 |
|
|
|
|
|
 |
 |
 |
 |
 |
|
Thyroid Fine Needle Aspiration Cytology
Follicular Lesions and the Gray Zone
Jonathan Somma, M.D., Nicolas F. Schlecht, Ph.D., Daniel Fink, M.D., Samer N. Khader, M.D., Richard V. Smith, M.D., and Antonio Cajigas, M.D.
Objective
To identify categories of potentially neoplastic, nonpapillary follicular lesions of the thyroid that would reduce the number of lobectomies for nonneoplastic disease.
Study Design The literature regarding fine needle aspiration (FNA)
of follicular lesions is difficult to interpret largely due to the poor
standardization of diagnostic categories. Based on our categorization
scheme, which is directly comparable to the new system proposed by the
National Cancer Institute (NCI), we quantitatively evaluated the
cytomorphologic features of 99 “atypical” cases.
Results Histologic evaluation revealed a 63% neoplasia rate and a 26% malignancy rate. Lesions with an abundance of macrophages and colloid on FNA were significantly less likely to be neoplastic or malignant. The strongest predictors of any neoplasia included absence of colloid and increased cellularity, which remained significant even after adjustment for all selected variables.
Conclusion Unlike similar studies, ours revealed statistically significant variables that may assist in determining how to manage follicular lesions better. Further, our current diagnostic scheme is equivalent to that recently proposed by the NCI, and, herein, we not only illustrate its application in clinical practice but also highlight potential pitfalls of this system. (Acta Cytol 2010;54:123–131)
Keywords: aspiration biopsy, fine-needle; “atypical”; follicular neoplasm; macrophage; thyroid.
 Our current diagnostic system at MMC works well for us and our clinicians only because of a solid history of interdepartmental communication. Thyroid nodules are common clinically; the vast majority of them are hyperplastic or colloid nodules in a multinodular goiter. It is because their malignancy rate may be as low as 5% that fine needle aspiration (FNA) of thyroid lesions has emerged as a necessary test to triage cases for either conservative management (e.g., watchful waiting) or excision.1,2 In its screening function for nonpapillary follicular lesions (i.e., multinodular goiter, follicular adenoma and follicular carcinoma), it is well accepted that FNA cannot distinguish between the follicular neoplasms, including the oncocytic (Hürthle cell) variants. As a consequence, it is expected that most excised follicular neoplasms will be benign on histology.
The use of diagnostic categories is central to the practice of cytopathology. Although they are critical for conveying the significance of a result to clinicians, in the case of follicular lesions of the thyroid, such categories are poorly standardized.3,4 Classification systems have been proposed with up to 6 or more categories that are meant to convey the degree of certainty regarding the presence of a significant follicular lesion.3 In other words, these categories are meant to deal with the known “gray zone” in thyroid FNA that exists due, in part, to a cytopathologist’s level of experience as well as the nature of these aspirates.5 Such schemes use various terms, including “atypical,” indeterminate and suspicious, and pathologists often also employ indefinite modifiers, like favor, cannot exclude, possible, or probable paired with diagnostic vocabulary including goiter, neoplasm and carcinoma. As clearly demonstrated by Redman et al,6 these systems do not clarify patient management and, additionally, cause confusion when patients are referred between institutions.
The literature regarding follicular lesions of the thyroid is understandably difficult to interpret due to the wide range of categories as well as study designs used by different authors. This prompted us to take a look at our experience at Montefiore Medical Center (MMC), a large (1,122-bed), urban, academic tertiary care facility, to see if we could shed any light on the use of diagnostic categories for follicular lesions, particularly for the gray zone. In addition, we hypothesized that, using cases which had been classified as “atypical,” we would be able to identify cytomorphologic features that would distinguish between nonneoplastic and neoplastic histologic outcomes. If so, FNA diagnoses could be refined to decrease the number of lobectomies for nonneoplastic disease.
