Lymphocyte subset is more suitable than systemic inflammatory response biomarker and immunoglobulin in constructing prognostic nomogram model for advanced gastric cancer

Abstract

The serum factors of inflammation are known to be useful prognostic indicators of gastric cancer (GC). However, few studies have made comparisons to screen out more suitable biomarkers for the construction of Nomogram models. In this study, 566 patients who underwent radical gastrectomy were randomly selected. We evaluated the prognostic value of markers of systemic inflammation, including WBC, NLR, PLR, circulating total T cells, CD4⁺T cells, CD8⁺T cells and CD19⁺B cells, serum IgA, IgM, IgE and IgG, and compared them with traditional tumor markers (CEA, CA19-9, CA72-4 and CA125). Kaplan‒Meier analysis was used to analyze the correlation between biomarkers and overall survival (OS). We used time-dependent ROC analysis to investigate the prognostic accuracy of each biomarker. The risk of death was evaluated by the Cox regression model, and the Nomogram model was constructed by R software. We found that circulating total T cells, CD8⁺T cells, CEA, and CA125 had statistical significance in predicting advanced GC prognosis. Circulating CD8⁺T cells and CA125 were continuously superior to circulating total T cells and CEA in the prediction of 5-year OS. Cox regression found that CA125, circulating CD8⁺T cells, sex, and lymph node metastasis rate were independent risk factors for advanced GC. Furthermore, we combined all these predictors to construct a nomogram, which can supplement the AJCC 8th system. According to the comparison with commonly used serum immune biomarkers, circulating CD8⁺T cells is more sensitive to advanced GC. The prediction function of the Nomogram will supplement the traditional AJCC system, which contributes to individual survival prediction.

1. Introduction

In East Asia, Eastern Europe, and South America, gastric cancer (GC) is an important cause of cancer deaths, with more than nine hundred thousand confirmed cases and more than seven hundred thousand deaths each year [1,2]. More than 70% of new cases and deaths come from developing countries, and 50% are from Asia, which suggests that GC is re-emerging as a serious public health issue [3,4]. Although molecular targeted therapy and immunotherapy have shown successes in recent years, the 5-year survival rate of patients with advanced GC is still lower than 10% [5,6]. At present, these traditional classification tools based on tumor burden, presence of cancer cells in regional lymph nodes, and metastasis provide limited information in estimating prognosis and ameliorating the survival rate of patients with GC individually [7]. Therefore, it is time to develop simple, inexpensive, rapid, robust, and powerful serum biomarkers to guide the clinical treatment direction and predict prognosis, especially for advanced patients.

Inflammation is a major driving factor in tumor development and progression, both in the blood and tumor microenvironment. Many recent studies have focused on the prognostic and predictive value of different immune cell types. From 2012 to 2018, Galon [8] first proposed to include TNM immunity (TNM-I) in the current tumor staging standard, and UICC included the infiltration degree of immune cells in the tumor microenvironment in colon cancer pathological staging [9]. However, due to the high heterogeneity of GC, traditional pathological immunity evaluation by immunohistochemical methods is defective in detecting the immune infiltration of pathological sections. The immune response of peripheral blood to cancer cells is an extremely sensitive defense system. Our previous studies have shown that the advantage of the peripheral blood immune response in the early diagnosis of GC is higher than that of traditional tumor markers [10]. It has also been reported that cancer cells in the tumor microenvironment can activate immune cells to different degrees when they enter peripheral blood. In addition, immune cells in the tumor microenvironment are related to the systemic inflammatory response [11,12]. Therefore, tumor progression and immune reactions are interrelated, and a comprehensive assessment of these factors is critical.

Innate immune cells, adaptive immune cells, and immunoglobulin in peripheral blood are commonly used to evaluate the immune response of patients [13]. Innate immune cells mainly include neutrophils and lymphocytes; adaptive immune cells mainly include T lymphocytes and B lymphocytes; immunoglobulin mainly includes IgG and IgM, which play an important role in humoral immunity, and IgE and IgA, which participate in the local immune response [[14], [15], [16], [17]]. In recent years, the NLR and PLR have also been shown to be suitable for predicting the prognosis of GC. Through the retrospective analysis of nearly 2000 cases of radical gastrectomy, Kim et al. [18] found that the NLR is more predictive of overall survival than the PLR. With the development of immunotherapy, circulating T cells and B cells have attracted the attention of researchers [19,20]. Yoon et al. [21] suggested that this simple and powerful peripheral blood biomarker could not only predict the prognosis but also the sensitivity of immunotherapy. However, in clinical practice, it is still difficult to find a certain immune biomarker of peripheral blood that can achieve reliable predictive value to be widely used. Benefiting from the progress of systems biology and a large amount of clinical data, it is expected to find more accurate indicators from current immune indicators to integrate the clinical pathological variables of patients to achieve a higher individual prediction value.

