Association of Ultrasound Characteristics With Extranodal Extension in Metastatic Papillary Thyroid Carcinoma

Importance  The presence of extranodal extension (ENE) in metastatic papillary thyroid carcinoma (PTC) has emerged as an important prognostic factor, independently associated with tumor persistence after initial resection, decreased likelihood of complete biochemical response, and worse cause-specific survival. Therefore, the ability to determine ENE before surgery is desirable and advantageous but to date has not yet been evaluated.

Objective  To evaluate the use of preoperative ultrasound characteristics in predicting pathologic ENE in patients with metastatic PTC.

Design, Setting, and Participants  Single-institutional, retrospective analysis of patients with metastatic PTC between December 1, 2007, and May 31, 2012. The dates of the analysis were September 1, 2014, to July 31, 2015. Clinicodemographic and histopathologic data were extracted. Preoperative ultrasound images were scored for characteristics of interest by 2 independent radiologists masked to radiology and pathology reports. The setting was an academic tertiary care center. Patients were excluded if they were younger than 18 years, did not have clinical or imaging follow-up after surgery, were found to have histologic diagnoses other than PTC, or were being treated for recurrent disease.

Exposures  Preoperative ultrasound and neck dissection.

Main Outcomes and Measures  Association of ultrasound characteristics with ENE.

Results  Data from 29 patients with metastatic PTC and available preoperative ultrasound images and pathologic data from neck dissection were included. The patients had a median age at diagnosis of 47 years (age range, 19-85 years); and 76% (22 of 29) were female. Among 29 patients, 11 (38%) had ENE. There were no significant differences in distributions of clinicodemographic or histopathologic characteristics between patients with vs without ENE. The following ultrasound characteristics were significantly associated with ENE positivity: node matting (odds ratio [OR], 6.67; 95% CI, 1.01-44.10; P = .049), presence of node matting or cystic areas (OR, 11.70; 95% CI, 1.85-74.19; P = .009), composite score of 3 or more ultrasound characteristics (OR, 14.00; 95% CI, 2.06-95.09; P = .007), and presence of node matting, perinodal edema, unclear margins, or cystic areas (OR, 10.00; 95% CI, 1.05-95.24; P = .045), as well as presence of node matting, perinodal edema, or unclear margins (OR, 7.07; 95% CI, 1.17-42.85; P = .03). A composite score of 3 or more ultrasound characteristics had the highest accuracy (79.3%) for predicting ENE positivity.

Conclusions and Relevance  This study identifies preoperative ultrasound characteristics of metastatic PTC that are significantly associated with pathologically determined ENE and supports prospective analyses to further evaluate this use of preoperative ultrasound.

Papillary thyroid carcinoma (PTC) is the most common thyroid malignancy, with an estimated incidence in the United States of 0.1% per year, amounting to more than 300 000 new cases diagnosed annually.¹ Despite its high incidence, PTC has a low mortality rate, which makes risk stratification especially important when considering treatment planning.^2,3 Prognostic factors well established to predict worse survival include age 45 years or older, tumor size, extracapsular invasion, nodal spread, and distant metastasis.⁴ Extranodal extension (ENE) in metastatic PTC has emerged as an important prognostic factor. It is associated with distant metastasis and independently predicts eventual progression to systemic disease.^5,6 Furthermore, ENE is associated with increased nodal persistence after initial resection and with reduced complete biochemical response to radioiodine treatment and is an independent marker of poor survival.^5-7 In these studies, the presence of ENE has been determined during surgery⁸ or after surgery.^9,10

Given its prognostic value in predicting survival, the ability to determine ENE before surgery is desirable, but whether this assessment can be accomplished by ultrasound remains unknown, to our knowledge. At present, neck ultrasound is routinely used in the preoperative setting to evaluate for nodal metastatic deposits of thyroid malignancy. Therefore, ultrasound (with concomitant cytological sampling) is used to determine the need for lateral neck dissections at the time of total thyroidectomy. Well-established and specific sonographic findings are associated with nodal metastasis, including the presence of cystic areas or calcification within the node, the loss of the normal fatty hilum, a round shape of the lymph node, and hypervascularity.¹¹ However, specific sonographic criteria for ENE have not been evaluated in patients with PTC and ENE, to our knowledge.

