To obtain the best evidence for nonsteroidal anti-inflammatory drug (NSAID) use in gastrointestinal (GI) bleeding, a detailed patient history was supplemented with objective tests of aspirin use, i.e., high-performance liquid chromatography of plasma and platelet cyclo-oxygenase inhibition, which detect aspirin intake within 24 and 120 hours, respectively. Seventy-one patients consecutively admitted for upper or lower GI bleeding and 138 age- and sex-matched controls were studied. Five bleeders were excluded for confounding factors, e.g., warfarin. Of the other 66 bleeders, 45 had upper GI bleeding (28 from peptic ulcer, 14 from duodenal ulcer, and 14 from gastric ulcer) and 21 lower GI bleeding. Evidence of current NSAID use (of which 89% was aspirin) was found in 80% of bleeders vs. 24.3% of controls (P less than 0.0001), for an odds ratio of 13.7 (95% confidence interval, 6.39-27.27). The cyclo-oxygenase test uncovered 21.5% more aspirin users than history alone. Severity of bleeding was not different in acetylsalicylic acid users. The surprisingly high association of current intake of NSAIDs, especially aspirin, with nonulcer GI bleeding including colonic bleeding, changes the conventional view of the following hierarchy of the risk: NSAID----peptic ulcer----bleeding to: NSAIDs----GI bleeding. This view has important implications for current ulcer cotherapy prophylactic strategies, which could fail to prevent greater than 50% of GI bleeding episodes.