To address these questions, we retrospectively analyzed patients who underwent radical gastrectomy at Harbin Medical University Cancer Hospital between October 2014 and June 2015. We evaluated the prognostic value of markers of systemic inflammation, including WBC, NLR, PLR, circulating total T cells, CD4⁺T cells, CD8⁺T cells and CD19⁺B cells, serum IgA, IgM, IgE and IgG, and compared them with the traditional tumor markers carcinoembryonic antigen (CEA), carbohydrate antigen 19–9 (CA19-9), carbohydrate antigen 125 (CA125), and carbohydrate antigen (CA72-4). Furthermore, we screened more sensitive immune biomarkers for predicting the prognosis of patients with advanced GC and combined them with clinical characteristics to construct a clinical prediction model.

2. Materials and methods

2.1. Patient characteristics

In this study, 566 GC patients who underwent radical gastrectomy were randomly selected at Harbin Medical University Cancer Hospital between October 2014 and June 2015. None of the patients had tumor invasion of the surrounding tissues. The diagnosis was based on the pathological report after surgery. All patients received an abdominal ultrasound, gastric computed tomography/magnetic resonance imaging, chest X-ray, and electrocardiogram, and some patients received positron emission tomography when necessary. Patients were classified according to the American Joint Committee on Cancer (AJCC)/Union for International Cancer Control (UICC) 8th staging system. The exclusion criteria were as follows: (a) preoperative radiotherapy and/or chemotherapy, (b) abdominal, pulmonary, intestinal, and other systemic infections, (c) severe cardiovascular disease, (d) antiplatelet therapy in three months, (e) steroid therapy in three months, (f) recurrent GC, (g) hematological malignancies and multiple myeloma, (h) severe cirrhosis, and (i) distant metastasis. More detailed medical records are included in the Gastric Cancer Information Management System v1.2 of Harbin Medical University Cancer Hospital (Copyright No. 2013SR087424, http://www.sgihmu.com/).

2.2. Ethics approval statement

This study was approved by the Ethics Committee of the Harbin Medical University Cancer Hospital (Approval Number: SHGC-1029). All experiments were performed in accordance with established ethical guidelines, and informed consent was obtained from the patients to confirm their voluntary participation.

2.3. Laboratory examination

Routine clinical laboratory analyses of peripheral blood samples were collected on the day of admission or the next day's morning. Two milliliters of peripheral blood collected from the elbow vein was used to ascertain leukocyte counts, neutrophil counts, lymphocyte counts, circulating total T cells, CD4⁺T cells, CD8⁺T cells and CD19⁺B cells. A Mindray automatic blood cell analyzer (BC-6800plus) was used for blood cell and platelet analysis, and original matching reagents (M − 68 P LH, M − 68 P DR, M − 68 P LD, M − 68 P LN, M − 68 P FN, M − 68 P FD, M − 68 P FR) and calibration reagents were used. The lymphocyte subsets were detected by flow cytometry (BD Canto II) with the detection reagent BD Multitest 6-Color TBNK Agent. The automatic blood count test was completed within 4 h. CEA, CA125, CA19-9, CA72-4, and IgG (20,152,400,549, Beckman Coulter, Inc.) were measured by enzyme-linked immunosorbent assay (ELISA); IgA (20,152,400,361, Beckman Coulter, Inc.), IgE (20,172,402,239, Beckman Coulter, Inc.), and IgM (20,152,402,660, Beckman Coulter, Inc.) levels were measured by enzyme immunoassay. The NLR was calculated by dividing the absolute neutrophil count (ANC) by the absolute lymphocyte count (ALC); the PLR was calculated by dividing the absolute platelet count (PLT) by the ALC.

2.4. Statistical analysis

We compared two data groups by t-test for continuous variables and the chi-square test for categorical variables. Continuous variables are expressed as the mean ± standard deviation. The Kaplan‒Meier method was used to analyze the correlation between biomarkers and overall survival probability, and the log-rank test was used to compare survival curves. We used R software version 4.0.1 and the “survivalROC” package to investigate the prognostic accuracy of each biomarker by time-dependent receiver operating characteristic (ROC) analysis. The cutoff value was defined to classify the patients into two groups (high vs. low) for each biomarker by ROC curve for survival in the five years, and the maximum value of sensitivity - (1-specificity) in ‘Youden index’ is the best cutoff value. The multivariable survival analysis was performed using the Cox regression model. The Nomogram model of risk assessment was constructed by the ‘SvyNom’ and 'rms' packages of R software. SPSS version 22.0 (SPSS Inc., Chicago, IL, USA) was used for data analysis and two-tailed P values < 0.05 were considered statistically significant.

