Rheumatoid arthritis (RA) is a chronic, progressive, inflammatory disease that affects approximately 0.5-1% of the adult population. The introduction of new disease-modifying antirheumatic drugs (DMARDs) such as leflunomide, anakinra and the tumour necrosis factor (TNF)-alpha antagonists (infliximab, etanercept and adalimumab) have transformed the management of RA. In particular, the last class of agents has generated substantial controversy. Costing between 16,000 US dollars and 20,000 US dollars per patient-year (2001 values), the potential greater efficacy of treatment with TNFalpha antagonists comes at much higher drug costs, making these agents natural candidates for cost-effectiveness analyses (CEAs).A MEDLINE search (until 31 January 2004) identified six original CEAs evaluating TNFalpha antagonists in RA. The aim of a CEA is to facilitate the allocation of scarce health resources and to inform policy decisions. However, to enhance the reliability and relevance of these analyses to policy makers, there must be similarity between the methodologies used. Recently, the OMERACT (Outcome Measures in Rheumatoid Arthritis Clinical Trials) group produced a document to define such a reference case; the OMERACT document was used as a foundation to structure comparisons and highlight discrepancies. The methodologies employed in each analysis differed; in particular, disparate time horizons, comparators, quantities of drug and treatment sequences prohibit the comparison of cost effectiveness between studies. Outcomes also differed between the analyses. Most reported health-related quality of life (HR-QOL) in quality-adjusted life-years (QALYs). The QALYs metric was based on preference scores that were typically derived from linear regressions using the Health Assessment Questionnaire (HAQ). However, models also used American College of Rheumatology (ACR) criteria, as well as the disease activity score (DAS). Common to all studies was the lack of data from long-term randomised studies where efficacy and resource consumption in comparison with standard care has been investigated. As such, investigators combined short-term randomised control trial data with that of a long-term observational cohort, and modelled cost effectiveness over an appropriate time horizon. In addition, most analyses lacked rigorous sensitivity analysis to examine the impact of uncertainty in the parameters. Those analyses that examined time horizons of 6 months and 1 year published incremental cost-effectiveness ratios (ICERs) of 34,800 US dollars per ACR 70% response criteria (ACR70) weighted response (duration 6 months, 1999 values) and 96,166 US dollars (duration 1 year, 2002 values). Analyses that modelled costs and health outcomes beyond the first year reported ICER estimates ranging between 26,800 US dollars (patients' lifetime, 1998 values) and 40,308 US dollars (10 years, 2002 values). In terms of HR-QOL, the analyses reported incremental QALYs that ranged from 0.116 (over 19 years) to 1.6 (over 10 years). Discounted costs of therapy ranged from 30,362 US dollars (10 years, 2002 values) to 93,000 US dollars (22 years, 1998 values), and comparator costs ranged from 22,593 US dollars (10 years, 2002 values) to 84,000 US dollars (22 years, 1998 values).