Background
Anemia is highly prevalent during critical illness; 60% of patients admitted to intensive care units (ICUs) are considered anemic, and 20
–30% have a first hemoglobin (Hb) concentration less than 9 g/dl [
1]. Moreover, after a 1-week ICU stay, up to 80% of patients will have Hb concentrations below this value [
2]. Cohort studies have suggested a strong association between anemia and poor outcome in heterogeneous populations of critically ill patients [
3,
4], especially among those with a history of cardiovascular disease [
5]. However, the optimal threshold to trigger red blood cell (RBC) transfusion in ICU patients is not clearly established and may depend on other confounders related to the underlying disease or to patient characteristics.
It has been shown that tissue oxygen delivery (DO
2) is dependent on organ blood flow and arterial oxygen content, which includes the Hb concentration and its oxygen saturation [
6]. Thus, during anemia, an increase in local blood flow or in oxygen extraction is necessary to maintain adequate DO
2 and avoid tissue hypoxia. In healthy conditions, brain tissue has a high oxygen extraction rate, which may limit its ability to further compensate for reduced DO
2 [
7]. Thus, cerebral vasodilation and increased cerebral blood flow (CBF) are the main adaptive responses observed in the brain during anemia. In healthy volunteers, acute isovolemic anemia with Hb concentrations around 5 g/dl was associated with cognitive deficits and impaired central processing, suggesting that maximal vasodilation had been achieved at these Hb concentrations and that no further increase in CBF could be obtained to compensate for reduced DO
2 [
8]. However, in acute brain injury, cerebral oxygen consumption may already be increased and cerebral vasodilation may be severely impaired, so that brain hypoxia could occur at higher Hb concentrations.
Several clinical studies have suggested that Hb concentrations less than 9 g/dl are associated with tissue hypoxia, metabolic crises, and poor outcomes among patients with traumatic brain injury (TBI), subarachnoid hemorrhage (SAH), or intracranial hemorrhage (ICH) [
9‐
11]. However, in the same setting, RBC transfusion has been associated with poor neurological recovery [
12,
13]. Thus, assessment of the risk/benefit ratio for transfusion is a key consideration in critically ill patients with acute brain injury [
14]. Transfusion practice has moved toward more restrictive strategies in critically ill patients (i.e., to keep Hb ≥7 g/dl) in recent years; however, more liberal targets (e.g., Hb ≥9 g/dl) are recommended in patients with underlying cardiovascular disease [
15]. Although this approach may also be logical in patients with acute brain injury, as was recommended in recent guidelines (i.e., to keep Hb between 8 and 10 g/dl) [
2,
16], patients with acute brain injury were poorly represented in large clinical trials of RBC transfusion strategies, and further clinical investigation is needed in these patients.
To better understand the feasibility of such studies and how to develop future therapeutic protocols, it is important to assess transfusion practice of ICU physicians who manage patients with acute brain injury. Two surveys of transfusion practice in patients with TBI and SAH have been published [
17,
18], but they were limited to cohorts of physicians in North America. Thus, the aim of this survey was to investigate, at a more global level, the management of anemia and RBC transfusion practice, including the Hb threshold for transfusion, in patients with acute brain injury.
Methods
This study was approved and endorsed by the European Society of Intensive Care Medicine (ESICM) Research Committee. No ethical approval was needed for the participating physicians.
Study design and administration
We conducted a web-based survey sent by email to all members of the ESICM, the Neurocritical Care Society, the Australian and New Zealand Intensive Care Society, the Società Italiana di Anestesia Analgesia Rianimazione e Terapia Intensiva, and the Brazilian Society of Critical Care (the questionnaire was not translated). The first email was sent on 21 May 2013, with two reminders sent 2 and 4 weeks thereafter, respectively. We did not specifically limit the survey to physicians working in academic institutions; our target sampling was intensivists who regularly care for patients with acute brain injury (i.e., neurointensivists, vascular neurosurgeons, neurologists) or with extensive interest and experience in the management of patients with acute brain injury. The survey was self-administered by the respondents, voluntary, and submitted online using a SurveyMonkey system (SurveyMonkey, San Mateo, CA, USA). No reimbursement was offered for questionnaire completion. All responses were anonymous.