3. Results

3.1. General patient characteristics

Table 1 shows the clinicopathological characteristics of the general patients, including sex, age, body mass index (BMI), tumor location, histological type, pTNM stage, vascular invasion, neural invasion, and lymph node metastasis ratio. Among the 566 patients, 406 (71.73%) patients were male, and 160 (28.27%) patients were female. Most of the patients were nonelderly patients (90.81%); nearly 70% of the tumors were located in the lower third of the stomach. According to the AJCC 8th staging system, 193 (34.10%) were stage I, 179 (31.63%) were stage II, and 194 (34.28%) were stage III. The median follow-up time was 51.45 months (range: 3.60–60.00 months).

Table 1.

Baseline characteristics in general gastric cancer patients.

Variables	Number of patients	Percentage of patients (%)
Sex
Male	406	71.73
Female	160	28.27
Age (years)
＜70	514	90.81
≥70	52	9.19
BMI (kg/m2)
<24	321	56.71
≥24	245	43.29
Tumor location
Lower third	396	69.96
Middle third	125	22.08
Upper third	42	7.42
Entire stomach	3	0.53
WHO classification
Well to moderately differentiated	258	45.58
Poorly differentiated	145	25.62
Signet ring cell	133	23.50
Mucinous	30	5.30
pTNM stagea
I	193	34.10
II	179	31.63
III	194	34.28
Vascular invasion
Absent	399	70.49
Present	167	29.51
Neural invasion
Absent	276	48.76
Present	290	51.24
Lymph node ratio (%)
0	262	46.29
>0 to ≤0.3	216	38.16
>0.3 to ≤0.6	67	11.84
>0.6	21	3.71

BMI Body Mass Index.

^aBased on the eighth edition of the AJCC Cancer Staging Manual of the American Joint Committee on Cancer.

3.2. Systemic inflammatory factors for risk of death in general patients

According to ROC analysis, NLR and PLR had statistical significance in predicting the prognosis (P = 0.010 and P < 0.001). The areas under the curve (AUCs) were 0.571 (95% CI: 0.518–0.625) and 0.609 (95% CI: 0.554–0.663), respectively. Furthermore, the Yodel index showed that 2.363 and 136.540 were the cutoff values. The sensitivities were 41.33% and 50.67%, respectively. The specificities were 72.84% and 69.23%, respectively (Table 2a, Fig. 1A).

Table 2.

aRelevant results of immune risk factors and tumor markers in general gastric cancer patients.

Factors	AUC	Cut-off value	Sensitivity	Specificity	P-value
WBC	0.541				0.141
NLR	0.571	2.363	41.33%	72.84%	0.010
PLR	0.609	136.540	50.67%	69.23%	< 0.001
Total T cells	0.552				0.057
Serum CD4	0.536				0.193
Serum CD8	0.592	18.850	77.33%	38.94%	0.001
Serum CD19	0.563	10.450	51.20%	62.00%	0.021
Serum IgM	0.563	1.145	30.05%	82.00%	0.022
Serum IgA	0.531				0.262
Serum IgG	0.537				0.182
Serum IgE	0.532				0.252
CEA	0.563	3.595	31.33%	84.13%	0.023
CA19-9	0.571	20.710	34.00%	81.49%	0.010
CA72-4	0.595	3.510	53.33%	63.94%	0.001
CA125	0.605	10.060	59.33%	60.82%	< 0.001

WBC White Blood Cell Count, NLR Neutrophil-to-Lymphocyte Ratio, PLR Platelet-to-Lymphocyte Ratio.

Fig. 1.

ROC curves of serum immune biomarkers and tumor markers in GC. WBC White Blood Cell Count, NLR Neutrophil-to-Lymphocyte Ratio, PLR Platelet-to-Lymphocyte Ratio.

3.3. Lymphocyte subsets for risk of death in general patients

According to ROC analysis, circulating CD8⁺T cells and CD19⁺B cells had statistical significance in predicting the prognosis (P = 0.001 and P = 0.021). The AUCs were 0.592 (95% CI: 0.538–0.645) and 0.563 (95% CI: 0.510–0.616), respectively. Furthermore, the Yodel index showed that 18.850 and 10.450 were the best cutoff values. The sensitivities were 77.33% and 51.20%, respectively. The specificities were 38.94% and 62.00%, respectively (Table 2a, Fig. 1B).

3.4. Immunoglobulin for risk of death in general patients

According to ROC analysis of immunoglobulin, IgM had statistical significance in predicting the prognosis (P < 0.022). The AUC was 0.563 (95% CI: 0.511–0.615). Furthermore, the Yodel index showed that 1.145 was the best cutoff value. The sensitivity was 30.05%, and the specificity was 82.00% (Table 2a, Fig. 1C).

