Transfusion thresholds and other strategies for guiding allogeneic red blood cell transfusion
Abstract
Background
There is considerable uncertainty regarding the optimal haemoglobin threshold for the use of red blood cell (RBC) transfusions in anaemic patients. Blood is a scarce resource, and in some countries, transfusions are less safe than others because of a lack of testing for viral pathogens. Therefore, reducing the number and volume of transfusions would benefit patients.
Objectives
The aim of this review was to compare 30‐day mortality and other clinical outcomes in participants randomized to restrictive versus liberal red blood cell (RBC) transfusion thresholds (triggers) for all conditions. The restrictive transfusion threshold uses a lower haemoglobin level to trigger transfusion (most commonly 7 g/dL or 8 g/dL), and the liberal transfusion threshold uses a higher haemoglobin level to trigger transfusion (most commonly 9 g/dL to 10 g/dL).
Search methods
We identified trials by searching CENTRAL (2016, Issue 4), MEDLINE (1946 to May 2016), Embase (1974 to May 2016), the Transfusion Evidence Library (1950 to May 2016), the Web of Science Conference Proceedings Citation Index (1990 to May 2016), and ongoing trial registries (27 May 2016). We also checked reference lists of other published reviews and relevant papers to identify any additional trials.
Selection criteria
We included randomized trials where intervention groups were assigned on the basis of a clear transfusion 'trigger', described as a haemoglobin (Hb) or haematocrit (Hct) level below which a red blood cell (RBC) transfusion was to be administered.
Data collection and analysis
We pooled risk ratios of clinical outcomes across trials using a random‐effects model. Two people extracted the data and assessed the risk of bias. We conducted predefined analyses by clinical subgroups. We defined participants randomly allocated to the lower transfusion threshold as 'restrictive transfusion' and to the higher transfusion threshold as 'liberal transfusion'.
Main results
A total of 31 trials, involving 12,587 participants, across a range of clinical specialities (e.g. surgery, critical care) met the eligibility criteria. The trial interventions were split fairly equally with regard to the haemoglobin concentration used to define the restrictive transfusion group. About half of them used a 7 g/dL threshold, and the other half used a restrictive transfusion threshold of 8 g/dL to 9 g/dL. The trials were generally at low risk of bias .Some items of methodological quality were unclear, including definitions and blinding for secondary outcomes.
Restrictive transfusion strategies reduced the risk of receiving a RBC transfusion by 43% across a broad range of clinical specialties (risk ratio (RR) 0.57, 95% confidence interval (CI) 0.49 to 0.65; 12,587 participants, 31 trials; high‐quality evidence), with a large amount of heterogeneity between trials (I² = 97%). Overall, restrictive transfusion strategies did not increase or decrease the risk of 30‐day mortality compared with liberal transfusion strategies (RR 0.97, 95% CI 0.81 to 1.16, I² = 37%; N = 10,537; 23 trials; moderate‐quality evidence) or any of the other outcomes assessed (i.e. cardiac events (low‐quality evidence), myocardial infarction, stroke, thromboembolism (high‐quality evidence)). Liberal transfusion did not affect the risk of infection (pneumonia, wound, or bacteraemia).
Authors' conclusions
Transfusing at a restrictive haemoglobin concentration of between 7 g/dL to 8 g/dL decreased the proportion of participants exposed to RBC transfusion by 43% across a broad range of clinical specialities. There was no evidence that a restrictive transfusion strategy impacts 30‐day mortality or morbidity (i.e. mortality at other points, cardiac events, myocardial infarction, stroke, pneumonia, thromboembolism, infection) compared with a liberal transfusion strategy. There were insufficient data to inform the safety of transfusion policies in certain clinical subgroups, including acute coronary syndrome, myocardial infarction, neurological injury/traumatic brain injury, acute neurological disorders, stroke, thrombocytopenia, cancer, haematological malignancies, and bone marrow failure. The findings provide good evidence that transfusions with allogeneic RBCs can be avoided in most patients with haemoglobin thresholds above 7 g/dL to 8 g/dL.
PICOs
Plain language summary
Is it safe to use lower blood counts as a trigger for blood transfusion in order to give fewer blood transfusions?
Background
Doctors and healthcare professionals often give blood transfusions to people after loss of blood from surgery, bleeding, or medical illnesses. Blood is a limited resource, so for this reason, and because some low‐income countries do not test the blood used in transfusions for the presence of dangerous viruses such as HIV or hepatitis, it is helpful to give blood transfusions only when they are really necessary.
A normal blood count is above 12. This review summarised all randomized controlled trials (RCTs) that investigated whether it is safe to give blood transfusions when the blood count drops to between seven and eight (thereby reducing the number of transfusions), rather than giving transfusions at higher blood counts of nine to 10.
Study characteristics
We examined the results of RCTs that randomly allocated participants to one of two groups. In one group, trial participants received blood at lower blood counts. In the other group, trial participants received blood at higher blood counts. The data are current up to May 2016.
Key results
We identified a total of 31 relevant trials, which involved 12,587 participants. All of the studies compared different policies for blood transfusions. We found that participants who were assigned to receive blood at lower blood counts were 43% less likely to receive a blood transfusion than those who were given blood at higher blood counts. The risk of dying within 30 days of the transfusion was the same whether the participants received transfusion at lower or higher blood counts. We also evaluated harmful events that occurred after participants received, or did not receive, blood transfusions, including infection (pneumonia, wound infection, and blood poisoning), heart attacks, strokes, and problems with blood clots, and found that there was no clear difference in the instance of these events between the group that received transfusions at lower blood counts and the group that received transfusions at higher blood counts.
Quality of evidence
We found that most of the RCTs provided a high quality of evidence, in that they were adequately conducted and used appropriate methods that minimised any possible biases that could make the validity of the results uncertain.
Authors conclusions
We concluded that it was not harmful to the participants' health status to give blood at lower or higher blood counts. If a policy of giving blood only at lower blood counts were followed routinely in clinical practice, it would reduce the amount of blood patients receive substantially and reduce the risk of patients receiving blood transfusions unnecessarily, as transfusions can have harmful effects. Additional studies are needed to establish the blood count at which a blood transfusion is needed in patients who have suffered a heart attack, brain injury, or have cancer.
Authors' conclusions
Summary of findings
| Restrictive compared with liberal transfusion protocols for guiding allogeneic red blood cell transfusion | ||||||
| Patient or population: Adults and children (haemodynamically stable) with potential need for RBC transfusion Settings: Inpatient Intervention: Restrictive transfusion protocol Comparison: Liberal transfusion protocol | ||||||
| Outcomes | Illustrative comparative risks* (95% CI) | Relative effect | Number of participants | Quality of the evidence | Comments | |
| Assumed risk | Corresponding risk | |||||
| Liberal transfusion (Hb 9 g/dL to 10 g/dL) | Restrictive transfusion (Hb 7 g/dL to 8 g/dL) | |||||
| People receiving blood transfusions | 841 per 1000 | 479 per 1000 | RR 0.57 (0.49 to 0.65) | 12,587 (31) | ⊕⊕⊕⊕ | ‐ |
| 30‐day mortality | 93 per 1000 | 90 per 1000 | RR 0.97 (0.81 to 1.16) | 10,537 (23) | ⊕⊕⊕⊝ | ‐ |
| Myocardial infarction | 17 per 1000 | 19 per 1000 | RR 1.08 (0.74 to 1.60) | 8303 (16) | ⊕⊕⊕⊕ | ‐ |
| Congestive heart failure | 36 per 1000 | 28 per 1000 | RR 0.78 (0.45 to 1.35) | 6257 (12) | ⊕⊕⊝⊝ | ‐ |
| Cerebrovascular accident (CVA) ‐ stroke | 17 per 1000 | 13 per 1000 | RR 0.78 (0.53 to 1.14) | 7343 (13) | ⊕⊕⊕⊕ | ‐ |
| Rebleeding | 163 per 1000 | 144 per 1000 | RR 0.75 (0.51 to 1.10) | 3108 (6) | ⊕⊕⊝⊝ | ‐ |
| Pneumonia | 82 per 1000 | 76 per 1000 | RR 0.94 (0.80 to 1.11) | 6277 (14) | ⊕⊕⊕⊕ | ‐ |
| Thromboembolism | 10 per 1000 | 8 per 1000 | RR 0.77 (0.41 to 1.45) | 4019 (10) | ⊕⊕⊕⊕ | ‐ |
| *The basis for the assumed risk (e.g. the median control group risk across studies) is provided in footnotes. The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). | ||||||
| GRADE Working Group grades of evidence | ||||||
| a. Downgraded for imprecision: could be up to 17 per 1000 more deaths in restrictive transfusion. | ||||||
Background
Description of the condition
Patients who are ill in hospital are frequently anaemic, with low haemoglobin concentrations. The causes of anaemia are diverse, including loss of blood from surgery, bleeding, excessive blood sampling for laboratory tests, or as a consequence of illness. Patients with cancer may develop anaemia because of the underlying disease or chemotherapy affecting production of red cells in their bone marrow. Anaemia both decreases the oxygen content of the blood supplied to the tissues, including the myocardium, and increases myocardial oxygen demand by requiring a higher cardiac output to maintain adequate systemic oxygen delivery (Sabatine 2005).
Anaemia has been associated with worse outcomes in patients who are anaemic prior to surgery or who have cardiovascular disease (Carson 1996; Shander 2014). However, it does not necessarily follow that the correction of anaemia by red blood cell (RBC) transfusion will improve outcomes. Anaemia is generally well tolerated by many people, and therefore, the benefits of potentially corrective treatments such as red cell transfusions need to be weighed against their risks.
Description of the intervention
The main treatment option for raising the haemoglobin concentration rapidly in patients with anaemia remains RBC transfusion. Red cells for transfusion are collected from whole blood donations that are centrifuged to concentrate them, prior to the addition of anticoagulant and storage solutions. However, there are recognised risks of blood transfusion, as with any medical intervention. These risks and the general availability of RBC transfusion vary throughout the world. In countries with well‐regulated blood supplies, the safety of allogeneic red cell transfusion has improved significantly over the past 30 years, and the overall risks are very low. For example, in the USA, the estimated risk per unit for HIV is 1:1,467,000 (Zou 2010); for the hepatitis C virus (HCV), 1:1,149,000 (Zou 2010); and for the hepatitis B virus (HBV), 1:282,000 to 1:357,000 (Zou 2009). This has been primarily due to improvements in donor blood screening policies and the implementation of more stringent quality control measures (Klein 2007).
