The value of the nomogram model based on acr ti-rads: prediction of the central cervical lymph node metastasis in thyroid papillary carcinoma

Objectives To explore the value of nomogram model (ATR model) based on American Society of Radiology thyroid imaging reporting and data system (ACR TI-RADS), computed tomography (CT), and cervical ultrasound (C-US) in predicting central cervical lymph node metastasis (CLNM) of thyroid papillary carcinoma (PTC). Materials 366 PTC nodules were used as the modeling cohort and 121 PTC nodules were included as the external validation cohort. According to the 7:3 ratio of the modeling cohort, the training set and test set were randomly generated. The ATR model was constructed in the training set, and the discrimination and consistency of the model were evaluated. In the external validation cohort, the diagnostic efficacy of the ATR model, CT, C-US, and C-US combined CT (CT+US) were compared and decision curve analysis (DCA) and clinical impact curve (CIC) were used to explore the clinical value. Results ATR model: Y= -3.548 + 0.923 × Gender + 1.065 × Multifocality + 0.092 × Maximum diameters + 0.235 × ACR TI-RADS score. In the training set, test set, and external validation cohort, the C-indexes (0.756, 0.729, 0.753) showed the ATR model has good discrimination. Hosmer-Lemeshow goodness of fit test (P=0.837, P=0.619, P=0.389) and the calibration curve results showed that the model has a good consistency. In the external validation cohort, the sensitivity of ATR model was significantly higher than that of CT, C-US and CT + US (66.1% vs. 9.7%, P<0.05; 66.1% vs. 16.1%, P<0.05; 66.1% vs. 24.2%, P<0.05), and the accuracy was better than that of CT and C-US (66.1% vs. 52.9%, P<0.05; 66.1% vs. 55.4%, P<0.05), The specificity was lower than that of the three (72.9% vs. 98.3%, P<0.05; 72.9% vs. 96.6%, P<0.05; 72.9% vs. 94.9%, P<0.05). DCA showed that the ATR model had the largest risk threshold range (0.3-0.8) and the highest net benefit (0.32). CIC showed that the ATR model had excellent accuracy when the predicted risk threshold was greater than 0.3. Conclusions The ATR model has good diagnostic performance and clinical value in predicting CLNM.