OBJECTIVES

To evaluate PFS as a surrogate endpoint for OS by modelling association between the treatment effects on each endpoint using RCTs investigating first-line (1L) therapies for advanced melanoma.

METHODS

A systematic literature review was conducted to identify trials reporting OS and PFS in 1L advanced melanoma evaluating any pharmacological interventions. The surrogacy relationship was assessed using bivariate random-effects meta-analysis and weighted linear regression. Validity and predictive accuracy of the estimated surrogacy equations were assessed via leave-one-out cross-validation (LOOCV). The primary analysis was conducted using all trials meeting the proportional hazards assumption (n=23) with crossover-adjusted hazard ratios (HRs) where available. Sensitivity analyses were also conducted by restricting the evidence base to: (1) phase II and III immune checkpoint inhibitor trials (n=5), (2) phase III trials (n=17), (3) trials with no crossover permitted or trials that provided adjusted crossover estimates (n=7), and (4) trials in the previous analysis set plus trials with unreported crossover descriptions (n=19).

RESULTS

Of 23 trials identified from the SLR and eligible for the primary analysis (9,737 patients), the estimated correlation coefficient was 0.72 (95% confidence interval [CI]: 0.50-0.85). Correlation estimates in the sensitivity analyses ranged from 0.71 to 0.88, with sensitivity analyses (2) and (3) reporting relatively stronger point estimates for correlations: 0.88 (95% CI: 0.71-0.95) and 0.87 (95% CI: 0.45-0.98), respectively. In LOOCV, alignment between observed and predicted OS HRs was ≥90% for the primary analysis and sensitivity analysis (2), and >95% in sensitivity analyses (1), (3), and (4).

CONCLUSIONS

Results suggest a moderate correlation between PFS and OS across all analyses. Given the clinical effects of subsequent treatments, the restriction of the evidence base to trials adjusting for crossover may be important in the 1L cohort, as supported by the improved correlation in the sensitivity analyses.