Related Experiment Video
Updated: May 29, 2026
Computer-Aided Three-Dimensional Visualization in the Treatment of Locally Advanced Thyroid Cancer
Published on: June 9, 2023
Building a predictive model and survival analysis for distant metastases in differentiated thyroid carcinoma: A
Jiuping Huang1, Peipei Yang2, Zhendong Wang3
1Department of Ultrasound, Peking University Third Hospital, Beijing, China.
Background:
Distant metastases (DM) in differentiated thyroid carcinoma (DTC) are uncommon but markedly worsen patient prognosis. Existing risk prediction models frequently show limited accuracy because key clinicopathological predictors are not fully integrated. This study aimed to identify risk factors for DM and to develop a decision tree-based predictive model.
Materials And Methods:
We retrospectively reviewed the medical records of 15,591 patients with DTC who underwent initial surgery at three tertiary centers between 2000 and 2018. Thirty-seven patients with DM and complete clinicopathological data constituted Cohort 1 (training set), whereas 14 comparable cases formed Cohort 2 (validation set). A 4:1 ratio-matched control group was generated by random sampling of patients with postoperative no evidence of disease (NED), matched for age, sex, and year of surgery.
Results:
Univariate and multivariate analyses demonstrated significant differences (P < 0.001) between the DM and NED cohorts in median age, extrathyroidal extension (ETE), AJCC stage, tumor location, histological subtype, and primary tumor diameter. Decision tree and random forest analyses identified AJCC stage and tumor diameter as the most influential predictors of DM. A predictive model incorporating these variables achieved perfect classification accuracy, which was confirmed through external validation. In cases with metachronous metastases, histological subtype and ETE independently predicted survival outcomes.
Conclusion:
This study statistically optimized the weighting of risk factors for DM prediction in DTC, emphasizing AJCC stage and tumor diameter as dominant determinants. The resulting model demonstrated high accuracy and may support clinical decision-making for personalized patient management as a risk stratification tool.
