Portions of this work were presented at the Annual Meetings of the American Society of Cytopathology, November 2010 (Boston MA) and November 2011 Baltimore MD).
Follicular lesion of undetermined significance in thyroid FNA revisited
Article first published online: 27 JUL 2013
Copyright © 2013 Wiley Periodicals, Inc.
Volume 42, Issue 1, pages 18–22, January 2014
How to Cite
Walts, A. E., Mirocha, J. and Bose, S. (2014), Follicular lesion of undetermined significance in thyroid FNA revisited. Diagn. Cytopathol., 42: 18–22. doi: 10.1002/dc.23019
- Issue published online: 26 DEC 2013
- Article first published online: 27 JUL 2013
- Manuscript Accepted: 7 MAY 2013
- Manuscript Received: 20 NOV 2012
- follicular lesion of undetermined significance (FLUS);
- papillary carcinoma;
- Bethesda System for Reporting Thyroid Cytopathology
Controversy exists regarding the validity of follicular lesion of undetermined significance (FLUS), an indeterminate diagnostic category of The Bethesda System for Reporting Thyroid Cytopathology (BSRTC). According to BSRTC, FLUS carries a 5–15% risk of cancer. This study was designed to determine if cytomorphology could stratify FLUS into subgroups with different risks of malignancy. Reports of 127 consecutive FNAs reported as FLUS with subsequent tissue diagnoses were evaluated for the presence of various cytologic features and the results were correlated with histological diagnoses. FLUS cases with focal nuclear atypia (nuclear overlap/crowding, nuclear grooves/membrane irregularities, nuclear enlargement, and/or nuclear pseudoinclusions) were more frequently malignant on excision whereas those with architectural atypia (microfollicles) were more often benign on excision (P < 0.05). The presence of any one or more of these nuclear features increased the risk of carcinoma in subsequent thyroid resection. Papillary carcinomas predominated in excised FLUS cases with focal nuclear atypia whereas most FLUS with architectural atypia were adenomas or hyperplastic nodules on histological evaluation. BSRTC recommends that thyroid aspirates containing follicular cell nuclear and/or architectural atypia insufficient for a diagnosis of suspicious for follicular neoplasm, suspicious for malignancy or malignant be classified as FLUS. Our findings indicate that FLUS cases with focal nuclear atypia carry a risk for malignancy that is substantially higher than that assigned to FLUS and are best classified as suspicious. FLUS cases lacking these atypical nuclear features have a risk for malignancy that approximates the risk BSRTC has assigned to FLUS. Diagn. Cytopathol. 2014;42:18–22. © 2013 Wiley Periodicals, Inc.
The Bethesda System for Reporting Thyroid Cytopathology (BSRTC) was developed in 2007 to standardize the diagnostic criteria and terminology utilized for thyroid cytopathology. BSRTC describes six categories for the diagnosis and reporting of thyroid FNAs, each with an assigned “risk of malignancy” and provides recommendations for clinical management.[1, 2] While thyroid FNA is an accurate test with high positive and negative predictive values for malignant and benign diagnoses, from 5 to 42% of thyroid FNA diagnoses are “indeterminate”. These diagnoses encompass three of the BSRTC categories, “atypia of undetermined significance/follicular lesion of undetermined significance (AUS/FLUS),” “follicular neoplasm (FN),” and “suspicious for malignancy” with assigned risks of malignancy of 5–15%, 15–30%, and 50–75%, respectively.
