Comparison Between Fine Needle Aspiration and Core Needle Biopsy for the Diagnosis of Thyroid Nodules: Effective Indications According to US Findings

Thyroid nodules are initially handled by fine needle aspiration (FNA). However, the stance of thyroid core needle biopsy (CNB) still is a challenge. This study aimed to compare the diagnostic performances and conclusive rates of FNA and CNB for the diagnosis of thyroid nodules and to define effective indications of CNB. This retrospective study enrolled 1,060 consecutive thyroid nodules in 1,037 patients who underwent FNA from January 2008 to May 2008, and 462 consecutive nodules in 453 patients who underwent CNB from January 2014 to December 2015 at our institution. Ultrasound (US) features were classified according to the American College of Radiology Thyroid Imaging Reporting and Data System (ACR TI-RADS) and Korean TIRADS (K-TIRADS). We compared diagnostic performances and conclusive rates between FNA and CNB groups. Propensity score matching was conducted to match FNA patients with CNB patients. After matching, the diagnostic performances for selecting surgical candidates and predicting malignancy were comparable between the two biopsy groups. Based on US findings, conclusive results were obtained significantly more in CNB than in FNA when thyroid nodules were classified as ACR TI-RADS or K-TIRADS category 4 and measured larger than 2 cm. Diagnostic performances between FNA and CNB were comparable. Superiority of CNB to FNA was found for thyroid nodules larger than 2 cm and classified as ACR TI-RADS or K-TIRADS category 4.


Results
Clinicopathological and US characteristics. Table 1 shows the clinicopathological and US characteristics of FNA and CNB patients before and after matching. For both groups, only age was comparable before matching (P = 0.149). In view of Bethesda or CNB diagnostic categories, none of the patients demonstrated non-diagnostic CNB results, while 13.9% of patients showed non-diagnostic FNA results. The incidences of the atypia/follicular lesion of undetermined significance (AUS/FLUS) or indeterminate, follicular neoplasm/suspicious for follicular neoplasm (FN/SFN), and suspicious for malignancy categories were higher in the CNB group than in the FNA group. The proportions of the benign and malignant categories were higher in the FNA group than in the CNB group (P < 0.001). CNB Patients had more nodules that were larger than 2 cm (P < 0.001), with solid composition (P < 0.001), ACR TI-RADS categories 3 and 4 (P < 0.001), and K-TIRADS category 4 (P < 0.001) compared with FNA patients. After FNA patients matching with CNB patients, regarding the matched variables, no differences were found between the two groups, except for previous FNA or CNB results (P < 0.001). Table 2 presents diagnostic performances of FNA and CNB to predict malignancy and to determine surgical indication (i.e. Bethesda or CNB diagnostic category IV, V, and VI).

Diagnostic performances of FNA and CNB.
Before matching, FNA had a significantly higher specificity, positive predictive value (PPV), accuracy (Ps < 0.001) and a lower sensitivity (P = 0.011) than CNB for selecting surgical candidates. Before matching, no differences were found in diagnostic performances between the two biopsy groups for predicting malignancy.
After matching, the diagnostic performances for selecting surgical candidates and predicting malignancy were comparable between the FNA and CNB groups.
Conclusive rates of FNA and CNB. Conclusive rates of FNA and CNB (Bethesda or CNB diagnostic categories II, IV, V, and VI vs. I and III) were also investigated and compared according to the final US category and nodule size. Depending on the degree of suspicion, conclusive results were obtained significantly more with CNB than with FNA when thyroid nodules were classified as ACR TI-RADS or K-TIRADS category 4 both before and after matching. For the subgroup analysis based on nodule size, the same results were also obtained for thyroid nodules larger than 2 cm after matching (Table 3).

