Metastasis-Free Survival, but Not Biochemical Failure, is a Strong Surrogate Endpoint for Overall Survival in Recurrent Prostate Cancer: Analysis of NRG Oncology/RTOG 9601

The Intermediate Clinical Endpoints in Cancer of the Prostate (ICECaP) working group identified metastasis-free survival (MFS) as a valid surrogate for overall survival (OS) in men treated for localized prostate cancer. Only 8% of men in the ICECaP MFS analysis were treated with prostatectomy and no salvage radiotherapy (SRT) trials were included. Surrogate endpoints are treatment and disease state specific, thus the performance of intermediate clinical endpoints (ICEs) to function as surrogate endpoints in the SRT setting are unknown. Leveraging NRG/RTOG 9601, we aimed to determine the performance of ICEs as surrogate endpoints for OS in men receiving SRT. Endpoints assessed for surrogacy were biochemical failure (BF), as defined in NRG/RTOG 9601(PSA nadir + 0.3-0.5 ng/mL, or initiation of salvage hormone therapy) and NRG/RTOG 0534 (PSA nadir + 2 ng/mL), distant metastasis (DM), and MFS (DM or all-cause death). Surrogacy was assessed by the Prentice criteria, however given its well-established limitations, the preferred methodology used in ICECaP of a two-stage meta-analytic approach was also utilized. The meta-analytic approach requires two conditions: (1) that the ICE be correlated with OS and (2) that the treatment effect on the ICE and OS be correlated. Condition one was assessed at the patient level with Kendall’s τ and condition two was assessed by randomly dividing the entire trial cohort into 10 pseudo trial centers and calculating the R2 between treatment hazard ratios for ICE and OS. BF defined in NRG/RTOG 9601 and by nadir + 2 ng/mL, DM, and MFS all satisfied each of the four Prentice criteria for surrogacy for OS. However, with the two-condition approach for assessment of surrogacy, there was strong correlation between MFS and OS (τ = 0.86), moderate correlation between DM and OS (τ = 0.66), and weaker correlation between BF (τ = 0.25) or PSA nadir + 2 ng/mL (τ = 0.40) and OS. Similarly, for condition two, the treatment effect of antiandrogen therapy on MFS and OS were correlated (R2 = 0.67), but this was not true for BF and OS (R2 = 0.09), PSA nadir + 2 ng/mL and OS (R2 = 0.12), or DM and OS (R2 = 0.18). We demonstrate that MFS is a strong surrogate for OS in men receiving SRT for recurrent prostate cancer after prostatectomy, consistent with data in the intact primary treatment setting. While BF and PSA nadir + 2 ng/mL both are prognostic, these PSA based ICEs were weakly correlated with OS at the patient level. Similarly, the treatment effect of antiandrogen therapy on these endpoints was only weakly correlated with treatment effect on OS. Thus, caution should be used when inferring clinical benefit from effects on BF as a surrogate for OS.