Coronary artery disease (CAD) has a polygenic basis, and identification of CAD susceptibility genes has the potential to aid the development of new treatments and enhance prediction of disease risk. Thus far, the strategy has firstly been to choose “candidate” genes coding for important “rate-limiting” proteins in the homeostatic systems involved in maintaining cardiovascular health; secondly to identify common variants in these candidate genes; thirdly to carry out genotyping and statistical analysis using genetic association studies; and finally to test the functional effects of the identified variants in vitro and in vivo. However, lack of reproducibility of genetic association studies has led to uncertainty about the nature and number of genes involved. In part this is because many of the studies conducted have not been adequately powered to detect small risk effects, or to permit adequate exploration of gene-gene or gene-environment interactions in a robust manner. Spurious positive and negative associations due to type I and type II statistical errors are likely to co-exist with real associations in the published literature. By utilising all available data to increase statistical power, meta-analysis of genetic association studies is increasingly being used to identify genotypic risk with a greater degree of precision. Though potentially powerful, this approach may be prone to publication bias. Therefore, very large genetic association studies will also be required to identify risk genes for CAD. This review lays out the framework for the candidate gene approach for CAD and illustrates this with published results from a UK prospective study of 3000 middle-aged men.