The combination of ATA classification and FNA results can improve the diagnostic efficiency of malignant thyroid nodules.

PURPOSE: To determine the diagnostic efficiency of the ATA classification and ultrasound-guided fine-needle aspiration (FNA) results in identifying the risk factors of malignancy, we analyzed the thyroid nodules of patients who underwent thyroidectomy and compared preoperative ATA classifications with FNA results.
METHODS: We retrospectively analyzed 274 nodules of 196 patients who underwent ultrasonography, FNA and thyroidectomy. Histopathological findings of thyroid nodules were considered as the Au standard in the analysis of the diagnostic efficiency of the ATA classification and FNA results. Univariate analysis and binary multivariate logistic regression analysis were applied to identify the ultrasound features associated with malignancy.
RESULTS: The overall malignancy rate of 274 nodules was 41.6%. The areas under the ROC curves (AUCs) for the ATA classification and FNA results were 0.88 and 0.878, respectively (P < 0.001). The sensitivity and specificity of the ATA classification were 86 and 86.9%, whereas those of FNA results were 68.5 and 91.4%, respectively. The specificity (98.7%) and sensitivity (94.3%) increased after the combined use of the ATA classification and FNA results. Taller-than-wide shape, microcalcifications, hypoechogenicity and irregular margins were independent risk factors for malignancy. Microcalcifications had the highest OR (7.58), and taller-than-wide shape had the highest specificity in BSRTC I, II, III and IV cytology.
CONCLUSION: The diagnostic efficiency of the ATA classification and FNA results in identifying malignant nodules was high, and the use of both criteria improved the diagnostic accuracy. Taller-than-wide shape, microcalcifications, hypoechogenicity and irregular margins were independent risk factors for malignancy.

Introduction

Nodular thyroid disease is common in endocrine practice. In iodine-sufficient areas, the thyroid nodule detection rate, as determined by palpation, is 5% in women and 1% in men (1, 2), whereas the detection rate, as determined by high-resolution ultrasonography, is 19–68%. Thyroid nodules are more common in women and the elderly (3, 4), with 5–15% of thyroid nodules characterized as malignant (5). Therefore, the accurate diagnosis of nodules and the prompt treatment of malignant nodules can improve patient outcomes.

Currently, fine-needle aspiration (FNA) is the most accurate diagnostic method in the identification of malignant thyroid nodules (6). The Bethesda System for Reporting Thyroid Cytopathology (BSRTC) has been widely used since 2009, and it classifies thyroid nodules into six categories based on FNA results, provides recommendations for managing malignant tumors and identifies the risk factors for malignancy (7). Although the sensitivity of FNA in the diagnosis of malignant nodules is 65–98%, the specificity is 72–100% and the accuracy is >95% (7, 8), approximately 20% of FNA results are inconclusive due to poorly preserved specimens (9) and 20–25% are indeterminate due to the presence of follicular lesions or follicular neoplasms (8, 10, 11), although most lesions with indeterminate cytology are surgically confirmed to be benign (7, 12). Thus, an improved awareness of risk factors for thyroid nodules, such as age, gender, and especially the ultrasound features, can reduce the frequency of unnecessary surgeries for benign lesions.

High-resolution ultrasonography is a simple, reliable and commonly used method that can describe a thyroid nodule’s morphology, characteristics and lymphadenopathy, which can facilitate diagnosis. In addition, ultrasonography can guide FNA, which can reduce the generation of non-diagnostic specimens. Ultrasound features, such as solidity, hypoechogenicity, microcalcifications, taller-than-wide shape and irregular margins, are useful in the diagnosis of malignant nodules (13, 14, 15, 16, 17, 18). Furthermore, other studies have reported that a simple cystic and spongiform appearance have a higher negative predictive value (NPV) for malignancy (13, 19, 20). However, there is overlap in the ultrasound features of benign and malignant nodules (21, 22), and the sensitivity and specificity of malignant nodules are also different when considering these features (23, 24, 25). The ‘2015 American Thyroid Association (ATA) Management Guidelines for Adult Patients with Thyroid Nodules and Differentiated Thyroid Cancer’ (referred to as the ATA classification) (26) more comprehensively defines the ultrasound features and risk factors of thyroid nodules, and characteristics of solidity, hypoechogenicity, irregular margins (i.e. infiltrative, microlobulated), taller-than-wide shape, microcalcifications, rim calcification interruption, and evidence of extrathyroidal extensions are signs of malignancy. Therefore, the ATA classification is useful in the diagnosis and treatment of nodules when FNA results are non-diagnostic or indeterminate.

