Efficient selection of controls for multi-centered collaborative studies of rare diseases.

An increasing number of epidemiologic studies of rare diseases are being conducted as collaborative ventures among several institutions in different geographic areas. Geographic area must be accounted for in the analysis, along with age and other potential confounders. This paper shows that it may not be possible to accurately predict the distribution of persons with rare diseases by age and geographic area. Therefore, selecting controls by matching on age group and geographic area is more efficient than selecting controls in all age and geographic area strata based on prior estimates of expected numbers of cases.