Prevention & lipidsCross-sectional study
Evaluating Performance and Agreement of Coronary Heart Disease Polygenic Risk Scores
- Cross-sectional study
- multicentre
Authors' conclusion
CHD PRSs that performed similarly at the population level demonstrated highly variable individual-level estimates of risk.
Study context
- Population
- Participants from diverse backgrounds in the All of Us Research Program (AOU), Penn Medicine BioBank (PMBB) and University of California, Los Angeles (UCLA) ATLAS Precision Health Biobank with electronic health record and genotyping data48 polygenic risk scores (PRSs), AOU
- Participants
- 171,095 participants
- Exposure
- Coronary heart disease (CHD) PRSs performing similarly at the population level, compared for individual-level agreement
- Outcome
- Population-level prediction of prevalent CHD (calibration and discrimination); individual-level agreement by ICC and Light κ
Main finding
As reported in the abstract
Scores with practically equivalent Brier scores and AUC in a model of prevalent CHD (region of practical equivalence ±0.02)
46
Participants with at least 1 score in both the top and bottom 5% of risk
Twenty percent
Continuous agreement of individual predictions poor: Intraclass correlation coefficient (ICC)
0.373 (95% CI 0.372-0.375)
Light κ (consistency of risk assignment) did not exceed
0.56
Observational data: associations, not effects.
Further findings
Replication in PMBB (n = 41 193) and ATLAS (n = 53 092)
- Different sets of equivalent scores, also lacking individual-level agreement
Original publication
Summarised from the published abstract; every number is checked against it