Diagnostic performance of ACR-TIRADS for thyroid nodule risk stratification in pediatric patients

The American College of Radiology Thyroid Imaging Reporting and Data System (ACR-TIRADS) is a widely used ultrasonographic risk-stratification system for thyroid nodules in adults and its use has been increasingly expanding in the pediatric population in recent years. Here, we evaluated the diagnostic performance of ACR-TIRADS in patients aged ≤ 18 years with thyroid nodules.

We performed a single-center retrospective cohort study of patients aged ≤ 18 years, followed by tertiary care for thyroid nodules. The ACR-TIRADS data were extracted from the image records by two radiologists with expertise in thyroid imaging. Malignancy rates were defined based on cytological examinations, histological diagnosis, or ultrasonographic follow-up concerning nodule characteristics and size.

The cohort comprised 58 patients (65 nodules). The majority were female (70.7%), with a mean age of 14.0 ± 3.4 years and 27.5% had at least one risk factor for thyroid malignancy. The malignancy rate was 20.7% (N = 12). We could not reassess the ultrasound images of 9 patients; therefore, for this analysis, 49 patients (56 nodules) were included. The TIRADS nodule classifications were as follows: 8 TR1 (14.3%), 18 TR2 (32.1%), 15 TR3 (26.8%), 7 TR4 (12.5%), and 8 TR5 (14.3%). The ACR-TIRADS interobserver agreement was high, with a free marginal kappa of 0.86 [95% confidence interval (CI): 0.75, 0.97]. All TR1, TR2, and TR3 nodules were benign, and 8 cases of thyroid malignant neoplasm in the TR4 (N = 1) and TR5 (N = 7) groups resulted in malignancy rates of 14.3 and 87.5%, respectively. Remarkably, the TR5 nodules exhibited a positive predictive value of 87.5%, negative predictive value of 97.9%, sensitivity of 87.5%, and specificity of 97.9% for predicting malignancy. We did not identify a cutoff of nodule size for predicting malignancy - area under the receiver operating characteristic curve (AUC) of 0.58 (95% CI 0.38-0.80).

ACR-TIRADS effectively stratifies malignancy risk in pediatric thyroid nodules, with TR5 nodules showing particularly high malignancy risk. Clinical risk factors combined with ultrasound characteristics provide better malignancy prediction than nodule size alone in this population.