American College of Radiology thyroid imaging report and data system combined with K-RAS mutation improves the management of cytologically indeterminate thyroid nodules

We investigated whether use of American College of Radiology thyroid imaging report and data system (ACR TIRADS) in combination with K-RAS mutation status may facilitate risk stratification of patients with cytological Bethesda Category III and IV thyroid nodules. Ultrasonographic, cytological, and histopathological diagnoses were retrospectively correlated with K-RAS mutation status in a series of 43 cytologically indeterminate thyroid nodules (CITNs) that were referred for surgical excision. K-RAS mutations were detected in 8/43 (18.6%) fine-needle aspiration (FNA) samples as against 11/43 (25.6%) surgical specimens. ACR TIRADS level (TR) TR3 lesions had a malignancy risk of 40%; the K-RAS mutation rate in FNA samples and surgical specimens of category TR3 lesions was 40% and 60%, respectively. K-RAS mutation-positive malignancy was significantly more frequently detected in follicular neoplasm/suspicious for follicular neoplasm (FN/SFN) lesions than that in atypia or follicular lesion of undetermined significance (AUS/FLUS) (P<0.01). Combined use of ACR TIRADS (TR5 as the diagnostic threshold) and K-RAS mutation status helped identify 83.3% (10/12) malignant nodules (58.6% specificity, 45.5% positive predictive value, 89.5% negative predictive value, and 65.9% accuracy). CITNs with ACR TIRADS category TR3 showed an unexpectedly high risk of malignancy. K-RAS mutation-positive FN/SFN nodules have a 50% risk of malignancy and surgery should be recommended. Combined use of ACR TIRADS and K-RAS mutation may facilitate risk-stratification of patients with CITNs. The high negative predictive value (NPV) for malignancy seems sufficient to allow conservative management of patients with active surveillance.

Introduction Thyroid nodules are common in the general population with a malignant rate of 5-15% [1]. Ultrasonography (US) is currently recommended as the initial modality for evaluation and workup of thyroid nodules. However, the diagnostic accuracy is heavily dependent on the experience and ability of the operator to describe and interpret suspicious features. In order to minimize the interobserver variability, use of a series of thyroid imaging report and data system (TIRADS) was proposed for classification and malignant risk-stratification of thyroid nodules [2][3][4]. Fine-needle aspiration (FNA) cytology is regarded as the most efficient method for evaluation of thyroid lesions [5]. However, the cytology results may be indeterminate in up to 25% of thyroid nodules [5][6]. There is no consensus on the diagnosis and management of cytological indeterminate thyroid nodules (CITNs), i.e., Bethesda Category III lesions [atypia of undetermined significance/follicular lesion of undetermined significance (AUS/FLUS)] and Category IV [follicular neoplasm/suspicious for follicular neoplasm (FN/SFN)]. Moreover, the reported risk of malignancy in CITNs shows wide variability [5][6][7]. Owing to this ambiguity, appropriate management of CITNs is a clinical challenge. Active surveillance or diagnostic surgical resection is recommended for such cases based on the clinical and US features [7]. Investigation of molecular makers is also a useful adjunctive tool for diagnosis of CITNs [7]. BRAF and RAS point mutations are currently the most widely used markers for CITNs. RAS mutations are not specific for malignancy and have a lower positive predictive value than BRAF; however, RAS mutation status may provide useful supplemental information for management of CITNs. However, to the best of our knowledge, no study has evaluated the application of ACR TIRADS in combination with RAS mutation analysis for evaluation of CITNs. Therefore, the purpose of this study was to evaluate the performance of the ACR TIRADS in combination with RAS mutation status for identification of potential malignant thyroid nodules with indeterminate cytology.

Study population
The study protocol was reviewed and approved by the Institutional Review Board of Jiangsu Institute of Nuclear Medicine. The requirement for written informed consent of patients was waived off. Between January 2017 and June 2018, 2074 patients underwent preoperative USguided fine-needle aspiration of thyroid nodules at the department of Ultrasound, Jiangyuan Hospital affiliated to Jiangsu Institute of Nuclear Medicine (Wuxi, China). Patients who qualified the following criteria were included in this study: (1) indeterminate thyroid cytology based on the Bethesda system; (2) histological diagnosis confirmed by surgical resection; (3) availability of BRAF and K-RAS mutation status; (4) patients received an ultrasound examination of the thyroid, which was classified by ACR TIRADS. Finally, a total of 43 patients (9 male, 34 female) with 43 CITNs were included in the analysis. Anonymized patient information was acquired and stored in a research database before accessed.

