Outcomes in Cohort Studies

This is the authors' abstract. We don't add key points for this paper.

The initial intellectual appeal of the cohort design for epidemiologic studies stems from the obvious fact that the method of subject selection leaves little doubt about the temporal relation between the exposure and the disease whose relation is under study. The inferential power of this fact clearly was realized by Doll and Hill in their landmark 1954 publication (1) in which they explicitly acknowledged the need to observe the future outcome (lung cancer) of disease in those whose exposure (cigarette smoking) is already known antecedent to the disease. Minimization of recall bias (in relation to exposure) and selection bias based on exposure, and to some extent on disease outcome, are benefits of cohort studies, although modern case-control studies that are based on techniques of risk-set sampling from cohorts or case-cohort designs have blurred some of the distinctions between the benefits of cohort versus case-control studies (2) with regard to issues of temporal sequence and selection bias. Cohort studies nonetheless retain the unique capacity, vis-a-vis any type of sampling design for casecontrol studies, to evaluate multiple outcomes as a consequence of given exposure, to study temporal processes of disease onset and progression (natural history study), and/or the biologic/physiologic processes that underlie the subsequent occurrence of a disease.