Continuous-Time Markov Modeling of Durvalumab Dynamics After Chemoradiotherapy in Stage III NSCLC (PACIFIC Era): Development, Validation, and Extrapolation. CTMC modeling of PACIFIC-era durvalumab

BackgroundImmunotherapy in unresectable stage III NSCLC shows delayed treatment effects and time-varying hazards that challenge proportional-hazards assumptions. We developed a continuous-time Markov chain (CTMC) model--with piecewise hazards--to capture the temporal dynamics of durvalumab after chemoradiotherapy (PACIFIC era), enabling transparent extrapolation beyond early trial landmarks. MethodsWe specified a multi-state CTMC with absorbing death, and clinically interpretable transient states (pre-progression disease control and post-progression). Piecewise transition rates were calibrated to early PACIFIC landmarks (e.g., 0-12-24-36 months) for overall survival (OS) and progression-free survival (PFS), using constrained optimization. External validity was assessed by comparing model-implied OS/PFS trajectories against published PACIFIC follow-ups and large real-world evidence. Uncertainty was quantified via parametric resampling of transition intensities and sensitivity to piecewise knots. ResultsThe CTMC reproduced the hallmark delayed separation of survival curves and the long-tail behavior typical of immune checkpoint inhibition. When calibrated only to early landmarks, the model generated plausible OS and PFS beyond 36 months, approaching 5-year targets reported in PACIFIC follow-ups (e.g., ~43% OS and ~33% PFS with durvalumab), while acknowledging mild underestimation of late PFS in sensitivity scenarios. Goodness-of-fit remained stable across alternative knot placements; uncertainty widened beyond 60 months, as expected from non-individual data calibration. ConclusionsA piecewise-hazard CTMC provides a transparent and reproducible framework to model non-proportional, time-dependent immunotherapy effects in stage III NSCLC, reconciling early trial information with clinically coherent long-term projections. This approach can support methodological rigor in survival analysis, scenario planning, and health-economic extrapolation in the PACIFIC setting and similar contexts.