Event-Free Survival, a Prostate-Specific Antigen-Based Composite End Point, Is Not a Surrogate for Overall Survival in Men With Localized Prostate Cancer Treated With Radiation

Abstract

Purpose: Recently, we have shown that metastasis-free survival is a strong surrogate for overall survival (OS) in men with intermediate- and high-risk localized prostate cancer and can accelerate the evaluation of new (neo)adjuvant therapies. Event-free survival (EFS), an earlier prostate-specific antigen (PSA)-based composite end point, may further expedite trial completion.

Methods: EFS was defined as the time from random assignment to the date of first evidence of disease recurrence, including biochemical failure, local or regional recurrence, distant metastasis, or death from any cause, or was censored at the date of last PSA assessment. Individual patient data from trials within the Intermediate Clinical Endpoints in Cancer of the Prostate-ICECaP-database with evaluable PSA and disease follow-up data were analyzed. We evaluated the surrogacy of EFS for OS using a 2-stage meta-analytic validation model by determining the correlation of EFS with OS (patient level) and the correlation of treatment effects (hazard ratios [HRs]) on both EFS and OS (trial level). A clinically relevant surrogacy was defined a priori as an R² ≥ 0.7.

Results: Data for 10,350 patients were analyzed from 15 radiation therapy-based trials enrolled from 1987 to 2011 with a median follow-up of 10 years. At the patient level, the correlation of EFS with OS was 0.43 (95% CI, 0.42 to 0.44) as measured by Kendall's tau from a copula model. At the trial level, the R² was 0.35 (95% CI, 0.01 to 0.60) from the weighted linear regression of log(HR)-OS on log(HR)-EFS.

Conclusion: EFS is a weak surrogate for OS and is not suitable for use as an intermediate clinical end point to substitute for OS to accelerate phase III (neo)adjuvant trials of prostate cancer therapies for primary radiation therapy-based trials.

FIG 1.

(A) Kaplan-Meier estimates of overall survival (OS), metastasis-free survival (MFS), and event-free survival (EFS) where non–prostate cancer deaths were counted as events. (B) Kaplan-Meier estimates of disease-specific survival (DSS), time to metastasis (TTM), and time to event (TTE) where non–prostate cancer deaths were censored. (C) Estimated hazard functions over time for OS, MFS, and EFS. (D) Estimated hazard functions over time for DSS, TTM, and TTE. Median follow-up was 10 years.

FIG 2.

Bubble plot and regression of (A) overall survival (OS) at 8 years on event-free survival (EFS) at 5 years; (B) metastasis-free survival (MFS) at 5 years on EFS at 3 years; (C) disease-specific survival (DSS) at 8 years on time to event (TTE) at 5 years; and (D) time to metastasis (TTM) at 5 years on TTE at 3 years. All rates were Kaplan-Meier estimates by trial and treatment arm. Circle size and regression were weighted by inverse variance of the 5- or 3-year estimate for EFS and TTE.

FIG 3.

Bubble plot and regression of (A) hazard ratio for overall survival (OS-HR) on hazard ratio for event-free survival (EFS-HR); (B) hazard ratio for metastasis-free survival (MFS-HR) on EFS-HR; (C) hazard ratio for disease-specific survival (DSS-HR) on hazard ratio for time to event (TTE-HR); (D) hazard ratio for time to metastasis (TTM-HR) on TTE-HR. Cox proportional hazards regression estimated HR for each study, and values were natural logarithm transformed. Circle size and regression were weighted by inverse variance of log(HR) estimates for EFS or TTE. The orange lines display surrogate threshold effect (STE), which is the intersection of the upper 95% prediction limit with the horizontal line representing an HR of 1 for the true end points.