Materials and Methods At MMC, the Cytopathology Laboratory receives thyroid FNA specimens that are generally obtained using multiple passes per lesion with 25-gauge needles with and without ultrasound guidance by a diverse group of physicians, including pathologists, otorhinolaryngologists, endocrinologists and radiologists. Cases include ethanol-fixed, Papanicolaou-stained smears and, depending on the physician who performed the aspirate, may also include a cytospin or cell block from needle rinses. If the FNA is pathologist performed or if microscopic adequacy assessment is provided, air-dried smears stained by a Romanowski- type method are prepared.
A retrospective search of the laboratory information system database was undertaken to identify 100 consecutive patients who had had a thyroid FNA collected on or before May 10, 2006, with a diagnosis of “atypical” and who subsequently underwent surgery. There were 5 full-time, board-certified cytopathologists signing out cases during the study period, all of whom completed their pathology residency in accredited United States programs and cytopathology fellowship training at MMC. Our consensus practice is that a cytologic category of “atypical” should include all potentially neoplastic follicular lesions that do not show overt papillary features with the expectation that these patients will be offered a hemithyroidectomy in the appropriate clinical setting. In general, “atypical” aspirates are relatively cellular specimens with scant colloid and frequently show uniform-appearing, crowded follicular cells arranged in 2-dimensional microfollicular structures (or 3-dimensional syncytial groups), each consisting of <15 cells and forming a spherical structure that is at least two-thirds complete. Importantly, nonpapillary follicular neoplasms are not categorized as “suspicious” or “positive”; these categories are reserved for cases that probably or definitely represent various malignancies.
Institutional review board approval was obtained, and all cases were rereviewed by 1 senior cytopathologist without knowledge of the detailed cytology or subsequent histology diagnoses in order to grade cytologic smears on a scale (0–3+) for colloid, cellularity, sheets of follicular cells (including macrofollicles), microfollicles and macrophages.
The grading scale was defined as follows. For colloid, 0 = no colloid, 1+ = <25%, 2+ = 25–50%, and 3+ = >50%, where % refers to percent of the diagnostic area of the slides. For cellularity, sheets of follicular cells and microfollicles, 0 = none, 1+ = <33%, 2+ = 33–66%, and 3+ = >66%. For macrophages, 0 = no macrophages, 1+ = <7 macrophages, 2+ = 7–14, and 3+ = >14; the number of macrophages was measured using 15
 oculars in five 40 fields containing the highest density of macrophages in each case. In addition, the presence or absence of nuclear atypia, including enlargement, crowding, grooves, inclusions, hyperchromasia, pleomorphism and/or prominent nucleoli, was noted. Nuclear:cytoplasmic ratio (N:C) was evaluated as normal, mildly increased or markedly increased, and the degree of nuclear enlargement relative to the size of a red blood cell was also noted.
Contingency tables were constructed from the categorical and ordinal data reflecting clinical and pathological correlates of disease, including demographic characteristics (age and gender) and cytologic characteristics (as above). Disease status was defined by histologic review to confirm the presence of neoplasia in the thyroid lesions. Pearson’s
 2 and Wilcoxon rank-sum exact tests were used to assess the relation between disease status and cytologic characteristics. To identify determinants of neoplasia or malignancy in “atypical” thyroid FNAs, we used logistic regression with stepwise selection. Parameters were evaluated against detection of malignant phenotype and presence of neoplasia separately. Cutoff criteria for inclusion in the multivariable model was based on p (entry) <0.05 and p (removal) ≥0.2 significance levels. The Wald test was used to measure the statistical significance of differences in odds. Variable categories with similar levels of association with disease were combined to maximize power. Assessment of confounding for the selected variables was also performed using a change in point estimate cutoff criterion of 10% in response to the inclusion of each of the remaining nonselected variables.7 All putative predictors and confounding variables associated with either neoplasia or malignancy were then incorporated into a final multivariable model.
