internet]. 2018 [cited 2020 July 1]; 6. Available
from: https://pubmed.ncbi.nlm.nih.gov/29780790/. https://doi.org/10.3389/fped.2018.00104 PMID:
29780790
22. Mariño R, Giacaman RA. Patterns of use of oral health care services and barriers to dental care among
ambulatory older Chilean. BMC Oral Health. 2017 January 9; 17(1). https://doi.org/10.1186/s12903-
016-0329-2 PMID: 28068973
23. Adunola F, Garcia I, Iafolla T, Boroumand S, Silveira ML, Adesanya M, et al. Self-perceived oral health,
normative need, and dental services utilization among dentate adults in the United States: National
Health and Nutrition Examination Survey (NHANES) 2011–2014. J Public Health Dent [Internet]. 2019
Dec 1 [cited 2020 Jul 1]; 79(1):79–90. Available from: https://pubmed.ncbi.nlm.nih.gov/30633355/.
https://doi.org/10.1111/jphd.12300 PMID: 30633355
24. Baldani MH, Mendes YBE, De Campos Lawder JA, De Lara API, Da Silva Rodrigues MMA, Antunes
JLF. Inequalities in dental services utilization among Brazilian low-income children: The role of individ-
ual determinants. J Public Health Dent. 2011 Dec; 71(1):46–53. https://doi.org/10.1111/j.1752-7325.
2010.00201.x PMID: 21667543
25. Pilotto LM, Celeste RK. The relationship between private health plans and use of medical and dental
health services in the Brazilian health system. Cienc e Saude Coletiva [Internet]. 2019 July 1 [cited
2021 January 27]; 24(7):2727–36. Available from: https://orcid.org/0000-0002-2468-6655. https://doi.
org/10.1590/1413-81232018247.24112017 PMID: 31340289
26. Wang TT, Mathur MR, Schmidt H. Universal health coverage, oral health, equity and personal responsi-
bility. Bull World Health Organ [Internet]. 2020 Oct 1 [cited 2021 Jan 18]; 98(10):719–21. Available
from: http://www.who.int/bulletin/volumes/98/10/19-247288/en/. https://doi.org/10.2471/BLT.19.
247288 PMID: 33177761
27. Andersen R, Newman JF. Societal and individual determinants of medical care utilization in the United
States. The Milbank Memorial Fund Quarterly. Health and Society. 1973; 51, 1: 95–124.
PLOS ONE | https://doi.org/10.1371/journal.pone.0254310 July 9, 2021
14 / 14
PLOS ONE |
10.1371_journal.pone.0257370 | RESEARCH ARTICLE
Short-term elevations in glucocorticoids do
not alter telomere lengths: A systematic
review and meta-analysis of non-primate
vertebrate studies
Lauren ZaneID
1*, David C. EnsmingerID
1
1,2, Jose´ Pablo Va´ zquez-MedinaID
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: Zane L, Ensminger DC, Va´zquez-Medina
JP (2021) Short-term elevations in glucocorticoids
do not alter telomere lengths: A systematic review
and meta-analysis of non-primate vertebrate
studies. PLoS ONE 16(10): e0257370. https://doi.
org/10.1371/journal.pone.0257370
Editor: Gabriele Saretzki, University of Newcastle,
UNITED KINGDOM
Received: April 28, 2021
Accepted: August 29, 2021
Published: October 1, 2021
Copyright: © 2021 Zane et al. This is an open
access article distributed under the terms of the
Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: This study was
registered in the Open Science Framework Registry
(https://osf.io/rqve6). The review protocol can be
accessed at https://bookdown.org/MathiasHarrer/
Doing_Meta_Analysis_in_R/. Data are available
from the Dryad Data Repository (https://datadryad.
org/stash/dataset/doi:10.6078/D10T42).
Funding: The authors received no specific funding
for this work.
Competing interests: The authors have declared
that no competing interests exist.
1 Department of Integrative Biology, University of California, Berkeley, CA, United States of America,
2 Department of Biological Sciences, San Jose State University, San Jose, CA, United States of America
* laurenzane@berkeley.edu
Abstract
Background
The neuroendocrine stress response allows vertebrates to cope with stressors via the acti-
vation of the Hypothalamic-Pituitary-Adrenal (HPA) axis, which ultimately results in the
secretion of glucocorticoids (GCs). Glucocorticoids have pleiotropic effects on behavior and
physiology, and might influence telomere length dynamics. During a stress event, GCs
mobilize energy towards survival mechanisms rather than to telomere maintenance. Addi-
tionally, reactive oxygen species produced in response to increased GC levels can damage
telomeres, also leading to telomere shortening. In our systematic review and meta-analysis,
we tested whether GC levels impact telomere length and if this relationship differs among
time frame, life history stage, or stressor type. We hypothesized that elevated GC levels are
linked to a decrease in telomere length.
Methods
We conducted a literature search for studies investigating the relationship between telomere
length and GCs in non-human vertebrates using four search engines: Web of Science, Goo-
gle Scholar, Pubmed and Scopus, last searched on September 27th, 2020. This review
identified 31 studies examining the relationship between GCs and telomere length. We
pooled the data using Fisher’s Z for 15 of these studies. All quantitative studies underwent a
risk of bias assessment. This systematic review study was registered in the Open Science
Framework Registry (https://osf.io/rqve6).
Results
The pooled effect size from fifteen studies and 1066 study organisms shows no relationship
between GCs and telomere length (Fisher’s Z = 0.1042, 95% CI = 0.0235; 0.1836). Our
meta-analysis synthesizes results from 15 different taxa from the mammalian, avian,
amphibian groups. While these results support some previous findings, other studies have
found a direct relationship between GCs and telomere dynamics, suggesting underlying
PLOS ONE | https://doi.org/10.1371/journal.pone.0257370 October 1, 2021
1 / 17
PLOS ONE Elevated glucocorticoids and telomere attrition
mechanisms or concepts that were not taken into account in our analysis. The risk of bias
assessment revealed an overall low risk of bias with occasional instances of bias from miss-
ing outcome data or bias in the reported result.
Conclusion
We highlight the need for more targeted experiments to understand how conditions, such as
experimental timeframes, stressor(s), and stressor magnitudes can drive a relationship
between the neuroendocrine stress response and telomere length.
Introduction
The vertebrate neuroendocrine stress response integrates external stimuli into a broad range
of physiological adjustments through the activation of the Hypothalamic-Pituitary-Adrenal
axis (HPA axis) and the concomitant secretion of glucocorticoids (GCs) [1, 2]. While the pri-
mary GC produced varies by taxa (e.g., cortisol in humans and corticosterone in birds and
other mammals [3]), the impacts of GCs on organismal physiology are remarkably similar.
Across species, an increase in GC secretion can typically be detected in 3–5 minutes following
interaction with a stressor [4]. Additionally, GCs are relatively easy to quantify because they
are present in all vertebrates and can be measured noninvasively in multiple matrices includ-
ing hair and feces using a variety of assays [5, 6]. Therefore, wildlife stress physiology studies
often rely on GC measurements as an indicator of the neuroendocrine stress response [7]. Fol-
lowing their secretion, GCs induce a myriad of acute behavioral and physiological effects to
prioritize immediate survival [8, 9].
In addition to allowing animals to cope with immediate stressors, GCs can influence other
cellular processes such as telomere length dynamics. Telomeres are evolutionarily conserved
caps that protect chromosomes against the loss of coding nucleotides during cell replication
and against chromosomal fusion [10]. Telomere shortening is associated with aging, the neu-
roendocrine stress response, and survival, and is thus of interest to several fields of biology [1,
11]. In humans, increased telomere loss predicts the onset of age-related diseases, cardiovascu-
lar complications, cellular senescence, and other aging phenotypes [12, 13]. Telomere attrition
can be attributed to several causes including the “end replication problem” in which the termi-
nal end of linear DNA cannot be completely replicated by the lagging strand [14]. Since the
end replication problem occurs at every cell division, telomeres continuously shorten with age
progression [15]. Other stressors such as inflammatory challenges erode telomeres regardless
of age [16].
In non-human vertebrates including birds, mammals, fish, amphibians and reptiles, expo-
sure to challenging environmental conditions correlates with shorter telomeres [17, 18].
Reproductive stressors such as an artificially increased brood size can also shorten telomeres
in zebra finch parents compared to controls and parents with a reduced brood size [19]. Early
telomere length is positively correlated with survival and lifetime breeding success in both wild
purple-crowned fairy wrens and zebra finches. Thus, individuals with longer telomeres are
more likely to survive and produce more offspring that survive to maturity [20, 21]. Therefore,
telomere dynamics—the change in telomere length attributed to processes of elongation and
shortening—is related to organismal fitness [22]. In addition to impacting telomere length,
stressors that lead to energy limitation such as psychological stress, disease, accelerated growth,
nutrient shortage and work load activate the HPA axis causing the release of GCs [23].
PLOS ONE | https://doi.org/10.1371/journal.pone.0257370 October 1, 2021
2 / 17
PLOS ONE Elevated glucocorticoids and telomere attrition
Thus, several hypothesized connections between GCs and telomere length exist. Firstly,
GCs are an essential part of the vertebrate stress response, and their primary function is to
mobilize energy [5]. Accordingly, the “metabolic telomere attrition hypothesis” proposes that
during events that require an increased amount of energy and metabolic rates, telomeres are
shortened as collateral [20]. As a result of the high energy expenditure, the energetically expen-
sive maintenance of telomeres cannot take place as an emergency survival mechanism due to a
shift in energy allocation [23]. In addition, GCs stimulate the generation of reactive oxygen
species (ROS) and subsequent oxidative damage to telomeres, which are particularly suscepti-
ble to oxidation due to a high guanine content [11, 24–26]. Finally, cortisol reduces telomerase
—the enzyme responsible for telomere maintenance—activity in human T lymphocytes [27].
This reduction in telomerase activity can result in excessive telomere attrition [28]. Since wild-
life face an array of stressors throughout their lifetime and these stressors can erode telomeres,
GCs may play a mechanistic role in telomere loss [1].
External stressors cause pleiotropic effects that can potentially influence telomere dynamics,
however the evidence for a causal relationship between GCs and telomere length is sparse.
Two recent literature reviews on the topic by Angelier et.al 2018 and Casagrande and Hau
2019 [11, 23] summarize the potential relationship between GCs and telomere length. How-
ever, it is essential to build a quantitative understanding of the relationships between the neu-
roendocrine stress response and its downstream effects. In this study, we review the existing
literature for empirical evidence of the relationship between GC secretion and telomere length
to better understand the underlying mechanism of telomere shortening as well as potential
consequences of the neuroendocrine stress response in non-primate vertebrates. Using a
meta-analytical framework, we tested whether GC levels impact telomere length and if this
relationship can differ among time frame, life history stage, or stressor type. We hypothesized
that elevated GC levels are linked to a decrease in telomere length.
Methods
Literature search and study selection
We conducted a literature search for studies investigating the relationship between telomere
length and GC levels in non-human vertebrates using four search engines: Web of Science,
Google Scholar, Pubmed and Scopus. Five subsets of the following keywords ‘reactive oxygen
species,’ ‘antioxidant,’ ‘glucocorticoid,’ ‘cortisol,’ ‘corticosterone,’ ‘telomere length,’ ‘chronic
stress,’ ‘oxidative stress,’ ‘acute stress,’ ‘chronic stress,’ ‘telomeres,’ and ‘HPA axis,’ were con-
ducted in each search engine. We did not specify a time frame in our literature search. Addi-
tional records were obtained from the reference section of studies included in the meta-
analysis. Our study includes a qualitative synthesis of 31 full-text, peer-reviewed studies, and
we report effect sizes for 15 of these studies.
Studies were excluded if (1) GCs were administered, but physiological measurements such
as feather or plasma GC levels were not taken. Such studies were excluded because it would
not be possible to calculate the appropriate effect size (Fisher’s Z) for correlation data. For
homogeneity in effect size calculation and statistical analysis, we did not include studies in
which (2) GCs and telomere length were not specifically measured at two different time points
(before or after treatment) (3) raw data was not accessible to use for the effect size calculation,
or (4) telomere length measurements or GC measurements were log transformed.
Statistical data analyses
Meta-analysis. We conducted statistical analyses exclusively on studies with raw data
available. When data was not publicly accessible, we contacted authors via email for consensual
PLOS ONE | https://doi.org/10.1371/journal.pone.0257370 October 1, 2021
3 / 17
PLOS ONE Elevated glucocorticoids and telomere attrition
access. For each study, the correlation coefficient (R2) was calculated by fitting a linear mixed
model using the “lme4” R package (version 3.6.1, R Development Core Team, Boston, MA).
When possible, random effects such as multiple blood draws from a single individual were
incorporated in the linear mixed model (LMER) to account for variability not captured by
explanatory parameters. For studies where a random effect could not be determined, a linear
model (LM) was fitted. From the LMs and LMERs, R2 values were obtained from the model
and converted into Fisher’s Z, then adjusted for sample size and combined into a pooled effect
size (Fisher’s Z; Z) using the R package “meta”. The random-effects model meta-analysis was
implemented in our study as this model accounts for the assumption that studies come from
different populations, rather than the same population. These pooled effect sizes were then
visualized in a forest plot.
The “meta” package was also used to assess the statistical difference between observed and
fixed effect model estimate of effect size (Cochrane’s Q) and the percent of variability in effect
sizes that is not caused by sampling error (I2). After estimating heterogeneity, we identified
potential outliers. Studies were classified as outliers if the study had an effect size with a confi-
dence interval that did not intersect with the confidence interval of the pooled effect size.
Since some studies can have a larger influence on the pooled effect size than others due to
its sample size or individual effect size, we conducted an influence analysis. The analysis was
conducted by omitting each study one at a time and simulating the pooled effect size, with a
confidence interval had the study not been included. This influence analysis was represented
in a Baujat plot, which shows the contribution of each study to heterogeneity as Cochrane’s Q,
and compares this to the study’s influence on the pooled effect size.
Subgroup analysis. Since experimental design can affect the outcome of a study, differ-
ences in effect size may be attributed to these variables. As such, further sources of between-
study heterogeneity were investigated through subgroup analysis and meta-regression. In the
subgroup analysis, studies were grouped based on different categorical experimental parame-
ters. We completed eight different subgroup analyses for the following parameters—duration
of stressor, type of GC assay, telomere assay, species, taxa, study type, life history stage, and
stressor type. For each subgroup analysis, a pooled effect size (Fisher’s Z) was calculated. We
then compared pooled effect sizes and tested for between-study subgroup differences. The
meta-regression was analogous to the subgroup analysis, except the parameter of investigation
is continuous rather than categorical. We conducted one meta-regression for publication year
and subsequently tested for between-study subgroup differences. For all analyses the signifi-
cance threshold was set at p<0.05.
In the subgroup analysis, studies included in the meta-analysis were clustered based on cat-
egorical grouping and represented as a pooled effect size with a 95% confidence interval. The
between study difference was indicated by Cochrane’s Q and the subsequent p-value for this
statistical measure. The first subgroup analysis “stressor duration” organized studies based on
the timeframe of the experiment—less than one week (n = 1), one to two weeks (n = 2), two to
three weeks (n = 7), three to four weeks (n = 1), or longer than four weeks (n = 4)—. The sec-
ond subgroup analysis, “type of stress” compared anthropogenic (n = 5) to naturally occurring
stress (n = 7), or if stress was simulated by GC administration (n = 3). The subsequent sub-
group analysis “life history stage, “differentiates studies based on pre-maturate study organ-
isms (n = 12), or post-maturate study organisms (n = 3). Next, the subgroup “GC assay,”
separates studies into those that quantified plasma GCs (n = 13) or non-plasma GCs (n = 2).
Similarly, by performing the subgroup analysis for the variable “telomere assay” we hoped to
parse out potential differences between the three methods of telomere quantification: qPCR
(n = 7), TeloTAGGG (n = 1), and Telomerase Restriction Fragment (TRF; n = 7). The fifth
subgroup analysis contrasts avian (n = 12) and non-avian (n = 3) studies. To explore the
PLOS ONE | https://doi.org/10.1371/journal.pone.0257370 October 1, 2021
4 / 17
PLOS ONE Elevated glucocorticoids and telomere attrition
relationship between individual species, we performed an additional subgroup analysis for
each species included in the study. Finally, the subgroup analysis “study type” distinguished
studies based on study design: cross-sectional (n = 5), repeated measures (n = 2), or within
individual (n = 8) design.
Publication bias. Published studies may not accurately represent the total studies investi-
gating an area of research due to selective outcome reporting, missing studies and a higher
likelihood of publication of studies reporting a significant (p<0.05) result. While proving
selective outcome reporting and other forms of publication biases is challenging, missing stud-
ies can be visually represented using a funnel plot. Commonly, studies with small effect sizes
and small sample studies are likely to be missing, which can be depicted with funnel asymme-
try or holes in the funnel plot. We created a funnel plot by graphing effect size against study
precision, defined as the standard error of the effect size to visualize potential publication bias.
We also report an Egger’s test, which is represented by the intercept, it’s confidence interval,
and the associated p-value to determine if publication bias was statistically significant.
Risk of bias in included studies. We assessed studies for missing outcome-level data,
measurement of the outcomes, and outcome reporting in each included study. For the missing
outcome-level data domain, we considered studies that could not report values for telomere
length or GCs in less than 10% of total study organisms as low risk. We designated studies that
did not report these values for 10–50% of study organisms as moderate risk and studies that
did not report values for over 50% of GCs or telomere length, as high risk. Secondly, we based
risk of bias in the measurement of outcome on the type of GC and telomere measurement.
Low risk studies utilized plasma GCs or salivary GCs because these quantifications capture ele-
vations related to a short-term stress event within minutes. Studies that measured GCs in fecal
matter received a ranking of some concern because fecal GCs typically encapsulate cumulative
stress over the day rather than GCs related to a particular environmental stressor. Fecal GCs
also received a ranking of some concern due to potential variations related to storage and col-
lection times, which can affect the concentration of fecal GC metabolites in a sample [29]. We
considered studies that measured GCs in feathers as high risk because feathers incorporate
GCs in over a month. Additionally, we considered feather GC quantification as high risk
because feather preparation and GC extraction can vary greatly [30]. Finally, for the risk of
bias due to outcome reporting we denoted studies that based results off a subset of time points
or measurements high risk. We denoted studies that report results based on all time points
with low risk. We took these three domains into consideration when assessing overall risk of
bias.
Results
Literature search and study selection
We electronically screened 789 records for relevance from the following databases: Google
Scholar (n = 512), Web of Science (n = 105), PubMed (n = 72), and Scopus (n = 100). 2113
additional records were hand screened from the reference section of the 31 studies used in
qualitative analysis. Of the total 2902 records that were screened for relevance, 78 were
removed as duplicates and 2,489 did not fit criteria for our study. For example, some excluded
studies include human trials, cell culture work, or studies that only assessed research questions
pertaining to either telomere length or GC levels, but not both (Fig 1; S1 Table). Of the 183
assessed full-text articles, we removed 152 studies that did not fulfill our inclusion criteria. We
statistically analyzed 15 of the remaining 31 studies, the ones that provided raw data for analy-
sis either within the manuscript or after contacting the corresponding author [16, 22, 29–42].
The other 16 studies appeared to fit criteria but did not provide raw data for analysis [26, 30,
PLOS ONE | https://doi.org/10.1371/journal.pone.0257370 October 1, 2021
5 / 17
PLOS ONE Elevated glucocorticoids and telomere attrition
Fig 1. PRISMA diagram. PRISMA diagram showing the selection process for references included in the meta-analysis
of the effects of GCs on telomere length.
https://doi.org/10.1371/journal.pone.0257370.g001
43–55]. The literature and study selection process is illustrated using a PRISMA diagram
(Fig 1).
Meta-analysis
The random-effects model meta-analysis is represented as a pooled effect size (Fisher’s Z) with
95% confidence intervals (Fig 2). No studies were removed as outliers. The model found no
relationship between GC levels and telomere length (Fisher’s Z = 0.1042, CI = 0.0235; 0.1836).
Both heterogeneity measures, Cochrane’s Q (Q = 11.31, p = 0.6615) and I2 with 95% confi-
dence intervals (I2 = 0.0%; CI = 0.0%; 42.6%) yielded similar results.
The influence analysis indicated that theoretically removing one study at a time did not
yield pooled effect sizes (Fisher’s Z = 0.09–0.11) that differed from the original pooled effect
size (Fisher’s Z = 0.11, S1 Fig). Additionally, the influence analysis demonstrated that certain
studies unevenly impacted the pooled effect size and/or overall heterogeneity (S2 Fig), but no
studies were removed as outliers.
Subgroup analysis
The subgroup analysis for “stressor duration” found no differences between any of the tested
time frames (Table 1). The difference between-studies was not statistically significant
Fig 2. Forest plot. Distribution of effect sizes of GCs on telomere length and 95% CI of effect size. Dashed lines
represent pooled effect sizes using a random and fixed effect model. Heterogeneity (I2), the percent of variability in
effect sizes that is not caused by sampling error indicates very little variability in effect size. Weight indicates the
influence the study has on the overall pooled effect.
https://doi.org/10.1371/journal.pone.0257370.g002
PLOS ONE | https://doi.org/10.1371/journal.pone.0257370 October 1, 2021
6 / 17
PLOS ONE Table 1. Pooled effect sizes with 95% CI of experimental parameters investigated during the five subgroup analyses for stressor duration, stressor type, life history
stage, GC assay and taxa group.
Experimental Parameter
Number of Studies
Effect Size
(Fisher’s Z)
95% CI
Elevated glucocorticoids and telomere attrition
Stressor Duration
Type of Stress
Life History Stage
GC Assay
Taxa Group
Species
Capreolus capreolus
Coturnix japonica
Fregata magnificens
Hydrobates pelagicus
Parus major
Phalacrocorax aristotelis
Rana temporaria
Rissa tridactyla
Sterna hirundo
Sturnus unicolor
Tachycineta bicolor
Turdus merula
Welsh pony
Telomere Assay
Study Type
< 1 week n = 1
1–2 weeks n = 2
2–3 weeks n = 7
3–4 weeks n = 1
> 4 weeks n = 4
Anthropogenic n = 5
Naturally occurring n = 7
GC administration n = 3
Pre-maturation n = 12
Post-maturation n = 3
Plasma GCs n = 12
Non-Plasma GCs n = 3
Avian n = 12
Non-Avian n = 3
n = 1
n = 1
n = 1
n = 1
n = 2
n = 1
n = 1
n = 1
n = 1
n = 1
n = 2
n = 1
n = 1
n = 7
qPCR
TeloTAGGG n = 1
TRF n = 7
n = 5
Cross sectional
Repeated measure n = 2
Within individual n = 8
0.1902
0.1425
0.1111
0.0843
0.1012
0.1161
0.1012
0.1183
0.0135
0.0959
0.0741
0.0957
0.2181
0.0451
0.0993
0.0596
0.1306
0.4651
0.0707
0.1852
0.0067
0.0826
0.0957
0.0088
0.1134
0.0539
0.2693
0.1186
0.1306
0.0909
0.1687
0.0271
0.0984
-0.2717; 0.2965
-0.1663; 0.3454
-0.0157; 0.1628
-0.2950; 0.4591
0.0101; 0.4080
0.0059; 0.3621
-0.0388; 0.1284
-0.0320; 0.3085
0.0185; 0.2019
-0.1183; 0.2800
0.0061; 0.1945
-0.0437; 0.2701
0.0088; 0.1919
-0.0600; 0.0936
[-0.2072; 0.3881]
[-0.1973; 0.3089]
[-0.1232; 0.3684]
[0.0857; 0.7267]
[-0.1322; 0.2678]
[-0.0600; 0.2893]
[-0.1962; 0.2090]
[-0.1686; 0.3238]
[-0.2950; 0.4591]
[-0.1182; 0.1356]
[-0.2110; 0.4153]
[-0.3004; 0.3952]
[0.0252; 0.4831]
[-0.0089; 0.2424]
[-0.1232; 0.3684]
[-0.0409; 0.2197]
[-0.0219; 0.3474]
[-0.1336; 0.1864]
[0.0040; 0.1910]
The meta-regression was performed for the continuous variable publication year and represented as Cochrane’s Q and the associated p = value. Publication dates ranged
from 2014–2021. Publication date was not a significant predictor of effect size (Q = 1.252, p = 0.2632).
https://doi.org/10.1371/journal.pone.0257370.t001
PLOS ONE | https://doi.org/10.1371/journal.pone.0257370 October 1, 2021
7 / 17
PLOS ONE Elevated glucocorticoids and telomere attrition
(Q = 1.86, p = 0.7594). Similarly, the subgroup analysis for “stressor type,” did not reveal a dif-
ference between types of stressors (Table 1). The between study difference was not significantly
different (Q = 2.56, p = 0.2783). Likewise, our subgroup “life history stage,” did not show dif-
ferences between effect sizes for pre- and post-maturation organisms (Table 1), and did not
indicate a difference between groups (Q = 0.06, p = 0.8119). The fourth subgroup analysis,
“GC assay” did not find a difference between plasma GCs and other GC measurements, yield-
ing a non-significant difference between studies (Q = 0.03, p = 0.8742) (Table 1). Additionally,
the between study difference for the telomere assay subgroup did not find a significant differ-
ence between the three telomere quantification methods (Q = 0.12, p = 0.9401; Table 1). Our
sixth subgroup analysis examined potential differences in effect size due to taxa, which could
be divided into the binary categories avian and non-avian (Table 1). There was no difference
between-studies (Q = 0.03, p = 0.8666). Our analysis further explored species-specific differ-
ences and accordingly did not find a significant difference between species (Q = 9.27,
p = 0.6797). Similarly, the final analysis investigated potential differences between study
designs and yielded a non-significant difference between cross-sectional, repeated measures,
or within individual designs (Q = 1.27, p = 0.5289).
Publication bias
We found publication bias against studies with small sample size and small effect size (S3 Fig;
Egger’s test for small sample bias: intercept = 1.420616, CI = 0.3753223; 2.465909,
p = 0.02064949).
Risk of bias in included studies
We represent the results of the risk of bias analysis in Table 2. Four of fifteen studies received a
risk of bias ranking of moderate concern. These studies had some missing values for GCs or
telomere length or selectively reported one time point in the results. The other eleven studies
received a ranking of low risk and accordingly reported nearly all values for physiological
parameters, measured GCs in plasma or saliva, and did not selectively report results.
Table 2. Overall risk of bias assessed based on missing outcome data, measure of outcome and in the selection of reported results.
Author
Year
Bias due to missing outcome data
Bias in measure of outcome
Bias in the selection of reported result
Overall Risk of Bias
Bauch et. al
Burraco et. al
Casagrande et. al
Gil et. al
Grunst et. al
Hau et. al
Herborn et. al
Injaian et. al
Lansade et. al
Lemaitre et. al
Pegan et. al
Sebastiano et. al
Stier et. al
Watson et. al
Young et. al
2016
2019
2020
2019
2020
2015
2014
2019
2018
2021
2019
2017
2020
2016
2017
high
low
high
low
low
low
low
high
low
low
high
low
low
low
moderate
https://doi.org/10.1371/journal.pone.0257370.t002
low
low
low
low
high
low
low
low
low
moderate
low
low
low
low
low
high
low
low
low
low
low
low
high
low
low
low
low
low
low
high
some concern
low risk
some concern
low risk
low risk
low risk
low risk
high risk
low risk
low risk
some concern
low risk
low risk
low risk
some concern
PLOS ONE | https://doi.org/10.1371/journal.pone.0257370 October 1, 2021
8 / 17
PLOS ONE Elevated glucocorticoids and telomere attrition
Discussion
External and internal stimuli can activate the neuroendocrine stress response in vertebrates,
resulting in the secretion of GCs, which induces multiple downstream physiological and
behavioral effects [8, 9]. GCs might directly or indirectly cause telomere erosion [1, 11, 32].
Therefore, our goal was to investigate the relationship between GCs and telomere length via
meta-analysis using data from empirical studies. Though our sample size was limited (n = 15),
our data do not support the hypothesis that elevated GC levels result in telomere shortening.
The empirical evidence for a relationship between GCs and telomere length is mixed, with
some studies showing that telomere shortening is directly related to GC levels, and other stud-
ies finding no relationship. For example, GCs influence telomere dynamics in wild roe deer
and great tits [32, 39], but not in red squirrels or magellanic penguins [46, 47]. These results
suggest that the relationship between GCs and telomere length is species-specific. Alterna-
tively, a potential relationship may be obscured by the methods used to measure GCs and telo-
mere length or by differences in experimental design including time frame. A differential
sensitivity of the HPA axis can also obscure conclusions made from GC measurements espe-
cially in free-ranging vertebrates that can potentially encounter a variety of external stimuli
[1]. For example, since GC levels in plasma remain elevated for several minutes after a stressor
subsides, it can be challenging to assess whether a measured GC increase results from the
stressor in question, the stress involved in obtaining a sample from the experimental subject,
or an unrelated event triggering HPA axis activation [6, 56]. As baseline plasma GC samples
must be collected quickly in many species, it can be logistically difficult to attain a true baseline
GC value in the field [57–60]. GCs can also be incorporated into other matrixes such as saliva,
feathers, and hair [4, 58]. The multitude of non-invasive GC sampling sources is advantageous
to conservation physiology as their quantification does not require capture [6]. However,
across tissues and fluids, the time required for GC incorporation varies. For example, eleva-
tions in plasma GCs can be detected within minutes of stressor exposure, whereas GCs inte-
grate into hair a week or more after stressor exposure [4]. Hence, there are caveats in the
interpretation of each measurement such as incongruencies between GC levels in plasma and
other tissues, hair and saliva [60]. Therefore, GC measurements in feces may be more repre-
sentative of accumulated stress, rather than the event in question [6].
GC quantification in tissues and feces can also present specific uncertainty and imprecision
during sampling, storage, and extraction. In fecal samples, GC metabolites can increase up to
92% in 120 days and provide an inaccurate assessment of GC levels [61, 62]. Excrement not
collected immediately or across different time scales can obscure potential differences since
exposure to abiotic factors like rainfall or extreme temperature can alter the concentration of
fecal glucocorticoid metabolites [63]. Moreover, diet can affect GC metabolites in fecal sam-
ples, since an increased amount of cellulose depresses fecal glucocorticoid metabolite concen-
trations [61]. Similarly, feather preparation and extraction can also affect GC levels [64].
Further, different parts of the feather yield different concentrations of GCs. Saliva based GC
extraction and quantification hosts similar shortcomings, though salivary GCs increase on a
similar timeline (5–10 minutes) to circulating plasma GCs and thus prove a close proxy for
plasma GC quantification [65]. Other factors such as time since last meal and recent activity
also impact salivary GC measurement [66].
Similar considerations must be taken into account when assessing telomere length. Since
telomere length can be influenced by environmental, maternal, and epigenetic effects, there is
a large inter-individual variability in telomere dynamics [11, 67]. Several factors may contrib-
ute to this variability including discrepancies between the repeatability of different telomere
measurement assays. Seven studies included in our meta-analysis utilized the telomere
PLOS ONE | https://doi.org/10.1371/journal.pone.0257370 October 1, 2021
9 / 17
PLOS ONE Elevated glucocorticoids and telomere attrition
restriction fragment (TRF) assay, which depends on the distribution of the terminal restriction
fragments to average the length of telomeres in a given cell population [68]. The other eight
studies used the quantitative polymerase chain reaction (qPCR, n = 7), which relies on the
quantification of the highly conserved (TTAGGG)n sequence for a Southern blot variation
(TeloTAGGG for telomere quantification (n = 1) [69]. TRF-based studies are highly repeatable
within individuals, whereas qPCR based studies are less repeatable and more variable than
TRF because they are more prone to measurement errors [70]. qPCR can also bias measure-
ments of telomere length because some species that exhibit interstitial telomeric repeats will
artificially enlarge telomere length [71, 72]. In addition to methodological differences, there is
large individual variability in telomere length based on tissue type [73]. In adult zebra finches,
telomere length in red blood cells is correlated with telomere length in the spleen, liver and
brain, but not muscle or heart [31]. While avian studies in our meta-analysis used red blood
cells for telomere measurement, telomere length was measured in tail muscle and liver in
mammals and amphibians, which could lead to discrepancies when comparing among studies
[31, 46, 57].
A variety of biological factors also contribute to the diversity of telomere dynamics observed
within a study and the large amount of observed inter-individual variability. The rate of telo-
mere shortening can be influenced by the life histories and environmental conditions [22]. In
accordance with the metabolic telomere attrition hypothesis, shortening is exacerbated by life
history stages requiring more energy, such as reproduction [32]. Within an energy intensive
process like reproduction, there can be a large inter-individual variability related to reproduc-
tive effort, which can be attributed to brood size and food availability [74]. Differences in
reproductive roles during the breeding season account for sex-specific telomere dynamics
which can contribute to differences in the variability of telomere dynamics within a study [75].
Finally, individuals respond differently to environmental challenges which can act synergisti-
cally with rapid growth or energy intensive life stages to magnify the rate of telomere shorten-
ing in non-model vertebrates [71].
Telomere dynamics can be complicated by the presence of telomerase which in some cases
can elongate telomeres [22, 76]. Typically, telomerase exhibits higher activity in developing
organisms as compared to adults [77]. Ectotherms such as amphibians and reptiles have telo-
merase that is active throughout adulthood while endotherms reduce telomerase expression
almost to non-detectable levels as they reach maturity [11, 70]. However, there is conflicting
evidence on these observations, as telomerase activity has been detected in adult common
terns and European Storm Petrels among other species [78, 79]. Nonetheless, adult telomere
shortening is observed in chickens, which have active telomerase in the adult life stage [26].
While there is an absence of empirical evidence on the long-term activity of telomerase in
many avian species, even adults exhibit general shortening trends [76].
Many factors influence GC and telomere measurements. During the subgroup analysis, we
attempted to disentangle the underlying causes of the variation in effect size. Ultimately, we
found no impact of stressor, taxa, type of GC assay, or life history stage on the heterogeneity of
the effect size. While no subgroup was identified as a predictor of heterogeneity in effect size,
pooled effect sizes in certain categories with the subgroup indicate a higher pooled effect size
than the overall pooled effect size. The small sample size for some parameters precluded fur-
ther statistical analysis, however, we found variables of interest that may play a large role in the
relationship between GCs and telomere length. For example, within “experimental time-
frame,” (n = 4) the group of studies with a timeframe above four weeks had a pooled effect size
of 0.2181, while all other groups’ pooled effect size was less than that of the overall pooled effect
size. Since most studies took place in less than four weeks, this suggests that while almost
immediate changes in GCs can be observed, the impact of GCs on telomere length cannot be
PLOS ONE | https://doi.org/10.1371/journal.pone.0257370 October 1, 2021
10 / 17
PLOS ONE Elevated glucocorticoids and telomere attrition
seen on short time scales. This idea is consistent with typical responses of telomere shortening
observed in studies that take place for more than a year [29, 54, 79–81]. More work is needed
to explore if long-term rather than short-term studies can be used to tease apart parameters
that underlie the connection between GCs and telomere length such as stressor type or
duration.
While GC secretion is often viewed as the endpoint of HPA axis activation in response to
external stimuli, GC manipulation is an oversimplification of the stress response which
involves a multitude of physiological mechanisms that can each impact energy allocation and
promote telomere erosion [8]. This highlights the problematic nature of the category “GC
stress” which was investigated as a category during the subgroup analysis, in which studies
subjected organisms to GC manipulation via an implant or oral administration. Since previous
research found that organismal stress can result in adverse physiological responses without the
involvement of the HPA axis, these results underscore the issue of using only GCs as a proxy
for stress [82, 83].
Overall, we found no relationship between GCs and telomere length across studies. Cur-
rently, the existing literature shows both a direct relationship and a lack of a relationship
between GCs and telomere dynamics, suggesting that the underlying mechanisms driving this
relationship are species-specific or altered by differences in experimental design. However,
due to limited sample size, we are unable to investigate the underlying variables that play a role
in this relationship. Here, we highlight the need for more studies with targeted experimental
parameters to understand how conditions, such as experimental timeframes, stressor(s), and
stressor magnitudes can drive a potential relationship between the neuroendocrine stress
response and cellular aging. Thus, we recommend the following research priorities to groups
studying similar questions.
1. Experimental timeframes and stressor magnitudes should be long enough to observe telo-
mere erosion in relation to stressors when studying GCs.
2. When possible, studies should use a repeated measures design to measure cortisol levels
and telomere lengths before and after stress exposure to account for individual variation.
3. While the avian taxa are well represented in this research topic, there is a dearth of informa-
tion on other taxa. It will be important to investigate the neuroendocrine stress response in
other vertebrates including mammals and reptiles to understand if similar principles hold
true in these taxa or if telomere dynamics differ across taxa.
4. If possible, future research should assess the functionality of the study organisms’ HPA axis
by ACTH/dexamethasone challenge prior to exposure to a stressor and completion of the
study.
Certainty of evidence
We utilized the applicable Cochrane/GRADE categories “risk of bias,” “inconsistency,” and
“publication bias,” for the determination of the certainty of evidence. Overall, we have a mod-
erate confidence in the certainty of evidence. While most studies received a low risk of bias
assessment, and had low heterogeneity, we report a considerable amount of publication bias as
evidenced by Egger’s test and an asymmetrical funnel plot.
Supporting information
S1 Checklist. PRISMA 2020 checklist.
(PDF)
PLOS ONE | https://doi.org/10.1371/journal.pone.0257370 October 1, 2021
11 / 17
PLOS ONE Elevated glucocorticoids and telomere attrition
S1 Fig. Influence analysis plot. The leave one out recalculation reveals a similar effect size
across studies and indicates that studies evenly contribute to the pooled effect size.
(TIF)
S2 Fig. Baujat plot. Studies can have an unequal influence on the pooled effect size and con-
tribute to the heterogeneity of effect sizes. The horizontal axis represents Cochrane’s Q and
influence on the pooled effect size on the vertical axis.
(TIF)
S3 Fig. Funnel plot. The lack of studies in the bottom left of the “funnel” demonstrates publi-
cation bias against studies with small sample sizes and small effect sizes.
(TIF)
S1 Table. Search strategy table. Details search term combinations used to search online data-
bases and websites.
(XLSX)
Acknowledgments
We would like to thank all the researchers who made their data freely available for this study.
In particular, we thank Christine Bauch (University of Groningen), Stefania Casagrande (Max
Planck Institute for Ornithology), Britt Heidinger (North Dakota State University), Marie-
Pierre Moisan (French National Institute for Agriculture, Food, and Environment), Teresa
Pegan (University of Michigan), Manrico Sebastiano (French National Centre for Scientific
Research), Mathilde Tissier (Bishop’s University), and Hannah Watson (Lund University).
Author Contributions
Conceptualization: Lauren Zane, David C. Ensminger, Jose´ Pablo Va´zquez-Medina.
Data curation: Lauren Zane.
Formal analysis: Lauren Zane.
Funding acquisition: Jose´ Pablo Va´zquez-Medina.
Investigation: Lauren Zane.
Methodology: Lauren Zane, David C. Ensminger.
Project administration: David C. Ensminger.
Resources: David C. Ensminger.
Software: Lauren Zane.
Supervision: David C. Ensminger, Jose´ Pablo Va´zquez-Medina.
Validation: David C. Ensminger, Jose´ Pablo Va´zquez-Medina.
Visualization: Lauren Zane.
Writing – original draft: Lauren Zane.
Writing – review & editing: David C. Ensminger, Jose´ Pablo Va´zquez-Medina.
References
1. Haussmann MF, Marchetto NM. Telomeres: Linking stress and survival, ecology and evolution. Current
Zoology. 2010; 56(6):714–27.
PLOS ONE | https://doi.org/10.1371/journal.pone.0257370 October 1, 2021
12 / 17
PLOS ONE Elevated glucocorticoids and telomere attrition
2. Wingfield JC, Sapolsky RM. Reproduction and Resistance to Stress: When and How: Reproduction
and resistance to stress. Journal of Neuroendocrinology. 2003; 1; 15(8):711–24. https://doi.org/10.
1046/j.1365-2826.2003.01033.x PMID: 12834431
3. Koren L, Whiteside D, Fahlman Å, Ruckstuhl K, Kutz S, Checkley S, et al. Cortisol and corticosterone
independence in cortisol-dominant wildlife. General and Comparative Endocrinology. 2012 May; 177
(1):113–9. https://doi.org/10.1016/j.ygcen.2012.02.020 PMID: 22449618
4. Gormally BMG, Romero LM. What are you actually measuring? A review of techniques that integrate
the stress response on distinct time-scales. Functional Ecology. 2020; 34(10):2030–44.
5. MacDougall-Shackleton SA, Bonier F, Romero LM, Moore IT. Glucocorticoids and “Stress” Are Not
Synonymous. Integrative Organismal Biology. 2019; 1(1).
6. Sheriff MJ, Dantzer B, Delehanty B, Palme R, Boonstra R. Measuring stress in wildlife: techniques for
quantifying glucocorticoids. Oecologia. 2011; 166(4):869–87. https://doi.org/10.1007/s00442-011-
1943-y PMID: 21344254
7. Arnold JM, Oswald SA, Voigt CC, Palme R, Braasch A, Bauch C, et al. Taking the stress out of blood
collection: comparison of field blood-sampling techniques for analysis of baseline corticosterone. Jour-
nal of Avian Biology. 2008; 39(5):588–92.
8. Angelier F, Wingfield JC. Importance of the glucocorticoid stress response in a changing world: Theory,
hypotheses and perspectives. General and Comparative Endocrinology. 2013; 190:118–28. https://doi.
org/10.1016/j.ygcen.2013.05.022 PMID: 23770214
9. Ricklefs RE, Wikelski M. The physiology/life-history nexus. Trends in Ecology & Evolution. 2002; 17
(10):462–8.
10. Young AJ. The role of telomeres in the mechanisms and evolution of life-history trade-offs and ageing.
Philosophical Transactions of the Royal Society B. 2018; 373(1741):20160452. https://doi.org/10.1098/
rstb.2016.0452 PMID: 29335379
11. Angelier F, Costantini D, Ble´ vin P, Chastel O. Do glucocorticoids mediate the link between environmen-
tal conditions and telomere dynamics in wild vertebrates? A review. General and Comparative Endocri-
nology. 2018; 256:99–111. https://doi.org/10.1016/j.ygcen.2017.07.007 PMID: 28705731
12. Campisi J, Andersen JK, Kapahi P, Melov S. Cellular senescence: A link between cancer and age-
related degenerative disease? Seminars in Cancer Biology. 2011; https://doi.org/10.1016/j.semcancer.
2011.09.001 PMID: 21925603
13. Donate LE, Blasco MA. Telomeres in cancer and ageing. Philosophical Transactions of the Royal Soci-
ety B. 2011; 366(1561):76–84. https://doi.org/10.1098/rstb.2010.0291 PMID: 21115533
14.
Levy MZ, Allsopp RC, Futcher AB, Greider CW, Harley CB. Telomere end-replication problem and cell
aging. Journal of Molecular Biology. 1992; 225(4):951–60. https://doi.org/10.1016/0022-2836(92)
90096-3 PMID: 1613801
15. Monaghan P. Organismal stress, telomeres and life histories. Journal of Experimental Biology. 2014;
217(1):57–66.
16. Sebastiano M, Eens M, Angelier F, Pineau K, Chastel O, Costantini D. Corticosterone, inflammation,
immune status and telomere length in frigatebird nestlings facing a severe herpesvirus infection. Con-
servation Physiology. 2017; 5(1):cow073. https://doi.org/10.1093/conphys/cow073 PMID: 28070333
17. Chatelain M, Drobniak SM, Szulkin M. The association between stressors and telomeres in non-human
vertebrates: a meta-analysis. Ecology Letters. 2020; 23(2):381–98. https://doi.org/10.1111/ele.13426
PMID: 31773847
18. Belmaker A, Hallinger KK, Glynn RA, Winkler DW, Haussmann MF. The environmental and genetic
determinants of chick telomere length in Tree Swallows (Tachycineta bicolor). Ecology and Evolution.
2019; 9(14):8175–86. https://doi.org/10.1002/ece3.5386 PMID: 31380080
19. Reichert S, Stier A, Zahn S, Arrive´ M, Bize P, Massemin S, et al. Increased brood size leads to persis-
tent eroded telomeres. Frontiers in Ecology and Evolution [Internet]. 2014; 2:9.
20. Eastwood JR, Hall ML, Teunissen N, Kingma SA, Hidalgo Aranzamendi N, Fan M, et al. Early-life telo-
mere length predicts lifespan and lifetime reproductive success in a wild bird. Molecular Ecology. 2019;
28(5):1127–37. https://doi.org/10.1111/mec.15002 PMID: 30592345
21. Heidinger BJ, Blount JD, Boner W, Griffiths K, Metcalfe NB, Monaghan P. Telomere length in early life
predicts lifespan. Proceedings of the National = Academy of Sciences. 2012; 109(5):1743–8. https://
doi.org/10.1073/pnas.1113306109 PMID: 22232671
22. Stier A, Metcalfe NB, Monaghan P. Pace and stability of embryonic development affect telomere
dynamics: an experimental study in a precocial bird model. Proceedings of the Royal Society B. 2020;
287(1933):20201378. https://doi.org/10.1098/rspb.2020.1378 PMID: 32842933
23. Casagrande S, Hau M. Telomere attrition: metabolic regulation and signalling function? Biol Lett. 2019;
15(3):20180885. https://doi.org/10.1098/rsbl.2018.0885 PMID: 30890069
PLOS ONE | https://doi.org/10.1371/journal.pone.0257370 October 1, 2021
13 / 17
PLOS ONE Elevated glucocorticoids and telomere attrition
24. Reichert S, Stier A. Does oxidative stress shorten telomeres in vivo? A review. Biol Lett. 2017; 13
(12):20170463. https://doi.org/10.1098/rsbl.2017.0463 PMID: 29212750
25. Costantini D, Marasco V, Møller AP. A meta-analysis of glucocorticoids as modulators of oxidative
stress in vertebrates. Journal of Comparative Physiology B. 2011; 181, 447–456. https://doi.org/10.
1007/s00360-011-0566-2 PMID: 21416253
26. Haussmann MF, Longenecker AS, Marchetto NM, Juliano SA, Bowden RM. Embryonic exposure to
corticosterone modifies the juvenile stress response, oxidative stress and telomere length. Proceedings
of the Royal Society B. 2012; 279(1732):1447–56. https://doi.org/10.1098/rspb.2011.1913 PMID:
22072607
27. Choi J, Fauce SR, Effros RB. Reduced telomerase activity in human T lymphocytes exposed to cortisol.
Brain, Behavior, and Immunity. 2008; 22(4):600–5. https://doi.org/10.1016/j.bbi.2007.12.004 PMID:
18222063
28. Wei YB, Backlund L, Wegener G, Mathe AA, Lavebratt C. Telomerase Dysregulation in the Hippocam-
pus of a Rat Model of Depression: Normalization by Lithium. International Journal of Neuropsychophar-
macology. 2015; 18(7):pyv002–pyv002. https://doi.org/10.1093/ijnp/pyv002 PMID: 25618407
29. Bauch C, Riechert J, Verhulst S, Becker PH. Telomere length reflects reproductive effort indicated by
corticosterone levels in a long-lived seabird. Molecular Ecology. 2016; 25(22):5785–94. https://doi.org/
10.1111/mec.13874 PMID: 27696588
30. Burraco P, Valde´ s AE, Orizaola G. Metabolic costs of altered growth trajectories across life transitions
in amphibians. Journal of Animal Ecology. 2020; 89(3):855–66. https://doi.org/10.1111/1365-2656.
13138 PMID: 31693168
31. Casagrande S, Stier A, Monaghan P, Loveland JL, Boner W, Lupi S, et al. Increased glucocorticoid con-
centrations in early life cause mitochondrial inefficiency and short telomeres. Journal of Experimental
Biology. 2020; 223(15):jeb222513. https://doi.org/10.1242/jeb.222513 PMID: 32532864
32. Gil D, Alfonso-Iñiguez S, Pe´ rez-Rodrı´guez L, Muriel J, Monclu´ s R. Harsh conditions during early devel-
opment influence telomere length in an altricial passerine: Links with oxidative stress and corticoste-
roids. Journal of Evolutionary Biology. 2019; 32(1):111–25. https://doi.org/10.1111/jeb.13396 PMID:
30387533
33. Grunst ML, Raap T, Grunst AS, Pinxten R, Parenteau C, Angelier F, et al. Early-life exposure to artificial
light at night elevates physiological stress in free-living songbirds. Environmental Pollution. 2020;
259:113895. https://doi.org/10.1016/j.envpol.2019.113895 PMID: 31926393
34. Hau M, Haussmann MF, Greives TJ, Matlack C, Costantini D, Quetting M, et al. Repeated stressors in
adulthood increase the rate of biological ageing. Frontiers in Zoology. 2015; 12(1):4. https://doi.org/10.
1186/s12983-015-0095-z PMID: 25705242
35. Herborn KA, Heidinger BJ, Boner W, Noguera JC, Adam A, Daunt F, et al. Stress exposure in early
post-natal life reduces telomere length: an experimental demonstration in a long-lived seabird. Proceed-
ings of the Royal Society B. 2014; 281(1782):20133151. https://doi.org/10.1098/rspb.2013.3151 PMID:
24648221
36.
37.
38.
Injaian AS, Gonzalez-Gomez PL, Taff CC, Bird AK, Ziur AD, Patricelli GL, et al. Traffic noise exposure
alters nestling physiology and telomere attrition through direct, but not maternal, effects in a free-living
bird. General and Comparative Endocrinology. 2019; 276:14–21. https://doi.org/10.1016/j.ygcen.2019.
02.017 PMID: 30796896
Lansade L, Foury A, Reigner F, Vidament M, Guettier E, Bouvet G, et al. Progressive habituation to sep-
aration alleviates the negative effects of weaning in the mother and foal. Psychoneuroendocrinology.
2018; 97:59–68. https://doi.org/10.1016/j.psyneuen.2018.07.005 PMID: 30005282
Lemaıˆtre J-F, Carbillet J, Rey B, Palme R, Froy H, Wilbourn RV, et al. Short-term telomere dynamics is
associated with glucocorticoid levels in wild populations of roe deer. Comparative Biochemistry and
Physiology Part A: Molecular & Integrative Physiology. 2021; 252:110836. https://doi.org/10.1016/j.
cbpa.2020.110836 PMID: 33144154
39. Pegan TM, Winkler DW, Haussmann MF, Vitousek MN. Brief Increases in Corticosterone Affect Mor-
phology, Stress Responses, and Telomere Length but Not Postfledging Movements in a Wild Songbird.
Physiological and Biochemical Zoology. 2019; 92(3):274–85. https://doi.org/10.1086/702827 PMID:
30840539
40.
Tissier ML, Williams TD, Criscuolo F. Maternal Effects Underlie Ageing Costs of Growth in the Zebra
Finch (Taeniopygia guttata). White SA, editor. PLoS ONE. 2014; 9(5):e97705. https://doi.org/10.1371/
journal.pone.0097705 PMID: 24828412
41. Watson H, Bolton M, Heidinger BJ, Boner W, Monaghan P. Assessing the effects of repeated handling
on the physiology and condition of semi-precocial nestlings. IBIS International Journal of Avian Science.
2016; 158(4):834–43. https://doi.org/10.1111/ibi.12402 PMID: 27708454
PLOS ONE | https://doi.org/10.1371/journal.pone.0257370 October 1, 2021
14 / 17
PLOS ONE Elevated glucocorticoids and telomere attrition
42. Young R, Orben R, Will A, Kitaysky A. Relationship between telomere dynamics and movement and
behavior during winter in the thick-billed murre. Marine Ecological Progress Series. 2017; 578:253–61.
43. Cerchiara JA, Risques RA, Prunkard D, Smith JR, Kane OJ, Boersma PD. Magellanic penguin telo-
meres do not shorten with age with increased reproductive effort, investment, and basal corticosterone.
Ecological Evolution. 2017; 7(15):5682–91. https://doi.org/10.1002/ece3.3128 PMID: 28811878
44. Dantzer B, van Kesteren F, Westrick SE, Boutin S, McAdam AG, Lane JE, et al. Maternal glucocorti-
coids promote offspring growth without inducing oxidative stress or shortening telomeres in wild red
squirrels. Journal of Experimental Biology. 2020; 223(1):jeb212373. https://doi.org/10.1242/jeb.212373
PMID: 31796605
45. Drury SS, Howell BR, Jones C, Esteves K, Morin E, Schlesinger R, et al. Shaping long-term primate
development: Telomere length trajectory as an indicator of early maternal maltreatment and predictor of
future physiologic regulation. Developmental Psychopathology. 2017; 29(5):1539–51. https://doi.org/
10.1017/S0954579417001225 PMID: 29162166
46. Gangoso L, Lambertucci SA, Cabezas S, Alarco´ n PAE, Wiemeyer GM, Sanchez-Zapata JA, et al. Sex-
dependent spatial structure of telomere length in a wild long-lived scavenger. Ecosphere. 2016; 7(10):
e01544.
47. Noguera JC, Velando A. Reduced telomere length in embryos exposed to predator cues. Journal of
Experimental Biology. 2019; 222(24):jeb216176. https://doi.org/10.1242/jeb.216176 PMID: 31796604
48. Noreikiene K, O¨ st M, Seltmann MW, Boner W, Monaghan P, Jaatinen K. Nest cover and faecal gluco-
corticoid metabolites are linked to hatching success and telomere length in breeding Common Eiders
(Somateria mollissima). Canadian Journal of Zoology. 2017; 95(9):695–703.
49. Ouyang JQ, Lendvai A´ Z, Moore IT, Bonier F, Haussmann MF. Do Hormones, Telomere Lengths, and
Oxidative Stress form an Integrated Phenotype? A Case Study in Free-Living Tree Swallows. Integra-
tive and Comparative Biology. 2016; 56(2):138–45. https://doi.org/10.1093/icb/icw044 PMID: 27252220
50. Quirici V, Guerrero CJ, Krause JS, Wingfield JC, Va´ squez RA. The relationship of telomere length to
baseline corticosterone levels in nestlings of an altricial passerine bird in natural populations. Frontiers
in Zoology 2016; 13(1):1. 22.
51. Schultner J, Moe B, Chastel O, Bech C, Kitaysky AS. Migration and stress during reproduction govern
telomere dynamics in a seabird. Biology Letters. 2014; 10(1):20130889. https://doi.org/10.1098/rsbl.
2013.0889 PMID: 24429681
52. Shen Q, Wu J, Ni Y, Xie X, Yu C, Xiao Q, et al. Exposure to jet lag aggravates depression-like behaviors
and age-related phenotypes in rats subject to chronic corticosterone. Acta Biochimica et Biophysica
Sinica. 2019; 51(8):834–44. https://doi.org/10.1093/abbs/gmz070 PMID: 31314053
53. Stevenson JR, McMahon EK, Boner W, Haussmann MF. Oxytocin administration prevents cellular
aging caused by social isolation. Psychoneuroendocrinology. 2019; 103:52–60. https://doi.org/10.1016/
j.psyneuen.2019.01.006 PMID: 30640038
54. Xie X, Shen Q, Ma L, Chen Y, Zhao B, Fu Z. Chronic corticosterone-induced depression mediates pre-
mature aging in rats. Journal of Affective Disorders; 229:254–61. https://doi.org/10.1016/j.jad.2017.12.
073 PMID: 29329057
55. Young RC, Welcker J, Barger CP, Hatch SA, Merkling T, Kitaiskaia EV, et al. Effects of developmental
conditions on growth, stress and telomeres in black-legged kittiwake chicks. Molecular Ecology. 2017;
26(13):3572–84. https://doi.org/10.1111/mec.14121 PMID: 28370751
56. Romero LM, Meister CJ, Cyr NE, Kenagy GJ, Wingfield JC. Seasonal glucocorticoid responses to cap-
ture in wild free-living mammals. American Journal of Physiology-Regulatory, Integrative and Compara-
tive Physiology. 2008; 294(2):R614–22. https://doi.org/10.1152/ajpregu.00752.2007 PMID: 18094060
57.
Tylan C, Camacho K, French S, Graham SP, Herr MW, Jones J, et al. Obtaining plasma to measure
baseline corticosterone concentrations in reptiles: How quick is quick enough? General and Compara-
tive Endocrinology. 2020; 287:113324. https://doi.org/10.1016/j.ygcen.2019.113324 PMID: 31733208
58. Dantzer B, Fletcher QE, Boonstra R, Sheriff MJ. Measures of physiological stress: a transparent or
opaque window into the status, management and conservation of species? Conservation Physiology.
2014; 2(1). https://doi.org/10.1093/conphys/cou023 PMID: 27293644
59. Reeder DM, Kramer KM. STRESS IN FREE-RANGING MAMMALS: INTEGRATING PHYSIOLOGY,
ECOLOGY, AND NATURAL HISTORY. Journal of Mammalogy. 2005; 86(2):225–35.
60. Keckeis K, Lepschy M, Scho¨ pper H, Moser L, Troxler J, Palme R. Hair cortisol: a parameter of chronic
stress? Insights from a radiometabolism study in guinea pigs. J Comp Physiol B. 2012; 182(7):985–96.
https://doi.org/10.1007/s00360-012-0674-7 PMID: 22592890
61. Goymann W. "Noninvasive Monitoring of Hormones in Bird Droppings: Physiological Validation, Sam-
pling, Extraction, Sex Differences, and the Influence of Diet on Hormone Metabolite Levels." Annals of
the New York Academy of Sciences (2005); 1046(1): 35–53.
PLOS ONE | https://doi.org/10.1371/journal.pone.0257370 October 1, 2021
15 / 17
PLOS ONE Elevated glucocorticoids and telomere attrition
62. Khan M. Z., Altmann J., Isani S. S. and Yu J. "A matter of time: evaluating the storage of fecal samples
for steroid analysis." General and Comparative Endocrinology (2002); 128(1): 57–64. https://doi.org/
10.1016/s0016-6480(02)00063-1 PMID: 12270788
63. MILLSPAUGH J.J. & WASHBURN B.E. Use of fecal glucocorticoid metabolite measures in conserva-
tion biology research: considerations for application and interpretation. General Comparative Endocri-
nology (2004); 138: 189–199. https://doi.org/10.1016/j.ygcen.2004.07.002 PMID: 15364201
64.
Freeman N. E. and Newman A. E. M. "Quantifying corticosterone in feathers: validations for an emerg-
ing technique." Conservation Physiology (2018); 6: 9.
65. Hammond TT, Au ZA, Hartman AC, Richards-Zawacki CL. Assay validation and interspecific compari-
son of salivary glucocorticoids in three amphibian species. Conservation Physiology. (2018); 6(1).
https://doi.org/10.1093/conphys/coy055 PMID: 30279992
66. Garde AH, Persson R, Hansen Å. M, O¨ sterberg K, Ørbæk P, Eek F, et al. Effects of lifestyle factors on
concentrations of salivary cortisol in healthy individuals. Scandinavian Journal of Clinical Lab Investiga-
tions (2009); 69(2): 242–250. https://doi.org/10.1080/00365510802483708 PMID: 18985537
67. Shalev I, Entringer S, Wadhwa PD, Wolkowitz OM, Puterman E, Lin J, et al. Stress and telomere biol-
ogy: A lifespan perspective. Psychoneuroendocrinology. 2013; 38(9):1835–42. https://doi.org/10.1016/
j.psyneuen.2013.03.010 PMID: 23639252
68. Mender I, Shay J. Telomere Restriction Fragment (TRF) Analysis. BIO-PROTOCOL. 2015; 5(22):
e1658. https://doi.org/10.21769/bioprotoc.1658 PMID: 27500189
69. Gomes NMV, Shay JW, Wright WE. Telomere biology in Metazoa. FEBS Letters. 2010; 584(17):3741–
51. https://doi.org/10.1016/j.febslet.2010.07.031 PMID: 20655915
70. Ka¨ rkka¨inen T, Briga M, Laaksonen T, Stier A. Within-individual repeatability in telomere length: a meta-
analysis in non-mammalian vertebrates. Preprints; 2021. https://doi.org/10.1111/mec.16155 PMID:
34455645
71. Rovatsos M, Kratochvı´l L, Altmanova´ M, Johnson Pokorna´ M. Interstitial Telomeric Motifs in Squamate
Reptiles: When the Exceptions Outnumber the Rule. Stanyon R, editor. PLoS ONE. 2015; 10(8):
e0134985. https://doi.org/10.1371/journal.pone.0134985 PMID: 26252002
72. Haussmann MF, Mauck RA. TECHNICAL ADVANCES: New strategies for telomere-based age estima-
tion: TECHNICAL ADVANCES. Molecular Ecology Resources. 2008; 8(2):264–74. https://doi.org/10.
1111/j.1471-8286.2007.01973.x PMID: 21585768
73. Reichert S, Criscuolo F, Verinaud E, Zahn S, Massemin S. Telomere Length Correlations among
Somatic Tissues in Adult Zebra Finches. Moreno E, editor. PLoS ONE. 2013; 8(12):e81496. https://doi.
org/10.1371/journal.pone.0081496 PMID: 24349076
74. Sudyka J, Arct A, Drobniak S, Dubiec A, Gustafsson L, Cichoń M. Experimentally increased reproduc-
tive effort alters telomere length in the blue tit (Cyanistes caeruleus). Journal of Evolutionary Biology.
2014; 27(10):2258–64. https://doi.org/10.1111/jeb.12479 PMID: 25228433
75. Parolini M, Romano A, Khoriauli L, Nergadze SG, Caprioli M, Rubolini D, et al. Early-Life Telomere
Dynamics Differ between the Sexes and Predict Growth in the Barn Swallow (Hirundo rustica). Lustig
AJ, editor. PLoS ONE. 2015; 10(11):e0142530. https://doi.org/10.1371/journal.pone.0142530 PMID:
26565632
76.
Taylor HA, Delany ME. Ontogeny of telomerase in chicken: Impact of downregulation on pre- and post-
natal telomere length in vivo. Development, Growth, and Differentiation. 2000; 42(6):613–21. https://
doi.org/10.1046/j.1440-169x.2000.00540.x PMID: 11142683
77. Gomes NMV, Ryder OA, Houck ML, Charter SJ, Walker W, Forsyth NR, et al. Comparative biology of
mammalian telomeres: hypotheses on ancestral states and the roles of telomeres in longevity determi-
nation: The comparative biology of mammalian telomeres. Aging Cell. 2011; 10(5):761–8. https://doi.
org/10.1111/j.1474-9726.2011.00718.x PMID: 21518243
78. Haussmann MF, Winkler DW, Huntington CE, Nisbet ICT, Vleck CM. Telomerase activity is maintained
throughout the lifespan of long-lived birds. Experimental Gerontology. 2007; 42(7):610–8. https://doi.
org/10.1016/j.exger.2007.03.004 PMID: 17470387
79. Angelier F, Vleck CM, Holberton RL, Marra PP. Telomere length, non-breeding habitat and return rate
in male American redstarts. Funct Ecol.2013; 27(2):342–50.
80. Asghar M, Hasselquist D, Hansson B, Zehtindjiev P, Westerdahl H, Bensch S. Hidden costs of infection:
Chronic malaria accelerates telomere degradation and senescence in wild birds. Science. 2015; 347
(6220):436–8. https://doi.org/10.1126/science.1261121 PMID: 25613889
81. Barrett ELB, Burke TA, Hammers M, Komdeur J, Richardson DS. Telomere length and dynamics pre-
dict mortality in a wild longitudinal study. Molecular Ecology. 2013; 22(1):249–59. https://doi.org/10.
1111/mec.12110 PMID: 23167566
PLOS ONE | https://doi.org/10.1371/journal.pone.0257370 October 1, 2021
16 / 17
PLOS ONE Elevated glucocorticoids and telomere attrition
82. Breuner CW, Delehanty B, Boonstra R. Evaluating stress in natural populations of vertebrates: total
CORT is not good enough. Functional Ecology. 2013; 27(1):24–36.
83. Gabor CR, Knutie SA, Roznik EA, Rohr JR. Are the adverse effects of stressors on amphibians medi-
ated by their effects on stress hormones? Oecologia. 2018; 186(2):393–404. https://doi.org/10.1007/
s00442-017-4020-3 PMID: 29222721
PLOS ONE | https://doi.org/10.1371/journal.pone.0257370 October 1, 2021
17 / 17
PLOS ONE |
10.1371_journal.pone.0258711 | RESEARCH ARTICLE
Single nucleotide polymorphisms and copy-
number variations in the Trypanosoma brucei
repeat (TBR) sequence can be used to
enhance amplification and genotyping of
Trypanozoon strains
Nick Van ReetID
Stijn DeborggraeveID
1
1¤, Philippe Bu¨ scherID
1*, Pati Patient Pyana2, Sara DehouID
1, Nicolas BebronneID
1,
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: Van Reet N, Patient Pyana P, Dehou S,
Bebronne N, Deborggraeve S, Bu¨scher P (2021)
Single nucleotide polymorphisms and copy-
number variations in the Trypanosoma brucei
repeat (TBR) sequence can be used to enhance
amplification and genotyping of Trypanozoon
strains. PLoS ONE 16(10): e0258711. https://doi.
org/10.1371/journal.pone.0258711
Editor: Maria Stefania Latrofa, University of Bari,
ITALY
Received: June 10, 2021
Accepted: October 4, 2021
Published: October 25, 2021
Peer Review History: PLOS recognizes the
benefits of transparency in the peer review
process; therefore, we enable the publication of
all of the content of peer review and author
responses alongside final, published articles. The
editorial history of this article is available here:
https://doi.org/10.1371/journal.pone.0258711
Copyright: © 2021 Van Reet et al. This is an open
access article distributed under the terms of the
Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
1 Department of Biomedical Sciences, Institute of Tropical Medicine, Antwerpen, Belgium, 2 De´ partement de
Parasitologie, Institut National de Recherche Biome´ dicale, Kinshasa, Democratic Republic of the Congo
¤ Current address: Me´ decins Sans Frontières—Access Campaign, Geneva, Switzerland
* nvanreet@itg.be
Abstract
The Trypanosoma brucei repeat (TBR) is a tandem repeat sequence present on the Trypa-
nozoon minichromosomes. Here, we report that the TBR sequence is not as homogenous
as previously believed. BLAST analysis of the available T. brucei genomes reveals various
TBR sequences of 177 bp and 176 bp in length, which can be sorted into two TBR groups
based on a few key single nucleotide polymorphisms. Conventional and quantitative PCR
with primers matched to consensus sequences that target either TBR group show substan-
tial copy-number variations in the TBR repertoire within a collection of 77 Trypanozoon
strains. We developed the qTBR, a novel PCR consisting of three primers and two probes,
to simultaneously amplify target sequences from each of the two TBR groups into one single
qPCR reaction. This dual probe setup offers increased analytical sensitivity for the molecu-
lar detection of all Trypanozoon taxa, in particular for T.b. gambiense and T. evansi, when
compared to existing TBR PCRs. By combining the qTBR with 18S rDNA amplification as
an internal standard, the relative copy-number of each TBR target sequence can be calcu-
lated and plotted, allowing for further classification of strains into TBR genotypes associated
with East, West or Central Africa. Thus, the qTBR takes advantage of the single-nucleotide
polymorphisms and copy number variations in the TBR sequences to enhance amplification
and genotyping of all Trypanozoon strains, making it a promising tool for prevalence studies
of African trypanosomiasis in both humans and animals.
Introduction
The subgenus Trypanozoon comprises various species, subspecies and subtypes of the unicellu-
lar protozoan Trypanosoma brucei (T.b.), all causing diseases in humans or animals [1, 2]. T.b.
PLOS ONE | https://doi.org/10.1371/journal.pone.0258711 October 25, 2021
1 / 14
PLOS ONE Data Availability Statement: All relevant data are
within the manuscript and its Supporting
Information files.
Funding: PB received grant OPP1174221 from The
Bill and Melinda Gates foundation
(gatesfoundation.org) and grant CHARHAT-RDC
from the Departement Economie, Wetenschap &
Innovatie (EWI-Vlaanderen.be). NVR received grant
1.5.093.16N from the Fonds Wetenschappelijk
Onderzoek (fwo.be). The funders had no role in
study design, data collection and analysis, decision
to publish, or preparation of the manuscript.
Competing interests: The authors have declared
that no competing interests exist.
TBR based Trypanozoon genotyping
gambiense (T.b.g.) is responsible for chronic human African trypanosomiasis (HAT), a disease
targeted for elimination by the World Health Organization, that still accounted for 977
patients reported in West and Central Africa in 2018 [3]. Annually, less than 100 cases are due
to T.b. rhodesiense (T.b.r.), which causes acute HAT in East Africa [4]. Trypanosoma brucei
brucei (T.b.b.) is a subspecies causing animal African trypanosomiasis (AAT) in an extensive
range of game and domestic animals in Sub-Saharan Africa, but is considered non-human
infective. All these T.b. subspecies are transmitted by tsetse flies (Glossina spp.), that also act as
vectors for other subgenera such as Dutonella and Nannomonas, which are other causative
agents of AAT [5]. The remaining species in Trypanozoon, T. evansi (T. ev.) and T. equiperdum
(T. eq.), cause non-tsetse transmitted African trypanosomoses (NTTAT). These parasites are
dyskinetoplastic mutants of T.b. that have largely or completely lost their mitochondrial
genome, and with it the ability to develop in and be transmitted by tsetse flies [6, 7]. Further
genetic studies have shown that some of the species and subspecies can be further divided in
types or groups that each have their own peculiarities [8–11]. In mammals infected with Try-
panozoon parasites, severity and disease progression may vary depending on the genotype of
the parasite and the host [12–14].
Molecular diagnosis of Trypanozoon infections often requires targeting multi-copy nucleic
acid sequences to increase the chance of detecting the sparse parasites in blood and other tis-
sues [15–18]. The molecular target with the highest known copy-number in Trypanozoon is
the 177 bp long Trypanosoma brucei repeat (TBR) sequence. TBR sequences are direct tandem
repeats that form the central core of the minichromosomes (MCs) and the few intermediate
chromosomes present in the nucleus [19]. Their organization as a large repetitive palindrome,
running from both subtelomeres to a central inversion point, indicate a role as origin of repli-
cation in these chromosome classes [19]. Around 100 MCs, sized 50–150 kb, are present in the
nuclear DNA of T. brucei and they represent almost 10% of the nuclear genome [19]. It is esti-
mated that roughly 55% of each MC, and thus 5.5% of the nuclear DNA in T. brucei, consists
of such TBR repeats [19, 20]. The non-repetitive DNA on MCs carries an important part of the
silent VSG gene repertoire, with most MCs having complete VSG genes that can be transposed
to the VSG expression site during the early stages of an infection [20]. In T.b.g., the average
lengths of the MCs are smaller, being 25 to 50 kb, and the estimated copy-numbers vary
between a few to up to 100 [21–24].
