Integrated multiomic scores for cardiovascular disease prediction across multiancestral populations

Background: Genomic, epigenomic, and proteomic data capture complementary aspects of cardiovascular disease biology, but the value of integrating these data for predicting multiple cardiovascular disease outcomes across ancestry groups remains uncertain. Methods: We derived disease-specific proteomic scores using 999 plasma proteins in 50,673 UK Biobank participants and validated them in 9,358 participants from the Prospective Urban and Rural Epidemiological (PURE) study. In PURE, we integrated these proteomic scores with 2,279 polygenic scores and 479 DNA methylation scores to develop 4 multiomic scores (MEGASCOREs) for myocardial infarction (MI), atrial fibrillation (AF), heart failure (HF), and cardiovascular mortality. Associations with incident cardiovascular disease, discrimination, calibration, and performance across ancestry groups were evaluated in PURE and 2 additional cohorts. Results: In pooled analyses, each standard deviation increase in the corresponding MEGASCORE was associated with MI (hazard ratio [HR], 2.13; 95% CI, 1.86-2.43), AF (HR, 1.94; 95% CI, 1.68-2.24), HF (HR, 3.33; 95% CI, 2.75-4.05), and cardiovascular mortality (HR, 3.28; 95% CI, 2.33-4.61). MEGASCOREs improved discrimination compared with clinical models, with strongest improvements for HF and cardiovascular mortality. MEGASCORE-CV-Death had a higher area under the receiver operating characteristic curve (AUC) than SCORE2 for cardiovascular mortality (0.83 [95% CI, 0.82-0.85] vs. 0.78 [95% CI, 0.77-0.80]). Its discrimination was also higher than that of SCORE2 across all 4 ancestry groups, with better calibration across non-European ancestry groups. MEGASCOREs remained associated with cardiovascular outcomes among participants without conventional cardiovascular risk factors. Conclusions: Integrated multiomic MEGASCOREs were associated with multiple cardiovascular outcomes and provided prognostic information beyond conventional clinical variables in PURE.