312 Background: The survival benefit and optimal timing of immune checkpoint inhibitors (ICIs) combined with definitive chemoradiotherapy (CRT) remain undefined in the immunotherapy era. This study evaluated the efficacy and safety of ICIs combined with definitive CRT for locally advanced esophageal cancer, and identified prognostic factors and optimal combination strategy. Methods: A total of 265 patients with unresectable locally advanced esophageal cancer receiving CRT between 2017 and 2024 were analyzed retrospectively from two Chinese institutions. Patients were categorized into the CRT group (n=144) and the CRT+ICI group (n=121). Propensity score matching (PSM) balanced baseline characteristics. Outcomes included progression-free survival (PFS), overall survival (OS), recurrence patterns, treatment-related adverse events (TRAEs), and dynamic peripheral blood indicators profiling. Survival analyses used Kaplan-Meier methods and Cox regression, flow cytometry quantified lymphocyte subsets. Subgroup analyses based on immunotherapy timing (concurrent, induction, or maintenance) were performed for multiple pairwise survival comparisons. Results: The addition of ICIs to CRT significantly improved median PFS (27.0 vs 18.0 months; HR 0.675, 95% CI 0.481-0.947; p = 0.023) and OS (54.0 vs 27.0 months; HR 0.494, 95% CI 0.329-0.743; p 0.71) as an independent predictor of improved OS (P=0.0019) in CRT+ICI group. No significant survival differences were observed between induction and concurrent immunotherapy strategies. However, maintenance immunotherapy demonstrated longer OS compared to non-maintaining group (54.0 vs 39.0 months). Concurrent and maintenance immunotherapy achieved superior outcomes. Conclusions: ICIs combined with CRT significantly enhance the survival of patients with locally advanced esophageal cancer, with acceptable TRAEs. CD4+CD45RO+ T cells may serve as predictive markers of efficacy. Concurrent-maintenance immunotherapy may represent the optimal combination strategy.
Lin et al. (Sat,) studied this question.