OBJECTIVES : Rheumatoid arthritis (RA) is a systemic chronic inflammatory autoimmune disease and is associated with extensive burden on health systems worldwide. The study objective was to review the existing literature and assess the approaches taken within decision-analytic modelling for appraising new interventions.

METHODS : Electronic databases (MEDLINE, Embase, The Cochrane Library and EconLit) and technology appraisal organisations in Europe and Australia were systematically searched for cost-effectiveness analyses in moderate-to-severe RA published from 2000 to 2020. Methodological data were extracted, assessed for applicability, and qualitatively synthesised.

RESULTS : The review identified 96 unique model analyses from 104 records, predominantly assessing biologics. The most common approaches were Markov cohort (n = 39) and patient-level simulations (n = 35). A long-term/lifetime horizon was generally used to capture the chronic, progressive nature of the disease. Disease progression was commonly characterised by the non-linear health assessment questionnaire disability index (HAQ-DI). Treatment response was frequently assessed through ACR (American College of Rheumatology) or EULAR (European League Against Rheumatism) response criteria. Treatment discontinuation occurred primarily due to loss of efficacy, and less frequently due to adverse events. Utilities were typically assigned to a patient’s HAQ-DI or ACR response-defined health state, using direct trial estimates or mapping algorithms. Direct costs generally consisted of drug acquisition, administration, and hospitalisation. Treatment-related adverse events were often excluded, with serious infection composite most commonly incorporated. Health technology assessment critiques noted Markov models were inappropriate for capturing non-linear disease progression and population heterogeneity. Discrete-event simulation was considered the best approach for incorporating long-term disease and treatment-related elements and patient-level variation.

CONCLUSIONS : This review suggests that long-term discrete event simulation modelling is the most appropriate method to accurately capture disease progression and treatment effects for evaluating the cost-effectiveness of RA treatment. Analyses should cover a patient’s lifetime, include adverse events, and capture disease progression using non-linear HAQ-DI.