Abstract

Identification of individuals at risk of cardiovascular disease is important for primary prevention strategies such as the prescription of statins to high risk groups. Epidemiological risk scores such as QRISK2 have proved clinically useful for such purposes, but the substantial heritability of cardiovascular disease suggests that genotypic data should provide substantial improvements in predictive accuracy. Genomewide association studies in large cohorts have discovered dozens of genetic variants associated with CVD , but their utility for individual risk prediction has so far proved modest. I will review recent studies of genetic prediction in CVD and present theoretical results suggesting that substantial improvements over epidemiological risk scores may soon be possible. I will discuss applications of genetic prediction to multiple traits at once, and propose some novel statistical measures of accuracy for this situation.