FLUS is currently the most frequent abnormal diagnosis rendered in thyroid FNAs. As currently defined by BSRTC, FLUS is reserved for specimens that “contain follicular cells with architectural and/or nuclear atypia that is not sufficient to be classified as suspicious for a follicular neoplasm, suspicious for malignancy, or malignant. Although BSRTC recommends that FLUS not exceed 7% of a pathologist's or a laboratory's thyroid FNA diagnoses,[1, 4] this percentage has varied from about 3–29% across laboratories and from 2.5% to 28.6% across cytopathologists.[5-8] In a recent review encompassing 8 series (30,466 cases) of thyroid FNAs classified by BSRTC, usage of the FLUS category ranged from 3.0% to 18.0% with a median rate of 9.9% while in another review of the literature the incidence of FLUS ranged from 0.7% to 18%.[9, 10] Others have noted that it is difficult to compare experiences across institutions and countries because differences in epidemiology, organized screening programs, analyses of outcome data, and/or clinical management can contribute to the marked variability in the incidence of FLUS and the frequency of malignancy reported following a diagnosis of FLUS.[10, 11] Nevertheless, it is the low levels of intra- and inter-observer diagnostic agreement associated with the indeterminate categories “FLUS” and “suspicious for malignancy” that reflect inconsistent application of subjective and overlapping criteria that have hampered the translation of these two categories into clinical practice.
A review of the recent cytology literature underscores the controversy regarding FLUS. Two schools of thought have emerged—one proposes to eliminate the FLUS category (or combine the FLUS and FN categories to form a five-tiered system) and the other suggests a further subdivision of FLUS into categories with different risk assignments.[12-19] Related issues involve determining whether nuclear atypia or architectural atypia (microfollicles) in an FNA is more reliably predictive of malignancy, quantifying the amount or extent of a “diagnostic” feature that should be required for inclusion in the suspicious for malignancy category,[16, 20-22] assessing the value of repeat FNA in nodules reported as FLUS and FN nodules,[23-25] requiring consensus review before a thyroid FNA is diagnosed as FLUS, and assessing the potential role of molecular studies as an adjunct to cytodiagnosis.[27-29] This study was designed to determine if cytomorphology could help further stratify lesions that are currently being reported as FLUS.
Materials and Methods
After IRB approval, all thyroid FNAs reported as FLUS between 9/1/2008 and 7/1/2012 were identified in our departmental cytopathology database. Our surgical pathology files were then searched to identify 127 consecutive thyroid nodules > 0.5 cm in diameter that had been diagnosed as FLUS on FNA and subsequently excised. A cutoff of 0.5 cm was chosen since these are the smallest lesions currently targeted for aspiration by our radiologists. Incidental lesions (measuring < 0.5 cm on imaging or histology) were not evaluated in this study. Demographic and pathological data were collected from patient medical records and pathology reports. The final diagnostic groups (benign and malignant) were compared for age, gender, size of lesion, and interval between FNA and excision. The benign and malignant groups were also compared for the presence of follicular cell sheets and clusters/aggregates, microfollicles, nuclear grooves/nuclear membrane irregularities (NG), nuclear overlap/crowding (NC), nuclear enlargement (NE), nuclear pseudoinclusions (NI), pale/powdery chromatin, and Hurthle cells as described in the cytology reports. The presence of colloid was also noted. None of the FNAs in our study contained an atypical lymphoid infiltrate, atypical cyst-lining cells, or other atypical cells. Each FNA consisted of at least 4 Papanicolaou stained direct smears and one liquid based (SurePath) Papanicolaou stained needle rinse preparation. The FNAs had been reported by our five cytopathologists in accordance with routine laboratory rotation assignments. Confounding effects of subjective terms such as “rare,” “few,” “some,” “focal,” “probable,” and “possible” that were used differently across cytopathologists and interpreted differently across the treating clinicians, were eliminated by scoring each feature on a binary scale (present or absent) with “possible” and “probable” recorded as present.
Numerical variables were summarized by Mean (SD) if normally distributed or by Median (range) if not normally distributed. Numerical variables were compared across the malignant and benign groups by the t test (normally distributed) or the Wilcoxon rank sum test (not normally distributed). Binary (presence or absence) variables were summarized by frequency (percent) and were compared across the groups by the Chi-Square (all expected cell counts in the cross-tabulation table were > 5) or Fisher exact test (any expected cell count ≤ 5). The Cochran-Armitage trend test (one-sided) was used to assess the presence of an increasing trend in the percent malignant as the number of clinically significant atypical nuclear features (NI, NG, NC, NE) increased from 0 to 4 (Table 3). Multivariable logistic regression was used to assess variables related to malignancy. The 0.05 significance level was used throughout. All statistical calculations were performed using SAS version 9.2 software (SAS Institute, Inc. Cary NC).
