Introduction

Nodular thyroid disease is a common endocrine disorder that is particularly prevalent in iodine-deficient regions [1, 2]. Although thyroid nodules are often benign, the malignancy rate is 5–17 % and is increasing steadily [3–7]. Malignancy risks in patients with thyroid nodules include radiation history to the head and neck, family history of medullary thyroid cancer, multiple endocrine neoplasia type II and papillary thyroid cancer, age <14 and >70 years old, male sex, growing and hard nodules, cervical lymphadenopathy, fixed nodules, as well as dysphonia, dysphagia, and dyspnea [8].

High-resolution (10 MHz) ultrasonography (USG) is a useful method for evaluating thyroid nodule morphology, lymph nodules, nodule follow-up, and guidance for fine-needle aspiration biopsy (FNAB) [9]. The American Radiology Association has defined the ultrasonographic features of nodules associated with malignancy as hypoechogenicity of the nodule, absence of a peripheral halo, presence of microcalcification, irregular border, anteroposterior diameter/transverse diameter >1, and increased central vascularity [10]. The sensitivity (Sn), specificity (Sp), and accuracy of these features vary among studies [3]. Therefore, the use of USG alone is not sufficiently sensitive or specific to identify malignant thyroid nodules. However, the nodule features mentioned above increase the malignancy rate [3].

Thyroid FNAB is the primary method for identifying malignancy in patients with thyroid nodules [11]. In iodine-sufficient regions, Sn and Sp of this method are both >90 % [9]. In 2007, the National Cancer Institute (NCI) organized the NCI Thyroid Fine Needle Aspiration State-of-the-Science Conference and discussed various aspects of thyroid FNAB [12]. Consequently, the Bethesda system (BS) for reporting thyroid cytopathology was described at this conference. According to this terminology, FNAB results are divided into six categories: (1) nondiagnostic, (2) benign, (3) atypia/follicular lesion of undetermined significance (AUS/FLUS), (4) follicular neoplasm/suspicious of follicular neoplasm, (5) suspicious of malignancy, and (6) malignant [13, 14]. By definition, AUS/FLUS involves a heterogeneous category that includes structural, cellular, and nuclear atypia of the samples [11]. The BS recommends that use of this category be limited and should not exceed 7 % of the diagnoses within a given laboratory [13, 15]. The malignancy rate in this category has been estimated to be 5–15 % [14]. The aim of this category was to group lesions that were not necessarily suspected of representing a follicular carcinoma (high cellularity/scant colloid, microfollicular/trabecular arrangements) but did not meet the criteria for the benign or malignant categories [16]. This category includes lesions that are not easily classified as benign, suspicious, or malignant [16]. This system makes it possible to repeat the biopsy for patients who have follicular atypia of a thyroid nodule. If the same diagnosis is determined (20–25 % of cases) then surgery is usually recommended [14].

