COMPARING NON-CANCER-SPECIFIC SURVIVAL ESTIMATED FROM LIFETABLES AND REPORTED TRIAL DATA: A CASE STUDY FROM EARLY STAGE RESECTED MELANOMA
Author(s)
Shubhram Pandey, MSc1, Sameer Mansoori, MSc1, Rashi Rani, MSc2, Barinder Singh, RPh1, Murat Kurt, BS, MS, PhD3.
1Pharmacoevidence Pvt. Ltd., SAS Nagar, Mohali, India, 2Heorlytics Pvt. Ltd., Mohali, India, 3Iovance Biotherapaeutics, Inc., Philadelphia, PA, USA.
1Pharmacoevidence Pvt. Ltd., SAS Nagar, Mohali, India, 2Heorlytics Pvt. Ltd., Mohali, India, 3Iovance Biotherapaeutics, Inc., Philadelphia, PA, USA.
OBJECTIVES: Survival extrapolations from early-stage cancer trials should account for cure. However, applying only general population mortality rates to cured patients may overlook residual risks from disease and treatment history, and bias from clinical trial patient selection. This case study compared NCSS derived from local lifetables and aggregate-level data for early-stage, resected melanoma population of CheckMate 238 trial to assess implications for long-term survival projections.
METHODS: For each arm, overall survival (OS) and melanoma-specific survival (MSS) data were reconstructed from published Kaplan-Meier curves from the trial. Minimum follow-up was 9 years. A non-parametric NCSS was derived as the ratio of OS to MSS and smoothed using isotonic regression to enforce temporal monotonicity. Justified by randomization, smoothed NCSS estimates were pooled across arms and extrapolated using standard parametric survival models (SPMs). Three best-fitting SPMs were blended using weights based on statistical goodness-of-fit criteria. NCSS estimates from blended SPMs (trial-based approach), a previously published model based on local lifetables matched to trial’s age, gender and enrollment distribution (lifetable-based approach [Weber et al, 2024]) and a conservative hybrid approach combining both were compared over observed and extrapolated periods.
RESULTS: Over observed follow-up, trial and lifetable-based NCSS showed no significant visual or statistical differences (p = 0.9426), with 0.08 year mean differential favoring trial-based approach. Over a lifetime horizon, hybrid approach yielded 0.7 additional years of mean NCSS than lifetable-based approach. Lifetable-based hazards underestimated trial-based hazards from 1.7 years onward, with divergence increasing over time. Beyond observed follow-up, NCSS rates from hybrid approach were on average 1.8% higher than those from lifetable-based approach.
CONCLUSIONS: Despite long-term trial follow-up, findings support a hybrid approach for estimating long-term NCSS and suggest that, contrary to prevailing belief, lifetables may not necessarily overestimate survival in cured patients. They also highlight the importance of incorporating trial-reported NCSS into survival analyses and economic evaluations.
METHODS: For each arm, overall survival (OS) and melanoma-specific survival (MSS) data were reconstructed from published Kaplan-Meier curves from the trial. Minimum follow-up was 9 years. A non-parametric NCSS was derived as the ratio of OS to MSS and smoothed using isotonic regression to enforce temporal monotonicity. Justified by randomization, smoothed NCSS estimates were pooled across arms and extrapolated using standard parametric survival models (SPMs). Three best-fitting SPMs were blended using weights based on statistical goodness-of-fit criteria. NCSS estimates from blended SPMs (trial-based approach), a previously published model based on local lifetables matched to trial’s age, gender and enrollment distribution (lifetable-based approach [Weber et al, 2024]) and a conservative hybrid approach combining both were compared over observed and extrapolated periods.
RESULTS: Over observed follow-up, trial and lifetable-based NCSS showed no significant visual or statistical differences (p = 0.9426), with 0.08 year mean differential favoring trial-based approach. Over a lifetime horizon, hybrid approach yielded 0.7 additional years of mean NCSS than lifetable-based approach. Lifetable-based hazards underestimated trial-based hazards from 1.7 years onward, with divergence increasing over time. Beyond observed follow-up, NCSS rates from hybrid approach were on average 1.8% higher than those from lifetable-based approach.
CONCLUSIONS: Despite long-term trial follow-up, findings support a hybrid approach for estimating long-term NCSS and suggest that, contrary to prevailing belief, lifetables may not necessarily overestimate survival in cured patients. They also highlight the importance of incorporating trial-reported NCSS into survival analyses and economic evaluations.
Conference/Value in Health Info
2026-11, ISPOR Europe 2026, Vienna, Austria
Value in Health, Volume 29, Issue 12S
Code
MSR261
Topic
Clinical Outcomes, Epidemiology & Public Health, Methodological & Statistical Research
Disease
No Additional Disease & Conditions/Specialized Treatment Areas, Oncology