Background: The purpose of pre-specifying primary outcomes in the systematic review process is to define the most clinically relevant outcomes and to protect against bias. Adding, omitting or changing review outcomes once the protocol is published can result in bias. Objectives: To investigate the discrepancies between primary outcomes listed in protocols and the subsequent reviews published in The Cochrane Library. To identify non-publication of review protocols and ill-defined primary outcomes. Methods: We examined new systematic reviews published between Issue 4, 2006 and Issue 2, 2007 of The Cochrane Library. For each review, discrepancies between the primary outcome(s) listed in the review protocol and the review itself were identified by a statistician and lead review authors were contacted to provide reasons for the discrepancies. Results: Atotal of 297 reviews were in the study cohort. For the primary outcome measures, there were 49 (16%) reviews that disagreed with the primary outcome(s) specified in the protocol; nine including at least one new primary outcome not specified in the protocol, five excluding at least one primary outcome specified in the protocol and 35 where a protocol outcome was either upgraded or downgraded to a primary/secondary review outcome respectively. A further 24 (8%) reviews had no protocol registered on the Cochrane Library while 14 (5%) reviews could not be assessed for outcome reporting bias due to poor primary outcome definition. Conclusions: Discrepancies between primary outcomes specified in the review and protocol are common. The most common reasons for these discrepancies were: 1) recommendation by editors/peer reviewers; and 2) recognition of the importance of the outcome before/after reading the results to the included trials. The seriousness of bias arising from the reasons provided by the review authors for such discrepancies, non-registration of review protocols and ill-defined outcome definition will be discussed.