BACKGROUND: Tranexamic acid reduces haemorrhage through its antifibrinolytic effects. In a previous version of the present review, we found that tranexamic acid may reduce mortality. The present review includes updated searches of randomised trials on tranexamic acid versus placebo, cimetidine or lansoprazole.

OBJECTIVES: To assess the effects of tranexamic acid for upper gastrointestinal bleeding.

SEARCH METHODS: Electronic searches (The Cochrane Library, MEDLINE, EMBASE, Science Citation Index) and manual searches were combined. The last search update was in October 2011.

SELECTION CRITERIA: Trials in which patients with upper gastrointestinal bleeding were randomised to receive either tranexamic acid or placebo, or any anti-ulcer drug, were included.

DATA COLLECTION AND ANALYSIS: Two authors independently extracted data. All-cause mortality was the primary outcome measure. Random-effects model meta-analyses were performed and results presented as relative risks (RR) with 95% confidence intervals (CI). Subgroup, sensitivity, regression and sequential analyses were performed to analyse sources of intertrial heterogeneity and the robustness of the overall result.

MAIN RESULTS: Seven double blind randomised trials on tranexamic acid versus placebo, cimetidine, or lanzoprazole were included. One trial offered endoscopic treatment to all patients that were randomised. Random-effects model meta-analysis found that tranexmic acid reduced mortality compared with placebo (41 of 829 versus 68 of 825 patients; RR: 0.61, 95% CI 0.42 to 0.89). The beneficial effect was not confirmed in subgroup analysis stratified for the quality of bias control, in worst case scenario analyses (in which 21% of the randomised patients were excluded), or in sequential analyses. No significant differences were found between tranexamic acid and placebo on bleeding, surgery, or transfusion requirements. No clear effects of tranexamic acid were identified in trials using endoscopic therapy or in the trials comparing tranexamic acid with cimetidine or lansoprazole. In the tranexamic acid group, five cases of serious thromboembolic events occurred (myocardial infarction, pulmonary embolism, and cerebral infarction). Overall, the number of patients with any thrombotic event was not significantly increased in the tranexamic acid group (RR 1.87, 95% CI 0.60 to 5.85).

AUTHORS' CONCLUSIONS: Considering the internal and external validity of the evidence, tranexamic acid cannot be recommended for routine use. Additional trials in which tranexamic acid is used in combination with the currently recommended interventions are required.