Proposed Models for Prediction of Mortality in Stage-I and Stage-II Gastric Cancer and 5 Years after Radical Gastrectomy

The current American Joint Committee on Cancer (AJCC) staging system provides limited information for patients with early death from stage-I and stage-II gastric cancer (GC) and death at >5 years after radical gastrectomy. The aim of this study was to construct nomogram models to predict the mortality risk of these patients. In this study, clinical and pathological data on patients who underwent curative gastrectomy at Harbin Medical University Cancer Hospital between 2000 and 2014 were retrospectively collected. Receiver operating characteristic analysis was used to screen for sensitive serum immune biomarkers to predict the risk of mortality death >5 years after radical gastrectomy (Group A) and risk of early death in stage-I and stage-II GC (Group B). The prediction model was constructed by combining serum immune markers with clinicopathological features by R Studio. We found that serum fibrinogen (F), systemic immune inflammation (SII), and pTNM stage were independent risk factors for prognosis in Group A (P < 0.05). F, SII, age, Borrmann type, and scope of gastrectomy were independent risk factors for prognosis in Group B (P < 0.05). The area under the curve of the predictive model in Groups A and B was 0.726 and 0.848, respectively. In conclusion, the predictive models of F and SII combined with clinicopathological features can predict high mortality risk in patients with stage-I and stage-II GC and >5 years after radical gastrectomy, which will contribute to the supplement of the traditional AJCC system and to individual survival prediction.


Introduction
With the enhanced awareness of public health protection in developing countries, the incidence rate of gastric cancer (GC) has been decreasing gradually, but it is still the sixth highest incidence and third highest mortality rate of malignant tumors worldwide [1]. TNM staging and histological subtype classification according to tumor infiltration, regional lymph nodes, and distant metastasis are the conventional criteria for evaluating prognosis and guiding treatment after surgery [2]. Radical surgery is the basic method of comprehensive treatment for GC [3]. In addition to the molecular targeted therapy, National Comprehensive Cancer Network (NCCN) guidelines also support the routine use of postoperative adjuvant chemotherapy for patients with stages II-IV GC [4]. However, it is well known that clinical outcome can significantly vary among patients receiving similar treatment regimens for the same histological tumor stage.
is situation indicates that although the traditional classification tools provide effective prognostic information, they are still unable to satisfy individual evaluation.
At present, it is generally believed that there are two limitations in the eighth edition of TNM staging system for GC [5]. Firstly, it provides deficient information for predicting recurrence and metastasis at 5 years after radical gastrectomy. Secondly, it has some defects in the individualized risk assessment of early recurrence and death in patients with stage-I and stage-II GC. Patients with survival time >5 years are often considered to have clinically complete tumor regression. e recurrence rate of these patients is considered to be ∼10% [6,7], and further postoperative review and follow-up are often omitted. For patients with stages IA, IB and II of GC based on the Japanese Gastric Cancer Association (JGCA) guidelines, the 5-year survival rate can reach 91.5%, 83.6%, and 70.6%, respectively [8]. As a result, clinicians pay less attention to these patients postoperatively than those with advanced-stage GC and ignore the possibility of early recurrence. In view of these two kinds of recurrence, it is necessary to supplement the traditional classification system for individualized risk prediction.
Clinicians have also made efforts for individualized tumor monitoring and construction of predictive models [9]. However, it is difficult for predictive models only based on tumor-related indicators to achieve satisfactory clinical application after further verification. Tumor immunity has an important role in controlling tumor progression, and more studies have shown that tumor progression or recurrence depends not only on tumor characteristics, but also on the immune response in the tumor microenvironment and peripheral blood. It has been proposed that adaptive immune response may play a critical role in preventing tumor recurrence [10]. In particular, memory T cells in peripheral blood can disseminate and are maintained for long periods of time and recognize the surface antigens of recurrent tumor cells. Researchers have combined the immune response in the tumor microenvironment with pTNM staging system to construct the TNM-I (TNM-immune) staging system and for guiding postoperative treatment of colon cancer, which has proved a reliable method for clinicians [11,12]. However, in GC with higher heterogeneity, the traditional immunohistochemical method is interfered by the complexity, heterogeneity, and inconsistency of regional tissue selection. e immune evaluation of peripheral blood may reduce the heterogeneity to provide an essential prognostic and potentially predictive tool [13]. Our previous studies have found that peripheral blood inflammatory factors are superior to traditional tumor markers in the early diagnosis of GC. erefore, introduction of peripheral immune response as a biomarker to classify cancer will facilitate clinical decision-making [14].
is study retrospectively analyzed patients who underwent radical gastrectomy in the Harbin Medical University Cancer Hospital between 2000 and 2014. Patients with death from recurrence >5 years after radical gastrectomy and patients with early death within 2 years from stage-I and -II GC were selected. Finally, we screened the serum immune factors related to recurrence and prognosis to construct a predictive model combined with clinicopathological characteristics. e diagnosis was based on paraffin sections obtained by electronic fiber gastroscopy before surgery and confirmed postoperatively by experienced pathologists. During hospitalization, all patients underwent hematological examination, abdominal ultrasonography, electrocardiography, stomach computed tomography (CT)/magnetic resonance imaging (MRI), chest radiography, and abdominal ultrasonography, and some patients underwent positron emission tomography (PET)/computed tomography (CT) when necessary.

