Ultrasonography (US) is the first-line and most preferred examination method for thyroid nodules. Conventional US can be used to detect nodules, and it also provides information on the composition, size, and number of nodules, as well as their anatomical relationships with surrounding structures. US findings that have been reported to be suggestive of malignant thyroid nodules include solid composition, hypoechogenicity or marked hypoechogenicity, a microlobulated or irregular margin, mixed calcification or microcalcifications, and a taller-than-wide shape [1,2]. However, a considerable number of US findings overlap between benign and malignant thyroid nodules, making it difficult to identify thyroid malignancies based on US imaging alone [3,4]. Fine-needle aspiration (FNA) can play an important role in differentiating thyroid nodules because it has a high sensitivity and specificity. However, the diagnostic performance of FNA depends on the operator’s experience, it is an invasive method, and it yields a high proportion of indeterminate (10%–20%) or nondiagnostic (10%–15%) results [5,6]. The recently published World FedE_ration for Ultrasound in Medicine and Biology guideline recommends US elastography as an additional modality in thyroid nodules to support surgical excision and periodic observation [7]. US elastography is a new sonographic imaging technology that estimates tissue hardness and helps to discriminate benign from malignant nodules. The basic concept of US elastography was first developed in 1991 by Ophir et al. [8]. Most malignant nodules have abnormally firm stroma because of the presence of myofibroblasts, which make it possible to differentiate thyroid cancer from benign nodules using elastography. Nell et al. [9] conducted a systematic review and meta-analysis of 20 studies. Their analysis demonstrated that unnecessary invasive diagnostic procedures such as FNA could be avoided in completely soft (Asteria elastography 1) thyroid nodules [10] with a 99% negative predictive value.

There are two kinds of US elastography: strain elastography and shear wave elastography (SWE) [11]. The limitations of strain elastography include difficulties with the quantitative estimation of tissue hardness and its low reliability as a result of operator dependency. In contrast, SWE uses transversely oriented shear waves created by mechanical vibration forces. The shear waves propagate faster in stiff tissue than in soft tissue, thus enabling a quantitative approach using color maps and elasticity indices including E_max, E_mean, E_min, and E_ratio (with units of kilopascals [kPa]) [11]. This new technique has made it possible to evaluate tissue stiffness in a manner that is more quantitative, more reproducible, and less operator-dependent [12,13]. Some comparative studies found that SWE had lower interobserver variability than strain elastography; thus, it was preferred to real-time elastography and regarded as a promising tool with higher diagnostic performance and the ability to provide quantitative results for differentiating thyroid nodules [14].

Until now, the majority of US elastography studies for predicting thyroid malignancies have investigated strain elastography [15-17]. To the best of our knowledge, only a few studies have investigated the use of SWE to identify thyroid malignancies, and those studies showed varying results in terms of diagnostic performance and the optimal cutoff values of elasticity parameters [18-21]. Therefore, the aim of this study was to compare clinicopathologic and radiologic factors between benign and malignant thyroid nodules and to evaluate and verify the diagnostic performance of a combination of SWE and B-mode US in distinguishing histologically proven malignant from benign thyroid nodules.

Our study revealed that patients with malignant thyroid nodules were significantly younger (mean age, 47.6 years) than those with benign nodules (mean age, 55.5 years). This is consistent with the results of a previous large-scale study of the use of SWE to detect cancer in thyroid nodules that found the mean age of patients with a thyroid malignancy to be 42.9 years, while the mean age of those with benign nodules was 52.2 years [23]. When assessing indeterminate or suspicious thyroid nodules, a more careful evaluation is required if the patient is relatively young.

We investigated B-mode US finding features that could be useful for differentiating malignant from benign thyroid nodules, and found that solid composition, hypoechogenicity, an irregular/microlobulated margin, a taller-than-wide shape, and the presence of calcifications were significantly more common in malignant thyroid nodules than in benign thyroid nodules. These results are consistent with those of previous studies on US findings suggestive of malignancy in thyroid nodules [24-26]. In the present study, we compared and evaluated the thyroid nodules using SWE. All elasticity parameters except E_ratio (E_max, E_mean, and E_min) on SWE were significantly higher for malignant thyroid nodules than for benign thyroid nodules. This observed relationship between elasticity and the likelihood of thyroid malignancy is consistent with several prior publications analyzing the use of SWE [11,18-21,23].

