Development and validation of a multivariable prediction model for lateral lymph node metastasis in papillary thyroid carcinoma.

Abstract

Background: Lateral lymph node metastasis (LLNM) is a critical determinant of surgical extent and a significant risk factor for locoregional recurrence and decreased survival in papillary thyroid carcinoma (PTC). However, accurate preoperative prediction of LLNM remains challenging due to the limited sensitivity of conventional ultrasound, which often leads to either unnecessary prophylactic lateral neck dissection or inadequate surgical management. This study aimed to develop and validate a nomogram incorporating preoperative clinical and ultrasound features to improve individualized LLNM risk prediction.

Methods: In this multicenter retrospective study, we included 1,195 consecutive PTC patients who underwent thyroid surgery at three centers (Beijing Tongren Hospital, Beijing Shijitan Hospital, and Qinhuangdao First Hospital) from January 2017 to June 2021.All patients had histopathologically confirmed PTC, and the reference standard for LLNM diagnosis was histopathological examination of dissected lateral neck lymph nodes. Patients with incomplete clinical or pathological data, a history of prior thyroid or neck surgery, other thyroid cancer subtypes, other malignancies, or preoperative thyroid function-related medication were excluded. Of the total patients, there were 1,080 were randomly split into a training cohort (n=756) and an internal validation cohort (n=324). An independent cohort of 115 patients was used for external validation. Candidate predictors including clinicodemographic and ultrasound features were screened via univariate analysis using Chi-squared or Wilcoxon rank-sum tests. Variables with P<0.05 in univariate analysis were entered into a multivariate logistic regression with backward stepwise selection to identify independent factors. Model performance was evaluated using area under the receiver operating characteristic curve (AUC), calibration curves, and decision curve analysis (DCA).

Results: The independent risk factors for LLNM were Hashimoto's thyroiditis (HT), tumor location (upper pole), size, multifocality, capsular invasion, extrathyroidal extension (ETE), and ≥5 central lymph node metastases (CLNM). The model achieved AUCs of 0.88, 0.89, and 0.82 in the training, internal validation, and external validation cohorts, respectively. The model demonstrated sensitivity of 84.2%, 81.6%, and 73.1%, and specificity of 90.1%, 88.9%, and 84.3% in the training, internal validation, and external validation cohorts, respectively. Calibration was good, and DCA showed net clinical benefit across a wide range of risk thresholds.

Conclusions: This nomogram, integrating readily available preoperative clinical and ultrasound features, demonstrates promising accuracy for stratifying LLNM risk in PTC patients. However, given the limited sample size of the external validation cohort (n=115; 26 events), these results should be interpreted with caution. Further large-scale, multicenter validation is warranted before routine clinical application. The model may serve as a useful adjunct in preoperative decision-making and personalized management.