To determine whether discontinuation patterns differed among nonsteroidal antiinflammatory drugs (NSAID) prescribed to treat osteoarthritis (OA). In a retrospective cohort study of Health Maintenance Organization enrollees, 1405 patients with OA aged 45 and older who received a new prescription for one of 4 NSAID were followed for 12 months. Survival analysis was used to evaluate time to discontinuation, used here as a relative measure of both drug efficacy and tolerability. Rates of NSAID discontinuation during the study period were high; only 15 to 20% of those started on a study NSAID were still using the same drug at the end of the 12 month followup period. Using a proportional hazards model to adjust for covariates, the risk of discontinuation did not differ when comparing the agent with the longest duration of use, piroxicam (the referent), to enteric coated aspirin [relative risk (RR) 1.10, 95% confidence interval (CI) 0.93 to 1.30]. Adjusted rates of discontinuation were significantly higher for ibuprofen (RR 1.43, 95% CI 1.22 to 1.69) and for naproxen (RR 1.40, 95% CI 1.19 to 1.65) when compared to piroxicam. NSAID discontinuation rates are high among patients with OA and risk of discontinuation differed between NSAID, even after controlling for the effects of such variables as age, disease severity, and concomitant therapy.