3.5. Tumor markers for risk of death in general patients

According to ROC analysis, CEA, CA125, CA19-9 and CA72-4 had statistical significance in predicting prognosis (P = 0.023, P = 0.010, P = 0.001 and P < 0.001). The AUCs were 0.563 (95% CI: 0.506–0.609), 0.571 (95% CI: 0.515–0.627), 0.595 (95% CI: 0.543–0.647) and 0.605 (95% CI: 0.550–0.659), respectively. Furthermore, the Yodel index showed that the best cutoff values were 3.595, 20.710, 3.510, and 10.060. The sensitivities were 31.33%, 34.00%, 53.33%, and 59.33%, respectively. The specificities were 84.13%, 81.49%, 63.94%, and 60.82%, respectively (Table 2a, Fig. 1D).

3.6. Survival analysis of systemic inflammatory factors in general patients

Survival analysis of the general patients according to the systemic inflammatory factors NLR and PLR revealed that both NLR and PLR were negatively correlated with survival (P = 0.001 and P < 0.001). The survival time of patients with a high NLR was 40.93 ± 20.12 months, and the five-year survival rate was 61.6%; the survival time of patients with a low NLR was 56.53 ± 18.48 months, and the five-year survival rate was 75.0% (Fig. 2A). The survival time of patients with high PLR was 41.62 ± 19.69 months, and the five-year survival rate was 59.1%; the survival time of patients with low PLR was 60.00 ± 18.47 months, and the five-year survival rate was 77.4% (Fig. 2B).

Fig. 2.

Overall survival probability of patients with gastric cancer stratified by serum immune biomarkers and tumor markers. WBC White Blood Cell Count, NLR Neutrophil-to-Lymphocyte Ratio, PLR Platelet-to-Lymphocyte Ratio.

3.7. Survival analysis of lymphocyte subsets in general patients

According to the survival analysis of circulating CD8⁺T cells was negatively correlated with survival (P < 0.001), and circulating CD19⁺B cells was positively correlated with survival (P = 0.004). The survival time of patients with high circulating CD8⁺T cells was levels was 47.52 ± 19.64 months, and the five-year survival rate was 65.6%; the survival time of patients with low circulating CD8⁺T cells was 60 ± 17.50 months, and the five-year survival rate was 80.7% (Fig. 2C). The survival time of patients with high circulating CD19⁺B cells was 60.00 ± 18.86 months, and the five-year survival rate was 76.5%; the survival time of patients with low circulating CD19⁺B cells was 43.65 ± 19.12 months, and the five-year survival rate was 65.9% (Fig. 2D).

3.8. Survival analysis of immunoglobulin in general patients

According to the survival analysis of IgM, there was a negative correlation between IgM and survival (P < 0.001). The survival time of patients with high IgM was 47.52 ± 19.64 months, and the five-year survival rate was 65.6%; the survival time of patients with low IgM was 49.30 ± 19.47 months, and the five-year survival rate was 67.2% (Fig. 2E).

3.9. Survival analysis of tumor markers in general patients

According to the survival analysis of CEA, CA19-9, CA72-4, and CA125, the levels of all tumor markers were negatively correlated with survival (P < 0.001). The survival time of high CEA patients was 36.60 ± 20.13 months, and the five-year survival rate was 55.5%; the survival time of low CEA patients was 53.57 ± 18.74 months, and the five-year survival rate was 74.9% (Fig. 2F). The survival time of high CA19-9 patients was 33.59 ± 20.03 months, and the five-year survival rate was 56.1%; the survival time of low CA19-9 patients was 60.00 ± 18.55 months, and the five-year survival rate was 75.1% (Fig. 2G). The survival time of high CA72-4 patients was 47.52 ± 20.24 months, and the five-year survival rate was 61.9%; the survival time of low CA72-4 patients was 57.10 ± 18.19 months, and the five-year survival rate was 77.0% (Fig. 2H). The survival time of high CEA patients was 42.65 ± 20.58 months, and the five-year survival rate was 61.6%; the survival time of low CEA patients was 60.00 ± 17.38 months, and the five-year survival rate was 78.2% (Fig. 2I).

3.10. Serum immunity factors for risk of death in advanced patients

There were 194 patients (34.28%) with stage III GC. According to ROC analysis, circulating total T cells, CD8⁺T cells, CEA and CA125 were significantly different (P = 0.034, P = 0.014, P = 0.041, and P = 0.021, respectively). For these values, the AUCs under the ROC curve were 0.589 (95% CI: 0.509–0.669), 0.603 (95% CI: 0.522–0.683), 0.585 (95% CI: 0.505–0.666), and 0.596 (95% CI: 0.516–0.676), respectively. Furthermore, the Yodel index showed that 69.500, 18.600, 1.105, and 10.755 were the cutoff values, respectively. The sensitivities were 56.19%, 80.00%, 80.00%, and 58.10%, respectively, and the specificities were 62.92%, 43.82%, 37.08%, and 60.67%, respectively (Table 2b, Fig. 3).

Table 2b.

Relevant results of immune risk factors and tumor markers in advanced gastric cancer patients.

