Incremental predictive value of 152 single nucleotide polymorphisms in the 10-year risk prediction of incident coronary heart disease: the Rotterdam Study.

Abstract

To examine the incremental predictive value of genetic risk scores of coronary heart disease (CHD) in the 10-year risk prediction of incident CHD. In 5899 subjects, we used 152 single nucleotide polymorphisms (SNPs) associated with coronary artery disease by the CARDIoGRAMplusC4D consortium to construct three weighted genetic risk scores: (i) GRS(gws) based on 49 genome-wide significant SNPs; (ii) GRS(fdr) based on 103 suggestively associated SNPs; and (iii) GRS(all) based on all 152 SNPs. We examined the changes in discrimination and reclassification of incident CHD when adding the genetic risk scores to models including traditional risk factors. We repeated the analysis for prevalent CHD. The genetic risk scores were associated with incident CHD despite adjustment for traditional risk factors and family history: participants had a 13% higher rate of CHD per standard deviation increase in GRS(all). GRS(all )improved the C-statistic by 0.006 [95% confidence interval (CI): 0.000, 0.013] beyond age and sex, 0.003 (95% CI: -0.001, 0.008) beyond traditional risk factors and 0.003 (95% CI: -0.001, 0.007) beyond traditional risk factors and family history. The genetic risk scores did not improve reclassification. GRS(all) strongly improved both discrimination and reclassification of prevalent CHD, even beyond traditional risk factors and family history, with a C-statistic improvement of 0.009 (0.003, 0.015). Although the genetic risk scores based on 152 SNPs were associated with incident CHD, they did not improve risk prediction. This discrepancy may be the result of SNP discovery for prevalent rather than incident CHD, since the SNPs do improve prediction for prevalent disease.