Nomogram to Differentiate Benign and Malignant Thyroid Nodules in the American College of Radiology Thyroid Imaging Reporting and Data System Level 5

Ting Hu; Zhengyi Li; Chuan Peng; Libing Huang; Huifang Li; Xu Han; Xingzhang Long; Wei Huang; Ruhai Zou


Clin Endocrinol. 2023;98(2):249-258. 

In This Article

Abstract and Introduction


Objectives: To develop and validate a nomogram for differentiating benign and malignant thyroid nodules of American College of Radiology (ACR) Thyroid Imaging Reporting and Data System (TI-RADS) level 5 (TR5) and improving the performance of the guideline.

Methods: From May 2018 to December 2019, 640 patients with TR5 nodules were retrospectively included in the primary cohort. Univariate and multivariable analyses were performed to determine the risk factors for thyroid cancer. A nomogram was established on the basis of multivariable analyses; the performance of the nomogram was evaluated with respect to discrimination, calibration, and clinical usefulness. The nomogram model was also compared to the ACR score model. External validation was performed and the independent validation cohort contained 201 patients from April 2021 to January 2022.

Results: Multivariable analyses showed that age, tumour location, multifocality, concomitant Hashimoto's disease, neck lymph node status reported by ultrasound (US) and ACR score were the independent risk factors for thyroid cancer (all p < .05). The nomogram showed good discrimination, with an area under the curve (AUC) of 0.786 (95% confidence interval [CI]: 0.742–0.830) and 0.712 (95% CI: 0.615–0.809) in the primary cohort and external validation cohort, respectively. Decision curve analysis demonstrated the clinical usefulness of the model. Compared to the ACR score model, the nomogram showed higher AUC (0.786 vs. 0.626, p < .001) and specificity (0.783 vs. 0.391).

Conclusions: The presented nomogram model, based on age, tumour features and ACR score, can differentiate benign and malignant thyroid nodules in TR5 and had a high specificity.


It has been reported that up to 68% of adults have thyroid nodules on high-resolution ultrasound,[1] but the reported incidence of malignant nodules in all thyroid nodules ranges only 1.6%–12%.[2] In the meantime, the mortality rate of thyroid cancer has remained relatively stable during the past few decades and the 30-year disease-specific mortality rate (<10%) is extremely low.[3]

Currently, ultrasound plays an essential role in the detection of thyroid nodules. According to 2015 American Thyroid Association Management Guidelines for Adult Patients with Thyroid Nodules and Differentiated Thyroid Cancer, for known or suspected thyroid nodules, nodular goiter, and other incidental imaging findings of thyroid nodules (computed tomography [CT], magnetic resonance imaging [MRI] and fluorodeoxyglucose [18F] positron emission tomography/CT [FDG-PET/CT]), ultrasound of the thyroid and cervical lymph nodes should be performed.[1] Typically, the diagnosis of thyroid cancer is made after an initial assessment of a thyroid nodule by ultrasound-guided fine-needle aspiration biopsy (FNAB) and interpretation using the Bethesda System for Reporting Thyroid Pathology (BSRTC). The system includes six categories: nondiagnostic (category I), benign (category II), atypia or follicular lesion of undetermined significance (category III), follicular neoplasm or suspicious for follicular neoplasm (category IV), suspicious for malignancy (category V), and malignant (category VI).[4]

With the increasing prevalence of differentiated thyroid cancer and the application of thyroid ultrasound around the world, thousands of patients are admitted to the hospital and further elect to undergo surgical intervention.[5,6] However, the trend of overdiagnosis and treatment modality has aroused wide concerns. In 2017, the American College of Radiology released the latest version of the Thyroid Imaging Reporting and Data System, which is a point-based risk stratification system that detects biologically significant thyroid cancers and guides decisions regarding FNAB and sonographic follow-up. Points are awarded on the basis of nodule sonographic features including composition, echogenicity, shape, margin and echogenic foci. According to summary points, the guideline can be used to categorise nodules and assign them to one of the five ACR TI-RADS risk levels: benign (TR1, 0 points), not suspicious (TR2, 2 points), mildly suspicious (TR3, 3 points), moderately suspicious (TR4, 4–6 points) and highly suspicious (TR5, 7 points or more). The ACR TI-RADS guideline recommends FNAB of TR5 nodules only if they are ≥10 mm due to an estimated risk of malignancy ≥20% in TR5 nodules.[7]

Although the ACR TI-RADS classification offers improved accuracy for ultrasound-based diagnosis of benign and malignant nodules, the malignant risk evaluation mainly depends on the performance of the equipment and the knowledge and experience of the physician. It has been reported that the aggregate risk level of TR5 lesions is 20.8%–68.4% for 10 points, indicating that ACR TI-RADS has poor specificity (44%–67.3%).[8–11] The dictionary of thyroid nodule ultrasound features presented in this report was used by our hospital after it was released. Some patients with TR5 nodules may prefer to be more positive about treatment and choose immediate FNAB or even surgery instead of active surveillance. Cytological assessment of thyroid nodules with US-guided FNAB is generally considered to be the standard diagnostic tool owing to its simplicity, safety, cost-effectiveness, and diagnostic accuracy. Studies report that that diagnostic sensitivity and specificity were 83%–98% and 70%–92%, respectively.[12] However, there are several limitations, such as false-negative results and a relatively high rate of nondiagnostic or indeterminate results, which correspond to categories I, III, and IV in the Bethesda System. Studies indicate that atypia of undetermined significance and follicular lesion of undetermined significance respectively occur in about 10%–33.6% and 15%–42% of all FNAB samples.[12] The American Thyroid Association guidelines recommend repeat FNAB or molecular testing for such nodules, but 10%–30% of category III are reclassified as category III in the repeat FNAB.[13] Unnecessary FNAB might place a substantial financial and medical burden and cause considerable anxiety to patients. The nomogram is a concise and intuitive graphical tool for risk prediction. We supposed that a nomogram could appropriately combine age and tumour features with the ACR TI-RADS score to obtain better predictive performance. Therefore, the aim of our study was to develop and validate a nomogram model that combines age, tumour features and ACR TI-RADS score for differentiating benign and malignant thyroid nodules in TR5 and improving the performance of the guideline.