The arguments made by Klatsky and colleagues point to the fact that some studies are not readily categorized into distinct epidemiologic study designs. Their study1 bears the hallmarks of a case-cohort design, but did not appear to have been analyzed as such.2 While the study of Klatsky et al1 did possess some elements of a cohort study, it is not "clearly a cohort study." It is worth noting that this group's earlier study,3 with a similar population, methods of ascertaining coffee drinking, subject follow-up, and identification of cases was categorized as a case-control study. The earlier study differed from the subsequent one only in the number and matching of control subjects to cases.
In fact, we did not omit the design of Klatsky et al1 from our meta-analysis.4 In accordance with standard meta-analyses, all possible relevant studies were carefully considered and analyzed. Those that