Prediction Model of Cervical Lymph Node Metastasis in Papillary Thyroid Carcinoma.

Abstract

Background: The objective of this study is to develop a predictive model for the assessment of cervical lymph node metastasis risk in papillary thyroid carcinoma (PTC).

Methods: A retrospective study was conducted on 212 patients with PTC who underwent initial surgical treatment from August 2022 to April 2023 in 2 hospitals. Data were randomly split into 7:3 training-validation sets. Logistic regression was used for feature selection and predictive model creation. Model performance was assessed using receiver operating characteristic (ROC) and calibration curves. Clinical utility was determined using decision curves.

Results: Among the 212 patients with PTC, 104 cases (49.1%) exhibited cervical lymph node metastasis, while 108 cases (50.9%) did not. Multivariate logistic regression analysis revealed that age (OR = 0.95), FT3 (OR = 0.41), tumor maximum diameter ≥0.9 cm (OR = 1.85), intratumoral microcalcifications (OR = 1.84), and suspicious lymph node on ultrasound (OR = 2.96) were independent risk factors for lymph node metastasis in PTC patients (P < 0.05). The constructed model for predicting the risk of cervical lymph node metastasis demonstrated a training set ROC curve area under the curve (AUC) of 0.742 (95% CI: 0.664 - 0.821), with a cut-off value of 0.615, specificity of 87.8%, and sensitivity of 51.4%. The validation set exhibited an AUC of 0.648 (95% CI: 0.501 - 0.788), with a cut-off value of 0.644, specificity of 91.2%, and sensitivity of 43.3%. Including the BRAF V600 E mutation did not improve the model's diagnostic performance significantly. Decision curve analysis indicated clinical feasibility of the model.

Conclusion: The predictive model developed in this study effectively predicts lymph node metastasis risk in PTC patients by incorporating ultrasound features, demographic characteristics, and serum parameters. However, including the BRAF V600 E mutation does not significantly improve the model's diagnostic performance.