Development of a novel biomarker model for predicting preoperative lymph node metastatic extent in esophageal squamous cell carcinoma1

The number and range of lymph node metastasis (LNM) are critical prognostic factors in esophageal squamous cell carcinoma (ESCC). Preoperative serum biomarkers are reported to be associated with LNM. However, whether these markers can precisely predict the extent of LNM is not known. The aim of this study was to evaluate the predictive value of preoperative serum SCC-Ag, Cyfra21-1, CEA, CA19-9 and CA72-4 for LNM number and range by retrospectively investigating 577 ESCC patients undergone esophagectomy from 2007-2010. In this study, the positive rate of SCC-Ag and CA19-9 were associated with pN stage. Significant differences were found in CEA and CA19-9 between pN0-1 stage patients and pN2-3 stage patients. However, in subgroup analysis (patients with pN0-1), significant difference was found only in SCC-Ag between pN0 and pN1 stage patients (P=0.003). Middle thoracic ESCC patients were Chosen to analyze the correlation between the range of LNM and biomarkers. SCC-Ag was correlated with paraesophageal and paracardial lymph nodes, but not correlated with subcarinal and left gastric artery lymph nodes. Interestingly, the results of CEA were opposite to that of SCC-Ag. CA19-9 was associated with subcarinal and paracardial LNM (P=0.000, P=0.038). Based on the results, a model incorporated SCC-Ag, CEA and CA19-9 was constructed. The rate of patients with pN2-3 stage was 15.4% and 54.4% in group 1 and 4 of our model. In summary, SCC-Ag was associated with early lymph node metastatic stage, and CEA and CA19-9 have a close relationship with advanced lymph node metastatic stage. The model combining SCC-Ag, CEA and CA19-9 might help identify the preoperative extent of LNM for a subgroup of ESCC patients that can be benefited from neoadjuvant therapy.


INTRODUCTION
The prognosis of esophageal squamous cell carcinoma (ESCC) depends on the clinical stage of the primary tumor and the extent of lymph node metastasis (LNM) [1].Lymph node status could be classified by LNM number or stations based on the latest version of the UICC/AJCC TNM classification ( 8 th edition) [2] or Japanese Classification of Esophageal Cancer ( 11 th edition) [3].These two LNM classification methods define the extent of LNM by analyzing the metastatic number and range.
Serum biomarkers are biologic or biochemical substances that are produced by tumor cells, and are commonly used for identifying cancer, indicating treatment outcomes and predicting prognosis.In ESCC, it has been reported that preoperative biomarkers have a close relationship with tumor burden, including tumor invasion, LNM and the survival rate [4,5].But there has been no studies focusing on the correlation between biomarkers and metastatic extent of lymph node which was more valuable for clinicians to evaluate operative indication.
In the present study, (1) we clarify which preoperative serum levels of the biomarkers were associated with LNM status.(2) SCC-Ag, Cyrfra21-1, CEA and CA19-9, which showed a significant correlation with LNM, were selected to evaluate the relationship with lymph node metastatic number and stations.(3) We analyzed the diagnostic values of the model incorporated preoperative serum SCC-Ag, CEA and CA19-9 for lymph node metastatic extent in ESCC.

Correlations between preoperative serum biomarkers and LNM number
SCC-Ag, Cyfra21-1, CEA and CA19-9 were chosen to analyze the correlations with pN stage.The positive rate of serum SCC-Ag and CA19-9 were associated with pN stage (Table 3; P=0.010 and P=0.004).To further examine the relationship between SCC-Ag, Cyfra21-1, CEA and CA19-9 and the LNM extent, patients with pN0 and pN1were selected (477 patients in total).In this subpopulation, the positive rate of SCC-Ag was much higher in pN1 stage patients compared to those in pN0 stage patients (23.9% vs. 13.2%,P=0.003), but no correlations were found in Cyfra21-1, CEA and CA19-9 (all P>0.05).(Table 4).
For analyzing the advanced stage of LNM, patients were divided into two subgroups, pN0-1 and pN2-3.Patients with pN0-1stage were considered earlier and less extensive LNM, while patients with pN2-3 stage were considered more extensive LNM.High positive rate of preoperative serum CEA and CA199 were associated with more extensive LNM (P=0.039,P=0.000, respectively).However, no correlation was found in SCC-Ag (P=0.105)(Table 5).