Misselt et al¹² studied the ability of ultrasound to predict ENE in axillary lymph nodes in patients with biopsy-proven breast cancer. In that study, specific sonographic features of lymph nodes such as unclear margins, node matting, and perinodal edema were found to predict ENE with a high degree of specificity. Building on this knowledge, the aim of this study was to evaluate the use of preoperative ultrasound characteristics in predicting pathologic ENE in patients with metastatic PTC using these same criteria. Furthermore, because one distinct feature of PTC nodal metastasis is the presence of cystic areas within the lymph node, the predictive value of this finding was also studied.

Patients who had a preoperative diagnosis of regionally metastatic PTC (by imaging or biopsy) and who had both preoperative neck ultrasound and neck dissection at the time of thyroidectomy performed at The Johns Hopkins Hospital (JHH) (December 1, 2007, to May 31, 2012) were eligible for this study. The dates of the analysis were September 1, 2014, to July 31, 2015. All patients of surgeons performing endocrine surgery at the JHH during the period examined were included if they met the eligibility criteria. Patients were excluded from this study if they were younger than 18 years, did not have preoperative ultrasound imaging performed in the Russell H. Morgan Department of Radiology and Radiological Science at the JHH, or were without clinical or imaging follow-up after surgery. Patients who had histologic diagnoses other than PTC or who were being treated for recurrent disease were also excluded. This study was a single-institution, retrospective analysis entitled, “Effect of Thyroid Cancer Preoperative Neck Sonography on Surgical Approach to Neck Dissection,” which was approved by The Johns Hopkins Hospital institutional review board.

Retrospective medical record abstraction of eligible patients was performed. Clinicodemographic variables of interest included age, sex, and race. Histopathologic variables of interest included TNM classification and overall stage, as defined by the American Joint Committee on Cancer (AJCC),¹³ as well as the presence of ENE.

Ultrasound of the neck was performed in accordance with the standardized institutional protocol. Representative images of nodal stations 1 through 7, as defined by Som et al,^14,15 were obtained bilaterally. If sonographically suspicious nodes were identified, they were measured in 3 dimensions (sagittal, anteroposterior, and transverse) and interrogated with color Doppler ultrasound. In selected cases, ultrasound cine images of suspicious nodes were obtained and stored alongside the static images on a picture archive and communications system (Imorgan Ultrasound PACS; Imorgan Medical). Sonography of the neck was performed by experienced sonographers specifically trained and familiar with the standardized institutional protocol.

Ultrasound images were reviewed by 2 independent board-certified fellowship-trained radiologists, one with more than 30 years’ experience (U.M.H.) and the other with 2 years’ experience (S.F.C.), who were masked to the radiology and pathology report results. Each examination was initially reviewed by each radiologist separately. Suspicious nodes were evaluated for the presence of hilar replacement, hilar effacement, unclear margins, node matting, perinodal edema, and cystic change. The first 5 sonographic findings were used as defined by Misselt et al.¹² If the node had a specific feature, it was considered present (and scored as 1), and if it did not display this feature, it was considered absent (and scored as 0). This grading was performed for each sonographic criterion. If there was disagreement between the radiologists as to the presence or absence of a finding, the case was rereviewed, and a consensus score was achieved. Consensus scoring was used for the final analysis.

A composite score was created from the 6 ultrasound findings and defined simply as the number of characteristics present. Descriptive statistics and ultrasound characteristics were reported as the number and frequency for categorical variables or as the median and range for continuous variables. χ² Tests were used for categorical data, and Wilcoxon rank sum tests were used for comparison of the medians. Association of ultrasound characteristics with ENE was reported as odds ratios and 95% CIs. Binomial logistic regression was performed to analyze the relationships among histopathologic variables of interest, ultrasound characteristics, and ENE. Two-sided P < .05 was considered statistically significant. The data analysis was performed using a spreadsheet (Microsoft Excel 2010; Microsoft Corporation).

Clinicodemographic and histopathologic characteristics of the study population are summarized in Table 1. In total, there were 29 cases eligible for analysis. The median age of patients at diagnosis was 47.0 years (age range, 19-85 years). Twenty-two patients (76%) were of white race, and 22 patients (76%) were female. At initial clinical presentation, 28 patients (97%) had N1b nodal disease. Lateral and central neck dissections were performed in all 29 patients. After surgery, 11 patients (38%) were determined to have ENE. There were no significant clinicodemographic differences between patients with vs without ENE.

