Preoperative High-Resolution Ultrasound for the Assessment of Macro-Metastatic Central Lymph Nodes Sublevel-By-Sublevel in Unilateral Papillary Thyroid Microcarcinoma

Wen Tan Sun Yat-sen University Cancer Center Lingli Xiao Sun Yat-sen University Cancer Center Shiwen Zhang Sun Yat-sen University Cancer Center Haoming Zheng Sun Yat-sen University Cancer Center Sheng Li Sun Yat-sen University Cancer Center Yu Zhang Sun Yat-sen University Cancer Center Ankui Yang Sun Yat-sen University Cancer Center Jianhua Zhou Sun Yat-sen University Cancer Center Longzhong Liu (  liulzh@sysucc.org.cn ) Sun Yat-sen University Cancer Center https://orcid.org/0000-0003-4639-1767


Introduction
The global incidence rate of thyroid cancer has increased rapidly by 3-15 times since the 1980s. Among them, the incidence rate of large papillary thyroid cancer (PTC) and other pathological types of thyroid cancer has remained the same compared with that before 1980. The vast majority of rapidly increasing cases are small papillary thyroid carcinoma (less than 15mm), half of which are papillary thyroid microcarcinoma (PTMC) 1, 2 , de ned as the longest diameter less than 10mm, regardless of whether there is lymph node metastasis 3 .
Generally, PTMC progresses slowly and has a good prognosis, but it is more often accompanied by central lymph node metastasis (CLNM) [4][5][6][7] , which is also proved to be an independent risk factor for recurrence in patients with thyroid cancer [8][9][10][11] . The 2015 ATA guideline suggested 12 that thyroidectomy without prophylactic central neck dissection may be is appropriate for small (T1 or T2), noninvasive, clinically node-negative PTC (cN0) (Strong Recommendation). However, according to the guideline, the patients with thyroid cancer accompanied by macro-metastatic cervical lymph nodes (the cancer foci in the lymph node longer than 2mm 13 ) have an intermediate risk of recurrence. It means that, it is necessary to perform central neck dissection even for PTMC patients if preoperative high-resolution ultrasound (HRUS) detects macro-metastatic central lymph nodes (macro-CLNM). Otherwise, there will be an increasing rate for the second surgery. Therefore, understanding the diagnostic e cacy of preoperative HRUS in the diagnosis of macro-CLNM will help surgeons decide whether to perform prophylactic lymph nodes (LNs) dissection in patients with PTMC. Although several studies indicate that preoperative ultrasound has low diagnostic e ciency for CLNM of PTMC 5,14 , whether the diagnostic e cacy of HRUS for macro-CLNM is also very low, there has not been any literature report as far as we know.
American Joint Committee on Cancer staging system 15 (AJCC, 8th edition) divides central lymph nodes (LNs) into four sublevels, including prelaryngeal LNs, pretracheal LNs, the left and the right paratracheal LNs. The dissection scope of central LNs for PTMC consisted of the LNs in prelaryngeal, pretracheal, and at least ipsilateral paratracheal sublevel 16 . Although it is di cult to correspond the cervical LNs detected by HRUS with that in pathology one by one, if the central compartment can be subdivided, and the LNs can be corresponded in ultrasonic examinations and pathology according to the four sublevels, the evaluation of HRUS diagnostic e cacy will be more accuracy.
To sum up, this study covered 220 consecutive PTMC patients, comparing their preoperative ultrasound and postoperative pathology of the central LNs in each sublevel to gure out the macro-CLNM rates and the diagnostic e cacy of preoperative ultrasound in each sublevel. It will be a reference for developing individualized surgical plans for PTMC patients.

Methods
This study was approved by our Institutional Review Board, and the requirement for informed consent was waived because of its retrospective nature. This article's authenticity was validated by uploading the key raw data onto the Research Data Deposit public platform (www.researchdata.org.cn), with the Research Data Deposit number RDDA2020001813.

Patients
In order to reduce the interference of the contralateral central lymph node and lateral cervical lymph node metastasis on the analysis of the ultrasound diagnosis performance in each sublevel, this study analyzed the preoperative ultrasound and postoperative pathological results of 220 patients with unilateral PTMC. These patients all underwent surgery at the Sun Yat-sen University Cancer Center from January 2015 to November 2019. The inclusion criteria are as follows: 1) PTMC Patients whose lymph nodes were nopalpable by physical examination; 2)Unilateral PTMC was con rmed by postoperative pathology; the surgery method was thyroidectomy total or lobectomy and ipsilateral central lymph node dissection; 3) preoperative HRUS examination and postoperative pathological results were reported sublevel-bysublevel. Patients were excluded if they met the following criteria: have received treatment before, with contralateral central or lateral cervical lymph node dissection, with distant metastases, were followed-up for less than one year, recurrence was found during the follow-up, had incomplete pathological data.

