Beyond PFS and OS: Recovering Pre- and Post-Progression Survival From Survival Curves
Author(s)
Iwona Zerda1, Tomasz Stanisz, Msc, PhD2, Emilie Clay, Msc, PhD3, Samuel Aballea, MSc, PhD4, Mondher Toumi, Sr., MSc, PhD, MD5.
1Poland, 2Clever-Access, Krakow, Poland, 3Clever-Access, Paris, France, 4Innovintel, Rotterdam, Netherlands, 5University Aix-Marseille, Marseille, France.
1Poland, 2Clever-Access, Krakow, Poland, 3Clever-Access, Paris, France, 4Innovintel, Rotterdam, Netherlands, 5University Aix-Marseille, Marseille, France.
OBJECTIVES: Progression-free survival (PFS) and overall survival (OS) are standard endpoints for assessing cancer treatment efficacy and are widely used in both clinical trials and health economic models. However, survival before and after progression—pre-progression survival (PrePS) and post-progression survival (PPS)—are rarely reported, despite their critical role in modelling disease progression and informing patients flows into subsequent treatment lines. As innovative drugs available at various disease stages may drive the assessment outcomes, precise information regarding treatment pathway is important. This study introduces a method to estimate PrePS and PPS from available PFS and OS data, enhancing model accuracy when direct estimates are unavailable.
METHODS: We developed a differential equations-based model to estimate PrePS and PPS by simulating patient transitions between health states over time. A two-step numerical optimization process was used to derive pre- and post-progression mortality rates that align model outputs with observed PFS and OS curves. To ensure identifiability, proportional hazards between PrePS and PPS were assumed. The model was validated using synthetic datasets under varied assumptions, including cases that violated proportional hazards, and further tested with real-world oncology data.
RESULTS: The method consistently generated accurate estimates across a range of PFS and OS curve shapes. In 72% of tested scenarios, the estimated curves differed from the originals by less than 10%, and in all cases, the difference remained under 20%. These results demonstrate the method’s robustness, flexibility, and applicability across diverse scenarios.
CONCLUSIONS: This approach provides a practical and reliable solution for estimating PrePS and PPS when not directly reported, enabling more accurate modelling of patient trajectories in semi-Markov models. While developed for oncology, the method is adaptable to other disease areas where similar survival modelling frameworks apply.
METHODS: We developed a differential equations-based model to estimate PrePS and PPS by simulating patient transitions between health states over time. A two-step numerical optimization process was used to derive pre- and post-progression mortality rates that align model outputs with observed PFS and OS curves. To ensure identifiability, proportional hazards between PrePS and PPS were assumed. The model was validated using synthetic datasets under varied assumptions, including cases that violated proportional hazards, and further tested with real-world oncology data.
RESULTS: The method consistently generated accurate estimates across a range of PFS and OS curve shapes. In 72% of tested scenarios, the estimated curves differed from the originals by less than 10%, and in all cases, the difference remained under 20%. These results demonstrate the method’s robustness, flexibility, and applicability across diverse scenarios.
CONCLUSIONS: This approach provides a practical and reliable solution for estimating PrePS and PPS when not directly reported, enabling more accurate modelling of patient trajectories in semi-Markov models. While developed for oncology, the method is adaptable to other disease areas where similar survival modelling frameworks apply.
Conference/Value in Health Info
2025-11, ISPOR Europe 2025, Glasgow, Scotland
Value in Health, Volume 28, Issue S2
Code
P16
Topic
Economic Evaluation, Study Approaches
Topic Subcategory
Trial-Based Economic Evaluation
Disease
No Additional Disease & Conditions/Specialized Treatment Areas, Oncology