US examination and ACR TIRADS
All US examinations were performed using a 5-12 MHz linear-array probe (iU22, Philips Healthcare, Bothell, WA, USA) by 1 of 10 radiologists who had 3 to 23 years of experience in thyroid imaging. A radiologist with 21 years of experience in thyroid imaging, blinded to cytopathological and histopathological diagnosis, retrospectively reviewed the US features of all CITNs.
The composition of the nodules was classified as solid or almost completely solid, mixed cystic and solid, cystic or almost completely cystic, and spongiform. Echogenicity was categorized as hyperechoic or isoechoic, hypoechoic and very hypoechoic. The shape of the nodule was classified as taller than wide (anteroposterior diameter greater than the transverse or longitudinal diameter) or wider than tall. The nodule margin was categorized as smooth, ill-defined, lobulated or irregular, or with extra-thyroid extension. Calcification was categorized as none or large comet-tail artifacts, macrocalcification, peripheral calcification, or punctuate echogenic foci. Subsequently, ACR TIRADS, a five-tier risk categorization system, was assigned to each nodule based on the aforementioned US features. TIRADS points were also recorded for each nodule.

US-guided fine-needle aspiration and detection of molecular markers
Written informed consent was obtained from all patients undergoing ultrasound-guided fine needle aspiration (UG-FNA). Following the recommendations given by surgeons, four patients with category TR2 nodules underwent UG-FNA because of enlargement of solid portion. All thyroid FNA specimens were collected under ultrasound guidance using a 23-gauge needle attached to a 2.5-mL disposable plastic syringe by a radiologist using the freehand technique. Three passes were performed for each thyroid nodule. The slides were immediately placed in 95% ethanol for Papanicolau staining. On-site cytological assessment was not available. Two of the three cytopathologists specializing in thyroid pathology interpreted the cytology reports based on the Bethesda System for Reporting Thyroid Cytopathology. After cytological preparation, the residual biopsy specimen in the syringe was used for BRAF V600E and K-RAS mutation analysis. BRAF V600E mutation analysis in FNA samples is routinely performed as an ancillary test at our institute, and K-RAS mutation analysis is performed only in case of indeterminate cytological samples.
Molecular biology techniques were processed as standard technique for clinical pathology lab. The needle tip was washed with 100 μL Buffer ATL and residual FNA material was pipetted directly into a microcentrifuge tube for molecular analysis. DNA was isolated using the QIAamp DNA Micro kit (Qiagen), according to the manufacturer's instructions. DNA extraction from formalin-fixed paraffin-embedded tissue specimens was performed as described elsewhere [8]. The methodology for analysis of BRAFV600E mutation is described elsewhere [9]. K-RAS mutation status was determined using real-time polymerase chain reaction (PCR) amplification followed by high resolution melting (HRM) analysis. The melting curve was analyzed by LightCycler 480 software (Roche Diagnostics). The position of the melting peaks was different for mutated-type and wild-type. Melting temperature (Tm) was identified by the peak position of the melting curve.

Statistical analyses
Sample size calculations had been made before gathering data. Finally, the sample size was determined by the feasibility of recruitment, and it was justified by minimally detectable effect sizes instead of power analysis. Comparisons of categorical variables were performed using Chi-squared test or Fisher's exact test, while independent two-sample t-test was used for comparison of continuous variables. The sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and accuracy of ACR TIRADS, molecular marker testing, and the combined use of both methods were calculated. Statistical analyses were performed with SPSS for Windows (ver. 23.0; IBM, Armonk, NY, USA). A p-value < 0.05 was considered indicative of statistically significant difference. diagnosed as benign (10 follicular adenomas, 17 hyperplastic nodules, 2 hurthle cell adenomas). Two (4.7%) nodules were borderline thyroid tumors, including one nodule with welldifferentiated tumor of uncertain malignant potential (WDT-UMP) and one nodule with follicular tumor of uncertain malignant potential (FT-UMP) (excluded from the statistical analysis because of indeterminate histology). Twelve (27.9%) nodules were malignant including 4 papillary thyroid carcinoma, 7 follicular thyroid carcinoma (FTC) (5 minimally invasive FTC, 2 widely invasive FTC) and 1 medullary thyroid carcinoma. The mean diameter of the benign nodules (25.5 ± 11.9 mm) was not significantly different from that of malignant nodules (26.0 ± 16.6 mm; P = 0.28). The cytological diagnoses for the 43 CITNs were AUS/FLUS (n = 18) and FN/SFN (n = 25). The malignancy rates between AUS/FLUS and FN/SFN were not significantly different [22.2% (4/18) versus 34.8% (8/23), respectively; P = 0.60].
On categorization of CITNs according to ACR TIRADS, the risk of malignancy in ACR TIRADS categories TR2, TR3, TR4, and TR5 was 0%, 40%, 6.7%, and 52.9%, respectively (P = 0.02). TIRADS TR3 and TR5 categories were more prevalent in malignant than benign group ( Table 2). The CITNs of category TR3 had a high risk of malignancy (40%), which is high relative to the generally accepted value for ACR TIRADS TR3 nodules. Representative cases are shown in Figs 1-3.