In resource‐poor countries, the supply of blood is inadequate and may not be safe because it is not often tested for viral pathogens. Blood donations are not routinely tested in 39 countries for transfusion‐transmissible infections that include HIV, hepatitis B, hepatitis C, and syphilis (WHO 2011). In 40 countries, less than 25% of the blood supply is collected from voluntary unpaid blood donors, with most coming from family or paid blood donors (WHO 2011). The prevalence of HIV in low‐income countries is 2.3% of blood donations compared with 0.001% in high‐income countries (WHO 2011).
Other general risks of transfusion have been described (although perhaps under‐reported) and include acute transfusion reactions, volume overload, bacterial contamination, infection with new ‐ but currently unknown ‐ blood‐borne pathogens, and transfusion‐related acute lung injury (Delaney 2016; Toy 2012). Additional possible adverse effects include loss of red cell nitric oxide production, which is thought to induce local vasodilatation; prothrombotic effects from factors in the supernatant; and variable immunomodulatory or proinflammatory effects from different cellular products in the red cell component. Overall, these harmful effects of red cell transfusions may be manifested as increased risks of infections in hospitals or cardiovascular events, including myocardial infarction or stroke.
There were concerns that the effects of storage on red cells may also render them less effective and potentially harmful. However, recent trials have not demonstrated clinical harm to blood stored for approximately 21 to 28 days compared with less than seven to 10 days (Dhabangi 2015; Fergusson 2013; Lacroix 2015; Steiner 2015).
Blood transfusion is expensive. The direct costs of each collected bag of red cells fail to capture the many associated costs related to hospital blood‐banking practice and safe patient administration. In 2008, the mean payment for one unit of leukoreduced RBCs in the USA was USD 223 (Whitaker 2011). However, if the costs of administration as well as the acquisition expenses of RBC transfusion are considered, the estimated cost derived from four USA and European hospitals rises to USD 761 per unit (standard deviation +/‐ USD 294) (Shander 2010).
Treatment options for anaemia other than red cell transfusions may include erythropoietin, and oral or intravenous iron therapy. These have been used in chronic anaemia, particularly in renal failure, for many years, but are the subjects of other systematic reviews, so we did not consider them further within this review.
How the intervention might work
The rationale for transfusing RBCs in anaemic patients is to improve oxygen delivery to the tissues and to the myocardium itself, to reduce the compensatory work done by the heart to increase cardiac output. Red blood cell transfusion is one of the few treatments that may restore tissue oxygenation adequately when oxygen demand exceeds supply (Klein 2007; Wang 2010).
Many randomized controlled trials that compared outcomes in participants allocated to different policies or schedules of using red cell transfusions have now been completed and reported. These studies presented results after randomising participants to either 'restrictive' triggers (typically, participants are transfused only when their haemoglobin concentration falls to around 7 g/dL to 8 g/dL) or 'liberal' triggers (participants are transfused at a higher haemoglobin concentration of around 9 g/dL to 10 g/dL). Historically, the widely accepted clinical standard was to transfuse patients when the haemoglobin level dropped below 10 g/dL or the haematocrit fell below 30%. Adams and Lundy first proposed this '10/30 rule' in 1942, and it served as a RBC transfusion trigger for decades (Madjdpour 2005; Wang 2010). However, the 1988 National Institutes of Health Consensus Conference in the USA reported that the evidence did not support a single criterion for transfusion (NIH 1988). Since then, most published guidelines have advised against a single threshold for RBC transfusion, recommending that a range of haemoglobin values between 6 g/dL and 10 g/dL can be used, depending on the presence of serious comorbidity (AAGBI 2008; ASA 2006; BCTMAG 2003; Carson 2012a; Napolitano 2009; NBUGI 2001). The American Association of Blood Banks (AABB) guidelines advise using a restrictive transfusion threshold of 7 g/dL to 8 g/dL in most clinical settings (Carson 2012a).
Why it is important to do this review
Much of the earlier evidence comparing restrictive and liberal thresholds comes from trials based in critical care. In 1999, the landmark TRICC trial (transfusion requirements in critical care) found a similar mortality in participants transfused at a restrictive trigger of less than 7 g/dL compared with a liberal trigger of less than 10 g/dL (Hébert 1999). Since the last review (Carson 2012b), the number of participants enrolled in trials has doubled from 6264 to 12,587. Therefore, there is a need to update this systematic review to ensure that new guidelines continue to be based on the most recent literature reporting on the effectiveness and safety of RBC transfusion.
The purpose of the review was to identify, appraise, and summarize the data from all randomized controlled trials (RCTs) that studied the clinical impact of varying thresholds for transfusion with RBCs. We were particularly interested in whether the results of RCTs support the trend for increasingly restrictive RBC transfusion practices across all patient groups and if RBC transfusions can be withheld in some circumstances without harming patients.
Objectives
The aim of this review was to compare 30‐day mortality and other clinical outcomes in participants randomized to restrictive versus liberal red blood cell (RBC) transfusion thresholds (triggers) for all conditions. The restrictive transfusion threshold uses a lower haemoglobin level to trigger transfusion (most commonly 7 g/dL or 8 g/dL), and the liberal transfusion threshold uses a higher haemoglobin level to trigger transfusion (most commonly 9 g/dL to 10 g/dL).
Methods
Criteria for considering studies for this review
Types of studies
To examine the evidence for the effect of transfusion thresholds on the use of red blood cell (RBC) transfusions and the evidence for any change in clinical outcomes, we included randomized controlled trials if the comparison groups were assigned on the basis of a transfusion 'threshold' (also known as a 'trigger'), defined as a haemoglobin or haematocrit level (with or without a specified level of haemodynamic instability) that had to be reached before a RBC transfusion was administered. We required that control group participants had to have been either transfused with allogeneic or autologous red blood cells, or both, at higher haemoglobin or haematocrit levels (transfusion threshold) than the intervention group, or transfused in accordance with current transfusion practices, which may not have included a well‐defined transfusion threshold, but involved liberal rather than restrictive transfusion practices. We excluded trials that were not designed to include any clinical outcomes relevant to this review.
Types of participants
We included trials of surgical or medical participants, involving adults or children, or both. We excluded studies enrolling neonates, given the distinct pathophysiology and clinical features of anaemia.
Types of interventions
The intervention considered was the use of transfusion thresholds ('triggers') as a means of guiding allogeneic or autologous RBC transfusion, or both. A liberal transfusion threshold most often refers to administration of blood transfusion when the haemoglobin level falls below 9 g/dL to 10 g/dL. A restrictive transfusion threshold most often refers to administration of blood transfusion when the haemoglobin level falls below 7 g/dL to 8 g/dL.
Types of outcome measures
Primary outcomes
In contrast to prior versions of this systematic review, the primary outcome for this analysis was 30‐day mortality. The primary outcome was changed because mortality is a more clinically relevant outcome and the number of participants enrolled in trials provided sufficient power to examine this outcome. Sample size calculations assuming baseline 30‐day mortality of 9% for restrictive transfusion, 90% power, alpha level of 0.05, indicate that to detect a 15%, 20%, or 25% relative decrease in mortality with the use of liberal transfusion, a study needs to enrol 17,500, or 9600, or 6000 participants, respectively.
Secondary outcomes
Other time periods we examined for mortality included: during hospital admission, at 90 days, and long term (median follow‐up of 3.1 years). We compared RBC transfusion use between the groups (listed below for the morbidity outcome) by proportion of participants exposed to transfusion, participants exposed to allogeneic or autologous transfusion, units of blood transfused, and units of blood transfused in those receiving any transfusion. We evaluated morbidity that occurred during hospitalisation, including cardiac events (composite of myocardial infarction, cardiac arrhythmias, cardiac arrest, pulmonary oedema, and angina), non‐fatal and fatal myocardial infarction, congestive heart failure, cerebral vascular accident (stroke), rebleeding, infection, thromboembolism, renal failure, mental confusion, function, and fatigue. Infection was defined in three ways: sepsis or bacteraemia, pneumonia alone, or pneumonia plus wound infection. We defined all morbidity outcomes according to their definitions in the individual trials.
As this review is an update, we have continued to include some of the secondary outcomes for historical reasons. As stronger evidence is accrued, we feel that in future updates of this review, some of these outcomes may need to be modified or omitted.
Search methods for identification of studies
Electronic searches
We searched the following databases and ongoing trial registries:
-
the Cochrane Central Register of Controlled Trials (CENTRAL; 2016, Issue 4) in the Cochrane Library;
-
MEDLINE via OvidSP (from 1946 to 27 May 2016);
-
Embase via OvidSP (from 1974 to 27 May 2016);
-
PubMed (for Epublications ahead of print only, on 27 May 2016);
-
Transfusion Evidence Library (www.transfusionevidencelibrary.com, 1950 to 27 May 2016);
-
Web of Science Conference Proceedings Citations Index (CPCI‐S, 1990 to 27 May 2016);
-
US National Institutes of Health Ongoing Trials Register ClinicalTrials.gov (www.clinicaltrials.gov; searched to 27 May 2016); and
-
World Health Organization International Clinical Trials Registry Platform (apps.who.int/trialsearch; searched to 27 May 2016);
-
ISRCTN Registry (www.isrctn.com; searched to 27 May 2016).
We combined searches in MEDLINE and Embase with adaptations of the Cochrane randomized controlled trial (RCT) search filter as detailed in Chapter 6 of the Cochrane Handbook for Systematic Reviews of Interventions (Lefebvre 2011). We did not restrict our search by date, language, or publication status. We present the search strategies for this update and the ongoing trial registries in Appendix 1 and Appendix 2 respectively. We present search strategies for the 2012 update in Appendix 3.
Searching other resources
We checked the references of all identified trials, relevant review articles, and current treatment guidelines for further literature. We limited these searches to the 'first‐generation' reference lists, i.e. reference lists of papers retrieved directly by the database searches.
We contacted experts in the field to identify information relevant to the review. Where possible and when necessary, we contacted authors of published studies for clarification of trial methodology and data. We emailed all authors of trials that did not report our primary outcome of 30‐day mortality, but this was not possible for older trials where contact information was not available. We searched the reference lists of relevant reviews and transfusion trials.