Soon after the discovery of the MCs as part of the African trypanosome satellite DNA [25,
26], the TBR sequence was chosen as target for diagnostic PCRs for screening Trypanozoon
infections in mammals and insects [27, 28]. Over the past years, several other TBR PCR were
developed for use in conventional and quantitative PCR [29–32]. Yet, despite suggestions by
Sloof et al. [25] and others [20, 33, 34], TBR sequence heterogeneity in Trypanozoon was never
extensively addressed. In this study, we provide evidence that TBR sequences are far more het-
erogenous than previously assumed. Furthermore, we show that single nucleotide polymor-
phisms and copy-number variations in the TBR sequences can be exploited to improve the
amplification of all Trypanozoon taxa using a newly developed quantitative TBR-PCR, called
qTBR, that may even allow to suggest the geographical origin of certain strains.
Materials and methods
Trypanozoon collection
Trypanosome sediments of Trypanozoon strains and cloned populations were available as
DEAE purified pellets kept at -80˚C [35–37] (S1 Table). They were prepared from in vivo
expansions in mice or rats for which clearance was issued by the Animal Ethics Committee of
the Institute of Tropical Medicine (DPU2017-1). DNA was extracted from 50 μl pellets,
PLOS ONE | https://doi.org/10.1371/journal.pone.0258711 October 25, 2021
2 / 14
PLOS ONE TBR based Trypanozoon genotyping
corresponding to 5 x 107 trypanosomes, using the Maxwell 16 Tissue DNA Purification Kit
(Promega), eluted in 300 μl of elution buffer, aliquoted at 10 ng/μl and stored at -20˚C. Most
populations were previously typed according to specific genetic markers for T.b.g. I [36], for T.
b.r. [37], for T. ev./eq. A [35] and for T. ev. B [38]. Strains negative for these (sub)species spe-
cific markers, yet positive for M18S II [15], were considered either as T.b.b., T.b.g. II or T. eq,
depending on host, geographical origin or described genetic background [10, 39]. None of the
Trypanozoon strains harboured mixed infections, as determined by testing the minisatellite
marker MORF2-REP [40].
BLAST search for TBR sequences
The TBR sequence [K00392.1] was queried in the Trypanosoma Blast Server using BLASTn on
the Sanger Institute website [https://www.sanger.ac.uk/resources/software/blast/] against the
T.b.b. EMBL data [/corebio/data/blastdb_web/tryppub/embl] and T.b.g. I "reads" database
[/corebio/data/blastdb_web/tryppub/TBGAMBIENSE.reads] in February 2017. The first 100
hits for T.b.b. and T.b.g. that had query scores above 800 were reverse complemented or not,
and aligned using MUSCLE in CLC SequenceViewer 8.0 (S1 File). Next, individual TBR
sequences were extracted from each hit using the restriction site HhaI as start point (S1 File).
Individual TBR sequences were realigned in MUSCLE and sorted according to the presence of
a few key SNPs into TBR sequence sets according to subspecies and sequence size. This
resulted in the construction of the Tbb177 and Tbbr176 TBR sequence sets from the T.b.b.
database and the Tbg177 and Tbg176 TBR sequence sets from the T.b.g. database (S1 File).
Consensus sequences for each TBR sequence set, representing 80% of the variants encoun-
tered, were aligned to the original TBR sequence and the 177-T1 and 177-T2 TBR variants
described by Wickstead et al. [33] using MUSCLE. The 177 bp TBR group gathers the
sequences from the Tbb177 and Tbg177 TBR sequence sets, while the 176 bp TBR group gath-
ers the Tbbr176 and Tbg176 TBR sequence sets.
Novel Trypanozoon qPCRs
All primers and probes for Trypanozoon detection are summarized in Table 1. We used IDT
PrimerQuest to design a hydrolysis probe based qPCR for the Trypanozoon specific single-
copy GPI-PLC gene (qGPI-PLC). The conventional M18S II PCR, as described in Deborg-
graeve et al. in [15], complemented with a hydrolysis probe for use in qPCR, as described by
Bendofil et al. in [41], targets the multi-copy 18S rRNA of Trypanozoon and was abbreviated
as q18S throughout this manuscript. IDT PrimerQuest was used to design a conventional
(c177) and quantitative (q177D) PCR based on the 80% consensus sequence of the Tbg177 set,
aiming to target the 177 bp TBR group. IDT PrimerQuest was also used to generate a primer
set for conventional PCR on the 80% consensus sequence of the Tbg176 set, called c176, with
the aim to target the 176 bp TBR group. To amplify target sequences of both the 176 bp and
177 bp TBR groups in a multiplexed reaction, we used AlleleID 7 (PREMIER Biosoft) to design
a common forward primer (qTBR-F), a 177-bp TBR group specific reverse primer and probe
(q177T), and a 176-bp TBR group specific reverse primer and probe (q176T) based on the con-
sensus sequences of the Tbg177 and Tbg176 sets. The position of the primers and probes tar-
geting TBR are shown in S1 Fig.
Conventional PCR
Conventional PCR was performed in a Biometra T3 using HotStarTaq Plus (Qiagen). Amplifi-
cation was performed in 1x Coral Load Buffer, using 500 nM of each forward and reverse
primer (IDT), 200 nM of each nucleotide (Eurogentec), 25 mM MgCl2 and 2 μl pure parasite
PLOS ONE | https://doi.org/10.1371/journal.pone.0258711 October 25, 2021
3 / 14
PLOS ONE TBR based Trypanozoon genotyping
Table 1. Primers and probes for PCR and qPCR detection of target sequences of GPI-PLC, 18S rDNA, and the 176 bp and 177 bp TBR groups in Trypanozoon.
PCR
qGPI-PLC
q18S
c177/q177D
c176
q177T
q176T
Oligo
qGPI-PLC-F
qGPI-PLC-R
qGPI-PLC-P
M18S II-F
M18S II-R
q18S-P
c177-F
c177-R
q177D-P
c176-F
c176-R
qTBR-F
q177T-R
q177T-P
qTBR-F
q176T-R
q176T-P
Sequence
CCCACAACCGTCTCTTTAACC
GGAGTCGTGCATAAGGGTATTC
FAM-ACACCACTTTGTAACCTCTGGCAGT-MGB
CGTAGTTGAACTGTGGGCCACGT
ATGCATGACATGCGTGAAAGTGAG
VIC-TCGGACGTGTTTTGACCCACGC-MGB
GCAACAAAGCTATTTAATGGTCCT
GCACACTTGTAATTAATATGGCACA
FAM-TGCGCAGTTAACGCTATTATACACA-MGB
GTGCAACAAAGCTAATAAATGGTTC
TAAAGAACAGCGTTGCAAACTT
CGCAGTTAACGCTATTATACA
GGACCATTAAATAGCTTTGTTG
NED-TGCCATATTAATTACAAGTGTGC-MGB
CGCAGTTAACGCTATTATACA
GAACCATTTATTAGCTTTGTTG
FAM-TGCAACGCTGTTCT-MGB
https://doi.org/10.1371/journal.pone.0258711.t001
Length (bp)
106
150
109
165
152
151
DNA (10 ng/μl) in a 20 μl reaction. PCR cycling consisted of 95˚C for 5 min, followed by 29
cycles of 94˚C for 30 seconds, 60˚C for 30 seconds and 72˚C for 30 seconds for c177, c176, and
M18S II [15]. Conventional PCR using TBR primers described in Masiga et al. [27], Mumba
et al. [30] and Becker et al. [29] was performed at annealing temperatures of 55˚C, 60˚C and
66˚C respectively. After a final extension of 5 minutes at 72˚C, amplification reactions (10 μl)
were visualized on 2% agarose after 135V for 30 minutes and stained in 0.5 mg/ml ethidium
bromide. A GeneRuler 100bp plus DNA ladder (Thermo Scientific) was used for amplicon
size estimation. Reactions were considered positive if bands of the expected length were
observed. Semi-quantitative conventional PCRs were performed by using 7 serial fivefold dilu-
tions: 1000 fg/μl, 200 fg/μl, 40 fg/μl, 8 fg/μl, 1.6 fg/μl, 0.32 fg/μl and 0.064 fg/μl of pure parasite
DNA of each Trypanozoon strain.
Quantitative PCR and RCN calculations
qPCR amplification was performed in a Quantstudio 5 (Applied Biosystems) using 1x Per-
feCTa qPCR Toughmix (Quantabio), 300 nM of each forward and reverse primer (IDT), 100
nM probe (Thermo Scientific) and 5 μl of pure parasite DNA in a total volume of 20 μl. qPCR
cycling consisted of 45˚C for 5 minutes, 95˚C for 10 minutes, followed by 35 cycles of 95˚C for
15 seconds and 60˚C for one minute. Analytical sensitivity and qPCR efficiency were calcu-
lated, using serial tenfold dilutions of pure parasite DNA: 100 pg/μl, 10 pg/μl, 1 pg/μl, 100 fg/
μl, 10 fg/μl and 1 fg/μl of two T.b.g. clones, LiTat 1.6 and AnTat 11.17. In duplexed qPCR reac-
tions, FAM-labelled probes for qGPI-PLC, q177D or the qPCR described by Mumba et al.
([30]), here abbreviated as qM, were combined with a VIC-labelled q18S probe to allow the
calculation of relative copy-numbers (RCN). These RCN, were calculated by subtracting the
Cq-values obtained in qGPI-PLC, q177D or qM from the Cq-value in q18S, resulting in a ΔCq-
value that was transformed to 2-ΔCq and averaged between replicates to yield the RCNs for
each of these targets. The qTBR reaction is performed as a triplex qPCR reaction with a NED-
labelled q177T probe, a FAM-labelled q176T probe and a VIC-labelled q18S probe. Here,
RCNs were calculated by subtracting the Cq-values obtained in q177T or q176T from the Cq-
PLOS ONE | https://doi.org/10.1371/journal.pone.0258711 October 25, 2021
4 / 14
PLOS ONE TBR based Trypanozoon genotyping
value in q18S, resulting in a ΔCq-value that was transformed to 2-ΔCq and averaged between
replicates to yield the RCNs for each TBR target. Graphical analysis was performed using R
(3.5.2) in RStudio (1.1.463) with packages “ggplot” (3.2.1), “ggrepel” (0.8.1) and “viridis”
(0.5.1).
Results
Novel TBR sequences identified by BLAST reveal the existence of two TBR
groups
BLAST analysis of the sole published TBR-sequence (K00392.1) resulted in 99 hits on T.b.b.
and 1 hit on T.b.r. in the T. brucei database, and 100 hits in the T.b.g. database. All hits were
tandem repeats of two to four TBR sequences with an average size of 552-bp in the T.b. data-
base and 761-bp in the T.b.g. database (S1 File). Alignment of HhaI extracted individual TBR-
repeats revealed that none were 100% identical to K00392.1 (S1 File). In addition to single
nucleotide polymorphisms (SNPs) and indels, also larger inserts and deletions were seen, as
previously reported by others, but these were not numerous enough for further analysis [20,
33, 34]. Remarkably, a few key SNPs permitted to sort these individual TBR sequences into
two major groups that were either 177 bp or 176 bp long. The 99 T.b.b. hits, contained 245
individual TBR sequences that formed the Tbb177 TBR sequence set. The 100 T.b.g. hits con-
tained 153 sequences that formed the Tbg177 TBR sequence set, and 141 sequences that
formed the Tbg176 set. Three individual TBR sequences obtained from the single T.b.r. hit
were joined with two 176 bp sequences from the T.b.b. database, to form the Tbbr176 sequence
set. For each of these four TBR sequence sets, 80% consensus sequences were generated and
aligned against K00392.1 and two previously reported TBR variants, 177-T1 and 177-T2 [20,
33, 34] (Fig 1, S1 File). The Tbb177 and Tbg177 TBR sequence sets, together with the 177-T2
sequence, can be gathered into a larger 177 bp TBR group, while the Tbbr176 and Tbg176 TBR
sequences sets share an indel and may be assembled into a broader 176 bp TBR group. Both
TBR groups differed in a few key SNPs with K00392.1, yet they differed even more with each
other.
Fig 1. Novel TBR groups identified by BLAST. Alignment of the TBR sequence [K00392.1], both TBR variants,
177-T1 and 177-T2, and the 80% consensus sequences derived from each of the four TBR sequence sets.
Polymorphisms in comparison to K00392.1 are indicated in green.
https://doi.org/10.1371/journal.pone.0258711.g001
PLOS ONE | https://doi.org/10.1371/journal.pone.0258711 October 25, 2021
5 / 14
PLOS ONE TBR based Trypanozoon genotyping
Existing TBR PCRs are biased towards amplification of the 176-bp TBR
group
For almost 40 years, the K00392.1 TBR sequence was the only template available for PCR
primer design. All 6 published TBR primer sets [27–32] match 100% with the Tbg176 and
Tbbr176 consensus sequences from the 176 bp TBR group (S2 Table). In contrast, most primer
sets contain one or several mismatches in at least one of the primer binding regions for the
Tbb177 and Tbg177 consensus sequences from the 177 bp TBR group (S2 Table). We devel-
oped a new primer set for conventional PCR, c177, to detect target sequences belonging to the
177 bp TBR group specifically, by intendedly mismatching both primers against the 176-bp
TBR group sequences. Similarly, we developed a new primer set for conventional PCR to
detect 176 bp TBR group, c176, that mismatches against the 177 bp TBR group sequences (S2
Table). We compared the analytical sensitivity in conventional PCR of c177 and c176 with the
TBR PCR previously published by Masiga et al. [27], and to conventional PCR adapted ver-
sions of the TBR PCRs published by Becker et al. [29] and Mumba et al. [30]. Amplification of
the 18S rDNA, using M18S II, as described in Deborggraeve et al. [15], was used as an external
standard for detection of Trypanozoon DNA. In total, we tested 36 different Trypanozoon
strains: 8 T.b.b., 10 T.b.g. I, 2 T.b.g. II, 9 T.b.r., 3 T. ev. A, 2 T. ev. B, 1 T.eq. B and 1 T.eq. O, in
semi-quantitative conventional PCR (S1 Table). None of the conventional TBR PCRs resulted
in the specific detection of a particular taxon, yet, not all TBR PCRs detected Trypanozoon
DNA with the same sensitivity (S2 Fig). For example, c177 outperformed all other TBR PCR
sets in T.b.g. I detection, while despite comparable DNA content in each dilution series accord-
ing to M18S II, some PCRs that target the 176 bp TBR group were not successful in amplifying
some of the T.b.g. I strains (Fig 2). Conventional PCR on TBR suffers from the difficult inter-
pretation of the electrophoretic patterns possibly causing confusion about the specificity of the
PCR. While the main amplicon of each TBR PCR had the expected length, many larger-sized
bands, including bands that were the size of one full TBR repeat larger than the main amplicon
occurred at higher amounts of template DNA.
A multiplex qPCR to ameliorate amplification and copy-number
calculation for target sequences of each of both TBR groups
We designed a multiplex qPCR, consisting of a common forward primer, qTBR-F, and specific
reverse primers, q177T-R and q176T-R, and probes, q177T-P and q176T-P, for simultaneous
amplification of target sequences of both the 177 bp and the 176 bp TBR group. When com-
bined with 18S rDNA qPCR amplification, this novel qPCR triplex, called qTBR, allows accu-
rate RCN calculation for each TBR target sequence using q18S as internal standard. We
compared the qTBR with two other TBR qPCRs: the qPCR described by Mumba and co-work-
ers [30], here, abbreviated as qM, and the q177D, using the primers of the conventional TBR
PCR, c177, complemented with a probe for qPCR, q177D-P. Combining q18S in a duplex
qPCR with qM or q177D allows to calculate the RCNs for these respective target TBR
sequences, while the RCNs of 18S rDNA were calculated using a duplex qPCR containing
qGPI-PLC and q18S. All novel qPCRs were first tested on a dilution series of DNA on two T.b.
g. type I strains. For both strains, the target sequences for the 177 bp group, q177D and q177T,
could detect 5 fg of DNA, while Cq-values varied little between replicates of the same dilutions
in simplex, duplex or triplex format (S3 Fig). In addition, qPCR efficiency had acceptable
slopes between -3.1 and -3.6 in most formats (S4 Fig). Next, we tested gDNA from 77 Trypano-
zoon strains and clones: 12 T.b.b., 35 T.b.g. I, 2 T.b.g. II, 11 T.b.r., 21 T. ev./eq. A, 3 T. ev. B, 1 T.
eq. B and 4 T. eq. O (S1 Table). We found that the RCNs of q18S showed little variation within
and between the different Trypanozoon taxa. In contrast, the RCNs of the TBR repertoire
PLOS ONE | https://doi.org/10.1371/journal.pone.0258711 October 25, 2021
6 / 14
PLOS ONE TBR based Trypanozoon genotyping
Fig 2. Semi-quantitative conventional TBR PCRs on Trypanozoon. Semi-quantitative conventional TBR PCRs using the c177, c176, Masiga, Becker, Mumba, and
M18S II primer sets on fivefold serial dilutions containing 2000, 400, 80, 16, 3.2, 0.64, 0.128 or 0 fg of pure genomic Trypanozoon DNA. Each of the six gels shows the
electrophoretic results of one conventional PCR tested on 3 T.b.r. (upper part of the gels) and 3 T.b.g. (lower part of the gels) strains, separated by 5 μl of the Generuler
100-bp DNA ladder (Thermo Scientific).
https://doi.org/10.1371/journal.pone.0258711.g002
varied greatly between and even within the different Trypanozoon taxa (Fig 3). The 177-bp
TBR group, assayed using q177D and q177T had RCNs ranging from tens to hundredths for
most Trypanozoon, including T.b.g. I, and up to thousands in the case of T.ev./eq. A. In
Fig 3. RCNs for 18S rDNA and the TBR repertoire within 77 Trypanozoon strains. All Trypanozoon strains were
tested at 50 ng of pure genomic DNA. Mean and standard deviation were calculated using three replicates for each
sample. RCNs for 18S rDNA were calculated using the ΔCq-method between q18S and qGPI-PLC. RCNs for the TBR
targets q177D, q177T, qM and q176T were calculated using the ΔCq-method with q18S as reference.
https://doi.org/10.1371/journal.pone.0258711.g003
PLOS ONE | https://doi.org/10.1371/journal.pone.0258711 October 25, 2021
7 / 14
PLOS ONE TBR based Trypanozoon genotyping
contrast, in qM, RCNs of the target sequence ranged from thousands in T.b.r, to fewer than 10
in most T.b.g. I strains. Similarly, the q176T target sequence RCNs ranged from thousands in
T.b.b. and T.b.r. to undetectably low in 15 out of 35 T.b.g. I and 10 out of 23 T. ev./eq. A
strains.
The qTBR can be used to assign Trypanozoon strains into TBR genotypes
By plotting the target sequence RCNs obtained for q177T against those obtained for q176T, the
77 Trypanozoon strains present in this collection are scattered along a continuum whereby
both ends can be named after the geographical region from where most strains were isolated
(Fig 4). The first TBR genotype, called TBR-East, is here defined as having at least 1.2-fold
higher RCNs observed in q176T than in q177T and consists of T.b.r. and T.b.b. strains that
mostly originate from East Africa. All the T.b.r. strains were isolated in Uganda, Rwanda or
Kenya, except TRPZ 210, which was isolated in Zambia. Six out of nine T.b.b. strains belong to
this TBR-East genotype. Five of them originated from East Africa (Uganda, Kenya and Tanza-
nia), yet AnTat 17.1 was isolated from a sheep in Kongo-Central province of The Democratic
Republic of the Congo (DRC). The second TBR genotype, called TBR-West, is here defined as
having higher or equal RCNs for the target sequence in q177T than in q176T. It comprises rep-
resentatives of all non-T.b.r. species and subspecies, including historical T.b.g. I strains from
Coˆte d’Ivoire, but also the most recent T.b.g. I strain in our collection, i.e. MM01, isolated in
2008 in Kwilu province in DRC. The T.b.g. II strains ABBA and LIGO and the remaining T.b.
b. strains were isolated in West Africa (Coˆte d’Ivoire, The Gambia, Nigeria), except T.b.b. J10,
which was isolated from a hyena in Zambia. All taxa of dyskinetoplastic trypanosomes are rep-
resented in this TBR-West genotype: 13 out of 23 T.ev./eq. A (China, Ethiopia, Kenya,
Morocco, unknown origin) and all four T. eq. O strains (RSA, Ethiopia, Venezuela), although
the three T.ev. B (Ethiopia and Kenya) and the single T.eq. B (Morocco) are positioned in the
middle between both TBR genotypes. A third TBR genotype, called TBR-Central, comprises
strains that have detectable RCN for q177T, yet remain negative for q176T. This genotype cor-
responds to T.b.g. I strains isolated mainly in Cameroon and in the East-Kasaï province in
DRC, but also holds 10 out of the 23 T.ev./eq. A strains (Brazil, Colombia, Indonesia, Kazakh-
stan, the Philippines, South-America, unknown origin).
Discussion
Over the last 40 years, TBR PCRs have increasingly been used for detecting Trypanozoon infec-
tions in vertebrates and insects. During this period, primer sets were often repositioned in dif-
ferent PCR formats, yet all those positional adaptations were based on the single TBR
sequence reported by Sloof et al. [25]. Here, we have shown that this TBR sequence is more
heterogenous than previously assumed. The rediscovered 177 bp TBR group allows to amelio-
rate amplification of most non-rhodesiense Trypanozoon, which is in particular relevant for T.
b.g. I and the dyskinetoplastic trypanosomes. Some strains belonging to these taxa are not
detected by primer sets that solely target the 176 bp TBR group. The dual TBR probes in the
qTBR improve amplification of all Trypanozoon taxa, which is particularly relevant for try-
panosomiasis infections caused by T.b.g I in insects and mammals [42]. Plotting the target
sequence RCNs of q177T versus those of the q176T classifies Trypanozoon strains into two
opposing genotypes: TBR-East and TBR-West, whose names roughly refer to the geographical
origin of the strains making up these opposites. This East-West dichotomy in Trypanozoon
has been observed earlier via various genotyping techniques such as zymodemes, VSG reper-
toire, microsatellites, and even genome-wide SNP analysis [11, 23, 39, 40, 43, 44]. These latter
techniques have some disadvantages such as the requirement of high amounts of input
PLOS ONE | https://doi.org/10.1371/journal.pone.0258711 October 25, 2021
8 / 14
PLOS ONE TBR based Trypanozoon genotyping
Fig 4. Three TBR genotypes among 77 Trypanozoon strains. All Trypanozoon stocks were tested at 50 ng of pure genomic DNA using the qTBR. Mean and standard
deviation were calculated using three replicates for each sample. RCNs for target sequences q177T and q176T were calculated using the ΔCq-method with q18S as
reference and plotted in a scatterplot. Two solid lines were drawn to divide the Trypanozoon strains into roughly 3 TBR-genotypes: East, West and Central. The line at
y = 1 separates TBR-Central from both TBR-West and TBR-East, while the line at y = 1.2x separates TBR-East from TBR-West and TBR-Central. The dotted line at
y = x represents equal RCNs for q177T and q176T.
https://doi.org/10.1371/journal.pone.0258711.g004
PLOS ONE | https://doi.org/10.1371/journal.pone.0258711 October 25, 2021
9 / 14
PLOS ONE TBR based Trypanozoon genotyping
material, multiple PCR reactions, difficult interpretation of banding patterns or requiring bio-
informatic analysis. In contrast, the qTBR allows a rough genotyping of strains by one single
multiplex qPCR. Because TBR sequences form the central core of MCs, absence of amplifica-
tion may therefore indicate loss of certain MC sets. According to this view, the TBR-Central
genotype may just appear as a degenerate form of the TBR-West genotype, composed of
strains that have lost MCs that contain 176 bp TBR sequences. The qTBR is one of the few
techniques that is able to demonstrate such microheterogeneity within both T.b.g. I and T.
evansi, two species assumed to be exclusively clonally propagated [10, 45]. T.b.g. I strains iso-
lated in Kwilu and Kasaï-Oriental provinces in the DRC suggest that different TBR genotypes
are circulating in foci separated less than 1000 km from each other. The 177 bp TBR group is
detectable in all T.b.g. I strains present in this collection. However, more strains from East
Africa and Central Africa ought to be included for a better geographical coverage of T.b.g. I
[46, 47]. In addition, another limit of our study is the absence of recently isolated T.b.b. strains
from West Africa and Central Africa in our Trypanozoon collection. Trypanosomes causing
HAT can be found among all three TBR genotypes and cannot unequivocally be differentiated
from those causing AAT or NTTAT. Nevertheless, presence of Trypanozoon DNA in human
clinical samples should always warrant special attention, since atypical, e.g. with T.b.g. II, or
even opportunistic Trypanozoon infections, e.g. with T.b.b. and T. ev., are known to occur [8,
48]. With the limited TBR sequence information available today, we cannot exclude that more
TBR groups may be discovered within larger and more diverse collections of Trypanozoon,
perhaps even specific for certain Trypanozoon taxa. Sequencing the Trypanozoon repetitive
DNA, preferentially using sequencing platforms that overcome the limitations imposed by tan-
dem repeat sequences, will be crucial to further understand the evolution of MC and the diver-
sity in TBR content of African trypanosomes [49].
Supporting information
S1 Fig. Position of the primers and probes targeting TBR. A representation of a TBR
sequence as tandem repeat (yellow) showing the relative position of primers and probes used
in conventional and quantitative PCR. Arrows indicate the 5’– 3’ direction of primers, while
for probes, circles indicate fluorophores and diamonds indicate quenchers. In A, blue indicates
the c177 and the q177D set, while red represents the c176 set. In B, the green arrow represents
the common primer qTBR-F, while the blue primer and probe indicate the q177T set, and the
red primer and probe indicate the q176T set.
(TIF)
S2 Fig. Conventional TBR PCR on 36 Trypanozoon strains. Semi-quantitative conventional
PCRs using the Masiga, Becker, Mumba, c177, c176 and M18S II primer sets on fivefold serial
dilutions containing 2000, 400, 80, 16, 3.2, 0.64, 0.128 or 0 fg of pure parasite DNA per lane. In
total, 36 Trypanozoon strains were tested. Each of the gels shows the electrophoretic results of
one of the conventional PCRs tested on 6 Trypanozoon strains (3 above and 3 below), sepa-
rated by 5 μl of the Generuler 100-bp DNA ladder (Thermo Scientific).
(TIF)
S3 Fig. qPCR efficiency of novel qPCRs in different multiplex formats. Cq-values obtained
from a tenfold dilution series from 500 pg down to 5 fg of pure parasite DNA (in elution
buffer) of two T.b.g. I clones: AnTat 11.17 and LiTat 1.6 in different qPCR formats: simplex,
duplex (in combination with q18S) or triplex (as qTBR). The slope of qPCR efficiency was esti-
mated by fitting a linear trendline on Cq values and the log transformed concentrations.
(TIF)
PLOS ONE | https://doi.org/10.1371/journal.pone.0258711 October 25, 2021
10 / 14
PLOS ONE TBR based Trypanozoon genotyping
S4 Fig. Analytical sensitivity of novel qPCRs in different multiplex formats. Cq-values
obtained from a tenfold dilution series from 500 pg down to 5 fg of pure parasite DNA (in elu-
tion buffer) of two T.b.g. I clones: AnTat 11.17 and LiTat 1.6 in different qPCR formats. Each
of the novel qPCRs was first tested individually in simplex format (A). qGPI-PLC, q177D or
qM, was combined with q18S in duplex format (B). q177T and q176T were combined with
q18S in triplex format, representing the qTBR (C).
(TIF)
S1 File. Extracted TBR sequences and arrangement into TBR sets. Tandem TBR sequences
obtained by running the TBR-sequence [K00392.1] on the Trypanosoma Blast Server using
BLASTn on the Sanger Institute website against the T.b.b. and T.b.g. database (sheets Hits T.b.
and Hits T.b.g.), individual TBR repeats extracted using HhaI (sheets HhaI T.b. and HhaI T.b.
g.) and TBR sequence sets, sorted per key SNPs and subspecies (sheets Tbb177, Tbg177,
Tbg176 and Tbbr176). The 80% consensus sequences for each TBR sequence set are described
on the last line of these sheets.
(XLSX)
S1 Raw images. Raw images of Fig 2.
(PDF)
S2 Raw images. Raw images of S2 Fig.
(PDF)
S1 Table. Trypanozoon strains and presumed taxon. Collection of Trypanozoon strains with
taxon, strain name, clone status, host, country, area and year of isolation and the results of the
PCR typing of this strains using TgsGP, SRA, RoTat 1.2, EVA B, MORF2-REP and their TBR
genotype according to qTBR.
(XLSX)
S2 Table. Existing TBR PCR sets and compatibility with the novel TBR sequence sets.
Existing TBR PCR sets with literature reference, type of PCR, primer names and sequence,
matching TBR sets and mismatching TBR sets.
(XLSX)
Acknowledgments
We would like to thank Jeroen Swiers (ITM, Antwerp) for excellent assistance in cryobiology.
Author Contributions
Conceptualization: Nick Van Reet, Philippe Bu¨scher.
Data curation: Nick Van Reet.
Formal analysis: Nick Van Reet.
Funding acquisition: Nick Van Reet, Philippe Bu¨scher.
Investigation: Nick Van Reet, Sara Dehou, Nicolas Bebronne.
Methodology: Nick Van Reet, Stijn Deborggraeve.
Project administration: Nick Van Reet, Sara Dehou, Nicolas Bebronne.
Resources: Pati Patient Pyana, Nicolas Bebronne, Stijn Deborggraeve, Philippe Bu¨scher.
Supervision: Nick Van Reet, Philippe Bu¨scher.
PLOS ONE | https://doi.org/10.1371/journal.pone.0258711 October 25, 2021
11 / 14
PLOS ONE TBR based Trypanozoon genotyping
Validation: Sara Dehou, Nicolas Bebronne.
Visualization: Nick Van Reet.
Writing – original draft: Nick Van Reet, Philippe Bu¨scher.
Writing – review & editing: Nick Van Reet, Pati Patient Pyana, Sara Dehou, Nicolas Beb-
ronne, Stijn Deborggraeve, Philippe Bu¨scher.
References
1. Hoare CA. The trypanosomes of mammals: a zoological monograph. The trypanosomes of mammals.
A zoological monograph. Blackwell Scientific Publications; 1972. 749 p.
2. Radwanska M, Vereecke N, Deleeuw V, Pinto J, Magez S. Salivarian Trypanosomosis: A Review of
Parasites Involved, Their Global Distribution and Their Interaction With the Innate and Adaptive Mam-
malian Host Immune System. Front Immunol. 2018 Oct 2;0(OCT):2253. https://doi.org/10.3389/fimmu.
2018.02253 PMID: 30333827
3.
Franco JR, Cecchi G, Priotto G, Paone M, Diarra A, Grout L, et al. Monitoring the elimination of human
African trypanosomiasis at continental and country level: Update to 2018. PLoS Negl Trop Dis. 2020
May 1; 14(5):1–18. https://doi.org/10.1371/journal.pntd.0008261 PMID: 32437391
4. Bu¨ scher P, Cecchi G, Jamonneau V, Priotto G. Human African trypanosomiasis. Lancet. 2017 Nov;
390(10110):2397–409. https://doi.org/10.1016/S0140-6736(17)31510-6 PMID: 28673422
5. Rotureau B, Van Den Abbeele J. Through the dark continent: African trypanosome development in the
tsetse fly [Internet]. Vol. 4, Frontiers in Cellular and Infection Microbiology. 2013. p. 53. https://doi.org/
10.3389/fcimb.2013.00053 PMID: 24066283
6.
Lai D-H, Hashimi H, Lun Z-R, Ayala FJ, Lukes J. Adaptations of Trypanosoma brucei to gradual loss of
kinetoplast DNA: Trypanosoma equiperdum and Trypanosoma evansi are petite mutants of T. brucei.
Proc Natl Acad Sci. 2008; 105(6):1999–2004. https://doi.org/10.1073/pnas.0711799105 PMID:
18245376
7. Kamidi CM, Saarman NP, Dion K, Mireji PO, Ouma C, Murilla G, et al. Multiple evolutionary origins of
Trypanosoma evansi in Kenya. PLoS Negl Trop Dis. 2017 Sep; 11(9):e0005895. https://doi.org/10.
1371/journal.pntd.0005895 PMID: 28880965
8.
Jamonneau V, Truc P, Gre´ baut P, Herder S, Ravel S, Solano P, et al. Trypanosoma brucei gambiense
Group 2: The Unusual Suspect. Trends Parasitol. 2019 Dec 1; 35(12):983–95. https://doi.org/10.1016/j.
pt.2019.09.002 PMID: 31668893
9. Birhanu H, Gebrehiwot T, Goddeeris BM, Bu¨ scher P, Van Reet N. New Trypanosoma evansi Type B
Isolates from Ethiopian Dromedary Camels. PLoS Negl Trop Dis. 2016; 10(4). https://doi.org/10.1371/
journal.pntd.0004556 PMID: 27035661
10. Carnes J, Anupama A, Balmer O, Jackson A, Lewis M, Brown R, et al. Genome and Phylogenetic Anal-
yses of Trypanosoma evansi Reveal Extensive Similarity to T. brucei and Multiple Independent Origins
for Dyskinetoplasty. Tschudi C, editor. PLoS Negl Trop Dis. 2015 Jan 8; 9(1):e3404. https://doi.org/10.
1371/journal.pntd.0003404 PMID: 25568942
11. Godfrey DG, Baker RD, Rickman LR, Mehlitz D. The distribution, relationships and identification of
enzymic variants within the subgenus Trypanozoon. Adv Parasitol. 1990; 29:1–74. https://doi.org/10.
1016/s0065-308x(08)60104-9 PMID: 2181826
12.
Losos GJ, Ikede BO. Review of Pathology of Diseases in Domestic and Laboratory Animals Caused by
Trypanosoma congolense, T. vivax, T. brucei, T. rhodesiense and T. gambiense. Vet Pathol. 1972 Jan
6; 9(1_suppl):1–79.
13. Berthier D, Brenière SF, Bras-Gonc¸alves R, Lemesre JL, Jamonneau V, Solano P, et al. Tolerance to
Trypanosomatids: A Threat, or a Key for Disease Elimination? Trends Parasitol. 2016; 32(2):157–68.
https://doi.org/10.1016/j.pt.2015.11.001 PMID: 26643519
14.
Jamonneau V, Solano P, Laveissière C, Bucheton B, Ilboudo H, Kabore´ J, et al. Untreated human infec-
tions by trypanosoma brucei gambiense are not 100% fatal. PLoS Negl Trop Dis. 2012; 6(6):e1691.
https://doi.org/10.1371/journal.pntd.0001691 PMID: 22720107
15. Deborggraeve S, Lejon V, Ekangu RA, Mumba Ngoyi D, Pati Pyana P, Ilunga M, et al. Diagnostic Accu-
racy of PCR in gambiense Sleeping Sickness Diagnosis, Staging and Post-Treatment Follow-Up: A 2-
year Longitudinal Study. Aksoy S, editor. PLoS Negl Trop Dis. 2011 Feb 22; 5(2):e972. https://doi.org/
10.1371/journal.pntd.0000972 PMID: 21364966
16. Mitashi P, Hasker E, Ngoyi DM, Pyana PP, Lejon V, Van der Veken W, et al. Diagnostic Accuracy of
Loopamp Trypanosoma brucei Detection Kit for Diagnosis of Human African Trypanosomiasis in
PLOS ONE | https://doi.org/10.1371/journal.pone.0258711 October 25, 2021
12 / 14
PLOS ONE TBR based Trypanozoon genotyping
Clinical Samples. Ndung’u JM, editor. PLoS Negl Trop Dis. 2013 Oct 17; 7(10):e2504. https://doi.org/
10.1371/journal.pntd.0002504 PMID: 24147176
17. Gonza´ lez-Andrade P, Camara M, Ilboudo H, Bucheton B, Jamonneau V, Deborggraeve S. Diagnosis of
Trypanosomatid Infections. J Mol Diagnostics. 2014 Jul; 16(4):400–4. https://doi.org/10.1016/j.jmoldx.
2014.02.006 PMID: 24814957
18. Chiweshe SM, Steketee PC, Jayaraman S, Paxton E, Neophytou K, Erasmus H, et al. Parasite specific
7SL-derived small RNA is an effective target for diagnosis of active trypanosomiasis infection. Caljon G,
editor. PLoS Negl Trop Dis. 2019 Feb 19; 13(2):e0007189. https://doi.org/10.1371/journal.pntd.
0007189 PMID: 30779758
19. Wickstead B, Ersfeld K, Gull K. The small chromosomes of Trypanosoma brucei involved in antigenic
variation are constructed around repetitive palindromes. Genome Res. 2004 Jun 12; 14(6):1014–24.
https://doi.org/10.1101/gr.2227704 PMID: 15173109
20. Cross GAM, Kim HS, Wickstead B. Capturing the variant surface glycoprotein repertoire (the
VSGnome) of Trypanosoma brucei Lister 427. Mol Biochem Parasitol. 2014 Jun 1; 195(1):59–73.
https://doi.org/10.1016/j.molbiopara.2014.06.004 PMID: 24992042
21. Gibson WC, Borst P. Size-fractionation of the small chromosomes of Trypanozoon and Nannomonas
trypanosomes by pulsed field gradient gel electrophoresis. Mol Biochem Parasitol. 1986 Feb; 18
(2):127–40. https://doi.org/10.1016/0166-6851(86)90033-2 PMID: 3960051
22. Dero B, Zampetti-Bosseler F, Pays E, Steinert M. The genome and the antigen gene repertoire of Try-
panosoma brucei gambiense are smaller than those of T. b. brucei. Mol Biochem Parasitol. 1987 Dec 1;
26(3):247–56. https://doi.org/10.1016/0166-6851(87)90077-6 PMID: 3431572
23. Kanmogne GD, Bailey M, Gibson WC. Wide variation in DNA content among isolates of Trypanosoma
brucei ssp. Acta Trop. 1997 Feb; 63(2–3):75–87. https://doi.org/10.1016/s0001-706x(96)00600-6
PMID: 9088421
24.
Jackson AP, Sanders M, Berry A, McQuillan J, Aslett MA, Quail MA, et al. The genome sequence of
Trypanosoma brucei gambiense, causative agent of chronic human African Trypanosomiasis. PLoS
Negl Trop Dis. 2010; 4(4). https://doi.org/10.1371/journal.pntd.0000658 PMID: 20404998
25. Sloof P, Menke HH, Caspers MPM, Borst P. Size fractionation of Trypanosoma brucei DNA: localization
of the 177-bp repeat satellite DNA and a variant surface glycoprotein gene in a mini-chromosomal DNA
fraction. Nucleic Acids Res. 1983 Jun 25; 11(12):3889–901. https://doi.org/10.1093/nar/11.12.3889
PMID: 6866765
26. Sloof P, Bos JL, Konings AFJM, Menke HH, Borst P, Gutteridge WE, et al. Characterization of satellite
DNA in Trypanosoma brucei and Trypanosoma cruzi. J Mol Biol. 1983 Jun 15; 167(1):1–21. https://doi.
org/10.1016/s0022-2836(83)80031-x PMID: 6345792
27. Masiga DK, Smyth AJ, Hayes P, Bromidge TJ, Gibson WC. Sensitive detection of trypanosomes in
tsetse flies by DNA amplification. Int J Parasitol. 1992 Nov; 22(7):909–18. https://doi.org/10.1016/0020-
7519(92)90047-o PMID: 1459784
28. Moser DR, Cook GA, Ochs DE, Bailey CP, McKane MR, Donelson JE. Detection of Trypanosoma con-
golense and Trypanosoma brucei subspecies by DNA amplification using the polymerase chain reac-
tion. Parasitology. 1989 Aug 6; 99(1):57–66. https://doi.org/10.1017/s0031182000061023 PMID:
2797872
29. Becker S, Franco JR, Simarro PP, Stich A, Abel PM, Steverding D. Real-time PCR for detection of Try-
panosoma brucei in human blood samples. Diagn Microbiol Infect Dis. 2004 Nov; 50(3):193–9. https://
doi.org/10.1016/j.diagmicrobio.2004.07.001 PMID: 15541605
30. Mumba Ngoyi D, Ali Ekangu R, Mumvemba Kodi MF, Pyana PP, Balharbi F, Decq M, et al. Perfor-
mance of Parasitological and Molecular Techniques for the Diagnosis and Surveillance of Gambiense
Sleeping Sickness. Ndung’u JM, editor. PLoS Negl Trop Dis. 2014 Jun 12; 8(6):e2954. https://doi.org/
10.1371/journal.pntd.0002954 PMID: 24921941
31. Ahmed HA, Picozzi K, Welburn SC, MacLeod ET. A comparative evaluation of PCR- based methods
for species- specific determination of African animal trypanosomes in Ugandan cattle. Parasit Vectors.
2013; 6(1):316. https://doi.org/10.1186/1756-3305-6-316 PMID: 24499678
32. Cunningham LJ, Lingley JK, Haines LR, Ndung’u JM, Torr SJ, Adams ER. Illuminating the Prevalence
of Trypanosoma brucei s.l. in Glossina Using LAMP as a Tool for Xenomonitoring. Van Den Abbeele J,
editor. PLoS Negl Trop Dis. 2016 Feb 18; 10(2):e0004441. https://doi.org/10.1371/journal.pntd.
0004441 PMID: 26890882
33. Wickstead B, Ersfeld K, Gull K. The frequency of gene targeting in Trypanosoma brucei is independent
of target site copy number. Nucleic Acids Res. 2003 Jul 15; 31(14):3993–4000. https://doi.org/10.1093/
nar/gkg445 PMID: 12853615
PLOS ONE | https://doi.org/10.1371/journal.pone.0258711 October 25, 2021
13 / 14
PLOS ONE TBR based Trypanozoon genotyping
34. Elhaig MM, Sallam NH. Molecular survey and characterization of Trypanosoma evansi in naturally
infected camels with suspicion of a Trypanozoon infection in horses by molecular detection in Egypt.
Microb Pathog. 2018 Oct; 123:201–5. https://doi.org/10.1016/j.micpath.2018.07.017 PMID: 30016680
35. Claes F, Radwanska M, Urakawa T, Majiwa PAO, Goddeeris B, Bu¨scher P. Variable surface glycopro-
tein RoTat 1.2 PCR as a specific diagnostic tool for the detection of Trypanosoma evansi infections.
Kinetoplastid Biol Dis. 2004; 3:1–6. https://doi.org/10.1186/1475-9292-3-1 PMID: 15142279
36. Radwanska M, Claes F, Magez S, Magnus E, Perez-Morga D, Pays E, et al. Novel primer sequences
for polymerase chain reaction-based detection of Trypanosoma brucei gambiense. Am J Trop Med
Hyg. 2002 Sep 1; 67(3):289–95. https://doi.org/10.4269/ajtmh.2002.67.289 PMID: 12408669
37. Radwanska M, Bu¨scher P, de Baetselier P, Claes F, Chamekh M, Vanhamme L, et al. The serum resis-
tance-associated gene as a diagnostic tool for the detection of Trypanosoma brucei rhodesiense. Am J
Trop Med Hyg. 2002 Dec 1; 67(6):684–90. https://doi.org/10.4269/ajtmh.2002.67.684 PMID: 12518862
38. Njiru ZK, Constantine CC, Masiga DK, Reid SA, Thompson RCA, Gibson WC. Characterization of Try-
panosoma evansi type B. Infect Genet Evol. 2006 Jul; 6(4):292–300. https://doi.org/10.1016/j.meegid.
2005.08.002 PMID: 16157514
39. Cuypers B, Van den Broeck F, Van Reet N, Meehan CJ, Cauchard J, Wilkes JM, et al. Genome-Wide
SNP Analysis Reveals Distinct Origins of Trypanosoma evansi and Trypanosoma equiperdum.
Genome Biol Evol. 2017 Aug 1; 9(8):1990–7. https://doi.org/10.1093/gbe/evx102 PMID: 28541535
40. Biteau N, Bringaud F, Gibson W, Truc P, Baltz T. Characterization of Trypanozoon isolates using a
repeated coding sequence and microsatellite markers. Mol Biochem Parasitol. 2000 Feb; 105(2):187–
202. PMID: 10745102
41. Benfodil K, Bu¨ scher P, Abdelli A, Van Reet N, Mohamed-herif A, Ansel S, et al. Comparison of serologi-
cal and molecular tests for detection of Trypanosoma evansi in domestic animals from Ghardaïa district,
South Algeria. Vet Parasitol. 2020 Apr 1; 280:109089. https://doi.org/10.1016/j.vetpar.2020.109089
PMID: 32222595
42. Bu¨ scher P, Bart J-M, Boelaert M, Bucheton B, Cecchi G, Chitnis N, et al. Do Cryptic Reservoirs
Threaten Gambiense-Sleeping Sickness Elimination? Informal Expert Group on Gambiense HAT Res-
ervoirs Can Cryptic Reservoirs in Humans and Animals Compromise the Sustainable Elimination of
gambiense-HAT? Trends Parasitol. 2017;
43. Balmer O, Beadell JS, Gibson W, Caccone A. Phylogeography and Taxonomy of Trypanosoma brucei.
Solano P, editor. PLoS Negl Trop Dis. 2011 Feb 8; 5(2):e961. https://doi.org/10.1371/journal.pntd.
0000961 PMID: 21347445
44. Gibson Stevens, Truc Philippe. Identification of trypanosomes: from morphology to molecular biology.
1999.
45. Weir W, Capewell P, Foth B, Clucas C, Pountain A, Steketee P, et al. Population genomics reveals the
origin and asexual evolution of human infective trypanosomes. Elife. 2016 Jan 26; 5(JANU-
ARY2016):1–14. https://doi.org/10.7554/eLife.11473 PMID: 26809473
46. Enyaru JCK, Allingham R, Bromidge T, Kanmogne GD, Carasco JF. The isolation and genetic hetero-
geneity of Trypanosoma brucei gambiense from north-west Uganda. Acta Trop. 1993; 54(1):31–9.
https://doi.org/10.1016/0001-706x(93)90066-k PMID: 8103626
47. Simo G, Njiokou F, Tume C, Lueong S, De Meeuˆs T, Cuny G, et al. Population genetic structure of Cen-
tral African Trypanosoma brucei gambiense isolates using microsatellite DNA markers. Infect Genet
Evol. 2010 Jan; 10(1):68–76. https://doi.org/10.1016/j.meegid.2009.09.019 PMID: 19819349
48. Van Vinh Chau N, Buu Chau L, Desquesnes M, Herder S, Phu Huong Lan N, Campbell JI, et al. A Clini-
cal and Epidemiological Investigation of the First Reported Human Infection With the Zoonotic Parasite
Trypanosoma evansi in Southeast Asia. Clin Infect Dis. 2016; 62(8):1002–8. https://doi.org/10.1093/
cid/ciw052 PMID: 26908809
49. Pita S, Dıaz-Viraque F, Iraola G, Robello C. The tritryps comparative repeatome: Insights on repetitive
element evolution in trypanosomatid pathogens. Genome Biol Evol. 2019; 11(2):546–51. https://doi.
org/10.1093/gbe/evz017 PMID: 30715360
PLOS ONE | https://doi.org/10.1371/journal.pone.0258711 October 25, 2021
14 / 14
PLOS ONE |
10.1371_journal.pone.0265361 | RESEARCH ARTICLE
Testing the reliability and ecological
implications of ramping rates in the
measurement of Critical Thermal maximum
Chi-Man LeongID*, Toby P. N. Tsang, Benoit Gue´ nard
School of Biological Sciences, The University of Hong Kong, Hong Kong SAR, China
* cmleong@connect.hku.hk
Abstract
Critical Thermal maximum (CTmax) is often used to characterize the upper thermal limits of
organisms and represents a key trait for evaluating the fitness of ectotherms. The lack of
standardization in CTmax assays has, however, introduced methodological problems in its
measurement, which can lead to questionable estimates of species’ upper thermal limits.
Focusing on ants, which are model organisms for research on thermal ecology, we aim to
obtain a reliable ramping rate that will yield the most rigorous measures of CTmax for the
most species. After identifying three commonly used ramping rates (i.e., 0.2, 0.5 and 1.0˚C
min-1) in the literature, we experimentally determine their effects on the CTmax values of 27
species measured using dynamic assays. Next, we use static assays to evaluate the accu-
racy of these values in function of the time of exposure. Finally, we use field observations of
species’ foraging activities across a wide range of ground temperatures to identify the most
biologically relevant CTmax values and to develop a standardized method. Our results dem-
onstrate that the use of a 1˚C min-1 ramping rate in dynamic assays yields the most reliable
CTmax values for comparing ant species’ upper thermal limits, which are further validated in
static assays and field observations. We further illustrate how methodological biases in
physiological trait measurements can affect subsequent analyses and conclusions on com-
munity comparisons between strata and habitats, and the detection of phylogenetic signal
(Pagel’s λ and Bloomberg’s K). Overall, our study presents a methodological framework for
identifying a reliable and standardized ramping rate to measure CTmax in ants, which can be
applied to other ectotherms. Particular attention should be given to CTmax values obtained
with less suitable ramping rates, and the potential biases they may introduce to trait-based
research on global warming and habitat conversion, as well as inferences about phyloge-
netic conservatism.
1. Introduction
Organisms are increasingly exposed to novel and warmer environmental conditions owing to
global changes such as deforestation, urbanization, and climate change. High temperatures, in
particular, limit species’ survival, reproduction, and foraging activities—especially for
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: Leong C-M, Tsang TPN, Gue´nard B
(2022) Testing the reliability and ecological
implications of ramping rates in the measurement
of Critical Thermal maximum. PLoS ONE 17(3):
e0265361. https://doi.org/10.1371/journal.
pone.0265361
Editor: Daniel E. Naya, Universidad de la Republica
Uruguay, URUGUAY
Received: August 24, 2021
Accepted: February 28, 2022
Published: March 14, 2022
Copyright: © 2022 Leong et al. This is an open
access article distributed under the terms of the
Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: Data deposited in the
Dryad Digital Repository: https://doi.org/10.5061/
dryad.jdfn2z3d0.
Funding: CML is supported by Fundac¸ão Macau
and Direcc¸ão dos Servic¸os do Ensino Superior. BG
is also supported by an HKU Early Career Scheme
2017/18 grant (27106417). The funders had no
role in study design, data collection and analysis,
decision to publish, or preparation of the
manuscript.
PLOS ONE | https://doi.org/10.1371/journal.pone.0265361 March 14, 2022
1 / 23
PLOS ONE Competing interests: The authors have declared
that no competing interests exist.
Ramping rate of Critical Thermal maximum
ectothermic organisms [1]. Therefore, measuring the upper thermal limits of ectotherms is key
to obtaining valuable information needed for forecasting changes in community composition
and species distributions in response to rising temperatures [2].
Thermal performance theory [3] represents a useful framework for describing an organ-
ism’s performance in function of the temperature experienced [4], and for delimiting the ther-
mal range within which an organism can remain active. The upper thermal maximum, also
termed the Critical Thermal maximum (CTmax), is a particularly important threshold that rep-
resents the temperature at which an organism is unable to withstand heat stress [5]. Investigat-
ing upper thermal limit is paramount for understanding how species’ fitness are impacted by
climate change [6]. However, it is challenging to forecast the impacts of rising temperatures on
species’ fitness due to the lack of standardized methods for measuring CTmax as well as their
incompatibility with field observations [7]. For instance, the ramping rate (i.e., the rate at
which temperature increases over time) used to measure CTmax varies substantially across
studies, and this can result in major differences in CTmax values for the same species [5, 7, 8].
Therefore, a biologically relevant and reliable method for measuring CTmax values that is
directly applicable to ecological research is urgently needed to provide meaningful estimates of
species’ maximal thermal limits.
CTmax was defined by Cowles and Bogert as “the thermal point at which locomotory activity
becomes disorganized and the animal loses its ability to escape from conditions that will promptly
lead to its death” [9]. The measurement of CTmax uses an experimental approach to determine
the upper thermal limit of an organism through a progressive increase of the environmental
temperature (i.e., at the ramping rate) until the organism experiences a loss of muscle control
or a heat-coma.
The use of ecological methodology should be standardized and comparative [10] to provide
a consistent method for a given taxon or across multiple taxa. Yet, the use of the ramping rate
has not been standardized across CTmax assays, resulting in problems when comparing CTmax
values obtained from different ramping rates. Misinterpretations can also emerge in ecological
studies that fail to consider this source of methodological bias [5, 7]. Although recent studies
have developed a biophysical model based on Drosophila flies [11, 12], it should be noted that
model exceptions are already known in other taxa such as ants [see 13 and S3 Appendix].
Additionally, it is crucial to select biologically relevant ramping rate(s) that will yield CTmax
values that best reflect a species’ functional thermal niche from the perspective of community
and functional ecology. Ultimately, the current lack of standardization and testing for ecologi-
cal relevance leaves the following question unresolved: how should one develop a framework
to test the reliability and ecological implications of the ramping rate used in the measurement
of CTmax?
In the present study, we use dynamic and static thermal assays, as well as field observations
to capture different aspects of the thermal tolerance of ants—model organisms for understand-
ing the ecophysiology of terrestrial ectotherms [5, 14, 15]—to investigate the correspondence
and biological relevance of experimental measurements of CTmax. Integrating both dynamic
and static thermal assays allows us to fully capture the thermal tolerance of an organism,
which depends on two main parameters: 1) the temperature to which the organism is exposed,
and 2) the duration for which the organism is exposed to the given temperature [11]. To pro-
vide a general CTmax framework for studying the upper thermal limits of different taxa, we
investigate the upper thermal limits of 27 ant species displaying different body sizes and which
are associated with different micro-habitats, phylogenetic clades, and biogeographic origins.
Our goals are to provide an overview of the limitations arising from the use of different ramp-
ing rates, and to identify a more reliable protocol for measuring biologically relevant and com-
parable CTmax values.
PLOS ONE | https://doi.org/10.1371/journal.pone.0265361 March 14, 2022
2 / 23
PLOS ONE Ramping rate of Critical Thermal maximum
First, we conduct a literature review to identify the different ramping rates that have been
used to measure the upper thermal limits of ants, and to select the most commonly used rates
(i.e., 0.2, 0.5, and 1.0˚C min-1) that we test in subsequent experiments (S1 Fig in S1 Appendix).
Second, we investigate how the ramping rate used in a dynamic assay affects the CTmax value
retrieved. Here, we predict strong positive correlations between ramping rates and CTmax values
(Fig 1A) [7, 16]. Third, we further test the exposure duration-based thermal tolerance of most
species, by using static assays to examine how species respond—in terms of their exposure dura-
tion—to the CTmax values retrieved from the dynamic essays at different ramping rates. Our lit-
erature review suggests that ant species tend to have relatively longer exposure duration-based
tolerance at slow ramping rates (e.g., 0.2˚C min-1) in comparison to faster ramping rates [e.g.
17–19]. We predict that for a given species, individuals exposed to the temperature of their
CTmax retrieved at a low ramping rate (i.e., 0.2˚C min-1) will show a more heterogenous and
lengthier duration of tolerance, while those exposed to the temperature of their CTmax retrieved
at a faster ramping rate (i.e., 0.5 and 1.0˚C min-1) will display a shorter duration of tolerance.
Fourth, we determine each species’ foraging temperature maximum (FTmax) and compare this
to their CTmax to investigate the concordance between laboratory- (i.e., CTmax) and field-based
(FTmax) estimates of thermal tolerance. The FTmax represents the upper thermal limit of an ant
species under natural conditions, and corresponds to a species’ thermal threshold prior to the
loss of muscle control. To examine whether ant species in the field cease their foraging activities
at the CTmax values retrieved with slow or fast ramping rates, we compare their CTmax values
retrieved from dynamic assays in three different treatments (each with a different ramping rate)
with the maximum temperature at which their activity was observed (i.e., FTmax) during year-
long field observations (Fig 1C). In theory, CTmax represents the maximum temperature to
which an individual of a species can be exposed before it loses muscle control. Thus, comparing
ant species’ CTmax values with their foraging performance in function of the temperature in the
field (FTmax) will allow us to test the ecological relevance of the CTmax values retrieved from dif-
ferent ramping rates. Finally, we investigate how the use of specific ramping rates can alter con-
clusions about the detection of phylogenetic signal (Fig 1D) as well as differences in species’ use
of habitats and microhabitats (Fig 1E) in empirical studies comparing multiple species.
2. Materials and methods
2.1 Literature collection
We identified published studies of ant species’ upper thermal limits and recorded the relevant
information describing the protocols that were used to measure CTmax in each study. This
information included: whether a dynamic and/or static assay was used, the starting tempera-
ture, ramping rate, and duration of exposure. We conducted the literature search in Google
Scholar (Google, USA), using one or combinations of the following keywords: “ant,” “CTmax,”
“Critical Thermal Maximum,” “Formicidae,” “ramping rate,” “thermal tolerance,” and “ther-
mal limit”. All articles published from January 1944 to June 2020 were considered. In addition,
we searched the CTmax database, GlobTherm [20], to obtain the methodological details (e.g.,
year, study species, locality, ramping rate, starting temperature) corresponding to each study.
2.2. Sampling ants for thermal tolerance measurements
We sampled a diversity of ant species with the goal to test whether the results retrieved could
be generalized to a majority of taxa or limited to a particular subset of species. Several factors
have shown to correspond to variation in upper thermal limits within a taxon [14]; four of
these were used to guide our selection of ant species (S1 Table in S1 Appendix). First, we tar-
geted ants from different microhabitats (i.e., subterranean ants, above-ground foraging ants,
PLOS ONE | https://doi.org/10.1371/journal.pone.0265361 March 14, 2022
3 / 23
PLOS ONE Ramping rate of Critical Thermal maximum
Fig 1. Study diagram of testing implications of ramping rates in the measurement of Critical Thermal maximum. (A)
The first hypothesis (H1) examines if a positive relationship between the ramping rates and CTmax values exists for each
ant species at the intraspecific level. (B) The second hypothesis (H2) examines the interspecific variations of exposure
duration-based tolerance in function of the temperature treatments, we hypothesize that species assemblages show
different exposure durations in their CTmax (0.2) values but presenting consistence in their CTmax (0.5) and CTmax (1) values.
(C) Thermal performance framework of ectotherm on the basis of foraging behavior illustrate species activity in function
of the temperature increase, FTmax recorded in the field presents critical and act as a thermal threshold for the organisms;
the thermal performance curve is predicted based on the ant foraging activity in function of temperature. Through the
comparison between CTmax and FTmax, the results can examine will the species stop at their CTmax if the environmental
temperature reached their CTmax and provide a biologically relevant ranking of the CTmax values retrieved by different
ramping rates (0.2, 0.5 and 1˚Cmin-1). (D) Examination of the effect of ramping rates used to measure CTmax values on
PLOS ONE | https://doi.org/10.1371/journal.pone.0265361 March 14, 2022
4 / 23
PLOS ONE Ramping rate of Critical Thermal maximum
the phylogenic signals using Pagel’s λ and Blomberg’s K. (E) Examination of the effects of the ramping rates used to
measure CTmax values on community comparisons from different strata and habitats.
https://doi.org/10.1371/journal.pone.0265361.g001
and arboreal ants) to account for the important differences in the magnitudes and fluctuations
in temperature experienced by organisms living in different microhabitats, which vary in their
sun exposure and buffering effects (e.g., from the canopy or soil), and which thus correspond
to different thermal niches in the ecosystem [21]. The thermal tolerances of ant species are
known to correlate with the fluctuations and extremes in temperatures observed in these
microhabitats [22, 23]. Similarly, we included species collected from urban areas and forests to
account for the wide range of conditions to which ants may be exposed across the thermal
landscape. Second, we targeted ant species displaying substantially different individual body
sizes (S1 Table in S1 Appendix) to account for the effect of size on thermal tolerance, which
results from thermal inertia [24] and heat dissipation [25]; and also because size is known to
affect other physiological processes in ants, such as desiccation [26]. Third, we targeted species
from different phylogenetic clades to represent a range of species with different evolutionary
histories for ultimately testing the impact of different CTmax treatments on the results retrieved
(see below). The importance of phylogenetic constraints on determining upper thermal limits
in ectotherms is debated, with studies presenting contradicting results [6, 27, 28]. Finally, bio-
geographic origin provides a wider range of natural history and microclimate with species
evolving under different constraints (e.g., the natural barrier of temperature may be stronger
in tropical environments) [29]. To ensure that the trends retrieved do not represent local adap-
tations, our sampling included several exotic species whose native range and thus evolutionary
center differed from local species.
For this study, 27 ant species were collected in the field (S2 Table in S1 Appendix). The sam-
pling was conducted in both secondary forests and urban habitats in Hong Kong SAR
(22.3193˚N, 114.1694˚E) and Macao SAR (22.1987˚N, 113.5439˚E), China, during the dry sea-
son of 2018 and the wet season of 2019 (characterized by a tropical monsoonal climate).
Research and collecting permits were obtained from the local governments of Hong Kong
SAR and Macao SAR, and no protected species were sampled. Ant colonies were collected by
hand, using an insect aspirator. Whenever possible, three hundred individuals were collected.
Dynamic and static assays of thermal tolerance were conducted at least two hours after the col-
onies were transported from the field to the laboratory, but not more than 72 hours so as to
prevent the individuals from acclimating to laboratory conditions under longer periods, which
could modify the CTmax values measured [30]. For species with small colonies, the maximum
number of workers available were collected; the individuals were later separated into groups of
equal size and allocated to the different treatments, as far as possible including 15 individuals
(minor workers in polymorphic species) in each treatment. We also included multiple colonies
of Crematogaster rogenhoferi (N = 2) and Solenopsis invicta (N = 3) to examine the effects of
different ramping rates on their CTmax in dynamic assays and on their exposure duration-
based tolerance in static assays. Upon collection, all individuals were transported to the labora-
tory (24 ± 2˚C; 57.5 ± 5% relative humidity) for CTmax assays, and provided with wet cotton.
The different treatments in each assay were run sequentially to limit the effects of acclimation.
2.3 Dynamic assays (continuous changes in temperature to determine
thermal limits)
To experimentally quantify the upper thermal limits of the ants, we conducted dynamic assays
of their Critical Thermal maximum (CTmax) based on a continuous increase in temperature
over time. Specifically, we placed individual ants in an environment in which the temperature
PLOS ONE | https://doi.org/10.1371/journal.pone.0265361 March 14, 2022
5 / 23
PLOS ONE Ramping rate of Critical Thermal maximum
was increased progressively and steadily according to the predefined ramping rate (˚C per
minute). We used a general protocol for measuring CTmax in ants adapted from [31], with
three different ramping rates set as the experimental treatments (see below and S2 Fig in S1
Appendix). We measured the CTmax of 27 species with different microhabitat preferences (i.e.,
associations with different vertical habitat and strata) using a digital dry bath (Benchmark—
BSH1004, advertised accuracy ± 0.2˚C). For most species, 15 individuals were tested in each
treatment (i.e., a total of 45 individuals per species, S3 Table in S1 Appendix); each replicate
comprised an individual ant worker placed within a 2.0 mL Eppendorf tube, with its cap filled
with cotton to prevent individuals from taking refuge at this location. In addition, to limit the
stress experienced by individual ants and the release of defensive chemicals (e.g., formic acid)
which in the closed environment could be harmful, we guided the ants into their respective
Eppendorf tubes instead of picking them up with forceps. Each individual was used only once,
as a repeated exposure to high temperatures in multiple assays could cause heat injury, result-
ing in lower CTmax values. To ensure that the temperature recorded as the CTmax was the
indeed the temperature experienced by the individual, an extra digital thermometer (UEi Test
Instruments DT302 Dual Input IP67) with its sensor placed inside a supplementary Eppendorf
tube (that was also in the dry bath) was used as a temperature control; the reading from this
thermometer represented the most accurate temperature corresponding to the loss of muscle
control.
Ramping rates of 0.2, 0.5, and 1.0˚C min-1 were selected based on a systematic review of
previous studies (see 2.1 literature collection and the list in S2 Appendix); these also reflect
environmental fluctuations observed within terrestrial ecosystems [21, 32]. The small body-
sizes of ants make them ideal model organisms for tracking the effects of changes in environ-
mental temperature on body temperature. For instance, a study conducting a similar CTmax
assay showed that an ant’s body temperature tracked the temperature of the inner surface of
the Eppendorf tube in which it was placed, as ant stood on the surface and had limited heat
buffering abilities [14]. To account for the similarity between the environmental and experi-
enced temperatures by the individual, we also conducted a preliminary study using an infrared
thermal camera to measure the body-surface temperature of ants, as this would allow us to
infer their body temperatures [33]. This experiment showed that emissions of environmental
heat are easily transferred to individual ants, with their body-surface temperatures increasing
by up to 9–10˚C within one minute. It thus indicated that ant bodies possessed limited heat
inertia in our thermal assays (S3 Fig in S1 Appendix), and could easily track the highest ramp-
ing rate we tested (i.e., 1˚C per minute).
The starting temperature of each dynamic assay was set at 36˚C, a common starting tem-
perature for CTmax assays in ants [34]. In keeping with the protocols used in previous CTmax
studies for ants [e.g. 18, 35], the individuals were directly exposed to the experimental temper-
ature without being subjected to long periods of acclimation in the laboratory. Depending on
the treatment tested, we gradually increased the temperature at either 0.2, 0.5, or 1.0˚C min-1
(Fig 1), until the individual was observed to display a loss of muscle control (i.e., the onset of
spasms), and the corresponding temperature was recorded as the CTmax value of that individ-
ual. The loss of muscle control was defined as the thermal limit of the individual, because it is
more relevant to biological survival than the lethal temperature [36].
2.4 Static assays (exposure duration-based tolerance at constant
temperature)
Here, a static assay refers to the experimental use of a constant temperature to measure an
individual’s thermal tolerance in terms of the duration for which it can withstand being
PLOS ONE | https://doi.org/10.1371/journal.pone.0265361 March 14, 2022
6 / 23
PLOS ONE Ramping rate of Critical Thermal maximum
exposed to that temperature until it experiences a loss of muscle control. It should be noted,
however, this exposure duration-based definition of CTmax has not been considered in the pre-
vious studies [e.g., 5]. We used static assays to investigate species’ exposure duration-based tol-
erances in function of their different CTmax values retrieved from the dynamic assays using the
three different ramping rates. Ant workers from the same colony were first used in the
dynamic assays (see above), while other individuals from the same colony were used for the
static assays. This experimental order eliminated inter-colony variation in thermal tolerance
measurements of static and dynamic assays. Three of the 27 species tested were not used in the
static assays due to the limited number of individuals available.
In the static assay, we placed an individual ant in an environment with a fixed temperature
and measured the duration for which it could tolerate that condition. The experimental tem-
peratures used for the static assays were determined from the CTmax values obtained in the
dynamic assays at different ramping rates (0.2, 0.5 or 1.0˚C min-1), such each species was mea-
sured in three separate static assays (treatments), each using a different temperature (i.e.,
CTmax(0.2), CTmax(0.5), or CTmax(1)) that was specific to that species (Fig 1). For most species, we
used 15 individuals in each of the three treatments and another 15 individuals as a control
group (S4 Table in S1 Appendix). We used the same experimental setup and the same proce-
dure to record the loss of muscle control as those used in the dynamic assays. In the static
assays, the temperature remained unchanged, and we checked for each individual’s loss muscle
control in one-minute intervals over a maximum duration of 30 minutes. A period of 30 min-
utes represents a relatively long duration, which in natural conditions should allow sufficient
time for an individual to locate a thermal refuge, thus avoiding exposure to its upper thermal
limit. For instance, individuals of the desert ant, Cataglyphis bombycina, can only tolerate their
upper thermal limits for about 10 minutes in the field [37, 38].
2.5 Estimating foraging temperatures
From the perspective of niche theory, the foraging temperature represents the realized niche—
the range of temperatures at which a given species can be active at in the field—while CTmax is
considered the fundamental niche of a species’ physiological response to temperature. Ants are
social insects living within a nest (a climatic refuge), and thus make decisions on whether to
forage based on various factors, including temperature [39], with many species presenting
recruitment behaviors. As such, ants are model species for studying the behavioral responses
of animals to changes in temperature [37]. In order to collect data on the temperatures at
which ants forage, we used baits to observe the occurrences and recruitment patterns for a
wide range of species during the day. Each baiting station was composed of a white disk (Ø
4.7cm) laid on the ground surface and on which a 4 mm slice of sausage (1Valley Chef) was
deposited in its centre. The sausage was used as its composition includes proteins, lipids and
carbohydrates, which are attractive to numerous species; moreover, the calibrated and circular
size provides a standardized amount of food available to the species between replicates. Spe-
cies’ foraging temperatures were recorded as the ground-surface temperatures at which they
were observed to recruit to the baits, and were measured using an infrared thermometer
(Fluke 62 MAX+) from one meter above the ground. We took the average of temperatures
measured at four cardinal locations from a distance of 2 to 5 cm from the edges of the white
disk (S5 Fig in S1 Appendix). Ground-surface temperatures are the most relevant environmen-
tal temperatures experienced by ants; with maximum foraging temperature representing the
upper thermal limit observed in the field [40].
From 2015 to 2018, > 11817 baits were installed in different localities in Hong Kong from
1000–1600 HRS, which corresponds to the warmest period of the day. The baits were
PLOS ONE | https://doi.org/10.1371/journal.pone.0265361 March 14, 2022
7 / 23
PLOS ONE Ramping rate of Critical Thermal maximum
positioned along transects, with a minimum distance of 10 m separating two baits. Baiting was
conducted throughout each year, and baits were placed predominantly within open urban and
suburban environments. The baits were left to operate for a period of 1 to 2 hours, and the ant
activity at each bait was noted every 15 minutes. Ant foraging activity was noted on 10,157
baits. We used the foraging temperature maximum of seven dominant species (Anoplolepis
gracilipes, Monomorium chinense, Paratrechina longicornis, Pheidole megacephala, P. parva,
Solenopsis invicta and Tapinoma melanocephalum), which each had at least 430 records (max.
= 1764). These seven species were observed at a total of 7,692 baits across multiple seasons in
Hong Kong. These species were selected on the basis of the large number of records available,
and their abundances in open habitats, where ground-surface temperatures can be very high
(the maximum value recorded during our study was 66.2˚C).
2.6 Statistical analyses
2.6.1 Ramping rates. First, individual CTmax was averaged by species identity and ramp-
ing rate. To assess the importance of the ramping rate in CTmax measurements, we used a lin-
ear mixed model, with species-average CTmax as the response, and ramping rate as the sole
predictor. To control for phylogenetic dependence, we specified species nested within genus
and subfamily as random intercepts. Second, we performed a Bartlett’s test to assess whether
the 24 ant species exhibited more heterogenous exposure duration-based tolerances of temper-
atures corresponding to their CTmax values measured at a low ramping rate in the static assays.
