METHODS FOR MODELING IMMATURE OVERALL SURVIVAL (OS) IN BREAST CANCER (BC) IN NATIONAL INSTITUTE FOR HEALTH AND CARE EXCELLENCE (NICE) TECHNOLOGY APPRAISALS (TAS)
Author(s)
Chloe Spalding, PhD1, ELEANOR PAUL, MSc2, Alicia N. Pepper, PhD2, Melody Zhao, MSc2, Eesha Kodi, MPH2, Rodrigo Martins de Almeida, MBA3.
1Daiichi Sankyo, Birmingham, United Kingdom, 2Precision AQ, Vancouver, BC, Canada, 3AstraZeneca, Cambridge, United Kingdom.
1Daiichi Sankyo, Birmingham, United Kingdom, 2Precision AQ, Vancouver, BC, Canada, 3AstraZeneca, Cambridge, United Kingdom.
OBJECTIVES: OS is often a key driver of cost-effectiveness analysis (CEA) results but can be uncertain given long-term extrapolations. This review identifies methods used to model immature OS in NICE TAs in BC and summarizes how these approaches were received.
METHODS: A targeted review was conducted in April 2026 leveraging the workbook published by NICE compiling cancer TAs published since 2000. To identify TAs relevant for review, the workbook was filtered using the following criteria: BC, single technology appraisals, 2020 or later, and exclusion of terminated or Cancer Drugs Fund exit TAs.
RESULTS: A total of 16 TAs were identified evaluating treatments for BC. Eleven TAs used partitioned survival models for advanced BC and modelled OS directly. Five used cohort state transition models for adjuvant and/or neoadjuvant BC and modelled OS indirectly. Within-trial crossover occurred in two trials; however, crossover adjustment was not explored. One TA adjusted for subsequent treatments using rank-preserving structural failure time methods. Five TAs discussed surrogacy relationships to justify long-term OS; however, none incorporated surrogacy assumptions when extrapolating OS. Extrapolated OS was validated against external trials, real-world evidence, and/or clinical expert opinion in all TAs, but structured expert elicitation was not employed. Treatment waning was applied in five TAs. EAG critiques on OS primarily focused on the choice of distribution for extrapolation (n=7), maturity of the primary trial data (n=5), treatment waning (n=5), and robustness of comparative analyses (n=3).
CONCLUSIONS: Based on a review of 16 BC TAs, CEA results were often considered highly uncertain by NICE in the absence of mature OS data. Primary trial OS data were preferred over external sources when the populations were not comparable; however, external sources were supportive when appropriately aligned and justified. Validation of OS extrapolations against external sources and clinical plausibility supported by expert opinion were critical.
METHODS: A targeted review was conducted in April 2026 leveraging the workbook published by NICE compiling cancer TAs published since 2000. To identify TAs relevant for review, the workbook was filtered using the following criteria: BC, single technology appraisals, 2020 or later, and exclusion of terminated or Cancer Drugs Fund exit TAs.
RESULTS: A total of 16 TAs were identified evaluating treatments for BC. Eleven TAs used partitioned survival models for advanced BC and modelled OS directly. Five used cohort state transition models for adjuvant and/or neoadjuvant BC and modelled OS indirectly. Within-trial crossover occurred in two trials; however, crossover adjustment was not explored. One TA adjusted for subsequent treatments using rank-preserving structural failure time methods. Five TAs discussed surrogacy relationships to justify long-term OS; however, none incorporated surrogacy assumptions when extrapolating OS. Extrapolated OS was validated against external trials, real-world evidence, and/or clinical expert opinion in all TAs, but structured expert elicitation was not employed. Treatment waning was applied in five TAs. EAG critiques on OS primarily focused on the choice of distribution for extrapolation (n=7), maturity of the primary trial data (n=5), treatment waning (n=5), and robustness of comparative analyses (n=3).
CONCLUSIONS: Based on a review of 16 BC TAs, CEA results were often considered highly uncertain by NICE in the absence of mature OS data. Primary trial OS data were preferred over external sources when the populations were not comparable; however, external sources were supportive when appropriately aligned and justified. Validation of OS extrapolations against external sources and clinical plausibility supported by expert opinion were critical.
Conference/Value in Health Info
2026-11, ISPOR Europe 2026, Vienna, Austria
Value in Health, Volume 29, Issue 12S
Code
HTA179
Topic
Health Technology Assessment, Methodological & Statistical Research
Topic Subcategory
Decision & Deliberative Processes
Disease
Oncology