Tests for molecular markers
All FNA specimens were adequate for successful DNA extraction and detection of molecular markers. Forty-two out of the 43 nodules were negative for BRAF V600E mutation, while one BRAF-positive nodule was histopathologically diagnosed as PTC.
We examined the concordance between FNA samples and surgical specimens with respect to K-RAS mutation status. Three patients (7.0%) had discordant mutation status; 2 hyperplastic nodules and 1 FTC tested negative for K-RAS mutation in FNA samples but tested positive in surgical specimens. For surgical specimens, after excluding one case of WDT-UMP with RAS-positive mutation, K-RAS mutations were detected in 4 out of the 12 (33.3%) malignant nodules (1 PTC and 3 FTCs) and in 6 out of 29 (20.7%) benign nodules, including 1 follicular adenoma, 1 hurthle cell adenoma, and 4 hyperplastic nodules. Coincidentally, K-RAS mutated malignancy was more frequently found in FN/SFN than in AUS/FLUS (50% vs. 0%, respectively; P < 0.01). The results of K-RAS mutation according to Bethesda category are presented in Table 3.  metastasis, while none had ETE or distant metastasis. There was no significant difference in aggressiveness between the two groups (ETE, P = 0.17; lymph node metastasis, P = 0.84).

Diagnostic performance
The diagnostic performance including sensitivity, specificity, PPV, NPV, and accuracy according to ACR TIRADS, K-RAS mutation analysis, and the combination are shown in Table 4.

