Introduction

Minimally invasive parathyroidectomy (MIP) is becoming the first choice of treatment of primary hyperparathyroidism (PHP).

Preoperative localization of the parathyroid lesions is crucial in clinical practice in order to perform minimally invasive surgical procedures such as minimally invasive or video-assisted parathyroidectomy [1, 2]. Identifying uniglandular disease allows unilateral neck exploration and reduces operation time and surgical morbidity. High-resolution B-mode gray-scale ultrasound (US) is highly accurate in localizing enlarged hyperplastic parathyroid glands and parathyroid adenomas, while power Doppler ultrasound (PD-US) is a supportive technique [3]. The preoperative combination of US and 99mTc-sestamibi scintigraphy increases the sensitivity of enlarged parathyroid localization to 95 % [4, 5].

Strain ultrasound elastography (USE) is a relatively new technique that can be used as an additional tool to US with the aim of improving the evaluation of superficial tissues such as the breast, prostate, testis, neck, and thyroid; however, variable results have been reported [6–11]. Other adjacent organs such as parathyroids can also be evaluated if sufficient external pressure can be applied. It is well known that normal parathyroid glands consist of 40–70 % adipose tissue, while parathyroid adenomas are characterized by a lack of adipose tissue and have thickened capsules [12, 13]. Since the thyroid gland is mainly composed of thyroid follicles consisting of colloid-filled areas surrounded by principal cells, their appearance is soft on USE; this can be used as a reference point during the evaluation of parathyroid lesions [14, 15].

Unlütürk U. et al. recently published the first paper presenting the USE features of parathyroid lesions [16]. Parathyroid adenomas were demonstrated to be stiff on USE, while almost half the cases of parathyroid hyperplasia were soft. The authors concluded that USE imaging of a parathyroid lesion may help in determining the surgical approach and could improve postoperative outcomes.

Despite these advances in scientific knowledge, the differential diagnosis of parathyroid lesions from other neck lesions such as thyroid nodules or lymph nodes can be difficult in many cases [17]. To date, no papers have been published [18] on the use of quasi-static USE techniques such as Elastoscan in parathyroid lesions, and the recent EFSUMB guidelines do not include parathyroid lesion assessment among the recommended uses of USE.

The aim of our prospective study was to assess whether semi quantitative US-elastography with Elastoscan (ECI) improves US accuracy in distinguishing between benign and malign parathyroid lesions and between parathyroid lesions and ectopic thyroid nodules or lymph nodes.

Materials and methods

The institutional review board of the Policlinico Umberto I Sapienza University Hospital of Rome approved the protocol, and all subjects gave their written informed consent. This prospective study used the STARD checklist flow diagram and was carried out from January 2014 to December 2015 at Sapienza University of Rome.

The study population included 85 consecutive patients (36 men and 49 women; mean age 55 years, range 20–81 years) with hyperparathyroidism (HPT) as defined by Mayo Clinic Staff (repeatedly high circulating intact parathyroid hormone (PTHi), hypercalcemia or normal blood calcium, low-normal blood phosphate and increased or high-normal urinary calcium and phosphate in the presence of vitamin D sufficiency and normal renal function) and controversial US neck lesions, recruited from the outpatient clinics of Policlinico Umberto I, Sapienza University Hospital of Rome. Of these, 6 patients with sustained tachycardia were excluded due to the high probability of its impact on USE evaluation.

The remaining 79 HPT patients with a follow-up of at least 12 months and candidates for surgery were included in the study (Fig. 1). Of these, 18 patients presented concomitant chronic autoimmune thyroiditis, based on high titers of anti-thyroid antibodies (anti-TPO and/or anti-Tg) and normal or reduced thyroid function. US exam showed a diffuse hypoechoic thyroid structure with increased thyroid volume (Hashimoto goiter), while other cases had a normal thyroid volume.

Fig. 1
figure 1

Flow diagram for the study

Full size image

Equipment and data acquisition

After US examination and USE assessment, all patients underwent fine needle aspiration biopsy and consequent surgery for the suspicious or malignant specimen. Cytology and post-surgical results were considered as the reference standard (mean length of time between cytology and USE = 2.6 months, range 2.2–2.9 months).

