Faut-il tenter de publier une étude qui ne conclut rien ?

Abstract

When a study finds that no exposure factor or therapy is significantly related to a given effect, researchers legitimately wonder if the results should be submitted for publication and to what journal. Clinical trials that report significant associations have a higher probability of publication, a phenomenon known as selective publication. The principal reasons of this selective publication include author self-censorship, peer-reviewing, trials not intended for publication, interpretation of the p value, cost of journal subscriptions, and policies. Subsequent reviews and meta-analyses are biased by the unavailability of nonsignificant results. Suggestions for preventing this risk include university training, trial registries, an international standard randomised controlled trial number (ISRCTN), Cochrane collaboration, and the gray literature. Journals (including electronic journals) interested in studies with nonsignificant results are listed. New technologies are changing the relations between publishers, libraries, authors and readers.