Development and Validation of a Prediction Model for Lymph Node Metastasis Based on Molecular Classification in Endometrial Carcinoma.

Qiuyue Han,Quanhong Jiang,Jiaqi Xu,Yuan Zhang,Li,Beihua Kong,Kun Song,Hualei Bu,Ruifen Dong

DOI: https://doi.org/10.1200/jco.2024.42.16_suppl.e17599

IF: 45.3

2024-01-01

Journal of Clinical Oncology

Abstract:e17599 Background: The aim of this study is to develop a robust preoperative model using molecular classification to accurately predict the likelihood of occult lymph node metastasis (LNM) in patients diagnosed with early-stage endometrial cancer (EC). Such a model could help guide surgeons in choosing between systematic lymph node dissection and sentinel lymph node biopsy. Methods: Patients with clinically early-stage EC were included from Qilu Hospital of Shandong University between January 2010 and December 2017. Univariate and multivariate analyses was performed to identify the risk factors associated with occult LNM. Different machine learning (ML) algorithm models were applied to construct several prediction models, based on factors screened by the least absolute shrinkage and selection operator (LASSO). These models were evaluated by the receiver operating characteristic (ROC) derived area under the curve (AUC) values and the optimal model was established to build the nomogram. Bootstrap was used to validate of this model. The AUC, decision-curve analysis, clinical impact curves and interpretability metrics were used to determine the clinical net benefit of the model. Furthermore, risk stratification of EC patients was created based on the nomogram's threshold values. Results: Out of 363 patients with early-stage EC, 35 (9.1%) presented with LNM. Serum carbohydrate antigen 125 (CA125) and depth of myometrial invasion were independent predictive factors of LNM. An optimal ML model based on logistic regression (AUC: 0.744) was established to predict the occult LNM, including five factors screened by LASSO and they are CA125, neutrophil-to-lymphocyte ratio (NLR), histological subtype, depth of myometrial invasion and molecular classification. This model displayed good discrimination ability (AUC: 0.843) and calibration ability. Conclusions: This new preoperative LNM prediction model based on molecular classification can precisely indicate the probability of LNM in patients with early-stage EC, which could provide recommendations for the treatment of systematic lymph node dissection and sentinel lymph node biopsy based on risk stratification.