Prognostic value of immunity change after treatment for patients with locally advanced gastric cancer: a retrospective study

Background Although the preoperative immune status is associated with the prognosis in some tumors, less is known about the prognostic signicance of immune status change during the treatment of patients with locally advanced gastric cancer (LAGC). Methods The records of 210 patients with LAGC were retrospectively analysed. The pre-, and post-treatment (after gastrectomy and three cycles of chemotherapy) values of lymphocyte-to-monocyte ratio (LMR) and change of LMR (cLMR) were evaluated. A novel immunity change score (ICS) incorporated both preoperative LMR (pLMR) and cLMR was developed and its prognostic value was evaluated. Results cLMR was an independent predictor and patients with cLMR >1 after treatment had a favorable survival compared with the others (51 vs 31 months, P < 0.001). Based on the cLMR and pLMR, the ICS was dened as follows: ICS=1 (pLMR ≤ 4.53 and cLMR ≤ 1); ICS=2 (pLMR ≤ 4.53 and cLMR>1, or pLMR>4.53 and cLMR ≤ 1); and ICS=3 (pLMR>4.53 and cLMR>1). Multivariate analysis revealed that the ICS was a signicant independent biomarker ( P < 0.001). The performances of ICS in terms of the time-dependent receiver operating characteristics (t-ROC) curve and concordance index (C-index) analysis were better than those of pLMR and cLMR. Then we established a nomogram incorporated the ICS, CEA, and TNM stage to predict the 3- and 5- year survival. Decision curve analysis and calibration curve demonstrated that the nomogram was clinically useful. Conclusion The dynamic change of immune status is signicantly associated with prognosis for LAGC patients. Combining with the cLMR and pLMR could improve the prognostication for LAGC patients.


Introduction
The prognosis of gastric cancer (GC) remains very poor, due to tumor load at the time of diagnosis and the limited e cacy of treatment [1][2] . For GC patients at an advanced stage, including stage -, the veyear survival rates are remaining at 30%-50% even if underwent radical resections [3] . Therefore, identifying patients at high risk of micrometastasis or local recurrence, and preventing deterioration are important to decrease cancer-related death [4] .
Recently, with the increasing understanding of the connection between the immune system and tumor, immunotherapy is receiving increasing attention and the prognosis value of systemic immune-related markers has been investigated in various tumors [5][6][7] . Because of the more convenient and affordable for detecting, absolute lymphocyte count (ALC) and monocyte count (AMC) in blood routine examination have been widely used for immune surveillance in several hematologic malignancies and solid tumors [8][9][10] . In GC, the value of ALC, AMC, and lymphocyte to monocyte ratio (LMR) had been reported to be associated with the prognosis both before and after gastrectomy, and a higher lymphocyte or lower monocyte were found to be associated with a longer survival time [9,[11][12][13] . However, previous studies isolated the prognostic values of ALC, AMC, and LMR at a single point in time, and ignored the alterations in immune statue during the treatment process, especially when the tumor load changes and chemotherapy applied [9,[12][13] .Thus, we hypothesis that adding the strati cation according to the change of LMR (cLMR) during treatment to the basis of preoperative LMR (pLMR) may help to discriminate patients accurately.
The aim of this study was to investigate the change of immune statue in response to surgery and adjuvant chemotherapy, and also assess its prognosis value. Besides, we try to establish an immunity changes score (ICS) based on both pLMR and cLMR to predict the prognosis of patients with stage -LAGC.

Methods
Study subjects.
From January 2013 to December 2017, patients with LAGC diagnosed and treated in Lanzhou University Second Hospital were retrospectively studied. All patients underwent D2 lymphadenectomy or extend D2 lymphadenectomy based on the guidelines of the Japanese Research Society for the Study of GC [14] , and suffered from stage pT2-4NxM0 according to the criteria of the 7th edition of the TNM classi cation [15] .
To ensure the homogeneity of treatment, patients who received any previous chemotherapy or radiotherapy, and received less than three cycles of adjuvant chemotherapy were excluded from this study. Finally, 210 consecutive patients were included.
All patients were followed up by giving a telephone call or posting letters. The last time of follow-up was 1 June 2019. All follow-up ndings were collected and recorded in the database. The current study was performed following the Declaration of Helsinki and approved by the institutional review board.

