Prognostic impact of postoperative systemic inflammatory response in patients with stage II/III gastric cancer

This study examined whether the systemic inflammatory response present in the early phase of the postoperative state correlates with long-term outcomes and to identify markers in patients with stage II/III gastric cancer. 444 consecutive patients who underwent radical gastrectomy for stage II/III gastric cancer were retrospectively reviewed. We evaluated maximum serum C-reactive protein (CRPmax) and white blood cell count (WBCmax), defined as the maximum serum CRP level and maximum WBC count during the interval from surgery until discharge, as systemic inflammation markers. In univariate analyses, CRPmax, WBCmax and infectious complications were significantly associated with both overall survival (OS) (p < 0.001, p < 0.001 and p = 0.011, respectively) and relapse-free survival (RFS) (p < 0.001, p = 0.001 and p < 0.001, respectively). Multivariate analysis revealed that high-CRPmax (> 9.2 mg/dL) was an independent prognostic factor for OS (hazard ratio (HR) 1.68, 95% confidence interval (CI) 1.19–2.36, p = 0.003) and RFS (HR 1.56, 95% CI 1.12–2.18, p = 0.009), while WBCmax and infectious complications were not. CRPmax, which reflects the magnitude of systemic inflammation induced by surgical stress and postoperative complications in the early phase after surgery, may be a promising prognostic indicator in patients with stage II/III gastric cancer who undergo curative resection.

Inflammation has been revealed to be closely related to the progression and metastasis of cancer cells 3 . Several studies have shown that postoperative infectious complications correlate with postoperative recurrence and poor prognosis in various malignancies, including gastric cancer [4][5][6][7] . Systemic inflammatory responses caused by infectious complications could induce not only the proliferation of residual cancer cells, but also declines in host immunity, which may lead to worsened prognosis 8,9 . On the other hand, in the postoperative early phase, surgical stress also induces systemic inflammatory responses. Several recent studies have focused on the association between long-term outcome and postoperative systemic inflammatory response caused by surgical stress and complications in gastroenterological cancer [10][11][12][13] . However, the significance of postoperative inflammatory response in the early phase after surgery and its markers in patients with gastric cancer remain unclear. In addition, appropriate markers of systemic inflammation that correlate with prognosis remain uncertain.
The aim of this study was to examine whether the systemic inflammatory response caused in the early phase of the postoperative state correlates with prognosis in advanced gastric cancer patients who undergo curative resection. We evaluated C-reactive protein (CRP) and white blood cell (WBC) count as simple and versatile markers of postoperative systemic inflammation in clinical practice that may be suitable for detecting the magnitude of inflammatory reaction.

Results
Predictive ability and cut-off values for CRP max and WBC max . Median CRP max and WBC max were 10.4 mg/dL (interquartile range [IQR], 7.3-14.9 mg/dL) and 11,000/mm 3 (IQR,9,,900/mm 3 ), respectively. Areas under the curve predicting 5-year OS were 0.615 for CRP max and 0.573 for WBC max , respectively. Values of 9.2 mg/dL for CRP max and 15,100/mm 3 for WBC max provided maximal Youden indices, and were thus set as the cut-off values. We classified 258 patients (58.1%) and 186 patients (41.9%) as having high-CRP max and low-CRP max , respectively, and 83 patients (18.7%) and 361 patients (81.3%) as having high-WBC max and low-WBC max , respectively.
Correlations between CRP max , WBC max and clinicopathological variables. Table 1 shows the associations between CRP max , WBC max and clinicopathological variables. CRP max was significantly associated with operative procedure, operation time, blood loss, tumor size, tumor location, and postoperative infectious complications (p < 0.001 each). WBC max was significantly associated with operative procedure, operation time, blood loss, tumor size, tumor location, and postoperative infectious complications (p < 0.001 each).
Thirteen patients were lost to follow-up within 5 years, with 16 months as the shortest follow-up period for survivors. Recurrence was observed in 162 cases, and median duration to recurrence was 15 months (IQR, 8-29 months). A total of 200 deaths were observed.
Five-year OS and RFS rates for the entire study population were 59.0% and 55.2%, respectively. OS and RFS rates in patients with high-CRP max were significantly poorer than those of patients with low-CRP max (p < 0.001 each) (Fig. 1). OS and RFS rates in patients with high-WBC max were also significantly poorer than those of patients with low-WBC max (p < 0.001 and p = 0.001, respectively) (Fig. 2).
Results of uni-and multivariate analyses for OS and RFS are summarized in Tables 2 and 3 Figure 3A and B show the Kaplan-Meier survival curves comparing OS for CRP max according to adjuvant chemotherapy. In patients with and without adjuvant chemotherapy, OS rates were significantly lower in the high-CRP max group than in the low-CRP max group (p = 0.002 and p < 0.001, respectively). Figure 3C and D show the Kaplan-Meier survival curves comparing OS for CRP max according to postoperative infectious complications. In patients without postoperative infectious complications, OS rates were significantly lower in the high-CRP max group than in the low-CRP max group (p < 0.001), whereas in the presence of postoperative infectious complications, no significant difference was observed between groups (p = 0.444); this was attributed to the fact that only 3 patients were included in the low-CRP max group. Therefore, we used another cut-off value of 26.2 mg/dL for patients with postoperative infectious complications, determined based on the same method used for the entire cohort. In patients with postoperative infectious complications, OS rates were significantly lower in this new high-CRP max (> 26.2 mg/dL) group than in the new low-CRP max (≤ 26.2 mg/dL) group (p = 0.015) (Fig. 4).