Ultrasound evaluation was performed using US B-mode and color Doppler (CD) and quasi-static USE with ACCUVIX A30, RS 80 A (Samsung Medison Co. Ltd., Seoul, Korea). The study was carried out with a high frequency (10–18 MHz) linear US transducer placed gently on the patient’s neck in hyperextension while in the supine position.

All US examinations were performed by a sonographer (VC) with more than 10 years of experience. Unenhanced B-mode and color Doppler US features were scored by independent readers (AMI and EG, sonographers with more than 10 years of experience, and DD, a radiologist with 15 years of experience), who were blinded to the patient’s history, according to a predefined protocol.

Ultrasound features were categorized using a binary score to enable statistical processing.

In the comparison between the US characteristics attributed to parathyroid vs. thyroid nodules a score of 1 was assigned to the features which generally characterize “parathyroid” and 0 for “thyroid” features as follows:

  • Shape: Oval = 1; Round = 0;

  • Echogenicity: Hypoechogenicity (more hypoechoic than the surrounding muscles)  = 1; Iso-hyperechoic or hyperechoic = 0;

  • Cleavage from thyroid: Yes = 1; No = 0;

  • Vascularization: Peripheral (external, independent)  = 1; Peri-intralesional (internal vessel from thyroid parenchyma) = 0.

US features discriminating between parathyroid vs. lymph nodes were also scored according to a predefined protocol, with a score of 1 assigned to the features which generally characterize “parathyroid”, and 0 for “lymph nodes” as follows:

  • Oval and hypoechoic = 1; Round or oval with hilum = 0.

  • Vascularization: Peripheral (external)  = 1 for parathyroid; Hilum (central)  = 0 for the lymph node

Quasi-static US elastography

Patients were asked to hold their breath for 3–4 s. Quasi-static USE was performed using carotid pulsation as the compression source (in vivo compression), in order to assess tissue strain in the region of interest. Elastosonogram superimposition (placed in the left part of the screen) on B-mode images and use of a polychrome scale (color map) enabled the evaluation of the lesion margins and the surrounding healthy parenchyma. The examination was performed on the axial and longitudinal diameter with manual placement of an ROI (region of interest) on the lesion. The largest diameter of the suspicious lesion was included in the image. The quasi-static USE was considered optimal when the screen quality indicator (multiple boxes that change color from white to red, yellow or green) on the upper right turned green. Also on the upper right, a color scale identified hard and soft tissues as red and blue respectively. Consequently, the tissue under examination changed color according to its stiffness. The calculated elasticity contrast index (ECI) was always displayed on the monitor, providing a semiquantitative stiffness evaluation.

For all patients, ECI values for lymph nodes, parathyroid and thyroid lesions were compared with laboratory data, clinical evaluation or histology when available.

Two blinded sonographers (both with >5 years of experience) performed all US and USE examinations and stored the images and loops for subsequent analysis by two other blinded readers (again, both with >5 years of experience).

Statistical evaluation

Statistical analysis was performed with Stata software (Stata version 12.0, Stata Corporation, College Station, TX, USA). Continuous variables are reported as mean ± 1 SD. Single US features, US score and ECI alone and summed in comparison with histology results, sensitivity, specificity, PPV (positive predictive value) and NPV (negative predictive value) were related to histology and estimated through a 2 × 2 contingency table, and the significance of these distributions was analyzed with Pearson’s χ 2 test (threshold: p < 0.05). For US features and ECI values, box-plot graphs were used to evaluate the difference in distribution of strain ratio values between benign and malignant nodules; a Mann-Whitney U-test was performed to evaluate the significance of distribution differences between the 2 groups (threshold: p < 0.05).

The Elastoscan diagnostic performance was assessed using ROC curves that were constructed for prediction of the parathyroid lesion type. Optimal cut-off values were selected to maximize the sum of sensitivity (Se) and specificity (Sp). Data were collected prospectively and recorded by each radiologist in a spreadsheet (Microsoft Excel).

