OBJECTIVES: Advanced gastric (GC), gastro-oesophageal junction (GOJC) and oesophageal cancer (OC) are characterised by poor survival. However, a small proportion of patients survive beyond 2 years, i.e., long-term remission. Partitioned-survival models (PSM) are the standard approach for economic evaluation of oncology therapies. However, they are unable to explicitly model populations with mixed outcomes. A semi-Markov model (SMM) can facilitate progression-specific outcomes modelling, including the impact of time since- and time of-progression on the rate of mortality. This method captures the expected benefits of immunotherapy, e.g., delayed progression incurring improved survival and benefits to HRQoL. This study assessed both modelling approaches and the impact on health economic outcomes.

METHODS: Two economic models were developed: a PSM reflecting pre-progression, post-progression and death using independent progression-free survival (PFS) and overall survival (OS) functions; and SMM reflecting pre-progression, post-progression, long-term remission, and death. Time on treatment Kaplan-Meier were used to inform duration of treatment. The inputs were informed by the CPS ≥ 5% subgroup of CheckMate 649. A payer perspective was adopted, with resource use derived for the NICE (England) setting. Costs and benefits were discounted at 3.5%.

RESULTS: The incremental cost-effectiveness ratios (ICERs) versus XELOX were £34,110 per quality-adjusted life year (QALY) and £45,383/QALY for the SMM and PSM approaches, respectively. However, the incremental first-line treatment costs and survival outcomes were consistent between models (1.8 and 1.7 life years, respectively), with ICER variance predominately due to differences in incremental post-progression health state costs.

CONCLUSIONS: Although providing similar incremental survival, the PSM could not explain the marginal mortality hazard or HRQoL benefits, whereas the SMM was able to model populations with mixed outcomes by reflecting progression-specific mortality rates and a long-term remission state. This demonstrated the long-term benefits of immunotherapy with improved ICER prediction and estimation of HRQoL, compared to the PSM.