Materials and Methods
Inclusion criteria were as follows. Group A: the survival time after radical gastrectomy was >5 years and the patients with long-term survival were followed up for at least 7 years. Group B: patients with pTNM stage-I and -II GC were followed up at least for 2 years (including loss to follow-up).
is study only focused on the recurrence of patients in two years and patients in Group B will be further followed up and treated. e exclusion criteria were as follows: (1) preoperative radiotherapy and/or chemotherapy; (2) antiplatelet therapy within the previous 3 months; (3) steroid therapy upon admission; (4) recurrent GC; (5) severe heart disease; (6) hematological malignancies, including multiple myeloma; (7)  e pTNM stage was according to the 8th edition American Joint Committee on Cancer (AJCC). All patients were reexamined by ultrasound, CT and gastroscopy, and tumor markers at least once a year, and PET/CT was performed as needed.

Laboratory Examination.
Complete blood count (CBC) with automated differential counts, serum fibrinogen (F), and traditional tumor markers were performed for all patients. On the day of admission or the morning of day, 2.2 ml peripheral fasting blood was collected from the cubital vein and the serum was separated for analysis. CBC was performed within 4 h. For the inflammatory index, neutrophil-lymphocyte ratio (NLR), platelet-lymphocyte ratio (PLR), and systemic immune inflammation (SII) � neutrophil count × platelet count/lymphocyte count.

Statistical Analysis.
Overall survival (OS) was defined as the time from surgery to death from any cause due to GC. If patients were alive at last follow-up, they were censored. Log-rank test and Kaplan-Meier method were used to analyze the survival curves. e survival time was presented as median ± standard deviation. e survival analysis of Group A only included patients who survived >5 years, so the starting point was 60 months. e prognostic significance of NLR, PLR, SII, and F for patients with GC was 2 Journal of Oncology calculated and compared according to receiver operating characteristic (ROC) curve analysis. e area under the curve (AUC) was calculated, and the optimal cut-off value was analyzed by the Youden index. e chi-square test also was used to analyze the association between blood test index and clinicopathological characteristics. P < 0.05 was considered statistically significant. Univariate and multivariate analyses based on cox regression were used to analyze the independent risk factors for prognosis. e variables with P < 0.05 in the univariate analysis were subsequently included in the multivariate analysis, and variables with P < 0.05 in the multivariate analysis were considered to be independent risk factors for prognosis. Odds ratios (ORs) and 95% confidence intervals (CIs) were estimated for each factor. R Studio was used to construct the nomogram model of risk assessment using the SvyNom and rms packages. e relation curve and scatter plot were drawn by GraphPad Prism 8. SPSS version 25.0 (Chicago, IL, USA) was used for analysis.

