Introduction

Gastric cancer is one of the most common causes of cancer-related death in the world [1, 2]. Radical gastrectomy with regional lymph node dissection is the only potentially curative treatment for patients with advanced gastric cancer. Gastrectomy with D2 lymph node dissection is the standard of practice for locally advanced gastric cancer in both Western and Asian countries. However, lymph node dissection around major blood vessels is technically demanding and challenging for surgeons, and excessive intraoperative blood loss (IBL) can occur even at the hands of experienced surgeons.

The impact of IBL on long-term postoperative outcomes in patients with cancer, not just gastric cancer, has long been of interest in the field of surgical oncology [3,4,5,6,7]. The adverse effect of IBL on the prognosis of patients with gastric cancer was first reported by Dhar et al. [8] in 2000. Their retrospective analysis of 152 patients with transmural advanced gastric cancer found that IBL > 500 ml was an independent prognostic factor of long-term survival. Several subsequent works also identified excessive IBL as predictive of a worse prognosis with different cutoff values assigned by each [9,10,11,12,13], while others did not associate IBL with outcome [14,15,16]. The issue of IBL and prognosis, while well studied, is unsettled. Common limitations of previous studies were their retrospective, single-center nature. The topic lacks a reliable report that utilizes a prospective cohort in which the patients, the surgical procedure, postoperative adjuvant chemotherapy, and follow-up were strictly defined.

The Japan Clinical Oncology Group study 1001 (JCOG1001) was a large-scale, multicenter, randomized phase III trial created to assess the efficacy of bursectomy in patients with locally advanced gastric cancer [17]. A total of 1204 patients were registered in this trial and randomly assigned to bursectomy or nonbursectomy treatment. While the trial failed to demonstrate survival superiority after bursectomy to nonbursectomy for cT3/4 gastric cancer, its homogeneity makes it an important resource for additional analysis of this population. The objective of the present study was to evaluate the impact of IBL on long-term outcomes after curative gastrectomy using data from the JCOG1001 phase III trial.

Methods

Patients

Between June 1, 2010, and March 30, 2015, 1204 patients registered in the JCOG1001 were randomly assigned (1:1) during surgery for either an omentectomy alone (nonbursectomy) or a bursectomy (n = 602 for each group). Trial eligibility criteria and the method of randomization have been previously reported in detail [17]. In brief, patients 20–80 years of age with histologically proven and resectable primary gastric carcinoma with an estimated depth of cT3(SS) or cT4(SE) were enrolled. Patients with distant metastases, bulky lymph nodes or Borrmann type 4 or large (≥ 8 cm) type 3 carcinomas were excluded. Fifty-seven Japanese hospitals participated in the trial. JCOG1001 was registered with UMIN-CTR, number UMIN000003688. The study protocol of JCOG1001 was approved by the JCOG Clinical Trial Review Committee and the institutional review boards of all participating institutions. The procedures were conducted in accord with the ethical standards of the Helsinki Declaration of 1975.

At the second planned interim analysis on September 17, 2016, the JCOG Data and Safety Monitoring Committee independently reviewed the results of the trial and recommended early termination of the study for futility because overall survival was lower in the bursectomy group than the nonbursectomy group. The study was terminated early at that time.

Of the 1204 patients, 1 patient did not receive a gastrectomy because of peritoneal metastases were identified after randomization. The remaining 1203 patients were included in the present analysis.

Surgical procedures and postoperative treatment

All patients received either a total or distal gastrectomy by laparotomy with D2 lymph node dissection as dictated by the Japanese gastric cancer treatment guidelines (third edition) [18]. Those patients who required a total gastrectomy for cT2 or deeper tumors in the proximal third of the stomach also had their spleen removed to complete a splenic hilar lymphadenectomy. An omentectomy was done in both groups in this trial. In the bursectomy group only the peritoneal lining of the bursa omentalis was removed en bloc to the greatest extent possible from the anterior plane of the transverse mesocolon and the pancreas. The bursa omentalis peritoneal lining was preserved as much as possible in the nonbursectomy group. Postresection reconstruction type was not specified in the protocol.

