SMP: Genetic Epidemiological Methods (GEM)

Introduction In cross-sectional epidemiological studies, a random sample is usually drawn from the general population in to make conclusions about the underlying general population. This approach assumes that the response probability is the same for all subjects in the population. However, it is known (1) that participants and non-participants differ in terms of their personal and disease-related characteristics. Such variation in response liability may impede the accuracy of a study and may bias any estimates obtained from it. This important problem has not hitherto received adequate attention, particularly not in genetic studies. In survey-based studies, characteristics of non-participants can be derived from information obtained during the contacting process or through a short interview. It has been proposed that participants who enter the study comparatively late show similar characteristics to non-participants (2). This hypothesis appears sensible since late participants would have turned non-participants if the contact period had been shorter or the effort had been less intense. However, whether investigators have to account for a response bias in genetic association studies, because of different allele frequencies in participants and nonparticipants, is as yet unclear. In genetic association studies between genetic variations and complex diseases, the problem of an inhomogeneous population with respect to allele frequencies is known as population stratification. Population stratification occurs, when the population of interest consists of subgroups, that have different allele frequencies for a gene on the chromosomal region of interest. If these subgroups also have different frequencies of a true risk factor, then subgroup membership is a confounder. In recent years, only a small fraction of significant association results has been replicated by other studies. Undetected genetic substructures in the population may be one of the reasons for spurious or biased results. Thus, this project is aimed at assessing the genetic heterogeneity between early and late participants in the S4 Survey (1999/2001) of the KORA study (Cooperative Health Research in the Region of Augsburg), in order to estimate the possible bias introduced into genetic association studies by low response rates. Other factors as age, origin of the parents and the urban-rural difference are also considered as confounding factors.