High‐risk group of upper and middle mediastinal lymph node metastasis in patients with esophagogastric junction carcinoma

Abstract Aim The aim of the present study was to clarify esophagogastric junction (EGJ) carcinoma patients who are at high risk of upper and middle mediastinal lymph node (MLN) metastasis. Methods This was a retrospective study and included 110 consecutive patients with EGJ carcinoma who underwent R0/R1 resection at Keio University Hospital between January 2000 and December 2013. Results Of the 110 patients, 18 (16.3%) had MLN metastasis, and the number increased to 23 (20.9%) when recurrence cases were added (adenocarcinoma, N = 11; squamous cell carcinoma, N = 12). Patients whose tumor epicenter was located above the EGJ had a significantly higher incidence of MLN metastasis/recurrence (18/51 [35.3%]) than those whose tumor epicenter was located below the EGJ (5/59 [8.5%]). The MLN metastasis/recurrence rate was particularly high when the distance from the EGJ to the proximal edge of the primary tumor was >3 cm for the upper and middle mediastinum (18.8%). Patients in a selected group (≥T2 and tumor epicenter located above the EGJ or below the EGJ with ≥3 cm esophageal invasion) showed 17.9% and 15.4% upper and middle MLN metastasis/recurrence rates, respectively. Therapeutic value of MLN dissection was relatively high (#105 + 106: 8.9, #110: 12.2). Conclusions Therapeutic value of MLN dissection to treat EGJ carcinomas was relatively high in patients with MLN metastasis. Our algorithm could select patients at high risk for MLN metastasis.

Usually, a transthoracic approach is carried out for patients with Siewert type I tumors, whereas the transhiatal approach is used for those with Siewert type III tumors. Both approaches have been used for Siewert type II tumors. A Dutch trial compared the right thoracic and transhiatal approaches for Siewert tumor types I and II. 4 Although the survival rate of patients with Siewert type II tumors was not different between the groups, a subgroup analysis of nodal metastatic patients showed that survival in patients who underwent the transthoracic approach was significantly better than that of patients who underwent the transhiatal approach. 4 Squamous cell carcinoma (SCC) of the EGJ is often observed in Asian populations.
Therefore, Nishi's classification of EGJ carcinoma has been used in Japan. 15 This definition describes EGJ carcinoma as a tumor whose tumor epicenter is located within a 2-cm area above and below the EGJ, regardless of histological type. Although this classification includes both histological types, most patients with AC have received a resection of the lower esophagus and stomach, whereas those with SCC have undergone subtotal esophagectomy, reflecting differences in the dominant tumor location.
Siewert et al 16 showed that R0 resection was a prognostic factor in patients with EGJ carcinoma. Thus, appropriate prophylactic mediastinal lymph node (MLN) dissection is necessary for EGJ carcinoma.
We usually use two major surgical approaches for EGJ carcinoma to retrieve MLN: the transthoracic approach and the transhiatal approach. The transthoracic approach is a more invasive procedure, which may increase morbidity and worsen the patient's quality of life after surgery. Although lymph nodes in the lower mediastinal region can be retrieved using both approaches, lymph node dissection of the upper-middle mediastinum can only be carried out with the transthoracic approach. However, the optimal surgical approach for patients with EGJ carcinoma remains unclear. In the present study, we aimed to explore high-risk EGJ carcinoma patients who have upper and middle MLN metastasis/recurrence and underwent the transthoracic approach. We retrospectively collected the medical records of patients with EGJ carcinoma (diagnosed according to Nishi's classification 15 ) who underwent surgical resection.

| Patients
This was a retrospective study and included 110 consecutive patients diagnosed with EGJ carcinoma who subsequently underwent curative surgical resection at Keio University Hospital (Tokyo, Japan) between January 2000 and December 2013. In an additional study, 22 EGJ carcinoma patients who were treated with surgical resection between January 2014 and November 2016 were included to prove the suitability of our algorithm, which was created according to the results of our retrospective study. EGJ carcinoma was defined according to Nishi's classification. 15 EGJ was identified from the resected specimen and was defined using the level of macroscopic caliber change. Tumors with epicenters located in the area of the EGJ, extending from 2 cm above to 2 cm below the EGJ, were designated as EGJ carcinomas.

| Follow up
For postoperative follow up, we usually carry out computed tomography (CT) scan every 6 months and endoscopic examination every year after operation.

| Evaluation and lymph node station
The 7th edition of the Union for International Cancer Control tumornode-metastasis classification of esophageal cancer was used for tumor staging, 17 and the Japanese classification of esophageal carcinoma was used to number the lymph node stations. 15

| Therapeutic value of lymph node dissection
To evaluate the therapeutic value of dissecting each lymph node station, we used the modified method presented in 1995 by Sasako et al 18 who used the therapeutic value index. The therapeutic value of nodal dissection was based on multiplication of the lymph node metastasis rate and the 3-year survival rate in patients with lymph node metastases (as a percentage). The rate of lymph node metastasis was calculated by multiplying the number of patients with lymph node metastases for each station and the number of those in whom that station was retrieved. The 3-year overall survival rates in patients with lymph node metastasis were calculated for each nodal station, regardless of lymph node metastasis for other stations.

