Correlation Analysis Between Preoperative Systemic Immune Inammation Index and Prognosis of Gastric Carcinoma Patients: Based on Propensity Score Matching Method

Background: To explore the correlation between the preoperative systemic immune inammation index (SII) and the prognosis of patients with gastric carcinoma (GC). Methods: The clinical data of 771 GC patients surgically treated in the Department of Gastrointestinal Surgery, Qinghai University Aliated Hospital from June 2010 to June 2015 were retrospectively analyzed, and their preoperative SII was calculated. The optimal cut-off value of preoperative SII was determined using the receiver operating characteristic (ROC) curve, the confounding factors between the two groups were eliminated using the propensity score matching (PSM) method, and the correlation between preoperative SII and clinicopathological characteristics was assessed by chi-square test. Moreover, the overall survival was calculated using Kaplan-Meier method, the survival curve was plotted, and log-rank test was performed for the signicance analysis between the curves. Univariate and multivariate analyses were also conducted using the Cox proportional hazards model. Results: It was determined by the ROC curve that the optimal cut-off value of preoperative SII was 489.52, based on which 771 GC patients were divided into high SII (H-SII) group and low SII (L-SII) group, followed by PSM in the two groups. The results of Kaplan-Meier analysis showed that before and after PSM, the postoperative 1-, 3- and 5-year survival rates in L-SII group were superior to those in H-SII group, and the overall survival rate had a statistically signicant difference between the two groups (P<0.05). Before PSM, preoperative SII [hazard ratio (HR) =2.707, 95% condence interval (CI): 2.074-3.533, P<0.001] was an independent risk factor for the prognosis of GC patients. After 1:1 PSM, preoperative SII (HR=2.669, 95%CI: 1.881-3.788, P<0.001) still independent risk factor for the prognosis of GC patients. The increase in preoperative SII in peripheral blood indicates a worse prognosis. the inuence of confounding factors on the prognosis, 1:1 PSM was performed for patients in high SII (H-SII) group and low SII (L-SII) group. Other covariates had no statistically signicant difference (P>0.05). 73 months (95%CI: 65.172–80.828) in L-SII group and 28 months (95%CI: 24.646–31.354) in H-SII group. The 1-, 3- and 5-year survival rates were 92.8%, 81.0% and 67.9%, respectively, in L-SII group, and 80.0%, 39.8% and 28.9%, respectively, in H-SII group. It can be seen that the survival rate in L-SII group was superior to that in H-SII group, and the OS rate had a statistically signicant difference between the two groups (P < 0.001). After PSM, the 1-, 3- and 5-year survival rates were 89.3%, 74.3% and 60.4%, respectively, in L-SII group, and 80.8%, 45.8% and 32.9%, respectively, in H-SII group. The mean survival time was 56.771 months (95%CI: 51.613–61.929) in L-SII group. In H-SII group, the mean survival time was 38.443 months (95%CI: 34.792–42.093), and the median survival time was 32 months (95%CI: 22.446–41.554). It can be seen that the survival rate in L-SII group was signicantly better than that in H-SII group, and the OS rate had a statistically signicant difference between the two groups (P < 0.001)


Introduction
Gastric carcinoma (GC) is a common digestive tract malignancy. In 2018, there were up to 782,000 deaths of GC, making it the third cause of death in malignancies [1] . China is a GC-prone country, and the new cases and deaths account for 44.1% and 49.9%, respectively, of the total globally, with an age-standardized 5-year survival rate of 27.4%. At the same time, GC patients have insidious early symptoms and the therapeutic effect is poor, leading to a poor prognosis [2] . Therefore, researchers have been constantly exploring the new simple, economical and accurate prognostic evaluation index similar to TNM stage currently, so as to better guide the clinical treatment [3] .
According to related studies, both in ammatory response and immune response are closely related to the occurrence and development of tumors [4,5] . Tumor-induced in ammatory response can cause corresponding changes in the blood neutrophil (NE), lymphocyte (LY) and platelet (PLT) counts [6,7,8] . On this basis, some studies have tried to discover new potential biomarkers related to the prognosis. For example, the preoperative systemic immune in ammation index (SII) calculated based on NE, LY and PLT is correlated with the prognosis of breast cancer, liver cancer and pancreatic cancer [9,10,11,12] . However, there are currently few studies on the correlation between SII and prognosis of GC. In the present study, therefore, the value of SII for the prognosis of GC patients was explored using propensity score matching (PSM), so as to provide references for clinical treatment.