Data collection and analysis
Selection of studies
Two authors (JLC and SS) independently screened the titles or abstracts of the search results, or both, and selected trials that met the inclusion criteria. We resolved disagreements by discussion until we reached consensus. We identified trials in which participants were randomized to a restrictive transfusion strategy (transfusion threshold or protocol, or both) or to a control group that was randomized to a liberal transfusion strategy.
Data extraction and management
Previously, JLC and Paul Carless (prior author) extracted all the data for the earlier versions of this review. For this 2016 update, using a data extraction form, JLC and SS independently extracted study characteristics and outcomes of new trials since the last review. Information recorded on the extraction form included study type, methodology descriptions, the presence of a transfusion threshold, transfusion protocol, the type of surgery involved, clinical setting, treatment outcomes, and general comments. JLC then entered data into Review Manager 5 (Review Manager 5a); NR checked data. We contacted authors of trials to request missing data.
We used a data extraction form to record data on the following outcomes: the number of participants exposed to allogeneic blood, the amount of allogeneic blood transfused, the number of participants receiving any transfusion (allogeneic blood, autologous blood, or both). For trials involving surgical participants, we recorded the following outcomes: postoperative complications (infection, haemorrhage, non‐fatal myocardial infarction, cardiac events, renal failure, stroke, thromboembolism, pulmonary oedema, mental confusion), mortality, and length of hospital stay (not reported in the review). We recorded data for blood loss and haemoglobin and hematocrit levels (on admission, pre‐ and post‐transfusion, and at discharge). We recorded information regarding demographics (age, sex), type of surgery, and medical condition on the data extraction form. We extracted data for allogeneic blood transfusion if it was expressed as packed RBCs. We documented information regarding the use of fresh frozen plasma or platelets, or both.
Assessment of risk of bias in included studies
We used the Cochrane tool for assessing risk of bias as described in section 8.5 of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011).
JLC and SS assessed the following domains for each study:
-
sequence generation;
-
allocation concealment;
-
blinding;
-
incomplete outcome data;
-
selective outcome reporting; and
-
other potential sources of bias.
We completed a 'Risk of bias' table for each study, incorporating a description of the study's performance against each of the above domains and our overall judgement of the risk of bias for each entry as follows: 'low', 'unclear' (indicating unclear or unknown risk of bias), or 'high' risk of bias.
Measures of treatment effect
We calculated the risk ratio (RR) for allogeneic blood transfusion in the intervention group compared with the control group and the corresponding 95% confidence intervals (CI) for each trial using a random‐effects model (Der Simonian 1986). We adopted a similar approach for other outcomes of transfusion. We also entered the mean number of units of RBCs transfused to each group and the corresponding standard deviations. We used the mean difference (MD) and 95% CI to express the average mean reduction in the number of units of RBC administered to the intervention group compared with the control group. When the event rate was low, we considered using the Peto odds ratio when criteria for this method were fulfilled.
Unit of analysis issues
The unit of analysis was the participant. In all of the trials except one (Jairath 2015), randomisation was at the individual participant level. In one trial in people with gastrointestinal bleeding (Jairath 2015), the randomisation was at the level of the hospital (cluster), but the analysis occurred at the level of the individual participant. The intraclass correlation coefficient (ICC) was very low (0.0001) for the outcome of mortality, and we therefore included the data considering the participant as the unit of randomisation and ignoring the clustering, but performed a sensitivity analysis excluding this trial, to see what effect, if any, it had on the analysis. We did not evaluate any outcomes with repeated measures.
Dealing with missing data
We performed all analyses on an intention‐to‐treat basis. We imputed no missing data. We received information on 30‐day mortality from three authors (DeZern 2016; Villanueva 2013; Webert 2008). The levels of missing data were acceptable.
Assessment of heterogeneity
We examined statistical heterogeneity using both the I² statistic and Chi² test. The I² statistic describes the percentage of total variation across studies due to heterogeneity rather than chance. A value of 0% indicates no observed heterogeneity, and larger values show increasing heterogeneity; moderate or substantial heterogeneity is considered to exist when the I² exceeds 50% or 85% respectively (Higgins 2011). For the Chi² test, we used a P value of < 0.10 to indicate the presence of statistically significant heterogeneity. Because of the anticipated significant clinical heterogeneity of the trials, we analyzed data using a random‐effects model. We also anticipated a high level of heterogeneity related to transfusion rates because practice in the different specialties of the trials would vary considerably according to speciality‐specific protocols. Therefore, as described later, we chose to provide a summary statistic for the outcomes of transfusion even when I² was very high, because of the clinically relevant information it provides.
Assessment of reporting biases
When there were more than 10 studies available, we examined funnel plots for the primary outcome of 30‐day mortality and the proportion of participants transfused for evidence of publication bias. We used the proportion of participants transfused because all trials reported this outcome, and it may reflect overall risk of publication bias better than 30‐day mortality, which was not reported in all of the trials.
Data synthesis
We performed all analyses using Review Manager software (Review Manager 5a). We entered data for the numbers of participants exposed to allogeneic blood and the numbers of participants in each treatment group into Review Manager. When studies presented transfusion volume as millilitres (mL), we converted these amounts to units by dividing by 300. We converted studies reporting haematocrit to haemoglobin concentration by dividing by three. We pooled the data for all outcomes and presented data stratified by subgroups for the primary outcome of 30‐day mortality and proportion of participants transfused. We used Peto odds ratios for the outcomes with event rates less than 1%. For continuous variables, we estimated the pooled mean difference and 95% CI using the generic inverse variance method.
Subgroup analysis and investigation of heterogeneity
The prespecified subgroups evaluated were clinical specialties (acute blood loss/trauma, cancer, cardiac surgery, critical care, orthopaedic surgery, myocardial infarction, vascular surgery, and haematological malignancies). We examined 30‐day mortality and the proportion of participants exposed to transfusion stratified by the transfusion threshold (difference between the liberal and restrictive transfusion thresholds: > 2 g/dL and < 2 g/dL) and restrictive transfusion threshold of less than 7 g/dL versus one of 8 g/dL to 9 g/dL. We also examined a post hoc subgroup of participants enrolled with myocardial infarction compared with all other clinical specialties, and we combined cardiac surgery with myocardial infarction because of emerging evidence that participants with acute myocardial infarction might differ from other anaemic participants (Carson 2013).
Where appropriate, as part of the exploration for clinical heterogeneity, we distinguished between adult and paediatric trials, for example, in analysis of the quantity of blood transfused (as this would not be directly comparable between adults and children), or in clinical settings where the widely used paediatric transfusion protocols differ.
For the primary outcome, we also compared findings between registered and unregistered trials.
Sensitivity analysis
We performed a sensitivity analysis to assess the effects of studies with a high risk of bias for allocation concealment and blinding of outcome assessment for the primary outcome only, as in earlier versions of the review, sensitivity analyses for secondary outcomes were not informative.
'Summary of findings' tables
We have presented the judgements about the quality of the evidence in a 'Summary of findings' table (according to guidelines developed by the GRADE Working Group) (Schünemann 2011). This table includes the following outcomes: number of people receiving blood transfusions, 30‐day mortality, myocardial infarction, congestive heart failure, cerebrovascular accident (stroke), rebleeding, pneumonia, and thromboembolism.
Results
Description of studies
Results of the search
In the first published version of this review (Hill 2002), we ran searches to 1999 and identified 10 included studies and one excluded study. We conducted the first update search in November 2004 but did not identify any new studies at the time. We conducted additional searches in 2009 and 2011 and identified nine new included studies. In the last published version (Carson 2012b), we included a total of 19 studies, excluded one study, listed one as awaiting classification, and identified two ongoing studies.
We conducted the most recent searches in April 2015, December 2015 and May 2016, which together retrieved 5727 records. After deduplication and screening, we identified 16 new, potentially relevant studies (Figure 1). On closer inspection two of these were new reports of studies already included in the 2012 review (So‐Osman 2013 and Villanueva 2013), one was previously awaiting classification (Cooper 2011), and two were previously listed as ongoing (Carson 2013/NCT01167582; Murphy 2015/ISRCTN70923932).
For this latest update we considered a total of 35 studies for inclusion; 14 new studies from searches run to May 2016 and 17 studies already included in the review. We excluded a total of four studies (two of these, Fortune 1987 and Zygun 2009, were previously included in the 2012 version of the review).
We also performed a search of the international trial registers in May 2016 and identified a further 11 studies of interest; nine are ongoing and two are completed but unpublished, so we have listed these as awaiting classification. Further updates of this review will incorporate the results of these studies (as appropriate).
PRISMA Flow Diagram
Included studies
Participants
Thirty‐one studies were eligible for inclusion in this review. The clinical settings were varied: 10 studies were in orthopaedic surgery (Carson 1998; Carson 2011; Fan 2014; Foss 2009; Gregersen 2015; Grover 2005; Lotke 1999; Nielsen 2014; Parker 2013; So‐Osman 2013); six were in critical care (de Almeida 2015; Hébert 1995; Hébert 1999; Holst 2014; Lacroix 2007; Walsh 2013); five were in acute blood loss or trauma (Blair 1986; Fisher 1956; Jairath 2015; Prick 2014; Villanueva 2013); five were in cardiac surgery (Bracey 1999; Hajjar 2010; Johnson 1992; Murphy 2015; Shehata 2012); two were in acute coronary syndrome (Carson 2013; Cooper 2011); two were in leukaemia and haematological malignancies (DeZern 2016; Webert 2008); and one was in vascular surgery (Bush 1997). One trial dealt with paediatric participants (Lacroix 2007).
Interventions
There was considerable variation in the definition of restrictive transfusion strategies specified in the protocols. These varied from 7.0 g/dL to 9.7 g/dL, with two further trials specifying haematocrit values of 25% or 30% (equivalent to haemoglobin levels of around 8 g/dL and 10 g/dL respectively). One trial administered blood for symptoms of anaemia (Parker 2013); and in another trial in postpartum haemorrhage (included in the acute blood loss or trauma grouping), no transfusion was administered in the restrictive group (Prick 2014), although red blood cell (RBC) transfusion was permitted for severe symptoms of anaemia or if physicians believed it was indicated. The liberal transfusion triggers varied: 100% of 'normal red cell volume' (Fisher 1956); two units of blood irrespective of clinical state (immediately in one trial (Blair 1986), postoperatively in another (Lotke 1999)); transfusion sufficient to maintain haemoglobin levels at or above 12 g/dL (Webert 2008), 11.3 g/dL (Gregersen 2015), 10 g/dL (Bush 1997; Carson 1998; Carson 2011; Carson 2013; Foss 2009; Grover 2005; Hébert 1995; Hébert 1999; Hajjar 2010; Jairath 2015; Parker 2013), 9.5 g/dL (Lacroix 2007; Shehata 2012), 9 g/dL (Bracey 1999; de Almeida 2015; Holst 2014; Murphy 2015; Villanueva 2013; Walsh 2013), 8.9 g/dL (Prick 2014), and 8 g/dL DeZern 2016. Three trials specified the liberal triggers as haematocrit levels of 32%, Johnson 1992, and 33%, Cooper 2011.
