Objective. To evaluate the cost effectiveness of the cyclooxygenase 2 (COX-2) selective nonsteroidal antiinflammatory drug (NSAID) rofecoxib compared with naproxen and the COX-2 NSAID celecoxib compared with ibuprofen and diclofenac. Methods. Cost-effectiveness analysis based on a 5-year Markov model. Probability estimates were derived from detailed data of 2 randomized trials and a systematic search of the medical literature. Utility estimates were obtained from 60 randomly selected members of the general public. Cost estimates were obtained from Canadian provincial databases. Incremental cost-effectiveness ratios were calculated for patients at average risk of upper gastrointestinal (UGI) events and for high-risk patients with a prior history of a UGI event. Subjects were patients with osteoarthritis or rheumatoid arthritis (RA) where a decision has been made to treat with NSAIDs but who do not require low-dose aspirin. Main outcome measures were proportion of patients with clinical or complicated UGI events, quality-adjusted life expectancy, and life expectancy. Results. Evaluation of rofecoxib versus naproxen in patients with RA at average risk resulted in costs per qualityadjusted life year (QALY) gained of $Can271,188. Celecoxib was dominated by diclofenac in average-risk patients. Both rofecoxib and celecoxib are cost-effective in high-risk patients. Analyses by age groups and assuming a threshold of Can$50,000 per QALY gained, suggest that rofecoxib or celecoxib would be cost-effective in patients aged over 76 and 81, respectively, without additional risk factors. Conclusion. Both rofecoxib and celecoxib are economically attractive in high risk and elderly patients. They are not economically attractive in patients at average risk. Coprescription of proton-pump inhibitors with COX-2 NSAIDs is not economically attractive for patients at high risk.