| Statistical analysis
Statistical analysis was carried out using SPSS statistical software (ver. 23; SPSS Inc., Chicago, IL, USA). Clinical and pathological variables were analyzed using Pearson's chi-squared and Mann-Whitney U-tests. Multivariate logistic regression analysis was carried out to identify the risk factors for MLN metastasis. P-value <0.05 was considered significant.

| Decision-making for operative procedures
Before 2014, when a surgical algorithm for EGJ carcinoma was published in the Japanese Gastric Cancer Guideline 4th edition, 19 clear algorithms in guidelines were lacking and surgery was carried out at the discretion of the operator or based on decisions made at a conference. Subtotal esophagectomy was the first choice of treatment for EGJ SCC as lower esophageal carcinomas. However, depending on the general condition of the patient, lower esophageal resection by a transhiatal approach was also carried out. Before 2004, the left transthoracic approach was selected when mediastinal anastomosis using the transhiatal approach was thought to be difficult. For AC, when CT showed mediastinal lymphadenopathy, lymph node dissection and esophagectomy were both carried out. The transhiatal approach was used in cases lacking indications for upper and/or middle MLN metastasis. However, even in such cases, a transthoracic approach was preferred when the anastomotic position was high and transhiatal anastomosis was expected to be difficult. After the guidelines were published, the procedures were basically determined according to the guidelines.  Table 1. A total of 110 patients were enrolled, consisting of 84 with AC and 26 with SCC. Although no significant differences were identified in terms of age, gender, or adjuvant therapy between the two groups, the surgical procedures and neoadjuvant therapy were found to vary. The transhiatal approach was mainly carried out in AC patients (80.9%), whereas only 11.5% of patients with SCC underwent this procedure.
Subtotal esophagectomy was primarily used in patients with SCC (80.8%), as compared to 14.3% of patients with AC. Total gastrectomy (40.4%) and proximal gastrectomy (39.2%) were mainly used in AC patients, whereas partial gastrectomy for making a gastric tube was common in patients with SCC (80.8%). Length from tumor center to EGJ was significantly different between the AC and the SCC groups (AC: 5.7 ± 9.5 mm vs SCC: -9.9 ± 8.7 mm; P < 0.001). Patients with SCC had significantly higher tumor center locations and oral tumor borders, as compared to AC patients

| Therapeutic value of lymph node dissection
Metastasis to MLN was detected in seven patients with AC (8.3%) and in eleven patients with SCC (42.3%). There were no patients who had lymph node metastasis to the pyloric region. The therapeutic value of extended nodal dissection of each lymph node station is shown in Table 2 Table 2.

| Mediastinal lymph node metastasis and recurrence
Mediastinal lymph node metastasis was observed in 18 patients, and the number increased to 23 when the recurrence cases were added (AC, N = 11; SCC, N = 12). Frequency of MLN metastasis or recurrence was higher in patients with SCC, as compared to those with AC (46.2% and 13.1%, respectively). Background characteristics and pathological findings of patients with and without MLN metastasis/ recurrences are shown in Table 3. Of the 23 patients, rates for MLN metastasis were significantly higher when the tumor epicenter was located above the EGJ (18/51: 35.3%), as compared to when the epicenter was located below the EGJ (5/59: 8.5%). In these patients, two pT1 cases presented MLN metastasis, and the tumor epicenter was located above the EGJ in both cases. In contrast, five patients whose tumor epicenter was located below the EGJ and who had MLN metastasis/recurrence were diagnosed with advanced carcinoma (pT2/T3/T4: 1/3/1).
Of the total of 110 patients, 12 were found to have upper-middle MLN metastasis/recurrence. In these patients, eight were AC and four were SCC, and they all had ≥pT2 tumor (pT2/T3/T4: 1/8/3). No significant difference was identified between the two histological types (Table 4). Among all patients, MLN metastasis and recurrence rates were found to be high when the distance from the EGJ to the proximal edge of the primary tumor was >3 cm for upper and middle mediastinum (18.8%) and >2 cm for lower mediastinum (15.0%) (

| DISCUSSION
In the present study, we created a surgical algorithm for the treat-  As the first two steps of our algorithm (location of the tumor epicenter and early/advanced carcinoma) are simple, the effect of the errors on surgical decision-making was small.
Third, in terms of the preoperative length of esophageal invasion, the precise length (in cm) was not noted prior to surgery in some cases, especially older cases; the correctness (or otherwise) of length evaluation is thus unknown.
In addition, in terms of the preoperative diagnostic accuracy of MLN metastasis, six cases showed CT findings of MLN metastasis prior to surgery, of whom five were ultimately found to have mediastinal metastases. The other 13 cases showed no CT findings of MLN metastasis, but such metastases were discovered on pathological evaluation. CT sensitivity for diagnosis of MLN metastasis was 27.8% (5/18) and the specificity was 98.9% (91/92) (Table S1). This means that patients with enlarged lymph nodes (thus clinically positive) were at high risk of metastasis, but clinically negative cases might nonetheless have metastases. Thus, estimation of the risk of MLN metastasis by reference to tumor location and depth, and the length of esophageal invasion, is necessary.
All deaths except two were caused by the original disease.
The two exceptions were deaths caused by postoperative complications (pneumonia). One patient had an AC and the other a SCC.
The prognosis did not differ significantly between the SCC and AC groups. However, SCC was associated with a lower 3-year survival rate, and the survival curve trended downward, reflecting the histological malignancy of SCC. However, if a tumor was present in the same location, the initial lymphatic ducts invaded