Trial design
In 34 out of 35 trials, the participant was the unit of randomisation and analysis. One trial used cluster randomisation by hospital (Jairath 2015). Ten trials included more than 100 participants. Four trials included over 900 participants (Carson 2011; Holst 2014; Jairath 2015; Murphy 2015). This systematic review included a total of 12,587 trial participants.
Excluded studies
From the earlier searches, we excluded one trial confined to participants with sickle cell disease, because the trigger was based on the level of sickle haemoglobin, not the haemoglobin or haematocrit level (Vichinsky 1995).
We excluded two trials included in earlier versions of this review (Fortune 1987; Zygun 2009), because they were not designed to evaluate clinical outcomes relevant to this review.
From the most recent searches, we excluded one trial because participants received concomitant erythropoietin (Robertson 2014), and hence, it did not fulfil the inclusion criteria.
Studies awaiting classification
Brief details of two completed but unpublished studies are shown in the Studies awaiting classification section.
Ongoing studies
Brief details of nine ongoing studies identified by searching the international trial registers to May 2016 are shown in the Ongoing studies section.
Risk of bias in included studies
The 'Risk of bias' tables detail the performance of the studies for each domain and are summarised in Figure 2 and Figure 3.
'Risk of bias' graph: review authors' judgements about each 'Risk of bias' item presented as percentages across all included studies. Thirty studies are included in this review.
'Risk of bias' summary: review authors' judgements about each methodological quality item for each included study.
Allocation
Sequence generation
We found low risk of bias for 20 trials, eight of which used computer randomisation and one of which used a table of random numbers to generate the allocation sequence for the participants. Another 11 trials used a variety of low‐risk methods. Two trials based the randomisation sequence on hospital record number, and we judged them to be at high risk of bias, while the remaining nine trials presented insufficient information to assess the adequacy of sequence generation, so we rated them as unclear.
Allocation concealment
We judged the risk of bias for this item to be low for 21 trials that used central allocation or sealed envelopes if appropriate safeguards (e.g. sequentially numbered envelopes) were used. We judged three trials to be at a high risk of bias; one of these trials used a cluster design, so everyone in hospital knew to which group all participants had been assigned (Jairath 2015). We rated seven studies as unclear because the publications did not provide any information about how allocation was concealed.
Blinding
Performance bias
The nature of the intervention means that blinding of clinicians involved in the care and administration of blood transfusions would not have been feasible. The extent to which this could have biased the results is unclear, but we rated 28 trials as being at unclear risk of bias for this domain. One trial reported that participants were blinded to treatment (Foss 2009). Thus, we have rated this study as being at low risk of bias for this domain. We rated one study as being at high risk because in a cluster design everyone knows the assigned group or cluster (Jairath 2015).
Detection bias
Outcomes are optimally assessed when assessors are blinded to assignment. It is possible to blind the assessment of many outcomes by using, for example, an adjudication committee. In contrast, for some outcomes such as death, blinded assessment is less useful. Since most trials evaluated multiple outcomes, it is possible that the potential bias for the assessment of one outcome would differ to that for another outcome. We classified risk of bias on the basis of the primary outcome of the trial. We judged the risk of bias to be low for 15 studies, high for three, and uncertain for 13 trials.
Incomplete outcome data
We rated 24 trials as being at low risk of bias for this domain as they either had no missing data or performed intention‐to‐treat analyses. Five trials reported a small number of exclusions, although the extent to which this may have introduced bias is uncertain; thus, we rated these trials as unclear. In two trials, the risk was high.
Selective reporting
We could not find any evidence of reporting bias. Although we did not have access to the trial protocols for the majority of trials, the results for the primary and secondary outcomes, as described in the methods sections of each trial, appeared clearly and concisely reported. Trial protocols were available for the trials with which the authors of this review had some degree of involvement (Carson 1998; Carson 2011; Hébert 1995; Hébert 1999; Lacroix 2007). We did communicate with one author who provided 30‐day mortality data (Villanueva 2013).
Other potential sources of bias
We identified no other sources of bias.
Effects of interventions
See: Summary of findings for the main comparison
There was substantial variation in the completeness of all review outcomes across the included trials. All of the trials contributed to the analysis comparing the proportion of participants transfused in the liberal and restrictive transfusion groups. Despite the heterogeneity in the methods and transfusion triggers reported in these randomized trials, it was possible to pool data, to varying degrees, for each of the review outcomes. See summary of findings Table for the main comparison.
Primary outcome
Thirty‐day mortality
Thirty‐day mortality is the primary outcome, and 23 trials reported data (N = 10,537 participants). There was no difference in the 30‐day mortality between restrictive and liberal transfusion strategies (risk ratio (RR) 0.97, 95% confidence interval (CI) 0.81 to 1.16; Analysis 1.1). Heterogeneity between these trials was not significant (Chi² = 29.75, df = 21 (P = 0.10); I² = 29%). The funnel plot demonstrates that the risk ratio for 30‐day mortality is symmetrically distributed, which indicates that there is not likely to be publication bias for this outcome (Figure 4).
Funnel plot of comparison: 1 Mortality, outcome: 1.1 30‐day mortality.
Subgroup analysis of 30‐day mortality: restrictive threshold of 8 g/dL to 9 g/dL versus 7 g/dL
We examined 30‐day mortality and stratified it by the restrictive transfusion threshold used in the trials. Fourteen trials, with 4772 participants, used a restrictive threshold of 8 g/dL to 9 g/dL. The risk ratio for 30‐day mortality was 1.05 (95% CI 0.78 to 1.40). Nine trials, with 5765 participants, used a 7 g/dL restrictive threshold (Analysis 1.2). The risk ratio for 30‐day mortality was 0.94 (95% CI 0.74 to 1.19). The test for subgroup differences was not significant (Chi² = 0.34, df = 1 (P = 0.56); I² = 0%), indicating that there was no difference in the mortality risk between the two thresholds.
Subgroup analyses of 30‐day mortality: clinical specialty
We examined 30‐day mortality and stratified it by the clinical speciality used in the trials: cardiac surgery, orthopaedic surgery, vascular surgery, acute blood loss or trauma (analyses for this grouping for 30‐day mortality included gastrointestinal (GI) bleeding only), critical care, acute myocardial infarction, or haematological malignancies. The overall risk ratio for 30‐day mortality stratified by clinical specialty was 0.97 (95% CI 0.81 to 1.16; Analysis 1.3). The test for differences in 30‐day mortality between the subgroups was not significant (Chi² = 9.78, df = 6 (P = 0.13); I² = 38.6%).
In two trials that reported mortality at 30 days in 2221 participants with acute blood loss or trauma (GI bleeding), the mortality was significantly lower using the restrictive strategy compared with the liberal strategy (RR 0.65, 95% CI 0.43 to 0.97; Analysis 1.3).
Two trials recruited 154 participants who had acute myocardial infarction and evaluated mortality after random allocation; for this subgroup, the mortality risk was higher in the restrictive strategy group than in the liberal strategy group (RR 3.88, 95% CI 0.83 to 18.13). We carried out a post hoc subgroup analysis that compared the 30‐day mortality in the two trials that included 154 acute myocardial infarction participants versus all other participants, but found no differences. The P value for subgroup differences was 0.08 (Chi² = 3.09, df = 1; I² = 67.6%; Analysis 1.4). Although we observed a very high risk ratio for the myocardial infarction participants (RR 3.88), the two included studies were very small, and hence, the pooled estimate is not robust.
In a separate subgroup, we combined five trials, with 3096 participants, those with myocardial infarction and those undergoing cardiac surgery; the risk ratio for 30‐day mortality was 1.28 (95% CI 0.80 to 2.06; Analysis 1.5), and there was very little heterogeneity (I² = 12%).
Subgroup analysis of 30‐day mortality: registered versus unregistered trials
Six of the included trials were not registered and were all conducted before 2000. This was expected, since more recent initiatives for research transparency dictate that all trials should be registered. There were no differences in the 30‐day mortality between the registered and unregistered trials (Analysis 2.1).
Sensitivity analysis
There were no differences in 30‐day mortality between trials with low versus unclear or high risk of bias in two bias domains, i.e. allocation concealment and blinding of outcome assessment (Analysis 3.1 and Analysis 4.1 respectively).
Secondary outcomes
Mortality at other time intervals
We analyzed mortality at hospital discharge (N = 5107, 10 trials; Analysis 5.1), 90 days (N = 3485, four trials; Analysis 5.2), and long term (N = 2016, one trial). There were no differences in mortality between transfusion strategies at each of the time points (mortality at hospital discharge: RR 0.86, 95% CI 0.73 to 1.01, Chi² = 8.67, df = 8 (P = 0.37); I² = 8%; 90‐day mortality: RR 1.15, 95% CI 0.95 to 1.40, Chi² = 3.76, df = 3 (P = 0.29); I² = 20%). In the one trial reporting long‐term mortality (Carson 2013), the hazard ratio was 1.09 (95% CI 0.95 to 1.25; P = 0.21). The results of mortality analyses at hospital discharge, 90 days, and long term are consistent with the results for mortality at 30 days.
Blood transfusions
Historically, exploration of differences in the proportion of participants transfused has been the analysis of preference in earlier versions of this Cochrane Review. As indicated in the Methods, we anticipated high levels of heterogeneity in the analysis of transfusion outcomes, for several reasons. In particular, standard 'control' rates of transfusion practice are very variable across the clinical specialties in which trials were identified for this update. These differing rates of transfusion policy reflect practice defined in speciality guidelines and recommendations. It is usually recommended that pooled estimates are not presented when there is such high heterogeneity. However, we have chosen to present the pooled results here, and our further justification for presenting these results for transfusion outcomes is presented in the Discussion.