There are no highly sensitive and specific diagnostic procedures available to evaluate indeterminate thyroid nodules [17]. In these cases, the radiologic and cytologic features as well as the expression of cell markers are proposed to serve as diagnostic adjunctive tools in the evaluation of such nodules, but they have limited relevance [18]. The NCI conference session on ancillary techniques provided no recommendations for indeterminate lesions [19]. Potential problems included the limited data available for immunohistochemistry (IHC) in large cohort studies and the different methods used to obtain cytological specimens for immunocytochemistry (ICC) [20, 21]. The conclusions from the conference were that IHC/ICC should be considered for some suspected malignancies and for some specific diagnoses (thyroid nodule vs. parathyroid nodule), but there was insufficient evidence to determine a specific recommendation for indeterminate/suspicious FNAB [20, 21]. However, several studies have shown that IHC markers may improve the accuracy of cytologic diagnosis. Fadda et al. showed that HBME-1 and galectin-3 are positive in 83.3 % of the cases resulting in malignancy and are negative in 87.5 % of cases resulting in benign histopathology [20]. Among the cases diagnosed as follicular neoplasia (FN) and AUS, a completely positive immunocytochemical panel was found in 76.9 % of cases resulting in malignancy, and a complete negative panel was found in 96.8 % of cases resulting in benign histology [20]. They found that HBME-1 had a better diagnostic accuracy than galectin-3 and suggested applying these markers in the cytological diagnosis of FN/AUS to determine the true preoperative diagnosis of thyroid nodules [20]. Cochand-Priollet et al. [21] found that 4 % of ICC results are false positives, but none false negative, for indeterminate thyroid lesions. In indeterminate thyroid lesions, they found that the Sn, Sp, negative predictive value (NPV), and positive predictive value (PPV) for malignant–benign lesions were 100, 85.2, 100, and 86.2 %, respectively. They suggested that ICC of thyroid FNAB using cytokeratin 19 and HBME-1 antibodies can reduce the false-positive and false-negative results of single morphological analyses, consequently improving the diagnostic accuracy and reducing the need for surgical controls [21]. Nikiforov et al. [6] studied 1,056 consecutive thyroid FNA samples with indeterminate cytology (to determine BRAF, RAS, RET/PTC, and PAX8/PPARγ) and found that detecting any mutations in the atypia/FLUS group was consistent with 88 % malignancy, whereas the risk of malignancy was 14 % when based only on cytology. Therefore, they recommended total thyroidectomy in mutation-positive cases and lobectomy in mutation-negative cases.

The AUS/FLUS category includes heterogeneous patterns of atypia. Some authors suggest that the different patterns are associated with a higher risk of malignancy [22–25]. In this study, we evaluated the malignancy rates in the AUS/FLUS category and determined whether there were different malignancy rates among the different patterns in this category. We also evaluated the presence of biochemical, clinical, and echographic features that are potentially predictive of malignancy related to AUS and FLUS.

Materials and Methods

Patients

Patients who visited our endocrinology clinics between January 2010 and March 2012, who were found to have at least one thyroid nodule by USG, underwent USG-guided FNAB because of solitary nodules or multinodular goiter according to current guidelines [8], and whose cytology was identified as AUS/FLUS according to the BS were evaluated retrospectively. Patients who had suspected ultrasonographic features or clinical features such as large nodule size or growth, mass effect, and a family history of thyroid cancer underwent surgery. Patients were divided into groups according to thyroid function test results: euthyroid, hypothyroid, hyperthyroid, subclinical hypothyroid, and subclinical hyperthyroid. Patients with overt hyperthyroidism were divided into two groups according to their scintigraphic results: toxic nodular or multinodular goiter, and toxic diffuse nodular or multinodular goiter.

Exclusion Criteria

Patients <15 years and those who had a history of thyroid operation, percutaneous intervention, or radiotherapy of the head and neck were excluded from the study. Patients who had a contraindication for surgery due to comorbid diseases (cardiovascular or respiratory system diseases) or who refused surgery were also excluded.

Laboratory

The levels of sensitive thyroid-stimulating hormone (sTSH), free triiodothyronine (fT3), free tetraiodothyronine (fT4), thyroid autoantibodies (thyroid peroxidase antibody [anti-TPO] and thyroglobulin antibody [anti-Tg]), and thyroglobulin were measured in all patients using chemiluminescence methods (Immulite 2000, Diagnostic Products Corp., Los Angeles, CA, USA and UniCel DXI 800, Beckman Coulter, Brea, CA). The normal ranges for sTSH, fT3, fT4, anti-Tg, and anti-TPO were 0.4–4 μIU/mL, 1.57–4.71 pg/mL, 0.61–1.12 ng/dl, <30 U/mL, <10 U/mL, and 0–55 ng/mL, respectively. Suppressed TSH levels with normal free thyroid hormone levels were described as “subclinical hyperthyroidism,” and suppressed TSH levels with high free hormone levels were described as “overt hyperthyroidism.” In contrast, elevated TSH with normal free thyroid hormone levels was categorized as “subclinical hypothyroidism,” and elevated TSH with reduced hormone levels was described as “over thypothyroidism.”

