Gastric cancer (GC) belongs to the most common malignant diseases worldwide with the highest incidence in Eastern Asia. Despite decreasing incidence in the West, it remains a therapeutic challenge. In the Western hemisphere, gastric malignancy is often diagnosed at an advanced stage and in contrast to Eastern Asia, it is preferably located in the proximal third of the stomach or the gastroesophageal junction (GEJ). Hence, multimodal treatment concepts have been introduced after demonstrating outcome benefit in randomized controlled trials.[3–5] These therapies not only influence patients’ outcome because of compromised immune functions and anemia after neoadjuvant chemotherapy, but also lead to the notion that oncologic resection as the only option for cure is technically demanding. Extended lymphadenectomy (D2) in advanced GC is considered a standard of care in specialized centers and surgical extension to the distal esophagus to achieve clear margins is sometimes crucial for the patient's survival. These factors in conjunction increase the risk for allogeneic blood transfusion during or after surgery in patients with advanced disease.
To date, several studies have investigated the influence of perioperative blood transfusion (PT) on clinical outcome following curative resection, resulting in conflicting results. A recent meta-analysis by Sun et al demonstrated that PT may be related to worsened prognosis after GC surgery. However, there was considerable heterogeneity and possible confounders that could not be adjusted for in the analysis. Further shortcomings of previous studies were that patients who were treated for GC before the year 2000 received non-leukocyte-depleted blood products and that multimodal therapies were not applied in clinical routine. In addition, most of the data published to date were derived from Eastern Asian patient cohorts, rendering conclusions on Western patients difficult. However, the most critical issue was that previous studies did not rule out the possibility of confounding factors leading to bias in the observed effects of PT on outcome. This bias arises because of the selection of patients with poorer clinical profiles or increased surgical complications to receive PT, thus artificially enhancing the association of PT with diminished clinical outcome. The aim of the present analysis was to compare the oncologic outcomes of Western patients who received leukocyte-depleted blood products versus those that did not in an era of multimodal treatment and modern perioperative patient care, and to adjust for selection-bias through propensity score matching.
2 Patients, materials and methods
2.1 Study subjects
Data from 610 patients who underwent curative surgery for GC at a tertiary referral hospital in southern Germany (Surgical Department of the Technische Universitaet Muenchen (TUM/MRI) from 2001 to 2013 were extracted from a prospectively documented database. Data was obtained from the medical records and transferred to the institutional database as soon as the patients were discharged from inpatient hospital care. Patients staged cT3/cT4cNany or cT2N1 received neoadjuvant chemotherapy after multidisciplinary team review. All other patients underwent primary surgical resection. Exclusion criteria were metastatic disease, gastric stump cancer, hospital mortality within 30 days, loss of follow-up within a 60-month period, and residual cancer after surgery (R1/R2). All surgical procedures were performed according to the Japanese guidelines for GC treatment including standardized D2 lymph node dissection. Perioperative period was defined as 3 days before and after surgery. In case of transfusion, patients received leukocyte-depleted packed red blood cells (PRBC). PRBCs were processed and leukocyte depleted according to the guidelines of the German Medical Association. Adjuvant chemotherapy was not applied in a standardized manner, but rather on a case by case basis after multidisciplinary team review. After oncologic surgery, all patients were followed-up every 6 to 12 months in an outpatient department (Roman Herzog Comprehensive Cancer Center) for the next 5 years according to local guidelines including physical examination, EGD, and annual computed tomography (CT) scans. Institutional Review Board (IRB) approval for this study was obtained according to local guidelines.
2.2 Statistical analysis
Wilcoxon and χ2 tests were used to compare continuous and categorical clinical characteristics, respectively, between patients who did and did not receive PT. Patients with missing data were omitted from this analysis. Overall survival (OS) and relapse-free survival (RFS, time to onset of recurrent disease) were graphed using empirical Kaplan–Meier curves with differences in 5-year survival rates among patient groups evaluated using the log-rank test. Associations between prognostic factors, including PT, and survival were estimated by univariable and multivariable Cox proportional hazards regression. Risk factors associated with reception of PT were calculated using logistic regression.
