Performance of Elastography for the Evaluation of Thyroid Nodules: A Prospective Study

Introduction

High-resolution ultrasound (US) is the most sensitive modality for the noninvasive identification and evaluation of thyroid nodules (TNs) (1,2). Specific sonographic features that are associated with a significantly increased malignancy risk include microcalcifications and irregular margins. The value of these features for predicting cancer is partially blunted by low sensitivities (3). Elastography is a promising new technique that may improve the US evaluation of TNs.

Elastography was developed at the University of Texas Medical School, Houston, by Ophir and colleagues in 1991 (4). This technology evaluates tissue stiffness. Malignant lesions tend to be stiffer than the surrounding benign tissues. There are two general US methods for determining TN stiffness; when a force is applied to the tissue, either the deformation parallel to the force direction (strain) or the deformation propagating perpendicular to the force direction (shear wave) is measured by using US signals (5).

Strain elastography

This method of elastography requires manual compression of the tissue being visualized. This manual compression then displaces the underlying tissue. The more elastic the tissue is, the more displacement is experienced. The displacement or distortion of the tissue is then measured and visualized as an elastogram (6).

Shear wave elastography

This method of elastography uses a patented technology called Sonic Touch to create a pushing beam. This pushing beam is a shear wave that travels perpendicular to the direction of propagation of the conventional US waves. This acoustic push pulse induces displacement of the tissue much like ripples in a pond (6). The stiffer the tissue is, the faster the shear wave will travel through the tissue. The wave is tracked, measured, and given a numerical value of the shear wave speed (7).

Acoustic radiation force impulse technology may be utilized to measure a numerical value of shear wave speed and tissue quantification (8). The quantifications of shear wave imaging were approved by the FDA on September 20, 2012 (9).

Several studies have investigated the performance of strain elastography (SE) in the evaluation of TNs. The results have been encouraging (10–15). One prospective study suggests that this technique is superior to conventional US for the evaluation of nodules selected for surgical resection (16), while two others suggest that it is inferior to conventional US (17,18). These studies have limitations. First, they have a selected patient population with a higher cancer prevalence than the general population (10–17). Second, the number of subjects is relatively small, so the elastography performance cannot be quantified relative to conventional sonographic characteristics of malignancy (10–16).

The objective of our study was to prospectively determine the performance characteristics of SE with manual compression alone and in conjunction with high-definition US characteristics of thyroid malignancy.

Methods

We prospectively evaluated 912 TNs in 706 patients from March 15, 2011, to March 15, 2012. All nodules were evaluated with high-resolution US, SE, fine-needle aspiration biopsy (FNAB), and, when indicated, by tissue pathology. All patients were evaluated by a single endocrinologist with >13 years experience in thyroid US and 15 months prior experience with elastography. The study protocol was approved by the Institutional Review Board of the New Hanover Regional Medical Center, Wilmington, NC. The inclusion criteria were the presence of a single or multiple TNs >5 mm and age ≥18 years. All patients gave written informed consent.

The TNs were evaluated for the following US characteristics: size, height, echogenicity, echogenic complexity, irregular margins, subcapsular location, isthmus location, macrocalcifications, microcalcifications, color Doppler, and power Doppler. The thyroid gland was evaluated for heterogeneity. TNs with microcalcifications were divided into two groups. Group 1 contained TNs with three or fewer foci of microcalcifications, and group 2 contained TNs with four or more foci.

Microcalcifications were defined as tiny flecks of calcium inside the TN without shadowing. Macrocalcification was defined as dense calcifications that create acoustic shadowing distally. Thyroid lesions were assessed for the presence of macrocalcification alone or the combination of macro- and microcalcifications. All thyroid lesions were also evaluated for power Doppler and color Doppler vascularity. The nodules were divided into four groups based on their Doppler vascular patterns. Group 1 had no blood flow; group 2 had peripheral blood flow only; group 3 had peripheral and central blood flow (peripheral>central); group 4 had primarily central blood flow (central>peripheral). The complex nodules were divided into four groups. Group 1 had <25% solid tissue; group 2 had 25–50% solid tissue; group 3 had 51–75% solid tissue; group 4 had >75% solid tissue. Subcapsular location was assigned to lesions <2 mm from the thyroid capsule.