FLUS comprised 10.5% of all 5595 thyroid FNAs reported by our cytology laboratory during the study period. Histological diagnoses of the 127 thyroid nodules > 0.5 cm diagnosed as FLUS by FNA included 84 benign nodules [46 follicular adenomas, 37 hyperplastic nodules, 1 nodular fibrosing Hashimoto's thyroiditis], 38 carcinomas [16 follicular variant of papillary thyroid carcinoma; FVPTC, 4 carcinomas with mixed papillary/follicular patterns, 7 conventional papillary carcinomas, 1 sclerosing papillary carcinoma, 7 minimally invasive follicular carcinomas, 1 anaplastic carcinoma, 1 follicular carcinoma, 1 medullary carcinoma], and 5 neoplasms of undetermined malignant potential. For statistical analysis, the 5 neoplasms of undetermined malignant potential were included in the malignant group bringing the total to 43 malignant nodules. Statistical analysis yielded similar results whether these 5 cases were classified as malignant or were excluded from the study. The 127 patients (100 females and 27 males) ranged from 23 to 84 years in age (median 55 years) at partial or total thyroidectomy. The interval from index FNA to excision ranged from 0.3 to 42 months (median 3 months). The aspirated nodules ranged in size from 0.6 to 10.0 cm (median 2.5 cm).
As shown in Table 1, when the FLUS cases with benign and malignant tissue diagnoses were compared, no significant difference was observed in age, gender distribution, or interval from FNA to thyroidectomy between the two groups. However, at excision the nodules in the malignant group were significantly smaller than those in the benign group.
|Malignant (n = 43)||Benign (n = 84)||P-value|
|Female||32 (74.4%)||68 (81.0%)||0.39|
|Age, Mean (SD)||51.6 yrs (15.8)||53.9 yrs (13.2)||0.37|
|Interval FNA to excision, Median (Range)||2.5 mos (0.3–18)||3 mos (0.5–42)||0.094|
|Nodule size, Median (Range)||1.8 cm (0.7–6)||2.5 cm (0.6–10)||0.048|
As shown in Table 2, when the FNAs from the benign and malignant cases were compared, 4 of the atypical nuclear features studied [nuclear pseudoinclusions (NI), nuclear overlap/crowding (NC), nuclear grooves/nuclear membrane irregularities (NG), and nuclear enlargement (NE) were each more frequent in the malignant group whereas architectural atypia (microfollicles) was more frequent in the benign group of cases. Nine (81.8%) of the 11 FNAs with NI, 24 (53.3%) of the 45 FNAs with NC, 16 (51.6%) of the 31 FNAs with NG, and 24 (45.3%) of the 53 FNAs with NE were malignant on excision whereas 51 (73.9%) of the 69 FNAs with microfollicles were benign on excision. No significant difference was observed in the frequencies of the other features between the two groups. When these five features were included as predictors of malignancy in a multivariable logistic regression model, two atypical nuclear features, namely, NI (odds ratio [OR] = 7.8, 95% confidence interval [CI] = 1.5–39.8 and NC (OR = 3.1, 95% CI = 1.4–7.0) remained as significant predictors.