Clinical Characteristics.
A total of 755 patients were selected in this retrospective research, and the process for inclusion of patients is shown schematically in Figure 1. According to the inclusion criteria, patients who survived >7 years or those with tumor-related death within 5-7 years after radical gastrectomy were designated Group A. Patients with pTNM stages I and II receiving radical gastrectomy were designated Group B. ere were 315 patients (238 male, 77 female) in Group A, with a median age of 60 years, and the number with stage-I, -II, and -III GC was 91, 103, and 121, respectively. ere were 440 patients (335 male, 105 female) in Group B, with a median age of 58 years, and the number with stage-I and -II GC was 160 and 280, respectively.

Survival Analysis of F-SII Score
System. According to the F-SII score system in Group A, patients with score 1 or 2 had no significant difference in OS, which was worse than that in patients with score 0. e OS of patients with scores 0, 1, and 2 was 84.0 ± 3.47, 84.0 ± 6.83, and 84.0 ± 6.18 months and the 5-7-year survival rate was 94.9%, 71.9%, and 68.4%, respectively ( Figure 4(a)). F-SII score had a significant association with tumor diameter and pTNM stage (P < 0.001 and P � 0.002) ( Table 2). In Group B, the F-SII score was negatively correlated with OS and patients with score 0 had the best survival. e OS of patients with scores 0, 1, and 2 was 24.0 ± 4.03, 24.0 ± 6.08, and 20.52 ± 7.88 months and the 2-year survival rate was 89.4%, 72.3%, and 48.2%, respectively ( Figure 4(b)). F-SII scoring system had a significant association with age, Borrmann type, tumor diameter, and pTNM stage (P � 0.024, P � 0.042, P < 0.001, and P < 0.001) ( Table 2).

Construction of Nomogram for Predicting Survival.
Nomogram models in predicting the prognosis of patients were constructed based on the independent risk factors for prognosis in Groups A and B (Figures 5(a) 5(c)). e AUC of the model in predicting prognosis within 5-7 years after radical gastrectomy in Group A was 0.726 (95% CI: 0.662-0.790), the sensitivity was 83.1%, and the specificity was 54.3% ( Figure 5(b)). e AUC of the model in predicting stage-I and -II GC patients prognosis within 2 years after radical gastrectomy in Group B was 0.848 (95% CI: 0.805-0.891), the sensitivity was 87.2%, and the specificity was 70.7% ( Figure 5(d)).

Discussion
In clinical practice, patients with a survival time >5 years are often considered not to need further follow-up and medical treatment, but there are still cases of recurrence and metastasis many years after tumor resection. According to Moon et al., the recurrence rate after long-term survival is about 10% [6]. In this study, the 5-7-year mortality rate due to GC was 18.7%. e high mortality rate may be related to the lack of follow-up after 5 years in our hospital, but it is also enough to show that clinicians should pay more attention to patients with a high risk of recurrence after 5 years. As early as 1982, Koga et al. [7] reported that 5.1% of patients had recurrence after 5 years of survival despite radical gastrectomy. Lee et al. [15] proposed that pathological stage T4a was a factor for predicting late recurrence and suggested that patients in T4a stage should be followed up and reexamined for longer after surgery.
is study proposes another viewpoint that may solve this problem. We found that F, SII, and pTNM stages were independent risk factors for death of GC patients 5-7 years after radical gastrectomy and could be used to construct a prognostic prediction model. e AUC was 0.726, the sensitivity was 83.1%, and the specificity was 54.3%. For similar studies, Katai et al. [8] pointed out that patients with a longer time between recurrence and surgery often had deeper infiltration of the primary tumor, and the main recurrence was always peritoneal metastasis. In our study, 59 patients died >5 years after surgery, and pT1 was seen in two cases (3.4%), pT2 in seven (11.9%), pT3 in seven (11.9%), and pT4 in 43 (72.9%). Huang et al. [16] reported that F could combine with carbohydrate antigen (CA)125 and CA19-9 to predict the risk of peritoneal metastasis in patients with GC after surgery. Recurrence after long-term survival may be related to the mechanism of tumor dormancy [17,18]. Early postoperative recurrence and metastatic dormancy are a particular mode of recurrence of malignant tumors. When the tumor cells break through the mechanical pressure and immune defense, they can enter the peripheral blood through the vascular endothelium and transform into circulating tumor cells (CTCs). However, some CTCs cannot adapt to the new microenvironment and the lack of adhesive factors and key signaling molecules can lead to tumor dormancy. With a change in the immune microenvironment of the tumor dormancy site, or another unknown stimulation, activated tumor cells can convert to  epithelial mesenchymal transition (EMT) through Smad2and β-catenin-related signaling pathways and develop powerful invasiveness and rapid proliferation, penetrate vascular tissue, and even progress to distant metastasis. In the present study, patients with high F-SII score had a significantly increased risk of distant dormant metastatic lesions, which should be paid more attention during postoperative follow-up and receive additional treatment [19,20]. With increased awareness of cancer prevention and improvement of detection rate of early GC through gastroscopy in recent years, the proportion of GC patients with early GC in developed countries such as South Korea has increased to 57.7%, and patients with stage-I and -II GC in Japan account for 72.3% [21][22][23]. According to epidemiological statistics, the recurrence rates of patients with stage-I and -II GC are about 10% and 30%, respectively [24]. At present, radical surgery, endoscopic mucosal resection (EMR), and endoscopic submucosal dissection (ESD) are alternative treatments for patients with GC invading the mucosa and submucosa [25][26][27]. However, developed countries such as South Korea and Japan are more inclined to ESD and EMR, while for developing countries in Southeast Asia and China, radical surgery is a more common is phenomenon may be due to the inevitable bleeding complications and residual tumor in gastrointestinal endoscopy in developing countries [28]. At the same time, these surgeons always choose radical resection of GC to avoid early occurrence of metastasis because it is difficult to find micrometastases by general imaging examination.