Correlations between preoperative serum biomarkers and LNM groups
Japanese Classification of Esophageal Cancer (11th edition) [3] graded pN stage(JES) based on the LNM groups which were defined according to the location of the tumor.For middle thoracic ESCC, middle thoracic paraesophageal lymph nodes and paracardial lymph nodes were involved in pN1 (JES), subcarinal lymph nodes and left gastric artery lymph nodes were involved into pN2 (JES).pN2 (JES) stage was more extensive lymphatic invasion than pN1 (JES) stage.
In middle thoracic ESCC, the positive rate of SCC-Ag was correlated with paraesophageal and paracardial lymph nodes in pN1 (JES) (all P<0.05), but not correlated

A biomarkers diagnostic model and risk groups to LNM
To better identify ESCC patients at high risk for LNM, we proposed a new diagnostic model by combining SCC-Ag, CEA and CA19-9, and stratified patients into four groups.Grade 1 group: negative SCC-Ag, CEA and CA19-9; Grade 2 group: positive SCC-Ag, negative CEA and negative CA19-9; Grade 3 group: negative SCC-Ag with positive CEA or positive CA19-9 or both; Grade 4 group: positive SCC-Ag with positive CEA or positive CA19-9 or both.In the patients reported here, there are 422, 91, 41 and 11 patients in the grade 1, 2, 3 and 4 groups, respectively.With this model, the analysis showed the rate of patients with pN2-3 stage in ESCC was 15.4% in grade 1 group.The rate of LNM patients was 81.8% and the rate of patients with pN2 -3 stage was 54.5% in grade 4 group.ROC curve analyses were performed, the area under the ROC curve (AUC) in predicting LNM with this model was 0.567 (P=0.006).The AUC in predicting LNM extent with this model was 0.563 (P=0.048)(Supplementary Figure 1) (Table 7).

DISCUSSION
To our knowledge, our report is the largest published retrospective study to date that analyzes the correlation between preoperative serum biomarkers (SCC-Ag, Cyfra21-1, CEA, CA19-9 and CA72-4) and LNM in ESCC.More importantly, there have been no studies investigating the predictive values of these preoperative serum biomarkers to the extent of LNM.The present study found that the serum SCC-Ag, CEA and CA19-9 were closely associated with lymph node metastatic extent.Different serum biomarkers revealed different stages of LNM.The model we built will help predict LNM and further define the precise status of pN stage (N0-1 or N2-3).
ESCC has a high prevalence in Asia and easily metastasizes to lymph nodes.Neoadjuvant therapy, especially preoperative chemoradiation therapy, could downstage the tumor, significantly reduce the 3-year mortality and locoregional recurrence compared with surgery alone [15][16][17].The clinical value of preoperative lymph node status identification is to decide whether patients should undergo neoadjuvant therapy first.Diagnostic imaging techniques, including computed tomography and endosonography, have limitation that cannot predict lymph node status sufficiently.Preoperative serum biomarkers, which are measured routinely prior to treatment, could provide information of LNM based on our study presented here.

