Treatment with non-steroidal anti-inflammatory drugs (NSAIDs) is associated with an increased risk of peptic ulcer complications, but it is not clear whether some drugs are more likely than others to cause such complications. We compared previous use of NSAIDs in 1144 patients aged 60 and older admitted to hospitals in five large cities with peptic ulcer bleeding and in 1126 hospital controls and 989 community controls matched for age and sex. Peptic ulcer bleeding was strongly associated with use of non-aspirin NSAIDs of any type during the 3 months before admission (411 cases, 351 controls; odds ratio 4.5 [95% CI 3.6 to 5.6]). The odds ratios for peptic ulcer bleeding were lowest for ibuprofen (2.0 [1.4-2.8]) and diclofenac (4.2 [2.6-6.8]), and intermediate for indomethacin, naproxen, and piroxicam (11.3 [6.3-20.3], 9.1 [5.5-15.1], and 13.7 [7.1-26.3]). Azapropazone and ketoprofen carried the highest risks (31.5 [10.3-96.9] and 23.7 [7.6-74.2]). Risks also increased with drug dose (low dose 2.5 [1.7-3.8], intermediate 4.5 [3.3-6.0], and high 8.6 [5.8-12.6]) for all drugs combined. Appropriate clinical strategies could prevent many episodes of peptic ulcer bleeding: NSAIDs should be used only in patients who do not respond to other analgesics; the lowest possible doses should be used; and the least toxic NSAIDs should be selected.