OBJECTIVES: COX-2 inhibitors (COXIBs) have been introduced in Italy in September 2000. We investigated and identified differences in history of claims for gastrointestinal disorders (GIDs) between COXIBs vs NSAIDs patients. METHODS: Prescriptions of antinflammatories made by 200 GPs to a population of 24,428 arthritis patients in a Northern Italian area were retrospectively investigated, focused on the first six months of COXIBs availability (1st October 2000-31st March 2001). COXIBs were prescribed to 6,204 patients and NSAIDs to 18,224 patients. We extracted the data of all 442 COXIB patients who received a co-prescription (i.e.simultaneously prescribed at least once in the observation period) of a gastroprotective agent (GPA). COXIB + GPAs patients were compared to a sample of 442 NSAIDs + GPAs patients, matched for age and gender. The two groups were compared in terms of history of claims for GIDs, including GPA prescription, diagnostic procedures, and hospitalizations, occurred in the course of the previous 2 years (1st October 1998-30th September 2000). Reimbursed prices, for drugs, and tariffs paid by NHS, for procedures and hospitalizations, were used to calculate costs. RESULTS: Prior to starting their COX-2 treatment, 84% of COXIB + GPAs patients vs 79% of NSAIDs + GPAs patients had a history of GID with significantly higher (p=.0026, U test) mean costs (€554.9 vs €362.6). All cost items were higher in the former group: hospitalizations (€185.8 vs 76.8), procedures (€52.5 vs 43.2), GPAs (€316.6 vs 242.6). CONCLUSIONS: Patients starting a treatment with COXIBs + GPAs are likely to have a previous history of GIDs significantly more severe and costly than patients who continue NSAIDs + GPAs. This constitutes a confounding factor when assessing therapy effectiveness and safety, in particular when evaluating co-prescription rates with GPAs in patients treated with antinflammatories.