Clinical Utility of Multimarker Genetic Risk Scores for Prediction of Incident Coronary Heart Disease

Abstract

Background— We evaluated whether including multilocus genetic risk scores (GRSs) into the Framingham Risk Equation improves the predictive capacity, discrimination, and reclassification of asymptomatic individuals with respect to coronary heart disease (CHD) risk. Methods and Results— We performed a cohort study among 51 954 European-ancestry members of a Northern California integrated healthcare system (67% female; mean age 59) free of CHD at baseline (2007–2008). Four GRSs were constructed using between 8 and 51 previously identified genetic variants. After a mean (±SD) follow-up of 5.9 (±1.5) years, 1864 incident CHD events were documented. All GRSs were linearly associated with CHD in a model adjusted by individual risk factors: hazard ratio (95% confidence interval) per SD unit: 1.21 (1.15–1.26) for GRS_8, 1.20 (1.15–1.26) for GRS_12, 1.23 (1.17–1.28) for GRS_36, and 1.23 (1.17–1.28) for GRS_51. Inclusion of the GRSs improved the C statistic (ΔC statistic =0.008 for GRS_8 and GRS_36; 0.007 for GRS_12; and 0.009 for GRS_51; all P <0.001). The net reclassification improvement was 5% for GRS_8, GRS_12, and GRS_36 and 4% for GRS_51 in the entire cohort and was (after correcting for bias) 9% for GRS_8 and GRS_12 and 7% for GRS_36 and GRS_51 when analyzing those classified as intermediate Framingham risk (10%–20%). The number required to treat to prevent 1 CHD after selectively treating with statins up-reclassified subjects on the basis of genetic information was 36 for GRS_8 and GRS_12, 41 for GRS_36, and 43 for GRS_51. Conclusions— Our results demonstrate significant and clinically relevant incremental discriminative/predictive capability of 4 multilocus GRSs for incident CHD among subjects of European ancestry.