At the time of the initial US examination, the following factors were ascertained in all study subjects: sex, age, and number of US-determined TNs ≥5 mm in diameter. Before FNAB, blood was drawn for the following measurements: thyrotropin, thyroid peroxidase antibody, antithyroglobulin antibody, thyroglobulin, and calcitonin. All measurements were performed by LabCorp (LabCorp Research Triangle Park-Center for Molecular Biology & Pathology, Esoteric Division, Research Triangle Park, NC).

Elastography

SE was performed using eSie Touch™ Elasticity Imaging Technology on the Siemens ACUSON S2000 ultrasound system with VB10D software and an 18L6 HD probe (Siemens, Erlangen, Germany). The tissue under investigation is lightly compressed with the transducer held in the longitudinal axis, while the scan line radiofrequency data are analyzed and compared with previous image frames. The same region of interest is kept within the frame from image to image, while slowly deforming the tissues with the probe. Small amounts of displacement, between 0.5% and 1%, are necessary between frames to allow estimation of stiffness (19,20). The elastography examination was the last part of the US examination before FNAB. Patients were asked to not swallow or breathe for few seconds. The image was then reproduced three times. When the calculated quality factor score was 40 or higher, the elastography image was then saved. The elastography score (ES) was assigned to the nodule before FNAB.

Elastography superimposes information in color on B-mode images. Each color represents a certain level of elasticity (21). In Color Map 1, the color purple represents soft tissue or elastic tissue, the color green represents median or isoelastic—representing the normal thyroid tissue—and the color red represents hard or less elastic tissue. The elastography image was displayed next to the B-mode image. Elastography in this study was rated with scores of ES 0-ES 3, and we used Color Map 1 to determine ES. Figure 1 depicts differences and ESs used for this study.

An ES 0 was assigned to purple nodules. These represented TNs that were more elastic than the surrounding tissue (soft tissue). ES 1 was assigned to green nodules, representing lesions with the same elasticity as the surrounding thyroid tissue. An ES 2 was given to yellow-orange nodules. These nodules were slightly less elastic (intermediate) than the surrounding thyroid tissue. Finally, ES 3 was assigned to red nodules, representing hard thyroid tissue with little to no elasticity. If a lesion had mixed elastography color, for example, 70% green and 30% red, this lesion was given the higher score as long as the color representing harder tissue was ≥30%.

FNAB procedure

Before the FNAB, we obtain written informed consent. The FNAB is performed under sterile conditions and with US guidance to confirm accurate needle placement. Three passes were made of each lesion using 27-gauge needles. If there was no sample on visual inspection, we then used a larger needle (25 gauge). Onsite adequacy was not performed in this study. The samples were submitted for cytology.

FNA cytology and histology

In this prospective study, 706 patients with 912 TNs underwent FNAB. There were 446 patients who had FNAB of one single TN and 260 patients who had FNAB of 2 or more TNs. On FNAB, 705 of the 912 lesions (77%) were diagnosed as benign nodules, 30 (3.28%) were nondiagnostic samples, 78 (8.55%) were diagnosed as atypia of undetermined significance or follicular lesions of undetermined significance, 31 (3.39%) were follicular neoplasm or suspicious for follicular neoplasm, 27 (2.96%) were diagnosed as suspicious for malignancy, and 41 (4.49%) were diagnostic of malignancy.

Surgical resection was recommended for patients with FNAB results that were positive for malignancy or highly suspicious for papillary thyroid cancer (PTC), or follicular neoplasm, or follicular cells of uncertain significance. Out of the 30 nondiagnostic nodules, 9 were surgically resected because of worrisome sonographic features and 21 are being followed with a repeat biopsy and/or a thyroid US examination. In addition, some patients chose to undergo a surgical resection because of personal concerns or compressive symptoms. Of the 173 patients meeting the criteria for a surgical resection, 167 elected surgical resections.

