Abstract
OBJECTIVES:
To compare the diagnostic efficacy of shear wave elastography (SWE) comnined with ACR TI-RADS categories for malignancy risk stratification of thyroid nodules with interminate FNA cytology.
METHODS:
The clinical data, sonographic features, ACR TI-RADS grading and shear wave elastography images of 193 patients of surgical pathologically proven thyroid nodules with interminate FNA cytology were retrospectively analyzed. The diagnostic efficacy of ACR TI-RADS categories, the maximum Young’s modulus (Emax) of SWE and the combination of the two were calculated respectively.
RESULTS:
The ROC curves were drawn using surgical pathology results as the gold standard. The ROC curves indicated that the cut-off value of ACR TI-RADS and Emax of SWE was TR5 and 41.2 kPa respectively, and the area under the ROC curve (AUC) was 0.864 (95% CI: 0.879–0.934) and 0.858 (95% CI: 0.796–0.920) respectively. The diagnostic sensitivity, specificity and accuracy of ACR TI-RADS was 81.4% (127/156), 84.8% (31/37), and 81.9% (158/193), respectively. That of SWE Emax was 80.8% (126/156), 78.4% (29/37), and 80.3% (155/193), respectively. After SWE combined with ACR TI-RADS, the sensitivity, specificity and accuracy was 94.2% (147/156), 75.7% (28/37), and 90.7% (175/193), respectively.
CONCLUSIONS:
ACR TI-RADS classification system and shear wave elastography had high diagnostic efficacy for thyroid nodules with interminate FNA cytology. The combination of the two could improve diagnostic sensitivity and accuracy, and could help to differentiate benign and malignant thyroid nodules with interminate FNA cytology.
Introdution
Elastography was first proposed and applied by Ophir in 1991 [1]. In recent years, it has been widely used in ultrasonic diagnosis of breast, thyroid, musculoskeletal system and prostate [2]. Elastography has some different typies. Among them, shear wave elastography (SWE) was the latest clinical elastic imaging technique which directly determined the shear wave velocity (m/s) or calculated the Young’s modulus (kPa) to display and superimpose on two-dimensional gray-scale images in real time [3–5]. In 2009, Horvath et al proposed the Thyroid Imaging Report and Data System (TI-RADS) firstly according to the model of Breast Imaging Report and Data System (BI-RADS) which developed by the American College of Radiology (ACR) in 2003 [6]. Since then, various versions have been proposed [7–13]. Based on large-scale evidence and clinical validation, the new version of TI-RADS were launched by ACR in 2017 [12]. It was mainly graded according to the five suspicious ultrasound features of the nodule. The TI-RADS level and the risk of malignancy of the nodule increased with the total score [12, 14]. Many reports confirmed this TI-RADS has high diagnostic accuracy [15–22]. Ultrasound-guided fine needle aspiration (FNA) has become a reliable method for the diagnosis of thyroid nodules, and its diagnostic sensitivity, specificity and accuracy could reach more than 90% [23, 24]. But its deficiency lied in the fact that some puncture specimens were not satisfied or the diagnosis was not clear, so repeated puncture, coarse needle biopsy or surgical resection were needed. The “indeterminate” categories in the Bethesda System included AUS/FLUS (Atypia of undetermined significance/Follicular lesion of undetermined significance), FN/SFN (Follicular Neoplasm/Suspicious for a Follicular Neoplasm) and SM (Suspicious for Malignancy) [25]. AUS/FLUS and FN/SFN occured in some throid nodules. For axample, AUS/FLUS occured in 3–18% of thyroid nodules, and had a malignant risk of 5–15% [26–27]. The purpose of this study was to investigate the value of SWE combined with ACR TI-RADS in the diagnosis of thyroid nodules with interminate FNA cytology by retrospectively analyzing the clinical data, sonographic features, ACR TI-RADS categories and shear wave elastography images of 193 patients of surgical pathologically proven thyroid nodules with interminate FNA cytology.
Materals and methods
Patients
The inclusion criteria of this study were as follow: 1) the maximal diameter of thyroid nodules were ≥5 mm; 2) nodules had been undergone FNA and the FNA cytology was Bethesda III, IV or V; 3) nodules were confirmed by surgical pathology; 4) the patients of nodules aged between 18 and 90 yrs old. The exclusion criteria were as follow: 1) the patients were lack of information on FNA or SWE; 2) the patients could not be undergone FNA or surgery. For multiple nodules in one patient, only one nodule was studied. The most suspicious malignant nodule was selected firstly, and the largest nodule was selected secondly. Finally, 193 patients of thyroid nodules from January 2014 to December 2019 were selected in this study. There were 41 men and 152 women. The mean age was (46.1±13.1) yrs (ranged, 19–75 yrs). The mean maximal diameter of nodules was (11.3±5.7) mm (ranged, 5.0–31.0 mm). The retrospective study was approved by the Ethics Committee of our hospital, and informed consent to include the data for analysis was obtained from all patients.
