pCR shows poor surrogacy for OS in neoadjuvant GEA trials overall, but strong for esophageal adenocarcinoma

BACKGROUND: Pathologic complete response (pCR) is frequently used as a surrogate endpoint in neoadjuvant gastroesophageal adenocarcinoma (GEA) trials, yet its validity in predicting overall survival (OS) remains uncertain. METHODS: This systematic review and meta-analysis, follows PRISMA and ReSEEM guidelines, and evaluates pCR as a surrogate for OS in randomized controlled trials (RCTs) of neoadjuvant therapy for GEA. Randomized trials enrolling patients with resectable distal esophageal, gastroesophageal junction (GEJ), or gastric adenocarcinoma were included if both pCR and OS were reported. Individual-level surrogacy was assessed by correlating pCR rates with median OS across trial arms; trial-level surrogacy was evaluated by correlating treatment effects on pCR (odds ratios) with treatment effects on OS (hazard ratios) using weighted linear regression and calculating the coefficient of determination (R²). RESULTS: Twenty-six RCTs including 7452 patients were analyzed. Overall, pCR showed no meaningful correlation with OS at the individual-level (R² = 0.006; p = 0.210), and treatment effects on pCR explained little of the variability in OS effects at the trial level (R² = 0.060; p = 0.206). Most trials evaluated cytotoxic chemotherapy or chemoradiotherapy. Surrogacy varied by subgroup, being strong for esophageal adenocarcinoma (R² = 0.886; p < 0.001), moderate for gastric cancer (R² = 0.386; p = 0.041), and absent for GEJ trials. Trials incorporating radiotherapy demonstrated strong surrogacy, whereas chemotherapy-only trials did not. CONCLUSIONS: The interpretation of pCR results in gastroesophageal adenocarcinoma trials requires caution. Surrogacy might need to be reassessed in the upcoming immunotherapy trials.