Guidelines on the management of anaemia and red cell transfusion in adult critically ill patients.

Abstract

This document aims to summarize the current literature guiding the use of red cell transfusion in critically ill patients and provides recommendations to support clinicians in their day-to-day practice. Critically ill patients differ in their age, diagnosis, co-morbidities, and severity of illness. These factors influence their tolerance of anaemia and alter the risk to benefit ratio of transfusion. The optimal management for an individual may not fall clearly within our recommendations and each decision requires a synthesis of the available evidence and the clinical judgment of the treating physician. This guideline relates to the use of red cells to manage anaemia during critical illness when major haemorrhage is not present. A previous British Committee for Standards in Haematology (BCSH) guideline has been published on massive haemorrhage (Stainsby et al, 2006), but this is a rapidly changing field. We recommend readers consult recent guidelines specifically addressing the management of major haemorrhage for evidence-based guidance. A subsequent BCSH guideline will specifically cover the use of plasma components in critically ill patients. The World Health Organization (WHO) defines anaemia in men and women as a haemoglobin (Hb) <130 and <120 g/l, respectively, (Beutler & Waalen, 2006; WHO, 2011) and severe anaemia as <80 g/l (Guralnik et al, 2004; WHO, 2011). Anaemia is highly prevalent among the critically ill; 60% of patients admitted to intensive care units (ICU) are anaemic and 20–30% have a first haemoglobin concentration (Hb) <90 g/l (Hebert et al, 2001a; Vincent et al, 2002; Corwin, 2004; Walsh et al, 2004a, 2006a). After 7 d 80% of ICU patients have an Hb <90 g/l. Cohort studies indicate a strong association between anaemia and inferior outcomes, especially amongst those with cardiovascular disease (Carson et al, 1996; Hebert et al, 1997; Kulier et al, 2007; Wu et al, 2007). Haemodilution, blood loss and blood sampling are the most important initial contributors to anaemia in critical care. Impaired erythropoiesis secondary to inflammation is increasingly important with prolonged illness (Walsh & Saleh, 2006). Depending upon casemix, 30–50% of ICU patients receive red cell (RBC) transfusions (Walsh et al, 2004b; Walsh & Saleh, 2006). Ten percent of all RBCs transfused nationally are given in general ICUs (Walsh et al, 2004a). Studies suggest that only 20% of transfusions are to treat haemorrhage (Vincent et al, 2002); the majority are given for anaemia. Mean blood consumption ranges from 2 to 4 units per admission. The writing group was selected by the BCSH Transfusion Task Force with input from the Intensive Care Society to provide expertise in relevant physiology, pathophysiology, general intensive care and specific subgroups of critically ill patients. We did not undertake a formal systematic literature review. We agreed a priori a range of issues relating to general intensive care patients, and specific sub-groups of patients with relevant co-morbidities. These were subcategories relating to general intensive care, weaning from mechanical ventilation, ischaemic heart disease (IHD), sepsis, and neurocritical care. A MEDLINE database search was conducted from its inception to December 2011 using a range of broad search terms relating to red cell transfusion, critical care, and intensive care. The search strategy is available from the authors on request. The search yielded 4856 papers. These were sub-divided according to the pre-defined subcategories and reviewed by sub-group members allocated to each part of the guideline. At least two group members contributed to each subcategory section. Using this approach a total of 508 relevant papers were extracted and reviewed in full. Recent systematic reviews and guidelines produced by other groups were also reviewed where available. The quality of evidence was judged by predefined Grades of Recommendation, Assessment, Development and Evaluation (GRADE) criteria (Jaeschke et al, 2008). Strong recommendations, grade 1, are made when the group was confident that the benefits do or do not outweigh the harm and burden of cost of a treatment. Where the magnitude of benefit is less certain, grade 2, or suggested recommendations are made. The quality of evidence is rated as A – high quality randomized control trials, B – moderate, C – low, D – expert opinion only. The GRADE system is summarized in Table 1. Global oxygen delivery (DO2) from the heart to tissues is the product of arterial O2 content and cardiac output (Barcroft, 1920). Arterial O2 content is calculated by the O2 carried by haemoglobin plus the dissolved O2; in health >99% of O2 is transported bound to haemoglobin. Tissue hypoxia can occur during critical illness as a result of problems at all stages in the O2 