Survey development
The survey was developed by two investigators (FST, MO) on the basis of a review of the relevant literature in this field. The questionnaire was presented at the Neuro-Intensive Care Section of the ESICM meeting in October 2012 and sent by email to all participants of this section to obtain further input. A consensus was then achieved, and the final version of the survey was sent to the ESICM Research Committee, which gave its final endorsement after a peer review process (i.e., two reviewers).
The survey was constructed using 24 multiple-choice questions (Additional file
1: Appendix 1) to evaluate physicians’ preferences for correcting anemia in patients with acute brain injury. We also recorded details of ICU and physician characteristics (i.e., country, specialty, position, years since completion of training, type of ICU, primary and secondary specialties, number of ICU beds, hospital size, and characteristics). We asked the respondents which Hb threshold they used to trigger RBC transfusion in patients with acute brain injury and whether the type of disease, patient characteristics, or specific conditions (e.g., increased intracranial pressure or vasospasm) would modify these thresholds. We also asked about the diagnosis and management of low iron levels and about the administration of erythropoietin. Finally, we asked clinicians about the potential risks and benefits of RBC transfusion in this setting and whether they believed a clinical trial should be conducted to specifically determine the optimal Hb threshold in this patient population. The survey was not specifically tested in a pilot cohort of potential respondents but underwent a peer review process within the ESICM Research Committee.
Definitions
We planned a priori to determine how transfusion practices might be influenced by geographical factors and categorized five world areas (Europe, North America, Central and South America, Asia/Africa, and Oceania). We also categorized Hb thresholds as a binary variable: less than 9 g/dl or greater than or equal to 9 g/dl. We defined the factors that could influence RBC transfusion policy as noncerebral (e.g., active bleeding, coronary artery disease[CAD], low mixed venous oxygen saturation [SvO2], age, lactate level greater than 2.5 mEq/L) or cerebral (e.g., low brain tissue pressure levels [PbtO2 less than 15 mmHg], delayed cerebral ischemia [DCI], data from other neuromonitoring tools [including electroencephalography, noninvasive cerebral oxygenation, transcranial Doppler], presence of intracranial hypertension, Glasgow Coma Scale score less than 9). Types of acute brain injury were separated as follows: TBI, SAH, ischemic stroke, ICH, postneurosurgery, seizure and status epilepticus (SE), hypoxic-ischemic encephalopathy (HIE), central nervous system (CNS) infections, and noninfectious and autoimmune encephalitis.
Statistical analysis
Discrete variables are expressed as count (percent) and continuous variables as median (IQR or range, as indicated). Differences between study groups were assessed using a chi-square test, Fisher’s exact test, Kruskal-Wallis test, or Mann-Whitney U test as appropriate. We performed multivariable logistic regression analyses to identify whether the following issues could be explained by the respondent’s primary specialty (internal medicine, intensive care, anesthesiology, neurology, surgery, or pediatrics), the period since completion of training (years of practice), the type of ICU (medical, neuro-ICU, surgical), or institution (university, university-affiliated, non-university-affiliated) rather than the geographical difference itself: (a) the Hb threshold used to initiate RBC transfusion after acute brain injury, (b) use of noncerebral vs. cerebral factors to adjust the Hb threshold for transfusion, and (c) the concentration of Hb used to initiate RBC transfusion in the presence of one or more influencing factors. The collinearity between variables was checked by inspection of the correlation between them, by looking at the correlation matrix of the estimated parameters, and by looking at the change of parametric estimates and at their estimated standard errors. No instability of parametric estimates or an excessive increase of standard errors was observed during the analyses. A Hosmer-Lemeshow goodness-of-fit test was considered to assess model calibration (agreement between observed outcomes and predictions). Q–Q plots were drawn to check for normality in the residuals. Results are given as ORs and their 95% CIs. Statistical analyses were performed using the IBM SPSS Statistics 24.0 for Windows NT software package (IBM, Armonk, NY, USA). All reported p values are two-sided. A p value less than 0⋅05 was considered to indicate statistical significance.
Discussion
This worldwide survey is the largest to provide data on the approach of physicians to anemia management and Hb thresholds for RBC transfusion in patients with acute brain injury. More than half of the respondents stated that they used an Hb threshold of 7–8 g/dl to initiate RBC transfusion after acute brain injury. A lower threshold for transfusion was observed in Africa/Asia and Oceania than in Europe. However, most respondents stated that they would increase the Hb trigger threshold in the presence of various factors, in particular CAD, active bleeding, and low SvO2. Most respondents said they felt an RCT was needed to compare different transfusion strategies in this patient population, in particular after SAH and TBI. Respondents’ characteristics that influenced transfusion policies in this setting included their origin, their background, and their experience, with the most liberal approaches used by respondents from Europe and North America, by anesthesiologists and neurologists/neurosurgeons, and by those with less than 5 years in practice.
