Background: Aprotinin has been shown to be effective in reducing peri-operative blood loss and the need for re-operation due to continued bleeding in cardiac surgery. The lysine analogues tranexamic acid (TXA) and epsilon aminocaproic acid (EACA) are cheaper, but it is not known if they are as effective as aprotinin. Methods: Studies were identified by searching electronic databases and bibliographies of published articles. Data from head-to-head trials were pooled using a conventional (Cochrane) metaanalytic approach and a Bayesian approach which estimated the posterior probability of TXA and EACA being equivalent to aprotinin; we used as a non-inferiority boundary a 20% increase in the rates of transfusion or re-operation because of bleeding. Results: Peri-operative blood loss was significantly greater with TXA and EACA than with aprotinin: weighted mean differences were 106 mls (95% CI 37 to 227 mls) and 185 mls (95% CI 134 to 235 mls) respectively. The pooled relative risks (RR) of receiving an allogeneic RBC transfusion with TXA and EACA, compared with aprotinin, were 1.08 (95% CI 0.88 to 1.32) and 1.14 (95% CI 0.84 to 1.55) respectively. The equivalent Bayesian posterior mean relative risks were 1.15 (95% Bayesian Credible Interval [BCI] 0.90 to 1.68) and 1.21 (95% BCI 0.79 to 1.82) respectively. For transfusion, using a 20% non-inferiority boundary, the posterior probabilities of TXA and EACA being non-inferior to aprotinin were 0.82 and 0.76 respectively. For re-operation the Cochrane RR for TXA vs. aprotinin was 0.98 (95% CI 0.51 to 1.88), compared with a posterior mean Bayesian RR of 0.63 (95% BCI 0.16 to 1.46). The posterior probability of TXA being non-inferior to aprotinin was 0.92, but this was sensitive to the inclusion of one small trial. Conclusions: The available data are conflicting regarding the equivalence of lysine analogues and aprotinin in reducing peri-operative bleeding, transfusion and the need for re-operation. Decisions are sensitive to the choice of clinical outcome and non-inferiority boundary. The data are an uncertain basis for replacing aprotinin with the cheaper lysine analogues in clinical practice. Progress has been hampered by small trials and failure to study clinically relevant outcomes.