Background

Gastric cancer (GC) is the sixth most common cancer and the second leading cause of cancer-related death worldwide [1]. Despite the recent advances in surgical procedures and chemotherapy, recurrence is still reported in a large number of patients with GC even after curative resection. Therefore, to improve the survival outcomes of GC, it is important to identify patients with a high risk of GC recurrence who need to be treated with adequate adjuvant treatments and meticulous surveillance even after curative gastrectomy.

Since inflammation and the immune system contribute to the development and progression of malignancies [2, 3], parameters that reflect a host’s immune-inflammatory status may be useful prognostic indicators of GC. Neutrophil (Neut), lymphocyte (Lymp), monocyte (Mono), and platelet (Plt) counts in the peripheral blood were previously shown to be convenient and helpful prognostic indicators of GC [4,5,6]. A systemic immune-inflammation index (SII), readily calculated by the combination of three complete blood count (CBC) parameters (Neut, Lymp, and Plt), has recently been attracting increasing attention as a powerful prognostic indicator in several malignancies, such as esophageal, colon, and pancreatic cancers [7,8,9]; however, the clinical utility of SII in GC remains unclear.

In the present study, the clinical significance of preoperative SII for predicting the postoperative survival outcomes of GC was compared with that of each CBC parameter (Neut, Lymp, Mono, or Plt). The relationship between preoperative SII and the type of GC recurrence was also examined. The aim of the present study was to establish a novel perioperative care system according to preoperative SII levels.

Methods

Patients

A total of 447 patients who underwent surgery for GC between January 2008 and June 2013 at the Kyoto Prefectural University of Medicine were included in the present study. The inclusion criteria were as follows: (1) patients pathologically diagnosed with gastric adenocarcinoma; (2) patients with pathological stages (pStage) I, II, and III; and (3) patients undergoing the curative resection of GC (R0). Patients who received neoadjuvant chemotherapy; those who underwent emergency surgery; those who underwent non-curative resection (R1/R2); those with hematological, chronic inflammatory, or autoimmune diseases that may affect preoperative SII values; and those with active infection or inflammatory diseases within 1 month before the blood examination were excluded. Patients with missing information on preoperative CBC and those with simultaneous malignancies other than GC were also excluded. Gastrectomy with lymph node dissection was performed based on the Japanese Gastric Cancer Treatment Guidelines (JGCTG) [10]. Tumor staging was performed according to the 8th edition of the Tumor, Node, Metastasis staging classification by the Union for International Cancer Control [11]. In the present study, grade 2 or higher postoperative complications according to the Clavien-Dindo Classification [12, 13] occurred in 60 patients (13.4%), and only one patient (0.22%) died of postoperative complications.

All patients provided written informed consent before surgery. The present study was approved by the Ethics Committee of the Kyoto Prefectural University of Medicine (approval no. ERB-C-1327).

Data collection of preoperative CBC and SII

Preoperative CBC (Neut, Lymp, Mono, and Plt) was collected from patients within 7 days before surgery, and SII was calculated as Plt (cell/mm3 × 103) × Neut (cell/mm3)/Lymp (cell/mm3).

Clinicopathological findings

The following clinicopathological data were reviewed from the medical record database of our institution: age, sex, body mass index (BMI), physical status (PS), comorbidities (hypertension, diabetes mellitus, heart disease, and chronic renal failure), tumor location, preoperative serum carcinoembryonic antigen (CEA), cancer antigen (CA) 19-9, albumin, C-reactive protein (CRP), preoperative CBC (Neut, Lymp, Mono, and Plt), pathological T stage (pT), pathological N stage (pN), lymphatic invasion, venous invasion, and tumor differentiation.

Adjuvant treatment and postoperative follow-up

After curative gastrectomy for GC, patients with pStage I generally received postoperative examinations alone without any adjuvant treatments. Patients with pStage II and III were basically treated with adjuvant chemotherapy based on JGCTG. The main chemotherapeutic regimens administered were S-1 alone [14], S-1 plus oxaliplatin (SOX) [15], or capecitabine plus oxaliplatin (CapeOX) [16], in line with JGCTG [10]. All patients were followed up at regular intervals with serum CEA and CA19-9; computed tomography (CT) of the chest, abdomen, and pelvis; and upper gastrointestinal endoscopy following the JGCTG. Most patients were postoperatively followed up for 5 years or until their death. GC recurrence was confirmed by imaging, such as CT and upper gastrointestinal endoscopy. If possible, recurrence was histologically confirmed via surgical biopsy, needle biopsy, or appropriate fluid cytology. Peritoneal recurrence was diagnosed by imaging alone, and diagnostic laparotomy was rarely performed.

