Choosing the best algorithm among five thyroid nodule ultrasound scores: from performance to cytology sparing—a single-center retrospective study in a large cohort

Abstract

Objective Incidental diagnosis of thyroid nodules, and therefore of thyroid cancer, has definitely increased in recent years, but the mortality rate for thyroid malignancies remains very low. Within this landscape of overdiagnosis, several nodule ultrasound scores (NUS) have been proposed to reduce unnecessary diagnostic procedures. Our aim was to verify the suitability of five main NUS. Methods This single-center, retrospective, observational study analyzed a total number of 6474 valid cytologies. A full clinical and US description of the thyroid gland and nodules was performed. We retrospectively applied five available NUS: KTIRADS, ATA, AACE/ACE-AME, EUTIRADS, and ACRTIRADS. Thereafter, we calculated the sensitivity, specificity, PPV, and NPV, along with the number of possible fine-needle aspiration (FNA) sparing, according to each NUS algorithm and to clustering risk classes within three macro-groups (low, intermediate, and high risk). Results In a real-life setting of thyroid nodule management, available NUS scoring systems show good accuracy at ROC analysis (AUC up to 0.647) and higher NPV (up to 96%). The ability in FNA sparing ranges from 10 to 38% and reaches 44.2% of potential FNA economization in the low-risk macro-group. Considering our cohort, ACRTIRADS and AACE/ACE-AME scores provide the best compromise in terms of accuracy and spared cytology. Conclusions Despite several limitations, available NUS do appear to assist physicians in clinical practice. In the context of a common disease, such as thyroid nodules, higher accuracy and NPV are desirable NUS features. Further improvements in NUS sensitivity and specificity are attainable future goals to optimize nodule management. Key Points • Thyroid nodule ultrasound scores do assist clinicians in real practice. • Ultrasound scores reduce unnecessary diagnostic procedures, containing indolent thyroid microcarcinoma overdiagnosis. • The variable malignancy risk of the “indeterminate” category negatively influences score’s performance in real-life management of thyroid lesions.

Diagnostic Performance of Kwak, EU, ACR, and Korean TIRADS as Well as ATA Guidelines for the Ultrasound Risk Stratification of Non-Autonomously Functioning Thyroid Nodules in a Region with Long History of Iodine Deficiency: A German Multicenter Trial

“…Therefore, to the best of our knowledge, the current data provide the most comprehensive results from an area with history of iodine deficiency. In a recently published Italian real-life setting study (single-center, retrospective, observational) that included 6474 cytologically investigated TNs and comprised five different RSSs, inferior sensitivities (50.1-94.5%), PPV (7.7-11.5%), and AUC values in ROC analyses (0.606-0.632) were reported [45]. Among other reasons, such as a different history of iodine supply between Germany and Italy [46], the superior performance of the RSSs in the current study may be due to the exclusion of non-autonomously functioning lesions.…”

Classic and Follicular Variant of Papillary Thyroid Microcarcinoma: 2 Different Phenotypes Beyond Tumor Size

“…For nonincidental microcarcinomas, preoperative US features and cytological results were also available. mPTCc more often showed a score ≥ TR4 at ACR-TIRADS evaluation [ 10 ] or were categorized as high, according to the AACE/ACE-AME classification [ 11 ] ( P = .002 for each). Notably, by ROC curve analysis, a high class of AACE/ACE-AME score showed a sensitivity of 82% and a specificity of 66% when identifying mPTCc (area under the curve [AUC] = 0.671; 95% CI 0.56-0.78, P = .004) ( Fig.…”

“…For each patient, we collected several clinical features, namely, age, gender, familial history of thyroid cancer, concurrent autoimmune thyroid disease, history of other cancer, and presurgical thyrotropin (TSH) levels. For nonincidental cases, we collected all the ultrasonographic (US) features of the investigated nodules (including size [3-dimensional]; composition [solid, mixed, or cystic]; position of the solid portion in case of a mixed nodule [eccentric or not]; echogenicity [anechoic, hyperechoic, or isoechoic; slightly hypoechoic, hypoechoic, or marked hypoechoic]; halo [present, absent, or present but discontinuous or thick]; margins [well defined or smooth, irregular, or blurred]; shape [taller than wide], presence of echogenic foci [hyperechoic spot, macro- and microcalcifications]; rim calcification with extrusive soft tissue component and type of vascularization [absence of flow signals; perinodular and absent or slight intranodular blood flow; marked intranodular blood flow or mixed]) and we scored each lesion according to the 2 highest performing nodule US scores in our institution's experience [ 10 ]: the American College of Radiology (ACR) Thyroid Imaging Reporting and Data Systems (TIRADS) [ 11 ] and the American Association of Clinical Endocrinologists (AACE/ACE-AME) [ 12 ], as previously reported [ 10 ]. Among the physicians, Cohen's κ statistics ranged from 0.5 to 0.73, according to different US scores [ 10 ].…”

Low-cost Thyroid Nodule Diagnosis Using a Supervised Autoencoder Neural network

“…Artificial Intelligence -beyond the classical methods mentioned previously -has proved to be very potent in medical applications in the last few years [30]. Deep neural networks are used in state-ofthe-art practices to perform both classification tasks (for example, on echography or cytology-based imagery) [24], [31] or feature selection (for example, on metabolomic data) [32]. Overall, they have proven their effectiveness in bioinformatics for classification and feature selection [33], [34].…”