Conventional USG and Scintigraphy

Esaote color Doppler USG (Model 796FDII; MAG Technology Co. Ltd., Yung-Ho City, Taipei, Taiwan) and standard USG with a superficial probe (Model LA523 13–4, 5.5–12.5 MHz) were used. The patient was in a supine position with their neck hyperextended, and the skin was coated with acoustic material. Localization, diameter (in milliliter), and volume of the nodule; components; echogenicity; border regularity; calcification; and the presence of a peripheral halo were evaluated by conventional USG. Parenchyma of the thyroid gland was evaluated as homogenous or heterogeneous. Echogenicity was evaluated as hypoechoic, isoechoic, iso-hypoechoic, or hyperechoic. Nodule components were described as solid (no cystic component), mixed (containing cystic part), or pure cystic (almost cystic or very little solid part). Calcification of the nodule was defined as microcalcification, macrocalcification, micro–macrocalcification, or no calcification.

99mTc pertechnetate was used for the scintigraphic evaluation of the thyroid gland. The patients were administered 5 mCi of 99mTc pertechnetate intravenously 20–40 min after radioiodine administration. Thyroid scintigraphy was carried out using a gamma camera with a pinhole collimator. Of the 276 patients, thyroid scintigraphy was applied to 71 (80.6 %) in the FLUS group and 57 (12.1 %) in the AUS group. According to the scintigraphic results, nodules were divided into hypoactive, hyperactive, and normoactive groups.

USG-Guided FNAB

FNAB was performed with USG using a General Logic Pro 200 system (model 2270968; GE Healthcare, Seoul, Korea and Seongnam SI, Gyeonggi-do, Korea) and a 5.5–7.5-MHz superficial probe. Written consent was obtained from all patients prior to FNAB. No anesthetics were administered before the procedure. USG-guided FNAB was performed in all patients by an experienced endocrinologist using a 23-gauge needle and 20-mL syringe.

Cytological and Histopathological Examination

Materials obtained by USG-guided FNAB were air dried, stained with May–Grünwald–Giemsa, and evaluated according to the BS classification [12]. The cytopathological diagnosis was made on conventional smears. Cell blocking was not performed. Accordingly, the cytology results were as follows: (1) nondiagnostic, (2) benign, (3) AUS/FLUS, (4) follicular neoplasm/suspicious for follicular neoplasm, (5) suspicious for malignancy, and (6) malignant [13, 14].

Although AUS/FLUS was categorized using the same classification as the BS, conventional smears of AUS/FLUS cases were subclassified into two patterns based on cytomorphology. AUS cases were defined as having an atypical cell pattern (could not exclude papillary thyroid carcinoma, such as the presence of occasional nuclear grooves, an abnormal chromatin pattern, or nuclear overlapping and crowding) (Fig. 1), and FLUS cases were defined as a microfollicular pattern with low cellularity and no or minimal colloid (could not exclude follicular and Hurthle cell neoplasm) (Fig. 2).

Fig. 1
figure 1

May–Grunwald–Giemsa (MGG) staining. Atypical cells with nuclear membrane irregularity and with nuclear larging (×40 objective)

Full size image
Fig. 2
figure 2

May–Grunwald–Giemsa (MGG) staining. Hypocellular and microfollicular pattern of the cells (×40 objective)

Full size image

The histopathological evaluation was made according to the 2004 World Health Organization criteria [26]. The differential diagnoses for the final histopathological diagnosis were made using specimens stained immunohistochemically for HBME-1, galectin-3, Ki-67, and/or CK19.

Statistical Analysis

Shapiro–Wilks tests were used for the constant variables. Descriptive statistics for the constant variables are expressed as mean ± standard deviation or median (range), and categorical variables are noted as numbers and percent.

