Central and lateral cervical compartments.
Soler ZM, Hamilton BE, Schuff KG, Samuels MH, Cohen JI. Utility of Computed Tomography in the Detection of Subclinical Nodal Disease in Papillary Thyroid Carcinoma. Arch Otolaryngol Head Neck Surg. 2008;134(9):973-978. doi:10.1001/archotol.134.9.973
Copyright 2008 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.2008
To characterize the ability of computed tomography (CT) to identify subclinical cervical metastatic disease in papillary thyroid carcinoma (PTC).
Tertiary academic center.
Consecutive patients undergoing neck dissection for PTC between July 1, 2004, and July 1, 2006.
Preoperative CT scans were reevaluated in a blinded fashion by a single head and neck radiologist. Positive criteria included node with size larger than 10 mm, round shape, calcification, cystic character, or abnormal enhancement.
Main Outcome Measure
Sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) were calculated separately for central compartment (level VI) and lateral compartment (levels III and IV) dissections.
One hundred four patients underwent selective neck dissection for PTC during the study period. Forty-three patients had disease that involved primary lymphadenectomy at the time of thyroidectomy, and 61 had persistent or recurrent disease. There were 111 lateral compartment dissections and 145 central compartment dissections. The overall sensitivity was 59% for both the central and lateral compartments, and the specificity was 76% and 71%, respectively. The PPV and NPV were 84% and 47% for the central compartment and 73% and 57% for the lateral compartment, respectively.
Computed tomography has a limited capability to identify subclinical metastatic cervical disease in PTC, with a sensitivity near 60% and an NPV near 50%. Sole reliance on CT findings will miss a significant portion of disease likely because of the high incidence of microscopic foci. However, using strict criteria, a positive finding on a CT scan provides useful information because it predicts with a fairly high assurance that disease will in fact be found in a specific compartment during surgical dissection.
The role of lymph node dissection in the management of papillary thyroid carcinoma (PTC) remains controversial given the overall excellent long-term survival and lack of documented survival benefit with prophylactic lymphadenectomy. When one considers removal of subclinical (nonpalpable) metastases, particularly in the central compartment, questions arise concerning the ability of available imaging studies to localize disease adequately for a curative surgical intent. Although computed tomography (CT) is routinely used to guide surgical planning in head and neck squamous cell carcinoma (HNSCC), its utility in PTC is less established. This is reflected in the 2006 American Thyroid Association Guidelines Taskforce recommendation in favor of preoperative imaging with ultrasonography.1 The taskforce recommended not using preoperative imaging with CT, noting that its sensitivity in the setting of PTC remains unknown. Despite the lack of data regarding its efficacy, CT offers several potential advantages over ultrasonography. Computed tomography provides static anatomical information in a fully retrievable format. This facilitates subsequent intraoperative correlation with surgical landmarks. The acquisition of images is also less operator dependent compared with ultrasonography, promoting reproducibility. Computed tomography may excel in specific settings, such as patients with a thick neck or for visualization of the low paratracheal nodes, which may be obscured by the sternum or tracheal air shadow when ultrasonography is used. In addition, the anatomical levels that form the basis of the current surgical technique have been formalized using CT-based criteria,2 which promotes consistent communication between the radiologist and the surgeon in regard to disease location. Disadvantages of CT include increased cost and radiation exposure.
Because of the potential advantages, it is important to determine the accuracy of CT in localizing metastatic disease. The aim of this study was to determine the ability of CT to accurately identify nonpalpable cervical metastatic disease from PTC and identify the factors that influenced its accuracy.
A retrospective review was performed of all patients seen in a multidisciplinary thyroid tumor clinic who had undergone neck dissection as part of the management of their PTC between July 1, 2004, and July 1, 2006. Prior approval was given by the hospital institutional review board. Patients were originally referred for either newly diagnosed PTC or suspected persistent and/or recurrent disease. All patients underwent cervical CT before their operation to identify nodal groups at risk of involvement by thyroid cancer. This included images of 3-mm-thick sections from the skull base to the carina using 1 of 2 multidetector scanners (Philips Brilliance 64-slice multidetector CT or Philips Brilliance 16-slice multidetector CT; Philips Healthcare, Andover, Massachusetts). Intravenous contrast was administered when the use of radioactive iodine (RAI) was not anticipated in the immediate postoperative period.
