Development and validation of a prediction model for lymph node metastasis based on molecular typing in clinically early-stage endometrial carcinoma.

Objective: To develop and externally validate a machine learning-based preoperative model integrating molecular typing and clinical features to predict lymph node metastasis (LNM) in patients with early-stage endometrial carcinoma (EC).

Methods: This retrospective study included 465 patients with clinically early-stage EC treated at Qilu Hospital of Shandong University. Tumors were classified into molecular subtypes using The Cancer Genome Atlas-based methods. Least Absolute Shrinkage and Selection Operator regression identified five preoperative predictors: molecular typing (CN-H vs. non-CN-H), histological subtype, depth of myometrial invasion, neutrophil-to-lymphocyte ratio, and CA125 levels. Multiple machine learning algorithms were evaluated, and logistic regression (LR) was selected based on optimal discrimination and clinical applicability. Model performance was assessed using area under the curve (AUC), calibration plots, and decision curve analysis (DCA). A web-based nomogram was developed for clinical use.

Results: The LR model demonstrated excellent discrimination, with AUCs of 0.843 in the training cohort and 0.809 in the testing cohort. The CN-H subtype was significantly associated with increased LNM risk. The model enabled effective risk stratification and calibration curves and DCA confirmed the model's accuracy and clinical utility.

Conclusion: By integrating molecular and preoperative clinical features, this model offers accurate LNM risk stratification for early-stage EC. It supports clinical decision-making and has been implemented as a user-friendly online tool. Further prospective multicenter validation is warranted.