Summary Background Aspirin products are known to cause irritation and injury to the gastric mucosa. The belief that enteric-coated and buffered varieties are less likely to occasion major upper-gastrointestinal bleeding (UGIB) than plain aspirin was tested in data from a multicentre case-control study. Methods 550 incident cases of UGIB admitted to hospital with melaena or haematemesis and confirmed by endoscopy, and 1202 controls identified from population census lists, were interviewed about use of aspirin and other non-steroidal anti-inflammatory drugs (NSAIDs) during the 7 days before the onset of bleeding (cases) or interview (controls). Relative risks of UGIB for each type of aspirin used regularly (at least every other day) were calculated overall, and according to dose, by multiple logistic regression, with control for age, sex, marital status, date, education, cigarette smoking, alcohol use, and use of NSAIDs. Findings The relative risks of UGIB for plain, enteric-coated, and buffered aspirin at average daily doses of 325 mg or less were 2·6, 2·7, and 3·1, respectively. At doses greater than 325 mg, the relative risk was 5·8 for plain and 7·0 for buffered aspirin; there were insufficient data to evaluate enteric-coated aspirin at this dose level. There were no important differences in risk attributable to the three aspirin forms according to bleeding site (gastric vs duodenal), or when users of NSAIDs were excluded. Interpretation Use of low doses of enteric-coated or buffered aspirin carries a three-fold increase in the risk of major UGIB. The assumption that these formulations are less harmful than plain aspirin may be mistaken.