Factors	AUC	Cut-off value	Sensitivity	Specificity	P-value
WBC	0.500				1.000
NLR	0.542				0.318
PLR	0.578				0.063
Total T cells	0.589	69.500	56.19%	62.92%	0.034
Serum CD4	0.506				0.895
Serum CD8	0.603	18.600	80.00%	43.82%	0.014
Serum CD19	0.554				0.195
Serum IgM	0.566				0.116
Serum IgA	0.500				0.995
Serum IgG	0.569				0.097
Serum IgE	0.504				0.919
CEA	0.585	1.105	80.00%	37.08%	0.041
CA19-9	0.547				0.256
CA72-4	0.519				0.650
CA125	0.596	10.755	58.10%	60.67%	0.021

WBC White Blood Cell Count, NLR Neutrophil-to-Lymphocyte Ratio, PLR Platelet-to-Lymphocyte Ratio.

Fig. 3.

ROC curves of serum immune biomarkers and tumor markers in advanced GC. WBC White Blood Cell Count, NLR Neutrophil-to-Lymphocyte Ratio, PLR Platelet-to-Lymphocyte Ratio.

3.11. Survival analysis of serum immune factors in advanced patients

According to the survival analysis of circulating total T cells, CD8⁺T cells, CEA, and CA125 in patients with stage III GC, these factors were negatively correlated with survival (P = 0.006, P = 0.001, P = 0.013, and P = 0.001). The survival time of patients with a high total circulating T cells was 22.47 ± 18.56 months, and the five-year survival rate was 28.4%; the survival time of patients with low circulating total T cells was 32.15 ± 19.75 months, and the five-year survival rate was 49.1% (Fig. 4A). The survival time of patients with high circulating CD8⁺T cells levels was 23.29 ± 18.27 months, and the five-year survival rate was 29.7%; the survival time of patients with low circulating CD8⁺T cells levels was 36.07 ± 20.51 months, and the five-year survival rate was 60.7% (Fig. 4B). The survival time of high CEA patients was 23.61 ± 19.15 months, and the five-year survival rate was 33.9%; the survival time of low CEA patients was 34.32 ± 19.56 months, and the five-year survival rate was 53.1% (Fig. 4C). The survival time of high CA125 patients was 20.51 ± 18.39 months, and the five-year survival rate was 29.1%; the survival time of low CA125 patients was 33.96 ± 19.34 months, and the five-year survival rate was 49.0% (Fig. 4D).

Fig. 4.

Overall survival probability of patients with advanced gastric cancer stratified by serum immune biomarkers and tumor markers.

3.12. Nomogram model for predicting the prognosis of advanced patients

The time-dependent ROC curve of circulating CD8⁺T cells and CA125 was continuously superior to that of total T cells and CEA in the prediction of 5-year OS postoperatively (Fig. 5A). By Cox regression analysis, it was found that sex, circulating CD8⁺T cells, CA125, and lymph node metastasis rate were independent risk factors for the prognosis of patients with stage III GC (P = 0.006, P = 0.018, P = 0.049 and P < 0.001, Table 3). To calculate the death risk of patients individually, a nomogram model was constructed by combining the bivariate factor sex and continuous variable factors circulating CD8⁺T cells, CA125, and lymph node metastasis rate (Fig. 5B). The AUC of the nomogram model in predicting the 3-year prognosis was 0.672 (95% CI: 0.597–0.747) (Fig. 5C). The AUC of the nomogram model in predicting the 3-year prognosis was 0.710 (95% CI: 0.638–0.782) (Fig. 5D).

Fig. 5.

A The time-dependent ROC curve of the independent serum biomarkers. B Nomogram of the Cox regression model. C ROC curve of the combined factors in predicting the prognosis for 3 years. D ROC curve of the combined factors in predicting the prognosis for 5 years.

Table 3.

Selected variables by Cox regression analysis in advanced gastric cancer patients.

Variables	OR	95% CI	P value
Sex			0.006
Male	reference
Female	3.046	1.387–6.692
Age (years)	0.995	0.960–1.030	0.759
BMI (kg/m2)	1.015	0.916–1.124	0.775
NLR	1.219	0.931–1.596	0.149
PLR	0.994	0.988–1.000	0.052
Serum CD4	0.991	0.951–1.032	0.652
Serum CD8	0.948	0.907–0.991	0.018
CA125	0.966	0.933–1.000	0.049
CA72-4	1.006	0.998–1.019	0.326
Serum IgE	1.001	1.000–1.002	0.235
Serum IgG	0.990	0.96–.058	0.762
Tumor location			0.297
Lower third + Entire stomach	reference
Middle third	1.941	0.827–4.555	0.127
Upper third	1.375	0.520–3.637	0.521
WHO classification			0.993
Well to moderately differentiated	reference
Poorly differentiated	0.942	0.400–2.220	0.891
Signet ring cell	0.887	0.373–2.111	0.786
Mucinous	0.878	0.242–3.176	0.842
Lymph node ratio (%)	0.031	0.005–0.212	< 0.001
Vascular invasion			0.742
Absent	reference
Present	1.123	0.562–2.245
Neural invasion			0.248
Absent	reference
Present	1.792	0.666–4.821

BMI Body Mass Index, NLR Neutrophil-to-Lymphocyte Ratio, PLR Platelet-to-Lymphocyte Ratio.

