Background: Selective reporting of outcomes of trials in systematic reviews could result in biased estimates of efficacy and safety. Access to trial protocols would facilitate detection of risk of bias due to selective reporting. Objectives: To evaluate the extent to which Cochrane systematic reviews seek and use trial protocols in risk of bias assessments of selective outcome reporting. Methods: All new intervention reviews in Issue 3, 2010 of the Cochrane Database of Systematic Reviews were assessed independently by three investigators for strategies to locate trial protocols. Characteristics of included studies, risk of bias and ongoing studies tables were evaluated for reference to trial protocols in assessing risk of bias. A subset of included trials published from 2005 was sought from the WHO search portal, clinicaltrials.gov, the Australian and New Zealand Clinical Trials Registry and the metaRegister of Controlled Trials (mRCT). Results: Twenty eight intervention reviews included 367 included studies, of which 333 were randomized controlled trials (RCTs); six additional reviews had no included studies. Only 14 (41.2%) reviews specifically searched trials registries, of which only three searched the WHO search portal; eight searched clincicaltrials.gov. Thirty nine ongoing trials were identified; 15 were from trials registries, the rest from conference proceedings, published protocols or drug-company registers. Eight reviews did not assess included trials for risk of selective reporting bias; sixteen assessed this from the study report, and five reviews used the protocols of 12 trials to assess selective reporting. Of 117 RCTs published from 2005, only two were identified in trials registries (both in clinicaltrials.gov and one in mRCT and the WHO search portal). Conclusions: The use of trial protocols to detect selective reporting bias is limited by the lack of uniform methods in systematic reviews to locate them and low rates of prospective registration of clinical trials.