The Diagnostic Performance of the American Thyroid Association (ATA) and American College of Radiology Thyroid Imaging and Reporting Data System (ACR TI-RADS) Risk Stratification Systems for Thyroid Nodules in Pediatric Patients

Background: The 2015 ATA and the 2017 ACR TI-RADS ultrasound-based malignancy risk stratification systems are the most widely utilized systems used in the United States to help guide the decision to perform biopsy. Both were developed for adult patients. Since thyroid nodules in children carry a greater risk of malignancy, the criteria for fine-needle aspiration biopsy (FNAB) may differ, thus calling into question the validity of these adult-based systems for the evaluation of nodules in children and adolescents. The purpose of our study is to assess and compare the diagnostic performance of the ATA and TI-RADS in a pediatric population.

Methods: We performed a single institution retrospective chart review of patients with thyroid nodules between April 2008 and June 2020. Ultrasound images were subjected to blinded review by both a pediatric radiologist and endocrine surgeon, who retrospectively assigned a TI-RADS and ATA category to each nodule. FNAB or surgical specimen pathology result was used as a reference standard for sampled nodules and excised glands, while for those followed clinically, the status of the nodule was determined with two years of clinical and radiologic follow-up. Test characteristics were calculated with a confidence interval of 95%. The receiver operating characteristic curves were generated and area under the curve calculated using a logistic regression computed by the statistical software Stata.

Results: Fifty-four thyroid nodules from 45 patients were studied; 41/54 (75.9%) were cytologically or histologically benign, while 13/54 (24.1%) were malignant. Using the decision to biopsy a nodule in accordance with ATA and TI-RADS size-based recommendations as a positive test for malignancy, the overall sensitivity was 100% for ATA and 84.6% for TI-RADS; the overall specificity was 26.8% for ATA and 73.2% for TI-RADS (Table 2a). Alternatively, using the two highest risk levels as a positive test for malignancy, the overall sensitivity was 76.9% for ATA and 100% for TI-RADS; the overall specificity was 48.8% for ATA and 51.2% for TI-RADS (Table 2b). The AUC of the ROC curves were 76.8% for ATA and 87.5% for TI-RADS.

Conclusion: Our initial results suggest that 100% sensitivity for malignancy is achieved either by using the two highest risk categories of TI-RADS (TR4-TR5) or the ATA adult-based size criteria for FNAB. However, the specificity of the former approach is higher at 51.2% versus 26.8%, suggesting the latter is likely overly conservative. By contrast, using the adult-based size criteria of TI-RADS and the two highest risk categories of ATA result in unacceptably low sensitivities of 76.9% and 84.6%, respectively. Thus, we recommend the use of TR4-TR5 as the cutoff for FNAB. Overall, these results highlight the problematic nature of a size criterion in children and the need to consider ultrasound characteristics and clinical context in the evaluation of pediatric thyroid nodules.