Subgroup analysis for OS according to adjuvant chemotherapy and postoperative infectious complications.
We classified patients according to CRP max and postoperative infectious complication status into three groups as follows: Group 1, high-CRP max with postoperative infectious complications; Group 2, either high-CRP max or the presence of postoperative infectious complications; and Group 3, low-CRP max without postoperative infectious www.nature.com/scientificreports/    Table 4. The proportion of patients with hematogenous recurrence was significantly higher in the high-CRPmax group (16.3%) and the high-WBCmax group (20.5%) than in the low-CRPmax group (3.8%, p < 0.001) and the low-WBCmax group (8.9%), respectively. Whereas no significant difference was evident between CRPmax and peritoneal recurrence, lymph node recurrence and local recurrence, and the same applied to WBC max .

Discussion
The present study investigated correlations between postoperative systemic inflammatory response and prognosis in advanced gastric cancer patients who underwent curative resection, using the maximum postoperative CRP level and WBC count during hospitalization after gastrectomy as markers of systemic inflammation. We found that CRP max with a cut-off of 9.2 mg/dL was an independent prognostic factor, but WBC max was not. Further, CRP max affected OS and RFS independent of postoperative infectious complications. These results suggest that systemic inflammatory response as represented by CRP max in the early phase after surgery induced by surgical stress and complications is associated with recurrence and survival in patients with stage II/III gastric cancer. This study revealed CRP max as an independent prognostic indicator for OS and RFS in patients with stage II/III gastric cancer who underwent curative resection. Consistent with our results, previous studies have demonstrated that postoperative systemic inflammatory response correlated with recurrence and poor prognosis in gastroenterological cancers including gastric cancer 11,12,14 , esophageal cancer 15,16 , and colorectal cancer 17 . However, the optimal marker remains uncertain. Because the magnitude of surgical stress and postoperative complications can differ markedly among cancers, optimal markers and the respective cut-off values should be determined for each cancer. Okumura et al. revealed postoperative prolonged hyperthermia, defined as a maximum body temperature > 38 °C for ≥ 4 days, as an independent prognostic factor for OS and RFS in patients with stage II/ III gastric cancer 11 . However, whether postoperative prolonged hyperthermia affected survival independent of the occurrence of postoperative infectious complications was unclear because postoperative complications were strongly associated with prolonged hyperthermia in that study. We evaluated WBC max and CRP max as markers that are not only commonly used to estimate the magnitude of inflammation in postoperative management, but also simple and readily available in daily clinical practice. Consequently, we found that CRP max was an independent prognostic factor for OS and RFS, but WBC max was not. Similarly, Saito et al. 12 demonstrated that postoperative CRP max (≥ 12 mg/dL) was an independent prognostic factor for RFS in advanced gastric cancer, although postoperative WBC max was not. Further, this finding was validated in their subsequent large-scale multicenter study 14 . Postoperative CRP max could provide a useful marker to predict prognosis in patients with stage II/III gastric cancer who undergo curative resection.
Interestingly, the high-CRP max group in the present study was significantly associated with hematogenous metastasis as the pattern of recurrence. Similar to our result, Kurokawa et al. reported that liver metastasis, but not peritoneal or lymph node metastasis, was significantly more frequent in the high-CRP max (≥ 12 mg/ dL) group than in the low-CRP max group. Although the exact mechanisms underlying this association between higher postoperative inflammation and hematogenous metastasis remain unclear, some potential explanations can be considered. First, host immunosuppressive influences such as impairment of cellular immunity caused by the surgical stress could negatively impact on circulating tumor cells (CTCs) and micro-metastatic cancer cells. Although natural killer cells and macrophages play important roles in eliminating CTCs and preventing the formation of metastases, both the cytotoxicity of natural killer cells and macrophage function are reportedly impaired in proportion to the extent and magnitude of surgery 18,19 . Extraperitoneal tumor growth was demonstrated to be accelerated accompanied by suppressed natural killer cell cytotoxicity after