OBJECTIVE The gastric safety of cyclooxgenase-2 inhibitors and prophylactic antisecretory therapy in high-risk arthritis patients is unclear. We studied the ulcer incidence and factors predicting ulcer recurrence in a prospective, double-blinded trial. METHODS We studied patients who presented with nonsteroidal anti-inflammatory drug-associated ulcer bleeding. After ulcer healing, patients who were negative for Helicobacter pylori were randomly assigned to celecoxib 200 mg twice a day plus omeprazole placebo once daily or diclofenac 75 mg twice daily plus omeprazole 20 mg once daily for 6 months. Patients underwent endoscopy if they developed recurrent bleeding. Those without recurrent events underwent endoscopy at their last follow-up visit. RESULTS Two hundred eighty-seven patients were enrolled; 24 had recurrent gastrointestinal complications. Among 259 patients without events, 222 underwent endoscopy (116 received celecoxib and 106 received diclofenac plus omeprazole). The probability of recurrent ulcers in 6 months was 18.7% in the celecoxib group and 25.6% in the diclofenac plus omeprazole group (difference, -6.7%; 95% CI: -17.8% to 3.9%) (P = 0.21). Combining bleeding and endoscopic ulcers, 24.1% in the celecoxib group and 32.3% in the diclofenac plus omeprazole group had recurrent ulcers in 6 months (difference, -8.2%; 95% CI: -19.5% to 2.9%) (P = 0.15). Treatment-induced significant dyspepsia (hazard ratio, 5.3; 95% CI: 2.6-10.8), age > or =75 (hazard ratio, 2.0; 95% CI: 1.1-3.5), and comorbidity (hazard ratio, 2.1; 95% CI: 1.2-3.7) independently predicted ulcer recurrence. CONCLUSIONS Among patients with previous ulcer bleeding, neither celecoxib nor diclofenac plus omeprazole adequately prevents ulcer recurrence. Treatment-induced significant dyspepsia is an indication for endoscopic evaluation.