Introduction
The immunomodulating qualities of allogenic blood transfusion have long been appreciated and have even been exploited to prevent renal allograft rejection in the era prior to the development of effective immunosuppressant drugs [
1]. The unintended clinical consequences of immune modulation by allogenic blood in the perioperative period include an increased susceptibility to infectious complications and cancer recurrence [
2],[
3]. As inpatient non-cardiac surgery has recently been associated with much higher than anticipated mortality rates, it is now imperative that all potentially avoidable causes of excess morbidity in this population are investigated and addressed [
4]. Whilst preoperative anaemia is associated with a doubling of inhospital mortality, it remains unclear whether the anaemia itself or its treatment with allogeneic blood is responsible [
5]. Furthermore, perioperative blood transfusions are not universally associated with an increase in complications particularly when patients receiving leukodepleted blood are included [
2],[
6]. Also lacking from the current literature is an analysis of alterations in key inflammatory pathways associated with the transfusion of leukodepleted blood following major elective surgery, although our group has recently described an association between blood transfusion following severe traumatic injury, an anti-inflammatory pattern of gene expression and an excess of infectious complications [
7].
It has been repeatedly demonstrated that messenger RNA (mRNA), assayed using real-time polymerase chain reaction (RT-PCR), can be used to accurately quantify patterns of gene expression [
8]-[
10]. Furthermore, careful selection of key regulatory and effector cytokines in conjunction with transcription factors specific to T cell subtypes allow inferences to be made as to the activity of specific inflammatory pathways related to T-helper (T
h) cell development [
11]. For example, an analysis that combines gene expression data relating to the key T
h1-promoting cytokine, interleukin (IL)-12, the main T
h1 effector cytokine, interferon gamma (IFNα) and the T
h1-specific transcription factor T-bet may be descriptive of the general activity of the T
h1 pathway. A similar approach can aid in the description of T-helper cell type 2 (T
h2), T
h17 and T regulatory (T
reg) cell pathways. This methodology has been previously utilised to predict perioperative infectious complications based on alterations in T cell pathways [
10].
In this study, we explore the hypothesis that blood transfusion during and following major elective gastrointestinal surgery is associated with an immunosuppressive pattern of gene expression, which may enhance susceptibility to later infectious complications.
Methods
This prospective cohort study was conducted at The Royal London Hospital, London UK and University College Hospital, London between June 2012 and November 2013. Ethical approval was granted by the North Wales Ethics Committee (Reference: 10/WNo03/25) and this research was conducted in compliance with the Helsinki Declaration. Each patient on a weekday elective operating list was screened. Eligible patients were then approached for written informed consent. Entry into the study did not influence the patient's clinical treatment course in any way. In our institution, scheduled postoperative intensive care unit (ICU) admission is mandated for specific surgical procedures. Based on an individual patient's co-morbidities the surgical or anaesthesia teams may also request postoperative ICU treatment. All patients received perioperative prophylactic antibiotic treatment as outlined in Additional file
1.
Patient selection
Patients aged over 45 years undergoing scheduled surgery involving the gastrointestinal tract, requiring a general anaesthetic and at least an overnight hospital stay were considered eligible for inclusion to this study. Exclusion criteria were emergency surgery and surgery that also involved access to the thoracic cavity.
Data collection
Data were collected on each patient until hospital discharge and included information on co-morbidities, American Society of Anesthesiology (ASA) physical status classification [
12] indication for surgery, cancer diagnosis, duration of the procedure, planned postoperative ICU admission and inhospital mortality. Preoperative immunosuppression was defined as the administration of chemotherapeutic agents and/or radiotherapy within the six months preceding surgery. The timing of any blood products administered was recorded for each patient during the first 24 hours. All packed red blood cells (PRBC) administered were leukocyte depleted prior to storage, included a standard sodium chloride, adenine, glucose, mannitol (SAGM) additive solution and had a maximum age of 35 days. Patients were examined daily by the clinical team for the presence of infection, which was then recorded by the research team following discussions with ICU physicians, surgeons or microbiologists where necessary. Definitions of infection were agreed prospectively by the investigators and were based on the Center for Disease Control and Prevention definitions and graded using the Clavien-Dindo classification (see Additional file
2) [
13].
