HEMOTION – transfusion thresholds in TBI

Liberal or Restrictive Transfusion Strategy in Patients with Traumatic Brain Injury

Turgeon AF. NEJM June 13 2024; doi:10.1056/NEJMoa240360

Clinical Question

  • In adult patients with moderate or severe traumatic brain injury, does a liberal transfusion strategy (transfusion if haemoglobin <10g/dl) compared to a restrictive strategy (transfusion if haemoglobin <7g/dl) reduce the rate of an unfavourable Glasgow Outcome Scale – Extended, at 6 months.


  • Traumatic Brain Injury accounts for the most number of trauma related deaths and has a high rate of morbidity in survivors
  • The optimal haemoglobin target in patients following traumatic brain injury is unknown. A Cochrane Review of the available evidence states that the optimal target is unclear
  • A trial in all ICU patients, the TRICC trial, suggested a restrictive transfusion practice was superior to a liberal strategy, however, this trial was published 15 years ago and was in the era where transfused blood was not leukodepleted. There were only small numbers of TBI patients included in the cohort
  • The injured brain is vulnerable to hypoxia and anaemia may exacerbate this state worsening the secondary injury


  • Patients were randomised 1:1 to a liberal or restrictive strategy
  • Allocation concealment and variable permuted blocks, stratified to centre
  • The treating clinicians were unblinded, but the follow-up assessors were blinded
  • The transfusion target was maintained until discharge from ICU
  • An independent data and safety monitoring committee reviewed data after each 25% increment of the target enrollment was reached. It also evaluated the results of a protocol-specified formal interim analysis that was performed when enrollment was at 50% of the target
  • Consent was obtained from next of kin or deferred when appropriate
  • Intention to treat analysis was performed


  • 34 centres in Canada, the UK, France, and Brazil were included
  • Patients were randomised between September 2017 and April 2023
  • A sample size of 712 patients provided 80% power to detect an absolute difference between groups of 10% for the primary outcome


  • Inclusion: adult patients with moderate or severe TBI (GCS 3-12) with anaemia (Hb<10g/dL)
  • Exclusion:
    • Patients who had received transfusion before randomisation.
    • Contraindication to transfusion or active bleeding
    • Patients who were brain dead or GCS 3 with fixed pupils
  • 6188 patients were assessed for eligibility with 742 randomised. 2 of these were randomised in error (had SAH) and 4 withdrew consent. This left 736 patients in the intention to treat analysis
  • Baseline characteristics
    • 72.7% of patients were men, the mean age was 48.7±18.9 years (+/-SD), and 73.2% had a severe traumatic brain injury (GCS <8)
    • The mean injury severity score was 30
    • Patients were assessed for chance of poor outcome using prognostic variables of the TBI-IMPACT calculator
    • There were some imbalances between groups
      • In the restrictive-strategy group there were more:
        • women (30% vs 24%)
        • patients with a GCS motor score of 1 (32.7% vs 26%)
        • patients with no pupillary reactivity (14.1% vs 12.4%)
        • patients who had an episode of hypotension (28.8% vs 22.7%)
      • In the liberal-strategy group there were more
        • episodes of low cerebral perfusion pressure (22.1% vs 17.3%)
        • patients with a Marshall score of V or VI (indicating any lesion that was surgically evacuated or a mass lesion of more than 25 ml, respectively) (37.1% vs 33.2%)
        • patients with a GCS motor score of 6 (7.9% vs 5.4%)


  • Liberal transfusion group
    • 369 patients were analysed in this group. Transfusion was triggered by a haemoglobin level of ≤10 g/dL
    • 98.9% of patients received at least 1 unit of packed cells with the median haemoglobin being 10.8g/dL


  • Restrictive transfusion group
    • 367 patients were analysed in this group. Transfusion was triggered by a haemoglobin level of ≤7 g/dL
    • 38.4% of patients received at least 1 unit of packed cells with the median haemoglobin being 8.8g/L

Management common to both groups

  • Transfusion was initiated within 3 hours of the trigger haemoglobin level
  • 1 unit of leukodepleted packed cells were administered at a time
  • Most patients had ICP monitoring
  • Few patients had CBO2 monitoring and decompressive craniectomy


  • Primary outcome:
    • GOS-E at 6 months was <4 (unfavourable) in 68.4% in the liberal-strategy group and 73.5% in the restrictive-strategy group
    • Adjusted absolute difference, restrictive strategy vs. liberal strategy, 5.4 percentage points; (95% CI, −2.9 to 13.7)
    • Findings were consistent across subgroups and also with the worst, intermediate, and best predicted prognoses
  • Secondary outcomes:
  • Comparing liberal vs. restrictive group
    • No significant difference in
      • Death in ICU (17.1% vs 15.3%), Death in Hospital (23% vs 21.5%), Death at 6 months (26.8% vs 26.3%)
    • No significant difference in functional independence measure or other measures of functional outcome
  • Tertiary Outcomes
    • Median (IQR) number of red-cell units transfused was 3 (2-5) in the liberal group and 0 (0-1) in the restrictive group
  • Safety
    • 6 patients in the liberal group had a mild reaction to transfusion, 1 in the restrictive group
    • VTE was the same in each group

Authors’ Conclusions

  • There was no significant difference in GOSE at 6 months between those with liberal or restrictive transfusion following moderate or severe traumatic brain injury


  • Allocation concealment, intention to treat analysis, minimal loss to follow-up
  • Multi-centre study in various countries gives the study good external validity
  • There was a clear separation between the groups with median daily haemoglobin levels
  • Although the bedside clinicians were unblinded, the outcome assessors of functional status were unaware of the allocation


  • The trial was powered to detect a 10% difference in unfavourable outcome
  • The result showed a 5.4% difference which may be clinically significant
  • There were baseline differences between the two groups
  • It is possible there may be up to a 13.7% absolute reduction in risk of an unfavourable outcome with a liberal transfusion strategy
  • The statistical analysis has taken a frequentist approach. The authors are also planning a Bayesian Analysis

The Bottom Line

  • Modern ICU practice has erred on the side of restrictive transfusion strategies based on mainly on the TRICC study which was published in the age of non-leukodepleted blood. In addition there were few patients with TBI in that landmark study
  • This study challenges the paradigm of following a restrictive transfusion regime in moderate-severe TBI patients.
  • I keenly await the Bayesian Analysis
  • I will aim closer to a haemoglobin of 10g/dL in this patient group

External Links


Summary author: Celia Bradford – @celiabradford
Summary date: June 21 2024
Peer-review editor: @davidslessor



One comment

  • David Slessor

    My take on this trial is that it it showed no difference in outcomes, and that the differences in baseline characteristics introduce bias. This trial doesn’t convince me that I should be using a higher transfusion threshold in this population. I keenly await the transfusion threshold studies that are currently ongoing in other neuro-critical care patients

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.