Journal of Oncology
With the increase in the proportion of patients with stage-I and -II GC, more attention should be paid to improving evaluation of these patients, along with appropriate treatment methods and prognosis of recurrence. In GC, an early stage is associated with long disease-free survival and OS, in addition to low risk of relapse and metastasis, which leads to neglect of their follow-up and treatment in clinical practice. erefore, there is an urgent requirement to find more sensitive markers for GC at an early clinical stage. To supplement the current staging system, we constructed a predictive model by screening for serum immune factors and combining with clinicopathological features. We found that F, SII, age, Borrmann type, and scope of gastrectomy were independent risk factors associated with the 2-year survival probability of stage-I and -II GC. e AUC of the constructed nomogram was 0.848,  Journal of Oncology the sensitivity was 87.2%, and the specificity was 70.7%. Previous studies have found that infiltration of lymphocytes and macrophages into tumor tissues increases significantly when gastric intraepithelial neoplasia develops into early GC [29]. is particular tumor environment also indirectly affects the immune status of the peripheral blood. For patients receiving proximal gastrectomy, the probability of postoperative acid reflux is significantly increased [30]. Patients receiving total gastrectomy also show poor prognosis due to 5%-19% weight loss and malnutrition [31,32]. erefore, the selection of surgical approach for GC will significantly affect the quality of life and survival rate, which needs to be carefully considered. According to SEER Medicare database, only 72.4% of patients with stage-IB GC and 50.6% of patients with stage-II GC received surgical treatment [33]. erefore, we suggest that nutritional support, follow-up observation, and adjuvant chemoradiotherapy should be emphasized in patients with high  nomogram model score after surgery. Similar conclusions have been confirmed by Datta et al. [34], who found significantly improved survival probability of patients with stage-IB and -II GC. In our study, high F-SII score indicated a high risk of early invisible metastasis, and more attention should be paid to improving evaluation of these patients. Another major concern is the impact of Helicobacter pylori on the occurrence and development of GC. Since this study is a retrospective analysis, most patients admitted to the hospital for surgical treatment were not routinely tested for H. pylori, which led to the missing data. However, it is undeniable that H. pylori plays an important role in peripheral blood immunity, tumor microenvironmental immune infiltration, and tumor progression [35]. According to data in the literature, approximately 1% to 3% of infected patients in the world will develop GC, while the proportion of noninfected patients is 0.13% [36]. For GC patients combined with H. pylori infection, anti-H. pylori treatment should be performed after surgery. A recent study pointed out that clarithromycin and levofloxacin have better anti-H. pylori efficacy. However, high BMI has been shown to be associated with high eradication failure rates [37]. erefore, the weight management should start after surgery, based on guidelines for lipid screening and blood pressure determination. Moreover, allium vegetables and their components have recently been intensively studied in digestive system tumors. Allium and its constituents have anti-inflammatory, immunomodulating, and anticancer effects, which have been observed in many in vivo and in vitro experiments [38].
ese substances may be used as an early prevention of the recurrence of GC, which may reduce the morbidity and mortality risks to a certain extent, while at the same time being more cost-effective. erefore, in addition to traditional drug treatment, postoperative patients should also be given dietary guidance.
In this study, we found that F and SII were independent risk factors for predicting prognosis and mortality whether in the early or subsiding stage of GC, which shared some similarities with our previous study [39].