Blood sampling and laboratory techniques
Blood samples were taken immediately before induction of anaesthesia (0 hours), at 24 and 48 hours following the operation. At each time point blood was collected in a PAXGene™ blood RNA tube (PreAnalytix, Hilden, Germany). Total RNA was extracted using PAXGene™ blood RNA kit (PreAnalytix). Samples were analysed for their integrity and reverse transcribed to complementary DNA (cDNA) [
11]. Genes were selected to reflect key elements of specific immune pathways (Table
1). Gene expression was quantified using quantitative real-time polymerase chain reaction (qRT-PCR). qRT-PCRs were performed using Taqman assays (Applied Biosystems, Foster City, CA, USA), which spanned the final two exons of the most common isoform of each gene, and were carried out on a 7900HT, Life Tech (Applied Biosystems, Foster City, CA, USA) as previously described [
7]. Each sample was assayed in triplicate. Reference genes (β2 microglobulin (
B2M) and ubiquitin C (
UBC)) were selected empirically from a panel of six [
14]. Relative quantification was calculated using the standard delta delta methodology [
15]. Results were expressed as a normalized ratio of candidate gene/reference gene.
Table 1
Selected cytokines and transcription factors and their related pathways
TNFα (TNF)b | Common end product of many innate and adaptive immune pathways |
+
|
IFNα (IFNG)b | Th1 effector cytokine |
+
|
IL-12 (IL12A)b | Promotes differentiation to Th1 effector cells |
+
|
T-bet (TBX21)b | Transcription factor utilised by Th1 cells |
+
|
IL-23 (IL23A)b | Promotes differentiation to Th17 phenotype |
+
|
IL-27 (IL27)b | Inhibits differentiation to Th17 phenotype |
-
|
RORγT (RORC)b | Transcription factor utilised by Th17 cells |
+
|
IL-10 (IL10)b | Anti-inflammatory cytokine produced by many T cell subtypes and some macrophages |
-
|
Foxp3 (FOXP3)b | Transcription factor utilised by naturally occurring CD4+ CD25+ Treg cells |
-
|
GATA3 (GATA3)b | Transcription factor utilised by Th2 cells |
-
|
Statistical analysis
Results are expressed as median and interquartile range (IQR) or odds ratios (OR) with 95% confidence intervals (CI). All statistical tests are two-sided and a P value of P <0.05 was considered significant. Differences in categorical variables were calculated using a chi-squared or Fisher's exact test as appropriate. The Kruskal-Wallis test was utilised for continuous variables.
Multivariable linear regression models assessed whether gene expression was independently associated with blood transfusion. A model was constructed in each case where gene expression was demonstrated to be associated with blood transfusion on a univariate analysis. Gene expression was then selected as the dependent variable following log transformation of the mRNA levels. A univariate analysis then assessed the relationship between gene expression and age, sex, duration of operation, ASA physical status classification, cancer diagnosis and preoperative immunosuppression. If this univariate analysis demonstrated a
P value <0.1 then that variable was added to the above model as an independent variable. Backwards elimination on non-significant variables was performed when appropriate (see Additional file
3).
A multivariable logistical regression analysis assessed whether infectious complications were independently associated with blood transfusion. The occurrence of infectious complications was the dependent variable and the threshold for inclusion of other variables was again
P <0.1, with variables selected from Table
2 (see Additional file
4).