cascade, including airway and pulmonary disease, inadequate cardiac function and reduced or maldistributed microvascular flow. Anaemia reduces O2 carrying capacity and there is strong biological plausibility in the belief that it causes tissue hypoxia. When tissue DO2 falls, O2 supply is maintained by compensatory mechanisms that increase O2 extraction. However, there is a critical DO2 at which these compensatory mechanisms are overwhelmed and O2 transport becomes directly proportional to O2 supply. In such circumstances, severe tissue hypoxia is much more likely to occur. Studies using normovolaemic haemodilution indicate that young adults can maintain an O2 supply at Hb concentrations of 40–50 g/l by increasing cardiac output and O2 extraction (Weiskopf et al, 2006). The heart and brain have high O2 extraction ratios, which limits these compensatory mechanisms. In addition, O2 consumption is increased in the critically ill. Therefore anaemia may be less well tolerated during critical illness. An assessment of the risk to benefit ratio of transfusion to improve O2 carrying capacity is a key consideration to optimise patient outcomes. The strongest evidence guiding transfusion policy in adult critically ill patients comes from the Transfusion Requirements In Critical Care (TRICC) study (Hebert et al, 1999). Patients with a Hb ≤90 g/l were randomized to either a relatively high Hb transfusion trigger of <100 g/l with a target of 100–120 g/l, the ‘liberal’ group, or a lower trigger of <70 g/l with a target of 70–90 g/l (the ‘restrictive’ group). Mortality was compared at 30 and 60 d, and a range of secondary outcomes compared. The restrictive group received 54% fewer units of blood and 33% received no blood transfusions in the ICU, whereas all of the liberal group were transfused. Thirty-day mortality in the liberal group was typical of general ICU populations (23·3%), but there was a non-significant trend towards lower mortality for the restrictive group (18·7%, P = 0·11). In two pre-defined subgroups, younger patients (aged < 55 years) and patients with lower illness severity [Acute Physiology and Chronic Health Evaluation (APACHE) II score < 20], the risk of death during 30-d follow up was significantly lower with the restrictive strategy. For patients aged <55 years those in the restrictive group had a 5·7% mortality vs. 13·0% for those in the liberal group [95% confidence interval (CI) for the absolute difference 1·1–13·5%; P = 0·028]. Similarly, for patients with an APACHE II score < 20, those in the restrictive group had an 8·7% mortality vs. 16·1% for the liberal group (95% CI for the absolute difference: 1·0–13·6%; P = 0·03). These differences represented a number needed to treat to benefit from restrictive over liberal transfusion of about 13 patients for these sub-groups. Overall, there were also lower rates of new organ failures in the restrictive group and a trend towards higher rates of Acute Respiratory Distress Syndrome in the liberal group (7·7% vs. 11·4%). These findings support using transfusions to maintain a Hb of 70–90 g/l. Concerns about the applicability of these results include the introduction of leucodepletion of red blood cells (RBCs), the storage age of RBCs, and risk of selection bias; few patients with cardiac disease were enrolled and there was a high clinician refusal rate. The results of the TRICC study have been corroborated by two recent studies. The Transfusion Requirements After Cardiac Surgery (TRACS) study found no difference in a composite end-point of 30-d mortality and severe comorbidity in cardiac patients prospectively randomized to a liberal or restrictive transfusion strategy (Hajjar et al, 2010). Most recently the ‘FOCUS’ (Transfusion Trigger Trial for Functional Outcomes in Cardiovascular Patients Undergoing Surgical Hip Fracture Repair) study of liberal or restrictive transfusion in high-risk patients after hip surgery showed no difference in mortality or mobility in the group assigned to the restrictive transfusion strategy (Carson et al, 2011). Importantly, although patients in the FOCUS trial were not critically ill, they were elderly and had a high prevalence of cardiovascular disease. Taken together the recent literature consistently shows no clear advantage with a liberal transfusion strategy. A suggested approach to transfusion in critical care is summarised in Fig 1. Critically ill patients do not generate a physiological increase in erythropoietin concentration in response to anaemia (Corwin et al, 1999, 2002, 2007; Hobisch-Hagen et al, 2001; Corwin, 2004; Shander, 2004; Hebert & Fergusson, 2006; Belova & Kanna, 2007; Arroliga et al, 2009; Bateman et al, 2009). Several trials have evaluated the efficacy and effectiveness of erythropoietin administration in critically ill patients. Methodological variations including different patient populations, and varying dosage regimens of both erythropoietin