Optimal multivariable predictive models for all endpoints were selected using stepwise regression with the Akaike Information Criterion (AIC) as the performance benchmark. To correct for the observational bias in receiving PT, a propensity score analysis adjusting for risk factors selecting for PT use was performed for the association of PT and overall and RFS following recommended statistical practice. Propensity scores were calculated using logistic regression with PT as the outcome variable and all available variables as predictors. Weighted Cox proportional hazards regression using the scores as inverse weights to assess the effect of PT on survival endpoints was then performed.
All statistical tests were performed at the two-sided 0.05 level of significance. Statistical analyses were performed using SAS-software (Version 9, SAS-Institute, Cary, NC) and R (Version 3.1.2, R-foundation, Vienna, Austria) together with the survival package (Version 2.37–7, Rochester, MN).
3.1 Patient data
Demographic data are shown in Table 1. Among 610 curatively resected patients (R0, pM0), 275 (45%) received at least 1 perioperative blood transfusion (Table 1). The median of transfused units was 2 [1–11]. The mean hemoglobin level before surgery in those patients receiving blood transfusion was 12.3 g/dL (SD 1.37). In the transfusion group, significantly more patients had lymph node metastasis (P = 0.005), had more extensive surgeries (P <0.001), and a higher frequency of tumor located in the proximal third of the stomach (P = 0.012). There were no significant differences in the distributions of pT- and UICC-stages, tumor grade, type of surgery, Lauren histotype, gender, age, neoadjuvant chemotherapy, adjuvant chemotherapy, histopathologic response rate, frequency of splenectomy, vessel infiltration, and dissected lymph nodes.
Median follow-up was 41 months [range: 0.1–153 months], comprising 61 months [range: 0.4–153 months] for survivors and 22.5 months [range: 0.1–124 months] for deceased patients. During the follow-up period, 258 patients (42.3%) died; the 5-year survival rate for patients not having been transfused was 62% compared with 53% in the transfused group (P = 0.025); the 10-year survival rates are 54% versus 38%, respectively (P = 0.003).
3.2 Predictors of overall and relapse-free survival
Univariate regression analysis revealed UICC stage, type of surgery, incidence of postoperative complications, PT, neoadjuvant chemotherapy, splenectomy, surgical extension, and tumor localization in the proximal third of the stomach to be significantly related to postoperative survival (Table 2). Multivariable analysis demonstrated UICC stage, severe postoperative complications (CD III–V), age, application of neoadjuvant chemotherapy, and PT to be significantly and independently related to worsened postoperative survival (Table 2, Supplementary Table 1, http://links.lww.com/MD/B137). The number of transfused blood units did not influence overall survival significantly among patients who received at least 1 transfusion (P = 0.20).
Regarding RFS, in univariable analysis UICC stage, type of surgery, incidence of postoperative complications, gender, neoadjuvant chemotherapy, histopathologic response, splenectomy, surgical extension, and tumor localization in the proximal third of the stomach were significant predictors (Supplementary Table 2, http://links.lww.com/MD/B137). In the multivariable analysis, UICC stage, type of surgery, and neoadjuvant chemotherapy remained independent predictors. PT was not associated with RFS, neither in univariable nor the multivariable analysis (Supplementary Table 3, http://links.lww.com/MD/B137, Supplementary Table 2, http://links.lww.com/MD/B137).
3.3 Predictors of PT
Factors predicting whether patients were more likely to receive PT (Supplementary Table 4, http://links.lww.com/MD/B137) were higher age (P = 0.05), surgical extension to adjacent organs/structures (P = 0.02), tumor localization at the proximal third of the stomach (P = 0.003), and female gender (P = 0.03).
