We conducted a case-control study in five general hospitals in the region of Antwerp, studying 161 patients (102 men, 59 women) and hospital control subjects matched for age and sex to explore the relation between drug use and upper gastrointestinal bleeding from 'erosive lesions' (peptic oesophagitis, gastric erosions, gastric ulcer(s), or duodenal ulcer(s]. There was a highly significant difference between cases and control subjects in the use of non-steroidal anti-inflammatory drugs (NSAIDs, excluding aspirin) (odds ratio 7.4, p less than 0.001; 95% confidence interval odds ratio 3.7 to 14.7). There also was a significant difference in the use of aspirin (odds ratio 2.2, p = 0.025; 95% CI odds ratio 1.3 to 4.0) and a highly significant difference regarding the presence of antecedents of peptic ulcer disease (odds ratio 5.5, p less than 0.001; 95% CI odds ratio 3.2 to 9.6). There was no significant difference in the use of other drugs, paracetamol and corticosteroids in particular, nor in the use of alcohol or tobacco. The patient group using NSAIDs was older, had more women, and had a higher mortality than the group not using NSAIDs. Among patients with bleeding gastric or duodenal ulcer(s), NSAID users were not more or less likely to have had symptoms of peptic ulcer disease, and had no higher frequency of multiple gastric or duodenal ulcers. The attributable risk for NSAID use was 0.30 (95% CI 0.23 to 0.37) and for aspirin use 0.14 (95% CI 0.08 to 0.20).