and iron therapy makes interpretation of these trials complicated. It appears on balance that a combination of iron supplementation and erythropoietin therapy can modestly decrease transfusion requirements, but the benefits become negligible when a transfusion trigger of 70 g/l is used (Corwin et al, 2007). No difference in patient outcomes has been demonstrated, except for a possible decrease in mortality among trauma patients. Erythropoietin therapy increases deep vein thrombosis, especially when prophylaxis is not used. Erythropoietin is not licenced for use in anaemic critically ill patients. The inflammatory response complicates the interpretation of iron indices in critical illness (Walsh & Saleh, 2006). Tests of iron status typically demonstrate an increased ferritin concentration whilst transferrin levels, the serum iron-to-iron binding ratio and transferrin saturation are decreased. Iron is shifted into macrophages resulting in a functional iron deficiency similar to the anaemia of chronic disease. Evidence of absolute iron deficiency is absent in most patients, and patients do not respond to iron supplementation alone (Walsh et al, 2006b; Munoz et al, 2008). There are no large randomized trials of iron monotherapy in critically ill patients, and excess iron may increase susceptibility to infection (Maynor & Brophy, 2007). The biochemical characteristics of anaemia in the critically ill are summarized in Table 2. Blood sampling contributes substantially to iatrogenic anaemia during critical illness (Smoller & Kruskall, 1986; Corwin et al, 1995; Zimmerman et al, 1997). Studies examining the magnitude of blood loss associated with routine phlebotomy indicate typical daily blood loss of approximately 40 ml (Foulke & Harlow, 1989; Fowler & Berenson, 2003; MacIsaac et al, 2003; Corwin, 2005; Chant et al, 2006; Harber et al, 2006; Sanchez-Giron & Alvarez-Mora, 2008). Available evidence suggests that blood conservation devices are infrequently used in the ICU, although few recent studies or surveys have been published (O'Hare & Chilvers, 2001). Several studies have assessed the impact of these devices. Two showed a significant reduction in blood loss, but without an effect on anaemia or RBC use (Foulke & Harlow, 1989; MacIsaac et al, 2003; Mukhopadhyay et al, 2010). One study (Mukhopadhyay et al, 2010) showed a reduction in the severity of anaemia and reduced RBC use with the Venous Arterial blood Management Protection (VAMP) system (Edwards Lifesciences, Irvine, CA, USA). Use of this device was associated with decreased requirements for RBC transfusion (control group 0·131 units vs. active group 0·068 units RBC/patient/d, P = 0·02). The intervention group also had a smaller reduction in Hb during ICU stay, 14·4 ± 20·8 vs. 21·3 ± 23·2 g/l; P = 0·02 (Mukhopadhyay et al, 2010). No cost-effectiveness evaluations of these systems in routine practice have been published. The use of small volume paediatric sampling bottles has also been consistently associated with reduced phlebotomy-related blood loss, without affecting assay quality (Harber et al, 2006; Sanchez-Giron & Alvarez-Mora, 2008). The need to ensure RBC transfusion is used only where appropriate is emphasized by concerns about adverse consequences (Fig 1). An increasing body of laboratory and clinical research has raised the possibility that stored RBCs have harmful effects, although the clinical consequences remain to be defined. Most cohort studies show associations between transfusion and adverse patient outcomes, including death, organ failure progression, infection and prolonged hospital stay (Marik & Corwin, 2008). However, the importance of residual confounding in these studies is uncertain. The risks of transfusion in the critically ill include those common to all blood transfusions (e.g. errors in administration) and those more specific to individual blood components (e.g. bacterial contamination in platelet transfusions). The key principles of safe administration of blood are summarized in the BCSH Guidelines for the Administration of Blood Components (Harris et al, 2009). In critically ill patients, transfusion-associated lung injury (TRALI) and transfusion-associated circulatory overload (TACO) are particularly relevant complications (Dara et al, 2005; Rana et al, 2006a; Gajic et al, 2007a,b; Khan et al, 2007). Any adverse events or reactions related to transfusion should be appropriately investigated and reported to local risk management, the Serious Hazards of Transfusion group (SHOT) and the Medicines and Healthcare products Regulatory Agency (MHRA) via the Serious Adverse Blood Reactions and Events (SABRE) system. SHOT defines TACO as acute respiratory distress with pulmonary oedema, tachycardia, increased blood pressure, and evidence of a positive fluid balance after a blood transfusion (Taylor et al, 2010). Assessing the true incidence of TACO is difficult due to