Preoperative B-type Raf kinase Val600Glu mutation, or BRAF(V600E), analysis has been proposed as a tool to guide initial surgery for indeterminate thyroid nodules. This study sought to determine if cytologic markers of malignancy are associated with the BRAF(V600E) mutation and if preoperative BRAF(V600E) testing would alter the initial management of patients with indeterminate nodules.
Patients who underwent surgery for a thyroid nodule between 2003 and 2012 at a tertiary care center were prospectively enrolled. Stored nodule samples were retrospectively genotyped for the BRAF(V600E) mutation. BRAF(V600E) status, demographics, cytologic and histopathologic findings, and choice of initial surgery were examined.
A total of 960 patients were enrolled, of which 310 (32%) had an indeterminate nodule. The BRAF(V600E) mutation was identified in 13 patients (4%), 12 of whom had either cytologic atypia or were Bethesda category V. Three percent of Bethesda category III or IV nodules that were malignant harbored the mutation compared with 42% of Bethesda category V malignancies. Nuclear grooves (P = .030), pseudoinclusions (P < .001), and oval nuclei (P = .022) were all more common among BRAF(V600E) mutants. The sensitivities of using BRAF testing alone, cytologic atypia/Bethesda category V classification, or both, were 15%, 73%, and 76%, respectively. Twelve of the 13 BRAF(V600E) mutants had total thyroidectomies initially due to worrisome cytologic features, and therefore the initial management of only one patient would have been altered if BRAF(V600E) testing had been performed preoperatively.
Thyroid nodules are a common clinical problem; 6% of the population has a palpable nodule and up to 65% have nodules diagnosed at autopsy.1 Ultrasound-guided fine-needle aspiration (FNA) is the preferred diagnostic test to assess the risk of malignancy of thyroid nodules.2-5 In 2009, the Bethesda system for reporting FNA cytopathology results replaced older systems with 6 categories. “Indeterminate” thyroid nodules (Bethesda categories III, IV, and V) pose a substantial dilemma for clinicians and patients, as they carry a risk of malignancy of approximately 5% to 15%, 15% to 30%, and 60% to 75%, respectively,6 and currently require at least a diagnostic hemithyroidectomy for histopathological examination to make a definitive diagnosis.
There is growing interest in adding preoperative molecular screening tests for markers such as BRAF, RAS, RET/PTC, and PAX8/PPARγ to improve the diagnostic accuracy of FNA.7-11 The most recent guidelines from the American Thyroid Association state that genetic mutation analysis “may be considered for patients with indeterminate cytology on FNA to help guide management.”12 Among the candidate genetic markers, the B-type Raf kinase Val600Glu mutation, or BRAF(V600E), has generated the greatest interest, because it is the most common single genetic alteration found in thyroid cancers, is present in 40% to 50% of all papillary thyroid carcinomas (PTCs), and can be easily assessed on preoperative FNAs.8, 11, 13-16 Its presence confers a 100% risk of malignancy and is associated with aggressive features such as extrathyroidal extension, lymph node metastases, advanced tumor stage, and radioactive iodine resistance.17, 18 Recently, several groups have developed more sophisticated screening tests that use panels of genetic mutations,10 gene expression,19 or micro-RNAs20 to improve the preoperative diagnosis of indeterminate nodules. Despite this, many authors continue to advocate for single mutation screening for the BRAF(V600E) mutation,21-25 and this practice continues to gain popularity despite the recent commercialization of at least one of the more sophisticated tests.19
As part of the conventional evaluation of FNA specimens, cytopathologists often note features of cytologic atypia such as nuclear grooves, pseudoinclusions, and oval nuclei that are not prominent enough to make the diagnosis of PTC. Such nodules are thereby categorized as a “follicular lesion of undetermined significance/atypia of undetermined significance (Bethesda category III).”6 This is a very heterogeneous group with an overall risk of malignancy ranging from 20% to 60%.26-30 Our group recently demonstrated that not all descriptors of atypia confer an equal risk of malignancy; the presence of nuclear grooves, pseudoinclusions, or both, increases the risk of malignancy to 55%, 65%, and 79%, respectively.31
The aims of this study were 3-fold: 1) to determine if certain features of cytologic atypia are independent risk factors for harboring the BRAF(V600E) mutation, 2) to determine if the additional knowledge of BRAF(V600E) status improves the sensitivity and specificity of FNA for detecting malignancy, and 3) to determine if performing preoperative BRAF(V600E) testing would alter the initial surgical management of patients with indeterminate thyroid nodules.
