OBJECTIVES: To construct a flexible and transparent pharmacoeconomic model to assess the value of COX-2 selective inhibitors (COX-2s) in the treatment of arthritis. METHODS: Literature searches and interviews with healthcare practitioners were conducted to identify critical issues for model construction. Model assumptions and parameters estimates for the new COSMO model were made based on these efforts. Convergent validity was assessed by comparing COSMO model outputs to results obtained using previously published models, when similar inputs were used in each model. RESULTS: Because arthritis patients often switch medications, the COSMO model simulates treatment strategies, rather than individual drugs. COSMO allows users to compare strategies under which patients start on an NSAID or COX-2 and switch medications twice. It was structured as a one-month cycle Markov model with the following disease states: GI discomfort, loss of efficacy, complicated ulcer, uncomplicated ulcer, CV event, no events, and death. Inputs allowed for different levels of complexity to address the difficulty in obtaining data sources in different countries. Multiple clinical (number of deaths, GI discomfort, ulcers, uncontrolled arthritis) and economic (drug costs, costs of managing GI and CV events, total costs) outputs were included. Monte Carlo simulations, acceptability curves, cost-effectiveness planes and extensive univariate and multi-variate sensitivity analysis can be performed with any input. For base case analyses, we assume a 3rd party payer perspective. In the validation exercise, we applied inputs from a published model (Maetzel et al., 2003), and obtained similar results (7% variation) and similar trends in sensitivity analyses, suggesting strong convergent validity. CONCLUSION: COSMO is a pharmacoeconomic model that assesses the value of different treatment strategies for osteoarthritis and rheumatoid arthritis. From a third party perspective, this model allows customization with respect to clinical and cost inputs, as well as treatment patterns.