Beyond traditional endpoints: A multi-state modelling approach to integrating morbidity and mortality in cancer...

Traditional benefit-harm assessments of cancer treatments evaluate morbidity and mortality separately, overlooking how patients experience these outcomes together. This study illustrates how a health profile-based approach using multi-state modelling can complement such assessments.

To compare the traditional and health profile-based approaches, we simulated a phase 3 trial in newly diagnosed glioblastoma. The traditional analysis estimated median progression-free survival (PFS), overall survival (OS), and grade 3 and 4 (G34) toxicity rates. The health profile-based approach used a multi-state Markov model (MSMM) to estimate time spent in different health states over 36 months. Quality-adjusted life years (QALYs) were calculated by combining time-in-state estimates with utility weights simulated to reflect a patient preference study.

The traditional analysis showed a 3.8-month increase in median PFS for the experimental treatment, accompanied by a substantially higher rate of G34 toxicity (48.4 % vs. 15.2 %). This allowed balancing benefits and harms on the basis of population-level effects of median PFS and cumulative incidence of toxicity. The MSMM approach offered a more granular view of the progression-toxicity trade-off, showing that patients in the experimental group spent less time with progressed disease (7.7 vs. 9.4 months), but more time progression-free with G34 toxicity (3.8. vs 0.3 months). When combined with simulated individual-level utility weights, QALY values could be assigned to the experimental treatment and the control.

Multi-state modelling offers a flexible way to integrate morbidity and mortality in evaluating cancer treatments. By capturing health trajectories and enabling utility-based assessments, it may complement traditional methods to inform treatment decisions.