IBL was measured according to the volume and weight of blood absorbed by surgical gauze and suction pumps during the gastrectomy. IBL was extracted from the operative and anesthesia records of each patient. Perioperative allogeneic blood transfusion was defined as the administration of blood cells from the start of surgery to the next morning. General indications for transfusion were hemoglobin concentration < 8.0 g/dl or hemodynamic changes during surgery, although transfusions were performed at the discretion of the anesthesiologist and the surgical team responsible for perioperative care.

Postoperative adjuvant chemotherapy with S-1 for 1 year was performed as a protocol treatment for patients who had complete tumor resection and pathologic Stage IIA, IIB, IIIA, IIIB or IIIC tumors except for pT1 and pT3 (SS) N0 tumors, according to the Japanese gastric cancer treatment guidelines (third edition) [18]. After protocol completion, no further treatment was given unless the tumor recurred. All clinical data were obtained from the JCOG1001 case report forms.

Follow-up

Postoperative follow-up was on a fixed schedule. A contrast-enhanced abdominal computed tomography was done every 6 months for 3 years, and yearly thereafter. Carcinoembryonic antigen and carbohydrate antigen 19–9 levels were drawn every 3 months for 3 years, and every 6 months thereafter. Upper endoscopy (only for distal gastrectomy cases) and chest X-ray were done every year. The median follow-up was 4.0 (interquartile range [IQR] 2.6–5.0) years, and the last follow-up date was November 29, 2016. Relapse-free survival (RFS) was the time from the date of randomization during surgery to the date of the first disease recurrence or death from any cause. Overall survival (OS) was the time from the date of randomization during surgery to the date of death from any cause. Data collection for patients who did not experience an event was completed on the date of the final observation.

Statistical analysis

Fisher’s exact test was performed for categorical variables and the Mann–Whitney U test was used for continuous variables. Survival was estimated using the Kaplan–Meier method, and differences between survival curves were compared using the log-rank test. The Cox proportional hazards model was used to evaluate the hazard ratio relative to each variable. Univariable and multivariable analyses with a Cox regression model were performed to identify variables that influenced RFS. Variables of interest included IBL (< 200/200 ≤ and < 400/ ≥ 400 ml), age (≤ 65/ > 65 years), sex (male/female), body mass index (BMI) (< 22.7/ ≥ 22.7 kg/m2), histological type (undifferentiated type/differentiated type), Eastern Cooperative Oncology Group performance status (PS) (0/1), surgical procedure (distal gastrectomy/total gastrectomy), combined organ resection except for cholecystectomy (performed/not performed), bursectomy (performed/not performed), extent of lymphadenectomy (≤ D2/ ≥ D2 +), operation time (< 238/ ≥ 238 min), blood transfusion (no/yes), tumor size (< 5.2/ ≥ 5.2 cm), pT (T1/2/3/4), pN (N0/1/2/3), and postoperative adjuvant chemotherapy (no/yes). All P values less than 0.05 were considered statistically significant. Statistical analyses were done with SAS version 9.2 or higher.

Results

Patient characteristics are shown in Table 1. A total gastrectomy was performed on 416 patients (34.6%). Fifty-six patients (4.7%) received blood transfusions. The median IBL was 285 ml (range 0–3068 ml). The first tertile (33.3 percentile) was 196 ml, and the second tertile (66.7 percentile) was 400 ml. On the basis of these tertiles, we divided all 1203 patients into 3 groups: IBL < 200 ml defined as small blood loss (SBL, n = 404), 200 ml ≤ IBL < 400 ml defined as medium blood loss (MBL, n = 393), and IBL ≥ 400 ml defined as large blood loss (LBL, n = 406). Median (range) IBL was 110 ml (0–199) in SBL, 283 ml (200–398) in MBL, and 575 ml (400–3068) in LBL. There were significant differences between the groups in age, sex, BMI, PS, surgical procedure, combined organ resection except for cholecystectomy, bursectomy, operative time, blood transfusions required, tumor size, and pathological T status. However, tumor histology, past surgical history, the extent of lymphadenectomy, pathological N status, pathological stage, and the proportion of patients receiving adjuvant chemotherapy were not different.

