Susceptibility and outcome in complex disorders such as asthma and cancer appear to be determined, at least in part, by genetic polymorphism. However, while our ability to identify new allelic variants and study them in case and control populations has greatly improved, considerable difficulties remain in elucidating how many genes determine particular clinical phenotypes. This is because most studies have concentrated on study of single genes in relatively small study groups. The important issues of gene-gene interactions (epistasis) and high-risk subgroups have not yet been adequately addressed. We now describe a general approach, using patients with head and neck cancers as an example. Our purpose is to demonstrate candidate gene selection, statistical approaches, and identification of patient subgroups.