Optimal Cut-off Value Of Sii
The ROC curve of SII was plotted, and its AUC and Youden index were 0.721 (95%CI: 0.681-0.761) and 0.384, respectively. The corresponding optimal cut-off value was 489.52, and the evaluation sensitivity and speci city were 55.5% and 82.9%, respectively ( Fig. 2). Based on the cut-off value, 771 patients were divided into L-SII group (SII ≤ 489.52, n = 531) and H-SII group (SII > 489.52, n = 240).
Correlation between SII and clinicopathological factors of GC patients before and after PSM Before matching, SII was related to the maximum diameter of tumor, histological type, serum albumin (ALb), white blood cell (WBC), carbohydrate antigen 125 (CA125) and TNM stage (P < 0.05), but not related to the age, gender, smoking, drinking, carcinoembryonic antigen (CEA), CA199 and operation method (P > 0.05). PSM was performed in L-SII group and H-SII group, and a total of 354 patients were eligible, including 177 cases in H-SII group and 177 cases in L-SII group. The above differences were evenly distributed after matching (P > 0.05) ( Table 1).  (Table 3). Besides, the correlation between statistically signi cant independent risk factors after PSM (P < 0.05) (except SII) and prognosis of GC patients was analyzed by Kaplan-Meier method. It was found that the larger maximum diameter of tumor, lower level of ALb, higher level of CA125 and higher TNM stage indicated a worse prognosis. Besides, the prognosis was better among patients undergoing DG than that among patients treated with PG and TG (Fig. 4).      (Table 6). a statistically signi cant difference between the two groups (P = 0.041) (Fig. 5).

