Background: Because of recent concerns about the safety of aprotinin, we updated our 2007 Cochrane review that compared the relative benefits and risks of aprotinin and the lysine analogues tranexamic acid and epsilon aminocaproic acid. Methods: We searched electronic databases, including CENTRAL, MEDLINE, EMBASE, Google and Google Scholar for trials of antifibrinolytic drugs used in adults scheduled for cardiac surgery. Searches were updated to January 2008. By comparing aprotinin and the 2 lysine analogues to control, we derived indirect head-to-head comparisons of aprotinin to the other drugs. We derived direct estimates of risks and benefits by pooling estimates from head-to-head trials of aprotinin and tranexamic acid or epsilon aminocaproic acid. Results: For indirect estimates, we identified 49 trials involving 182 deaths among 7439 participants. The summary relative risk (RR) for death with aprotinin versus placebo was 0.93 (95% confidence interval [CI] 0.69-1.25). In the 19 trials that included tranexamic acid, there were 24 deaths among 1802 participants. The summary RR was 0.55 (95% CI 0.24-1.25). From the risk estimates derived for individual drugs, we calculated an indirect summary RR of death with use of aprotinin versus tranexamic acid of 1.69 (95% CI 0.70-4.10). To calculate direct estimates of death for aprotinin versus tranexamic acid, we identified 13 trials with 107 deaths among 3537 participants. The summary RR was 1.43 (95% CI 0.98-2.08). Among the 1840 participants, the calculated estimates of death for aprotinin compared directly to epsilon aminocaproic acid was 1.49 (95% CI 0.98-2.28). We found no evidence of an increased risk of myocardial infarction with use of aprotinin compared with the lysine analogues in either direct or indirect analyses. Compared with placebo or no treatment, all 3 drugs were effective in reducing the need for red blood cell transfusion. The RR of transfusion with use of aprotinin was 0.66 (95% CI 0.61-0.72). The RR of transfusion was 0.70 (95% CI 0.61-0.80) for tranexamic acid, and it was 0.75 (95% CI 0.58-0.96) for use of epsilon aminocaproic acid. Aprotinin was also effective in reducing the need for reoperation because of bleeding (RR 0.48, 95% CI 0.34-0.67). Interpretation: The risk of death tended to be consistently higher with use of aprotinin than with use of lysine analogues. Aprotinin had no clear advantages to offset these harms. Either tranexamic acid or epsilon aminocaproic acid should be recommended to prevent bleeding after cardiac surgery.