Patients with Adenocarcinoma of The Esophagogastric Junction and Upper Gastric Cancer

Xin Yin Harbin Medical University Third Clinical College: Tumor Hospital of Harbin Medical University Tianyi Fang Harbin Medical University Third Clinical College: Tumor Hospital of Harbin Medical University Xuan Lin Harbin Medical University Second A liated Hospital Department of Cardiology Xiliang Cong Harbin Medical University Third Clinical College: Tumor Hospital of Harbin Medical University Yimin Wang Harbin Medical University Third Clinical College: Tumor Hospital of Harbin Medical University Chunfeng Li Harbin Medical University Third Clinical College: Tumor Hospital of Harbin Medical University Yan Ma Harbin Medical University Third Clinical College: Tumor Hospital of Harbin Medical University Yingwei Xue (  xueyingwei@hrbmu.edu.cn ) Tumor Hospital of Harbin Medical University https://orcid.org/0000-0002-8427-9736


Introduction
Gastric cancer (GC) is still the third leading cause of cancer mortality worldwide [1]. With the increasing awareness of malignant tumor prevention in countries with high incidence, such as Japan, Korea and China, GC has been decreasing. However, the incidence of gastroesophageal junction cancer (AEG) and upper gastric cancer (UGC) has increased annually, especially in Japan, where it has increased from 2.3% to 10% in the past 40 years [2]. For AEG, according to Siewert type, 93% of type I patients are mainly treated by esophageal and surrounding lymph node dissection, and >66% of Siewert type II and 90% of Siewert type III patients are mainly treated by radical gastrectomy [3][4][5][6]. AEG and UGC have the clinical features of strong invasion, poor prognosis, advanced clinical stage and high rate of postoperative recurrence [7,8]. Regarding the surgical method, there is debate about total gastrectomy (TG) and proximal gastrectomy (PG). Golematis et al. [9] found that TG can ensure a su cient distal margin and extended lymph node dissection, which can brings survival bene t. However, Harrison et al. found that patients who underwent TG and PG had no difference in overall survival (OS) rate, although PG led to better postoperative nutritional status [10]. Therefore, in addition to tumor location, size and clinical stage, it is important to nd suitable clinical prognostic factors to help surgeons choose a suitable method of gastrectomy as well as evaluate prognosis after surgery.
Researchers have proposed use of the American Joint Committee on Cancer (AJCC) eighth edition for treatment guidelines for AEG. It is recommended that the distance from the tumor center to the gastroesophageal junction should be >2 cm and the gastric cardia should not be injured, as in GC treatment. However, these recommendations do not entirely meet the clinical requirements [11]. We found that the current clinical guidelines are only based on assessment of the tumor's general progression and do not consider the immune response caused by tumor cells. In 2014, Galon et al. [12] rst proposed pTNM-I, which combines the immune response in the tumor microenvironment and pTNM stage. In 2018 [13], pTNM-I staging was used to guide postoperative chemotherapy in patients with colon cancer. We consider that tumor immunity is also important in assessing tumor progression. However, the high degree of heterogeneity makes it di cult for pathologists to assess the immune status of patients individually and the randomness of eld selection of para n sections also limits the study of pTNM-I in GC. Reichert et al. [14] found that changes in the peripheral immune microenvironment of the tumor can be re ected by circulating immune-related cells such as neutrophils (N), platelets (P), monocytes (M) and lymphocytes (L). Systemic immune in ammation score (SII) [15][16][17], neutrophil-lymphocyte ratio (NLR) [18,19], platelet-lymphocyte ratio (PLR) [20], lymphocyte-monocyte ratio (LMR) [21], and scoring systems that combine with in ammation index, such as C-reactive protein (CRP)-NLR [22], NLR-PLR [23], Fibrinogen (F)-NLR [24] and SIRI-PLR [25] have been con rmed to evaluate accurately the prognosis of GC and other malignant tumors. Therefore, as a new comprehensive in ammatory index scoring system combining lymphocytes, neutrophils, monocytes and platelets, the systemic in ammatory response index (SIRI) and platelet-lymphocyte ratio (PLR) are expected to predict the prognosis of GC more accurately, and even help clinicians choose appropriate operation methods for AEG and UGC.
In this study, 371 patients with AEG and UGC who underwent radical surgery at the Cancer Hospital of Harbin Medical University were randomly selected consecutively. The relationship between SIRI-PLR and pathological factors was investigated by cohort study to explore the clinical signi cance of SIRI-PLR scoring.