Data collection and de nition
Blood routine examinations were tested conventionally during the perioperative period and the following course of adjuvant chemotherapy. The values of ALC and AMC within 3 days before the operation and the fourth course of adjuvant chemotherapy were collected. The normal value of peripheral blood ALC and AMC were 0.8-4x10 9 /L and 0.12-0.8x10 9 /L, respectively. Immune statue was quantized by LMR, which was equivalent to the ratio of ALC to AMC.
Additional variables on clinicopathologic characteristics including age, sex, Eastern Cooperative Oncology Group (ECOG) performance status, tumor location, tumor size, Lauren classi cation, degree of differentiation, nerve invasion, vessel invasion, pathological TNM stage (pTNM) and the preoperative values of Carbohydrate antigen 199 (CA-199) as well as Carcinoembryonic Antigen (CEA) were obtained.
All the patients in this study underwent total, subtotal, or distal gastrectomy based on the primary tumor size and location, and also accepted at least three cycles of adjuvant chemotherapy, which including FLOT, SOX, and FOLFOX6 regimen [15] .

De nition of immune-related indicators and ICS
The changes in overall immune status throughout the treatment process including surgery and adjuvant was evaluated by the cLMR, which was determined by dividing the LMR before the fourth chemotherapy (fLMR) by the preoperative one (pLMR), and immunity increased was de ned as cLMR > 1. The optimal cut-off value for continuous variables including pLMR in terms of survival was calculated by the software X-tile (Tale University, New Haven, CT, the United States of American) [16] . According to the cut-off value of pLMR, all patients were divided into four groups: HG, patients with high pLMR and cLMR greater than one; HL: patients with high preoperative LMR and cLMR equal to or lesser than one; LG: patients with low preoperative LMR and cLMR greater than one; and LL: patients with low preoperative LMR and cLMR equal to or lesser than one. Different degree of ICS will be given to the above four groups based on survival.

Statistical analysis
Paired samples Wilcoxon test were used to compare the value of pLMR and fLMR. For categorical variables, the Chi-square test or Fisher's exact test were used. Overall survival (OS) was de ned as the time between diagnosis and last contact or death, and which between different subgroups were plotted using the Kaplan-Meier curves and compared with the log-rank tests. The prognosis abilities of pLMR and ICS were compared by generating time-dependent receiver operating characteristic (t-ROC), which evaluated the discriminatory of continuous variables for time-dependent disease outcomes [17] .The area under the curve (AUC) represents the discriminative power of these two variables and can be calculated to comparing the ROC curves visually [18] . Besides, to quantify the discrimination performance of pLMR and ICS, Harrell's concordance-index (C-index) was measured. Univariate analysis was performed rstly to screen out the potential prognostic factors, and then the COX proportional hazards regression analysis was used to identify the independent factors. Finally, a nomogram based on the basic of multivariable COX analysis and ICS was built to assist the clinician in predicting prognosis individually. Calibration curves and C-index were used to assess the calibration and quantify the predictive accuracy of the predictive nomogram. To investigate the clinical usefulness of the nomogram for predicting 3-year survival, the decision curve analysis (DCA) was performed, which could quantify the net the bene ts at different threshold probabilities in the validation dataset [19] . All hypothesis tests were two-sides and P < 0.05 was recognized as statistical signi cance. All statistical analysis was conducted with Statistical Product for Social Sciences (SPSS) software (version 24.0; Inc, Chicago, IL, United States) and R ver. 3.5.3 (R Foundation for Statistical Computing, Vienna, Austria). The R packages used in this study were including "rms", "timeROC", "stdca", and "survival".

Patients characteristics
Two hundred and ten consecutive stage II-III LAGC patients who received surgery and followed by adjuvant chemotherapy were meeting our inclusion criteria. The median age was 57.6 years (range, 31-80 years), and 61(29.05%) patients were female. Patients demographics are described in SupplementaryTable S1. According to the pathological con rmation, 92(43.81%) patients were at stage II and 118(56.19%) patients were at stage III. The surgical procedures included total gastrectomy in 68(32.38%), subtotal gastrectomy in 44(20.95%), and distal gastrectomy in 98(46.67%). Besides, 138(65.71%) patients received adjuvant chemotherapy with docetaxel.