Proportion of participants transfused
This analysis demonstrates the difference in the proportion of participants transfused in the liberal and restrictive arms of the trials. Data on the proportion of transfused participants were available from 31 trials (12,547 participants). The implementation of a restrictive transfusion trigger across all trials reduced the relative risk of receiving a RBC transfusion by 43% (RR 0.57, 95% CI 0.49 to 0.65; Analysis 6.1). Heterogeneity between these trials was large and significant (Chi² = 948.58, df = 30 (P < 0.00001); I² = 97%).
The proportion of participants transfused in the liberal and restrictive arms of the trials were very different across the different clinical specialities, with the largest difference evident in the subgroup of participants with acute blood loss (Analysis 6.2). The test for subgroup differences was significant: Chi² = 175.97, df = 6 (P < 0.00001); I² = 96.6%. The acute blood loss/trauma subgroup included diverse underlying illnesses for haemorrhage, including comorbidities. For example, Prick 2014 recruited young (otherwise healthy) women with postpartum haemorrhage, while Jairath 2015 enrolled older participants with gastrointestinal bleeding, characterised by many comorbidities. Prick 2014 contributed to a large extent to the high heterogeneity in this subgroup, and temporarily removing it from the analysis reduced heterogeneity to 77%. By contrast, participants enrolled in the subgroup of cardiac surgery trials demonstrated less variability in risk of transfusion across trials, and in this subgroup, we observed no heterogeneity. In the subgroup of critical care trials, the high heterogeneity (I² = 88%) was substantially reduced to 35% by temporarily removing the paediatric trial (Lacroix 2007). This sensitivity analysis, although post hoc, highlights how transfusion policies in this setting differed from adult protocols in a critical care setting.
The relative risk of transfusion was higher when the differences in haemoglobin transfusion thresholds between the restrictive and liberal transfusion arms were 2 g/dL or more, compared with those in which the difference was less than 2 g/dL (test for overall effect: Z = 2.17 (P = 0.03); Analysis 6.3).
There was no difference in the proportion of participants transfused between trials that used a restrictive transfusion threshold of 8 g/dL to 9 g/dL versus less than 7 g/dL (test for subgroup differences: Chi² = 0.13, df = 1 (P = 0.72); I² = 0%; Analysis 6.4).
The funnel plot for the proportion of participants transfused displays a grouping of trials with a risk ratio around 0.5 for receiving a transfusion in the restrictive transfusion arm (Figure 5), which is consistent with the overall observation that participants in the restrictive arm were transfused approximately half as often as those in the liberal arm. As expected, there were no studies in which participants in the restrictive arm were transfused more than the liberal arm.
Funnel plot of comparison: 2 Blood transfusions, outcome: 2.1 Participants exposed to blood transfusion (all studies).
Quantity of RBCs transfused
Twelve trials reported the quantities of blood transfused. The use of a restrictive transfusion trigger resulted in an average saving of 1.30 units of RBCs per transfused participant (mean difference (MD) ‐1.30, 95% CI ‐1.85 to ‐0.75; Analysis 6.5). Heterogeneity between these trials was, again, large and significant (Chi² = 139.91, df = 11 (P < 0.00001); I² = 92%).
Haemoglobin or hematocrit concentration
Sixteen trials reported the difference in haemoglobin or haematocrit levels between the liberal and restrictive transfusion arms. The timing of measurement varied. When we pooled data (without regard to timing, which was consistent within studies), participants assigned to a restrictive strategy had a haemoglobin concentration on average 1.32 g/dL lower than participants assigned to a liberal transfusion strategy (MD ‐1.32, 95% CI ‐1.64 to ‐0.99; Analysis 7.1). Heterogeneity between these trials was large and significant (Chi² = 867.08, df = 15 (P < 0.00001); I² = 98%).
Clinical outcomes
Cardiac events
Nine trials reported data on postenrolment cardiac events in 4849 participants. The risks of cardiac events (myocardial infarction, cardiac arrhythmias, cardiac arrest, pulmonary oedema, and angina) were not increased by the use of restrictive transfusion strategies (RR 1.04, 95% CI 0.79 to 1.39; Analysis 8.1). Heterogeneity between these trials was moderate and significant (Chi² = 21.25, df = 8 (P = 0.0007); I² = 62%). It is possible that participants were counted in more than one category of this composite outcome because these disorders are clinically inter‐related (for example, a participant could have angina that might lead to pulmonary oedema).
Myocardial infarction
Sixteen trials reported the outcome data on myocardial infarction (fatal and non‐fatal) in 8303 participants after random allocation to the liberal or restrictive transfusion arms. There was no difference between the restrictive and liberal transfusion strategies (RR 1.08, 95% CI 0.74 to 1.60; Analysis 8.2). There was no evidence of heterogeneity between trials (Chi² = 16.63, df = 15 (P = 0.34); I² = 10%).
Congestive heart failure
Twelve trials reported data for congestive heart failure in 6257 participants. There was no significant difference between the restrictive and liberal transfusion strategies (RR 0.78, 95% CI 0.45 to 1.35; Analysis 8.3). Heterogeneity between the trials was moderate and significant (Chi² = 20.98, df = 10 (P = 0.02); I² = 52%).
Cerebrovascular accident: stroke
Thirteen trials reported data for stroke in 7343 participants. There was no difference between transfusion strategies (RR 0.78, 95% CI 0.53 to 1.14; Analysis 8.4). Heterogeneity between the trials was not significant (Chi² = 9.28, df = 12 (P = 0.67); I² = 0%).
Rebleeding
Six trials reported data for rebleeding in 3108 participants. There was no difference between the restrictive and liberal transfusion strategies (RR 0.75, 95% CI 0.51 to 1.10; Analysis 8.5). Heterogeneity between these trials was statistically significant (Chi² = 11.95, df = 5 (P = 0.04); I² = 58%). In participants with acute blood loss or trauma, the risk of developing recurrent bleeding associated with restrictive transfusion was about half that of liberal transfusion (RR 0.75, 95% CI 0.51 to 1.10).
Sepsis/bacteraemia
Seven trials reported data for sepsis/bacteraemia in 3963 participants. There was no difference between the restrictive and liberal transfusion strategies (RR 1.03, 95% CI 0.79 to 1.35; Analysis 8.6). Heterogeneity between these trials was not statistically significant (Chi² = 6.28, df = 5 (P = 0.28); I² = 20%).
Pneumonia
Fourteen trials reported data for pneumonia in 6277 participants. There was no difference between the restrictive and liberal transfusion strategies (RR 0.94, 95% CI 0.80 to 1.11; Analysis 8.7). Heterogeneity between these trials was not statistically significant (Chi² = 9.83, df = 13 (P = 0.71); I² = 0.0%).
Pneumonia or wound infection
Fourteen trials reported data for infections in 9574 participants. The definition of infection was pneumonia or wound infection. There was no difference between the restrictive and liberal transfusion strategies (RR 0.96, 95% CI 0.86 to 1.07; Analysis 8.8). Heterogeneity between these trials was not statistically significant (Chi² = 19.00, df = 13 (P = 0.12); I² = 32%). Combining pneumonia with wound infections in the same analysis group has been the norm historically, but this was primarily due to a lack of sufficient data to separate these analyses. With the current number of trials included in the review and the large amount of participants, we had the ability to examine pneumonia separately rather than as a composite with wound infection ‐ and demonstrated that the results are similar.
Thromboembolism
Ten trials reported data for thromboembolism in 4019 participants. We calculated the risk ratio using the Peto method because the risk of thromboembolism was less than 1%. There was no difference between the restrictive and liberal transfusion strategies (RR 0.76, 95% CI 0.40 to 1.45; Analysis 8.9). Heterogeneity between these trials was not statistically significant (Chi² = 5.78, df = 9 (P = 0.76); I² = 0%).
Renal failure
Ten trials reported data for renal failure in 5929 participants. There was no difference between the restrictive and liberal transfusion strategies (RR 1.04, 95% CI 0.92 to 1.18; Analysis 8.10). Heterogeneity between these trials was not statistically significant (Chi² = 9.13, df = 9 (P = 0.43); I² = 1%).
Mental confusion
Six trials reported data for mental confusion in 1344 participants. There was no difference between the restrictive and liberal transfusion strategies (RR 0.92, 95% CI 0.65 to 1.30; Analysis 8.11). Heterogeneity between these trials was not statistically significant (Chi² = 5.53, df = 5 (P = 0.36); I² = 10%).
Functional recovery
Four trials reported functional outcomes in orthopaedic surgery participants. The functional measures were different in the trials, and hence, we could not pool them into a meta‐analysis, and only the numerical values from each trial are presented in graphs. Death or inability to walk at 30 days (RR 1.04, 95% CI 0.95 to 1.14) or 60 days (RR 0.99, 95% CI 0.87 to 1.11) was not significant between transfusion strategies (Analysis 9.1 ‐ pooled results not shown). No other measures of function were reported as significant between transfusion strategies in the single studies for each of the categories (Analysis 9.2 ‐ pooled results not shown).
Discussion
Summary of main results
We identified 31 randomized controlled trials that compared outcomes in participants allocated to receive transfusions of red blood cells (RBCs) at different haemoglobin concentration thresholds. These trials were undertaken between 1956 and 2016 and enrolled 12,587 participants across diverse patient populations. In the past four years, the number of participants in published trials evaluating transfusion thresholds has doubled.
The results of the meta‐analyses indicated that restrictive transfusion strategies led to a reduction of more than one‐third (43%) in the number of participants receiving at least one unit of blood, a red blood cell (RBC) transfusion requirement that was approximately 1.3 units lower, and a haemoglobin concentration that was around 1.32 g/dL lower than in the liberal transfusion groups. Most importantly, the meta‐analyses provided no evidence that restrictive transfusion policies harmed the participants or that they benefited from the use of liberal transfusion policies, within the parameters defined in the trials. Put another way, there was no evidence of an impact on clinically important outcomes when following a restrictive RBC transfusion policy compared with a liberal RBC transfusion policy. These findings may encourage the wider adoption of restrictive transfusion strategies, which would contribute to preserving blood supply.