In regard to both radiologic examination and operative planning, the neck was considered to have 4 compartments, including the left and right lateral compartments and the left and right central compartments (Figure). The lateral compartment consisted of the lymph nodes along the jugular vein from the level of the hyoid superiorly to the level of the clavicle inferiorly (levels III and IV). The level VI compartment was split into a right central compartment and a left central compartment. Each central compartment included the paratracheal nodes from the lateral margin of the carotid artery to the midline of the trachea and the cricoid superiorly to the innominate artery inferiorly (level VI).
For newly diagnosed cancers, the ipsilateral central compartment was always dissected electively at the time of thyroidectomy in addition to any other compartment believed to be involved according to imaging results. Suspected recurrent or persistent disease was based on elevated thyroglobulin levels and/or imaging in cases in which antithyroglobulin antibodies interfered. Imaging findings were used to guide surgical planning. Principles guiding the surgical approach included recognition of the high incidence of initial central node metastases and the high incidence of bilateral involvement when the ipsilateral nodes contain disease, a significant but lesser initial involvement of lateral nodes, evidence that “skip” metastases (eg, lateral disease without central disease) are rare, and poor efficacy of “berry picking.” Because of the high incidence of microscopic disease and the difficulty with subsequent operations, we tended to be inclusive rather than exclusive in the decision making and were likely to extend the operation at the time of surgery if unexpected disease was identified in a compartment. Any compartment that, according to imaging results, contained a suspicious node was dissected completely within the mentioned anatomical limits. Surgical planning involved a systematic review for each patient of the location of the original tumor, prior lymph node resection and involvement, evaluation with anatomical imaging (usually CT), and occasional functional imaging (positron emission tomography [PET] with fludeoxyglucose F 18) to attempt to identify residual thyroid tissue or disease. Applying these principles, selective neck dissections were generally planned to include all compartments with evidence of disease and the compartments on either side if they had not previously undergone complete dissection. For example, if the right central compartment was involved, the right lateral and left central compartments were also dissected; if both central compartments were involved, all 4 compartments were dissected. Extending surgery beyond the compartment with known disease represents an aggressive approach. Surgery was performed with the understanding that no current evidence supports improved survival with prophylactic neck dissection, although it may diminish regional recurrence.
For the purposes of this study, CT scans were retrieved retrospectively and then reevaluated in a blinded fashion by a single head and neck radiologist (B.E.H.) to identify metastatic cervical nodes. Positive criteria included any 1 of the following features: size greater than 10 mm, spherical shape, calcification, cystic character, or abnormal enhancement. Results were reported by formal neck compartment level.2 The level VI compartment was split at the midline and reported separately as right and left. To identify subtleties of diagnosis, the radiologist was also asked to classify each compartment as “positive,” “probable,” or “negative.” This analysis was independent of the criteria mentioned in the preceding paragraphs and represents the radiologist's general impression. The positive and probable classifications were then combined and considered a “radiologic positive” result. These findings are reported separately.
Neck dissections were performed by a single surgeon (J.I.C.) using the compartment-oriented approach. Surgical specimens were separated into individual nodal levels at the time of dissection and then given to the pathology laboratory. In the surgical pathology laboratory, each anatomical level was serially sectioned, and a representative complete cross-section of each candidate lymph node was submitted for histopathologic review. The surgical pathologic findings were then correlated with the CT findings. This correlation was performed on a compartment basis. The levels III and IV compartments were grouped together and considered the lateral compartment. The level VI compartment was split at the midline into a right and left central compartment. Sensitivity, specificity, PPV, and NPV were calculated separately for the central compartment (level VI) and the lateral compartment (levels III and IV) dissections. Subgroup analysis was performed comparing primary disease vs recurrent and/or persistent disease, prior RAI use vs being RAI naive, and contrast-enhanced CT vs noncontrast CT. We used χ2 tables to test for differences between subgroups. P ≤ .05 was considered statistically significant.