Discussion
We compared the diagnostic performances and conclusive rates of FNA and CNB performed at our institution for patients who had thyroid nodules equal to or larger than 1 cm. For diagnostic performance, FNA had a significantly higher specificity, PPV, and accuracy and a lower sensitivity than CNB for selecting surgical candidates before matching. In the other conditions, however, no significant differences were found in the diagnostic performances of the two biopsy groups. Meanwhile, the conclusive rates were significantly higher in CNB than in FNA when thyroid nodules were classified as ACR TI-RADS or K-TIRADS category 4 and were larger than 2 cm.
Previous literatures demonstrated significantly lower non-diagnostic result rates or higher accuracies for malignancy in CNB groups compared to FNA groups when the nodules showed previous non-diagnostic or previous indeterminate FNA results 8,10,13,20 . In addition, several studies have demonstrated that CNB was more effective even for initially detected thyroid nodules on US 18,21,22 . In this study, consistent with those studies, there was no non-diagnostic case of CNB, while non-diagnostic rates for FNA were 13.9% and 3.5% before and after matching, respectively. In addition, the diagnostic performances for malignancy were comparable between the two biopsy methods.
Conversely, other studies have reported that CNB has comparable or lower accuracy and sensitivity compared to FNA 17,23 . Among these studies, a recent study by Kim et al. comparing the diagnostic performances of 3,048 FNA and 144 CNB cases concluded that CNB may not be helpful in diagnosing papillary carcinomas and neoplasms 17 . However, despite their large sample size, only 16.3% of all patients underwent the procedures at the authors' institution, while the remaining 2,672 patients (87.4% in total; 83.3% of FNA and 93.1% of CNB) were diagnosed by reviewing submitted slides after procedures at other hospitals. As is well known, the technique of FNA and CNB is one of the most important factors affecting the diagnostic yield of biopsy. In this study, we included only FNA or CNB cases performed at our institution in order to consistently compare the quality of the procedure. In addition, they did not include any nondiagnostic lesions by either FNA or CNB after matching for their comparative analysis of FNA and CNB groups. These two might be the significant limitations of their research.   24 . In particular, with CNB, the main concerns most commonly include bleeding and hematoma 24,25 . The CNB complication rates have been reported between 0.2% and 1.0% [25][26][27][28] , and patient discomfort and tolerability levels were not significantly different between FNA and CNB 28,29 . Fortunately, in our results, the superiority of CNB to FNA was apparent in nodules more than 2 cm with US findings categorized as intermediate suspicion. This indication reduces the risk of overuse of CNB in small thyroid nodules.
This study has several limitations. First, the retrospective data collection may have resulted in case selection bias. Second, there was a timing difference between the two biopsy groups, because CNB started to be actively performed in the middle of 2013 and the time when FNA had the least influence on CNB was 2008. Third, the proportion of 'previous Bethesda or CNB diagnostic category III' in CNB group was decreased after matching, which was still significantly higher than that in FNA group. As widely known, Bethesda or CNB diagnostic category III at initial biopsy is a risk factor leading repeated category III biopsy results. However, in this study, these did not affect the interpretation of our results. Forth, we did not compare the frequency and degree of complications with FNA and CNB groups. The frequency and degree of complications in FNA cases have been reported to be lower or similar compared with CNB cases. In our study, the complication rate of CNB was 0%. Last, in this study, we defined benign thyroid nodules as nodules with no significant changes in one-year follow-up and with benign FNA or CNB results. However, one year may not be enough to detect significant changes in certain nodules to distinguish benign nodules from malignant nodules.
In conclusion, our study revealed that CNB showed comparable diagnostic performance to FNA. Meanwhile, the conclusive rates were significantly higher in CNB than in FNA when thyroid nodules were classified as ACR TI-RADS or K-TIRADS category 4 and measured larger than 2 cm. Therefore, we can predict the effective indications of CNB for thyroid nodules that are larger than 2 cm and show US findings corresponding to ACR TI-RADS or K-TIRADS category 4.

Materials and Methods
Patient population. The Institutional Review Board at Samsung Medical Center, Seoul, Korea, approved this retrospective study, and patient approval and informed consent were not required for the review of US images and medical data. However, written informed consent was acquired from patients before undergoing the US-guided FNA and US-guided CNB procedures. In addition, all methods were conducted according to relevant guidelines and regulations.
From January 2008 to May 2008, 1,096 consecutive patients underwent FNA for 1,109 thyroid nodules (≥1 cm) at our institution (Fig. 1). First, for the evaluation of conclusive rates, we excluded 49 palpable cyst cases with typical colloid cystic appearance on US to avoid false negative results. Second, for the assessment of diagnostic performance, we excluded 439 nodules additionally as follows: (1) 320 nodules with no follow-up after being diagnosed as benign by FNA in order to avoid false negative cases, (2) 22 nodules showing significantly increased size after benign FNA results without subsequent confirmation to avoid false negative cases, (3) 26 nodules with suspicious for malignancy (n = 1) or malignant (n = 25) FNA results without subsequent confirmation, and (4)   www.nature.com/scientificreports www.nature.com/scientificreports/ nodules in 615 patients (489 women and 126 men; mean age ± standard deviation 49.82 ± 11.67, range 14-87) for diagnostic performance.
A total of 452 consecutive patients with 462 thyroid nodules (≥1 cm) underwent CNB at our institution between January 2014 and December 2015 (Fig. 1). First, we included all 452 patients for the evaluation of conclusive rates. Second, for the assessment of diagnostic performance, we excluded 132 patients who had an imaging or clinical follow-up duration less than one year or were lost to follow-up after CNB (n = 75), who had cases with indeterminate (n = 4), or FN/SFN (n = 46) results on CNB without subsequent confirmation, and who had cases showing significantly increased size 27 after benign CNB results without subsequent confirmation (n = 7). Finally, the study enrolled 462 nodules in 452 patients (333 women and 119 men; mean age ± standard deviation 49.98 ± 13.20, range 14-83) for the conclusive rates, and 330 nodules in 322 patients (236 women and 86 men; mean age ± standard deviation 49.39 ± 13.23, range 14-81) for the diagnostic performance.
For the assessment of diagnostic performance, final diagnoses of malignant nodules (n = 252) were decided based on the pathological results by surgery (n = 244) or CNB (n = 8). Final diagnoses of benign nodules (n = 699) were decided based on the pathological results by surgery (n = 196, 28.0%), benign FNA or CNB results repeated at least twice (n = 317, 45.4%) and a concordant benign result of FNA or CNB and a decreased or stable nodule size at US or clinical follow-up of at least one year (n = 186, 26.6%).

