Development and validation of a dynamic nomogram for predicting central lymph node metastasis in papillary thyroid carcinoma patients based on clinical and ultrasound features.

Background: Prophylactic cervical lymph node dissection (CLND) for patients with papillary thyroid carcinoma (PTC) has long been a subject of controversy. To accurately perform preoperative staging and risk stratification of PTC patients, this study developed and validated a preoperative nomogram model for predicting central lymph node metastasis (CLNM) based on clinical and ultrasound features, thereby guiding surgical resection and postoperative adjuvant therapy. Methods: Patients with PTC (n=409), as confirmed by surgery and histopathology combined with CLND, were divided into training and validation groups. Clinical information, ultrasound features, American College of Radiology Thyroid Imaging Reporting and Data System (ACR TI-RADS) scores and Chinese version of the Thyroid Imaging Reporting and Data System (C TI-RADS) scores were collected. The features in the training group were selected by least absolute shrinkage and selection operator (LASSO) regression. These potential features were included in a multivariate logistic regression analysis to identify independent risk factors for CLNM and to develop a dynamic nomogram. In both the training and validation groups, the nomogram was evaluated for discrimination, calibration and clinical utility. Results: It was found that sex, age, multifocality, capsule contact, margin, micro-calcification, and ultrasound-based CLNM status were independent risk factors of CLNM, and a dynamic nomogram was used to develop a prediction model. The prediction model showed good discriminability, with an area under the receiver operating characteristic curve of 0.905 (95% confidence interval: 0.870-0.940) in the training group and 0.865 (95% confidence interval: 0.799-0.932) in the validation group. Based on the calibration curve and Hosmer-Lemeshow test, the prediction model was found to have good concordance in both the training group (P=0.6259) and validation group (P=0.1182). Decision curve analysis and clinical impact curve analysis demonstrated that the prediction model had good net clinical benefit. Conclusions: Dynamic nomograms developed using clinical and ultrasound characteristics can predict CLNM risk in PTC patients, thereby providing valuable support to clinicians for making personalized treatment decisions.