Prediction of disease risk using common genetic variants

Abstract

Abstract A frequently proposed application of knowledge on inherited predisposition to cancers is the possibility that susceptible individuals could be targeted for enhanced surveillance, lifestyle advice, or chemoprevention, and thus have reduced risk of diagnosis or mortality from cancer. Action based on knowledge of mutations in high penetrance genes such as BRCA1 in breast cancer, or hMSH2 and colorectal cancer has become established practice in high risk clinics. With the advent of genome-wide association studies (GWAS) dozens of common genetic variants (mostly SNPs) have been discovered to be associated with cancer at various sites, with approximately 30 discovered for prostate cancer alone. In general, the relative risks associated with these risk variants are an order of magnitude or more less than those associated with high penetrance gene mutations, and thus, the risk information contributed by a single common risk variant contributes minimally to risk prediction. However, when cumulated into multi-SNP risk scores, large relative risks are predicted for the small number of persons in the tail of the risk distribution (i.e. those who have inherited mostly risk variants as opposed to “protective” variants). For this small number of people, the estimated relative risks approach those for the high penetrance single gene mutations, but at this stage of our knowledge the risk scores are unstable and may change as more common variants are discovered. For the majority of people, with either modestly elevated or modestly reduced predicted risks, the clinical application of this knowledge is even less secure. However, combinations of risk variants have been shown at some cancer sites to be at least as predictive as available risk scores based on demographic and lifestyle factors. As more variants are discovered, the predictive value of these risk scores will increase, and thus the clinical utility of the information should be rigorously tested.