A meta-analysis of randomized, controlled trials in metastatic melanoma establishes progression-free survival as a surrogate for overall survival

Abstract

Background: Four randomized, phase 3 metastatic melanoma trials have shown a statistically significant overall survival (OS) benefit over the last few years. This provides the first opportunity to investigate progression free survival (PFS) as a potential surrogate for OS, taking into account a large number of recently conducted melanoma trials. Materials and Methods: A meta-analysis was performed of 10 randomized controlled trials enrolling 4,215 metastatic melanoma patients. From initially 48 published trials identified, those were selected which included DTIC as control arm, reported both PFS and OS, conform to the convention of reporting hazard ratio (HR) demonstrating benefit as <1. Experimental arm therapies included vemurafenib, dabrafenib, trametinib, sorafenib, ipilimumab, intetumumab, bosentan, oblimersen, temozolomide, and nabpaclitaxel. HRs for OS and PFS were correlated using several weighting strategies: by sample size or by precision of the HR estimate assuming fixed and random effects. Sensitivity analysis was performed including or excluding trials with crossover upon progression, phase 3 trials only, large trials, and only those with DTIC 1,000 mg/m2 dosing. Results: There was a statistically significant correlation between PFS and OS regardless of weighting strategy. Correlation coefficients (R) ranged from 0.74 (95% CI 0.26-0.93) for random effects assumption to 0.90 (95% CI 0.66-0.97) for sample size weighting. Considering only trials without crossover, R was 0.97 (95% CI 0.82-1.0) and only slightly lower after including 2 additional trials with less than 50% crossover (R = 0.94; 95% CI 0.74-0.99). Similarly, the correlation strength was nonsignificantly affected by restricting to phase 3, large, or DTIC 1,000 mg/m2 (0.95, 0.94, and 0.92, respectively). Only inclusion of mature follow-up data after >50% crossover (vemurafenib and dabrafenib phase 3 trials) significantly weakened PFS/OS correlation (R=0.58; 95% CI -0.03-0.88). We considered the inclusion of trials with no or limited crossover using the random effects assumption to yield a conservative and appropriate statement of the PFS/OS correlation: R=0.93 (95% CI 0.70-0.99). Conclusions: Based on a meta-analysis of all reported randomized trials in melanoma for which DTIC was the control arm therapy, we found a strong correlation between PFS and OS regardless of mechanistic class of therapy. PFS should be considered as a robust surrogate for OS in subsequent randomized trials.