MATERIALS AND METHODS
Patient Selection and Clinical Data Collection
Patients aged 18 years and older who had either a hemi- or total thyroidectomy at a single tertiary referral center between January 2003 and March 2012 were prospectively enrolled. Patients whose preoperative FNAs identified an indeterminate thyroid nodule (Bethesda categories III, IV, or V) and had both cytologic and histopathologic reports available were included in the final study population. The Institutional Review Board of Weill Cornell Medical College approved this study, and informed, written consent was obtained from all patients.
Patient age, sex, preoperative FNA cytology, and final histopathological diagnosis were evaluated. All preoperative FNAs and surgical specimens were reviewed by a board-certified cytopathologist or surgical pathologist, respectively. Original FNA cytology reports were used to maintain generalizability to other referral centers that review reports from multiple outside institutions. For cases where multiple nodules were present, the FNA result was matched by location and size to the correct corresponding nodule on the histopathology report. If the Bethesda classification was not explicitly stated in the biopsy report (ie, biopsies performed prior to 2009 and biopsies performed at institutions that do not use the Bethesda system in their reports), the appropriate Bethesda category was assigned based on the findings noted in the report. For this reason, Bethesda categories III and IV were grouped together, because it was often difficult to distinguish these 2 groups on the basis of the terminology used in those reports.
The following features of cytologic atypia were tabulated for each biopsy: psammoma bodies, clear chromatin, nuclear grooves, pseudoinclusions, enlarged nuclei, pinpoint nuclei, peripheral nucleoli, pleomorphic nuclei, and atypia not otherwise specified.
BRAF(V600E) Mutation Analysis
BRAF(V600E) mutation status was determined retrospectively, using Sanger sequencing of frozen tissue samples that were collected immediately after surgical resection. All tissue samples were snap-frozen in liquid nitrogen and stored at −80°C until analysis. Genomic DNA was extracted using the QIAamp DNA Mini Kit (Qiagen, Valencia, Calif) according to manufacturer instructions. The following primers were used as a template for a polymerase chain reaction (PCR) reaction under standard conditions using the AmpliTaq Gold PCR Kit (Applied Biosystems, Foster City, Calif): forward 5′-TGCTTGCTCTGATAGGAAAATG-3′, reverse 5′-GACTTTCTAGTAACTCAGCAGC-3′. The PCR product was purified using the QIAQuick PCR Purification Kit (Qiagen) and visualized on a 2% agarose gel, which confirmed the presence of a single band of approximately 238 base pairs. The PCR product was then direct-sequenced on an Applied Biosystems Automated 3730xl DNA analyzer (Biotechnology Resource Center of Cornell University, Ithaca, NY) and read at position 1799 for the presence of a mutant (A) allele. All malignant nodules and 18 of the 221 benign nodules were genotyped. The remaining benign nodules were assumed to be wild-type, because the BRAF(V600E) mutation has never been detected in a benign nodule.
Univariate analysis was performed comparing demographic and cytologic features of BRAF(V600E) mutants versus the wild type. Features of cytologic atypia that were identified on univariate analysis as having a significant association with the BRAF(V600E) mutation were then entered into a multivariate logistic regression model to assess their influence on harboring the BRAF(V600E) mutation. The sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) of, 1) BRAF(V600E) mutation testing, or 2) cytologic atypia or Bethesda category V classification as surrogate markers for malignancy, were calculated both alone and in combination to determine if the additional knowledge of the former would have improved the diagnostic accuracy of the latter.