the lack of a consensus definition. A single large study evaluating the incidence of TACO in critically ill patients, defined the condition as the onset of pulmonary oedema within 6 h of transfusion with a PaO2:FiO2 ratio of <300 mmHg or SaO2 of <90% on room air, bilateral infiltrates on a chest radiograph in the presence of clinically evident left atrial hypertension (Rana et al, 2006b). The different criteria used in these studies may account for the reported differences in incidence, varying from one in every 357 units of RBCs transfused, to the 2009 SHOT report (Taylor et al, 2010), which identified 34 cases of TACO, 33 attributable to RBCs, but only five cases confirmed as highly likely. TRALI is defined as the onset of pulmonary oedema within 6 h of transfusion with a PaO2:FiO2 ratio of <300 mmHg in room air, bilateral infiltrates on a chest radiograph in the absence of left atrial hypertension. TRALI was first reported in 1951 but did not receive widespread recognition until more aggressive transfusion support became established (Silliman et al, 2005). It is difficult to recognize and can occur after transfusion of plasma, platelets or RBCs. Rana et al (2006a) estimated the incidence of TRALI as in one in every 1271 transfusions. Blood donors in confirmed cases are typically multiparous women who have developed leucoagglutinins during pregnancy. Many Blood Transfusion Services have introduced a policy of sourcing plasma from male donors, which has reduced the incidence of TRALI (Chapman et al, 2009). When suspected, TRALI should be investigated systematically; a suggested procedure is summarized in Table 3. Cohort studies have explored the relationship between the age of blood and clinical outcomes, including hospital-acquired infections and mortality. Interpretation of these studies is difficult because of the problems of confounding and also lack of control of the RBC storage duration. Several, but not all, studies have found associations between the transfusion of older RBCs and adverse clinical outcomes (Zallen et al, 1999; Mynster & Nielsen, 2001; Offner et al, 2002; Koch et al, 2008; Petillä et al, 2011). There are no completed randomized trials comparing standard issue RBCs with either fresher RBCs or older RBCs; several are in progress (Lacroix et al, 2011). Current storage regulations are based on RBCs recovering effective O2-carrying function within 24 h of transfusion into the patient. The maximum duration of storage varies from 35 to 42 d between countries. Typically, ICU patients receive RBCs stored for 2–4 weeks, in part because blood banks often issue older RBCs as they tend to be transfused shortly after issue. RBC storage results in changes that potentially impair O2 release (2,3 DPG depletion) and limit capillary transit (decreased nitric oxide production; membrane changes; decreased deformability; increased adherence to endothelium). Accumulation of bioactive substances (cytokines, lipid mediators) in the supernatant could also have adverse effects, especially in countries transfusing non-leucodepleted RBCs (Tinmouth et al, 2006). Severe sepsis is the commonest reason for admission to the ICU in the UK, accounting for 30% of cases (Harrison et al, 2006), with mortality ranging from 10% to 40% (Angus et al, 2001). Sepsis is associated with impaired tissue DO2 through a range of mechanisms, including respiratory failure, poor cardiac function and abnormalities of microvascular flow. The physiological rationale for using blood transfusions is to correct reductions in O2-carrying capacity in anaemic patients. Tissue hypoxia is common during the early stages of sepsis. Resuscitation strategies include respiratory and cardiovascular support. The aim is to correct a low DO2 and meet tissue O2 demands. Evidence of benefit from RBC transfusion in early sepsis comes from a single centre study of goal-directed resuscitation (Rivers et al, 2001). Both groups in the study received fluid boluses and vasopressor drugs to achieve resuscitation targets comprising a central venous pressure ≥8 cm H2O and mean arterial pressure ≥65 mmHg. The goal-directed therapy group were monitored during the first 6 h of treatment by measuring the central venous oxygen saturation (ScvO2). In cases where the ScvO2 was <70%, patients received blood transfusions to maintain a haematocrit (Hct) of 0·30 (Hb ≈ 100 g/l) and/or dobutamine to increase cardiac output (Rivers et al, 2001). This intervention decreased the absolute risk of death in hospital by 16% (30·5% vs. 46·5%). One major difference between the groups was the early use of blood (64·1% vs. 18·5%). As this was a complex intervention it is difficult to attribute clinical benefit to a single component. However, when patients are anaemic and there is evidence of inadequate DO2 during early sepsis, a target Hb of 100 g/l is probably advisable. In early sepsis, ScvO2 <70%, mixed venous oxygen