Results During the period between October 28, 2002, and May 10, 2006, 1,737 thyroid FNAs were reported at MMC on 1,362 patients; on average 1.3 FNAs were performed per patient. These FNAs were reported as follows: 279 “inadequate,” 1,118 “negative,” 275 “atypical,” 32 “suspicious” and 33 “positive.” The “atypical” results were from 249 different patients, and, of these, 100 (40%) had had histologic follow-up at MMC at the time of data collection.
A total of 99 of 100 “atypical” cases were analyzed; cytology slides for 1 case could not be located. The most common lesion types found in this sample were nodular goiter (n=35), followed by adenoma (follicular [26] and Hürthle cell adenoma [10]), and carcinoma (including papillary [9], follicular [8] and follicular variant of papillary [6]) (Table I).
 The mean (and median) age of the patients included in this study was 49 years (SD = 14.3) with a range from 20 to 81 years. The majority (78%) were female. Cytologic characteristics are shown in Table II. Histologic evaluation revealed 62 (63%) neoplastic cases and 26 (26%) malignant ones. Lesions with increased abundance of macrophages and colloid on cytologic evaluation were significantly less likely to be neoplastic or malignant. Whereas neoplastic lesions were more likely to have increased cellularity, the trend appeared to be reversed for malignant lesions.
 Table III shows the relative odds of predicting neoplasia and malignancy based on cytologic characteristics. The strongest predictors for any neoplasia included absence of colloid and increased cellularity, which remained significant even after adjustment for all selected variables. The observed univariate association with increased microfollicles and neoplasia did not remain after adjustment for absence of colloid and nuclear atypia. In comparison, absence of colloid was the strongest cytologic predictor of malignancy. Other putative predictors of neoplasia and malignancy included lack of macrophages, nuclear enlargement and normal nuclear:cytoplasmic ratio, although none were significant at the p<0.05 level.
 In order to assess the potential impact of misclassification in disease status, we further restricted our analyses to those cases that underwent surgery and histologic confirmation within 3 or 6 months of their FNA (n=73 and 87, respectively) and found little change in the associations with colloid or cellularity. Furthermore, inclusion of additional variables in the multivariable models did not significantly improve outcome prediction by colloid alone for either neoplasia or malignancy.
Discussion There are two types of studies in the literature that evaluate potentially neoplastic, nonpapillary follicular lesions of the thyroid using FNA. The majority are diagnostic category based, with histologic outcome as the gold standard and utilize a variety of overlapping cytologic categories for cytomorphologically similar follicular lesions (e.g., “atypical,” “indeterminate,” “suspicious”); in addition, the diagnoses are variably tailored using the aforementioned indefinite modifiers.6,8 The reported malignancy rates range from <10% up to 30%.9 This wide variation may reflect the lack of standardized diagnostic terminology, especially within the gray zone of thyroid FNA, and, consequently, the ability to draw good comparisons between these studies is limited.
The minority of studies in the literature are cytomorphology based, like ours, where all potentially neoplastic follicular lesions are examined without subclassification in an attempt to be more objective and quantitative. In our study, the malignancy rate of 26% falls within the range seen by other authors, suggesting that our data set probably consists of comparable cases. Interestingly, we are picking up a relatively high percentage of malignant cases at MMC using only 3 diagnostic categories for nonpapillary follicular lesions (e.g., “negative,” “no header” and “atypical”—see below).
Greaves et al10 reviewed 96 FNA cases of “follicular lesion”; these cases had initially been diagnosed using a wide range of terminology. They reported a 30% malignancy rate, and, after grading cytologic parameters similar to ours, they did not find any distinguishing parameters predictive of histologic outcome. Ersöz et al11 examined 56 FNA cases of “solitary cellular nodules” and “cellular microfollicular lesions” without defining these groups further. They reported a 23% malignancy rate, and, again, no distinguishing cytologic features predictive of histologic outcome were found using cytomorphologic criteria similar to those in our study.
