Building a predictive model for the lymphatic skip metastasis of papillary thyroid carcinoma based on machine learning

WANG Shuai; YU Yaocheng; YU Zhiwei; WANG Ting

doi:10.3969/j.issn.1674-4136.2024.04.003

Abstract：Objective To explore the incidence and risk factors of skip metastasis in papillary thyroid carcinoma and establish a predictive model using machine learning algorithms.Methods A retrospective analysis was conducted on the clinical and pathological data of 730 patients with papillary thyroid carcinoma admitted to the First Affiliated Hospital of Air Force Medical University from June 2013 to December 2021.The patients were divided into a positive group and a negative group based on the presence of skip metastasis.Univariate and multivariate analyses were used to determine the independent factors associated with skip metastasis.A predictive model was built using six machine learning algorithms including Logistic regression,k-nearest neighbors,decision tree,random forest,extreme gradient boosting,and support vector machine.Results The incidence of skip metastasis in papillary thyroid carcinoma was 11.51%(84/730).Multivariate analysis indicated that age＞45 years,microcarcinoma,tumor involvement of the upper pole,extrathyroidal extension,number of central compartment lymph nodes,and number of lateral compartment lymph node metastases were independent factors associated with skip metastasis.The receiver operating characteristic curve showed that the k-nearest neighbor algorithm had the best predictive performance,with an area under the curve of 0.813 in the training set and 0.808 in the validation set.Conclusions Skip metastasis in papillary thyroid carcinoma is associated with age,papillary thyroid microcarcinoma,tumor involvement of the upper pole,extrathyroidal extension,number of central compartment lymph nodes,and number of lateral compartment lymph node metastases.The independent-factor-based machine learning algorithm model has high predictive performance and can guide clinical decision-making.