Statistical analysis

Associations between the thyroid malignancy status and dichotomized potential covariates, including elastography, hypoechogenicity, isoechogenicity, color Doppler, power Doppler, solid lesion, complex lesion, irregular margins, isthmus location, calcifications, and subcapsular location, were determined using chi-squared tests. Bivariate associations between the malignant thyroid status and various cutoffs of elastography were established using a logistic regression. Multivariate associations between the malignancy status and the covariates were determined through a stepwise logistic regression. In addition, sensitivity, specificity, and positive and negative predictive values (PPV and NPV) were determined for the variables significant in the bivariate associations. All statistical analysis was conducted using SAS 9.3 (SAS Institute, 2011).

Results

The mean size of all TNs selected for the FNAB was 15 mm. There were 391 of all selected nodules that were <10 mm in size, and the remaining 521 nodules were ≥10 mm in size. The mean age of the male patients was 47.7 and the mean age of female patients was 48.5. There were a total of 158 nodules with an ES of 3, 156 nodules with an ES of 2, 565 nodules with an ES of 1, and 33 nodules with an ES of 0. A total of 167 patients underwent a thyroidectomy and 86 malignant nodules were discovered by either FNAB or surgical pathology. On final surgical pathology, 80 malignant nodules were confirmed in 71 patients. Of these, 45 were PTC, 32 were follicular variants of PTC, and 3 were follicular carcinoma. To date, the other six nodules that were malignant on FNAB have not had surgery. Among this group, four patients are awaiting surgery and two have refused surgery. These six patients were included in this study because of their FNAB pathology.

Figure 2 shows the number of nodules with each elastography category and the risk of malignancy for that category. The stiffest nodules (ES 3) carry the greatest risk of malignancy, while the most flexible nodules carry the least risk of malignancy. The bivariate analyses and their statistical significance are summarized in Table 1. There is also an association between malignancy and elastography with scores 0, 1, and 2 having fewer malignancies than score 3, while ES 3 is most indicative of the malignancy group (p<0.0001). In addition, thyroid malignancy was associated with hypoechoic nodule echogenecity (65.1% vs. 43.0%; p=0.0001), irregular nodule margins (58.1% vs. 22.5%; p=0.0001), isthmus location (15.1% vs. 7.8%, p=0.0193), presence of both macro- and microcalcifications (17.4% vs. 4.0%; p=0.0001), microcalcification level 2 (32.6% vs. 6.1%; p=0.0001), subcapsular location (64.0% vs. 34.1%; p=0.0001), and heterogeneous versus homogenous thyroid gland (51.2% vs. 40.2%; p=0.0492). In this study, vascularity (as measured by power Doppler or color Doppler) was not a significant risk factor for thyroid cancer (TC).

The results of logistic regression calculations for various values of elastography predicting malignancy are shown in Table 2. While grouping of 0 and 1 vs. 3 gives the greatest odds of elastography, with ES 3 predicting malignancy, the analysis excludes the ES 2 elastography group. Elastography grouping of 0 and 1 vs. 2 and 3 gives the optimal prediction.

Table 3 shows the results of the multivariate stepwise logistic regression model. In the final model, the presence of elastography (odds ratio [OR] 10.1 [95% confidence interval (CI) 5.5, 18.5]), hypoechogenicity (OR 3.3 [CI 2.0, 5.6]), isthmus location (OR 3.7 [CI 1.7, 8.1]), and microcalcification level 2 (OR 4.2 [CI 2.3, 7.8]) are all significantly associated with TC.