saturation (SvO2) <65%, or lactate concentration >4 mmol/l are widely considered consistent with the existence of tissue hypoxia although this may not be the case for patients with later, more established sepsis. Ongoing clinical trials are evaluating the importance of early goal-directed therapy in sepsis. The evidence base for RBC transfusions in patients managed during the later stages of critical illness resulting from sepsis is complex. The use of intravenous fluids, RBC transfusion and inotropic and/or vasopressor therapies to achieve ‘supra-normal’ values for DO2 has been discredited (Hayes et al, 1994; Gattinoni et al, 1995). Current evidence suggests that using RBC transfusions to achieve a Hb higher than 70–90 g/l has no clinical benefit once the patient has established organ failure beyond the early resuscitation period. A subgroup analysis of patients with severe infection in the TRICC trial failed to show benefit from liberal transfusion, with more deaths in the liberally transfused group (30-d mortality: restrictive group 22·6% vs. liberal group 29·7%; Hebert et al, 1999). Cohort studies have also examined the association between RBC transfusion and clinical outcomes in septic patients. Studies carried out before the introduction of leucodepletion reported associations between RBC transfusion and higher mortality, whereas those performed after leucodepletion have reported lower mortality, raising the possibility that this may influence the risk to benefit profile of transfusion in these patients (Hebert et al, 1999; Vincent et al, 2002, 2008; Corwin et al, 2004). Ongoing trials are comparing restrictive versus liberal transfusion practice for patients with sepsis. Best transfusion practice when a patient with established critical illness develops a second episode of severe sepsis during an ICU stay, such as bacteraemia or ventilator-associated pneumonia, is uncertain – no prospective trials are available to guide management in this situation. Under these circumstances clinicians should use changes in available physiological indicators of O2 supply-demand balance, such as lactate, acid-bases status, ScvO2 and SvO2 together with clinical judgement to guide transfusion practice. Current evidence does not support transfusion to a Hb > 90–100 g/l. Cerebral DO2 is derived from the cerebral blood flow (CBF) and the arterial O2 content. Following brain injury, several factors converge to impair cerebral DO2, including hypoxaemia, hypovolaemia, raised intra-cranial pressure (ICP), vasospasm, failure of cerebral autoregulation and disruption of flow-metabolism coupling (Mendelow, 1988). The cerebral tissues compensate for a fall in DO2 by increasing their oxygen extraction ratio (O2ER), but this compensatory mechanism has limits and damaged brain tissue with a high O2ER is particularly vulnerable to ischaemia and secondary injury. Measurement of brain tissue O2 partial pressure (PbtO2) confirms that cerebral ischaemia is consistently associated with poor outcomes following brain injury and maintaining adequate DO2 to prevent cerebral ischaemia is central to the management of critically ill neurological patients. Although anaemia is common in patients admitted to the ICU following brain injury, the manipulation of the Hct to maintain cerebral DO2 remains contentious. While increasing the Hct increases O2 carrying capacity, there is an inverse relationship between Hct and blood viscosity and high Hct levels have been shown to reduce CBF and may predispose to cerebral ischaemia (Pendem et al, 2006). There are few prospective studies that have attempted to define the optimal Hct in critically ill neurological patients and current understanding is largely drawn from single centre observational studies and expert opinion. While the use of a restrictive transfusion strategy may improve outcomes in most critically ill adults, it remains unclear whether these findings can safely be applied to neurocritical care patients (Hebert et al, 1999; Vincent et al, 2002; Corwin et al, 2004). There is little evidence that blood transfusion improves outcome in anaemic patients with brain injury and transfusion itself appears to be associated with unfavourable outcomes in several studies. The evidence is considered in the context of traumatic brain injury (TBI), subarachnoid haemorrhage (SAH) and ischaemic stroke. Delayed cerebral ischaemia is a major cause of secondary injury following TBI (Dhar et al, 2009). Clinical markers of cerebral oxygenation are predictive of unfavourable outcome in these patients (Gopinath et al, 1994; Valadka et al, 1998; van den Brink et al, 2000). Maintenance of adequate cerebral DO2 and prevention of cerebral ischaemia is essential (Elf et al, 2002; Patel et al, 2002; Al Thanayan et al, 2008). Strategies to maintain CBF focus largely on maintaining adequate cerebral perfusion pressure and the avoidance of excessively raised ICP. The