Diagnostic performance of the ACR TI-RADS classification in identifying and excluding thyroid malignancy: a multi-correlative retrospective study in a South African tertiary hospital

Abstract Background Thyroid nodule evaluation is essential to exclude malignancy. The American College of Radiology Thyroid Imaging Reporting and Data System (ACR TI-RADS) stratifies malignancy risk based on sonographic features, guiding biopsy decisions and reducing unnecessary invasive procedures. Fine-needle aspiration cytology remains the preferred diagnostic tool due to its safety profile and cost-effectiveness; however, non-diagnostic and indeterminate results often necessitate repeat biopsies, which in resource-limited settings contribute to poor patient follow-up and missed malignancies. Core needle biopsy, with histological evaluation, is increasingly recognized as the gold standard for definitive diagnosis, reducing the need for repeat sampling. This study retrospectively assessed the diagnostic accuracy of ACR TI-RADS in identifying and excluding malignant thyroid lesions using histology as the reference standard, and exploring secondary correlation with cytology, biochemistry, and nuclear scintigraphy which were limited by availability. Methods This retrospective study reviewed records of patients undergoing thyroid ultrasound with ACR TI-RADS classification at a tertiary referral hospital between January 2020 and May 2024. Sonographic features, ACR TI-RADS scores, and histology results were analyzed, with additional correlation to cytology, biochemistry, and nuclear scintigraphy. Diagnostic accuracy of ACR TI-RADS was calculated against histology. For ease of analysis, ACR TI-RADS categories 1–3 were grouped as benign (test negative) and categories 4–5 as malignant (test positive). Results Sixty-eight patients were included (mean age 52.6 years; range 27–82), with female predominance and a 16.2% thyroid malignancy rate. Comparison of ACR TI-RADS with histology demonstrated sensitivity of 63.6%, specificity of 38.6%, positive predictive value of 16.7%, and negative predictive value of 84.6%. Receiver operating characteristic curve analysis showed an area under the curve of 0.51. Among sonographic features, intralesional vascularity was significantly associated with malignancy (p < 0.05), advocating its inclusion into a modified ACR TI-RADS scoring system. Conclusion ACR TI-RADS is a valuable tool for thyroid nodule risk stratification but demonstrates limitations in sensitivity and specificity within our setting. Discrepancies with histology highlight potential over or underestimation of malignancy risk, influenced by interoperator variability and inconsistent reporting. Standardized reporting protocols, ongoing training, and the incorporation of additional sonographic features, such as vascularity assessment, may improve diagnostic performance, thereby reducing patient morbidity and mortality.