Predictors for malignancy risk in subcentimeter thyroid nodules categorized as atypia/follicular lesion of undetermined significance by fine needle aspiration

Little work has been done on the prediction of malignancy risk in patients with subcentimeter thyroid nodule (TN) categorized as atypia/follicular lesion of undetermined significance (AUS/FLUS). We performed a retrospective analysis on the medical records of subcentimeter TNs whose initial fine-needle aspiration (FNA) diagnosis was AUS/FLUS at our center between November 2013 and August 2018. Univariate analysis and multivariate logistic regression analysis were used to select independent factors associated with malignancy. Of the 324 patients who were classified as AUS/FLUS on initial FNA, 153 patients underwent surgical procedures and showed an associated malignancy rate of 45.10% (69/153). The malignancy rates in AUS/FLUS settings with and without repeat FNA were 38.30% (18/47), and 48.11% (51/106), respectively, p = 0.260. Multivariate logistic regression analysis revealed that age < 55 (OR 3.015, 95% CI 1.196–7.596), microcalcification (OR 9.162, 95% CI 3.332–25.916) and taller than wide shape (OR 10.785, 95% CI 4.108–28.319) were three independent predictors for malignancy. The malignancy rates in the patients with one or none of predictor and patients with two or three above predictors were 20.5% (17/83) and 74.3% (52/70), respectively, p < 0.001 (OR 11.216, 95% CI 5.266–23.885). In conclusion, our study showed that for subcentimeter TNs with AUS/FLUS category, patient’s age, taller than wide shape and microcalcification were three independent predictive factors for malignancy, which was helpful for decision-making of surgery or observation in such patient population.

nodule <1 cm (also named as subcentimeter TN) was not routinely recommended for FNA in the American Thyroid Association (ATA) guideline 10 .
Subcentimeter TN with any suspicious ultrasound characteristic for malignancy is routinely recommended to undergo FNA in our center, if patients prefer to exclude thyroid malignancy or have high risk factors such as family history of thyroid cancer, radiation history and evidence of lymph node metastasis. For those subcentimeter TNs with the cytological diagnosis of AUS/FLUS, the further managements represent an ongoing challenge, because no research has been carried out to support any of the following managements: clinical observation, ultrasound follow-up, repeat FNA or surgery.
A wide range of 6-48% malignancy risk in specimens categorized as AUS/FLUS was described in previous studies [11][12][13][14][15] . However, for subcentimeter TNs with suspicious US features, the accurate malignancy rate is unclear, and whether there is any predictive factor for malignancy is uncertain. Therefore, our study aims to investigate the rate of malignancy in subcentimeter TNs categorized as AUS/FLUS, and further to explore predictive factors for malignancy in patients who underwent operation.

Material and Methods
Patients. After approval from the Institutional Ethics Committee, we retrospectively analyzed the data from thyroid aspirates at Ningbo Medical Center Lihuili Eastern Hospital between November 2013 and August 2018. Thyroid nodules with one or more of the following suspicious US characteristics: (1) poorly-defined margin; (2) taller than wide shape; and (3) microcalcification, were recommended to undergo FNA under US guidance regardless of the nodule size. For cases with multiple nodules, the specimen was obtained from the lesion that was suspicious for malignancy. Nodule size, location, composition, echogenicity and vascularity of the nodule were all evaluated. Repeat FNA was performed for a proportion of referral case in our institution. Written informed consent was obtained from every patient.
Exclusion criteria included: (1) patient with thyroid nodule ≥1 cm in greater diameter; (2) patient with a history of thyroid carcinoma; (3) patient with the evidence of neck lymph node metastasis; (4) patient without the final histopathology evaluation. Cystic or mixed nodules, isoechoic or hyperechoic nodules were also excluded from the study.
Statistical analysis. T test for continuous variables, Chi-squared test or Fisher's exact test for categorical variables were used to detect predictors for malignancy in univariate analysis. Then, multivariate logistic regression analysis, including all variables from univariate analysis that were associated with malignancy, was performed to test factors' independence. P value of <0.05 was considered to have statistical significance; hazard ratio (HR) and 95% confidence intervals (CI) were also calculated. Statistical tests were two-sided, and analyses were performed using SPSS v. 20.0 Software (SPSS, Chicago, IL, http://www.spss.com).

US characteristics.
Restricted with inclusion criteria, the associations between echo structures (solid/cystic/ mixed), echogenicity (hypoechogenicity/isoechogenicity/hyperechogenicity) and malignancy risk were not explored in the study. Univariate analysis found that microcalcification, taller than wide shape and poorly-defined margin were associated with malignancy ( Table 2).
The malignancy risk on different subgroup. According to the presence or absence of the above three predictive factors, 153 patients could be classified into eight groups ( Table 4). The highest rate of malignancy (8/8, 100%) was found in the group with three predictive factors, followed by 69.2-72.2% in the group with two predictive factors, then 13.6-33.3% in the group with one predictive factor, and the lowest was (3/16, 18.8%) in the group without any predictive factor. When patients with two or three predictive factors were classified into one group, patients with one or none of predictive factor were classified into the other group, there was a significant difference on malignancy between two groups (Table 5).