Thirdly, we performed ANOVA and Tukey’s tests to assess if increasing the ramping rate led
to a higher observed CTmax for each species; and preformed Kruskal-Wallis and Dunn’s tests
to assess whether each species had a shorter exposure duration-based tolerance of the tempera-
ture corresponding to its CTmax that was measured using a fast ramping rate. Finally, we per-
formed 27 simple linear models for dynamic assays to examine the CTmax impacted by the
ramping rate in the supporting information, and reported the adjusted R2 values in explaining
intraspecific variations within the dataset (see S5 Table in S1 Appendix).
2.6.2 Critical Thermal maximum vs. foraging temperature maximum. For comparing
CTmax and FTmax in each species, we used their absolute difference and calculated the mean
and standard deviation for each ramping rate (FTmax of each species was defined as the maxi-
mum foraging temperature observed across all the foraging temperatures recorded). In addi-
tion, comparisons between CTmax and the top 1% of FTs was also conducted to account for the
variation in foraging temperature maximum that can be observed within different ant popula-
tions. We also used a linear model with the ramping rate and species identity as the predictors
and the absolute differences as the response to examine how both factors (i.e., ramping rate
and interspecific variation) affected species’ CTmax and FTmax. Because we only had sufficient
foraging records (N > 500) for seven species, and these species generally represented different
genera and subfamilies, we used a simple linear model for the comparative analyses.
2.6.3 Phylogenetic signal analyses. To test if the choice of ramping rate would affect the
results of a phylogenetic analysis, we generated a genus-level phylogeny comprising the genera
of our study species. Here, we used a backbone tree from a published genus-level phylogeny
[41] and applied tree pruning to keep a single species for every genus. Then we simulated
1,000 species-level phylogenies using a Yule (pure-birth) process with the function genus.to.
species.tree in the Rpackage “phytools” (Revell 2012). For each species-level phylogeny, we
used Pagel’s λ and Blomberg’s K to examine the phylogenetic signals in CTmax generated from
the different treatments (i.e., ramping rates). We also tested whether λ and K were significantly
different from random using a likelihood ratio test and a randomization test (1,000 randomi-
zations), respectively.
PLOS ONE | https://doi.org/10.1371/journal.pone.0265361 March 14, 2022
8 / 23
PLOS ONE Ramping rate of Critical Thermal maximum
2.6.4 Habitat and microhabitat comparisons. Ant species experience different variations
in temperature based on the habitats and microhabitats in which they live [14, 35]. For habitat
and microhabitat comparisons of species’ CTmax, we again used linear mixed models to ana-
lyze how a change in the ramping rate used could alter conclusions about the CTmax of species
from different habitats. We built one model to compare species from different vertical strata
(i.e., subterranean ants, above-ground foraging ants, and arboreal ants) and another to com-
pare those from different habitats (i.e., urban and forest). Each model included CTmax as the
response, and ramping rate, strata (or habitat), and their interaction as the predictors. We
included species nested within genus and subfamily as random intercepts. We further con-
ducted pairwise comparisons between strata or habitats within the same ramping rate and
obtained the Tukey-adjusted p-value, which indicated the detection of a significant interaction.
Finally, we assessed whether habitat/strata differences in CTmax varied with the use of different
ramping rates by setting the null value as the weakest effect size (instead of zero).
All statistical analyses were performed in R version 3.6.2 [42]. Bartlett’s test of homogeneity
of variances was conducted using the function bartlett.test. Linear models were performed
using the function lm. Linear mixed models were performed using lme4 and lmerTest, with
Tukey’s post-hoc comparisons performed using emmeans [43]. White-adjusted p-values were
obtained using the function Anova in R-package “car” [44] to control for the effect of a vari-
ance in heteroscedasticity.
3. Results
3.1 Literature review of studies on ant species’ upper thermal limits
We retrieved a total of 51 publications (49 studies using dynamic assays and two using static
assays) investigating ant species’ upper thermal limits between January 1944 to June 2020 (see S2
Appendix). In total, 20 different values of ramping rate were used, with 0.2˚C min-1 (13/49; 27%)
and 1.0˚C min-1 (22/49; 45%) being the most widely used ramping rates (S1 Fig in S1 Appendix).
3.2 Dynamic assays (Critical Thermal maximum CTmax)
A total of 1,743 individuals from 27 species were used for dynamic assays. In the intraspecific
models, the CTmax values retrieved were dependent on the ramping rate used; with fast ramp-
ing rates resulting in significantly higher CTmax values than slow ramping rates for all species
(p-value < 0.001, Fig 2, S3 Table in S1 Appendix). Differences in the CTmax values retrieved
between the slow ramping rate (i.e., 0.2˚C min-1) and the fast ramping rate (i.e., 1˚C min-1)
averaged 4.13˚C, ranging from 1.40˚C in Aenictus sp. laeviceps gp. to 6.47˚C in Crematogaster
rogenhoferi (S3 Table in S1 Appendix). The CTmax values for most (i.e., 26 out of 27) species
were correlated with the ramping rates used in the intraspecific models (adjusted R2 = 0.629
[0.229–0.95], p-value < 0.05; S5 Table in S1 Appendix), with the exception of Anochetus risii
(Adjusted R2 = 0.090, p-value = 0.052; S5 Table in S1 Appendix). Mixed models including data
of all species revealed that CTmax values were positively correlated with ramping rate (Marginal
R2 = 0.126, p-value < 0.001; Table 1). Species identity was a strong predictor of variation in
CTmax values; including this variable as random effects in the linear mixed model led to a
marked improvement in explanatory power (Conditional R2 = 0.942; Table 1).
3.3 Static assays (exposure duration-based tolerance to a constant
temperature)
A total of 1,191 individuals from 24 species were used for static assays. The loss of muscle con-
trol was observed within 30 minutes in all individuals of the 24 species exposed to the
PLOS ONE | https://doi.org/10.1371/journal.pone.0265361 March 14, 2022
9 / 23
PLOS ONE Ramping rate of Critical Thermal maximum
Fig 2. Results of dynamic assays. Line plots of CTmax values measured in function of three ramping rates used for 27 ant species found in
function of their vertical stratification (arboreal, ground and subterranean strata).
https://doi.org/10.1371/journal.pone.0265361.g002
temperatures corresponding to their CTmax(0.5) and CTmax(1.0). However, this was not the case
in static assays for temperatures corresponding to species’ CTmax (0.2), where for four species,
only 40–96% of the 15 individuals tested displayed this condition (S4 Table in S1 Appendix).
Table 1. Outcomes of the linear mixed model examining the relationship between parameters (CTmax, ramping rate and species identity as well as the genus and
subfamily identity) by linear mixed model for the dynamic assay with white-adjusted p-value.
CTmax ~ Ramping rate + (1| Subfamily /Genus/ Species)
Ramping rate
Marginal R2: 0.126 and Conditional R2: 0.942
181.65
Chi square
Groups
Species: (Genus: Subfamily)
Genus: Subfamily
Subfamily
Residuals
Intercept
Ramping rate– 0.5˚C min-1
Ramping rate– 1˚C min-1
Intercept
Intercept
Intercept
Estimate
43.7481
1.6325
3.8025
https://doi.org/10.1371/journal.pone.0265361.t001
DF
2
p-value
<0.001
Random effects
Variance
Std. Dev.
1.707
7.702
6.542
1.122
Fixed effects
Std. Error
1.1766
0.2831
0.2831
1.307
2.775
2.558
1.059
df
6.805
55.018
55.018
t-value
37.182
5.767
13.433
p-value
<0.001
<0.001
<0.001
PLOS ONE | https://doi.org/10.1371/journal.pone.0265361 March 14, 2022
10 / 23
PLOS ONE Ramping rate of Critical Thermal maximum
Fig 3. Results of static assays. Mean exposure duration-based tolerance values (±SE) of 24 ant species for three temperatures based on
the values retrieved in the CTmax (0.2, 0.5, and 1.0) treatments. Right y-axis refers to the duration tolerance the ants were maintaining their
muscle control, and left y-axis refers to the log-transformed duration tolerance values.
https://doi.org/10.1371/journal.pone.0265361.g003
The exposure duration-based tolerances of 24 species were affected by the specific ramping
rates used to measure their CTmax, with different ramping rates leading to significantly differ-
ent exposure durations (p-value < 0.001, Kruskal-Wallis test and Dunn’s test). At CTmax(0.2),
species’ exposure durations (i.e., the duration required to initiate a loss of muscle control) ran-
ged from 2.3 min. (A. risii) to 17.8 min. (Paratrechina longicornis) (Mean ± SD: 8.5 ± 4.3 min);
this represented the largest interspecific variation in exposure duration observed among the
three temperature treatments tested (CTmax (0.2, 0.5 and 1.0), S4 Fig in S1 Appendix). At CTmax
(0.5), species’ exposure durations ranged from 1.5 min. (A. risii) to 5.8 min. (Solenopsis invicta)
(Mean ± SD: 3.6 ± 1.5 min). At CTmax (1.0), almost all species showed similar exposure dura-
tions (Mean ± SD: 2.2 ± 0.7 min), with the longest exposure duration being 3.6 min. (Pheidole
parva) (S4 Table in S1 Appendix and Fig 3). For each static assay, the responses of the 24 spe-
cies tested showed the smallest interspecific variation in terms of exposure tolerance duration
(SD: 0.685 min.) for values retrieved from the CTmax (1.0) treatment. The 24 ant species dis-
played high interspecific variation in exposure duration in static assays for temperatures corre-
sponding to their CTmax (0.5) and CTmax (0.2) (SD: 1.5 at CTmax (0.5) and SD: 4.3 at CTmax (0.2), S4
Table in S1 Appendix). Additionally, their exposure durations displayed unequal variances
across the three static assays corresponding to their CTmax (1.0), CTmax (0.5) and CTmax (0.2) (p-
value < 0.05 in the Bartlett’s test of homogeneity of variances, S4 Fig in S1 Appendix).
3.4 Foraging temperature maximum vs. Critical Thermal maximum
For five out of the seven species tested, CTmax (1.0) was the closest to their FTmax value mea-
sured (Fig 4) as well as within the top 1% of their FTs. Specifically, absolute differences
between FTmax and CTmax values were lowest when a 1˚C min-1 ramping rate was used to
PLOS ONE | https://doi.org/10.1371/journal.pone.0265361 March 14, 2022
11 / 23
PLOS ONE Ramping rate of Critical Thermal maximum
Fig 4. Critical Thermal maximum vs. foraging temperature records. Upper plot showing the difference between
FTmax and CTmax (0.2, 0.5, and 1.0) and errors bars as standard deviation of the CTmax values. Lower plot shows the range
and distribution frequency of all surface temperatures measured near baiting stations during the sampling period,
independently of the presence of ants or not. Vertical lines indicate the FTmax values measured for each species in the
field.
https://doi.org/10.1371/journal.pone.0265361.g004
measure CTmax in the dynamic assay (Mean ± SD: CTmax (1.0): 2.39 ± 1.41; CTmax (0.5):
3.79 ± 2.13; CTmax (0.2): 5.60 ± 2.57, Fig 4 and Table 2); a similar trend was observed when
FTmax was defined as the top 1% of a species’ FTs (Table 2). Overall, the comparison between
FTmax/1% top FT and CTmax values was significantly different between the three ramping rates
(p-value < 0.05 and adjusted R2 = 0.489 for FTmax; p-value < 0.05 and adjusted R2 = 0.471 for
1% top FT; Table 2), indicating a better overall performance in reconciling field and laboratory
data.
3.5 Phylogenetic signals
Methodological approaches used for measuring CTmax resulted in differences in the signifi-
cance of phylogenetic signals. Species’ CTmax values retrieved at 0.2, 0.5, and 1.0˚C min-1 dis-
played different patterns of variation across the ant phylogeny (Fig 5). This effect of ramping
rate on phylogenetic patterns was further confirmed in a genus-level polytomy tree and phylo-
genetic analyses. Specifically, CTmax displayed a stronger phylogenetic signal when measured
PLOS ONE | https://doi.org/10.1371/journal.pone.0265361 March 14, 2022
12 / 23
PLOS ONE Ramping rate of Critical Thermal maximum
Table 2. The comparsion between CTmax values retrevied from three different ramping rates and FTmax /1% top
FT (sample size of each species shown on Fig 4), the lightly gray color background refers the CTmax value, closer to
the FTmax for those species.
| FTmax—CTmax | (˚C)
0.5˚C min-1
1.1
1˚C min-1
4.4
Species
Anoplolepis gracilipes
Monomorium chinense
Paratrechina longicornis
Pheidole megacephala
Pheidole parva
Solenopsis invicta
0.2˚C min-1
0.1
7.1
4.4
7.7
8.1
6.8
5.8
1
6.3
5.9
4
2.4
Tapinoma melanocephalum
Mean of | FTmax—CTmax | (˚C):
Standard deviation
| FTmax—CTmax | ~ Ramping rate + Species identity (p-value < 0.05, adjusted R2 = 0.4885)
2.57
2.32
5.60
3.79
5
Anoplolepis gracilipes
Monomorium chinense
Paratrechina longicornis
Pheidole megacephala
Pheidole parva
Solenopsis invicta
| 1% top FT—CTmax | (˚C)
5.8
1
6.3
5.9
4
1.1
7.1
4.4
7.7
8.1
6.8
0.1
7.1
Tapinoma melanocephalum
Mean of | 1% top FT—CTmax | (˚C):
Standard deviation
| 1% top FT—CTmax | ~ Ramping rate + Species identity (p-value < 0.05, adjusted R2 = 0.4708)
4.27
2.61
2.15
5.90
5.8
3
1.4
3
3.7
0.8
0.4
2.39
1.41
3
1.4
3
3.7
0.8
4.4
3
2.76
1.16
https://doi.org/10.1371/journal.pone.0265361.t002
at a ramping rate of 1.0˚C min-1, as indicated by higher values of Pagel’s λ (Mean = 0.991) and
Blomberg’s K (Mean = 0.906) (Table 3), but a weaker phylogenetic signal when the other ramp-
ing rates (i.e., 0.2 and 0.5˚C min-1) were used for measurement, as indicated by the lower val-
ues of Pagel’s λ (0.827–0.862) and Blomberg’s K (0.768–0.781) (Table 3). In addition, the
proportion of simulated trees that detected significant phylogenetic signals increased when a
higher ramping rate was used to measure CTmax (S6 Fig in S1 Appendix).
3.6 Habitat and microhabitat comparisons
The measured CTmax values of ant species were important for distinguishing their ecology. For
instance, for all three ramping rates, the CTmax values of ant species occupying the subterra-
nean, ground, and arboreal strata were significantly different (p-value < 0.01; Table 4). Specifi-
cally, arboreal species showed consistently higher thermal tolerances than ground and
subterranean species (p-value < 0.05; Table 4). The ramping rate used did not influence the
relationships between the CTmax values of species from different vertical strata, as no signifi-
cant interaction between strata and ramping rate was detected (p-value = 0.559). There was,
however, a significant interaction between habitat and ramping rate; CTmax values of species
collected within urban habitats (N = 9) were significantly higher than those of species collected
in forested habitats (N = 18, p-value < 0.001) for all three ramping rates tested. Nevertheless,
we found that when a ramping rate of 1.0˚C min-1 was used, the effect size of a difference in
habitat on species’ CTmax (i.e., 5.16˚C) was significantly larger than that observed when a
ramping rate of 0.2˚C min-1 was used (i.e., 3.42˚C) (p-value = 0.003), which in turn was
PLOS ONE | https://doi.org/10.1371/journal.pone.0265361 March 14, 2022
13 / 23
PLOS ONE Ramping rate of Critical Thermal maximum
Fig 5. Critical Thermal maximum (CTmax) of 27 ant species in function of the phylogeny. Color shading corresponds with the magnitude
of thermal tolerance measured with different ramping rates, A. 0.2˚C min-1, B. 0.5˚C min-1, C. 1.0˚C min-1. Ant illustrations credited to Mr
Runxi Wang with permission.
https://doi.org/10.1371/journal.pone.0265361.g005
marginally different (p-value = 0.07) from the effect observed when a 0.5˚C min-1 ramping
rate was used (i.e., 4.34˚C).
4. Discussion
There has been an increasing interest the forecasting of species’ tolerances to warming envi-
ronments based on their CTmax [19, 45]. Although design of ramping rate has been discussed
since the development of CTmax [5, 7, 20], there is no consensus on what constitutes a suitable
Table 3. Phylogentic signals, Pagel’s λ and Blomberg’s K, for dynamic assay 0.2˚C min-1, 0.5˚C min-1, 1.0˚C min-1 (1000 times).
Pagel’s lambda
Blomberg’s K
Pagel’s λ:
p value:
p value < 0.05:
p value < 0.01:
Blomberg’s K:
p value:
p value < 0.05:
p value < 0.01:
https://doi.org/10.1371/journal.pone.0265361.t003
0.2˚C min-1
0.700–1.151 (Mean: 0.862)
0.5˚C min-1
0.669–1.139 (Mean: 0.827)
1.0˚C min-1
0.879–1.191 (Mean: 0.991)
0.0013–0.0644
0.0008–0.0521
0.0017–0.0215
96.30%
7.20%
99.60%
8.90%
100%
42.10%
0.002–0.980 (Mean: 0.781)
0.002–0.965 (Mean: 0.768)
0.014–1.023 (Mean: 0.906)
0.0003–0.5588
0.0001–0.5768
0.0004–0.2678
75.50%
31.20%
72.60%
31.70%
94.10%
85.80%
PLOS ONE | https://doi.org/10.1371/journal.pone.0265361 March 14, 2022
14 / 23
PLOS ONE Table 4. Comparison of species CTmax values from habitats and microhabitats respectively, the linear mixed models examine the relationship between parameters
(CTmax, ramping rate and species identity) by linear mixed model for the dynamic assay with white-adjusted p-value; with species identify nested within genus and
subfamily identity.
CTmax ~ Strata + Ramping rate + (1| Subfamily/ Genus/ Species)
Ramping rate of Critical Thermal maximum
Intercept
Strata
Ramping rate
Strata: Ramping rate
Ramping rate 0.2˚C min-1
A.–G
A.–S.
G.–S.
Ramping rate 0.5˚C min-1
A.–G
A.–S.
G.–S.
Ramping rate 1˚C min-1
A.–G
A.–S.
G.–S.
Intercept
Habitat
Ramping rate
Habitat: Ramping rate
Ramping rate 0.2˚C min-1
Forest—Urban
Ramping rate 0.5˚C min-1
Forest—Urban
Ramping rate 1˚C min-1
Forest—Urban
Chi square
1609.786
18.021
60.72
3.018
Estimate
5.465
7.269
1.804
Estimate
5.573
6.508
0.935
Estimate
5.845
8.194
2.349
Df
1
2
2
4
SE
1.55
2.24
2
SE
1.55
2.24
2
SE
1.55
2.24
2
p-value
<0.001
<0.001
<0.001
0.555
Df
18.6
23.6
24.6
Df
18.6
23.6
24.6
Df
18.6
23.6
24.6
CTmax ~ Habitat + Ramping rate + (1| Subfamily/ Genus/ Species)
Chi square
2352.643
32.831
216.135
16.933
Estimate
3.42
Estimate
4.34
Estimate
5.16
Df
1
1
2
2
SE
0.611
SE
0.611
SE
0.611
p-value
<0.001
<0.001
<0.001
<0.001
Df
63.2
Df
63.2
Df
63.2
p-value
0.0062
0.0093
0.6454
p-value
0.0053
0.024
0.8875
p-value
0.036
0.034
1.172
Null
3.42
Null
3.42
Null
3.42
For the strata, A. = Arboreal, G. = Ground, S. = Subterranean.
https://doi.org/10.1371/journal.pone.0265361.t004
p-value
0.5012
p-value
0.0691
p-value
0.003
ramping rate, and arguments for slow as well as fast ramping rates have been made from vari-
ous ecological, physiological and methodological aspects [5, 7, 8, 11, 46]. Here, using a combi-
nation of dynamic and static assays on a wide range of ant species with distinct ecological,
morphological, phylogenetic and biogeographic characteristics, our results evidence a consis-
tent trend between the ramping rate used and the CTmax values retrieved, suggesting that a
major part of the variation observed in (and among) species’ CTmax values results from differ-
ences in the methodological approaches that have been used [5, 7, 11]. Furthermore, we pro-
pose that CTmax values retrieved from a fast ramping rate (1.0˚C min-1) are the most
biologically relevant, and evidence this using an additional experimental approach (i.e., static
assays) as well as field observations of species’ FTmax values. Overall, our study provides impor-
tant experimental and field-based evidence to guide the selection of a reliable ramping rate for
CTmax measurements of ant species. This approach may also be applied to numerous other ter-
restrial ectotherms.
PLOS ONE | https://doi.org/10.1371/journal.pone.0265361 March 14, 2022
15 / 23
PLOS ONE Ramping rate of Critical Thermal maximum
4.1 Literature review of Critical Thermal maximum in ants
In reviewing the literature on ant species’ upper thermal limits, we find that an overwhelming
number of studies (49/51) have used the dynamic assay approach, confirming previous obser-
vations of studies across a wide range of taxa [5]. Our review also shows that an extensive vari-
ety of ramping rates (0.05–2.0˚C min-1) have been used to measure ant species’ CTmax.
Furthermore, the two most frequently used ramping rates—0.2 and 1.0˚C min-1—exhibit a
fivefold difference in magnitude (S1 Fig in S1 Appendix). One emerging issue with such meth-
odological differences among studies is that the results obtained from assessments using differ-
ent ramping rates are not directly comparable [10]. This also implies that unless CTmax values
are somehow corrected for ramping rate, the conclusions of meta-analyses may be unreliable,
as the trends observed are likely to be strongly impaired by methodological artifacts.
4.2 Use of ramping rate in CTmax assay
Since the implementation of CTmax measurements, there has been much debate over the
ramping rate used, as well as the tradeoffs between the use of slow versus fast ramping rates [5,
20, 46]. The use of different ramping rates to measure species CTmax is controversial and has
not been standardized [17]. Across studies of ant species’ thermal tolerances, the most fre-
quently used ramping rates have been 0.2, 0.5, and 1˚C min-1 (S1 Fig in S1 Appendix). Among
these, the results from our experiments on 27 ant species clearly support the use of the faster
ramping rates (i.e., 0.5 and 1˚C min-1) rather than the slow ramping rate of 0.2˚C min-1
(Table 4). The use of a slow ramping rate (i.e., 0.2˚C min-1) to measure species’ CTmax resulted
in a failure to forecast their activities in response to ground temperatures (see: foraging tem-
perature section 3.4 and Fig 4). In addition, when species were exposed to temperatures corre-
sponding to their CTmax measured at this slow ramping rate in the static assays, most species
remained active even after long periods of exposure (> 10 min., static assay section 3.3 and Fig
3), suggesting that those temperatures were not representative of species’ critical limits [see 36
for Critical Thermal Maximum]. A biologically relevant CTmax of a species should force indi-
viduals of that species to seek thermal refuges as soon as possible, and therefore correspond to
a relatively short exposure duration during a static assay.
For instance, when exposed to very hot temperatures in the field (67–70˚C), individuals of
the Saharan Silver Ant (Cataglyphis genus) limit their foraging activities to approximately 10
minutes before returning to thermal refuges [37, 38]. A period of several minutes should repre-
sent a critical duration for individual ants to be exposed to their thermal limits, as heteroge-
neous habitats provide ample opportunities for individuals to locate thermal refuges. Our
static assays also revealed high heterogeneity in the CTmax of individual ant species, with some
individuals of some species remaining active even after being exposed to temperatures corre-
sponding to their CTmax for over 30 minutes. Except in a few completely open habitats (e.g.,
deserts), such long durations of exposure are unlikely to be a major constraint on the foraging
activities of small ectotherms. Furthermore, in the static assays, the ant species showed impor-
tant variation in their exposure duration-based tolerance under temperatures corresponding
to their CTmax(0.2) (Fig 1B and Fig 3). If species’ CTmax(0.2) were used to compare their thermal
limits, the 24 species exposed to their CTmax(0.2) temperatures in the static assays would display
high interspecific variation in exposure duration-based tolerance (S4 Table in S1 Appendix).
Such interspecific variation in exposure duration-based tolerance was substantial, even
between species that displayed the most similar CTmax values across dynamic essays using dif-
ferent ramping rates. For instance, while A. risii and Ooceroea biroi displayed similar CTmax
(0.2) values in dynamic assays (at 39.8˚C and 40.1˚C, respectively), they differed extensively in
their exposure duration-based tolerances of these temperatures in static assays; O. biroi could
PLOS ONE | https://doi.org/10.1371/journal.pone.0265361 March 14, 2022
16 / 23
PLOS ONE Ramping rate of Critical Thermal maximum
tolerate 40.1˚C for 17.7 minutes but A. risii could only tolerate 39.8˚C for 2.3 minutes (S4
Table in S1 Appendix). In contrast, when species were exposed to temperatures corresponding
to their CTmax (0.5) and CTmax (1) in the static assays, less variation in their exposure duration-
based tolerance was observed, allowing for a more direct comparison of their upper thermal
limits (Fig 3).
Some studies measured environmental temperature of some species and have justified the
use of slow ramping rates by referring to temperature fluctuation of some species micro-habi-
tat [5, 7, 14, 15], and proposed that in interspecific comparisons of ectotherms, measurements
of species’ upper thermal limits should relate to their thermal niches [46]. However, when
slower (as compared to faster) ramping rates are used in dynamic assays, species are usually
exposed to increasing temperatures for longer periods of time before reaching their upper
thermal limits [5]. We therefore recommend the use of ramping rates on the basis of exposure
duration-based tolerance to examine ramping rate effects and the measured CTmax (Fig 1A &
1B) including static assays (Fig 3) and a comparison with field data on observed foraging tem-
peratures when available (Table 2, Fig 4).
Ideally, a species’ CTmax should correspond to the temperature that induces a heat-coma in
individuals of that species [25, 36]. This conceptualization of CTmax will provide a functional
trait that is useful for interpreting species’ use of habitats and microhabitats. Our results show
that the ramping rates used in dynamic assays of ant species’ CTmax can directly shape the rela-
tionships observed among the CTmax of different species assemblages that are classified based
on habitats or microhabitats in linear mixed models (Table 4). In particular, the ramping rates
used in measurements of CTmax can strongly influence conclusions about the differences
between the upper thermal limits of species from urban habitats and those from forest habitats.
Such methodological issues can bias general conclusions about the ecology of species based on
species’ CTmax in future studies.
4.3 Foraging temperature maximum
Foraging temperature maximum (FTmax) is one of the most intuitive measures of species’
upper thermal limits. It is derived from field observations and the identification of a maximum
temperature threshold after which individuals of a species suspend their foraging activity [37].
Here, we examined species’ behaviors in the field and compared those with their CTmax values
retrieved using different ramping rates. Although the concepts of CTmax and FTmax both relate
to species’ upper thermal limits, they are markedly different. CTmax is supposed to represent a
species’ maximum physiological threshold [36]; and thus it is expected that a species’ CTmax
should exceed its FTmax, at which individuals cease to actively forage. While a species’ FTmax is
affected by both its abiotic and biotic interactions [39], its CTmax is not. At a temperature
exceeding a species’ CTmax, individuals of that species should lose muscle control, and display
an onset of spasms and heat-shock [5]. If a species’ CTmax is substantially lower than its FTmax,
it can likely forage at temperatures exceeding its CTmax; this represents a biological underesti-
mation of a species’ thermal limit under laboratory conditions. Our results show discrepancies
between species’ FTmax and CTmax values, which are most extensive when the slowest ramping
rate (0.2˚C min-1) is used to measure CTmax. At this ramping rate, six species (out of 7) display
CTmax values that are 4.4˚C to 8.1˚C lower than their FTmax. Such gaps question the biological
relevance of using a slow ramping rate to measure CTmax. In contrast, the majority of CTmax
values retrieved using the fastest ramping rate (1˚C min-1) aligned more precisely with species’
FTmax values (Fig 4 and Table 2).
The ramping rate used affects the values of CTmax measured and the forecasting of species’
activities. Therefore the use of a reliable CTmax is paramount and should refer to a biologically
PLOS ONE | https://doi.org/10.1371/journal.pone.0265361 March 14, 2022
17 / 23
PLOS ONE Ramping rate of Critical Thermal maximum
relevant thermal limit. Using an unreliable CTmax may result in a mischaracterization of the
activity patterns and distribution of a species. We illustrate this problem with the case of the
Red Imported Fire Ant, S. invicta, which has established populations in Hong Kong following
introductions from the USA [47]. The CTmax of this species in the USA has been measured
with the use of slow ramping rates such as 0.12 or 0.2˚C min-1 [16, 48] as well as a faster ramp-
ing rate of 1.0˚C min-1 [16, 49]. Coincidently, our observations of the foraging activity of this
species in the field (N = 1,398) suggest a thermal threshold (FTmax) that corresponds with its
CTmax measured at 1.0˚C min-1 in both the USA and in Hong Kong (this study), with a mar-
ginal difference of 0.71˚C on average. In comparison, the species’ FTmax exceeds values of its
CTmax measured at slower ramping rates by an average of 4.77˚C. While few intensive studies
of ectotherm species’ activity-temperature relationships—such as those of S. invicta—are avail-
able [48, 50, 51], our field observations of this species demonstrate the importance of identify-
ing biologically relevant CTmax values that can predict species’ activity patterns. Such
approaches are not only important for understanding the ecology of individual species but also
for characterizing whole assemblages (Table 2, Fig 3).
Measuring FTmax is challenging because it is difficult to control the environmental condi-
tions and to control for the effects of biotic interactions such as competition, which may ulti-
mately affect the values measured [52]. Including comparisons between CTmax and FTmax in
our CTmax framework (Fig 1) allow to capture field and biological relevance of upper thermal
limit. Often, the habitat in which a species is most encountered is unlikely to experience tem-
peratures that are close to that species’ thermal limit; this is especially true for species living in
the leaf-litter layer of closed-canopy forests. For instance, the lowest CTmax value we observe
(independently of the ramping rate used) among the ant species from Hong Kong is 38.3˚C
(i.e., a soil/litter-dwelling ant, Strumigenys minutula). This temperature remains 0.5˚C higher
than the highest air temperature recorded in Hong Kong’s history (Hong Kong observatory).
Moreover, in the absence of direct solar radiation, soil temperatures are similar or lower than
air temperatures [53]. Thus, for some habitats or microhabitats, air temperatures measured in
the field cannot be used directly as maximum temperature thresholds of the species present
and the microclimates they experience. Knowledge of microclimates is paramount for under-
standing species’ thermal niches [54]. Our observations of ant species’ foraging activity pat-
terns across a wide range of surface temperatures (i.e., 10–66.2˚C) provide crucial information
on the microclimates that these species experience. Therefore, measurements of species’ upper
thermal limits using experimental approaches (i.e., dynamic assays for CTmax) remain neces-
sary so long as they can estimate biologically relevant thresholds for these limits.
4.4 Phylogenetic signals
The methods used to measure eco-physiological traits can strongly influence the values
retrieved, bias empirical findings, and cause ecological phenomena to be misinterpreted. Our
study shows that CTmax is strongly affected by the ramping rate used, and that the use of a spe-
cific ramping rate use can induce significant biases in subsequent analyses of species’ phyloge-
netic relationships (Table 3). A number of studies have shown that upper thermal limit is
phylogenetically conserved in ectotherms, such as in ants [34, 55], fruit flies [56], and lizards
[57]. In contrast, other studies retrieved no evidence for a relationship between phylogeny and
thermal tolerance. Notably, a slow ramping rate of 0.2˚Cmin-1 was used in these studies [28,
58]. To the best of our knowledge, our study represents the first to compare phylogenetic sig-
nal in species’ CTmax to the ramping rate used in CTmax measurements. Our results show that
the detection of a phylogenetic signal (i.e., Pagel’s λ and Bloomberg’s K) is directly influenced
by the methodology used to measure CTmax (Table 3). The findings also show that apart from
PLOS ONE | https://doi.org/10.1371/journal.pone.0265361 March 14, 2022
18 / 23
PLOS ONE Ramping rate of Critical Thermal maximum
differences in the topography of the phylogenetic tree and the species pool, differences in
ramping rate could also explain these inconsistencies of the phylogenetic signals. Although
upper thermal limit has been shown to be strongly constrained by evolutionary history [55],
the ramping rates used to collect the data should also be considered as an important cofound-
ing factor in tests for various evolutionary hypotheses. In such analyses, one should avoid the
use of a slow ramping rate (e.g., 0.2˚C min-1) to measure CTmax values (at least for ants), as
well as the combination of CTmax values originating from different methodologies.
4.5 Predicted CTmax from biophysical model vs. experimental CTmax from
physiological measurement
Recent studies have provided a mathematical model, based on thermal tolerance landscapes,
to predict the CTmax value for a given species independently of the ramping rate used [11, 12].
As mentioned by the authors [12], this model has been developed based on the study of eleven
Drosophila species, and should therefore be tested with other ectotherms. While not central to
our study, our dataset provides a good opportunity to test this model and understand its gener-
ality. Our results, however, provide rather limited support to the model, with predicted and
observed values diverging substantially in most cases tested (see details results in S3 Appen-
dix). It thus appears that the biophysical model should be used cautiously, and may not be suit-
able for a majority of the ant species tested here. Although this model can provide important
insights into species’ ecophysiology, it has limited value in helping to identify a satisfactory
ramping rate that will facilitate biologically relevant CTmax measurements. Ultimately, our
study supports the need for further validation and examination in other ectotherm groups
[12], with further research needed for identifying pertinent ramping rates in ecological studies.
5. Conclusion
The use of CTmax to study ectotherms has significantly increased in the past decades, and its
application has yielded multiple predictions about the impacts of global change [17, 59, 60]. As
shown here as well as in other studies of ectotherms (Ants: 13; 16. Drosophila fly: 12, 16), the
CTmax values observed are, however, largely affected by the ramping rates used and identity of
species used in dynamic assays. Ideally, CTmax values should facilitate comparisons between
studies and be grounded in biological relevance. Our study thus establishes a new and hybrid
method to address this goal, integrating dynamic and static assays in addition to comparisons
with field data (FTmax), to identify a reliable ramping rate for ant species. Each approach we
use serves to test whether a given ramping rate produces biologically relevant CTmax values.
Our results indicate that a ramping rate of 1.0˚C min-1 is the most appropriate for measuring
the CTmax of ant species. This new methodological framework can be used to detect the limita-
tions of particular ramping rates and help to identify more reliable CTmax values for trait-
based studies in functional ecology. In particular, studies using CTmax should carefully con-
sider the ramping rate used, as our results show that different ramping rates can lead to differ-
ent conclusions about relationships within and between ecological communities (Table 4) and
also bias the detection of phylogenetic signal (Table 3 and Fig 5). Our results support the use of
a ramping rate of 1˚Cmin-1 over relatively slower ramping rates. This ramping rate provides a
more reliable measure of ant species CTmax that aligns with assumptions about ant species’
thermal adaptions and observations of their foraging activities within natural environments
[61]. Although we only applied the integrated framework for the three most frequently used
ramping rates among studies of ant species’ CTmax, our study represents the most comprehen-
sive investigation of ant species’ CTmax thus far, and demonstrates the limitations associated
with the use of particular ramping rates and their consequences for conclusions relating to
PLOS ONE | https://doi.org/10.1371/journal.pone.0265361 March 14, 2022
19 / 23
PLOS ONE Ramping rate of Critical Thermal maximum
species’ habitat use or phylogenetic signal. We believe that this integrated framework should
also be applied for other terrestrial ectotherms. The framework we use to identify a suitable
ramping rate should produce reliable CTmax values that hold the potential to reveal crucial
information about species’ upper thermal limits. Such information will be useful for various
studies on climate warming, land-use change, pest control and trait-based ecology.
Supporting information
S1 Appendix.
(DOCX)
S2 Appendix.
(XLSX)
S3 Appendix.
(DOCX)
Acknowledgments
We are grateful to the HKU IBBL members, Runxi Wang (illustrations on Fig 5), Mark Wong,
Francois Brassard, Roy Cheung, Roger Lee, Kin Chan, Mac Pierce, Brian Worthington and
Maria Lo for their help and suggestions; Timothy Bonebrake, Lisa Bjerregaard Jørgensen,
Jelena Bujan, Bayden D. Russell, Tommy TY Hui for valuable suggestions. Lastly, we thank
two reviewers for the helpful comments.
Author Contributions
Conceptualization: Chi-Man Leong, Benoit Gue´nard.
Data curation: Chi-Man Leong.
Formal analysis: Chi-Man Leong, Toby P. N. Tsang.
Funding acquisition: Benoit Gue´nard.
Investigation: Chi-Man Leong, Toby P. N. Tsang, Benoit Gue´nard.
Methodology: Chi-Man Leong, Benoit Gue´nard.
Project administration: Benoit Gue´nard.
Resources: Benoit Gue´nard.
Supervision: Benoit Gue´nard.
Visualization: Chi-Man Leong, Toby P. N. Tsang.
Writing – original draft: Chi-Man Leong, Toby P. N. Tsang, Benoit Gue´nard.
Writing – review & editing: Chi-Man Leong, Toby P. N. Tsang, Benoit Gue´nard.
References
1. Deutsch CA, Tewksbury JJ, Huey RB, Sheldon KS, Ghalambor CK, Haak DC, et al. Impacts of climate
warming on terrestrial ectotherms across latitude. Proc Natl Acad Sci. 2008; 105(18):6668–72. https://
doi.org/10.1073/pnas.0709472105 PMID: 18458348
2. Hoffmann AA, Chown SL, Clusella-Trullas S. Upper thermal limits in terrestrial ectotherms: how con-
strained are they? Funct Ecol. 2013; 27(4):934–49. https://doi.org/doi:10.1111/j.1365-2435.2012.
02036.x.
PLOS ONE | https://doi.org/10.1371/journal.pone.0265361 March 14, 2022
20 / 23
PLOS ONE Ramping rate of Critical Thermal maximum
3. Kaspari M, Bujan J, Roeder KA, de Beurs K, Weiser MD. Species Energy and Thermal Performance
Theory predict 20-year changes in ant community abundance and richness. Ecology. 2019;0(ja).
https://doi.org/10.1002/ecy.2888.
4. Sinclair BJ, Marshall KE, Sewell MA, Levesque DL, Willett CS, Slotsbo S, et al. Can we predict ecto-
therm responses to climate change using thermal performance curves and body temperatures? Ecol
Lett. 2016; 19(11):1372–85. https://doi.org/10.1111/ele.12686 PMID: 27667778
5.
Lutterschmidt WI, Hutchison VH. The critical thermal maximum: history and critique. Can J Zool. 1997;
75(10):1561–74. https://doi.org/10.1139/z97-783.
6. Kellermann V, Overgaard J, Hoffmann AA, Flojgaard C, Svenning JC, Loeschcke V. Upper thermal lim-
its of Drosophila are linked to species distributions and strongly constrained phylogenetically. Proc Natl
Acad Sci. 2012; 109(40):16228–33. Epub 2012/09/19. https://doi.org/10.1073/pnas.1207553109
PMID: 22988106
7.
Terblanche JS, Deere JA, Clusella-Trullas S, Janion C, Chown SL. Critical thermal limits depend on
methodological context. Proceedings Biological sciences. 2007; 274(1628):2935–42. Epub 2007/09/
18. https://doi.org/10.1098/rspb.2007.0985 PMID: 17878142
8. Chown SL, Jumbam KR, Sørensen JG, Terblanche JS. Phenotypic variance, plasticity and heritability
estimates of critical thermal limits depend on methodological context. Funct Ecol. 2009; 23(1):133–40.
https://doi.org/10.1111/j.1365-2435.2008.01481.x.
9. Cowles RB, Bogert CM. A preliminary study of the thermal requirements of desert reptiles. Bull Am Mus
Nat Hist. 1944; 83(5):261–96.
10. Krebs CJ. Ecological methodology. New York: Harper Collins Publishers; 1989.
11.
Jørgensen LB, Malte H, Overgaard J. How to assess Drosophila heat tolerance: Unifying static and
dynamic tolerance assays to predict heat distribution limits. Funct Ecol. 2019; 33(4):629–42. https://doi.
org/doi:10.1111/1365-2435.13279.
12. Rezende EL, Bozinovic F, Szila´gyi A, Santos M. Predicting temperature mortality and selection in natu-
ral Drosophila populations. Science. 2020; 369(6508):1242–5. https://doi.org/10.1126/science.
aba9287 PMID: 32883867
13. Ribeiro PL, Camacho A, Navas CA. Considerations for assessing maximum critical temperatures in
small ectothermic animals: insights from leaf-cutting ants. PLoS ONE. 2012; 7(2):e32083. https://doi.
org/10.1371/journal.pone.0032083 PMID: 22384147
14. Kaspari M, Clay NA, Lucas J, Yanoviak SP, Kay A. Thermal adaptation generates a diversity of thermal
limits in a rainforest ant community. Global Change Biol. 2015; 21(3):1092–102. https://doi.org/doi:10.
1111/gcb.12750.
15. Schumacher A, Whitford WG. The foraging ecology of two species of Chihuahuan desert ants: Formica
perpilosa and Trachyrmyrmex smithi neomexicanus (Hymenoptera Formicidae). Insectes Soc. 1974;
21(3):317–30. https://doi.org/10.1007/BF02226923.
16. Bentley MT, Hahn DA, Oi FM. The thermal breadth of Nylanderia fulva (Hymenoptera: Formicidae) is
narrower than that of Solenopsis invicta at three thermal ramping rates: 1.0, 0.12, and 0.06˚C min-1.
Environ Entomol. 2016; 45(4):1058–62. https://doi.org/10.1093/ee/nvw050 PMID: 27252409
17. Roeder KA, Roeder DV, Bujan J. Ant Thermal Tolerance: A Review of Methods, Hypotheses, and
Sources of Variation. Ann Entomol Soc Am. 2021; 114(4):459–69. https://doi.org/10.1093/aesa/
saab018.
18. Bujan J, Yanoviak SP, Kaspari M. Desiccation resistance in tropical insects: causes and mechanisms
underlying variability in a Panama ant community. Ecol Evol. 2016; 6(17):6282–91. Epub 2016/09/21.
https://doi.org/10.1002/ece3.2355 PMID: 27648242
19. Diamond SE, Sorger DM, Hulcr J, Pelini SL, Toro ID, Hirsch C, et al. Who likes it hot? A global analysis
of the climatic, ecological, and evolutionary determinants of warming tolerance in ants. Global Change
Biol. 2012; 18(2):448–56. https://doi.org/doi:10.1111/j.1365-2486.2011.02542.x.
20. Bennett JM, Calosi P, Clusella-Trullas S, Martinez B, Sunday J, Algar AC, et al. GlobTherm, a global
database on thermal tolerances for aquatic and terrestrial organisms. Scientific Data. 2018; 5:180022.
Epub 2018/03/14. https://doi.org/10.1038/sdata.2018.22 PMID: 29533392
21. Stark AY, Adams BJ, Fredley JL, Yanoviak SP. Out on a limb: Thermal microenvironments in the tropi-
cal forest canopy and their relevance to ants. Journal Thermal Biology. 2017; 69:32–8. Epub 2017/10/
19. https://doi.org/10.1016/j.jtherbio.2017.06.002 PMID: 29037401
22. Baudier KM, Mudd AE, Erickson SC, O’Donnell S. Microhabitat and body size effects on heat tolerance:
implications for responses to climate change (army ants: Formicidae, Ecitoninae). J Anim Ecol. 2015;
84(5):1322–30. https://doi.org/10.1111/1365-2656.12388 PMID: 26072696
23. Kaspari M, Weiser M. The size–grain hypothesis: do macroarthropods see a fractal world? Ecol Ento-
mol. 2007; 32(3):279–82. https://doi.org/10.1111/j.1365-2311.2007.00870.x.
PLOS ONE | https://doi.org/10.1371/journal.pone.0265361 March 14, 2022
21 / 23
PLOS ONE Ramping rate of Critical Thermal maximum
24. Stevenson RD. Body size and limits to the daily range of body temperature in terrestrial ectotherms. Am
Nat. 1985; 125(1):102–17. https://doi.org/10.1086/284330.
25. Bergmann C. U¨ ber die Verha¨ltnisse der Wa¨rmeo¨ konomie der Thiere zu ihrer Gro¨sse. Gottinger Studien.
1848; 3:595–708.
26. Hood WG, Tschinkel WR. Desiccation resistance in arboreal and terrestrial ants. Physiol Entomol.
1990; 15(1):23–35. https://doi.org/doi:10.1111/j.1365-3032.1990.tb00489.x.
27. Nowakowski AJ, Watling JI, Thompson ME, Brusch IV GA, Catenazzi A, Whitfield SM, et al. Thermal
biology mediates responses of amphibians and reptiles to habitat modification. Ecol Lett. 2018; 21
(3):345–55. https://doi.org/10.1111/ele.12901 PMID: 29314479
28. Bishop TR, Robertson MP, Rensburg BJ, Parr CL. Coping with the cold: minimum temperatures and
thermal tolerances dominate the ecology of mountain ants. Ecol Entomol. 2017; 42(2):105–14. https://
doi.org/doi:10.1111/een.12364.
29.
Janzen DH. Why mountain passes are higher in the tropics. Am Nat. 1967; 101(919):233–49.
30. Garcı´a-Robledo C, Kuprewicz EK, Staines CL, Erwin TL, Kress WJ. Limited tolerance by insects to high
temperatures across tropical elevational gradients and the implications of global warming for extinction.
Proc Natl Acad Sci U S A. 2016; 113(3):680–5. Epub 2016/01/04. https://doi.org/10.1073/pnas.
1507681113 PMID: 26729867
31. Bujan JF, Kaspari M. Nutrition modifies critical thermal maximum of a dominant canopy ant. J Insect
Physiol. 2017; 102:1–6. https://doi.org/10.1016/j.jinsphys.2017.08.007 PMID: 28830761
32. Allen JL, Chown SL, Janion-Scheepers C, Clusella-Trullas S. Interactions between rates of temperature
change and acclimation affect latitudinal patterns of warming tolerance. Conserv Physiol. 2016; 4(1):
cow053. https://doi.org/10.1093/conphys/cow053 PMID: 27933165
33. Woods HA. Ontogenetic changes in the body temperature of an insect herbivore. 2013; 27(6):1322–31.
https://doi.org/https://doi.org/10.1111/1365-2435.12124.
34. Bujan JF, Roeder KA, de Beurs K, Weiser MD, Kaspari M. Thermal diversity of North American ant
communities: Cold tolerance but not heat tolerance tracks ecosystem temperature. Global Ecol Bio-
geogr. 2020; 00:1–9. https://doi.org/10.1111/geb.13121.
35. Diamond SE, Chick LD, Perez A, Strickler SA, Zhao C. Evolution of plasticity in the city: urban acorn
ants can better tolerate more rapid increases in environmental temperature. Conserv Physiol. 2018; 6
(1):coy030-coy. https://doi.org/10.1093/conphys/coy030 PMID: 29977563
36.
Lutterschmidt WI, Hutchison VH. The critical thermal maximum: data to support the onset of spasms as
the definitive end point. Can J Zool. 1997; 75(10):1553–60. https://doi.org/10.1139/z97-782.
37. Wehner R, Wehner S. Parallel evolution of thermophilia: daily and seasonal foraging patterns of heat-
adapted desert ants: Cataglyphis and Ocymyrmex species*. Physiol Entomol. 2011; 36(3):271–81.
https://doi.org/doi:10.1111/j.1365-3032.2011.00795.x.
38. Shi NN, Tsai C-C, Camino F, Bernard GD, Yu N, Wehner R. Keeping cool: Enhanced optical reflection
and radiative heat dissipation in Saharan silver ants. Science. 2015; 349(6245):298–301. https://doi.
org/10.1126/science.aab3564 PMID: 26089358
39. Cerda´ X, Retana J, Manzaneda A. The role of competition by dominants and temperature in the forag-
ing of subordinate species in Mediterranean ant communities. Oecologia. 1998; 117(3):404–12. https://
doi.org/10.1007/s004420050674 PMID: 28307920
40. Cerda´ X, Retana J, Cros S. Critical thermal limits in Mediterranean ant species: trade-off between mor-
tality risk and foraging performance. Funct Ecol. 1998; 12(1):45–55. https://doi.org/doi:10.1046/j.1365-
2435.1998.00160.x.
41. Economo EP, Narula N, Friedman NR, Weiser MD, Gue´nard B. Macroecology and macroevolution of
the latitudinal diversity gradient in ants. Nature Communications. 2018; 9(1):1778. https://doi.org/10.
1038/s41467-018-04218-4 PMID: 29725049
42. R Development Core Team. R: A language and environment for statistical computing. Vienna, Austria:
R Foundation for Statistical Computing; 2020.
43.
Lenth R, Singmann H, Love J. Emmeans: Estimated marginal means, aka least-squares means. R
package version. 2018; 1(1).
44.
Fox J, Weisberg S. An R companion to applied regression: Sage publications; 2018.
45. Diamond SE, Chick LD, Perez A, Strickler SA, Martin RA. Evolution of thermal tolerance and its fitness
consequences: parallel and non-parallel responses to urban heat islands across three cities. Proc Biol
Sci. 2018; 285(1882):20180036. Epub 2018/07/28. https://doi.org/10.1098/rspb.2018.0036 PMID:
30051828
PLOS ONE | https://doi.org/10.1371/journal.pone.0265361 March 14, 2022
22 / 23
PLOS ONE Ramping rate of Critical Thermal maximum
46.
Terblanche JS, Hoffmann AA, Mitchell KA, Rako L, le Roux PC, Chown SL. Ecologically relevant mea-
sures of tolerance to potentially lethal temperatures. J Exp Biol. 2011; 214(22):3713–25. https://doi.org/
10.1242/jeb.061283 PMID: 22031735
47. Ascunce MS, Yang CC, Oakey J, Calcaterra L, Wu WJ, Shih CJ, et al. Global invasion history of the fire
ant Solenopsis invicta. Science. 2011; 331(6020):1066–8. Epub 2011/02/26. https://doi.org/10.1126/
science.1198734 PMID: 21350177
48. Roeder KA, Roeder DV, Kaspari M. The role of temperature in competition and persistence of an
invaded ant assemblage. Ecol Entomol. 2018; 43(6):774–81. https://doi.org/doi:10.1111/een.12663.
49. Wendt CF, Verble-Pearson R. Critical thermal maxima and body size positively correlate in red imported
fire ants, Solenopsis invicta. Southwest Nat. 2016; 61(1):79–83. https://doi.org/10.1894/0038-4909-61.
1.79.
50. Vogt JT, Smith WA, Grantham RA, Wright RE. Effects of temperature and season on foraging activity of
red imported fire ants (Hymenoptera: Formicidae) in Oklahoma. Environ Entomol. 2003; 32(3):447–51.
https://doi.org/10.1603/0046-225X-32.3.447.
51. Drees BBM, Summerlin B, Vinson SB. Foraging activity and temperature relationship for the red
imported fire ant. Southwest Entomol. 2007; 32(3):149–55. https://doi.org/10.3958/0147-1724-32.3.
149.
52. Bruno JF, Stachowicz JJ, Bertness MD. Inclusion of facilitation into ecological theory. Trends Ecol Evol.
2003; 18(3):119–25.
53. Alam MR, Zain MFM, Kaish ABMA, Jamil M. Underground soil and thermal conductivity materials
based heat reduction for energy-efficient building in tropical environment. Indoor Built Environ. 2015; 24
(2):185–200. https://doi.org/10.1177/1420326X13507591.
54. Kaspari M. In a globally warming world, insects act locally to manipulate their own microclimate. Proc
Natl Acad Sci. 2019:201901972. https://doi.org/10.1073/pnas.1901972116.
55. Diamond SE, Chick LD. The Janus of macrophysiology: stronger effects of evolutionary history, but
weaker effects of climate on upper thermal limits are reversed for lower thermal limits in ants. Curr Zool.
2018; 64(2):223–30. https://doi.org/10.1093/cz/zox072 PMID: 30402063
56. Kellermann V, Loeschcke V, Hoffmann AA, Kristensen TN, Fløjgaard C, David JR, et al. Phylogenetic
constraints in key functional traits behind species’ climate niches: patterns of desiccation and cold resis-
tance across 95 Drosophila species. Evolution. 2012; 66(11):3377–89. https://doi.org/10.1111/j.1558-
5646.2012.01685.x PMID: 23106704
57. Grigg JW, Buckley LB. Conservatism of lizard thermal tolerances and body temperatures across evolu-
tionary history and geography. Biol Lett. 2013; 9(2):20121056–. https://doi.org/10.1098/rsbl.2012.1056
PMID: 23325735
58. Roeder KA, Bujan J, de Beurs KM, Weiser MD, Kaspari M. Thermal traits predict the winners and losers
under climate change: an example from North American ant communities. Ecosphere. 2021; 12(7):
e03645. https://doi.org/https://doi.org/10.1002/ecs2.3645.
59. Kellermann V, van Heerwaarden B. Terrestrial insects and climate change: adaptive responses in key
traits. Physiol Entomol. 2019; 44(2):99–115. https://doi.org/10.1111/phen.12282.
60. Stillman JH. Acclimation capacity underlies susceptibility to climate change. Science. 2003; 301
(5629):65–. https://doi.org/10.1126/science.1083073 PMID: 12843385
61. Angilletta MJ Jr. Thermal adaptation: a theoretical and empirical synthesis: Oxford University Press;
2009. https://doi.org/10.1890/08-2224.1 PMID: 19886501
PLOS ONE | https://doi.org/10.1371/journal.pone.0265361 March 14, 2022
23 / 23
PLOS ONE |
10.1371_journal.pone.0259135 | RESEARCH ARTICLE
Magnetic resonance colonography
assessment of acute trinitrobenzene sulfonic
acid colitis in pre-pubertal rats
Claire Dupont-LucasID
Christine Boˆ le-Feysot1, Emmeline Salameh1, Alexis Goichon1, Lionel Nicol2,
Moutaz Aziz3, Ce´ line Savoye-Collet4, Guillaume Savoye5
1¤*, Rachel Marion-Letellier1, Mathilde Pala1, Charlène Guerin1,
1 INSERM UMR 1073, Institute for Biomedical Research, Rouen University, Rouen, France, 2 INSERM
UMR 1096, Institute for Biomedical Research, Rouen University, Rouen, France, 3 Department of Pathology,
Rouen University Hospital, Rouen, France, 4 QUANTIF-LITIS EA 4108, Rouen University, Rouen, France,
5 Department of Gastroenterology, Rouen University Hospital, Rouen, France
¤ Current address: Department of Pediatrics, Caen University Hospital, Caen, France
* dupont-c@chu-caen.fr
Abstract
Pre-pubertal murine models of acute colitis are lacking. Magnetic resonance colonography
(MRC) is a promising minimally invasive tool to assess colitis. We aimed to: 1/ Adapt a model
of acute experimental colitis to pre-pubertal rats and determine whether MRC characteristics
correlate with histological inflammation. 2/ Test this model by administering a diet supple-
mented in transforming growth factor β2 to reverse inflammation. Twenty-four rats were ran-
domized at weaning to one of 3 groups: Trinitrobenzene Sulfonic Acid (TNBS) group (n = 8) fed
a standard diet, that received an intra-rectal 60 mg/kg dose of TNBS-ethanol; Control group (n
= 8) fed standard diet, that received a dose of intra-rectal PBS; TNBS+MODULEN group (n =
8) that received a dose of TNBS and were exclusively fed MODULEN-IBD® after induction of
colitis. One week after induction of colitis, rats were assessed by MRC, colon histopathology
and inflammation markers (Interleukin 1β, Tumor necrosis factor α, Nitric Oxide Synthase 2
and Cyclooxygenase 2). TNBS induced typical features of acute colitis on histopathology and
MRC (increased colon wall thickness, increased colon intensity on T2-weighted images, target
sign, ulcers). Treatment with MODULEN-IBD® did not reduce signs of colitis on MRC. Inflam-
matory marker expression did not differ among study groups.
Introduction
Inflammatory bowel disease (IBD), including Crohn’s disease and Ulcerative colitis, is a
chronic relapsing disease affecting the digestive tract. The incidence and prevalence of these
diseases are increasing worldwide [1]. A systematic worldwide review showed that the highest
reported prevalence values for IBD were in Europe (UC: 505 per 100,000 persons, CD: 322 per
100,000 persons) [1]. An estimated ten percent of new IBD cases patients are children, with a
steady increase of incidence of pediatric IBD worldwide [2, 3]. Crohn’s disease beginning in
childhood has several specificities compared to adult-onset disease among which a higher inci-
dence of complicated phenotypes and of growth failure [4–6]. Focusing on the mechanisms of
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: Dupont-Lucas C, Marion-Letellier R, Pala
M, Guerin C, Boˆle-Feysot C, Salameh E, et al.
(2021) Magnetic resonance colonography
assessment of acute trinitrobenzene sulfonic acid
colitis in pre-pubertal rats. PLoS ONE 16(11):
e0259135. https://doi.org/10.1371/journal.
pone.0259135
Editor: Rosanna Di Paola, University of Messina,
ITALY
Received: July 11, 2021
Accepted: October 14, 2021
Published: November 4, 2021
Peer Review History: PLOS recognizes the
benefits of transparency in the peer review
process; therefore, we enable the publication of
all of the content of peer review and author
responses alongside final, published articles. The
editorial history of this article is available here:
https://doi.org/10.1371/journal.pone.0259135
Copyright: © 2021 Dupont-Lucas et al. This is an
open access article distributed under the terms of
the Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: The dataset is
available as supplementary material: https://
figshare.com/s/01b2dd941d08b0fa6d9e).
PLOS ONE | https://doi.org/10.1371/journal.pone.0259135 November 4, 2021
1 / 13
PLOS ONE Funding: This work was supported by the
European Union and Normandy Regional Council.
Europe contributes to Normandy through the
European Regional Development Fund (ERDF).
One of the authors (MP) received a grant from
Nestle´ Health Science given by SFNCM (Socie´te´
Francophone Nutrition Clinique et Me´tabolisme).
The funders had no role in study design, data
collection and analysis, decision to publish, or
preparation of the manuscript.
Competing interests: No authors have competing
interests.
TNBS acute colitis in prepubertal rats
the similarities and differences between pediatric and adult IBD could help unravel some
aspects of IBD pathogenesis.
Although the cause of IBD remains unknown, studies provide evidence that pathogenesis
of disease involves a loss of immune tolerance to the gut microbiota in a genetically susceptible
host, exposed to environmental factors [7]. Animal models of intestinal inflammation mimick-
ing Inflammatory Bowel Disease (IBD) are widely used to better understand the cellular and
molecular pathways of inflammation and fibrosis, and target these pathways to develop new
drugs [8–11]. Among these, the 2,4,6 Trinitrobenzene Sulfonic Acid Colitis (TNBS) model is
frequently used and many variations exist [12]. Intra-rectal instillation of TNBS causes an IL-
12-driven Th1 T-cell mediated immune response in the colonic mucosa [13]. Briefly, TNBS is
instilled intra rectally mixed with ethanol, which acts as a mucosal barrier breaker allowing the
hapten TNBS to interact with colonic proteins and elicit an immunological response by ren-
dering these proteins immunogenic to the host immune system. The stimulated antigen
producing cells secrete IL-12, causing induction of IFN-γ by the T cells, which stimulates mac-
rophages to produce inflammatory mediators such as TNF-α, IL-6 and IL-1β. The resulting
colonic inflammation resembles Crohn’s disease, with transmural inflammation, ulcers and
granulomas. The clinical picture of TNBS colitis is also similar to Crohn’s disease, associating
weight loss and bloody diarrhea [13]. The model has been used to test anti-inflammatory prop-
erties of dietary compounds, and drugs on colonic inflammation through inhibition of the
NF-κB pathway [11, 14].
Most studies on acute TNBS colitis in rats have been carried out on adult animals. However
weanling and pre-pubertal rats have several distinct features that would require adapting the
TNBS model, such as rapid growth and susceptibility to growth failure, increased mucosal per-
meability and increased susceptibility to drugs [15, 16].
In order to follow non-invasively the effect of interventions and reduce number of animals
needed for longitudinal studies, small animal imaging techniques have been developed. In a
model of acute TNBS colitis in adult rats we previously showed that magnetic resonance colo-
nography (MRC) could accurately evaluate inflammation, compared to histopathology [17].
The recommended treatment for inducing remission in children with IBD is exclusive
enteral nutrition [18]. Several enteral diets have been compared, among which a polymeric
diet enriched in transforming growth factor beta 2 (TGF-beta 2): MODULEN-IBD1 (Nestle,
Vevey, Switzerland). Administered as sole food source for 8 weeks, MODULEN-IBD1 can
induce clinical remission in 79% children and mucosal healing in 31% [19].
The objectives of our study were: 1/ to adapt a model of acute experimental colitis to pre-
pubertal rats and determine whether MRC characteristics could be correlated to histopathol-
ogy. 2/ To test this model by treating the rats with MODULEN-IBD1 in the aim of reversing
inflammation.
Materials and methods
Study design
Twenty-four Sprague Dawley male rats were purchased at weaning (postnatal day 21) from
Janvier labs (Le Genest Saint Isle, France). Mean baseline weight was 77.4 ± 3.9 g. They were
randomly allocated to one of three study groups: Control (n = 8), TNBS (n = 8) and TNBS
+ MODULEN (n = 8) (Fig 1, Panel A). The rats were housed 4 per standard cage to provide
for their interaction needs, were exposed to light / dark cycles of 12 hours each and provided
with water ad libitum. After induction of colitis, the TNBS + MODULEN group received
MODULEN-IBD1 powder as sole food source. The other groups received a standard rat
breeding diet (A03, SAFE) in powdered form.
PLOS ONE | https://doi.org/10.1371/journal.pone.0259135 November 4, 2021
2 / 13
PLOS ONE TNBS acute colitis in prepubertal rats
Fig 1. Experimental design for induction of acute colitis in pre-pubertal rats, and treatment with
MODULEN-IBD1. A. Experimental design. B. Body weight by group, in the days following randomization (��
P<0.01 compared to the 2 other groups). C. Tail length at the end of the study, by group (The groups did not differ
statistically). CTL: controls, TNBS: 2,4,6-trinitrobenzene sulfonic acid colitis group, MOD: TNBS + MODULEN
group.
https://doi.org/10.1371/journal.pone.0259135.g001
PLOS ONE | https://doi.org/10.1371/journal.pone.0259135 November 4, 2021
3 / 13
PLOS ONE TNBS acute colitis in prepubertal rats
All sections of this report adhere to the ARRIVE Guidelines for reporting animal research
[20]. Sample size was calculated using G�Power Software [21]. Based on our previous study
[17], a difference in colon wall thickness on MRC between control and TNBS group was
shown with an effect size of 3.7, and mean mortality rate was 30%. We estimated that effect
size would be 50% lower (= 1.9) using a dose of TNBS reduced by 50% to take into account
higher susceptibility to drugs in young animals. To achieve power = 0.80 and alpha = 0.05 to
detect this difference would require 6 animals per group, which we increased to 8 based on a
predicted attrition rate of 30%.
Induction of colitis
After one week of acclimation, colitis was induced by intra-rectal instillation of 2,4,6-trinitro-
benzene sulfonic acid (TNBS, Sigma-Aldrich, Saint-Quentin Fallavier, France). After fasting
for 24 hours, rats were anesthetized by intra-peritoneal injection of ketamine 8 mg/kg (Pan-
pharma) and Chlorpromazine 1 mg/kg (Sanofi-Aventis). A polyurethane cannula was inserted
3 cm into the rectum and TNBS—ethanol injected, at a dose of 60 mg/kg of body weight.
TNBS was mixed with a 50% Ethanol vehicle, for a total volume of 250 μL. The rats were main-
tained in a Trendelenburg position for 5 minutes after the injection to minimize leaks. Rats in
the control group received an intra-rectal instillation of 250 μL of phosphate-buffered saline
(PBS). During the week following induction of colitis, the rats were observed for signs of pain
or significant weight loss.
Magnetic resonance colonography
MRC was performed within the week following TNBS or PBS administration (between Day 4
and Day 7), using a small animal machine: Bruker BioSpec 47 / 40 USR, 4.7 Tesla (Bruker Bios-
pin, Ettlingen, Germany). Rats were anesthetized by intra-peritoneal injection of thiopental 90
mg/kg (Panpharma). Cardiac rhythm was monitored by surface electrodes. Rats were installed
in the cradle in a supine position. There was no injection of antispasmodic or contrast agent.
Respiratory movements were corrected using the Intragate™ technique.
Parameters for the T2 rapid acquisition with relaxation enhancement (RARE) sequence were:
Repetition time (RT) 5443 ms, Echo time (ET) 34 ms, matrix 320x224, slice 1 mm, Number of
excitations (NEX) 3, flip angle 180˚, field of view (FOV) 5.0 x 5.3 cm, acquisition time: 10 minutes,
RARE factor 8. Parameters for T2 RARE with fat suppression (FAT-SAT): ET 39 ms, RT 6027
ms, RARE factor 8, NEX 3, acquisition time 8 minutes, flip angle 180˚, slice 1 mm, FOV 5.0 x 5.3
cm, matrix 320x224. T1 sequence with intragate fast low angle shot (FLASH); RT 413 ms, ET 2.8
ms, matrix 256x256, slice 1.1 mm, flip angle 80˚, FOV 4.5 x 4.5 cm; acquisition time: 14 minutes.
MRC images were analysed in DICOM, using ParaVision 5.0 software. A senior radiologist (C.
S.-C.) who was blind to allocation group interpreted all images.
Image quality taking into account respiratory and bowel wall movements was assessed on a
scale of 0 (poor quality) to 3 (excellent quality). Measures were made in the descending colon.
In order to assess inflammation, the criteria used were: maximal colon wall thickness (aver-
age of 3 measures), minimal colon wall thickness (average of 3 measures), colon wall thickness
at splenic angle (average of 3 measures), colon wall signal intensity in Regions of Interest
(ROI) on T2w sequences (average of 2 measures), target sign, colon mucosa irregularities
suggesting ulcerations, spontaneous enhancement of colon wall T1w signal [17].
Killing and samples
On Day 8 after colitis induction, rats were killed by a lethal dose of intraperitoneal thiopental
and then decapitated. Tail was measured as a proxy for growth. Colon was sampled, washed
PLOS ONE | https://doi.org/10.1371/journal.pone.0259135 November 4, 2021
4 / 13
PLOS ONE TNBS acute colitis in prepubertal rats
with PBS to remove feces, measured and weighed. Six one-centimeter samples were taken
from the colon, starting from the rectum, one of which was fixed in 10% neutral buffered for-
malin (Sigma-Aldrich) for histopathology, the others stored at– 80˚C until analysis.
Histopathology
Histological analyses were made by a senior pathologist (M.A.), blinded to allocation group.
The formalin-fixed samples were embedded in paraffin, and 5-micrometer sections were col-
ored with hematoxylin/eosin (H&E; Merck, Darmstadt, Germany) for standard histopathol-
ogy. Samples were studied on 3 levels of cut. Inflammation was scored using a semi-
quantitative score previously used by our team: from 0 (no inflammation) to 3 (severe inflam-
mation) [17, 22]. Fibrosis was scored from 0 (no fibrosis) to 3 (severe fibrosis). Images were
taken by standard light microscopy using a Leica microscope.
Colon expression of cyclooxygenase-2 (COX-2) by Western Blot
Frozen colon samples were homogenized in PBS with 0.1% protease inhibitor cocktail (Sigma)
and phosphatase inhibitor cocktail (Sigma). Homogenates were centrifuged (12 000g, 15 min,
4˚C) and supernatants were collected and stored at -80˚C. Protein concentration was deter-
mined following Bradford’s colorimetric method and Western blot for COX-2 was performed
as previously [23]. Protein expression was quantified by densitometry with the ImagQuantTL
software (GE Healthcare, USA). To check equal loading, the blots were analyzed for glyceral-
dehyde-3-phosphate dehydrogenase (GAPDH) expression.
mRNA levels for Interleukin-1 β (IL-1 β), Tumor necrosis factor α (TNF-α)
cytokines and inflammatory marker Nitric Oxide Synthase 2 (NOS2)
Quantitative reverse transcription PCR (RT-qPCR) was performed as previously [24]. Briefly,
colon samples were frozen in liquid nitrogen and stored at -80˚C before ribonucleic acid
(RNA) preparation. Total RNA was isolated using guanidium isothiocyanate method and
reverse transcribed into complementary deoxyribonucleic acid (cDNA). PCR was performed
with CFX96 Real-Time System (Bio-Rad, Marnes-la-Coquette, France).
Data analyses
Characteristics of the rats were compared between groups. Since group sizes were < 30, non-
parametric methods were used for analyses: qualitative variables were compared using Fisher’s
exact test, quantitative variables were compared using Wilcoxon Mann-Whitney with exact
correction for small samples. Missing data (following premature death of an animal) was
excluded from analyses.
Statistical analyses were performed using SAS 9.2 (Cary, NC, USA) and Graph Pad Prism 5
(San Diego, CA, USA). A two-sided p-value < 0.05 was considered significant.
Compliance with ethical standards
Animal care and experimentation complied with French and European Community regula-
tions (Directive 2010/63/UE). Study protocol was approved by the Institutional Care and Use
Committee (Comite´ d’Ethique Normande en Matière d’Expe´rimentation Animale, CENO-
MEXA). Painful procedures were carried out under deep sedation and all efforts were made to
minimize suffering.
PLOS ONE | https://doi.org/10.1371/journal.pone.0259135 November 4, 2021
5 / 13
PLOS ONE TNBS acute colitis in prepubertal rats
Results
Validation of the acute colitis model
Growth and histopathology. Body weight growth curve during study protocol and tail
length at the end of the protocol did not differ between TNBS and Control groups (Fig 1,
Panel B, C). Mortality rate was 25% (2/8) in the control group, and 38% (3/8) in the TNBS
group.
Histological inflammation score was significantly higher in the TNBS group compared to
control group (p = 0.015) (Fig 2, Panel D). TNBS group had a higher fibrosis score but did not
reach statistical significance (p = 0.06) (Fig 2, Panel E). Colon weight/length, a marker of
inflammation, did not differ between TNBS and Control groups (p = 0.08) (Fig 2, Panel F),
although colon weight was significantly increased in TNBS group compared to Controls
(2.3 ± 1.1 g vs. 1.4 ± 0.3 g, p = 0.009).
