Evidence for the effectiveness of peer review

Abstract

Scientific editors’ policies, including peer review, are based mainly on tradition and belief. Do they actually achieve their desired effects, the selection of the best manuscripts and improvement of those published? Editorial decisions have important consequences—to investigators, the scientific community, and all who might benefit from correct information or be harmed by misleading research results. These decisions should be judged not just by intentions of reviewers and editors but also by the actual consequences of their actions. A small but growing number of studies has put editorial policies to a strong scientific test. In a randomized, controlled trial. blinding reviewers to author and institution was usually successful and improved the quality of reviews. Two studies have shown that, contrary to conventional wisdom, reviewers early in their careers give better reviews than senior reviewers. Many studies have shown low agreement between reviewers but there is disagreement about whether this represents a failing of peer review or the expected and valuable effect of choosing reviewers with complementary expertise. In a study of whether manuscripts are improved by peer review and editing, articles published in Annals of Internal Medicine were improved in 33 of 34 dimensions of reporting quality, but published articles still had room for improvement. Because of the central place of peer review in the scientific community and the resources it requires, more studies are needed to define what it does and does not accomplish.