Statistical analysis

Differences between the two groups for categorical and continuous variables were analyzed by the chi-squared test and Mann-Whitney U test, respectively. The optimal cutoff value for each immune-nutritional parameter (Neut, Lymp, Mono, Plt, or SII) was selected according to the receiver operating characteristic (ROC) curve for overall survival (OS) with the maximal Youden index based on the sum of sensitivity and specificity [17, 18]. The cutoff values for serum albumin and CRP were set at 3.5 g/dl and 0.5 mg/dl, respectively, with reference to the modified Glasgow Prognostic Score, which is regarded as a useful prognostic indicator in various malignancies [19, 20]. OS and recurrence-free survival (RFS) were generated using the Kaplan-Meier method, and the differences between the two groups were assessed with the log-rank test. Parameters showing significance in univariate analyses were further assessed using multivariate Cox’s models. We used two types of multivariate Cox’s models. In one model (model #1), Neut, Lymp, or Plt was separately incorporated as explanatory variables. In the other model (model #2), SII was incorporated as an explanatory variable instead of Neut, Lymp, and Plt. Hazard ratios (HR) and 95% confidence intervals (CI) were subsequently calculated. A p value < 0.05 was considered to be significant. Statistical analyses were performed with the software package JMP software version 10 (JMP, Cary, NC, USA).

Results

Clinical characteristics

The clinicopathological characteristics of 447 patients are shown in Table 1. The median age was 67 years (range 29–89), with 289 male (64.7%) and 158 female patients (35.3%). The median (range) values were 3690 (1250–15,850) for Neut, 1780 (570–4770) for Lymp, 330 (50–1400) for Mono, 22.5 × 104 (8.7 × 104–57.0 × 104) for Plt, and 474 (105–4455) for SII.

Table 1 Clinicopathological characteristics of the patients
Full size table

ROC curve analysis

The area under the curve (AUC) value showing the predictive power of SII for 5-year OS was 0.650, and the Youden index was 0.238. As shown in Fig. 1, the optimal cutoff value for preoperative SII was 395 (sensitivity 41.0%, specificity 82.8%). All patients were then divided into the high SII (SII ≥ 395) and low SII (SII < 395) groups. The optimal cutoff values of other preoperative CBC parameters for predicting 5-year OS are shown in Additional file 1: Figure S1 (3690 for Neut (AUC 0.612), 1860 for Lymp (AUC 0.587), 320 for Mono (AUC 0.523), and 27.2 × 104 for Plt (AUC 0.581)).

Fig. 1
figure 1

ROC curve analysis of preoperative SII for predicting OS in patients with GC

Full size image

Relationship between preoperative SII and clinicopathological factors

The differences in clinicopathological characteristics between the high and low SII groups are shown in Table 1. Preoperative Neut, Lymp, Mono, and Plt were significantly higher (p < 0.001, < 0.001, = 0.008, and < 0.001, respectively), and BMI was significantly lower (p = 0.021) in the high SII group than in the low SII group. In addition, hypoalbuminemia (albumin < 3.5), pN+, positive lymphatic invasion, and undifferentiated adenocarcinoma were frequent among patients in the high SII group (p = 0.014, 0.003, 0.004, and 0.022, respectively). No significant differences were observed in age, sex, PS, comorbidities, tumor location, preoperative CEA, CA 19-9, and CRP, pT, or venous invasion between the two groups.

Prognostic significance of preoperative SII

Survival curves for OS stratified by preoperative SII and each CBC parameter (Neut, Lymp, Mono, and Plt) are shown in Fig. 2A and Additional file 2: Figure S2, respectively. The 5-year OS rate of the high SII group was 80.0%, which was significantly worse than that (92.7%) of the low SII group (p < 0.001). Univariate and multivariate survival analyses of OS are shown in Table 2. In the univariate analysis, PS ≥ 2, hypertension, heart disease, upper GC, CEA ≥ 5, CA19-9 ≥ 37, albumin< 3.5, CRP ≥ 0.5, pT4, pN+, positive lymphatic invasion, positive venous invasion, Neut ≥ 3690, Lymp < 1860, Plt ≥ 27.2 × 104, and SII ≥ 395 correlated with poor OS. In multivariate Cox’s model #1, Neut ≥ 3690 (HR 1.97; 95% CI 1.08–3.68; p = 0.027), heart disease, CRP ≥ 0.5, pT4, and pN+ were identified as independent risk factors for poor OS. In multivariate Cox’s model #2, SII ≥ 395 (HR 2.95; 95% CI 1.49–6.39; p = 0.001), heart disease, CRP ≥ 0.5, pT4, and pN+ were identified as independent risk factors for poor OS.

