Utility and significance of clinical risk factor scoring model in predicting central compartment lymph node metastasis (CLNM) in patients with papillary thyroid cancer (PTC)

ABSTRACT Objectives: To establish and discuss the significance of a clinical risk factor scoring model in predicting central compartment lymph node metastasis (CLNM) (level VI) in patients with papillary thyroid cancer (PTC). Methods: A retrospective analysis was performed on 412 patients who underwent surgical treatment for PTC who were admitted to the Second Hospital of Hebei Medical University between July 2016 and May 2017, with the patients being divided into a CLNM group and a non-metastasis (NM) group. Risk factors such as sex, age, tumor diameter, capsular invasion, multifocality, and tumor location were recorded for scoring via maximum likelihood estimation (MLE)-based discriminant analysis. The scoring model was used for prospective analysis of CLNM in another 104 patients. Besides, the discriminant function that was developed using the risk factors based on the retrospective data derived from the 412 patients was evaluated by plugging the retrospective data in for specified variables, with a higher score indicating a greater risk of developing CLNM. Clinical diagnosis of CLNM was based on postoperative paraffin section pathology, which was adopted as the criterion to assess discriminative accuracy in the prospective and retrospective groups. Results: The discriminative accuracy of the scoring model was 71.8% in the retrospective group and 72.2% in the prospective group. Conclusions: The scoring model enables simplified, quantitative analysis of CLNM in PTC patients. The scoring model has clinical significance in that it provides a basis for the choice of operation, personalized postoperative treatment, and prognosis of PTC.


INTRODUCTION
Thyroid cancer (TC) is a common and frequently occurring disease in otorhinolaryngology -head and neck surgery, and papillary thyroid cancer (PTC) accounts for 80% to 90% of the overall incidence. In some PTC patients, cervical lymph node or distant metastasis is present at diagnosis. Metastasis does not necessarily first appear but is most likely to occur in the central compartment (level VI) of the affected side. 1 Ultrasonography is the main imaging technique for the detection of thyroid and regional lymph nodes, which exhibits excellent sensitivity and accuracy in the differential diagnosis between benign and malignant thyroid nodules. Compared with lateral neck ultrasound, which is shown to have a sensitivity of up to 93.8%, ultrasonography of central compartment lymph nodes (CLNs) displays a relatively low diagnostic sensitivity (about 31.3%) due to anatomical and technical reasons. 2 This study presented a maximum likelihood estimation (MLE)-based retrospective analysis of a group of PTC patients to calculate the scores of relevant risk factors for central compartment lymph node metastasis (CLNM), and the scoring criteria were also applied to the prospective screening for CLNM in another group of PTC patients, aiming to assess the significance of this risk factor scoring model as a clinical guidance tool in predicting CLNM in PTC patients.   (Table-I). Statistical Analysis: The frequency of occurrence P(X kj /Y g ) of each clinical risk factor was calculated using the data available in Table-I to determine the presence of metastasis in the central compartment, where Y represents the type of CLNs, with g =1 meaning positive (+) and g =2 negative (-) for metastatic lymph nodes, and X kj (k =1, 2, 3, 4, 5, 6; j =1, 2) stands for the jth class of the kth risk factor. An MLE-based discriminant analysis of qualitative data was conducted to calculate the probability of CLNs being classified as g when a patient falls in the jth class of the kth risk factor: P g =P(X 1j /Y g ) ×P(X 2j /Y g ) ... ×P(X 6j / Y g) . Log probability can be calculated using the following equation: lgP g =lgP(X 1j /Y g ) ×P(X 2j /Y g ) ... ×P(X 6j /Y g ). The score of each risk factor can be obtained from the following equation:[lgP(X kj /Y g ) +1] ( Table-II). The scoring formula is written as follows:

