Practical Considerations in Choosing between the Case-Cohort and Nested Case-Control Designs

When the outcome is rare, both the nested case-control and case-cohort designs can provide economical esti mates of relative risk parameters under the proportional hazards model, while requiring exposure and covariate information on only a small subset of the cohort. Comparisons of the statistical efficiency of the case cohort and nested case-control designs for studies with substantial late entry or censoring suggest a small to moderate advantage for the case-control study1-3; in contrast, the results for studies with little late entry or censoring1'4,5 slightly favor the case-cohort design. Below, I discuss some practical points that have been relevant in several studies of etiologic factors for cancer where the analysis requires tight control or matching on a time variable. When control for time is not essential, as in studies of perinatal mortality, the case-cohort design is identical to the case-base design.6'7 The primary consideration for choosing between a case-control and a case-base design seems to be whether the odds ratio or the risk ratio is