Transfusion after traumatic brain injury: seeking a target for a magic bullet?

Traumatic brain injury (TBI) is one of the leading causes of death and disability globally [1]. Management of TBI is focused primarily on preventing secondary brain injury through cerebral protection and prioritising brain tissue oxygen delivery [2]. Anaemia is common after TBI, especially in the context of surgical intervention or associated extracranial injury, and may increase the risk of cerebral ischaemia and worsen outcomes. As a result, it has been suggested that the transfusion threshold may be different in patients with TBI. In a global study of transfusion practices, half of respondents used a higher threshold in patients with acute brain injury than the 70–80 g.l-1 that is used typically in patients with critical illness and no neurological injury [3]. The transfusion trigger for patients with critical illness and no brain injury is well established. Evidence from the Transfusion Requirements in Critical Care trial showed a restrictive transfusion threshold of 70 g.l-1 was associated with a similar 30-day mortality as a liberal transfusion threshold of 100 g.l-1, whilst reducing the risk of complications such as secondary infection or lung injury [4]. Following this landmark study, there have been around 20 randomised controlled trials assessing transfusion targets in patients with critical illness. A consensus panel recommended that for most patients who are critically ill, a restrictive transfusion strategy is preferable to a permissive approach because this significantly decreases transfusion rates with no increase in the risk of death or complications [5]. For patients with acute coronary syndrome, a restrictive transfusion strategy was not advised, whilst guidance for patients with TBI was limited due to the analysed trials including very few patients with neurological injury. In this editorial, we will review the evidence for transfusion thresholds after severe TBI, with a focus on recently published randomised controlled trials identified by a targeted literature review (online Supporting Information Table S1). The premise that anaemia is deleterious to the brain following an acute brain injury implies that transfusion should promote brain tissue oxygenation. It is possible to assess brain tissue oxygen and metabolic state directly via a tissue oxygen monitor and cerebral microdialysis catheter [6]. In patients with TBI, the increase in haemoglobin concentration following a transfusion is positively associated with brain tissue oxygen tension (PbtO2) [7]. The lactate-to-pyruvate ratio is an indicator of the metabolic state of tissue, with higher values consistent with ischaemia and anaerobic metabolism. The relationship between changes in haemoglobin concentration and PbtO2 was absent in patients with a normal baseline lactate-to-pyruvate ratio. However, in patients with a starting lactate-to-pyruvate ratio > 25, PbtO2 increased significantly with a change in haemoglobin concentration following transfusion. This suggests that transfusion may be beneficial when signs of cerebral ischaemia or mitochondrial dysfunction are present [7]. In a second study, the use of a restrictive transfusion threshold of < 70 g.l-1 led to a significantly greater incidence of brain tissue hypoxia, with low PbtO2 values in 25% of patients allocated to the restrictive transfusion group, compared with 10% of those allocated to a threshold of 100 g.l-1 [8]. These studies suggest a potential beneficial effect of red blood cell transfusion with regard to tissue oxygenation following TBI. However, these studies did not measure long-term neurological follow-up. There have been several key studies reviewing liberal vs. restrictive transfusion strategies in patients with acute TBI. A large European multicentre observational study analysed data from 1590 patients, finding that 18% received at least one packed red blood cell transfusion within the first 7 days of admission, with a median haemoglobin before transfusion of 84 g.l-1. Requirement for transfusion was associated with a higher injury severity score; prehospital hypoxaemia or hypotension; and a lower Glasgow Coma Scale motor score. Multivariable regression showed that a higher haemoglobin concentration was independently associated with decreased mortality and improvement in functioning at 6 months [9]. However, the detrimental effects of packed red blood cells may occur when transfusion occurs at haemoglobin > 100 g.l-1, as this has been associated with increased mortality [10]. Two large multicentre randomised controlled trials assessed transfusion thresholds predominantly following TBI with associated anaemia [11, 12]. The transfusion thresholds for the restrictive group were similar at < 70 g.l-1, whilst the liberal transfusion threshold was 100 g.l-1 in the