Operation methods and postoperative followup:
The retrospective and prospective groups consisted of 516 patients in total. All patients had received standard surgical treatment, including 214 cases of thyroid lobectomy with isthmusectomy and homolateral CLND, 218 cases of total thyroidectomy with homolateral CLND, and 84 cases of total thyroidectomy with bilateral CLND. Levothyroxine sodium tablets were administered orally after surgery, in combination with thyroidstimulating hormone (TSH) suppression therapy. Thyroid function was tested a month after surgery to adjust the dose of levothyroxine sodium tablets according to the TSH level. During the adjustment period, thyroid function tests were taken every one and a half months. Thyroid and cervical lymph node ultrasound was performed six months after surgery. Subsequent reexaminations were recommended every six months, including physical checkup, thyroid function test, B-mode ultrasound, electrocardiogram (ECG), and chest X-ray. The patients were advised to seek prompt medical attention if any symptoms occurred.
Except two patients who were lost to follow-up because they had changed their contact information without prior notice, all the other patients had completed follow-up for one to four years, with the mean follow-up period reaching 28.3 months. Postoperative hypocalcemia and hoarseness were observed in some cases, and full recovery was achieved within six months after surgery. No serious complications were detected. Table-II. The results of discriminant analysis showed that in the retrospective group, correct diagnoses were made in 129 positive (+) and 168 negative (-) cases of CLNM, with the coincidence rates of 65.5% and 78.1%, respectively, and the mean coincidence rate was 71.8% (Table-III); as to the prospective group, correct diagnoses were made in 28 positive (+) cases and 48 negative (-) cases of CLNM, with the coincidence rates of 68.2% and 76.2%, respectively, and the mean coincidence rate was 72.2% (Table-IV).

DISCUSSION
PTC is a clinically common disease, with B-mode ultrasound as the mainstay of differential diagnosis between malignant and benign thyroid nodules. It was shown in Dan HJ et al. 3 that high-frequency color Doppler ultrasound could produce a diagnostic accuracy of up to 87.64% in the diagnosis of PTMC. However, when applied to the diagnosis of CLNM, color ultrasound only has a sensitivity of 31.3% 2 , substantially lower than that in the diagnosis of lateral lymph node metastasis (up to 93.8% 4 ). This is probably associated with the unique anatomical position (adjacent to the thyroid and the air-containing trachea) of the CLNs or other technical factors.
CLNM risk may differ by diverse factors, six of which were chosen for discriminant analysis: (1) Sex: male patients were at higher risk of lymph node metastasis than female patients 5 ; among the 516 patients, 66 out of the 114 male patients were diagnosed with CLNM, and 172 out of the 402 female patients had CLNM, with the metastatic rates of 57.9% and 42.8%, respectively. (2) Age: According to the Chinese guidelines on tumornode-metastasis (TNM) staging (2012 edition) and foreign studies, an age ≥45 is considered as a highrisk factor for thyroid cancer. 6 Notably, it was reported in Ito et al. 7 that a younger age entailed a higher metastatic rate, which was up to 50% in TC patients under 20. (3) Tumor diameter: In the NCCN clinical practice guidelines in oncology, a tumor diameter smaller than 1-cm was considered as a high-risk factor of neck lymph node metastasis 8 .
PTC patients with a tumor diameter greater than 2 cm are shown to have higher rates of CLNM and lateral cervical lymph node metastasis (LCLNM) compared with those with a tumor diameter of or smaller than 2-cm. 9 In some studies, the threshold tumor diameter is set as 0.5 cm or 0.6 cm to analyze CLNM in PTMC patients. In a preliminary study 10 , it was proposed that in PTMC, a tumor diameter of or greater than 0.7 cm had statistical significance in predicting CLNM. (4) Capsular invasion: Radowsky et al. viewed capsular invasion by tumor cells was a key risk factor for the prognosis of PTC patients. 11 When capsular invasion occurs, the invasiveness of PTC cells increases, which is associated with the reduced inhibitory effect of the thyroid extracellular matrix (ECM) on lymph node metastasis in PTC, as well as the invasion of the thyroid's compact lymphoreticular system by PTC cells. 12 (5) Multifocality: Kuo et al. 13 reported that compared with unifocal PTMC, multifocal PTMC exhibited a significant increase in the metastatic rate, consistent with the findings in their preliminary study. (6) Tumor location: Zhang et al. 14 pointed out that when the foci occurred in the upper pole of the thyroid, there was an increased risk of LCLNM, but the risk of CLNM was reduced. Wang et al. 15 found that tumors in the mid-and lower pole of the thyroid were more likely to induce CLNM. Based on the data given in Table-

CONCLUSIONS
Indications of prophylactic CLND for cN0 PTC patients are still controversial at home and abroad. The present study used retrospective data to obtain frequency distributions of clinical risk factors for PTC and carried out an MLE-based discriminant analysis of the prospective group; the discriminant analysis was evaluated using the retrospective data. With the discriminant rules remaining unchanged when the actual probability was converted to log probability, the risk factor scoring criteria were established accordingly. For example, if a 28-year-old male patient is diagnosed with unifocal PTC with a tumor diameter of 1.6 cm, and the tumor is located at the lower pole of the lateral thyroid without any signs of capsular invasion, his scores for the six risk factors are as follows: S positive =37, S negative =30.2, S positive >S negative . The discriminant results show that this patient has CLNM.