Survival outcomes of neoadjuvant chemotherapy versus chemoradiotherapy in locally advanced esophageal cancer: a SEER-based propensity score analysis

Abstract This study aimed to compare postoperative overall survival (OS) and cancer-specific survival (CSS) between neoadjuvant chemotherapy (nCT) and neoadjuvant chemoradiotherapy (nCRT) in patients with locally advanced esophageal cancer using a real-world cohort. This retrospective study enrolled patients with primary esophageal adenocarcinoma or squamous cell carcinoma staged cT1N+M0 or cT2–4aNanyM0 from 2000 to 2022. The cohort was stratified by histology into esophageal adenocarcinoma (EAC, primary analysis) and esophageal squamous cell carcinoma (ESCC, exploratory analysis). Within each population, dedicated propensity scores were derived from demographic and clinical variables including age, sex, race, marital status, income, urban/rural status, T stage, N stage, tumor grade, primary tumor site, tumor size, and year of diagnosis. IPTW was applied with stabilized weights truncated at the 1st–99th percentiles. IPTW-weighted Kaplan–Meier and doubly robust multivariable Cox regression estimated treatment effects. A total of 2,775 patients were included (nCT = 189; nCRT = 2,586). The EAC cohort comprised 2,273 patients (nCT = 168; nCRT = 2,105) and the ESCC cohort comprised 502 patients (nCT = 21; nCRT = 481). Post-weighting covariates were well balanced (maximum |SMD| ≤ 0.04 in the EAC cohort). In the EAC cohort, IPTW-weighted multivariable Cox models showed no statistically significant association between nCRT and OS (HR = 1.17, 95% CI 0.95–1.46, P = 0.145) or CSS (HR = 1.26, 95% CI 0.96–1.65, P = 0.090). Point estimates trended toward a higher hazard with nCRT but did not reach statistical significance. IPTW-weighted Kaplan–Meier OS comparison was non-significant (log-rank P = 0.29), consistent with the Cox regression findings. The exploratory ESCC analysis (OS HR = 1.38, P = 0.293; CSS HR = 1.20, P = 0.650) and a pooled sensitivity analysis yielded concordant results. Subgroup analyses within the EAC cohort revealed no statistically significant interaction between treatment and age (interaction P = 0.918 for OS) or N stage (interaction P = 0.701 for OS), indicating that the treatment effect did not differ significantly across these subgroups. In this large real-world cohort, adding radiotherapy to neoadjuvant chemotherapy was not associated with improved survival in either EAC or ESCC populations. These findings are hypothesis-generating and require validation in prospective studies.