Methodological Problems in Multigenerational Epidemiology
While epidemiology has typically focused on how exposures impact the individuals directly exposed, recent interest has been shown in investigating exposures with multigenerational effects—ones that affect the children and grandchildren of those directly exposed. For example, a recent motivating study examined the association between maternal in-utero diethylstilbestrol exposure and ADHD in the Nurses Health Study II. Such multigenerational studies, however, are susceptible to informative cluster size, occurring when the number of children to a mother (the cluster size) is related to their outcomes. But what if some women have no children at all? We first consider this problem of informatively empty clusters. Second, observing populations across multiple generations can be prohibitively expensive, so multigenerational studies often measure exposures retrospectively—and hence are susceptible to misclassification and recall bias. We thus study the impact of exposure misclassification when cluster size is potentially informative, as well as when misclassification is differential by cluster size. Finally, outside the relative control of laboratory settings, population-based multigenerational studies have had to entertain a broad range of study designs. We show that these designs have important implications on the scope of scientific inquiry, and we highlight areas in need of further methodological research.