OBJECTIVES: To compare the analytic judgements, data and assumptions of different models used in the economic evaluation of infliximab, one of a new class of drugs for rheumatoid arthritis (RA). The purpose was to understand why different models give such varying results. METHODS: A detailed assessment was made of three models, one submitted (in a reimbursement dossier) by the manufacturer, one produced by an independent academic group and one published in the literature. Factors considered included the key data inputs, assumptions about the sequencing of treatments for RA, the estimation of cost offsets and the modelling of the maintenance of treatment effect for patients continuing or discontinuing infliximab. RESULTS: Two of the models, although embodying different methodological approaches, gave fairly similar results (approximately £30,000-£40,000 cost per additional quality-adjusted life-year gained). The third model, by the independent academic group, gave much higher estimates, around £100,000 per QALY. The differences were mainly because of the assumptions about the positioning of infliximab in the treatment sequence and assumptions about the long-term effect of therapy. CONCLUSIONS: Economic models of treatments for rheumatoid arthritis incorporate several key analytic judgements, which can have major impacts on cost-effectiveness. Two of the three models examined gave similar results, which suggests that consensus can be reached on several of the main methodological issues.