OBJECTIVES: To estimate the cost-utility of adjuvant UC treatment with nivolumab versus observation after radical resection from a Canadian public-payer perspective.
METHODS: A three health-state Markov model consisting of disease-free (DF), recurred disease (RD), and death states was developed to evaluate discounted total costs and quality-adjusted life-years (QALYs) over a 30-year time horizon. The annual discount rate on costs and outcomes was 1.5%. The model included costs of drug acquisition, administration, monitoring, subsequent treatment, adverse events, routine disease management, and end-of-life care. Transitions from the DF state were estimated using DF survival data and the distribution of first-recurrence events from CheckMate 274 and published data from the control arm of the EORTC 30994 study in adjuvant treatment of UC. Model assumed functional cure for DF patients starting from year 5. Published literature informed subsequent treatment-specific survival estimations from the RD state. Treatment-specific survival estimations were weighted with corresponding usage rates in Canada. Health-state utilities were derived from CheckMate 274 EQ-5D-3L data. Costs [2021 Canadian dollars (CAD)] were obtained from publicly available local sources and published literature. Incremental cost-utility ratio (ICUR) was the primary outcome of the analysis and assessed probabilistically.
RESULTS: Nivolumab was associated with an additional 1.37 life-years compared to observation which translated to 1.18 incremental QALYs. An incremental cost of CAD 75,361 over observation produced an ICUR of CAD 64,046/QALY. Nivolumab had almost 100% probability of cost-effectiveness for all willingness-to-pay thresholds ≥ CAD 76,000/QALY. Key drivers of ICURs included annual discount rate and utility values in DF state. Alternative approaches to estimating transition probabilities had only marginal impact on the results, with ICURs ranging from CAD 61,936-97,683.
CONCLUSION: Nivolumab is estimated to be a life-extending treatment option compared with observation for UC in Canada, which results in a compelling ICUR that is robust to uncertainties in the data.