Preoperative ultrasound examination and diagnosis
Patients included in this study underwent ultrasound examinations one week before the surgery, using a high-resolution sonography system (GE, Mindray, SIEMENS, Toshiba, Super Sonic, Hitachi, PHILIPS, etc.) equipped with an at least 10MHz linear array transducer. Central compartment (level VI) lymph nodes were divided into prelaryngeal, pretracheal, and ipsilateral paratracheal sublevel by AJCC (8th edition).
During the preoperative ultrasound examination, the largest and/or the most suspicious LNs detected in each sublevel should be marked down, and the following sonographic features of lymph nodes should be described: the length of the long and short axis, the shape (oval, round, or irregular), the margin (regular or irregular), corticomedullary demarcations (clear or unclear), hilar structure (detected or not detected) and blood ow signals (hilar, peripheral, or mixed type). Besides, for suspicious LNs, the features supporting the diagnosis of metastasis should be recorded. The 2015 ATA guidelines 12 pointed out that the cervical LNs should be considered metastatic if the following sonographic features occurred: round shape, microcalci cations, cystic aspect, hyperechogenicity, and peripheral vascularization.

Surgery and pathological data
All patients included in this study underwent unilateral central compartment lymph nodes dissection, and the surgical area covered from prelaryngeal level, pretracheal level to ipsilateral paratracheal level, and the excised lymph nodes were marked by sublevels and sent for pathological examination. By reviewing the pathological slices, two experienced pathologists divided the central LNs in each sublevel into no, micro (the cancer foci was shorter than 2mm), and macro-CLNM (the cancer foci was 2mm or longer than 2mm) according to the classi cation of metastatic LNs in the 8th AJCC 13 . In this study: a sublevel was de ned as no-metastasis sublevel when there were no metastatic LNs; a sublevel was de ned as micrometastasis sublevel when there was one or more micro-metastatic LNs without any macro-metastatic LNs; a sublevel was de ned as macro-metastasis sublevel when there was at least one macrometastatic LN. If the two pathologists had different opinions on the diagnosis of the LNs, the two will give a consensus diagnosis after consultation.

Statistical analysis
We used MedCalc® statistical software to analyze the statistics. The independent-samples t-test was used to compare continuous variables. Categorical variables were compared using the Chi-squared (χ 2 ) test. We calculated the ultrasound diagnostic accuracy (ACC), sensitivity (SENS), speci city (SPEC), positive predictive value (PPV), negative predictive value (NPV), and the area under the ROC curve (AUC) to nd out the ultrasound diagnostic e cacy for macro-CLNM or CLNM in each sublevel. P-values < 0.05 were considered indicative of a statistically signi cant difference.

