OBJECTIVE: Selective COX-2 inhibitors (coxibs) provide comparable efficacy with less gastrointestinal (GI) adverse events compared to the conventional non-selective non-steroid anti-inflammatory drugs (NSAIDs) in patients with arthritis. We conducted an economic analysis, focusing specifically on differences in GI-related event rates between the coxibs and conventional NSAIDs. METHODS: We developed a decision model, using Microsoft Excel(r) and Decisioneering Crystal Ball(r) , which focused on three areas of potential economic differentiation between treatment with COX-2 inhibitors and conventional non-selective NSAIDs; GI-related complications, uncomplicated GI ulcers, and GI-related adverse effects. The model was populated with published data describing resource implications and mortality risks, unit costs, underlying NSAID GI-event risks and relative GI-event risks for coxibs. We considered two treatment options (i) celecoxib and (ii) a single NSAID drug based on naproxen, ibuprofen, or diclofenac (as observed in the CLASS study). Sensitivity analyses considered variation in the underlying GI-event risks alongside general uncertainty in resource usage and drug cost data. RESULTS: Under baseline GI complication annual risk assumptions (1.5% for NSAIDs), cost savings for celecoxib ($10,000 per 100 patients) through avoided GI events were dominated by the additional drug costs ($66,000 per 100 patients). This relationship held true even when higher costs NSAIDs, based solely on either diclofenac or naproxen, and higher underlying rates of 6% were considered. Cost effectiveness ratios were calculated at $41,824 per life year gained under baseline conditions. Sensitivity analysis showed, however, that underlying annual risk of GI-related complication had a strong influence on the cost-effectiveness of the COX-2 inhibitors. At 3% per year risk levels, the cost per LYG reduced to $17,107. CONCLUSION: The analysis suggests that the coxibs have an attractive cost-effectiveness profile when patients have an underlying annual risk of GI-related complications on NSAIDs of at least 2.5% (equivalent to a patient having at least two recognised risk factors).