Patients
Tumors that were mainly located in the upper third of the stomach were designated UGC. Tumors that were mainly located at 2-5 cm from the gastroesophageal junction were designated Siewert type GC Total patients were divided into two independent cohorts according to admission time, 194 patients from 2003 to 2010 were the training set, and 177 patients from 2011 to 2014 were the validation set. Over the all known died patients, 186 patients died due to GC and 21 patients (11 patients in the training set, 10 patients in the validation set) died from other causes (7 deaths because of heart disease, 4 patients died naturally, 2 patients died from accidents and 8 deaths from unknown causes).
Operation methods and postoperative chemotherapy standards are based on The Japanese gastric cancer treatment guidelines [27]. The range of gastrectomy is determined according to the patient's clinical stage and tumor location. TG is for clinically node-positive (cN+) or T2-T4a. PG is for the patients with acceptable proximal resection margin (>5cm or frozen section examination of the resection line is desirable). The invasion of the pancreas by tumors requiring pancreaticosplenectomy must also be performed by TG, regardless of the location of the tumor and distance of the surgical margin. TG includes all gastric tissues, including pylorus and cardia, combined with D2 lymph node dissection range (No. [1][2][3][4][5][6][7]9,10,11p,11d,12a). Digestive tract reconstruction methods are Roux-en-Y esophagojejunostomy, Jejunal interposition and Double tract. PG involves proximal 2/3 of the gastric tissues, with gastroesophageal junction and pylorus retained, in combination with the D1+ lymph node dissection area (No. 1,2,3a,4sa,4sb,7 and No. 8a,9,11p,and No.110 dependent on the need for surgery), and the methods of digestive tract reconstruction include Esophagogastrostomy, Jejunal interposition and Double tract method. There were 38 patients with T4b, of which 7 patients had intragastric resection of adjacent organs (2 patients with partial liver resection, 1 with transverse colon, 2 with spleen, and 2 with pancreatic tail). In addition, oxaliplatin+capecitabine (XELOX) or oxaliplatin+S-1 (SOX) are the main treatment options for patients with postoperative pathological stages II-III, which were 336 patients in the study. In order to ensure the accuracy of this study, we included all patients in our institution with a total of 117 patients. The remaining 219 patients were not included in the postoperative chemotherapy patient group. This is because these patients did not complete all postoperative chemotherapy regimens in our institution. Most of the patients returned to the local hospital for treatment after surgery and did not have complete chemotherapy records.

Statistical analysis
Overall survival (OS) was calculated as the time from surgery to death from any cause. If patients were alive at the last follow-up, they were censored. Patients survival time in each group were shown as median±standard deviation. We used R software version 3.6.1 and the 'survivalROC' package to investigate the prognostic or predictive accuracy of each variable by time-dependent receiver operating characteristic (ROC) analysis. An optimal cutoff value was de ned to classify the patients into two groups (high vs low) for each variable with use of the receiver operating characteristic curve for survival in 5-year, and the maximum value of sensitivity -(1speci city) in 'Youden index' is the best cutoff value. Delong nonparametric method was used to estimate the AUC con dence interval. Kaplan-Meier method and Log-rank test were used to analysis survival curves. Median follow up was calculated by the reverse Kaplan-Meier analysis. The chi-square test was used to analyze the association between SIRI-PLR and patients characteristics. Univariate and multivariate analyses based on the competitive risk regression model were used to analyze independent predictors for high risk of death due to AGC and UGC. The R software was used to construct the Nomogram model of risk assessment by 'SvyNom' and 'rms' packages. Standardized Hazard Ratio and 95% CIs were estimated for each factor. SPSS version 22.0 (SPSS Inc., Chicago, IL, USA) was used for analysis and two-tailed P values <0.05 were considered statistically signi cant.

SIRI-PLR
The time-dependent ROC curve shows that these variables were continuously keep satisfactory signi cance (Fig.1 a). For the SIRI and PLR, 0.82 and 134.62 were the cutoff value. The area under the curve (AUC) was 0.677 [95% con dence interval (CI): 0.602-0.752] and 0.678 (95% CI: 0.602-0.754). Similarly, the optimal cutoff values of SII and NLR were 464.23 and 2.46 by ROC curve analysis ( Fig.1 b-d). And patients with SIRI ≤0.82 and PLR ≤134.62 were in the score 0 group, patients with SIRI ≤0.82 and PLR >134.62 and SIRI >0.82 and PLR ≤134.62 were in the score 1 group, and patients with SIRI >0.82 and PLR >134.62 were in the score 2 group.

