# Radiographic progression-free survival as a surrogate endpoint for overall survival in first-line ARPi naïve metastatic castration-resistant prostate cancer

## TL;DR

This study shows that radiographic progression-free survival can reliably predict overall survival in a specific group of prostate cancer patients, reducing the need for long-term follow-up.

## Contribution

The study validates radiographic progression-free survival as a surrogate endpoint for overall survival in first-line ARPi naïve metastatic castration-resistant prostate cancer.

## Key findings

- Radiographic progression-free survival showed medium to strong correlations with overall survival across multiple analyses.
- A surrogate threshold effect of 0.83 was identified, allowing inference of overall survival benefits from radiographic progression-free survival results.
- Leave-one-out cross-validation confirmed the robustness of the predictive accuracy of radiographic progression-free survival as a surrogate endpoint.

## Abstract

Overall survival (OS) is the gold standard endpoint in oncology trials but requires long follow-up and may be confounded by post-protocol treatments. Radiographic progression-free survival (rPFS) is used as an earlier endpoint in metastatic castration-resistant prostate cancer (mCRPC). This study evaluated the validity of rPFS as a surrogate for OS in first-line, asymptomatic/mildly symptomatic, androgen receptor pathway inhibitor (ARPi) naïve, mCRPC using methods recommended by Germany’s Institute for Quality and Efficiency in Health Care (IQWiG).

A systematic search in Ovid® identified randomized controlled trials reporting both rPFS and OS. Trial-level rPFS-OS correlations of hazard ratios (HRs) were calculated using bivariate random-effects meta-analysis (BRMA) and weighted linear regression (WLR). Correlation strength was interpreted per IQWiG criteria. The surrogate threshold effect (STE) was estimated to assess surrogacy. Leave-one-out cross-validation (LOOCV) assessed model robustness. The primary analysis included trials meeting the proportional hazards (PH) assumption. Sensitivity analyses included trials violating PH and further excluded outliers outside the 95% confidence intervals (CIs) in the correlation plot.

Eleven RCTs were identified. The primary analysis (n = 10 trials) yielded medium correlations (BRMA R2: 0.78 [95% CI: 0.53-0.90]; WLR R2: 0.65 [0.40-0.90]; STE: 0.83). Sensitivity analyses yielded medium (n = 11 trials) and strong (n = 8 trials) correlations. LOOCV showed good predictive accuracy (75%-82%).

Results suggest rPFS is a valid surrogate for OS in first-line ARPi naïve mCRPC per IQWiG criteria. A statistically significant OS effect can be inferred for a trial demonstrating an upper confidence limit of HR < 0.83 in rPFS.