Presently, the ATA classification is not commonly used in the stratification of risk factors for thyroid nodule malignancy in China. In this study, we analyzed 274 nodules of 196 patients with ultrasound-guided FNA results, compared ATA classification with histopathology results, and discussed the diagnostic efficiency of the ATA classification and FNA results for thyroid nodules.

Materials and methods

Patients

This study retrospectively analyzed 274 nodules (160 benign and 114 malignant) of 196 patients (28 male and 168 female) pathologically confirmed from June 2016 to June 2018 at the First Affiliated Hospital of China Medical University. The average age was 47.24 ± 12.15 years, and the average nodule size was 19.18 ± 11.82 mm. All patients underwent ultrasonography, FNA and thyroidectomy and received complete ultrasonography, FNA cytology and pathology results. Patients with a history of thyroid surgery, thyroid metastasis, and those with surgically removed nodules that were not one-to-one matched with the ultrasound findings were excluded from the study.

The study was approved by Ethics Institutional Review Board of China Medical University. Data collection followed the principles outlined in the Declaration of Helsinki. All patients provided their written informed consent to share their own anonymous information to participate in our study.

Ultrasound findings and evaluation

Toshiba color Doppler ultrasound instruments (models ssa-680 and aplio770, Japan) were used with a 7- to 10-MHz linear array transducer. The ultrasonography reported several parameters as follows: number, size, location, morphology, margins, the presence or absence of taller-than-wide shape, echogenicity, calcifications and vascularity. Clinical results were stored in workstations connected to the ultrasound instruments. According to the ATA classification (26), the ultrasound malignant risk classifications were as follows: high suspicion, intermediate suspicion, low suspicion, very low suspicion and benign.

Ultrasound-guided fine-needle aspiration and cytopathological evaluation

The ultrasound-guided FNA was performed on nodules with at least one suspicious ultrasound characteristic or on the largest lesion when none of the thyroid nodules had any suspicious ultrasound features. FNA was performed with a 25-gauge needle attached to a 2 mL syringe. Each nodule was aspirated at least twice, and the aspirated specimens were expelled onto glass slides, which were then placed into 95% alcohol for Papanicolaou staining. FNA cytopathological results were classified according to BSRTC (7) as follows: (i) non-diagnostic or unsatisfactory (DN/UNS); (ii) benign; (iii) atypia (or follicular lesion) of undetermined significance (AUS/FLUS); (iv) follicular neoplasm or suspicious for a follicular neoplasm (FN/SFN); (v) suspicious for malignancy (SPUS); and (vi) malignant.

Statistical analyses

Analyses were performed with SPSS version 25.0 software. Differences in nodule characteristics between benign and malignant nodules were analyzed using t-test or χ2 test. The diagnostic efficiency was expressed as the sensitivity, specificity and accuracy. Receiver operating characteristic (ROC) curves were constructed, and the areas under the curves (AUCs) were calculated. The predetermined level of significance was set at P < 0.05. The binary multivariate logistic regression analysis was performed to identify the independent factors associated with malignancy, and their diagnostic value in the prediction of thyroid cancer was then tested.

Results

ATA classification, FNA results and histopathological findings of thyroid nodules

Among 274 nodules, 43.4% (119) were ATA classified as high suspicion, 17.9% (49) as intermediate suspicion, 27.7% (76) as low suspicion, and 10.9% (30) as very low suspicion. The malignancy rates in these categories were as follows: 82.4 (98), 16.3 (8), 9.2 (7) and 3.3% (1), respectively, and the malignancy rate was significantly different among the groups (χ2 = 137.31, P < 0.001). FNA cytopathological classifications were as follows: 6.6% (18) were BSRTC I, 36.1% (99) were BSRTC II, 23% (63) were BSRTC III, 2.9% (8) were BSRTC IV, 20.4% (56) were BSRTC V and 10.9% (30) were BSRTC VI. The malignancy rates in these classifications were as follows: 50 (9), 6.1 (6), 39.7 (25), 12.5 (1), 78.6 (44) and 96.7% (29), respectively, and the malignancy rate was significantly different among the groups (χ2 = 123.8, P < 0.001). For BSRTC I nodules, all very low and low suspicion nodules were benign, and most (90%) intermediate and high suspicion nodules were malignant. For BSRTC III and IV nodules, 78.6% high suspicion nodules were malignant, and most (90.7%) other nodules were benign (Table 1).

Table 1

ATA classification, FNA results and histopathological findings of thyroid nodules.