ROC curve of in ammation index
In the training set patients, we compared the SIRI-PLR, NLR and SII by ROC. The relationship between SIRI-PLR score and clinical and pathological factors is shown in Table 3. In the training set, SIRI-PLR score had a signi cant association with tumor size, surgical procedure, NLR, SII and pTNM stage (P=0.004, P=0.014, P=0.002, P<0.001 and P=0.037). In the validation set, SIRI-PLR had a signi cant association with tumor size, NLR and SII (P=0.003, P<0.001 and P<0.001).

SIRI-PLR and postoperative chemotherapy and surgical method
In all patients. There was no signi cant difference between patients with and without postoperative chemotherapy in the score 0 and 1 groups (P=0.958), and patients without postoperative chemotherapy had shorter survival than patients with postoperative chemotherapy in the score 2 group (P=0.002) (Fig.4 a and b). In the score 0 and 1 groups, patients with and without postoperative chemotherapy had the survival times of 60.00±18.62 and 60.00±21.23 months, respectively, and 5-year survival rates were 53.6% and 51.1%, respectively. In the score 2 group, the survival times were 37.57±18.80 and 16.02±19.23 months, respectively, and 5-year survival rates were 33.3% and 14.9%, respectively.
There was also no signi cant difference between patients with PG and TG in the low PLR, low SIRI and high SIRI patients (P=0.271, P=0.271 and P=0.260), and patients with TG had shorter survival than patients with PG in the high PLR group (P=0.049) (Fig.4 c-f). In the high PLR group, patients with PG and TG had the survival times of 33.52±21.07 and 18.40±21.43 months, respectively, and 5-year survival rates were 36.4% and 25.0%, respectively.

Univariate and multivariate regression analyses
To identify the independent predictors for OS of patients with AGC and UGC, univariate and multivariate analyses based on competitive risk regression model in the training set. According to univariate analysis, age (P=0.001), NLR (P<0.001), SIRI (P<0.001), PLR (P<0.001) and pTNM stage (P=0.048) were signi cant. According to multivariate analyses, age (P=0.010), SIRI (P=0.036), PLR (P=0.045) and pTNM stage (P=0.001) were independent predictors for high risk of death due to AGC and UGC (Table 4).

Prognostic Nomogram for OS
Because age, SIRI, PLR and pTNM stage are independent predictors for patients with AGC and UGC in the training set, we rst combined these clinical features in the training set and constructed a Nomogram model of continuous variable data to predict the 3-year and 5-year prognosis (Fig.5 a). Concordance was 0.710 and standard error was 0.017. Calibration curves for predicting survival at 3 and 5-year are shown in Fig.5 b and c. At the same time, all patients were scored and ROC analysis was performed. The AUC were 0.767 [95% con dence interval (CI): 0.699-0.835] (Fig.5 d) and 0.754 [95% con dence interval (CI): 0.687-0.822], respectively (Fig.5 f). The sensitivity were 88.8% and 84.8%, respectively, and the speci city were 57.9% and 60.0%, respectively.