The sensitivity and specificity of US and Elastoscan were compared using Fisher’s exact test and ROC curves. Histopathological findings were used to evaluate effective sensitivity, specificity, positive predictive value, negative predictive value and accuracy, which were analyzed for statistical significance according to the χ 2 test.

Considering the organ, the disease and the spread of ECI values, which did not follow a bell-shaped distribution, a nonparametric test (Kruskal-Wallis) was used to evaluate the difference of the medians between more than two groups, adjusted with the Bonferroni correction. ANOVA test was performed to confirm the results of the non-parametric test. For each test, the results were statistically significant when the p value of the two-tailed test was less than 0.05.

Results

Histological examination of the cohort of 79 patients gave the following results: lymph node and thyroid group: 14 reactive lymph nodes, 16 benign thyroid nodules and 2 thyroid carcinomas; parathyroid group: 29 hyperplasia, 15 parathyroid adenomas and 3 parathyroid carcinomas (Table 1).

Table 1 Clinical data (median and interquartile range) of all patients and lesions according to the reference diagnosis
Full size table

The mean ECI values in the two groups were as follows:

  • - lymph node and thyroid: 1.12 ± 0.58 for reactive lymph node; 1.72 ± 0.66 for benign thyroid nodule; 3.01 ± 0.37 for thyroid carcinoma (see Table 2 and Fig. 2 panels C, D, E and F);

    Table 2 ECI data in neck lesions. Values are mean ± standard deviation (SD)
    Full size table
    Fig. 2
    figure 2

    Baseline US and qualitative elastography of neck lesions. Panels A, C, and E show the US scan respectively of a parathyroid adenoma, a reactive lymph node and a papillary thyroid cancer. Panels B, D, and F show the quasi-static USE and ECI for the same lesions

    Full size image

  • - parathyroid: 1.35 ± 0.61 for hyperplasia; 1.77 ± 0.54 for adenoma; 3.47 ± 0.77 for carcinoma (see Table 2 and Fig. 2 panels A and B).

In all, 18 thyroid lesions, 14 reactive lymph nodes and 47 parathyroid lesions were identified. A confidence interval of 95 % was used for the parameters below.

Parathyroid vs. thyroid

Shape correctly identified all the parathyroid lesions (47/47) and 9/18 thyroid lesions (Se: 100 %, Spe: 50 %, ROC area: 0.75; PPV: 83.9 %; NPV: 100 %).

Echogenicity identified 46/47 parathyroid lesions and 6/18 thyroid lesions (Se: 97.9 %, Spe: 33.3 %, ROC area: 0,656; PPV: 79.3 %; NPV: 85.7 %).

Cleavage identified 40/47 parathyroid lesions and 14/18 thyroid lesions (Se: 85.1 %, Spe: 77.8 %, ROC area: 0,814; PPV: 90.9 %; NPV: 66.7 %).

Vascularization identified 43/47 parathyroid lesions and 13/18 thyroid lesions (Se: 91.5 %, Spe: 72.2 %, ROC area: 0,819; PPV: 89.6 %; NPV: 76.5 %).

These results are summarized in Table 3.

Table 3 US score. 95 % CI are shown in parenthesis
Full size table

The optimal ECI cut-off was 1.28 (Se: 46 %, Spe: 77 %, ROC area: 0.605), identifying 22/47 parathyroid lesions and 14/18 thyroid lesions.

ROC curves were used to compare the sensitivity, specificity and accuracy of US, CD, and ECI.

Parathyroid vs. Lymph node

Shape identified all parathyroid lesions (47/47) and 3/14 lymph nodes (Se: 0 %, Spe: 78.6 %, ROC area: 0.393; PPV: 0%; NPV: 19 %).

Echogenicity identified 46/47 parathyroid lesions and no lymph nodes (0/14) (Se: 2.13 %, Spe: 100 %, ROC area: 0.511; PPV: 100 %; VPN: 23.3 %).