BSRTC	ATA classification										Total (%)	Malignancy rate (%)a	Malignancy rate (%)b
	Very low		Low		Intermediate		High		All
	Ben-	Mal-	Ben-	Mal-	Ben-	Mal-	Ben-	Mal-	Ben-	Mal-
I (DN/UNS)	2	0	6	0	0	2	1	7	9	9	18 (6.6)	50.0	1–4
II (Benign)	20	0	37	0	24	2	12	4	93	6	99 (36.1)	6.1	0–3
III (AUS/FLUS)	5	0	16	1	11	2	6	22	38	25	63 (23)	39.7	5–15
IV (FN/SFN)	1	0	5	0	1	1	0	0	7	1	8 (2.9)	12.5	15–30
V (SFM)	1	1	5	4	4	1	2	38	12	44	56 (20.4)	78.6	60–75
VI (Malignant)	0	0	0	2	1	0	0	27	1	29	30 (10.9)	96.7	97–99
All	29	1	69	7	41	8	21	98	160	114	274	41.6
Total (%)	30 (10.9)		76 (27.7)		49 (17.9)		119 (43.4)		274
Malignancy rate (%)c	3.3		9.2		16.3		82.4		41.6
Malignancy rate (%)d	<3		5–10		10–20		70–90

aThe malignancy rate of the FNA results in this study, bThe malignancy rate of BSRTC (7), cThe malignancy rate of the ATA classification in this study and dThe malignancy rate of the ATA classification (26).

Ben, benign; Mal, malignant.

Diagnostic efficiency of the ATA classification and FNA results

When a nodule classified according to the ATA classification was considered highly suspicious, the AUC was 0.880 (95% CI 0.837–0.923, P < 0.001) (Fig. 1A); the Youden index (0.729) was highest; the sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV) and accuracy (AC) were 86, 86.9, 82.4, 89.7 and 86.5%, respectively (Table 2). After BSRTC I specimens were excluded, the best cutoff point for diagnosis was BSRTC V; the AUC was 0.878 (95% CI 0.834–0.921, P < 0.001) (Fig. 1B); the Youden index (0.609) was highest; the sensitivity, specificity, PPV, NPV and AC were 69.5, 91.4, 84.9, 81.2 and 82.4%, respectively (Table 2).

Figure 1

AUCs of the ATA classification (A) and FNA results (B).

Table 2

Diagnostic efficiency of the ATA classification, FNA results and both criteria in combination.

Method	SEN	SPE	PPV	NPV	AC	AUC	95% CI	P value of the AUC
ATA	86	86.9	82.4	89.7	86.5	0.880	0.837–0.923	< 0.001
FNA	69.5	91.4	84.9	81.2	82.4	0.878	0.834–0.921	< 0.001
Combination 1	94.3	77.5	74.4	95.1	84.4	0.859	0.811–0.907	< 0.001
Combination 2	61.9	98.7	97	78.8	83.6	0.803	0.742–0.864	< 0.001

Combination 1: Conditions under which a nodule was diagnosed as malignant according to one criterion (the ATA classification or FNA results). Combination 2: Conditions under which a nodule was diagnosed as malignant according to both criteria (the ATA classification and FNA results).

AC, accuracy; NPV, negative predictive value; PPV, positive predictive value; SEN, sensitivity; SPE, specificity.

When a nodule classified according to one criterion (either the ATA classification or FNA results) was considered malignant, and then both criteria were used to re-classify the nodule, the sensitivity, specificity, PPV, NPV and AC were 94.3, 77.5, 74.4, 95. and 84.4%, respectively. The AUC was 0.859 (95% CI 0.811–0.907). When a nodule classified according to both criteria (the ATA classification and FNA results) was considered malignant, the sensitivity, specificity, PPV, NPV and AC were 61.9, 98.7, 97, 78.8 and 83.6%, respectively. The AUC was 0.803 (95% CI 0.742–0.864) (Table 2).

Analysis of nodule characteristics

The clinical characteristics of patients and the ultrasound features of nodules were examined by univariate analysis (Table 3). Younger patients with single and smaller nodules had a higher malignancy rate (P < 0.05), although no significant differences in gender and nodule location were noted (P = 0.437 and 0.895, respectively). With regard to the ultrasound features of nodules, those that were solid and hypoechoic, with irregular margins, taller-than-wide shape, microcalcifications, and posterior echo attenuation were malignant, whereas those with regular halo were benign, and the differences were significant (P < 0.001). In binary multivariate logistic regression analysis, hypoechogenicity (OR = 2.74, 95% CI 1.421–5.29), taller-than-wide shape (OR = 5.81, 95% CI 1.79–18.86), microcalcifications (OR = 7.58, 95% CI 3.89–14.77) and irregular margins (OR = 3.87, 95% CI 2.0–7.46) were independent risk factors for thyroid nodule malignancy (Table 4).