Magnetic resonance colonography. Bowel wall thickness measured in descending colon
on axial plane, T2 sequence, was significantly increased in the TNBS group compared to con-
trols (Table 1 and Fig 3, Panel A). MRC signs suggestive of inflammation, such as a target sign,
wall ulcers, and increased wall signal intensity on T2w images were significantly more preva-
lent in the TNBS group than in controls (Table 1). There was no significant increase of T1w
Fig 2. Validation of the acute colitis model: Histopathology, hematoxylin-eosin stain. A. H&E histology from the control group showing normal
histology. B. Representative histology from the TNBS group. The arrow indicates severe transmural inflammation. C. Representative histopathology from the
TNBS + MODULEN. The arrow indicates an ulceration. D. Histology inflammation score by group. E. Histological fibrosis score by group. F. Colon weight/
length ratio by group. CTL: controls, TNBS: 2,4,6-trinitrobenzene sulfonic acid colitis group, MOD: TNBS + MODULEN group. � P < 0.05.
https://doi.org/10.1371/journal.pone.0259135.g002
PLOS ONE | https://doi.org/10.1371/journal.pone.0259135 November 4, 2021
6 / 13
PLOS ONE TNBS acute colitis in prepubertal rats
Table 1. Magnetic resonance colonography characteristics—TNBS induced acute colitis model.
CTL (n = 8)
TNBS (n = 5) MOD (n = 6)
TNBS/ CTL (p) MOD/ TNBS (p) MOD/ CTL (p)
Maximum colon wall thickness (axial T2) (mm)
Minimal wall thickness (mm)
Wall thickness at kidney hilum level (mm)
Increased wall signal intensity on T2w (n, %)
0.36 ±0.04
0.36±0.05
0.42 ±0.06
0 (0)
Target sign—yes (n, %)
Spontaneous T1w hypersignal intensity (n, %)
Irregular patterns of mucosal wall (ulcers) (n, %)
Stenosis (n, %)
Mucosal flap (n, %)
https://doi.org/10.1371/journal.pone.0259135.t001
0 (0)
0 (0)
0 (0)
0 (0)
0 (0)
1.03 ±0.36
0.63 ±0.14
0.72 ±0.14
4 (0.80)
4 (0.80)
2 (0.40)
5 (1.0)
3 (0.60)
0 (0)
1.26 ±0.48
0.69 ±0.19
0.53 ±0.22
5 (0.83)
3 (0.50)
1 (0.17)
6 (1.0)
6 (1.0)
2 (0.33)
0.002
0.002
0.002
0.007
0.007
NS
<0.001
0.04
-
NS
NS
NS
NS
NS
NS
-
NS
NS
<0.001
<0.001
0.06
0.003
0.055
NS
<0.001
<0.001
NS
wall signal intensity, luminal stenosis or mucosal flap. Image quality was good, except for 1 rat
in the TNBS group that had wall artifacts.
Inflammatory markers and cytokine expression. The mRNA levels of 3 inflammatory
markers genes (IL1β, TNFα and NOS2) were not significantly different between TNBS and
control groups (Fig 4). Colon expression of COX-2 was not significantly different between
groups (Fig 4).
Effect of MODULEN-IBD1 in an acute TNBS colitis model
Growth. Final body weight in the TNBS+MODULEN group was significantly lower than
TNBS group (154.6 ± 12.9 g vs. 183.8 ± 14.6, p = 0.004) and control group (mean weight
193.6 ± 11.9 g, p = 0.0007) (Fig 1, Panel B). Tail length did not differ significantly between
groups (Fig 1, Panel C).
Histopathology. Histological inflammation and fibrosis scores did not differ significantly
between TNBS+MODULEN and TNBS group (Fig 2, Panel D). Colon weight/length ratio, a
marker of inflammation, was significantly lower in the TNBS+MODULEN group than in the
TNBS group (p = 0.03) (Fig 2, Panel F).
Magnetic resonance colonography. There was no significant difference in MRC charac-
teristics between TNBS and TNBS+MODULEN groups (Table 1). Rats from the TNBS+-
MODULEN group differed from controls on the following characteristics: increased wall
thickness, increased wall intensity on T2w images, presence of ulcers (100% of rats) and lumi-
nal stenosis (100% of rats). A mucosal flap was observed in 33% of rats from the TNBS+-
MODULEN group, but not in other groups. Image quality was good, except for 1 rat that had
wall artifacts and 1 that had respiratory artifacts.
Inflammatory marker and cytokine expression. The colon mRNA expression of IL-1β,
TNF-α and NOS2 was not significantly different between MODULEN and TNBS or control
groups (Fig 4). Colon expression of COX-2 was not significantly different between groups.
Discussion
In this study we have shown that a single intra-rectal instillation of TNBS was able to induce
acute colitis in pre-pubertal rats, and that typical features of acute colitis could be observed on
MRC and confirmed by histopathology.
The TNBS-hapten model of colitis first described in 1989 by Morris et al. [12] has been
shown to mimic IBD by eliciting a Th1 T-cell based response. Although this model has been
frequently used for studies in adult rats, few protocols exist in pre-pubertal rats. Hence our
PLOS ONE | https://doi.org/10.1371/journal.pone.0259135 November 4, 2021
7 / 13
PLOS ONE TNBS acute colitis in prepubertal rats
Fig 3. Validation of acute colitis model: Magnetic resonance colonography. A. Colon wall thickening: maximum thickness, by group (�� P < 0.01). B.
MRC representative image from the TNBS group. The arrow indicates an irregular colon wall suggestive of an ulcer. C. MRC representative image from
the TNBS group: the arrow indicates increased wall signal intensity on T2-weighted sequence. D. MRC representative image from the TNBS group: the
arrow indicates narrowing of the colon lumen suggestive of a stenosis and the star, a pre-stenotic dilation. MRC: magnetic resonance colonography. CTL:
controls, TNBS: 2,4,6-trinitrobenzene sulfonic acid colitis group, MOD: TNBS + MODULEN group.
https://doi.org/10.1371/journal.pone.0259135.g003
first objective was to adapt this model of acute experimental colitis to pre-pubertal rats. There
is no standardized protocol for inducing TNBS colitis in rats, and various doses of TNBS have
been used, putting into balance the severity of colitis induced with the mortality rate [25]. Low
doses of 50 mg/kg cause mild colitis (Wallace score of 2), whereas high doses of up to 150 mg/
kg cause severe colitis (minimal Wallace score of 5) but are associated with mortality rates of
30% caused by excessive inflammation. The only pre-pubertal rat models reported to date
used a fixed dose of 8 mg of TNBS (between 80 and 133 mg/kg depending on animal’s weight)
[26, 27]. In our experimental protocol we chose an individually weight-adapted dose of 60 mg/
kg, to take into account the increased susceptibility caused by young age [15]. This induced
mild to severe histological colitis in 4/5 rats of the TNBS group. Despite the low dose chosen,
our team’s expertise using this model with older animals and precautious manipulations, there
was significant mortality (between 2 and 3 rats per group), which underlines the challenge of
developing a reproducible model of acute bowel inflammation in pre-pubertal rats with the
least invasive procedures possible.
PLOS ONE | https://doi.org/10.1371/journal.pone.0259135 November 4, 2021
8 / 13
PLOS ONE TNBS acute colitis in prepubertal rats
Fig 4. Expression of inflammatory markers in colonic tissue homogenates, TNBS-induced acute colitis model in pre-pubertal rats. Gene expression
measured by RT-PCR: IL-1β (Panel A), NOS2 (Panel B) and TNF-α (Panel C); Expression of COX-2 measured by Western Blot (Panel D). None of the
groups were statistically different. CTL: controls, TNBS: 2,4,6-trinitrobenzene sulfonic acid colitis group, MOD: TNBS + MODULEN group.
https://doi.org/10.1371/journal.pone.0259135.g004
Magnetic resonance imaging is an important technique to follow progression of inflamma-
tion and tissue damage in IBD. Our team showed the feasibility of MRC to assess acute TNBS
colitis in adult rats [17]. Consistently with our previous study, we obtained good quality
images, without requiring anti spasmodic or contrast agents, and observed typical signs of
colon inflammation in TNBS treated rats, but not in controls: ulcers, target sign and bowel
wall thickening.
We were not able to replicate our previous findings of increased colon production of IL-1β
and COX-2 following TNBS administration. A possible explanation for this discrepancy was
that we chose to wait one week after induction of colitis instead of 2 days, hoping to observe
the maximum severity of lesions. However, this delay might have allowed the pro-inflamma-
tory markers to normalize, contemporary to recovery process, although histological lesions
might take longer to.
We were surprised by a signal of fibrosis observed on histopathology, since rats in the TNBS
group had a higher although not significant histological fibrosis score than controls. Intestinal
fibrosis is a dynamic process, which is thought to arise as a consequence of chronic uncontrolled
inflammation. Fibrosis in TNBS models has been observed after 4 to 8 doses of TNBS [27, 28].
PLOS ONE | https://doi.org/10.1371/journal.pone.0259135 November 4, 2021
9 / 13
PLOS ONE TNBS acute colitis in prepubertal rats
Presence of fibrosis as soon as 1 week after the first dose of TNBS would suggest that tissue repair
pathways might be activated very early in inflammation or might be an independent phenome-
non. Brenna et al. showed temporal changes in gene expression in acute TNBS colitis between 0
and 12 days after induction of colitis, with a decrease in regulation of metabolism and an increase
of the regulation of tissue remodeling genes [29]. They conclude that mucosal cells might be
exposed to pro-fibrotic signaling cascades during inflammatory response.
The anti-inflammatory effect of MODULEN-IBD1 has been shown in children based on
clinical, biological and endoscopic endpoints [30–33]. In our pre-pubertal rat model, we
showed that the histological inflammation scores of TNBS rats treated with MODULEN-IBD1
did not differ from controls or from TNBS group. In addition, the reduced colon weight/
length ratio in the TNBS+MODULEN group is in favor of an anti-inflammatory effect. It must
be noted that reduced weight/length ratio cannot be attributed to growth failure, since colon
and tail lengths were not different between groups. We did not observe different final weights
between TNBS and control groups, based on the body weight at the end of the protocol, 8 days
after induction of colitis. Weight loss is one of the clinical features of the TNBS colitis model.
However, weight loss reaches a peak 3 days after induction of colitis but can be recovered by
Day 7 [29], which could explain the absence of weight difference between TNBS and control
groups. However, final weight was lower in the TNBS + MODULEN group than in the two
others, indicating that the intake of MODULEN-IBD1 might have been insufficient to meet
the energy requirements. The macronutrient composition of MODULEN-IBD1 has several
differences with a standard growing rodent diet. Energy content of MODULEN-IBD1 is 5000
kcal / kg of powder, of which 42% is brought by lipids, 44% by carbohydrates and 14% by pro-
tein. In contrast, the standard A03 “rodent diet for growing and breeding” (SAFE) nutritional
composition is: 13.5% lipids, 61.3% carbohydrates, 25.2% protein with a total of 3395 kcal / kg
of powder form. Given as a sole source of nutrition, and ad libitum instead of by oral gavage,
the effect might also have been hampered by low palatability. In order to control for feeding
difficulties due to the powdered form of the feed, we fed the control and TNBS groups the stan-
dard rodent diet in a powdered form and not as pellets.
We did not observe a difference in levels of mRNA for inflammatory cytokines (IL-1β and
TNFα) or NOS 2 and colon COX-2 expression between the 3 groups, although this could be
due to a lack of power due to small size of study groups. These inflammatory markers are part
of the NF-κB signaling pathway [34]. In our previous studies with older rats, we found that
acute TNBS colitis was associated with colon NF-κB activation with upregulated downstream
molecules [35]. This was in accordance with findings from pre-clinical studies using chemical-
induced colitis [11, 36]. As in the present study, we previously did not observe a significant dif-
ference in colon COX-2 expression in rats with TNBS-induced colitis 7 days after the TNBS
injection while colon COX-2 expression was upregulated in rats with TNBS-induced colitis
after 2 days [17]. The timing of the present study (7 days after) may have contributed to this
discrepancy by a partial recovery.
Treatment with MODULEN-IBD1 did not reverse the intestinal damage seen on MRC.
This might have been affected by the timing of MRI. Rimola et al. showed in adult Crohn’s dis-
ease patients persistent damage on MR-enterography (MRE) despite endoscopic remission
after one year of anti-TNF or stem cell transplantation treatment [37]. The MRE abnormalities
that persisted were: mural hyperenhancement, mural thickness and strictures. One might
argue that the persistence of strictures reflects fibrotic sequelae of inflammation. In animal
models, the minimal delay in which improvement and resolution of lesions on MRC can be
expected remains to be determined, using longitudinal repeated MRC measures. Future stud-
ies could also incorporate analysis of gut microbiota, in particular since MODULEN-IBD1
has been shown to modify gut microbiota composition in children [33].
PLOS ONE | https://doi.org/10.1371/journal.pone.0259135 November 4, 2021
10 / 13
PLOS ONE TNBS acute colitis in prepubertal rats
In conclusion, we have shown that a single intra-rectal installation of TNBS in prepubertal
rats causes colonic inflammation that can be observed on histopathology and non-contrast
MRC. Treatment with MODULEN-IBD1 showed a mild anti-inflammatory effect on histopa-
thology, but not on MRC. Optimal timing of MRC remains to be determined before imple-
menting MRC as routine non-invasive assessment tool of therapeutic interventions efficacy in
TNBS colitis.
Acknowledgments
The authors would like to thank Elodie Colasse and Amelyne David for their assistance in data
acquisition.
Author Contributions
Conceptualization: Claire Dupont-Lucas, Rachel Marion-Letellier, Guillaume Savoye.
Data curation: Claire Dupont-Lucas, Rachel Marion-Letellier.
Formal analysis: Claire Dupont-Lucas, Mathilde Pala.
Investigation: Mathilde Pala, Charlène Guerin, Christine Boˆle-Feysot, Emmeline Salameh,
Alexis Goichon, Lionel Nicol, Moutaz Aziz, Ce´line Savoye-Collet.
Methodology: Claire Dupont-Lucas, Rachel Marion-Letellier, Guillaume Savoye.
Project administration: Charlène Guerin.
Supervision: Rachel Marion-Letellier, Guillaume Savoye.
Writing – original draft: Claire Dupont-Lucas.
Writing – review & editing: Claire Dupont-Lucas, Rachel Marion-Letellier, Mathilde Pala,
Charlène Guerin, Christine Boˆle-Feysot, Emmeline Salameh, Alexis Goichon, Lionel Nicol,
Moutaz Aziz, Ce´line Savoye-Collet, Guillaume Savoye.
References
1. Molodecky NA, Soon IS, Rabi DM, Ghali WA, Ferris M, Chernoff G, et al. Increasing incidence and prev-
alence of the inflammatory bowel diseases with time, based on systematic review. Gastroenterology.
2012; 142(1):46–54 e42; quiz e30. Epub 2011/10/18. https://doi.org/10.1053/j.gastro.2011.10.001
PMID: 22001864
2. Ghione S, Sarter H, Fumery M, Armengol-Debeir L, Savoye G, Ley D, et al. Dramatic Increase in Inci-
dence of Ulcerative Colitis and Crohn’s Disease (1988–2011): A Population-Based Study of French
Adolescents. Am J Gastroenterol. 2018; 113(2):265–72. Epub 2017/08/16. https://doi.org/10.1038/ajg.
2017.228 PMID: 28809388
3. Sykora J, Pomahacova R, Kreslova M, Cvalinova D, Stych P, Schwarz J. Current global trends in the
incidence of pediatric-onset inflammatory bowel disease. World J Gastroenterol. 2018; 24(25):2741–
63. Epub 2018/07/12. https://doi.org/10.3748/wjg.v24.i25.2741 PMID: 29991879
4.
5.
de Bie CI, Paerregaard A, Kolacek S, Ruemmele FM, Koletzko S, Fell JM, et al. Disease phenotype at
diagnosis in pediatric Crohn’s disease: 5-year analyses of the EUROKIDS Registry. Inflamm Bowel Dis.
2013; 19(2):378–85. Epub 2012/05/11. https://doi.org/10.1002/ibd.23008 PMID: 22573581
Fumery M, Pariente B, Sarter H, Savoye G, Spyckerelle C, Djeddi D, et al. Long-term outcome of pedi-
atric-onset Crohn’s disease: A population-based cohort study. Dig Liver Dis. 2019; 51(4):496–502.
Epub 2019/01/07. https://doi.org/10.1016/j.dld.2018.11.033 PMID: 30611597
6. Pigneur B, Seksik P, Viola S, Viala J, Beaugerie L, Girardet JP, et al. Natural history of Crohn’s disease:
comparison between childhood- and adult-onset disease. Inflamm Bowel Dis. 2010; 16(6):953–61.
Epub 2009/10/17. https://doi.org/10.1002/ibd.21152 PMID: 19834970
7. Guan Q. A Comprehensive Review and Update on the Pathogenesis of Inflammatory Bowel Disease. J
Immunol Res. 2019; 2019:7247238. Epub 2019/12/31. https://doi.org/10.1155/2019/7247238 PMID:
31886308
PLOS ONE | https://doi.org/10.1371/journal.pone.0259135 November 4, 2021
11 / 13
PLOS ONE TNBS acute colitis in prepubertal rats
8. Bramhall M, Florez-Vargas O, Stevens R, Brass A, Cruickshank S. Quality of methods reporting in ani-
mal models of colitis. Inflamm Bowel Dis. 2015; 21(6):1248–59. https://doi.org/10.1097/MIB.
0000000000000369 PMID: 25989337
9. Eichele DD, Kharbanda KK. Dextran sodium sulfate colitis murine model: An indispensable tool for
advancing our understanding of inflammatory bowel diseases pathogenesis. World J Gastroenterol.
2017; 23(33):6016–29. Epub 2017/10/04. https://doi.org/10.3748/wjg.v23.i33.6016 PMID: 28970718
10. Gadaleta RM, Garcia-Irigoyen O, Moschetta A. Exploration of Inflammatory Bowel Disease in Mice:
Chemically Induced Murine Models of Inflammatory Bowel Disease (IBD). Curr Protoc Mouse Biol.
2017; 7(1):13–28. Epub 2017/03/03. https://doi.org/10.1002/cpmo.20 PMID: 28252200
11. Siracusa R, Fusco R, Peritore AF, Cordaro M, D’Amico R, Genovese T, et al. The Antioxidant and Anti-
Inflammatory Properties of Anacardium occidentale L. Cashew Nuts in a Mouse Model of Colitis. Nutri-
ents. 2020; 12(3). Epub 2020/04/05. https://doi.org/10.3390/nu12030834 PMID: 32245085
12. Morris GP, Beck PL, Herridge MS, Depew WT, Szewczuk MR, Wallace JL. Hapten-induced model of
chronic inflammation and ulceration in the rat colon. Gastroenterology. 1989; 96(3):795–803. Epub
1989/03/01. PMID: 2914642
13. Neurath M, Fuss I, Strober W. TNBS-colitis. Int Rev Immunol. 2000; 19(1):51–62. Epub 2000/03/21.
https://doi.org/10.3109/08830180009048389 PMID: 10723677
14. Rashidian A, Muhammadnejad A, Dehpour AR, Mehr SE, Akhavan MM, Shirkoohi R, et al. Atorvastatin
attenuates TNBS-induced rat colitis: the involvement of the TLR4/NF-kB signaling pathway. Inflammo-
pharmacology. 2016; 24(2–3):109–18. Epub 2016/04/04. https://doi.org/10.1007/s10787-016-0263-6
PMID: 27038922
15. Shimizu T, Suzuki M, Fujimura J, Hisada K, Yoshikazu O, Obinata K, et al. The relationship between
the concentration of dextran sodium sulfate and the degree of induced experimental colitis in weanling
rats. J Pediatr Gastroenterol Nutr. 2003; 37(4):481–6. https://doi.org/10.1097/00005176-200310000-
00015 PMID: 14508220
16. Vicario M, Crespi M, Franch A, Amat C, Pelegri C, Moreto M. Induction of Colitis in Young Rats by Dex-
tran Sulfate Sodium. Dig Dis Sci. 2005; 50(1):143–50. https://doi.org/10.1007/s10620-005-1292-y
PMID: 27837390
17. Charpentier C, Marion-Letellier R, Savoye G, Nicol L, Mulder P, Aziz M, et al. Magnetic resonance colo-
nography in rats with TNBS-induced colitis: a feasibility and validation study. Inflamm Bowel Dis. 2012;
18(10):1940–9. https://doi.org/10.1002/ibd.22897 PMID: 22262626
18.
19.
van Rheenen PF, Aloi M, Assa A, Bronsky J, Escher JC, Fagerberg UL, et al. The Medical Management
of Paediatric Crohn’s Disease: an ECCO-ESPGHAN Guideline Update. Journal of Crohn’s & colitis.
2020. Epub 2020/10/08. https://doi.org/10.1093/ecco-jcc/jjaa161 PMID: 33026087
Fell JM, Paintin M, Arnaud-Battandier F, Beattie RM, Hollis A, Kitching P, et al. Mucosal healing and a
fall in mucosal pro-inflammatory cytokine mRNA induced by a specific oral polymeric diet in paediatric
Crohn’s disease. Alimentary pharmacology & therapeutics. 2000; 14(3):281–9. https://doi.org/10.1046/
j.1365-2036.2000.00707.x PMID: 10735920
20. Kilkenny C, Browne WJ, Cuthill IC, Emerson M, Altman DG. Improving bioscience research reporting:
the ARRIVE guidelines for reporting animal research. PLoS biology. 2010; 8(6):e1000412. https://doi.
org/10.1371/journal.pbio.1000412 PMID: 20613859
21.
Faul F, Erdfelder E, Lang AG, Buchner A. G*Power 3: a flexible statistical power analysis program for
the social, behavioral, and biomedical sciences. Behavior research methods. 2007; 39(2):175–91.
https://doi.org/10.3758/bf03193146 PMID: 17695343
22. Melchior C, Loeuillard E, Marion-Letellier R, Nicol L, Mulder P, Guerin C, et al. Magnetic resonance
colonography for fibrosis assessment in rats with chronic colitis. PloS one. 2014; 9(7):e100921. https://
doi.org/10.1371/journal.pone.0100921 PMID: 25000184
23. Charpentier C, Chan R, Salameh E, Mbodji K, Ueno A, Coeffier M, et al. Dietary n-3 PUFA May Attenu-
ate Experimental Colitis. Mediators of inflammation. 2018; 2018:8430614. https://doi.org/10.1155/2018/
8430614 PMID: 29670469
24. Salameh E, Meleine M, Gourcerol G, do Rego JC, do Rego JL, Legrand R, et al. Chronic colitis-induced
visceral pain is associated with increased anxiety during quiescent phase. American journal of physiol-
ogy Gastrointestinal and liver physiology. 2019; 316(6):G692–G700. https://doi.org/10.1152/ajpgi.
00248.2018 PMID: 30735453
25. Antoniou E, Margonis GA, Angelou A, Pikouli A, Argiri P, Karavokyros I, et al. The TNBS-induced colitis
animal model: An overview. Annals of medicine and surgery. 2016; 11:9–15. https://doi.org/10.1016/j.
amsu.2016.07.019 PMID: 27656280
26. Ballinger AB, Azooz O, El-Haj T, Poole S, Farthing MJ. Growth failure occurs through a decrease in
insulin-like growth factor 1 which is independent of undernutrition in a rat model of colitis. Gut. 2000; 46
(5):694–700. https://doi.org/10.1136/gut.46.5.695 PMID: 10764714
PLOS ONE | https://doi.org/10.1371/journal.pone.0259135 November 4, 2021
12 / 13
PLOS ONE TNBS acute colitis in prepubertal rats
27.
28.
Fitzpatrick LR, Meirelles K, Small JS, Puleo FJ, Koltun WA, Cooney RN. A new model of chronic hap-
ten-induced colitis in young rats. J Pediatr Gastroenterol Nutr. 2010; 50(3):240–50. https://doi.org/10.
1097/MPG.0b013e3181cb8f4a PMID: 20118800
Lawrance IC, Wu F, Leite AZ, Willis J, West GA, Fiocchi C, et al. A murine model of chronic inflamma-
tion-induced intestinal fibrosis down-regulated by antisense NF-kappa B. Gastroenterology. 2003; 125
(6):1750–61. https://doi.org/10.1053/j.gastro.2003.08.027 PMID: 14724828
29. Brenna O, Furnes MW, Drozdov I, van Beelen Granlund A, Flatberg A, Sandvik AK, et al. Relevance of
TNBS-colitis in rats: a methodological study with endoscopic, histologic and Transcriptomic [corrected]
characterization and correlation to IBD. PloS one. 2013; 8(1):e54543. https://doi.org/10.1371/journal.
pone.0054543 PMID: 23382912
30. Beattie RM, Schiffrin EJ, Donnet-Hughes A, Huggett AC, Domizio P, MacDonald TT, et al. Polymeric
nutrition as the primary therapy in children with small bowel Crohn’s disease. Alimentary pharmacology
& therapeutics. 1994; 8(6):609–15. https://doi.org/10.1111/j.1365-2036.1994.tb00338.x PMID:
7696450
31. Beaupel N, Brouquet A, Abdalla S, Carbonnel F, Penna C, Benoist S. Preoperative oral polymeric diet
enriched with transforming growth factor-beta 2 (Modulen) could decrease postoperative morbidity after
surgery for complicated ileocolonic Crohn’s disease. Scandinavian journal of gastroenterology. 2017;
52(1):5–10. https://doi.org/10.1080/00365521.2016.1221994 PMID: 27553420
32. Hartman C, Berkowitz D, Weiss B, Shaoul R, Levine A, Adiv OE, et al. Nutritional supplementation with
polymeric diet enriched with transforming growth factor-beta 2 for children with Crohn’s disease. The
Israel Medical Association journal: IMAJ. 2008; 10(7):503–7. PMID: 18751627
33. Pigneur B, Lepage P, Mondot S, Schmitz J, Goulet O, Dore J, et al. Mucosal Healing and Bacterial
Composition in Response to Enteral Nutrition Vs Steroid-based Induction Therapy-A Randomised Pro-
spective Clinical Trial in Children With Crohn’s Disease. Journal of Crohn’s & colitis. 2019; 13(7):846–
55. Epub 2018/12/13. https://doi.org/10.1093/ecco-jcc/jjy207 PMID: 30541015
34.
Liu T, Zhang L, Joo D, Sun SC. NF-kappaB signaling in inflammation. Signal Transduct Target Ther.
2017; 2. Epub 2017/11/22. https://doi.org/10.1038/sigtrans.2017.23 PMID: 29158945
35. Mbodji K, Charpentier C, Guerin C, Querec C, Bole-Feysot C, Aziz M, et al. Adjunct therapy of n-3 fatty
acids to 5-ASA ameliorates inflammatory score and decreases NF-kappaB in rats with TNBS-induced
colitis. J Nutr Biochem. 2013; 24(4):700–5. Epub 2012/07/31. https://doi.org/10.1016/j.jnutbio.2012.03.
022 PMID: 22841543
36. Di Paola R, Fusco R, Gugliandolo E, D’Amico R, Cordaro M, Impellizzeri D, et al. Formyl peptide recep-
tor 1 signalling promotes experimental colitis in mice. Pharmacol Res. 2019; 141:591–601. Epub 2019/
02/04. https://doi.org/10.1016/j.phrs.2019.01.041 PMID: 30711419
37. Rimola J, Alfaro I, Fernandez-Clotet A, Castro-Poceiro J, Vas D, Rodriguez S, et al. Persistent damage
on magnetic resonance enterography in patients with Crohn’s disease in endoscopic remission. Alimen-
tary pharmacology & therapeutics. 2018; 48(11–12):1232–41. https://doi.org/10.1111/apt.15013 PMID:
30345577
PLOS ONE | https://doi.org/10.1371/journal.pone.0259135 November 4, 2021
13 / 13
PLOS ONE |
10.1371_journal.pone.0266065 | RESEARCH ARTICLE
Numerical investigation of the effect of
spanwise length and mesh density on flow
around cylinder at Re = 3900 using LES model
Haider Ali1, Niaz Bahadur KhanID
Muhammad Sajid1, Adnan Munir1, A. El-Sayed Ahmed3, Khalid Abdulkhaliq M. AlharbiID
Ahmed M. Galal5,6
1*, Muhammad Jameel2, Azam Khan1,
4,
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: Ali H, Khan NB, Jameel M, Khan A, Sajid
M, Munir A, et al. (2022) Numerical investigation of
the effect of spanwise length and mesh density on
flow around cylinder at Re = 3900 using LES
model. PLoS ONE 17(4): e0266065. https://doi.
org/10.1371/journal.pone.0266065
Editor: Ahad Javanmardi, Fuzhou University,
CHINA
Received: December 21, 2021
Accepted: March 14, 2022
Published: April 8, 2022
Copyright: © 2022 Ali et al. This is an open access
article distributed under the terms of the Creative
Commons Attribution License, which permits
unrestricted use, distribution, and reproduction in
any medium, provided the original author and
source are credited.
1 National University of Sciences and Technology (NUST), Islamabad, Pakistan, 2 Department of Civil
Engineering, College of Engineering, King Khalid University, Abha, Saudi Arabia, 3 Mathematics
Department, Faculty of Science, Taif University, Taif, Saudi Arabia, 4 Mechanical Engineering Department,
College of Engineering, Umm Al-Qura University, Makkah, KSA, 5 Mechanical Engineering Department,
College of Engineering, Prince Sattam Bin Abdulaziz University, Wadi addawaser, Saudi Arabia,
6 Production Engineering and Mechanical Design Department, Faculty of Engineering, Mansoura University,
Mansoura, Egypt
* n_bkhan@yahoo.com
Abstract
Flow around circular cylinder has been extensively studied by researchers for several
decades due to its wide range of engineering applications such as in heat exchangers,
marine cables, high rise building, chimneys, and offshore structures. The lack of clear
understanding of the unsteady flow dynamics in the wake of circular cylinder and high
computational cost are still an area of high interest amongst the researchers. The aim of the
current study is to investigate the effect of variation in spanwise length and grid resolution in
the spanwise direction on the recirculation length, separation angle of wake flow by perform-
ing large eddy simulations (LES). This study is an extension to previous work by Khan, NB
et al, 2019 in which the spanwise length is restricted to 4D only. In current study, the span-
wise length is changed from 0.5D to 8D where D is diameter of cylinder and mesh resolution
in the spanwise direction is changed from 1 to 80 elements in the present study. The recircu-
lation length, separation angle and wake characteristics are analyzed in detail. It is con-
cluded that after getting optimize spanwise length, mesh resolution in the spanwise
direction is the only parameter contributing toward better result.
Data Availability Statement: All relevant data is
available within the paper.
Introduction
Funding: The authors extend their appreciation to;
The Deanship of Scientific Research at Umm Al-
Qura University for supporting this work by Grant
Code: (22UQU4310392DSR02), Taif University
Researchers concerning the support of project
number (TURSP-2020/159), Taif University, Saudi
Arabia & the deanship of scientific research at King
Khalid University for funding this work through
Investigating the unsteady nature of flow around cylinder is one of the highly research topics
in the field of offshore engineering and fluid-structure interaction (FSI). This unsteady nature
is very sensitive in terms of flow separation, boundary layer, wake region characteristics, drag
coefficient and Strouhal number. Primarily, the analysis has been performed using the three
main turbulent model i.e., direct numerical simulation (DNS), Reynolds-averaged Numerical
simulation (RANS) model, and large eddy simulation (LES) model. Each of these models dif-
fers in terms of solving the Navier-Stokes equation and presenting the effects of the flow. DNS
PLOS ONE | https://doi.org/10.1371/journal.pone.0266065 April 8, 2022
1 / 13
PLOS ONE research group program under grant number (R.G.
P. 2/93/43).
Competing interests: The authors have declared
that no competing interests exist.
Abbreviations: Cd, Drag coefficient; CL, Lift
coefficient; LES, Large eddy simulation; Lx, Length
whereas subscript x, y and z represents the
coordinate direction; Mz, Mesh resolution whereas
subscript represent coordinate; N, Number of
elements and subscript R and C stand for radial
and circumference; Re, Reynolds number; St,
Strouhal number; U, Inlet velocity.
Impact of spanwise length and mesh density on flow around cylinder at Re = 3900
is the only numerical method in which all spatial and temporal scales of turbulence are
resolved in case of flow around a cylinder. However, numerical cost is the major obstacle in
using DNS method. RANS model unable to capture the transitional flow characteristics,
boundary layer and separation region accurately [1, 2]. Due to limitation of DNS and RANS
model, LES is most attractive option for investigating the unsteady nature of flow around cyl-
inder in subcritical regime. The flow around a fixed cylinder at Re = 3900 come under the
lower subcritical flow regime, which is highly sensitive in terms of flow separation, boundary
layer nature, recirculation length, wake characteristics, and other hydrodynamic coefficients
[3]. The availability of large-scale experimental and numerical data [4–10] for the flow around
cylinders at Re = 3900 makes it an excellent benchmark case for assessing the capability of
computational tools and process.
Breuer [11] performed large eddy simulations at Re = 3900 and investigated the effect of
near-wall modeling, sub-grid scale modeling and spanwise resolution on the accuracy of
numerical model. In another study Breuer [11] computed the low recirculation length (of less
than 1.1) with a Smagorinsky model and comparatively high value of recirculation length
when using the dynamic model. Unlike dynamic Smagorinsky model, the traditional Smagor-
insky model over-predicted the importance of drag coefficient and separation angle. While
systematically investigating the accuracy of LES model on flow past a cylinder at Re = 3900,
[11] found that the resolution in spanwise direction effects the three-dimensionality of flow
and hence, accuracy of results significantly. Tremblay et al. [12] studied the effects of the SGS
model and grid resolution on large eddy simulations using technique known as cartesian grid.
The length of recirculation and profiles of mean velocity could not be reliably predicted in this
analysis. Lysenko et al. [13] used OpenFOAM tool to investigate the flow aroud a cylinder
using a dynamic k-equation SGS model and the large eddy simulations model. Parnaudeau
et al. [14] used LES with a high order scheme to simulate a flow around cylinder. The numeri-
cal analysis resulted in power spectra and turbulence statistics up to 10 D. Moreover, Wissink
and Rodi, Dong et al. [15, 16] used direct numerical simulations (DNS) to investigate flow
around cylinder.
Dong et al. [16] and MA, Karamanos, and Karniadakis [17] studied flow around cylinder
with spanwise length of 10D using Direct Numerical simulations. At a moderate to high values
of Reynolds number, Rajani et al. [2] investigated the limitation of URANS in measuring drag
forces, skin friction coefficients and mean pressure coefficients. Zhang et al. [18] investigated
the both the effects of infinite and finite cylinders on flow characteristics and observed that
free end of cylinder has significant impact on wake characteristics. Wissink et al. [15, 19] used
direct numerical simulations in order to run a series of simulations at Reynolds number = 3300
and compared the findings to experimental studies at Reynolds number = 3900. The roll-up of
the splitting shear layer, which transformed to turbulence, was observed. Even at low Reynolds
numbers, direct numerical simulations are costly and provides accurate and reliable perfor-
mance. Due to the deficiencies in the URANS method and the high computational cost of
direct numerical simulations, large eddy simulations (smagorinsky model) with fixed coeffi-
cient but without model, and with a dynamic model is most attractive choice to analyze the
unsteady nature of flow around cylinder.
Franke and Frank [20] used a cell-centered volume code to perform a sequence of large
eddy simulation at Reynolds number = 3900. The study concluded that small value of recircu-
lation length computed during analysis is mainly due to short-time averaging data. Krav-
chenko and Moin [21] used large eddy simulations and a high-order precise numerical model
which is based on B-splines to investigate the flow around a cylinder at Reynolds num-
ber = 3900. They concluded that low mesh resolution in the shear layer causes limited recircu-
lation lengths values and vice versa. At Reynolds number = 5800, Shao and Zhang [22]
PLOS ONE | https://doi.org/10.1371/journal.pone.0266065 April 8, 2022
2 / 13
PLOS ONE Impact of spanwise length and mesh density on flow around cylinder at Re = 3900
compared the Reynolds-averaged Navier–Stokes (RANS) and large eddy simulation results
(same sub-critical regime). The momentum equations were solved using a second-order
upwind scheme and bounded central differencing scheme. Khan et al. [23] used LES and Sma-
gorinsky SGS model to numerically simulate the flow over cylinder at Reynolds number = 3900.
The simulation was run for 60 non-dimensional time steps before the time statistical data were
collected to ensure that the flow was free of initial conditions and completely formed. The data
was collected for 30 vortex-shedding periods. Feng et al. [5] uses LES model to compute hydro-
dynamic coefficients, pressure distribution, velocity profiles, Reynold’s stress distribution in
wake, flow separation and recirculation length. The aim of this paper is to use a benchmark
problem to compare the performance of the Open FOAM sub grid model quantitatively, as
well as to address some key factors that affect predictive performance. Korinek et al. [7] used
LES code and the Smagorinsky—Lilly subgrid-scale model to identify the effect of spanwise
length and mesh resolution on measuring recirculation length and angle of separation around
a fixed structure at Reynolds number = 3900. Most recently, Korinek et al. [9] used partially
average NS-bridging technique to investigate the flow over same Reynolds number at different
ratio of resolved and unresolved turbulent kinetic energy. A shorter recirculation length is
obtained at low Reynolds number due to earlier transition. Filipe et al. [24] examined the
modelling accuracy of distinct RANS equations and SRS method at Re = 3900. They concluded
that SRS model is more accurate than RANS. Jiang and Cheng [25] investigated the flow
around cylinder at Re = 400 to 3900 using opensource tool with emphasize on generalize mesh
density, domain size and other parameters. Lekkalla et al. [10] performed two-dimensional
numerical simulations at Reynold number of 3900 for rotating cylinder using k-e model. They
investigated the vortex patterns and drag coefficients in the wake of the cylinder. Tian and
Xiao [26] recently studied flow around cylinder using LES with conclusion that weak produc-
tion rates due to shear layers delay the downward movement of the mean flow, resulting in a
longer recirculation region. Most recently [27–31], the flow around cylinder (in laminar and
turbulent regime) numerically and analytically has been investigated by numerous researchers
for better understanding of the unsteady behavior [18, 29, 32–39].
Current study, which is an extension of the previous study (Khan, Ibrahim, Ali, et al. 2019;
Khan, Ibrahim, Bin Mohamad Badry, et al. 2019), the impact of spanwise length (0.5D, 1D,
2D, πD, 4D, 8D to find optimum spanwise length), mesh resolution in spanwise direction (1
to 80 elements to find optimum mesh design) on the recirculation length, angle of separation,
hydrodynamic coefficients and wake characteristics in detail at Re = 3900. Earlier the study
was performed with spanwise length 4D-8D only.
Computational domains, boundary conditions and mesh
Fluid flow around cylinder is highly dependent on the flow domain size. In the past studies, as
referred earlier, size of the domain varies from 15D to 70D in crossflow (Y) and streamflow
(X) directions. To resolve the wake region and boundary layers, grid points are being clustered
in wake region and over the cylinder. In past studies, range of crossflow (Ly) domain is kept
between 80D to 20D, while range of streamflow (Lx) domain is usually kept from 40D to 20D.
In several cases, mesh is designed in such a way that it is divided into several regions i.e.,
O-Grid meshing is surrounded over cylinder and remaining region will then meshed using
structured method. In the current study, domain size of 40D×20D (inflow x cross flow) is
used, whereas the spanwise length is varied from 8D to 0.5D.
Comparatively, a small variation in results are observed while varying the spanwise length
from 4D to 8D [15, 19]. There is only 1% influence on solution by increasing spanwise length
beyond 4D [40]. In order to minimize cylinder response effects, blockage ratio of 5% (cylinder
PLOS ONE | https://doi.org/10.1371/journal.pone.0266065 April 8, 2022
3 / 13
PLOS ONE Impact of spanwise length and mesh density on flow around cylinder at Re = 3900
Fig 1. CAD model and boundary conditions.
https://doi.org/10.1371/journal.pone.0266065.g001
diameter / domain width) is suitable [41], [4, 42]. Due to this reason, computational domain
size of 40D×20D (streamflow ×crossflow) with varying spanwise lengths (8D, 4D, πD, 2D, 1D,
0.5D) is used in the current study. Flow parameters are highly influenced by aspect ratio.
Therefore, boundary conditions are periodically assigned to bottom and top of the wall so they
can reduce the effect of impact ratio on the numerical results. Left side inlet boundary of the
domain is placed 10D from center of the cylinder. Right side outlet boundary of the domain is
placed 30D from center of the cylinder as shown in Fig 1. Lower and upper walls of this
computational domain is 10D from center of the cylinder and are symmetric in nature. In
spanwise direction, periodic boundaries are assigned with multiple spanwise lengths (8D, 4D,
πD, 2D, 1D, 0.5D).
Velocity at inlet is 0.6 m/s, which is on right side of domain, maintaining Reynolds number
Re = 3900 by keeping diameter of cylinder D = 0.1m, viscosity = 0.000016 kg/ms, density = 1.04
kg/m3. At the outflow border, a static reference pressure of 0 Pa is applied on average. A sym-
metric wall condition is applied to both upper and lower walls of flow domain. To investigate
the boundary layer separation and vortex generation phenomenon, a no-slip condition is
given to cylinder wall. The modeling and analysis are performed using ANSYS tool (design
modeler and ICEM CFD respectively). All the meshes are designed using structured method
of meshing and the computational domain is then divided into number of regions in a manner
that O-Grid near the boundary wall of cylinder and then structured meshed away from the cyl-
inder, as shown in Fig 2. Rectangular domain is used in all numerical studies. Greifzu [6] con-
cluded that, value of y+ must be smaller or equal to unity in order to ensure the proper
resolutions of grids near the wall of cylinder. In all case studies, y+ value equal or less than 1 is
maintained near the cylinder wall. First node is placed at 0.002D in all numerical studies, in
order to completely ensure y+ less than unity. Fig 2 gives details of overall mesh design and
view near the wall of cylinder respectively. Multiple meshes are designed with different span-
wise lengths and in some cases, study with different mesh resolution is performed in order to
PLOS ONE | https://doi.org/10.1371/journal.pone.0266065 April 8, 2022
4 / 13
PLOS ONE Impact of spanwise length and mesh density on flow around cylinder at Re = 3900
Fig 2. Mesh details.
https://doi.org/10.1371/journal.pone.0266065.g002
validate previous study. Table 1 review the mesh type used in literature to perform flow around
cylinder study.
Results and discussions
Table 2 shows case studies (A to X) with spanwise length ranging from 8D to 0.1D (8D, 4D,
πD, 2D, 0.5D and 0.1D) and mesh density ranging from 0.0125 to 2. In all these case studies,
number of elements on cylinder circumference (Nc) × number of elements on cylinder in
radial direction (NR) are 240×60 respectively. 240x60 is taken from the study of Khan et al. [7],
who concluded that reduction in number of circumferential and radial nodes will overestimate
the value of Cd, reduction in recirculation length, and a delay in flow separation.
In case studies (A to I) having spanwise length (Lz) of 8D, 4D and πD, it is clearly observed
that results are well converged when the Mz is 0.1. Further improvement in Mz value (by
increasing the number of elements in spanwise length) is only resulting in increasing computa-
tional cost. It is also found that drag coefficient and St number are less sensitive compared to
separation angle and recirculation length. Even with higher value of Mz = 0.2 (Case A, D and
G), drag coefficient and St number are well captured, however, deficiency in flow separation
Table 1. Mesh details and boundary conditions used in literature.
NT×106
Mesh type
[15] 62
[14] 45.8
[12] 7.7
[13] 5.76
[43] 13.5
[44] 5.5
[4] 1.7
[45] 6
[46] 1.04
[11] 1.74
[7] 2.031
[7] 3.064
[7] 4.063
O-Grid
Hybrid
Hybrid
O-type
Hybrid
O-type
Hybrid
Hybrid
Hybrid
O-type
Hybrid
Hybrid
Hybrid
https://doi.org/10.1371/journal.pone.0266065.t001
Lx × Ly
25D×20D
20D×20D
20D×20D
50D
32D×16D
35D
24D×8D
30D×20D
32D×16D
30D
40D×20D
40D×20D
40D×20D
Lz
4D-8D
πD
πD
πD
4D
πD
10D
4D
3.288D
πD
4D
4D
8D
PLOS ONE | https://doi.org/10.1371/journal.pone.0266065 April 8, 2022
5 / 13
PLOS ONE Impact of spanwise length and mesh density on flow around cylinder at Re = 3900
Table 2. Results comparison with grid variation in spanwise direction.
Case NT×106
2.51
A
B
C
D
E
F
G
H
I
J
K
L
M
N
O
P
Q
R
S
T
U
V
W
X
4.90
9.813
1.316
2.51
4.9033
1.0171
1.97
2.51
0.17
0.41
0.71
1.31
2.51
0.41
0.71
1.01
2.51
0.17
0.29
0.35
0.41
1.31
2.51
Lz
8D
8D
8D
4D
4D
4D
πD
πD
πD
2D
2D
2D
2D
2D
1D
1D
1D
1D
0.5D
0.5D
0.5D
0.5D
0.5D
0.5D
Nz
40
80
160
20
40
80
15
31
40
1
5
10
20
40
5
10
15
40
1
3
4
5
20
40
Mz = Lz/Nz
0.2
0.1
0.05
0.2
0.1
0.05
0.2
0.1
0.0785
2
0.4
0.2
0.1
0.05
0.2
0.1
0.06
0.025
0.5
0.166
0.125
0.1
0.025
0.0125
https://doi.org/10.1371/journal.pone.0266065.t002
Nc×NR
240×60
240×60
240×60
240×60
240×60
240×60
240×60
240×60
240×60
240×60
240×60
240×60
240×60
240×60
240×60
240×60
240×60
240×60
240×60
240×60
240×60
240×60
240×60
240×60
Cd
St
1.03
0.98
0.98
1.0
0.98
0.9866
1.0413
0.96
0.96
1.42
1.22
1.04
0.97
0.97
1.12
1.01
1.00
0.99
1.59
1.05
1.07
1.00
1.02
1.04
ϴs
89.19
87.66
87.66
89.19
87.66
87,66
89.19
87.66
87.66
104.44
95.29
89.19
87.66
87.66
90.73
87.66
87.66
87.66
101.3
89.19
89.19
87.66
87.66
87.66
0.21
0.21
0.203
0.187
0.201
0.200
0.182
0.202
0.202
0.198
0.199
0.201
0.203
0.197
0.194
0.212
0.212
0.212
0.198
0.199
0.198
0.197
0.215
0.205
Lr/D
1.42
1.70
1.70
1.11
1.70
1.70
1.1132
1.71
1.71
0.38
0.72
1.36
1.70
1.70
1.10
1.60
1.60
1.60
0.32
1.34
1.25
1.42
1.42
1.42
angle (ϴs) and recirculation length (Lr/D) is observed. Higher value of ϴs and shorter Lr/D is
observed when the value of Mz is increased. It is concluded form case studies (A to I) that ϴs
and Lr/D should be the focused more to get accurate result.
In case studies (J to N) having constant spanwise length of 2D whereas the number of ele-
ments in spanwise direction are varying from 1 to 40, resulting in mesh density (length/NO. of
elements) ranging from 0.05 to 2. It is clearly observed from the result of case J that coarse
mesh (1 element in spanwise direction = Nz) does not capture the flow behavior. Higher value
of hydrodynamic coefficients, shorter recirculation length and delay in flow separation is
observed. With increase in number of elements in spanwise direction, the results improve and
converge. In case M, where number of elements are 20 (or Nz = 0.1) the hydrodynamic coeffi-
cients, strouhal number, recirculation length and angle of separation are computed well. Fur-
ther increase in number of elements only results in increasing the computational costs. These
case studies (J to N) show that having spanwise length of 2D, the results are converged at mesh
density of 0.1. This behavior is observed at spanwise length of 8D, 4D, πD, and 2D having total
number of elements of 4.90, 2.51, 1.97 and 1.31 million, respectively.
Additional studies are performed to further investigate the impact of spanwise length and
mesh density on flow behavior. Further reduction in spanwise length (1D and 0.5D) with
mesh density ranging from 0.2 to 0.025 have been investigated (Case J to Case X). Shorter
recirculation length and delay in flow separation is observed in all the cases, irrespective of the
mesh density, with spanwise length less than 2D. Although, drag and Strouhal number is well
captured at high mesh density cases, however, the most sensitive parameters like recirculation
PLOS ONE | https://doi.org/10.1371/journal.pone.0266065 April 8, 2022
6 / 13
PLOS ONE Impact of spanwise length and mesh density on flow around cylinder at Re = 3900
Table 3. At Re = 3900, comparison between numerical and experimental results.
Case Cd
[47], experiment 0.98±0.005
[14], experiment
- - -
Laurenco and shih, experiment 0.99
[48], experiment 0.98±0.005
[16], DNS - - -
[49] 0.88
[13], LES 0.97
[12], LES 1.15
[40], LES 1.02
[50], LES 0.99
[7], Case D, LES 0.98
[7], Case L, LES 0.986
[7], Case O, LES 0.982
Present Case A, 8D, LES 0.98
Present Case E, 4D, LES 0.98
Present Case H, πD, LES 0.966
Present Case M, 2D, LES 0.98
Present Case P, 1D, LES 1.012
Present Case W, 0.5D, LES 1.0
ϴs
- - - -
- - -
86
- - -
- - -
911
89
86.5
86
- - -
86.18
86.18
86.18
86.77
86.77
86.77
86.77
86.77
86.77
Lr/D
1.33±0.2
1.560.21
St
0.215±0.005
0.21
1.19
- - -
1.47
1.04
1.67
1.02
1.49
1.37
1.68
1.70
1.73
1.70
1.70
1.71
1.70
1.61
1.43
0.22
0.21
0.203
0.250
0.209
0.215
0.207
0.212
0.218
0.205
0.21
0.21
0.201
0.202
0.203
0.212
0.204
https://doi.org/10.1371/journal.pone.0266065.t003
length and flow separation angle are not converged. From all these studies, it is observed that
minimum spanwise length should be equal or greater than 2D in order to well capture the
hydrodynamic coefficient, recirculation length, flow separation and Strouhal number. Further-
more, at spanwise length of 2D, the mesh density should be kept equal to or higher than 0.1.
Table 3 show comparison between experimental [14], numerical [7, 13, 16, 20, 40], and cur-
rent results (spanwise length 8D, 4D, πD, 2D, and 0.5D with mesh density of 0.1.
Fig 3 depicts pressure distribution over the cylinder surface. This research demonstrates
that the findings are in strong accordance with the experimental and numerical results.
The mean streamwise velocity profiles in the wake of the cylinder are shown in Fig 4. The
mean velocity is calculated from the cylinder’s centerline to a 10-diameter distance behind the
cylinder, as shown in Fig 5. In Fig 6, the results show that the detached eddy simulation (DES)
has a slightly shorter recirculation length in comparison with experimental results. There is a
small disparity between the experimental results and the findings of this research study. This
disparity may be due to Lourenco and Shih’s usage of the PIV process, where an external dis-
ruption causes the separating shear layer to transition early [21]. It is also observed that coarse
mesh in spanwise direction result in v-shape profile of mean streamwise velocity which is
improved to U-shaped with improvement in mesh density. Overall, this research study vali-
dates previous numerical and experimental results. This LES case study’s recirculation length
matches well with Khan [7] numerical results.
Fig 6 compares the mean streamwise velocity profiles in the wake region to previous studies
(at x/D = 1, x/D = 3, and x/D = 5). The data ranges from y/d = –3 to y/d = 3. By dividing the
mean streamwise velocity by the inlet velocity, the mean streamwise velocity is normalized.
Present results agrees well with Khan [7] results. Also, there is a slight difference in peak at x/
d = 1, this discrepancy may be attributable to Lourenco and Shih’s different experimental
methods, where some external disruption will cause early transition and shear layers separa-
tion. At x/d = 1, a U-shape profile is observed, indicating that the current data are reliable and
PLOS ONE | https://doi.org/10.1371/journal.pone.0266065 April 8, 2022
7 / 13
PLOS ONE Impact of spanwise length and mesh density on flow around cylinder at Re = 3900
Fig 3. Coefficient of pressure (Cp) vs Angle (ϴ).
https://doi.org/10.1371/journal.pone.0266065.g003
accurate. When the grid resolution is coarse enough in the spanwise direction, a V-shaped pro-
file is observed, according to Khan [7] and Kravchenko and Moin [21].
At x/D = 1, x/D = 3, x/D = 5, the mean crossflow velocity profiles in the wake region are
shown in Fig 7. There is a difference between present study results and Lourenco’s and Shih’s
experimental and Trembley [12] results in the wake region near the cylinder at x/D = 1. How-
ever, the current LES study is in accordance with the profile of the mean crossflow velocity
Fig 4. Mean stream velocity profile at centerline, behind the cylinder.
https://doi.org/10.1371/journal.pone.0266065.g004
PLOS ONE | https://doi.org/10.1371/journal.pone.0266065 April 8, 2022
8 / 13
PLOS ONE Impact of spanwise length and mesh density on flow around cylinder at Re = 3900
Fig 5. Vertical profiles and centerline sketch behind the cylinder.
https://doi.org/10.1371/journal.pone.0266065.g005
component of simulations by Khan [7]. This LES analysis captures the results well, away from
the cylinder. Norberg [47] investigated the wake flow near a cylinder with various geometrical
parameters and found that the spanwise end condition has a major impact on the onset shear
layer instability. The periodic boundary condition is used in simulations in the spanwise faces,
so the aspect ratio near the cylinder wake has no effect on the results. However, near the wake,
a grid independence analysis is needed to obtain accurate results.
Conclusions
In this study, flow around cylinder at Re = 3900 has been investigated using LES model, with
main focus on optimization of spanwise length and mesh distribution. This study mainly
extended the previous studies by analyzing the case studies from (8D to 0.5D) with mesh den-
sity in the range of (2 to 0.0125). The study concluded that irrespective of the spanwise length,
mesh density should not be kept greater than 0.1. Mesh density greater than 0.1 result in delay
in separation angle, shorter recirculation length and over-predicted value of hydrodynamic
Fig 6. Comparison of mean stream velocity profiles of flow over cylinder at X/D = 1,3 and 5 for present study and
other numerical and experimental results.
https://doi.org/10.1371/journal.pone.0266065.g006
PLOS ONE | https://doi.org/10.1371/journal.pone.0266065 April 8, 2022
9 / 13
PLOS ONE Impact of spanwise length and mesh density on flow around cylinder at Re = 3900
Fig 7. Comparison of mean crossflow velocity profiles of flow over cylinder at X/D = 1,3 and 5 for present study and other
numerical and experimental results.
https://doi.org/10.1371/journal.pone.0266065.g007
coefficients. It is also observed that coarse mesh in spanwise direction result in V-shape profile
of mean streamwise velocity which is improved to U-shaped with improvement in mesh den-
sity. With mesh density equal to 0.1, it has been concluded that optimum spanwise length of
2D is able to extract the reliable results of hydrodynamic coefficients, Strouhal number, separa-
tion angle and recirculation length. It is also observed that further reducing the spanwise
length results in shorter recirculation length, even with high mesh density in spanwise length.
It is also concluded that drag coefficient and Strouhal number are comparatively less sensitive
parameters as the change is spanwise length and mesh resolution have minor impact on the
result.
Author Contributions
Conceptualization: Haider Ali, Niaz Bahadur Khan, Muhammad Jameel, Azam Khan,
Muhammad Sajid, Adnan Munir, A. El-Sayed Ahmed, Khalid Abdulkhaliq M. Alharbi,
Ahmed M. Galal.
Data curation: Haider Ali, Niaz Bahadur Khan, Muhammad Jameel, Azam Khan, Muham-
mad Sajid, Adnan Munir, A. El-Sayed Ahmed, Khalid Abdulkhaliq M. Alharbi, Ahmed M.
Galal.
Formal analysis: Haider Ali, Niaz Bahadur Khan, Muhammad Jameel, Azam Khan, Muham-
mad Sajid, Adnan Munir, A. El-Sayed Ahmed, Khalid Abdulkhaliq M. Alharbi, Ahmed M.
Galal.
Funding acquisition: Haider Ali, Niaz Bahadur Khan, Muhammad Jameel, Azam Khan,
Muhammad Sajid, Adnan Munir, A. El-Sayed Ahmed, Khalid Abdulkhaliq M. Alharbi,
Ahmed M. Galal.
PLOS ONE | https://doi.org/10.1371/journal.pone.0266065 April 8, 2022
10 / 13
PLOS ONE Impact of spanwise length and mesh density on flow around cylinder at Re = 3900
Investigation: Haider Ali, Niaz Bahadur Khan, Muhammad Jameel, Azam Khan, Muhammad
Sajid, Adnan Munir, A. El-Sayed Ahmed, Khalid Abdulkhaliq M. Alharbi, Ahmed M.
Galal.
Methodology: Haider Ali, Niaz Bahadur Khan, Muhammad Jameel, Azam Khan, Muhammad
Sajid, Adnan Munir, A. El-Sayed Ahmed, Khalid Abdulkhaliq M. Alharbi, Ahmed M.
Galal.
Project administration: Haider Ali, Niaz Bahadur Khan, Muhammad Jameel, Azam Khan,
Muhammad Sajid, Adnan Munir, A. El-Sayed Ahmed, Khalid Abdulkhaliq M. Alharbi,
Ahmed M. Galal.
Resources: Haider Ali, Niaz Bahadur Khan, Muhammad Jameel, Azam Khan, Muhammad
Sajid, Adnan Munir, A. El-Sayed Ahmed, Khalid Abdulkhaliq M. Alharbi, Ahmed M.
Galal.
Software: Haider Ali, Niaz Bahadur Khan, Muhammad Jameel, Azam Khan, Muhammad
Sajid, Adnan Munir, A. El-Sayed Ahmed, Khalid Abdulkhaliq M. Alharbi, Ahmed M.
Galal.
Supervision: Haider Ali, Niaz Bahadur Khan, Muhammad Jameel, Azam Khan, Muhammad
Sajid, Adnan Munir, A. El-Sayed Ahmed, Khalid Abdulkhaliq M. Alharbi, Ahmed M.
Galal.
Validation: Haider Ali, Niaz Bahadur Khan, Muhammad Jameel, Azam Khan, Muhammad
Sajid, Adnan Munir, A. El-Sayed Ahmed, Khalid Abdulkhaliq M. Alharbi, Ahmed M.
Galal.
Visualization: Haider Ali, Niaz Bahadur Khan, Muhammad Jameel, Azam Khan, Muhammad
Sajid, Adnan Munir, A. El-Sayed Ahmed, Khalid Abdulkhaliq M. Alharbi, Ahmed M.
Galal.
Writing – original draft: Haider Ali, Niaz Bahadur Khan, Muhammad Jameel, Azam Khan,
Muhammad Sajid.
Writing – review & editing: Haider Ali, Niaz Bahadur Khan, Muhammad Jameel, Azam
Khan, Muhammad Sajid, Adnan Munir, A. El-Sayed Ahmed, Khalid Abdulkhaliq M.
Alharbi, Ahmed M. Galal.
References
1. Deng G., et al., Vortex-shedding flow predictions with eddy-viscosity models. Eng. Turb. Modelling and
Experiments, 2014. 2: p. 143–152.
2. Rajani B., Kandasamy A., and Majumdar S., On the reliability of eddy viscosity based turbulence mod-
els in predicting turbulent flow past a circular cylinder using URANS approach. Journal of Applied Fluid
Mechanics, 2012. 5(11): p. 67–79.
3. Sumer B.M. and Fredsøe J., Hydrodynamics around cylindrical structures. 1997: World Scientific.
4.
5.
Fang, Y.Y. and Z.L. Han. Numerical experimental research on the hydrodynamic performance of flow
around a three dimensional circular cylinder. in Applied Mechanics and Materials. 2011. Trans Tech
Publ.
Feng Z., et al. Comparisons of Subgrid-Scale Models for OpenFoam Large-Eddy Simulation. in Journal
of Physics: Conference Series. 2021. IOP Publishing.
6. Greifzu F., et al., Assessment of particle-tracking models for dispersed particle-laden flows imple-
mented in OpenFOAM and ANSYS FLUENT. Engineering Applications of Computational Fluid
Mechanics, 2016. 10(1): p. 30–43.
PLOS ONE | https://doi.org/10.1371/journal.pone.0266065 April 8, 2022
11 / 13
PLOS ONE Impact of spanwise length and mesh density on flow around cylinder at Re = 3900
7. Khan N.B., et al., Numerical simulation of flow with large eddy simulation at Re = 3900: A study on the
accuracy of statistical quantities. International Journal of Numerical Methods for Heat & Fluid Flow,
2019.
8. Khan N.B. and Ibrahim Z., Numerical investigation of vortex-induced vibration of an elastically mounted
circular cylinder with One-degree of freedom at high Reynolds number using different turbulent models.
Proceedings of the Institution of Mechanical Engineers, Part M: Journal of Engineering for the Maritime
Environment, 2019. 233(2): p. 443–453.
9. Korinek, T., et al., Partially-Averaged Navier-Stokes simulation of the flow around circular cylinder at Re
= 3900: comparison of low and high Reynolds number approaches, in Proceedings Topical Problems of
Fluid Mechanics 2020. 2020. p. 106–113.
10.
Lekkala, M.R., et al. Numerical Assessment of Flow Around Circular Cylinder. in Proceedings of the
International Conference on Civil, Offshore and Environmental Engineering. 2021. Springer.
11. Breuer M., Large eddy simulation of the subcritical flow past a circular cylinder: numerical and modeling
aspects. International journal for numerical methods in fluids, 1998. 28(9): p. 1281–1302.
12.
13.
Tremblay F., Manhart M., and Friedrich R., LES of flow around a circular cylinder at a subcritical Rey-
nolds number with cartesian grids, in Advances in LES of Complex flows. 2002, Springer. p. 133–150.
Lysenko D.A., Ertesvåg I.S., and Rian K.E., Large-eddy simulation of the flow over a circular cylinder at
Reynolds number 3900 using the OpenFOAM toolbox. Flow, turbulence and combustion, 2012. 89(4):
p. 491–518.
14. Parnaudeau P., et al., Experimental and numerical studies of the flow over a circular cylinder at Rey-
nolds number 3900. Physics of Fluids, 2008. 20(8): p. 085101.
15. Wissink J.G. and Rodi W., Large-scale computations of flow around a circular cylinder, in High perfor-
mance computing on vector systems 2007. 2008, Springer. p. 71–81.
16. Dong S., et al., A combined direct numerical simulation-particle image velocimetry study of the turbulent
near wake. Journal of Fluid Mechanics, 2006. 569: p. 185.
17. MA X., Karamanos G.-S., and Karniadakis G., Dynamics and low-dimensionality of a turbulent near
wake. Journal of fluid mechanics, 2000. 410: p. 29–65.
18.
Zhang H., et al., Large-eddy simulation of the flow past both finite and infinite circular cylinders at Re =
3900. Journal of Hydrodynamics, 2015. 27(2): p. 195–203.
19. Wissink J. and Rodi W., Numerical study of the near wake of a circular cylinder. International journal of
heat and fluid flow, 2008. 29(4): p. 1060–1070.
20.
Franke J. and Frank W., Large eddy simulation of the flow past a circular cylinder at ReD = 3900. Jour-
nal of wind engineering and industrial aerodynamics, 2002. 90(10): p. 1191–1206.
21. Kravchenko A.G. and Moin P., Numerical studies of flow over a circular cylinder at Re D = 3900. Physics
of fluids, 2000. 12(2): p. 403–417.
22. Shao J. and Zhang C., Numerical analysis of the flow around a circular cylinder using RANS and LES.
International Journal of Computational Fluid Dynamics, 2006. 20(5): p. 301–307.
23. Young, M. and A. Ooi, Comparative assessment of LES and URANS for flow over a cylinder at a Rey-
nolds number of 3900. 2007.
24. Pereira F.S., Vaz G., and Ec¸a L. J.O.E., Evaluation of RANS and SRS methods for simulation of the
flow around a circular cylinder in the sub-critical regime. 2019. 186: p. 106067.
25.
26.
Jiang H. and Cheng L. J.P.o.F., Large-eddy simulation of flow past a circular cylinder for Reynolds num-
bers 400 to 3900. 2021. 33(3): p. 034119.
Tian G. and Xiao Z. J.A.A., New insight on large-eddy simulation of flow past a circular cylinder at sub-
critical Reynolds number 3900. 2020. 10(8): p. 085321.
27. Chu Y.-M., et al., Enhancement in thermal energy and solute particles using hybrid nanoparticles by
engaging activation energy and chemical reaction over a parabolic surface via finite element approach.
2021. 5(3): p. 119.
28.
Zhao, T.H., M.I. Khan, and Y.M.J.M.M.i.t.A.S. Chu, Artificial neural networking (ANN) analysis for heat
and entropy generation in flow of non-Newtonian fluid between two rotating disks. 2021.
29. Nazeer M., et al., Theoretical study of MHD electro-osmotically flow of third-grade fluid in micro channel.
2022. 420: p. 126868.
30.
Zha T.-H., et al., A fuzzy-based strategy to suppress the novel coronavirus (2019-NCOV) massive out-
break. 2021: p. 160–176.
31. Chu Y.-M., et al., Combined impact of Cattaneo-Christov double diffusion and radiative heat flux on bio-
convective flow of Maxwell liquid configured by a stretched nano-material surface. 2022. 419: p.
126883.
PLOS ONE | https://doi.org/10.1371/journal.pone.0266065 April 8, 2022
12 / 13
PLOS ONE Impact of spanwise length and mesh density on flow around cylinder at Re = 3900
32. Pereira F., et al., Simulation of the flow around a circular cylinder at Re = 3900 with Partially-Averaged
Navier-Stokes equations. 2018. 69: p. 234–246.
33. Chu S., et al., Three-dimensional spectral proper orthogonal decomposition analyses of the turbulent
flow around a seal-vibrissa-shaped cylinder. 2021. 33(2): p. 025106.
34. Siddiqui A.A. and Turkyilmazoglu M. J.M., A new theoretical approach of wall transpiration in the cavity
flow of the ferrofluids. 2019. 10(6): p. 373.
35.
Turkyilmazoglu M. J.T.E.P.J.P., Exact solutions concerning momentum and thermal fields induced by a
long circular cylinder. 2021. 136(5): p. 1–10.
36. Mustafa, T.J.I.J.o.N.M.f.H. and F. Flow, Eyring–Powell fluid flow through a circular pipe and heat trans-
fer: full solutions. 2020.
37. Munir A., et al., Numerical investigation of the effect of plane boundary on two-degree-of-freedom of
vortex-induced vibration of a circular cylinder in oscillatory flow. 2018. 148: p. 17–32.
38. Munir A., et al., Flow-induced vibration of a rotating circular cylinder at high reduced velocities and high
rotation rates. 2021. 238: p. 109562.
39. Munir A., et al., Effects of gap ratio on flow-induced vibration of two rigidly coupled side-by-side cylin-
ders. 2019. 91: p. 102726.
40. Afgan I., et al., Large eddy simulation of the flow around single and two side-by-side cylinders at subcrit-
ical Reynolds numbers. Physics of Fluids, 2011. 23(7): p. 075101.
41. Mittal S., Free vibrations of a cylinder: 3-D computations at Re = 1000. Journal of Fluids and Structures,
2013. 41: p. 109–118.
42.
Zdravkovich M., Conceptual overview of laminar and turbulent flows past smooth and rough circular cyl-
inders. Journal of wind engineering and industrial aerodynamics, 1990. 33(1–2): p. 53–62.
43. Prsic M.A., et al., Large Eddy Simulations of flow around a smooth circular cylinder in a uniform current
in the subcritical flow regime. Ocean engineering, 2014. 77: p. 61–73.
44. Mani A., Moin P., and Wang M., Computational study of optical distortions by separated shear layers
and turbulent wakes. Journal of fluid mechanics, 2009. 625: p. 273.
45. Meyer M., Hickel S., and Adams N., Assessment of implicit large-eddy simulation with a conservative
immersed interface method for turbulent cylinder flow. International Journal of Heat and Fluid Flow,
2010. 31(3): p. 368–377.
46. Heggernes, K.K., Numerical simulation of three-dimensional viscous flow around marine structures.
2005.
47. Norberg C., Effects of Reynolds number and a low-intensity freestream turbulence on the flow around a
circular cylinder. Chalmers University, Goteborg, Sweden, Technological Publications, 1987. 87(2): p.
1–55.
48. Ong L. and Wallace J., The velocity field of the turbulent very near wake of a circular cylinder. Experi-
ments in fluids, 1996. 20(6): p. 441–453.
49.
Liaw, K.F., Simulation of flow around bluff bodies and bridge deck sections using CFD. 2005, University
of Nottingham Nottingham.
50. Park N., et al., A dynamic subgrid-scale eddy viscosity model with a global model coefficient. Physics of
Fluids, 2006. 18(12): p. 125109.
PLOS ONE | https://doi.org/10.1371/journal.pone.0266065 April 8, 2022
13 / 13
PLOS ONE |
10.1371_journal.pone.0253387 | RESEARCH ARTICLE
Classification of cannabis strains in the
Canadian market with discriminant analysis of
principal components using genome-wide
single nucleotide polymorphisms
Dan Jin1,2, Philippe HenryID
3,4, Jacqueline Shan2, Jie ChenID
1,5*
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: Jin D, Henry P, Shan J, Chen J (2021)
Classification of cannabis strains in the Canadian
market with discriminant analysis of principal
components using genome-wide single nucleotide
polymorphisms. PLoS ONE 16(6): e0253387.
https://doi.org/10.1371/journal.pone.0253387
Editor: Tzen-Yuh Chiang, National Cheng Kung
University, TAIWAN
Received: November 10, 2020
Accepted: June 3, 2021
Published: June 28, 2021
Copyright: © 2021 Jin et al. This is an open access
article distributed under the terms of the Creative
Commons Attribution License, which permits
unrestricted use, distribution, and reproduction in
any medium, provided the original author and
source are credited.
Data Availability Statement: Raw genome
sequencing data for 23 strains are available from
the NCBI with BioProject PRJNA683613.
