Development and optimisation strategies for a nomogram-based predictive model of malignancy risk in thyroid nodules

This study aimed to develop and validate a clinical prediction model to assist radiologists in optimising the diagnostic classification of the Chinese Thyroid Imaging Reporting and Data System (C-TIRADS).

A total of 1659 patients from two hospitals were included in this study. The derivation cohort comprised 909 patients for model development and internal validation, while 750 patients formed the external validation cohort. A binary logistic regression model was constructed. Model performance in the derivation set was evaluated using receiver operating characteristic (ROC) curves and visualised with a nomogram. In the external validation set, ROC and calibration curves were used to assess discrimination and calibration.

The original C-TIRADS category, abnormal cervical lymph node sonographic findings, and changes in thyroid nodule size emerged as significant predictors of C-TIRADS optimisation. The optimised nomogram demonstrated an area under the ROC curve (AUC) of 0.730 (95% confidence interval=0.697-0.762), with a sensitivity of 63.2%, specificity of 74.9%, and overall accuracy of 67.7% for predicting optimisation. Using probability thresholds of ≥60% to recommend an upgrade and <30% to recommend a downgrade, the calibration curve showed good agreement, and decision curve analysis demonstrated a favourable net clinical benefit. External validation confirmed excellent discrimination (AUC=0.865; 95% confidence interval=0.839-0.891).

An optimised C-TIRADS model that integrates imaging features of thyroid nodules with clinical risk factors may aid radiologists in improving the diagnostic efficiency and clinical utility of the TIRADS classification.