3.13. Comparison of survival analysis between the nomogram model and TNM staging system at three years

According to the prognosis of the patients in three years, we divided the nomogram model into low (81), medium (46), and high (67) groups by the “Youden” index. The cutoff values were 61.32 and 76.66, respectively. ROC analysis showed that the AUC of the TNM staging system was 0.595 (95% CI: 0.515–0.675), less than the AUC of the nomogram model: 0.680 (95% CI: 0.604–0.756) (Fig. 6A).

Fig. 6.

A ROC curve of the AJCC 8th staging system and Nomogram model in predicting the prognosis for 3 years. B ROC curve of the AJCC 8th staging system and Nomogram model in predicting the prognosis for 5 years. C Overall survival probability of patients with advanced gastric cancer stratified by AJCC 8th staging system. The red line is IIIC patients, the green line is IIIB patients and the blue line is IIIA patients. D Overall survival probability of patients with advanced gastric cancer stratified by Nomogram model. The red line is high points patients (>76.66), the green line is medium points patients (>46.39, ≤76.66) and the blue line is low points patients (>0, ≤46.39).

3.14. Comparison of survival analysis between the nomogram model and TNM staging system at five years

According to the prognosis of the patients in five years, we divided the nomogram model into low (30), medium (97), and high (67) groups by the “Youden” index. The cutoff values were 46.39 and 76.66, respectively. ROC analysis showed that the AUC of TNM staging was 0.604 (95% CI: 0.525–0.684), less than the AUC of the nomogram model: 0.710 (95% CI: 0.638–0.782) (Fig. 6B). According to the survival analysis of the nomogram model, the survival time of the low-scoring nomogram model was 40.07 ± 17.77 months, and the five-year survival rate was 73.7%; the survival time of the medium-scoring nomogram model was 32.20 ± 20.57 months, and the five-year survival rate was 47.0%; the survival time of the high-scoring nomogram model was 19.50 ± 15.04 months, and the five-year survival rate was 13.4% (Fig. 6C). According to the survival analysis of the TNM staging system, the survival time of patients in stage IIIA was 31.33 ± 19.20 months, and the five-year survival rate was 45.8%; the survival time of patients in stage IIIB was 23.97 ± 19.86 months, and the five-year survival rate was 39.6%; the survival time of patients in stage IIIC was 20.27 ± 17.39 months, and the five-year survival rate was 23.2% (Fig. 6D). The scoring system of the nomogram model based on circulating CD8⁺T cells combined with CA125, sex, and lymph node metastasis rate was more significant for the prognosis of stage III GC than the current TNM staging system.

4. Discussion

GC is a common malignant tumor in developing countries, and most patients have advanced stages. At present, radical surgery is recognized as the major method to improve prognosis. However, the impact of tumor burden and surgery often leads to damage to the immune defense and changes in the inflammatory response. It always leads to the low immunity of patients, who are easily infected. On the other hand, the changes in the immune system also show clinical prediction biomarkers, which are expected to evaluate tumor progression and predict the survival rate after surgery. Considering the important role of the immune response in tumors, it is imperative to initiate the incorporation of tumor immunity as a component of cancer classification and a prognostic tool.

Researchers initially realized that tumor immunity can be developed by enteral nutrition to improve the survival of patients after surgery. This kind of immunonutrients plays an important role in regulating the immune system. Omega-3 fatty acids can improve immune function, promote operative effects, and reduce inflammatory reactions [22]. Klek et al. [23] noted that immunonutrients can improve the short-term survival rate of GC in stage IV and contribute to the regulation of inflammation and the enhancement of systemic immunity. In addition, other studies have also concluded that immunonutrition formula can increase the levels of immune cytokines in peripheral blood, such as CD4þ, CD4þ/CD8þ, IgG, and IgM [24]. However, the predictive value of serum immune biomarkers for tumors has not received attention from clinicians until recent years [25].

