Background: The Cochrane Handbook for Systematic Reviews of Interventions notes that excluded studies are those appearing to meet the inclusion criteria (which others might believe to be relevant), that upon closer inspection were excluded. It also notes that the table of excluded studies should provide a brief summary of why studies were excluded (e.g. inappropriate control group), and that a single reason for exclusion is usually sufficient.

Objectives: To study the distributions in numbers of included and excluded studies and reasons for exclusions in Cochrane reviews.

Methods: Cross-sectional survey of the 2619 Cochrane reviews published in The Cochrane Library 2006 Issue 1.

Results: The reviews had a median of seven included studies and eight excluded studies. The median difference between number of included and excluded studies was 0 (80% central range -30 to 11). Seventy-three reviews, representing 26 of the 50 different review groups, had more than 100 excluded studies, 15 had more than 200, and the maximum was 607. Many of the excluded articles were irrelevant as they would not have met the inclusion criteria even without closer inspection (e.g. reviews, animal studies, wrong intervention, wrong disease, cohort studies, uncontrolled studies and retrospective studies). At the Nordic Cochrane Centre, we have tried to use a more restrictive approach, listing only those studies which others might believe to be relevant, in the 12 reviews (from six review groups) we have published. These reviews have a median of 13 included studies and five excluded studies, and the median difference between number of included and excluded studies is 8.5 (80% central range -9 to 47). The difference between number of included and excluded studies in our reviews is considerably higher than in other reviews (p<0.001).

Conclusions: Cochrane reviews list far too many excluded studies.