|Feature present||Malignant (n=43)||Benign (n=84)||P-value|
|Sheets||13 (30.2%)||26 (31.0%)||0.93|
|Aggregates/clusters||39 (90.7%)||76 (90.5%)||>0.99|
|Microfollicles||18 (41.9%)||51 (60.7%)||0.044|
|Nuclear grooves/irregular nuclear membranes||16 (37.2%)||15 (17.9%)||0.016|
|Nuclear overlap/crowding||24 (55.8%)||21 (25.0%)||0.0006|
|Nuclear enlargement||24 (55.8%)||29 (34.5%)||0.021|
|Nuclear pseudoinclusions||9 (20.9%)||2 (2.4%)||0.0001|
|Pale/powdery chromatin||12 (27.9%)||15 (17.9%)||0.19|
|Hurthle cells||10 (23.3%)||24 (28.6%)||0.52|
|Colloid||35 (81.4%)||71 (84.5%)||0.80|
|Significant cytological features in FLUS FNA||No. of FNAs||No. Malignant on excision (%)|
|All 4 features present||5||5 (100)|
|Any 3 of the 4 features present||15||5 (33.3)|
|Any 2 of the 4 features present||20||12 (60.0)|
|Any 1 of the 4 features present||35||14 (40.0)|
|0 (none) of the 4 features present||52||7 (13.5)|
|Cochran-Armitage Trend Test: P < 0.0001|
Malignancy was present in 14 (40%) of the 35 thyroid lesions >0.5 cm whose FNAs exhibited one of four atypical nuclear features (NI, NC, NG, NE). Thus, the presence of even one of these atypical nuclear features was associated with a risk of malignancy that is substantially higher than that currently assigned to FLUS by BSRTC (Table 3). In contrast only 7 (13.5%) of the 52 FLUS that lacked all four of these atypical nuclear features were malignant on excision, yielding a risk of malignancy that is within the 5–15% currently assigned to the FLUS category by BSRTC. Of the 38 thyroid carcinomas diagnosed following an FNA reported as FLUS, 28 (73.7%) were papillary thyroid carcinomas (PTC) and only 8 (21.1%) were follicular carcinomas. FVPTC comprised 57.1%, classical PTC comprised 25%, and PTC with both patterns comprised 14.3% of the excised PTCs. The remaining PTC was a sclerosing variant.
In summary, among nodules that were diagnosed as FLUS and then excised, the overall rate of a histological diagnosis of malignancy was 29.9% (or 33.9% if the 5 nodules that were indeterminate on histological examinations are included), substantially higher than the 5–15% risk of malignancy currently assigned to the FLUS category by BSRTC. Our study indicates that FLUS could be stratified into two groups. One group exhibited any one or more of four atypical nuclear features [nuclear pseudoinclusions (NI), nuclear overlap/crowding (NC), nuclear grooves/irregular nuclear membranes (NG), and nuclear enlargement (NE)] and had a 33.3–100% risk of malignancy. Even the minimum risk calculated (33.3%) exceeds the risk currently assigned to FLUS. In our cohort when two of the atypical nuclear features were present, the risk of malignancy was similar to that BSRTC assigns to the suspicious for malignancy category and when all four of the atypical nuclear features were present, the risk of malignancy reached 100%. The other group which lacked these atypical nuclear features but included cases with architectural atypia (microfollicles) was associated with a lower risk of malignancy (13.5%) appropriate to the FLUS category as assigned by BSRTC. Based on our findings, we propose that FNAs exhibiting one or even equivocal for one or more of these atypical nuclear features be assigned a higher risk of malignancy than cases of FLUS with architectural atypia. One of the concerns with the current use of FLUS is that it has become a waste basket category that might even include cases more appropriately reported as unsatisfactory due to scant cellularity and/or poor preservation but designated as FLUS in response to pressure exerted on cytopathologists to minimize the number of cases reported as unsatisfactory. While we agree that adherence to established criteria for adequate cellularity and preservation are important determinants of the clinical utility of FLUS as an FNA diagnosis, the subjective evaluation of nuclear features remains a major issue. Review of individual cases by several cytopathologists whether independently or at a multiheaded microscope in order to arrive at a consensus diagnosis would improve interobserver agreement and in at least some cases also improve diagnostic accuracy, but this is often impractical. The large number of thyroid FNAs performed, the relatively small number of cells of interest that are often distributed over numerous smears, and/or difficulties in securing the time required for careful multihead review of multislide cases present problems in large departments and the presence of only one cytopathologist on site precludes this kind of case review in small hospital or non-hospital laboratory settings. Hence our study reflects the daily sign-out situation in which cytologists often find themselves – uncertain as to whether or not there are a sufficient number of nuclear grooves to diagnose “suspicious for malignancy” or uncertain as to whether or not a “rare pseudoinclusion” is present, etc. and issuing reports that convey uncertainty and equivocation through the use of “rare,” “focal,” “possible,” and “probable”.