Given the significant increase (a doubling) in the number of participants enrolled in transfusion threshold trials since this Cochrane Review was last published, this review has sufficient power to allow the primary outcome to be changed to 30‐day mortality. This increase also provided the opportunity to explore subgroups of participants with different underlying diseases, to ascertain whether the effects of RBC transfusion by haemoglobin thresholds are consistent across clinical specialities. This is important because there are pathophysiological reasons to postulate why transfusion might impact clinical outcomes differently in different patient populations, due to factors such as duration of anaemia (short term in critical illness versus long‐term transfusion dependence in bone marrow failure) or presence or not of an underlying restriction in cardiac function. Overall, across nearly all subgroups, the results indicated that risk of death and other adverse events were not impacted by either liberal or restrictive transfusion thresholds. Transfusion strategy did not influence the risk of cardiovascular events, including myocardial infarction, congestive heart failure, or stroke, although heterogeneity was observed in the trials that evaluated congestive heart failure (P = 0.01; I² = 57%).
In two subgroups of participant – those with acute blood loss and those with acute myocardial infarction, mortality may be influenced by a liberal or restrictive transfusion strategy, though the test for differences in 30‐day mortality between the subgroups was not significant (P = 0.13; I² = 41.2%). To be specific, in three trials (N = 1522) in participants with gastrointestinal bleeding (included in the acute blood loss or trauma grouping), a restrictive transfusion strategy was associated with a 35% lower risk of 30‐day mortality than a liberal transfusion strategy. The mechanism responsible for this significantly reduced risk of death may be due to a lower risk of rebleeding under restrictive transfusion regimens (risk ratio (RR) 0.65, 95% confidence interval (CI) 0.43 to 0.97; Analysis 1.3.4). The reason for this effect is not known, but may reflect higher vascular pressures following transfusion in the liberal transfusion group compared with the restrictive transfusion group. Participants enrolled with acute myocardial infarction constituted the other subgroup with apparently different results for 30‐day mortality. Two small trials included these participants (N = 154) for which the 30‐day mortality was 3.88 times higher in the restrictive transfusion group than in the liberal transfusion group (95% CI 0.83 to 18.13). These findings only provide a signal, since the results were not statistically significant and were based on a small number of participants.
Studies that used a restrictive threshold of 7 g/dL enrolled approximately half of the trial participants, and the other half used 8 g/dL as the restrictive threshold for transfusion. Most participants in the 7 g/dL restrictive transfusion threshold trials were based in critical care settings. The specialties were more varied in trials that tested an 8 g/dL restrictive transfusion threshold and included orthopaedic and cardiac surgery, gastrointestinal bleeding, and acute myocardial infarction. However, there was no apparent difference in the risk of death at 30 days between the two strata.
We compared 30‐day mortality in trials where the difference between the liberal and restrictive transfusion thresholds was at least 2 g/dL with those trials where the difference was less than 2 g/dL. Again, there was no evidence of dose‐effect of RBC transfusion by different trigger levels of haemoglobin concentration on clinical outcomes.
In this review, we compared the risk of infection in three ways but did not find evidence of a reduced risk of infection associated with restrictive transfusion. We combined pneumonia with wound infection (because they were the most common infections) and also examined sepsis or bacteraemia, and pneumonia (alone); the comparative risks of infection between the two transfusion strategies were nearly identical for all of these analyses. These results varied significantly from our prior analyses that had reported an elevated risk of infection in the liberal transfusion group (Rohde 2014). The change may be due to the incorporation of information from a recently reported cardiac surgery trial that had a large number of infections distributed equally between the liberal and restrictive transfusion strategies (Murphy 2015). Future trials may contribute to the power needed to examine subgroups.
Four trials assessed functional recovery, which used different measures, so meta‐analysis was not appropriate.
Overall completeness and applicability of evidence
With the expanding number of trials, the completeness of the evidence is increasing, and clinical trials have now evaluated many of the most common clinical specialities in which RBCs are transfused. Thus, the findings from this review are widely applicable to most clinical subgroups. However, we lack knowledge about the safety of different transfusion thresholds in groups of patients who frequently receive transfusion, but trials have not included. These understudied groups include those people with acute coronary syndrome, myocardial infarction, neurological injury/traumatic brain injury, acute neurological disorders, stroke, thrombocytopenia, cancer, haematological malignancies, and bone marrow failure.
Quality of the evidence
Overall, the quality of the evidence across the trials is good and shows improvement over time. The number of trials and participants enrolled has increased substantially, and the precision of the effect of transfusion has improved since the 2012 update of this review. We found relatively little heterogeneity for each clinical outcome across all analyses. However, we observed a significant amount of heterogeneity in the analyses evaluating the proportion of participants transfused, quantity of RBCs transfused, and differences in haemoglobin/haematocrit concentrations. It is conventional practice not to pool studies when there is such a large amount of heterogeneity; however, we chose to present the pooled results for these transfusion outcomes for several reasons. Firstly, the impact of restrictive transfusion on the proportion of participants transfused only varied by the magnitude of the reduction in transfusion, not the direction. In all of the trials, participants in the restrictive transfusion group received fewer transfusions, although the amount varied because the transfusion protocols were different and the clinical specialties required different frequencies of transfusion. Secondly, we expected the heterogeneity because of the variety of specialities for the clinical trials, including age; degree of comorbidities; and importantly, the policies for standard transfusion practice, which in turn reflect speciality‐specific guidelines and recommendations. At one extreme, nearly all participants, if not all, with leukaemia and cancer were transfused (DeZern 2016). Transfusion risk in participants in critical care, Hébert 1995; Hébert 1999; Lacroix 2007, or with acute blood loss, Villanueva 2013, was about 50% at the time of the studies.
In summary, we have chosen to present the pooled results for outcomes of transfusion because we are evaluating the effect of restrictive transfusion practice, and all study estimates for changes in transfusion are consistently in the same direction, and the substantial heterogeneity reflects the diversity in the strength of the estimates, rather than the efficacy of the policy. The reasons for the diversity in the strength of the trial estimates lies in the known and expected clinical specialities and the different practice guidelines used by different specialties. The subgroup explorations for transfusion outcomes reported earlier demonstrated these differences. The recognised differences in transfusion risk and amount by clinical specialities are a further justification for the more detailed analysis of all outcomes by clinical subgroups in this review, which we prespecified.
Potential biases in the review process
We performed extensive searches in an attempt to identify all eligible trials irrespective of publication status. Inspection of the funnel plots did not identify a major risk of publication bias (Figure 4; Figure 5).
'Risk of bias' evaluations revealed varying methodological issues between trials. We applied Cochrane methodology for defining high or low risk of bias to all trials, but acknowledged a number of challenges, including how to assign a single level of bias for multiple outcomes, for example, incomplete data or blinding (masking). Blinding the use of transfusion at the bedside is difficult to achieve unless study personnel are assigned to each participant, which would be an expensive procedure. The importance of blinding will differ according to the choice of primary trial outcome. For example, mortality is a hard endpoint and less open to bias than other functional outcomes.
Outcome assessment by observers who are blind to the treatment group is probably the most rigorous practical approach. Fourteen of the trials reviewed here reported this approach. Maintaining the integrity of the randomisation process becomes important if the trial is not to over‐estimate the benefit of the intervention (Schulz 1995). Some studies in this review did not report the methods used to conceal the allocation sequence from the treating clinicians. Ten trials used centralised allocation, and eight others used randomisation codes in sealed envelopes.
It is important to acknowledge limitations that apply to the findings of this review. The randomized trials may not have adequately evaluated some clinical outcomes specifically relevant to the use of RBC transfusions. Different grades of severity of cardiovascular events, such as myocardial infarction, congestive heart failure or stroke, or risk of overall infection, will occur in participants and may present in ways that are not always clinically overt and so are more subjective in interpretation. This is important because RBC transfusions may have both harmful and beneficial effects on the risk of these outcomes, for example, balancing prothrombotic tendencies against protective mechanisms to limit restrictions in myocardial oxygen delivery. Future studies need to establish robust definitions of all outcomes. The identified trials only evaluated the effect of transfusion in participants in hospital and not in outpatients in whom function and fatigue may be more important endpoints. Despite the large number of participants included in these trials, there remains inadequate power for many outcomes. Finally, a core rationale for RBC transfusion is to improve tissue and cellular oxygenation, but technologies for monitoring this directly are not available routinely; therefore, haemoglobin concentration is applied as a surrogate marker of need for transfusion, but it may not be a reliable biomarker.
Agreements and disagreements with other studies or reviews
The results of this review are consistent with other recent published systematic reviews. Holst 2015 recently identified 31 trials, with a total of 9813 randomized participants. We included 31 trials in our review, including a large, recently published trial in cardiac surgery (Murphy 2015), and a small pilot study in participants with haematological malignancies (DeZern 2016), but unlike Holst, we excluded studies in neonates (in whom the physiological responses to anaemia are considered very different from children and adults). The authors of Holst 2015 reported that restrictive transfusion strategies were associated with a reduction in the number of RBC units transfused and number of participants being transfused, with no effect on outcomes, including mortality, overall morbidity, and myocardial infarction. Restrictive compared with liberal transfusion strategies were not associated with risk of death (RR 0.86, 95% CI 0.74 to 1.01; 5707 participants), overall morbidity (RR 0.98, 95% CI 0.85 to 1.12; 4517 participants), or fatal or non‐fatal myocardial infarction (RR 1.28, 95% CI 0.66 to 2.49; 4730 participants). The inclusion of trials with unclear or high risk of bias did not affect results.
A number of systematic reviews have specifically examined participant subgroups. A review by Fominskiy aimed to assess the effect of liberal and restrictive RBC transfusion strategies on mortality in perioperative and critically ill adult participants through a meta‐analysis of relevant trials (Fominskiy 2015). Seventeen studies enrolled 7552 participants in perioperative settings, while 10 trials enrolled 3469 participants in critically ill settings. Participants in the perioperative period in the liberal transfusion strategy group had lower all‐cause mortality when compared with those allocated to the restrictive transfusion strategy group (odds ratio (OR) 0.81, 95% CI 0.66 to 1.00; P = 0.05; I² = 25%; number needed to treat for additional beneficial outcome = 97). There was no difference in mortality among critically ill participants receiving a liberal transfusion strategy when compared with the restrictive transfusion strategy (OR 1.10, 95% CI 0.99 to 1.23; P = 0.07; I² = 34%). The interpretation of systematic reviews that report only subgroups of the wider trial literature should be approached with some caution, to ensure that the patterns of findings are not inappropriately selective.
