Thyroid Imaging Reporting and Data System Score Combined with the New Italian Classification for Thyroid Cytology Improves the Clinical Management of Indeterminate Nodules

The new Italian cytological classification (2014) of thyroid nodules replaced the TIR3 category of the old classification (2007) with two subclasses, TIR3A and TIR3B, with the aim of reducing the rate of surgery for benign diseases. Moreover, thyroid imaging reporting and data system (TI-RADS) score appears to ameliorate the stratification of the malignancy risk. We evaluated whether the new Italian classification has improved diagnostic accuracy and whether its association with TI-RADS score could improve malignancy prediction. We retrospectively analyzed 70 nodules from 70 patients classified as TIR3 according to the old Italian classification who underwent surgery for histological diagnosis. Of these, 51 were available for cytological revision according to the new Italian cytological classification. Risk of malignancy was determined for TIR3A and TIR3B, TI-RADS score, and their combination. A different rate of malignancy (p = 0.0286) between TIR3A (13.04%) and TIR3B (44.44%) was observed. Also TI-RADS score is significantly (p = 0.003) associated with malignancy. By combining cytology and TI-RADS score, patients could be divided into three groups with low (8.3%), intermediate (21.4%), and high (80%) risk of malignancy. In conclusion, the new Italian cytological classification has an improved diagnostic accuracy. Interestingly, the combination of cytology and TI-RADS score offers a better stratification of the malignancy risk.


Introduction
Palpable thyroid nodules are present in about 4-7% of the adult population in countries with adequate iodine intake and up to 20% in iodine insufficient areas [1,2]. Highresolution ultrasound can detect thyroid nodules in 19-68% of randomly selected individuals with higher frequencies in women and elderly [2,3]. Since about 5% of patients affected by thyroid nodules harbour a malignant lesion, the first aim in their evaluation is to exclude malignancy [4].
To this end, fine-needle aspiration cytology (FNAC) represents the diagnostic cornerstone because of its diagnostic accuracy, reproducibility, and cost-effectiveness [5][6][7][8]. However, FNAC is characterized by a grey diagnostic area in which the indeterminate cytology precludes a distinction between benign and malignant lesions [9]. Surgical excision, frequently necessary to obtain a definitive diagnosis, shows that about 80% of these patients harbour a benign lesion [9]. In order to reduce unnecessary thyroidectomy, a number of instrumental and molecular diagnostic approaches have been proposed [10][11][12][13][14][15][16][17][18][19][20][21][22][23][24][25][26][27]. In addition, new classification systems for thyroid cytology have been designed. In particular, in the Bethesda System for Reporting Thyroid Cytopathology (BSRTC), indeterminate nodules have been subcategorized in atypia or follicular lesions of undetermined significance (AUS/FLUS) and in follicular neoplasms or suspicious for follicular neoplasms (FN/ SFN) which are expected to correspond to different rate of malignancy with different clinical action required [23]. In 2009, the British Thyroid Association-Royal College of Pathologists (BTA-RCPath) revised the previous British reporting system already in use in UK since 2007 along the lines of the BSRTC and split the Thy 3 category in Thy 3a and Thy 3f [28]. In 2014, the Italian Society for Anatomic Pathology and Cytology (SIAPEC) together with the Italian Thyroid Association (AIT) modified the previous thyroid cytology classification of 2007, by replacing the TIR3 class with two new subclasses, TIR3A and TIR3B [24][25][26]. The latter are comparable both to the BSRTC AUS/FLUS and FN/SFN classes and to the BTA-RCPath classes Thy 3a and Thy 3f [26]. However, unlike the BSRTC and BTA-RCPath, the SIAPEC extends TIR3B to include those cases with "mild/focal nuclear atypia" suggestive of papillary carcinoma that are expected to be at higher risk of malignancy [24].
The present study was conducted with the aim of evaluating whether the malignancy rate of TIR3A and TIR3B differs, thus improving the clinical management of patients with lesions classified in the TIR3 category in the previous classification. To this end, cytological smears of 51 nodules from patients thyroidectomized following TIR3 diagnosis were independently reevaluated by the same three cytopathologists, who made the initial TIR3 cytological diagnosis, according to the new SIAPEC 2014 classification. In addition, we evaluated whether the thyroid imaging reporting and data system (TI-RADS) score, either alone or in combination with the SIAPEC 2014 cytological diagnosis, could be of any value in predicting malignancy.

Materials and Methods
2.1. Patients. The series comprises 70 patients (52 females and 18 males, median age 58, range 13-77 yr) who underwent fine-needle aspiration cytology (FNAC) in the period between January 2005 and December 2013. Patients provided their written informed consent. All of them had a TIR3 diagnosis based on the old Italian cytological classification (SIAPEC 2007) and were submitted to surgical excision for histological diagnosis, as shown in Table 1. Of the 70 cases, cytological smears were available for 51 patients. Of the latter, one patient with histological diagnosis of uncertain malignant potential was excluded from the analysis. All smears were independently reevaluated by the same three cytopathologists (FN, VA, and DB), who made the initial TIR3 cytological diagnosis, according to the new SIAPEC 2014 classification, blinded for histology.

