ECE2021 Audio Eposter Presentations Late Breaking (114 abstracts)
1Ankara City Hospital, Endocrinology and Metabolism, Turkey; 2Ankara Yildirim Beyazit University, Faculty of Medicine, Endocrinology and Metabolism, Turkey; 3Ankara City Hospital, General Surgery, Turkey; 4Ankara Yildirim Beyazit University, Faculty of Medicine, Biostatistics, Turkey
Background
Our aim was to compare the effectiveness of five different risk-stratification systems using the ultrasonographical features of thyroid nodules in determining malignant histopathology.
Methods
The preoperative ultrasonography (US) features of 6925 nodules of 3030 patients who underwent thyroidectomy between 2007 and 2014 were analyzed retrospectively. The nodules were classified according to Kwak-TIRADS (Thyroid imaging reporting and data system), European Thyroid Association-TIRADS (EU-TIRADS), Korea-TIRADS, American Society of Clinical Endocrinologists (AACE), American Thyroid Association (ATA).1362 nodules (1042 patients) that could be scored in all of these systems were included. Malignancy rate, specificity, sensitivity, positive predictive value (PPV) and negative predictive value were calculated.
Results
In the nodules with high risk scored, malignant histopathology rate were 50%, 27.7%, 32.2%, 29.9%, 31.9% in Kwak-TIRADS, EU_TIRADS, Kore-TIRADS, AACE, ATA, respectively (Table 1). Area under curve (AUC) was higher than 0.5 for all systems (P <0.001, Table 2). When determining optimal cut-off for each system, sensitivity, specificity, PPV, and NPV were about 60%, 75%, 2730% and 92%, respectively (Table 2). EU-TIRADS had significantly lower AUC (corrected P<0.05). AUC of Korea- TIRADS was lower than that of ATA (corrected-P<0.001). AUCs of Kwak-TIRADS, AACE, and ATA were similar and significantly higher than the others (Table 2).
Benign (n = 1179) | Malignant (n = 183) | |
n (%1/%2) | n (%1/%2) | |
Kwak-TIRADS | ||
Possibly benign | 137 (97.9/11.6) | 3 (2.1/1.6) |
Low | 774 (91.8/65.7) | 69 (8.2/37.7) |
Intermediate | 61 (82.4/5.2) | 13 (17.6/7.1) |
Moderate | 195 (69.4/16.5) | 86 (30.6/47.0) |
High | 12 (50.0/1.0) | 12 (50.0/6.6) |
EU-TIRADS | ||
Benign | 97 (95.1/8.2) | 5 (4.9/2.7) |
Low | 729 (91.8/61.9) | 65 (8.2/35.5) |
Intermediate | 58 (100.0/4.9) | 0 (0.0/0.0) |
High | 295 (72.3/25.0) | 113 (27.7/61.8) |
Kore-TIRADS | ||
Benign | 129 (94.9/10.9) | 7 (5.1/3.8) |
Low | 787 (92.4/66.8) | 65 (7.6/35.5) |
Intermediate | 57 (81.4/4.8) | 13 (18.6/7.1) |
High | 206 (67.8/17.5) | 98 (32.2/53.6) |
AACE | ||
Low | 187 (96.4/15.9) | 7 (3.6/3.8) |
Intermediate | 729 (91.9/61.8) | 64 (8.1/35.0) |
High | 263 (70.1/22.3) | 112 (29.9/61.2) |
ATA | ||
Bening | 129 (94.9/10.9) | 7 (5.1/3.8) |
Very low | 58 (100.0/4.9) | 0 (0.0/0.0) |
Low | 729 (91.8/61.9) | 65 (8.2/35.5) |
Intermediate | 54 (80.6/4.6) | 13 (19.4/7.1) |
High | 209 (68.1/17.7) | 98 (31.9/53.6) |
1 Row percentage; 2Column percentage. |
Cut-off | Specificity | Sensitivity | PPV | NPV | AUC | 95% CI | |
Kwak-TIRADS | intermadiate | 60.7% | 77.3% | 29.3% | 92.7 | 0.717 | 0.680–0.754 |
EU-TIRADS | high | 61.7% | 75.0% | 27.7% | 92.7 | 0.680 | 0.641–0.719 |
Kore-TIRADS | intermadiate | 60.7% | 77.7% | 29.7% | 92.7 | 0.705 | 0.666–0.744 |
AACE | high | 61.2% | 77.7% | 29.9% | 92.8 | 0.710 | 0.674–0.747 |
ATA | intermadiate | 60.7% | 77.7% | 29.7% | 92.7 | 0.713 | 0.675–0.751 |
Conclusion
Risk-stratification systems have variable efficiency in predicting malignancy risk since they are operator dependent and require experience in that field. The sensitivity was found to be lower than the previous studies which might be due to retrospective evaluation of US reports rather than real time risk scoring.