Characteristics
No Larger number or range of LNM stands for more extensive metastasis and later stage.Among the 4 biomarkers, only high positive rate of CEA and CA19-9 were demonstrated to associate with advanced stage LNM.There was no significant difference for SCC-Ag (P=0.140).These results showed SCC-Ag would be elevated at early stage of LNM and could predict early LNM status.On the contrary, CEA and CA19-9 could predict advanced LNM status.
It should be noted that our purpose to analyze the correlation between lymph node stations and biomarkers was not to predict the specific lymph node station metastatic status with markers.In Japanese Classification of Esophageal Cancer, some specific lymph node stations stand for more extensive metastasis.To confirm our thoughts above, we chose middle thoracic esophageal carcinoma patients and analyzed the status of middle thoracic paraesophageal lymph nodes (pN1, JES) and paracardial lymph nodes (pN1, JES), subcarinal lymph nodes (pN2, JES), and left gastric artery lymph nodes (pN2, JES).Interestingly, SCC-Ag was significantly associated with pN1 (JES) stage LNM (P<0.05), but not pN2 (JES) stage (P>0.05).Opposite relationships with lymph node stations were found in CEA.Similar results of CA19-9 were verified in the thoracic lymph node stations.In abdominal lymph node stations, CA19-9 was correlated with paracardial LNM.A probable reason for this discrepancy may be explained as follows.Wang et al. [26]argued that most patients had abdominal LNM accompanied by thoracic (mediastinal) LNM at the same time and the 5-year survival rate was lower than that of patients with thoracic LNM.CA19-9 could not indentify well in abdominal LNM whose lymph metastasis extent was more severe compared with thoracic LNM.The positive rate of CEA in ESCC ranges from 9.1% to 23% [4,18,27].CEA has been reported to own the function of an adhesion molecule.CEA-containing tumor cells could preferentially go into metastatic cascade [28,29].Kosugi et al. [4] reported that CEA had no correlation with LNM, but with distant metastasis in ESCC.However, another study revealed significant differences were found between CEA and LNM in ESCC [18].CA19-9 is also called sialylated Lewis Antigen which is speculated to play roles in the extravasation of cancer cells from blood and the promotion of metastatic spread to distant organs [30].CA19-9 was widely used in pancreatic cancer, colon cancer, gastric cancer, and other gastrointestinal tumors [11,12,[31][32][33].In a meta-analysis of gastric cancer, there were the correlation between elevated CA19-9 and LNM, T stage, TNM stage, vessel invasion and poor survival Several limitations to our study exist.Since the analyses were performed retrospectively on a single institutional database, selection biases might be underestimated.Additional validation of the models using different data sets might further prove the clinical value of the model.
In conclusion, this study demonstrates the correlation between LNM and the positive rate of preoperative serum SCC-Ag, Cyfra21-1, CEA and CA19-9.SCC-Ag was associated with early lymph node metastatic stage.CEA and CA19-9 have a close relationship with advanced lymph node metastatic stage.The model incorporated SCC-Ag, CEA and CA19-9 could further increase the predictive value for the LNM status of esophageal squamous cell carcinoma.

Patients and eligibility criteria
Among patients with esophageal squamous cell carcinoma undergone esophagectomy in Cancer Institute and Hospital of Tianjin Medical University from January 2007 to December 2010, the study was conducted on 577 patients whose preoperative serum SCC-Ag, Cyfra21-1, CEA, CA19-9 and CA72-4 were examined.The cases exclude from the current study based on the following criteria: (1) patients who had history of malignant disease; (2) patients who had received preoperative treatment (chemotherapy and/or radiotherapy); (3) patients who had other malignant tumors except esophageal squamous carcinoma.
The serum Cyfra21-1, CEA, CA19-9 and CA72-4 were detected within 7 days before surgery with electrochemiluminescence immunoassay (Roche Elecsys E170, Germany).The SCC-Ag was detected by chemical luminescence immunoassay (Abbott Architect i2000, America).The normal upper limit was 1.5 ug/L, 3.3 ug/L, 5 ug/L, 39 U/ml and 6 U/ml for SCC-Ag, Cyfra21-1, CEA, CA19-9 and CA72-4 respectively.The esophagectomy with 2-to 3-field lymph node dissection was taken for standard treatment.The depth of primary tumor, degree of lymph node (pN stage) and TNM staging were defined according to UICC/AJCC TNM classification (7th edition) [14].For analyzing the relationship between lymph node metastatic range and biomarkers, the LNM grading method of Japanese Classification of Esophageal Cancer (11th edition) [3] also were used and defined as "pN stage (JES)".JES is the abbreviation for Japanese Esophageal Society.In present study, we defined pN1 and pN2-3 (UICC/AJCC TNM classification or Japanese Classification of Esophageal Cancer) as early stage and advanced stage LNM to analyze the correlation between biomarkers and LNM extent.

Statistical analysis
The statistical analyses were performed with the SPSS 21.0 software (ver.21 SPSS Inc., Chicago, IL).The correlations between serum level of biomarkers and clinicopathological characteristics were assessed using the Chi-square or Mann-Whitney-U test.Significant risk factors identified by univariate analysis were further assessed by multivariate analysis using logistic regression.A receiver operating characteristic (ROC) curve was constructed.P < 0.05 was considered statistically significant.

Table 1 : Clinical and biological characteristics of 577 patients in ESCC
(JES).Interestingly, the association between the positive rate of CEA and LNM groups were opposite to that in SCC-Ag.Significant results of CEA were found in pN2 group (JES) (all P<0.05) and not in pN1 group (JES).The positive rate of serum CA19-9 was associated with subcarinal and paracardial LNM (P=0.000,P=0.038)(Table6).