OBJECTIVES: In the phase 3 clinical trial, CheckMate 577, patients with resected EC/GEJC treated with nivolumab had statistically significant and clinically meaningful improvement, versus placebo, in the primary endpoint, disease-free survival (DFS). Overall survival (OS) data is currently not available. The objective of this study was to evaluate approaches for estimating post-recurrence survival (PRS) to populate a cost-effectiveness model (CEM) for CM577 and assess sensitivity.

METHODS: To determine the ICUR over 30-years in Canada, the CEM adopted a three-health state (3-HS) Markov model (pre-recurrence, post-recurrence, death) allowing flexibility to include external data sources to estimate PRS. Two PRS data sources were explored: (i) adjuvant patients with EC/GEJC from the Netherlands Comprehensive Cancer Organization (IKNL) registry matched to the CM 577 population and (ii) first-line (1L) patients with advanced EC/GEJC OS pooled from clinical trials CM 649 and KEYNOTE- 590. The IKNL registry data were also analyzed by type of recurrence (locoregional and distant) to explore model sensitivity to a 3-HS approach (pooling type of recurrence) versus a 4-HS approach with patients separated by type of recurrence. The CEM assumed the same PRS for each treatment arm, based on clinical opinion.

RESULTS: Using adjuvant IKNL registry PRS data in the 3-HS model, the ICUR of nivolumab versus surveillance was $42,733/QALY gained. Using IKNL data in the 4-HS model, the ICUR was $42,920/QALY gained. Using 1L advanced EC/GEJC trial data in the 3-HS model, the ICER was $45,961/QALY gained.

CONCLUSIONS: The CEM for nivolumab in the adjuvant setting for patients with resected EC/GEJC in Canada is not sensitive to the data source for PRS or analysis by type of recurrence (3-HS or 4-HS) in the absence of OS data. Introducing additional complexity to the analysis by type of recurrence did not produce varying results in the adjuvant EC/GEJC setting, and therefore 3 HS model is preferred.