FNA and cytopathological classification
According to the recommendation of ACR TI-RADS, FNA was performed for the TR3 nodules≥25 mm, the TR4 nodules≥10 mm, and the TR5 nodules≥10 mm. For those patients with heavy mental burden, the nodules≥5 mm could be punctured. FNA was performed by two operators who had engaged in FNA for more than 3 years.
The pre-puncture preparation included the assessment of patients and thyroid nodules, and signing of informed consent. After skin local anesthesia with 2% lidocaine, FNA was performed by a 22-gauge PTC needle under ultrasound guidance. Three to four smears were obtained for evey nodule, and they were collected with 95% ethanol. FNA cytopathologies were classified in accordance with TBSRTC as follows: Bethesda I (nondiagnostic or unsatisfactory), Bethesda II (benign), Bethesda III (AUS/FLUS, atypia of undetermined significance/follicular lesion of undetermined significance), Bethesda IV (FN/SFN, follicular neoplasm/suspicious for follicular neoplasm), Bethesda V (suspicious for malignancy), and Bethesda VI (malignant) [25]. Bethesda VI nodules belonged to malignant nodules confirmed by FNA cytology, while Bethesda II nodules which did not change for 6 months follow-up belonged to benign nodules confirmed by FNA cytology [28]. In this study thyroid nodules with interminate FNA cytology included Bethesda III, IV, and V nodules.
Shear wave elastography (SWE)
SWE and 2D-US were performed by two ultrasound operators who had engaged in thyroid ultrasound for more than 8 years. SWE was performed using three Super-Sonic Imaging Ultrasonic Instruments (France, 4–15 MHz linear transducer). During SWE operation, the probe was placed lightly without pressure. SWE mode was switched when 2D gray-ultrasonic (2D-US) image was satisfied, so that 2D-US image and SWE image were displayed up and down in the same screen. Secondly, the region of interest (ROI) was chosen to cover the target nodule and a few peripheral normal thyroid tissue. SWE image was frozen and stored after image stabilization. Thirdly, the Q-Box was placed inside the target nodule (avoiding cystic and coarse calcified sites) for measurements of Young’s modulus Emax (kPa), Emin (kPa), Emean (kPa), SD and shear wave velocity, and then the images were frozen and stored. Forthly, Q-Box Ratio measurements were performed: the first Q-Box (2 mm size) was taken to place the highest hardness of the target nodule (except the calcification), and the 2nd Q-Box was placed in the normal thyroid tissue at the same depth. Young’s modulus, Ratio and other indexes were measured, and then images were freezen and stored. The transverse and longitudinal sections of each nodule were measured for 5 times, and the patient was told to hold his/her breath for 2∼3 s when measuring. This study included only the Young’s modulus Emax (kPa) of SWE, and the average values of 10 measurements (5 transverse and 5 longitudinal sections) were taken statistically.
ACR TI-RADS categories
According to ACR TI-RADS categories, nodule composition, echogenicity, margin, shape, and echogenic foci were evaluated and a score was assigned to each ultrasound feature. For composition (choosing one), “cystic or almost comletely cystic” or “spongiform” awarded to 0 point, “mixed cystic and solid” awarded to one point, and “solid or almost comlietely solid” awarded to 2 points. For echogenicity (choosing one), “anechoic”, “hyperechoic or isoechoic”, “hypoechoic”, and “marked hypoechoic” awarded to 0, 1, 2, and 3 points, respectively. For margin (choosing one), “smooth or ill-defined”, “lobulated or irregular”, and “extra-thyroidal extension” awarded to 0, 2, and 3 points, respectively. For sharp (choosing one), “wider-than-tall” and “taller-than-wide” awarded to 0 and 3 points respectively. For echogenic foci (choosing all that apply), “none or large comet-tail artifact”, “macrocalcification”, “peripheral (rim) calcification”, and “microcalcification” awarded to 0, 1, 2, and 3 points, respectively. ACR TI-RADS classified according to the sum of points as follow: TR1 : 0 point, TR2 : 2 points, TR3 : 3 points, TR4 : 4–6 points, and TR5: ≥7 points [12].
