In a recent article on the risk of selected gastrointestinal cancers in relation to ranitidine use, Adami and colleagues took their (seemingly “negative”) results to imply that they “should be reassuring for millions of concerned past users of ranitidine” (1). In my view, however, the study suffered from some important limitations complicating the causal-inference process and making the results much less reassuring. Although the authors state that the design of their “cohort study … emulated a randomized trial,” I do not believe the emulation attempt was successful. While using active-comparator control in nonexperimental studies does help prevent confounding, it can hardly achieve this goal as well as a trial using not only active-comparator control but also randomization coupled with blinding. Thus, a nonexperimental study aiming to minimize confounding would, in addition, require an extremely thorough statistical confounding control. However, the authors seemed to have data on only a...

You do not currently have access to this content.