Whether or not controls should be totally unselected on the basis of other phenotypes (i.e. derived from the general population), or should (also) comprise a mixture of other case groups with unrelated conditions, is currently a matter for debate; it is likely that the optimum solution will vary from one disease to the other. When cases are recruited from clinics, the use of clinic controls diagnosed with other conditions has the advantage of matching for health-seeking and socio-economic characteristics that may be associated with population substructure; however, selective inclusion of other diseases amongst controls has the potential to increase the false positive rate. When centre-based recruitment of controls is not feasible, investigators often use groups of already genotyped ‘common’ controls. In particular, genome-wide association studies (e.g. the Wellcome Trust Case Control Consortium – http://www.wtccc.org.uk/ 14) are likely to use this approach as it is much more economical. It is important that basic characteristics of such panels are known, such as ethnicity, sex, and age, and if possible area of recruitment, so that they can be matched for in the design