One hundred four patients underwent selective neck dissection for PTC during the study period. All patients had preoperative CT scans available for review. Contrast was used in 33 patients, with the remainder undergoing noncontrast CT. The average patient age was 46.5 years. Four patients had clinically palpable disease in a total of 5 compartments. In each case, the CT scan demonstrated true-positive results. These compartments were then excluded from further analysis. Forty-three patients had primary lymphadenectomy at the time of thyroidectomy, and 61 had persistent or recurrent disease. There were 111 lateral compartment dissections and 145 central compartment dissections. The overall sensitivity was 59% for both the central and lateral compartments, and the specificity was 76% and 71%, respectively. The PPV and NPV were 84% and 47% for the central compartment and 73% and 57% for the lateral compartment, respectively. The results of subgroup analysis and the radiologist's general impression are reported in Table 1. A statistically significant difference was found in the lateral compartment when comparing patients with prior RAI use with RAI-naive patients (P = .01) (Table 2). Specifically, the CT scan demonstrated higher sensitivity and PPV in patients with prior RAI use. Similarly, a higher sensitivity and PPV were seen in patients who underwent surgery for recurrent and/or persistent disease (P = .01) (Table 2). Compared with using strict criteria, the radiologist's general impression was more sensitive (84%) but less specific (36%) for the central compartment (P = .001) compared with the objective criteria.
Computed tomography is commonly used to identify metastatic cervical disease in HNSCC. In HNSCC, the practice of comprehensive (levels I-V) neck dissection provided substantial pathological data to correlate with the preoperative CT findings, which has allowed validation of criteria such as size and shape and determination of sensitivity and specificity.3 Sensitivity in this regard is approximately 80% and represents an improvement over clinical examination. Elective comprehensive neck dissection has not been widely advocated for PTC given the overall excellent prognosis. As such, pathological data have not been readily available to correlate with preoperative CT findings. The utility of CT to identify PTC metastatic to cervical nodes thus remains poorly characterized. The perfect study design to answer this question would involve obtaining CT scans from all patients who present with PTC, followed by surgical removal of levels I through VI metastases. This is the only design that would define both the prevalence of disease and the accuracy of CT to detect such disease. However, given the overall good prognosis of PTC, such a study would obviously not be ethically justifiable.
The recent shift toward the surveillance of patients with PTC through the use of serum thyroglobulin measurement and imaging has led to an increase in the removal of subclinical (nonpalpable) nodes. A selective neck dissection is usually performed, wherein the lymph nodes thought to be involved are removed specifically and those not believed to be involved are left in situ. Rarely is a dissection undertaken in a comprehensive (levels I-V) manner for limited disease. It is within this clinical paradigm of selective neck dissection for subclinical disease that our study took place. The selective nature of lymph node removal introduces a certain degree of bias, likely in favor of a higher prevalence of metastases in the compartments dissected. This limitation is an inherent factor of the study and must be taken into consideration when interpreting statistics such as PPV and NPV, which are influenced by the prevalence of disease. Studies attempting to evaluate ultrasonography, magnetic resonance imaging, and combined PET and CT in the setting of PTC also suffer from this same limitation.4- 7
Research that seeks to measure the performance of an imaging modality must define what exactly constitutes a positive test result. Unlike HNSCC, widely accepted criteria do not exist to define lymph node positivity in PTC. For the purposes of this research, a positive test result was defined as the presence of any 1 of the following findings: size larger than 10 mm, spherical shape, calcification, cystic character, or abnormal enhancement. These criteria are based on findings reported by Som et al4,8 in their review of CT findings in well-differentiated thyroid carcinoma and those used in similar studies. Further study is necessary to determine whether these criteria maximize the utility of CT scanning or whether additional criteria, such as clustering of nodes or a combination thereof, would be superior.