P values were calculated using Fisher's exact test (dichotomous variables) or Student t test (continuous variables) as appropriate. All continuous variables followed a normal distribution and are presented as mean ± standard deviation. P values < .05 were considered statistically significant. All statistical analyses were performed using STATA 12.0 (Stata Corporation, College Station, Tex).
A total of 960 patients were enrolled in this study, of which 310 (32%) had indeterminate nodules (Bethesda categories III, IV, or V) and were included in the final study population (Fig. 1). Among this population, 274 (88%) nodules were Bethesda category III or IV, and 36 (12%) were Bethesda category V. In addition, 89 of the indeterminate nodules were malignant, making the overall risk of malignancy 29%. A total of 145 nodules (47%) had at least 1 feature of cytologic atypia, which increased the risk of malignancy to 42%. Thirteen BRAF(V600E)-mutant nodules were identified, all of which were malignant. There were no significant differences in age, sex, or nodule size between BRAF(V600E) mutants and the wild type (Table 1).
Table 1. Patient and Nodule Characteristics Grouped by BRAF(V600E) Status
BRAF Mut (n = 13)
BRAF WT (n = 297)
Mean ± standard deviation.
Percents given for these 2 rows are within-row percent (all others are within-column percent). Abbreviations: BRAF(V600E), B-type Raf kinase Val600Glu mutation; Mut, mutant; WT, wild-type.
The various histologic subtypes identified among the total study population as well as several subgroups are shown in Table 2. The most common malignant histologic subtype overall was follicular variant of PTC (fvPTC). Bethesda category III or IV nodules that were malignant were most likely to be of the fvPTC subtype, whereas Bethesda category V malignant nodules were most likely to be classic PTCs. Eleven (85%) of the BRAF(V600E)-mutant nodules were classic PTCs, whereas 2 (15%) were fvPTCs.
Table 2. Final Histopathologic Diagnoses by Subgroup
Eleven of the BRAF(V600E)-mutant nodules had cytologic atypia, whereas 2 did not (P = .005). Nodules with cytologic atypia were 6.7 times more likely to harbor the BRAF(V600E) mutation than those without atypia (95% confidence interval [CI] = 1.5-30.7, P = .004). Similarly, nodules with cytologic atypia were 3.6 times more likely to be malignant than nodules without atypia (95% CI = 2.1-6.0, P < .001).
Nuclear grooves (P = .030), intranuclear pseudoinclusions (P < .001), and oval nuclei (P = .022) were all more common in BRAF(V600E)-mutant nodules compared with wild-type nodules (Table 1). Multivariate logistic regression analysis revealed that pseudoinclusions and oval nuclei were independent risk factors for harboring the BRAF(V600E) mutation, with odds ratios = 11.2 (95% CI = 3.0-42.2) and 10.0 (95% CI = 1.4-72.0), respectively (Table 3). Nuclear grooves showed a trend toward increased risk but did not reach significance (odds ratio = 2.3, 95% CI = 0.6-8.3). The area under the receiver operating curve (ROC) for a multivariate model including all 3 descriptors was 0.749, but the area decreased to 0.705 when the factor of nuclear grooves was removed, thus indicating that nuclear grooves contributed to the overall predictive value of the model despite not reaching statistical significance independently.
Table 3. Multivariate Logistic Regression Analysis for the Presence of the BRAF(V600E) Mutationa
95% Confidence Interval
Overall P value of the model is <.001 and the receiver operating characteristic area under the curve is 0.749.
The incidence of the BRAF(V600E) mutation among Bethesda category III or IV nodules that were found to be malignant on final pathologic analysis was 3%, whereas the incidence among Bethesda category V nodules that were malignant was 42% (P < .001, Table 4). The sensitivity of BRAF(V600E) testing for detecting malignancy in all indeterminate nodules was 15%. The sensitivity among Bethesda III or IV nodules was 3.2% and among Bethesda V nodules was 42%. Among nodules with cytologic atypia, the sensitivity was 18%, whereas among Bethesda category III or IV nodules without atypia, the sensitivity was only 4%. The specificity of BRAF(V600E) testing was 100% for all subgroups.
