Prognostic impact of treatment modalities and examined lymph nodes count on survival in T3-4 esophageal cancer: a retrospective cohort and predictive modeling study from the Surveillance, Epidemiology, and End Results database.

Esophageal carcinoma (EC) is a malignancy of significant global burden, characterized by high incidence and mortality rates. The majority of patients are diagnosed with advanced disease. While surgery-based multimodality therapy is standard for resectable disease at this stage, the optimal strategy requires further definition. Concurrently, the impact of the number of examined lymph nodes (ELNs) on patient prognosis across different treatment modalities remains to be further explored. This study aimed to investigate the prognostic impacts of different treatment modalities and the corresponding number of ELNs in patients with T3-4 EC. This retrospective cohort study analyzed patients with pT3-4 EC from the Surveillance, Epidemiology, and End Results (SEER) database who underwent radical esophagectomy. Participants were categorized into four treatment groups: neoadjuvant therapy plus surgery (NS), adjuvant therapy plus surgery (AT), surgery alone (SA), and perioperative therapy (PT). Overall survival (OS) was compared using Kaplan-Meier analysis. Propensity score matching (PSM) was applied to balance baseline characteristics across treatment groups and across subgroups stratified by the number of ELNs. For prognostic modeling, patients were randomly divided into training and validation sets (7:3 ratio). The variables ultimately used to construct the nomogram were progressively refined through univariate Cox analysis, least absolute shrinkage and selection operator (LASSO) regression, and multivariable Cox analysis. Model performance was assessed using the concordance index (C-index), receiver operating characteristic (ROC) curves, calibration plots, and decision curve analysis (DCA). A total of 4,485 eligible patients were included in the study. With a median follow-up of 24 months, 2,796 deaths were recorded. After PSM, OS was significantly better in the NS and PT groups than in the SA and AT groups (P<0.05), while no significant difference was found between NS and PT (P=0.21). The AT group also showed superior OS compared to SA (P<0.001). ELNs demonstrated treatment-specific prognostic value, with optimal thresholds identified at 9 for NS, 16 for AT, 17 for SA, and 19 for PT. Exceeding these thresholds was associated with significantly improved OS, and this association remained robust in sensitivity analyses. A nomogram was developed using independent prognostic factors from multivariate Cox analysis, including age, node (N) stage, metastasis (M) stage, ELNs, and treatment modality. The model showed good discrimination, with a C-index of 0.608 in the training set and 0.616 in the validation set. The areas under the curves (AUCs) for predicting 1-, 3-, and 5-year OS were 0.645, 0.620, and 0.618 in the training set, and 0.658, 0.619, and 0.631 in the validation set. Calibration plots and DCA confirmed good model fit and clinical utility. This study confirms that NS provides significant survival benefits in advanced EC. For NS-ineligible patients, AT is superior to SA. The proposed treatment specific ELNs thresholds offer an objective surgical quality benchmark, enabling balanced therapeutic risk assessment. The integrated prognostic model supports personalized EC management.