In this study, we chose to report the findings based on compartments rather than individual levels. This was done to mimic clinical practice wherein a compartment is usually dissected in its entirety. This has been referred to as a compartment-oriented surgical approach. Levels III and IV were grouped together in this study and considered the lateral compartment. Typically, lymph nodes at these levels are dissected together as a single unit because they represent the first echelon of lateral nodal spread in thyroid carcinoma.9,10 Grouping them together for analysis serves to minimize imprecision inherent in separating these levels on the operating room table. Although level IIA is considered part of the lateral compartment, this level was not included in the review. Few patients had surgical dissection of lymph nodes at this level (unless there was macroscopic involvement of level III), and thus scant pathological data were available to correlate with the CT findings.
The overall sensitivity of CT to detect metastatic lymphadenopathy was approximately 60% for both the central and lateral compartments. Sole reliance on CT findings would thus miss 40% of metastatic disease. This low sensitivity is likely explained by the high prevalence of micrometastasis.11,12 Overall, occult cervical metastases are estimated to occur in 25% to 90% of patients.13,14 This wide variation is likely explained by differences in patient populations and the extent of surgical dissection. Microscopic foci of PTC are unlikely to significantly change the size or architecture of a lymph node as visualized by CT.3 Thus, if the goal of preoperative imaging is to identify the entirety of microscopic disease, then CT suffers from inherent limitations. This limitation, however, likely applies to all available imaging modalities, including ultrasonography.4,5 In perhaps the largest review of preoperative imaging followed by modified radical dissection, Ito et al6 demonstrated a sensitivity of only 39% for ultrasonography. Kouvaraki et al7 found the sensitivity of ultrasonography to be 52% for the central compartment and 77% for the lateral compartment. With its combined functional and anatomical information, PET-CT holds theoretic promise for improved sensitivity in PTC. However, 1 prior study demonstrated a sensitivity of 66%, whereas another showed no difference compared with ultrasonography or CT.4,5 It must be kept in mind that making comparisons of imaging modalities across studies is problematic given differences in patient populations, practice patterns, and imaging technique.
In the present study, the overall specificity of CT was 76% for the central compartment and 71% for the lateral compartment. Thus, a positive finding on a CT scan is unlikely to represent a false-positive result. This may be of clinical utility, guiding the surgeon toward certain compartments likely to harbor disease. The PPV was fairly high at 84% and 73% for the central and lateral compartments, respectively. The NPV was low at 47% and 57%, respectively. This finding probably also reflects the high incidence of metastatic lymphadenopathy because the PPV and NPV are influenced by the prevalence of disease. Again, a positive finding is unlikely to represent a false-positive result and will likely lead to a fruitful surgical dissection, but a negative finding misses a significant portion of disease.
Subgroup analysis was performed to determine whether certain variables, such as prior surgery, prior RAI, or contrast enhancement, affect CT utility. In patients with an existing thyroid gland, the thyroid itself could potentially obscure adjacent nodes because of its location and natural enhancement. However, no appreciable difference was found in CT accuracy for the central compartment between patients with a history of prior thyroidectomy and those without. In addition, RAI could theoretically alter nodal appearance and CT accuracy. If this was the case, one would expect to find differences in both the central and lateral compartments. The only statistically significant differences were found in the lateral compartment when comparing patients who disclosed prior RAI use with those who were RAI naive (P = .01) and patients with primary disease vs those with recurrent and/or persistent disease (P = .01) (Table 2). The sensitivity and PPV were higher in groups with a history of prior RAI use and those presenting with recurrent and/or persistent disease. In reality, these 2 subgroups are not independent of one another. Often, patients presenting with recurrent or persistent disease had undergone prior RAI therapy. It is possible that these differences are not related to the imaging technique but instead are a function of a higher pretest probability of a positive finding in patients with recurrent disease as opposed to those presenting with a new diagnosis. A significant difference was not found between contrast-enhanced CT and noncontrast CT. However, the sensitivity of contrast-enhanced CT for the lateral compartment was 75% vs 47% for noncontrast CT. Given the small number of patients undergoing contrast-enhanced scans, the possibility of a type II error exists and definitive conclusions should not be made.
