The value of ultrasonographic scoring method and nomogram in assessing cervical lymph node metastasis of papillary thyroid carcinoma

Abstract

Background: The preoperative cervical lymph node metastasis (CLNM) status in patients with papillary thyroid carcinoma (PTC) critically determines the type of lymph node dissection performed. Currently, ultrasonography is the primary method for initial CLNM screening in PTC patients. This study aims to analyze the efficacy of an ultrasonic-characteristics-based scoring system in diagnosing lymph node metastasis in PTC and construct a predictive nomogram.

Methods: The imaging findings, fine-needle aspiration (FNA) results, and surgical pathology data from 269 suspected CLNM cases at Peking University Shenzhen Hospital, spanning from July 2021 to October 2022, were retrospectively analyzed. We identified specific ultrasound characteristics and assigned scores based on our clinical experience. The diagnostic performance of the ultrasound scoring system was assessed by plotting receiver operating characteristic (ROC) curves and calculating the area under the curve (AUC). Additionally, a nomogram was developed using least absolute shrinkage and selection operator (LASSO)-logistic regression. The nomogram's discrimination was evaluated using ROC analysis, its accuracy was assessed with calibration curves, and its clinical utility was determined by decision curve analysis (DCA).

Results: In this study, factors such as age, sex, lymph node length, thickness, aspect ratio, shape, hilum status, echogenicity, microcalcification, cystic necrosis, blood flow pattern, and the ultrasonic score were included in the analysis. The ultrasound score had the highest (AUC = 0.914, 95% confidence interval [CI]: 0.880-0.950), with an optimal cutoff value of 2.5. A score of 3 or higher had a diagnostic sensitivity for CLNM of 81.1%, specificity of 85.2%, positive predictive value (PPV) of 83.1%, negative predictive value (NPV) of 83.4%, and Kappa value of 0.664. Subsequent LASSO regression analysis identified sex, hyperechogenicity, peripheral disordered blood flow, and the ultrasonic score as independent predictors of CLNM, which were incorporated into a logistic regression-based predictive nomogram. The model exhibited strong discriminatory performance in both the training set (AUC = 0.933, 95% CI: 0.820-0.910) and the test set (AUC = 0.958, 95% CI: 0.790-0.890) for distinguishing PTC with and without CLNM. Furthermore, calibration curves and decision curve analysis (DCA) confirmed the model's good fit and favorable clinical net benefit.

Conclusion: The ultrasonic scoring method and the Nomogram have significant clinical utility in the preoperative assessment of CLNM in PTC, reducing unnecessary FNA procedures, and are simple and practical for clinical application.

Clinical trial number: Not applicable.

Keywords: Cervical lymph node metastasis(CLNM); Nomogram; Papillary thyroid carcinoma(PTC); Ultrasonography.

Fig. 1

The patient selection flowchart

Fig. 2

ROC curves for risk factors

Fig. 3

A 40-year-old woman presented with a suspicious lymph node in the neck. Ultrasound scoring of the lymph node: irregular shape (1 point), aspect.ratio = 1.2 (< 2, 1 point), hilum poorly displayed(1 point), non-homogeneous slightly higher echo (1 point, A), localized cystic necrosis (2 points, B), CDFI revealed disordered blood flow signals within and around the lymph node (1 point, C), yielding a total score of 7 points. Both the patient’s FNAB and post-operative pathology confirmed lymph node metastasis of thyroid papillary carcinoma

Fig. 4

Coefficient path diagram: the regression coefficients gradually contract as λ increases. Eventually only a few coefficients remain non-zero and the coefficients of the remaining variables are compressed to zero

Fig. 5

Cross validation curve: LASSO regression showed the minimum λ value on the right side was chosen, and 4 variables were selected for further logistic regression analysis

Fig. 6

Nomogram for predicting CLNM in PTC

Fig. 7

ROC curve for training set

Fig. 8

ROC curve for test set

Fig. 9

Calibration curve of the nomogram for CLNM in PTC; The x-axis shows the predicted risk for the nomogram and the y-axis represents the actual CLNM risk. “Apparent” indicates the performance of the model on the training set,“bias-corrected” indicates the performance of the model after bias correction, and the third bar indicates the case of perfect calibration.The “bias-corrected” in this figure is closer to “Ideal”, indicating a better calibration of the nomogram

Fig. 10

DCA for CLNM in PTC. The horizontal coordinate shows the threshold probability and the vertical coordinate indicates the net benefit. The red solid line shows the benefit of the plotted nomogram and the blue solid line shows the diagnostic benefit of the ultrasound score.Within most thresholds(0.05 to 0.90), decision making using nomogram and ultrasound score is more beneficial than a strategy that assumes everyone receives the intervention or no one receives the intervention

Fig. 11

A 28-year-old male presented with a suspicious lymph node in the neck. The node exhibited a hyperechoic mass (2 points) (A). CDFI showed abundant and disordered blood flow signals within the hyperechoic area (1 point) (B). Additionally, the lymph node demonstrated regular shape (0 points), an aspect ratio > 2 (0 points), a normal lymph node hilum structure (0 points), and no microcalcifications (0 points) or cystic necrosis (0 points) observed, resulting in a total score of 3 points. According to the nomogram, the probability of CLNM was predicted to be > 95% (C). The patient’s final pathological diagnosis confirmed metastatic papillary thyroid carcinoma in the lymph node