Failures in research ethics review examinedThe recent tragic death of Ellen Roche1 provides valuable lessons for research ethics review. The reasons for the wrongful administration of hexamethonium stem from researchers failing to act in certain ways, not from deliberate malicious or negligent actions.FIRST FAILING AND FIRST LESSON: PUBLICATION BIASThe first major failing was the failure of researchers who conducted the 1978 San Francisco study of hexamethonium to report similar adverse reactions.The tendency of researchers to fail to publish disappointing results2 or (...) results which are unattractive to them or their sponsors' interests has been well documented.3 It is called publication bias. An important part of publication bias is the failure to report all relevant or potentially relevant results. As this case illustrates, researchers must report or make available all research results, even negative results, and all potentially relevant outcomes of that research. If researchers believe an adverse event is unrelated to the study, they should explain why they believe this rather than omitting the event from their report. Elimination of publication bias is essential to scientific progress, to the protection of participants in research and in maintaining public faith in the integrity of scientists and doctors. Ethics committees and journals should make it clear that all adverse events—related or unrelated—to a project must be acknowledged, evaluated and further details made available to other researchers or the public if requested.SECOND FAILING: FAILURE TO CONDUCT A SYSTEMATIC REVIEWBut there is a more important failing. The most disturbing feature of this case was that literature documenting the pulmonary toxicity of hexamethonium was available before the trial began.In the study protocol submitted to the ethics committee, the researcher cited four reports involving 20 patients given inhaled hexamethonium without adverse event. However, Johns Hopkins investigators found a number of reports between 1953 and 1962, …. (shrink)