Objective - To see whether fibrinolytic inhibitors are of value when given to patients with upper gastrointestinal haemorrhage. Design - Meta-analysis of six randomised double blind placebo controlled trials. Two methods used for obtaining an overall estimate of effect, including a random effects model incorporating any heterogeneity of outcome in the estimate of the overall treatment effect. Setting - Inpatient care in hospitals in the United Kingdom, Sweden, and Australia. Patients - 1267 Patients admitted to hospital with primary diagnosis of acute upper gastrointestinal haemorrhage. Five of the six trials included a high proportion of elderly patients. Most patients were bleeding from peptic ulcers in the stomach and duodenum (43-88%) or gastric erosions (4-23%). A variable proportion had a degree of clinical shock at entry. Interventions - Tranexamic acid 3-6 g/day given intravenously for two or three days followed by 3-6 g/day by mouth for a further three to five days (four trials) or 4.5-12 g/day by mouth for two to seven days (two trials). End points - Frequency of recurrent haemorrhage, need for surgery, and death. Main results - Treatment with tranexamic acid was associated with a 20-30% reduction in the rate of rebleeding, a 30-40% reduction in the need for surgery, and a 40% reduction (95% confidence interval 10% to 60%) in mortality. Conclusions - Treatment with tranexamic acid may be of value to patients considered to be at risk of dying after an upper gastrointestinal haemorrhage.