The immune system recognizes normal cells with malignant tendencies [26]. Innate and adaptive immune systems often have lethal effects in the early stage of malignant cell transformation [27]. At present, research on tumor immune biomarkers mainly focuses on peripheral blood and the tumor microenvironment, including macrophages, neutrophils, mast cells, dendritic cells, natural killer cells, and T and B lymphocytes. Similar studies have shown that tumor-related T lymphocytes, tumor-related macrophages, and tumor-infiltrating leukocytes, which can recognize tumor-specific antigens, accumulate in tumor tissues [28,29]. Although these immune cells have immune activity, the immune response to cancer cells seems to be dysfunctional and always promotes tumor progression [30,31]. Because of the correlation between tumor molecular features and immune reactions, this kind of immune damage seems to have been regarded as a biomarker to predict prognosis. In 2018, the AJCC/UICC included the immunoscore in the tumor microenvironment in the traditional pathological stage of colon cancer for the first time. The prediction of tumor-related immune cells on prognosis has been a popular topic in domestic and international studies [9]. However, due to the higher heterogeneity of GC, the difference in detection technology, the complex participation of immune factors in immune responses, and the inconsistency of paraffin section selection criteria, it seems difficult to propose a common or unified immune prediction model. Therefore, it is possible to obtain more accurate cancer information from blood samples by using relatively convenient detection methods with uniform measurement standards.

There have been preliminary studies on the clinical significance of serum immune biomarkers in gastrointestinal cancers. For example, NLR and PLR are considered to be independent prognostic factors of liver cancer and are highly correlated with postoperative recurrence of colon cancer [32]; serum IgG and IgM of anti-CEA have better diagnostic significance than CEA [33], and the scoring system combining the serum CD4/CD8 ratio with CA125 can further predict the prognosis of patients with advanced pancreatic cancer [34]. However, because of the high heterogeneity of GC, although serological immune biomarkers for prognosis and chemosensitivity are also of concern, they do not meet the requirements of clinical application. For clinical patients in our region, traditional tumor markers are still widely used, although they cannot achieve a satisfactory diagnosis. CEA is a polysaccharide-protein complex that mainly exists in the rectum, colon cancer tissue, and embryonic intestinal mucosa. As a widely used tumor marker, CEA is closely related to gastrointestinal malignancies, such as colorectal cancer, pancreatic cancer, gastric cancer, esophageal cancer, and gallbladder cancer. The positive rates of CEA in malignant tumors are colon cancer (70%), gastric cancer (60%), pancreatic cancer (55%), lung cancer (50%), breast cancer (40%), ovarian cancer (30%), and uterine cancer (30%). CA125 is a glycoprotein antigen secreted by coelomic epithelial cells, and its content is highly correlated with the staging of epithelial ovarian cancer. Serum CA125 is effective in diagnosing advanced ovarian cancer but is not effective in diagnosing early stage. The specificity of CA125 is poor, and it will increase to varying degrees in cervical cancer, breast cancer, pancreatic cancer, gastric cancer, lung cancer, and colorectal cancer. CA19-9 is a related tumor marker for pancreatic cancer, gastric cancer, colorectal cancer, and gallbladder cancer. Many studies have proven that the level of CA19-9 is related to the size of these tumors, and it is the most sensitive marker for pancreatic cancer. The positive rate of CA72-4 in the diagnosis of gastric cancer was 65–70%. The level of CA72-4 was significantly correlated with the stage of gastric cancer. For patients with gastric cancer metastasis, the serum level of CA72-4 is much higher than that of nonmetastatic patients. Cytokines are involved in all immune responses in a complex manner [35]. When cytokines directly or indirectly interfere with each other, it is difficult for researchers to choose one cytokine to represent the progression of all GC patients unless patients are grouped according to clinical characteristics. Meanwhile, there are few studies in the same group of patients with GC to compare a variety of serum immune indicators to screen out more sensitive markers.

Therefore, this study is the first to compare all commonly used immunology-related serum biomarkers with traditional tumor markers and analyze their significance in predicting the prognosis of GC. First, we found that NLR, PLR, circulating CD8⁺T cells, CD19⁺B cells and IgM have similar significance to traditional tumor markers in general patients, which is similar to most of the current research. However, when we analyzed advanced GC patients with a lower 5-year survival rate, we found that only circulating total T cells and CD8⁺T cells, CEA, and CA125 had significance. Combined with our previous studies, we found that changes in the NLR and PLR are more sensitive to the diagnosis of early-stage GC than tumor markers [10]. For advanced GC, the change in lymphocyte subsets has a more obvious effect on the survival period. Given the low 5-year survival rate (less than 35%) [36], the examination of lymphocyte subsets is worth popularizing. As an index of autoimmunity, lymphocyte subsets are an important means to evaluate the immune function of tumor patients. It can evaluate the level of cellular immunity and humoral immunity, reflect the current immune function, state, and balance level, and assist in diagnosing certain diseases (such as autoimmune diseases, immunodeficiency diseases, malignant tumors, blood diseases, and allergic diseases). When the number or ratio of lymphocyte subsets changes abnormally, the body will have a series of pathological changes and immune disorders, which can also reflect the progress of the disease. To some extent, the occurrence of tumors is due to low immune function in the body and decreased ability to kill tumor cells. However, the process of tumor treatment is a process of killing and immune function reconstruction. That is, killing tumor cells, increasing the attack of the immune system, and killing more tumor cells. After long-term antitumor treatment, the immune system of patients will be affected, and lymphocyte subsets may change, leading to immunodeficiency disease or autoimmune disease, increasing the probability of malignant tumor metastasis and virus infection and even shortening the life of patients. Therefore, it is of great significance to detect the peripheral blood lymphocyte subsets of patients to formulate further treatment plans. At the same time, our study also found that although the changes in the immune system response in the peripheral blood are more significant than tumor markers, the types of immune cells are different for each stage of GC. Therefore, it is necessary to discuss the immune factors combined with the tumor stage for prognosis prediction, which may be helpful to improve the accuracy of clinical application.