Discussion
AEG and UGC have shown an increasing ratio in all GC patients annually, and >70% of them are diagnosed with advanced GC [28], which has the characteristics of high rate of lymph node metastasis, poor prognosis and worse survival time. Although the 8th edition of the AJCC has classi ed the tumors in detail based on tumor location, there is debate in gastrectomy, and suitable clinical prognostic factors that can help surgeons choose the appropriate treatment for individual patients are still needed. With recognition of the role of systemic in ammation in promoting tumor growth, progression and metastasis of malignant tumors, cytokines, in ammatory proteins and immune cells in the systemic in ammatory response are also considered as potential clinical markers to guide treatment decisions [29,30]. Hu et al. [31] found that high-sensitivity C-reactive protein in non-obese radiotherapy patients with malignant tumors could predict grade 4 skin toxicity, and advise doctors to change treatment options. Recently, the prognostic importance of SIRI for patients with renal, pancreatic and nasopharyngeal cancers has been con rmed [25,32-36]. The higher the SIRI score is, the shorter the disease-free survival and OS. SIRI score can also guide appropriate treatment. For patients undergoing thoracoscopy for lung cancer, SIRI >0.99 has the greatest survival bene t, and for those receiving mFOLFIRINOX chemotherapy in pancreatic cancer, SIRI ≥2.3 has the greatest survival bene t. Studies of upper tract urothelial carcinoma also have shown that combination of SIRI and PLR scoring evaluates prognosis more effectively, hence, evaluation of SIRI-PLR for other malignant tumors has developed.
In this study, the AUC of SIRI and PLR curve are 0.677 and 0.678, respectively, and the cut-off values are 0.82 and 134.62, respectively, which is similar to the research by Zheng et al. [25]. But we found the AUC of PLR in our study is higher than that reported by Zhang et al. [37]. According to chi-square analysis between SIRI-PLR with clinical and pathological features it could be rstly found that it was correlated with tumor size, NLR and SII. Many previous studies con rmed that NLR and SII could be used as in ammatory indexes for evaluating prognosis of GC patients, which means that SIRI-PLR is expected to become a more signi cant clinical biomarker. In our study, ROC analysis of SIRI-PLR, NLR and SII showed that the AUC of SIRI-PLR was not only larger than that of NLR and SII, but also had better speci city. We can conclude that the combination of SIRI and PLR is better to other in ammatory indexes in predicting the prognosis, which was consistent with the study by Li et al. The tumor immune microenvironment of GC patients can be re ected by postoperative pathological tissue sections or immune indexes. Postoperative pathological tissue sections can be used to analyze the condition of the tumor immune microenvironment, which can be individually evaluated precisely by immunohistochemistry. In this way, it had high speci city. However, the randomness of tissue site selection and development of preoperative adjuvant therapy reduce the accuracy of the tests. On the other hand, the immune indexes constructed by circulating immune cells were calculated based on preoperative hematological tests, whose clinical application is convenient and can re ect the in ammatory condition of the body. The immune indexes were obtained by retrospective analysis of clinical data from a large number of patients, although the speci city was low and couldn't re ect the immune environment of individual patients. However, immune index accuracy was high, because it was in line with the in ammatory condition in most patients. During treatment, hematological examination also was routine texts for GC patients before surgery, and the samples would be sent to hematology laboratory immediately after be collected. Soon, doctors would get reports. The immune index could be calculated by the corresponding mathematical model easily based on these reports. So, it is easy for clinical application [41,42].
The SIRI-PLR scoring system included circulatory neutrophils, platelets, monocytes and lymphocytes, which comprehensively covered the circulating immune cells for calculation of the in ammatory index. This enabled SIRI-PLR to evaluate more comprehensively the physical in ammatory status of patients than NLR and SII did. The study of tumor-related circulating immune cells and in ammation as important parts of the tumor microenvironment plays an important role in measuring tumor progression. GC cells can induce enrichment of neutrophils in tumors, especially GC positive for programmed death ligand 1 and correlative with Epstein-Barr virus [43][44][45]. Formation of neutrophils and secretion of interferon-γ and tumor necrosis factor-α can inhibit proliferation of lymphocytes and biological activity of CD4+ and CD8+ T cells, lead to immune escape of cancer cells and distant micro-metastasis of tumor cells. Nie et al. [46] showed that expression of a large number of monocytes can inhibit the immune response of T cells near the tumor and promote immune escape of cancer cells by increasing cyclo-oxygenase-2 expression. Senescence-associated secretory phenotype interaction between platelets also plays an important role in metastasis and invasion of cancer cells [47]. Tumor occurrence, development and metastasis lead to changes in immune cells and in ammation in the tumor microenvironment, which would affect the circulating immune cells with disease progression. Changes in immune cells can be indirectly detected by rapid hematological methods, and SIRI-PLR can be calculated to evaluate immune status and prognosis of patients.
As we all know, choosing appropriate individualized treatment for each patient can greatly improve the quality of life even survival time. Surgery is still generally the rst choice intervention in patients with GC [48]. At present, clinicians mainly choose TG and PG methods for gastrectomy according to tumor size, tumor location, clinical stage and surgical experience. However, there is still a part of GC which is consistent with both surgical indications of TG and PG. Moreover, the different surgical methods of gastrectomy affect the quality of life and survival time after operation. This has led to a controversy about how to choose the operation method for such patient. Postoperative body mass index, albumin and nutrition index of patients treated with PG are higher than in those treated with TG, but there is no difference in postoperative nutritional quality [49,50]. PG patients have a high risk of gastroesophageal re ux, which may be related to the high rate of tumor recurrence. Golematis et al. [9] found that TG can ensure a su cient distal surgical margin, extended lymph node dissection and dissection of tissues of organs surrounding the stomach, which can reduce the risk of recurrence and prolong survival. However Pu et al. [51] found that there was no difference in survival time between patients with PG and TG. Therefore, this study suggests that in addition to the conventional surgical guidelines, surgeons can also use an additional in ammatory index to help choose the method of gastrectomy. We found that for patients with high PLR patients.
By comparing the postoperative survival of patients who received TG and PG, patients who underwent PG had better prognosis than those who underwent TG. These two methods can effectively remove tumor lesions and reduce the tumor burden, but Rashid et al. [52] showed that effective preservation of lesion-peripheral CD4+ and CD8+ T cells could retain tumor immunity potential after surgery, prolong survival, and reduce postoperative recurrence rate. This might explain the difference in postoperative survival between TG and PG patients. In addition, researchers worldwide have developed abundant preoperative and postoperative treatments for GC, such as neoadjuvant chemotherapy, postoperative chemotherapy, postoperative radiotherapy and targeted therapy. However, it was di cult to assess the sensitivity of patients to different treatment methods. Yuka et al. found that preoperative PLR can evaluate the sensitivity of postoperative chemotherapy [53]. This study also found that patients with SIRI-PLR score 2 had better prognosis than those who did not receive postoperative chemotherapy.
Although patients with high SIRI score are reported not to be suitable for chemotherapy [31], we focused on the scoring system that combined SIRI with PLR, and on GC patients who also had AGC and UGC, which might lead to different results.
Previous studies using serum immune index alone to predict the prognosis of patients or guide therapy have unsatisfactory sensitivity and speci city, which is di cult to individualize evaluation. With the development of Real World Study and Big Data for Cancer Research, the mode combining clinicopathological features and serum immunity markers to predicting the prognosis of GC is widely used in clinical. Fanotto found that the patients with higher ECOG score and lower lactate dehydrogenase (LDH) NLR had better prognosis after chemotherapy [54] Huang et al. found that CA125, CA19-9, CA72-4 and Fibrinogen-to-Lymphocyte are risk factors for peritoneal dissemination among patients with GC, and used these indicators to constructe the Nomogram model [55]. We analyzed the clinical signi cance of the SIRI-PLR score through the The Gastric Cancer Information Management System v1.2 of Harbin Medical University Cancer Hospital database. According to multivariate analysis, it was found that age, SIRI, PLR, and pTNM were independent predictors for high risk of death due to AGC and UGC in the training set. Then, a Nomogram model was constructed in the training set to predict the prognosis for 3 and 5-year. Through ROC analysis, it was found that AUC was 0.767 and 0.754, sensitivity was 88.8% and 84.8%, and speci city was 57.9% and 60.0%. And the constructed Nomogram model also can be used well in the veri cation set. Although this result may be limited by the small number of patients and the cohort grouped by time of admission, it can still indicate that for the patients with AEG and UGC, the prediction model established by age, SIRI, PLR and pTNM is worth further clinical veri cation and application.