Statistical analysis
Statistical data were analyzed using a version 20.0 of SPSS software. Quantitative data were expressed as the mean±standard deviation (SD). The independent two-sample t-test was used to compare the patient age and the nodule size. The gender ratio of patients, diagnosis sensitivity, specificity, and accuracy were compared by Chi-squared tests. Receiver operator characteristics (ROC) curves were drawn to determine the cut-off values of ACR TI-RADS and SWE Emax. When SWE Emax was higher than the cut-off value, the grade level of ACR TI-RADS TR3 or TR4 were up-regulated. After combination, TR 5 was used as the cut-off value to judge the benign and malignant nodules. The difference was statistically significant with P < #x003C;< #x200A;0.05.
Results
Pathological findings
Of 193 nodules in this study, 37 nodules (19.2%) and 156 nodules (80.8%) were diagnosed as benign and malignant by surgery pathology respectively. There were 150 papillary thyroid carcinomas, 4 follicular thyroid carcinomas, one hyalinizing trabecular tumor (HTT), one medullary carcinoma, 36 nodular goiters, 2 chronic lymphocytic thyroiditis, one subacute thyroiditis, one follicular adenoma, 7 Hashimoto’s nodules, and 11 adenomatous goiters (Fig. 1).
Ultrasound images, SWE images and pathological images of nodules. Image A: a 7 mm nodule in a 62-yrs-old women, ACR TI-RADS TR5 (Image A1). SWE Emax = 31.5 kPa <41.2 kPa (Image A2). Surgery pathology comfirmed a papillary thyroid carcinoma (Image A3, ×400). Image B: a 24 mm nodule in a 33-yrs-old women, ACR TI-RADS TR4 (Image B1). SWE Emax = 74.9 kPa >41.2 kPa (Image B2). Surgery pathology comfirmed a papillary thyroid carcinoma (Image B3, ×400).
The mean age of the patients of benign nodules (50.5±13.2 yrs) was older than that of malignant ones (45.1±12.9 yrs) (t = 2.28, P = 0.024). There was no significant difference in the gender ratio of the patients between benign nodules and malignant ones (10/33 vs. 32/124, x2 = 0.152, P = 0.696). There was no significant difference in the average of max diameter between benign nodules (12.0±6.7 mm) and malignant ones (11.2±5.5 mm) (t = 0.83, P = 0.406). There were significant differences in composition, echogenicity, shape, margin, and microcalcification between benign nodules and malignant ones (all P < #x003C;< #x200A;0.05, Table 1)
Clinic and ultrasound characteristics for the patients with nodules
Clinic and ultrasound characteristics for the patients with nodules
There was significant difference in SWE Emax between benign nodules (35.65±13.01 kPa) and malignant ones (67.99±38.85 kPa) (t = 4.99, P < #x003C;< #x200A;0.001).The ROC curves indicated that the cut-off value of ACR TI-RADS and SWE Emax was TR5 and 41.2 kPa respectively, and the area under the ROC curve (AUC) was 0.864 (95% CI: 0.879–0.934) and 0.858 (95% CI: 0.796–0.920) respectively (Fig. 2). The diagnostic sensitivity of ACR TI-RADS and SWE Emax was 81.4% (127/156) and 80.8% (126/156) respectively. The specificity of ACR TI-RADS and SWE Emax was 84.8% (31/37) and 78.4% (29/37) respectively. The accuracy of ACR TI-RADS and SWE Emax was 81.9% (158/193) and 80.3% (155/193) respectively (Table 2).
ROC curves of SWE Emax and ACR TI-RADS. The area under the ROC curve (AUC, 95% CI) of SWE Emax and ACR TI-RADS was 0.858 (0.796–0.920) and 0.864 (0.787–0.940) respectively.
Diagnostic efficiency of SWE, ACR TI-RADS and the combination
After SWE combined with ACR TI-RADS categories, the diagnostic sensitivity, specificity and accuracy was 94.2% (147/156), 75.7% (28/37), and 90.7% (175/193), respectively (Table 2). The sensitivity of the combination was higher than that of SWE (x2 = 12.923, P < #x003C;< #x200A;0.001) and ACR TI-RADS categories (x2 = 11.986, P = 0.001). Compared with SWE (x2 = 0.076, P = 0.782) and ACR TI-RADS categories (x2 = 0.753, P = 0.386), the combination had no significant difference in diagnostic specificity. The diagnostic accuracy of the combination was higher than that of SWE (x2 = 8.355, P = 0.004) and ACR TI-RADS categories (x2 = 6.321, P = 0.012) (Table 3).