Table 4 gives the sensitivity, specificity, and predictive values of the significant predictors of malignancy. Hypoechogenicity, isthmus location, microcalcification level 2, and elastography (0, 1 vs. 2, 3; 0, 1 vs. 3; 0, 1, 2 vs. 3) all have high NPVs (94%, 91.3%, 93%, and 97.2%, respectively), while isthmus location and microcalcification level 2 have high specificity (92.3% and 93.9%, respectively). Elastography 0 and 1 vs. 3 is associated with higher sensitivity and specificity (77% and 85.2%).

The mean age of patients with malignant nodules was 43.8 years. Of the 86 malignant nodules, 39 were <10 mm and 47 were ≥10 mm. At the time this manuscript was submitted, among the ES 3 group, 57 TNs of the 158 (36.1%) were malignant nodules. Among the ES 2 group, 12 TNs of the 156 (7.7%) were malignant. In the ES 1 group, 16 TNs of the 565 (2.8%) were malignant. Among the ES 0 group, only one TN of the 33 (3%) was malignant. This patient had a very large complex lesion that was >80% cystic and the histology report showed a follicular variant of PTC. Figure 2 demonstrates the prevalence of TC in study patients stratified by the ES.

The sensitivity and specificity for TNs with ES 3 for the diagnosis of malignancy versus ES 1 and ES 0 was 77.0% and 85.2%, respectively. The PPV for this group was 36.1%, and the NPV was 97.2%. See Figures 3–5 for examples of malignant lesions with ES 3.

Discussion

The primary goal of this study was to evaluate the performance of real-time elastography alone and in conjunction with known sonographic characteristics of malignancy in an unselected group of patients with TNs. Elastography can be easily integrated into a routine US examination. Our results are encouraging and additional prospective studies are needed to confirm our findings.

The presence of microcalcifications in malignant nodules is often attributed to psammoma bodies in PTC and is frequently seen in medullary thyroid carcinoma. In vitro studies with various tumors show a 10-fold greater stiffness of malignant neoplasms compared with normal tissues (11).

This study evaluated the performance of SE alone and in conjunction with conventional US criteria for evaluating the risk of malignancy in TNs. Our results indicate that elastography alone has a PPV that is equal to that of microcalcification level 2 and greater compared with echogenicity and isthmus location. The NPV is similar for each of these criteria, in part, because most nodules are not malignant. Furthermore, multivariate regression analysis indicates that the elastography prediction is independent of conventional US criteria. Therefore, elastography provides additional new information for the evaluation of TNs and has the potential for being combined with conventional US criteria to provide a more accurate prediction of malignancy.

The strengths of this study include the prospective design, the inclusion of all nodules under investigation, and a sample size large enough to allow for comparison with conventional US characteristics by multivariate regression analysis. Other investigations have suggested that SE may be useful in the evaluation of TNs. Some of these studies included only a small subset of patients being evaluated for malignancy, so that the predictive values are not applicable to the general population of nodules. They do not have the statistical power to allow comparison of elastography to conventional US (10–16).

One large study (17) found grayscale US to be superior to elastography. Differences between this study and ours include the method of statistical analysis, the frequency of malignancy within the population, and the specific technology employed. Moon et al. compared elastography to US. We view these technologies as complimentary, not competing. Therefore, we used multivariate regression analysis to precisely define the key independent variables that predict malignancy. The risk of malignancy in patients in the study by Moon et al. was 30%, while ours was 11.40%, indicating that very different populations were being evaluated.

A recent article in Thyroid by Moon et al. (18) suggested that elastography was not superior to grayscale US. However, their study found a 25% cancer rate. They likely have such a high cancer rate due to their exclusion criteria, not including complex lesions with >20% cystic component or lesions containing macrocalcifications. The specific population of people created by their exclusion criteria may not be an accurate representation of the population as a whole. These two studies (17,18) used the same US system.

