Utility of recurrence-based surrogate endpoints for survival in adjuvant cancer trials: evidence from melanoma
Menée à partir de données portant sur 10 379 patients atteints d'un mélanome de stade I-III traité entre 2000 et 2023, cette étude internationale évalue la possibilité d'utiliser, dans les essais cliniques de traitements adjuvants, la survie sans maladie et la survie sans métastases distantes comme critères de jugement de substitution à la survie globale et survie spécifique
Purpose : Surrogate endpoints are widely used to support interim analyses and regulatory approval of adjuvant therapies in resected solid tumours, yet many lack formal trial-level validation. We evaluated disease-free survival (DFS) and distant metastasis-free survival (DMFS) as surrogate endpoints for disease-specific survival (DSS) and overall survival (OS) in resected melanoma and assessed the suitability of OS as primary endpoint.
Patients and Methods : A total of 10,379 patients with AJCC stage I–III melanoma treated between 2000 and 2023 at five international centres were analysed; all underwent sentinel node biopsy. Trial-level surrogacy was assessed using centre-specific log hazard ratios and restricted mean survival time at prespecified surrogate horizons (12 to 48 months), with primary endpoints truncated at 60 months. Individual-level surrogacy was evaluated using Kendall’s
τ. Prespecified subgroup analyses compared pT-defined IB
–IIA and IIB–IIC disease.
Results : There were 2,418 DFS events, 1,833 DMFS events, and 1,235 melanoma-specific deaths. Trial-level surrogacy for DFS with DSS was strongest at 24 to 36 months (R2 up to 0.95). DMFS showed consistently high surrogacy for DSS from 24 to 48 months (R2 0.87 to 0.91). Correlations with OS were weak at all horizons (R2 < 0.40). Surrogacy was stage-dependent, with robust performance in pT-defined IIB–IIC disease and limited surrogacy in IB–IIA disease.
Conclusion : DFS and DMFS are valid surrogate endpoints for DSS in resected melanoma at later time horizons, particularly in higher-risk disease. OS shows attenuated correlation with recurrence-based endpoints, limiting its utility for early endpoint evaluation in adjuvant trials.
Journal of the National Cancer Institute , article en libre accès, 2026