Prediction of Patients With Anaplastic Thyroid Carcinoma With Bone Metastasis: A Population-Based Study

Background: Bone metastasis (BM) is a significant risk factor for the survival and prognosis of patients with anaplastic thyroid carcinoma (ATC). The aim of this study was to predict BM in patients with ATC. Methods: Demographic and clinicopathological data of patients with ATC were extracted from the Surveillance, Epidemiology, and End Results database between 2010 and 2020. Logistic regression (LR) was used to identify the linear influencing factors for BM. We developed prediction models for BM using six machine learning models: support vector machine (SVM), LR, adaptive boosting (AD), decision tree (DT), eXtreme Gradient Boosting (XGB), and random forest (RF). The area under the receiver operating characteristic curve (AUC) values, accuracy, recall rate, precision, F1 scores, calibration curves, and precision-recall curves were used to determine the best model and evaluate its effectiveness. The SHapley Additive exPlanations algorithm was used to reveal the interpretability of the prediction model. Results: This study included 781 patients with ATC, of whom 78 (9.99%) patients occurred BM and 703 (90.01%) patients were free of BM. The XGB model significantly outperformed the other models, with the highest F1 (0.897), accuracy (0.878), precision (0.924), recall (0.900), and AUC (0.897) values. The results of the LR model showed that age, gender, lung metastasis, and liver metastasis were linear influencing factors. According to XGB model, metropolitan area, median household income, N stage, and race were also strongly associated with BM among patients with ATC. Conclusion: We explored influencing factors for BM and established a prediction model based on XGB that yielded excellent results in predicting BM in patients with ATC. This study provides a theoretical basis for early decision making in clinical practice.