AHA Scientific Sessions 2026 Abstract · Nov 2, 2026
Performance of Five Atrial Fibrillation Prediction Polygenic Risk Scores Across Sex and Race/Ethnicity Subgroups in the All of Us Research Program
By Le Q, Gupta A, Sileshi R et al.
Background: The comparative performance of atrial fibrillation (AF) polygenic risk scores (PRS) remains understudied in diverse populations. We compared five published AF PRS within the All of Us Research Program. Methods: In this retrospective case-control study, AF cases were matched 1:1 to controls on logit propensity score derived from all 11 CHARGE-AF clinical components (caliper 0.2 SD) in the All of Us Controlled Tier Dataset v8 with whole-genome sequencing (total 368,388 participants). Five PRS were evaluated: Miyazawa AF-PRS MA (4,520 variants, cross-ancestry), Roselli PRS-CS MA (1,113,668 variants, PRS-CS), Lubitz 12-SNP score (12 variants, curated loci), Khera GPS AF (6,728,187 variants, LDpred), and Weng PRS AF (1,168 variants, pruning-and-thresholding). Discrimination of cases from controls by each PRS was quantified as the odds ratio (OR) per 1 SD increase and areas under the ROC curves (AUC) were compared using DeLong tests. Subgroup analyses were performed by sex and race/ethnicity. Results: Propensity matching yielded 19,395 matched pairs (N = 38,790, females 45.4%, non-Hispanic Blacks 13.5%, Hispanics 10.0%). Overall, all five PRS were associated with AF (all P < 0.001). Miyazawa (OR per SD 1.48 [95% CI 1.45-1.52], AUC 0.607) and Roselli (1.47 [1.44-1.50], 0.606) demonstrated the strongest discrimination (ΔAUC P = 0.58), and outperformed 12-SNP (AUC 0.565), Khera (0.562), and Weng (0.531) (all ΔAUC P < 0.001). Discrimination by PRS was attenuated for females (best Miyazawa 1.40 [1.36-1.45], 0.593 and Roselli 1.39 [1.35-1.43], 0.591), non-Hispanic Blacks (best Miyazawa 1.22 [1.16-1.29], 0.557 and Roselli 1.21 [1.14-1.27], 0.554) and Hispanics/Latino (best Roselli 1.44 [1.35-1.54], 0.597 and Miyazawa 1.40 [1.31-1.50], 0.592). Conclusions: In a diverse population, cross-ancestry large-scale GWAS-derived AF PRS outperform earlier scores, providing risk stratification independent of clinical factors. Discrimination of AF by PRS was weaker in females, Blacks, and Hispanics, underscoring the need for female- and race/ethnicity-specific PRS.