Ersöz et al,11 however, did note certain nonsignificant trends, including that the likelihood of accurately detecting neoplastic lesions increased with increasing cellularity, nuclear size and microfollicles. Our data, while only significant with regard to the former, concurred for both cellularity and nuclear size; the observed univariate association with microfollicles was not observed in the multivariable model. Ersöz and coworkers also noted, as we did, that increased nuclear size was associated with malignancy. Last, in Ersöz’s study, specimens with abundant colloid were least likely to have been derived from a neoplastic lesion. In our study, this association was not only confirmed but also found to be statistically significant for both neoplastic as well as malignant outcomes.
Barbaro et al1 looked exclusively at cases with microfollicular architecture and evaluated anisokaryosis, amount of colloid and nuclear overlapping. Their study did not include gray zone cases and found an unusually low malignancy rate of 7.6% for a group with cytomorphologic features suggestive of follicular neoplasm. Additionally, they did not report their neoplasia rate or provide enough data to calculate one. However, they did show a statistically significant association between both lack of colloid and nuclear overlapping with carcinoma.
A noteworthy finding in our analysis was the association of decreased cellularity on FNA with malignant outcome. This relationship is unexpected given the traditional teaching that follicular neoplasms are hypercellular relative to goiters.12 It is possible that malignant lesions are relatively hypervascular, and, as a result, these aspirates tend to be more bloody and less cellular than FNAs of both adenomas and goiters. Also, we commonly observe that physicians who were not trained by someone experienced with the optimal FNA technique may actually intentionally continue an aspirate until blood is seen in the hub of the needle, incorrectly believing that this is the indication of a successful aspirate. Another explanation for the association may be that physicians, particularly if operating without the aid of microscopic adequacy assessments, may perform fewer passes for lesions that are more clinically suspicious for malignancy.
There have been prior attempts to standardize diagnostic terminology for reporting thyroid FNA, including the recommendation of the Papanicolaou Society of Cytopathology (PSC) in 1996 to utilize a five-category system including “inadequate,” “benign,” “atypical cells present,” “suspicious for malignancy” and “malignant.”13 However, such attempts have not resulted in an appreciable change in consistent FNA reporting.3,6 More recently, the PSC has revised their diagnostic scheme, and, in 2007, the National Cancer Institute (NCI) coordinated a “State of the Science” conference for thyroid FNA.8,9 Both have proposed similar 6-category systems with notable modifications intended to clarify cases of nonpapillary follicular lesions by subdividing the “atypical cells present” category into two, “follicular lesion (atypia) of undetermined significance” and “(suspicious for) follicular neoplasm.”9
At MMC, we are already using a 6-category scheme that is equivalent to the new proposals. We classify thyroid lesions as “inadequate,” “negative,” “no header,” “atypical,” “suspicious for malignancy” and “malignant.” However, for nonpapillary follicular lesions, we only utilize 3 of these 6 categories: “negative,” “no header” and “atypical.” Figure 1 illustrates the classic appearance of a “negative” FNA. This smear, which did not originate from our “atypical” data set, is hypocellular (1+), showing honeycomb sheets of follicular cells (i.e., macrofollicular architecture) and abundant colloid (3+). In addition to the “negative” header, the report included the comment, “Nodular goiter showing abundant colloid and benign follicular cells.” Figure 2 is from a case in our data set that, after lobectomy, was shown to be a follicular adenoma. It demonstrates the prototypical “atypical” smear pattern—hypercellular (3+) microfollicular architecture (3+) with no background colloid. Figure 3 is from another “atypical” case, which also was a follicular adenoma on histology, and shows several well-formed microfollicles at high power. These cases both included the comment, “Smear pattern consistent with a follicular neoplasm.”
