Accounting for Cure Rates Estimated from Progression-Free Survival (PFS) in Long-Term Overall Survival (OS) Projections: A Case Study from Previously Treated Advanced Cancers

Author(s)

Shubhram Pandey, MSc1, Supreet Kaur, MSc1, Barinder Singh, RPh2, Murat Kurt, BS, MS, PhD3.
1Pharmacoevidence Pvt. Ltd., SAS Nagar, Mohali, India, 2Pharmacoevidence Pvt. Ltd., London, United Kingdom, 3Iovance Biotherapaeutics, Inc., Levittown, PA, USA.
OBJECTIVES: In health technology assessments, using external data can improve reliability of long-term survival projections from the trials. Due to advancements in treatment of metastatic cancers with modern agents, PFS data often reveal earlier signs for plausibility of cure by reaching statistical maturity sooner than OS data. The impact of incorporating cure rates estimated from PFS data into OS modeling was assessed through a case study in relapsed/refractory advanced cancers treated with Pembrolizumab, focussing on long-term OS projections.
METHODS: PFS and OS data were reconstructed from the clinical publications of registrational trials in head & neck squamous cell carcinoma, cervical cancer, esophageal cancer, hepatocellular carcinoma. For each tumor, reconstructed OS data were extrapolated using standard parametric models accounting for background mortality rates with no possibility of cure and using mixture cure models (MCMs) with fixed cure rates derived from PFS data. MCMs were also applied to PFS data to derive cure rates for each tumor. For each trial, background mortality rates were derived using publicly available local lifetables from the countries with majority of enrolment. Model selection guided by Akaike and Bayesian Information Criteria (AIC & BIC), visual fits to reported survival curves and underlying hazard trends. Estimated mean OS over a lifetime horizon was compared between the approaches.
RESULTS: Modeling OS data with fixed cure rates estimated from PFS data improved AIC and BIC scores in all instances. While estimated cure fractions from the best-fitting models to the PFS data were modest and ranged between 3.3%-11.9% across all tumors accounting for these fractions in OS modeling led to 1.64 (range: 0.53-3.16) years of higher mean OS across all tumors.
CONCLUSIONS: Accounting for cure rates obtained from PFS data in OS projections while OS data are maturing with no visible plateaus aligns long-term projections with expected mortality patterns and reduce numerical and structural uncertainty.

Conference/Value in Health Info

2025-11, ISPOR Europe 2025, Glasgow, Scotland

Value in Health, Volume 28, Issue S2

Code

MSR11

Topic

Economic Evaluation, Health Technology Assessment, Methodological & Statistical Research

Disease

Oncology

Your browser is out-of-date

ISPOR recommends that you update your browser for more security, speed and the best experience on ispor.org. Update my browser now

×