Familial aggregation study designs: Causes of discrepancies in case-control and reconstructed cohort effect estimates.

摘要

The overall objective of my dissertation is to examine the validity of different study designs used to study familial aggregation of disease with a particular focus on how the reconstructed cohort compares with the traditional cohort and its counterpart, the case-control design. I hypothesize that the family history measure and not bias in the case control design, is responsible for differences in the effect estimates derived from the case-control and reconstructed cohorts studies. I do this because previous researchers have concluded that bias in the case-control design, and possible confounding by family size and age, is responsible for differences in effect estimates derived from then case-control and reconstructed cohort designs. They recommended that the reconstructed cohort design be used as a standard approach for assessing familial aggregation when family history data is collected for cases and controls. I also hypothesize that neither family size nor age of family members defining exposure are potential confounders in the case-control design. I do this because confounding by family size and age was a major rationale for using the reconstructed cohort in lieu of the case-control design when it was initially introduced into the Epidemiological literature.; My findings are consistent with my hypotheses in all three assessments; in a conceptual review of the family history measure and its use in familial aggregation study designs as well as a theoretical demonstration of why family size and age are unlikely confounders in the case-control design. And again in both a series of simulations comparing the traditional cohort, case control and, reconstructed cohort and an assessment of empirical data using the case-control and reconstructed cohort study designs. The results of these assessments demonstrate that the case-control design can consistently provide an equally or more valid estimate of the effect of familial aggregation on disease compared to the reconstructed cohort. This is entirely due to the flexibility of the family history measure used in the case-control design.