Presentation (CTI)

OBJECTIVES: To examine heterogeneity in the cost-effectiveness of sequential targeted therapies for moderate-to-severe rheumatoid arthritis (RA) in China.
METHODS: A lifetime probabilistic discrete event simulation model compared four-line targeted treatment sequences in 10,000 hypothetical Chinese patients. TNF-α inhibitors, tofacitinib, and tocilizumab, each combined with methotrexate, were used as first-line therapies, with subsequent treatments determined by treatment history and response. The model accommodated patient heterogeneity regarding age, sex, serology, weight, disease activity, pain, and joint damage. Treatment effects were derived from network meta-analyses of American College of Rheumatology ordered categorical responses and mapped to Disease Activity Score-28 to inform disease progression. Treatment discontinuation was determined by stop reason, with dose adjustment modelled for patients achieving remission. Mortality hazard was linked to time-varying disease activity. Utilities and costs were estimated from patient characteristic-based algorithms in Chinese RA patients. Optimal sequences were identified by patient characteristics. The expected value of individual care (EVIC) and perfect information (EVPI) were calculated.
RESULTS: At the willingness to pay threshold of 200,000 CYN, the expected net monetary benefit of first-line TNF-α inhibitors, tofacitinib, and tocilizumab were 1,264,840, 1,256,598, and 1,230,587, respectively, at the population level. The optimal sequence was initiated with TNF-α inhibitors for male and seronegative patients, tofacitinib for older patients (aged >60 years) and those with lower baseline disease activity, and tocilizumab for female seropositive patients aged <60 years. The total EVIC and EVPI was 3,602 and 26,143 CYN per patient, respectively. Variation in results was observed when treatment sequences were defined at the individual drug rather than drug-class level.
CONCLUSIONS: The model estimated subgroup-specific cost-effectiveness of sequential treatments across patient groups in RA. By distinguishing uncertainty attributable to patient heterogeneity from stochastic uncertainty, the model supports personalised treatment decision-making. This approach offers a transferable methodology for health economic evaluations requiring flexible treatment sequencing and heterogeneity-informed inference.