Fate of biomedical research protocols and publication bias in France: retrospective cohort study.

OBJECTIVES To describe the fate of protocols approved by the French research ethics committees, a national system created by the French 1988 Huriet-Sérusclat Act; to assess publication bias at a national level. DESIGN Retrospective cohort study. SETTING Representative sample of 25/48 French research ethics committees in 1994. PROTOCOLS: 649 research protocols approved by committees, with follow-up information. MAIN OUTCOME MEASURES Protocols' initial characteristics (design, study size, investigator) abstracted from committees' archives; follow-up information (rates of initiation, completion, and publication) obtained from mailed questionnaire to principal investigators. RESULTS Completed questionnaires were available for 649/976 (69%) protocols. Of these, 581 (90%) studies were initiated, 501/581 (86%) were completed, and 190/501 (38%) were published. Studies with confirmatory results were more likely to be published as scientific papers than were studies with inconclusive results (adjusted odds ratio 4.59, 95% confidence interval 2.21 to 9.54). Moreover, studies with confirmatory results were published more quickly than studies with inconclusive results (hazard ratio 2.48, 1.36 to 4.55). CONCLUSION At a national level, too many research studies are not completed, and among those completed too many are not published. We suggest capitalising on research ethics committees to register and follow all authorised research on human participants on a systematic and prospective basis.