The first step in the design of a case-control study is to define the disease or phenotype of interest as accurately and specifically as possible. This is important because non-specific case definitions will increase both the genetic and environmental heterogeneity in underlying causal factors, and can therefore drastically decrease the power of detection of an effect. Secondly, replication of the study (a crucial part of the validation of the results found) will become impossible if the phenotype has not been adequately defined. Often a balance has to be achieved between phenotype definitions that are seen as clinically relevant (but which may not be highly specific) and those that are seen biologically relevant (which may be more specific but less clinically relevant). Such definitions are likely to change historically, as more clinical and biological information becomes available. For example, the diagnosis of the main subtypes of diabetes has grown more specific over the years, from ‘early-onset vs. adult-onset’ to ‘insulin-dependent vs non-insulin dependent’ to Type I/Type II, and most recently Type1/Type 2 15;16. Even the most recent definition of Type 2
