Ultrasound Score to Select Subcentimeter-sized Thyroid Nodules Requiring Ultrasound-guided Fine Needle Aspiration Biopsy in Eastern China

As the widespread use of routine ultrasound survey during physical examination, the number of patients with thyroid cancer has been rising steeply during recent years in eastern China (Xiang et al., 2010). Up to now, ultrasound-guided FNAB is the least invasive preoperative modality for discriminating benign from malignant thyroid nodules (Nixon et al., 2013). However, only a small proportion of thyroid nodules turn out to be malignant (Hegedüs, 2004). The practical problem arises as ultrasound-guided FNAB being a costly diagnostic item with a low yield in identifying the tiny proportion of nodules that actually represent malignant disease (Iannuccilli et al., 2004). Several US features have been reported as predictors of malignancy (Iannuccilli et al., 2004; Kang et al., 2004; Kim et al., 2008; Lee et al., 2011;Vinayak et al., 2012; Kim et al., 2013a; Kim et al., 2013b). But currently there is no acknowledged guideline of selecting which thyroid nodules for biopsy, especially for the subcentimeter-sized ones (Iannuccilli et al., 2004; Moon et al., 2012; Grani et al., 2013; Kim et al., 2013a; Nixon et al., 2013). Our aim was to obtain an ultrasound (US) score for selecting thyroid nodules ≤ 10 mm in size requiring ultrasound-guided FNAB in eastern China.


Introduction
As the widespread use of routine ultrasound survey during physical examination, the number of patients with thyroid cancer has been rising steeply during recent years in eastern China (Xiang et al., 2010). Up to now, ultrasound-guided FNAB is the least invasive preoperative modality for discriminating benign from malignant thyroid nodules (Nixon et al., 2013). However, only a small proportion of thyroid nodules turn out to be malignant (Hegedüs, 2004). The practical problem arises as ultrasound-guided FNAB being a costly diagnostic item with a low yield in identifying the tiny proportion of nodules that actually represent malignant disease (Iannuccilli et al., 2004). Several US features have been reported as predictors of malignancy (Iannuccilli et al., 2004;Kang et al., 2004;Kim et al., 2008;Lee et al., 2011;Vinayak et al., 2012;Kim et al., 2013a;Kim et al., 2013b). But currently there is no acknowledged guideline of selecting which thyroid nodules for biopsy, especially for the subcentimeter-sized ones (Iannuccilli et al., 2004;Moon et al., 2012;Grani et al., 2013;Kim et al., 2013a;Nixon et al., 2013). Our aim was to obtain an ultrasound (US) score for selecting thyroid nodules ≤ 10 mm in size requiring ultrasound-guided FNAB in eastern China.

Materials and Methods
Between January 2006 and March 2012, 622 patients underwent ultrasound-guided FNAB for a total of 665 thyroid nodules ≤ 10 mm in size at our hospital. A maximum of three suspicious nodules were examined for a single patient when multiple nodules were found.
All nodules were evaluated by US using an Acuson Sequoia and 128XP sonographic scanners (Siemens Medical Solutions, Mountain View, CA) equipped with commercially available 8-to 13-MHz linear probes. Ultrasound images were reviewed by the same experienced ultrasound radiologist who was blinded to clinical and pathologic data. The US variables recorded were nodule size, the presence of multiple nodules, nodule location, shape, composition, echogenicity, echogenic density, margin, calcification, posterior shadowing, capsule, vascularization pattern and the ratio between antero-posterior and transversal diameters (AP/TR).
Specimens obtained from ultrasound-guided FNAB were smeared on a slide, immediately fixed in 95% ethanol and stained by the HE method. Two senior pathologists interpreted cytology in blind. Final cytological diagnoses were divided into five groups: 1. non-diagnostic; 2. benign; 3. follicular lesion; 4. suspicious of malignancy; Logistic regression analysis was fitted to data to assess independent associations of malignancy with all variables above. The differences were considered statistically significant with P-value < 0.05. Independent predictors according to multiple logistic regression analysis were given different points for constructing a scoring system. Moreover, the receiver-operating characteristic curves (ROC curves) was used to identify the optimal point with high sensitivity and low falsenegative rate (1-specificity).
There were 270 nodules removed surgically, including 143 benign nodules and 127 malignant lesions according to the pathological results which consisted with the cytological findings in 232 nodules (85.9%) and disagreed in 38 (14.1%).
Comparison was made to evaluate the clinical and US differences between benign and malignant nodules in Table 1. Irregular shape, hypoechogenicity, solid composition, no well-defined margin, absence/incomplete capsule, calcifications and the AP/TR ratio ≥ 1 were strongly significant with malignant (P < 0.001). Male (P=0.004), single nodule (P=0.006), heterogeneous echo (P=0.01), absence/degenerate posterior shadowing (P=0.01) and vascularization pattern (P=0.044) were also significant variables. Age, nodule size and location were not statistically significant.
Logistic regression analysis showed a significant predictive role for Irregular shape P < 0.001),   US score > 2 was the optimal point for distinguishing between benign and malignant lesions. The sensitivity and specificity of US score > 2 reached 80.3% and 72.7%, respectively, while the positive and negative predictive values were 72.3% and 80.6% hypoechogenicity (P =0.014), absence/incomplete capsule (P < 0.001), the presence of calcifications (P < 0.001) and AP/TR ratio ≥ 1 (P < 0.001) as in Table 2.
We then graded the nodules by giving an index score to variables that were significantly associated with malignant outcome. Table 3 shows the index score for each US feature. Mean US score was 1.7±1.0 in benign group and 3.4 ±1.1 in malignant group, which was statistically significant (P < 0.001). By using ROC curves, we found US score > 2 was the optimal point for distinguishing between benign and malignant lesions (Figure 1). The sensitivity and specificity of US score > 2 reached 80.3% and 72.7%, respectively, while the positive and negative predictive values were 72.3% and 80.6%.

