WHEN THE TAIL WAGS THE DOG V: RELATIVE SURVIVAL DISCIPLINES AN IMPLAUSIBLE OVERALL-SURVIVAL EXTRAPOLATION IN CURATIVE-INTENT BLADDER CANCER (KEYNOTE-905/EV-303)
Author(s)
Andrew Briggs, DPhil1, Alex Hill, PhD1, Anna Forsythe, MBA, MSc, PharmD2.
1London School of Hygiene & Tropical Medicine, London, United Kingdom, 2Founder, Oncoscope, Miami, FL, USA.
1London School of Hygiene & Tropical Medicine, London, United Kingdom, 2Founder, Oncoscope, Miami, FL, USA.
OBJECTIVES: In immature, curative-intent trials patients may be cured and track population mortality, yet unconstrained extrapolation can project wildly divergent lifetime benefits. Using KEYNOTE-905/EV-303 (perioperative enfortumab vedotin plus pembrolizumab vs surgery in cisplatin-ineligible muscle-invasive bladder cancer), we tested whether a relative-survival floor disciplines it, applying the Extrapolation Benefit Share (EBS: the share of incremental mean survival arising beyond the observed data) and Survival Ratio Shift (SRS: whether the advantage is tail-concentrated).
METHODS: Pseudo-individual patient data were reconstructed (SurvdigitizeR; Guyot algorithm) from the published event-free survival (EFS) and overall survival (OS) Kaplan-Meier curves. Eight parametric distributions were fitted jointly. Mean survival was decomposed into observed and extrapolated components over a 40-year horizon. OS was additionally modelled as relative survival - excess hazard plus a UK male population-mortality floor.
RESULTS: OS was significant (HR 0.50) but immature; the all-cause extrapolation was implausible and divergent: incremental survival ranged 47-104 months, with decreasing-hazard fits (Gompertz 104, Royston-Parmar 90) extrapolating the plateau as a near-permanent survival fraction. The relative-survival floor disciplined this - clipping Gompertz from 104 to 38 months and narrowing the OS range to 20-38 - as the fitted excess hazards converged onto the population line (a long-term-survivor fraction). EBS stayed high (75-83%; OS immature), but the floor bounded the magnitude and resolved the divergence. The surrogate EFS, with no floor, extrapolated freely (47-145 months) - what OS would look like unconstrained. The advantage was modestly tail-loaded (SRS 1.1-2.0).
CONCLUSIONS: In a curative-intent, high-cure-fraction setting the population-mortality floor is not cosmetic: it diminishes the implausible tails and turns a divergent OS extrapolation into a bounded range. Where disease-specific mortality no longer dominates, relative survival earns its keep - the counterpoint to settings where it correctly does nothing. Costless where excess mortality dominates, decisive where it does not, the floor should be standard.
METHODS: Pseudo-individual patient data were reconstructed (SurvdigitizeR; Guyot algorithm) from the published event-free survival (EFS) and overall survival (OS) Kaplan-Meier curves. Eight parametric distributions were fitted jointly. Mean survival was decomposed into observed and extrapolated components over a 40-year horizon. OS was additionally modelled as relative survival - excess hazard plus a UK male population-mortality floor.
RESULTS: OS was significant (HR 0.50) but immature; the all-cause extrapolation was implausible and divergent: incremental survival ranged 47-104 months, with decreasing-hazard fits (Gompertz 104, Royston-Parmar 90) extrapolating the plateau as a near-permanent survival fraction. The relative-survival floor disciplined this - clipping Gompertz from 104 to 38 months and narrowing the OS range to 20-38 - as the fitted excess hazards converged onto the population line (a long-term-survivor fraction). EBS stayed high (75-83%; OS immature), but the floor bounded the magnitude and resolved the divergence. The surrogate EFS, with no floor, extrapolated freely (47-145 months) - what OS would look like unconstrained. The advantage was modestly tail-loaded (SRS 1.1-2.0).
CONCLUSIONS: In a curative-intent, high-cure-fraction setting the population-mortality floor is not cosmetic: it diminishes the implausible tails and turns a divergent OS extrapolation into a bounded range. Where disease-specific mortality no longer dominates, relative survival earns its keep - the counterpoint to settings where it correctly does nothing. Costless where excess mortality dominates, decisive where it does not, the floor should be standard.
Conference/Value in Health Info
2026-11, ISPOR Europe 2026, Vienna, Austria
Value in Health, Volume 29, Issue 12S
Code
MSR194
Topic
Clinical Outcomes, Economic Evaluation, Methodological & Statistical Research
Disease
Oncology