Dataset Viewer
Auto-converted to Parquet Duplicate
effect
stringlengths
6
23
effectName
stringlengths
6
37
severity
stringclasses
3 values
clinicalPooledRatePct
float64
0.1
20.5
βŒ€
clinicalCiLowPct
float64
0
6
βŒ€
clinicalCiHighPct
float64
1
50
βŒ€
clinicalStatedRates
int64
0
250
clinicalDistinctSources
int64
0
85
clinicalSourceDiversity
stringclasses
3 values
clinicalRateKind
stringclasses
2 values
clinicalIntervalBasis
stringclasses
1 value
clinicalPhase1Sources
int64
0
7
clinicalPhase2Sources
int64
0
3
clinicalPhase3Sources
int64
0
69
clinicalPhase4Sources
int64
0
4
communityReportingFrequencyPct
float64
0
35.7
communityMentions
int64
0
15
communityTotalReports
int64
42
42
nausea
Nausea
mild
20.5
6
50
250
85
very_high
incidence
sampling_plus_between_study
7
3
69
4
33.3
14
42
diarrhea
Diarrhoea
mild
14.7
5
35
237
75
very_high
incidence
sampling_plus_between_study
6
2
64
3
28.6
12
42
constipation
Constipation
mild
11.7
3
36
219
64
very_high
incidence
sampling_plus_between_study
6
1
54
2
21.4
9
42
reduced_appetite
Reduced appetite
mild
8.6
3
23
213
62
very_high
incidence
sampling_plus_between_study
6
1
51
3
31
13
42
headache
Headache
mild
8.4
3
19
198
59
very_high
incidence
sampling_plus_between_study
6
1
49
3
9.5
4
42
pancreatitis
Pancreatitis (rare)
severe
0.1
0
1
13
4
moderate
mixed_incidence_and_serious_ae
sampling_plus_between_study
0
0
4
0
0
0
42
fatigue
Fatigue / Low energy
mild
6.5
2
20
112
34
very_high
incidence
sampling_plus_between_study
2
2
28
2
14.3
6
42
vomiting
Vomiting
moderate
10.4
4
27
236
75
very_high
incidence
sampling_plus_between_study
7
2
62
3
35.7
15
42
abdominal_pain
Abdominal pain
moderate
5.7
2
15
172
42
very_high
incidence
sampling_plus_between_study
6
2
33
1
19
8
42
acid_reflux
Acid reflux (GERD)
mild
5.1
3
10
86
28
very_high
incidence
sampling_plus_between_study
1
1
25
1
19
8
42
gallstones
Gallstones (cholelithiasis)
severe
1.7
1
4
36
13
high
mixed_incidence_and_serious_ae
sampling_plus_between_study
0
0
12
0
9.5
4
42
hair_loss
Hair loss (thinning)
mild
5
2
11
22
11
high
incidence
sampling_plus_between_study
0
0
9
1
2.4
1
42
injection_site_reaction
Injection site reaction
mild
4.5
1
15
20
9
moderate
incidence
sampling_plus_between_study
0
1
8
0
28.6
12
42
dizziness
Dizziness
mild
6.8
3
14
137
41
very_high
incidence
sampling_plus_between_study
4
2
34
1
7.1
3
42
emotional_blunting
Emotional blunting / Reduced pleasure
moderate
null
null
null
0
0
null
null
null
0
0
0
0
2.4
1
42

GLP-1 Side-Effect Pooled Rates (Magistra aggregate table)

A per-effect summary of how often 15 side effects are reported for GLP-1 medications: a pooled clinical rate with its 95% interval, the number of rates and distinct sources behind it, and, separately, how often each effect is mentioned in a screened set of community reports. Built by Magistra from its curated corpus of clinical-trial registry results, papers and regulatory reports. Snapshot as of 2026-10-29 (version 1.5).

Data, not medical advice. Reporting frequency is not incidence β€” do not compare it to the clinical rate as if the two measured the same thing.

