Model Conceptualization for IDH Mutant Diffuse Glioma: A Case Study Exploring Two Alternative Structures

OBJECTIVES: Health economic models of interventions for early-stage cancers (e.g., IDH-mutant diffuse glioma), are associated with several interconnected challenges that influence their conceptualisation. These include limited long-term data, lack of evidence on surrogacy relationships, and challenges in translating trial outcomes into lifetime outcomes. Some challenges may be addressed through considered model design incorporating multiple data sources. Using a case study in IDH-mutant diffuse glioma, we explored two candidate modelling approaches.

METHODS: Cohort-level state-transition (STM) and patient-level simulation (PLS) models were deemed suitable following initial conceptualisation. Both models included nine health states, spanning three treatment lines and best supportive care, informed using the same input data. Critically, the PLS allowed for time-varying transitions between latter health states. Median and mean survival (expressed as total estimated life-years [LYs]) for a hypothetical treatment strategy were compared, and practical considerations for both models were noted.

RESULTS: Median survival estimates were similar across both models (14.03 versus 13.95 years for the PLS versus STM, respectively). Mean LYs were lower for the PLS (16.03 years) versus the STM (17.38 years). The STM was faster in terms of both build and run time (particularly for sensitivity analysis), but imposed simplifying assumptions for long-term survival (i.e., constant probabilities), and therefore may have omitted important aspects of IDH-mutant diffuse glioma (such as accurately capturing disease progression).

CONCLUSIONS: While the STM may be preferred in terms of its relative simplicity, this would need to be weighed against the risk of ‘incorrect’ decision making based on inaccurate modelling of survival outcomes (which would need to be determined based on inspection of the input survival data). Echoing sentiments raised in published guidance, consideration of multiple model structures highlights fundamental structural uncertainties when addressing a decision problem. Considering alternatives better encapsulates prospective cost-effectiveness of new interventions.