Changes in immune status during treatment
The median LMR increased from 4.816 (range 0.636-22.75, IQR3.68-6.408) at baseline to 5.358 (range 0.547-24.9, IQR3.943-7.114) (P = 0.039) after radical resection and three cycles of chemotherapy, and 111 (52.86%) patients showed an increase in LMR. The relationship between cLMR and clinicopathological variables were shown in Table 1. Patients with a well histologic type or intestinal type of Lauren classi cation were more likely to have an elevated immunity after treatment.
Additionally, the relationship between clinicopathologic characteristics and ICS was investigated (Table   3). There were 34 (16.19%) patients in the ICS = 1 group, 119(56.67%) patients in the ICS = 2 group, and 57(27.14%) patients in the ICS = 3 group. An elevated ICS was found to be signi cantly associated with intestinal Lauren classi cation and low tumor size (all P < 0.05).

Predictive nomogram based on ICS
To provide a quantitative method for prognosis prediction, the model that incorporated the independent factors from the multivariate analysis including CEA, pTNM stage, and ICS was developed and presented as the nomogram ( Figure 3A). The calibration curves of the nomogram for the probabilities of 3-year survival and 5-year survival both demonstrated optimal agreements between nomogram prediction and actual observation in the internal validation ( Figure 3B, C). In addition, the C-index of the nomogram based on the ICS, CEA, and pTNM stage (model1; 0.816; 95% CI, 0.778-0.855) was both signi cantly higher than the prognostic model merely based on pTNM stage and CEA (model2; 0.767; 95% CI 0.713-0.82; P = 0.008) and the model based on the pLMR, CEA, and pTNM stage (model3; 0.792; 95% CI, 0.748-0.835; P = 0.035).
The decision curves for models 1, 2, and 3 to predict the correct 3-year outcome in patients with AGC were shown in Figure 4. All models were useful between threshold probabilities of 0.45-0.72, and the net bene t of model 1 was better than the other 2 models when the threshold probabilities between 0.26 and 1.