HEMOTION trial [11] and < 90 g.l-1 in the TRAIN trial, respectively [12]. The transfusion thresholds were maintained for the full duration of the ICU stay in the HEMOTION trial (mean 15 days) and for up to 28 days or until hospital discharge (mean 22 days) in the TRAIN trial. In the HEMOTION trial (n = 742), 73% of patients had a severe TBI, and the majority had extracranial injuries [11]. Mean haemoglobin concentrations at randomisation were 91 g.l-1. Patients were categorised at baseline into either a poor or good prognostic group based on a prognostic score defined previously for grading TBI severity (IMPACT score [13]). Functional outcomes were measured at 6 months using the extended Glasgow Outcome Scale (GOS-E) score. An unfavourable outcome was defined as a GOS-E score ≤ 3 for the poor-prognosis group and ≤ 5 for the good-prognosis group. Overall, 249/364 patients (68%) in the liberal-strategy group had an unfavourable outcome, compared with 263/358 (74%) in the restrictive-strategy group, with no significant difference noted between the two transfusion strategies (relative risk 0.93, 95%CI 0.83–1.04). Findings were consistent across those with the best and worst predicted neurological outcomes at baseline. No difference in mortality was observed. In the TRAIN study (n = 820), the pathology of the acute brain injury was more varied, with 61% of patients having suffered TBI, with the remainder having aneurysmal subarachnoid (22%) or non-traumatic intracerebral haemorrhage (18%) [12]. Mean haemoglobin at randomisation was 85 g.l-1. An unfavourable 6-month GOS-E score (≤ 5) occurred in 300/413 (73%) patients allocated to the restrictive group and in 246/393 (63%) those allocated to the liberal group (relative risk was 0.86, 95%CI 0.79–0.94, p = 0.002). Preceding these two studies, several smaller studies showed equivocal results [14, 15] (online Supporting Information Table S1). Meta-analyses including recent studies were unable to show a benefit with regard to neurological outcome with a liberal transfusion threshold [16, 17]. However, both meta-analyses, despite using different methodologies, identified that a study by Robertson et al. [14] was an outlier and omitted this study. Larcipretti et al., using a leave-one-out analysis, showed a significant improvement in neurological functioning with a liberal transfusion strategy (relative risk 1.24, 95%CI 1.06–1.45, p < 0.01) [17]. By using a predefined sensitivity analysis including only low risk-of-bias studies, the study by Blot et al. resulted in a relative risk of 0.74 (95%CI 0.63–0.87, p = 0.005) in favour of liberal transfusion [16]. There are two features from the study by Robertson et al. [14] which may contribute to its outlier status. First is that patients were allocated randomly to the respective transfusion groups when the median haemoglobin was > 140 g.l-1. Second, the haemoglobin concentration remained high even in patients allocated to the restrictive transfusion group, with a median haemoglobin > 95 g.l-1 throughout the study period. This may have led to less cerebral hypoxia as patients were less anaemic than in other restrictive transfusion groups. Although transfusion may offer potential benefits in terms of improved cerebral oxygen delivery, this must be balanced with the risk of the deleterious effects of packed red blood cell administration (e.g. immunomodulation; infection risk; transfusion-related lung injury). Adverse events in the TRAIN study were relatively well-balanced between groups, except for a higher likelihood of a cerebral ischaemic event in patients allocated to the restrictive group (57/423, 13.5% vs. 35/397, 8.8%). In the HEMOTION trial, there were no significant differences between groups for the pre-specified adverse outcomes (transfusion reactions, venous thromboembolic events, sepsis and acute respiratory distress syndrome). Both studies also assessed the adverse effects of transfusion, with no significant differences in infection or thromboembolic complications between the groups; however, an increased prevalence of acute respiratory distress syndrome was noted in one meta-analysis [17], but not the other [16]. In summary, combining the outcomes from the two most recent randomised controlled trials assessing a restrictive vs. liberal transfusion strategy in acute brain injury points towards a potential benefit in adopting a more liberal transfusion strategy, although it should be noted that neither trial was conclusive [11, 12]. This uncertainty was reflected in the transfusion guidelines from the European Society of Intensive Care Medicine, where no recommendation was made in relation to patients with acute neurological injury, although restrictive transfusion targets were recommended in patients who were critically ill with sepsis, acute coronary syndrome or after cardiac surgery [18]. In