Brain Trauma Foundation (BTF) has published widely adopted guidelines on the management of the above parameters (BTF, 2007). These guidelines make no recommendation on the optimal Hb target to maximize cerebral DO2. A number of observational studies suggest that anaemia is associated with poor outcomes following TBI (Angus et al, 2001; Rivers et al, 2001; Hollenberg et al, 2004; Harrison et al, 2006; Dellinger et al, 2008; Sanchez-Giron & Alvarez-Mora, 2008; Bennett-Guerrero et al, 2009) but the association of anaemia with mortality is not a universal finding (Carlson et al, 2006; Schirmer-Mikalsen et al, 2007). Whilst RBC transfusion improves cerebral oxygenation in most anaemic patients with TBI, the increment is frequently small and PbtO2 actually appears to decrease in some patients following transfusion (Smith et al, 2005; Leal-Noval et al, 2006; Zygun et al, 2009). It has been speculated that this variation in clinical effect may be attributable to the storage age of blood, but this remains unproven (Leal-Noval et al, 2008). The influence of RBC transfusion on the outcome of TBI is unclear. Transfusion itself is associated with poor outcome, but in cohort studies this could be due to confounding (Carlson et al, 2006; Salim et al, 2008). A retrospective subgroup analysis of the TRICC study, which included 67 patients with moderate to severe TBI, suggested no significant improvement in mortality in patients randomized to a liberal (Hb 100–120 g/l) as compared to restrictive (Hb 70–90 g/l) transfusion strategy (Hebert et al, 1999; McIntyre et al, 2006). Although underpowered, this suggests a restrictive transfusion strategy may be safe in this group of patients. The Lund approach to the management of TBI uses a combination of measures to preserve the normal colloid and osmotic pressure across the disrupted blood brain barrier following TBI, including RBC transfusion to maintain a Hb > 100 g/l; a small single-centre non-randomized study has suggested improved outcomes using this approach, but the use of this technique remains controversial (Eker et al, 1998). In summary, there is insufficient evidence to reach an evidence-based conclusion on the optimal Hb target. Anaemia is consistently associated with unfavourable outcome in patients with SAH and it is uncertain whether transfusion improves outcome (Naidech et al, 2006, 2007; Wartenberg et al, 2006; Kramer et al, 2008). While transfusion improves cerebral DO2 in anaemic patients with SAH, it may decrease brain tissue oxygenation in others (Smith et al, 2005). Transfusion has been associated with reduced mortality in two observational studies (Dhar et al, 2009; Sheth et al, 2011). A small prospective randomized feasibility study, in which patients with SAH were randomized to a Hb target of either >100 or >115 g/l, has suggested only a trend towards improved secondary outcomes, reduced infarction rate and greater rates of functional independence with restrictive transfusion, but large randomized studies are lacking. Retrospective studies have suggested an association between RBC transfusion and poor outcome (Smith et al, 2004; De Georgia et al, 2005; Kramer et al, 2008; Tseng et al, 2008). Conversely haemodilution, targeting a Hct of approximately 0·30, has been used in combination with induced hypertension and hypervolaemia (triple-H therapy) in the treatment and prevention of cerebral vasospasm following SAH (Lee et al, 2006). Definitive studies demonstrating the efficacy of triple-H therapy are lacking, and it is unclear whether reduced blood viscosity and/or reduced Hb are responsible for the benefits reported (Dankbaar et al, 2010). The optimal Hb in patients with SAH has not been defined. It remains unclear whether the use of RBC transfusion improves (or worsens) outcomes. Observational studies in patients with ischaemic stroke suggest that the effect of Hct on outcome is u-shaped, with both high and low Hb associated with unfavourable outcome (Diamond et al, 2003; Kramer et al, 2008). Although high Hcts predispose to cerebral ischaemia and reduced reperfusion, RCTs have failed to show significant benefit from modest haemodilution (Asplund, 2002; Allport et al, 2005). An observational study examining CBF in patients with ischaemic stroke suggests that cerebral DO2 is maximal with a Hct of 0·40–0·45, a similar range to that in healthy volunteers (0·42–0·45; Kusunoki et al, 1981; Gaehtgens & Marx, 1987). Diamond's study of 1012 patients with ischaemic stroke demonstrated that the most favourable outcomes occurred in patients with Hcts of approximately 0·45 (Diamond et al, 2003). The impact of transfusion in anaemic patients admitted to the ICU following ischaemic stroke has not been evaluated. There is insufficient evidence to recommend a specific lower Hb target (or transfusion trigger) in patients admitted to neurocritical care following ischaemic stroke. Anaemia is a risk factor