Anticoagulants are effective for the prevention of venous thromboembolism (VTE) but cause bleeding. Interpretation of the risks and benefits of new anticoagulant regimens for VTE prevention is complicated by a lack of standardized definitions and reporting of bleeding. We reviewed the reporting of bleeding in randomized controlled trials of new anticoagulants compared with standard doses of enoxaparin in hip and knee arthroplasty, and examined the possible impact of differences in the definition of major bleeding on interpretation of the trial results. Electronic searches identified 16 phase III trials published between 2001 and 2010 involving 41,265 patients comparing one of five new anticoagulants with a common comparator, enoxaparin. Major bleeding rates in patients treated with enoxaparin ranged from 0.1% to 3.1% in hip arthroplasty trials and from 0.2% to 1.4% in knee arthroplasty trials. In studies that excluded surgical-site bleeding from the definition, major bleeding rates were about 10-fold lower than in those which included surgical-site bleeding. Within the individual trials, the choice of bleeding definition and the methods of assessment of bleeding influenced the conclusions regarding the risk of bleeding with new anticoagulant regimens relative to enoxaparin. Eight of the 16 studies demonstrated a ≥ 40% relative risk differences in major bleeding between treatment groups but the difference was statistically significant in only two of these trials. Randomized VTE prevention trials report markedly different rates of major bleeding despite similar patient populations and doses and durations of anticoagulant prophylaxis and were underpowered to detect modest differences in patient-important bleeding events. Standardization of bleeding definitions and reporting seems desirable. © 2010 International Society on Thrombosis and Haemostasis.