xpertsystems/hc-end-007-sample

HC-END-007 — Osteoporosis & Metabolic Bone Disease Synthetic Dataset (Sample)

XpertSystems.ai · Synthetic Data Factory · Endocrinology Vertical

A deeply trial-calibrated synthetic cohort of osteoporosis and metabolic bone disease patients spanning DXA bone mineral density (lumbar/femoral neck/total hip), WHO classification, FRAX & Garvan fracture risk, bone turnover markers (CTX, P1NP, osteocalcin, BSAP, sclerostin, RANKL/OPG), treatment efficacy across nine drug classes, secondary osteoporosis workup, fall-risk & neuromuscular assessment, and incident-fracture clinical outcomes. This repository contains a 500-row, single-seed sample. The full commercial product scales to 20,000+ patients with 20-year longitudinal trajectories and CSV / Parquet / JSON / FHIR R4 delivery.

Per-drug BMD gains and fracture risk ratios are hard-anchored to landmark RCTs (FIT, HORIZON, FREEDOM, FRAME, ARCH) and applied directly — so drug-specific benchmarks reproduce the trial literature precisely.

• SKU: HC-END-007
• Sample size: 500 patients × 144 columns
• License (sample): CC-BY-NC-4.0 — commercial license available for the full product
• Contact: pradeep@xpertsystems.ai · https://xpertsystems.ai

---

Validation

This sample passes XpertSystems Grade A+ validation (overall 10.000 / 10) with deterministic reproduction across all six canonical seeds [42, 7, 123, 2024, 99, 1].

Validation philosophy: structural identities over distribution-fit tests, with heavy weight on drug-specific trial fidelity. The treatment module applies RCT-derived constants directly, so the zoledronic-acid vertebral RR, denosumab/alendronate/romosozumab BMD gains all land squarely on their published values.

Calibration anchors

Metric	Sample value	Target range	Source
Osteoporosis prevalence	61.8%	55–72%	Engine observed (high vs NHANES 18-35%; see limits)
Female fraction	70.8%	65–78%	Osteoporosis female predominance ~72%
Vitamin D deficiency	30.0%	28–55%	US 50+ population ~30-55%
Treatment rate	60.0%	50–80%	Treatment among eligible ~55-75%
Zoledronic acid vertebral RR	0.30	0.25–0.38	HORIZON 2007 (~0.30)
Denosumab spine BMD gain (3yr)	9.1%	7.5–11.5%	FREEDOM 2009 (~9.2%)
Alendronate spine BMD gain (3yr)	8.0%	6.5–10.5%	FIT 1996 (~8%)
Romosozumab spine BMD gain	13.5%	10–16.5%	FRAME 2017 (~13.3%)
T-score femoral neck in [-5,3.5]	100%	≥1.0	DXA physiology bounds
FRAX hip in [0.1,35]%	100%	≥1.0	FRAX probability bounds
BMD lumbar in [0.45,1.60]	100%	≥1.0	DXA physiology bounds
Column count	144	≥138	Schema completeness (8 modules)

---

Schema highlights by module (144 columns)

Demographics. Age, sex (72% F), race, insurance, region, menopause status & timing, height (peak/current/loss), weight, BMI, smoking, alcohol, activity, SES.

Bone mineral density. Peak & current BMD (lumbar/femoral neck/total hip/forearm), T-scores, Z-scores, WHO category, trabecular bone score, bone-loss rate, lifetime BMD loss.

Fracture risk. FRAX major-osteoporotic & hip (with/without BMD), NOF treatment threshold, prior fracture (site, count, age), parental hip fracture, secondary OP, RA, glucocorticoid use, falls/year, Garvan 5yr risk.

Secondary osteoporosis. Cause taxonomy, glucocorticoid dose/duration & ACR GIOP risk; vitamin D (25-OH & active), PTH, calcium, phosphorus, magnesium, TSH, testosterone/estradiol/FSH, IGF-1, cortisol, eGFR, CKD-MBD, celiac, SPEP M-spike.

Bone turnover markers. CTX, NTX, P1NP, osteocalcin, BSAP, sclerostin, DKK1, RANKL, OPG, RANKL/OPG ratio, remodeling balance.

Treatment. Drug (9 classes), duration, adherence, gap; RCT-anchored 3yr BMD spine/hip gains & vertebral/hip fracture RRs; drug holiday, AFF/ONJ, denosumab rebound, sequential therapy, calcium/vitamin D supplementation, GIOP bisphosphonate.

Fall risk & neuromuscular. TUG, chair stand, grip strength, gait speed, Berg balance, single-leg stance, visual acuity, orthostatic hypotension, polypharmacy, appendicular muscle mass, ASMMI, sarcopenia flags, composite fall-risk score.