Fig. 2
figure 2

Survival curves of patients stratified by preoperative SII. A OS. B RFS

Full size image
Table 2 Univariate and multivariate survival analyses for OS
Full size table

Similar to the results of analyses of OS, the 5-year RFS rate of the high SII group was 77.7%, which was significantly worse than that (90.2%) of the low SII group (p = 0.001) (Fig. 2B). The multivariate Cox’s model identified SII ≥ 395 as an independent risk predictor of 5-year RFS (HR 2.36; 95% CI 1.31–4.48; p = 0.004) (Additional file 3: Table S1).

Relationship between preoperative SII and the type of GC recurrence

GC recurrence was detected in 51 out of 447 patients. Cumulative recurrence rates stratified by preoperative SII were examined according to the type of GC recurrence (peritoneal, hematogenous, and lymph node recurrence) (Fig. 3). Peritoneal recurrence was more frequent in the high SII group than in the low SII group (p = 0.028), whereas no significant difference was observed in the cumulative recurrence rate for hematogenous or lymph node recurrence between the high and low SII groups (p = 0.248 and 0.096, respectively).

Fig. 3
figure 3

Cumulative recurrence rate for each recurrence pattern stratified by preoperative SII. A Peritoneal recurrence. B Lymph node recurrence. C Hematogenous recurrence

Full size image

Discussion

In the present study, we examined the clinical significance of preoperative SII to predict postoperative survival outcomes in GC. The optimal cutoff value of preoperative SII for predicting OS was set at 395 according to the ROC curve analysis. The results obtained demonstrated that SII, a combination of three CBC parameters (Neut, Lymp, and Plt), was an independent risk factor for poor OS and RFS as well as pT and pN. Of special note was that high SII correlated with peritoneal recurrence. These results suggested that preoperative SII may contribute to perioperative precise care and adjuvant treatments for patients with GC undergoing curative gastrectomy.

In the present study, in addition to Neut, Lymp, Mono, and Plt, preoperative serum albumin and BMI correlated with preoperative SII. A relationship was previously suggested between obesity and chronic inflammation [21]; therefore, patients with high BMI may have a stronger inflammatory response. However, the present study showed that BMI was significantly lower in the high SII group. In cancer patients, inflammation is induced by inflammatory cytokines as cancer progresses, and, thus, patients with a high inflammatory response may lose weight [22]. In the present study, high SII correlated with low albumin levels, suggesting that cachexia had an influence on the results obtained. Hirahara et al. also reported that BMI was significantly lower in the high SII group [23].

Neutrophilia, lymphopenia, and thrombocytosis each affect a host’s immune-inflammatory status and tumor progression and have been implicated in poor cancer survival [24,25,26,27,28,29]. In addition, several indexes calculated by combining these factors, such as neutrophil-to-lymphocyte ratio, platelet-to-lymphocyte ratio, and monocyte-to-lymphocyte ratio, have been used to predict survival outcomes of GC [30,31,32]. Accordingly, high SII, resulting from neutrophilia, lymphopenia, and thrombocytosis, may also be a useful prognostic indicator. In the present study, univariate survival analyses showed that Neut, Lymp, and Plt were significant prognostic factors in patients with GC; however, the multivariate survival analysis (model #1) only identified Neut ≥ 3690 (HR 1.97; 95% CI 1.08–3.68; p = 0.027) as an independent risk factor of poor OS. The multivariate survival analysis (model #2) also identified SII ≥ 395 (HR 2.95; 95% CI 1.49–6.39; p = 0.001) as an independent risk factor of poor OS, with a higher HR and lower p value than those of Neut in model #1. The AUC value showing the predictive power of SII for OS (AUC 0.650) was higher than that (AUC 0.612) of Neut.

The negative impact of postoperative complications on survival outcomes has recently been clarified [33, 34]. Accordingly, worse OS may be attributed to a higher incidence of postoperative complications. However, in the present study, the incidence of postoperative complications of Clavien-Dindo grade II or higher in the high SII group was 13.9% (39/280), which was not significantly different from that (21/167, 12.6%) in the low SII group (p = 0.684, data not shown). Correlations between preoperative inflammatory indices and the occurrence of postoperative complications remain controversial [35, 36], and, thus, further studies are needed to confirm these relationships.