Funding: PBG BioPharma Inc. (https://
pbgbiopharma.com/) provided funding support in
the form of salaries for authors DJ and JS. PBG
BioPharma Inc. also provided financial support for
genome sequencing and travel expenses. JS is the
founder and CEO of PBG BioPharma Inc., and
reviewed the manuscript. Labs-Mart Inc. (http://
labs-mart.ca/) provided chemical standards and
1 Department of Biomedical Engineering, University of Alberta, Edmonton, Alberta, Canada, 2 PBG
BioPharma Inc., Leduc, Alberta, Canada, 3 Egret Bioscience Ltd., West Kelowna, British Columbia, Canada,
4 Lighthouse Genomics Inc., Salt Spring Island, British Columbia, Canada, 5 Department of Electrical and
Computer Engineering, University of Alberta, Edmonton, Alberta, Canada
* jc65@ualberta.ca
Abstract
The cannabis community typically uses the terms “Sativa” and “Indica” to characterize drug
strains with high tetrahydrocannabinol (THC) levels. Due to large scale, extensive, and
unrecorded hybridization in the past 40 years, this vernacular naming convention has
become unreliable and inadequate for identifying or selecting strains for clinical research
and medicinal production. Additionally, cannabidiol (CBD) dominant strains and balanced
strains (or intermediate strains, which have intermediate levels of THC and CBD), are not
included in the current classification studies despite the increasing research interest in the
therapeutic potential of CBD. This paper is the first in a series of studies proposing that a
new classification system be established based on genome-wide variation and supple-
mented by data on secondary metabolites and morphological characteristics. This study
performed a whole-genome sequencing of 23 cannabis strains marketed in Canada, aligned
sequences to a reference genome, and, after filtering for minor allele frequency of 10%,
identified 137,858 single nucleotide polymorphisms (SNPs). Discriminant analysis of princi-
pal components (DAPC) was applied to these SNPs and further identified 344 structural
SNPs, which classified individual strains into five chemotype-aligned groups: one CBD dom-
inant, one balanced, and three THC dominant clusters. These structural SNPs were all mul-
tiallelic and were predominantly tri-allelic (339/344). The largest portion of these SNPs
(37%) occurred on the same chromosome containing genes for CBD acid synthases
(CBDAS) and THC acid synthases (THCAS). The remainder (63%) were located on the
other nine chromosomes. These results showed that the genetic differences between mod-
ern cannabis strains were at a whole-genome level and not limited to THC or CBD produc-
tion. These SNPs contained enough genetic variation for classifying individual strains into
corresponding chemotypes. In an effort to elucidate the confused genetic backgrounds of
commercially available cannabis strains, this classification attempt investigated the utility of
DAPC for classifying modern cannabis strains and for identifying structural SNPs.
PLOS ONE | https://doi.org/10.1371/journal.pone.0253387 June 28, 2021
1 / 14
PLOS ONE instrumentation support for chemical testing, but
did not have any additional role in study design,
data collection and analysis, decision to publish, or
preparation of the manuscript. Egret Bioscience
Ltd. (https://egret.bio) and Lighthouse Genomics
Inc. (https://lighthousegenomics.com/) provided
support in the form of salaries for author PH, but
did not have any additional role in the study design,
data collection and analysis, decision to publish, or
preparation of the manuscript. The specific roles of
these authors are articulated in the ‘author
contributions’ section.
Competing interests: The funder provided support
in the form of salaries for authors DJ, PH, and JS.
This does not alter our adherence to PLOS ONE
policies on sharing data and materials.
Genome-wide single nucleotide polymorphisms to classify cannabis strains
Introduction
Cannabis has a complex breeding history. Whether its botanical classification is monotypic
(sativa) or polytypic (sativa and indica) remains controversial [1]. Since the 1980s, breeding
for high psychoactive THC content has occurred very aggressively in North America [2].
Nearly all drug-type cannabis currently cultivated in the USA, Canada, and Europe are hybrid-
ized, resulting in thousands of strains [3]. Recent genetic studies focused on validating the ver-
nacular classification of “Sativa” and “Indica” [4–7]. However, this terminology is inadequate
for identifying or selecting strains for clinical research and medicinal production due to the
misuse of the botanical nomenclature, extensive cross-breeding, and unreliable labelling dur-
ing unrecorded hybridization [2]. One genetic study found that the reported ancestry percent-
age of “Sativa” vs. “Indica” for 81 drug stains is only moderately correlated with the calculated
genetic structure (r2 = 0.36) [5]. In addition, CBD dominant strains and balanced strains
(THC � CBD), which have gained increasing attention due to CBD’s use as a therapeutic [8–
12], have been omitted in recent classification studies.
Cannabis has a diploid genome (2n = 20) with nine autosomal chromosomes and one pair
of sex chromosomes [13]. The length of the haploid genome size is 818 Mbp for females and
843 Mbp for males [14]. An SNP is a variation of a single nucleotide at a specific position in
the genome, and it is useful for understanding the genetic basis of diversity among populations
[15]. SNPs are usually bi-allelic, with two alleles observed in the population [16]. Multiallelic
SNPs have more than one alternative allele for that locus. Tri-allelic SNPs, which have three
nucleotide substitution-based alleles at the same position, are relatively rare but are being con-
sidered of great relevance in epidemiological studies [17], in disaster victim identification
using mixed and/or degraded DNA samples [18], and in animals pedigree accuracy studies
[19]. Tri-allelic SNPs are reported to have a higher power of discrimination than bi-allelic
SNPs requiring fewer markers and lowering costs [18, 20]. However, tri-allelic SNPs have been
excluded in cannabis population structural analysis in the current literature [6, 21].
Cannabis classification studies that employ SNPs generally used partial genome informa-
tion with few or no overlap sequences between datasets [22]. Whole-genome sequencing is
used less often in the literature, but is preferable despite its higher cost because it enables com-
parison of genome datasets from different sources [22]. It also provides comprehensive genetic
information [22], as studies showed that differences between fiber- and drug-type cannabis are
at a genome-wide level and not necessarily limited to genes involved in THC production [5].
The recent release of the 10-chromosome map of the cannabis genome [23–27] may improve
the understanding of the genetic architecture, identify a superior set of SNPs associated with
interesting traits, and reduce future targeted genotyping costs by using fewer but more accu-
rate SNPs [28].
Several approaches are now available for the analysis of population genetic structure. One
of these approaches is the DAPC, which is a multivariate clustering method that combines the
merits of both principal component analysis (PCA) and discriminant analysis (DA) [7, 29–31].
PCA is a multivariate analysis that can be applied to large datasets to reduce dimensions, but
does not provide a group assessment, which is essential for investigating genetic structures of
biological populations [32]. DA achieves the best classification of individuals into pre-defined
groups by maximizing between-group variation and minimizing within-group variation, but
the number of variables (alleles) needs to be fewer than the number of observations (individu-
als), which is generally not the case for SNP data [29]. DAPC first uses PCA to transform raw
data (genome-wide identified SNPs) into principal components (PC), which are mutually
orthogonal linear combinations of the original variables. This ensures that variables submitted
to DA are perfectly uncorrelated and that there are fewer variables than number of individuals.
PLOS ONE | https://doi.org/10.1371/journal.pone.0253387 June 28, 2021
2 / 14
PLOS ONE Genome-wide single nucleotide polymorphisms to classify cannabis strains
Then, linear discriminant functions, which are synthetic variables of linear combinations of
these SNPs, are constructed to maximize inter-cluster differences and minimize intra-cluster
variation [29]. By combining the advantages of PCA and DA, DAPC can identify groups,
assign individuals to groups, visualize between-population differentiation, and identify indi-
vidual alleles that have contributed to population structuring.
The objectives of this study are to:
1. investigate whether modern cannabis strains can be classified and differentiated at the
whole-genome level, and
2. investigate the chromosomal location and putative functions of identified structural SNPs.
This study is a part of an integrated cannabis strain classification project utilizing genetic,
chemical, and morphological profiles, wherein plants were grown in a commercial greenhouse
under the same condition.
Materials and methods
DNA extraction and whole genome sequencing
This study included 23 commercially available cannabis strains, and the research was carried
out under a cannabis research license issued by Health Canada. Where possible, the reported
ancestry (“Sativa”, “Indica”, or “Sativa-dominant” and “Indica-dominant”) was obtained from
the licensed producer providing the strain or from an online strain database (https://www.
leafly.ca) (Table 1). Each strain was analyzed for chemical composition using methods estab-
lished in our previous study [33] and labelled as “THC dominant”, “balanced”, and “CBD
dominant”. DNA was extracted from 100 mg of fresh leaves for each strain using a Qiagen
DNeasy Plant Mini Kit (QIAGEN, Canada). DNA concentrations were determined using a
Qubit Fluorometer (Thermo Fisher Scientific, US). DNA integrity was tested by agarose gel
electrophoresis. Library construction and sequencing were performed by BGI (USA) using
DNBseq™ sequencing technology to a depth of 30x. DNBseq™ is a high-throughput sequencing
solution, where DNA is fragmented into 100–300 bp and made into DNA nanoballs (DNB™),
which are continuous DNA molecule with multiple head-to-tail copies of the same DNA frag-
ment by linear isothermal rolling-circle replication. They are loaded onto high-density
sequencing templates and sequenced by combinatorial probe-anchor synthesis (cPAS), where
fluorescently tagged nucleotides complete for addition to the growing chain. After the addition
of each nucleotide, high-resolution digital imaging is carried out where the DNB clusters are
excited by a light source and a characteristic fluorescent signal is emitted. Hundreds of and
thousands of clusters are sequenced in a massively parallel process. The emission wavelength,
along with the signal intensity, determines the base call and the number of the cycles deter-
mines the length of the read. Sequence reads were then aligned to the reference genome assem-
bly ASM23057v4 of a drug type strain Purple Kush (PK) in the NCBI BioProject database
under accession number PRJNA73819 [34] using Burrows-Wheeler Alignment (BWA) tool
[35]. New assignments of chromosomes numbers (1–10) were used as in ASM23057v5 [36].
The first step of SNP calling is marking duplications in BAM format files, and selected duplica-
tions are included in SNP calling by GATK (Genome Analysis Toolkit) (https://www.
broadinstitute.org/gatk/). Local realignment around inDels is performed to avoid the bias of
SNP calling, and the variation sites around inDel are identified as SNPs. A total of 235,334
SNPs was identified, including 225,046 bi-allelic and 10,288 multiallelic SNPs. After filtering
for SNPs with no missingness by locus and a minor allele frequency less than 10% using
PLOS ONE | https://doi.org/10.1371/journal.pone.0253387 June 28, 2021
3 / 14
PLOS ONE Table 1. Strain information of 23 strains and preassigned clusters by DAPC.
Strain number
Strain name
Chemotypes
Clusters (W-SNPs)
Clusters (I-SNPs)
"Sativa" or "Indica"
Genome-wide single nucleotide polymorphisms to classify cannabis strains
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
Lemon Garlic OG
Royal Medic
Blue Hawaiian
Kandy Kush
Special
NN
Dance World
Treat
High
CB7
33˚
Banana Cake
Bananium
Burmese Blueberry
Divine Banana
Granddaddy Purple
Lemon Love
Lemon Sorbet
MeatHead
Nanitro
Platinum Jelly Punch
SBSK2 (Lemon Thai)
Super sherbet
1-Balanced
2-Balanced
3-CBD
4-CBD
5-CBD
6-CBD
7-Balanced
8-CBD
9-Balanced
10-CBD
11-THC
12-THC
13-THC
14-THC
15-THC
16-THC
17-THC
18-THC
19-THC
20-THC
21-THC
22-THC
23-THC
C1
C3
C3
C3
C3
C3
C3
C3
C3
C3
C1
C2
C3
C2
C2
C2
C1
C1
C2
C1
C1
C3
C1
C4
C2
C1
C1
C1
C1
C2
C1
C2
C1
C4
C5
C3
C5
C4
C5
C5
C4
C5
C4
C4
C3
C4
"Indica" dominant
"Sativa" dominant
"Sativa" dominant
"Sativa" dominant
Not provided
Not provided
"Sativa" dominant
Not provided
Not provided
Not provided
Not provided
"Indica" dominant
"Indica" dominant
"Indica" dominant
"Indica" dominant
"Indica" dominant
"Indica" dominant
"Indica" dominant
"Indica" dominant
"Indica" dominant
"Indica" dominant
50/50 hybrid
"Indica" dominant
�The column of clusters W-SNPs was obtained using the whole set of 137,858 filtered SNPs. The column of clusters I-SNPs was obtained using 344 structural SNPs.
https://doi.org/10.1371/journal.pone.0253387.t001
VCFtools, 137,858 SNPs, including 128,810 bi-allelic and 9,048 multiallelic SNPs, remained
for analysis.
Analysis of population structure and identification of structural SNPs
The population structure in this work was analyzed by DAPC using the adegenet package [37]
in R software [38]. First, the find.clusters function ran successive K-means [39] for a range of k
values (where the number of clusters k = K), and identified the optimal number of clusters by
comparing the Bayesian Information Criterion (BIC) [40] of the corresponding models. After
groups were assigned, a cross-validation function (xvalDapc) was used to determine the opti-
mal number of PCs to avoid over-sacrificing information or over-fitting in the subsequent
DAPC. In cross-validation, the data were divided into a training set (90% of the data) and a
validation set (10% of the data) by default. DAPC was carried out on the training set and the
accuracy of predicting the membership of individuals in the validation set was used to identify
the number of PCs. The sampling and DAPC were repeated 30 times by default at each level of
PC retention. After assigning individuals to clusters, DA was carried out on the retained PCs
and contributions of the alleles to each discriminant function were stored. An SNPZIP analysis
(snpzip) in R was then used to provide objective delineation between structural and non-struc-
tural SNPs, as identified by DAPC, to determine which SNPs contribute significantly to the
between-population structure [41].
PLOS ONE | https://doi.org/10.1371/journal.pone.0253387 June 28, 2021
4 / 14
PLOS ONE Genome-wide single nucleotide polymorphisms to classify cannabis strains
First, the whole set of 137,858 SNPs were applied to DAPC to identify SNPs that contrib-
uted most to the identified clusters. DAPC was carried out again using the identified SNPs to
validate their differentiation efficiency by confirming the separation of the 23 strains into their
preassigned clusters. A short sequence (about 600 nt) around each one of these identified SNP
was searched using the BLAST software (https://blast.ncbi.nlm.nih.gov) against Cannabis
sativa Annotation Release 100 [42]. In addition to DAPC, other clustering methods, including
PCA, neighbor-joining (NJ) tree [43], and hierarchical dendrogram using Ward’s minimum
variance method [44], were also employed to assess the robustness of the final inferred clusters.
PCA and NJ tree were plotted using R. The hierarchical dendrogram was plotted using JMP
14.0.0.
Results and discussions
Discriminant analysis of principal components using 137,858 SNPs
As indicated by the elbow in the curve of BIC values as a function of k in Fig 1(a), the optimal
number of identified clusters was three, corresponding to the lowest BIC values. The number
of PCs retained for DAPC analysis was four, as calculated by cross-validation in Fig 1(b),
where it had 100% predictive success, and 0% associated root mean squared error (RMSE). In
this study, the number of PCs associated with the highest mean success was also associated
with the lowest MSE, which made it easier to choose the number of PCs to retain. For the sub-
sequent DAPC analysis, four PCs and two discriminant functions were retained. The DAPC
Fig 1. DAPC for 23 cannabis genotypes. (a) The x-axis is the number of clusters k and the y-axis is the corresponding value of
BIC. (b) The plot of DAPC cross-validation. The x-axis is the number of PCA axes retained for DAPC, and the y-axis is the
proportion of successful outcome prediction. Individual replicates appear as points, and the density of those points in different
regions of the plot is displayed in blue. (c) DAPC plot for 23 cannabis genotypes along two linear discriminants (LD 1 and LD 2).
https://doi.org/10.1371/journal.pone.0253387.g001
PLOS ONE | https://doi.org/10.1371/journal.pone.0253387 June 28, 2021
5 / 14
PLOS ONE Genome-wide single nucleotide polymorphisms to classify cannabis strains
plot of 23 cannabis genotypes is shown in Fig 1(c). The grouping assignment for individual
strains by DAPC is listed in Table 1 (as W-SNPs). C1 is a THC dominant cluster and includes
six THC dominant strains (11, 17, 18, 20, 21, and 23-THC) and one balanced strain (1-bal-
anced). C2 is another THC dominant cluster and includes five THC dominant strains (12, 14,
15, 16, and 19-THC). C3 is a cluster dominated by CBD dominant and the balanced strains
which includes six CBD dominant strains (3, 4, 5, 6, 8, and 10-CBD), three balanced strains (2,
7, and 9-balanced), and two THC dominant strains (13 and 22-THC). While C2 is closer to C3
and is more distant to C1, C1and C3 are clearly separated along linear discriminant 1 (LD1).
While C1 and C3 are roughly at the same level with respect to linear discriminant 2 (LD2), C2
is separated from both. PCA was also carried out on the same set of SNPs and results are
shown in S1 Fig. Twenty-three cannabis strains are plotted along pair-wise PCs of the first 4
PCs, which account for 18.4%, 11.5%, 9.5%, and 8.7% of the total variance, respectively. Simi-
larly, the first PC suggests the existence of a relatively compact CBD & balanced clade on the
left side of the plot and a more dispersed THC dominant clade on the right side of the plot. Bal-
anced strains share a closer gene pool with CBD dominant strains, while the THC gene pool is
more dispersed. Because THC is psychoactive and its potency can be readily assessed through
consumption, selection for increasing THC content started early and widely for recreational
purposes by traditional breeding [45]. In contrast, CBD is non-psychoactive and must be ana-
lyzed in a laboratory for potency, and therefore breeding for high CBD concentrations began
later [45]. A complete genome assembly implied that CBD dominant varieties were generated
by integrating hemp-type CBD acid synthase gene clusters into a background of drug-type
cannabis to elevate CBDA production [24]. These balanced strains may have been created by
crossing purebred THC dominant types with CBD dominant types [46]. Therefore, there may
be a relatively limited selection of CBD dominant strains for breeding balanced strains.
Discriminant analysis of principal components using 344 structural SNPs
DAPC was repeated using identified 344 structural SNPs. The optimal number of identified
clusters was five, corresponding to the lowest BIC values (Fig 2(a)). Two PCs were retained for
the following DAPC analysis in Fig 2(b), where it had 98.9% predictive success and 0.04%
RMSE. For the subsequent DAPC analysis, two PCs and two discriminant functions were
retained. The grouping assignment for individual strains by DAPC is listed in Table 1 (as
I-SNPs). Within the five clusters (Fig 2(c)), C1 is a CBD dominant cluster that includes six
strains (3, 4, 5, 6, 8, and 10-CBD), C2 includes three balanced strains (2, 7, and 9-balanced),
and C3, C4, and C5 are THC dominant clusters that include two (13 and 22-THC), seven
(1-balanced, 11, 15, 18, 20, 21, 23-THC), and five (12, 14, 16, 17, and 19-THC) strains,
respectively.
These multiallelic SNPs were also subjected to PCA, NJ tree, and hierarchical clustering
analysis. In Fig 3, the 23 cannabis strains are plotted along PC1 and PC2, which account for
44.5% and 10.0% of the total variance, respectively. The proportions of explained variance are
higher compared to the previous PCA results (18.4% and 11.5%) obtained using the whole set
of SNPs. CBD dominant cluster C1 and balanced cluster C2 are on the left side of the scatter
plot (PC1<0) and the THC dominant clusters C3, C4, and C5 are on the right side of the scat-
ter plot (PC1>0). Notably, six CBD dominant strains are separated from three balanced
strains, while they were previously combined in the analysis using the whole set of SNPs. In
addition, two THC dominant strains 13-THC and 22-THC are separated from the CBD and
balanced cluster, and instead placed closer to other THC dominant strains. Strain 1-balanced
is closer to THC dominant strain regardless of whether the whole set of SNPs or 344 identified
SNPs were used.
PLOS ONE | https://doi.org/10.1371/journal.pone.0253387 June 28, 2021
6 / 14
PLOS ONE Genome-wide single nucleotide polymorphisms to classify cannabis strains
Fig 2. DAPC of 23 cannabis genotypes using 344 multiallelic structural SNPs. Clusters indicated as C1, C2, C3, C4, and C5 corresponds to the
I-SNPs in Table 1.
https://doi.org/10.1371/journal.pone.0253387.g002
The genetic structure from NJ-tree and hierarchical clustering using the 344 multiallelic are
displayed in Fig 4, mostly congruent with that of DAPC. In the NJ-tree, all six CBD dominant
strains are clustered together, with three balanced strains clustered closer on the same branch
(Fig 4(a)). Most THC dominant strains are also clustered adjacent to strains within their own
clusters. The dendrogram using hierarchical clustering by Ward’s method reveals two major
groups, where one group is comprised of CBD dominant & balanced strains, and the other of
THC dominant strains (Fig 4(b)). They are further separated into five subclusters, where CBD
dominant and balanced clusters are consistent with the DAPC grouping results, and several
THC dominant strains clustered differently. Two strains, 15-THC and 18-THC, were assigned
to C4 using DAPC but are assigned closer to C5 in the dendrogram. Two other strains,
14-THC and 16-THC, were assigned to C5 in DAPC but are assigned closer to C3 in the den-
drogram. The clustering results are congruent between DAPC and hierarchical clustering with
an assignment agreement rate of 83% (19/23).
Allele frequencies for 344 multiallelic SNPs in three chemotypes
DAPC identified 344 highly contributing SNPs (S1 Table). All the structural SNPs are multial-
lelic, among which 98.5% (339/344) are tri-allelic and the remainder 1.5% (5/344) are tetra-
allelic. The dendrogram of 23 strains using hierarchical clustering based on the allele counts in
the 344 structural SNPs (S2 Table) separated the strains into CBD dominant, balanced, and
THC dominant strains, mostly corresponding to the grouping results of DAPC (Fig 5). The
allele frequency was calculated by dividing the counts of that allele for all strains within the tar-
geted group by the sum of the counts for each allele for that SNP within the targeted group.
Allele frequencies of the structural SNPs were calculated for three major branches, each corre-
sponding one of three chemotypes. (S1 Table). If 1-balanced strain was assigned to the THC
PLOS ONE | https://doi.org/10.1371/journal.pone.0253387 June 28, 2021
7 / 14
PLOS ONE Genome-wide single nucleotide polymorphisms to classify cannabis strains
Fig 3. Scatter plot of 23 cannabis strains on PC1 & PC2 using 344 structural SNPs. Clusters indicated as C1, C2, C3,
C4 and C5 correspond to I-SNPs in Table 1.
https://doi.org/10.1371/journal.pone.0253387.g003
dominant group as indicated by DAPC for allele frequency calculation, there are 87% (300/
344) SNPs in CBD dominant clusters, 46% (157/344) SNPs in balanced clusters, and 11% (39/
344) SNPs in THC dominant clusters that have one allele with allele frequencies > 80% (S1
Table). Among them, 140 SNPs shared same alleles with allele frequencies > 80% in CBD
dominant strains (140/300) and balanced strains (140/157), which further indicated that CBD
dominant strains and balanced strains closely share a gene pool. There are 38 SNPs that have
one allele present in CBD dominant strains with allele frequencies > 80% and are not detected
in THC dominant strains. There are 322 SNPs whose alleles that are present in THC dominant
strains but were not detected in CBD dominant strains.
If the 1-balanced strain is assigned to the balanced group for allele frequency calculation,
there are 87% (300/344) SNPs in CBD dominant clusters, 10% (36/344) SNPs in balanced clus-
ters, and 13% (44/344) SNPs in THC dominant clusters that have one allele with allele
frequencies > 80% (S2 Table). Among them, 32 SNPs shared same alleles with allele
frequencies > 80% in CBD dominant strains (32/300) and balanced strains (32/36). There are
38 SNPs that have one allele present in CBD dominant strains with allele frequencies > 80%
and are not detected in THC dominant strains. There are 321 SNPs whose alleles are present
in THC dominant strains but were not detected in CBD dominant strains. Assigning the 1-bal-
anced strain to the balanced group added more genetic diversity to the balanced group, and
the effect of adding or deleting this strain for the THC dominant group in terms of allele fre-
quency is small and can be neglected.
PLOS ONE | https://doi.org/10.1371/journal.pone.0253387 June 28, 2021
8 / 14
PLOS ONE Genome-wide single nucleotide polymorphisms to classify cannabis strains
Fig 4. NJ-tree and hierarchical clustering using the 344 multiallelic SNPs (a) NJ-tree and (b) The dendrogram using hierarchical clustering by
Ward’s method for 23 cannabis genotypes. Clusters indicated as C1, C2, C3, C4, and C5 corresponds to I-SNPs in Table 1.
https://doi.org/10.1371/journal.pone.0253387.g004
Fig 5. Hierarchical clustering of 23 strains based on the allele counts for 344 structural SNPs identified by DAPC.
https://doi.org/10.1371/journal.pone.0253387.g005
PLOS ONE | https://doi.org/10.1371/journal.pone.0253387 June 28, 2021
9 / 14
PLOS ONE Genome-wide single nucleotide polymorphisms to classify cannabis strains
Fig 6. Features of 344 multiallelic SNPs (a) Distribution of structural SNPs on chromosome 1–10 and unplaced
scaffolds. (b) BLAST results for structural SNPs against a fully annotated genome.
https://doi.org/10.1371/journal.pone.0253387.g006
BLAST analysis of 344 multiallelic SNPs
These 344 SNPs were spread across all 10 chromosomes (Fig 6(a)), indicating that commer-
cially available cannabis strains in North America are significantly differentiated at a genome-
wide level. The number of identified SNPs ranged from 7 to 127 on each genome, with 37% of
the genetic variation occurring (127 SNPs) on chromosome 6, where CBDAS and THCAS are
located [13]. The rest SNPs were spread over the remaining nine chromosomes. All ten chro-
mosomes have genes related to the biochemical pathways of secondary metabolites, including
cannabinoids, monoterpenes, and sesquiterpenes [13, 24, 47–51]. BLAST results showed that
90% (310/344) of these structural SNPs had no feature, 7% (24/344) are uncharacterized loci
with unknown functions, and 3% (10/344) are predicted for certain functions (Fig 6(b)).
Conclusions
Although the cannabis industry is rapidly advancing after the relaxation of legal restrictions in
North America, the increasing number of THC dominant strains, CBD dominant strains, and
balanced strains only adds confusion to the currently poorly understood genetic background
of the thousands of varieties already in existence. Although there were only 23 strains included
in this study, they covered the three typical chemotypes of cannabis strains currently available
in the market. Leveraging as much genetic variation as possible using whole-genome sequenc-
ing, we identified 344 multiallelic SNPs that were used to investigate the genetic structure of 23
cannabis genotypes using DAPC, PCA, NJ tree, and hierarchical clustering, which provided
consistent observations and groupings despite the differences in algorithms. The clustering
results revealed that these 23 strains could be separated into five clusters, with one cluster
PLOS ONE | https://doi.org/10.1371/journal.pone.0253387 June 28, 2021
10 / 14
PLOS ONE Genome-wide single nucleotide polymorphisms to classify cannabis strains
containing six CBD dominant strains, another cluster containing three balanced strains, and
the remaining three clusters containing 13 THC dominant strains and one balanced strain.
CBD dominant strains and the balanced strains are closer genetically. This may be attributed
to how medical interest in breeding for non-psychoactive, CBD-elevated strains (CBD domi-
nant and balanced strains) has only recently been in vogue, resulting in an overlapping and
less diverse gene pool for CBD dominant and balanced strains compared to the longer breed-
ing history for THC strains. Some alleles are only present in CBD dominant strains or in THC
dominant strains. More alleles present in balanced strains are shared with CBD dominant
strains. One third of these structural SNPs are located on the chromosome containing THCAS
and CBDAS. The remaining SNPs are located on the other nine chromosomes. An area of
potential investigation is how the identified structural SNPs are associated with the production
of other cannabinoids, mono- and sesquiterpenes, flavonoids, other compounds, or morpho-
logical characteristics.
Since the late 20th century, genetic methodologies have been developed for separating
industrial hemp from drug-type cannabis for forensic purposes, thus differentiating CBD
dominant and THC dominant strains [52–56]. For the past 20 years, with the extensive hybrid-
ization of THC dominant strains, many classification studies have focused on separating
“Sativa” and “Indica” strains and many have suggested abolishing this vernacular [5–7]. The
genotyping results of this study indicate that modern, extensively hybridized strains can still be
separated using genome-wide information. As a powerful multivariate approach that investi-
gates population structures based solely on genetic information, DAPC separated strains into
clusters aligned with their chemotypes. Additionally, DAPC has the potential to sort the disor-
dered genetic background of thousands of THC dominant strains by identifying the number
of genetic clusters within THC dominant strains, describing clusters by interpreting group
memberships, and identifying the contributing SNPs that have the potential to be used as
genetic markers for strain classification and identification. This would require a concerted
effort from the cannabis industry by contributing whole genome sequence data to public data-
bases and by building a common taxonomy based on genomics. Optimally, the identified
genetic markers can be used as genomic fingerprints in combination with chemical finger-
prints and morphological characteristics for strain identification. These markers can be lever-
aged for strain selection in clinical trials and for manufacturing cannabis-based products and
medicines.
Supporting information
S1 Fig. PCA of 23 strains using whole set of SNPs.
(PDF)
S1 Table. 344 multiallelic SNPs identified by DAPC.
(XLSX)
S2 Table. Allele counts for 344 structural SNPs identified by DAPC.
(XLSX)
Acknowledgments
The authors are grateful to licensed grower Emerald Flower Farm who provided commercial
greenhouse to cultivate cannabis. The authors are also grateful to Dr. Limin Wu for assisting
DNA extraction, Dr. Jie Zeng for assisting BLAST analysis, and Shengxi Jin for proofreading
the manuscript.
PLOS ONE | https://doi.org/10.1371/journal.pone.0253387 June 28, 2021
11 / 14
PLOS ONE Genome-wide single nucleotide polymorphisms to classify cannabis strains
Author Contributions
Conceptualization: Dan Jin.
Data curation: Dan Jin.
Formal analysis: Dan Jin, Philippe Henry.
Funding acquisition: Dan Jin, Jacqueline Shan.
Investigation: Dan Jin.
Methodology: Dan Jin, Philippe Henry.
Project administration: Dan Jin.
Resources: Dan Jin.
Software: Dan Jin, Philippe Henry.
Supervision: Jie Chen.
Validation: Dan Jin, Philippe Henry.
Visualization: Dan Jin, Philippe Henry.
Writing – original draft: Dan Jin.
Writing – review & editing: Philippe Henry, Jacqueline Shan, Jie Chen.
References
1. Hillig KW (2005) A systematic investigation of Cannabis, PhD thesis, Indiana University. PhD Thesis
2. McPartland JM (2017) Cannabis sativa and Cannabis indica versus “Sativa” and “Indica”. In: Cannabis
sativa L.-Botany and Biotechnology. Springer, pp 101–121
3. McPartland JM, Guy GW (2017) Models of Cannabis Taxonomy, Cultural Bias, and Conflicts between
Scientific and Vernacular Names. The Botanical Review 4:327–381
4. Knight G, Hansen S, Connor M, Poulsen H, McGovern C, Stacey J (2010) The results of an experimen-
tal indoor hydroponic Cannabis growing study, using the ‘Screen of Green’(ScrOG) method—Yield,
tetrahydrocannabinol (THC) and DNA analysis. Forensic science international 202:36–44 https://doi.
org/10.1016/j.forsciint.2010.04.022 PMID: 20462712
5. Sawler J, Stout JM, Gardner KM, Hudson D, Vidmar J, Butler L, et al. (2015) The genetic structure of
marijuana and hemp. PloS one 10:e0133292 https://doi.org/10.1371/journal.pone.0133292 PMID:
26308334
6.
Lynch RC, Vergara D, Tittes S, White K, Schwartz CJ, Gibbs MJ, et al. (2016) Genomic and chemical
diversity in Cannabis. Critical Reviews in Plant Sciences 35:349–363
7. Henry P (2015) Genome-wide analyses reveal clustering in Cannabis cultivars: the ancient domestica-
tion trilogy of a panacea. https://doi.org/10.7287/peerj.preprints.1553v2
8. McGuire P, Robson P, Cubala WJ, Vasile D, Morrison PD, Barron R, et al. (2018) Cannabidiol (CBD) as
an adjunctive therapy in schizophrenia: a multicenter randomized controlled trial. Am J Psychiatry
175:225–231 https://doi.org/10.1176/appi.ajp.2017.17030325 PMID: 29241357
9.
French J, Thiele E, Mazurkiewicz-Beldzinska M, Benbadis S, Marsh E, Joshi C, et al. (2017) Cannabi-
diol (CBD) significantly reduces drop seizure frequency in Lennox-Gastaut syndrome (LGS): results of
a multi-center, randomized, double-blind, placebo controlled trial (GWPCARE4)(S21.001). Neurology
88:S21–001
10. Avraham Y, Grigoriadis NC, Poutahidis T, Vorobiev’ L, Magen I, Ilan Y, et al. (2011) Cannabidiol
improves brain and liver function in a fulminant hepatic failure-induced model of hepatic encephalopathy
in mice. British journal of pharmacology 162:1650–1658 https://doi.org/10.1111/j.1476-5381.2010.
01179.x PMID: 21182490
11. Bloomfield MAP, Green SF, Hindocha C, et al (2020) The effects of acute cannabidiol on cerebral blood
flow and its relationship to memory: An arterial spin labelling magnetic resonance imaging study. Jour-
nal of psychopharmacology (Oxford, England) 269881120936419 https://doi.org/10.1177/
0269881120936419 PMID: 32762272
PLOS ONE | https://doi.org/10.1371/journal.pone.0253387 June 28, 2021
12 / 14
PLOS ONE Genome-wide single nucleotide polymorphisms to classify cannabis strains
12. Upton R, Craker L, ElSohly M, Romm A, Russo E, Sexton M (2014) Cannabis inflorescence: cannabis
spp.; standards of identity, analysis, and quality control. American Herbal Pharmacopoeia, Scotts Val-
ley, CA https://doi.org/10.4081/ijfs.2020.8581 PMID: 32913724
13. Kovalchuk I, Pellino M, Rigault P, et al (2020) The genomics of cannabis and its close relatives. Annu
Rev Plant Biol 71:713–739 https://doi.org/10.1146/annurev-arplant-081519-040203 PMID: 32155342
14. Sakamoto K, Akiyama Y, Fukui K, Kamada H, Satoh S (1998) Characterization; Genome Sizes and
Morphology of Sex Chromosomes in Hemp (Cannabis sativa L.). Cytologia 63:459–464
15. Shastry BS (2007) SNPs in disease gene mapping, medicinal drug development and evolution. Journal
of Human Genetics 52:871–880 https://doi.org/10.1007/s10038-007-0200-z PMID: 17928948
16. Sachidanandam R, Weissman D, Schmidt SC, et al (2001) A map of human genome sequence varia-
tion containing 1.42 million single nucleotide polymorphisms. Nature 409:928–933 https://doi.org/10.
1038/35057149 PMID: 11237013
17. Casci T (2010) SNPs that come in threes. Nature Reviews Genetics 11:8–8 https://doi.org/10.1038/
nrg2725 PMID: 20050277
18. Westen AA, Matai AS, Laros JFJ, Meiland HC, Jasper M, de Leeuw WJF, et al. (2009) Tri-allelic SNP
markers enable analysis of mixed and degraded DNA samples. Forensic Science International: Genet-
ics 3:233–241
19. Kalbfleisch T, Petersen JL, Tait RG Jr., Qiu J, Basnayake V, Hackett PH, et al. (2020) Using triallelic
SNPs for determining parentage in North American yak (Bos grunniens) and estimating cattle (B. tau-
rus) introgression. F1000Res 9:1096 https://doi.org/10.12688/f1000research.25803.2 PMID:
33163159
20. Phillips C, Amigo J, Tillmar AO, et al (2020) A compilation of tri-allelic SNPs from 1000 Genomes and
use of the most polymorphic loci for a large-scale human identification panel. Forensic Science Interna-
tional: Genetics 46:102232 https://doi.org/10.1016/j.fsigen.2020.102232 PMID: 31986343
21. Henry P, Khatodia S, Kapoor K, et al (2020) A single nucleotide polymorphism assay sheds light on the
extent and distribution of genetic diversity, population structure and functional basis of key traits in culti-
vated north American cannabis. Journal of Cannabis Research 2:26 https://doi.org/10.1186/s42238-
020-00036-y PMID: 33526123
22. Vergara D, Baker H, Clancy K, Keepers KG, Mendieta JP, Pauli CS, et al. (2016) Genetic and genomic
tools for Cannabis sativa. Critical Reviews in Plant Sciences 35:364–377
23.
Laverty KU, Stout JM, Sullivan MJ, et al (2019) A physical and genetic map of Cannabis sativa identifies
extensive rearrangements at the THC/CBD acid synthase loci. Genome Res 29:146–156 https://doi.
org/10.1101/gr.242594.118 PMID: 30409771
24. Grassa CJ, Wenger JP, Dabney C, Poplawski SG, Motley ST, Michael TP, et al. (2018) A complete
Cannabis chromosome assembly and adaptive admixture for elevated cannabidiol (CBD) content.
bioRxiv 458083
25.
26.
27.
Jenkins C, Orsburn B (2019) Constructing a Draft Map of the Cannabis Proteome. bioRxiv 577635
Jenkins C, Orsburn B (2019) The Cannabis Multi-Omics Draft Map Project. bioRxiv 753400
Jenkins C, Orsburn B (2019) The First Publicly Available Annotated Genome for Cannabis plants. bioR-
xiv 786186
28. Benevenuto J, Ferrão LFV, Amadeu RR, Munoz P (2019) How can a high-quality genome assembly
help plant breeders? Gigascience. https://doi.org/10.1093/gigascience/giz068 PMID: 31184361
29.
Jombart T, Devillard S, Balloux F (2010) Discriminant analysis of principal components: a new method
for the analysis of genetically structured populations. BMC Genetics 11:94 https://doi.org/10.1186/
1471-2156-11-94 PMID: 20950446
30. Henry P (2017) Cannabis chemovar classification: terpenes hyper-classes and targeted genetic mark-
ers for accurate discrimination of flavours and effects. PeerJ Preprints 5:e3307v1
31. Henry P (2018) The genetic basis of the human-cannabis relationship. bioRxiv 287938
32.
33.
Jin D, Jin S, Yu Y, Lee C, Chen J (2017) Classification of Cannabis Cultivars Marketed in Canada for
Medical Purposes by Quantification of Cannabinoids and Terpenes Using HPLC-DAD and GC-MS. J
Anal Bioanal Tech 8:2
Jin D, Dai K, Xie Z, Chen J (2020) Secondary Metabolites Profiled in Cannabis Inflorescences, Leaves,
Stem Barks, and Roots for Medicinal Purposes. Scientific Reports 10:3309 https://doi.org/10.1038/
s41598-020-60172-6 PMID: 32094454
34. ASM23057v4—Genome—Assembly—NCBI. https://www.ncbi.nlm.nih.gov/assembly/GCA_
000230575.4/.
35.
Li H, Durbin R (2010) Fast and accurate long-read alignment with Burrows–Wheeler transform. Bioinfor-
matics 26:589–595 https://doi.org/10.1093/bioinformatics/btp698 PMID: 20080505
PLOS ONE | https://doi.org/10.1371/journal.pone.0253387 June 28, 2021
13 / 14
PLOS ONE Genome-wide single nucleotide polymorphisms to classify cannabis strains
36. ASM23057v5—Genome—Assembly—NCBI. https://www.ncbi.nlm.nih.gov/assembly/GCA_
000230575.5.
37.
Jombart T (2008) adegenet: a R package for the multivariate analysis of genetic markers. Bioinformat-
ics 24:1403–1405 https://doi.org/10.1093/bioinformatics/btn129 PMID: 18397895
38. R Core Team (2018) R: A language and environment for statistical computing. R Foundation for Statis-
tical Computing, Vienna, Austria. https://www.R-project.org/.
39.
Liu N, Zhao H (2006) A non-parametric approach to population structure inference using multilocus
genotypes. Human Genomics 2:353 https://doi.org/10.1186/1479-7364-2-6-353 PMID: 16848973
40. Konishi S, Kitagawa G (2008) Information Criteria and Statistical Modeling. https://doi.org/10.1007/978-
0-387-71887-3
41. Deperi SI, Tagliotti ME, Bedogni MC, Manrique-Carpintero NC, Coombs J, Zhang R, et al. (2018) Dis-
criminant analysis of principal components and pedigree assessment of genetic diversity and population
structure in a tetraploid potato panel using SNPs. PLoS One 13:e0194398 https://doi.org/10.1371/
journal.pone.0194398 PMID: 29547652
42. Cannabis sativa Annotation Report. https://www.ncbi.nlm.nih.gov/genome/annotation_euk/Cannabis_
sativa/100/.
43. Saitou N, Nei M (1987) The neighbor-joining method: a new method for reconstructing phylogenetic
trees. Mol Biol Evol 4:406–425 https://doi.org/10.1093/oxfordjournals.molbev.a040454 PMID: 3447015
44. Ward JH Jr (1963) Hierarchical Grouping to Optimize an Objective Function. Journal of the American
Statistical Association 58:236–244
45. Clarke RC, Merlin MD (2016) Cannabis domestication, breeding history, present-day genetic diversity,
and future prospects. Critical Reviews in Plant Sciences 35:293–327
46.
de Meijer EPM, Bagatta M, Carboni A, Crucitti P, Moliterni VMC, Ranalli P, et al. (2003) The inheritance
of chemical phenotype in Cannabis sativa L. Genetics 163:335–346 PMID: 12586720
47. Stout JM, Boubakir Z, Ambrose SJ, Purves RW, Page JE (2012) The hexanoyl-CoA precursor for can-
nabinoid biosynthesis is formed by an acyl-activating enzyme in Cannabis sativa trichomes. The Plant
Journal. https://doi.org/10.1111/j.1365-313x.2012.04949.x PMID: 22353623
48. Gagne SJ, Stout JM, Liu E, Boubakir Z, Clark SM, Page JE (2012) Identification of olivetolic acid
cyclase from Cannabis sativa reveals a unique catalytic route to plant polyketides. Proceedings of the
National Academy of Sciences 109:12811–12816 https://doi.org/10.1073/pnas.1200330109 PMID:
22802619
49. Van Bakel H, Stout JM, Cote AG, Tallon CM, Sharpe AG, Hughes TR, et al. (2011) The draft genome
and transcriptome of Cannabis sativa. Genome biology 12:R102 https://doi.org/10.1186/gb-2011-12-
10-r102 PMID: 22014239
50.
Zager JJ, Lange I, Srividya N, Smith A, Lange BM (2019) Gene Networks Underlying Cannabinoid and
Terpenoid Accumulation in Cannabis. Plant Physiology 180:1877–1897 https://doi.org/10.1104/pp.18.
01506 PMID: 31138625
51. Booth JK, Page JE, Bohlmann J (2017) Terpene synthases from Cannabis sativa. PLOS ONE 12:
e0173911 https://doi.org/10.1371/journal.pone.0173911 PMID: 28355238
52. Piluzza G, Delogu G, Cabras A, Marceddu S, Bullitta S (2013) Differentiation between fiber and drug
types of hemp (Cannabis sativa L.) from a collection of wild and domesticated accessions. Genetic
resources and crop evolution 60:2331–2342
53. Datwyler SL, Weiblen GD (2006) Genetic variation in hemp and marijuana (Cannabis sativa L.) accord-
ing to amplified fragment length polymorphisms. J Forensic Sci 51:371–375 https://doi.org/10.1111/j.
1556-4029.2006.00061.x PMID: 16566773
54. Gilmore S, Peakall R, Robertson J (2003) Short tandem repeat (STR) DNA markers are hypervariable
and informative in Cannabis sativa: implications for forensic investigations. Forensic Sci Int 131:65–74
https://doi.org/10.1016/s0379-0738(02)00397-3 PMID: 12505473
55. Hakki EE, Kayis SA, Pinarkara E, Sag A (2007) Inter simple sequence repeats separate efficiently
hemp from marijuana (Cannabis sativa L.). Electronic Journal of Biotechnology 10:570–581
56. Hilyard A, Lewin S, Johnson S, Henry P, Orser C (2019) Application of a Simple Genetic Assay to
Discriminate Hemp from Drug-Type Cannabis. In: Cannabis Science Tech. https://www.
cannabissciencetech.com/view/application-simple-genetic-assay-discriminate-hemp-drug-type-
cannabis.
PLOS ONE | https://doi.org/10.1371/journal.pone.0253387 June 28, 2021
14 / 14
PLOS ONE |
10.1371_journal.pone.0269555 | RESEARCH ARTICLE
Grapevine trunk diseases of cold-hardy
varieties grown in Northern Midwest
vineyards coincide with canker fungi and
winter injury
David H. DeKreyID
1
1*, Annie E. Klodd2, Matthew D. Clark3, Robert A. BlanchetteID
1 Department of Plant Pathology, University of Minnesota, St. Paul, Minnesota, United States of America,
2 University of Minnesota Extension, Farmington, Minnesota, United States of America, 3 Department of
Horticultural Science, University of Minnesota, St. Paul, Minnesota, United States of America
* dekre004@umn.edu
Abstract
Grapevine trunk diseases make up a disease complex associated with several vascular fun-
gal pathogenic species. Surveys to characterize the composition of grapevine trunk dis-
eases have been conducted for most major grape growing regions of the world. This study
presents a similar survey characterizing the fungi associated with grapevine trunk diseases
of cold-hardy interspecific hybrid grape varieties grown nearly exclusively in the atypical
harsh winter climate of Northern Midwestern United states vineyards. From the 172 samples
collected in 2019, 640 isolates obtained by culturing were identified by ITS sequencing and
represent 420 sample-unique taxa. From the 420 representative taxa, opportunistic fungi of
the order Diaporthales including species of Cytospora and Diaporthe were most frequently
identified. Species of Phaeoacremonium, Paraconiothyrium, and Cadophora were also
prevalent. In other milder Mediterranean growing climates, species of Xylariales and Botryo-
sphaeriales are often frequently isolated but in this study they were isolated in small num-
bers. No Phaeomoniellales taxa were isolated. We discuss the possible compounding
effects of winter injury, the pathogens isolated, and management strategies. Additionally,
difficulties in researching and understanding the grapevine trunk disease complex are
discussed.
Introduction
Grapevine trunk diseases (GTDs) make up a disease complex most often associated with sev-
eral wood-inhabiting fungal species [1] and more recently possibly some bacterial species [2].
Sub-groups of these diseases are frequently categorized by symptomology and or taxonomic
designation of causal fungal agents. Common names given to GTDs include Esca [3], folletage
or berry shrivel [4], Petri disease, young esca, young vine decline [5], hoja de malvo´n [6],
Botryosphaeria dieback, bot canker, black goo [7], slow stroke [8], eutypiosis, Eutypa dieback
[9], black dead arm, dying arm, dead arm [10], swelling arm [11], grapevine leaf stripe disease
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: DeKrey DH, Klodd AE, Clark MD,
Blanchette RA (2022) Grapevine trunk diseases of
cold-hardy varieties grown in Northern Midwest
vineyards coincide with canker fungi and winter
injury. PLoS ONE 17(6): e0269555. https://doi.org/
10.1371/journal.pone.0269555
Editor: Hernaˆni Gero´s, Universidade do Minho,
PORTUGAL
Received: January 24, 2022
Accepted: May 23, 2022
Published: June 3, 2022
Copyright: © 2022 DeKrey et al. This is an open
access article distributed under the terms of the
Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: All relevant data are
within the paper and its Supporting Information
files.
Funding: MDC, RAB, AEK, DHD. This research was
funded by the Minnesota Agricultural Rapid
Response Fund and USDA Hatch Project MIN-22-
081 and MIN-22-089. The funders had no role in
study design, data collection and analysis, decision
to publish, or preparation of the manuscript.
Competing interests: The authors have declared
that no competing interests exist.
PLOS ONE | https://doi.org/10.1371/journal.pone.0269555 June 3, 2022
1 / 25
PLOS ONE Grapevine trunk diseases of the Northern Midwest
[12], Phomopsis dieback, black spot [13], black measles [14], and black foot disease [15]. These
diseases can be difficult to diagnose due to their sporadic symptom display and similarity of
external and internal symptoms. Such symptoms may include interveinal foliar chlorosis and
necrosis or tiger striping, generalized dieback, apoplexy or sudden death, gummosis, vascular
streaking, wedge- or V-shaped vascular discoloration, cankers, and wood decay (Fig 1) [16].
GTDs cause serious grapevine health and economic problems and can be found in all grape
growing regions of the world [17–19]. In the Northern Midwest United States (NMW), grow-
ers often struggle with unproductive cordon sections commonly referred to as “skips in the
cordon” or “blind wood” (Fig 1A and 1B). Cordon skips come at the cost of vineyard managers
with lower yields and require retraining new cordons. In the past, chemicals such as sodium
arsenate was used to control GTDs but health [20] and environmental concerns [21] have
eliminated its widespread use. Often the dramatic increase in the incidence of GTDs in the last
two decades is associated with the 2003 ban on sodium arsenate [22]. However, the increasing
incidence of GTDs in countries which have never used sodium arsenate points to other factors
being involved [22]. To date, very few chemical options are available to growers but recent
research in the use of biological control agents has shown some promise for controlling spe-
cific fungal GTD pathogens [23]. In most situations, development of best practices for GTD
management remains the best option for growers. Management strategies can include prac-
tices such as variety selection, rootstock selection, training system, pruning timing, double-
pruning, wound-protection, multi-trunking, trunk renewal, trunk surgery, debris removal,
tool sterilization, and other practices [24]. Management options of GTD pathogens tend to be
region specific with considerations to climate, weather, cultural practices, and varieties grown.
In the NMW, wine grape growing is a relatively new industry that is increasing at a consider-
able pace. According to the 2016 University of Minnesota Extension vineyards and grapes sta-
tus report, planted cold-hardy grapevine varieties increased from 5900 acres to 7580 acres
from 2011 to 2015 [25]. However, Tuck et al. also reported an average decrease in yield of 3.5
to 3.2 tons per acre from 2011 to 2015 which indicates a need for better-informed, variety and
region-specific GTD management practices. To accomplish this, it is important to identify the
GTDs responsible for the problems.
Traditional European Vitis vinifera cultivars are not often grown in the NMW due to diffi-
culties brought on by harsh winters and a short growing season. Instead, own-rooted cold-
hardy interspecific hybrid grape (CIHG) varieties are widely and often exclusively grown in
the region. The genetic contribution of the native riverbank grape (V. riparia) provide CIHG
varieties developed in Minnesota their cold-hardiness (rated down to -30˚C) [26, 27] and
some resistance to endemic diseases and insect pests like phylloxera [28, 29]. Over the past
four decades, the University of Minnesota has become a leader in the development of several
CIHG wine and table grape varieties. The varieties most produced in the region include Mar-
quette, Frontenac blanc, Frontenac, La Crescent, Petite Pearl, Brianna, and Frontenac Gris
[30].
As many NMW vineyards are now reaching a decade in age since their first vines were
planted, the characteristic cordon skip (Fig 1A and 1B) and dieback (Fig 1D) symptoms of
GTDs have begun to appear. In addition, the compounding effect that GTDs and winter injury
have on vines is becoming a major concern (Fig 1L). In many other parts of the world where
grapes are grown, surveys have been conducted to characterize the region-specific composi-
tion of GTD pathogens. In Europe and nearby Mediterranean countries where GTDs were
first reported, major causal agents include fungal species of the genera Eutypa, Diplodia,
Botryosphaeria, and Phaeomoniella [19]. Similar fungal species have also been identified as
major causal agents of GTDs in Australia, New Zealand [31], South Africa [32], China [33–
35], and Chile [36, 37] as well as southern US and west coast US [38, 39]. Species of
PLOS ONE | https://doi.org/10.1371/journal.pone.0269555 June 3, 2022
2 / 25
PLOS ONE Grapevine trunk diseases of the Northern Midwest
Fig 1. Symptoms of grapevine trunk diseases in Northern Midwest vineyards. Cankers were often associated with skips in the cordons but had
rarely wedge-shaped discoloration (A). Cankers more often had irregular shaped xylem reactions (B). Pycnidia were sometimes observed fruiting
from cankers (B) and bleached canes (C). Dieback symptoms are common and the result of successive skips starting from tips of cordons (D).
Pruning wounds were associated with minor (E), moderate (F), and severe (G) vascular streaking symptoms. Near completely healthy vascular
tissue observed in wild Vitis riparia vines (H). Infrequent shallow cracks (I), several shallow cracks (J), and deep cracks (K and L) were associated
with minor to moderate (I), moderate to severe (J), and severe (K) vascular symptoms. Winter injury often results in deep cracks on the trunk (L).
Observations included cankers (ck), skips in the cordons (sk), bleached canes (bc), dieback (db), pruning wounds (pw), shallow cracks (sc), deep
cracks (dc), black spotting (bs), black lines (bl), brown-red wood streaking (br), brown to black necrotic streaking (bn), discolored xylem (dx),
sometimes healthy tissue (h), and white rot (wr). Bars = 1 cm.
https://doi.org/10.1371/journal.pone.0269555.g001
Fomitiporia are often the main white-rot pathogen found in older vines in most of these
regions as well [19]. Species of Phaeoacremonium have been identified in grapevines and other
woody hosts in several countries around the world [40]. Species of Cadophora are on occasion
identified as well notably found in Canada [41]. Species of Diaporthe and Cytospora have also
PLOS ONE | https://doi.org/10.1371/journal.pone.0269555 June 3, 2022
3 / 25
PLOS ONE Grapevine trunk diseases of the Northern Midwest
been identified in most of these regions though often to a lesser extent and usually in more
humid growing regions [32].
However, no surveys have been conducted in the NMW or exclusively on CIHG cultivars.
The objective of this study was to identify the major GTD species throughout the grape-grow-
ing regions of Minnesota and Wisconsin. Three hypotheses were explored in this study. First,
NMW GTDs will have a regionally distinct composition compared to other grape-growing
regions of the world given the harsh growing climate and CIHG varieties grown. Second, die-
back symptoms and internal vascular streaking can be associated with pruning wounds and
winter injury. Third, isolation frequency of fungal genera will significantly differ compared to
sample variety, variety berry color, sample section type, and sample county origin.
Methods
Sample collection
Our sample collection was targeted towards symptomatic grapevines showing skips in the cor-
dons, generalized dieback, reduced productivity, vascular discoloration, vascular decay, or
apoplexy (Fig 1). A few externally asymptomatic vines were also collected for comparison. In
2019, a total of 172 samples were collected and brought to the laboratory. Samples were col-
lected throughout both the dormant and growing season of 2019. Some samples were shipped
by priority mail. Most samples collected were woody sections of grapevines, especially of cor-
dons and trunks. It is important to note that regular re-trunking is frequently practiced in
NMW vineyards and therefore main woody trunks of vines rarely, if ever, exceed ten years in
age. Samples were stored at -20˚C until processed. Samples were acquired from 34 vineyards
in Minnesota and Wisconsin from a total of 21 counties (Fig 2). However, data reported in this
study is down to the county level to conserve anonymity of contributing vineyards. Primarily
named CIHG varieties were collected as well as a few wild vines and genetically unique breed-
ing lines.
Sample processing
Large diameter vine samples were cut and 3–5 mm3 chips excised from the margins of discol-
ored or decayed internal vascular wood tissue or from the edge of cankers. For smaller diame-
ter vine samples, the bark was peeled off and 3–5 mm thick discs were cut. Some disks were
kept whole while others cut in half or in fourths depending on diameter of the sample. Excised
chips where surface sterilized for 30 sec in an aqueous 10% sodium hypochlorite solution, fol-
lowed by two washes in sterile distilled H2O, and one wash in 70% EtOH then left to dry in a
clean air cabinet prior to plating. Between 3–5 chips per plate were then semi-embedded into 3
different culturing medias including malt extract agar (MEA; 15 g of Difco Bacto-agar, 15 g of
Difco Bacto malt extract, and 1 L of deionized water with 0.1 g streptomycin sulphate dissolved
in a small amount of 95% EtOH added post-autoclaving once cooled to 50˚C), basidiomycete
semi-selective agar (BSA; same as MEA recipe plus 2 g of Difco yeast extract and 0.06 g Aldrich
benomyl dissolved in a small amount of 95% EtOH added pre-autoclaving with 2 mL 85% lac-
tic acid added post-autoclaving once cooled to 50˚C, adapted from Worrall, 1991) [42], and
sabouraud dextrose agar (SDA; same as MEA recipe with 0.1 g Aldrich cycloheximide dis-
solved in a small amount of deionized water added post-autoclaving once cooled to 50˚C,
adapted from Harrington, 1981) [43]. Plates were left to incubate at 20–23˚C in darkness and
checked daily. Emerging fungi were transferred onto fresh MEA. All cultures were maintained
and stored in plastic bins at 20–23˚C.
PLOS ONE | https://doi.org/10.1371/journal.pone.0269555 June 3, 2022
4 / 25
PLOS ONE Grapevine trunk diseases of the Northern Midwest
Fig 2. Map of counties sampled. Grapevine wood samples collected from Minnesota and Wisconsin counties traced in yellow included Blue
Earth (BE), Carver (C), Crow Wing (CW), Douglas (D), Fillmore (F), Goodhue (G), Jackson (J), Kanabec (K), Lac Qui Parle (LQP), Le Sueur
(LS), Meeker (Mk), Mower (Mw), Murray (Mr), Pine (Pn), Polk (Pl), Trempealeau (T), Vernon (V), Wabasha (Wb), Walworth (Wl),
Washington (W), and Winona (Wn). Color scale indicates elevation in meters. Pink points denote locations of the University of Minnesota
(UMN) St. Paul campus and the UMN Horticultural Research Center (HRC) where the grape breeding program is located and where several
samples were collected. Map constructed in R with the public domain map collections Natural Earth (https://www.naturalearthdata.com/) and
Terrain Tiles (https://registry.opendata.aws/terrain-tiles/).
https://doi.org/10.1371/journal.pone.0269555.g002
Isolate selection
Fungal isolates for each sample were selected by culture macro-morphology on MEA and
genetic identification by sequencing the internal transcribed spacer (ITS) genomic region. The
primary macro-morphological characteristics considered included isolate color, growth rate,
hyphal branching, hyphal depth, hyphal extension, hyphal margin, fruiting, sporulation, and
metabolite staining of media. At the start of this study all unidentified cultures with unique
morphologies isolated from a single sample were selected for sequencing. Isolates were later
selected by macro-morphology in a more targeted manner as the study progressed by choosing
unique cultures or cultures similar to known pathogens we previously identified in the isolate
collection. Any isolates with questionable, non-descript, or similar culture macro-morphology
were sequenced to be sure of their identity.
DNA extraction and amplification
The DNA of select isolates was extracted using the NaOH protocol according to Osmundson
et al. (2013) [44]. Hyphae were scraped using a sterile scalpel from cultures of select isolates on
MEA that had grown out larger than 2.5 cm in diameter. Hyphal tissue was transferred to a 1.7
mL microcentrifuge tube with 300 μL of 5 mM NaOH and 3 to 5 3.5 mm glass beads. The sam-
ples where then vortexed for 1 to 5 min and centrifuged for 30 sec at 10,000 rpm. Then 5 μL of
supernatant was transferred to new tubes containing 495 μL Tris-HCL 5 mM, pH 8.0.
PLOS ONE | https://doi.org/10.1371/journal.pone.0269555 June 3, 2022
5 / 25
PLOS ONE Grapevine trunk diseases of the Northern Midwest
The ITS region of the isolated DNA was targeted for PCR amplification using the ITS1F/4
primer pair [45] according to Blanchette et al. (2016) [46]. Each PCR had a final volume of
25.5 μL consisting of 12.5 μL GoTaq1 Green Master Mix, 9.5 μL molecular grade water, 1 μL
of each primer at 10 μM, and 0.5 μL bovine serum albumin. The ITS locus was amplified using
a Bio-Rad T100™ Thermal Cycler following a program of 94˚C for 5 min, 35 cycles of 94˚C for
1 min, 50˚C for 1 min, and 72˚C for 1 min, followed by a final extension step of 72˚C for 5
min. Locus amplification was confirmed by gel electrophoresis of SYBR stained PCR products
prior to sequencing. Crude PCR products were Sanger sequenced by ABI 3730xl DNA
sequences, Applied Biosystems, Foster City, CA.
Molecular identification
Sequences were processed using Geneious v9.0. The processed sequences where then identified
with the basic local alignment search tool algorithm program for nucleotide sequences
(BLASTn) initially against the TrunkDiseaseID.org [47] database and also against the standard
complete NCBI GenBank. Best sequence identity match was selected for by consideration of
highest score for published data as denoted in GenBank at the time of BLASTn searches. Iden-
tity of isolates were matched to published sequences from taxonomic studies and identified to
the species level whenever possible. Isolates with greater than 97% sequences identity match
were considered homologous. Pathogenicity of identified fungal species on grapevines were
denoted according initially to TrunkDiseaseID.org [47] and then confirmed and expanded by
an assortment of grapevine pathogenicity trials found in published literature. However, most
pathogenicity trials for these fungi were conducted on traditional V. vinifera grapevine culti-
vars. Samples were scored as GTD+ upon sequence confirmation of at least one known patho-
genic species. Additional isolation and sequencing was discontinued once a sample was
designated GTD+.
Data analysis
Data were analyzed using the R statistical programming language in the RStudio integrated
development environment using an assortment of packages but most notably the collection of
Tidyverse Packages (v1.3.0) [48], the iNEXT package (v2.0.20) [49] to analyze sample coverage,
and the vcd package (v1.4–9) for multivariant analysis. The vcd package was used to explore
potential differences in isolation frequencies of genera-level taxa compared to a few variables
of interest that are descriptive of the 168 collected woody samples. In brief, Hill numbers are
used in the iNEXT package to estimate and then visualized sample completeness [50]. Addi-
tionally, diversity Pearson residuals statistics were used to analyze the measure of discrepancy
between observed and expected values within the vcd package. For each statistical comparison,
a p-value is returned from a corresponding Chi-square test and a residual shaded mosaic plot
was produced. Mosaic plots are graphs used for visualizing the comparison of multi-categori-
cal data where both the x- and y-axis are sized proportionally to the input data, i.e. the sum
area of the blocks represent 100% of the data and individual blocks are size proportionally to
the frequency with which the categories are observed.
Results
Internal symptoms of GTDs following the terminology of Mugnai et al. 1999 [16] included
brown-red wood streaking in a clearly defined wedge-shape from the cambium to pith which
is indicative of canker fungi were observed in few of our samples. Cankers more often
occurred in irregular forms and were associated with skips in the cordons (Fig 1A and 1B).
Centrally diffuse brown-red wood as well as brown to black necrotic streaking originating
PLOS ONE | https://doi.org/10.1371/journal.pone.0269555 June 3, 2022
6 / 25
PLOS ONE Grapevine trunk diseases of the Northern Midwest
from the pith was often associated with pruning wounds (Fig 1E–1G) and cracks (Fig 1I–1L).
All samples collected had discolored xylem to some extent and nearly completely healthy vas-
cular tissues were only observed in some cross-sections of wild riverbank grapevines in forest
and urban environments not included in this survey (Fig 1H). Cross-sections near pruning
spurs often showed discolored wood symptoms without being preceded by diffuse brown-red
wood or brown to black streaking (Fig 1E). Concentric black spotting, the result of longitudi-
nal streaking, was also frequently observed, and in some cases, black spotting would begin to
coalesce into shorter black lines (Fig 1F). Most samples had severe mottled expression of vas-
cular symptoms especially for brown-red wood streaking, brown to black necrotic streaking,
and discolored xylem (Fig 1G). Severe symptoms were sometimes associated with several
points of origin from shallow cracking from winter injury or hail damage (Fig 1J). Rarely, if
ever, have GTD foliar symptoms been observed in the NMW which possibly may be the result
of our overall young vineyards or different climate. Foliar symptoms are more often observed
in older vines under particular seasonal conditions [51, 52]. Moreover, lack of foliar symptoms
may also be the result of regular re-trunking, a common cultural practice in the NMW. It is
uncommon in the NMW for grapevine trunk wood to exceed ten years in age.
All wood samples collected had some degree of internal vascular symptoms including exter-
nally healthy samples (Fig 1). From 172 samples with various symptoms that included cankers
and vascular discoloration, dieback as well as pruning wounds and cracks from cold injury or
other environmental stresses yielded 640 isolates. These isolates represented 420 species-level
taxa unique to individual samples. Rarefication using the 420 representative taxa estimate a
sample coverage of 83% that reached to 90% by doubling the number of representative taxa
(Fig 3).
We found 32 of the 34 sampled vineyard locations, 20 of 21 counties, in this survey to have
at least one GTD+ sample. Of the 172 samples we collected, 142 (83%) had taxa reported as
pathogens associated with GTDs. Most samples were cordon sections and of the Marquette
variety (Tables 1 and 2). Most taxa are of the phylum Ascomycota (398 isolates, 94.76%) which
encompass 19 different orders, 38 genera, and potentially 89 species (Table 3).
The most frequently isolated genera obtained in this study that were known to be associated
with GTDs from previous reports included Cytospora, Phaeoacremonium, Diaporthe,
Fig 3. Rarefaction sample coverage curve. Observed sample coverage reaches 83% for the 420 sample representative
taxa. Extrapolated sample coverage reaches 90% by doubling the number of sample representative taxa.
https://doi.org/10.1371/journal.pone.0269555.g003
PLOS ONE | https://doi.org/10.1371/journal.pone.0269555 June 3, 2022
7 / 25
PLOS ONE Grapevine trunk diseases of the Northern Midwest
Table 1. Type of samples collected from the Northern Midwest in 2019. The majority of the 172 samples included
woody sections of cordon, trunk, root, sucker, shoots, and unknown.
Type
cordon
trunk
root
sucker
shoot
unknown
slime flux
bark
basidiocarp
Sample no.
113
32
11
5
4
3
2
1
1
Percent
66%
19%
6%
3%
2%
2%
1%
1%
1%
https://doi.org/10.1371/journal.pone.0269555.t001
Cadophora, Pestalotiopsis, Diatrypella, Diplodia, and Botryosphaeria, respectively (Fig 4). Of
these genera the most frequent species level sequence matches associated with GTDs included
Cy. viticola, Ph. fraxinopennsylvanicum, Ph. minimum, Dpr. ampelina, Cd. luteo-olivacea, Ps.
neglecta, Dt. verruciformis, Dpl. seriata, and Bt. dothidea (Table 3).
Table 2. Varieties of the 172 samples collected in the Northern Midwest in 2019.
Variety
Marquette
La Crescent
Frontenac
St. Pepin
Frontenac Blanc
Brianna
Frontenac Gris
Edelweiss
Itasca
Marechal Foch
unknown
Petite Pearl
Prairie Star
Valiant
MN1069
MN1016
Sabrevois
slime flux
St. Croix
MN43765
basidocarp
Millot
MN1005
Osceola Muscat
Riverbank Grape
Sauvignon
Virginia Creeper
Sample no.
42
28
20
Percent
24%
16%
12%
9
8
7
7
6
6
6
6
3
3
3
2
2
2
2
2
1
1
1
1
1
1
1
1
5%
5%
4%
4%
3%
3%
3%
3%
2%
2%
2%
1%
1%
1%
1%
1%
1%
1%
1%
1%
1%
1%
1%
1%
https://doi.org/10.1371/journal.pone.0269555.t002
PLOS ONE | https://doi.org/10.1371/journal.pone.0269555 June 3, 2022
8 / 25
PLOS ONE Table 3. Taxonomy, isolation frequency, and pathogenicity of fungi identified by ITS with greater than 97% homology match.