Clinical outcomes. Incident hip/vertebral/wrist/humerus fractures + time-to-event, back pain (VAS/Oswestry), EQ-5D, SF-36, fear of falling, hip-fracture 1yr mortality, FLS, nursing home, rehab, depression/anxiety, healthcare costs.

Coding. ICD-10, LOINC; FHIR R4 DiagnosticReport bundle (full product).

---

Files

• hc_end_007_sample.csv — 500-patient sample (144 columns)
• generate_sample_dataset_hc_end_007.py — reproducible generator + validation harness
• validation_report.json / validation_report.md — full scorecard
• sweep_summary.json — 6-seed determinism results

Loading

import pandas as pd
df = pd.read_csv("hc_end_007_sample.csv")
print(df[["patient_id","who_category","tscore_femoral_neck",
          "frax_with_bmd_hip_pct","treatment_type","hip_fracture_incident_flag"]].head())

from datasets import load_dataset
ds = load_dataset("csv", data_files="hc_end_007_sample.csv")

Use cases

• Fracture-risk prediction (FRAX/Garvan replication, BMD + clinical risk factors)
• Treatment-efficacy and comparative-effectiveness modeling across drug classes
• Time-to-event / survival analysis on incident hip & vertebral fractures
• Fall-risk and sarcopenia screening tooling
• Health-economics modeling (fracture episode costs, FLS impact)
• ML training where real osteoporosis EHR + DXA data is PHI-restricted

---

Honest limitations & disclosed generator behavior

This engine has standout drug-level trial fidelity; the main caveat is cohort composition.

1. Osteoporosis prevalence runs high (~62%). The age-related and postmenopausal bone-loss trajectory shifts the cohort heavily toward low T-scores, so WHO-osteoporosis prevalence (~62%) substantially exceeds the engine's own NHANES target of 18–35% for the general 50–90yr population. Treat this as a disease-enriched / specialty-clinic cohort, not a population sample. The median femoral-neck T-score is ~−1.9 (osteopenic-to-osteoporotic range).
2. Drug benchmarks are constants, not learned. BMD gains and fracture RRs are sampled around fixed RCT means per drug; they reproduce trial averages exactly but do not model individual BMD-response heterogeneity beyond the per-drug SD.
3. Longitudinal is summary-level. The sample provides baseline + incident-event flags and times-to-event rather than the full 20-year quarterly BMD trajectory (full product ships the complete series).
4. Some flags drawn independently. Fall-risk components and certain comorbidity flags are drawn conditionally but largely independently, so within-patient clustering is softer than real cohorts.

General caveat: cross-field correlations beyond those explicitly modeled may be weaker than in real cohorts. Not for clinical decision-making — research/development use only.

---

Commercial product comparison

Capability	This sample	Full HC-END-007 product
Patients	500	20,000+ (configurable)
Longitudinal	Baseline + incident events	20-year quarterly BMD + fracture trajectory
Seeds / cohorts	1	Multi-seed, reproducible
Formats	CSV	CSV + Parquet + JSON + FHIR R4 Bundle
Cohort composition	Disease-enriched (~62% OP)	Tunable to population NHANES distribution
BMD response	Per-drug constant + SD	Individual response-heterogeneity model
License	CC-BY-NC-4.0	Commercial
Support & SLA	—	Included

Full product, custom cohorts, or other endocrinology SKUs: pradeep@xpertsystems.ai

---

Citation

@dataset{xpertsystems_hc_end_007_2026,
  title        = {HC-END-007: Osteoporosis & Metabolic Bone Disease Synthetic Dataset},
  author       = {XpertSystems.ai},
  year         = {2026},
  publisher    = {XpertSystems.ai Synthetic Data Factory},
  url          = {https://xpertsystems.ai},
  note         = {Synthetic; CC-BY-NC-4.0 (sample). Drug efficacy hard-anchored to: FIT
                  (alendronate; Black et al. 1996); HORIZON (zoledronic acid; Black et al.
                  2007, NEJM); FREEDOM (denosumab; Cummings et al. 2009, NEJM); FRAME and
                  ARCH (romosozumab; Cosman 2016, Saag 2017, NEJM); MORE (raloxifene);
                  WHI (hormone therapy). BMD reference: NHANES III (Looker et al. JBMR 1995).
                  Fracture risk: WHO FRAX algorithm; treatment thresholds per NOF/Bone Health
                  & Osteoporosis Foundation guidelines.}
}

Synthetic data generated by XpertSystems.ai. Not derived from real patient records. Not for clinical use.