Clinical Value of CTR, CEA, Histological Type, Ki-67 and EGFR in Predicting pN in Clinical Stage IA Lung Adenocarcinoma and Nomogram Construction

To investigate the clinical value of CTR, CEA, histological type, Ki-67 and EGFR in detecting pathological lymph node metastasis (pN) in clinical stage IA (cIA) lung adenocarcinoma and to construct a pN Nomogram model. A total of 374 cIA lung adenocarcinoma patients who had undergone thoracoscopic radical resection with Systematic mediastinal lymph node dissection (SMLD) in the Department the Aliated Hospital of University between January 2018 to January 2020 were retrospectively reviewed. The patients were divided into pN(+) and pN(-) groups. Univariate and multivariate Logistic regression analyses were used to analyze the independent risk factors of pN in lung cancer patients. The ROC curve was used to compare the accuracy of CTR, CEA and Ki-67 in predicting pN. R software was used to construct a Nomogram prediction model based on multivariate Logistic regression analysis of the pN risk. The C-index was calculated, and a calibration curve was drawn to judge the calibration degree of the model. The preoperative and intraoperative examinations showed that CTR (OR 570.406, P (cid:0) 0.001), CEA (OR 1.239, P (cid:0) 0.001) and micropapillary adenocarcinoma (OR 86.712, P (cid:0) 0.001) were independent risk factors of pN. Immunohistochemical analysis and gene detection showed that Ki-67 index (OR 4.832, P (cid:0) 0.001) and EGFR mutations, such as exon 19 (OR 10.319, P (cid:0) 0.001), exon 21 (OR 7.163, P (cid:0) 0.001) and exon 19+20 mutations (OR 570.406, P (cid:0) 0.001), were signicant factors in predicting pN. CTR, CEA, histological type, Ki-67 index, and EGFR mutations are the predictive factors of pN in cT1a-3aN0M0 lung adenocarcinoma patients. SMLD is recommended to improve patients’ postoperative survival rate when preoperative CTR ≥ 0.775, CEA (cid:0) 2.52μg/L or intraoperative rapid freezing pathology shows micropapillary components. on multivariate Logistic regression analysis model


Introduction
More and more early non-small cell lung cancers (NSCLCs) have been identi ed, especially in cIA, due to the introduction of chest high-resolution computed tomography (HRCT) and health awareness improvement.
Anatomical pulmonary lobectomy with SMLD is the standard procedure for NSCLC. However, pathological metastatic positive lymph node affects prognosis [1] and can signi cantly reduce the survival rate of lung cancer patients [2,3]. The lymph node metastasis rate of NSCLC (diameter≤2cm) is between 15%-20% [4,5]. Some scholars have shown that SMLD has greater surgical trauma, a higher incidence rate of postoperative complications, longer postoperative catheterization and hospitalization. Besides, extensive lymphadenectomy can decrease local immune function, leading to local recurrence and distant metastasis, indicating that SMLD is not necessary for all early NSCLCs [6,7]. Metastase-negative lymph node dissection is ine cient and can prolong operation or increase the perioperative complication risks [7,8]. Therefore, the clinicopathological data of 374 cT1a-3aN0M0 primary lung adenocarcinoma patients were retrospectively analyzed, the risk factors of pathologically metastatic positive lymph nodes were explored, and an individualized Nomogram model was constructed to predict the pN occurrence rate. This study provides a standardized theoretical basis for the formulation of intraoperative lymph node dissection and postoperative adjuvant therapy for cIA lung adenocarcinoma patients.

Methods
Patients A retrospective case-control study was conducted. The clinicopathological data of 374 cT1a-3aN0M0 lung adenocarcinoma patients who underwent anatomical pulmonary lobectomy with SMLD in the Department of Thoracic Surgery of the A liated Hospital of Qingdao University between January 2018 to January 2020 were collected (8th edition of TNM). A preoperative evaluation was conducted on all patients via thyroid ultrasound, chest CT, abdominal CT, craniocerebral CT and bone scanning. Chest CT showing lymph node with short axial ≥1cm indicated metastatic positive. Inclusion criteria were: age 70 years old, single nodule ≤3cm in diameter, hilar and mediastinal lymph nodes with long axis diameter ≤1cm in HRCT, non-distant metastasis, and pathologically con rmed primary lung adenocarcinoma. Exclusion criteria included preoperative immunotherapy, targeted therapy or chemoradiotherapy, a history of other malignant tumor and incomplete clinicopathological data.