e similarities between the two conclusions should be verified. Neutrophils, lymphocytes, and platelets are likely to be important for tumor recurrence and tumor dormancy [40,41]. F, which plays a role in coagulation, cell adhesion, and inflammation, also promotes tumor progression and distant metastasis and is considered to be a potential molecular marker of various malignant tumors [42][43][44][45]. Mucosal epithelial cells can produce F by inflammatory stimuli. F may convert to insoluble fibrin and exert a considerable influence on cancer cell progression. Similar studies have demonstrated that serum platelets and F enhance metastasis of tumor cells by impeding intravascular tumor cell clearance by natural killer (NK) cells. At the same time, platelet expression of aIIbβ3 may interact with tumor cells through an F bridge and contribute to metastasis. In the immune response of peripheral blood, lymphocytes kill CTCs and inhibit distant metastasis. Proliferating neutrophils can secrete cytokines such as interleukin-10 and tumor necrosis factor-α, inhibit activity of lymphocytes (CD4 + and CD8 + T cells) and NK cells, and further promote the immune escape of CTCs. At the same time, tissue destruction and nonspecific inflammatory reaction were shown to be dependent on neutrophil activity around cancer cells. e cascade effect caused by neutrophils is considered to promote the distant implantation and metastasis of CTC. e predictive accuracy of this traditional AJCC staging system relies on the assumption that the risk of tumor recurrence decreases with survival time, and tumor progression depends only on the characteristics of the tumor cells. erefore, it is inevitable that this staging system pays less attention to cancer in early stage or subsiding stage. anks to the promotion of Real-World Research and Big Data for Cancer Research [46], we can integrate and comprehensively analyze the clinicopathological characteristics and immune status of these two groups of patients. e predictive model based on immune response, pathological stage, and clinical characteristics can predict the risk of postoperative recurrence in patients who are difficult to assess [47]. Our study was based on the clinical limitations of the TNM staging system and proposed a more accurate individualized predictive model, which is worthy of further promotion and validation in clinical practice.
is was a retrospective study that had some limitations. Firstly, this study focused on Asian patients in a single center, and whether the results are widely applicable to white or black populations needs further study. Secondly, the recurrence patterns of the two groups, such as primary recurrence, peritoneal recurrence, or distant metastasis and H. pylori infection, could not be completely recorded, which also needs to be explored in future research.

Conclusion
In summary, we developed predictive models based on F and SII combined with clinicopathological features to predict early mortality of stage-I and -II GC, as well as patients with cancer regression. is study suggests that radical resection and appropriate postoperative treatment can be performed for patients with high probability of recurrence of stage-I and -II GC. Further follow-up and imaging examination should be carried out for patients with high probability of recurrence whose survival time is > 5 years. At the same time, proper weight management and dietary guidance are also essential for patients.

Data Availability
e datasets used in this study are available from the corresponding author on reasonable request. More information can also be obtained from the Gastric Cancer Information Management System v1.2 of Harbin Medical University Cancer Hospital (copyright no. 2013SR087424, http://www. sgihmu.com/).

Ethical Approval
All programs followed were according to the ethical standards of the Human Subjects Responsibility Committee (institutions and countries), as well as the 1964 Helsinki Declaration and subsequent editions.
is research was approved by Harbin Medical University Cancer Hospital Ethics Committee of (approval no. SHGC-1029).