For further clinical application, we screened CA125 and circulating CD8⁺T cells by time-dependent ROC and constructed a prediction model with independent factors related to the prognosis for advanced GC, including sex and lymph node metastasis rate. In contrast to other studies, we used CA125, circulating CD8⁺T cells, and lymph node metastasis rate as continuous variables to construct the nomogram model, which may score each patient more accurately and estimate the risk of death for three and five years. At the same time, for patients with advanced gastric cancer, we did not include the TNM stage (IIIA, IIIB, IIIC) in the multivariate analysis but chose the lymph node metastasis rate, which is more sensitive for the prognosis of advanced gastric cancer, as an important consideration. Compared with the AJCC staging system, the accuracy of the nomogram model in predicting 3-year and 5-year survival rates was significantly better, which will supplement the AJCC 8th system. Similar to the current studies, we included sex, which is considered to be highly related to the immune system response, in the evaluation factors [37,38], and previous studies also showed that serum inflammatory biomarkers were more valuable for the early diagnosis of GC in men than in women. This kind of nomogram prediction model, which combined the immune response and clinicopathological characteristics, has clinical application prospects in a variety of gastrointestinal tumors [39,40].

Although the current nomogram model shows high accuracy in its prediction function, some limitations of this study must be considered. GC is one of the most widely distributed cancers, which makes it difficult to reach a general conclusion in a single center [41]. At the same time, all the patients in our study were Asian race. Whether the results of this study can be extended to the European and American population needs further study. In addition, this study did not discuss the influence of Helicobacter pylori infection and erosive gastritis on serum immune factors. However, some current research on “noninflammatory tumors” such as breast cancer and thyroid cancer shows that the NLR and PLR are also related to tumor progression, which may indicate that cancer cells entering peripheral blood cause an immune response without requiring local inflammation [42,43]. Inevitably, due to the retrospective nature of this study, some important data from the included patients might have inevitably been missing, thus reducing the number of enrolled cases. For more developed countries and with the improvement of diagnostic level, the proportion of early stage gastric cancer patients is increasing. Future research should pay more attention to the changes of early cancer immune status. The cooperative research between departments of endoscopic and gastrointestinal surgery will get twice the result with half the effort. Although our research has a long way to go to supplement international guidelines for predicting the prognosis of GC, it does not prevent studies from applying inflammatory biomarkers as a method for the screening and identification of high-risk patients. We suggest that further studies should not only focus solely on finding peripheral cancer cells or their “secretion” as a prognostic tool but also include the peripheral blood immune response caused by tumor cells in the staging system for GC. What's more, different stage patients have different levels of tumor markers expression and immune response and the exploration of tumor markers should distinguish the stages of patients, rather than generalize all patients.

5. Conclusions

We compared the significance of commonly used serum immune biomarkers for prognosis and screened out circulating CD8⁺T cells, which is more sensitive to advanced GC with a low survival rate, to construct a nomogram model with CA125, lymph node metastasis rate, and sex. The prediction function of the nomogram will contribute to supplementing the traditional AJCC staging system, which will contribute to individual survival prediction.

Author contribution statement

Tianyi Fang; Xin Yin: Conceived and designed the experiments; Analyzed and interpreted the data; Wrote the paper.

Yufei Wang: Conceived and designed the experiments; Analyzed and interpreted the data.

Lei Zhang; Yimin Wang; Xuan Lin: Contributed reagents, materials, analysis tools or data.

Xinghai Zhang; Xudong Zhao: Analyzed and interpreted the data.

Yingwei Xue: Conceived and designed the experiments; Wrote the paper.

Funding statement

This work was supported by Nn10 program of Harbin Medical University Cancer Hospital from Yingwei Xue (No. Nn10 PY 2017-03), Haiyan Research Fund of Harbin Medical University Cancer Hospital from Tianyi Fang (No. JJQN2021-06) and Fundamental Research Funds for the Provincial Universities from Tianyi Fang (No. 2022-KYYWF-0286).

Data availability statement

Data will be made available on request