Histopathological assessment
The resected tumor tissue and lymph nodes were stained using hematoxylin and eosin. Immunohistochemical and genetic tests were performed, and two pathologists reviewed the biopsy specimens. Histopathological information, Ki-67 index and EGFR mutation status were noted. Besides, the resected lung tissues were classi ed into acinardominant adenocarcinoma (APA), papillary type-dominant adenocarcinoma (PPA), micropapillary type-dominant adenocarcinoma (MPA), solid type-dominant adenocarcinoma (SPA) and anchorage type-dominant adenocarcinoma (LPA), according to the International Association for the Study of Lung Cancer (IASLC)/The American Thoracic Society (ATS)/the European Respiratory Society (ERS).

Statistical analyses
IBM SPSS 23.0 software was used for data analysis. The measurement data (skewed distribution) was expressed as a median. Mann-Whitney U test was used to compare the two groups. Chi-square test or Fisher's exact test were used to compare the enumeration data. P 0.05 was considered statistically signi cant. Univariate analysis was used to determine pN risk factors. Multivariate Logistic regression analysis of statistically signi cant factors was conducted to identify the independent risk factors for lymph node metastasis. The SPSS software was used to draw the receiver operating characteristic (ROC), including CTR, CEA, Ki-67, CTR+CEA combined prediction probability and CTR+CEA+Ki-67 combined prediction probability. The area under the curve (AUC) and the best cut-off value of ROC were then calculated. The MedCalc software was used to construct the ROC curves of CTR, CEA and Ki-67 to compare and evaluate their differences in predicting pN. The R software was used to construct an individualized Nomogram model of CTR, CEA, and histological types based on multivariate Logistic regression analysis of preoperative risk assessment of pN. The C-index was then calculated, and the calibration degree of the model was plotted.

Results
Clinicopathological characteristics of patients A total of 374 patients met the inclusion criteria, 171(45.7%) male and 203(54.3%) female. The average age was 60 years old. About 76.7% were non-smokers, 294 had no alcohol history, 222 had lung cancer on the right side (59.4%), and 152(40.6%) on the left side, with the right upper lobe being the most common site (32.6%). Postoperative pathological diagnosis revealed that 145 patients had lymph node metastasis with a metastasis rate of 38.8%. Most of the 145 pN(+) patients had no history of smoking or drinking. Preoperative imaging examination showed that the median tumor size and CTR were 2.50 cm (0.70-3.00cm) and 0.90, respectively. Moreover, the histological types were micropapillary type (35.2%), followed by acinus type and papillary type, and EGFR exon 21 mutation was the most common, according for 48.2% (Table 1).

ROC curve differences between CTR, CEA and Ki-67 in predicting pN
MedCalc software was used to compare the ROC curve differences between CTR, CEA, and Ki-67 in predicting pN. The AUC of CEA was not signi cantly different from CTR and Ki-67, indicating that CEA was as accurate as CTR and Ki-67 in predicting pN. However, CTR was statistically different in predicting lymph node metastasis (P=0.03) and had higher accuracy and greater clinical value in predicting pN than Ki-67 ( Fig.2 and Table 4).
Construction and veri cation of Nomogram prediction model R software was used to establish a Nomogram model to predict pN risk based on the multivariate Logistic regression analysis results of CTR, CEA and histological types (Fig.3). The model indicated that the total score of the three risk factors decreased, and the risk of pN increased, with increasing CTR or CEA. The C-index of the predicted model was 0.922 (95% CI 0.896-0.948). The calibration curve revealed that the model had a good calibration degree. Furthermore, the average absolute error between the actual occurrence risk and the predicted occurrence risk was 0.01 (Fig.4).