Discussion
Indeterminate thyroid cytology is a gray zone wherein the cytological evaluation is inconclusive. Up to 25% of thyroid FNAs cannot be categorized as either benign or malignant according to the Bethesda System for Reporting Thyroid Cytopathology [5]. The risk of malignancy is varied, ranging from 5%-15% for AUS/FLUS to 15%-30% for FN/SFN diagnosed by examination of surgical specimens [5]. Therefore, tools that help identify or exclude malignant CITNs may facilitate appropriate management and treatment. In recent years, molecular marker testing has become available as an aid to refine the diagnosis of CITNs. Even though molecular markers help in risk stratification of nodules with indeterminate cytology [10][11], there is no current consensus on the role of RAS mutation as a predictor of malignancy. In the present study, we evaluated the US features according to ACR TIRADS and examined the frequency of K-RAS mutation in CITNs. The objective was to determine whether the use of ACR TIRADS in combination with K-RAS mutation status can help provide an optimal management decision for thyroid nodules with indeterminate cytology. In a multicenter prospective study by Ha et al., presence of microcalcification, taller-thanwide tumors, and spiculated/microlobulated margins were independent predictors of malignancy [12]. However, these suspicious US features were not significantly associated with follicular thyroid carcinoma. In this study, a greater percentage of malignant nodules were widerthan-tall, had no punctuate echogenic foci, or presented with larger comet-tail artifact. Theoretically, thyroid nodules can be stratified using ACR TIRADS according to malignancy risk based on suspicious US features [4]. In the present study, use of ACR TIRADS for risk stratification of CITNs using TR5 as the diagnostic threshold was associated with 75.0% sensitivity and 72.4% specificity. Suspicious US features did not show good performance in predicting malignancy in CITNs. Utility of ACR TIRADS combined K-RAS Generally, the risk of malignancy tends to increase with advancing ACR TIRADS category. However in the present study, the malignancy risk of category TR3 lesions was higher than the indicated malignancy risk of 4.8% [95% confidence interval (CI), 3.4-6.5%] for TR3 proposed by ACR TIRADS [13], as well as the malignancy risk of category TR4 thyroid nodules. Some cases of FTCs among malignant nodules may be the underlying cause of increased malignancy risk. Two malignant CITNs categorized as TR3 were FTCs with no suspicious US features, which is an indication for active surveillance or UG-FNA follow-up; however, in the present study, these patients were referred for surgery owing to pressure symptoms or increase in size. We believe that the risk of malignancy in CITNs with category TR3 based on surgical specimens may be higher than that of ACR TIRADS TR3 nodules reported in previous reports. We believe that US follow-up is insufficient for CITNs with ACR TIRADS category TR3; however, our findings should be validated in a larger study.
Ancillary molecular analyses for common somatic mutations of thyroid cancer have been developed to improve the diagnostic accuracy of FNA for CITNs [10][11][14][15][16][17][18]. Follicular thyroid tumors commonly harbor RAS mutations, and in general, the presence of RAS mutation is associated with good prognosis. In the present study, K-RAS mutations were identified in 18.6% of the CITNs; this percentage is comparable to that reported from a large cohort study by Nikiforov et al. [9], but is lower than that reported from other series [19][20][21]. Positive K-RAS mutation may facilitate clinical management of indeterminate thyroid cytology, which is associated with a moderate risk of malignancy according to previous reports [5]; this represents a dilemma in treatment decision-making. Ten K-RAS mutated surgical specimens were found in our study; of these, 2 were diagnosed as AUS/FLUS and 8 as FN/SFN. In addition, 50% (4/8) of K-RAS mutated FN/SFN were found to be malignant at surgery. Therefore, K-RAS mutation was more frequently associated with malignancy in FN/SFN as compared to AUS/FLUS. This indicates that resection may be a better approach for such patients.
As suggested by a previous study [22], RAS mutation-positive tumors have limited aggressive behavior, including extra-thyroidal extension, lymph node metastasis, or distant metastasis. Our study also demonstrates that RAS-positive thyroid carcinoma is no more aggressive than the wild-type malignancy. However, ETE is more frequently detected in RAS-positive CITNs. The low detection rate of BRAF V600E mutation in this study is similar to that in a previous study [23] in which 5.6% (1/18) of patients with AUS/FLUS and 0% (0/25) of those with FN/SFN had BRAF V600E mutation. Although with negative BRAF mutations, we could not exclude that poor tumor behavior is associated with the coexistence of TERT promoter mutation, which was not tested in this study.
Our study affirms the results of previous studies that RAS mutation has a low sensitivity and a high specificity for detection of malignant CITNs [11,[19][20]. In this study, K-RAS mutation was found in 21.1% of benign CITNs, and a number of malignant CITNs did not harbor K-RAS mutation, which may be associated with the low sensitivity. Because of low sensitivity, K-RAS mutation is often recommended as a supplemental aid to management of CITNs. We found that use of ACR TIRADS criteria in conjunction with K-RAS mutation increased the sensitivity of K-RAS mutation from 25% to 83.3%. In other words, use of ACR TIRADS in combination with K-RAS mutation increased the sensitivity by 58.3% as compared to that of K-RAS mutation alone. We observed that the NPV of K-RAS mutation alone for malignant diagnosis was 73.5%, which increased to 89.5% when used in combination with Utility of ACR TIRADS combined K-RAS ACR TIRADS; this suggests that combined use of both methods may help avoid unnecessary thyroidectomy. The results of our small-scale preliminary study indicate that use of K-RAS analysis in conjunction with ACR TIRADS may be potentially useful for excluding CITNs that require resection. Therefore, follow-up may be recommended for patients with category TR2, TR3, and TR4 CITNs who test negative for K-RAS mutation. However, for CITNs with ACR TIRADS scores >6, further work up should be performed owing to the high risk of malignancy.
In this study, discordant false negative results for K-RAS status between FNA samples and surgical specimens were identified in samples from 3 (7.0%) patients. The false negative results may be due to insufficient number of tumor cells in the leftover material after preparation of cytological smears [24]. However, none of the cases in our study were affected by insufficient and/or poor quality DNA. Studies have shown that some genetic mutations are inhomogeneously distributed in tumor cells and that the mutated allele percentage is liable to vary within the tumor [25][26]. This may be attributable to greater heterogeneity among larger nodules with respect to mutation when compared with nodules of small size. Since only a portion of the tumor is sampled during FNA, the FNA results are less likely to sufficiently represent the surgical results. The issue of tumor heterogeneity or focal presentation suggests that due caution should be exercised during clinical decision-making solely based on molecular markers.
Some limitations of our study should be acknowledged. First, our sample size was relatively small. In contrast to BRAF mutation, K-RAS mutation is only available for CITNs. As a cancer  referal center, indeterminate cytological diagnosis is observed in up to 17% of thyroid FNAs performed at our institution. Most patients with CITNs are suggested to undergo active surveillance or repeat FNA. Second, for precise identification, we only included nodules for whom surgical results were available. So the relatively higher malignancy rate among CITNs of category TR3 may not be representative. In addition, small sample size tends to lead to overestimation of malignancy rate. Third, owing to the retrospective study design, our results may have been affected by selection bias. Additionally, in the present study, only K-RAS mutaion testing was performed. The cost-effective difference between limited mutation analysis compared to full gene expression classifer or next generation sequencing is needs to be further investigated. Although our initial results are promising, further large-scale prosepctive studies are required to validate the role of K-RAS mutation in routine clinical use.

Conclusion
In conclusion, CITNs with ACR TIRADS category TR3 were unexpectedly associated with a high risk of malignancy. Patients with K-RAS mutation-positive FN/SFN nodules exhibited a 50% risk of malignancy; therefore, resection should be considered in such cases to obtain a definitive diagnosis. In this study, use of ACR TIRADS in combination with K-RAS mutation was associated with an 89.5% NPV for malignant CITDs. Our results suggest that use of a combination of TIRADS and K-RAS mutation status can improve thyroid management and help avoid unnecessary surgery.