for adverse cardiovascular events and death for patients with acute and chronic IHD (Hajjar et al, 2010; Carson et al, 2011). It is unknown if RBC transfusion modifies this relationship. Coronary perfusion occurs primarily during diastole, especially to the left ventricle, which has highest O2 demand. The O2ER of the coronary system is high, meaning that matching the increased O2 demand requires an increase in coronary blood flow. Anaemia decreases the O2 content of blood per unit volume and occlusive coronary disease restricts blood flow; these factors increase the risk of ischaemia. During critical illness, cardiac work can also be significantly increased as a result of the increased global O2 requirements, while hypotension and tachycardia may reduce diastolic coronary blood flow. There is, therefore, biological plausibility that anaemia is tolerated poorly by patients with IHD. Two cohort studies of perioperative and critically ill patients found an association between anaemia and mortality in patients with IHD (Carson et al, 1996; Hebert et al, 1997). In both studies a Hb below 90–100 g/l was associated with excess mortality. These observations were corroborated by others demonstrating associations between anaemia and higher mortality in general surgical populations, particularly among older patients (Wu et al, 2001; Kulier et al, 2007). In the TRICC trial, there were no excess adverse cardiac events in the patients managed with a restrictive transfusion strategy. The proportion of patients who suffered a myocardial infarction (MI) post-randomization was higher in the liberal group (0·7% vs. 2·9%, P = 0·02), and overall cardiac adverse events were also higher (13·2% vs. 21·0%, P < 0·01). In a post hoc subgroup analysis of 257 patients who were documented as suffering from IHD at baseline, there was a non-significant trend towards lower 30-d mortality among patients managed with the liberal strategy (difference in 30-d survival 4·9% (95% CI 15·3% to −5·6%); these data suggested possible benefit from liberal blood use in patients with known IHD, but the sub-group analysis was underpowered. In contrast, the recently published FOCUS study in elderly patients undergoing hip fracture surgery, which compared a liberal strategy (Hb < 100 g/l) with a restrictive strategy (symptomatic anaemia or Hb < 80 g/l), found no difference in mortality or cardiovascular complications despite 40% of patients having IHD (Carson et al, 2011). Similarly, the TRACs study compared similar liberal and restrictive transfusion strategies in patients undergoing elective cardiac surgery and found no differences in 30-d mortality or severe morbidity between the groups (Hajjar et al, 2010). Although these trials were not in critically ill patients, both included patients at high risk of coronary events. There are no large randomized trials of transfusion strategies for patients with acute coronary syndromes (ACS). A recent small pilot study in 45 patients compared liberal and conservative transfusion approaches in patients with an acute myocardial investigation (Cooper et al, 2011). The primary outcome measure of in-hospital death, recurrent MI or worsening of congestive cardiac failure occurred in eight patients in the liberal group and three in the conservative arm (38% vs. 13%; P = 0·046). The majority of our current evidence is based on the physiological rationale for maintaining a higher blood O2 content, and data from cohort studies. Anaemic patients developing an ACS have worse outcomes (Guralnik et al, 2004). An older population, the widespread use of antiplatelet therapy, together with potential blood loss during percutaneous revascularization procedures, have increased the prevalence of anaemia among patients with ACS. Wu et al (2001) analysed approximately 79 000 US patients in the Medicare database aged >65 years presenting with acute MI. After statistical adjustment for confounding, transfusion improved 30-d mortality for patients with a Hct < 0·33%; the benefit of transfusion appeared highest among those patients with most severe anaemia (Wu et al, 2001). In separate cohort studies, data were analysed from trials of non-transfusion interventions for ACS (Rao et al, 2004; Yang et al, 2005). Although anaemia was associated with worse patient outcomes, these studies found no benefit from transfusion at lower Hb, and transfusion was associated with worse outcomes. Wu et al (2001) compared the impact of transfusion in patients with ST segment elevation myocardial infarction (STEMI) and non-STEMI. In this cohort, anaemia (Hb < 140 g/l) was associated with increased mortality in STEMI and RBC transfusions were associated with decreased risk. Conversely, for non-STEMI cases, anaemia (<110 g/l) was associated with increased mortality, but RBC transfusions were associated with increased risk. More recent cohort studies also did not find clinical benefit from transfusion