Surrogacy of intermediate clinical endpoints for overall survival in BCG-naïve high-risk non-muscle-invasive bladder cancer.

INTRODUCTION

Recent trials in BCG-naïve high-risk non-muscle-invasive bladder cancer (HR-NMIBC) have added immunotherapy to BCG, often using event-free survival (EFS) as the primary endpoint. However, neither EFS nor progression-free survival (PFS) has been validated as a surrogate for overall survival (OS) in this setting.

METHODS

We retrospectively analyzed data from 18 international centers, including patients with BCG-naïve HR-NMIBC treated between 2001 and 2025. Adequate BCG exposure (International Bladder Cancer Group criteria) was the treatment contrast for its effects on intermediate clinical endpoints (ICEs) and OS, using a 12-month landmark approach and inverse probability of treatment weighting (IPTW). Surrogacy was assessed with: (1) adapted Prentice criteria, with time-varying ICEs in IPTW-adjusted Cox models for OS; and (2) an emulated two-stage meta-analytic framework estimating individual-level association (Kendall's τ) and pseudo-trial-level surrogacy (R2) across 500 random pseudo-trial replications.

RESULTS

Of 3384 patients, 3153 entered the landmark analysis; 2697 (85.5%) received adequate BCG. Median follow-up was 50 months (IQR 17-77). Adequate BCG was associated with improved OS (HR 0.63, 95% CI 0.47-0.83) and lower risks of PFS events (HR 0.46, 95% CI 0.37-0.58) and EFS events (HR 0.48, 95% CI 0.40-0.58). Both ICEs were independently associated with OS, but adjusting for each did not fully attenuate the BCG effect. Individual-level association with OS was strong for PFS (τ = 0.79, 95% CI 0.75-0.83) and EFS (τ = 0.71, 95% CI 0.66-0.75). Pseudo-trial-level surrogacy was moderate, with median R2 of 0.60 (PFS) and 0.51 (EFS) and wide dispersion across replications, both below the prespecified threshold for adequate surrogacy (≥ 0.70).

CONCLUSIONS

In this exploratory, hypothesis-generating analysis, PFS and EFS were strongly associated with OS at the individual level but failed to demonstrate robust pseudo-trial-level surrogacy. These endpoints may not fully capture treatment effects on OS, warranting caution when interpreting ICE-driven trial results in BCG-naïve HR-NMIBC.