OBJECTIVES: To evaluate upper GI (UGI) events, use of GI medications, and healthcare costs among arthritis patients in a managed care setting. METHODS: Commercial claims data for three million health maintenance (HMO) and preferred provider organization (PPO) members in Southeast U.S. were used to identify new users of COX-2s and NSAIDs in 2002. Patients had =1 arthritis-related claim followed by an index claim for COX-2 (rofecoxib, valdecoxib or celecoxib) or NSAID (ibuprofen, naproxen, diclofenac, and nabumetone) and were continuously enrolled for = one year pre- and post-index date. Patients dispensed either a COX-2 or NSAID during one year pre-index and patients with claims for both COX-2s and NSAIDs were excluded. Multiple logistic regression was used to model UGI events (ulcers and bleeds) and GI medication use (proton pump inhibitors and H2-antagonists), and a log transform model was used for total healthcare costs (medical and prescription) at 1 year controlling for age, gender, health status, medication persistence, and baseline utilization. RESULTS: In total, 3449 arthritis patients were included: 47% COX-2 (26% rofecoxib, 15% celecoxib, 7% valdecoxib) and 53% NSAID. Patients in the COX-2 group were significantly older, taking more medications and more persistent, more likely to be female or belong to a PPO, and had more comorbidities, GI events, and higher costs at baseline compared to the NSAID group. After adjusting for these factors, no significant differences were observed in the rate of GI events (1% overall), rate of GI medication use (5%) total healthcare costs (mean=$1712), or medical costs (mean=$1513) after 1 year. Prescription drug costs were 38% and 51% higher for rofecoxib and celecoxib patients respectively compared to the NSAID group (p<.0001). CONCLUSION: In contrast to initial marketing information, in this observational study, we found no significant difference in GI-related outcomes or total healthcare costs between the two groups.