The results of these trials need to be viewed against large observational studies that compared clinical outcomes at varying haemoglobin levels in transfused and non‐transfused participants, finding conflicting results. In a study of 2202 participants undergoing coronary bypass surgery, the liberal transfusion group had a higher incidence of myocardial infarction than the conservative transfusion group (Spiess 1998). In a study of 8787 hip fracture participants, there was no difference in short‐ or long‐term mortality between participants transfused and not transfused (Carson 1998). In a study of 4470 intensive care unit participants, mortality was reduced in those receiving transfusion of up to six units of blood (Hébert 1997). A retrospective study of 78,974 Medicare beneficiaries found that blood transfusion was associated with a lower short‐term mortality rate among elderly patients with acute myocardial infarction if the haematocrit on admission was 30% or lower, and that blood transfusion may be effective with a haematocrit level as high as 33% on admission (Wu 2001). A study of 310,311 participants 65 years of age or older who underwent major non‐cardiac surgery found a 1.6% increase in 30‐day postoperative mortality for each 1% decrease in preoperative haematocrit (Wu 2007). Another study of 239,286 participants 65 years of age or older who underwent major non‐cardiac surgery found intraoperative blood transfusion was associated with a reduction in mortality in those with preoperative haematocrit levels of less than 24%, or in those with blood loss exceeding 500 mL (Wu 2010). The main limitation of these observational studies is that there may be residual confounding by indication, despite the extensive statistical adjustment of the results. It is possible that differences in participant characteristics between those that were transfused and non‐transfused may not have been identified or adequately adjusted for. This point is emphasised by the fact that a randomized controlled trial and an observational study (Hébert 1999 and Hébert 1997 respectively) in intensive care participants, performed by the same group of investigators, came to opposite conclusions. Even more convincing is how different ‐ and potentially biased ‐ the results of the meta‐analyses of observational data are; Chatterjee 2013 and Marik 2008 found a large increase in risk of death associated with transfusion. In contrast, the results of the meta‐analysis of clinical trials performed in this review showed no increased risk of death with liberal transfusion thresholds compared with restrictive transfusion thresholds. Despite recent assertions to the contrary (Benson 2000; Concato 2000), we believe that adequately powered, rigorously performed, randomized clinical trials are the only way of overcoming these limitations.
The transfusion policies reviewed here represent fairly small modifications to routine clinical practice. They are consistent with the recommendations of published clinical practice guidelines (AAGBI 2008; ASA 2006; BCTMAG 2003; Carson 2012a; Napolitano 2009; NBUGI 2001; Retter 2013; STSBCGTF 2011). The transfusion triggers (in terms of haemoglobin levels) were most often in the range of 7 g/dL to 10 g/dL. In fact, the 'restrictive' transfusion triggers in some trials were equivalent to the 'liberal' triggers used in other trials. Nevertheless, the trials documented significant reductions in the risk of RBC transfusion and worthwhile blood conservation. These effects are similar to what has been documented in meta‐analyses of trials of blood‐sparing techniques, such as cell salvage, Carless 2010a, and antifibrinolytic drugs, Henry 2011. Adoption of a conservative transfusion threshold appears to be as effective as these technologies in avoiding the need for transfusion and is likely to cost less.
Some guidelines have recommended RBC transfusion for symptoms or haemodynamic instability, rather than for a specific trigger haemoglobin level (AAGBI 2008; ASA 2006; Napolitano 2009; NBUGI 2001). A pilot study involving 84 participants tested this approach to transfusion (Carson 1998), as well as a trial involving 2016 participants (Carson 2011), and a 110‐participant trial in acute myocardial infarction (Carson 2013), in which participants could be transfused if they exhibited symptoms or had a haemoglobin concentration less than 8 g/dL. These studies found no difference in functional recovery, mortality, or morbidity in participants in the restrictive (symptomatic) transfusion group in the orthopaedic surgery trials (Carson 1998; Carson 2011); though in the trial involving participants with acute myocardial infarction (Carson 2013), there was a tendency toward worse outcomes in the restrictive transfusion group.
'Risk of bias' graph: review authors' judgements about each 'Risk of bias' item presented as percentages across all included studies. Thirty studies are included in this review.
'Risk of bias' summary: review authors' judgements about each methodological quality item for each included study.
Funnel plot of comparison: 1 Mortality, outcome: 1.1 30‐day mortality.
Funnel plot of comparison: 2 Blood transfusions, outcome: 2.1 Participants exposed to blood transfusion (all studies).
Comparison 1 Mortality at 30 days, Outcome 1 30‐day mortality.
Comparison 1 Mortality at 30 days, Outcome 2 30‐day mortality subgroup by restrictive haemoglobin level.
Comparison 1 Mortality at 30 days, Outcome 3 30‐day mortality subgroup analysis by clinical specialities.
Comparison 1 Mortality at 30 days, Outcome 4 30‐day mortality by clinical specialities: myocardial infarction versus all others.
Comparison 1 Mortality at 30 days, Outcome 5 30‐day mortality in participants with cardiac surgery and myocardial infarction.
Comparison 2 Subgroup analysis by registered trials, Outcome 1 30‐day mortality.
Comparison 3 Sensitivity analysis by allocation concealment, Outcome 1 30‐day mortality.
Comparison 4 Sensitivity analysis by blinding of outcome assessment, Outcome 1 30‐day mortality.
Comparison 5 Mortality: other time intervals, Outcome 1 Hospital mortality.
Comparison 5 Mortality: other time intervals, Outcome 2 90‐day mortality.
Comparison 6 Blood transfusions, Outcome 1 Participants exposed to blood transfusion (all studies).
Comparison 6 Blood transfusions, Outcome 2 Participants exposed to blood transfusion by clinical specialties.
Comparison 6 Blood transfusions, Outcome 3 Participants exposed to blood transfusion (by transfusion threshold).
Comparison 6 Blood transfusions, Outcome 4 Participants exposed to blood transfusion by transfusion threshold.
Comparison 6 Blood transfusions, Outcome 5 Units of blood transfused.
Comparison 7 Haemoglobin concentration, Outcome 1 Haemoglobin concentration.
Comparison 8 Morbidity: clinical outcomes, Outcome 1 Cardiac events.
Comparison 8 Morbidity: clinical outcomes, Outcome 2 Myocardial infarction.
Comparison 8 Morbidity: clinical outcomes, Outcome 3 Congestive heart failure.
Comparison 8 Morbidity: clinical outcomes, Outcome 4 Cerebrovascular accident (CVA) ‐ stroke.
Comparison 8 Morbidity: clinical outcomes, Outcome 5 Rebleeding.
Comparison 8 Morbidity: clinical outcomes, Outcome 6 Sepsis/bacteraemia.
Comparison 8 Morbidity: clinical outcomes, Outcome 7 Pneumonia.
Comparison 8 Morbidity: clinical outcomes, Outcome 8 Pneumonia or wound infection.
Comparison 8 Morbidity: clinical outcomes, Outcome 9 Thromboembolism.
Comparison 8 Morbidity: clinical outcomes, Outcome 10 Renal failure.
Comparison 8 Morbidity: clinical outcomes, Outcome 11 Mental confusion.
Comparison 9 Function and fatigue, Outcome 1 Function.
Comparison 9 Function and fatigue, Outcome 2 Function and fatigue.
| Restrictive compared with liberal transfusion protocols for guiding allogeneic red blood cell transfusion | ||||||
| Patient or population: Adults and children (haemodynamically stable) with potential need for RBC transfusion Settings: Inpatient Intervention: Restrictive transfusion protocol Comparison: Liberal transfusion protocol | ||||||
| Outcomes | Illustrative comparative risks* (95% CI) | Relative effect | Number of participants | Quality of the evidence | Comments | |
| Assumed risk | Corresponding risk | |||||
| Liberal transfusion (Hb 9 g/dL to 10 g/dL) | Restrictive transfusion (Hb 7 g/dL to 8 g/dL) | |||||
| People receiving blood transfusions | 841 per 1000 | 479 per 1000 | RR 0.57 (0.49 to 0.65) | 12,587 (31) | ⊕⊕⊕⊕ | ‐ |
| 30‐day mortality | 93 per 1000 | 90 per 1000 | RR 0.97 (0.81 to 1.16) | 10,537 (23) | ⊕⊕⊕⊝ | ‐ |
| Myocardial infarction | 17 per 1000 | 19 per 1000 | RR 1.08 (0.74 to 1.60) | 8303 (16) | ⊕⊕⊕⊕ | ‐ |