Table 1 Clinicopathological characteristics of SBL, MBL and LBL
Full size table

A total of 283 recurrences were reported. The most frequent first recurrent site was the peritoneum, followed by the lymph nodes and liver (Table 2). RFS curves of the three IBL groups are shown in Fig. 1. The 3-year RFS after SBL, MBL, and LBL were 81.7% (95% confidence interval [CI] 77.3–85.2), 74.8% (95% CI 70.0–79.0), and 70.6% (95% CI 65.7–74.9) respectively, which were significantly different between the three groups (log-rank two-sided P = 0.002). Compared with SBL as a reference, the hazard ratio (HR) for RFS of MBL was 1.351 (95% CI 1.016–1.796, P = 0.038) and that of LBL was 1.635 (95% CI 1.243–2.150, P < 0.001). The 3-year OS after SBL, MBL, and LBL were 88.9% (95% CI 85.2–91.7), 83.6% (95% CI 79.2–87.1), and 81.3% (95% CI 77.0–84.9), which were significantly different among the three groups (log-rank two-sided P = 0.014) (Online Resource 1).

Table 2 First recurrent sites
Full size table
Fig. 1
figure 1

Relapse-free survival by intraoperative blood loss

Full size image

Multivariable analysis revealed that IBL was an independent risk factor for recurrence, together with PS, pT status, pN status, and postoperative adjuvant chemotherapy. Bursectomy, blood transfusion, and combined organ resection except for cholecystectomy were not independent risk factors (Table 3). Compared with SBL as a reference, the HR of MBL was 1.461 (95% CI 1.088–1.962, P = 0.012) and that of LBL was 1.520 (95% CI 1.108–2.083, P = 0.009).

Table 3 Results of unevaluable and multivariable analysis for relapse-free survival (RFS)
Full size table

Discussion

The impact of IBL on the long-term outcome of gastric cancer patients has long been the focus of attention, and several retrospective data have been reported. We sought to elucidate whether IBL was associated with the long-term outcome of patients with locally advanced gastric cancer using data from the JCOG1001, a large-scale, multicenter, randomized phase III trial. The JCOG1001 was terminated early after the second interim analysis, which was conducted 1.5 years after the last patient was enrolled, and the follow-up period was not long enough. Therefore, multivariable analyses were performed using RFS as the endpoint, which can assess long-term outcomes in a shorter period. Our results indicated that an IBL ≥ 200 ml was significantly associated with a shorter RFS, and IBL was an independent risk factor for disease recurrence after curative gastrectomy. This is the first study to report a negative impact of IBL on the long-term outcome of patients with gastric cancer using data from a large-scale, multicenter, prospective cohort.

The median IBL for all 1203 patients was 285 ml, which was similar to what was reported by previous studies [10, 12, 13, 15]. Increased IBL was associated with male sex, higher BMI, larger tumor size, total gastrectomy, and cases where combined organ resection was required, correlations also similar to previous reports. We selected 200 ml and 400 ml as IBL cutoff points for our analyses because these were nice round numbers for surgeons to recognize and were also located around our predicted optimal cutoff value, which we estimated would be in the range of 200–500 ml as identified in previous studies [8,9,10,11,12,13]. IBL ≥ 200 ml was identified as an independent risk factor for reduced RFS, which is the lowest IBL threshold ever reported.

The JCOG1001 showed that a bursectomy did not provide a survival advantage over nonbursectomy in the treatment of patients with resectable advanced gastric cancer [17]. OS and RFS were slightly lower in the bursectomy group than the nonbursectomy group, and a pancreatic fistula was significantly more common in the bursectomy group than in the nonbursectomy group (5% vs. 2%; P = 0.032). We speculated that a possible cause of the negative results of this trial was the growth stimulation of residual cancer cells by inflammatory cytokines and growth factors induced by surgical stress or postoperative complications [17, 19]. In fact, in a prior exploratory analysis of the JCOG1001 we reported that postoperative complications adversely affected long-term survival [20]. On the other hand, the median IBL in the bursectomy group was 330 ml (IQR 183–530), which was 100 ml greater than that in the nonbursectomy group (230 ml [IQR 130–410]; P < 0.0001). One of the causes of the negative results in JCOG1001 might be the greater IBL in the bursectomy group.

There were several studies that reported on the adverse effects of IBL on long-term outcomes, for which they discussed several possible causes. One possible cause is the spillage of cancer cells into the peritoneal cavity due to significant bleeding during surgery. Peritoneal recurrence is thought to be caused by this mechanism. Kamei et al. [9] reported that an IBL of ≥ 475 mL was specifically associated with the development of a peritoneal recurrence after curative gastrectomy in 146 patients with advanced gastric cancer. Arita et al. [11] investigated 540 patients undergoing curative gastrectomy and reported that a high IBL correlated with a higher risk of peritoneal recurrence. Moreover, in the laboratory setting they confirmed that the ability of gastric cancer cells and mesothelial cells to adhere to each other was increased in the presence of blood plasma.