addition to a lack of definitive evidence, there are several other caveats that prevent a firm recommendation regarding the optimal transfusion trigger in patients with TBI. First is the conceptualisation of TBI as a single disease entity; it is a collection of heterogeneous disease types, which will often have different causes, severity, management and outcomes. For example, TBI may include one or more of the following: subdural haematoma; extradural haematoma; cerebral contusional injury; traumatic subarachnoid haemorrhage; diffuse axonal injury; or intraparenchymal haemorrhage. The underlying neuronal susceptibility to hypoxia, anaemia and perfusion pressure targets will likely be different in these subgroups and indeed the transfusion target may be patient- or disease-specific. Second, is the chronological history of TBI. It is unlikely that a liberal or restrictive transfusion target will be appropriate at all stages of a patient's ICU stay. Conceptually, anaemia may be most detrimental in the first few days following the injury. This has been noted in analyses from large multicentre databases [9, 19], which showed lower mortality in the first 7 days after injury with liberal transfusion. A restrictive transfusion target of ≤ 70 g.l-1 may be appropriate in the plateau stage of recovery, but this is yet to be clarified. Third, the duration of follow-up to determine ultimate neurological recovery has not been elucidated. Follow-up was limited to 6 months for the TRAIN and HEMOTION studies. However, it is thought that neurological recovery can continue up to 2 years following injury, with one study showing 54% of survivors moving from an unfavourable to a favourable neurological outcome over a period from 6 months to 24 months following TBI [20]. If neurological outcomes in TBI are not final at 6 months, there could still be divergence between restrictive and liberal transfusion strategies with respect to functional outcomes that occur beyond this follow-up period. We believe that there is mounting evidence that a more liberal transfusion strategy may offer outcome benefits, at least in the acute phase of TBI management. However, we acknowledge that there is heterogeneity within TBI with respect to pathology and that the optimal transfusion threshold may vary over the time-course of ICU admission. Randomised controlled trials have often shown limited evidence of outcome benefit for any intervention in the ICU when applied unselectively across a spectrum of patients and disease states, and these studies have stressed the need for a greater focus on individualised care. For example, the negative findings from randomised controlled trials assessing immunomodulation in sepsis may not reflect therapeutic failure, but rather the inability to identify a subset of patients who could benefit from therapies targeting the dysregulated host response [21]. Within TBI, there will be specific patients who would benefit from a higher haemoglobin concentration and a subset who would be better served by following a more conservative transfusion strategy. The decision of whether a transfusion is indicated should still be made on an individual basis, as we have seen that transfusion only improves PbtO2 in patients with evidence of cerebral ischaemia or mitochondrial dysfunction [7]. It is impossible to reduce the complexity of neuro-intensive care to a series of fixed targets, even for apparently arbitrary values such as haemoglobin concentration, mean arterial pressure or cerebral perfusion pressure. This is reflected in the results of surveys that show wide variation in current practice [3, 22]. Further research in personalising treatment may come from multimodal monitoring, such as pressure reactivity index for measuring cerebral autoregulation and the optimal cerebral perfusion pressure, and may aid clinicians in delivering transfusions to those patients who would benefit most [23]. In conclusion, due to the complexity of neuro-intensive care, clinicians will still make transfusion decisions on a case-by-case basis, weighing up the risks and benefits of transfusion. It is our hope that through the wider adoption of multimodal neurological monitoring, direct information from PbtO2 and cerebral metabolic status can be used to target transfusions to those patients who will benefit most. In the meantime, during acute neuroprotection in the early stages of TBI management, a transfusion trigger of < 90 g.l-1 will remain our default, given the lack of evidence of harm and the increasing suggestion of improvements in neurological recovery, which remains the most important outcome to both patients and their families. MW is the Editor-in-Chief of Anaesthesia. Table S1. Key randomised controlled trials assessing transfusion targets in traumatic brain injury. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.