Time-dependent effects of BCG plus immune checkpoint inhibitors in BCG-naive high-risk non-muscle-invasive bladder cancer: a reconstructed patient-level analysis.

PURPOSE: Three phase 3 trials (CREST, POTOMAC, ALBAN) adding an immune checkpoint inhibitor (ICI) to Bacillus Calmette-Guérin (BCG) in BCG-naive high-risk non-muscle-invasive bladder cancer (NMIBC) reported different results. We used reconstructed individual patient-level data (rIPD) to examine the overall event-free survival (EFS) effect and its change over time. METHODS: This post hoc analysis used data reconstructed from published Kaplan-Meier curves rather than original trial datasets, harmonized as EFS. A trial-stratified Cox model estimated the pooled hazard ratio (HR). Two-parameter Weibull models described the timing of events (scale, β) and the temporal pattern of risk (shape, α), with proportional hazards assessed by Schoenfeld residuals and log-log plots. The trials differed in agent, endpoint definition, and BCG schedule. RESULTS: Reconstructed event counts and HRs closely matched reported values. The pooled EFS HR was 0.76 (95% confidence interval [CI] 0.63-0.91), and 0.68 (95% CI 0.54-0.85) after excluding ALBAN. Weibull-based cumulative HRs suggested differing time patterns: the fitted cumulative HR decreased over time in CREST, remained broadly stable in POTOMAC, and approached 1 in ALBAN. Model-based 5-year EFS differences (11.5%, 8.4%, and 0.4%) were exploratory, relying on extrapolation in CREST and sparse tail data in ALBAN. CONCLUSION: ICI plus BCG was associated with improved EFS, mainly reflecting CREST and POTOMAC. A single-pooled HR may obscure regimen-level differences in how the benefit evolves over time. Given the reconstructed data, between-trial differences, and restriction to efficacy, these findings are hypothesis-clarifying rather than practice-changing.