Recurrence-free survival as a surrogate endpoint for overall survival in resectable esophageal cancer: An individual patient data analysis of phase III RCTs.

4068 Background: Overall survival (OS) is regarded as the gold standard efficacy endpoint but requires long follow-up. This study aimed to determine the validity of recurrence-free survival (RFS) as a surrogate endpoint for OS in resectable esophageal cancer. Methods: A systematic review of phase III randomized controlled trials (RCTs) comparing perioperative treatments for resectable advanced esophageal and gastroesophageal junction cancer was conducted. Individual patient data (IPD) were requested from all included trials. Surrogacy between RFS and OS was assessed at the individual level using the Kendall rank correlation coefficient ( τ ) and at the trial level using the coefficient of determination ( R² ) from a meta-regression model. A τ of 0.8 and an R² of 0.65 were considered thresholds indicative of a good surrogate endpoint. Results: Twenty-two eligible trials were identified by the systematic review, and IPD were available from 10 RCTs (JCOG1109, JCOG9907, JCOG9204, FFCD9901, FFCD9102, SAKK75/08, CROSS, KOK, NeoRes2 and CMISG1701), including 2,145 patients who underwent R0 resection (cStage IV, cT1N0 and cT4b excluded). Of these, 1563 patients had squamous cell carcinoma, and 575 patients had adenocarcinoma. The 5-year OS and RFS rates were 53.2% and 46.2%, respectively, with a median OS of 6.2 years and a median RFS of 3.6 years. For individual-level surrogacy, Kendall’s τ was 0.823 (95% CI: 0.807–0.839). Subgroup analysis based on treatment modality revealed τ values of 0.830 (95% CI: 0.800–0.861) for patients receiving neoadjuvant chemotherapy (NAC; n = 586), 0.827 (95% CI: 0.803–0.850) for those receiving neoadjuvant chemoradiotherapy (NACRT; n = 982), 0.770 (95% CI: 0.713–0.828) for the surgery-alone group (n = 320), and 0.861 (95% CI: 0.824–0.898) for the adjuvant chemotherapy group (n = 257). Trial-level surrogacy analysis across all 22 trials demonstrated an R 2 of 0.735 (95% CI: 0.512–0.939). The surrogate threshold effect was 0.929, indicating the minimum RFS treatment effect required to predict a nonzero effect on OS. Conclusions: This study demonstrated strong individual-level and trial-level surrogacy between RFS and OS in surgically resectable esophageal cancer across all perioperative treatment modalities. These findings hold promise for expediting the development of novel perioperative treatment by shortening the follow-up of clinical trials on esophageal cancer.