Ultrasound and
Color-Flow Doppler. Thyroid ultrasonography (US) was performed in 69 patients using a Toshiba Aplio XV system equipped with a linear transducer (PLT-805AT). The nodule classification was based on echogenicity and echostructure: solid hypoechoic, solid isoechoic, solid hyperechoic, mixed, or anechoic. Anteroposterior (APD), transverse (TD), and longitudinal (LD) diameters of the nodules were used to obtain the volume of the nodules, based on the formula of ellipsoid: Volume = APD × TD × LD × π/6. Nodule margin was defined regular or irregular. Microcalcifications, defined as hyperechoic spots < 22 mm, were recorded. The pattern of nodular vascular signal was evaluated by color-flow Doppler (CFD) and defined as CFD 1, as an absent signal; CFD 2, as a perinodular spot signal; and CFD 3, as a perinodular and/or intranodular signal. Ultrasound examinations were performed by two observers (AN, EDA) with an agreement on the US and CFD characteristics greater than 95%. US features, such as hypoechogenicity, irregular margins, microcalcifications, and taller-than-wide shape, were used to calculate the TI-RADS score as described by Kwak et al. [16].
2.3. Fine-Needle Aspiration Cytology. All patients were instructed not to take aspirin or any other anticoagulants 5 days prior to thyroid nodule aspiration. A 23-27 gauge needle, attached to 20 ml plastic syringes, was used to aspirate nodules. All aspirates were smeared directly on 4-6 glass slides and stained by May-Grunwald-Giemsa and Papanicolaou. Cytological specimens of the 70 patients had been evaluated by three cytopathologists (FN, VA, and DB) from the same institution. All 70 nodules had a TIR3 cytological diagnosis, based on the SIAPEC 2007 classification: TIR1, nondiagnostic; TIR2, negative for malignant cells; TIR3, indeterminate (follicular lesion); TIR4, suspicious for malignancy; and TIR5, diagnostic of malignancy [24]. Of these 70 cases, cytological smears were available for 51 nodules, which were reevaluated collegially by the same three cytopathologists according to the new SIAPEC 2014 classification [24]. In particular, these cases were reevaluated and classified either as TIR3A (low-risk indeterminate lesion) or as TIR3B (highrisk indeterminate lesion). Discordant diagnosis was resolved by a consensus review.

Histological Outcome.
Histological diagnoses of patients who had undergone surgery were used as gold standard for correlation with the cytological interpretations according to the WHO classification currently in use (Table 1) [29].
2.5. Statistical Analysis. Patients' age and mean nodular volume in benign versus malignant thyroid nodules were compared by the nonparametric Mann-Whitney U test, while clinical, US, and CFD characteristics were compared by the χ 2 test or the Fisher exact test. The statistical significance was set at p < 0 05.

Results
Following total thyroidectomy, histological diagnosis showed that, of the 70 patients with indeterminate lesions, 17 (24.3%, 13 females and 4 males) had a malignant lesion. Of these, 2 had follicular thyroid carcinoma (FTC) and 15 papillary thyroid carcinoma (PTC), of which 12 classical and 3 follicular variants, and 1 had a well-differentiated tumor of uncertain We initially evaluated the role of gender, patient's age at diagnosis, size of the lesion, ultrasound, and color-flow Doppler features in predicting malignancy. No significant association was observed for gender, median age at diagnosis, echostructure, nodularity, and color-flow Doppler between benign and malignant lesions ( Table 2). On the other hand, lower median nodule's volume (p = 0 016), taller-than-wide nodule's shape (p = 0 046), irregular margin (p = 0 008), microcalcifications (p = 0 043), and hypoechogenicity (p = 0 021) are associated with nodule malignancy ( Table 2).
We next evaluated whether the TI-RADS score is associated with lesion's malignancy and found a positive correlation between TI-RADS score and risk of malignancy (p = 0 003) ( Table 3).
Since cytological smears were available for 51 of the 70 patients, they were independently reevaluated by the same three cytopathologists who provided the initial diagnosis according to the new SIAPEC 2014 classification, in which the TIR3 category was replaced by two subclasses, TIR3A and TIR3B. A concordant diagnosis was reached in 32 cases. The 19 (37.2%) cases with discordant diagnosis were resolved by a consensus review. From the 51 cases, however, the single patient with histological diagnosis of uncertain malignant potential was excluded from the analysis. The results showed a difference in the rate of malignancy (p = 0 0286) between the TIR3A and TIR3B lesions (Table 4).
In the attempt to stratify indeterminate lesions according to the risk of malignancy, we combined cytology and TI-RADS score. In particular, the TIR3A and TIR3B categories were dichotomized based on TI-RADS score, as reported in Table 5. As it may be noticed, the TIR3A combined with the TI-RADS scores 3, 4a, and 4b showed a low malignancy rate (8.3%). On the other hand, the TIR3A combined with the TI-RADS scores 4c and 5, as well as the TIR3B combined with the TI-RADS scores 3, 4a, and 4b, showed an intermediate risk of malignancy (20.9%). Lastly, the TIR3B combined with the TI-RADS scores 4c and 5 showed a high risk of malignancy (80%). These results lead us to propose a new stratification of the risk of malignancy for indeterminate lesions in low, intermediate, and high as reported in Table 6.