Table 4. Correlation Between BRAF(V600E) Status and Outcome Among Various Subgroups of the Study Population
Malignant (n = 89)
Benign (n = 221)
Abbreviations: AUS, atypia of undetermined significance; BRAF(V600E), B-type Raf kinase Val600Glu mutation; FLUS, follicular lesion of undetermined significance; NPV, negative predictive value; PPV, positive predictive value.
All indeterminate nodules
(Bethesda III, IV or V)
Malignant (n = 63)
Benign (n = 211)
Bethesda III or IV
(FLUS/AUS or follicular neoplasm)
Malignant (n = 26)
Benign (n = 10)
Bethesda V (suspicious for PTC)
Malignant (n = 61)
Benign (n = 84)
Malignant (n = 23)
Benign (n = 135)
Bethesda III or IV without atypia
The sensitivity, specificity, PPV, and NPV of BRAF(V600E) mutation testing compared with, or in combination with, use of cytologic atypia or Bethesda category V classification as surrogate markers for malignancy are displayed in Figure 2. When BRAF(V600E) testing was added to these conventional cytologic markers of malignancy, there was no meaningful improvement in sensitivity, specificity, NPV, or PPV.
Finally, the extent of initial surgical resection performed on the 13 patients with the BRAF(V600E) mutation was reviewed to determine if the initial operative plan would have changed if the presence of the BRAF(V600E) mutation had been known preoperatively (ie, if a total thyroidectomy would have been recommended instead of a hemithyroidectomy initially). Eleven of the 13 BRAF(V600E) mutant patients had Bethesda category V nodules and thus already underwent a total thyroidectomy as their initial operation, as is the standard recommendation at our institution. One patient had a Bethesda category III nodule with pseudoinclusions and enlarged nuclei and, therefore, that patient was also advised to undergo a total thyroidectomy due to the increased risk of malignancy that those atypical features imparted. One patient underwent a hemithyroidectomy and isthmusectomy for a 2.7-cm follicular neoplasm (Bethesda category IV) with no cytologic atypia that was found to be a BRAF(V600E)-mutant fvPTC. Therefore, the initial surgical plan would have been altered in only 1 of the 310 patients with indeterminate nodules if BRAF(V600E) testing had been performed preoperatively.
Preoperative screening for the BRAF(V600E) mutation as a way to refine the accuracy of FNA of indeterminate thyroid nodules is rapidly gaining popularity.11, 16, 21, 22, 32 However, in this study, we found that 92% of indeterminate nodules that harbor the BRAF(V600E) mutation have other conventional cytologic markers of malignancy (cytologic atypia or Bethesda V classification) that already warrant a total thyroidectomy, and therefore, BRAF(V600E) screening is unlikely to alter their initial management. Furthermore, the sensitivity of BRAF(V600E) screening is lowest among nodules for which it would be hoped to be most helpful (Bethesda III or IV nodules without atypia which normally warrant only a diagnostic hemithyroidectomy), thus limiting its potential value as a preoperative screening test. As such, this is the first study to report that cytologic atypia predicts the presence of the BRAF(V600E) mutation, and also that preoperative BRAF(V600E) screening does not improve the evaluation of indeterminate nodules if one considers cytologic atypia.
The low sensitivity of BRAF(V600E) screening is the main limitation of single-mutation analysis for detecting malignancy among indeterminate FNAs. To overcome this limitation, Nikiforov and colleagues have developed a genetic screening panel for molecular analysis of preoperative FNAs that detects not only BRAF(V600E), but also point mutations of NRAS codon 61, HRAS codon 61, and KRAS codons 12/13, and rearrangements of RET/PTC1, RET/PTC3, and PAX8/PPARγ.10 Testing positive for any mutation on the panel has a sensitivity of 62%, specificity of 99.7%, and PPV of 97% for malignancy. However, very few centers as of yet have the ability to screen for the full panel of mutations, and most that have begun or are considering adding mutation analysis only test for BRAF(V600E).
