Advertisement

CN4 INVESTIGATING PROGRESSION-FREE SURVIVAL AS A POTENTIAL SURROGATE ENDPOINT FOR OVERALL SURVIVAL IN FIRST-LINE TREATMENT FOR GLIOBLASTOMA MULTIFORME: A META-ANALYSIS OF RANDOMIZED CONTROLLED TRIALS

      Objectives

      Due to unmet therapeutic needs in glioblastoma multiforme (GBM), identification of earlier endpoints for evaluating new therapies is desirable. Given that progression-free survival (PFS) is an earlier assessable endpoint than overall survival (OS), we evaluated whether PFS could act as a surrogate for OS among first-line therapies for GBM.

      Methods

      Phase 3 RCTs on methylated, unmethylated, and mixed (methylated, unmethylated, undetermined) populations from a systematic literature review informed the evidence base. Correlation between the trial-level log-hazard ratios (HRs) of PFS and OS was explored via bivariate random-effects meta-analyses (BRMA), and weighted linear regression (WLR) based on trial size. WLR was also used to derive estimated surrogacy equations and compute the surrogate threshold effect (STE), which is the minimal treatment effect on PFS required to predict a significant treatment effect on OS.

      Results

      Five, seven, and ten trials were considered for methylated, unmethylated, and mixed populations, respectively. The correlation coefficients between log(HRPFS) and log(HROS) in the BRMA for the methylated, unmethylated, and mixed populations were 0.73 (95% confidence interval: 0.01,0.95), 0.88 (0.58,0.97), and 0.71 (0.34,0.89). WLR analyses showed slightly greater correlation coefficients. Pairs of slope and intercept of the surrogacy equations expressing log(HROS) as a linear function of log(HRPFS) were (0.81,0.13), (1.20,0.04), and (0.72,0.02) for methylated, unmethylated, and mixed populations, respectively. The STEs implied by the surrogacy equations from the WLR were 0.34, 0.81, and 0.66 for methylated, unmethylated, and mixed populations, respectively.

      Conclusions

      Estimated correlations between treatment effects were only moderate and do not support PFS as a surrogate for OS. With sparse evidence and lack of individual-level data, further evidence is required to support PFS as a surrogate for OS in this context.