Discussion
In recent years, related studies have demonstrated that the in ammatory response (tissue necrosis, activation of immune response, secretion of cytokines) plays an important role in the occurrence, development and metastasis of tumors [15,16] . In the early stage, the in ammatory response is induced and the anti-tumor immune response is inhibited, thereby promoting growth of tumor cells. With the growth of tumor, the in ammatory response is further enhanced. To eliminate in ammation, the recruitment of immune cells to tumor tissues is reduced, resulting in immune escape of tumor cells. In the late stage, the body's immunity is almost lost, and the in ammatory response is further worsened, thus facilitating the tumor progression [17] . SII, a new in ammatory index integrating the NE, LY and PLT counts [18] , can re ect the balance status between tumor immunity and in ammatory response more comprehensively than platelet-to-lymphocyte ratio (PLR), neutrophil-to-lymphocyte ratio (NLR) and monocyte-to-lymphocyte ratio (MLR), and better predict the prognosis of cancer patients [18,19,20 ] .
In the present study, the optimal cut-off value of SII was determined using the ROC curve, based on which the patients were divided into H-SII group and low L-SII group. In terms of the correlation between SII and the clinicopathological characteristics of patients, SII was related to the maximum diameter of tumor, histological type, ALb, WBC, CA125 and TNM stage (P < 0.05). In addition, the survival time of GC patients in H-SII group was signi cantly shortened, and the 5-year survival rate was signi cantly lower than that in L-SII group. Wang [21] et al. retrospectively analyzed the clinical data of 444 GC patients treated with gastrectomy and found that the 5-year survival rate is 29.4% and 46.1%, respectively, in H-SII group and L-SII group, suggesting that the SII level is related to the prognosis of GC patients, consistent with the results of this study. Han BL [22] also showed that SII is closely related to the prognosis of GC patients, and high SII indicates a poor prognosis of GC patients. Its mechanism can be explained by the functions of NE, PLT and LY. In the case of tumor invasion, the in ammatory response can promote the increase in NE and the secretion of NO, reactive oxygen species, arginase and other active substances, which can regulate the activity of NK cells and LY, thereby facilitating tumor growth and metastasis [23,24] . In addition, the PLT count rises in in ammation, and PLT promotes tumor angiogenesis via releasing vascular endothelial growth factor and platelet-derived growth factor [25] , creating favorable conditions for tumor growth. Moreover, PLT can form colonizing tumor thrombi with tumor cells, thus promoting the further invasion of tumor cells into the body. On the contrary, LY can effectively suppress the occurrence and development of tumors, and induce the death of tumor cells. The long-term in ammatory stimulus will lead to decline in LY subsets, thereby causing immune dysfunction, and the decrease in LY count can raise the risk of immune escape of tumor cells [26,27] . Therefore, SII can re ect the relation between systemic in ammatory response and body's immunity during tumor progression. The higher the SII is, the severer the in ammatory response and immunosuppression in GC patients will be.
In this study, it was con rmed by univariate analysis before and after PSM that age, maximum diameter of tumor, SII, ALb, CA125, operation method and TNM stage were risk factors for the prognosis of GC patients. Before PSM, the results of multivariate analysis showed that age, histological type, SII, ALb, CEA, CA125, WBC, operation method and TNM stage were independent risk factors for the prognosis of GC patients. After PSM, the results of multivariate analysis revealed that the maximum diameter of tumor, SII, ALb, CA125, operation method and TNM stage were independent risk factors for the prognosis of GC patients. Then the statistically signi cant factors after PSM (except SII) were analyzed. It was found that the larger maximum diameter of tumor, lower level of ALb, higher level of CA125 and higher TNM stage indicated a worse prognosis. The prognosis was better among patients undergoing DG than that among patients treated with PG and TG. Wang et al. [28] found in the study on the correlation between SII and the prognosis of GC patients that SII and TNM stage are independent risk factors for the prognosis of GC patients, consistent with the results of this paper. As the main component of plasma protein, ALb can re ect the nutritional status of patients, and it has also been widely used as an antidote and transporter in predicting the survival status and disease progression of cancer patients [29,30] . CA125 is mainly present in epithelial tissues and serum of patients, and its expression level in the serum of GC patients is higher than that in normal people, which is positively correlated with the TNM stage and negatively correlated with the prognosis of GC patients, consistent with the results of this study. This indicates that CA125 can be used to assess the severity of GC [31] . In addition, the multivariate analysis before and after PSM demonstrated that no positive result was obtained in age after PSM, but other scholars argued that there is a correlation between age and the prognosis of GC patients [32,33] . The reasons why age was not an independent risk factor for the prognosis of GC patients in this study are as follows: PSM was used in this study to validate the impact of SII on the prognosis on the basis of excluding other confounding factors, which stressed the effect of SII on the prognosis, thus affecting the prognostic assessment by other indexes to a certain extent. Besides, it is possibly because the sample size was small after PSM and this study itself had limitations, rather than simply considering age unrelated to the prognosis of patients. Therefore, the sample size remains to be further expanded in the future. To further explore the correlation between age and the prognosis of GC patients, age was further subjected to strati cation analysis to exclude the value deviation in a certain stage caused by the small sample size or the bias of age strati cation in this study. It is reported in the literature that elderly patients are aged ≥ 60 years old [34,35] . In this study, the median age of 354 patients was 59.0 years old after PSM. Therefore, patients aged ≥ 60 years old were de ned as elderly patients in this study. The results revealed that the mean SII of elderly patients (495.73) was signi cantly higher than that of non-elderly patients (443.21), suggesting the correlation between SII and patient's age. In addition, the risk factors for the prognosis of GC patients were analyzed under age strati cation. It was found that the high SII was correlated with the poor prognosis of elderly patients with GC, and the 5-year OS rate of elderly patients was lower than that of non-elderly patients in H-SII group (31.9% vs. 62.1%). The speci c mechanism remains to be further con rmed by a large number of prospective studies.
There were certain limitations in this study. First, this was a small-sample retrospective study on GC patients in high-altitude areas. To better control the bias, PSM was used for data analysis, but all the impact of covariates on the outcome failed to be fully eliminated. Second, this was a single-center study, so the conclusion cannot fully represent the characteristics of other study centers and populations, limiting its popularization. Therefore, the preliminary results need to be validated by larger-sample randomized controlled studies in the future, so as to offer more convincing theoretical support to the existing conclusion.
In conclusion, SII, that remains simple, universal, non-invasive, cheap and reproducible, is expected to be an index for assessing the prognosis of GC patients.

Data availability
The datasets generated during and/or analysed during the current study are available from the corresponding author on reasonable request.  ROC curves were plotted to determine the optimal cut-off value of SII.

Figure 3
Postoperative OS curve in the two groups before and after PSM. a: Before PSM, the 5-year OS rate in L-SII group was signi cantly higher than that in H-SII group (67.9% vs. 28.9%) (P<0.001). b: After PSM, the 5-year OS rate in L-SII group was signi cantly higher than that in H-SII group (60.4% vs. 32.9%) (P<0.001).  Correlation between the prognosis of non-elderly and elderly patients and SII after PSM. a: The 5-year OS rate in L-SII group was signi cantly higher than that in H-SII group among the non-elderly patients (62.1% vs. 31.9%) (P<0.001). b: The 5-year OS rate in L-SII group was signi cantly higher than that in H-SII group among the elderly patients (43.8% vs. 19.4%) (P=0.041).