Diagnostic Value of TI-RADS Classification System and Next Generation Genetic Sequencing in Indeterminate Thyroid Nodules

Abstract

This study aims to evaluate the diagnostic accuracy, inter-reader, and intra-reader variability of the ACR Thyroid Imaging Reporting and Data System (TI-RADS) for risk-stratification of indeterminate thyroid nodules using next generation genetic sequencing and tissue histology as a reference standard. Retrospective chart review was performed on all patients who underwent thyroid ultrasound for a nodule with subsequent fine-needle aspiration ± surgical resection from January 2017 to August 2018. Four radiologists with expertise in thyroid ultrasound assessed imaging twice, ≥1 month apart. Results of cytology and next generation genetic sequencing were used as a reference standard for high versus low risk of malignancy in each nodule. Inter-reader reliability between readers and intra-reader reliability between replicate self-reads for TI-RADS categorization were assessed. Univariate analysis, kappa statistics, and receiver operating characteristic curve were calculated. One hundred and thirty six nodules across 121 patients met inclusion criteria. 84.6% of patients were female and average age was 55.8 ± 14.1 years. One hundred and eighteen of 135 nodules (87%) had indeterminate cytology (Bethesda III or IV). One of 23 high-risk mutations was identified in 30.1% (42) of the nodules. Of the 52 patients who had surgery, 24 (47.1%) had confirmed malignant disease on surgical pathology. Inter-reader reliability between the four radiologists was marginal, κ = 0.293. Intra-reader reliability ranged from marginal to good, κ = 0.337 to κ = 0.560, respectively. The area under the receiver operating characteristic curve was 0.509, and no optimal TI-RADS Level for identifying high-risk nodules existed. The ACR TI-RADS classification system performs with low inter-reader and intra-reader reliability when assessing the genetic risk of nodules with indeterminate cytology.