Using Structured Expert Elicitation in Cost-Effectiveness Analysis: An Exploratory Case Evaluating Selpercatinib for Thyroid Cancer in Norway

OBJECTIVES: To explore the feasibility of using structured expert elicitation (SEE) to supplement single-arm evidence in a cost-effectiveness analysis comparing selpercatinib to cabozantinib or vandetanib for RET mutant medullary thyroid cancer (MTC) patients.

METHODS: We developed two partitioned survival models to compare the clinical benefit and cost-effectiveness (CE) of selpercatinib with cabozantinib or vandetanib. In the model based on phase II data, we extrapolated progression-free survival (PFS) from the single-arm 'LIBRETTO-001' for the intervention arm. To inform overall survival (OS) in both treatment arms and PFS in the standard-of-care arm, SEE was used. In the model utilizing phase III data, PFS and OS were based on the randomized trial 'LIBRETTO-531'. We probabilistically evaluated CE across both models over lifetime from a Norwegian healthcare perspective, with costs and effects discounted at 4% per year. Incremental cost-effectiveness ratios (ICERs) were calculated per life-year (LY) saved and per quality-adjusted life year (QALY) gained.

RESULTS: Both models indicated an improvement in PFS and OS in patients treated with selpercatinib compared to standard-of-care. Furthermore, the model based on phase II data provided 1.562 LYs saved and 1.913 QALYs gained for an additional cost of €380,520, yielding ICERs of €243,652 per LY saved and €198,960 per QALY gained. Analysis of phase III data gave higher LYs saved (2.582) and QALYs gained (2.486) for incremental cost of €451,431, yielding ICERs of €174,829 per LYs saved and €181,592 per QALY gained. At the current cost-effectiveness threshold (€72,960 per QALY gained), the models consistently demonstrated that selpercatinib was not cost-effective compared to cabozantinib and vandetanib.

CONCLUSIONS: Despite similar reimbursement recommendations across the phase II and phase III models, SEE may not precisely predict future outcomes as illustrated by the differences on disaggregated level outcomes. Further research should investigate the use of SEE for nearly cost-effective innovative treatments that address small patient populations.