Journal Article

Abtract Background: This study compares the American Thyroid Association (ATA) classification system with the 2017 Thyroid Imaging Reporting and Data System (TI-RADS) for predicting cancer risk in thyroid nodules. METHODS:This is a retrospective review of ultrasound imaging of all adult patients with thyroid nodules >5mm who underwent thyroidectomy at a tertiary care hospital in 2016. We assessed the ability of either system to predict malignancy based on surgical histopathology. Sensitivity, specificity, negative (NPV) and positive predictive values (PPV), and area-under-the-curve (AUC) were calculated and compared using McNemar's, Fisher's exact, or DeLong's tests. RESULTS:323 nodules from 213 adults were included. Median patient age was 55 years; 75.6% were female. 27.2% nodules were malignant. Both ATA and ACR TI-RADS provide effective diagnostic performance, sensitivity of 77.3% vs 78.4%, respectively; specificity of 76.6% vs 73.2%; PPV of 55.3% vs 52.3%, and NPV of 90% for both. The level of agreement between the two classification systems was almost perfect [κ= 0.93, AUC 0.77 ATA vs 0.76 TI-RADS (p=0.18)]. However, of the 40 TR 3 nodules (<2.5 cm), 10% were malignant, and of the 31 TR4 nodules (<1.5 cm), 38% were malignant. CONCLUSION:The ATA and TI-RADS classification systems appear to have similar diagnostic value for predicting thyroid cancer. However, sub-analysis of TR3 and TR4 nodules with consideration of size criteria showed that there is a higher risk of missing a malignancy if ACR TI-RADS recommendation is followed. These results should be validated in a different patient cohort with a lower incidence of cancer.

Full Text

Duke Authors

Cited Authors

  • Ahmadi, S; Oyekunle, T; Jiang, XS; Scheri, R; Perkins, J; Stang, M; Roman, S; Sosa, JA

Published Date

  • January 18, 2019

Published In

PubMed ID

  • 30720343

Pubmed Central ID

  • 30720343

Electronic International Standard Serial Number (EISSN)

  • 1934-2403

International Standard Serial Number (ISSN)

  • 1530-891X

Digital Object Identifier (DOI)

  • 10.4158/ep-2018-0369


  • eng