Discussion
The overall AUS/FLUS utilization rate is 6.34% (324/5109) in our study, which is within the recommended 7% utilization for this diagnostic category. It is implied that there was a cancer risk range from 5% to 15% for this category. However, in the present study, the malignancy rate is 45.10%, which is much higher than the acceptable threshold. The characteristics of TNs in our study are specific, which may contribute to this result. Firstly, the included TNs are subcentimeter in size, which is always evaluated in China if there is any suspicious US characteristic. Small size of TNs decreases the satisfaction and accuracy of FNA, which may increase the inconclusive diagnosis especially unsatisfactory and AUS/FLUS category. Secondly and the most importantly, besides hypoechogenicity and solid characteristics, the included TNs also have one or more following suspicious US characteristics: microcalcification, poorly-defined margins, taller than wide shape. That is to say, the TNs in our study www.nature.com/scientificreports www.nature.com/scientificreports/ contain three or more suspicious characteristics, and should be categorized as TI-RADS 4b, 4c even 5. Therefore, the nodules in the current study should have greater risk of malignancy irrespective of small size.
For AUS/FLUS category, the common recommendation in ATA is to repeat US-FNA. Kuru B. et al. 16 observed a significantly increasing malignicancy rate in patients with AUS/FLUS category when a repeat FNAB was performed compared with those without repeat FNAB. However, in other studies, there was no significant difference between patients with repeat FNAB and without repeat FNAB 17,18 . We found the similar result, which showed that the malignancy rate was 38.30% (18/47) in patients with repeat FNAB, while 48.11% (51/106) in patients without   18 also did not find a significant difference in patients who had one AUS diagnosis and two successive AUS diagnosis. Repeat FNA may make a conclusive diagnosis in a proportion of cases 16 , therefore, it could be a reasonable choice for initial AUS diagnosis.
However, we found the conclusive diagnosis could be determined only in 23.40% (11/47) patients, including 3malignancy diagnosis in repeat FNA and 8 benign diagnosis. In the remaining 36 inconclusive diagnosis after repeat FNA, there were 2 (22.22%) patients with unsatisfactory diagnosis and 8 (36.36%) patients with AUS/ FLUS, 5 (100.00%) patients with sM were finally determined to be malignancy after surgery. That is to say, the TNs diagnosed with malignancy and sM after repeat FNA should be surgically removed, and there was still a malignancy risk of 22.22-36.36% in unsatisfactory and AUS/FLUS category.
We find that younger age, microcalcification and taller than wide shape are three independent predictors for malignancy. Age is determined as one of prognostic factors for differentiate thyroid carcinoma (DTC), and the most important parameter in TNM staging system for DTC. However, little attention was made previously to its influence on the risk of malignancy. Recently, the eighth (8 th ) American Joint Commission on Cancer (AJCC) staging system recommended 55-year age as an ideal cut-off threshold for thyroid cancer staging 19 . In our study, we stratify the included patients into two groups: younger than 55-year age group and 55-year age or older group. For patients with AUS/FLUS, the rate of malignancy is 50.9% in the younger group, and 28.2% in the older group, which suggests that TN with an AUS/FLUS diagnosis in younger people is associated with a higher risk of malignancy when compared with older people. Todorovic, E. et al. 20 also found that age <55 was a risk factor for malignancy in thyroid nodule with AUS/FLUS.
Ultrasound characteristics are always used to estimate the malignancy risk for thyroid nodules, and well-accepted worldwidely ultrasound characteristics for malignancy include solid structure, hypoechogenicity, irregular margins, microcalcification, and taller than wide shape, which are the foundations of TI-RADS. In the present study, for subcentimeter TNs with AUS/FLUS category, microcalcification and taller than wide shape are associated with the higher risk of malignancy. Chng. et al. 21 and Maia. et al. 22 reported that irregular margin of the TN was the utmost predictor for malignancy. However, we could only observe a tendency for association between irregular margin of TN and the risk of malignancy. The predictive values of solid structure and hypoechogenicity of nodule are not assessed in this study, because the included nodules are solid and hypoechoic TNs.
The highlights of the study should be acknowledged. Firstly, it is the first study with regard to the predictors for malignancy in subcentimeter TNs. The management of subcentimeter TNs is challenging, because widely-used TI-RADS and BSRTC after FNA are performed to estimate malignancy risk for the nodules ≥1 cm. Malignancy risk evaluation for subcentimeter TNs is rarely done in practice, because TN <1 cm is not routinely recommended for FNA in ATA guideline 10 .
Additionally, the decreased FNA accuracy in lower size nodules also aggravates the difficulty. The findings in our study would provide some evidence for malignancy risk evaluation in subcentimeter TNs. Secondly, microcalcification and taller than wide shape are found the more powerful predictive factors than patient's age, which indicates that physicians should pay more attention to suspicious US characteristics than patient's age when malignancy risk of thyroid nodule is assessed. Thirdly and the most importantly, we could stratify the subcentimeter TNs into two groups with different risk of malignancy: patients with two or more predictors have a 74.29% of malignancy risk, while the patients with one or none of predictor only have a 20.48% of malignancy risk. Our findings suggest that for subcentimeter TN with two or more predictors, even if the initial FNA diagnosis of AUS/ FLUS, thyroidectomy should be a suitable choice; while for that with one or none of above predictor, repeat FNA or observation sounds reasonable.  Table 4. The rates of malignancy in different groups stratified by three independent predictive factors. A: stands for age < 55 year old; B: stands for taller than wide; C: stands for microcalcification.

Hazard ratio 95%CI
Group with two or three predictors 52 18 <0.001 11.216 5.266-23.885 Group with one or none of predictors 17 66 Table 5. Comparison of malignancy rate between subgroups with different number of predictive factors.