ESMO-MCBS: DO CURRENT CRITERIA UNDERVALUE CLINICAL BENEFIT IN POOR-PROGNOSIS POPULATIONS
Author(s)
Caroline M.J. van der Meijden, MSc, PhD1, Isaac Corro Ramos, MSc, PhD2, Carin Uyl-De Groot, Sr., PhD3.
1Valence Consultancy & iMTA Erasmus University, Rotterdam, Netherlands, 2iMTA Erasmus University, Rotterdam, Netherlands, 3ESHPM/iMTA Erasmus University, Rotterdam, Netherlands.
1Valence Consultancy & iMTA Erasmus University, Rotterdam, Netherlands, 2iMTA Erasmus University, Rotterdam, Netherlands, 3ESHPM/iMTA Erasmus University, Rotterdam, Netherlands.
OBJECTIVES: The European Society for Medical Oncology Magnitude of Clinical Benefit Scale (ESMO-MCBS) uses hazard ratios (HR) and median survival to grade the clinical benefit of anticancer therapies. However, HRs are relative measures whose translation into absolute survival gains depends on baseline survival and survival curve shape, typically interpreted under proportional hazards. This study examines whether current MCBS criteria adequately capture clinical benefit, using number needed to treat (NNT) as an anchor of absolute benefit.
METHODS: We modelled the HR required to achieve absolute survival gains corresponding to NNTs of 10, 20, and 33 across baseline survival values of 10-90%, using Survival_treatment(t)=Survival_comparator(t)^HR. Required HRs were compared with ESMO-MCBS thresholds for curative (Form 1) and non-curative (Form 2a) settings. Form 1 applies HR thresholds to point estimates and grades benefit based on absolute gains in OS or DFS. Form 2a applies thresholds to the lower bound of the 95% CI for HR and grades benefit based on absolute OS gain, stratified by control arm median OS. To illustrate findings, we reviewed published MCBS scores, alongside trial-reported HRs and survival gains for selected therapies.
RESULTS: For baseline survival of 10-60%, even an NNT of 10 does not yield HRs meeting the curative threshold of 0.65 or lower. NNTs of 20 and 33 require HRs only achievable at baseline survival more than 70%. The HR-baseline survival relationship is non-linear, with greatest disadvantage at 30-40% survival. Empirical review confirmed that poorer prognosis settings require smaller NNTs and achieve lower scores.
CONCLUSIONS: ESMO-MCBS criteria structurally disadvantage poor-prognosis populations, with the HR threshold (0.65 or lower) rarely attainable. These findings raise questions about how clinical benefit should be measured in poor prognosis populations and whether alternative absolute effect measures, including NNT and restricted mean survival time (RMST), better capture clinical benefit.
METHODS: We modelled the HR required to achieve absolute survival gains corresponding to NNTs of 10, 20, and 33 across baseline survival values of 10-90%, using Survival_treatment(t)=Survival_comparator(t)^HR. Required HRs were compared with ESMO-MCBS thresholds for curative (Form 1) and non-curative (Form 2a) settings. Form 1 applies HR thresholds to point estimates and grades benefit based on absolute gains in OS or DFS. Form 2a applies thresholds to the lower bound of the 95% CI for HR and grades benefit based on absolute OS gain, stratified by control arm median OS. To illustrate findings, we reviewed published MCBS scores, alongside trial-reported HRs and survival gains for selected therapies.
RESULTS: For baseline survival of 10-60%, even an NNT of 10 does not yield HRs meeting the curative threshold of 0.65 or lower. NNTs of 20 and 33 require HRs only achievable at baseline survival more than 70%. The HR-baseline survival relationship is non-linear, with greatest disadvantage at 30-40% survival. Empirical review confirmed that poorer prognosis settings require smaller NNTs and achieve lower scores.
CONCLUSIONS: ESMO-MCBS criteria structurally disadvantage poor-prognosis populations, with the HR threshold (0.65 or lower) rarely attainable. These findings raise questions about how clinical benefit should be measured in poor prognosis populations and whether alternative absolute effect measures, including NNT and restricted mean survival time (RMST), better capture clinical benefit.
Conference/Value in Health Info
2026-11, ISPOR Europe 2026, Vienna, Austria
Value in Health, Volume 29, Issue 12S
Code
PT4
Topic
Clinical Outcomes, Health Technology Assessment, Methodological & Statistical Research
Topic Subcategory
Decision & Deliberative Processes, Value Frameworks & Dossier Format
Disease
Oncology