In a hospital-based case-control study, the risk of a first episode of major upper gastrointestinal tract bleeding in subjects now known to be predisposed was assessed in relation to the use of nonnarcotic analgesics. For aspirin use within the week before the onset of symptoms, the rate ratio estimates, adjusted for potential confounding, were 15 (95% confidence interval, 6.4 to 34) for regular use (at least four days a week) and 5.6 (confidence interval, 2.7 to 12) for occasional use. For aspirin use discontinued at least one week earlier, the estimate was 1.6 (confidence interval, 0.6 to 4.2). There was no evidence that acetaminophen use increased the risk. For the regular use of other analgesics in the week before onset, the adjusted rate ratio estimate was 9.1 (confidence interval, 2.7 to 31); there were insufficient data to evaluate occasional use. The findings suggest that the risk of bleeding is increased substantially by aspirin, even when used occasionally. With the exception of acetaminophen, other nonnarcotic analgesics may also increase the risk, but they remain to be evaluated individually.