Presentation (CTI)

OBJECTIVES: Partitioned survival models (PSM) are well accepted in advanced oncology indications. New therapies are increasingly assessed in earlier indications, including neoadjuvant and perioperative settings. These earlier indications present modelling challenges and additional data requirements to capture long-term outcomes and costs. This research aims to evaluate perioperative and neoadjuvant models submitted to the National Institute for Health and Care Excellence (NICE) to assess trends and acceptability of methods.
METHODS: We reviewed all NICE submissions in the neoadjuvant and perioperative space (up to April 2025). Data were extracted for model structure, number of health states (HS), methods for estimating transitions, sources informing transitions, implementation of cure assumptions, limitations from the External Assessment Group (EAG), and reimbursement recommendations.
RESULTS: Five early oncology submissions were identified, two neoadjuvant and three perioperative. All submissions evaluated targeted or immunotherapies, two for breast cancer and three for non-small cell lung cancer. All submissions used a cohort state transition model (STM). Three models implemented four HS: event-free (EF), local recurrence (LR), distant metastasis (DM), and death. Two models further separated DM into pre- and post-progression. Three models applied a cure assumption to EF and one captured cure as a separate HS; the cure timepoint varied from 5-7 years. For transitions out of EF, the relevant pivotal trials were used most often, incorporating indirect treatment comparisons for external comparators if needed. Transitions out of LR/DM were informed by external clinical trials in later stages and/or real-world databases. Complexity in DM varied, including treating as an absorbing HS, using constant transitions, or using a nested PSM. EAG noted limitations with assuming constant transitions, uncertainty with extrapolations, cure timepoints, and immunotherapy retreatment.
CONCLUSIONS: There was a consensus in STM structure, with variability in number of HS and approaches to modelling transitions. Limitations and critiques from reviewers did not contribute to any negative recommendations.