Table 2
Characteristics of patients developing infections and those remaining infection-free following scheduled abdominal surgery
Age (years) | 66 (59-75) | 64 (56-71) | 0.19 |
Male sex (%) | 61 | 63 | >0.99 |
Diabetes (%) | 18 | 16 | 0.80 |
Current smokers (%) | 23 | 19 | 0.64 |
Smoking history (%) | 48 | 57 | 0.34 |
Cancer diagnosis (%) | 55 | 71 | 0.10 |
Preoperative immunosuppression (%) | 14 | 14 | >0.99 |
Duration of operation (minutes) | 243 (176-313) | 195 (142-295) | 0.06 |
Endoscopic surgery (%) | 18 | 32 | 0.13 |
Planned postoperative intensive care unit admission (%) | 77 | 66 | 0.22 |
ASA grade 3 or 4 (%) | 30 | 31 | >0.99 |
By surgical specialty n (%)
|
General surgery | 4 (44) | 5 (55) | |
Upper gastrointestinal | 9 (33) | 18 (67) | |
Colorectal | 18 (37) | 31 (63) | |
HPB | 11 (37) | 19 (63) | |
HPB + colorectal | 1 (33) | 2 (67) | |
General surgery + colorectal | 1 (100) | 0 (0) | 0.84a |
Intraoperative blood transfusion (%) | 14 | 5 | 0.17 |
Blood transfusion in the first 24 hours (%) | 23 | 7 | 0.02 |
Inhospital death n (%) | 1 (2%) | 2 (2.5%) | >0.99 |
Data analysis was performed using the JMP (version 11) statistical software package (SAS Institute Inc., Cary, NC, USA).
Discussion
In this study we determined levels of mRNA for a panel of interlinked cytokines and related transcription factors and demonstrated that those patients receiving a perioperative blood transfusion following major gastrointestinal surgery displayed a pattern of gene expression consistent with greater immunosuppression when compared with a cohort not receiving a blood transfusion. This observation was independent of variables descriptive of the extent of surgical trauma. The gene expression data presented suggests that blood transfusion in this setting may be associated with specific immune defects in innate immunity and in Th1 and Th17 pathways. These patients also had an excess of postoperative infectious complications when compared with patients that did not receive a blood transfusion; an observation that was again independent of the duration of the surgical procedure.
The genes selected as candidates in this study are well described as surrogate measures of the specific immune pathways outlined in Table
1[
10],[
11],[
16]. Furthermore, the construction of a ratio of gene expression that incorporates the prototypical proinflammatory cytokine, TNFα, and the archetypal anti-inflammatory cytokine, IL-10, permits the concurrent assessment of both major facets of the immune response and has previously been utilised to predict postoperative outcome [
17]. Whilst the observed reductions in TNFα and IL-12 gene expression are indicative of impaired innate immunity and T
h1 pathways respectively and the reduction in IL-23 and RORγt is indicative of reduced T
h17 activity, these data should not be interpreted as being conclusive. Flow cytometry experiments would be necessary to confirm these initial data but present significant logistical barriers and are ideally not performed on stored samples [
18]. The T
h1 pathway is an essential link between innate and adaptive immunity and T
h1 effector cells are particularly important for effective bactericidal activity against intracellular pathogens [
19]. T
h17 cells are a branch of the adaptive immune system and appear to deal primarily with organisms inadequately subdued by T
h1 or T
h2 immunity and that seem to require a very robust inflammatory response [
20]. It is biologically plausible that an acquired immune deficit involving these pathways could lead to an increased susceptibility to opportunistic infections with the pathogens encountered in the perioperative period.
Although numerous small studies have suggested that perioperative allogenic blood transfusion may induce immune modulations that create an environment that promotes both microbial growth and tumorigenesis their conclusions are limited by small cohort size,
in vitro and
ex vivo assays and insensitive protein assays [
3],[
21]-[
26]. In addition, these studies largely predate the routine leukodepletion of transfused blood and suggest a causal relationship between immune modulation and the infusion of allogeneic leukocytes; a biological effect that could be ameliorated by pre-storage leukodepletion [
22]. It cannot therefore be inferred that a similar immunological response will be observed following exposure to a leukodepleted product. In contrast, the packed red cells transfused in the study reported here were universally leukodepleted. Furthermore, the use of qRT-PCR in this study to assay gene expression allows for a validated ultrasensitive technique to assess immune pathways in the
in vivo state [
8]-[
11]. The results we present suggest that the immunosuppression observed following perioperative transfusion may not be solely dependent on the presence of allogeneic leukocytes in the transfused blood. However, despite a 3 log reduction in leukocyte numbers achieved following current filtration techniques some leukocytes do persist and these retain the ability to influence host immune responses [
22].