Among the elasticity parameters, only E_max was found to be an independent risk factor for malignancy in thyroid nodules in the multivariate logistic regression analysis (OR, 1.22; 95% CI, 1.20 to 1.57, P=0.021). E_max had the best diagnostic performance, with a sensitivity, specificity, and AUC of 53.7%, 70.4%, and 0.633, respectively. In concordance with the present results, Wang et al. [27] also reported, in a study of 322 thyroid nodules, that E_max was the only independently significant SWE imaging parameter, with a sensitivity of 59.40%, a specificity of 89.8%, and an AUC of 0.796 for differentiating between malignant and benign thyroid nodules.

Several studies have reported a somewhat variable range of values of SWE imaging parameters that could be used to predict thyroid malignancies, especially regarding the optimal cutoff value of E_max. In our study, in which 43% of the nodules were malignant, the optimal cutoff value of E_max was 47.1 kPa, while those of E_mean and E_min were 54.9 kPa and 40.4 kPa, respectively. A prior study reported similar results to ours, demonstrating that the optimal cutoff value of E_max was 47.0 kPa for predicting malignancy in 322 nodules, of which 33% were malignant [27]. However, Liu et al. [21], in an analysis of 313 nodules, of which 194 were malignant, reported that the optimal cutoff value of E_max was 51.9 kPa, with a sensitivity of 81.4% and a specificity of 83.2%. This may reflect the higher proportion of malignant nodules than in our study, because malignant nodules show higher SWE values than benign nodules. In contrast, Veyrieres et al. [19] analyzed 297 nodules, of which only 12% were malignant, and reported a high threshold cutoff value of E_max (66 kPa) with high sensitivity (80.0%), specificity (90.5%), and negative predictive value (97.1%), but with a relatively low positive predictive value (52.8%). Thyroid malignancies with a maximum diameter equal to or less than 1 cm showed significantly lower SWE values than nodules larger than 1 cm, probably because larger nodules might be subjected to higher precompression due to their convex bulging through the skin [28]. While the mean diameter of the thyroid malignancies studied by Veyrieres et al. [19] was 25.0±13.8 mm, the mean diameter was 15.7±10.2 mm in our study. As discussed above, the optimal cutoff values of SWE parameters vary considerably due to differences in the number of enrolled patients, the proportion of malignant thyroid nodules, and nodule size. Furthermore, several other characteristics of nodules, such as the presence of calcifications or cystic changes and their position, could affect the results [28]. Therefore, it is necessary to further assess the optimal cutoff values for predicting malignant thyroid nodules in a general screening population in a future study.

A previous study including 313 thyroid nodules reported that combining B-mode US findings with SWE parameters improved diagnostic performance for detecting thyroid malignancy, with the sensitivity and negative predictive value increasing from 89.7% and 84.0% to 94.6% and 90.0%, respectively [21]. However, the specificity and positive predictive value slightly decreased when B-mode US findings and SWE were combined. Another previous large-scale study involving 476 thyroid nodules showed similar results. The combination of B-mode US findings and elasticity parameters significantly increased the sensitivity, negative predictive value, and AUC from 92.9%, 68.7%, and 0.769 to 95.0%, 74.0%, and 0.834, respectively, but with slight decreases in the specificity and positive predictive value [18]. In our study, we observed slight improvements in all parameters of diagnostic performance as a result of combining the B-mode US category of high suspicion with each elasticity parameter, compared with using only the B-mode US category. It was especially meaningful that the sensitivity and negative predictive value increased from 95.1% and 95.8%, respectively, when using only the B-mode US category, to 100% and 100%, respectively, when combining the B-mode US category with E_max. Screening thyroid US requires a high sensitivity and a high negative predictive value in order to avoid unnecessary biopsies and surgical resections. However, in this present study, no statistically significant added benefit in diagnostic performance was achieved by adding SWE. This may have been due to the excellent diagnostic performance of the B-mode US category of high suspicion by itself, with a sensitivity of 95.1%, a specificity of 85.12%, a positive predictive value of 82.9%, a negative predictive value of 95.8%, and an accuracy of 89.5%.

There were several limitations of this study. First, selection bias was inevitable due to its retrospective design. Second, this study was based on results obtained by two operators, and we did not assess interobserver variability. However, SWE is known to be relatively reproducible and minimally operator-dependent. All thyroid nodule data were obtained using the same protocol to minimize operator bias. Third, this study included a relatively small number of thyroid nodules and only presented experiences from a single center. Therefore, our findings should be verified by a prospective study with a large population. Lastly, there was an unusually high proportion of malignant nodules in our study group. The patients included in the present study who had indeterminate or suspicious lesions underwent US FNA or surgical removal.

SWE is a valuable additional diagnostic tool for accurately stratifying the risk of thyroid nodules in routine clinical settings. A combination of B-mode US with SWE yielded improved diagnostic performance compared to B-mode US findings only, although the differences in diagnostic performance were statistically non-significant.