|| Checking for direct PDF access through Ovid
Candidate gene studies have become very popular but some of their implicit constraints, such as the familial risk and the population attributable fraction (PAF) conferred by the gene under study, are poorly understood. We model here these parameters for susceptibility genes in terms of genotype relative risk (GRR), allele frequency and statistical power in simulated genetic association studies, assuming 500 or 2000 case–control pairs and different modes of inheritance. The results show that the common association studies on genes with minor allele frequency >10% have sufficient power to detect disease-causing variants conferring PAFs >10%, which can be compared to known genes, such as BRCA1 with a PAF of 1.8%. Yet, common low-risk variants confer low familial relative risks (FRRs), typically <1.1. The models show that candidate gene studies may be able to identify genes conferring close to 100% of the PAF, but they may not explain the empirical FRRs. In order to explain FRRs, rare, high-penetrant genes or interacting combinations of common variants need to be uncovered. However, the candidate gene studies for common alleles do not target this class of genes. The results may challenge the common disease–common variant hypothesis, which posits common variants with low GRRs and large PAFs, however failing to accommodate the empirical FRRs.