Comparison of neoadjuvant treatment strategies in resectable locally advanced esophageal squamous cell carcinoma: A real-world study
Abstract
The optimal neoadjuvant regimen for resectable locally advanced esophageal squamous cell carcinoma (LA-ESCC) remains uncertain, and pretreatment indicators are lacking. To evaluate and compare the efficacy and safety of three major neoadjuvant strategies in current clinical practice for patients with resectable LA-ESCC. Retrospective real-world cohort study. We compared the outcomes of patients who received neoadjuvant chemoradiotherapy (nCRT), immunochemotherapy (nICT), or immunotherapy combined with chemoradiotherapy (nICRT), followed by surgery. The primary endpoint was pathological complete response (pCR); the secondary endpoints included major pathological response (MPR), progression-free survival (PFS), overall survival (OS), and safety. Generalized overlap weighting based on multinomial propensity scores was used to balance measured pretreatment characteristics. Exploratory proteomic analysis of the pretreatment biopsy specimens was also performed to identify response-associated proteins. After overlap weighting, the pCR rates differed among the regimens (nCRT, 44.9%; nICT, 19.6%; nICRT, 35.9%; p < 0.001), with a similar pattern for MPR (74.3%, 38.6%, and 67.7%, respectively; P < 0.001). At a median follow-up of 28.63 months, PFS and OS did not differ significantly. Grade 3–4 hematologic toxicity and treatment-related pneumonitis differed among the three groups, whereas overall postoperative complication rates were comparable after overlap weighting. No individual protein remained significant after false discovery rate correction. TMED10 was nominally prioritized as a shared response-associated candidate. In this retrospective real-world cohort, pCR and toxicity profiles differed among the three neoadjuvant strategies, whereas no statistically significant differences in PFS or OS were detected after overlap weighting. These findings represent observational associations and do not establish causal superiority. The TMED10 finding is hypothesis-generating and requires independent validation.