To develop and validate a nomogram prediction model for local recurrence in early-stage gastrointestinal cancer patients after endoscopic submucosal dissection (ESD), providing a quantitative tool for postoperative recurrence risk assessment.

A retrospective analysis was conducted on 377 patients with early-stage gastrointestinal cancer who underwent ESD at Gaochun People's Hospital of Nanjing between January 2020 and December 2022. Patients were divided into recurrence (46 cases) and non-recurrence (331 cases) groups based on recurrence status within 12 months postoperatively. Univariate and multivariate logistic regression analyses were performed to identify independent risk factors. A nomogram prediction model was constructed and internally validated using the Bootstrap resampling method (1 000 iterations). The model's discrimination and calibration were evaluated using receiver operating characteristic (ROC) curve, calibration curve, and the Hosmer-Lemeshow test.

Univariate analysis revealed significant associations between recurrence and age ≥65 years, tumor size ≥2 cm, poor differentiation, submucosal invasion depth reaching the SM2 layer, lymphovascular invasion, perineural invasion, positive resection margin, abnormal preoperative carcino-embryonic antigen levels, and depressed tumor morphology (all P<0.05). Multivariate analysis ultimately identified positive resection margin (OR=13.135), tumor size ≥2 cm (OR=8.244), lymphovascular invasion (OR=7.802), poor differentiation (OR=6.556), abnormal preoperative carcino-embryonic antigen (OR=5.496), submucosal invasion depth reaching the SM1 or SM2 layer (OR=4.649 and 1.667, respectively), and depressed tumor morphology (OR=4.851) as independent risk factors. The nomogram achieved an area under the ROC curve (AUC) of 0.955 (95% CI: 0.928-0.982) with 89.1% sensitivity and 91.5% specificity. Bootstrap validation yielded a C-index of 0.932, and the calibration curve demonstrated good agreement between predicted and actual risks (P=0.261).

The nomogram model, incorporating tumor biological characteristics and surgical-pathological parameters, exhibits high predictive accuracy for local recurrence in early-stage gastrointestinal cancer patients after ESD. It significantly outperforms traditional single-indicator assessment systems and provides an objective quantitative tool for early identification of recurrence and personalized intervention. This model holds substantial clinical value for optimizing treatment decisions, reducing secondary surgery rates, and improving patient prognosis.