Significant differences between the means of different groups were evaluated by Student’s t tests, and significant differences in median values were evaluated by Mann–Whitney U tests. Categorical variables were evaluated by Pearson’s chi-square or Fisher’s tests. The odds ratio and 95 % confidence intervals of all factors that could affect malignancy rates were evaluated.

Following the univariate statistical analysis, risk factors with p < 0.025 were subjected to a multivariate model. According to the gradual logistic regression analysis, the most predictive risk factors for malignancy were determined. p values ≤ 0.05 were considered statistically significant.

Results

In total, 276 nodules from 268 patients were evaluated; 219 (81.7 %) patients were female and 49 (18.3 %) were male. The mean age (range) of the patients was 47.85 ± 11.64 (15–74) years. FNAB samples were divided into the AUS and FLUS groups according to the BS. The mean age of the patients was 48.38 ± 11.41 (15–74) years in the AUS group and 46.76 ± 12.1 (17–74) years in the FLUS group. In the AUS group, 150 (83.3 %) patients were female and 30 (16.7 %) were male, and in the FLUS group, 69 patients (78.4 %) were female and 19 (21.6 %) were male. No significant differences were observed between the two groups in terms of age or gender.

According to the thyroid function test and USG results, euthyroid multinodular goiter was the most common finding. In the AUS group, 185 nodules from 180 patients, and in the FLUS group, 91 nodules from 88 patients were evaluated retrospectively. According to the thyroid scintigraphy results, more hypoactive nodules tended to be seen in the FLUS group than those in the AUS group, but the difference was not significant. The sonographic features of the nodules were not different between the two groups. In the AUS group, the mean transverse diameter of nodules was 21.73 ± 13.5 mm in the benign group and 19.09 ± 9.79 mm in the malignant group; in the FLUS group, the values were 21.98 ± 11.02 and 27.65 ± 18.4 mm, respectively. These differences were not significant.

Histopathologic Results

The histopathologic findings are listed in Table 1.

Table 1 Comparison of histopathologic results between groups
Full size table

The most common malignancy was papillary thyroid carcinoma (PTC) (Fig. 3), and the most common PTC subtype was follicular variant of PTC in all groups. No significant differences in vascular, capsular, extracapsular, lymph node metastasis, or multicentricity were observed in the malignant histopathologic groups among AUS, FLUS, or AUS + FLUS. Mean tumor size was 1.64 ± 1.39 (0.1–6) cm. No significant difference in tumor size was observed between AUS and FLUS. No significant difference was observed between single and repeat FNAB after comparing the benign (Fig. 4) and malignant histopathologic results according to repeated FNAB.

Fig. 3
figure 3

Classical variant of papillary thyroid carcinoma (with hematoxylin–eosin staining) (×40 objective)

Full size image
Fig. 4
figure 4

Colloidal nodule with hyperplastic changes (with hematoxylin–eosin staining) (×40 objective)

Full size image

Mean age was significantly lower in the malignant group than that in the benign group in all histopathologic groups (p = 0.01). No significant differences were observed for sex, thyroid function tests, or the scintigraphic results.

Hypoechoic nodule rates and peripheral vascularization rates were significantly higher in the malignant group than those in the benign group according to the ultrasonographic features (p values = 0.02 and 0.011, respectively). Isoechoic nodule rates were significantly higher in the FLUS group (77.8 %) than those in the AUS group (48.9 %) (p = 0.036). The ultrasonographic features of the malignant and benign groups are presented in Table 2.

Table 2 Ultrasonographic features of the malignant and benign histopathologic results
Full size table

Predictive factors for distinguishing between malignant and benign histopathologic results were evaluated by multivariate logistic regression analysis. Age, sex, thyroid function tests, thyroid antibody positivity, nodule size, and ultrasonographic features of the nodules were evaluated. The results are given in Table 3.

Table 3 Predictive factors for differentiating between malignant and benign histopathologic results
Full size table

According to the multivariate logistic regression analysis, the only predictive features of malignancy were hypoechogenicity in the AUS group (PR 4.714, 95 % CI 1.732–12.835, p = 0.004), and peripheral vascularization in the FLUS group (odds ratio 3.460, 95 % CI 1.0007–11.886, p = 0.049).