Grapevine trunk diseases of the Northern Midwest
Family (n)
Genus (n)
Isolate Host variety
Origin
GenBank
Phylum (n)
Order (n)
Ascomycota (398)
Diaporthales (98)
Valsaceae (61)
Diaporthaceae (37)
Togniniales (40)
Togniniaceae (40)
Pleosporales (99)
Didymosphaeriaceae (31)
Didymellaceae (28)
Pleosporaceae (26)
Astrosphaeriellaceae (3)
Coniothyriaceae (3)
Phaeosphaeriaceae (3)
undefined family (3)
Cucurbitariaceae (1)
Species (n) [pathogenicity studies]
Cytospora (60)
Cy. viticola (57) [53]
Cy. piceae (2)
Cy. mali (1)
GI-174
Frontenac Gris
Blue Earth, MN
OM307727
GI-847 Virginia Creeper
Carver, MN
GI-89 Marquette
Meeker, MN
OM307728
OM307729
Frontenac
Crow Wing, MN
OM307730
Frontenac Gris
Douglas, MN
OM307731
GI-75
GI-31
GI-384
GI-856
GI-413
GI-347
GI-741
GI-212
Valsa (3)
Vl. sordida (2)
Vl. salicina (1)
Diaporthe (38)
Dpr. ampelina (31) [54–56]
Dpr. eres (7) [54]
Phaeoacremonium (41)
Ph. fraxinopennsylvanicum (22) [41]
Ph. minimum (15) [3, 38]
Ph. amstelodamense (1)
Ph. angustius (1) [57]
Ph. canadense (1) [41]
Ph. hungaricum (1)
Paraconiothyrium brasiliense (30) [58, 59] GI-516
Paraphaeosphaeria sporulosa (1)
GI-157
Didymella (16)
Dd. pinodella (11)
Dd. glomerata (2)
Dd. pomorum (2)
Dd. bellidis (1)
Epicoccum (10)
Ep. nigrum (9)
Ep. sorghinum (1)
Nothophoma spiraeae (2)
Alternaria (26)
Al. alternata (19)
Al. tenuissima (5)
Al. arborescens (1)
Al. infectoria (1)
Pithomyces chartarum (3)
Coniothyrium palmicola (3)
GI-162
GI-885
GI-879
GI-190
GI-389
GI-837
GI-882
GI-878
Frontenac
Carver, MN
GI-236 Marquette
GI-422 Marquette
St. Pepin
Edelweiss
Goodhue, MN
Goodhue, MN
Douglas, MN
Goodhue, MN
La Crescent
Carver, MN
St. Pepin
Pine, MN
Osceola Muscat Wabasha, MN
Frontenac
Edelweiss
Douglas, MN
Wabasha, MN
GI-257 Marquette
GI-116 MN43765
GI-787 MN1016
Wabasha, MN
Carver, MN
Carver, MN
La Crescent
Walworth, WI
La Crescent
Carver, MN
GI-223 Marquette
Blue Earth, MN
Riverbank Grape Carver, MN
La Crescent
La Crescent
Frontenac
GI-829 Marquette
Carver, MN
Blue Earth, MN
Wabasha, MN
Blue Earth, MN
Frontenac Gris
Blue Earth, MN
La Crescent
Blue Earth, MN
OM307732
OM307733
OM307734
OM307735
OM307736
OM307737
OM307738
OM307739
OM307740
OM307741
OM307742
OM307743
OM307744
OM307745
OM307746
OM307747
OM307748
OM307749
OM307750
OM307751
OM307752
OM307753
OM307754
OM307755
OM307756
Sclerostagonospora (2)
Sc. cycadis (1)
Sc. lathyri (1)
Neosetophoma cerealis (1)
Microsphaeropsis olivacea (3) [60]
Neocucurbitaria quercina (1)
GI-141 Marquette
Meeker, MN
GI-140 Marquette
Blue Earth, MN
GI-480
Frontenac Blanc
Crow Wing, MN
OM307757
GI-505 Marquette
Lac Qui Parle, MN
OM307758
GI-37 Marquette
Meeker, MN
OM307759
(Continued )
PLOS ONE | https://doi.org/10.1371/journal.pone.0269555 June 3, 2022
9 / 25
PLOS ONE Grapevine trunk diseases of the Northern Midwest
Family (n)
Genus (n)
Isolate Host variety
Origin
GenBank
Table 3. (Continued)
Phylum (n)
Order (n)
Hypocreales (49)
Nectriaceae (29)
Bionectriaceae (17)
Hypocreaceae (2)
Xylariales (25)
Sporocadaceae (11)
Diatrypaceae (9)
Hypoxylaceae (2)
Apiosporaceae (1)
Xylariaceae (2)
Helotiales (21)
undefined family (18)
Dermateaceae (1)
Porodiplodiaceae (1)
Sclerotiniaceae (1)
Dothideales (14)
Saccotheciaceae (14)
Botryosphaeriales (10)
Botryosphaeriaceae (10)
Species (n) [pathogenicity studies]
Fusarium (26)
Fs. acuminatum (5)
Fs. equiseti (3)
Fs. solani (3)
Fs. culmorum (1)
Ilyonectria liriodendri (1) [61–63]
Scolecofusarium ciliatum (1)
Thyronectria austroamericana (1)
Clonostachys (18)
Cln. rosea (16) [60]
Cln. byssicola (1)
Trichoderma (2)
Trc. atroviride (1)
Trc. deliquescens (1)
Pestalotiopsis (10)
Ps. neglecta (6)
Ps. uvicola (2) [38, 56]
Ps. brassicae (1)
Ps. chamaeropis (1)
Neopestalotiopsis mesopotamica (1)
Seimatosporium lichenicola (1)
Seiridium rosarum (1)
Diatrypella
Dt. verruciformis (8) [64, 65]
Dt. pulvinata (1)
Diatrype stigma (1) [65]
Hypomontagnella submonticulosa (1)
Hypoxylon invadens (1)
Arthrinium arundinis (1)
Rosellinia corticium (1)
Cadophora (16)
Cd. luteo-olivacea (13) [41, 66–68]
Cd. melinii (2) [68]
Cd. ferruginea (1)
Discohainesia oenotherae (1)
Porodiplodia vitis (1)
Botrytis cinerea (1)
Aureobasidium pullulans (14)
Diplodia (7)
Dpl. seriata (6) [38, 56, 58, 64, 69]
Dpl. corticola (1) [38, 56]
Botryosphaeria dothidea (2) [38, 56, 70]
Phaeobotryon negundinis (1)
GI-820
Frontenac
GI-95 MN1005
Fillmore, MN
Carver, MN
OM307760
OM307761
GI-376
Frontenac
Crow Wing, MN
OM307762
GI-151
La Crescent
Carver, MN
GI-322 Marechal Foch
Goodhue, MN
GI-149
La Crescent
GI-796
slime flux
Carver, MN
Carver, MN
OM307763
OM307764
OM307765
OM307766
GI-870
La Crescent
Polk, MN
OM307767
GI-874 Marquette
Trempealeau, WI
OM307768
GI-795
slime flux
Carver, MN
GI-351
Frontenac Blanc
Fillmore, MN
OM307769
OM307770
GI-491
Edelweiss
Trempealeau, WI
OM307771
GI-738
Edelweiss
Trempealeau, WI
OM307772
GI-403 Marquette
Lac Qui Parle, MN
OM307773
GI-231
St. Pepin
Goodhue, MN
GI-220
Frontenac Gris
Blue Earth, MN
GI-99 Marquette
GI-352
La Crescent
Carver, MN
Murray, MN
OM307774
OM307775
OM307776
OM307777
GI-464 Marquette
GI-416 Valiant
GI-895 Valiant
Carver, MN
Douglas, MN
OM307778
OM307779
Crow Wing, MN
OM307780
GI-817
Frontenac Blanc
Blue Earth, MN
GI-350
La Crescent
Blue Earth, MN
GI-70
GI-62
Frontenac
St. Croix
Carver, MN
Wabasha, MN
GI-370
La Crescent
Mower, MN
GI-316 Marechal Foch
Goodhue, MN
GI-328
La Crescent
Goodhue, MN
GI-442
Frontenac Gris
Winona, MN
GI-269
Frontenac
Blue Earth, MN
GI-386
Frontenac
Walworth, WI
GI-886
La Crescent
Blue Earth, MN
OM307781
OM307782
OM307783
OM307784
OM307785
OM307786
OM307787
OM307788
OM307789
OM307790
OM307791
GI-408
La Crescent
Blue Earth, MN
OM307792
GI-373
Petite Pearl
Crow Wing, MN
OM307793
GI-225 Marquette
Blue Earth, MN
GI-131 MN1005
Carver, MN
OM307794
OM307795
(Continued )
PLOS ONE | https://doi.org/10.1371/journal.pone.0269555 June 3, 2022
10 / 25
PLOS ONE Table 3. (Continued)
Phylum (n)
Order (n)
Family (n)
Genus (n)
Isolate Host variety
Origin
GenBank
Species (n) [pathogenicity studies]
Grapevine trunk diseases of the Northern Midwest
Cladosporiales (11)
Cladosporiaceae (11)
Eurotiales (8)
Aspergillaceae (4)
Trichocomaceae (4)
Sordariales (6)
Chaetomiaceae (4)
Sordariaceae (1)
Trichosphaeriales (3)
Trichosphaeriaceae (3)
Chaetomellales (1)
Chaetomellaceae (1)
Chaetothyriales (1)
Herpotrichiellaceae (1)
Coniochaetales (1)
Coniochaetaceae (1)
Glomerellales (1)
Glomerellaceae (1)
Saccharomycetales (1) Dipodascaceae (1)
Thelebolales (1)
Thelebolaceae (1)
Valsariales (1)
Valsariaceae (1)
Basidiomycota (15)
Polyporales (9)
Phanerochaetaceae (3)
Irpicaceae (2)
Polyporaceae (2)
Cerrenaceae (1)
Meruliaceae (1)
Russulales (2)
Peniophoraceae (1)
Stereaceae (1)
Agaricales (2)
Physalacriaceae (1)
Schizophyllaceae (1)
Hymenochaetales (1)
Hymenochaetaceae (1)
Cystofilobasidales (1) Mrakiaceae (1)
Mucoromycota (7)
Mucorales (7)
Mucoraceae (7)
Cladosporium (11)
Cld. cladosporioides (9)
Cld. anthropophilum (1)
Cld. westerdijkiae (1)
Talaromyces amestolkiae (4)
Penicillium (5)
Pn. pulvillorum (2)
Pn. raistrickii (1)
Pn. simplicissimum (1)
Pn. sumatraense (1)
Ovatospora (3)
Ov. brasiliensis (2)
Ov. mollicella (1)
Chaetomium concavisporum (1)
Sordaria fimicola (1)
Nigrospora oryzae (3)
Chaetomella raphigera (1)
Rhinocladiella quercus (1)
Coniochaeta velutina (1)
Colletotrichum acutatum (1)
Geotrichum candidum (1)
Thelebolus microsporus (1)
Valsaria spartii (1)
Bjerkandera adusta (2)
Hyphodermella rosae (1)
Irpex lacteus (2)
Trametes versicolor (2)
Cerrena unicolor (1)
Phlebia radiata (1)
Peniophora cinerea (1)
Stereum complicatum (1)
Cylindrobasidium laeve (1)
Chondrostereum purpureum (1)
Phellinus conchatus (1)
Tausonia pullulans (1)
Mucor (7)
Mucr. circinelloides (6)
Mucr. moelleri (1)
GI-866 Marquette
GI-209 Marquette
GI-194 Marquette
GI-801 Vitis spp.
Blue Earth, MN
Blue Earth, MN
Blue Earth, MN
Carver, MN
GI-337 Marquette
Blue Earth, MN
GI-309
St. Pepin
Goodhue, MN
GI-210 Marquette
Blue Earth, MN
GI-36 Marquette
Meeker, MN
GI-828 Marquette
Blue Earth, MN
GI-881
La Crescent
Blue Earth, MN
GI-865 Marquette
Blue Earth, MN
GI-848 Virginia Creeper
Carver, MN
GI-846
Edelweiss
GI-217 Marquette
Winona, MN
Goodhue, MN
GI-486 Marquette
Blue Earth, MN
GI-472
Brianna
GI-287
La Crescent
GI-164
Edelweiss
Winona, MN
Goodhue, MN
Goodhue, MN
GI-88
Frontenac Blanc
Pine, MN
GI-43 Marquette
Meeker, MN
OM307796
OM307797
OM307798
OM307799
OM307800
OM307801
OM307802
OM307803
OM307804
OM307805
OM307806
OM307807
OM307808
OM307809
OM307810
OM307811
OM307812
OM307813
OM307814
OM307815
GI-417 Marquette
Trempealeau, WI
OM307816
GI-823
Frontenac
GI-806 Vitis spp.
GI-68
La Crescent
GI-198
Prairie Star
GI-798 Vitis spp.
GI-342
La Crescent
GI-200 Marquette
GI-263
Itasca
GI-444
Itasca
GI-805 Vitis spp.
slime flux
GI-61
Fillmore, MN
Carver, MN
Wabasha, MN
Carver, MN
Carver, MN
Blue Earth, MN
Goodhue, MN
Wabasha, MN
Le Sueur, MN
Carver, MN
Carver, MN
OM307817
OM307818
OM307819
OM307820
OM307821
OM307822
OM307823
OM307824
OM307825
OM307826
OM307827
GI-71
La Crescent
GI-365 Marquette
Vernon, WI
Mower, MN
OM307828
OM307829
Taxonomic rankings from order to species are denoted followed by isolation frequency in parenthesis. The isolation frequency is the count of samples each taxa was
isolated from a possible 172 samples. Representative isolates deposited to GenBank are listed for each species along with the sample variety and county origin of that
isolate. Pathogenicity studies conducted for each species are listed in brackets following species. See references for complete citations. Highlighted species have
associated pathogenicity studies. Highlighted isolates pictured in Fig 4.
https://doi.org/10.1371/journal.pone.0269555.t003
PLOS ONE | https://doi.org/10.1371/journal.pone.0269555 June 3, 2022
11 / 25
PLOS ONE Grapevine trunk diseases of the Northern Midwest
Fig 4. Fungal genera isolated from more than one grapevine sample. Individual culture image areas are relatively proportional to the number of samples
each fungal genus was isolated from. All pictured fungal isolates grown on MEA. Cytospora (Cy = 60, GI-174, 80 dpi); Phaeoacremonium (Ph = 41, GI-422, 94
dpi); Diaporthe (Dpr = 38, GI-449, 47 dpi); Paraconiothyrium (Pr = 30, GI-516, 47 dpi); Alternaria (Al = 27, GI-879, 36 dpi); Fusarium (Fs = 27, GI-151, 17 dpi);
Clonostachys (Cln = 18, GI-870, 27 dpi); Cadophora (Cd = 16, GI-370, 94 dpi); Didymella (Dd = 16, GI-257, 21 dpi); Aureobasidium (Ar = 14, GI-886, 21 dpi);
Cladosporium (Cld = 11, GI-866, 21 dpi); Epicoccum (Ep = 10, GI-837, 19 dpi); Pestalotiopsis (Ps = 10, GI-491, 40 dpi); Diatrypella (Dt = 9, GI-464, 15 dpi);
Diplodia (Dpl = 7, GI-408, 20 dpi); Mucor (Mucr = 7, GI-71, 20 dpi); Talaromyces (Tl = 6, GI-801, 55 dpi); Penicillium (Pn = 5, GI-337, 14 dpi); Coniothyrium
(Cn = 3, GI-882, 21 dpi); Microsphaeropsis (Mcrs = 3, GI-505, 20 dpi); Nigrospora (Ng = 3, GI-846, 14 dpi); Ovatospora (Ov = 3, GI-828, 19 dpi); Pithomyces
(Pt = 3, GI-162, 28 dpi); Valsa (Vl = 3, GI-384, 35 dpi); Bjerkandera (Bj = 2, GI-417, 28 dpi); Botryosphaeria (Bt = 2, GI-225, 35 dpi); Irpex (Ir = 2, GI-806, 55
dpi); Nothophoma (Nt = 2, GI-878, 31 dpi); Sclerostagonospora (Sc = 2, GI-140, 17 dpi); Trametes (Trm = 2, GI-68, 9 dpi); and Trichoderma (Trc = 2, GI-795, 14
dpi).
https://doi.org/10.1371/journal.pone.0269555.g004
There were 15 taxa of Basidiomycota (3.57%) which encompass 5 orders, 11 genera, and 12
species. Bjerkandera adusta, Irpex lacteus, and Trametes versicolor where the most frequently
identified Basidiomycota isolated. These fungi were present in 5 samples but were found in
PLOS ONE | https://doi.org/10.1371/journal.pone.0269555 June 3, 2022
12 / 25
PLOS ONE Grapevine trunk diseases of the Northern Midwest
counties not adjacent to one another. Very few taxa of Mucoromycota (7 isolates, 1.68%) were
identified. Mucor circinelloides, not considered associated with GTD, was isolated from 6 sam-
ples which originated from different vineyards in nonadjacent counties.
There were no significant differences in isolation frequencies of genera based on sample
berry color (S1 Fig). There is some indication of significant isolation frequency differences of
genera by sample variety (S2 Fig) or sample county origin (S3 Fig). For most genera not
enough sample representative taxa were obtained to be sure these isolation frequency differ-
ences between counties are truly significant. However, Cadophora obtained in this study nota-
bly had significant differences in isolation frequency between sample section types with a p-
value of 0.013. Cadophora spp. were less often isolated from cordons with a Pearson residual of
-2.9 and more frequently isolated from root, trunk, and sucker sections with Pearson residuals
between 2.2 to 3.9 (S4 Fig).
Discussion
The vast majority of the fungal taxa isolated in this study are of the phylum Ascomycota, sev-
eral of which are considered pathogenic to grapevines (references cited in Table 3). Cytospora
spp. and Diaporthe spp. of the order Diaporthales as well as Phaeoacremonium spp. make up
the largest majority of isolates identified and are known to be pathogenic on grapevines. Fre-
quent species identification of Phaeoacremonium in this study aligns similarly with most other
GTD surveys. However, the other major results confirm our first hypothesis that the composi-
tion of GTDs for the NMW is different in comparison to most other studied grape growing
regions. These fungi in the Diaporthales typically are considered a minor or secondary group
of causal agents in other regions were GTD surveys were completed [55, 71–74] but in the
study presented here they are the most commonly found GTD fungi in the NMW. The GTD
species genera of Diaporthe was previously named Phomopsis by some other investigators [55,
75]. Species of Cytospora have been reported previously as prevalent in cold climate regions
[74] and areas of high humidity [32].
The etiology of these Diaporthales pathogens on grapevines has been previously well char-
acterized [76]. In addition to symptoms of internal wood discoloration, pathogenic species of
Cytospora and Diaporthe can induce symptoms of cane bleaching (Fig 1C). They also produce
asexual fruiting bodies known as pycnidia which can be found on all affected tissues (Figs 1B,
1C and 4). On succulent green tissue, pycnidia may be surrounded by a halo or half halo of
chlorotic tissues or may reside hidden just under the bark of infected vines. Conidio-spores
that ooze out from pycnidia serve as a major source of inocula that can re-infect the same host
or infect other nearby hosts. Conidia are most often disseminated by rain or irrigation splash
but also by contaminated tools and more rarely by wind alone. These species overwinter in col-
onized wood of canes, spurs, pruning debris, and dormant buds [77]. However, symptoms
appear to differ from locations of sample collection and isolation particularly for Diaporthe
spp. [11, 55, 78, 79]. Symptom differences may be explained by genetic differences of local fun-
gal populations due to the result of horizontal gene transfer of transposable elements for the
acquisition or loss of pathogenicity [80, 81]. However, horizontal gene transfer has never been
studied in fungal GTD pathogens. Many of the Diaporthales are assumed opportunistic patho-
gens, causing disease only in stressed or weakened hosts or may live endophytically without
causing disease [82]. Since these fungi can colonize wounds, the prevalence of these fungi
found may be a result of wounds caused by cold injury.
In January of 2019, an atypical polar vortex occurred in the NMW. In Minnesota on Janu-
ary 30, 2019, the temperature dropped to -33˚C (-48˚C with wind chill) for the Minneapolis-
St. Paul area while the lowest temperature recorded in the state was -39˚C (-53˚C with wind
PLOS ONE | https://doi.org/10.1371/journal.pone.0269555 June 3, 2022
13 / 25
PLOS ONE Grapevine trunk diseases of the Northern Midwest
chill) [83]. The polar vortex temperatures were well below the lowest ratings for most of the
CHIG varieties in many counties. Winter injury was most apparent on Marquette variety
grapevines and in vineyards with little wind protection. Exposure of grapevines to these
extreme weather conditions resulted in frost cracks of woody tissues and damage to dormant
buds. However, winter injury of grapevines more typically occurs by means of sun exposure.
Injury occurs when both direct and snow reflected sunlight warms trunks to above freezing
during the day followed by a rapid decrease in temperature to below freezing at night. The sud-
den drop in temperature ruptures just the outer most layer of phloem cells for mild cases while
more serious cases kill cambial cells and damage xylem tissues. When this occurs on trees in
the NMW it is often referred to as sunscald. On grapevines this could be considered “winter
sunscald”, not to be confused with sunscald of grape berries in the summer. Like extreme cold
weather exposure, winter sunscald can also result in both shallow and deep frost cracks
depending on severity (Fig 1I–1L). Additionally, winter sunscald of grapevines results in a
blackened appearance of the bark on the south to southwest facing side of the vine (Fig 1L). In
either case of winter injury often the roots and the lower trunk of vines are protected by the
insulating snow covering. Thus, trunk replacement by sucker is a viable and common manage-
ment strategy in the NMW [84]. Regardless, associated observations of winter injury, vascular
discoloration, and identification of fungi suggests our second hypothesis is true and that the
polar vortex likely predisposed grapevines to these opportunistic canker pathogens.
Wounds, perhaps from winter injury or mechanical pruning, serve as portals for infection
by GTD pathogens under conducive weather conditions such as cool spring or fall rains [32,
85–87]. Fungal spores then colonize and spread through the vascular tissue either by hyphal
growth or additional sporulation. Many canker pathogens secrete cell wall degrading enzymes
or other compounds to spread laterally through xylem tissues eventually circumnavigating and
killing the entire cambium. However, the grapevine host produces tyloses, gels, phenolics, and
suberin to compartmentalize the damaged tissues and invading microorganisms [88]. How-
ever, restricted balanced production of these defensive structures and compounds is essential.
Overproduction of occlusions in response to pathogenic infections can lead to extensive
hydraulic failure resulting in external foliar symptoms and often vine death [89, 90]. In cross
sections the defense response of the grapevine host is seen as a continuum of brown-red wood
to brown-black necrotic tissue (Fig 1). Lighter vascular discoloration indicates more recent
responding tissues and likely the front of pathogen spread. Darker vascular discoloration indi-
cates long responding tissues and the probable point of pathogen entry [91]. Alternatively, pri-
marily pectinolytic active pathogens degrade gels in xylem vessels and spread longitudinally by
spores through the small spaces between tyloses partially occluding xylem conduits [92]. Lon-
gitudinal spread of these pathogens is seen in cross section by the host defense response as
black spotting and black lines (Fig 1F) [16]. Genomes of Cytospora spp. and Diaporthe spp.
reveal these fungi employ an abundance of cell wall degrading enzymes [93, 94].
Xylem vessel anatomy likely influences host resistance, pathogen spread, and environmen-
tal resilience. In the Dutch elm disease pathosystem, smaller diameter xylem vessels appeared
to confer some level of resistance to the causal fungal agents Ophiostoma ulmi and O. novo-
ulmi [95]. Reduced vessel diameter permits a more energetically conserved faster occlusion of
tissues adjacent to damaged or infected xylem tissues. Pouzoulet et al. (2014; 2017; 2020) con-
ducted histological and pathogenicity studies comparing a few V. vinifera grapevine cultivars
that had varying susceptibility to GTDs and showed cultivars with smaller diameter xylem ves-
sels may likely confer some resistance [88, 96, 97]. Unfortunately, histological pathogenicity
studies of grapevines against vascular pathogens are few and completely lacking for hybrid
varieties. However, hybrid varieties as well as traditional cultivars have been studied and show
links between vessel anatomy and environmental resilience against freezing and drought
PLOS ONE | https://doi.org/10.1371/journal.pone.0269555 June 3, 2022
14 / 25
PLOS ONE Grapevine trunk diseases of the Northern Midwest
conditions, though more research is needed [98, 99]. Interestingly, developmental histological
studies of grapevine xylem tissues have shown plasticity of vessel diameter even within a vari-
ety or individual based on early season precipitation [100]. Therefore, abundant early irriga-
tion influence vines to develop larger xylem vessels that can allow for vigorously growing
higher yielding grapevines but may also render vines more susceptible to biotic vascular patho-
gens and abiotic environmental stresses. Moreover, environmental stresses, including extreme
weather events, will be more frequent due to climate change which may provide more oppor-
tunities for some vascular pathogens [101]. However, the effects of climate change are likely
dependent on the pathogens, cultivars, and environments in question. For example, intensive
drought conditions have shown both positive [102] or negative [103] effects for grapevines suf-
fering from GTDs.
Species of Phaeoacremonium are found in many grape growing countries and often associ-
ated with GTDs of young vines. Phaeoacremonium spp. are often found to spread through
wounds, nursery propagation and grafting [104–112], see review by Gramaje et al. (2015) [40].
Ph. fraxinopennsylvanicum is widespread throughout the world on other hosts and has been
found in many other investigations in the Midwestern United States [113] but Ph. minimum
appears to be the most widespread Phaeoacremonium spp. throughout grape growing
countries.
Cd. luteo-olivacea has been isolated from many substrates including soils [114], decaying
wood [115–117], and grapevines [68] as well as from various grafting tools [118] and pruning
shears [119]. Cd. luteo-olivacea is often referred to as a weak pathogen and this fungus was not
recognized as pathogenic on grapevine until extended grapevine inoculation studies were con-
ducted (see Table 3). Interestingly, Cadophora was the only genus in this study that showed
some differences in isolation frequency in comparison with four tested criteria of hypothesis
three. No significant differences were observed for any fungal genera isolated compared to
variety berry color (S1 Fig). Some researchers have indicated suspicions that red cultivars are
more susceptible to GTD pathogens, although much more research is needed [120]. Some sig-
nificant differences were observed for a few genera compared to sample variety (S2 Fig) or
sample county origin (S3 Fig). However, the residuals were only slight for the variety and
county origin comparisons and more research is needed to be sure of these correlations. Yet,
Cadophora spp. were found to be significantly less isolated from cordon sample sections and
significantly more isolated from woody sections of trunk, roots, and suckers (S4 Fig).
Increased isolation of Cadophora spp. from the more central main trunk of the vine may be an
indication that infection occurs from the soil or possibly the infection was acquired prior to
planting. Additional research of vineyard soils and nursery stock materials would better eluci-
date the origin of Cadophora spp. in NMW grapevines. Additionally, Ilyonectria liriodendri,
isolated just once in this study, is another weak pathogen often associated with GTDs of roots
and often found in nurseries [121, 122].
Fungi considered nonpathogenic to grapevines according to TrunkDiseaseID.org [47]
included Penicillium, Alternaria, Didymella, Epicoccum, and Paraconiothyrium which were
some of the more frequently isolated genera in this study. However, few GTD pathogenicity
studies have tested Pleosporales fungi. Paraconiothyrium spp. have been demonstrated to be
pathogenic on fruit trees and other woody species [105] and potentially pathogenic on grape-
vines [59]. In the NMW, Paraconiothyrium spp. could be a potential pathogen. This fungus
was recently found associated with the emerald ash borer and found to cause small cankers on
healthy ash trees [113, 123]. Further investigation of the pathogenesis of Pleosporales may
prove interesting considering the persistence in isolation and sequencing of these fungi.
Basidiomycota have been also found to play a role in GTDs [124–126]. Often the presence
of these wood decay fungi are found mainly in older vines following the colonization of faster
PLOS ONE | https://doi.org/10.1371/journal.pone.0269555 June 3, 2022
15 / 25
PLOS ONE Grapevine trunk diseases of the Northern Midwest
growing host detoxifying pioneering ascomycota [16]. In many parts of the grape growing
regions of the world, the primary Basidiomycota associated with GTDs are Fomitporia medi-
terriana [127] and Stereum hirsutum [128]. However, in our study no species of Fomitporia
were isolated but a different species of Stereum was one of the more frequently isolated Basi-
diomycota. Additionally, Trametes versicolor and Cerrena unicolor were also isolated. These
fungi are commonly found on forest and shade trees locally (personal observations). Fruiting
bodies of these fungi have been found on trunks of grapevines that had advanced stages of
GTD symptoms. These Basidiomycota have not been tested for pathogenicity on grapevines
but Trametes versicolor causes cankers and decay on fruit trees [129] and Cerrena unicolor is
well characterized on hardwoods where it is an aggressive canker rot pathogen [130].
GTD fungi in the Xylariales, Botryosphaeriales, and Phaeomoniellales are of concern in
many grape growing countries and this includes fungi in the families Diatrypaceae, Botryo-
sphaeriaceae, and Phaeomoniellaceae [73]. In our study, only 10 Diatrypaceae, and 10 Botryo-
sphaeriaceae isolates were identified. No Phaeomoniellaceae isolates were identified. No
species of Eutypella or Eutypa (both Diatrypaceae), common in some other grape growing
regions, were isolated. However, a few isolates not included in analysis closely matched with
Eutypa but with less than 97% and may prove to be new species following additional detailed
taxonomic studies (See S1 Table). Pestalotiopsis spp., Neopestalotiopsis spp., Diatrypella verru-
ciformis, and Diatrype stigma of the Xylariales as well as Diplodia seriata, Botryosphaeria dothi-
dea, Diplodia corticola, and Phaeobotryon negundinis of the Botryosphaeriales represented a
minority of the GTD pathogens isolated in our study (Table 3).
At the start of our research project, many grape growers expressed concerns about GTDs in
their young vineyards. There was also considerable concern about Botryosphaeriaceae GTD
often called “Bot-rot” or otherwise known as Bot canker. Grape growers in the NMW regularly
associate any wedge-shaped discoloration of cross-sections of grapevines as Bot-rot. Based on
this survey, Botryosphaeriaceae GTD is rare in the NMW. Confusion and concerns of growers,
viticultural professionals, and even fellow researchers is understandable given the complexity
of GTDs in addition to the many various names used in an attempt to sub-categorize GTDs.
Many of these sub-categorized GTDs have been associated with irregular generalized symp-
toms of grapevines influenced by the cultivar or variety, climate, and environmental condi-
tions [32, 118, 131]. Moreover, many GTD designations rapidly become obsolete with each
taxonomic recategorization of fungal species. Fungal taxonomy will likely continue to change
as more genetic information is gathered into databases and mycologist strive to dissolve the
two-name system for fungi [132].
The isolation frequency differences of these typically important GTD groups in other grape
growing regions is especially notable. Such differences could possibly be correlated with the
different climate of the NMW as compared to the many other grape growing regions which
typically have more seasonally mild, often Mediterranean climates. Several spore trapping
studies from various countries have attempted to characterize sporulation events of various
GTD pathogens in correlation with varying weather measurements [87, 133–135]. Given the
drastically different GTDs composition of the NMW, additional studies using spore trapping
would prove insightful to obtain a better understanding of the GTD pathogens in the NMW.
Notably, culturing methods may present bias as some faster growing fungal species such as
those of Cytospora, Diaporthe, Diplodia, and Botryosphaeria may outgrow slower growing spe-
cies such as those of Phaeoacremonium, Phaeomoniella, and many basidiomycota. This bias of
culture-based studies would benefit being paired with modern metagenomic techniques to
characterize all potential microorganismal species present in a substrate. However, many
metagenomic techniques also present bias such as detection of non-viable organisms or unin-
tended preferential over-identification. Thus, metagenomic techniques also benefit being
PLOS ONE | https://doi.org/10.1371/journal.pone.0269555 June 3, 2022
16 / 25
PLOS ONE Grapevine trunk diseases of the Northern Midwest
paired with classical culture-based techniques. Additionally, inclusion of culture-based meth-
ods allows for the curation of live microbial collections for use in future pathogenicity or char-
acterization studies. Therefore, future studies which use both classical culturing and
metagenomic techniques would better elucidate the NMW GTD complex. This combination
of classical and modern techniques has been demonstrated effective in recent local studies of
Heterobasidion Root Rot [136]. Regardless, in the study we present here the sample coverage
curve (Fig 3) was observed to have reached a plateau providing confidence all major fungal
species were identified in this study. Moreover, the use of three types of media allowed for fre-
quent isolation of slow growing fungal species such as those of Phaeoacremonium as well as
infrequent isolation of fast growing fungal species such as those of the Botryosphaeriales.
Therefore, confidence is assured that the isolation frequency of these fungi identified is repre-
sentative for grapevines of the NMW.
In this study we revealed a large diversity of fungal species associated with cold-hardy
hybrid grapevines in the NMW. A handful of these isolates (those with less than 97% sequence
match provided in S1 Table) could potentially be revealed as new species following additional
detailed taxonomic studies. However, the majority of fungal species we identified show Dia-
porthales predominate GTDs in the NMW. Diaporthales GTD species, Cytospora and Dia-
porthe, are generally opportunistic fungi and largely spread to new hosts within short distances
by asexual conidia via rain splash or contaminated tools. Basic understanding of these oppor-
tunistic pathogens lifecycles emphasizes the benefit growers would gain from more intentional
phytosanitary practices such as prompt removal and destruction of pruning debris as well as
the regular sanitization of tools. Pruning debris and diseased canes left unpruned have recently
been shown to be a major source of Diaporthe GTD Inoculum [135]. Our current recommen-
dation for grape growers in the NMW is to prune in the dormant winter season during a
period of cold and dry weather. Recommendations on pruning timing could be fine-tuned by
epidemiological spore trapping studies in NMW vineyards and may possibly allow for some
degree of GTDs forecasting. Vineyard spore trapping could also provide the opportunity for
broader biosurveillance of invasive pathogenic microbial species of forest, shade, and orchard
trees.
Knowing the prevalence of GTDs in the NMW provides insight for the development of
improved management practices. Similar studies of GTD pathogens spread from nurseries of
cold-hardy grapevine hybrid varieties would also provide insight to improved propagation
practices as well as yield less stressed, higher quality, and more vigorous growing nursery stock
plants for growers. Assessment of these hybrid varieties against a panel of GTD pathogens may
reveal novel evidence of resistance or susceptibility that would be useful for grape breeders.
The development of cost-effective rapid molecular assays for the most prevalent GTDs in the
NMW would be a useful tool to measure the effectiveness of practices or the variability of vari-
ety susceptibility to GTDs.
Supporting information
S1 Table. Isolates with less than 97% homology match. Taxonomic rankings from order to
species are denoted followed by isolation frequency in parenthesis. The isolation frequency is
the count of samples each taxa was isolated from a possible 172 samples. Isolates deposited to
GenBank are listed for each species along with the sample variety and county origin of that iso-
late. Pathogenicity studies conducted for each species are listed in brackets following species.
See references for complete citations. Highlighted species have associated pathogenicity stud-
ies.
(DOCX)
PLOS ONE | https://doi.org/10.1371/journal.pone.0269555 June 3, 2022
17 / 25
PLOS ONE Grapevine trunk diseases of the Northern Midwest
S1 Fig. Mosaic plot of genus level taxa isolated from sample varieties.
(TIF)
S2 Fig. Mosaic plot of genus level taxa isolated from sample variety berry color.
(TIF)
S3 Fig. Mosaic plot of genus level taxa isolated from sample section types.
(TIF)
S4 Fig. Mosaic plot of genus level taxa isolated from counties.
(TIF)
Acknowledgments
We thank Owen Geier and Brian Prior for technical support and the staff and management of
all the contributing Northern Midwestern vineyards that provided samples for this study.
Author Contributions
Conceptualization: David H. DeKrey.
Data curation: David H. DeKrey.
Formal analysis: David H. DeKrey.
Funding acquisition: Matthew D. Clark.
Investigation: David H. DeKrey.
Methodology: David H. DeKrey.
Project administration: David H. DeKrey.
Resources: David H. DeKrey, Robert A. Blanchette.
Software: David H. DeKrey.
Supervision: Matthew D. Clark, Robert A. Blanchette.
Validation: David H. DeKrey.
Visualization: David H. DeKrey.
Writing – original draft: David H. DeKrey.
Writing – review & editing: David H. DeKrey, Annie E. Klodd, Matthew D. Clark, Robert A.
Blanchette.
References
1. Bertsch C, Ramı´rez-Suero M, Magnin-Robert M, Larignon P, Chong J, Abou-Mansour E, et al. Grape-
vine trunk diseases: complex and still poorly understood. Plant Pathology. 2013; 62: 243–265. https://
doi.org/10.1111/j.1365-3059.2012.02674.x
2. Haidar R, Yacoub A, Vallance J, Compant S, Antonielli L, Saad A, et al. Bacteria associated with wood
tissues of Esca-diseased grapevines: functional diversity and synergy with Fomitiporia mediterranea
to degrade wood components. Environmental Microbiology. 2021; 23: 6104–6121. https://doi.org/10.
1111/1462-2920.15676 PMID: 34288352
3.
Larignon P, Dubos B. Fungi associated with esca disease in grapevine. European Journal of Plant
Pathology. 1997; 103: 147–157. https://doi.org/10.1023/A:1008638409410
4. Galloway BT, Fairchild DG. Diseases of the Grape in Western New York. The Journal of Mycology.
1891; 6: 95–99. https://doi.org/10.2307/3752563
PLOS ONE | https://doi.org/10.1371/journal.pone.0269555 June 3, 2022
18 / 25
PLOS ONE Grapevine trunk diseases of the Northern Midwest
5. Scheck H, Vasquez S, Fogle D, Gubler WD. Grape growers report losses to black-foot and grapevine
decline. Cal Ag. 1998; 52: 19–23. https://doi.org/10.3733/ca.v052n04p19
6.
The Gatica M. “Hoja de Malvo´ n” grape disease in Argentina. Phytopathologia Mediterranea. 2000; 1–
5. https://doi.org/10.1400/57809
7. Pascoe IG. Grapevine trunk diseases in Australia: diagnostics and taxonomy. Symptoms and occur-
rence of grape declines. Fort Valley, VA: International Ampelography Society and International Coun-
cil on Grapevine Trunk Diseases; 1998. pp. 56–77.
8.
Fontaine F, Gramaje D, Armengol J, Smart R, Nagy ZA, Borgo M, et al. Grapevine trunk diseases. A
review. OIV Publications; 2016.
9. Moller WJ, Kasimatis AN. Dieback of grapevines caused by Eutypa armeniacae. Plant Disease
Reporter. 1978; 62: 254–258.
10.
Lehoczky J. Black dead-arm disease of grapevine caused by Botryosphaeria stevensii infection. Acta
Phytopathologica Academiae Scientiarum Hungaricae. 1974; 9: 319–327.
11. Kajitani Y, Kanematsu S. Diaporthe kyushuensis sp. nov., the teleomorph of the causal fungus of
grapevine swelling arm in Japan, and its anamorph Phomopsis vitimegaspora. Mycoscience. 2000;
41: 111–114. https://doi.org/10.1007/BF02464318
12.
Fontaine F, Pinto C, Vallet J, Cle´ ment C, Gomes AC, Spagnolo A. The effects of grapevine trunk dis-
eases (GTDs) on vine physiology. Eur J Plant Pathol. 2016; 144: 707–721. https://doi.org/10.1007/
s10658-015-0770-0
13. Hewitt WB. Phomopsis cane and leaf spot. Compendium of grape diseases. APS Press; 1988. pp.
17–18.
14. Chiarappa L. Wood decay of the grapevine and its relationship with black measles disease. Phytopa-
thology. 1959;49.
15. Halleen F, Schroers HJ, Groenewald JZ, Crous PW. Novel species of Cylindrocarpon (Neonectria)
and Campylocarpon gen. nov. associated with black foot disease of grapevines (Vitis spp.). Studies in
Mycology. 2004; 50: 431–455.
16. Mugnai L, Graniti A, Surico G. Esca (black measles) and brown wood-streaking: two old and elusive
diseases of grapevines. Plant Disease. 1999; 83: 404–418. https://doi.org/10.1094/PDIS.1999.83.5.
404 PMID: 30845530
17. Kaplan J, Travadon R, Cooper M, Hillis V, Lubell M, Baumgartner K. Identifying economic hurdles to
early adoption of preventative practices: The case of trunk diseases in California winegrape vineyards.
Wine Economics and Policy. 2016; 5: 127–141. https://doi.org/10.1016/j.wep.2016.11.001
18. Baumgartner K, Hillis V, Lubell M, Norton M, Kaplan J. Managing grapevine trunk diseases in Califor-
nia’s southern San Joaquin Valley. Am J Enol Vitic. 2019; 70: 267–276. https://doi.org/10.5344/ajev.
2019.18075
19. Guerin-Dubrana L, Fontaine F, Mugnai L. Grapevine trunk disease in European and Mediterranean
vineyards: occurrence, distribution and associated disease-affecting cultural factors. Phytopathologia
Mediterranea. 2019; 58: 49–71. https://doi.org/10.13128/Phytopathol_Mediterr-25153
20.
21.
Lee T-C, Ho I-C. Modulation of cellular antioxidant defense activities by sodium arsenite in human
fibroblasts. Arch Toxicol. 1995; 69: 498–504. https://doi.org/10.1007/s002040050204 PMID: 8526746
Tanner CC, Clayton JS. Persistence of arsenic 24 years after sodium arsenite herbicide application to
Lake Rotoroa, Hamilton, New Zealand. New Zealand Journal of Marine and Freshwater Research.
1990; 24: 173–179. https://doi.org/10.1080/00288330.1990.9516412
22. Rezgui A, Vallance J, Ben Ghnaya-Chakroun A, Bruez E, Dridi M, Demasse RD, et al. Study of Lasi-
diodiplodia pseudotheobromae, Neofusicoccum parvum and Schizophyllum commune, three patho-
genic fungi associated with the grapevine trunk diseases in the North of Tunisia. Eur J Plant Pathol.
2018; 152: 127–142. https://doi.org/10.1007/s10658-018-1458-z
23. Van Jaarsveld WJ, Halleen F, Mostert L. In vitro screening of Trichoderma isolates for biocontrol of
black foot disease pathogens. Phytopathologia Mediterranea. 2020; 59: 465–471.
24. Mondello V, Songy A, Battiston E, Pinto C, Coppin C, Trotel-Aziz P, et al. Grapevine trunk diseases: a
review of fifteen years of trials for their control with chemicals and biocontrol agents. Plant Disease.
2018; 102: 1189–1217. https://doi.org/10.1094/PDIS-08-17-1181-FE PMID: 30673583
25.
Tuck B, Gartner W, Appiah G. Vineyards and grapes of the North. University of Minnesota; 2016 Dec.
Available: http://conservancy.umn.edu/handle/11299/197814.
26. Hemstad PR, Luby JJ. Utilization of Vitis riparia for the development of new wine varieties with resis-
tance to disease and extreme cold. VII International Symposium on Grapevine Genetics and Breeding
528. 2000. pp. 487–496. https://doi.org/10.17660/ActaHortic.2000.528.70
PLOS ONE | https://doi.org/10.1371/journal.pone.0269555 June 3, 2022
19 / 25
PLOS ONE Grapevine trunk diseases of the Northern Midwest
27. Hemstad PR, Luby JL. La Crescent, a new cold hardy, high quality, white wine variety. VIII Interna-
tional Conference on Grape Genetics and Breeding 603. 2003. pp. 719–722. https://doi.org/10.
17660/ActaHortic.2003.603.100
28. Clark MD. Development of Cold Climate Grapes in the Upper Midwestern U.S. Plant Breeding
Reviews. John Wiley & Sons, Ltd; 2019. pp. 31–60. https://doi.org/10.1002/9781119616801.ch2
29. Yin L, Karn A, Cadle-Davidson L, Zou C, Underhill A, Atkins P, et al. Fine mapping of leaf trichome
density revealed a 747-kb region on chromosome 1 in cold-hardy hybrid wine grape populations. Fron-
tiers in Plant Science. 2021;12. https://doi.org/10.3389/fpls.2021.587640 PMID: 33746993
30. Clark M, Tuck B. Minnesota grape production statistics: 2020—estimates for the yield, production, and
pricing data of the Minnesota grape industry. 2021. Available: http://conservancy.umn.edu/handle/
11299/224882.
31. Billones-Baaijens R, Savocchia S. A review of Botryosphaeriaceae species associated with grapevine
trunk diseases in Australia and New Zealand. Australasian Plant Pathology. 2019; 48: 3–18. https://
doi.org/10.1007/s13313-018-0585-5
32. Van Niekerk JM, Bester W, Halleen F, Crous PW, Fourie PH. The distribution and symptomatology of
grapevine trunk disease pathogens are influenced by climate. Phytopathologia Mediterranea. 2011;
50: S98–S111.
33. Ye Q, Jia J, Manawasinghe IS, Li X, Zhang W, Mugnai L, et al. Fomitiporia punicata and Phaeoacre-
monium minimum associated with Esca complex of grapevine in China. Phytopathology Research.
2021; 3: 11. https://doi.org/10.1186/s42483-021-00087-w
34. Yan J-Y, Xie Y, Zhang W, Wang Y, Liu J-K, Hyde KD, et al. Species of Botryosphaeriaceae involved in
grapevine dieback in China. Fungal Diversity. 2013; 61: 221–236. https://doi.org/10.1007/s13225-
013-0251-8
35.
Li H, Li RY, Wang H. New Disease for Wine-making Grape-Eutypa Dieback. Liquor Making Science
and Technology. 2007; 5: 48.
36. Dı´az GA, Latorre BA. Infection Caused by Phaeomoniella chlamydospora Associated with Esca-like
Symptoms in Grapevine in Chile. Plant Disease. 2014; 98: 351–360. https://doi.org/10.1094/PDIS-12-
12-1180-RE PMID: 30708428
37. Dı´az GA, Auger J, Besoain X, Bordeu E, Latorre BA. Prevalence and pathogenicity of fungi associated
with grapevine trunk diseases in Chilean vineyards. Ciencia e investigacio´ n agraria. 2013; 40: 327–
339. https://doi.org/10.4067/S0718-16202013000200008
38. Urbez-Torres JR, Peduto F, Striegler RK, Urrea-Romero KE, Rupe JC, Cartwright RD, et al. Charac-
terization of fungal pathogens associated with grapevine trunk diseases in Arkansas and Missouri.
Fungal Diversity. 2012; 52: 169–189. https://doi.org/10.1007/s13225-011-0110-4
39. Gubler W, Rolshausen P, Trouillas F, U´ rbez-Torres JR, Voegel T, Leavitt GM, et al. Grapevine trunk
diseases in California: Research update. Practical Winery & Vineyards. 2005; 1–9.
40. Gramaje D, Mostert L, Groenewald JZ, Crous PW. Phaeoacremonium: From esca disease to phaeo-
hyphomycosis. Fungal Biology. 2015; 119: 759–783. https://doi.org/10.1016/j.funbio.2015.06.004
PMID: 26321726
41. U´ rbez-Torres JR, Haag P, Bowen P, O’Gorman DT. Grapevine trunk diseases in British Columbia:
incidence and characterization of the fungal pathogens associated with esca and petri diseases of
grapevine. Plant Disease. 2014; 98: 469–482. https://doi.org/10.1094/PDIS-05-13-0523-RE PMID:
30708729
42. Worrall JJ. Media for selective isolation of hymenomycetes. Mycologia. 1991; 83: 296–302. https://doi.
org/10.1080/00275514.1991.12026013
43. Harrington TC. Cycloheximide sensitivity as a taxonomic character in ceratocystis. Mycologia. 1981;
73: 1123–1129. https://doi.org/10.1080/00275514.1981.12021447
44. Osmundson TW, Eyre CA, Hayden KM, Dhillon J, Garbelotto MM. Back to basics: an evaluation of
NaOH and alternative rapid DNA extraction protocols for DNA barcoding, genotyping, and disease
diagnostics from fungal and oomycete samples. Molecular Ecology Resources. 2013; 13: 66–74.
https://doi.org/10.1111/1755-0998.12031 PMID: 23121735
45. Gardes M, Bruns TD. ITS primers with enhanced specificity for basidiomycetes—application to the
identification of mycorrhizae and rusts. Molecular Ecology. 1993; 2: 113–118. https://doi.org/10.1111/
j.1365-294x.1993.tb00005.x PMID: 8180733
46. Blanchette RA, Held BW, Hellmann L, Millman L, Bu¨ ntgen U. Arctic driftwood reveals unexpectedly
rich fungal diversity. Fungal Ecology. 2016; 23: 58–65. https://doi.org/10.1016/j.funeco.2016.06.001
47.
Lawrence DP, Travadon R, Nita M, Baumgartner K. TrunkDiseaseID.org: a molecular database for
fast and accurate identification of fungi commonly isolated from grapevine wood. Crop Protection.
2017; 102: 110–117. https://doi.org/10.1016/j.cropro.2017.08.017
PLOS ONE | https://doi.org/10.1371/journal.pone.0269555 June 3, 2022
20 / 25
PLOS ONE Grapevine trunk diseases of the Northern Midwest
48. Wickham H, Averick M, Bryan J, Chang W, McGowan LD, Franc¸ois R, et al. Welcome to the Tidy-
verse. Journal of Open Source Software. 2019; 4: 1686. https://doi.org/10.21105/joss.01686
49. Hsieh TC, Ma KH, Chao A. iNEXT: an R package for rarefaction and extrapolation of species diversity
(Hill numbers). Methods in Ecology and Evolution. 2016; 7: 1451–1456. https://doi.org/10.1111/2041-
210X.12613
50. Chao A, Gotelli NJ, Hsieh TC, Sander EL, Ma KH, Colwell RK, et al. Rarefaction and extrapolation with
Hill numbers: a framework for sampling and estimation in species diversity studies. Ecological Mono-
graphs. 2014; 84: 45–67. https://doi.org/10.1890/13-0133.1
51. Sosnowski MR, Shtienberg D, Creaser ML, Wicks TJ, Lardner R, Scott ES. The Influence of Climate
on Foliar Symptoms of Eutypa Dieback in Grapevines. Phytopathology®. 2007; 97: 1284–1289.
https://doi.org/10.1094/PHYTO-97-10-1284 PMID: 18943686
52. Marchi G, Peduto F, Mugnai L, Di Marco S, Calzarano F, Surico G. Some observations on the relation-
ship of manifest and hidden esca to rainfall. Phytopathologia Mediterranea. 2006; 45: S117–S126.
53.
54.
Lawrence DP, Travadon R, Pouzoulet J, Rolshausen PE, Wilcox WF, Baumgartner K. Characteriza-
tion of Cytospora isolates from wood cankers of declining grapevine in North America, with the
descriptions of two new Cytospora species. Plant Pathology. 2017; 66: 713–725. https://doi.org/10.
1111/ppa.12621
Lawrence D, Travadon R, Baumgartner K. Diversity of Diaporthe species associated with wood can-
kers of fruit and nut crops in northern California. Mycologia. 2015; 107: 926–940. https://doi.org/10.
3852/14-353 PMID: 26240309
55. U´ rbez-Torres JR, Peduto F, Smith RJ, Gubler WD. Phomopsis dieback: a grapevine trunk disease
caused by Phomopsis viticola in California. Plant Disease. 2013; 97: 1571–1579. https://doi.org/10.
1094/PDIS-11-12-1072-RE PMID: 30716818
56. U´ rbez-Torres JR, Adams P, Kamas J, Gubler WD. Identification, incidence, and pathogenicity of fun-
gal species associated with grapevine dieback in texas. Am J Enol Vitic. 2009; 60: 497–507.
57. Aroca A, Raposo R. Pathogenicity of Phaeoacremonium Species on Grapevines. Journal of Phytopa-
thology. 2009; 157: 413–419. https://doi.org/10.1111/j.1439-0434.2008.01513.x
58. Cloete M, Fourie PH, Damm U, Crous PW, Mostert L. Fungi associated with die-back symptoms of
apple and pear trees, a possible inoculum source of grapevine trunk disease pathogens. Phytopatho-
logia Mediterranea. 2011; 50: S176–S190.
59. Elena G, Bruez E, Rey P, Luque J. Microbiota of grapevine woody tissues with or without esca-foliar
symptoms in northeast Spain. Phytopathologia Mediterranea. 2018. https://doi.org/10.14601/
Phytopathol_Mediterr-23337
60. Bahmani Z, Abdollahzadeh J, Amini J, Evidente A. Biscogniauxia rosacearum the charcoal canker
agent as a pathogen associated with grapevine trunk diseases in Zagros region of Iran. Sci Rep. 2021;
11: 14098. https://doi.org/10.1038/s41598-021-93630-w PMID: 34239007
61. Berlanas C, Ojeda S, Lo´ pez-Manzanares B, Andre´ s-Sodupe M, Bujanda R, del Pilar Martı´nez-Diz M,
et al. Occurrence and diversity of black-foot disease fungi in symptomless grapevine nursery stock in
Spain. Plant Disease. 2020; 104: 94–104. https://doi.org/10.1094/PDIS-03-19-0484-RE PMID:
31738690
62. Cabral A, Cecı´lia R, CROUS P, Oliveira H. Virulence and cross-infection potential of Ilyonectria spp. to
grapevine. Phytopathologia Mediterranea. 2012; 51: 340–354.
63. Probst CM, Ridgway HJ, Jaspers MV, Eirian Jones E. Pathogenicity of Ilyonectria liriodendri and Dac-
tylonectria macrodidyma propagules in grapevines. Eur J Plant Pathol. 2019; 154: 405–421. https://
doi.org/10.1007/s10658-018-01664-0
64.
65.
Lawrence DP, Travadon R, Baumgartner K. Novel Seimatosporium species from grapevine in north-
ern california and their interactions with fungal pathogens involved in the trunk-disease complex. Plant
Disease. 2018; 102: 1081–1092. https://doi.org/10.1094/PDIS-08-17-1247-RE PMID: 30673434
Trouillas FP, Gubler WD. Pathogenicity of diatrypaceae species in grapevines in California. Plant Dis-
ease. 2010; 94: 867–872. https://doi.org/10.1094/PDIS-94-7-0867 PMID: 30743554
66. Halleen F, Mostert L, Crous PW. Pathogenicity testing of lesser-known vascular fungi of grapevines.
Australasian Plant Pathology. 2007; 36: 277–285. https://doi.org/10.1071/AP07019
67. Raimondo ML, Carlucci A, Ciccarone C, SADDALLAH A, Francesco L. Identification and pathogenicity
of lignicolous fungi associated with grapevine trunk diseases in southern Italy. Phytopathologia Medi-
terranea. 2019; 58: 639–662.
68.
Travadon R, Lawrence DP, Rooney-Latham S, Gubler WD, Wilcox WF, Rolshausen PE, et al. Cado-
phora species associated with wood-decay of grapevine in North America. Fungal biology. 2015; 119:
53–66. https://doi.org/10.1016/j.funbio.2014.11.002 PMID: 25601149
PLOS ONE | https://doi.org/10.1371/journal.pone.0269555 June 3, 2022
21 / 25
PLOS ONE Grapevine trunk diseases of the Northern Midwest
69. U´ rbez-Torres JR, Leavitt GM, Guerrero JC, Guevara J, Gubler WD. Identification and pathogenicity of
Lasiodiplodia theobromae and Diplodia seriata, the causal agents of bot canker disease of grapevines
in Mexico. Plant Disease. 2008; 92: 519–529. https://doi.org/10.1094/PDIS-92-4-0519 PMID:
30769642
70. Marsberg A, Kemler M, Jami F, Nagel JH, Postma-Smidt A, Naidoo S, et al. Botryosphaeria dothidea:
a latent pathogen of global importance to woody plant health. Molecular Plant Pathology. 2017; 18:
477–488. https://doi.org/10.1111/mpp.12495 PMID: 27682468
71. Baranek M, Armengol J, Holleinova´ V, Pečenka J, Calzarano F, Peňa´zova´ E, et al. Incidence of symp-
toms and fungal pathogens associated with grapevine trunk diseases in Czech vineyards: first exam-
ple from a north-eastern European grape-growing region. Phytopathologia Mediterranea. 2018; 57:
449–458. https://doi.org/10.14601/Phytopathol_Mediterr-22460
72. Mundy DC, Brown A, Jacobo F, Tennakoon K, Woolley RH, Vanga B, et al. Pathogenic fungi isolated
in association with grapevine trunk diseases in New Zealand. New Zealand Journal of Crop and Horti-
cultural Science. 2020; 48: 84–96. https://doi.org/10.1080/01140671.2020.1716813
73. Kaliterna J, Miličević T, Cvjetković B. Grapevine trunk diseases associated with fungi from the dia-
porthaceae family in Croatian vineyards. Arhiv za higijenu rada i toksikologiju. 2012; 63: 471–478.
https://doi.org/10.2478/10004-1254-63-2012-2226 PMID: 23334042
74. Rolshausen PE, Wilcox W, Baumgartner K. Distribution and occurrence of fungi associated with
grapevine trunk diseases in Northeastern American vineyards. Phytopathologia Mediterranea. 2010;
49: 105.
75.
van Niekerk JM, Groenewald JZ, Farr DF, Fourie PH, Halleen F, Crous PW. Reassessment of Pho-
mopsis species on grapevines. Austral Plant Pathol. 2005; 34: 27. https://doi.org/10.1071/AP04072
76. Mostert L, Crous PW, Petrini O. Endophytic fungi associated with shoots and leaves of Vitis vinifera,
with specific reference to the Phomopsis viticola complex. Sydowia. 2000; 52: 46–58.
77. Anderson HW, Colby AS. Dead arm disease of grape in Illinois. The Plant Disease Reporter. 1943; 27:
245–246.
78. Baumgartner K, Fujiyoshi PT, Travadon R, Castlebury LA, Wilcox WF, Rolshausen PE. Characteriza-
tion of species of Diaporthe from wood cankers of grape in eastern North American vineyards. Plant
Disease. 2013; 97: 912–920. https://doi.org/10.1094/PDIS-04-12-0357-RE PMID: 30722541
79. Dissanayake AJ, Liu M, Zhang W, Chen Z, Udayanga D, Chukeatirote E, et al. Morphological and
molecular characterisation of Diaporthe species associated with grapevine trunk disease in China.
Fungal Biology. 2015; 119: 283–294. https://doi.org/10.1016/j.funbio.2014.11.003 PMID: 25937058
80.
Fitzpatrick DA. Horizontal gene transfer in fungi. FEMS Microbiology Letters. 2012; 329: 1–8. https://
doi.org/10.1111/j.1574-6968.2011.02465.x PMID: 22112233
81. Casacuberta E, Gonza´ lez J. The impact of transposable elements in environmental adaptation. Molec-
ular Ecology. 2013; 22: 1503–1517. https://doi.org/10.1111/mec.12170 PMID: 23293987
82. Christensen CM. Studies on the biology agric. Valsa sordida and Cytospora chrysosperma. Phytopa-
thology. 1940;30. Available: https://www.cabdirect.org/cabdirect/abstract/19401101375.
83. Cold Outbreak: January 27–31, 2019. In: Minnesota Department of Natural Resources [Internet].
[cited 19 Dec 2021]. Available: https://www.dnr.state.mn.us/climate/journal/cold-outbreak-january-27-
31-2019.html.
84. Svyantek A, Ko¨se B, Stenger J, Auwarter C, Hatterman-Valenti H. Cold-hardy grape cultivar winter
injury and trunk re-establishment following severe weather events in North Dakota. Horticulturae.
2020; 6: 75. https://doi.org/10.3390/horticulturae6040075
85.
van Niekerk JM, Calitz FJ, Halleen F, Fourie PH. Temporal spore dispersal patterns of grapevine trunk
pathogens in South Africa. Eur J Plant Pathol. 2010; 127: 375–390. https://doi.org/10.1007/s10658-
010-9604-2
86. Ayres M, Billones -Baaijens Regina, Savocchia S, Scott E, Sosnowski M. Trunk diseases: susceptibil-
ity of pruning wounds to grapevine trunk disease pathogens. Wine & Viticulture Journal. 31: 48–50.
https://doi.org/10.3316/informit.523883064121148
87. Kraus C, Voegele RT, Fischer M. Temporal development of the culturable, endophytic fungal commu-
nity in healthy grapevine branches and occurrence of GTD-associated fungi. Microb Ecol. 2019; 77:
866–876. https://doi.org/10.1007/s00248-018-1280-3 PMID: 30397796
88. Pouzoulet J, Scudiero E, Schiavon M, Rolshausen PE. Xylem vessel diameter affects the compart-
mentalization of the vascular pathogen Phaeomoniella chlamydospora in grapevine. Frontiers in Plant
Science. 2017; 8: 1442. https://doi.org/10.3389/fpls.2017.01442 PMID: 28871268
89. Bortolami G, Gambetta GA, Delzon S, Lamarque LJ, Pouzoulet J, Badel E, et al. Exploring the Hydrau-
lic Failure Hypothesis of Esca Leaf Symptom Formation. Plant Physiology. 2019; 181: 1163–1174.
https://doi.org/10.1104/pp.19.00591 PMID: 31455632
PLOS ONE | https://doi.org/10.1371/journal.pone.0269555 June 3, 2022
22 / 25
PLOS ONE Grapevine trunk diseases of the Northern Midwest
90. Bortolami G, Farolfi E, Badel E, Burlett R, Cochard H, Ferrer N, et al. Seasonal and long-term conse-
quences of esca grapevine disease on stem xylem integrity. Journal of Experimental Botany. 2021;
72: 3914–3928. https://doi.org/10.1093/jxb/erab117 PMID: 33718947
91. Agrios GN. Plant Pathology. fifth. Elsevier Inc.; 2005.
92. Valtaud C, Larignon P, Roblin G, Fleurat-Lessard P. Developmental and ultrastructural features of
Phaeomoniella chlamydospora and Phaeoacremonium aleophilum in relation to xylem degradation in
esca disease of the grapevine. Journal of Plant Pathology. 2009; 91: 37–51.
93. Yin Z, Liu H, Li Z, Ke X, Dou D, Gao X, et al. Genome sequence of Valsa canker pathogens uncovers
a potential adaptation of colonization of woody bark. New Phytologist. 2015; 208: 1202–1216. https://
doi.org/10.1111/nph.13544 PMID: 26137988
94. Morales-Cruz A, Amrine KCH, Blanco-Ulate B, Lawrence DP, Travadon R, Rolshausen PE, et al. Dis-
tinctive expansion of gene families associated with plant cell wall degradation, secondary metabolism,
and nutrient uptake in the genomes of grapevine trunk pathogens. BMC Genomics. 2015; 16: 469.
https://doi.org/10.1186/s12864-015-1624-z PMID: 26084502
95. Blanchette RA. Degradation of the lignocellulose complex in wood. Can J Bot. 1995; 73: 999–1010.
https://doi.org/10.1139/b95-350
96. Pouzoulet J, Pivovaroff AL, Santiago LS, Rolshausen PE. Can vessel dimension explain tolerance
toward fungal vascular wilt diseases in woody plants? Lessons from Dutch elm disease and esca dis-
ease in grapevine. Frontiers in Plant Science. 2014; 5: 253. https://doi.org/10.3389/fpls.2014.00253
PMID: 24971084
97. Pouzoulet J, Rolshausen PE, Charbois R, Chen J, Guillaumie S, Ollat N, et al. Behind the curtain of
the compartmentalization process: Exploring how xylem vessel diameter impacts vascular pathogen
resistance. Plant, Cell & Environment. 2020; 43: 2782–2796. https://doi.org/10.1111/pce.13848
PMID: 32681569
98. Knipfer T, Eustis A, Brodersen C, Walker AM, Mcelrone AJ. Grapevine species from varied native hab-
itats exhibit differences in embolism formation/repair associated with leaf gas exchange and root pres-
sure. Plant, Cell & Environment. 2015; 38: 1503–1513. https://doi.org/10.1111/pce.12497 PMID:
25495925
99. Smith MS, Centinari M. Young grapevines exhibit interspecific differences in hydraulic response to
freeze stress but not in recovery. Planta. 2019; 250: 495–505. https://doi.org/10.1007/s00425-019-
03183-6 PMID: 31089803
100.
Lovisolo C, Schubert A. Effects of water stress on vessel size and xylem hydraulic conductivity in Vitis
vinifera L. Journal of Experimental Botany. 1998; 49: 693–700. https://doi.org/10.1093/jxb/49.321.693
101. Garrett KA, Nita M, De Wolf ED, Esker PD, Gomez-Montano L, Sparks AH. Chapter 24—Plant patho-
gens as indicators of climate change. In: Letcher TM, editor. Climate Change (Third Edition). Elsevier;
2021. pp. 499–513. https://doi.org/10.1016/B978-0-12-821575-3.00024–4
102. Bortolami G, Gambetta GA, Cassan C, Dayer S, Farolfi E, Ferrer N, et al. Grapevines under drought
do not express esca leaf symptoms. Proceedings of the National Academy of Sciences. 2021;118.
https://doi.org/10.1073/pnas.2112825118 PMID: 34675082
103. Galarneau ERA, Lawrence DP, Travadon R, Baumgartner K. Drought Exacerbates Botryosphaeria
Dieback Symptoms in Grapevines and Confounds Host-based Molecular Markers of Infection by Neo-
fusicoccum parvum. Plant Disease. 2019; 103: 1738–1745. https://doi.org/10.1094/PDIS-09-18-1549-
RE PMID: 31082329
104. Whiteman SA, Stewart A, Ridgway HJ, Jaspers MV. Infection of rootstock mother-vines by Phaeomo-
niella chlamydospora results in infected young grapevines. Australasian Plant Pathology. 2007; 36:
198–203. https://doi.org/10.1071/AP07012
105. Damm U, Verkley GJM, Crous PW, Fourie PH, Haegi A, Riccioni L. Novel Paraconiothyrium species
on stone fruit trees and other woody hosts. Persoonia. 2008; 20: 9–17. https://doi.org/10.3767/
003158508X286842 PMID: 20467483
106. Groenewald M, Kang J-C, Crous PW, Gams W. ITS and β-tubulin phylogeny of Phaeoacremonium
and Phaeomoniella species. Mycological Research. 2001; 105: 651–657. https://doi.org/10.1017/
S0953756201004282
107. Gramaje D, Alaniz S, Pe´rez-Sierra A, Abad-Campos P, Garcı´a-Jime´nez J, Armengol J. First report of
Phaeoacremonium mortoniae causing Petri disease of grapevine in Spain. Plant Disease. 2007; 91:
1206–1206. https://doi.org/10.1094/PDIS-91-9-1206A PMID: 30780685
108. Mohammadi H. First report of Phaeoacremonium mortoniae associated with grapevine decline in Iran.
Plant Disease. 2011; 95: 1034–1034. https://doi.org/10.1094/PDIS-05-11-0361 PMID: 30732074
PLOS ONE | https://doi.org/10.1371/journal.pone.0269555 June 3, 2022
23 / 25
PLOS ONE Grapevine trunk diseases of the Northern Midwest
109. Crous PW, Gams W, Wingfield MJ, van Wyk PS. Phaeoacremonium gen. nov. associated with wilt
and decline diseases of woody hosts and human infections. Mycologia. 1996; 88: 786–796. https://doi.
org/10.1080/00275514.1996.12026716
110. Gatica M, Dupont J, Magnin S, Ce´ sari C. Phaeoacremonium species and Phaeomoniella chlamydos-
pora in vines showing “hoja de malvo´ n” and young vine decline symptoms in Argentina. Phytopatholo-
gia Mediterranea. 2001; 1000–1008. https://doi.org/10.1400/14644
111. Cottral E, Pascoe I. Developments in grapevine trunk diseases research in Australia. Phytopathologia
Mediterranea. 2000; 39: 68–75. https://doi.org/10.1400/57813
112. Essakhi S, Mugnai L, Crous PW, Groenewald JZ, Surico G. Molecular and phenotypic characterisation
of novel Phaeoacremonium species isolated from esca diseased grapevines. Persoonia. 2008; 21:
119–134. https://doi.org/10.3767/003158508X374385 PMID: 20396582
113. Held BW, Simeto S, Rajtar NN, Cotton AJ, Showalter DN, Bushley KE, et al. Fungi associated with gal-
leries of the emerald ash borer. Fungal Biology. 2021; 125: 551–559. https://doi.org/10.1016/j.funbio.
2021.02.004 PMID: 34140151
114. Arenz BE, Held BW, Jurgens JA, Farrell RL, Blanchette RA. Fungal diversity in soils and historic wood
from the Ross Sea Region of Antarctica. Soil Biology and Biochemistry. 2006; 38: 3057–3064. https://
doi.org/10.1016/j.soilbio.2006.01.016
115. Blanchette RA, Held BW, Jurgens JA, McNew DL, Harrington TC, Duncan SM, et al. Wood-destroying
soft rot fungi in the historic expedition huts of Antarctica. Applied and Environmental Microbiology.
2004 [cited 21 Jan 2022]. https://doi.org/10.1128/AEM.70.3.1328-1335.2004 PMID: 15006750
116. Blanchette RA, Held BW, Jurgens J, Stear A, Dupont C. Fungi attacking historic wood of Fort Conger
and the Peary Huts in the High Arctic. PLOS ONE. 2021; 16: e0246049. https://doi.org/10.1371/
journal.pone.0246049 PMID: 33497418
117. Rusman Y, Held BW, Blanchette RA, He Y, Salomon CE. Cadopherone and colomitide polyketides
from Cadophora wood-rot fungi associated with historic expedition huts in Antarctica. Phytochemistry.
2018; 148: 1–10. https://doi.org/10.1016/j.phytochem.2017.12.019 PMID: 29366851
118. Gramaje D, Mostert L, Armengol J. Characterization of Cadophora luteo-olivacea and C. melinii iso-
lates obtained from grapevines and environmental samples from grapevine nurseries in Spain. Phyto-
pathologia Mediterranea. 2011; 50: S112–S126.
119. Agustı´-Brisach C, Leo´ n M, Garcı´a-Jime´nez J, Armengol J. Detection of grapevine fungal trunk patho-
gens on pruning shears and evaluation of their potential for spread of infection. Plant Disease. 2015;
99: 976–981. https://doi.org/10.1094/PDIS-12-14-1283-RE PMID: 30690978
120. Murolo S, Romanazzi G. Effects of grapevine cultivar, rootstock and clone on esca disease. Austral-
asian Plant Pathology. 2014; 43: 215–221. https://doi.org/10.1007/s13313-014-0276-9
121. Agustı´-Brisach C, Mostert L, Armengol J. Detection and quantification of Ilyonectria spp. associated
with black-foot disease of grapevine in nursery soils using multiplex nested PCR and quantitative
PCR. Plant Pathology. 2014; 63: 316–322. https://doi.org/10.1111/ppa.12093
122.
del Pilar Martı´nez-Diz M, Andre´ s-Sodupe M, Berbegal M, Bujanda R, Dı´az-Losada E, Gramaje D.
Droplet digital PCR technology for detection of Ilyonectria liriodendri from grapevine environmental
samples. Plant Disease. 2020; 104: 1144–1150. https://doi.org/10.1094/PDIS-03-19-0529-RE PMID:
32053059
123. Rajtar NN, Held BW, Blanchette RA. Fungi from galleries of the emerald ash borer produce cankers in
ash trees. Forests. 2021; 12: 1509. https://doi.org/10.3390/f12111509
124. Biodiversity Fischer M. and geographic distribution of basidiomycetes causing esca-associates white
rot in grapevine: a worldwide perspective. Phytopathologia Mediterranea. 2006; 1–13.
125.
Fischer M, Garcı´a G. V. An annotated checklist of European basidiomycetes related to white rot of
grapevine. Phytopathologia Mediterranea. 2015; 54: 281–298.
126. Brown AA, Lawrence DP, Baumgartner K. Role of basidiomycete fungi in the grapevine trunk disease
esca. Plant Pathology. 2020; 69: 205–220. https://doi.org/10.1111/ppa.13116
127.
128.
Fischer M, Edwards J, Cunnington JH, Pascoe IG. Basidiomycetous pathogens on grapevine: a new
species from Australia-Fomitiporia australiensis. Mycotaxon. 2005; 92: 85–96.
Tomoiaga LL, Chedea VS. Grapevine trunk diseases, anagement in vineyards from central Transylva-
nia. BUASVMCN-HORT. 2020; 77: 117. https://doi.org/10.15835/buasvmcn-hort:2019.0041
129. Dilley MA, Covey RP. Survey of wood decay and associated hymenomycetes in central Washington
apple orchards. Plant Disease. 1980; 64: 561.
130. Enebak SA, Blanchette RA. Canker formation and decay in sugar maple and paper birch infected by
Cerrena unicolor. Canadian Journal of Forest Research. 2011 [cited 21 Jan 2022]. https://doi.org/10.
1139/x89-031
PLOS ONE | https://doi.org/10.1371/journal.pone.0269555 June 3, 2022
24 / 25
PLOS ONE Grapevine trunk diseases of the Northern Midwest
131. Claverie M, Notaro M, Fontaine F, Wery J. Current knowledge on grapevine trunk diseases with com-
plex etiology: a systemic approach. Phytopathologia Mediterranea. 2020; 59: 29–53.
132. Wingfield MJ, De Beer ZW, Slippers B, Wingfield BD, Groenewald JZ, Lombard L, et al. One fungus,
one name promotes progressive plant pathology. Molecular Plant Pathology. 2012; 13: 604–613.
https://doi.org/10.1111/j.1364-3703.2011.00768.x PMID: 22146077
133. Silva F, Santos K, Tamiris R, Armengol J, Rossi V, Michereff S, et al. Temporal conidial dispersal pat-
tern of Botryosphaeriaceae species in table-grape vineyards in Northeastern Brazil. Phytopathologia
Mediterranea. 2018; 57: 547–556. https://doi.org/10.14601/Phytopathol_Mediterr-24240
134. Urbez-Torres JR, Gispert C, Trouillas FP. Spore dispersal of Eutypella species under desert grape-
growing conditions of southern California. Phytopathologia Mediterranea. 2020; 59: 485–501. https://
doi.org/10.14601/Phyto-11807
135. Gonza´lez-Domı´nguez E, Caffi T, Languasco L, Latinovic N, Latinovic J, Rossi V. Dynamics of Dia-
porthe ampelina conidia released from grape canes that overwintered in the vineyard. Plant Disease.