Table 3

Clinical characteristics of patients and ultrasound features of thyroid nodules.

Characteristics		Benign (160)	Malignancy (114)	χ2/Z value	P value
Sex	Male	20 (52.6%)	18 (47.4%)	0.603	0.437
	Female	140 (59.3%)	96 (40.7%)
Age		53 (46.25, 58.75)	40 (32, 51)	2.524	<0.001
Size (mm)		18.55 (11.5, 31.68)	13.1 (9.08, 19)	33.107	<0.001
Location	Isthmus	3 (50%)	3 (50%)	0.222	0.895
	Left	77 (59.2%)	53 (40.8%)
	Right	80 (58%)	58 (42%)
Number	Sigle	11 (31.4%)	24 (68.6%)	12.009	0.001
		149 (62.3%)	90 (37.7%)
Structure	Solid	62 (43.7%)	80 (56.3%)	27.407	<0.001
	Mainly solid	85 (72.6%)	32 (27.4%)
	Mainly Cystic	13 (86.7%)	2 (13.3%)
Hypoechoic	Present	59 (41.3 %)	84 (58.7%)	36.148	<0.001
	Absent	101 (77.1%)	30 (22.9%)
Margins	Regular	129 (78.2%)	36 (21.8%)	66.846	<0.001
	Irregular	31 (28.4%)	78 (71.6%)
Taller-than-wide shape	Present	5 (22.73%)	17 (77.3%)	12.525	0.001
	Absent	155 (61.5%)	97 (38.5%)
Calcification	Micro	23 (23.2%)	76 (76.8%)	78.88	<0.001
	Macro	36 (60%)	24 (40%)	0.082	0.775
	Comet tail-like	35 (63.6%)	20 (36.4%)	0.778	0.378
Halo	Regular	35 (87.5%)	5 (12.5%)	16.332	<0.001
	Irregular	2 (33.3%)	4 (66.7%)	1.586	0.238
Echotexture	Homogeneous	31 (66%)	16 (34%)	1.336	0.248
	Heterogeneous	129 (56.8%)	98 (43.2%)
Posterior echo attenuation	Present	13 (29.5%)	31 (70.5%)	17.956	<0.001
Attenuation	Absent	147 (63.9%)	83 (36.1%)

Table 4

Binary multivariate logistic regression analysis of ultrasound features for the detection of malignant thyroid nodules.

US characteristics	β	OR (95% CI)	P value
Hypoechoic	1.008	2.74 (1.421–5.29)	0.003
Taller-than-wide shape	1.760	5.81 (1.79–18.86)	0.003
Microcalcification	2.025	7.58 (3.89–14.77)	<0.001
Irregular margins	1.352	3.87 (2.0–7.46)	<0.001

β, regression coefficient; OR, odds ratio.

We evaluated the diagnostic efficiency of the malignant ultrasound features for nodules classified according to FNA results as BSRTC I, II, III and IV (Table 5). Taller-than-wide shape showed the highest specificity (95.6–100%) for all nodules, and BSRTC II showed the highest accuracy (91.9%), but its sensitivity was poor. The sensitivity of each ultrasound feature was low (<70%). In the presence of microcalcifications or hypoechogenicity, the sensitivity of all nodules was the highest (83.3–92.3%).

Table 5

Diagnostic efficacy of ultrasound features for the detection of malignant thyroid nodules.

US characteristics	BSRTC I			BSRTC II			BSRTC III and IV
	SEN	SEP	AC	SEN	SEP	AC	SEN	SEP	AC
Hypoechoic	77.8	88.9	83.3	66.7	64.5	64.6	69.2	62.2	64.8
Taller-than-wide shape	22.2	100	61.1	16.7	96.8	91.9	11.5	95.6	64.8
Microcalcification	55.6	88.9	72.2	50	85	82.8	61.5	86.7	77.5
Irregular margins	66.7	88.9	77.8	50	77.4	75.8	65.4	84.4	77.5
Either microcalcification or hypoechoic	88.9	77.8	83.3	83.3	55.9	57.9	92.3	57.8	70.4

Discussion

A variety of methods are available for the preoperative examination of thyroid nodules. The ultrasonography is a commonly used imaging technology that utilizes a high-resolution probe to identify thyroid nodule features (27). FNA is one of the most accurate methods for differentiating between the benign and malignant thyroid nodules (6); however, the FNA is often performed by experienced personnel at medical centers in China. Therefore, the diagnosis of thyroid nodules as benign or malignant based on ultrasound features is important for the subsequent treatment of the disease.

