Video Experimental Relacionado
Updated: Jan 7, 2026
Comparison of Predictive Performance of Three Lymph Node Staging Systems in Colorectal Signet Ring Cell Carcinoma Based on Machine Learning Model
Published on: April 18, 2025
Factores de riesgo de metástasis de ganglios linfáticos cervicales laterales en el carcinoma papilar de tiroides y
Lei An1, Anran Du1, Jiayi Wang1
1Department of General Surgery, Beijing Friendship Hospital, Capital Medical University, Beijing, China.
Objective:
To identify risk factors for lateral lymph node metastasis (LLNM) in papillary thyroid carcinoma (PTC) and to establish clinical prediction models.
Methods:
We retrospectively collected clinical data from 249 patients with PTC and suspected LLNM, 222 patients met the inclusion criteria. Based on postoperative pathology of the lateral compartment, 145 patients without metastasis were classified as the non-metastasis group, 77 patients with metastasis were classified as the metastasis group. All included patients were randomly assigned to training set and validation set. Univariate and multivariate logistic regression analyses were performed to screen predictors of LLNM and construct nomogram models for preoperative and postoperative prediction. Model performance was evaluated using the Hosmer-Lemeshow goodness-of-fit test, calibration curves with bootstrap resampling, receiver operating characteristic (ROC) curves and the area under the curve (AUC), as well as decision curve analysis (DCA).
Results:
In preoperative analyses, age, maximum tumor diameter ≥1 cm on ultrasound, hyperechoic area in the lateral cervical lymph node, and lateral cervical lymph nodes perinodal vascularity were independent predictors of LLNM. In postoperative analyses, age, multifocality, pathological maximum tumor diameter ≥1 cm, and concomitant central lymph node metastasis were independent predictors. The AUCs for the preoperative model were 0.805 (training set) and 0.719 (validation set), and for the postoperative model were 0.885 (training set) and 0.762 (validation set). After 1,000 bootstrap resamples, the mean absolute errors (MAE) of the calibration curves were 0.047 and 0.066 for the preoperative model (training set and validation set), and 0.021 and 0.046 for the postoperative model.
Conclusion:
DCA showed a higher net clinical benefit of both models than the treat-all or treat-none strategies, indicating good predictive accuracy and clinical utility.