Discussion
Due to the prevalence of thyroid nodules, it is clinically important to predict the risk of malignancy of a given nodule (Yang et al., 2011). Ultrasound-guided FNAB is the least invasive preoperative method for the differential diagnosis of thyroid nodules (Nixon et al., 2013). However, it is blatantly inappropriate to conduct ultrasound-guided FNAB for all thyroid nodules since the ratio of malignancy is relatively low (Hegedüs., 2004;Iannuccilli et al., 2004). US is commonly used as an effective noninvasive technique for the diagnosis of benign and malignant thyroid nodules. Many scholars have focused on the valuable role of US features as predictors of malignancy (Iannuccilli et al., 2004;Kang et al., 2004;Kim et al., 2008;Lee et al., 2011;Vinayak et al., 2012;Kim et al., 2013a;Kim et al., 2013b). Currently, there is no internationally acknowledged guideline of selecting which thyroid nodules for biopsy, especially for the subcentimeter-sized ones (Iannuccilli et al., 2004;Grani et al., 2013;Kim et al., 2013a;Nixon et al., 2013). Many institutions now apply ultrasound-guided FNAB just on nodules larger than 1 cm in diameter. This occurs as a result of the popular recognition that the mortality rate of thyroid microcarcinoma (defined as ≤ 1 cm in maximum diameter) is fairly low (Giordano et al., 2010). However, there is still a vital significance for differential diagnosis of benign and malignant subcentimeter-sized thyroid nodules, as cases with regional lymph nodes and distant metastases have been well documented in the literature.
In this study, we attempted to determine whether US features could be utilized to select subcentimeter-sized thyroid nodules for ultrasound-guided FNAB. For nodules < 1 cm in our hospital, ultrasound-guided FNAB was performed only if there were suspicious US features for malignancy. This led to the high proportion of malignancy and suspicious of malignancy (20.3%) compared with other reports (Cavaliere et al., 2009). It has been reported that thyroid nodules with a diameter <10 mm were more frequently non-diagnostic (Grani et al., 2013). In view of this, the number of non-diagnostic nodules in our study (20.9%) was receivable.
Irregular shape, hypoechogenicity, solid composition, no well-defined margin, absence/incomplete capsule, calcifications, the AP/TR ratio ≥ 1, male, single nodule, heterogeneous echo, absence/degenerate posterior shadowing and vascularization pattern were significant characteristics of malignancy on univariate analysis in our study. The result is generally similar to several earlier studies (Kang et al., 2004;Kim et al., 2008;Cavaliere et al., 2009;Vinayak and Sande., 2012;Wang et al., 2012;Kim et al., 2013a;Kim et al., 2013c). However, there was little standardization among authors of the vascularization pattern, which was statistically significant in present result. Multivariate analysis showed that irregular shape, hypoechogenicity, absence/incomplete capsule, the presence of calcifications and AP/TR ratio ≥ 1 were independent predictors for malignant pathology. In some articles solitary nodule showed a strong predictive value for positive cytology (Cavaliere et al., 2009;Petrone et al., 2012) which was inconsistent in our paper. Similar situation was observed on vascularization pattern (Vinayak et al., 2012;Lingam et al., 2013). This could be explained by the small sample size and possible collinearity between variables.
As the poor discriminal efficiency of using US independently, we made a US scoring system on the basis of multiple logistic regression analysis. The area under the ROC curves was of high value (0.85, 95%CI: 0.81-0.90) in predicting the final pathological findings. We identified the US index points > 2 as the best cut-off for distinguishing between benign and malignant lesions with a sensitivity of 80.3%, specificity of 72.7%, positive predictive value of 72.3% and negative predictive value of 80.6%. According to our result, we suggest FNAB for nodules when the US score is higher than 2. For nodules with US score lower than 2, observation of disease and regular follow-up were recommended.
Ultrasound-guided FNAB is the most accurate method available to distinguish benign from malignant thyroid nodules (Cooper et al., 2009). And there is a clear need to select nodules for ultrasound-guided FNAB to maximize benefits and minimize cost. US carries important implication in the treatment of thyroid diseases as an important means to help make a judgment on FNAB guidance. Up to now, the guideline of selecting which thyroid nodules for biopsy, especially for the subcentimeter-sized ones, is still controversial. Our study was designed to build and assess an ultrasound (US) score system for selecting thyroid nodules ≤ 10 mm in size requiring ultrasound-guided FNAB. Our US scoring system may allow surgeons to diagnose thyroid diseases more accurately, decrease patients' financial burden and enhance the utilization ratio of health resources. However, it should be noted that any good technique could be more effective with the help of others. Therefore, independently of the US score, clinical judgment is always recommended for patients run the risk of malignancy.