Abstract
Pharmacoepidemiological studies are often conducted to evaluate the association between the event with low incidence and exposure with low prevalence. To overcome the difficulty due to low event incidence and low exposure prevalence, White proposed a two-stage design in 1982 and Cain and Breslow further developed this design.In this article, the case-cohort and nested case-control designs are reviewed in contrast with the two-stage design. In addition, the usefulness of the case-cohort design in pharmacoepidmiology is assessed in comparison with that of the nested case-control design. In the nested case-control study, the control subjects are selected only after cases occur irrespective of whether the method of “risk set matching” or “unmatched density sampling” is used. Therefore, all of the events evaluated in the study must be clearly defined in advance. On the other hand, in a prospective case-cohort study, a single subcohort, selected independent of cases, is used to analyze multiple outcomes. Owing to this feature, the case-cohort design may be useful to study unknown adverse events that have not been specified as a target event prior to the study but are recognized as a problem that requires in-depth evaluation during or after the study is conducted. Weaknesses and limitations of the case-cohort design as compared to the nested case-control design are also discussed.Data yielded by simulations for hypothetical case-cohort studies are analyzed by the SAS PHREG procedure with the robust variance estimation. The program used and results of simulations are presented.The case-cohort study design may be useful for various pharmacoepidemiological studies in Japan where no large medical database is available.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have