2021; 105: 3092–3100. https://doi.org/10.1094/PDIS-12-20-2639-RE PMID: 33755509
136. Otto EC, Held BW, Gould TJ, Blanchette RA. Fungal Diversity in Multiple Post-harvest Aged Red Pine
Stumps and Their Potential Influence on Heterobasidion Root Rot in Managed Stands Across Minne-
sota. Frontiers in Fungal Biology. 2021;2. Available: https://www.frontiersin.org/article/10.3389/ffunb.
2021.782181.
PLOS ONE | https://doi.org/10.1371/journal.pone.0269555 June 3, 2022
25 / 25
PLOS ONE |
10.1371_journal.pone.0265477 | RESEARCH ARTICLE
Interactive tool for clustering and forecasting
patterns of Taiwan COVID-19 spread
Mahsa Ashouri, Frederick Kin Hing PhoaID*
Institute of Statistical Science, Academia Sinica, Taipei, Taiwan
* fredphoa@stat.sinica.edu.tw
Abstract
The COVID-19 data analysis is essential for policymakers to analyze the outbreak and man-
age the containment. Many approaches based on traditional time series clustering and fore-
casting methods, such as hierarchical clustering and exponential smoothing, have been
proposed to cluster and forecast the COVID-19 data. However, most of these methods do
not scale up with the high volume of cases. Moreover, the interactive nature of the applica-
tion demands further critically complex yet compelling clustering and forecasting tech-
niques. In this paper, we propose a web-based interactive tool to cluster and forecast the
available data of Taiwan COVID-19 confirmed infection cases. We apply the Model-based
(MOB) tree and domain-relevant attributes to cluster the dataset and display forecasting
results using the Ordinary Least Square (OLS) method. In this OLS model, we apply a
model produced by the MOB tree to forecast all series in each cluster. Our user-friendly
parametric forecasting method is computationally cheap. A web app based on R’s Shiny
App makes it easier for practitioners to find clustering and forecasting results while choosing
different parameters such as domain-relevant attributes. These results could help in deter-
mining the spread pattern and be utilized by medical researchers.
Introduction
The Coronavirus Disease 2019 (COVID-19) from Wuhan (Hubei, China), which started
spreading quickly in late December 2019, was announced as an outbreak by the public health
emergency of international in January 2020 and a pandemic by the World Health Organization
(WHO) on March 11, 2020. It transmits from person to person and causes symptoms like high
fever, cough, and shortness of breath after a 2-to-14-day infection period [1]. On December 15,
2020, more than 72.8 million people were confirmed by COVID-19, with 742 cases confirmed
in Taiwan. Confirmed cases grew exponentially across all continents [2]. The world has
changed dramatically ever since the first case broke out, and many countries have encountered
multiple crises, such as health crises, financial crises, and economic collapses [3]. At that time,
Taiwan had successfully curbed the spread for more than a year since the outbreak started.
Taiwan center of disease control reported the first confirmed infection case on January 21,
2020, a 50-year-old woman who was a teacher in Wuhan. Due to early responses and active
contact tracing policies, Taiwan managed to contain the spread successfully with a record of
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: Ashouri M, Phoa FKH (2022) Interactive
tool for clustering and forecasting patterns of
Taiwan COVID-19 spread. PLoS ONE 17(6):
e0265477. https://doi.org/10.1371/journal.
pone.0265477
Editor: Chun-Hsi Huang, Southern Illinois
University, UNITED STATES
Received: August 23, 2021
Accepted: March 2, 2022
Published: June 30, 2022
Copyright: © 2022 Ashouri, Kin Hing Phoa. This is
an open access article distributed under the terms
of the Creative Commons Attribution License,
which permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: The raw data was
collected from Taiwan Centers for Disease Control
(CDC) and it is available at www.cdc.gov.tw/En.
The tool and dataset are publicly available at github.
com/mahsaashouri/Taiwan-COVID-19-Interactive-
tool.
Funding: FKHP, AS-TP-109-M07, Academia Sinica,
https://www.sinica.edu.tw/ FKHP, 107-2118-M-
001-011-MY3 and 109-2321-B-001-013, Ministry
of Science and Technology (Taiwan), https://www.
most.gov.tw/.
PLOS ONE | https://doi.org/10.1371/journal.pone.0265477 June 30, 2022
1 / 11
PLOS ONE Competing interests: The authors have declared
that no competing interests exist.
Interactive tool for clustering and forecasting patterns of Taiwan COVID-19 spread
250 consecutive days without any locally transmitted cases. However, Taiwan started to face a
sharp surge of confirmed cases in late April 2021 [4]. The policy-making and spread patterns
of the disease depend on many factors (such as environmental factors [5]), which may not fol-
low the previously available models. Therefore, creating a more efficient and accurate interac-
tive analytical tool is essential in identifying the spread pattern and providing helpful
information to enact effective policies.
Time series clustering is essential to determine similarities and/or differences in the behav-
ior of COVID-19 across cities, states, or countries, and it is advantageous in selecting forecast-
ing models. [6] measured the similarity of the COVID-19 time series between states using the
dynamic time warping distance (DTW) as the similarity matrix and applied a hierarchical clus-
tering approach to analyze the behavioral relationships in the United States (US) pandemic. As
a result, they found different pandemic behaviors in eastern and western zones. [7] suggested a
non-negative matrix factorization (NMF) followed by a k-means clustering procedure on the
coefficients of the NMF basis to cluster the US states into different communities. Their method
not only has the advantage of capturing patterns, but it has also reflected the spread and con-
trol of the pandemic by July 25, 2020.
[8] used an unsupervised machine learning technique to identify COVID-19 cases. They
applied a lung radiography dataset to the Robust Continuous Clustering algorithm (RCC) to
identify confirmed patients.
Forecasting the pattern of the COVID-19 pandemic is critical to health services, health
policymakers, healthcare providers, and epidemiologists. Various time series approaches
aim to forecast the COVID-19 pandemic using statistical modeling. For example, [9] pro-
posed a time series statistical approach to predict the short-term behavior of COVID-19.
They applied multiplicative trend to forecast the number of confirmed cases and deaths
globally and presented a 10-day-ahead competitive forecast over four months. [10] intro-
duced an objective approach to predict the continuation of COVID-19. They produced fore-
casts using models from the exponential smoothing family suitable for the short-term time
series. [2] presented a simple interactive non-linear method to forecast the number of con-
firmed cases. Their method took the expected recoveries and deaths into account to deter-
mine the maximum daily growth rate. Finally, [11] suggested a simplified and accurate
method using fast linear regressions with only a few parameters to forecast deaths, which
can consider the effect of many complexities of the epidemic process.
R’s Shiny app
R’s Shiny app [12] is a package from RStudio [13] developed for an easier and more efficient
result visualization. This web-based application allows users to change the model parameters
and interact with results. In the COVID-19 subject, many researchers published medical and
epidemiological research regarding interactive data analysis and visualization with the R Shiny
framework.
For instance, the COVID-19 tracker [14] in R’s Shiny package provides more context for
daily headlines and a fresh perspective of historical turning points.
[15] developed a COVID-19 worldwide web-based application using R’s Shiny package.
They design the tool for the country-specific analysis to visualize epidemiological pandemic
indicators. [16] suggested a COVID-19 watcher of the updated information for medical and
public use. Their tool aggregated the data from different resources and visualized them using
an online dashboard.
This research proposes an interactive web-based R’s Shiny app to cluster and forecast Tai-
wan COVID-19 time series while benefiting from domain-relevant attributes. Our tool helps
PLOS ONE | https://doi.org/10.1371/journal.pone.0265477 June 30, 2022
2 / 11
PLOS ONE Interactive tool for clustering and forecasting patterns of Taiwan COVID-19 spread
users choose from various parameters to interact with results. For example, users can identify
possible domain-relevant splitting variables of interest.
The dataset contains the number of confirmed cases in the cities, townships, and districts of
Taiwan. The data was collected from Taiwan Centers for Disease Control (CDC)and contains
183 daily series with a length of 155 from January 1 to June 4, 2021. We assume that this gov-
ernmental data is legitimate and trustworthy.
Our app can also be used to analyze COVID-19 time series data of any places in the rest of
the world, only if the dataset follows the same structure below. There should be eight columns
in the dataset corresponding to administrative types, city name, a YES/NO on whether the city
has an airport, a YES/NO on whether the case is imported or local otherwise, the number of
cases, the region category, the number of population, and the date. Among them, all except the
number of cases, the number of population, and the dates are categorical entries. In addition,
all rows are arranged in the ascending order of the dates for each city. Finally, the first row
should be the name of the titles of these eight columns.
Methodology
To cluster and forecast COVID-19 time series, we applied the method suggested by [17], and
we will briefly explain it in this section. This clustering approach applies domain-relevant attri-
butes and time series temporal patterns (trend, seasonality, and autocorrelation). Domain-rel-
evant attributes are cross-sectional attributes that link time series into sub-groups. For
example, the sales volume of items in a supermarket can be divided into different sub-groups.
Similarly, the COVID-19 cases can also be grouped based on geographical features.
This method based on the model-based partitioning tree (MOB) [18] is automated for clus-
tering large collections of time series. It consists of fitting local parametric models into differ-
ent subsets based on a recursive partitioning algorithm. The parameters and split points are
estimated using an objective function and a greedy forward search. To determine which vari-
able should be used for partitioning, we test each model score for parameter instability in each
node. Each node of the resulting tree is associated with a parametric statistical model. When
using the MOB algorithm, we need to specify the outcome, the predictors, the splitting vari-
ables, and the ‘fit’ function. The next part will discuss how the ordinary least squares (OLS)
model is used as the ‘fit’ function within the MOB framework.
To capture time series temporal patterns, [17] suggested an OLS model with predictors to
model their trend, seasonality, and autocorrelation. This model is parametric and flexible in
trend shapes (e.g., linear, quadratic) and seasonal patterns (e.g., seasonal dummies or a smooth
function for slowly changing seasonality). These predictors allow incorporating external attri-
butes valuable for clustering or forecasting time series. For instance, we can include the ‘Easter’
dummy variable indicating the timing of Easter.
Y ¼ Trend þ Season þ ARðpÞ þ External data þ error;
ð1Þ
Where AR(p) is a weighted average of lags in order p, and p can be equal to seasonality order or
specified based on the data type and domain knowledge. As an example, Eq 1 can be written as:
yt ¼ a0 þ a1f ðtÞ
þb1Season1t þ b2Season2t þ � � � þ bm(cid:0) 1Seasonðm(cid:0) 1Þt
þg1yt(cid:0) 1 þ g2yt(cid:0) 2 þ � � � þ gpyt(cid:0) p
þdzt þ �t;
PLOS ONE | https://doi.org/10.1371/journal.pone.0265477 June 30, 2022
ð2Þ
3 / 11
PLOS ONE Interactive tool for clustering and forecasting patterns of Taiwan COVID-19 spread
where yt (t = 1, 2, . . ., T) is the value of series at time t, f(t) is a function of the time index that
captures trend (e.g., linear, quadratic), Seasonjt is a dummy variable taking value 1 if time t is in
season j, m is the number of seasons (e.g., for a daily time series with day-of-week seasonality,
m = 7), and zt is the external data at time t. Furthermore, yt−j is the jth lagged value. One advan-
tage of OLS models is the interpretability of coefficients. The contribution of each feature to the
output will be equal to its coefficient. For example, if there is a linear trend, α1 measures the
changes in yt from one period to the next due to the passage of time while holding other vari-
ables in the model constant. As another example, with quadratic trend, α1 f(t) would be a0
1t þ
1t2 means when a0
a00
1 are positive, the trend is increasing while holding other variables in
the model constant.
1 and a00
Using the MOB partitioning tree and pseudo-R notations with partitioning variables
[Z1, . . ., Zq], Eq 2 can be written as:
yt ¼ a0 þ a1f ðtÞ
þb1Season1t þ b2Season2t þ � � � þ bm(cid:0) 1Seasonðm(cid:0) 1Þt
ð3Þ
þg1yt(cid:0) 1 þ g2yt(cid:0) 2 þ � � � þ gpyt(cid:0) p
þdztjZ1 þ � � � þ Zq:
This approach creates clusters with the same domain-relevant attribute profile and the simi-
lar trend, seasonality, and autocorrelation pattern. Based on this approach, we can cluster the
time series using Algorithm 1.
Algorithm 1: MOB time series clustering algorithm
• Zero time series: separate ‘all zero’ time series
• Normalize the series: subtract the mean and divide the standard
deviation
• MOB tree: run the tree on the series using Eq 3
• Prune the MOB tree: stop the tree when reaching the best improvement
on Mean Square Error (MSE), tree simplicity, and AIC [19] or BIC [20]
• Coefficient plot: compare OLS models in non-neighboring clusters and
check their differences/similarities
Finally, we computed forecasts by one linear model in each cluster produced by the MOB
partitioning tree. We apply the same linear model for series in the same cluster to produce
forecasts. We generate forecasts at fixed time t with h steps ahead (the lagged values of y are
replaced by their forecasted values if they occur in periods after the forecast origin). We also
compare our OLS forecast results with the Exponential Smoothing (ETS) approach. For run-
ning ETS, we applied functions ‘est’ forecast package [21] in R. We run this function indepen-
dently on each series. Then, we use the average of Root Mean Square Errors (RMSEs) and
Mean Absolute Error (MAE) across all series and display box and density plots for forecast
errors. We define the forecast error as the difference between the observed value and its fore-
cast. For better visualization, we do not plot the outliers.
Clustering and forecasting Taiwan COVID-19 confirmed cases
The collected dataset includes 183 daily series (cities, townships, and districts) with a length of
161 from January 1 to June 10, 2021, and the number excluding zero time series is one. Before
running the clustering method, we scaled the data by subtracting the mean and dividing the
standard deviation. Additionally, we partition the data into training and test sets, with the last
PLOS ONE | https://doi.org/10.1371/journal.pone.0265477 June 30, 2022
4 / 11
PLOS ONE Interactive tool for clustering and forecasting patterns of Taiwan COVID-19 spread
Table 1. Domain-relevant attribute categories used in Taiwan COVID-19 confirmed infection cases.
Domain-relevant attributes
Categories
Region
Administrative
Population
Imported
Airport
north, east, west, south, null (imported cases)
township/city, district, null (imported cases)
numeric—no categories
yes, no (local cases)
yes, no (the city has an international airport or not)
https://doi.org/10.1371/journal.pone.0265477.t001
7 days as our test set and the rest as the training set. Then we combine the training and test sets
and update the model and forecast one-week-ahead of the confirmed cases. Note that we
update the model in each cluster while keeping the clustering results unchanged.
For Taiwan COVID-19 daily dataset, we included the following predictors in the MOB-
based clustering and forecasting OLS model (‘fit’ function): a linear trend, six seasonal dum-
mies, and lags 1 to 7. Also domain-relevant splitting variables includes geographical division,
including ‘region’ (6 categories), ‘administrative’ (3 categories), ‘population’ (numeric),
‘imported’ (2 categories), and ‘airport’ (2 categories) (Table 1).
Interactive tool
Table 2 demonstrates the interactive panel inside our tool with three options to choose from,
the MOB depth (number of splits +1), prune option, and domain-relevant attributes (splitting
variables). Additionally, the ‘choose file to upload’ button lets users upload the desired dataset.
Our web-based interactive tool consists of eight parts (displays in Figs 1 to 6). For better
visualization, we divide the results into six figures.
The number on the top shows the MSE for all splits in the MOB partitioning tree. The first
MOB-heatmap includes two parts (Fig 1). The right part displays the MOB tree, which helps
users see domain-relevant attributes and split order accessioned with each cluster based on the
specified depth, prune options, and domain-relevant attributes. The left part is the time series
heatmap of all clusters, displaying time series patterns. Each row represents one series, and
darker color means higher values (color pallet: white, green, and red). Vertical stripes specify
similarities among the series in each cluster.
The second MOB-heatmap (Fig 2) is similar to the first plot, except it combines periods
into seasonal aggregations to highlight the seasonal effects. In both heatmaps, we order series
based on their values. For example, series with higher values in a similar period gather in the
same area. Also, based on the number of series in each cluster, the size of the cluster box would
be different.
Table 2. Taiwan COVID-19 interactive tool panel.
Categories
Application
Choose file to upload
let users upload the Taiwan COVID-19 dataset
MOB depth (number of
splits + 1)
Prune option
Splitting variables
changes from ‘no split’ to ‘full tree’, which controls the tree simplicity
AIC or BIC
include all available options for domain-relevant attributes (splitting variables).
Options are ‘region’, ‘administrative’, ‘population’, ‘imported’, and ‘airport’
Screenshot
let users screenshot the result
https://doi.org/10.1371/journal.pone.0265477.t002
PLOS ONE | https://doi.org/10.1371/journal.pone.0265477 June 30, 2022
5 / 11
PLOS ONE Interactive tool for clustering and forecasting patterns of Taiwan COVID-19 spread
Fig 1. Clustering and forecasting Taiwan COVID-19 confirmed infection cases—Part 1.
https://doi.org/10.1371/journal.pone.0265477.g001
Fig 2. Clustering and forecasting Taiwan COVID-19 confirmed infection cases—Part 2.
https://doi.org/10.1371/journal.pone.0265477.g002
Fig 3. Clustering and forecasting Taiwan COVID-19 confirmed infection cases—Part 3.
https://doi.org/10.1371/journal.pone.0265477.g003
The following plot shows the time series line chart in gray and the average line in red (Fig
3). The coefficient plot displays OLS coefficients for predictors in all clusters (Fig 4). In other
words, each line represents one model connecting the coefficients for each predictor. This plot
is useful for users to choose the number of clusters. Also, by clicking on the coefficient points,
its value will appear in the box below. The final plots, the forecast error box, and density plots,
display forecast errors for the OLS and ETS methods on a one-week test set (Fig 5). In the
PLOS ONE | https://doi.org/10.1371/journal.pone.0265477 June 30, 2022
6 / 11
PLOS ONE Interactive tool for clustering and forecasting patterns of Taiwan COVID-19 spread
Fig 4. Clustering and forecasting Taiwan COVID-19 confirmed infection cases—Part 4.
https://doi.org/10.1371/journal.pone.0265477.g004
following tables, we first examine the linear models in each cluster and OLS results by comput-
ing Pearson [22] and concordance correlation coefficient [23] (between forecasted and
observed values). Then we compare OLS and ETS approaches using RMSE and MAE across
all series. Lastly, we presented the one-week-ahead forecast results (by updated model on com-
bined training and test sets) of all cities, townships, and districts in Taiwan computed by OLS
and ETS models (Fig 6). Users can download the forecasting results in an excel file by clicking
on the ‘Excel’ button next to tables.
Fig 5. Clustering and forecasting Taiwan COVID-19 confirmed infection cases—Part 5.
https://doi.org/10.1371/journal.pone.0265477.g005
PLOS ONE | https://doi.org/10.1371/journal.pone.0265477 June 30, 2022
7 / 11
PLOS ONE Interactive tool for clustering and forecasting patterns of Taiwan COVID-19 spread
Fig 6. Clustering and forecasting Taiwan COVID-19 confirmed infection cases—Part 6.
https://doi.org/10.1371/journal.pone.0265477.g006
Figs 1 to 6 demonstrate the screenshots of our interactive tool results of Taiwan COVID-19
confirmed cases.
In Fig 1, in the top left side panel, we chose three as the MOB tree depth (two splits), AIC as
the prune option, and all splitting variables as domain-relevant attributes, which resulted in
three clusters differing in terms of ‘population’ and ‘region’. Changing options in the panel
update results shown in Figs 1 to 6. In Fig 1, the first split divides the series into population
more than 198795 and population less than 198795, and for more populated areas, there is no
further splits while in the less populated area there is one further split on the ‘region’, shows
series in central, east, south, islands (up) behave differently from north, null (imported cases)
(down).
Table 3 represents the final clusters of confirmed cases with the number of series.
PLOS ONE | https://doi.org/10.1371/journal.pone.0265477 June 30, 2022
8 / 11
PLOS ONE Interactive tool for clustering and forecasting patterns of Taiwan COVID-19 spread
Table 3. Cluster categories of Taiwan COVID-19 confirmed infection cases by choosing three as the MOB depth,
AIC as pruning option, and region, population, imported, administrative, and airport as domain-relevant
attributes.
Cluster 1
Cluster 2
Cluster 3
Cluster categories
Population: more than 198795
Population: less than 198795
Region: central, east, south and islands
Population: less than 198795
Region: north and null (imported cases)
https://doi.org/10.1371/journal.pone.0265477.t003
Number of series
26
103
54
The heatmap in this figure shows in early June—when pick started—the number of con-
firmed cases is higher and more frequent (frequent dark green and red points—‘spiky’ series)
in the more populated areas (cluster 1), while the number of confirmed cases is fewer and less
frequent (frequent light green points) in the less populated areas (clusters 2 and 3). Also, the
diverse distribution of cases (time series temporal patterns), based on populations and regions,
is visible between the final clusters.
The heatmap in Fig 2 shows changes in the number of confirmed cases on different days of
the week. Based on this plot, the number of reported cases in all clusters is lower on Mondays
and Tuesdays and slightly higher on Sundays. Fig 3 shows the line chart of all series with their
average (red line) in each cluster. The comparison of the series and the average line in different
clusters shows the visible between-cluster variability. Clusters 1 (more populated areas) show
more confirmed cases. In cluster 3, the imported case series demonstrates a continuous report
of confirmed cases from the beginning of the year. Another series (Wanhua District) shows a
high jump in early June when the breakdown started in Taiwan.
Based on the coefficient plot in Fig 4, coefficients in all clusters differ mainly in terms of
lags (daily autocorrelation coefficients). The trend and seasonal dummies do not seem to vary
across clusters.
The final part of our web-based interactive tool is the forecast results displayed in Figs 5
and 6. We presented the forecasting performance on a one-week test set, using one OLS model
in each cluster, and compared it with forecasts generated by ETS, a more complex method.
Error box and density plots in Fig 5 show the one-week-ahead forecast errors of three clus-
ters using OLS and ETS models. Based on the error distribution of these two approaches, we
can see that for the Taiwan COVID-19 dataset, OLS performs significantly better. We compute
Pearson and concordance correlation coefficients between observed and forecasted values to
evaluate the OLS performance and forecast precision on each cluster. Based on these coeffi-
cients, the forecasting result, computed by three OLS models, is precise. We also compared
their performances using RMSE and MAE, and the results are the same as in plots. Lastly, we
present two tables in Fig 6 that indicate the one-week-ahead forecast for all cities, townships,
and districts in Taiwan using updated OLS and ETS models on combined training and test
sets in each cluster.
Conclusion
This research proposes an interactive web-based Shiny app for clustering and forecasting Tai-
wan COVID-19 confirmed infection cases. This tool is designed based on the MOB partition-
ing tree, cross-sectional attributes called domain-relevant attributes, and time series temporal
patterns (trend, seasonality, and autocorrelation). Our tool helps users analyze Taiwan
COVID-19 data via changing factors, including MOB depth, model complexity parameter
(AIC or BIC), and domain-relevant attributes.
PLOS ONE | https://doi.org/10.1371/journal.pone.0265477 June 30, 2022
9 / 11
PLOS ONE Interactive tool for clustering and forecasting patterns of Taiwan COVID-19 spread
One advantage of our tool is grouping the series into interpretable clusters in which we can
label a certain cluster by its corresponding domain-relevant attributes. This MOB-based clus-
tering approach results in a single parametric OLS model in each cluster used to forecast all
series in that cluster. Clustering series into groups with similar temporal patterns led us to
enough accurate forecasts of Taiwan COVID-19 confirmed cases. This OLS forecasting
approach has low computational complexity in forecasting these cases.
Our clustering results determine the different spread patterns of confirmed infection cases
in the least populated in different regions and most populated areas. For example, the number
of confirmed cases in populated areas is higher than in other places. Also, the COVID-19 time
series shows different seasonality patterns on certain days of the week, higher on Sundays and
lower on Mondays and Tuesdays.
Another advantage of our tool is its usefulness in handling the existence of missing values
(missing completely at random (MCAR) or missing at random (MAR) variables)—displayed
in gray in heatmaps. In addition, users can have the most updated results of the COVID-19
transmission in Taiwan by simply updating the dataset in the tool. Although this tool is specifi-
cally designed for Taiwan COVID-19 confirmed cases, it can be easily applied to other regions
and/or countries with few changes and updates.
The OLS and ETS forecast results show an increase in infected cases in different cities. Note
that these results are before vaccine rollout, and we need to adjust the model to consider the
vaccination effect on the forecasting results. In addition, the concordance of our forecast is not
studied in this work, and we expect that the forecast has a no-more-than moderate concor-
dance. It is a future task to improve the concordance of our forecast.
Supporting information
S1 File.
(PDF)
Acknowledgments
The authors would like to thank Ms. Ula Tzu-Ning Kung for providing English editing service
in this paper.
Author Contributions
Conceptualization: Mahsa Ashouri, Frederick Kin Hing Phoa.
Formal analysis: Mahsa Ashouri.
Funding acquisition: Frederick Kin Hing Phoa.
Investigation: Mahsa Ashouri.
Methodology: Mahsa Ashouri, Frederick Kin Hing Phoa.
Software: Mahsa Ashouri.
Supervision: Frederick Kin Hing Phoa.
Validation: Mahsa Ashouri, Frederick Kin Hing Phoa.
Visualization: Mahsa Ashouri.
Writing – original draft: Mahsa Ashouri, Frederick Kin Hing Phoa.
Writing – review & editing: Frederick Kin Hing Phoa.
PLOS ONE | https://doi.org/10.1371/journal.pone.0265477 June 30, 2022
10 / 11
PLOS ONE Interactive tool for clustering and forecasting patterns of Taiwan COVID-19 spread
References
1. Anastassopoulou C, Russo L, Tsakris A, Siettos C. Data-based analysis, modelling and forecasting of
the COVID-19 outbreak. PloS one. 2020; 15(3):e0230405. https://doi.org/10.1371/journal.pone.
0230405 PMID: 32231374
2. Perc M, Gorisˇek Miksić N, Slavinec M, Stozˇer A. Forecasting Covid-19. Frontiers in Physics. 2020;
8:127. https://doi.org/10.3389/fphy.2020.00127
3. Gopinath G. The great lockdown: Worst economic downturn since the Great Depression. IMFBlog–
Insights & analysis on economics & finance. 2020;.
4.
Is Taiwan’s COVID success story in jeopardy?;.
5. Aabed K, Lashin MM. An analytical study of the factors that influence COVID-19 spread. Saudi journal
of biological sciences. 2021; 28(2):1177–1195. https://doi.org/10.1016/j.sjbs.2020.11.067 PMID:
33262677
6. Rojas I, Rojas F, Valenzuela O. Estimation of COVID-19 dynamics in the different states of the United
States using Time-Series Clustering. medRxiv. 2020;.
7. Chen J, Yan J, Zhang P. Clustering US States by Time Series of COVID-19 New Case Counts with
Non-negative Matrix Factorization. arXiv preprint arXiv:201114412. 2020;.
8. Householder J, Householder A, Gomez-Reed JP, Park F, Zhang S. Clustering COVID-19 Lung Scans.
arXiv preprint arXiv:200909899. 2020;.
9. Petropoulos F, Makridakis S, Stylianou N. COVID-19: Forecasting confirmed cases and deaths with a
simple time-series model. International Journal of Forecasting. 2020;.
10. Petropoulos F, Makridakis S. Forecasting the novel coronavirus COVID-19. PloS one. 2020; 15(3):
e0231236. https://doi.org/10.1371/journal.pone.0231236 PMID: 32231392
11. Srivastava A, Xu T, Prasanna VK. Fast and Accurate Forecasting of COVID-19 Deaths Using the SIkJα
Model. arXiv preprint arXiv:200705180. 2020;.
12. RStudio, Inc. Shiny: web application framework for R.; 2013. Available from: http://www.rstudio.com/
shiny/.
13. RStudio Team. RStudio: Integrated Development Environment for R; 2020. Available from: http://www.
rstudio.com/.
14. Parker E. COVID-19 tracker.; 2019. Available from: https://shiny.rstudio.com/gallery/covid19-tracker.
html.
15.
Tebe´ C, Valls J, Satorra P, Tobı´as A. COVID19-world: a shiny application to perform comprehensive
country-specific data visualization for SARS-CoV-2 epidemic. BMC Medical Research Methodology.
2020; 20(1):1–7. https://doi.org/10.1186/s12874-020-01121-9 PMID: 32958001
16. Wissel BD, Van Camp P, Kouril M, Weis C, Glauser TA, White PS, et al. An interactive online dashboard
for tracking covid-19 in us counties, cities, and states in real time. Journal of the American Medical Infor-
matics Association. 2020;. https://doi.org/10.1093/jamia/ocaa071 PMID: 32333753
17. Ashouri M, Shmueli G, Sin CY. Tree-based methods for clustering time series using domain-relevant
attributes. Journal of Business Analytics. 2019; p. 1–23. https://doi.org/10.1080/2573234X.2019.
1645574
18.
Zeileis A, Hothorn T, Hornik K. Model-based recursive partitioning. Journal of Computational and
Graphical Statistics. 2008; 17(2):492–514. https://doi.org/10.1198/106186008X319331
19. Akaike H. Information theory and an extension of the maximum likelihood principle. In: Selected papers
of hirotugu akaike. Springer; 1998. p. 199–213.
20. Schwarz G, et al. Estimating the dimension of a model. The annals of statistics. 1978; 6(2):461–464.
https://doi.org/10.1214/aos/1176344136
21. Hyndman RJ, Athanasopoulos G. Forecasting: principles and practice. OTexts; 2018.
22.
23.
Freedman D, Pisani R, Purves R. Statistics (international student edition). Pisani, R Purves, 4th edn
WW Norton & Company, New York. 2007;.
Lawrence I, Lin K. A concordance correlation coefficient to evaluate reproducibility. Biometrics. 1989; p.
255–268.
PLOS ONE | https://doi.org/10.1371/journal.pone.0265477 June 30, 2022
11 / 11
PLOS ONE |
10.1371_journal.pone.0267550 | RESEARCH ARTICLE
Anatomy Nights: An international public
engagement event increases audience
knowledge of brain anatomy
Katherine A. SandersID
Claire L. Cunningham4, Jason M. OrganID
1☯*, Janet A. C. PhilpID
3,5
2☯, Crispin Y. Jordan2, Andrew S. Cale3,
1 Centre for Anatomical and Human Sciences, Hull York Medical School, University of Hull, Hull, United
Kingdom, 2 Deanery of Biomedical Sciences, Edinburgh Medical School, University of Edinburgh, Edinburgh,
United Kingdom, 3 Department of Anatomy, Cell Biology & Physiology, Indiana University School of
Medicine, Indianapolis, Indiana, United States of America, 4 Centre for Anatomy and Human Identification,
Medical Sciences Institute, University of Dundee, Dundee, United Kingdom, 5 Department of Communication
Studies, Indiana University Purdue University Indianapolis, Indianapolis, Indiana, United States of America
☯ These authors contributed equally to this work.
* Kat.Sanders@hyms.ac.uk
Abstract
Anatomy Nights is an international public engagement event created to bring anatomy and
anatomists back to public spaces with the goal of increasing the public’s understanding of
their own anatomy by comparison with non-human tissues. The event consists of a 30-min-
ute mini-lecture on the anatomy of a specific anatomical organ followed by a dissection of
animal tissues to demonstrate the same organ anatomy. Before and after the lecture and
dissection, participants complete research surveys designed to assess prior knowledge and
knowledge gained as a result of participation in the event, respectively. This study reports
the results of Anatomy Nights brain events held at four different venues in the UK and USA
in 2018 and 2019. Two general questions were asked of the data: 1) Do participant post-
event test scores differ from pre-event scores; and 2) Are there differences in participant
scores based on location, educational background, and career. We addressed these ques-
tions using a combination of generalized linear models (R’s glm function; R version 4.1.0 [R
Core Team, 2014]) that assumed a binomial distribution and implemented a logit link func-
tion, as well as likelihood estimates to compare models. Survey data from 91 participants
indicate that scores improve on post-event tests compared to pre-event tests, and these
results hold irrespective of location, educational background, and career. In the pre-event
tests, participants performed well on naming structures with an English name (frontal lobe
and brainstem), and showed signs of improvement on other anatomical names in the post-
test. Despite this improvement in knowledge, we found no evidence that participation in
Anatomy Nights improved participants’ ability to apply this knowledge to neuroanatomical
contexts (e.g., stroke).
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: Sanders KA, Philp JAC, Jordan CY, Cale
AS, Cunningham CL, Organ JM (2022) Anatomy
Nights: An international public engagement event
increases audience knowledge of brain anatomy.
PLoS ONE 17(6): e0267550. https://doi.org/
10.1371/journal.pone.0267550
Editor: Ebrahim Shokoohi, University of Limpopo,
SOUTH AFRICA
Received: May 17, 2021
Accepted: April 12, 2022
Published: June 9, 2022
Peer Review History: PLOS recognizes the
benefits of transparency in the peer review
process; therefore, we enable the publication of
all of the content of peer review and author
responses alongside final, published articles. The
editorial history of this article is available here:
https://doi.org/10.1371/journal.pone.0267550
Copyright: © 2022 Sanders et al. This is an open
access article distributed under the terms of the
Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: All relevant data are
within the manuscript and its Supporting
Information files. We have clearly described all of
the statistical tests used as well as which functions
PLOS ONE | https://doi.org/10.1371/journal.pone.0267550 June 9, 2022
1 / 12
PLOS ONE in R were used. Furthermore, we provide
supplementary files that include all the raw data
from the surveys (S3), all estimates from
generalized linear mixed models (S2), and the R
scripts used to analyze the data (S4) so that
anyone interested can re-run the analyses if they
wish.
Funding: The authors received no specific funding
for this work.
Competing interests: The authors have declared
that no competing interests exist.
Anatomy Nights events and public knowledge of anatomy
Introduction
The public has always had a fascination with the human body. Public dissections were histori-
cally led by experts of anatomy, including Mondino De Luzzi (14th Century) [1], and Andreas
Vesalius (16th Century) [2, 3]. More recently, the presentation of previously dissected human
bodies by BODY WORLDS has become phenomenally popular, each adding something to the
public understanding of anatomy [4, 5]. Here, we describe Anatomy Nights, a new public
engagement format that returns the art of dissection to public audiences.
Anatomy Nights was created to bring anatomy and anatomists back to public spaces and
audiences. It is an event series coordinated by a central team and provides the necessary tools
to enable anatomists to engage with a local public audience in local public spaces [6]. Through
the presentation of human anatomical concepts and demonstration of these concepts via dis-
section of animal tissue, Anatomy Nights’ goal is to increase the public’s understanding of
their own anatomy by reference to non-human tissues (e.g., lamb, pig), and to link this ana-
tomical knowledge to common health conditions such as stroke.
Anatomy has long been considered a cornerstone of medical education, and the develop-
ment of public knowledge of anatomy can be considered important in promoting health liter-
acy [7]. This is particularly significant as low health literacy is associated with poorer health
outcomes [8]. Over 100,000 and 795,000 people are affected by stroke each year in the UK [9]
and the USA [10], respectively, and these figures are expected to rise. Stroke is an anatomically
related medical condition that affects a large proportion of our population, and yet the public’s
understanding of their own bodies has been demonstrated multiple times to be lacking [11,
12]. Whilst members of the public can generally correctly identify that the brain is located
within the skull [11], how it works and how injuries can affect it are not so widely understood.
Coupled with an overestimation of the publics’ medical knowledge by the clinical professions
[13, 14], this lack of understanding of their own bodies can lead to communication issues
about medical diagnoses and treatment procedures leaving patients, and their families con-
fused and anxious [15], impacting patient care.
As previously stated, part of the Anatomy Nights event includes dissection of non-human
tissue to demonstrate key, homologous anatomical structures. The absence of dissection of
human tissue is due to moral and legal considerations around what constitutes appropriate use
of donated human tissue, and this can be a barrier to anatomists being able to take part and
host their own Anatomy Nights events. These considerations are exemplified by the existence
of national legislation within Anatomy Nights’ host countries. In the United Kingdom (UK),
the use and display of human tissue are regulated by the Human Tissue Act 2004 and the
Human Tissue (Scotland) Act 2006. These acts were preceded by The Anatomy Act [16] in
1832, which allowed cadaveric specimens to be used only by approved medical schools. This
restriction still exists, and it is this caveat that stops the British public from accessing education
of human anatomy through viewing anatomical dissection by an expert. Similarly, in the
United States of America (USA), the Uniform Anatomical Gift Act legislation was adopted by
26 of 50 states in 1968 [17], and revisions in 1987 and 2006 have resulted in 48 states adopting
uniform laws related to body and organ donation [18].
Regulation of anatomical specimens for public display, combined with fewer opportunities
for the public to engage around this material with qualified anatomists, has resulted in a lack
of public access to anatomy. Therefore, non-academic demonstrations of anatomical prepara-
tions without proper educational context, such as the traveling plastination exhibitions, have
become popular throughout Europe and the United States. The spate of non-contextualized
information (or even misinformation) can be confusing, leaving the public with a misunder-
standing of anatomical structure and function similar to that seen in patients who forego
PLOS ONE | https://doi.org/10.1371/journal.pone.0267550 June 9, 2022
2 / 12
PLOS ONE Anatomy Nights events and public knowledge of anatomy
medical advice in favor of internet self-diagnosis [19]. While some establishments have tried to
address this [20], no data has been collected as to the effect of the courses, and the high cost of
such events is a limiting factor for engagement. To minimize the impact of financial barriers,
Anatomy Nights events are accessible to the public for a nominal ticket fee (around the price
of a coffee in each host’s country). The majority of the fee is donated to a relevant charity, and
the rest used for funding the next events.
Here we present our evaluation of a brain-focused Anatomy Nights event and its impact on
increasing public knowledge of anatomy. Specifically, the research aims were to determine if
the Anatomy Nights events had a positive impact on the public audiences’ knowledge of brain
anatomy. In particular, the goals were to establish which anatomical features of the brain are
common knowledge, where there is a deficit, and whether audiences could take this knowledge
and apply it to neuroanatomical concepts.
Materials and methods
Event format
Data were collected at four different venues at events during October 2018—October 2019
(Dundee, Edinburgh, Hull, UK, and Indianapolis, USA). Each event was hosted by a different
anatomist using a template presentation covering the anatomical knowledge that was tested.
The venues were all public settings, specifically in venues not associated with universities to
encourage attendance from anyone who was interested but may not feel comfortable in a uni-
versity setting [6]. The events were advertised as a short talk about the brain followed by a dis-
section of a lamb or pig’s brain.
Each of the Anatomy Nights events followed a standard format. This started with a 30-min-
ute talk on the anatomy of the brain by the hosting anatomist/s, including where it is; menin-
ges; white and grey matter; lobes and the cerebellum; functional areas and the homunculus;
decussation of fibers; basic blood supply; CSF and the ventricles. Following this talk, a dissec-
tion of a non-human brain was conducted. To ensure all members of the audience could see
this clearly, a camera-projector rig was set up. In the dissection, the audiences were familiar-
ized with the external structure, including lobes and brainstem. The brain was then cut into
sections to demonstrate white vs. grey matter, points of decussation, and the ventricles. Any
significant deviations from human anatomy (e.g., the olfactory lobe seen on pigs brains) were
highlighted to the audience.
Data collection
Everyone in attendance aged 16 and over was invited to participate in research surveys
designed to assess baseline and acquired knowledge as a result of participation in Anatomy
Night. Data were collected through two separate instruments—before and after the events—
which enabled assessment of the existing knowledge gap and whether the learning program
was effective. For those attendees to choose to participate in the research surveys, each received
a participant information sheet alongside the survey for them to retain for their records. At the
top of the surveys, participants were notified that submission of the surveys at the conclusion
of the event constituted informed consent. This process of gaining informed consent from par-
ticipants aged 16 years and over was explicit in the ethics applications to the institutional com-
mittees, who thus waived the need for parental consent for participants aged 16 and 17 years
old. The study was granted ethical approval by the Hull York Medical School Ethics Commit-
tee (reference number 17 26) and was granted exempt status by the Indiana University School
of Medicine Institutional Review Board (reference number 1901221393).
PLOS ONE | https://doi.org/10.1371/journal.pone.0267550 June 9, 2022
3 / 12
PLOS ONE Anatomy Nights events and public knowledge of anatomy
As the audience entered the event venue, they were presented with a test sheet (S1 File) that
tested their knowledge of the location of 7 different brain regions and structures, and a further
question on stroke tested whether they could extrapolate from a damaged brain area to the
physical consequences in the body with four options to select from. The maximum score an
individual could receive was eight from eight questions. This sheet also asked for some demo-
graphic details. This was used to identify the composition of the audience attracted to the
event and, for some characteristics, to explore whether this had an effect on performance in
the test. Age and gender were not analyzed in the context of performance on the test, as it was
not deemed appropriate to use data on protected characteristics for the purpose of this study.
After the talk and dissection, the participants completed a post-event test which asked the
same questions. The pre-and post-test sheet answers keys were set up with different answer
coding to ensure that anyone who simply copied between tests, rather than engaging with the
activity, would be clear for data analysis and could be removed from the dataset. The pre-and
post-event tests were on either side of a single sheet of paper, allowing the individual change in
performance to be determined. Participation in the study aspect of the Anatomy Nights event
was completely anonymous and voluntary, and completed survey sheets were given a random
Participant ID to compare individuals’ change in scores.
Data analysis
We asked two general questions about the data. First, do scores differ between testing pre- vs.
post- the Anatomy Nights talks and dissection? We addressed this question using a generalized
linear model (R’s glm function; R version 4.1.0 [21]) that assumed a binomial distribution and
implemented a logit link function. Here, and throughout, we present back-transformed (gen-
eralized) means, standard errors and 95% confidence intervals (hereafter, 95% CI) for each
treatment level. We present the magnitude of the effect of pre- vs. post- testing on scores as an
odds ratio, (with standard error and 95% CI), calculated with R’s ‘emmeans’ package (V 1.6.1).
Throughout, odds ratio estimates (and associated error) are also based on generalized means.
Throughout, 95% confidence intervals for effect sizes are not adjusted for multiple
comparisons.
Second, we investigated whether the participants’ Academic Qualification (school, under-
graduate, postgraduate, none), employment in the healthcare sector (Yes, No), and location
(Dundee, Hull, Edinburgh, Indianapolis) affected test scores. Specifically, we tested whether
each of these factors affected i) the magnitude of the change in score between pre- and post-
sessions (i.e., quantified by an interaction term between test timing and the focal factor), and
ii) the average score attained. We implemented three models for each of these three factors
(i.e., education, employment type, location). Model type (a) included a term for the test timing
(pre- vs. post-), a term for the given factor (e.g., location), and a term accounting for the inter-
action between the given factor and timing; model type (b) was identical to type (a), but lacked
the interaction term; model type (c) included a term for timing, only. We compared models
using likelihood ratio tests to determine whether a focal factor influenced test scores. For
example, for a given factor type (e.g., employment type), we compared model type (a) vs. (b) to
determine whether the interaction term affected test scores and (b) vs. (c) to test whether the
focal factor affected average scores. Note that two participants had no Academic Qualification
(category ‘none’). We excluded these two participants from the analysis of Academic Qualifi-
cation because the sample size for this group (‘none’) was too small to effectively compare it to
the remaining three Academic Qualification categories.
We used the glmmTMB function [22] (Version 1.1.2) to model test performance with gen-
eralized mixed-effects models, implementing a binomial response distribution and logit-link
PLOS ONE | https://doi.org/10.1371/journal.pone.0267550 June 9, 2022
4 / 12
PLOS ONE Anatomy Nights events and public knowledge of anatomy
function. Models included the fixed effects terms described, above (models a—c), Participant
ID as a random effect, and a second random effect (‘units’) that modeled overdispersion.
As described in Results, likelihood ratio tests revealed little evidence for interactions
between test timing and each of the three focal factors (Academic Qualification, Employment
type, Location). Therefore, we calculated mean test performance (averaging over test timing)
and effect size for each level of each focal factor using models that lack an interaction (i.e.,
model type (b)). S1 Table presents these estimates and effect sizes (again) on the latent scales.
We used binomial tests to determine whether the probability of correctly answering the
question regarding stroke differed significantly from the random expectation of 25%; ‘pre-’
and ‘post-’ data were analyzed separately.
We do not test whether performance changed between ‘pre-’ and ‘post-’ time periods for
individual questions because exceedingly high test performance led to poor performance by
Generalized Linear Models. In particular, all participants answered Question 3 correctly post-
lecture and dissection (i.e., 91 / 91 answered correctly), which led to nonsensical effect size esti-
mates (and 95% CI’s) for this question. Therefore, we report results for individual questions
descriptively.
All data and R scripts are available (S1 Data and S2 File) to allow readers to replicate our
analyses.
Results
Audience demographics
All individuals attending the Anatomy Night event were invited to participate in the research.
From those in attendance, a total of 102 participants sheets were collected. Nine of these were
removed from the analysis because eight of them had not completed both sides of the test
sheet, and one of them had copied the answers from the pre-test; two were removed due to
being too young (<16 years) to meet requirements of ethical approval. The demographics of
the 91 participants who completed the test sheets are shown in Table 1.
Our analyses revealed no evidence that any of the three factors (academic qualification,
experience working in healthcare, Location) affected the extent to which scores changed
between pre- and post-educational activity (i.e., test of interaction term; likelihood ratio tests,
all p-values > 0.26).
Similarly, we found no evidence that academic qualification and experience working in
healthcare affected the average test score (likelihood ratio tests, all p-values > 0.23; see Tables 2
and S1 for effect size estimates). In contrast, our models revealed strong evidence that mean
test scores differed among locations (likelihood ratio test, p = 0.0059). Examination of effect
sizes and their 95% CI’s (Tables 2 and S1) suggests that Edinburgh and USA both tended to
have higher scores than Dundee and Hull, but little difference occurred between Edinburgh
and USA and between Dundee and Hull. Note that our models of Academic Qualification,
experience working in healthcare, and Location also analyze effects of test timing on scores:
these results (p-values, effect size with 95% CI’s; not shown) are consistent with the results pre-
sented in “Anatomical Knowledge”, below.
Anatomical knowledge
Overall, our analyses reveal strong evidence that Anatomy Nights events increased overall test
scores (Generalized linear model; z value = -9.325, p < 2e-16) from (generalized mean
proportion ± SE) 0.639 ± 0.018 (95% CI: 0.603, 0.673) prior to the activity to 0.857 ± 0.013
(95% CI: 0.830, 0.881) (Fig 1). This result corresponds to an odds ratio (post / pre) of
3.39 ± 0.445 (95% CI: 2.62, 4.39). (These results correspond to estimated mean ± SE of (pre-)
PLOS ONE | https://doi.org/10.1371/journal.pone.0267550 June 9, 2022
5 / 12
PLOS ONE Table 1. Audience demographics of the Anatomy Nights brain event from 2018–19.
Demographic
Location of event
Age
Gender
Highest educational qualification
Works in healthcare
https://doi.org/10.1371/journal.pone.0267550.t001
Category
Dundee, UK
Edinburgh, UK
Hull, UK
Indianapolis, US
16–17 years
18–34
35–50
>50
Man
Woman
Gender diverse
Prefer not to say
School
Undergraduate
Postgraduate
None
Yes
No
Anatomy Nights events and public knowledge of anatomy
% participants (n = 91)
26.4
26.4
22.0
25.3
2.2
72.5
19.8
5.5
34.1
64.8
0
1.1
11.0
38.4
49.4
1.1
28.6
71.4
0.570 ± 0.077 (95% CI’s: 0.419, 0.721) and (post-) 1.792 ± 0.106 (95% CI’s 1.58, 2.00) on the
logit scale, and to an estimated effect size of (post—pre) 1.22 ± 0.13 (95% CI’s 0.96, 1.48) on
the log odds ratio scale.)
Most questions contributed to the overall improvement between pre- and post-test scores.
Fig 2, however, highlights three questions that likely contributed little to overall score improve-
ment. Specifically, pre-test scores were very high for questions 1 and 3, leaving little scope for
improvement, whereas scores improved and declined to similar degrees for question 8 (scores
improved, declined, and remained unchanged for 12, 20, and 59 participants, respectively).
Question 8, regarding stroke (S1 File) had few correct answers; of 91 participants, 39 and 31
correctly answered this question at ‘pre-’ and ‘post-’ stages, respectively (Fig 2, Question 8).
However, binomial tests revealed moderate to strong evidence that participants tended to cor-
rectly answer this question more often than expected due to chance (25%) at both the ‘pre’
(p = 0.00022) and ‘post’ (p = 0.05238) stages.
Discussion
Audience demographics
Overall, the results demonstrate the Anatomy Nights brain event increased the audience’s
knowledge of brain anatomy (Fig 1). The majority of the audiences were aged 18–34 (72.5%),
were women (64.8%), had a university education (undergraduate 38.4%, postgraduate 49.4%),
and did not work in healthcare (71.4%) (Table 1). The predominance of women and university
graduates in the audience aligns with research that shows women and people with higher edu-
cation are more likely to actively seek out information relating to their health [23]. Our audi-
ence proportions are also similar to those of Science Cafe´ events; however, Anatomy Nights
reaches a younger age group than such events (18–34 vs. 40+) [24]. This younger demographic
more closely aligns to that seen at annual Pint of Science events [25] and shows that Anatomy
Nights joins a growing number of events reaching a younger audience.
PLOS ONE | https://doi.org/10.1371/journal.pone.0267550 June 9, 2022
6 / 12
PLOS ONE Table 2. Estimates from Generalized Linear Mixed Models for effects of academic qualfication, empoyment in healthcare, and location on test performance. All
results based on generalized means, averaged over effect of test timing (pre-, post-). 95% CI’s for contrasts are not adjusted for multiple comparisons.
Anatomy Nights events and public knowledge of anatomy
Academic Qualification
Level
Post
School
Under
Contrast
post / school
post / under
school / under
Work in Healthcare
Level
No
Yes
Contrast
No / Yes
Location
Level
Dundee
Edinburgh
Hull
USA
Contrast
Dundee / Edin
Dundee / Hull
Dundee / USA
Edin / Hull
Edin / USA
Hull / USA
Proportion answers correct
0.794
0.797
0.774
Odds ratio
0.982
1.127
1.148
Proportion answers correct
0.773
0.816
Odds ratio
0.766
Proportion answers correct
0.726
0.837
0.715
0.835
Odds ratio
0.516
1.060
0.525
2.054
1.018
0.495
https://doi.org/10.1371/journal.pone.0267550.t002
SE
0.0240
0.0492
0.0287
SE
0.329
0.242
0.394
SE
0.0212
0.0291
SE
0.172
SE
0.0354
0.0261
0.0395
0.0272
SE
0.133
0.277
0.137
0.554
0.275
0.135
95% CI
0.743, 0.837
0.683, 0.877
0.712, 0.825
df
174
174
174
95% CI
0.729, 0.812
0.752, 0.867
df
177
95% CI
0.651, 0.790
0.779, 0.883
0.631, 0.786
0.774, 0.882
df
175
175
175
175
175
175
95% CI
0.507, 1.90
0.737, 1.72
0.583, 2.26
95% CI
0.493, 1.19
95% CI
0.310, 0.859
0.633, 1.774
0.313, 0.880
1.206, 3.499
0.598, 1.733
0.289, 0.849
As well as analysis of audience demographics, data on educational background and employ-
ment in the healthcare sector were used to ascertain if these had an effect on baseline knowl-
edge and performance after the event. Interestingly, neither of these had a significant impact.
There was, however, a significant difference between locations; reasons for this may be due to
differences in the presentation styles of the hosts and/or the demographics of the audiences
between locations. Regarding educational background, information on the field studied at the
university level was not collected, so no conclusion on previous anatomical knowledge could
be derived. However, it is notable that this group’s performance after an educational activity
was no different from those without a university qualification.
There was no evident difference in baseline and improvement in the group that works in
the healthcare sector, which given the presence of anatomy in most health professions training,
is surprising. This may be for two reasons: the vast differences in levels of anatomy education
in allied health professions [26], and individuals not directly involved in patient care (e.g., hos-
pital administrators) could reasonably have said they work in the healthcare sector. Without
further detail on the attendees’ role in healthcare, the effect of prior exposure to anatomical
education cannot be ascertained.
PLOS ONE | https://doi.org/10.1371/journal.pone.0267550 June 9, 2022
7 / 12
PLOS ONE Anatomy Nights events and public knowledge of anatomy
Fig 1. Overall test scores. Test scores obtained (out of eight) by study participants (n = 91) before (“Pre”) and after
(“Post”) experiencing the educational activity component of Anatomy Nights. Lines connect pre- and post-test scores
for individual participants.
https://doi.org/10.1371/journal.pone.0267550.g001
Public anatomical knowledge
From the results, it could be said that the general public has a reasonable baseline knowledge
with regards to brain anatomy with an average score of 65% in the pre-test. However, while
Fig 2. Test scores for individual questions. Test scores obtained for each question by study participants (n = 91) before (“pre”) and after (“post”) experiencing
the educational activity component of Anatomy Nights. Lines connect pre- and post-scores for individual participants.
https://doi.org/10.1371/journal.pone.0267550.g002
PLOS ONE | https://doi.org/10.1371/journal.pone.0267550 June 9, 2022
8 / 12
PLOS ONE Anatomy Nights events and public knowledge of anatomy
this is by no means a “fail” grade on the test, the performance in individual questions high-
lights the role that anatomical language may be playing in the differences between perfor-
mance on questions. Indeed, all the questions except for 1, 3, and 8 appear to demonstrate
improvement after the Anatomy Nights event (Fig 2). Questions 1 and 3 (S1 File) asked the
audience to identify the “brain stem” and “frontal lobe” and had little room for improvement,
with 98% and 94% answering the questions correctly in the pre-test, respectively. Compared to
the other questions, 1 and 3 included words that are familiar to a non-scientific audience,
namely “stem” and “frontal”. The anatomical structures themselves reflect the normal defini-
tions of these words: the brain stem descends from under the brain, like a plant’s stem is under
the flower; the frontal lobe is at the front of the brain. While it cannot be stated with certainty
that it is the public’s familiarity with the words rather than already knowing what these struc-
tures are, these were the only questions without anatomical jargon in their name. Within all
science communication, jargon remains a significant barrier between scientists and the public
[15], and anatomical sciences are no exception. These two questions account for 25% of the
test, and if they were removed, then baseline knowledge of the general public audience would
be verging on a 50% pass. It can be concluded from this that the brain stem and frontal lobe
form part of the common knowledge of the general public; however, all other structures asked
for in the test are not.
Application of anatomical knowledge
Question 8 also showed no apparent change in performance between pre- and post-tests; how-
ever, this was a more complex question than the other seven. This question required the appli-
cation of knowledge. An area on the brain was identified with an “X” and they were asked
which area of the body would be affected by damage there. They were given four options with
a combination of right or left and upper or lower limbs. To correctly answer this question, the
participants needed to identify the side of the brain shown in the diagram, apply knowledge of
decussation, and then overlay the map of the motor homunculus onto the diagram to deter-
mine if the upper or lower limb would be affected.
While performance on Question 8 appeared to change little between test periods (Fig 2),
the audience tended to perform marginally better than would be expected by chance at both
periods. It is interesting to note that the participants’ responses to this question did not remain
static despite similar overall performance. It can be seen in Fig 2, Question 8, that roughly sim-
ilar proportions of participants either changed their answer or kept it the same between pre-
and post-test, including changing from a correct answer to an incorrect one. The content of
the Anatomy Nights event challenged the cognitive capacity of the audience, and Question 8
demonstrates the limits of that. The presentation of new information and the requirement to
first select the appropriate content and then integrate different aspects of it to reach the correct
answer may exceed a reasonable expectation of cognitive load for an audience presented with
extensive volume of novel information (intrinsic load) in a short timeframe (extrinsic load)
[27]. Indeed, the improvement in the overall test score demonstrates a net gain in knowledge,
but most participants were unable to apply the new knowledge to different neuroanatomical
contexts.
Confounding factors
There are several factors, which could not be fully controlled for, and should be considered in
the interpretation of the results presented here. While every effort was made to standardize the
event across the four locations, including the provision of a template presentation, each pre-
senter inevitably added their own style of public engagement. This is inherent in the design of
PLOS ONE | https://doi.org/10.1371/journal.pone.0267550 June 9, 2022
9 / 12
PLOS ONE Anatomy Nights events and public knowledge of anatomy
the Anatomy Nights format, and each anatomist should continue to be encouraged to develop
their own style so the event can educate and develop faculty at the same time as educating the
public. However, this difference in delivery could not be factored into the analysis.
Further factors, including delivery of the event in the evening, variable alcohol consumption at
the venues, attendees potentially working with others or using personal smart devices to search
for answers, filling in the pre-test during or after the talk and dissection, are all possible confound-
ers and will have played some part in the performance on the pre- and post-event tests. Regard-
less, an increase in test performance after the event was seen and each of these factors would be
considered likely to diminish this effect. Therefore, we have confidence in the positive educational
influence of the Anatomy Nights event on short-term anatomical knowledge.
Finally, it is also not possible to ascertain whether the audience’s new knowledge was
derived from the presentation or from the dissection. It is probably a combination of the two,
with the presentation being the primary source of information gain and the dissection acting
as a way to consolidate this information with a 3D structure. The act of dissection is a novelty
in science communication, and this distinctive way of engaging the audience will have created
a learning landmark [28] for some audience members. Irrespective of whether the lecture or
the dissection provided the more significant component of knowledge acquisition, the promise
of the dissection component inevitably attracted some attendees and is, therefore, a crucial
component of a successful Anatomy Nights event.
Conclusion
The Anatomy Nights brain event comprising a short presentation and dissection led by an
expert anatomist significantly increases the public’s knowledge of the anatomy of the brain
irrespective of location, educational background, and employment in the healthcare sector.
However, we found no evidence that participation in Anatomy Nights improved participants’
ability to apply this knowledge to neuroanatomical contexts (e.g., stroke).
Baseline anatomical knowledge most strongly aligns with structures that have a standard
English name. Future events should cement this knowledge and introduce more anatomical
structures to become common knowledge. A follow-up of audience members after the event
could also be used to determine if acquired knowledge is retained over time.
Supporting information
S1 File. Pre- and post-test sheet. The double-sided test that each participant completed before
and after the event.
(PDF)
S2 File. R Scripts from statistical analyses of the results.
(TXT)
S1 Table. Estimates from Generalized Linear Mixed Models for effects of academic qualifi-
cation, employment in healthcare, and location on test performance.
(DOCX)
S1 Data. Raw data from the pre- and post-tests.
(CSV)
Acknowledgments
The authors would like to acknowledge all the faculty and volunteers involved in the events,
the venues and management, and the attendees without whose participation, the analysis
PLOS ONE | https://doi.org/10.1371/journal.pone.0267550 June 9, 2022
10 / 12
PLOS ONE Anatomy Nights events and public knowledge of anatomy
would not have been possible. The event in Indianapolis was organized by the Science Out-
reach Community of Indiana University School of Medicine (SOCI) graduate student group.
The event in Hull was organized by the Hull and East Yorkshire (HEY) Science group.
Author Contributions
Conceptualization: Katherine A. Sanders, Janet A. C. Philp.
Data curation: Janet A. C. Philp, Andrew S. Cale, Claire L. Cunningham, Jason M. Organ.
Formal analysis: Katherine A. Sanders, Janet A. C. Philp, Crispin Y. Jordan, Jason M. Organ.
Investigation: Katherine A. Sanders.
Methodology: Katherine A. Sanders, Janet A. C. Philp, Crispin Y. Jordan.
Project administration: Katherine A. Sanders, Janet A. C. Philp, Andrew S. Cale, Claire L.
Cunningham, Jason M. Organ.
Writing – original draft: Katherine A. Sanders, Janet A. C. Philp, Crispin Y. Jordan, Andrew
S. Cale, Claire L. Cunningham, Jason M. Organ.
Writing – review & editing: Katherine A. Sanders, Janet A. C. Philp, Andrew S. Cale, Claire L.
Cunningham, Jason M. Organ.
References
1. Mavrodi A, Paraskevas G. Mondino de Luzzi: a luminous figure in the darkness of the Middle Ages.
Croat Med J. 2014; 55(1):50–3. Epub 2014/03/01. https://doi.org/10.3325/cmj.2014.55.50 PMID:
24577827; PubMed Central PMCID: PMC3944418.
2.
Zampieri F, ElMaghawry M, Zanatta A, Thiene G. Andreas Vesalius: Celebrating 500 years of dissect-
ing nature. Glob Cardiol Sci Pract. 2015; 2015(5):66. Epub 2017/01/28. https://doi.org/10.5339/gcsp.
2015.66 PMID: 28127546; PubMed Central PMCID: PMC4762440.
3. Ghosh SK. Human cadaveric dissection: a historical account from ancient Greece to the modern era.
Anat Cell Biol. 2015; 48(3):153–69. Epub 2015/09/30. https://doi.org/10.5115/acb.2015.48.3.153
PMID: 26417475; PubMed Central PMCID: PMC4582158.
4.
Fonseca GR, Finn GM. The impact of Body Worlds on adult visitors’ knowledge on human anatomy: A
preliminary study. Clin Anat. 2016; 29(4):439–45. Epub 2016/01/21. https://doi.org/10.1002/ca.22692
PMID: 26789643.
5. Walter T. Body Worlds: clinical detachment and anatomical awe. Sociol Health Illn. 2004; 26(4):464–88.
Epub 2004/07/23. https://doi.org/10.1111/j.0141-9889.2004.00401.x PMID: 15268703.
6.
Tan SZK, Perucho JAU. Bringing Science to Bars: A Strategy for Effective Science Communication.
Science Communication. 2018; 40(6):819–26. https://doi.org/10.1177/1075547018808298.
7. Rosenbaum AJ, Pauze D, Pauze D, Robak N, Zade R, Mulligan M, et al. Health Literacy in Patients
Seeking Orthopaedic Care: Results of the Literacy in Musculoskeletal Problems (LIMP) Project. Iowa
Orthop J. 2015; 35:187–92. Epub 2015/09/12. PMID: 26361464; PubMed Central PMCID:
PMC4492130.
8. Berkman ND, Sheridan SL, Donahue KE, Halpern DJ, Crotty K. Low health literacy and health out-
comes: an updated systematic review. Ann Intern Med. 2011; 155(2):97–107. Epub 2011/07/20. https://
doi.org/10.7326/0003-4819-155-2-201107190-00005 PMID: 21768583.
9. Seshadri S, Beiser A, Kelly-Hayes M, Kase CS, Au R, Kannel WB, et al. The lifetime risk of stroke: esti-
mates from the Framingham Study. Stroke. 2006; 37(2):345–50. Epub 2006/01/07. https://doi.org/10.
1161/01.STR.0000199613.38911.b2 PMID: 16397184.
10. Benjamin EJ, Blaha MJ, Chiuve SE, Cushman M, Das SR, Deo R, et al. Heart Disease and Stroke Sta-
tistics-2017 Update: A Report From the American Heart Association. Circulation. 2017; 135(10):e146–
e603. Epub 2017/01/27. https://doi.org/10.1161/CIR.0000000000000485 PMID: 28122885; PubMed
Central PMCID: PMC5408160.
11.
Taylor AM, Diggle P, Wessels Q. What do the public know about anatomy? Anatomy education to the
public and the implications. Anat Sci Educ. 2018; 11(2):117–23. Epub 2017/11/08. https://doi.org/10.
1002/ase.1746 PMID: 29112336.
PLOS ONE | https://doi.org/10.1371/journal.pone.0267550 June 9, 2022
11 / 12
PLOS ONE Anatomy Nights events and public knowledge of anatomy
12.
Taylor AM. Anatomy Education to the Public. In: Chan LK, Pawlina W, editors. Teaching Anatomy: A
Practical Guide. Cham: Springer International Publishing; 2020. p. 73–84.
13. Kelly PA, Haidet P. Physician overestimation of patient literacy: a potential source of health care dispari-
ties. Patient Educ Couns. 2007; 66(1):119–22. Epub 2006/12/05. https://doi.org/10.1016/j.pec.2006.10.
007 PMID: 17140758.
14. Dickens C, Lambert BL, Cromwell T, Piano MR. Nurse overestimation of patients’ health literacy. J
Health Commun. 2013; 18 Suppl 1:62–9. Epub 2013/10/08. https://doi.org/10.1080/10810730.2013.
825670 PMID: 24093346; PubMed Central PMCID: PMC3814908.
15. Hoffmann-Longtin K, Organ JM, Helphinstine JV, Reinoso DR, Morgan ZS, Weinstein E. Teaching
advocacy communication to pediatric residents: the efficacy of applied improvisational theater (AIT) as
an instructional tool. Communication Education. 2018; 67(4):438–59. https://doi.org/10.1080/
03634523.2018.1503314.
16.
The National Archives. The Anatomy Act 1832 [19 June 2020]. Available from: https://www.
nationalarchives.gov.uk/education/resources/body-snatchers/source-four-the-anatomy-act-1832/.
17. Uniform Law Commission. Anatomical Gift Act [19 June 2020]. Available from: https://www.
uniformlaws.org/committees/community-home?CommunityKey=015e18ad-4806-4dff-b011-
8e1ebc0d1d0f.
18. Verheijde JL, Rady MY, McGregor JL. The United States Revised Uniform Anatomical Gift Act (2006):
new challenges to balancing patient rights and physician responsibilities. Philos Ethics Humanit Med.
2007; 2:19. Epub 2007/09/14. https://doi.org/10.1186/1747-5341-2-19 PMID: 17850664; PubMed Cen-
tral PMCID: PMC2001294.
19. Au A. Why Physician Guidance Matters: A Night of Neuralgia, Meningitis, and WebMD. Ann Fam Med.
2019; 17(5):462–4. Epub 2019/09/11. https://doi.org/10.1370/afm.2414 PMID: 31501210; PubMed
Central PMCID: PMC7032901.
20. Gillingwater TH, Findlater GS. Anatomy: back in the public spotlight. Lancet. 2015; 385(9980):1825.
Epub 2015/05/20. https://doi.org/10.1016/S0140-6736(15)60916-3 PMID: 25987145.
21. R Core Team. R: A language and environment for statistical computing. Vienna, Austria: URL http://
www.R-project.org/; 2014.
22. Brooks M, Kristensen K, van Benthem K, Magnusson A, Berg C, Nielsen A, et al. glmmTMB Balances
Speed and Flexibility Among Packages for Zero-inflated Generalized Linear Mixed Modeling. R Journal.
2017; 9(2):378–400. https://doi.org/10.32614/RJ-2017-066.
23.
Tu HT, Cohen GR. Striking jump in consumers seeking health care information. Track Rep. 2008;
(20):1–8. Epub 2008/09/06. PMID: 18770913.
24. Ahmed S, DeFino MC, Connors ER, Kissack A, Franco Z. Science cafes: engaging scientists and com-
munity through health and science dialogue. Clin Transl Sci. 2014; 7(3):196–200. Epub 2014/04/11.
https://doi.org/10.1111/cts.12153 PMID: 24716626; PubMed Central PMCID: PMC4410806.
25. Pint of Science. Pint of Science 2021 [cited 2021 10 May]. Available from: https://pintofscience.co.uk/.
26. Chan LK, Pawlina W. Teaching anatomy: A practical guide. 2 ed. Switzerland: Springer International
Publishing; 2020. 546 p.
27. Sweller J, van Merrienboer JJG, Paas FGWC. Cognitive Architecture and Instructional Design. Educa-
tional Psychology Review. 1998; 10:251–96. https://doi.org/10.1023/A:1022193728205.
28.
Finn GM, McLachlan JC. A qualitative study of student responses to body painting. Anat Sci Educ.
2010; 3(1):33–8. Epub 2009/11/18. https://doi.org/10.1002/ase.119 PMID: 19918974.
PLOS ONE | https://doi.org/10.1371/journal.pone.0267550 June 9, 2022
12 / 12
PLOS ONE |
10.1371_journal.pone.0262844 | RESEARCH ARTICLE
Improvements following multimodal pelvic
floor physical therapy in gynecological cancer
survivors suffering from pain during sexual
intercourse: Results from a one-year follow-
up mixed-method study
Marie-Pierre CyrID
Paul Bessette2,5, Annick Pina6,7, Walter Henry GotliebID
He´ lène Mayrand7,10, Me´ lanie Morin1,2*
1,2, Rosalie Dostie1,2, Chantal Camden1,2, Chantale DumoulinID
3,4,
8,9, Korine Lapointe-Milot2,5, Marie-
1 Faculty of Medicine and Health Sciences, School of Rehabilitation, University of Sherbrooke, Sherbrooke,
Quebec, Canada, 2 Research Center of the Centre Hospitalier Universitaire de Sherbrooke, Sherbrooke,
Quebec, Canada, 3 Faculty of Medicine, School of Rehabilitation, University of Montreal, Montreal, Quebec,
Canada, 4 Research Center of the Institut Universitaire de Ge´ riatrie de Montre´al, Montreal, Quebec, Canada,
5 Faculty of Medicine and Health Sciences, Division of Gynecologic Oncology, Department of Obstetrics and
Gynecology, University of Sherbrooke, Sherbrooke, Quebec, Canada, 6 Faculty of Medicine, Division of
Gynecologic Oncology, Department of Obstetrics and Gynecology, University of Montreal, Montreal, Quebec,
Canada, 7 Research Center of the Centre Hospitalier de l’Universite´ de Montre´al, Montreal, Quebec,
Canada, 8 Faculty of Medicine, Division of Gynecologic Oncology, Department of Obstetrics and
Gynecology, McGill University, Montreal, Quebec, Canada, 9 Lady Davis Institute of the Jewish General
Hospital, Montreal, Quebec, Canada, 10 Faculty of Medicine, Departments of Obstetrics and Gynecology
and Social and Preventive Medicine, University of Montreal, Montreal, Quebec, Canada
* melanie.m.morin@usherbrooke.ca
Abstract
Background
A large proportion of gynecological cancer survivors suffer from pain during sexual inter-
course, also known as dyspareunia. Following a multimodal pelvic floor physical therapy
(PFPT) treatment, a reduction in pain and improvement in psychosexual outcomes were
found in the short term, but no study thus far has examined whether these changes are sus-
tained over time.
Purpose
To examine the improvements in pain, sexual functioning, sexual distress, body image con-
cerns, pain anxiety, pain catastrophizing, painful intercourse self-efficacy, depressive symp-
toms and pelvic floor disorder symptoms in gynecological cancer survivors with dyspareunia
after PFPT, and to explore women’s perceptions of treatment effects at one-year follow-up.