A total of 274 nodules from 196 patients were included in this study, and the overall malignancy rate was 41.6%. Potential selection bias may have contributed to the higher malignancy rate. According to the different ultrasound features, the ATA classification categorized thyroid nodules into five categories, namely high suspicion, intermediate suspicion, low suspicion, very low suspicion and benign, and the malignancy rates were 70–90, 10–20, 5–10, <3 and <1%, respectively (26). By contrast, the malignancy rates were 82.4, 16.3, 9.2 and 3.3%, respectively, when we re-classified nodules according to the ATA classification. The malignancy rate of the FNA results was higher than that of the BSRTC classification, except for BSRTC IV and VI nodules (7). Among BSRTC I nodules, all very low and low suspicion were benign, and 90% of intermediate and high suspicion nodules were malignant. Furthermore, 78.6% of high suspicion BSRTC III and IV nodules were malignant, and most (90.7%) other nodules were benign. These findings were consistent with those reported by Tang et al. (28, 29). Collectively, these results indicate that the ATA classification, when used in combination with FNA results, can improve the detection accuracy of benign and malignant nodules (30, 31). In our study, 18 (6.6%) nodules were classified as BSRTC I according to the FNA results, with a high malignancy rate of 50%, which was caused by selection bias. Among these nodules, 61.1% (11/18) nodules were <10 mm in size, and 44.4% (8/18) nodules were classified as high suspicion nodules according to the ATA classification. Chinese guidelines recommend nodules <10 mm but with malignant ultrasound characteristics should undergo FNA as well, which increased the proportion of non-diagnostic results of FNAB and their malignancy rate in this study. Furthermore, 99 (36.1%) nodules were classified as BSRTC II according to the FNA results, with a malignancy rate of 6.1%, and these patients underwent surgery for two main reasons. First, intermediate and high suspicion nodules accounted for 42.4% of all nodules, and among the low and very low suspicion nodules, more than half measured >2 cm. Therefore, clinicians in China are more likely to adopt and to devise treatment protocols using the ultrasound features of nodules. Secondly, surgery was preferred by patients.

The AUCs of the ATA classification and FNA results were 0.88 and 0.878, respectively (P < 0.001), indicating a high diagnostic value. The ATA classification showed a sensitivity of 86% and a specificity of 86.9%. For the ATA classification, a number of studies reported a sensitivity of 77.3–98.2% and a specificity of 37.4–76.6% (32, 33, 34, 35), consistent with the results of this study, whereas FNA results showed a sensitivity of 68.5% and a specificity of 91.4%. The low sensitivity indicated a high rate of misdiagnosis. In this study, the optimal cutoff point of the FNA results was BSRTC V, indicating that BSRTC III nodules were classified as benign in the calculation of sensitivity. Because the malignancy rate of BSRTC III nodules was 39.7% and these nodules were considered benign, the misdiagnosis rate increased. On the other hand, FNA results showed a high specificity (91.4%), indicating that the misdiagnosis rate was low, and a NPV of 81.2%, suggesting that the FNA results contributed to the diagnosis of thyroid cancer.

After combining the ATA classification and FNA results, if a nodule was considered malignant by both criteria, the specificity and PPV increased to 98.7 and 97%, respectively, indicating that when both criteria indicate malignancy or a suspicious malignancy, the malignant possibility of such a nodule is very high, and the probability of misdiagnosing a malignant nodule as a benign nodule is very low. Therefore, surgery is required for these patients. If a nodule was considered malignant by one criterion, the sensitivity and NPV increased to 94.3 and 95.1%, respectively, indicating that when both criteria indicate non-malignancy, the malignant possibility of such a nodule is low. Therefore, follow-up is recommended for these patients.