Discussion
Over the last years, major efforts have been made to generate new classification systems for thyroid cytology in order to ameliorate the diagnostic stratification of nodules with indeterminate cytology, a grey area of fine-needle aspiration cytology [23][24][25][26][27][28]. In particular, all the new systems have  [23][24][25][26][27][28]. However, a study from Brophy and colleagues, using the BTA-RCPath classification system on 151 Thy 3 nodules with histological diagnosis, found no difference in the malignancy rate of lesions classified as Thy 3a and Thy 3f [30]. In particular, although the authors observed a slightly higher malignancy rate in Thy 3f cases (17.9%) than in Thy 3a cases (13.4%), this difference was not significant. In addition, similar findings emerged from a recent meta-analysis of 51 studies, using the BSRTC classification system on a total of 4475 AUS/FLUS and 3202 FN/ SFN nodules, showing a 27% rate of malignancy for the AUF/FLUS FNAC and 31% for the FN/SFN FNACs [31].
In the latter of relevance is the high rate of malignancy observed in the AUS/FLUS categories, with respect to the expectations of BSRTC. In this context, studies attempting to perform a 2-tier subclassification of the AUS/FLUS categories in order to achieve a more accurate estimate of the risk of malignancy have been reported [32,33]. The data reported here are not in agreement with the above studies because in our series, a significant difference could be observed between the rates of malignancies of TIR3A and TIR3B cases. In fact, in TIR3A category, we found a slight higher rate of malignancy (13%), with respect to the expected one (<10%), while TIR3B category had a higher malignant rate (44%), with respect to the expected one (20%) [26]. These data demonstrated that the new SIA-PEC 2014 provides a better stratification of the malignancy risk for indeterminate lesions, with respect to previous SIA-PEC 2007 classification [26]. With respect to the BSRTC and BTA-RCPath, it must be emphasized that, in the SIAPEC 2014 Italian classification, the TIR3B subcategory includes samples characterized by nuclear alterations suggestive of papillary carcinoma that are too mild or focal to be included in the TIR4 category. Even if our data derive from a relatively small series of cases, they are in line with the referred increasing evidence that the AUS/FLUS with cytological atypia is the AUS/FLUS subcategory most frequently associated with malignancy [34]. It has to be said, however, that the consensus review of the cytological smears by three different cytopathologists (also suggested by the BSRTC in challenging cases) could have affected the outcome of TIR3 subclassification in TIR3A and TIR3B, representing a possible confounding   factor with respect to one observer performance [35,36]. In this context, it is also worth mentioning that, in 37.2% (in 19 out of 51) of cases, a discordant diagnosis was made by the three different cytopathologists, which was then resolved by a consensus review of the cytological smears. This is not surprising since it is well known in the literature that the major interobserver discordance is observed in the indeterminate lesions [37]. However, also with this classification, more than half of TIR3B patients undergo unnecessary surgery. The ability of TI-RADS score in predicting thyroid nodule malignancy has been demonstrated [17]. This was confirmed in the present study also for indeterminate lesions, in which a significant association between high-risk TI-RADS score and malignancy is observed. For this reason, we attempted to ameliorate the accuracy of the SIAPEC 2014 thyroid cytological classification by combining it with the TI-RADS score. In particular, dichotomizing the TIR3A and TIR3B categories based on low-risk TI-RADS score (3, 4a, and 4b) and high-risk TI-RADS score (4c and 5), the risk of malignancy for indeterminate lesions could be stratified in three classes: low (below 10%), intermediate (about 20%), and high (about 80%). Similar results were recently reported by Maia and colleagues using the Bethesda system and by Chng and colleagues using the BTA-RCPath system [38,39].

Conclusions
Compared to the old SIAPEC 2007, the new SIAPEC 2014 thyroid cytological classification has improved diagnostic accuracy, reducing the numbers of patients with indeterminate lesions requiring surgery. The combination of SIAPEC 2014 thyroid cytological classification and TI-RADS score could offer a better stratification of the malignancy risk suggesting a conservative approach for low-risk class and a surgical approach for high-risk class. For patients with intermediate risk, a careful evaluation of risk factors for thyroid malignancy and a close follow-up is recommended.