Prediction model construction of cervical central lymph node metastasis in papillary thyroid carcinoma combined with Hashimoto's thyroiditis utilizing conventional ultrasound and elastography.

Abstract

Background: When papillary thyroid carcinoma (PTC) is accompanied by Hashimoto's thyroiditis (HT), it is often challenging for preoperative ultrasound to distinguish between central lymph node enlargement caused by PTC metastasis and inflammatory reaction due to HT. However, central lymph node metastasis (CLNM) is closely associated with the risk of PTC recurrence after surgery. In this study, we developed a model to predict in patients with PTC combined with HT, based on conventional ultrasound characteristics and shear wave elastography (SWE) quantitative parameters of the primary lesion. We aimed to evaluate its predictive value to provide a useful reference for clinical decisions regarding central lymph node dissection.

Methods: This retrospective study included ultrasound data for 181 PTC patients with concurrent HT (totaling 215 nodules), confirmed by surgical pathology at our hospital and routinely undergoing central neck lymph node dissection. All enrolled PTC patients were randomly divided into training and test groups at a 7:3 ratio. Then, patients in each group were further segregated into two distinct cohorts: the CLNM group and the non-CLNM group as per the gold standard of pathology assessment. Subsequent statistical analysis of conventional ultrasound characteristics pertaining to primary foci alongside quantitative parameters derived from SWE, facilitated the identification of independent risk factors associated with CLNM. Then, a nomogram model was constructed, and its predictive value was evaluated. The test group was used for internal validation.

Results: Univariate analysis results in the training group indicated that nodule size, multiplicity, location, capsular invasion, and Emax were significantly associated with CLNM (all P<0.05). Multivariate analysis further identified nodule size, multiplicity, location, capsular invasion, and Emax as independent risk factors for CLNM (all P<0.05). Based on the multivariate analysis results, a nomogram model was developed to predict the occurrence of CLNM in PTC patients with HT. Receiver operating characteristic (ROC) curve analysis showed high predictive accuracy for CLNM, with an area under the ROC curve (AUC) of 0.837 in the training group and 0.882 in the test group. Calibration curves demonstrated good fit, closely aligning with the diagonal, indicating strong consistency in predicting CLNM.

Conclusions: The nomogram model, based on primary lesion ultrasound characteristics and SWE quantitative parameters in PTC patients with HT, may aid clinicians in preoperatively predicting the likelihood of CLNM in PTC patients.