Development and Validation of a Web-Based Malignancy Risk-Stratification System of Thyroid Nodules

Bin Zhang; Shufang Pei; Qiuying Chen; Yuhao Dong; Lu Zhang; Xiaokai Mo; Shuzhen Cong; Shuixing Zhang


Clin Endocrinol. 2020;93(6):729-738. 

In This Article

Abstract and Introduction


Objectives: Previous publications on risk-stratification systems for malignant thyroid nodules were based on conventional ultrasound only. We aimed to develop a practical and simplified prediction model for categorizing the malignancy risk of thyroid nodules based on clinical data, biochemical data, conventional ultrasound and real-time elastography.

Design: Retrospective cohort study.

Patients: A total of 2818 patients (1890 female, mean age, 45.5 ± 13.2 years) with 2850 thyroid nodules were retrospectively evaluated between April 2011 and October 2016. 26.8% nodules were malignant.

Measurements: We used a randomly divided sample of 80% of the nodules to perform a multivariate logistic regression analysis. Cut-points were determined to create a risk-stratification scoring system. Patients were classified as having low, moderate and high probability of malignancy according to their scores. We validated the models to the remaining 20% of the nodules. The area under the curve (AUC) was used to evaluate the discrimination ability of the systems.

Results: Ten variables were selected as predictors of malignancy. The point-based scoring systems with and without elasticity score achieved similar AUCs of 0.916 (95% confidence interval [CI]: 0.885–0.948) and 0.906 (95% CI: 0.872–0.941) when validated. Malignancy risk was segmented from 0% to 100.0% and was positively associated with an increase in risk scores. We then developed a Web-based risk-stratification system of thyroid nodules (

Conclusion: A simple and reliable Web-based risk-stratification system could be practically used in stratifying the risk of malignancy in thyroid nodules.


Thyroid nodules are a common and challenging clinical problem, the incidence of which has been increasing worldwide, largely due to more sensitive thyroid nodule screening[1,2] such as ultrasonography (US)-based screening. Approximately 20%-67% of the general population is estimated to have nodules that cannot be detected by palpation.[3] Thyroid cancer is currently the fifth leading cancer diagnosis in women. By the year 2030, it may become the second leading cancer diagnosis in women and the ninth leading cancer diagnosis in men.[4] Nevertheless, only 5%-15% of nodules are malignant.[5] This highlights the importance of reliable and highly accurate diagnostic methods that can differentiate malignant nodules from benign ones.

Conventional US is the preferred imaging tool for diagnosing thyroid diseases because it is noninvasive, convenient and inexpensive, and it does not expose the patient to damaging radiation. Nodule features detected through thyroid US help clinicians stratify patients according to their risk of cancer before fine needle aspiration (FNA) biopsy is performed. However, US features show wide-ranging diagnostic sensitivity for detecting nodule malignancy.[6] This reflects the complex structure of thyroid nodules and the many features common to benign and malignant nodules. Additionally, there was difference in US feature definitions. As a result, US features alone for diagnosing malignant nodules remain subjective and dependent on clinician experience.

Risk-stratification systems for thyroid nodules on US have been developed for the effective management of thyroid nodules, which could facilitate the effective interpretation and communication of thyroid US findings among referring physicians and cytopathologists. Currently, there are two types of risk-stratification system—a qualitative grading system or a quantitative scoring system.[7] The qualitative grading system such as American Thyroid Association (ATA) Management Guidelines was based on US feature pattern to stratify the malignancy risk of each nodule.[8] Most recently, Thyroid Imaging Reporting and Data Systems (TI-RADS), based on quantitative scoring, have been proposed to stratify the risk of malignancy in thyroid nodules and standardize the reporting system in thyroid US.[9] Several forms of TI-RADS have been proposed by calculating the number of suspicious US features to categorize thyroid nodules and recommend cytological diagnosis.[10–17] However, all risk-stratification systems were developed utilizing a combination of US features.

Hence, we sought to propose a freely available Internet Web-based program based on clinical, biochemical and US characteristics to stratify the malignancy risk of thyroid nodules. This practical prediction model may minimize potential harm from overuse of FNA and reduce the risk of unnecessary surgery for benign conditions, reducing healthcare costs and psychological burden and risk for patients.