Identification of recurrence risk factors and development of personalized follow-up strategies for esophageal squamous cell carcinoma patients after neoadjuvant therapy and surgery: a retrospective cohort study.

Background: Recurrence after neoadjuvant therapy and R0 resection remains common in esophageal squamous cell carcinoma (ESCC), yet validated baseline models for predicting recurrence using routinely available clinical and inflammatory/nutritional markers are limited, and how recurrence risk evolves with additional recurrence-free survival time has rarely been described in this setting. This study aimed to identify high-risk factors for recurrence in ESCC and to construct a prognostic prediction model. We also aimed to evaluate the dynamic survival and recurrence risks of ESCC and, as a complementary descriptive analysis, to formulate a hypothesis-generating personalized follow-up framework using conditional survival (CS). Methods: We retrospectively analyzed 408 ESCC patients who underwent neoadjuvant therapy followed by R0 resection at Fujian Cancer Hospital. Candidate predictors were screened by univariate logistic regression (P<0.05) and entered into multivariate logistic regression to identify independent recurrence risk factors and construct a nomogram; performance was assessed within the derivation cohort using receiver operating characteristic (ROC) curve, calibration curve and decision-curve analysis. Kaplan-Meier analysis compared overall survival (OS), progression-free survival (PFS), local-regional recurrence-free survival (LRRFS) and distant metastasis-free survival (DMFS) across nomogram-based risk groups, and CS was used to describe how recurrence-free probabilities changed with additional survival time. Results: During a median follow-up of 28 months, recurrence occurred in 156 patients (38.2%). Multivariate Cox analysis identified age at diagnosis, tumor length, tumor thickness, N stage, and lymphocyte-to-monocyte ratio (LMR) as independent risk factors for recurrence. The time-dependent area under the curve (AUC) was 0.723. Calibration curves showed good agreement. Decision curve analysis (DCA) confirmed net clinical benefit. Patients were stratified into low-, moderate-, and high-risk groups with significant survival differences (all P<0.001). CS analysis revealed increasing recurrence-free probability over time, most pronounced in high-risk patients surviving beyond 2 years. We propose a risk-adapted follow-up framework as a hypothesis for future prospective testing. Conclusions: This study established a clinically applicable risk stratification system for post-surgical ESCC patients, integrating baseline risk factors with dynamic survival analysis. The model has not been externally validated, and the proposed follow-up framework is hypothesis-generating and requires prospective validation before clinical use. This work provides a foundation for developing personalized surveillance protocols that could optimize resource utilization and potentially improve early recurrence detection.