Recently, using a similar methodology we described an excess of infectious complications amongst severely injured trauma patients receiving leukodepleted blood transfusion in conjunction with a gene expression profile characterised by specific deficits in T
h1 and T
h17 pathways [
7]. The similarities between these results are notable given that there were important differences in the patient populations. The trauma cohort were younger (median age 41), had a higher incidence of early blood transfusion (64%) and received greater volumes of PRBCs within the initial 24 hours (median 4 units) in conjunction with additional blood products. In addition, the trauma cohort are likely to have endured a greater physiological insult given a higher mortality rate (19%) and also a greater incidence of nosocomial infections (63%). Clearly, patients that receive a blood transfusion outside of the context of a randomised controlled trial will differ from those that do not require transfusion and it is important to attempt to control for these factors. In the previously reported trauma cohort the described observations were independent of the severity of injury and degree of shock at presentation [
7]. Similarly, in this perioperative cohort, although on univariate analysis there was a trend towards longer and therefore potentially more complicated surgery being associated with postoperative infections and to a lesser degree with blood transfusion, a multivariable analysis confirmed that the association between blood transfusion and both infectious complications and gene expression patterns were independent of the duration of surgery. Although we describe an association between blood transfusion, an immunosuppressive pattern of gene expression and excess infectious complications, it may be that, despite adjustment for known confounders, the need for blood transfusion is a surrogate marker of severity of surgical stress or an unmeasured premorbid condition and thus a causative role in postoperative immune phenotype is suggested but not proven. However, the consistency of these data from two distinct patient populations, post major gastrointestinal surgery patients and trauma patients, strengthens our underlying hypothesis.
The association between blood transfusion and infectious complications in the perioperative period has been well described previously [
2],[
3],[
27]. This association appears to be consistent amongst cohorts that received either whole blood or leukodepleted blood. The magnitude of the effect size is potentially quite substantial with one recent estimate quantifying the increased risk of infection at 29% for each unit of blood transfused [
27]. That we were able to describe a similar association between blood transfusion and nosocomial infection in a smaller cohort of patients may be related to this effect size, in addition to the selection of an at-risk population using patient age, gastrointestinal surgery under general anaesthesia and the requirement for inpatient treatment as inclusion criteria. We did not, however, record blood transfusions that were administered following the initial 24 hours postoperatively. The postoperative infection rates we report are consistent with similar patient populations [
28]. This study also described an association between blood transfusion and mortality. However, whilst the mortality rate is consistent with similar populations [
28], with only three deaths occurring in the cohort as a whole, no conclusions regarding causation should be inferred from this association.
It is unsurprising that we did not observe an association between increasing age of the transfused blood and either perioperative infectious complications or patterns of gene expression as this study was underpowered to detect such an association. There is much current debate surrounding the potential pathogenicity associated with blood stored for excessive periods prior to transfusion. In the perioperative period, blood stored for longer periods has been linked to an excess of infectious complications [
29]-[
31]. The results of a randomised controlled trial comparing blood stored for under a week with older blood in intensive care patients are awaited and will provide important and relevant data [
32].
Acknowledgements
The authors are grateful to the Adult Critical Care Research Team of the Royal London Hospital, Barts Health NHS Trust for data collection and blood sampling and to Dr Charles A Mein DPhil, Genome Centre, William Harvey Research Institute, Charterhouse Square, Barts and the London School of Medicine and Dentistry for his technical advice and assistance.
Funding
This study was supported by a grant from the National Institute of Academic Anaesthesia (British Journal of Anaesthesia / Royal College of Anaesthetists Project Grant).
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
PCF contributed to the data collection, laboratory analysis, and manuscript preparation. HDTT contributed to the data analysis, laboratory analysis, and manuscript preparation. RMP contributed to the study design, patient recruitment, and manuscript preparation. GLA contributed to the study design, patient recruitment, and manuscript preparation. JRP contributed to the data analysis, and manuscript preparation. HCO contributed to the laboratory analysis, data analysis, and manuscript preparation. CJH contributed to the study design, data analysis, and manuscript preparation. MJO'D contributed to the study design, patient recruitment, data analysis, and manuscript preparation. All authors read and approved the final manuscript.