Second possible way that IBL can influence survival is through antitumor immunosuppression induced by excessive IBL. Bruns et al. [4] reported that the activity of natural killer cells was significantly decreased in patients with an IBL of more than 700 ml during upper gastrointestinal surgery compared to those who lost less than 500 ml. Miki et al. [21] reported that interleukin (IL)-6 and IL-6-triggered tumor growth factors were increased in patients with colorectal cancer receiving blood transfusion because of excessive IBL, which was associated with a poor prognosis. In several previous studies that reported a relationship between IBL and survival immunosuppression was thought to be the primary reason why bleeding was associated with recurrence, especially via hematogenous spread [9, 10, 13].

Third, perioperative blood transfusions and postoperative complications may adversely affect long-term outcomes as confounders of excessive IBL. In general, excessive IBL increases the need for blood transfusions, and patients with increased IBL are at greater risk of subsequent postoperative complications. Several studies have reported on the negative impact of blood transfusions and complications on RFS [14, 16, 22,23,24], and Nakanishi et al. reported in their review article that postoperative complications and blood transfusions may have a negative impact on long-term outcomes [25]. In the present study less than 5% of all patients required a blood transfusion, with 2.3% (9/393) of the MBL group requiring a transfusion even though it was above the IBL threshold (≥ 200 ml) predictive of a negative survival impact. In multivariable analysis, transfusion was not a significant prognostic factor (HR 0.919, P = 0.730). These results negatively indicate that transfusions may have affected long-term outcomes and support the conclusion that excessive IBL itself had an adverse effect on long-term outcome.

In contrast, in the present study, the group with greater IBL had more postoperative complications (Online Resource 2). In our other exploratory analysis, Clavien–Dindo (C–D) complications of Grade III and above were associated with poor RFS and were identified as the most suitable definition of complication for predicting negative long-term survival outcomes [20]. Excessive IBL may have induced complications, which in turn may have led to poor long-term survival. It is also possible that there were other confounding factors that may affect both excessive IBL and postoperative complications, such as surgical difficulty or tumor factors other than tumor size, pT or pN status. Further research is needed to answer these questions.

Reducing the incidence of postoperative complications may be more difficult for surgeons than reducing IBL. In the JCOG1001, 67% of the total patients had an IBL ≥ 200 ml, and 23% and 10% had C–D complications equal to or more than Grade II and Grade III, respectively. Since the number of patients with complications was not as large as those with greater IBL, and since surgeons are constantly careful and make efforts during surgery to avoid complications, it is never easy to improve long-term outcomes by further reducing complications. It is possible to reduce the IBL through the use of sealing devices, such as ultrasonically activated shears, [26, 27] or careful procedures even if they require longer operation time. Laparoscopic approaches including a robotic gastrectomy could also reduce IBL [28,29,30,31].

The present study has several limitations. First, it was unclear whether IBL adversely affects long-term survival in patients with poor prognostic factors, such as bulky lymph nodes or Borrmann type 4 or large type 3 carcinomas, which were excluded in the JCOG1001. Second, the mechanism behind RFS and excessive IBL has not been elucidated, and it is unclear whether IBL itself causes frequent relapses or adversely affects RFS through postoperative complications, or whether IBL is a surrogate for the other factors impacting RFS, although the present study showed that IBL during a curative gastrectomy was associated with more frequent recurrence. Further basic and clinical investigations to elucidate the mechanism are needed. Third, the present study did not include patients who underwent minimally invasive surgeries (MIS), such as a laparoscopic or robotic gastrectomy. MIS can be expected to reduce the amount of bleeding, and it is therefore of great interest to see whether MIS actually has less of an adverse effect on long-term outcomes due to reduced IBL.

In conclusion, the negative impact of IBL on long-term outcome after curative gastrectomy was confirmed by analyzing data from a large-scale prospective study. An IBL of 200 ml or more was significantly associated with poor long-term survival in this patient group. While external validation is required, we encourage surgeons to reduce IBL below 200 ml as much as possible to achieve a better outcome for patients with advanced gastric cancer.