Comparison of diagnosis efficiency of the combination with SWE and ACR TI-RADS
Comparison of diagnosis efficiency of the combination with SWE and ACR TI-RADS
Ultrasonic elastography was applied to the diagnosis of thyroid nodules, and most reports confirmed its clinical application value. In our study, the AUC of SWE Emax was 0.858 (95% CI: 0.796–0.920), and the diagnostic sensitivity, specificity, and accuracy of SWE Emax was 80.9%, 78.4%, and 80.3%, respectively. Aghaghazvini L et al reported that real time SWE was a promising test for preoperative malignant tumor risk stratification in patients, and the maximum velocity had the strongest predictive value for both conventional and elastic imaging variables [29]. The study of Shang H et al suggested that SWE could be used to diagnose small thyroid nodules, which was closely related to the size of PTMC and thyroid capsule infiltration [30]. The study of Filho RHC et al reported 2D SWE was an important tool that supported ultrasound in clinical practice in the diagnosis of thyroid nodules [31]. Liao LJ et al reported that SWE was an independent predictor of malignant thyroid nodules [32]. There have also some reports of dissent. For example, Bardet S et al found that despite the high elastic values of classical PTC, conventional SWE indices could not differentiate the cytologically indeterminate thyroid nodules [33].
ACR TI-RADS was an ultrasound classification system for thyroid nodules which was proposed by ACR in 2017 [12], and many literatures had reported its clinical application value [15–18]. In our study, the AUC of ACR TI-RADS was 0.864 (95% CI: 0.879–0.934), and the iagnostic sensitivity, specificity, and accuracy were all more than 80%. Compared with other classification systems, ACR TI-RADS had advantages and was widely recognized [19–22]. Tappouni RR et al proposed that the adoption of ACR TI-RADS might require changes at the practice level, including image acquisition and workflow, interpretation and reporting, and substantial resources should be invested to train ultrasound examiners and radiologists to accurately identify features that contribute to the nodule scoring [15].
The study of Hang J et al showed that combining SWE and TI RADS improved the specificity and positive predictive value of TI RADS-alone in differentiating between benign and malignant nodules [34]. Liu BJ et al reported SWS and TI-RADS were beneficial to differentiate between follicular thyroid carcinoma and follicular adenoma by analyzing retrospectively conventional ultrasound and point-shear wave elastography (p-SWE) of 28 follicular thyroid carcinomas and 67 follicular adenomas which were proven by surgery pathology [35]. Li X et al reported that VTIQ combined with ACR TI-RADS could improve the specificity of the differential diagnosis of thyroid nodules without decreasing the sensitivity [36]. Jin ZQ et al reported that the modified TI-RADS based on ACR TI-RADS + SWE + CEUS could reduce the number of FNA performed on benign nodules and implement the consistent follow-up in clinical practice [37]. Du YR et al considered that TI-RADS in combination with VTI and VTQ could significantly improve the ability of diagnosis by analyzing TI-RADS, ultrasound elastography, VTI and VTQ manifestations of 142 thyroid nodules with ≤10 mm [38]. But there were also some reports of differing views. For example, the study of Yang BR et al found that it was limited to differentiate thyroid nodules of intermediate-suspicion in ATA guidelines using elastic score and strain ratio of elastography [39]. The study of Swan KZ et al showed that SWE had low diagnostic accuracy in prediction of thyroid carcinoma, and could not be an important method to evaluate thyroid nodules [40].
In our study, the combination of SWE and ACR TI-RADS had higher diagnostic sensitivity and accuracy than SWE-alone and ACR TI-RADS-alone, but also the specificity was not decreased. There were also some reports about SWE or ultrasound classification system for the cytologically indeterminate thyroid nodules. Barbosa TLM et al reported that ACR TI-RADS and ATA guidelines were useful to guide the management of indeterminate thyroid nodules [41]. Rocha TG et al reported that ACR TI-RADS was useful for predicting malignancy specifically in thyroid nodules >10 mm with indeterminate cytology [42]. Ahmadi S et al reported that using either ATA guidelines or ACR TI-RADS to risk-stratify the cytologically indeterminate thyroid nodules might be helpful to plan better for diagnosis and treatment for the clinicians [43]. Liu JF et al reported that SWE could effectively improve the diagnostic performance of ATA guidelines in differentiation malignant from benign of Bethesda III thyroid nodules [44]. But the study of Bardet S et al concluded that the traditional SWE index could not differentiate the cytologically indeterminate thyroid nodule despite the high elastic values of classical PTC [45].
There were some deficiencies in our study: First, this study was a retrospective study, and there was possible selection bias. Second, all shear wave elastography images and 2D ultrasound images were static images, which might cause to misdiagnosis. Third, the sample size of this study was small, and needed to be further studied by increasing the sample size. Fourth, This study did not analyze the intra- and inter- observer difference.
Conclusion
ACR TI-RADS classification system and shear wave elastography had high diagnostic efficacy for thyroid nodules with interminate FNA cytology. The combination of the two could improve diagnostic sensitivity and accuracy, and could help to differentiate benign and malignant thyroid nodules with interminate FNA cytology.