AI3 Disease-Free Survival (DFS) As a Surrogate Endpoint for Overall Survival (OS) in Adults with Resectable Esophageal or Gastroesophageal Junction Cancer: A Correlation Meta-Analysis


      To evaluate the appropriateness of using DFS as a surrogate for OS in trials investigating adults with resectable esophageal or gastroesophageal junction cancer (EC/GEJ) receiving therapies in the (neo)adjuvant and perioperative settings.


      A systematic literature review was conducted to identify EC/GEJ trials that reported OS and DFS, or progression-free survival (PFS) compatible with the definition of DFS. The primary analysis was restricted to studies meeting the proportional hazards (PH) assumption which was assessed statistically after digitization of the reported KM curves. Secondary analysis was restricted to the adjuvant setting only. Sensitivity analyses consisted of removing studies with outlier characteristics and other potential biases, and estimating differences between restricted mean survival times to assess the impact of including studies with non-proportional hazards. The surrogacy relationship was assessed using bivariate random-effects meta-analysis and weighted linear regression. Surrogate threshold effect (STE), the minimum DFS/PFS benefit that would translate into OS benefit, was also estimated. The robustness of the models was tested via leave-one-out cross-validation (LOOCV).


      The primary analysis included 26 trials after removing 7 trials not meeting the PH assumption. The estimated correlation coefficient was 0.83 (95% CI: 0.70-0.90). The resulting surrogacy equation was log(HROS) = 0 + 0.80 × log(HRDFS/PFS) with a corresponding STE of 0.82. LOOCV verified the stability of the results with 95% alignment between observed and estimated HROS. Correlation estimates across secondary analyses and sensitivity scenarios were similar to those from the primary analysis.


      The results and their validation point out a correlation between DFS and OS in EC/GEJ. The estimated surrogacy equation and the corresponding STE can enable HRDFS/PFS as a surrogate predictor of HROS in EC/GEJ trials. These findings can be useful when evaluating the long-term efficacy of treatments where OS may not be immediately available.