| Congestive heart failure | 36 per 1000 | 28 per 1000 | RR 0.78 (0.45 to 1.35) | 6257 (12) | ⊕⊕⊝⊝ | ‐ |
| Cerebrovascular accident (CVA) ‐ stroke | 17 per 1000 | 13 per 1000 | RR 0.78 (0.53 to 1.14) | 7343 (13) | ⊕⊕⊕⊕ | ‐ |
| Rebleeding | 163 per 1000 | 144 per 1000 | RR 0.75 (0.51 to 1.10) | 3108 (6) | ⊕⊕⊝⊝ | ‐ |
| Pneumonia | 82 per 1000 | 76 per 1000 | RR 0.94 (0.80 to 1.11) | 6277 (14) | ⊕⊕⊕⊕ | ‐ |
| Thromboembolism | 10 per 1000 | 8 per 1000 | RR 0.77 (0.41 to 1.45) | 4019 (10) | ⊕⊕⊕⊕ | ‐ |
| *The basis for the assumed risk (e.g. the median control group risk across studies) is provided in footnotes. The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). | ||||||
| GRADE Working Group grades of evidence | ||||||
| a. Downgraded for imprecision: could be up to 17 per 1000 more deaths in restrictive transfusion. | ||||||
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 1 30‐day mortality Show forest plot | 23 | 10537 | Risk Ratio (M‐H, Random, 95% CI) | 0.97 [0.81, 1.16] |
| 2 30‐day mortality subgroup by restrictive haemoglobin level Show forest plot | 23 | 10537 | Risk Ratio (M‐H, Random, 95% CI) | 0.97 [0.81, 1.16] |
| 2.1 Restrictive < 8 g/dL to 9 g/dL | 14 | 4772 | Risk Ratio (M‐H, Random, 95% CI) | 1.05 [0.78, 1.40] |
| 2.2 Restrictive < 7 g/dL | 9 | 5765 | Risk Ratio (M‐H, Random, 95% CI) | 0.94 [0.74, 1.19] |
| 3 30‐day mortality subgroup analysis by clinical specialities Show forest plot | 23 | 10537 | Risk Ratio (M‐H, Random, 95% CI) | 0.97 [0.81, 1.16] |
| 3.1 Cardiac surgery | 3 | 2942 | Risk Ratio (M‐H, Random, 95% CI) | 1.18 [0.77, 1.81] |
| 3.2 Orthopaedic surgery | 6 | 2831 | Risk Ratio (M‐H, Random, 95% CI) | 1.27 [0.72, 2.25] |
| 3.3 Vascular | 1 | 99 | Risk Ratio (M‐H, Random, 95% CI) | 0.98 [0.26, 3.70] |
| 3.4 Acute blood loss/trauma | 3 | 1522 | Risk Ratio (M‐H, Random, 95% CI) | 0.65 [0.43, 0.97] |
| 3.5 Critical care | 6 | 2840 | Risk Ratio (M‐H, Random, 95% CI) | 0.97 [0.75, 1.25] |
| 3.6 Acute myocardial infarction | 2 | 154 | Risk Ratio (M‐H, Random, 95% CI) | 3.88 [0.83, 18.13] |
| 3.7 Haematological malignancies | 2 | 149 | Risk Ratio (M‐H, Random, 95% CI) | 0.37 [0.07, 1.95] |
| 4 30‐day mortality by clinical specialities: myocardial infarction versus all others Show forest plot | 21 | 10388 | Risk Ratio (M‐H, Random, 95% CI) | 0.99 [0.82, 1.18] |
| 4.1 Myocardial infarction | 2 | 154 | Risk Ratio (M‐H, Random, 95% CI) | 3.88 [0.83, 18.13] |
| 4.2 All but myocardial infarction | 19 | 10234 | Risk Ratio (M‐H, Random, 95% CI) | 0.96 [0.81, 1.15] |
| 5 30‐day mortality in participants with cardiac surgery and myocardial infarction Show forest plot | 5 | 3096 | Risk Ratio (M‐H, Random, 95% CI) | 1.28 [0.80, 2.06] |
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 1 30‐day mortality Show forest plot | 21 | 10350 | Risk Ratio (M‐H, Random, 95% CI) | 0.98 [0.82, 1.18] |
| 1.1 Registered trials | 15 | 8773 | Risk Ratio (M‐H, Random, 95% CI) | 1.06 [0.84, 1.35] |
| 1.2 Unregistered trials | 6 | 1577 | Risk Ratio (M‐H, Random, 95% CI) | 0.80 [0.63, 1.02] |
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 1 30‐day mortality Show forest plot | 22 | 10477 | Risk Ratio (M‐H, Random, 95% CI) | 0.98 [0.82, 1.18] |
| 1.1 Low risk | 15 | 8936 | Risk Ratio (M‐H, Random, 95% CI) | 1.01 [0.82, 1.24] |
| 1.2 Unclear or high risk | 7 | 1541 | Risk Ratio (M‐H, Random, 95% CI) | 0.87 [0.56, 1.34] |
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 1 30‐day mortality Show forest plot | 21 | 10388 | Risk Ratio (M‐H, Random, 95% CI) | 0.99 [0.82, 1.18] |
| 1.1 Low risk | 16 | 9356 | Risk Ratio (M‐H, Random, 95% CI) | 1.00 [0.81, 1.24] |
| 1.2 Unclear or high risk | 5 | 1032 | Risk Ratio (M‐H, Random, 95% CI) | 0.93 [0.56, 1.55] |
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 1 Hospital mortality Show forest plot | 10 | 5107 | Risk Ratio (M‐H, Random, 95% CI) | 0.86 [0.73, 1.01] |
| 2 90‐day mortality Show forest plot | 4 | 3485 | Risk Ratio (M‐H, Random, 95% CI) | 1.15 [0.95, 1.40] |
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 1 Participants exposed to blood transfusion (all studies) Show forest plot | 31 | 12587 | Risk Ratio (M‐H, Random, 95% CI) | 0.57 [0.49, 0.65] |
| 2 Participants exposed to blood transfusion by clinical specialties Show forest plot | 31 | 12587 | Risk Ratio (M‐H, Random, 95% CI) | 0.57 [0.49, 0.65] |
| 2.1 Cardiac surgery | 5 | 3021 | Risk Ratio (M‐H, Random, 95% CI) | 0.67 [0.64, 0.70] |
| 2.2 Orthopaedic surgery | 10 | 3907 | Risk Ratio (M‐H, Random, 95% CI) | 0.50 [0.38, 0.67] |
| 2.3 Vascular | 1 | 99 | Risk Ratio (M‐H, Random, 95% CI) | 0.91 [0.77, 1.08] |
| 2.4 Acute blood loss/trauma | 5 | 2416 | Risk Ratio (M‐H, Random, 95% CI) | 0.39 [0.23, 0.67] |
| 2.5 Critical care | 6 | 2840 | Risk Ratio (M‐H, Random, 95% CI) | 0.63 [0.55, 0.72] |
| 2.6 Acute myocardial infarction | 2 | 155 | Risk Ratio (M‐H, Random, 95% CI) | 0.40 [0.19, 0.82] |
| 2.7 Haematological malignancies | 2 | 149 | Risk Ratio (M‐H, Random, 95% CI) | 1.00 [0.95, 1.05] |
| 3 Participants exposed to blood transfusion (by transfusion threshold) Show forest plot | 26 | Risk Ratio (M‐H, Random, 95% CI) | Subtotals only | |
| 3.1 Difference between liberal and restrictive haemoglobin thresholds ≥ 2 g/dL | 19 | 7683 | Risk Ratio (M‐H, Random, 95% CI) | 0.56 [0.48, 0.65] |
| 3.2 Difference between liberal and restrictive haemoglobin thresholds < 2 g/dL | 7 | 3903 | Risk Ratio (M‐H, Random, 95% CI) | 0.79 [0.63, 0.98] |
| 4 Participants exposed to blood transfusion by transfusion threshold Show forest plot | 31 | 12587 | Risk Ratio (M‐H, Random, 95% CI) | 0.57 [0.49, 0.65] |
| 4.1 Restrictive < 7 g/dL | 12 | 6453 | Risk Ratio (M‐H, Random, 95% CI) | 0.58 [0.47, 0.72] |
| 4.2 Restrictive < 8 g/dL to 9 g/dL | 19 | 6134 | Risk Ratio (M‐H, Random, 95% CI) | 0.55 [0.46, 0.67] |
| 5 Units of blood transfused Show forest plot | 12 | 4674 | Mean Difference (IV, Random, 95% CI) | ‐1.30 [‐1.85, ‐0.75] |
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 1 Haemoglobin concentration Show forest plot | 16 | 7791 | Mean Difference (IV, Random, 95% CI) | ‐1.32 [‐1.64, ‐0.99] |
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 1 Cardiac events Show forest plot | 9 | 4849 | Risk Ratio (M‐H, Random, 95% CI) | 1.04 [0.79, 1.39] |
| 2 Myocardial infarction Show forest plot | 16 | 8303 | Risk Ratio (M‐H, Random, 95% CI) | 1.08 [0.74, 1.60] |
| 3 Congestive heart failure Show forest plot | 12 | 6257 | Risk Ratio (M‐H, Random, 95% CI) | 0.78 [0.45, 1.35] |
| 4 Cerebrovascular accident (CVA) ‐ stroke Show forest plot | 13 | 7343 | Risk Ratio (M‐H, Random, 95% CI) | 0.78 [0.53, 1.14] |
| 5 Rebleeding Show forest plot | 6 | 3108 | Risk Ratio (M‐H, Random, 95% CI) | 0.75 [0.51, 1.10] |
| 6 Sepsis/bacteraemia Show forest plot | 7 | 3963 | Risk Ratio (M‐H, Random, 95% CI) | 1.03 [0.79, 1.35] |
| 7 Pneumonia Show forest plot | 14 | 6277 | Risk Ratio (M‐H, Random, 95% CI) | 0.94 [0.80, 1.11] |
| 8 Pneumonia or wound infection Show forest plot | 14 | 9574 | Risk Ratio (M‐H, Random, 95% CI) | 0.96 [0.86, 1.07] |
| 9 Thromboembolism Show forest plot | 10 | 4019 | Peto Odds Ratio (Peto, Fixed, 95% CI) | 0.76 [0.40, 1.45] |
| 10 Renal failure Show forest plot | 10 | 5929 | Risk Ratio (M‐H, Random, 95% CI) | 1.04 [0.92, 1.18] |
| 11 Mental confusion Show forest plot | 6 | 1344 | Risk Ratio (M‐H, Random, 95% CI) | 0.92 [0.65, 1.30] |
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 1 Function Show forest plot | 1 | Risk Ratio (M‐H, Random, 95% CI) | Totals not selected | |
| 1.1 Inability to walk or death at 30 days | 1 | Risk Ratio (M‐H, Random, 95% CI) | 0.0 [0.0, 0.0] | |
| 1.2 Inability to walk or death at 60 days | 1 | Risk Ratio (M‐H, Random, 95% CI) | 0.0 [0.0, 0.0] | |
| 2 Function and fatigue Show forest plot | 3 | Mean Difference (IV, Random, 95% CI) | Totals not selected | |
| 2.1 Lower extremity physical activities of daily living at 30 days | 1 | Mean Difference (IV, Random, 95% CI) | 0.0 [0.0, 0.0] | |
| 2.2 Lower extremity physical activities of daily living at 60 days | 1 | Mean Difference (IV, Random, 95% CI) | 0.0 [0.0, 0.0] | |
| 2.3 Instrumental activities of daily living at 30 days | 1 | Mean Difference (IV, Random, 95% CI) | 0.0 [0.0, 0.0] | |
| 2.4 Instrumental activities of daily living at 60 days | 1 | Mean Difference (IV, Random, 95% CI) | 0.0 [0.0, 0.0] | |
| 2.5 Energy/fatigue at 30 days | 1 | Mean Difference (IV, Random, 95% CI) | 0.0 [0.0, 0.0] | |
| 2.6 Energy/fatigue at 60 days | 1 | Mean Difference (IV, Random, 95% CI) | 0.0 [0.0, 0.0] | |
| 2.7 EuroQol (EQ‐5D) | 1 | Mean Difference (IV, Random, 95% CI) | 0.0 [0.0, 0.0] | |
| 2.8 'Timed up and go' test | 1 | Mean Difference (IV, Random, 95% CI) | 0.0 [0.0, 0.0] | |
