Thyroid nodule malignancy risk stratification by numerous TIRADS systems – can we define the best one?

Background. Thyroid ultrasound (US) is a key examination for management of thyroid nodules. Numerous TIRADS systems aim to improve interobserver reproducibility of descriptions, but the existing different stratification criteria maintain accuracy in suboptimal levels.
Aim. The purpose of the study was to assess, which of the four TIRADS systems (L-TIRADS, EU- TIRADS, K-TIRADS or ACR-TIRADS) is more accurate and sensitive, and to evaluate, which of the systems have the strongest association with malignancy and in which categories of these systems is the stronger association with malignancy.
Methods. The prospective study in which thyroid ultrasound and FNA biopsy were done to 200 patients (23 to 85 years, the mean age 54 years) with 213 nodules according to the national thyroid nodule biopsy guidelines. The majority of patients were women 186 (87.3%) and 27 (12.7%) men. Each nodule was classified by all four TIRADS systems and nodule malignancy risk stratification comparison was accomplished based on malignant pattern defined by certified radiologist and approved by cytology results (Bethesda).
Results. Of the 213 thyroid nodules, 149 nodules (69.95%) were benign, 9 (4.23%) nodules were suspicious of malignancy and 20 (9.39%) were malignant. All of the systems showed high sensitivity (100%), however, L-TIRADS appeared with higher accuracy (68%) and area under ROC curve (81%). All systems had a high sensitivity (100%), however, L-TIRADS showed a better accuracy (ACC=68%, p<0.001) than other systems – EU-TIRADS (ACC=40%, p<0.001), K-TIRADS (ACC=47%, p<0.001) and ACR-TIRADS (ACC=39%, p<0.001). Evaluation of L-TIRADS 4A-5 categories individually reveals that there is a statistically significant association between L-TIRADS 4C and malignancy (p <0.001) with a moderately high AUC value (0.63) and between L-TIRADS 5 and malignancy (p<0.001) with AUC of 0.60 L-TIRADS 4A (one sign) and 4B (two signs) categories did not show difference in malignancy risk (p=0.63 and p=0.99), respectively.
Conclusion. Although different TIRADS systems show high sensitivity, there are significant differences in accuracy among the intermediate and malignant subcategories (TIRADS 3–5). Simplified L-TIRADS system categorization by count of malignant signs was recognized as the most accurate in comparison to EU-TIRADS, K-TIRADS or ACR-TIRADS systems.