The problem of biomedical fraud: A model for retrospective and prospective action

Peer review of scientific manuscripts is geared towards the evaluation oj individual submissions for their merit, novelty, and potential importance; it is not designed to detect scientific fraud. The peer-review process is particularly insensitive to the detection of fraud when papers are individually credible but combine to form a body of work that is suspect when taken as a whole. Rhetorical analysis provides a tool for the investigation of suspected fraud in such situations, though it is labourintensive enough that it cannot be undertaken lightly and thus cannot alone serve as an effective fraud deterrent. However, we suggest that it can be combined with measures designed to make fraud more liable to detection, such as the routine agreement of a submitting author to submit to data audit if the authenticity of reported work is challenged. Finally, when the circumstances suggest impropriety, but fraud cannot be proven beyond a doubt - or cannot be proven for the specific papers a journal has published - withdrawal of journal aegis provides a mechanism short of outright retraction for making the concern public.