Discussion

We investigated the malignancy rate of AUS/FLUS thyroid nodules and whether there were different malignancy rates among the different patterns in this category. We also evaluated the presence of biochemical, clinical, and echographic features that were potentially predictive of malignancy related to AUS and FLUS. The BS recommends that use of this category be limited and should not exceed 7 % of the diagnoses within a given laboratory [13, 15]. In another study which is carried out in our center, revealed that our institution’s overall AUS/FLUS diagnostic rate was 4.8 % [27]. We found malignancy rates of 24.3 % in AUS, 19.8 % in FLUS, and 22.8 % in the AUS + FLUS groups. The malignancy rate was not different between the groups. The most common malignancy was PTC in the AUS group, whereas it was PTC and well-differentiated thyroid neoplasm in the FLUS group. The most common PTC subtype in both groups was the follicular variant.

The malignancy rate for AUS/FLUS is thought to be 5–15 % [11], but the true rate is uncertain because not all nodules are subjected to histopathological evaluation. In previous studies that included a cytologic follow-up, the malignancy rate was low, but in a large series, it was 27.5 % [28]. The malignancy rate was 6–48 % among operated cases [11, 29]. In one study, the malignancy rate was 25 % in indeterminate cytology [30], yet in another study, it was 21.8 % [2]. Because fewer patients underwent surgery, the actual malignancy rate in the FLUS group is conjectural, although it has been reported to be 0–35 % [18, 31]. Nayar et al. reported a malignancy rate of 6 % [32], Yassa et al. reported 24 % [33], and Yang et al. reported 19 % [34] in a FLUS group. In a AUS group, the malignancy rate was 25 % in patients who underwent operations. Nevertheless, the actual incidence is 5–10 % in nonoperated cases [31]. The malignancy rates of AUS/FLUS reported in the literature are listed in Table 4.

Table 4 Malignancy rates of thyroid nodules defined as AUS/FLUS by FNAB
Full size table

In our study, the malignancy rate of AUS/FLUS was 22.8 %, which was similar to some values in the literature [2, 16, 30, 38], although this ratio was higher than that determined by the BS. Our malignancy rate for FLUS (19.8 %) was higher than that reported by Nayar et al. [32], lower than that of Yassa et al. [33], and similar to that of Yang et al. [34]. Our rate was higher than the BS probably because we studied patients who underwent surgery. Another factor contributing to the high ratio may be the heterogeneity of AUS/FLUS. Additionally, some studies predate the BS and then adopted the BS [18]. Furthermore, specimen preparation methods could contribute to such differences [28], and different patterns of AUS/FLUS may have different malignancy risks [11]. Some authors have suggested that different patterns are associated with a higher risk of malignancy [22–25]. Önder et al. evaluated the frequency and malignancy outcomes of indeterminate categories and classified AUS/FLUS into four subcategories: PTC pattern, microfollicular pattern, Hurthle cell pattern, and atypical cell pattern [39]. They found that the malignancy rates for the PTC, atypical cell, and microfollicular pattern were 28, 22.2, and 6.9 %, respectively. The Hurthle cell category did not show any malignant lesions [39]. Renshaw et al. [23] reported a malignancy rate of 38 % in AUS/FLUS with PTC features, whereas Luu et al. [22] and Olson et al. [24] reported malignancy rates of 45.8 and 48 %, respectively, for this category. Choi et al. [40] subclassified the AUS/FLUS category into two subcategories according to previous studies [24, 31, 35, 41]: nuclear atypia (AUS group) and low cellularity with a predominant microfollicular pattern and no or minimal colloid (FLUS group). They showed that the AUS group had a significantly higher risk of malignancy (65 vs. 14.3 %) as well as malignant USG findings (76.5 vs. 52.7 %) compared with those in the FLUS group. The total malignancy rate was 32.4 % [40]. They recommended that clinical follow-up may be adequate for managing FLUS nodules [40]. They also suggested that the guidelines for managing the AUS/FLUS category clinically may be further reevaluated because of the different malignancy risks, management recommendations, and malignant USG findings between the two groups [40]. Considering all of these factors, we conclude that the definition of this category in the BS does not reflect agreement between pathologies.

