From 1979 to 1985 2435 patients having had transient ischaemic attacks (TIAs) or minor ischaemic strokes, were enrolled in the UK TIA trial and were randomised to receive either aspirin 300 mg, daily or aspirin 1200 mg or placebo. Analysis of reported upper gastrointestinal bleeding events (defined as haematemesis or melaena, or both) showed a risk of bleeding in a dose dependent manner, odds ratios (95% CI) for 300 mg of aspirin = 3.3 (1.2 to 9.0) and for 1200 mg = 6.4 (2.5 to 16.5) and, as would be expected, an increased risk of hospitalisation because of bleeding also in a dose dependent manner, odds ratio = 3.6 (0.7 to 17.2) for 300 mg and 8.7 (2.0 to 37.6) for 1200 mg. Further analysis suggested greater risks of bleeding from duodenal ulcers than gastric ulcers and that bleeding is more likely early in the course of treatment with aspirin used as secondary prevention. There was also an increased risk of lower gastrointestinal bleeding, defined as fresh blood per rectum for both doses of aspirin, odds ratio 1.8 (0.5 to 6.1) for 300 mg of aspirin, and 1.5 (0.4 to 5.3) for 1200 mg of aspirin.