The comparative safety of nabumetone (1,000-2,000 mg/day) versus diclofenac (100-200 mg/day), naproxen (500-1,500 mg/day), piroxicam (10-20 mg/day), and ibuprofen (1,200-3,200 mg/day) was evaluated in a 12-week, randomized, open-label, multicenter study. Patients with osteoarthritis (OA) or rheumatoid arthritis (RA) were enrolled in a 3:1 ratio (nabumetone:one of the four comparator NSAIDs). The incidence of > or = 1 adverse event considered by the investigator to be related or probably related to therapy was similar in all groups. However, significantly (p < 0.02) more diclofenac-treated patients experienced abdominal pain and/or gastritis than nabumetone-treated patients. Naproxen-treated patients experienced significantly (p < 0.002) more dyspepsia as compared with patients treated with nabumetone or ibuprofen and significantly (p < or = 0.001) more nabumetone-treated patients experienced diarrhea than patients treated with naproxen, ibuprofen, or piroxicam. Ulcers occurred in one (0.03%) nabumetone-treated patient versus six (0.5%) patients treated with one of the comparator NSAIDs (p = 0.001). A decrease in hemoglobin > or = 1.5 g/dL occurred in fewer nabumetone-treated patients than in patients treated with diclofenac (p < 0.04), ibuprofen (p < or = 0.04), or piroxicam (p = 0.055). Finally, a similar percentage of patients in all treatment groups withdrew from the study because of adverse events related or probably related to treatment. More (p < 0.001) diclofenac-treated patients withdrew because of elevated hepatic transaminases than patients treated with the other agents. Withdrawal because of gastritis was also noted for more diclofenac-treated patients than nabumetone-treated patients (p < 0.04). In conclusion, nabumetone was demonstrated to be at least as safe as diclofenac, piroxicam, ibuprofen, and naproxen as related to subjective complaints, such as dyspepsia or gastritis. However, more serious events, such as ulcers or meaningful decreases in hemoglobin, seem to occur less often with nabumetone.