We compared the findings of repeat FNAB and found no statistically significant difference between the benign and malignant histopathological groups. In a previous study, patients with a single AUS diagnosis who went directly to surgery showed similar malignancy rates to those who had two consecutive AUS diagnoses [28]. Another study reported that the malignancy rate was 19 % in AUS/FLUS and 27 % in the repeat FNAB group [38].

We found that the mean age of the patients was younger in the malignant group than that of the benign group in the total AUS + FLUS and AUS groups, whereas no significant difference was observed in the FLUS group. No significant differences were observed between histopathologic groups and sex, thyroid function tests, Tg levels, or scintigraphic features of nodules in the AUS, FLUS, and AUS + FLUS groups. In a previous study, which was similar to our study, no differences between histopathology results and sex and thyroid function tests were observed [31]. Nevertheless, there is no agreement among studies for the relationship between age and malignancy rate. In one study, the malignancy rate was higher in patients <40 years old [42]. Similarly, in another study [43], patients with thyroid malignancy were younger than patients in the benign group. However, other studies have suggested the opposite, reporting higher rates of thyroid malignancy with increasing age [5, 44].

Studies have suggested that nodule size may assist in cancer risk assessment, although data are conflicting, and some studies have claimed that nodule size is not associated with malignancy rate [45–50]. Although similar to previous findings [18, 30, 31, 42, 51, 52], we did not observe significant differences between nodule diameter and malignancy rate. In a large series of nodules with indeterminate or suspicious cytology, a threshold effect was detected at a nodule diameter of approximately 2.5 cm, and it has been suggested that the malignancy risk increases by up to 39 % for every 1 cm increase in nodule size [53].

Gray-scale ultrasonographic patterns suggesting malignancy include solid hypoechoic appearance, presence of microcalcification, and being anteroposterior or with a transverse diameter of >1 [43]. Our results revealed that among the ultrasonographic features in the AUS/FLUS group, hypoechogenicity and peripheral vascularization were significantly higher in the malignant group. In the AUS group, hypoechogenicity, peripheral vascularization, and the solid component were significantly higher in the malignant group. Isoechoic nodule rates were higher in the FLUS group than in the AUS group, but the difference was not statistically significant. Rago et al. [2] reported that the malignancy rate in nodules with indeterminate cytology was 21.8 %, and that gray-scale USG alone was insufficient for distinguishing between malignant and benign lesions. In that study, the absence of a halo and the presence of microcalcification features were the most relevant to malignancy. The main objective of our study was to evaluate the biochemical, clinical, and echographic features one by one in AUS, FLUS, and in both groups related to malignancy.

The limitations of the present study are that it was retrospective and we selected patients who had undergone an operation. The predictive features of malignancy included hypoechogenicity and peripheral vascularization of the nodule in the total AUS + FLUS group, but only hypoechogenicity in AUS and peripheral vascularization in the FLUS group. These findings support that ultrasonographic features alone may be insufficient for predicting malignancy; therefore, all clinical and ultrasonographic features must be considered. Another restriction of our study was the lack of immunohistochemical markers. As mentioned above, markers may be promising for diagnosing thyroid nodules, particularly in cytologically indeterminate lesions.

The suggested recommendation is repeat FNAB for cases that have AUS/FLUS cytology, due to the low malignancy rates, except pure cystic cases [11]. This approach is considered safe and cost effective [11]. Clinical and radiological correlations are also important [11]. However, we revealed no difference in malignancy rate between single and repeat FNAB, and that the overall malignancy rate was higher than expected. Therefore, we suggest that surgery may be conceivable earlier instead of repeat FNAB especially with suspected ultrasonographic features.