3.4 Results of propensity score weighting analysis
In order to balance for potential confounders, the adjusted propensity score was calculated using the following variables: UICC stage, grading, type of surgery, Lauren histotype, CD grades, gender, age, neoadjuvant treatment, adjuvant chemotherapy, lymphatic vessel infiltration, splenectomy, surgical extension, number of dissected lymph nodes, and tumor localization (Table 3). The propensity score weighting for those variables demonstrated that PT was associated with an increased risk of death (hazard ratio (HR): 1.31, 95%CI: 1.01–1.69, P = 0.04). As for the unadjusted analysis, adjusted propensity score analysis on RFS revealed an increased risk of recurrence without statistical significance (HR: 1.19, 95%CI: 0.88–1.61, P = 0.27). Kaplan–Meier analysis for overall survival after propensity score weighting revealed that patients who received PT demonstrated a 9.4% lower 5-year (P = 0.07) and a 19.7% lower 10-year survival rate (P = 0.01). The weighted log-rank test indicated a significant difference in the Kaplan–Meier estimates (P = 0.04, Fig. 1). The results of a propensity score analysis may be sensitive to the choice of variables used in the calculation of the propensity scores. Use of all baseline characteristics is most often recommended, but to assess the robustness of conclusions the choice of predictors to use in the propensity score calculation varied. Results showed that PT remained statistically significant in the majority of alternative analyses for overall and RFS (Supplementary Fig. 1, http://links.lww.com/MD/B137 and Supplementary Tables 5 and 6, http://links.lww.com/MD/B137).
This propensity score weighted retrospective analysis of 610 patients undergoing curative GC surgery demonstrates an important influence of perioperative blood transfusion on OS but not on RFS. Several studies investigating the effect of perioperative transfusion (PT) were published over the recent years with conflicting results making it difficult to draw definite conclusions. A recent meta-analysis concluded that the negative effects of PT prevail; however, the authors also stated that confounders could not be ruled out and that heterogeneity jeopardizes clear conclusions from their results.
There are several reasons that should be considered in the context of previous analyses. Most of the so far published analyses date back to a period from 1980 to 2000 during which transfusions guidelines were almost nonexistent.[12–20] Clear transfusion triggers were not defined at that time and transfusion was left to the surgeon's discretion. Previous publications reported on patient data from that specific period. Another issue to be considered is that many of those papers reported on long-time periods between 1980 and 2000. During those periods, perioperative care may have improved considerably and therefore influenced the outcome. Furthermore, whole blood transfusions (buffy coat depleted) were applied before the year 2000. Leukocyte-depleted PRBCs are known to possess less immunogenic properties. Most of the published studies did not report on how the respective transfusions were prepared, which may have led to the contradictory results. Lastly, most of the analyses originated in Eastern Asia, where GC incidence is known to be the highest worldwide. Recently, new insights in different molecular properties between Asian and Caucasian patients were reported, identifying various genes that may be responsible for the considerably better oncologic outcome in Eastern Asian patients.
Previous analyses of Western patients could not conclusively determine a negative influence of perioperative blood transfusions.[16,17,22–24] The present analysis therefore aimed to eradicate the above-mentioned limitations. In this respect the unicentric nature of the present analysis may be advantageous. Only patients having been surgically treated in a curative manner (R0 resection, M0 stage) after the year 2000 were included in this analysis. All transfused patients received leukocyte-depleted blood products only. The frequency of transfusion in this analysis was comparable to previously published data and can be considered quite high. However, a decrease in transfusion frequency was noticed over the investigated period, which is related to introduction of specific transfusion triggers in the institution in 2007.[25,26] The transfusion rate was reduced from 62% (period 2001–2006) to 27% (period 2007–2013). This was achieved by introducing stricter transfusion triggers allowing for lower hemoglobin values.
Similar to previous publications the present analysis reveals considerable differences in the baseline characteristics between transfused and nontransfused patients, which are related to tumor localization, surgical extent, and postoperative complications. The distribution of tumor stages was not significantly different between the groups. However, multivariable analysis demonstrated UICC stage, postoperative complications, PTs, application of neoadjuvant chemotherapy, and age to be significantly related to overall survival, which is comparable to previous data.
Previous studies primarily made inference from the results of multivariable cox regression analysis. This analysis in addition implemented propensity score weighting to rule out possible confounders. Propensity score weighting is a statistical method applied to reduce possible selection-bias in observational/nonrandomized studies, which was initially proposed by Rosenbaum and Rubin in 1983. It remains the state-of-the art for adjusting for confounding in nonrandomized studies.[29,30] In this analysis, inversion of the propensity score was used to calculate weights because the patient cohorts were almost comparable in size. After propensity score adjusting of both groups, perioperative blood transfusion was still predictive for overall survival and revealed a pronounced difference in overall postoperative survival.