Want per-drug rates?

This table pools across drugs. If you need rates per drug (semaglutide oral and injectable, tirzepatide, liraglutide, dulaglutide, exenatide, lixisenatide) with the trial arm behind every figure, see the GLP-1 competitor tolerability benchmark: a free gastrointestinal sample is on the page and as a PDF; the full benchmark (every drug and effect in our corpus, insufficient-data cells marked rather than filled) is a paid product prepared to order. Custom data work and the free API: https://magistra.health/en/data-api

Example row

effect clinicalPooledRatePct 95% interval clinicalStatedRates clinicalDistinctSources
nausea 20.5 6–50 250 85

Columns (glp1-aggregate-rates.csv)

Column Meaning
effect, effectName Effect id and display name
severity Magistra's severity tag: mild / moderate / severe
clinicalPooledRatePct Weighted random-effects pooled rate (%) across eligible clinical and regulatory rate points, one entry per distinct source. Empty when the corpus has no eligible clinical rate for the effect
clinicalCiLowPct, clinicalCiHighPct 95% interval bounds (%)
clinicalStatedRates Rate rows behind the estimate (a registry row is one arm of one trial β€” rows, not independent observations)
clinicalDistinctSources Distinct sources (studies) behind the estimate
clinicalSourceDiversity Bucket of distinct-source count (e.g. moderate / high / very_high)
clinicalRateKind incidence; serious_ae_floor when every source is a registry serious-adverse-event rate (a floor on all-cause incidence); mixed_incidence_and_serious_ae when the pool contains both
clinicalIntervalBasis sampling_plus_between_study normally; sampling_only_single_source when one source contributes (heterogeneity term is 0 by construction β€” do not read the narrow interval as stronger evidence)
clinicalPhase1Sources … clinicalPhase4Sources Distinct ClinicalTrials.gov sources by trial phase. Non-registry sources count toward none, so these need not sum to clinicalDistinctSources
communityReportingFrequencyPct Share of the screened community reports that mention the effect. Not an incidence rate
communityMentions Reports mentioning the effect
communityTotalReports Denominator: distinct screened community reports, the same for every effect (42 in this snapshot)

The JSON file carries the same rows under effects, plus methodologyNote (full definitions), evidentiaryBase (corpus counts), knownDefect, countCaveat, disclaimer and the licence fields.

Corpus behind this snapshot

3,649 data points across the tracked effects; 2,592 state a rate; 1,951 pass the eligibility screen; 88 distinct studies. The point-level corpus itself is not included and is not under this licence.

Limitations

  • Pooled across drugs, doses and populations. These are not per-drug rates and are not placebo-adjusted.
  • Rows are not participants. Some crossover and dose-escalation trials post the same people in several arms; see countCaveat in the JSON for the studies affected and their effect on the counts.
  • Registry thresholds. Registries list non-serious adverse events only above each trial's reporting threshold (usually 5%), which biases low-frequency rates upward.
  • A corrected registry sampling bias (disclosed 2026-09-10, corrected 2026-09-16) is described in knownDefect in the JSON.
  • Community track is small and not growing. Patient-community collection (Reddit) has been blocked since 2026-05-28, so reporting frequencies rest on a fixed, screened set of 42 reports and carry self-selection bias.
  • Emotional blunting has no citable clinical rate; its clinical columns are empty.
  • Dated snapshot. Live figures: https://magistra.health/api/data?q=overview

Method: https://magistra.health/en/methodology Β· Code (Apache-2.0): https://github.com/saurabhgoyal75/magistra-predictor

Licence and citation

CC BY 4.0. Attribution: Magistra, magistra.health. Suggested citation (from the JSON's citation field):

Magistra Health. "GLP-1 Aggregate Pooled Side Effect Rates." magistra.health/data/glp1-aggregate-rates.json. Accessed 2026-10-29.

Published by Phlo Systems B.V. (Magistra). Contact: saurabh@magistra.health

Downloads last month
45