Published studies that address the pros and cons of further stratifying FLUS, preserving the current BSRTC, or combining FLUS and FN have yielded conflicting results, and differences in study design and/or in criteria applied confound cross-study comparisons. For the most part, these studies involved retrospective review and reclassification of cases that had been previously reported as FLUS and an assessment of the effects of this reclassification on the prediction of malignancy in thyroid tissue excised following the FNA. Our study did not involve reclassification of cases. Instead, we utilized the presence of nuclear and architectural features as described in the original FNA reports and histological outcome to investigate the feasibility of and potential contribution that further stratification of FLUS could make to improve management of thyroid nodules > 0.5 cm. While there was no significant difference in the interval from FNA to excision in the two groups, the excised carcinomas were significantly smaller than the excised benign nodules. This could reflect the slow growth rate of most of these differentiated carcinomas, effective utilization of clinical and/or imaging parameters to identify FLUS cases at highest risk for malignancy and prioritize them for surgery, and/or the tendency of benign nodules to undergo spontaneous hemorrhage and/or cystic degeneration resulting in a rapid and substantial increase in size. We chose 0.5 cm as the threshold size of the aspirated nodules in this study because our radiologists regard 0.5 cm as the lower threshold for reliable visualization by imaging and accurate sampling by ultrasound guided FNA. In so doing we excluded an additional 11 thyroid FNAs that had been reported as FLUS and subsequently excised during the study period. Nine of these eleven thyroid glands contained a microscopic PTC from 0.1 to 0.4 cm in diameter and each of the remaining 2 glands contained a benign 0.5 cm in diameter nodule.
In our study, papillary carcinoma accounted for 73.7% of the thyroid carcinomas that followed a diagnosis of FLUS and the follicular variant (FVPTC) comprised 57.1% of these PTCs. In a separate study VanderLaan noted that 90% of the cancers that followed an FNA diagnosis of FLUS were PTCs among which 85% were FVPTC. In contrast follicular carcinomas accounted for 50% of the malignancies that followed an FNA diagnosis of FN in our laboratory during the same study period suggesting that FLUS and FN identify different histologic subsets of thyroid cancer. Our findings are also in accord with Luu et al. who recently reported a significantly higher risk of malignancy in FLUS cases “with equivocal features of PTC” when compared with FLUS exhibiting “other patterns” and with other studies indicating that nuclear features (irregular nuclear membranes, nuclear overlapping, altered chromatin) alone or in combination with architectural features are more predictive of malignancy than the presence of microfollicles alone[17, 21, 22] which are more predictive of follicular adenoma. Stratifying cases that are currently being signed out as FLUS and distinguishing those that exhibit the focal atypical nuclear features that are suggestive of papillary carcinoma from those that exhibit architectural atypia (microfollicles) appears warranted. This information should be communicated to clinicians because it could substantially improve the selection of cases that warrant surgical treatment.
- 4The Bethesda system for reporting thyroid cytopathology. Definitions, criteria, and explanatory notes. New York: Springer, 2010., .
- 5The indeterminate thyroid fine-needle aspiration: experience from an academic center using terminology similar to that proposed in the 2007 National Cancer Institute Thyroid Fine Needle Aspiration State of the Science Conference. Cancer Cytopathol 2009;117:195–202., .
- 11The Bethesda system for reporting thyroid fine needle aspiration cytology: A study comparing the results of two centers from two different countries. Ann Pahol 2012;32:e29–e34, 415–420., .
- 17Atypia of undetermined significance: Institutional experience with the Bethesda System. CytoJournal 2011;8:S65., , , .
- 18Review of the National cancer institute/Bethesda terminology system for thyroid cytology specimens: The New York university experience, with special focus on the indeterminate category follicular lesion of undetermined significance. CytoJournal 2011;8:S61., , , , , .
- 24Subclassification of atypical cells of undetermined significance in direct smears of fine-needle aspirations of the thyroid: Distinct patterns and associated risk of malignancy. Cancer Cytopathol 2011;119:322–327..
- 25Fine-needle aspiration of follicular patterned lesions of the thyroid: Diagnosis, management, and follow-up according to National Cancer Institute Recommendations. Diagn Cytopathol 2010;38:731–739., .