Only two other surveys have evaluated transfusion practice in patients with acute brain injury. In the first study, Sena et al. [
17] evaluated responses of 312 physicians, including trauma surgeons, neurosurgeons, and intensivists from level I trauma centers in the United States. The authors used two clinical scenarios, focusing only on TBI. Neurosurgeons reported the highest Hb thresholds for initiating RBC transfusion compared with trauma surgeons and intensivists in patients with normal (8.3 vs. 7.5 vs. 7.5 g/dl, respectively) and elevated (8.9 vs. 8.0 vs. 8.4 g/dl, respectively) intracranial pressure. Interestingly, fewer neurosurgeons than other physicians reported that they used systemic and cerebral factors to influence the decision to transfuse. In a second study, Kramer et al. [
18] performed a survey of 282 academic neurointensivists, neurosurgeons, and multidisciplinary intensivists from the United States and Canada to evaluate transfusion policies for patients with SAH. The mean Hb concentration used to initiate RBC transfusion was higher in low-grade patients (8.2 g/dl) and in those with DCI (8.6 g/dl), with large variability in the proposed ranges. Again, neurosurgeons expressed the highest minimum Hb goals compared with other physicians. Respondents were more likely to transfuse patients with low PbtO
2 values and high brain lactate-to-pyruvate ratios. Similarly, we also found that most of the respondents used an Hb threshold of 7–8 g/dl to initiate RBC transfusion in patients with acute brain injury, although nearly 60% of these respondents used a higher threshold in patients with TBI, SAH, and ischemic stroke.
Nevertheless, in our study, we made some interesting observations that were not reported in the previous surveys. First, Hb thresholds were more liberal among anesthesiologists than among intensivists. This may appear to be in contrast to recent guidelines from the American Society of Anesthesiologists suggesting the need for restrictive transfusion criteria to minimize RBC use with no additional risks for poor outcome or cardiac, neurological, or pulmonary complications [
19]. Differences in practice between these two groups of physicians may explain these findings, with more than half of transfusions given by anesthesiologists partially initiated as a result of a physiological trigger (e.g., hypotension, tachycardia, preexisting anemia) rather than the absolute value of Hb [
20]. Moreover, one study also showed that, in 48,086 surgical patients at a tertiary U.S. academic medical center, the target Hb value after transfusion ordered by an anesthesiologist was 11.7 ± 1.3 g/dl [
21]. Second, previous surveys have been focused only on physicians in North America, whereas we report that a conservative (i.e., 7–8 g/dl) threshold of Hb was more often used in Asia/Africa and Oceania than in Europe to initiate transfusion. The lower threshold used in African and Asian countries may be due to limited resources and less blood availability, as well as to shortage of other supplies (e.g., bottles, bags) and greater risk of blood contamination [
22]. The differences between Europe and Oceania may be explained by national and statewide initiatives in Australia/New Zealand to encourage implementation of patient blood management programs as a cost-effective standard of care in their public health system, whereas this approach is not entirely integrated into routine management in Europe [
23,
24].
In this survey, we observed that many clinicians did not initiate RBC transfusion at a fixed Hb threshold, but adjusted their practice according to the presence of other factors. In particular, compared with intensivists, neurologists more frequently used noncerebral trigger factors, and neurosurgeons used more frequently used cerebral factors. The presence of systemic factors such as low SvO
2 values has been widely used to trigger RBC transfusion in critically ill patients with sepsis [
25]. RBC transfusion may also improve microvascular abnormalities associated with sepsis or may contribute to improved tissue oxygenation in patients with high lactate levels [
26,
27]. Similarly, Oddo et al. [
10] showed that only the combination of low Hb concentrations with reduced PbtO
2 values negatively affected the outcome of patients with TBI. Thus, decisions regarding when to transfuse patients with acute brain injury remain difficult and should be titrated on an individual patient basis. Biomarkers of impaired systemic or tissue DO
2 may help identify patients who are most likely to benefit from RBC transfusion in this setting. Heterogeneity in clinical practice among physicians with different backgrounds also suggests the need to better evaluate and identify which biomarkers could be used in this context.
