Improving Genetic Risk Prediction of CAD in Chinese by Multi-ancestry and Multi-trait GWAS Integration

While many polygenic risk scores (PRSs) of coronary artery disease (CAD) have been developed to stratify disease risks in Europeans, their performances in the Chinese population are suboptimal due to population heterogeneity. Considering the complex genetic architecture of CAD, we train a multi-ancestry multi-trait PRS for CAD, termed PRSCAD+, which is optimized in a prospective Chinese cohort to integrate information from large-scale genome-wide association studies (GWASs) of CAD and 15 related traits in East Asians and Europeans. The hazard ratio (HR) for incident CAD is 1.26 (95% confidence interval: 1.21-1.31) per standard deviation of PRSCAD+, which is stronger than published PRSs, East Asian-specific multi-trait PRS, and the single-trait PRS of CAD. PRSCAD+ also increases the concordance index (C-index) by an average of 1.1% over 14 published PRSs (ΔC ranges from 0.4% to 1.6%; P < 0.05). Addition of PRSCAD+ to traditional clinical risk factors led to a significant improvement in the C-index by 1.3% (P < 0.05). In an external validation set (mean age 58 years), PRSCAD+ achieved an odds ratio of 2.40 (2.18-2.65) and an area under the receiver operating characteristic curve of 0.799 (0.782-0.816) for predicting early-onset CAD. Furthermore, we observed significant gradients across the quintiles of PRSCAD+ in both datasets. These results demonstrate that PRSCAD+ can improve risk prediction and stratification of CAD in the Chinese population by incorporating genetic information of related traits from both European and East Asian studies.