There are multiple recommendations for determining which variables should be included in a propensity score analysis, ranging from all available to only those predictive of outcome. In this study all available patient factors were used. However, additional sensitivity analyses demonstrated that the choice of factors to use in the propensity scores had negligible influence on results. Exploratory statistical analyses following adjustment by propensity scores demonstrated that patient characteristics were balanced between those who received and did not receive transfusions. After adjustment, the statistically significance and magnitude of the influence of PTs on overall survival remained, but were dampened.
RFS was not significantly affected by PT, suggesting that the influence of PT on survival did not operate via this intermediary pathway. Factors affecting overall and RFS showed considerable differences. Whereas UICC stage, application of neoadjuvant chemotherapy, and extended surgery were associated with relapse risk, UICC stage, severe postoperative complications (CD III–V), age, application of neoadjuvant chemotherapy, and PT were associated with overall survival. These prognosticators suggest that relapse risk is primarily associated with advanced tumors, whereas overall survival is also influenced by perioperative factors, such as blood transfusions and severe complications. Under this scenario, factors related to postoperative complications and blood transfusion could have an impact on immune functions, allowing for dissemination of tumor cells. Patients having undergone transfusion and subsequent relapse were more prone to experience earlier mortality in the further course of disease. Nonetheless, a benefit for RFS could be detected (HR: 1.19, 95%CI: 0.88–1.61, Fig. 2), which was just not statistically significant. In prior investigations, the influence of splenectomy was argued as significantly influencing the detrimental effects of allogeneic blood transfusions. The splenectomy rate was considerably low in this patient cohort, prohibiting conclusions as to its effect.
Regarding the present data, in concordance with the previously published results, definitive conclusions on perioperative blood transfusion cannot be drawn. Although there are several strengths in this analysis, such as propensity weighting of confounders, inclusion of only curatively treated patients in an experienced center, and application of leukocyte-depleted blood products only, the following limitations have to be taken into account: propensity score weighting is not able to balance for unmeasured factors such as surgical quality and interindividual biologic or genetic differences. Further unmeasured factors such as surgical quality, influence of surgical trauma, the influence of improved postoperative intensive care, and general improvements in a time period over 12 years could not be assessed in this analysis. A randomized controlled trial is crucial in order to shed light on the influence of perioperative blood transfusion in GC patients. To the authors knowledge no such trial has been conducted to date. Therefore, patient blood management programs are encouraged to be applied until further data is available on the possibly negative influence of allogeneic blood transfusion in GC patients.
The authors thank Dr Frick from the Department of Clinical Chemistry, Klinikum Rechts der Isar Muenchen for supporting this study by providing database access and his ambitious efforts for patient data retrieval.
1. Guggenheim DE, Shah MA. Gastric cancer
epidemiology and risk factors. J Surg Oncol
2. Lagarde SM, ten Kate FW, Reitsma JB, et al Prognostic factors in adenocarcinoma of the esophagus or gastroesophageal junction. J Clin Oncol
3. Cunningham D, Allum WH, Stenning SP, et al Perioperative chemotherapy versus surgery alone for resectable gastroesophageal cancer. N Engl J Med
4. Ychou M, Boige V, Pignon J-P, et al Perioperative chemotherapy compared with surgery alone for resectable gastroesophageal adenocarcinoma: an FNCLCC and FFCD multicenter phase III Trial. J Clin Oncol
5. van Hagen P, Hulshof MC, van Lanschot JJ, et al Preoperative chemoradiotherapy for esophageal or junctional cancer. N Engl J Med
6. Songun I, Putter H, Kranenbarg EM-K, et al Surgical treatment of gastric cancer
: 15-year follow-up results of the randomised nationwide Dutch D1D2 trial. Lancet Oncol
