STATE TRANSITION OR PARTITIONED SURVIVAL MODEL? MODEL SELECTION FRAMEWORKS FOR EARLY-STAGE ONCOLOGY
Author(s)
Sam Sutton, MSc1, Rebecca Beale, MA2, Kenny Suen, MSc1, Naomi van Hest, MSc, MBA2.
1Costello Medical, Manchester, United Kingdom, 2Costello Medical, Cambridge, United Kingdom.
1Costello Medical, Manchester, United Kingdom, 2Costello Medical, Cambridge, United Kingdom.
OBJECTIVES: Partitioned survival (PSM) and state transition models (STM) are two frequently adopted decision-analytic frameworks used to assess the cost-effectiveness of oncology therapies. Factors informing model choice have been well-studied in advanced oncology. However, given differences in disease stage, clinical outcomes assessed and data maturity, there is a need to understand key considerations to inform this decision in early oncology settings.
METHODS: Individual PSMs and STMs were developed to assess the cost-effectiveness of bortezomib, melphalan and prednisone (BMP) versus lenalidomide plus dexamethasone (Ld) in early multiple myeloma. The structure of the STM comprised progression-free (PF), second-line (2L), third-line (3L), progressed disease (PD) and death, and the PSM comprised PF, PD and death states. Each model was informed by the same efficacy (BMP versus Ld), utility and cost data, but parameterised in accordance with the model structures. External data informed efficacy for subsequent treatment lines in the STM.
RESULTS: The cost-effectiveness of BMP versus Ld was superior in the STM (incremental cost-effectiveness ratio [ICER]: £5,193) versus the PSM (£5,961). While the PSM estimated higher total costs, the STM predicted greater total LYs and QALYs across interventions. Additionally, the STM yielded higher incremental costs (£6,244 vs. £4,789) and QALYs (1.20 vs. 0.80) vs the PSM. The STM demonstrated lower sensitivity to both cost and utility fluctuations than the PSM; varying the discount on subsequent treatment costs by ±25% resulted in ±43% ICER fluctuations, versus ±24% in the STM. Utility adjustments yielded narrower ICER variations in the STM (-5% to +3%) versus the PSM (-12% to +6%).
CONCLUSIONS: Our findings suggest key considerations when selecting a STM or PSM in early oncology should include data maturity of the main trial and external data informing subsequent therapies given their substantial impact on results, plus the need for post-progression treatment pathway granularity, and sensitivity to downstream cost and utilities.
METHODS: Individual PSMs and STMs were developed to assess the cost-effectiveness of bortezomib, melphalan and prednisone (BMP) versus lenalidomide plus dexamethasone (Ld) in early multiple myeloma. The structure of the STM comprised progression-free (PF), second-line (2L), third-line (3L), progressed disease (PD) and death, and the PSM comprised PF, PD and death states. Each model was informed by the same efficacy (BMP versus Ld), utility and cost data, but parameterised in accordance with the model structures. External data informed efficacy for subsequent treatment lines in the STM.
RESULTS: The cost-effectiveness of BMP versus Ld was superior in the STM (incremental cost-effectiveness ratio [ICER]: £5,193) versus the PSM (£5,961). While the PSM estimated higher total costs, the STM predicted greater total LYs and QALYs across interventions. Additionally, the STM yielded higher incremental costs (£6,244 vs. £4,789) and QALYs (1.20 vs. 0.80) vs the PSM. The STM demonstrated lower sensitivity to both cost and utility fluctuations than the PSM; varying the discount on subsequent treatment costs by ±25% resulted in ±43% ICER fluctuations, versus ±24% in the STM. Utility adjustments yielded narrower ICER variations in the STM (-5% to +3%) versus the PSM (-12% to +6%).
CONCLUSIONS: Our findings suggest key considerations when selecting a STM or PSM in early oncology should include data maturity of the main trial and external data informing subsequent therapies given their substantial impact on results, plus the need for post-progression treatment pathway granularity, and sensitivity to downstream cost and utilities.
Conference/Value in Health Info
2026-11, ISPOR Europe 2026, Vienna, Austria
Value in Health, Volume 29, Issue 12S
Code
EE82
Topic
Economic Evaluation, Health Technology Assessment
Disease
Oncology