Typically, a GWAS uses a case–control design, in which cases are ascertained based on the trait of interest [48]. The choice of control subjects is often less obvious. A control subject should be disease free, but must also be free of other traits that are not shared by cases. For example, if the phenotype of interest is known to manifest within a certain age, it is tempting to choose controls that are much older than the cases to guarantee that they never develop the phenotype. This choice for controls would introduce confounding, and the discovered associations may be related to aging rather than the trait of interest. In general, matching controls on variables that are not of interest in the study, such as exposure to some environmental conditions, is important to avoid introducing spurious associations. However, some caution is necessary to avoid overmatching and losing of generality of the results or missing important associations [49]. Controls selected from families of the cases often offer protection against confounding by matching exposures to risk factors, or genetic background when controls are genetically related to the cases, this being the rationale of family based association studies [50].
