OBJECTIVE: Given the sensitivity of findings to the gastroprotective agent (GPA) rate assumption used in COX-2 cost-effectiveness models, the purpose of this study is to verify the GPA rate assumptions and to re-estimate model outcomes from one published COX-2 cost-effectiveness study using GPA rates from actual practice. METHODS: Prescription and medical claims data from a large preferred provider organization (PPO) located in the Midwest were used to estimate GPA rates within three samples of adult patients new to non-selective nonsteroidal antiinflammatory drugs (NSAIDs) and COX-2 therapy: all new NSAID users, new NSAID users with a diagnosis of arthritis, and a cohort matched on GI risk. Members were continuously eligible over the study period of January 1, 1999 through May 31, 2001. RESULTS: Of the more than 319,000 members with at least 1 day of eligibility, the number of members meeting inclusion criteria in each of the three samples were 1,900 for new NSAID users, 289 with a diagnosis of arthritis, and 1,386 in the matched cohort sample. GPA estimates for non-selective NSAID and COX-2 users were consistent across all 3 samples with COX-2 GPA rates of 22%, 21% and 20% and nonselective NSAID GPA rates of 15%, 15%, and 18%, for new NSAID users, those with a diagnosis of arthritis, and the matched cohort, respectively. Re-estimation of the cost-effectiveness model using the most conservative GPA rates increased the cost per year of life saved for COX-2s from $18,614 to over $100,000. CONCLUSIONS: Contrary to COX-2 cost-effectiveness model assumptions, the rate of GPA use is positive and marginally higher among COX-2 users than among nonselective NSAID users. These findings call into question the validity of assumptions regarding patterns of use when made prior to a product's use in the real world. Given these findings, a re-evaluation of the cost-effectiveness of COX-2 therapies should be considered.