   While the usual “atypical” aspirate ascribes to the general guidelines stated and illustrated above and is comparable to the proposed “follicular neoplasm” category, cytopathologic diagnoses are not simply about applying diagnostic criteria but also comparing any given case to one’s practice experience and always doing so in the clinical context. As such, there are a minority of “atypical” cases that do not perfectly comply with the guidelines, which are not intended to be absolute criteria, and these are reviewed at an intradepartmental conference for consensus opinion prior to sign-out. Ultimately, all “atypical” cases are diagnosed as such because they are thought to be potential neoplasms requiring lobectomy to rule out malignancy.2
Our “no header” group parallels the proposed “follicular lesion of undetermined significance” category (i.e., the gray zone), and we agree that the use of this category should be minimized.9 The common feature of the “no header” group is that the neoplastic potential of the lesion cannot reliably be determined. Classically, this is due to a mixed pattern showing features of both a goiter and follicular neoplasm where neither can be favored, but this category also includes cases where assessment of neoplastic features is quantitatively or qualitatively (i.e., by artifact) limited. Figure 4 illustrates this mixed pattern of moderately cellular smears (2+), including areas with prominent microfollicular architecture (2+), in a background of colloid (2+). The histology revealed a follicular carcinoma.
 For such “no header” cases, we remove the usual diagnostic header from the cytology report and provide a descriptive sign-out with a differential diagnosis. For internal, departmental quality-control purposes, these cases are still assigned to the “atypical” category even though our clinicians do not see the “atypical” header, and, thus, we have captured these cases in the current data set. They represent only a very small proportion, 7 cases (7%) of the sample, which limited our ability to investigate this subgroup in any detail. However, 4 of these “no headers” were neoplastic on excision, and 3 were malignant (3 follicular carcinomas and 1 follicular variant of papillary). In fact, the malignancy rate in the gray zone may be 10% or greater, and the neoplasia rate may be >50%.4,8,9,14 We think that the use of a descriptive sign-out for these cases without prescribed categorical terminology allows the clinician the latitude to decide between excision, repeat FNA or watchful waiting.
Regardless of the specific terminology, we are enthusiastic about the much-needed move towards standardizing thyroid FNA. Such a change will minimize the variation between different pathologists in diagnosing cytomorphologically similar lesions as well as the ensuing confusion among clinicians, especially as patients move between institutions. However, we think there are certain points that require particular emphasis. Importantly, the gray zone, whether it is dealt with as a “no header” or with the newly proposed “follicular lesion of undetermined significance” category, should not be a “wastebasket” category. With experience, it should constitute only a very low proportion of those nonpapillary follicular lesions that are not clearly multinodular goiters on FNA, and these cases can then be triaged based largely on their clinical settings.
Second, as an endocrine organ, there is a certain degree of baseline nuclear atypia in the thyroid gland that should be accepted when seen focally in a setting of otherwise nonneoplastic cytomorphology. This concept may in fact explain why nuclear atypia was 5 times more likely to be present with nonneoplastic outcomes in our study and why, while absolute nuclear enlargement was associated with malignant outcome, cases with a normal nuclear:cytoplasmic ratio were 3 times more likely to be malignant. Further, assessment of atypia is subjective and quite dependent on a pathologist’s level of experience. As such and in contrast to the implication of some authors, we emphasize that focal atypia should be regarded cautiously to prevent overinterpretation of nonneoplastic lesions and unnecessary surgery.8,14,15
In addition, there is a great deal of emphasis on malignancy rate in the thyroid FNA literature, and it is important to remember that, assuming the basic tenet that cytology cannot distinguish between follicular adenoma and carcinoma, we should also focus on neoplasia rates associated with screening for nonpapillary follicular lesions.1,4,5,10,14,15 Similarly, using the terminology “suspicious,” which in cytology traditionally implies a likely malignancy, to describe a possible follicular neoplasm on FNA seems misleading, given that 70–80% of these cases are benign on excision.9
Since we performed a retrospective study, we were limited in our ability to control for certain preanalytic variables. Many of these, notably ultrasound guidance, microscopic adequacy assessments and the number of passes per FNA, which conceivably could cause a bias, have their largest impact in ensuring that a good sample was obtained. Since all our cases were diagnosed as “atypical” rather than “inadequate,” the degree of bias is probably limited. In addition, there is the potential that our sample of 99 cases is not representative of the 249 patients who had “atypical” results during the study period. However, we feel that the missing follow- up is due to random reasons not likely to generate a significant bias; examples include: (1) patient lost to follow-up, (2) lobectomy not performed because either the patient declined or physician chose watchful waiting in the larger clinical context, and (3) lobectomy performed at another institution because either the patient was “self-referred” or physician was not an in-house physician (i.e., we have a large outreach volume and function in a capacity similar to a reference laboratory). Further, as compared to similar studies, our malignancy rate was appropriate, our sample size was large, and our limitations were analogous.1,9,10,11 Ultimately, all of our cases crossed the necessary combined clinical and cytologic threshold requiring surgery to better define the lesion.