US examination and US-Guided biopsy procedures.
All US examinations were operated with a 5-to 12-MHz linear-array transducer and a real-time US system (HDI 5000, Philips Medical Systems, Bothell, WA, USA; iU22, Philips Medical Systems, Bothell, WA, USA). The US examinations and US-guided biopsy procedures were performed by four faculty radiologists who specialize in thyroid imaging (more than seven years of experience), fellows, or residents. Faculty radiologists supervised all procedures of the fellows or residents.
US-guided FNA was operated with a 23-gauge needle attached to a 2 mL disposable plastic syringe routinely and a 21-gauge needle attached to a 2 mL or 5 mL disposable plastic syringe selectively. US-CNB was performed with a disposable 18-gauge, double-action spring-activated needle with an 11 mm excursion (TSK Ace-cut; Create Medic, Yokohama, Japan) after local anesthesia. After biopsy, every patient was observed with local compression of the biopsy site for 10 to 20 minutes.
In our institution, we have used a CNB technique to contain the nodule, nodular margin, and surrounding parenchyma in at least one core specimen since June 2013 with a minimum of two cores 14,30 . Among the CNB patients, there was no case of technical failure or complication in obtaining biopsy specimens during the study period.  www.nature.com/scientificreports www.nature.com/scientificreports/ US features were retrospectively classified according to the ACR TI-RADS and K-TIRADS 3,7 . All thyroid nodules were recorded for composition, echogenicity, shape, orientation, margin, and echogenic dot (calcification) according to the ACR TI-RADS or K-TIRADS. In the ACR TI-RADS, US findings are given 0-3 points corresponding to their association with malignancy, and points were added to assess the final risk stratification level. In the K-TIRADS, malignancy risk stratification was finally assessed into five categories based on the US characteristics of the thyroid nodules. The nodule size was determined as the maximum diameter measured on the static images, regardless of the acquisition plane.
Cytology and histology analysis. During the study period, one of seven experienced pathologists reassessed the results of FNA and CNB according to the Bethesda System for Reporting Thyroid Cytopathology and the research of the Korean Endocrine Pathology Thyroid Core Needle Biopsy Study Group, respectively 15,16 . Statistical analysis. Significant differences were found in the characteristics and numbers between CNB and FNA nodules. Therefore, propensity score matching was conducted to match FNA patients to CNB patients on sex and age, previous inconclusive results (i.e., I [non-diagnostic] or III [AUS/FLUS or indeterminate]) of FNA or CNB, present FNA or CNB results, nodule size, nodule composition on US, ACR TI-RADS risk category, and K-TIRADS category 31 . Clinicopathological and US features between the two biopsy groups were compared with the independent two-sample t test (before matching) and the paired t test (after matching) for continuous variables, and the χ 2 test or Fisher's exact test (before matching) and McNemar's test (after matching) for categorical variables.
The sensitivities, specificities, PPVs, negative predictive values (NPV), and accuracies of FNA and CNB were calculated for the diagnosis of thyroid malignancy (i.e. Bethesda or CNB diagnostic category V and VI vs. II) and for the determination of surgical indication (i.e. Bethesda or CNB diagnostic category IV, V, and VI vs. I, II, and III) 32 , and we compared them using the χ 2 test or Fisher's exact test in the non-matched population and generalized estimating equations in the matched population.
Conclusive rates for FNA and CNB (i.e. Bethesda or CNB diagnostic category II, IV, V, and VI vs. I and III) were also investigated and compared according to the final US category and nodule size using the χ 2 test or Fisher's exact test in the non-matched population and McNemar's test for categorical variables in the matched population 32 .
Statistical significance was accepted with a two-tailed P value < 0.05. All statistical analyses were conducted using SAS version 9.4 (SAS Institute, Cary, NC, USA) and R 3.2.2 (Vienna, Austria; http://www.R-project.org/).