Selecting control groups for studies of familial aggregation of disease

In genetic-epidemiologic studies to determine the association between the disease status of family members, this association is measured by comparing rates of the disease in relatives of probands (index cases) with the disease, with the rates of the disease among individuals in a control group. Either of two types of control groups are generally used: (1) a control group consisting of a random sample from the population or the entire population if available or (2) a control group consisting of relatives of individuals without the disease under study. We examine the advantages and disadvantages of using these different types of control groups. We show two major results for family studies: (1) when there are no other factors associated with the disease status of an individual other than the disease status of a family member, both types of control groups will give a valid test of the null hypothesis of no familial aggregation. However, tests using a population control group will always be less efficient statistically, than those performed with a control group of relatives of probands without the disease under study, the degree of efficiency decreasing with increasing population prevalence of the disease. (2) When factors other than the disease status of a family member are also associated with the disease status of an individual, if this factor is a proband characteristic (which is not shared by relatives) population control groups cannot be adjusted to eliminate possible bias due to the potential confounding effect of this factor (unlike control groups consisting of relatives of probands without the disorder).