Results
Among the 220 patients with PTMC included in this study, 217 cases were pathologically con rmed as classical variants, while 3 cases were follicular variants. There were 95 patients with metastatic central compartment LNs, which accounted for about 43%. The included cases were divided into 220 sublevels of prelaryngeal LNs, 220 sublevels of pretracheal LNs, 104 sublevels of left paratracheal LNs, and 116 sublevels of right paratracheal LNs (Table 1). The number of LNs and macro-metastasis rate in each sublevel according to pathological results The postoperative pathological results (Table 2 and Table 3) showed a tiny number of prelaryngeal LNs, with only 188 LNs removed in 220 patients. Meanwhile, the LN macro-metastasis rate in prelaryngeal sublevel was also the lowest among all sublevels, only about 0.91%. Next was the pretracheal LNs, with an average number of about 1.9 ± 0.1 LNs per case; the macro-metastasis rate in the pretracheal sublevel was higher than that of the prelaryngeal LNs, reaching 6.36%. It was worth mentioning that the number of LNs in the right paratracheal was signi cantly beyond that in the left (P<0.0001), whose average values were 4.6 ± 0.3 and 3.3 ± 0.2, respectively. The macro-metastasis rate of the right paratracheal LNs was also higher than that of the left side, accounting for approximately 11.21% and 6.73%, respectively.   11.21% (13) The size and the HRUS detective rate of macro-metastatic LNs in each sublevel.
As shown in the Table 4, among the 660 sublevels, the detective rate was extremely low (0-21%) in the prelaryngeal and pretracheal sublevels no matter there was metastatic LNs or not. However, as for ipsilateral paratracheal sublevels, the detective rate of macro-CLNM was higher than non-macro-CLNM, with the macro-CLNM detective rate in left and right paratracheal of 85.71% and 76.92%, respectively, the non-macro-CLNM detective rate of 38.14% and 32.04%, respectively. However, there was no statistical signi cance in the HRUS detective rate of no-and micro-CLNM in the left and right paratracheal sublevels. In Table 5, the mean long-axis diameter of the macro-CLNM in the prelaryngeal, pretracheal, left paratracheal, and right paratracheal was 4mm, 5.4mm, 4.3mm, and 4.9mm, respectively; the mean longaxis diameter of the non-macro-CLNM was 3mm, 3.5mm, 3.6mm, and 3.1mm, respectively. the long and short-axis diameters of the macro-CLNM were longer than that of the non-macro-CLNM in each sublevel, and there was statistical signi cance in the pretracheal and left paratracheal sublevels (P<0.05).  The ultrasound diagnostic e cacy for macro-CLNM or CLNM in each sublevel As revealed in Table 6 and Table 7, no matter for macro-CLNM or CLNM, the HRUS diagnostic e cacy in the prelaryngeal and pretracheal sublevels was low, with the AUC of 0.51(95% CI 0.44-0.58) and 0.55 (95% CI 0.48-0.62), respectively. Moreover, the false-negative rate of diagnosing prelaryngeal and pretracheal LNs by HRUS reached 100% (2/2) and 86% (12/14), respectively. Among them (Table 4), the proportion of LNs that were not detected by ultrasound was very high, 100% (2/2), 92% (11/12), respectively. Compared with the prelaryngeal-pretracheal sublevels, HRUS was more effective in diagnosing macro-CLNM in the paratracheal sublevels, with the AUC of 0.80 (95% CI 0.71-0.87) in the left and 0.75 (95% CI 0.67-0.83) in the right. Meanwhile, the diagnostic e cacy of macro-CLNM was also signi cantly higher than that of CLNM in the paratracheal sublevels (0.51 on the left and 0.54 on the right). When the diagnosis object was changed from CLNM to macro-CLNM, the negative predictive value (NPV) was also increased considerably. The NPV of the ultrasound diagnosis of macro-CLNM was 99% of the left paratracheal and 95% of the right side of the paratracheal.   19 . On the other hand, there is no de nite evidence to prove that PCND can improve the prognosis or prolong the survival of patients with thyroid cancer, and the expansion of the dissection scope will increase the incidence of postoperative complications 19,20 . Lymph node macro-metastasis is a vital risk strati cation factor for PTMC patients according to the 2015 ATA guidelines 12,21,22 . Therefore, preoperative identi cation of macro-CLNM can provide a certain reference for the management of PTMC patients.
The number and macro-metastasis rate of lymph nodes in PTMC patients are, from high to low, the right paratracheal, the left paratracheal, the pretracheal, and the prelaryngeal level. The right paratracheal level has more LNs and a higher macro-metastasis rate than the left side, which may relate to anatomy. The angle between the right RLN and the carotid artery is larger than the angle on the left side, where the RLN is parallel to the tracheoesophageal sulcus 23 .
This study demonstrated that the preoperative ultrasound had an extremely high false-negative rate for assessing macro-metastatic LNs in prelaryngeal-pretracheal sublevel, there were 14 macro-CLNM cases in pretracheal and prelaryngeal sublevel which were diagnosed as negative by HRUS. Among them, HRUS failed to detect any LN in 13 cases. The LNs cannot be observed by ultrasound because the prelaryngealpretracheal LNs are located between the anterior cervical muscles and the trachea. The relatively small space in this area makes the number and the size of lymph nodes small. There were only 1-2 prelaryngeal-pretracheal LNs per patient (Table 2); the average short-xis diameter of the prelaryngeal and pretracheal LNs was less than 3mm and 2-4mm, respectively (Table 5), and even the smallest nodule detected by HRUS is only 2-3mm 24 . The extremely low detective rate of HRUS in prelaryngeal-pretracheal LNs caused many false-negative cases, with the AUC of only 0.51-0.55 (Table 6). We suggested that preoperative HRUS did not help identify the macro-CLNM in prelaryngeal-pretracheal sublevel.
Compared with the prelaryngeal-pretracheal sublevel, dissection of LNs in the paratracheal levels is more likely to cause complications such as hoarseness and hypocalcemia common in post thyroid cancer operation because of the injury of recurrent laryngeal nerve and parathyroid 19 . Therefore, the preoperative diagnosis of the paratracheal LNs is of great signi cance. The diagnostic e cacy of HRUS for CLNM in paratracheal sublevels is low, with the AUC of only 0.51-0.54, which is consistent with the conclusions of other previous studies 5,14 . For macro-CLNM, the AUC of the paratracheal sublevels has been signi cantly improved, with 0.80 on the left and 0.75 on the right. This is probably because the diameter of macro-CLNM is larger than that of the non-macro-CLNM (including no-and micro-CLNM) ( Table 5), which is more easily to be detected by HRUS (Table 4). In addition, micro-CLNM has only pathological changes below 2 mm, which is di cult to distinguish by ultrasound, while cancer foci in macro-CLNM are larger than 2 mm, which are more likely to show abnormalities in ultrasound images. Moreover, the NPV of ultrasound diagnosis of macrometastasis in paratracheal sublevels reached 99% (left) and 95% (right). In other words, when the paratracheal sublevel is negative under HRUS, the probability of macrometastasis is extremely low, only 1%-5%. Therefore, this study believes that preoperative HRUS has an exclusive diagnostic effect on macrometastasis in paratracheal sublevels.
It is worth mentioning that macrometastasis in prelaryngeal-pretracheal sublevel is a high risk factor for paratracheal macrometastases, with an OR value of 9.1 (95% CI 2.8-29.2, P = 0.0002), And among the patients with macrometastases in prelaryngeal-pretracheal sublevel, nearly 40% of them accompanied by paratracheal macrometastases (6/15). As a result, macro-CLNM in prelaryngeal-pretracheal sublevel has a certain prompting effect on paratracheal LNs macrometastasis. If PTMC patients are found with prelaryngeal-pretracheal macro-CLNM during surgery, it is necessary to be alert to ipsilateral paratracheal macro-CLNM.
This study creatively analyzed the preoperative ultrasound and postoperative pathological results of patients with PTMC sublevel-by-sublevel, guring out the number of LNs and macro-metastatic rate in each sublevel and discussing the diagnostic e cacy of ultrasound. Compared with the traditional rough analysis of central LNs, the ultrasound diagnostic performance obtained in this paper is more accurate and closer to the real situation. However, there are some limitations to our study. Firstly, in order to reduce the interference of contralateral central LN and lateral cervical LN to the analysis of ultrasound diagnostic performance in each sublevel, this retrospective study design only focus on unilateral PTMC, so there is an unavoidable selection bias. Secondly, this study can only analyze LNs sublevel-by-sublevel rather than single LN one-to-one. It remains a clinically signi cant topic that urgently needs research on whether combining with other imaging tests or arti cial intelligence can improve diagnosis e ciency.