Recent studies have indicated that no single ultrasound feature can directly and accurately predict the malignancy of a nodule (36, 37); however, some characteristics, such as hypoechogenicity, irregular margins, taller-than-wide shape, microcalcifications and solidity, may predict thyroid cancer (14, 16, 17, 19, 20, 38). In this study, solidity, hypoechogenicity, irregular margins, taller-than-wide shape, microcalcifications and posterior echo attenuation associated with malignant nodules, whereas regular halo associated with benign nodules. Furthermore, taller-than-wide shape, microcalcifications, hypoechogenicity and irregular margins were independent risk factors for malignancy, consistent with the results of previous studies (15, 18, 39, 40, 41). Taller-than-wide shape is a highly specific indicator of thyroid cancer, with studies reporting a specificity of 81.5–98.5% for malignant thyroid nodules with a taller-than-wide shape (15, 23, 42). In this study, the OR value of taller-than-wide shape was 5.81, which significantly increased the risk of malignancy. The specificity was >95% for all nodules, and the accuracy of BSRTC II nodules was 91.9%. Microcalcifications are another important feature in predicting malignant thyroid nodules. The specificity of microcalcifications was high (90%), but the sensitivity was low (35.4–44.2%) (24, 42). In this study, 65.8% of malignant nodules had microcalcifications, which increased the risk of malignancy by 7.6 times. Furthermore, the sensitivity was >50%, whereas the specificity was >85%, indicating that the predictive value was high. Hypoechogenicity is another important indicator of malignancy, but the ability to predict malignancy was poor compared with taller-than-wide shape and microcalcifications. BSRTC I nodules showed high sensitivity (77.8%), specificity (88.9%) and accuracy (83.3%). In the presence of microcalcifications or hypoechogenicity, the sensitivity increased and the misdiagnosis rate decreased for various nodules.

Age and nodule size are also important predictors of malignancy. We found that patients with malignant nodules were younger than those with benign nodules, consistent with the results of previous studies (14, 43, 44). The research by Kwong et al. found an increased prevalence of thyroid nodules with older patients ,but it decreases the risk that such nodules will be malignant with advancing age (44). Although young patients have a high malignant of thyroid nodules, they have a better prognosis. The 8th Edition of AJCC for TNM staging suggested that extending the cutoff age to 55 years would prevent over-staging in low-risk patients and prevent over-aggressive treatment (45). However, the usefulness of nodule size in the diagnosis of malignancy is still controversial. A previous study has revealed a nonlinear correlation between size and malignant tumor risk. For nodules measuring 2.5 cm, the malignancy risk was the lowest. For nodules measuring <2.5 cm, the risk increased 53% for every centimeter decrease in size, and for nodules measuring >2.5 cm, the risk increased 39% for every centimeter increase in size (46). Kamran reported that the malignant risk of nodules measuring 1.0–1.9 cm was 10.5%, whereas that of nodules measuring >2 cm were 15%. However, the malignant risk of nodules measuring >2 cm was not increased (47). In a study of 1003 nodules from 659 patients, nodules measuring <2 cm (~30%) had the highest malignancy risk with no graded decrease in risk for nodules measuring >2 cm (~20%), indicating that thyroid nodule size is inversely related to malignancy risk, as larger nodules showed lower malignancy rates (48), consistent with the results of this study. However, in our study, 75.5% of nodules were <2.5 cm in diameter, and considering that patients with larger nodules may have underwent surgery in the absence of FNA, there was selection bias, which increased the malignancy rate of small nodules.

There were several limitations in our study. First, the number of cases and samples were small, and further studies with larger sample sizes are needed. Secondly, there was selection bias, because patients included in this study underwent FNA and surgery, indicating that patients were not representative of the whole population. Therefore, the malignancy rate may be higher than that reported in this study. Thirdly, this was a single-center, retrospective study, which may have reduced the statistical significance. Therefore, multi-center and prospective studies from different institutions and regions are necessary to verify our findings.

In conclusion, both the ATA classification and FNA results had high value in detecting benign and malignant nodules. The combined use of both criteria was more effective determining the nature of thyroid nodules. Analysis of ultrasound features revealed that taller-than-wide shape, microcalcifications, hypoechogenicity and irregular margins were the main predictive factors of malignancy. For BSRTC I, III and IV nodules, taller-than-wide shape showed the highest specificity, which greatly reduced the misdiagnosis rate. In the presence of microcalcifications or hypoechogenicity, the sensitivity improved, thereby reducing the misdiagnosis rate. Therefore, the combined use of the ATA classification, FNA results and specific ultrasound features can determine the nature of nodules more accurately, provide more valuable reference information for the selection of the treatment protocol, and guide clinicians to make correct decisions.

Declaration of interest

The authors declare that there is no conflict of interest that could be perceived as prejudicing the impartiality of the research reported.

Funding

This study was supported by the National Natural Science Foundation of China (Grant No. 81770784).