There is a continuing debate in the popular and medical literature about the merit of reporting positive results—that is, findings supporting an association between an exposure and a health outcome—generated by observational epidemiologic studies. Articles in the popular press have alleged that the resolving power of observational epidemiology is severely limited, with the result that positive results are often wrong and lead to unnecessary treatments.1,2 Preferential publication of all null results, while positive findings are published “sparingly” and in anticipation of refutation, has been suggested to counter the inherent limitations of epidemiological research.3

Discussion of the role of positive results in the …