Diagnostic model incorporating clinicopathological characteristics of Delphian lymph node metastasis risk profiles in papillary thyroid cancer

Background Delphian lymph node (DLN), also known as prelaryngeal node, is the component of central lymph node (Level VI). DLN has been well studied in laryngeal cancer while less in papillary thyroid cancer (PTC). However, increasing attention has been payed to its clinical value in PTC lymph node metastasis. The aim of this study was to assess the clinicopathologic risk factors for DLN metastasis and to calibrate an informative diagnosis model for predicting risk stratification in DLN metastasis. Methods Retrospectively analysis 936 PTC patients who underwent thyroidectomy by a single surgeon, in the department of Head and Neck Surgery of Tianjin Cancer Hospital from 2017 to 2019. Moreover, 250 PTC patients underwent thyroidectomy by another surgeon at Tianjin Cancer Hospital between January 2019 and April 2019 were used as a validation cohort. Univariate analysis and multivariable logistic regression were used to identify clinicopathologic risk factors associated with DLN metastasis. A diagnostic model was generated using logistic regression, and internal validation and external validation were also performed. Predictive performance was assessed using area under the receiver operating characteristic curve (AUROC) and calibration plot. A nomogram plot was drawn to predict individual risk based on the diagnosis prediction model. Results In 936 cases, 581 (62.1%) show DLN, and 177 (177/581,30.5%) were verified metastasis. DLN metastasis was significantly associated with gender, age, tumor size, bilateral cancer, multifocality, extrathyroidal extension, lymphovascular invasion and central and lateral neck lymph node metastasis. Multivariate analysis revealed that age, gender, tumor size, extrathyroid extension, lymphovascular invasion and central lymph node metastasis were independent risk factor of DLN metastasis. The nomogram included age, gender, tumor size, extrathyroid extension, lymphovascular invasion and central lymph node metastasis, and all these features had a specificity of 100%. Particularly, tumor size was the best single predictor. The diagnostic model had an area under the curve (AUC) of 0.829 (95% confidence interval, 0.804–0.854). Internal and external validation of nomogram was 0.819 and 0.745, respectively. Conclusions DLN metastasis appears to be an important parameter that predict metastatic disease of the central compartments. To assess DLN metastasis in more precise criterions, has great clinical significance for PTC treatment selection.