when the Hb was >80–90 g/l (Aronson et al, 2008; Alexander et al, 2009). All cohort studies are limited by confounding, and the quality of evidence is low. Weaning consists of liberation from mechanical ventilation and extubation. Strategies to improve the speed and success of weaning are of particular relevance because they are likely to be both clinically effective and cost saving. Depending on case-mix, up to 25% of patients exhibit delayed weaning, and 5–10% continue to require ventilation at 30 d (Make, 1995; Ely et al, 1996). Extubation failure is associated with a sevenfold increase in mortality (Epstein et al, 1997; Jurban & Tobin, 1997). Weaning failure can be associated with an imbalance in O2 supply and demand. As weaning commences, DO2 maybe reduced by a lower Hb and a lower cardiac output while increases in maximal O2 uptake (VO2) occur due to the extra work of independent breathing (Walsh & Maciver, 2009). Two studies have shown an association between anaemia and a failure to wean (Ouellette, 2005; Silver, 2005). Increasing DO2 by increasing the Hb using transfusion potentially improves arterial O2 content and is the physiological basis of using RBC transfusion to assist weaning. Schonhofer et al (1998) studied normal and chronic obstructive pulmonary disease (COPD) patients and noted that transfusion reduced the work of breathing in the COPD group. A small, five patient case series by the same author suggested transfusion may be beneficial in weaning ventilated anaemic COPD patients (Schonhofer et al, 1998). However, larger studies in a more heterogeneous group of ventilated patients have shown either no benefit from transfusion, or suggested that it is associated with a worse outcome (Schonhofer et al, 1998; Hebert et al, 2001b; Vamvakas & Carven, 2002; Levy et al, 2005). The two largest studies are subgroup analyses of other studies – TRICC and CRIT (Hebert, 1998; Hebert et al, 2001b; Corwin et al, 2004). Both provide weak evidence because they were not designed to evaluate weaning or the effect of RBC transfusion on weaning duration. Vamvakas and Carven (2002), suggested that transfusion was associated with an increased duration of mechanical ventilation. Available evidence does not allow strong recommendations specific to transfusion and weaning from mechanical ventilation, but existing data do not support the use of a liberal transfusion strategy. Anaemia is prevalent in the critically ill and is associated with adverse outcomes. At present there are no clinically or cost-effective alternatives to RBC transfusion for rapidly increasing the Hb and restoring O2 carrying capacity. The prospective and observational data that is available consistently suggests that transfusion of RBCs when the Hb is within the 70–90 g/l range has no beneficial effect on clinical outcomes either in the general critical care population, or in specific patient sub-groups for whom a physiological rationale for reduced anaemia tolerance exists. Importantly, it is currently uncertain whether the lack of effectiveness of blood transfusions in this population is because anaemia itself does not affect outcomes or because the risks associated with current stored red cell transfusions out weigh physiological benefits. In the future, large well designed, prospective randomized control trials are required to further evaluate the risk to benefit balance of RBC transfusion in a range of acute conditions resulting in critical illness. While the advice and information in these guidelines is believed to be true and accurate at the time of going to press, neither the authors, the British Society for Haematology nor the publishers accept any legal responsibility for the content of these guidelines. Summary of the literature used to write the guideline and provide the key recommendations Single centre, retrospective cohort study: 31 patients Non-leucocyte deplete blood Prospective double-blind randomized control trial: 22 patients Patients randomized to either units ≤5 d old or units ≥20 d old, if Hb <90 g/l Double-blind, multicentre, randomized controlled study: 57 patients Patients randomized to receive RBCs ≤8 d old versus conventional therapy Single centre, retrospective study: 2732 patients Retrospective study 2872 patients received blood <14 d old 3130 patients received blood >14 d old Prospective case-series: 19 patients Transfusion of 2 units RBCs in patients on a surgical ICU Randomized controlled trial: 838 patients Randomization to one of two transfusion strategies Liberal – Hb maintained above 100 g/l Restrictive – Hb maintained at 70–90 g/l Single centre, randomized control study: 263 patients Combined series of interventions, including targeting Hct > 0·30% Prospective observational study: 35 patients Transfusion of 1–2 units of RBCs Tertiary referral centre: 20 patients Patients received 1 unit of RBCs for each 10 g/l that their Hb was <110 g/l