Cleavage identified 40/47 parathyroid lesions and no lymph node (0/14) (Se: 14.9 %, Spe: 100 %, ROC area: 0.574; PPV: 100 %; NPV: 25.9 %).

Vascularization identified 43/47 parathyroid lesions and 13/14 lymph nodes (Se: 91.5 %, Spe: 92.9 %, ROC area: 0.922; PPV: 97.7 %; NPV: 76.5 %).

These results are summarized in Table 3.

The ECI optimal cut-off was 1 (Se: 78 %, Spe: 71 %, ROC area: 0.705) (Fig. 3), identifying 37/47 parathyroid lesions and 10/14 lymph nodes.

Fig. 3
figure 3

ROC curve analysis of parathyroid pathology prediction. The diagnostic performance of Elastoscan was assessed using ROC curves. The optimal ECI cut-off value of 2.58 was selected to discriminate between parathyroid carcinoma and hyperplastic lesion or parathyroid adenoma

Full size image

ROC curves were used to compare the sensitivity, specificity and accuracy of US, CD and ECI.

Parathyroid

An ECI cut-off of 2.58 was the most accurate in discriminating between malignant and benign nodules (Se: 100 %; Spe: 95.4 %; ROC area: 0.973) (Fig. 3): 3/3 parathyroid carcinomas had an ECI above 2.58, while 43/44 hyperplastic lesions or parathyroid adenomas had an ECI below 2.58. Shape identified 3/3 parathyroid carcinomas and 44/44 benign hyperplastic lesions and parathyroid adenomas (Se: 100 %; Spe: 100 %; ROC area: 1.0; PPV: 83.9 %; NPV: 100 %). Margins identified 2/3 malignant lesions and 36/44 hyperplastic lesions and parathyroid adenomas (Se: 66.7 %; Spe: 81.8 %; ROC area: 0.742; PPV: 20 %; NPV: 97.3 %). Calcifications identified 2/3 parathyroid carcinomas and 44/44 benign parathyroid lesions (Se: 66.7 %; Spa: 100 %; ROC area: 0.833; PPV: 100 %; NPV: 97.8 %).

The results are summarized in Table 3. ROC curves were used to compare the sensitivity, specificity, and accuracy of the US, CD and ECI.

Discussion

The aim of our prospective study was to assess if semiquantitative USE with Elastoscan (ECI) improves US accuracy in distinguishing between normal and pathological parathyroid glands and in differential diagnosis against ectopic thyroid nodules and lymph nodes. We identified the optimal ECI cut-off values in the analyses of parathyroid vs. thyroid nodules and lymph nodes. ECI improved US accuracy in differentiating parathyroid lesions from lymph nodes. Cleavage and extra thyroidal vascularization on US improved differentiation between parathyroid lesions and thyroid nodules. Furthermore, analysis of parathyroid lesions revealed that the ECI was significantly higher in malignant than in benign lesions. Parathyroid carcinoma is in fact stiffer than thyroid carcinoma, as shown in Table 1. Analyzing our results we established a mean cut-off value of 2.58 to more accurately distinguish benign parathyroid diseases (hyperplasia and adenoma) from malignant diseases. Combining ECI index with conventional US (shape, margins, calcifications) and in particular with shape, sensitivity, and specificity reached 100 %.

Our results thus suggest the importance of B-mode and demonstrate that combining all main parameters of B-mode (echogenicity, cleavage, and vascularization) with ECI produces a marked gain in sensitivity in the differential diagnosis of neck lesions. This can clearly be seen in the thyroid and parathyroid analyses shown in Fig. 2, in which echogenicity, cleavage, and vascularization have a crucial role in discriminating between the different lesions, while ECI alone is inadequate. In fact, morphologic characteristics (shape, margin, echogenicity, and cleavage) and vascularization seen using B-mode identify and characterize thyroid and parathyroid lesions better than ECI.

