Biostatistics: how to detect, correct and prevent errors in the medical literature.

Abstract

Approximately half the articles published in medical journals that use statistical methods use them incorrectly. These errors are so widespread that the present system of peer review has not been able to control them. This article presents a few rules of thumb to help readers identify questionable statistical analysis and estimate what the authors would have concluded had they used appropriate statistical methods. To prevent such errors from appearing, journals should secure review by someone knowledgeable in statistics before accepting a manuscript. In addition, human research committees should insist that an investigator define an appropriate strategy for data analysis before approving a protocol. Such policies should quickly and effectively increase the reliability of the clinical and scientific literature.