Peer Review in Epidemiology Cannot Accomplish Its Ostensible Goals Due to Incomplete Reporting and Unverifiable Analyses

摘要

The accuracy and appropriateness of an analysis often cannot be verified by the contemporary peer review process. Peer review is also unlikely to identify possible publication bias in situ (PBIS) (running many different statistical models but only reporting the results of one, often an outlier result). A review of articles reporting analyses of a cohort of Swedish construction workers revealed unacknowledged and unexplained variations in methodology, including the use of different variables measuring tobacco use, age and body mass index and a failure to adequately and accurately reference previous related articles about the cohort. Seemingly minor changes in methodology, such as the cutoffs used to convert continuous variables to categorical variables may result in significant changes to the results. These inconsistencies were likely not discovered during the peer review process as it would have required that reviewers conduct a systematic review of previous analyses of the dataset. Practical solutions to this dilemma include enhanced post publication review and ensuring that data are available for secondary analysis.