surgery 20 . Second, systemic inflammation could accelerate the adhesion of CTCs to distant organs. E-selectin up-regulation, induced by inflammatory cells and pro-inflammatory cytokines such as interleukin (IL)-1 and tumor necrosis factor (TNF)-α, has been shown to promote recruitment of CTCs to the vascular endothelium [21][22][23] . Third, growth factors and cytokines such as TNF-α, vascular endothelial growth factor and IL-6 induced by the inflammatory response could promote the proliferation and metastasis of residual cancer cells 24 . Thus, not only optimizing surgical procedures, but also immunomodulatory approaches and anti-inflammatory approaches in the perioperative period might improve oncological outcomes.
In the present study, postoperative infectious complications were not an independent prognostic factor for OS and RFS, despite significant associations with those outcomes in univariate analyses. Some studies have demonstrated that postoperative complications correlate with prognosis in gastric cancer patients 4,5 , whereas others have not 25,26 . This inconsistency may be attributable to differences in the definition and frequency of postoperative complications among studies. Moreover, most studies did not enter postoperative inflammatory responses (including surgical stress) into multivariate analyses for survival. Our findings suggest that systemic inflammatory response induced by not only postoperative complications, but also surgical stress is more important than the actual postoperative complications as a prognostic factor in patients with advanced gastric cancer.
In the present study, because the 71 patients with postoperative infectious complications included only three patients with high-CRP max , we evaluated another cut-off value of CRP max for patients with postoperative infectious complications. Consequently, patients with postoperative infectious complications and a high-CRP max of > 26.2 mg/dL revealed poorer OS than those with low-CRP max ≤ 26.2 mg/dL. This finding suggests that the magnitude of systemic inflammatory response caused by infection could affect OS. When postoperative infectious complications occur, early diagnosis and appropriate treatment for infectious complications may be important to reduce systemic inflammatory response and improve long-term outcomes in patients with stage II/III gastric cancer.
From Japan, some important surgical randomized controlled trials for advanced gastric cancer have been reported by the Japan Clinical Oncology Group (JCOG). The JCOG9501 study compared D2 lymphadenectomy alone with D2 lymphadenectomy plus para-aortic nodal dissection for advanced gastric cancer without clinical para-aortic lymph node metastasis 27 . The JCOG0110 study compared spleen preservation with splenectomy for advanced proximal gastric cancer not involving the greater curvature 28 . The JCOG1001 study compared nonbursectomy with bursectomy for advanced gastric cancer with cT3 (SS)-cT4b (SI) 29 . However, none of these studies demonstrated the prognostic efficacy of extended surgery. One reason why extended surgery uniformly failed to improve survival may be that the negative impact on residual cancer cells of the systemic inflammatory www.nature.com/scientificreports/ response involved in surgical stress and postoperative complications may offset any positive impact of extended surgery. Furthermore, the JCOG9502 study compared an abdominal-transhiatal approach with a left thoracoabdominal approach for advanced gastric cancer with esophageal invasion of ≤ 3 cm revealed worse survival in the left thoracoabdominal approach group 30 . In the JCOG9502 study, the rate of postoperative complications was higher in the left thoracoabdominal approach group and surgical stress was obviously larger in that same group. The impact of postoperative systemic inflammatory response induced by surgical stress and postoperative complications on survival may warrant more attention from general surgeons. This study has some potential limitations that should be considered when interpreting the results. First, this retrospective study was conducted at a single institution and sample size was relatively small. Second, the present study showed heterogeneity in the adjuvant chemotherapy regimens, because the indications for and standard regimens of adjuvant chemotherapy had not been established until 2007, when the results of the ACTS-GC trial confirmed the efficacy of S-1 as adjuvant chemotherapy for stage II/III gastric cancer 31 . However, in the subgroup analysis with or without adjuvant chemotherapy, OS rates were significantly lower in the high-CRP max group than in the low-CRP max group. Prospective large-scale validation studies are needed to confirm our findings.