Previous studies demonstrated the negative impact of high preoperative SII on the survival outcomes of GC [23, 37, 38]; however, the cutoff values of SII differed between these studies. Wang et al. [37] reported that the optimal cutoff value of SII for predicting OS was 660 (AUC 0.612) according to the ROC curve analysis, and SII ≥ 660 was an independent predictor of OS in GC patients. Hirahara et al. [23] showed that the optimal cutoff value of SII for predicting OS was 661.9 (AUC 0.584) according to the ROC curve analysis, and SII ≥ 661.9 was an independent prognostic indicator in GC patients, particularly in the elderly population. In the present study, although an ROC curve analysis was also used, the optimal cutoff value of SII for predicting OS was 395, which was markedly lower than 600. Differences in the pStage of GC patients may affect SII cutoff values. Although Wang et al. only targeted pStage III GC patients who had undergone R0, R1, or R2, the present study examined pStage I, II, and III patients undergoing R0 [37]. Similar to the present results, Shi et al. only targeted patients with pStage I–III GC who had undergone R0 and demonstrated that the optimal SII cutoff value was 320 (AUC 0.64) and that SII > 320 was an independent prognostic indicator in these patients [38]. Therefore, further studies are needed to validate the optimal cutoff value of SII.

The present study demonstrated that preoperative SII was a risk factor for poor RFS independent of pT and pN and that SII ≥ 395 correlated with GC recurrence, particularly peritoneal recurrence. To the best of our knowledge, this is the first study to demonstrate the impact of high SII on the specific recurrence pattern of GC. A previous study showed that neutrophils promoted the migration and invasion of GC cells by activating the ERK pathway and inducing epithelial-mesenchymal transition [27]. Other studies indicated that lymphopenia and thrombocytosis were predictive factors for peritoneal dissemination [39, 40]. In addition, Nakamura et al. [41] reported that the preoperative neutrophil-to-lymphocyte ratio was a significant independent predictor of the presence of peritoneal metastasis during staging laparoscopy in patients with advanced GC. Since high SII results from neutrophilia, lymphopenia, and thrombocytosis, it may also be a useful predictor of peritoneal recurrence. In the present study, from a pathological viewpoint, pN+ and undifferentiated adenocarcinoma were often observed in the high SII group. Peritoneal recurrence was also frequently detected in patients with pN+ (pN+, 25/124; pN−, 3/320; p < 0.001) and undifferentiated adenocarcinoma (differentiated type, 10/231; undifferentiated type, 18/216; p = 0.081); therefore, the incidence of peritoneal recurrence may be high in patients with high SII.

According to JGCTG, the current indication of postoperative adjuvant chemotherapy is assessed only by pStage (pT and pN). However, the present results suggest that patients with high preoperative SII may also be considered for adjuvant chemotherapy. In Western countries, perioperative chemotherapy (preoperative plus postoperative FLOT) is the standard treatment [42, 43]. Although postoperative adjuvant chemotherapy alone is the standard treatment in Eastern Asia, its effects are inadequate in many cases; therefore, patients with poor prognostic factors, such as high SII, may be good candidates for neoadjuvant chemotherapy. Perioperative interventions, such as immunonutrition, may be another treatment option for improving the survival outcomes of patients with high SII. However, there is currently no evidence to show that perioperative immunonutrition actually improves the long-term outcomes of GC patients. It also currently remains unclear whether preoperative immunonutrition reduces morbidity rates after gastrectomy for GC [44,45,46]. Xin et al. [47] has recently shown by a systematic review and a meta-analysis that enteral nutrition feeding tube support is an essential intervention to elevate patients’ immunity, depress levels of inflammation, and reduce the risk of complications after gastrectomy for GC. Therefore, further research and the development of novel treatment methods that fundamentally suppress cancer-induced inflammation and boost host immunity are desired.

The present study had several limitations. It was a retrospective study with a small sample size from one institution, which may have limited its statistical power and generated statistical biases. Furthermore, since the cutoff values of Neut, Lymp, Mono, Plt, and SII were only calculated by a mathematical method, further validations are needed to confirm whether these values are clinically meaningful and applicable to other cohorts. In addition, the present study only examined CBC parameters; however, serum immune-nutritional markers may also be useful predictors of survival outcomes in GC. Dynamic changes in the host’s immune-inflammatory status during the perioperative period may provide more important information about survival outcomes, but the present study failed to evaluate dynamic changes in the SII values [30]. Relationships between tumor microenvironment and the mechanisms of immune dysfunction should be uncovered to develop promising immunotherapy and improve survival outcomes of GC patients [48,49,50]; however, no molecular biological examinations have been done in this study. Nevertheless, the present study clearly demonstrated the novel potential of preoperative SII for predicting postoperative survival outcomes and recurrence patterns in GC over that of each CBC parameter (Neut, Lymp, Mono, and Plt). The results of the present study and the optimal cutoff value of SII need to be validated in further studies with large sample sizes in order to establish a novel perioperative care system according to preoperative SII levels.

Conclusion

Preoperative SII may be a useful predictor of postoperative survival outcomes in GC. Appropriate postoperative adjuvant treatments and the meticulous surveillance of GC relapse, particularly peritoneal dissemination, are necessary for patients with SII ≥ 395 even after curative gastrectomy.