Bayesian Inference for Matched Case-Control Studies

Case-control studies occupy a central place in epidemiological research. The basic idea here is the comparison of a group (the “cases”) having the outcome of interest (normally some disease) with respect to a control group (those without disease) regarding one or more characteristics (exposure to certain external events). One of the early examples, which sparked a lot of debate in the fifties between Fisher and Cornfield, is the comparison of persons affected with lung cancer (cases) with persons not affected (controls) in order to find out the effect of smoking (exposure) on lung cancer. Over the years, other important real life examples have emerged, and there is a continuing development of new statistical methodology to address these problems. A very succinct description of many of the important statistical tools for solving case-control problems is given in Breslow and Day (1980). A more recent review article on this subject is due to Breslow (1996) which includes many of the important contributions in this area with an extensive reference list.