Unlike similar studies in the literature of potentially neoplastic follicular lesions including the gray zone, ours revealed statistically significant associations that, upon further study, may allow increased precision and reduction of the gray zone in thyroid FNA, thereby reducing the number of lobectomies for nonneoplastic disease. This gray zone is to some degree unavoidable because there is a cytomorphologic spectrum of nonpapillary follicular lesions that can be variable from case to case, and, even with experience, it can be challenging to separate probable goiter from likely neoplasm in a minority of cases. However, this distinction is worsened by the current, variable use of diagnostic categories, which causes confusion among treating physicians and does not clarify patient management decisions. Thus, there is a clear need for the adoption of standardized categories as proposed by the PSC and the NCI.
We think that there are several caveats that should be emphasized in any universal scheme. First, with increasing familiarity with thyroid FNA, an individual cytopathologist’s gray zone becomes progressively smaller and may be minimized further by exercising caution regarding overinterpreting focal atypia within an otherwise nonneoplastic pattern. Additionally, consideration should be paid to possible issues, including misleading clinicians with inconsistent terminology like “suspicious for follicular neoplasm” and limiting clinical management options for FNA diagnoses like “follicular lesion of undetermined significance,” where clinical judgment should play a large role in deciding to watch, repeat FNA or even excise in certain circumstances.
Our current diagnostic system at MMC works well for us and our clinicians only because of a solid history of interdepartmental communication. However, we are in support of the adoption of a universal classification system not only to maximize the clarity of thyroid FNA reports for improved patient care but also so that future studies from different institutions can more easily and accurately be compared in order to assess FNA diagnoses vs. histologic outcome among larger populations. This would greatly assist in determining how to manage nonpapillary follicular lesions better in the future.
References
From the Departments of Pathology, Otorhinolaryngology and Clinical Surgery, Montefiore Medical Center, and the Departments of Epidemiology and Population Health and Medicine, Albert Einstein College of Medicine, Bronx, New York, U.S.A.
Dr. Somma was Cytopathology Fellow, Department of Pathology, Albert Einstein College of Medicine, and currently is Assistant Professor, State University of New York Downstate Medical Center, Brooklyn, New York. Dr. Schlecht is Assistant Professor, Departments of Epidemiology and Population Health and of Medicine, Albert Einstein College of Medicine.
Dr. Fink was Medical Student, Albert Einstein College of Medicine, and currently is Resident Physician, Department of Otolaryngology, Massachusetts Eye and Ear Infirmary, Boston, Massachusetts. Dr. Khader is Assistant Professor, Department of Pathology, Montefiore Medical Center.
Dr. Smith is Associate Professor, Departments of of Otorhinolaryngology and Clinical Surgery, Montefiore Medical Center. Dr. Cajigas is Associate Professor, Department of Pathology, Montefiore Medical Center.
Address correspondence to: Jonathan Somma, M.D., Department of Pathology, State University of New York Downstate, 450 Clarkson Avenue, Box 37, Brooklyn, New York 11203, U.S.A. (jonathan.somma@downstate.edu). Financial Disclosure: The authors have no connection to any companies or products mentioned in this article.
Received for publication March 26, 2009. Accepted for publication May 26, 2009.
 0001-5547/10/5402-0123/$21.00/0 © The International Academy of Cytology
Acta Cytologica Send website comments to Webmaster@acta-cytol.com |
|||||||||||