More than 60% of respondents thought that an RCT comparing a restrictive and a liberal transfusion strategy was needed, in particular in patients with SAH and TBI. Interestingly, whereas 40% indicated that one group should be guided by neuromonitoring, only 13% of all respondents stated that neuromonitoring influenced their RBC transfusion policy and that transfusions were guided mainly by noncerebral events. This finding underlines that the availability of cerebral trigger factors as a tool to guide transfusion policy may be logistically difficult because these tools are not widely available. Also, the optimal threshold to initiate RBC transfusion in critically ill patients remains unclear, and many physicians still use predefined Hb values in their practice. Finally, it is also possible that the limited options provided in the survey format prevented respondents from providing other alternatives to an RCT comparing restrictive vs. liberal transfusion. Although the willingness to evaluate transfusion practices in the context of a randomized trial has also been suggested in previous surveys [
17,
18], it remains unclear which threshold could accurately define a restrictive or a liberal approach. Current guidelines for RBC transfusion in the critically ill suggest initiating transfusion when the Hb is less than 7 g/dl because higher thresholds do not provide additional benefit [
28]. Nevertheless, most practitioners would probably adopt a higher transfusion Hb threshold in patients with acute coronary syndrome, as well as in those at risk of secondary brain injury, such as patients with SAH with DCI or in patients with severe TBI [
11,
29,
30]. However, it remains unclear whether using RBC transfusion to increase Hb concentrations to 9–11 g/dl is a logical therapeutic decision to improve cerebral oxygenation and neurological recovery in these patients. Indeed, in patients with TBI, the increase in PbtO
2 after RBC transfusion was generally limited [
31,
32]. In patients with low-grade SAH, each 1.0 g/dl increase in Hb concentration after transfusion was associated with an increase of 1.4 mmHg in PbtO
2, without significant effects on cerebral metabolism [
33]. In another study, RBC transfusion resulted in a significant improvement in cerebral DO
2, in particular in those cerebral territories with the lowest baseline DO
2, and these effects were more significant than those produced by fluid expansion or vasopressors [
34,
35].
Although we observed some differences in Hb thresholds and RBC transfusion practices across regions and specialties, we could not determine from our survey whether this was due to discrepancies in local protocols, lack of compliance with general recommendations, or other factors that we did not investigate. For example, we showed that working in large hospitals (for the use of noncerebral triggers) and a greater degree of seniority (for the use of a conservative threshold) may influence the decision to transfuse. Whether these observations reflect a different familiarity with published literature on these issues or just personal practice is difficult to determine.
This survey has several limitations that need to be acknowledged. First, the validity of the survey depends on a high response rate among target physicians. Because the questionnaire was distributed using email addresses of members of different medical societies, we could not determine the potential denominator, which may have resulted in an underpowered analysis. The total number of potential respondents (i.e., considering the members of participating critical care societies; a list of respondents per countries is provided in Additional file
1: Appendix 2) probably exceeded 15,000 people, and this might significantly limit the external validity/generalizability of our findings. We did not specifically contact only physicians who were experts in neurocritical care, as was the case in previous surveys [
17,
18]; we also allowed other physicians working in general or mixed ICUs treating these patients to respond, and this may provide a broader picture of the general management of these patients who are admitted to nonspecialized ICUs in many countries. Second, survey development and testing should follow a process of item generation through a process of a systematic review; item reduction; pretesting; and pilot testing for clinical sensibility, reliability, and validity [
36]. Our survey was based on a consensus within experts in the Neuro-Intensive Care Section of the ESICM, and the preparation process might have affected comprehensiveness, clarity, and face validity as well as respondents’ interpretation of questions. Third, as is true for all surveys, the responses do not necessarily reflect local practice but personal opinions of the strategy used to manage RBC transfusion in this setting. A prospective audit collecting data on this topic would be welcome to provide more real-world data for this therapeutic intervention. Fourth, we had few respondents from Africa and needed to consider Asia and Africa as a single group, although practices may be different in these two areas. Fifth, our survey was conducted before the publication of a recent U.S. study evaluating RBC transfusion management in patients with TBI [
37], which showed no differences between a liberal (Hb less than 10 g/dl) and a restrictive (Hb less than 7 g/dl) transfusion strategy. It is impossible to know whether these results would have influenced the responses in our survey. Finally, because the survey was conducted anonymously, we could not control whether several respondents worked at the same center, which may have influenced the variability and generalizability of the results.