Methods
This mixed-method study included 31 gynecological cancer survivors affected by dyspareu-
nia. The women completed a 12-week PFPT treatment comprising education, manual
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: Cyr M-P, Dostie R, Camden C, Dumoulin
C, Bessette P, Pina A, et al. (2022) Improvements
following multimodal pelvic floor physical therapy
in gynecological cancer survivors suffering from
pain during sexual intercourse: Results from a one-
year follow-up mixed-method study. PLoS ONE
17(1): e0262844. https://doi.org/10.1371/journal.
pone.0262844
Editor: Diego Raimondo, Dipartimento di Scienze
Mediche e Chirugiche (DIMEC), Orsola Hospital,
ITALY
Received: September 10, 2021
Accepted: January 6, 2022
Published: January 25, 2022
Copyright: © 2022 Cyr et al. This is an open access
article distributed under the terms of the Creative
Commons Attribution License, which permits
unrestricted use, distribution, and reproduction in
any medium, provided the original author and
source are credited.
Data Availability Statement: All relevant data are
within the paper.
Funding: The Quebec Network for Research on
Aging funded the current study. The Fonds de
recherche du Que´bec – Sante´ granted a
scholarship to Marie-Pierre Cyr and salary awards
to Me´lanie Morin, Chantal Camden and Marie-
He´lène Mayrand. The Canadian Research Chair Tier
PLOS ONE | https://doi.org/10.1371/journal.pone.0262844 January 25, 2022
1 / 20
PLOS ONE II on Urogynecological Health and Aging supported
Chantale Dumoulin. The laboratory infrastructures
were funded by the Canadian Foundation for
Innovation. The funders had no role in study
design, data collection and analysis, decision to
publish, or preparation of the manuscript.
Competing interests: The authors have declared
that no competing interests exist.
Multimodal pelvic floor physical therapy for cancer survivors suffering from pain during sexual intercourse
therapy and pelvic floor muscle exercises. Quantitative data were collected using validated
questionnaires at baseline, post-treatment and one-year follow-up. As for qualitative data,
semi-structured interviews were conducted at one-year follow-up to better understand wom-
en’s perception and experience of treatment effects.
Results
Significant improvements were found from baseline to one-year follow-up on all quantitative
outcomes (P � 0.028). Moreover, no changes were found from post-treatment to one-year
follow-up, supporting that the improvements were sustained at follow-up. Qualitative data
highlighted that reduction in pain, improvement in sexual functioning and reduction in urinary
symptoms were the most meaningful effects perceived by participants. Women expressed
that these effects resulted from positive biological, psychological and social changes attrib-
utable to multimodal PFPT. Adherence was also perceived to influence treatment
outcomes.
Conclusions
Findings suggest that the short-term improvements following multimodal PFPT are sus-
tained and meaningful for gynecological cancer survivors with dyspareunia one year after
treatment.
Introduction
An increasing number of women live with the deleterious, long-term consequences of cancer
[1,2]. Alongside urinary incontinence, chronic pain during sexual intercourse, also known as
dyspareunia, is one of the most common sexual health issues, affecting more than half of gyne-
cological cancer survivors [3,4]. Dyspareunia is recognized as resulting from the complex
interaction of anatomical, physiological, psychological and relationship factors related to can-
cer and oncological treatments [5], in line with the biopsychosocial model [6,7]. Vaginal steno-
sis, impaired tissue flexibility, heightened pelvic floor muscle tone and contractility
impairments as well as vaginal dryness [5,8] may contribute to experiencing pain during inter-
course. These biological factors interplay with pain anxiety (i.e., fear of pain), pain catastro-
phizing [9] and low pain self-efficacy [10], thereby intensifying the pain [11]. Gynecological
cancer survivors are also at risk of depressive symptoms and body image concerns [12,13],
which may disturb how they perceive themselves as women [14–16]. These pain and psycho-
logical factors may contribute to sexual distress [17,18]. Moreover, women who have been
treated for gynecological cancer are often affected by other sexual dysfunctions such as loss of
libido or sexual desire [17]. All this can lead to relationship difficulties [12,13], disrupting their
quality of life [19–21].
Despite the high prevalence of dyspareunia, there are limited treatment options supported
by empirical evidence. Clinical survivorship guidelines suggest multimodal pelvic floor physi-
cal therapy (PFPT) as a nonhormonal, non-pharmacological and non-invasive first-line treat-
ment to alleviate dyspareunia in cancer survivors [22–24]. Through psychosexual education,
manual therapy techniques and pelvic floor muscle exercises, PFPT targets the consequences
of oncological treatments by restoring the pelvic floor tissues [8] while providing support and
guidance to women to resume painless sexual activities [25,26]. So far, only one recent
PLOS ONE | https://doi.org/10.1371/journal.pone.0262844 January 25, 2022
2 / 20
PLOS ONE Multimodal pelvic floor physical therapy for cancer survivors suffering from pain during sexual intercourse
multicenter prospective study conducted by our team investigated a 12-week PFPT treatment
in this population [27]. Significant changes in biological and psychosexual outcomes were
found following treatment [27–29]. Using a comprehensive assessment combining intra-vagi-
nal dynamometry and ultrasound imaging, pelvic floor muscle tone was significantly reduced
while tissue flexibility, muscle contractile properties, control as well as endurance significantly
improved immediately after treatment [28]. An increase in vaginal dimensions and a reduction
in vaginal atrophy signs were also measured [28]. Concurrently, pain during intercourse, sex-
ual distress, body image concerns, pain anxiety, pain catastrophizing, depressive symptoms,
urinary symptoms, vaginal symptoms and sexual matters decreased while sexual functioning
and pain self-efficacy improved after PFPT [27,29].
To date, no study has examined whether the short-term improvements following PFPT in
gynecological cancer survivors with dyspareunia are sustained over time. Long-term treatment
effects have important socioeconomic implications [30,31], and evaluating them may provide
critical insights beyond those assessed in the short term [32]. More importantly, using only
quantitative methods may not be sufficient to fully capture the extent of PFPT effects as these
are multidimensional and likely depend on the interaction of multiple factors [6]. Further-
more, it has been recently recognized that PFPT is not only a physical treatment but it is also a
behavioral treatment, which emphasizes the relevance of investigating physical, cognitive and
behavioral outcomes associated with PFPT [33]. Combining quantitative and qualitative meth-
ods would therefore provide a better understanding of the treatment effects and how they
influence each other considering the clinical context of multimodal PFPT [34,35]. This mixed-
method study aimed to examine the improvements in pain, sexual functioning, sexual distress,
body image concerns, pain anxiety, pain catastrophizing, pain self-efficacy, depressive symp-
toms and pelvic floor disorder symptoms in gynecological cancer survivors with dyspareunia
after multimodal PFPT, and to explore women’s perceptions of treatment effects at one-year
follow-up.
Materials and methods
Design and methodology
This study is a planned follow-up study of a multicenter prospective interventional study
investigating the treatment effects of multimodal PFPT for gynecological cancer survivors
with dyspareunia [27]. Our intent was to follow the whole cohort instead of a subsample in
order to most closely match the primary trial (mainly in terms of participant characteristics
and study outcomes) [32]. This research was conducted in Sherbrooke and Montreal (Can-
ada). Changes from baseline to post-treatment have been published elsewhere [27–29], and
changes from baseline and post-treatment to one-year follow-up will be the focus of the
present manuscript. The participants underwent baseline, post-treatment and one-year fol-
low-up assessments. Quantitative data were collected at all time points. To ascertain and
advance our understanding of treatment effects at one-year follow-up, individual semi-
structured telephone interviews were carried out to collect qualitative data [34,35]. This
research was approved by the Ethics Review Board of the CIUSSS de l’Estrie–CHUS (MP-
31-2016-1322) and was registered on ClinicalTrials.gov (NCT03935698). Participants pro-
vided written informed consent.
Participants
Women were included according to the following criteria: (i) all planned oncological treat-
ments for either endometrial or cervical cancer (stages ranging from I to IV) completed for
at least three months; (ii) in remission given the absence of disease on radiologic imaging
PLOS ONE | https://doi.org/10.1371/journal.pone.0262844 January 25, 2022
3 / 20
PLOS ONE Multimodal pelvic floor physical therapy for cancer survivors suffering from pain during sexual intercourse
for at least three months; (iii) moderate to severe vulvovaginal pain during sexual inter-
course (i.e., pain at the entry of the vagina and at the mid-vagina, at the level of the pelvic
floor muscles), corresponding to a pain intensity of 5 or more on a Numerical Rating Scale
(NRS) ranging from 0 (no pain) to 10 (worst pain); (iv) vulvovaginal pain experienced in
more than 80% of sexual intercourse for at least three months; (v) a stable sexual partner;
and (vi) willingness to attempt vaginal penetrations. A gynecologic oncologist of our team
at each site performed a standardized gynecological examination to rule out other condi-
tions possibly causing dyspareunia (e.g., vaginitis, cystitis or dermatitis). Exclusion criteria
were: (i) inability to communicate in French or English; (ii) dyspareunia prior to cancer or
pelvic pain unrelated to intercourse; (iii) other pelvic conditions including urinary tract or
vaginal infection, deep pelvic pain (i.e., pain experienced in the abdomen with deep pene-
tration), chronic constipation, severe pelvic organ descent based on the Pelvic Organ Pro-
lapse–Quantification system (stage III or more); (iv) other primary pelvic cancer or breast
cancer; (v) any history of vulvar, vaginal or pelvic surgery unrelated to cancer; (vi) PFPT in
the last year; (vii) changes in the use or dosage of menopausal hormone therapy in the last
six months; (viii) a major medical or psychological condition likely to interfere with study
procedures; or (ix) refusal to abstain from using other treatments for dyspareunia until the
post-treatment assessment.
Treatment content
The treatment protocol was designed by a multidisciplinary team consisting of experts in gyne-
cologic oncology, physical therapy, psychology and sexual health. The treatment included 12
weekly sessions of 60 minutes with a physical therapist certified and experienced in pelvic and
women’s health. The treatment components were chosen to reflect practice in a clinical setting
[36]. At each session, the physical therapist provided information, advice and support to
women. She explained the underlying mechanisms of chronic pain experienced during sexual
intercourse after gynecological cancer including the role of the pelvic floor muscles and how
the treatment could help to reduce the pain. She gave additional information about how to
manage chronic pain and other pelvic floor disorder symptoms (e.g., bladder training). The
use of relaxation techniques using deep breathing as well as the application of vaginal lubri-
cants and moisturizers were encouraged. The physical therapist also helped the participants
gain more knowledge about sexual functioning (i.e., physiology of desire, excitation and
orgasm) and guided them into resuming non-painful sexual activities with their partners. The
latter was invited to participate in the treatment to help his partner in this process. Moreover,
the physical therapist was available to further discuss topics with the participants who were
invited to reflect on their sexual difficulties in order to overcome them with the help of their
therapist. At each session, manual therapy techniques (i.e., stretching, myofascial release and
tissue desensitization) and pelvic floor muscle exercises with electromyography biofeedback
(i.e., relaxation, motor control, strength and endurance) using a small intra-vaginal probe
were used. Women were also asked to perform home exercises resembling those performed
under supervision five times per week as well as auto-insertion exercises with a finger or
graded vaginal dilator in addition to desensitization techniques three times per week.
Throughout the treatment, the physical therapist supervised each woman’s progress and pro-
vided feedback. Additionally, modalities were intensified (e.g., more pressure applied to stretch
the tissues, longer duration of the technique or exercise and greater dilator size) following each
woman’s progress. At the end of the treatment, women were encouraged to pursue home exer-
cises two to three times per week to maintain the effects of treatment. Further details of the
treatment modalities are presented elsewhere [27].
PLOS ONE | https://doi.org/10.1371/journal.pone.0262844 January 25, 2022
4 / 20
PLOS ONE Multimodal pelvic floor physical therapy for cancer survivors suffering from pain during sexual intercourse
Data collection
Participants were assessed at baseline, post-treatment and one-year follow-up. Sample charac-
teristics were collected at baseline. At each time point, quantitative outcomes were assessed
using validated scales and questionnaires. After the collection of quantitative data at one-year
follow-up, an individual semi-structured telephone interview was conducted in French or in
English to further explore women’s perceptions of treatment effects. Participants were also
asked if there were any changes regarding their health (e.g., cancer recurrence), if they were
pursuing the home exercises, if they had attempted other treatments for pain or sexual dys-
function and if their relationship status had changed during the follow-up period.
Study outcomes
Quantitative. The NRS was used to evaluate the average intensity of pain during inter-
course [37]. The McGill Pain Questionnaire (MPQ) was used to qualify the pain according to
its sensory, affective and evaluative dimensions, with higher scores corresponding to more sig-
nificant pain [38]. The Female Sexual Function Index (FSFI) was used to examine sexual func-
tioning including desire, arousal, lubrication, orgasm, satisfaction and pain, with higher total
scores representing a better sexual function [39,40]. The Female Sexual Distress Scale-Revised
(FSDS-R) was used to assess sexual distress, with higher scores relating to more sexual distress
[41,42]. The Body Image Scale (BIS) was administered to evaluate body image concerns, with
higher scores indicating greater concerns [43]. The Pain Anxiety Symptom Scale (PASS),
which is an indirect measure of fear of pain during intercourse, was used to assess pain-related
anxiety, with higher scores indicating more severe pain anxiety [44]. The Pain Catastrophizing
Scale (PCS) was used to evaluate the exaggerated negative cognitions and emotions regarding
pain, with higher scores pointing to greater pain catastrophizing [45]. The Painful Intercourse
Self-Efficacy Scale (PISES) was used to assess pain self-efficacy associated with painful sexual
intercourse, with higher scores representing better self-efficacy [46]. The Beck Depression
Inventory-II (BDI-II) was used to evaluate depressive symptoms, with higher scores corre-
sponding to higher severity of symptoms [37]. Pelvic floor disorder symptoms including uri-
nary symptoms, vaginal symptoms and sexual matters were assessed with the International
Consultation on Incontinence Questionnaire (ICIQ) modules. The ICIQ-Urinary Inconti-
nence Short Form (ICIQ-UI SF) was used for urinary symptoms [47] and the ICIQ-Vaginal
Symptoms (ICIQ-VS) for vaginal symptoms and sexual matters [48], with higher scores repre-
senting more symptoms or sexual matters [47,48]. In addition, the Patient Global Impression
of Change (PGIC) allowed the participants to self-report their perceived improvement (catego-
ries ranging from very much improved to very much worse) [49].
Qualitative. Prior to their individual semi-structured telephone interview, participants
were informed of the interview topics and invited to reflect on the treatment effects they per-
ceived and how these effects evolved over time during the follow-up period. Each interview
lasted approximately 70 minutes. The first author (MPC) underwent qualitative research train-
ing to conduct all the interviews. She was not involved in participants’ care and was blinded to
the participants’ responses in the questionnaires to avoid any preconceived ideas about the
treatment effects. Before conducting the interviews, the interviewer reconfirmed the women’s
consent to participate in the interviews and for recording the conversation. She used a non-
judgmental approach and created a trustful and respectful relationship to ease the discussion
of what could be perceived by participants as sensitive topics. Interviews followed a semi-struc-
tured guide co-constructed by the first author (MPC), the principal investigator (MM) and
another research team member who has extensive experience conducting qualitative research
(CC) (see S1 File for the interview guide). The interview questions related to this manuscript’s
PLOS ONE | https://doi.org/10.1371/journal.pone.0262844 January 25, 2022
5 / 20
PLOS ONE Multimodal pelvic floor physical therapy for cancer survivors suffering from pain during sexual intercourse
research objective focused on women’s perceptions of treatment effects and their hypotheses
about factors influencing these effects. Probing questions aimed to obtain in-depth informa-
tion about participants’ perceptions of treatment effects, exploring short-term effects previ-
ously documented in quantitative research [27–29] and using a biopsychosocial approach of
health to explore any further effects and factors perceived to influence these effects [6,7]. The
semi-structured guide was pilot-tested with a patient partner under the supervision of the
principal investigator (MM) and the other research team member (CC).
Sample size
An a priori sample size was calculated for the multicenter prospective interventional study
based on the proportion of completed home exercises (80%) as adherence was suggested as
being important to perceive significant effects in physical therapy [50]. With a confidence level
of 95%, an interval width of 30%, and to account for potential dropouts over time (15%), a
total of 31 women were initially recruited for quantitative purposes (further details are avail-
able elsewhere) [27]. All these women were invited to take part in an individual semi-struc-
tured telephone interview to explore all of the various perceptions of treatment effects.
Data analysis
Quantitative data analysis was performed using IBM SPSS Statistics 27 (IBM Corporation,
Armonk, N.Y., USA). Descriptive statistics were used to present baseline and one-year sample
characteristics as well as PGIC results. Intention-to-treat analyses (i.e., all participants are
included in the statistical analysis, regardless of their level of adherence) were conducted to
explore whether the improvements in all outcomes were sustained at one-year follow-up. Out-
comes at baseline and one-year follow-up as well as the changes from baseline and post-treat-
ment to one-year follow-up are reported and expressed as mean estimated values (95%
confidence interval) according to linear mixed modeling with Bonferroni correction [51–53].
Models included time as the fixed effect and random intercepts for each subject to account for
repeated measures (i.e., to accommodate within-subject correlation). Statistical significance
was set at p-value < 0.05 (two-tailed).
Qualitative data analysis was based on the audio-recorded interviews which were transcribed
and analyzed by the first author (MPC) using NVivo (version 12) software. A thematic analysis
was adopted to ensure data-driven analyses and interpretations [54]. Specifically, an inductive
approach was used when the first author (MPC) coded key ideas and started identifying emerg-
ing themes. Subsequently, another team member (RD) reviewed the codes. Coding disagree-
ments were discussed until a consensus was achieved. Codes were reviewed by two research
team members (MM and CC), and several meetings were held to regroup codes into themes.
Relationships between themes were explored by observing patterns across themes. As most of
the original quotations used in this manuscript were in French, they were translated into English
and revised by a certified translator. Field notes were used to explore researcher reflexivity and
further support the data interpretation. It should be noted that results from quantitative and
qualitative methods were integrated during the interpretation phase of the study.
Results
Participant characteristics
Thirty-one women enrolled initially in this study. Fig 1 shows the flow of participants through
the study. Additional details on screening and eligibility assessments are available elsewhere
[27].
PLOS ONE | https://doi.org/10.1371/journal.pone.0262844 January 25, 2022
6 / 20
PLOS ONE Multimodal pelvic floor physical therapy for cancer survivors suffering from pain during sexual intercourse
Fig 1. Flow of participants through the study.
https://doi.org/10.1371/journal.pone.0262844.g001
Baseline sample characteristics (n = 31) are presented in Table 1. Before the multimodal
PFPT treatment, women had an average pain intensity of 7.3 (6.7 to 8.0) on the NRS and
the median duration of pain was approximately three years. Of the 29 women assessed at
one-year follow-up, three reported having had a cancer recurrence or another cancer dur-
ing the follow-up period, and one was recovering from a severe upper urinary tract
infection.
Study outcomes
Quantitative. The quantitative outcomes assessed at baseline and one-year follow-up as
well as the changes from baseline and post-treatment to one-year follow-up are presented in
Table 2. Significant improvements were found from baseline to one-year follow-up on all out-
comes (P � 0.028). Moreover, changes from post-treatment to one-year follow-up were statis-
tically non-significant (P � 0.084), suggesting that the improvements were maintained over
time. Of the 29 women assessed at one-year follow-up, 25 (86%) reported being very much or
much improved. The others reported minimal improvements (7%), no changes (3%) or being
minimally worse (3%) compared to baseline. Concerning the adherence to home exercises, 18
(62%) performed the pelvic floor muscle exercises during the follow-up period, with a median
frequency of three times (two to eight) per month. Moreover, 10 (34%) participants performed
the auto-insertion exercises, with a median frequency of three times (one to five) per month.
PLOS ONE | https://doi.org/10.1371/journal.pone.0262844 January 25, 2022
7 / 20
PLOS ONE Multimodal pelvic floor physical therapy for cancer survivors suffering from pain during sexual intercourse
Table 1. Sample characteristics at baseline.
Characteristics
Age (years), mean (SD)
Body mass index (kg/m2), mean (SD)
Cancer type, n (%)
Endometrial
Cervical
Disease stage, n (%)
I
II
III
IV
Time since oncological treatments (months), median (Q1 to Q3)
Oncological treatments, n (%)
Surgery alone
Surgery + brachytherapy or external beam radiation therapy
Surgery + brachytherapy + external beam radiation therapy + chemotherapy
Surgery + chemotherapy
Brachytherapy + external beam radiation therapy + chemotherapy
SD, standard deviation; n, number of participants; Q1, first quartile; Q3, third quartile.
https://doi.org/10.1371/journal.pone.0262844.t001
Value
55.9 (10.8)
28.5 (5.3)
20 (64.5)
11 (35.5)
19 (61)
6 (19)
5 (16)
1 (3)
38 (9 to 70)
9 (29)
6 (19)
7 (23)
2 (6)
7 (23)
No women stated having attempted other treatments for pain or sexual dysfunction during
this period, and only one reported being no longer with her partner at one-year follow-up.
Qualitative. Three main themes were described by participants as the most meaning-
ful treatment effects for them in terms of symptoms or functioning: (a) reduction in pain
during intercourse; (b) improvement in sexual functioning; and (c) reduction in urinary
symptoms. These themes are detailed below along with participants’ perceived modulating
and contributing factors. Modulating factors were defined as the factors altering the mag-
nitude of the main effects (e.g., adherence) while contributing factors were those
described as other treatment effects which influenced positively the main effects (e.g.,
reduction in muscle tensions). Fig 2 illustrates how the main treatment effects (in black)
interacted and were influenced by various biological, psychological or social factors (in
grey).
THEME 1. Reduction in pain during intercourse. All participants reported experiencing less
pain during intercourse, with several stating having no pain at all since the end of the PFPT
treatment. Although the majority expressed that this effect was maintained, a small number of
women said that the pain reduction was attenuated at one-year follow-up. Among the poten-
tial explanations, some of them suggested that discontinuing home exercises or stopping regu-
lar sexual intercourse with vaginal penetration might have contributed to this depletion effect.
“It fixed my pain problem and it lasted over time.”–C02
“I would say that it has deteriorated a bit since, but it’s my fault because I didn’t keep doing
the exercises long enough. I know if I resumed the exercises it would get better. However, it
[the pain] hasn’t come back to how it was before; in other words what has been done has been
of benefit. Having sexual intercourse regularly helps to ensure these gains are maintained in a
way.”–C12
PLOS ONE | https://doi.org/10.1371/journal.pone.0262844 January 25, 2022
8 / 20
PLOS ONE Multimodal pelvic floor physical therapy for cancer survivors suffering from pain during sexual intercourse
Table 2. Outcomes at baseline (n = 31) and one-year follow-up (n = 29) and changes from baseline and post-treatment to one-year follow-up.
Pain intensity NRS (0–10)
7.3 (6.7 to 8.0)
2.7 (2.0 to 3.3)
-4.6 (-5.7 to -3.6)
Baseline
One-year follow-up Changes from baseline to
follow-up
Pain quality MPQ (0–78)
21.1 (17.6 to 24.6)
6.7 (3.1 to 10.4)
-14.4 (-20.5 to -8.3)
Sexual function FSFI (2–36)
Sexual distress FSDS-R (0–52)
18.9 (16.3 to 21.4)
(n = 20)b
26.7 (22.3 to 31.1)
23.4 (20.8 to 26.0)
(n = 18)b
16.6 (12.1 to 21.1)
4.6 (1.0 to 8.1)
-10.0 (-15.7 to -4.4)
Body image concerns BIS (0–30)
6.4 (4.8 to 7.9)
3.0 (1.4 to 4.6)
-3.4 (-5.4 to -1.3)
Pain anxiety PASS (0–100)
37.5 (32.4 to 42.7)
23.7 (18.4 to 28.9)
-13.9 (-21.6 to -6.2)
Pain catastrophizing PCS (0–52)
20.9 (16.6 to 25.2)
8.3 (3.9 to 12.7)
-12.6 (-18.1 to -7.1)
Painful intercourse self-efficacy
PISES (10–100)
63.6 (58.1 to 69.0)
80.6 (75.0 to 86.2)
17.1 (10.1 to 24.1)
Depressive symptoms BDI-II (0–63)
10.9 (8.0 to 13.9)
7.5 (4.5 to 10.5)
Urinary symptoms ICIQ-UI (0–21)
3.8 (2.5 to 5.2)
1.8 (0.4 to 3.3)
Vaginal symptoms ICIQ-VS (0–53)
13.5 (11.5 to 15.4)
7.2 (5.2 to 9.2)
-3.5 (-6.6 to -0.3)
-2.0 (-3.3 to -0.6)
-6.3 (-8.6 to -4.0)
Sexual matters ICIQ-VS (0–58)
43.7 (37.7 to 49.7)
(n = 24)c
20.9 (14.8 to 27.0)
(n = 23)c
-22.8 (-32.3 to -13.4)
Pa
<
0.001
<
0.001
0.009
<
0.001
<
0.001
<
0.001
<
0.001
<
0.001
0.028
0.002
<
0.001
<
0.001
Changes from post-treatment to
follow-up
1.0 (-0.1 to 2.0)
-0.5 (-6.7 to 5.6)
-2.8 (-6.2 to 0.5)
2.7 (-2.9 to 8.4)
0.1 (-1.9 to 2.1)
2.8 (-5.0 to 10.5)
0.6 (-5.0 to 6.1)
-6.3 (-13.4 to 0.7)
1.1 (-2.1 to 4.2)
-0.5 (-1.8 to 0.9)
-0.4 (-2.7 to 1.9)
1.2 (-8.0 to 10.3)
Pa
0.084
1.000
0.119
0.708
1.000
1.000
1.000
0.095
1.000
1.000
1.000
1.000
The data shown are the mean estimated values (95% confidence interval) derived from the linear mixed models.
a P-values extracted from the linear mixed modeling with Bonferroni correction.
b Eleven women at baseline and 11 women at one-year follow-up did not engage in sexual activities including vaginal penetration in the last month and thereby, due to
the one-month time frame used in the FSFI questionnaire, their total score could not be compilated. Reasons for not engaging in such activities at one-year follow-up:
4 = partner-related reasons including lack of sexual desire or medical problems such as erectile problems; 4 = participant-related reasons including lack of sexual desire
(n = 2) or pain during intercourse (n = 2) although they reported a pain reduction of 4.5 and 5 on the NRS from baseline to one-year follow-up; 2 = relationship-related
difficulties; 1 = medical indication to not engage due to vaginal bleeding unrelated to PFPT.
c Seven participants at baseline and six at one-year follow-up did not engage in any form of sexual activities in the last month (time frame of ICIQ-VS for sexual
matters).
https://doi.org/10.1371/journal.pone.0262844.t002
Every participant associated the pain reduction with pelvic floor tissue changes. They
noticed that the muscle tensions decreased while the tissue flexibility increased, attributing this
to the manual techniques and the exercises. Some emphasized that relaxation techniques such
as deep breathing promoted muscle relaxation, reduction of tensions, and hence, a pain relief.
Overall, the women related these tissue changes to a less tense or deeper vagina, which allowed
them to be more at ease and helped them to have a more complete and comfortable vaginal
penetration with less or no pain.
“All the exercises [contraction and stretching] I had done and what the physical therapist had
done removed the tension and loosened me up. It felt good. Penetration was easier.”–C01
“The stretching we did reduced my pain because when it stretches better, it’s less painful. Oth-
erwise, I felt like the skin inside wanted so badly to split because, before, it wouldn’t stretch.”–
C18
“Breathing helps because I think when you calm down, it’s less contracted and there’s more
flexibility for the activity.”–C16
PLOS ONE | https://doi.org/10.1371/journal.pone.0262844 January 25, 2022
9 / 20
PLOS ONE Multimodal pelvic floor physical therapy for cancer survivors suffering from pain during sexual intercourse
Fig 2. Relationships between treatment effects that emerged from the interviews.
https://doi.org/10.1371/journal.pone.0262844.g002
Many women also observed becoming more aware of the pelvic floor musculature and its
relationship with pain. During the PFPT treatment, they recalled gaining control over their
muscles and developing muscle awareness. Motor control was noted as being important by the
participants to break a chain of events involving the pelvic floor muscles and pain.
“When you are calmer, it [the pelvic floor muscle] is less contracted, so it is more flexible. [. . .]
Before the treatments, I didn’t know how to do [relax my muscles], I was tense. Now, I have
techniques that last over time. [. . .] I have gone from. . . not hysteria, but from an uncon-
trolled fear to something more serene. I am calmer when considering having sex, I am more
welcoming.”–C16
Our participants often mentioned being reassured knowing how to influence the pain.
They frequently expressed being less afraid of pain because they understood what led to their
symptoms and were taught relevant and effective tools to reduce it.
“After cancer treatments, you feel diminished. Will it come back as before? I was starting to be
afraid. With physical therapy, you feel less diminished. It seemed as if it was finally possible
that things could get better. When I got into the program, it was another story as I realized it
was possible to improve, and it was much less upsetting, less scary. It’s because we found
PLOS ONE | https://doi.org/10.1371/journal.pone.0262844 January 25, 2022
10 / 20
PLOS ONE Multimodal pelvic floor physical therapy for cancer survivors suffering from pain during sexual intercourse
where it hurt most. It’s about understanding. . . It’s partly confidence, partly the fear that’s
gone.”–C124
Consequently, they explained that they were feeling more in control, self-efficient and
hopeful while being less anxious about their pain. Some participants even emphasized that
they were no longer afraid to undergo gynecological examinations. Experiencing less pain dur-
ing intercourse also enhanced these feelings, which in turn amplified their self-esteem and
confidence to engage in sexual activities. They felt less distressed, with several highlighting that
fact they were less depressed and more positive in their everyday lives.
“And what I also learned was that I felt that I could influence my pain. When it’s less painful,
less tight, you are more relaxed, you have more confidence and you let go more easily. Psycho-
logically, I could say that I felt I was moving further away from the operation and its negative
side. I found that I was getting closer to a more normal life, as it was before, in a sense. . . with-
out much difficulty. Yes, it’s vague, isn’t it? Well, normal life. . . having sex again, get away
from the cancer thing.”–C115
THEME 2. Improvement in sexual functioning. All women reported improvement in their
sexual functioning following PFPT. Although a low proportion of participants did not perceive
changes in their sexual functioning in terms of lubrication and libido or sexual desire, the vast
majority mentioned their vagina being less dry and more naturally lubricated during sexual
activities. Among other things, several women emphasized not needing to use vaginal products
anymore and reported being less stressed and more interested in engaging in sexual activities.
“The lubrication. . . it all came basically together after the treatment. Sure, at first I needed
some lubricant, but little by little, as I worked, it just faded so I didn’t need the lubricant any-
more.”–C09
The perceptions relating to pain reduction described previously could also suggest how par-
ticipants felt about sexuality. Many of them reported being more interested in engaging given
the pain reduction and the positive emotions and thoughts they developed about their sexual
identity. Some women associated their increased sexual desire to the improved perception of
their body, which defined them as women. They grew to accept themselves, sensed that their
body belonged to them and reclaimed it. Participants specified that this body re-appropriation
helped them to express themselves sexually as women. They were able to have sexual inter-
course with vaginal penetration rather than endure the barriers induced by cancer, which
hampered them. Consequently, they referred to being complete women and having a more
normal life. Participants related that regaining the capacity of having intercourse helped them
initiate and engage in sexual activities, which in turn increased their femininity.
“I could see that there were still defects in my body since the operation and all that, and psy-
chologically it disturbed me. Now, I let myself go more. There is a connection that has been
made with my body and my whole person. I participate more with my body now, which I
didn’t before. I had an easier time opening up to sexuality. That’s why I say it really. . .
changed my life. Physical therapy is beneficial, it is a psycho-unblocker.”–C100
“Knowing what to do to have intercourse and being able to have it [sexual intercourse] really
made me feel like a woman. I am very happy to have learned to control my body better and to
be able to have a more fulfilling sex life. It’s like. . . I feel like more of a complete woman, I
don’t know. . . entirely a woman.”–C11
PLOS ONE | https://doi.org/10.1371/journal.pone.0262844 January 25, 2022
11 / 20
PLOS ONE Multimodal pelvic floor physical therapy for cancer survivors suffering from pain during sexual intercourse
“Basically, sexuality is more about being a complete woman, [. . .] Now, if I feel like having
sex, I can have it. [. . .] So, life for me is much more normal than it used to be. It changed my
life, it gave me back intimacy. So, we’re less active than we were, but at least if we want to, we
can! So that’s the difference.”–C10
Participants also recognized that they were more comfortable talking about sexuality. They
stressed that this led them to communicate more about their feelings and difficulties to their
partner. As a result, participants and their partners were more capable of adapting their behav-
ior, and when considering physical intimacy, it was therefore less stressful and more pleasur-
able. Furthermore, participants said that because they had less pain during intercourse, their
partner was less afraid to hurt them, and this dynamic was helpful for the couple to be physi-
cally intimate.
“I was also able to talk about it [thoughts about intercourse] with my partner because I had
not talked about it before. When I had intercourse before, it was because I felt obliged. It was
very rare that we had any. With the study, it was like day and night, winter and summer. It
was like having sex two or three times a week by the end of the study.”–C06
“I was no longer in pain. . . well, for sure in our intimate relationship and all that there was a
letting go so that was really amazing. Less fear, less apprehension. Yes, I think it reassured my
husband a lot to see that it was going well, that it was getting better. He was also less afraid of
hurting me and he was more reassured that there were two of us in this sexual activity.”–C08
Because they were more communicative, most women acknowledged that they and their
partner discussed their sexuality and intimacy more openly. Those who did not report any
changes in this regard claimed their relationship was already strong and without issues before
enrolling in the study. The former noticed that they and their partner were closer to each
other, discovered and tried new ways to express their love. Several participants spoke of how it
became more affectionate than sexually demonstrative with intercourse during the one-year
follow-up period. For a handful of women, this was accentuated if there had been a significant
event (e.g., cancer recurrence), low sexual desire, pain during intercourse or a medical condi-
tion of the partner.
“It helped me to understand how my body reacted to a lot of things, to understand that I was
not alone and it helped me to accept myself and accept living my sex life in a different way. It
[the treatment] allowed us to make different connections. There is a lot, really a lot of affec-
tion. It starts slowly, and, in the end, it becomes intense. This is what is new, this is what we
learned.”–C17B
THEME 3. Reduction in urinary symptoms. Half of the sample experienced either stress uri-
nary incontinence, urgency urinary incontinence or symptoms of urinary urgency before the
study and all women reported significant improvements following PFPT. Participants
observed that the pelvic floor muscle exercises in addition to bladder training increased their
muscle awareness, strength and endurance to activate their pelvic floor muscles when needed.
For instance, it gave them the means to delay the urge to urinate or to hold the urine for longer
periods.
“Before, I used to go to the bathroom. . . a lot! Almost every hour, and now I go like three or
four times a day and that’s enough. So, for sure, there is a difference there as well.”–C14
PLOS ONE | https://doi.org/10.1371/journal.pone.0262844 January 25, 2022
12 / 20
PLOS ONE Multimodal pelvic floor physical therapy for cancer survivors suffering from pain during sexual intercourse
“I used to go to the bathroom all the time, all the time, and she [the physical therapist] gave
me some tips for the bladder and exercises, and it’s getting better in that respect too.”–C111
“All the exercises, the squeezing and all that helped. You squeeze and it calms your bladder. I
didn’t think it would work. Listen, I can even hold my urine when I go to the bathroom. . .
Before, when I saw the toilet, I had to run and when I saw the toilet bowl, I leaked two or
three drops. But now, I am able to hold it. I know what to do.” –C10
Interestingly, two women said that having had painful urination and difficulty retaining
high volumes of urine since the oncological treatments and they explained that, by releasing
tensions in the pelvic area, the PFPT modalities such as manual therapy and auto-insertion
exercises helped them to resolve these issues.
“It was stiff near the bladder and it hurt. I felt the bladder was jammed, it was like there was
no room for it to fill up. So, the physical therapy helped to relax the tensions and my bladder
had more room so I needed to urinate less often. At night, I used to get up every three hours, I
get up less now, so I sleep better. Everything is going in the right direction.”–C17B
Discussion
This mixed-method study provides evidence that the improvements in pain, sexual function-
ing, sexual distress, body image concerns, pain anxiety, pain catastrophizing, painful inter-
course self-efficacy, depressive symptoms, urinary symptoms, vaginal symptoms and sexual
matters following multimodal PFPT can be sustained at one-year follow-up in gynecological
cancer survivors with dyspareunia. Furthermore, reduction in pain during sexual intercourse,
improvement in sexual functioning and reduction in urinary symptoms were reported by par-
ticipants as the most meaningful effects during the interviews. In addition, participants
expressed these treatment effects in relation to adherence. They also emphasized that the treat-
ment led to positive biological, psychological and social changes which contributed to the
improvements in dyspareunia and sexual functioning.
This is the first study to examine whether the short-term improvements following multi-
modal PFPT are maintained over time in gynecological cancer survivors affected by dyspareu-
nia [55]. Interventional studies conducted to date in women who had been treated for
gynecological cancer were not specific to dyspareunia (e.g., urinary incontinence, vaginal atro-
phy or low sexual desire) [56–64]. To our knowledge, only a few cohort studies included a fol-
low-up assessment beyond six months [60,62,65,66]. Improvement in sexual functioning have
been seen following interventions integrating psychosexual education and unsupervised pelvic
floor exercises in gynecological cancer survivors [60,62], which is consistent with the current
study. However, their target population was different as women with or without symptoms
were included immediately after oncological treatments. The experimental interventions were
also designed to prevent or address common symptoms in gynecological cancer survivors
while not specifically targeting dyspareunia [60,62]. In contrast, our sample was probably
more affected at baseline as all women presented a pain intensity of more than 5 on the NRS
for a median duration of three years, representing chronic moderate-to-severe dyspareunia
[67,68]. Despite this chronicity and severity, it is noteworthy that participants still observed
and reported sustained significant effects one year later.
The women in the present study expressed meaningful improvements in pain during inter-
course, sexual functioning and urinary symptoms that lasted one year after PFPT. Similar find-
ings were found in studies investigating multimodal PFPT effects in younger women suffering
PLOS ONE | https://doi.org/10.1371/journal.pone.0262844 January 25, 2022
13 / 20
PLOS ONE Multimodal pelvic floor physical therapy for cancer survivors suffering from pain during sexual intercourse
from vulvar pain with no history of cancer, although the available data is limited to a six-
month follow-up in this population [69,70]. Morin et al. [70] in a large multicenter random-
ized controlled trial (n = 212) revealed reductions in pain and sexual distress with improved
sexual functioning from baseline to six-month follow-up, compared to topical lidocaine, a fre-
quent first-line treatment. Moreover, a recent Cochrane meta-analysis concluded that pelvic
floor muscle training can reduce or cure urinary symptoms in women without a history of can-
cer [71], which is in line with our results. It is worth noting that the majority of studies con-
ducted in women affected by dyspareunia with no history of cancer applied quantitative
methods to evaluate the effects of multimodal PFPT [69,70,72]. A quantitative research design
could only provide a narrow view of PFPT effects, as demonstrated in the current study.
Quantitative results combined with the participants’ inputs suggest that multimodal PFPT
improved multiple dimensions of the biopsychological framework of dyspareunia [6,9,11,73],
and these improvements remained at one-year follow-up. More precisely, the effects on pain
during intercourse, sexual functioning and urinary symptoms were explained by gynecological
cancer survivors through biological, psychological and social changes attributable to PFPT
modalities. Gynecological cancer survivors emphasized the role of multimodal PFPT in the
effects perceived and how it helped them to achieve pain-free sexual activities or improve their
sexual functioning or behavior. It is notable that the treatment not only improved the pelvic
floor tissues, as observed in short-term studies using objective tools [28,72,74], but also had a
direct or indirect positive impact on psychological and social dimensions according to our
cohort. Qualitative data suggested that performing PFPT exercices or having sexual inter-
course regularly could be important to retain the biological changes related to pain for certain
women. These details show that treatment effects over time could depend on adherence in the
long term. Comparing our results to the studies conducted in women with no history of can-
cer, only two studies [75,76] to date have investigated the improvements following myofascial
release techniques [76] and multimodal PFPT at three-month follow-up using a shorter inter-
view [75] for dyspareunia in young women. The latter study reported similar effects in regard
to muscle awareness, knowledge and communication about pain, self-efficacy, self-esteem,
sexual confidence, attitudes about sexuality and relationship with the partner [75]. However, it
should be underlined that our group of participants was still experiencing substantial effects at
one-year follow-up after PFPT even though they had been treated for cancer, were older and
had had dyspareunia for a median duration of three years. As opposed to previous work
[75,76], our study is the first to triangulate data from different methods and to present exten-
sively qualitative findings about multimodal PFPT effects by reporting the participants’ inputs
that supported our interpretation while providing a deeper understanding. Overall, our find-
ings suggest multimodal PFPT as a biopsychosocial treatment for reducing dyspareunia and
improving sexual functioning.
The main strength of this study is the integration of quantitative and qualitative methods to
allow data triangulation and complementarity to fully capture the treatment effects [77–79].
Validated scales and questionnaires were used to assess the quantitative outcomes. Intention-
to-treat analyses were conducted and considered multiple comparisons as well as missing data.
The high participation rate in qualitative interviews promoted a wide range of perspectives
and shed light on how multimodal PFPT could have influenced dimensions other than the
well-known biological dimension. The mixed-method design has allowed us to illustrate elo-
quently the quantitative findings supported by statistics and through the perceptions of
women. Our results should, however, be interpreted within the context of certain limitations.
The absence of a control group limits the causal inference. Nonetheless, the women’s percep-
tions support the role of PFPT in leading to these effects. They also did not attempt other treat-
ments during the follow-up period. Moreover, it is unlikely that they would have improved
PLOS ONE | https://doi.org/10.1371/journal.pone.0262844 January 25, 2022
14 / 20
PLOS ONE Multimodal pelvic floor physical therapy for cancer survivors suffering from pain during sexual intercourse
without any treatment, given that they were suffering from dyspareunia for a median time of
approximately three years and that sexual issues tend to persist over time [80,81]. Even though
these aspects are suggestive of a causal inference of PFPT on outcomes, a randomized con-
trolled trial is ultimately required to confirm the long-term efficacy of this treatment. As the
PFPT treatment combined multiple modalities, it is difficult to isolate their respective effect on
the outcomes. Moreover, determining precisely how the treatment effects (i.e., reduction in
pain during intercourse, improvement in sexual functioning and reduction in urinary symp-
toms) and their modulating and contributing factors (i.e., adherence as well as biological, psy-
chological and social changes) interacted was not feasible. It is worth mentioning that it has
frequently been reported that these may overlap and influence each other dynamically and dif-
ferently among gynecological cancer survivors [18,82,83]. A biopsychosocial treatment
approach could have contributed to the magnitude of the effects [26].
Conclusions
Findings of this one-year follow-up mixed-method study suggest that the short-term improve-
ments in pain during sexual intercourse, sexual functioning and urinary incontinence follow-
ing PFPT can be sustained over time in gynecological cancer survivors with dyspareunia.
Although a randomized controlled trial is still required to confirm the efficacy, multimodal
PFPT showed beneficial effects of treating dyspareunia in this population through biological,
psychological and social changes after one year. The study therefore supports the biopsychoso-
cial role of multimodal PFPT in gynecological cancer survivors who are frequently affected by
pain and other types of sexual dysfunction. This treatment could be implemented in multidis-
ciplinary cancer care.
Supporting information
S1 File. Semi-structured interview guide.
(DOCX)
Acknowledgments
We would like to extend our gratitude to the physical therapists involved in the treatments and
assessments. We would also like to thank all the study participants for their support and dedi-
cation to this research project.
Author Contributions
Conceptualization: Marie-Pierre Cyr, Chantal Camden, Chantale Dumoulin, Me´lanie Morin.
Data curation: Marie-Pierre Cyr.
Formal analysis: Marie-Pierre Cyr, Rosalie Dostie, Chantal Camden, Me´lanie Morin.
Funding acquisition: Marie-Pierre Cyr, Chantale Dumoulin, Paul Bessette, Walter Henry
Gotlieb, Me´lanie Morin.
Investigation: Marie-Pierre Cyr, Chantale Dumoulin, Paul Bessette, Annick Pina, Walter
Henry Gotlieb, Korine Lapointe-Milot, Me´lanie Morin.
Methodology: Marie-Pierre Cyr, Chantal Camden, Chantale Dumoulin, Paul Bessette, Annick
Pina, Walter Henry Gotlieb, Marie-He´lène Mayrand, Me´lanie Morin.
Project administration: Marie-Pierre Cyr, Chantale Dumoulin, Me´lanie Morin.
PLOS ONE | https://doi.org/10.1371/journal.pone.0262844 January 25, 2022
15 / 20
PLOS ONE Multimodal pelvic floor physical therapy for cancer survivors suffering from pain during sexual intercourse
Resources: Marie-Pierre Cyr.
Supervision: Chantale Dumoulin, Me´lanie Morin.
Validation: Marie-Pierre Cyr, Me´lanie Morin.
Visualization: Marie-Pierre Cyr.
Writing – original draft: Marie-Pierre Cyr.
Writing – review & editing: Marie-Pierre Cyr, Rosalie Dostie, Chantal Camden, Chantale
Dumoulin, Paul Bessette, Annick Pina, Walter Henry Gotlieb, Korine Lapointe-Milot,
Marie-He´lène Mayrand, Me´lanie Morin.
References
1. Howlader N, Noone AM, Krapcho M, Miller D, Brest A, Yu M, et al. SEER Cancer Statistics Review,
1975–2017, National Cancer Institute. Bethesda, 2020.
2.
Torre LA, Islami F, Siegel RL, Ward EM, Jemal A. Global cancer in women: burden and trends. Cancer
Epidemiol Biomarkers Prev. 2017; 26:444–57. https://doi.org/10.1158/1055-9965.EPI-16-0858 PMID:
28223433
3. Rutledge TL, Heckman SR, Qualls C, Muller CY, Rogers RG. Pelvic floor disorders and sexual function
in gynecologic cancer survivors: a cohort study. Am J Obstet Gynecol. 2010; 203:514 e1–7. https://doi.
org/10.1016/j.ajog.2010.08.004 PMID: 20869691
4. Stinesen Kollberg K, Waldenstrom AC, Bergmark K, Dunberger G, Rossander A, Wilderang U, et al.
Reduced vaginal elasticity, reduced lubrication, and deep and superficial dyspareunia in irradiated
gynecological cancer survivors. Acta Oncol. 2015; 54:772–9. https://doi.org/10.3109/0284186X.2014.
1001036 PMID: 25761090
5. Coady D, Kennedy V. Sexual health in women affected by cancer: focus on sexual pain. Obstet Gyne-
col. 2016; 128:775–91. https://doi.org/10.1097/AOG.0000000000001621 PMID: 27607852
6. Bergeron S, Corsini-Munt S, Aerts L, Rancourt K, Rosen NO. Female sexual pain disorders: a review of
the literature on etiology and treatment. Curr Sex Health Rep. 2015; 7:159–69.
7. Engel GL. The need for a new medical model: a challenge for biomedicine. Science. 1977; 196:129–36.
https://doi.org/10.1126/science.847460 PMID: 847460
8. Cyr MP, Dumoulin C, Bessette P, Pina A, Gotlieb WH, Lapointe-Milot K, et al. Characterizing pelvic
floor muscle function and morphometry in survivors of gynecological cancer who have dyspareunia: a
comparative cross-sectional study. Phys Ther. 2021;101. https://doi.org/10.1093/ptj/pzab042 PMID:
33522584
9.
10.
Thomte´ n J, Linton SJ. A psychological view of sexual pain among women: applying the fear-avoidance
model. Women’s health (London, England). 2013; 9:251–63. https://doi.org/10.2217/whe.13.19 PMID:
23638781
Lemieux AJ, Bergeron S, Steben M, Lambert B. Do romantic partners’ responses to entry dyspareunia
affect women’s experience of pain? The roles of catastrophizing and self-efficacy. J Sex Med. 2013;
10:2274–84. https://doi.org/10.1111/jsm.12252 PMID: 23809759
11. Corsini-Munt S, Rancourt KM, Dube´ JP, Rossi MA, Rosen NO. Vulvodynia: a consideration of clinical
and methodological research challenges and recommended solutions. J Pain Res. 2017; 10:2425–36.
https://doi.org/10.2147/JPR.S126259 PMID: 29070953
12.
Juraskova I, Butow P, Robertson R, Sharpe L, McLeod C, Hacker N. Post-treatment sexual adjustment
following cervical and endometrial cancer: a qualitative insight. Psychooncology. 2003; 12:267–79.
https://doi.org/10.1002/pon.639 PMID: 12673810
13. Cleary V, Hegarty J. Understanding sexuality in women with gynaecological cancer. Eur J Oncol Nurs.
2011; 15:38–45. https://doi.org/10.1016/j.ejon.2010.05.008 PMID: 20584629
14. Bowes H, Jones G, Thompson J, Alazzam M, Wood H, Hinchliff S, et al. Understanding the impact of
the treatment pathway upon the health-related quality of life of women with newly diagnosed endome-
trial cancer—a qualitative study. Eur J Oncol Nurs. 2014; 18:211–7. https://doi.org/10.1016/j.ejon.2013.
10.007 PMID: 24290535
15. Reis N, Beji NK, Coskun A. Quality of life and sexual functioning in gynecological cancer patients:
results from quantitative and qualitative data. Eur J Oncol Nurs. 2010; 14:137–46. https://doi.org/10.
1016/j.ejon.2009.09.004 PMID: 19836305
PLOS ONE | https://doi.org/10.1371/journal.pone.0262844 January 25, 2022
16 / 20
PLOS ONE Multimodal pelvic floor physical therapy for cancer survivors suffering from pain during sexual intercourse
16. Sekse RJ, Raaheim M, Blaaka G, Gjengedal E. Life beyond cancer: women’s experiences 5 years after
treatment for gynaecological cancer. Scand J Caring Sci. 2010; 24:799–807. https://doi.org/10.1111/j.
1471-6712.2010.00778.x PMID: 20487404
17. Vermeer WM, Bakker RM, Kenter GG, Stiggelbout AM, Ter Kuile MM. Cervical cancer survivors’ and
partners’ experiences with sexual dysfunction and psychosexual support. Support Care Cancer. 2016;
24:1679–87. https://doi.org/10.1007/s00520-015-2925-0 PMID: 26412245
18. Bakker RM, Kenter GG, Creutzberg CL, Stiggelbout AM, Derks M, Mingelen W, et al. Sexual distress
and associated factors among cervical cancer survivors: a cross-sectional multicenter observational
study. Psychooncology. 2017; 26:1470–7. https://doi.org/10.1002/pon.4317 PMID: 27862635
19. Stabile C, Gunn A, Sonoda Y, Carter J. Emotional and sexual concerns in women undergoing pelvic
surgery and associated treatment for gynecologic cancer. Transl Androl Urol. 2015; 4:169–85. https://
doi.org/10.3978/j.issn.2223-4683.2015.04.03 PMID: 26816823
20. Abbott-Anderson K, Kwekkeboom KL. A systematic review of sexual concerns reported by gynecologi-
cal cancer survivors. Gynecol Oncol. 2012; 124:477–89. https://doi.org/10.1016/j.ygyno.2011.11.030
PMID: 22134375
21.
Izycki D, Wozniak K, Izycka N. Consequences of gynecological cancer in patients and their partners
from the sexual and psychological perspective. Prz Menopauzalny. 2016; 15:112–6. https://doi.org/10.
5114/pm.2016.61194 PMID: 27582686
22. Carter J, Lacchetti C, Andersen BL, Barton DL, Bolte S, Damast S, et al. Interventions to address sexual
problems in people with cancer: American Society of Clinical Oncology Clinical Practice Guideline
Adaptation of Cancer Care Ontario Guideline. J Clin Oncol. 2018; 36:492–511. https://doi.org/10.1200/
JCO.2017.75.8995 PMID: 29227723
23. Huffman LB, Hartenbach EM, Carter J, Rash JK, Kushner DM. Maintaining sexual health throughout
gynecologic cancer survivorship: a comprehensive review and clinical guide. Gynecol Oncol. 2016;
140:359–68. https://doi.org/10.1016/j.ygyno.2015.11.010 PMID: 26556768
24. Crean-Tate KK, Faubion SS, Pederson HJ, Vencill JA, Batur P. Management of genitourinary syndrome
of menopause in female cancer patients: a focus on vaginal hormonal therapy. Am J Obstet Gynecol.
2020; 222:103–13. https://doi.org/10.1016/j.ajog.2019.08.043 PMID: 31473229
25. Morin M, Carroll MS, Bergeron S. Systematic review of the effectiveness of physical therapy modalities
in women with provoked vestibulodynia. Sex Med Rev. 2017; 5:295–322. https://doi.org/10.1016/j.
sxmr.2017.02.003 PMID: 28363763
26. Wijma AJ, van Wilgen CP, Meeus M, Nijs J. Clinical biopsychosocial physiotherapy assessment of
patients with chronic pain: the first step in pain neuroscience education. Physiother Theory Pract. 2016;
32:368–84. https://doi.org/10.1080/09593985.2016.1194651 PMID: 27351769
27. Cyr MP, Dumoulin C, Bessette P, Pina A, Gotlieb WH, Lapointe-Milot K, et al. Feasibility, acceptability
and effects of multimodal pelvic floor physical therapy for gynecological cancer survivors suffering from
painful sexual intercourse: a multicenter prospective interventional study. Gynecol Oncol. 2020;
159:778–84. https://doi.org/10.1016/j.ygyno.2020.09.001 PMID: 33010968
28. Cyr MP, Dumoulin C, Bessette P, Pina A, Gotlieb WH, Lapointe-Milot K, et al. Changes in pelvic floor
morphometry and muscle function after multimodal physiotherapy for gynaecological cancer survivors
suffering from dyspareunia: a prospective interventional study. Physiotherapy. 2021;In press.
29. Cyr MP, Dumoulin C, Bessette P, Pina A, Gotlieb WH, Lapointe-Milot K, et al. A prospective single-arm
study evaluating the effects of a multimodal physical therapy intervention on psychosexual outcomes in
women with dyspareunia after gynecologic cancer. J Sex Med. 2021; 18:946–54. https://doi.org/10.
1016/j.jsxm.2021.02.014 PMID: 33931347
30. Hlatky MA, Owens DK, Sanders GD. Cost-effectiveness as an outcome in randomized clinical trials.
Clinical trials (London, England). 2006; 3:543–51. https://doi.org/10.1177/1740774506073105 PMID:
17170039
31. Edmunds K, Ling R, Shakeshaft A, Doran C, Searles A. Systematic review of economic evaluations of
interventions for high risk young people. BMC Health Serv Res. 2018; 18:660. https://doi.org/10.1186/
s12913-018-3450-x PMID: 30139384
32.
33.
Fitzpatrick T, Perrier L, Shakik S, Cairncross Z, Tricco AC, Lix L, et al. Assessment of long-term follow-
up of randomized trial participants by linkage to routinely collected data: a scoping review and analysis.
JAMA. 2018; 1:e186019-e. https://doi.org/10.1001/jamanetworkopen.2018.6019 PMID: 30646311
Frawley HC, Dean SG, Slade SC, Hay-Smith EJC. Is pelvic-floor muscle training a physical therapy or a
behavioral therapy? A call to name and report the physical, cognitive, and behavioral elements. Phys
Ther. 2017; 97:425–37. https://doi.org/10.1093/ptj/pzx006 PMID: 28499001
34. Abildgaard JS, Saksvik PØ, Nielsen K. How to measure the intervention process? An assessment of
qualitative and quantitative approaches to data collection in the process evaluation of organizational
interventions. Front Psychol. 2016; 7:1380. https://doi.org/10.3389/fpsyg.2016.01380 PMID: 27713707
PLOS ONE | https://doi.org/10.1371/journal.pone.0262844 January 25, 2022
17 / 20
PLOS ONE Multimodal pelvic floor physical therapy for cancer survivors suffering from pain during sexual intercourse
35. Blackman T, Wistow J, Byrne D. Using qualitative comparative analysis to understand complex policy
problems. Evaluation. 2013; 19:126–40.
36. Hartmann D, Strauhal MJ, Nelson CA. Treatment of women in the United States with localized, pro-
voked vulvodynia: practice survey of women’s health physical therapists. J Reprod Med. 2007; 52:48–
52. PMID: 17286069
37. Dworkin RH, Turk DC, Wyrwich KW, Beaton D, Cleeland CS, Farrar JT, et al. Interpreting the clinical
importance of treatment outcomes in chronic pain clinical trials: IMMPACT recommendations. J Pain.
2008; 9:105–21. https://doi.org/10.1016/j.jpain.2007.09.005 PMID: 18055266
38. Melzack R. The McGill Pain Questionnaire: major properties and scoring methods. Pain. 1975; 1:277–
99. https://doi.org/10.1016/0304-3959(75)90044-5 PMID: 1235985
39. Wiegel M, Meston C, Rosen R. The female sexual function index (FSFI): cross-validation and develop-
ment of clinical cutoff scores. J Sex Marital Ther. 2005; 31:1–20. https://doi.org/10.1080/
00926230590475206 PMID: 15841702
40. Meyer-Bahlburg HF, Dolezal C. The female sexual function index: a methodological critique and sug-
gestions for improvement. J Sex Marital Ther. 2007; 33:217–24. https://doi.org/10.1080/
00926230701267852 PMID: 17454519
41. Santos-Iglesias P, Mohamed B, Walker LM. A systematic review of sexual distress measures. J Sex
Med. 2018; 15:625–44. https://doi.org/10.1016/j.jsxm.2018.02.020 PMID: 29576431
42. Derogatis L, Clayton A, Lewis-D’Agostino D, Wunderlich G, Fu Y. Validation of the female sexual dis-
tress scale-revised for assessing distress in women with hypoactive sexual desire disorder. J Sex Med.
2008; 5:357–64. https://doi.org/10.1111/j.1743-6109.2007.00672.x PMID: 18042215
43. Hopwood P, Fletcher I, Lee A, Al Ghazal S. A body image scale for use with cancer patients. Eur J Can-
cer. 2001; 37:189–97. https://doi.org/10.1016/s0959-8049(00)00353-1 PMID: 11166145
44. McCracken LM, Dhingra L. A short version of the Pain Anxiety Symptoms Scale (PASS-20): preliminary
development and validity. Pain Res Manag. 2002; 7:45–50. https://doi.org/10.1155/2002/517163
PMID: 16231066
45. Sullivan MJL, Bishop SR, Pivik J. The Pain Catastrophizing Scale: development and validation. Psychol
Assess. 1995; 7:524–32.
46. Desrochers G, Bergeron S, Khalife S, Dupuis MJ, Jodoin M. Fear avoidance and self-efficacy in relation
to pain and sexual impairment in women with provoked vestibulodynia. Clin J Pain. 2009; 25:520–7.
https://doi.org/10.1097/AJP.0b013e31819976e3 PMID: 19542801
47. Avery K, Donovan J, Peters TJ, Shaw C, Gotoh M, Abrams P. ICIQ: a brief and robust measure for eval-
uating the symptoms and impact of urinary incontinence. Neurourol Urodyn. 2004; 23:322–30. https://
doi.org/10.1002/nau.20041 PMID: 15227649
48. Price N, Jackson SR, Avery K, Brookes ST, Abrams P. Development and psychometric evaluation of
the ICIQ Vaginal Symptoms Questionnaire: the ICIQ-VS. BJOG. 2006; 113:700–12. https://doi.org/10.
1111/j.1471-0528.2006.00938.x PMID: 16709214
49.
Farrar JT, Young JP Jr., LaMoreaux L, Werth JL, Poole RM. Clinical importance of changes in chronic
pain intensity measured on an 11-point numerical pain rating scale. Pain. 2001; 94:149–58. https://doi.
org/10.1016/S0304-3959(01)00349-9 PMID: 11690728
50. Venegas M, Carrasco B, Casas-Cordero R. Factors influencing long-term adherence to pelvic floor
exercises in women with urinary incontinence. Neurourol Urodyn. 2018; 37:1120–7. https://doi.org/10.
1002/nau.23432 PMID: 29095511
51. Schielzeth H, Dingemanse NJ, Nakagawa S, Westneat DF, Allegue H, Teplitsky C, et al. Robustness of
linear mixed-effects models to violations of distributional assumptions. Methods Ecol Evol. 2020;
11:1141–52.
52. Ashbeck EL, Bell ML. Single time point comparisons in longitudinal randomized controlled trials: power
and bias in the presence of missing data. BMC Med Serv Res. 2016; 16:43. https://doi.org/10.1186/
s12874-016-0144-0 PMID: 27068578
53. Harrell FJ. Regression modeling strategies: with applications to linear models, logistic and ordinal
regression, and survival analysis. 2nd ed. New York: Springer; 2015.
54. Braun V, Clarke V. Using thematic analysis in psychology. Qualitative research in psychology. 2006;
3:77–101.
55. Brennen R, Lin KY, Denehy L, Frawley HC. The effect of pelvic floor muscle interventions on pelvic floor
dysfunction after gynecological cancer treatment: a systematic review. Phys Ther. 2020; 100:1357–71.
https://doi.org/10.1093/ptj/pzaa081 PMID: 32367126
56. Yang EJ, Lim JY, Rah UW, Kim YB. Effect of a pelvic floor muscle training program on gynecologic can-
cer survivors with pelvic floor dysfunction: a randomized controlled trial. Gynecol Oncol. 2012;
125:705–11. https://doi.org/10.1016/j.ygyno.2012.03.045 PMID: 22472463
PLOS ONE | https://doi.org/10.1371/journal.pone.0262844 January 25, 2022
18 / 20
PLOS ONE Multimodal pelvic floor physical therapy for cancer survivors suffering from pain during sexual intercourse
57. Rutledge TL, Rogers R, Lee SJ, Muller CY. A pilot randomized control trial to evaluate pelvic floor mus-
cle training for urinary incontinence among gynecologic cancer survivors. Gynecol Oncol. 2014;
132:154–8. https://doi.org/10.1016/j.ygyno.2013.10.024 PMID: 24183730
58. Bober SL, Recklitis CJ, Michaud AL, Wright AA. Improvement in sexual function after ovarian cancer:
effects of sexual therapy and rehabilitation after treatment for ovarian cancer. Cancer. 2018; 124:176–
82. https://doi.org/10.1002/cncr.30976 PMID: 28881456
59. Damast S, Jeffery DD, Son CH, Hasan Y, Carter J, Lindau ST, et al. Literature review of vaginal steno-
sis and dilator use in radiation oncology. Pract Radiat Oncol. 2019; 9:479–91. https://doi.org/10.1016/j.
prro.2019.07.001 PMID: 31302301
60. Bakker RM, Mens JW, de Groot HE, Tuijnman-Raasveld CC, Braat C, Hompus WC, et al. A nurse-led
sexual rehabilitation intervention after radiotherapy for gynecological cancer. Support Care Cancer.
2017; 25:729–37. https://doi.org/10.1007/s00520-016-3453-2 PMID: 27787681
61. Carter J, Goldfarb S, Baser RE, Goldfrank DJ, Seidel B, Milli L, et al. A single-arm clinical trial investigat-
ing the effectiveness of a non-hormonal, hyaluronic acid-based vaginal moisturizer in endometrial can-
cer survivors. Gynecol Oncol. 2020; 158:366–74. https://doi.org/10.1016/j.ygyno.2020.05.025 PMID:
32522420
62. Carter J, Stabile C, Seidel B, Baser RE, Goldfarb S, Goldfrank DJ. Vaginal and sexual health treatment
strategies within a female sexual medicine program for cancer patients and survivors. J Cancer Surviv.
2017; 11:274–83. https://doi.org/10.1007/s11764-016-0585-9 PMID: 27868156
63.
64.
Li J, Huang J, Zhang J, Li Y. A home-based, nurse-led health program for postoperative patients with
early-stage cervical cancer: a randomized controlled trial. Eur J Oncol Nurs. 2016; 21:174–80. https://
doi.org/10.1016/j.ejon.2015.09.009 PMID: 26482004
Zhu G, Li X, Yang S. Effect of postoperative intervention on the quality of life of patients with cervical
cancer. 2016; 28:819–22.
65. Bahng AY, Dagan A, Bruner DW, Lin LL. Determination of prognostic factors for vaginal mucosal toxicity
associated with intravaginal high-dose rate brachytherapy in patients with endometrial cancer. Int J
Radiat Oncol Biol Phys. 2012; 82:667–73. https://doi.org/10.1016/j.ijrobp.2010.10.071 PMID:
21300451
66. Gondi V, Bentzen SM, Sklenar KL, Dunn EF, Petereit DG, Tannehill SP, et al. Severe late toxicities fol-
lowing concomitant chemoradiotherapy compared to radiotherapy alone in cervical cancer: an inter-era
analysis. Int J Radiat Oncol Biol Phys. 2012; 84:973–82. https://doi.org/10.1016/j.ijrobp.2012.01.064
PMID: 22898381
67.
Turner JA, Franklin G, Heagerty PJ, Wu R, Egan K, Fulton-Kehoe D, et al. The association between
pain and disability. Pain. 2004; 112:307–14. https://doi.org/10.1016/j.pain.2004.09.010 PMID:
15561386
68. Dworkin RH, Turk DC, Peirce-Sandner S, Baron R, Bellamy N, Burke LB, et al. Research design consid-
erations for confirmatory chronic pain clinical trials: IMMPACT recommendations. Pain. 2010; 149:177–
93. https://doi.org/10.1016/j.pain.2010.02.018 PMID: 20207481
69. Goldfinger C, Pukall CF, Thibault-Gagnon S, McLean L, Chamberlain S. Effectiveness of cognitive-
behavioral therapy and physical therapy for provoked vestibulodynia: a randomized pilot study. J Sex
Med. 2016; 13:88–94. https://doi.org/10.1016/j.jsxm.2015.12.003 PMID: 26755091
70. Morin M, Dumoulin C, Bergeron S, Mayrand MH, Khalife´ S, Waddell G, et al. Multimodal physical ther-
apy versus topical lidocaine for provoked vestibulodynia: a multicenter, randomized trial. Am J Obstet
Gynecol. 2021; 224:189.e1–.e12. https://doi.org/10.1016/j.ajog.2020.08.038 PMID: 32818475
71. Dumoulin C, Cacciari LP, Hay-Smith EJC. Pelvic floor muscle training versus no treatment, or inactive
control treatments, for urinary incontinence in women. Cochrane Database Syst Rev. 2018; 10:
Cd005654. https://doi.org/10.1002/14651858.CD005654.pub4 PMID: 30288727
72. Gentilcore-Saulnier E, McLean L, Goldfinger C, Pukall CF, Chamberlain S. Pelvic floor muscle assess-
ment outcomes in women with and without provoked vestibulodynia and the impact of a physical ther-
apy program. J Sex Med. 2010; 7:1003–22. https://doi.org/10.1111/j.1743-6109.2009.01642.x PMID:
20059663
73.
Thomte´ n J, Lundahl R, Stigenberg K, Linton S. Fear avoidance and pain catastrophizing among women
with sexual pain. Women’s health (London, England). 2014; 10:571–81. https://doi.org/10.2217/whe.
14.51 PMID: 25482484
74. Mercier J, Morin M, Tang A, Reichetzer B, Lemieux MC, Samir K, et al. Pelvic floor muscle training:
mechanisms of action for the improvement of genitourinary syndrome of menopause. Climacteric.
2020; 23:468–73. https://doi.org/10.1080/13697137.2020.1724942 PMID: 32105155
75. Goldfinger C, Pukall CF, Gentilcore-Saulnier E, McLean L, Chamberlain S. A prospective study of pelvic
floor physical therapy: pain and psychosexual outcomes in provoked vestibulodynia. J Sex Med. 2009;
6:1955–68. https://doi.org/10.1111/j.1743-6109.2009.01304.x PMID: 19453890
PLOS ONE | https://doi.org/10.1371/journal.pone.0262844 January 25, 2022
19 / 20
PLOS ONE Multimodal pelvic floor physical therapy for cancer survivors suffering from pain during sexual intercourse
76. Mackenzie N. A phenomenological study of women who presented to a physiotherapy-led continence
service with dyspareunia and were treated with trigger point massage. J Assoc Chart Physiother Wom-
en’s Health. 2009; 105:24–39.
77. Schoonenboom J, Johnson RB. How to construct a mixed methods research design. Kolner Z Soz
Sozpsychol. 2017; 69:107–31. https://doi.org/10.1007/s11577-017-0454-1 PMID: 28989188
78.
Irvine FE, Clark MT, Efstathiou N, Herber OR, Howroyd F, Gratrix L, et al. The state of mixed methods
research in nursing: a focused mapping review and synthesis. J Adv Nurs. 2020; 76:2798–809. https://
doi.org/10.1111/jan.14479 PMID: 32896959
79. O’Cathain A, Murphy E, Nicholl J. The quality of mixed methods studies in health services research. J
Health Serv Res Policy. 2008; 13:92–8. https://doi.org/10.1258/jhsrp.2007.007074 PMID: 18416914
80. Pieterse QD, Kenter GG, Maas CP, de Kroon CD, Creutzberg CL, Trimbos JB, et al. Self-reported sex-
ual, bowel and bladder function in cervical cancer patients following different treatment modalities: longi-
tudinal prospective cohort study. Int J Gynecol Cancer. 2013; 23:1717–25. https://doi.org/10.1097/IGC.
0b013e3182a80a65 PMID: 24172106
81. DeSimone M, Spriggs E, Gass JS, Carson SA, Krychman ML, Dizon DS. Sexual dysfunction in female
cancer survivors. Am J Clin Oncol. 2014; 37:101–6. https://doi.org/10.1097/COC.0b013e318248d89d
PMID: 22643563
82. Carpenter KM, Andersen BL, Fowler JM, Maxwell GL. Sexual self schema as a moderator of sexual
and psychological outcomes for gynecologic cancer survivors. Arch Sex Behav. 2009; 38:828–41.
https://doi.org/10.1007/s10508-008-9349-6 PMID: 18418707
83. Andersen BL, Woods XA, Copeland LJ. Sexual self-schema and sexual morbidity among gynecologic
cancer survivors. J Consult Clin Psychol. 1997; 65:221–9. https://doi.org/10.1037//0022-006x.65.2.221
PMID: 9086685
PLOS ONE | https://doi.org/10.1371/journal.pone.0262844 January 25, 2022
20 / 20
PLOS ONE |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.