Only 4 patients had palpable cervical nodes at the time of their preoperative imaging for a total of 3.9%. This number is low compared with prior series that report palpable nodes 10% to 30% of the time.15 This number is probably a reflection of the new paradigm of PTC follow-up, with emphasis on chemical markers such as thyroglobulin level to define recurrence. In this setting, persistent or recurrent disease is identified much earlier compared with clinical examination alone. In each case of clinically palpable disease, the CT scan demonstrated positive findings in the appropriate nodal compartment. These compartments were excluded from further analysis.
The radiologist's general impression is worth noting. The radiologist was asked to call the study positive, probable, or negative. This portion was included to capture subtleties of radiologic diagnosis perhaps missed by strict criteria. This may more closely mimic actual clinical practice, wherein the impression of the radiologist carries substantial weight. If one considers both study results labeled positive and probable as a positive study, then the sensitivity was increased to 84% in the central compartment. This was statistically significant compared with the use of strictly defined criteria (P = .001). However, this increase is at the expense of specificity, which decreased to 36%. Thus, a conservative reading will pick up more disease but also lead to additional false-positive results. This further supports the idea of using strictly defined criteria to delineate positivity in CT of PTC.
Although not the aim of this study, the overall question remains whether preoperative CT plays any role in the management of PTC. The reality is that metastatic nodes likely fall into 3 categories. The first category includes large, palpable nodes readily picked up on physical examination. The other extreme is microscopic foci, which are likely to be discovered only on final pathological testing. The second group represents nodes that are abnormal enough to be picked up on imaging but too subtle to be noticed on clinical examination. In HNSCC, this intermediate group represents approximately 40% of occult metastatic nodes.16 Without comparisons of CT findings with comprehensive (levels I-VI) dissections, it is impossible to accurately define this number for PTC. In our series, physical examination results were positive in only 3.9% of patients. Disease considered positive according to CT findings but not detected on physical examination was found in 157 of 257 compartments (61.1%). Kouvaraki et al,7 in their review of preoperative ultrasonography, found ultrasonography-positive disease not detectable by physical examination in 52 of 151 compartments (34.4%). Given the overall excellent prognosis of PTC, it is debatable whether most patients require imaging to identify this intermediate group. Third, certain patient subgroups display a significant risk of disease recurrence, with increased morbidity and mortality. Wang et al11 have suggested that attempts to detect and remove occult lateral cervical lymph node metastases might be considered in these high-risk subgroups. Perhaps it is these high-risk patients for whom imaging such as CT may be particularly useful to guide surgical planning.
In conclusion, CT has a limited capability to identify subclinical metastatic cervical disease in PTC, with sensitivity near 60% and NPV near 50%. Sole reliance on CT will miss a significant portion of disease, likely because of the high incidence of microscopic foci. However, by using strict criteria, a positive finding on CT provides useful information because it predicts with a fairly high assurance that disease will in fact be found in a specific compartment during surgical dissection.
Correspondence: James I. Cohen, MD, PhD, Department of Otolaryngology/Head and Neck Surgery, Oregon Health & Science University, 3181 SW Sam Jackson Park Rd, Campus Box PV-01, Portland, OR 97239-3098 (email@example.com).
Submitted for Publication: October 3, 2007; final revision received October 28, 2007; accepted October 30, 2007.
Author Contributions: Drs Soler, Samuels, and Cohen had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Soler, Schuff, and Cohen. Acquisition of data: Soler. Analysis and interpretation of data: Soler, Hamilton, Schuff, Samuels, and Cohen. Drafting of the manuscript: Soler and Cohen. Critical revision of the manuscript for important intellectual content: Hamilton, Schuff, Samuels, and Cohen. Statistical analysis: Soler. Administrative, technical, and material support: Soler. Study supervision: Hamilton, Schuff, Samuels, and Cohen.
Financial Disclosure: None reported.
Previous Presentation: This study was presented at the American Head and Neck Society Annual Meeting; April 28, 2007; San Diego, California.