Discussion
The pathological status of lymph nodes is essential for the formulation of postoperative treatment strategy and prognosis of NSCLC [9]. Some scholars have shown that clinicopathological factors of patients can predict the lymph node metastasis status in early NSCLC [10]. Similarly, in this study, CTR, preoperative serum CEA level, histological type, Ki-67 index and EGFR mutations were key in predicting pathological lymph node metastasis.
Tumor size is a risk factor for pN in clinically node-negative NSCLC [11][12][13]. However, the predictive effect of tumor size on lymph node metastasis in NSCLC≤3cm is unknown. Yu et al. [14] indicated that 53 patients (13.4%) had lymph node metastasis in early NSCLC (diameter, 1-2cm), while no metastasis in tumor 1cm. Zhang [15] reported that lymph node metastasis rates of NSCLC of diameter≤1cm and 1-2cm were 3.8% and 7.4%, respectively. However, in this study, multivariate Logistic regression analysis showed that tumor size was not an independent pN risk factor (P=0.089). Similarly, Haruki [16] showed that larger tumors (diameter 2.0cm, ≤5.0cm) was not associated with mediastinal lymph node metastasis (P=0.158). Lung adenocarcinoma growth is characterized by diameter and solid components increasement. GGO-dominant tumor patients mainly show solid components increasement, indicating that tumor size generally does not affect pN. The high CTR ratio-tumors are mainly characterized by diameter increasement, and tumor size is related to lymph node metastasis. Therefore, another possible reason is that the included patients could be GGO-dominant.
Preoperative serum CEA level might be an indicator of lymph node dissection. Haruki showed that the elevated CEA level in primary lung adenocarcinoma in the lower lobe of both lungs was signi cantly correlated with upper mediastinal lymph node metastasis (P 0.001) [16]. Koike also indicated that CEA was a risk factor for lymph node metastasis in cIA NSCLC [17]. In this study, the AUC, sensitivity, speci city and best cut-off value of CEA were 0.823, 0.786, 0.760 and 2.52 μg/L, respectively, indicating good pN predicting ability. The lymph node metastasis risk increases with increasing CEA levels when CEA 2.52 μg/L preoperatively. Moreover, some scholars have suggested that CEA 5 μg/L belongs to normal range and is not associated with pN. The single-center and small sample study could cause the error. Therefore, the relationship between CEA an pN should be further studied. Several studies have shown that preoperative serum CEA level is a risk factor for N2 lymph node metastasis [18-21] but is not associated with N1 lymph node metastasis [18].
Zhang et al. [18] indicated that tumor consolidation is a risk factor for N1 and N2 lymph node metastasis. Meanwhile, Haruki et al. [16] showed that the mediastinal lymph node metastasis rate of adenocarcinoma is 9.9% (74/744). Besides, hilar and mediastinal lymph node metastasis does not occur in GGO-dominated lung cancer patients. In this study, the ROC curve was used to determine CTR predictive e cacy of lymph node metastasis. CTR was shown to be a preoperative pN predictor, with no statistical difference in predictive power between with CEA (P=0.336). The cut-off value of CTR was calculated (0.775). The lymph node metastasis rate increased when the cut-off value 0.775. In consequence, SMLD should be intraoperatively adopted when HRCT indicates CTR≥0.775 before operation.
In addition, adenocarcinoma in situ and microinvasive adenocarcinoma rarely have lymph node metastasis [22].
However, invasive adenocarcinoma, especially MPA, is associated with a higher lymph node metastasis rate and poor prognosis [11,23]. The analysis of 297 cN0-1 lung adenocarcinoma patients showed that MPA was related with pathological N2 lymph node metastasis, and the rate increased with increasing micropapillary components [24]. In our study, we also found that AIS and MIA were not associated with lymph node metastasis. Besides, the proportion of PPA and MPA was higher in pN(+) than in pN(-). Multivariate analysis indicated that MPA was an independent risk factor for lymph node metastasis. Therefore, lymph node dissection or sampling is not required for AIS and MIA, while SMLD is recommended for invasive adenocarcinoma, especially MPA.
Nomogram is a simple visualization model used to predict diseases based on multiple factors. The predictive model for postoperative anastomotic leakage in colorectal cancer has been widely reported [28,29]. However, it has rarely been reported on lymph node metastasis risk in lung cancer. In our study, a personalized prediction model was constructed using available clinicopathological data to analyze and evaluate the pN risk. Preoperative and intraoperative risk indicators can help to select the appropriate method of lymph node dissection. Postoperative indicators can also guide clinicians to formulate reasonable postoperative adjuvant treatment plans.
However, this study has some limitations: First, it is a retrospective case-control study, and there could be selection bias. Secondly, this is a small sample and single-center study so that a large-sample randomized multicenter prospective study is necessary to provide strong evidence. Finally, this research did not consider postoperative survival and recurrence, which may lead to ignore some lymph node dissection indicators.

Conclusions
In conclusions, 374 cIA lung adenocarcinoma patients were retrospectively analyzed. The pathological lymph node metastasis risk signi cantly increased when CTR≥0.775, CEA 2.52μg/L, rapid freezing pathology examination showed micropapillary type, Ki-67 index≥12.5%, or EGFR exon 19, 21, 19+20 co-mutation occurred. The established Nomogram model can be used for individualized prediction of preoperative lymph node metastasis. Furthermore, the combined Ki-67 index and EGFR mutation status analysis can guide surgical planning and postoperative adjuvant therapy development. However, further studies are needed to explore other potential risk factors for predicting pathological lymph node metastasis. Date availability The data used and/or analyzed during the current study are available by contacting the corresponding author with reasonable request.

Declarations
Compliance with ethical standards Con ict of interest The author declare no con icts of interests.
Consent for publication The suthors declare that they consent to the manuscript publication.   A Nomogram prediction model of pN established based on the CTR,CEA and Histology subtypes of the patients. Each score was added to get the total score and the pN risk