The frequency of ultrasound characteristics of interest is summarized in Table 2. The most common characteristic observed was hilar replacement, found in 21 patients (72%). Although the presence of cystic nodes in thyroid disease is pathognomonic for PTC, it was noted in only 9 patients (31%).

Ultrasound characteristics were compared by the presence or absence of ENE (Table 3). Hilar effacement, perinodal edema (Figure, A), and greater than 50% cystic areas (Figure, B) were more common in ENE-positive patients than ENE-negative patients, although nonsignificantly (P > .10 for all). Node matting (Figure, C) was found to be significantly associated with ENE-positivity (odds ratio [OR], 6.67; 95% CI, 1.01-44.10; P = .049).

To determine whether the presence of more than 1 ultrasound characteristic is associated with ENE, several groupings of features were considered (Table 3). The presence of node matting, perinodal edema, unclear margins, or cystic areas was associated with ENE (OR, 10.00; 95% CI, 1.05-95.24; P = .045). Similarly, the presence of node matting, perinodal edema, or unclear margins was strongly associated with ENE (OR, 7.07; 95% CI, 1.17-42.85; P = .03), as was the presence of node matting or cystic areas (OR, 11.70; 95% CI, 1.85-74.19; P = .009). When 3 or more ultrasound characteristics were observed (ie, composite score of ≥3), the odds of ENE were significantly increased (OR, 14.00; 95% CI, 2.06-95.09; P = .007).

For each ultrasound characteristic or grouping of characteristics that was found to be significantly associated with ENE on univariable analysis, a bivariable analysis was performed adjusting for the AJCC cancer stage. For all bivariable analyses, stage was not statistically significant, while the ultrasound characteristics remained significant.

Performance characteristics of the ultrasound descriptors were explored. Table 4 summarizes the statistical performance of ultrasound descriptors for predicting ENE. The sensitivity was highest (90.9%) for the variable that allowed the presence of any feature (presence of node matting, perinodal edema, unclear margins, or cystic areas). The absence of any of these characteristics also had the highest negative predictive value (90.0%). The specificity was highest (94.4%) for nodes observed to be greater than 50% cystic. The ultrasound observation of perinodal edema had the highest positive predictive value (83.3%) for ENE. The accuracy of most descriptors was 65% or greater, with hilar replacement and unclear margins having lower accuracy. The composite score of 3 or more ultrasound characteristics had the highest accuracy (79.3%).

In this analysis, preoperative ultrasound characteristics of nodal metastases were significantly associated with pathologically determined ENE. Although this investigation was a small retrospective pilot study, these data suggest that ultrasound may assist with preoperative risk stratification and warrants further evaluation.

Preoperative neck ultrasound for the detection of nodal metastasis in the central and lateral compartments has become the standard of care for patients with PTC. Beyond ascertainment of potential occult nodal metastases, preoperative neck ultrasound images have not been used to further characterize nodes (ie, evaluate for ENE), to our knowledge. Based on this analysis, we propose that ENE may be another clinically significant characteristic of nodal metastases that in the future can be determined before surgery pending further information.

The criteria investigated were based on a previously published study¹² of metastatic breast cancer that evaluated the performance of several sonographic features for predicting ENE. In breast cancer, features significantly associated with ENE included node matting, perinodal edema, unclear margins, and hilar replacement on univariable analysis, with both node matting and unclear margins being independently associated on multivariable analysis.¹² The additional ultrasound characteristic of interest evaluated in the present study was the presence of cystic areas, which is classically associated with PTC. However, other features may also be associated with ENE in metastatic PTC, including high lymph node burden.^16,17

The presence of pathologically determined ENE reflects a subset of PTC with biologically aggressive behaviors. Extranodal extension has been shown to correlate with posttreatment nodal persistence, an increased 10-year recurrence rate, and worse disease-specific survival.^5,6,9 One analysis found similar survival between patients with stage IVa or IVb PTC without ENE and patients with stage III PTC with ENE.⁹ The same study⁹ demonstrated that ENE is a stronger predictor of survival than location of metastatic lymph nodes (ie, N1a vs N1b). The results of these studies suggest that N1 disease represents a heterogeneous population with risk-stratification nuances not captured by the AJCC staging system,¹³ which does not account for the presence of ENE. Some authors have argued that ENE status differentiates patients with metastatic PTC into prognostically distinct groups and should be considered in future updates of the AJCC’s staging system for thyroid cancer,^6,9,18 which highlights the importance of this study. Such recommendations would be based on pathologic staging, but ultrasound could help in preoperative clinical staging and prognostication.

