Background Due to the scarcity of resources (new) treatments need not only to proof that they are effective, but also cost effective. Especially in the case of new more costly treatment options as for rheumatoid arthritis, establishing the cost-effectiveness is relevant. Sometimes data about the costs and effects (in the long term) are absent, especially in the case of new treatment options. A method to overcome the problem of incomplete data and to extrapolate short-term endpoints to long-term outcomes is modelling.

Objectives To develop and validate a Markov model for the economic evaluation of (new) treatments in RA.

Methods A Markov model was developed with a cycle time of 3 months and a time-horizon of 5 years. In a Markov model patients are classified in a number of states (Markov states), defined by the severity of the disease. Development of the disease is defined as transitions between these states. Markov states were defined using the Disease Activity Score (DAS), and were valued in utility and medical and total costs using a dataset from a 48-week trial with methotrexate. The model was calculated for ?usual care? for patients in the first 5 years of their disease, using transition probabilities calculated from the 5-years follow-up data of patients participating in an open prospective study of early (disease duration < 1 year) RA. To test the validity of the model, the structure of the model, the inputs of the model, the results of the model, and the value of the model to the decision-maker were assessed.

Results The costs and the utility showed a clear relation with the Markov states. The expected medical and total costs per patient over 5 years were DFL. 14.896, and DFL 27.889 respectively and the expected Quality Adjusted Life Years (QALY?s) were 3,27. Differences in patient characteristics like age, gender, disease duration, used DMARD?s, number of underwent orthopaedic procedures, and education had a moderate effect on the valuation of the Markov states in costs and utility. The input data for the Markov model was appropriate for the Dutch situation. The expected total costs seemed somewhat lower than the costs found in another Dutch study in the first six years of the disease, but the mean utility was comparable to utilities found in other studies. It appeared possible to validly extrapolate the course of the disease over 5 years with the Markov model, using transition probabilities calculated from only the first year of follow-up of the patients in the open study. In this study the value of the model for the decision-maker could not be well established.

Conclusion The developed Markov model is valid for use in economic evaluations in RA.