Conclusion
CRP max , which reflects the magnitude of systemic inflammation induced by surgical stress and postoperative complications in the early phase after surgery, was associated with oncologic outcome in patients with stage II/ III gastric cancer who underwent curative resection. Our findings suggest that surgeons should not underestimate the prognostic impact of surgical stress and postoperative complications in the management of advanced gastric cancer. To improve long-term outcomes for advanced gastric cancer patients, reducing surgical stress and postoperative complications may be important.

Methods
Patients. This retrospective analysis investigated consecutive patients who underwent radical gastrectomy for gastric cancer at Osaka City University Hospital (Osaka, Japan) between January 2000 and December 2013. Patients diagnosed with stage II/III gastric cancer on postoperative pathological examination were enrolled in this study. We excluded 23 patients who underwent neoadjuvant chemotherapy, 33 patients with concomitant multiple cancers, 27 patients with R1/2 resection, 9 patients with specific histological type, 6 patients with perioperative death, and 13 patients for whom the full set of clinical data was not available. Ultimately, 444 patients were included in this study. The Osaka City University Ethics Committee approved this retrospective study of clinical data (approval no. 4386), which was conducted in accordance with the principles of the Declaration of Helsinki.
Data collection and evaluation of postoperative inflammatory response. We evaluated clinicopathological characteristics including age, sex, body mass index (BMI), Eastern Cooperative Oncology Group performance status (PS), blood test examination data, operative approach and procedure, operation time, blood loss, tumor size, depth of invasion, lymph node metastasis, pathological stage, tumor location, histology, macroscopic type, lymphatic invasion, venous invasion, postoperative infectious complication, and adjuvant chemotherapy from medical records. Tumors were staged according to the third English edition of the Japanese classification of gastric carcinoma 32 . Subjects were categorized according to age as elderly (≥ 75 years) and non-elderly (< 75 years), and according to BMI as underweight (BMI < 18.5 kg/m 2 ) and normal or overweight (BMI ≥ 18.5 kg/ m 2 ). In the present study, postoperative infectious complications included intra-abdominal abscess, anastomotic leakage, pancreatic fistula, pneumonia, surgical site infection, acute cholecystitis, and enteritis of Grade 2 or higher according to the Clavien-Dindo classification 33 .
Serum CRP and WBC count were measured routinely on postoperative day (POD) 1, POD 3, and POD 6 or 7. In addition, serum CRP and WBC count were assessed when postoperative complications were suspected and when treatment efficacy was evaluated. We defined the maximum serum CRP level (CRP max ) and maximum WBC count (WBC max ) as the highest levels of those parameters identified during the entire interval from surgery until discharge. To set cut-off values for CRP max and WBC max , time-dependent receiver operating characteristic (ROC) curve analyses for 5-year overall survival (OS) as the endpoint were calculated, and maximal Youden indices were estimated. All patients were classified as having high or low values according to these cut-offs. In the present study, postoperative infectious complications were defined as postoperative complications accompanied by elevated CRP needing antibiotics, drainage and surgery. To diagnose infectious complications, blood examination, measurement of amylase in drain fluid, and imaging studies, such as computed tomography, ultrasonography and contrast swallow, were performed based on clinical suspicion. The attending surgeons performed each examination and recorded the results. Consequently, we included 24 cases of anastomotic leakage, 24 cases of pancreatic fistula, 7 cases of intra-abdominal abscess, 7 cases of surgical site infection, 3 cases of pneumonia, 2 cases of acute cholecystitis, 1 case of acute pancreatitis, 1 case of enteritis, 1 case of urinary tract infection, and 1 case of catheter-related blood stream infection as Grade 2 or higher according to the Clavien-Dindo classification 15 .
Treatment and follow-up. Surgical procedures were determined according to tumor size, location, and the status of resection margins. Laparoscopic surgery was preferred when clinical stage was less than stage IB. In principle, adjuvant chemotherapy with oral fluoropyrimidines (5-fluorouracil, uracil-tegafur, doxifluridine, or S-1) was administered after obtaining written informed consent, except for patients with pathological T1. Patients were routinely followed-up every 4 months for the first 2 years, every 6 months for the next 3 years, and annually thereafter. Each follow-up examination included physical examinations, routine blood tests, measurements of tumor marker levels, and contrast-enhanced computed tomography of the abdomen. These same