XGBoost-based nomogram for predicting lymph node metastasis in endometrial carcinoma

This study aims to construct and optimize risk prediction models for lymph node metastasis (LNM) in endometrial carcinoma (EC) patients, thus improving the identification of patients at high risk of LNM and further providing accurate support for clinical decision-making. This retrospective analysis included 541 cases of EC treated at The First Affiliated Hospital, Jinan University between January 2017 and January 2022. Various clinical and pathological variables were incorporated, including age, body mass index (BMI), pathological grading, myometrial invasion, lymphovascular space invasion (LVSI), estrogen receptor (ER) and progesterone receptor (PR) levels, and tumor size. Multivariate Logistic regression analysis was used to identify independent risk factors for LNM. Subsequently, the Least Absolute Shrinkage and Selection Operator (LASSO), Extreme Gradient Boosting (XGBoost), RandomForest, and Support Vector Machine (SVM), all machine-learning algorithms, were adopted to select features and build models. The XGBoost model gave the best performance among all models, with areas under the curve (AUCs) of 0.876 and 0.832 for training and validation sets, respectively, suggesting its high discriminatory ability and prediction accuracy. Moreover, the calibration curve analysis further verified the consistency of the model-predicted values with the actual results, indicating the model's good applicability at various risk levels. According to the decision curve analysis, the XGBoost model showed high net benefits within most risk-threshold ranges, indicating its substantial practical value in clinical applications. Conclusively, this study successfully builds machine-learning models based on multiple clinical and pathological features, which can effectively predict the LNM risk in EC patients. The model is expected to provide important references for clinicians in surgical decision-making and the formulation of individualized treatment plans, thereby enhancing patient outcomes.