Presentation (CTI)

OBJECTIVES: Indirect treatment comparisons (ITCs) are often used to evaluate treatments not directly compared in head-to-head clinical trials. Traditional ITC methods may need to rely on the proportional hazards (PH) assumption for time-to-event endpoints. In such cases, a single hazard ratio (HR) may not capture the true treatment effect, necessitating more flexible approaches. Multi-level network meta-regression (ML-NMR) extends ITC methodology by incorporating both aggregate and patient-level data in a network, while adjusting for treatment-effect modifiers. This study explores the application of non-proportional hazards (NPH) ML-NMR models to enhance survival extrapolations in oncology, which is often a key component required for health technology assessments (HTA).
METHODS: ML-NMR models were developed for progression-free survival (PFS). Several parametric survival distributions were tested. To accommodate NPH, a regression model was applied to the shape of the baseline hazard, allowing treatment-specific variations while maintaining the ability to predict absolute treatment effects in any target population. Model convergence was assessed using effective sample size (ESS) and Rhat diagnostics. Predicted survival probabilities and hazards were extrapolated over a lifetime horizon for economic modeling.
RESULTS: Results demonstrated that NPH ML-NMR provides enhanced flexibility in estimating long-term survival, allowing for survival extrapolations in any target population while adjusting for differences in treatment-effect modifiers. These results may be advantageous when the PH assumption is not met, and a HR cannot be reliably applied from traditional ITC methods in an economic model.
CONCLUSIONS: ML-NMR demonstrates promising advancements in ITC methodology. This case study highlights the advantage of ML-NMR to obtain more flexible survival extrapolations in oncology by directly using predictions from NPH ML-NMR models when the PH assumption is not met. These methodological advancements have potential to strengthen the evidence base and decision-making for HTA evaluations, particularly in contexts where traditional ITC methods are not suitable.