Critics in the media have become wary of exaggerated research claims from clinical trials and epidemiological studies. Closer to home, reviews of published studies find a high frequency of poor quality in research methods, including those used for statistical analysis. The statistical literature has long recognized that questionable research findings can occur when investigators fail to set aside their own outcome preferences as they analyse and interpret data. These preferences can be related to financial interests, a concern for patients, peer recognition, and commitment to a hypothesis. Several analyses of published papers provide evidence of an association between financial conflicts of interest and reported results. If we are to regain professional and lay confidence in research findings some changes are required. Clinical journals need to develop more competence in the review of analytic methods and provide space for thorough discussion of published papers whose results are challenged. Graduate schools need to prepare students for the conflicting interests that surround the practice of statistics. Above all, each of us must recognize our responsibility to use analytic procedures that illuminate the research issues rather than those serving special interests.
Copyright 2003 John Wiley & Sons, Ltd.