7. Kutup A, Nentwich MF, Bollschweiler E, et al What should be the gold standard for the surgical component in the treatment of locally advanced esophageal cancer: transthoracic versus transhiatal esophagectomy. Ann Surg
8. Sun C, Wang Y, Yao HS, et al Allogeneic blood transfusion and the prognosis
of gastric cancer
patients: systematic review and meta-analysis. Int J Surg
9. Austin PC. The use of propensity score methods with survival or time-to-event outcomes: reporting measures of effect similar to those used in randomized experiments. Stat Med
11. Xie J, Liu C. Adjusted Kaplan–Meier estimator and log-rank test with inverse probability of treatment weighting for survival data. Stat Med
12. Blumberg N, Heal JM, Murphy P, et al Association between transfusion of whole blood and recurrence of cancer. Br Med J (Clin Res Ed)
13. Moriguchi S, Maehara Y, Akazawa K, et al Lack of relationship between perioperative blood transfusion
and survival time after curative resection for gastric cancer
14. Fong Y, Karpeh M, Mayer K, et al Association of perioperative transfusions with poor outcome in resection of gastric adenocarcinoma. Am J Surg
15. Choi JH, Chung HC, Yoo NC, et al Perioperative blood transfusions and prognosis
in patients with curatively resected locally advanced gastric cancer
16. Sanchez-Bueno F, Garcia-Marcilla JA, Perez-Abad JM, et al Does perioperative blood transfusion
influence long-term prognosis
of gastric cancer
? Dig Dis Sci
17. Heiss MM, Allgayer H, Gruetzner KU, et al Prognostic influence of blood transfusion on minimal residual disease in resected gastric cancer
patients. Anticancer Res
18. Dhar DK, Kubota H, Tachibana M, et al A tailored perioperative blood transfusion
might avoid undue recurrences in gastric carcinoma patients. Dig Dis Sci
19. Hyung WJ, Noh SH, Shin DW, et al Adverse effects of perioperative transfusion on patients with stage III and IV gastric cancer
. Ann Surg Oncol
20. Ojima T, Iwahashi M, Nakamori M, et al Association of allogeneic blood transfusions and long-term survival of patients with gastric cancer
after curative gastrectomy. J Gastrointest Surg
21. Lin SJ, Gagnon-Bartsch JA, Tan IB, et al Signatures of tumour immunity distinguish Asian and non-Asian gastric adenocarcinomas. Gut
22. Weitz J, D’Angelica M, Gonen M, et al Interaction of splenectomy and perioperative blood transfusions on prognosis
of patients with proximal gastric and gastroesophageal junction cancer. J Clin Oncol
23. Pacelli F, Rosa F, Marrelli D, et al Do perioperative blood transfusions influence prognosis
of gastric cancer
patients? Analysis of 927 patients and interactions with splenectomy. Ann Surg Oncol
24. Rausei S, Ruspi L, Galli F, et al Peri-operative blood transfusion in gastric cancer
surgery: prognostic or confounding factor? Int J Surg
2013; 11 (suppl 1):S100–S103.
25. Muller MM, Geisen C, Zacharowski K, et al Transfusion of packed red cells. Dtsch Arztebl Int
26. Carson JL, Carless PA, Hebert PC. Transfusion thresholds and other strategies for guiding allogeneic red blood cell transfusion. Cochrane Database Syst Rev
2012; 4: CD002042.
27. Reim D, Loos M, Vogl F, et al Prognostic implications of the seventh edition of the international union against cancer classification for patients with gastric cancer
: the Western experience of patients treated in a single-center European institution. J Clin Oncol
28. Rosenbaum PR, Rubin DB. The central role of the propensity score in observational studies for causal effects. Biometrika
29. Kuss O, Legler T, Borgermann J. Treatments effects from randomized trials and propensity score analyses were similar in similar populations in an example from cardiac surgery. J Clin Epidemiol
30. Lonjon G, Boutron I, Trinquart L, et al Comparison of treatment effect estimates from prospective nonrandomized studies with propensity score analysis and randomized controlled trials of surgical procedures. Ann Surg
31. Austin PC, Schuster T. The performance of different propensity score methods for estimating absolute effects of treatments on survival outcomes: a simulation study. Stat Methods Med Res
32. Shander A, Van Aken H, Colomina MJ, et al Patient blood management in Europe. Br J Anaesth
gastric cancer; perioperative blood transfusion; prognosis
Supplemental Digital Content
Copyright © 2016 The Authors. Published by Wolters Kluwer Health, Inc. All rights reserved.