Divergence between progression-free and overall survival treatment effect sizes in antibody-drug conjugate trials: A meta-epidemiologic study.

Abstract

e13076 Background: Progression-free survival (PFS) is frequently used as a primary endpoint in contemporary oncology trials, including antibody–drug conjugate (ADC) studies in metastatic breast cancer (MBC) settings, despite uncertainty regarding its validity as a surrogate for overall survival (OS). A meta-epidemiological analysis was performed to quantify the systematic differences between the treatment effects on PFS and OS in ADC randomized clinical trials (RCTs). Methods: Phase II–III ADC RCTs with MBC patients reporting hazard ratios (HRs) with 95% confidence intervals (CIs) for both progression-free survival (PFS) and overall survival (OS) were included. Differences in treatment effects within each trial were quantified using the ratio of hazard ratios (rHR = HR_PFS/HR_OS). Random-effects models were used to pool the rHRs across the trials. Prespecified subgroup and meta-regression analyses were used to evaluate the trial characteristics, clinical context, and regulatory status. Sensitivity analyses assessed the robustness across varying assumptions of the within-trial correlation between the endpoints. Results: Twenty-one RCTs evaluating nine ADCs in patients with MBC were included, comprising three phase II (14%) and 18 phase III (86%) studies. The pooled rHR demonstrated significantly larger treatment effects for PFS than for OS (rHR, 0.79; 95% CI, 0.73–0.86), with no between-trial heterogeneity (I² = 0%, τ² = 0). Sensitivity analyses assuming within-trial correlations between PFS and OS ranging from 0 to 0.95 yielded consistent results. Greater divergence between endpoints was observed in trials enrolling patients with triple-negative breast cancer (rHR, 0.72; 95% CI, 0.61–0.85) and HER2-low disease (rHR, 0.77; 95% CI, 0.62–0.96). Phase III trials showed larger differences than phase II trials (rHR, 0.79 vs. 0.89). When OS was a secondary endpoint, divergence was greater than when OS was primary (rHR, 0.77 vs. 0.87). Across trials, 16 of 21 studies (76%) demonstrated concordant effect directions between PFS and OS: 11 (52%) with a statistically significant benefit for both endpoints and 5 (24%) with neither endpoint significant. Discordance occurred in five trials (24%), all characterized by a statistically significant PFS benefit without a corresponding OS improvement. No trial demonstrated an OS benefit in the absence of a PFS benefit. Conclusions: Across ADC trials, PFS systematically overestimated OS treatment effects, underscoring the limitations of PFS as a standalone surrogate endpoint. These findings support the cautious interpretation of PFS-driven efficacy and reinforce the need for endpoint validation frameworks tailored to the mechanism, disease context, and trial design.