Limitatons
This retrospective study still had some limitations. First, it focused on AGC and UGC in an Asian population, therefore, prevalence of AGC and UGC in non-Asian populations needs further exploration. Second, it was di cult to determine whether preoperative gastritis and Helicobacter pylori infection affected circulatory immune cells in patients with GC, and this also needs further exploration. Because there are only 35 patients in stage I, it is not enough to analyze separately. Therefore, we combined 117 patients with stage II and stage I patients into this study. Therefore, whether the constructed Nomogram model has the same clinical signi cance for patients with stage I and stage II should be further studied.
Conclusion SIRI-PLR score is an independent predictor of survival in patients who undergo curative surgery for AEG and UGC. It can further subgroup patients with stage / and to supplement the eighth edition of the AJCC guidelines.
Patients with SIRI-PLR score 2 with postoperative chemotherapy have better survival than patients with SIRI-PLR score 0 or 1. Patients with high PLR with PG have better survival than patients with TG. SIRI-PLR may help clinicians to decide upon individualized treatment for patients with AEG or UGC. The Nomogram with combination of SIRI, PLR, age and pTNM can predict postoperative survival.

Declarations Ethics approval
Approved by Harbin Medical University Cancer Hospital Ethics Committee of (Approval Number: SHGC-1029).

Consent to participate
Because this study is a retrospective analysis, the informed consent form has been signed when collecting patients' general information and clinicopathological data. This study has protected patient privacy under the guidance of the ethics committee.

Availability of data
Patients' data were saved in the Gastric Cancer Information Management System v1.2 of Harbin Medical University Cancer Hospital (Copyright No. 2013SR087424, http:www.sgihmu.com).

Con ict of Interest
The authors declare that they have no con ict of interest.        Relationship between SIRI-PLR score and bene t from postoperative chemotherapy in total patents with SIRI-PLR score 0,1 (a) and 2 (b). Relationship between SIRI, PLR and bene t from surgical method in patents with low PLR (c), high PLR (d), low SIRI (e) and high SIRI (f).