The conventional US characteristics of malignancy observed in this study are similar to those described by others and include the presence of microcalcifications, irregular margins, and hypoechogenecity (22,23), although there are some subtle differences. Vascularity did not quite reach statistical significance as a predictor of malignancy in the bivariate analysis, and it did not have a significant relationship with malignancy in the multivariate analysis. After multivariate analysis, subcapsular location, heterogeneous gland, and irregular margins were no longer predictors of malignancy.

We divided microcalcifications into two different groups. Those with level 1 microcalfications had ≤3 microcalcifications per nodule, while those with level 2 microcalcifications had ≥4 microcalcifications. Only level 2 microcalcifications were associated with an increased malignancy risk. To our knowledge, this stratification of microcalcifications has not been previously reported. Figure 3 shows an example of a 6-mm PTC in a hypoechoic lesion with microcalcifications and ES 3.

We included all nodules with macrocalcifications in our study. We had 57 TNs with macrocalcification, 8 (14%) of which were malignant. In the ES 3 group, 15 patients had macrocalcification and 7 (46.7%) had cancer (Fig. 4). In the ES 1 and ES 2 groups, we had 42 TNs with macrocalcification, but only one TN was malignant. In smaller TNs (<10 mm), the presence of macrocalcification/microcalcifications can falsely elevate ES and subsequent risk assessment for cancer.

We also carefully evaluated complex TNs in our elastography examination. There were 304 TNs that were described as complex nodules in this study. Of these, 23 (7.5%) nodules were malignant. Sixteen malignant complex nodules were in the >75% solid group and six malignant complex nodules were in the 50–75% solid group. One patient had cancer in the <25% solid group. There was no cancer in the 25–50% solid group. The presence of cancer was similar in complex TNs compared with solid TNs in the ES 3 group, 34.9% versus 36.5%. In the ES 2 group, 9.3% of complex lesions were malignant as compared to 6.9% of solid lesions. Among the ES 1 group, 1.1% of complex lesions were malignant versus 3.7% of solid lesions. Figure 5 is an example of an 18-mm FVPTC in a 35-mm complex TN.

It was surprising to find that nodules in the thyroid isthmus had a greater chance of being malignant than nodules found elsewhere in the thyroid gland. To our knowledge, this has not been reported. This was a small effect that might have been missed in other studies. At the present time, there is no obvious reason why isthmus nodules may have a higher risk for thyroid malignancy. It will be of interest to determine if this relationship is reproducible in other studies. Finally, since the conventional US findings are similar to those found in other studies, they act as internal controls to validate the expertise of the ultrasonographer (G.A.).

Elastography measures stiffness of the TN relative to the surrounding thyroid gland. Therefore, there are potential interferences that may reduce its accuracy. In Hashimoto's disease, the thyroid gland is usually stiffer than a normal thyroid gland (24). In this setting, elastography measurements may inaccurately estimate the risk of malignancy of TNs.

There can be considerable interobserver variability when elastography is used to evaluate TNs (25). It should be noted that all US examinations, elastography determinations, and FNABs were performed by a single clinician/investigator (G.A.) with documented skill and experience in TN evaluation (26,27). This individual had 15 months of experience with TN elastography before initiating the study. Therefore, it is anticipated that experience in elastography and thyroid US are necessary to reproduce the associations described here.

In summary, we investigated the performance of SE in an appropriately powered prospective study and demonstrated that the ES as a single variable and in conjunction with other known US criteria is associated with the risk of TC. We conclude that SE stratifies the malignancy risk of TNs independently and with a PPV equal to microcalcification and greater than other conventional US criteria associated with the malignancy risk. The NPV of elastography was greater than all conventional US criteria.

Acknowledgments

For performing the statistical analysis and for expert statistical advice, we acknowledge Maggie M. Kuchibhatla, PhD (Duke University, Durham, NC). For assisting in reviewing the literature, we would like to thank Dr. Annett Chua (University of Connecticut Health Center). We also thank the Wilmington Endocrinology staff, especially Brandy Lundberg and Selby Li, for their tireless effort and assistance in helping with this study.

Disclosure Statement

No competing financial interests exist.