In contrast, ECI discriminated between parathyroid and lymph nodes better than B-mode features, except for vascularization, which showed a good accuracy. The optimal ECI cut-off value of 1 showed high sensitivity and specificity in discriminating parathyroid lesions from lymph nodes. The combination of vascularization and ECI produced a further increase in sensitivity.

These data offer important advances in clinical practice. The accurate assessment of parathyroid diseases to enable the use of minimally invasive procedures such as minimally invasive parathyroidectomy or, in some cases, total video-assisted parathyroidectomy is one of the biggest challenges emerging in medical imaging [19]. Several authors have questioned the actual value of US as a first instance diagnostic tool [1, 2], especially when compared to more advanced techniques such as magnetic resonance imaging [20], computed tomography, [21–23] and 99m-MIBI scintigraphy [24–26] that is crucial in the diagnosis of primary hyperparathyroidism, due to the high percentage of patients with ectopic parathyroid lesions (such as of the mediastinum) where ultrasound is not feasible [4]. Recent years have seen the rise of multiparametric ultrasound, where CD [27] is joined by other diagnostic tools such as USE [14, 15, 28–33] and contrast enhanced ultrasound (CEUS) [34]. At present there are no specific guidelines for the use of USE or CEUS for the evaluation of parathyroid glands [18], and to date only three papers using USE in this specific setting have been published [16, 35, 36]. These studies found that USE may not only help distinguish between parathyroid and thyroid conditions [35], but might also be useful in the investigation of different parathyroid diseases [16] for differentiating parathyroid adenomas from benign and malignant thyroid nodules, including posteriorly located nodules [36]. Specifically, Azizi G. et al. [35] prospectively examined 42 patients underwent parathyroidectomy for a parathyroid adenoma showing that tissue elasticity for all parathyroid adenomas was significantly lower than thyroid tissue [median share wave velocity (SWV): 2.02 m/s (range 1.53, 2.50) vs. 2.77 m/s (1.89, 3.70)]. Authors explained that increased tissue stiffness in the thyroid gland may be due to inflammation and fibrosis often occurring in chronic thyroiditis.

Ünlütürk U. et al. [16] analyzing 93 parathyroid lesions undergoing parathyroidectomy, showed that all parathyroid adenomas had a greater stiffness, while most parathyroid hyperplasia (63 %) had significantly higher elasticity [median strain ratio (min–max) : 3.56 (0.47–60) vs. 1.49 (0.24–8.56)] [16]. Also in our study hyperplastic parathyroid lesions were more elastic than parathyroid adenomas, but statistically significant differences were measured between benign and stiffer malignant lesions.

In the differential diagnosis between parathyroid and thyroid lesions, our results are consistent with those sowed in the study of Batur A. et al. [36] that analyzing 21 parathyroid adenomas and 71 thyroid nodules (40 benign lesions; 31 malignant lesions) showed that parathyroid adenomas had higher stiffness levels compared to benign thyroid nodules (mean SWV ± SD: 3.09 ± 0.75 vs. 2.20 ± 0.39 m/s: p < 0.001) and significant lower stiffness levels compared to malignant thyroid nodules (mean SWV: 3.09 ± 0.75 vs. 3.9 ± 0.43 m/s: p < 0.001).

New insights could be arise from the use of 3-D ultrasound in distinguishing between parathyroid lesions from other neck lesions. Indeed nowadays, three-dimensional US was demonstrated to be highly accurate in diagnosing thyroid nodules and in the differential diagnosis between benign and malignant thyroid nodules [37].

Our study has some limitations. (1) The use of ECI was not effective in all patients: in 6 patients a reliable elastogram could not be obtained due to tachycardia, therefore these patients were not included in the study. (2) There is a limited caseload: a study on a larger population is necessary for greater statistical value. Further multicenter studies, and hopefully, international guidelines to be applied in the future are warranted.

In conclusion, our preliminary results have important implications for patient care: we demonstrated that quasi-static USE with ECI could be a useful additional tool to improve the preoperative US detection of parathyroid lesions, as well as in discriminating between benign and malignant lesions. The new 3-D probes might offer further improvements.