Conclusion
1. The number and macro-metastasis rate of lymph nodes in unilateral PTMC patients are, from high to low, the right paratracheal, the left paratracheal, the pretracheal, and the prelaryngeal level. 2. Preoperative HRUS has a high diagnostic e ciency for macro-CLNM in the ipsilateral paratracheal area (AUC is 0.80 on the left and 0.75 on the right, respectively); when the ultrasound diagnosis of the ipsilateral paratracheal area is negative, the probability of macrometastasis is extremely low. This study was approved by Sun Yat-sen University Cancer Center Institutional Review Board, and the requirement for informed consent was waived because of its retrospective nature.

Consent for publication
All the patients involved in this study were informed and the agreed that their personal data would be used for publication for the purpose of scienti c research.

Availability of data and materials
This article's authenticity was validated by uploading the key raw data onto the Research Data Deposit public platform (www.researchdata.org.cn), with the Research Data Deposit number RDDA2020001813.

Competing interests
Wen Tan, Lingli Xiao, Shiwen Zhang, Haoming Zheng, Sheng Li, Yu Zhang, Ankui Yang, Jianhua Zhou, and Longzhong Liu have no con icts of commercial interest in the subject of study and the source of any nancial or material support to disclose.

Funding
This study has received funding from Natural Science Foundation of Guangdong Province, and the funding number is 2021A1515010546.

Authors' contributions
Longzhong Liu contributed substantially to the conception of the work; Longzhong Liu, Jianhua Zhou, Wen Tan, and Lingli Xiao participated in the study design; Wen Tan and Lingli Xiao drafted the work; Longzhong Liu, Wen Tan, and Lingli Xiao revised it; Shiwen Zhang, Yu Zhang, and Haoming Zheng participated in the acquisition of the data of the work and offered statistically analysis; Sheng Li and Yu Zhang offered imporment comments on the methology; Ankui Yang and Jianhua Zhou provided critical comments on the work. All the authors had read and approved the nal article.