Betting Odds and Genetic Associations

Open Access
Publisher Website
Google Scholar
Abstract
“Multiple Comparisons? No Problem!” read the title of an editorial in an issue of Epidemiology in 1991 (1). Attitudes toward multiple comparisons have swung like a pendulum, the view quoted above representing one end of a cycle that has been reversed in recent years. Few classical risk factor epidemiology studies test more than a few hundred associations between exposures and diseases, even in an exploratory mode. In molecular epidemiology, however, the growing feasibility of testing many thousands, or even millions, of single nucleotide polymorphisms (SNPs) associations in genome-wide studies have forced even the most liberal investigators to re-examine their positions on multiple hypothesis testing (2,3). These concerns have been reinforced by the observation that an alarming proportion of reported associations between genetic variants and diseases are not replicated (4–6).