Discussion
It is well accepted that LMR as one of the immunoin ammatory markers was playing an important role in immune surveillance and tumor progression. Previously published data had demonstrated the correlation between survival and LMR from different points in time, especially in the preoperative setting. However, there were no reports concerning the prognostic role of the kinetics of this biomarker during treatment. Indeed, this is the rst study to investigate the prognostic value of LMR and its kinetic after gastrectomy and adjuvant chemotherapy in patients with LAGC. We found that LMR increased after treatment was an independent favorable factor of prognosis. In addition, patients with intestinal Lauren classi cation or well-differentiated might be more susceptible to enhance immunity after treatment.
As the basis for the adaptive and innate immune system, lymphocytes participate and enhance immunesurveillance, thereby mediating cytotoxic cell and inhibiting tumor cell proliferation, invasion, and metastasis [20] . It has been reported that the presence of tumor-in ltrating lymphocytes was associated with improving the prognosis in a variety of tumors, possibly owing to tumor-in ltrating, lymphocyteinduced, and anti-tumor activity as well as angiogenesis [21] . Therefore, lymphocytopenia weakens the immune response and represents a decrease of anti-tumor ability and a poor outcome in patients with cancer. Inversely, the circulating monocytes may contribute to tumor growth and inhibit immunosurveillance according to the previous nding [22] . Furthermore, tumor-associated macrophages derived from circulating monocytes have been proven to be associated with tumor metastasis, enhanced angiogenesis, and poor outcome [23]. Thus, LMR, which is calculated based on the values of ALC and AMC, can amplify the ability of ALC and AMC to re ect the immune response and play an important role in the prognosis of GC.
Several studies have determined the prognosis value of LMR in the preoperative and postoperative setting [13,[24][25][26] .However, there was an important limitation that the cut-off value for LMR differed among those researches and, thus limited patient strati cation. In addition to this, the outcome may be similar even if the patients were at different groups according to pLMR. In our study, the optimal cutoff value for pLMR is 4.53, and the 5-year OS rate between patients in HL group and LG group are indiscrimination. Previous studies focused on the dynamic changes in terms of the LMR after treatment showed that compared with the pretreatment LMR, the LMR elevated at 4 weeks after the start of nivolumab monotherapy was signi cantly associated with a prolonger progression-free survival and OS in non-small-cell lung cancer [27] . Moreover, LMR increased more than 1.8 from day 15 to day 100 in patients with Hodgkin Lymphoma who underwent autologous stem cell transplantation was correlated with a lower risk of relapse [28] . In this study, we found that increased cLMR was a favorable prognostic factor for A GC. Although the prognostic accuracy of cLMR was not as well as pLMR, the combination of cLMR could further assist pLMR in identifying outcome and selecting the patients who should receive individualized treatment. Additionally, cLMR was an independent prognostic factor and an increased cLMR was associated with a well histological type or intestinal Lauren classi cation. These data support the notion that cLMR, which re ected a dynamic reaction of immune response caused by treatment and tumor load, was also a meaning measurement. Therefore, we developed a novel prognostic score, called ICS, based on the combination of pLMR and cLMR, and found that the ICS was also an independent prognostic factor for patients with LAGC. Moreover, as an integrated indicator, the clinical value, discrimination and prognostic accuracy of ICS were signi cantly better than that of pLMR and cLMR. Thus, we recommended that patients with lower ICS should reexamine regularly and try to receive some molecular targeted agents.
To date, the most important and common predictive systems for GC are the AJCC TNM staging system, which is based on the depth of tumor invasion (T), the number of lymph node metastasis (N), and the distant metastases (M) [15] . However, due to the variety of patient prognosis, it might be ambiguous for patient selection solely depending on the TMM classi cation. To assist with clinical practice and decision making, a nomogram, currently an accurate and discriminatory tool for predicting prognosis among patients with cancer [29] , was built by incorporating ICS into CEA and TNM staging. It is essential to validate the nomogram to avoid over tting of the model and determine generalizability [30] . In this study, the calibration plots for 3-year and 5-year survival showed a favorable agreement between prediction and actual obversion in the internal validation, which guarantees the reliability of the constructed nomogram. Moreover, the nomogram exhibited better discrimination power and accuracy than that of non-ICS or pLMR based nomogram. Therefore, in clinical practice, the ICS can be used as a supplement to the TNM staging system to identify high-risk subgroups of patients, to assess the individual clinical outcome, and to provide a basis for guiding follow-up and treatment.
Our study does have some limitations and de ciencies. First, we lacked the data to examine all patients at all time points during treatment. Second, this study with a retrospective nature was performed in a single center, that inevitably led to potential biases and a relatively small sample size. However, the sample size of this article is su cient to effectively show the prognosis value of LMR kinetics on survival. Moreover, due to the patients were consecutively enrolled, some patients had a follow-up period less than ve years and no outcome was observed, which may have an impact on the results of the survival analysis. Finally, because the ICS in our nomogram is calculated based on LMR after three cycles of chemotherapy, so it doesn't apply to the prediction of outcomes in the initial three months after surgery. Future studies will need to elucidate the mechanism of changes in immune status of GC patients with different tumor stages after surgery or chemotherapy to deduce whether cLMR can better stratify patients for additional treatment.

Conclusion
Our study is the rst to show that immune status change is signi cantly associated with prognosis for GC patients. Considering that, we developed a novel prognostic score named ICS, which could discriminate the prognosis of patients effectively. Additionally, we established and validated a novel nomogram that incorporates ICS, CEA, and TNM staging system to predict prognosis for LAGC patients. Through this model, clinicians could estimate the effect of treatment, predict the outcome of individual patients, and identify patients who should receive speci c treatment strategy more precisely.

Declaration
Ethical approval and consent to participate All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards. Due to the retrospective nature of this study, the Ethics Committee of Lanzhou University Second Hospital approved the study and determined that written informed consent was not required.
Con ict of Interest: All authors declare no con icts of interest.
Availability of data and materials:The datasets used and/or analysed during the current study available from the corresponding author on reasonable request.
Funding: This research did not receive any speci c grant from funding agencies in the public, commercial, or not-for-pro t sectors.     Time-dependent ROC curves for the pLMR, cLMR, and ICS. The x-axis represents year after diagnosis, whereas the y-axis represents the estimated AUC at a single point in time. The red, blue, and green full lines represent the AUCs for the pLMR, cLMR, and ICS, and dash lines represent the 95% con dence intervals for each AUC. Abbreviation: ROC: receiver operating characteristic; AUC: area under the curve; pLMR: preoperative lymphocyte-monocyte ratio; cLMR: change of lymphocyte-monocyte ratio; ICS: immunity change score.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.