Preoperative knowledge of ENE would necessitate a comprehensive neck dissection at the time of surgery and definitely exclude any possibility of the American Thyroid Association’s guideline reference to “berry picking.”¹⁹ Given the prognostic implications of ENE, this subset would require meticulous and thorough removal of lymphadenopathy. Although the PTC patterns of lymphatic spread to the lateral neck are predictable, the need for central neck dissection remains an area of controversy. The present recommendations are for therapeutic central neck dissection for patients with evidence of suspicious or malignant nodes in the central compartment, with prophylactic neck dissection recommended for advanced primary tumors.²⁰ The morbidities of central neck dissection mandate caution, and the American Thyroid Association advises interpretation of guidelines based on available surgical expertise.²⁰ The determination of ENE on preoperative ultrasound may represent a clinical feature to improve preoperative risk stratification by helping to delineate the need for additional surgical expertise and evaluation for central neck dissection. Although our study evaluated the use of preoperative ultrasound in identifying lateral compartment nodes suspicious for ENE, one would expect the sonographic features associated with ENE studied herein would also apply to central compartment nodes.

To our knowledge, this study is the first analysis to identify ultrasound characteristics associated with ENE in PTC. These methods involved retrospective data collection, as is common for pilot analyses. Sonographic features such as node matting, unclear margins, and perinodal edema may be difficult to capture on static images compared with cine clips, requiring that the sonographer or sonologist be aware of these characteristics to image a node appropriately. As such, some of these features might have been noticeable on cine images (performed in only a small subset of the patients) had they been obtained, and the present study may underestimate the value of these features. In addition, further use of 3-dimensional ultrasound of the neck may be applicable in these patient populations. Although prospectively collected data might be more robust, ultrasound evaluations were reviewed by 2 experienced sonologists masked to clinical or pathologic data, who reached consensus for each criterion. The few included cases herein is a reflection of the eligibility criteria, which required imaging performed by the JHH department of radiology before surgery and after surgery, in addition to the availability of surgical pathologic data. While the clinical volume of patients is higher, clinical care does not require ultrasound by the department of radiology at the JHH. Many patients are seen with ultrasound images from outside centers and referring endocrinologists and may not follow up at the JHH for routine surveillance. While this limitation may introduce selection bias, the limited data set was useful for the present pilot study to determine whether the hypothesis is worth pursuing prospectively.

Our analysis provides compelling evidence for future studies to evaluate the use of ultrasound for detecting ENE before surgery. Whether it has a role as another preoperative risk-stratification tool requires further prospective investigations. In addition, because the clinical significance of ENE as a prognostic marker is also being reported in head and neck squamous cell carcinoma,²¹ this analysis provides preliminary evidence that it is worth investigating this hypothesis in future prospective studies.

This pilot study identifies characteristics of metastatic PTC on preoperative ultrasound that are significantly associated with pathologically determined ENE. Node matting was significantly associated with ENE. In addition, the presence of 3 or more ultrasound characteristics was significantly associated with ENE. Nodes with greater than 50% cystic content were found to be highly specific for ENE. The absence of node matting, unclear margins, perinodal edema, and cystic change had a high negative predictive value when considering pathologic ENE. These data provide strong justification for future analyses to further evaluate this use of preoperative ultrasound of the neck nodes in patients with PTC. Whether ultrasound has a role as another preoperative risk-stratification tool requires further prospective investigations.

Submitted for Publication: September 28, 2015; final revision received November 11, 2015; accepted December 2, 2015.

Corresponding Author: Carole Fakhry, MD, MPH, Department of Otolaryngology–Head and Neck Surgery, The Johns Hopkins University School of Medicine, 601 N Caroline St, Johns Hopkins Outpatient Center, Sixth Floor, Baltimore, MD 21287 (cfakhry@jhmi.edu).

Published Online: February 4, 2016. doi:10.1001/jamaoto.2015.3558.

Author Contributions: Drs Qualliotine and Fakhry had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study concept and design: Coquia, Fakhry.

Acquisition, analysis, or interpretation of data: Qualliotine, Coquia, Hamper.

Drafting of the manuscript: All authors.

Critical revision of the manuscript for important intellectual content: All authors.

Statistical analysis: Qualliotine.

Administrative, technical, or material support: Fakhry.

Study supervision: Fakhry.

Conflict of Interest Disclosures: None reported.