Evaluating the potential of candidate surrogate endpoints for overall survival in treatment-naive metastatic melanoma treated with anti-PD-1-based regimens with or without anti-CTLA-4 therapy.

Using surrogates for overall survival (OS) may expedite the development of, and patient access to, novel treatments. We assessed potential surrogates for OS in patients with metastatic melanoma treated with nivolumab-containing regimens in the first-line treatment setting. We used individual-patient data from 1865 patients enrolled in four randomized controlled trials studying single-agent nivolumab or combinations of nivolumab and ipilimumab against dacarbazine or immunotherapy. Using the two-level meta-analytic framework, we evaluated three candidate surrogates: objective response rate (ORR), progression-free survival (PFS), and time to next treatment or death (TNTD). We measured the patient-level associations between candidates and OS using ORs in the case of ORR and Spearman's correlation coefficient (ρ) in the case of time-to-event surrogate endpoints. We used R2 to measure the trial-level association between ORs or HRs for each surrogate and the HRs for OS. For ORR, at the individual-level, OR of survival was equal to 12.29 (95% CI 9.78 to 14.80), and at the trial-level R2 was equal to 0.62 (95% CI 0 to 1.00). For PFS, at the individual-level ρ was equal to 0.72 (95% CI 0.70 to 0.73), and at the trial-level R2 was equal to 0.73 (95% CI 0.27 to 1.00). For TNTD, at the individual-level ρ was equal to 0.77 (95% CI 0.76 to 0.78), and at the trial-level R2 was equal to 0.77 (95% CI 0.37 to 1.00). In cross-validation, the 95% prediction intervals for HRs for OS predicted by regression models always contained the observed HRs for OS, indicating the stability of the models. At the individual-level, ORR exhibited a strong correlation with OS, whereas PFS and TNTD showed a moderate level correlation with OS. At the trial level, the key requirement for validating surrogates, all candidate surrogates demonstrated moderate predictive abilities for OS in future trials. These findings should be interpreted within the context of anti-PD-1-based therapies, with or without anti-CTLA-4 combinations, consistent with the trial evidence base included in this study.