Correlation analysis to assess surrogate endpoints for overall survival (OS) in triple-negative breast cancer (TNBC).

Abstract

e13089 Background: The use of surrogate endpoints can help reduce the trial duration and cost which is important for aggressive diseases such as TNBC. Previous studies suggest progression-free survival (PFS) may serve as a surrogate for OS in metastatic TNBC. We aim to evaluate PFS and additional surrogate endpoints for their correlation with OS in localized and advanced TNBC. Methods: TNBC phase 3 trials and pooled analysis as well as trials including TNBC subgroups, were identified through a systematic search using LARVOL CLIN, PubMed, clinicaltrials.gov, and conference proceedings. Out of these, trials reporting median survival and hazard ratios (HR) for both OS and surrogate endpoints were considered for further statistical analysis. Surrogate endpoints reported by fewer than 4 trials were excluded from the study. Pearson correlation (r) and determination coefficient (R2) were calculated to assess surrogacy between Median OS and Median PFS, HR OS and HR PFS, HR disease-free survival (DFS), and invasive disease-free survival (IDFS). Weighted regression analysis was conducted controlling for sample size among trial arms. Results: 26 TNBC cohorts and 2 pooled analyses evaluating PARP, PD-1, PD-L1 inhibitors, and chemotherapies were included. The association analysis results are shown. There was a significant correlation between median OS and median PFS (r = 0.73, p value = 1.026×10−6) as well as between HR OS and HR PFS (r = 0.61, p value = 0.009). There was also a significant correlation between HR DFS and HR IDFS with HR OS (r = 0.92, p value = 0.003; r = 0.99, p value = 0.014). Conclusions: PFS is significantly correlated with OS in TNBC across diverse therapies. Also, HR PFS, HR DFS, and HR IDFS are significantly correlated with HR OS. Overall, these results illustrate that PFS could be used as a surrogate for OS in TNBC. Subsequent studies should include early-phase trials to increase association robustness and enable correlation comparison between the diverse mechanism of actions and other surrogate endpoints. [Table: see text]