Several other groups have recently developed molecular profiling tests that use multiple molecular markers to refine the preoperative diagnosis of indeterminate nodules. The Veracyte Afirma thyroid FNA analysis system measures expression of 142 genes to classify indeterminate nodules as benign or suspicious with 92% sensitivity and 52% specificity.19 Our group has recently developed a panel of 4 micro-RNAs (miR-222, miR-328, miR-197, and miR-21) that has been preliminarily shown to discriminate between benign and malignant indeterminate nodules with 100% sensitivity and 86% specificity.20 Future studies that combine existing knowledge of conventional cytologic risk factors and multiple molecular markers of thyroid cancer are warranted.
Nuclear grooves, pseudoinclusions, and oval nuclei were found to be associated with the BRAF(V600E) mutation on univariate analysis, and pseudoinclusions and oval nuclei maintained significance on multivariate logistic regression. These findings are consistent with our group's recent study that showed that grooves and pseudoinclusions are independent risk factors for malignancy.31 It also supports the findings of Adeniran et al. who found that the incidence of the BRAF(V600E) mutation among indeterminate nodules with cytologic atypia was 23% whereas the incidence among nodules without atypia was 0%.21 This suggests that these features of cytologic atypia may be used as surrogate markers for the BRAF(V600E) mutation.
The incidence of the BRAF(V600E) mutation varies greatly according to Bethesda category; the incidence in Bethesda category III and IV nodules is < 5% whereas the rate among Bethesda category VI (malignant) nodules is roughly 70%.11, 22, 23, 32 One explanation for this difference is that the majority of Bethesda category III and IV nodules that are found to be malignant are of the fvPTC subtype (as was the case in this study), and only approximately 10% to 15% of fvPTCs harbor the BRAF(V600E) mutation.8, 17, 33 Therefore, the likelihood of finding the BRAF(V600E) mutation among Bethesda category III or IV nodules is quite low which limits the potential contribution of preoperative mutation analysis for these nodules.
A similar yet distinct research question is whether preoperative BRAF(V600E) screening of indeterminate nodules would influence the decision to perform a prophylactic central compartment lymph node dissection in addition to a total thyroidectomy at the time of the initial operation. Two recently published studies found that the BRAF(V600E) mutation is an independent risk factor for occult central compartment lymph node metastases, and therefore, the authors suggest that preoperative BRAF(V600E) screening should be used to guide the extent of initial surgery.24, 25 Both of these studies, however, only included Bethesda category VI (malignant) nodules, a subgroup characterized by a significantly higher incidence of the BRAF(V600E) mutation (discussed above) and a larger proportion of aggressive histologic subtypes than Bethesda categories III, IV, and V nodules.34 Furthermore, it remains unclear whether performing a prophylactic central compartment lymph node dissection improves outcomes or if the overall benefits of the procedure outweigh the increased risk of complications such as permanent hypocalcemia.35-39
There are at least 2 limitations to this study. First, this was a single-center study and, as such, the choice of initial surgery is reflective of the practice patterns of the 2 participating surgeons (T.J.F. and R.Z.). At our center, we typically recommend a total thyroidectomy for all patients whose risk of malignancy is ≥ 70%, which includes Bethesda category V nodules and nodules with nuclear grooves and pseudoinclusions. Centers that follow a more thyroid-preserving approach may observe a more frequent change in initial operative plans if preoperative BRAF testing is performed. Second, BRAF(V600E) testing was performed retrospectively using frozen tissue samples and not prospectively using cytologic material from an FNA biopsy. However, several authors have previously demonstrated the feasibility of performing mutation analysis using FNA material.8, 11, 15, 16
In conclusion, the BRAF(V600E) mutation is most common in nodules with other cytologic risk factors for malignancy, which already warrant a total thyroidectomy. Therefore, single-mutation screening for BRAF(V600E) does not meaningfully improve preoperative risk stratification and is unlikely to alter the initial management of patients with indeterminate nodules.
This study was supported in part by grant TL1RR024998 of the Clinical and Translational Science Center at Weill Cornell Medical College, and by a donation from the Dancers Care Foundation.