Case-control and case-cohort study designs: methodological considerations

Abstract

The "case-control" study is not a homogeneous entity in epidemiology. This article presents a narrative review of characteristics that differentiate "traditional case-control", "population-based case-control", "nested case-control", and "case-cohort" studies. The use of a secondary basis for the choice of controls makes case-control studies particularly vulnerable to selection biases and limits the representativeness of the control group relative to the study population base. The choice of prevalent cases, although useful for maximizing statistical power, undermines the interpretation of associations since determinants of incidence are mixed with determinants of disease duration. "Nested case-control" and "case-cohort" studies partly overcome these limitations because they are based on a true cohort and often use incident cases. "Case-cohort" studies give emphasis to the contrast between case exposure experience and population-based exposure experience, providing additional benefits when there is temporal matching between cases and the reference group. This article provides insights into the methodological arsenal of case-control studies and highlights some of the methodologies discussed - particularly those in which controls are included - favors the use of secondary databases without the need for a key variable that correlates them.