OBJECTIVES To quantify and compare the overall adverse event (AE) related costs and their constituents associated with celecoxib and traditional NSAID treatment in osteoarthritis (OA). METHODS Using an amended version of a recent NICE Markov model (five-year horizon, three-month cycles) evaluating NSAID treatment in OA, the overall AE related costs and their constituents associated with celecoxib and ibuprofen treatment were quantified and compared. The individual costs (11 in total) were aggregated to four categories relevant to OA treatment: (i) costs directly attributable to clinically significant gastrointestinal (GI) AEs; (ii) costs directly attributable to dyspepsia; (iii) costs associated with myocardial infarction (MI); and (iv) co-medication (mainly proton pump inhibitors) costs. The model, which has a healthcare perspective, was populated with UK data and a discount rate of 3.5% per year was used. Except for the MI risks which were adjusted for age, all absolute and relative risks were taken from a meta-analysis in a recent NICE HTA report. Costs were aggregated to 1000 patients on treatment. RESULTS Although 9 of the 11 adverse event related individual costs were lower with celecoxib than ibuprofen treatment, the overall cost difference (≤42,640 in favor of celecoxib) was attenuated by the relatively high MI related costs associated with celecoxib treatment (≤92,734). CONCLUSIONS Whilst the GI safety profile of celecoxib yields cost advantages, the MI costs associated with celecoxib treatment weakens the effect on overall adverse event related costs resulting from the beneficial GI safety profile.