Skip to main content

Foundational Trials

HEMOTION

Image: Shutterstock / ChaNaWiT

Publication

  • Title: Transfusion Strategies in Traumatic Brain Injury
  • Acronym: HEMOTION (Hemoglobin Transfusion Threshold in Traumatic Brain Injury Optimisation)
  • Year: 2024
  • Journal published in: New England Journal of Medicine
  • Citation: Turgeon AF, Lauzier F, Simard JF, et al. Transfusion Strategies in Traumatic Brain Injury. N Engl J Med. 2024;391:723-735.

Context & Rationale

  • Background
    Anaemia is common after moderate–severe traumatic brain injury (TBI) and may aggravate secondary brain injury by reducing oxygen delivery.
    Red-cell transfusion can increase oxygen-carrying capacity but is associated with immunomodulation and transfusion-related complications; neurocritical care clinicians have therefore varied widely in transfusion thresholds.
    Prior evidence in TBI was dominated by observational associations (anaemia/transfusion and outcome) and small/heterogeneous trials, leaving major uncertainty about an optimal haemoglobin trigger.
  • Research Question/Hypothesis
    In ICU patients with moderate–severe blunt TBI and anaemia (haemoglobin ≤10 g/dL; 100 g/L), does a restrictive transfusion strategy (trigger 7 g/dL) improve 6‑month neurological outcome compared with a liberal strategy (trigger 10 g/dL)?
  • Why This Matters
    TBI is a high-burden condition where small shifts in long-term functional outcomes are clinically meaningful.
    Transfusion thresholds are an everyday, modifiable ICU decision, but “one-size-fits-all” restrictive strategies derived from general critical care may not apply to injured brain tissue.

Design & Methods

  • Research Question: Whether a restrictive (7 g/dL) versus liberal (10 g/dL) red-cell transfusion trigger changes 6‑month neurological outcome after moderate–severe TBI with anaemia.
  • Study Type: Pragmatic, multicentre, randomised, controlled, parallel-group trial in 34 centres (Canada, United Kingdom, France, Brazil); ICU setting; open-label transfusion strategy with blinded outcome assessment.
  • Population:
    • Adults with moderate or severe blunt TBI admitted to ICU.
    • Anaemia with haemoglobin ≤10 g/dL (≤100 g/L) during the early post-injury ICU course (timing window defined in the published protocol).1
    • Key exclusions included conditions where protocolised thresholds were inappropriate or infeasible (e.g., ongoing life-threatening bleeding, transfusion refusal), per protocol.1
  • Intervention:
    • Restrictive strategy: transfuse red cells if haemoglobin ≤7 g/dL (≤70 g/L); reassess haemoglobin before additional units.
    • Applied while in ICU (or until protocol end-point per trial procedures).
  • Comparison:
    • Liberal strategy: transfuse red cells if haemoglobin ≤10 g/dL (≤100 g/L); reassess haemoglobin before additional units.
    • Co-interventions otherwise per usual care; rescue transfusion for active bleeding/clinical necessity permitted.
  • Blinding: Treating teams were unblinded (transfusion triggers), but 6‑month outcome assessment was intended to be blinded; objective mortality outcomes reduce detection bias risk.
  • Statistics: Sample size planning (protocol) assumed ~40% unfavourable outcome and sought a 10% absolute reduction with 80% power at two-sided α=0.05; an ordinal “sliding dichotomy” approach for GOSE was prespecified to increase efficiency; primary analysis was intention-to-treat.1
  • Follow-Up Period: 6 months for the primary neurological outcome and key patient-centred secondary outcomes; in-hospital and ICU outcomes captured during index admission.

Key Results

This trial was not stopped early. A prespecified interim analysis was performed; stopping criteria were not met.

Outcome Restrictive (Hb ≤7 g/dL) Liberal (Hb ≤10 g/dL) Effect p value / 95% CI Notes
Primary: Unfavourable neurological outcome at 6 months (GOSE “sliding dichotomy”) 263/358 (73.5%) 249/364 (68.4%) Adjusted absolute difference (restrictive − liberal): +5.4 percentage points 95% CI −2.9 to 13.7 Denominators reflect available 6‑month GOSE (small missingness).
Median haemoglobin during ICU stay 8.8 g/dL 10.8 g/dL Not reported Not reported Demonstrates achieved biological separation (≈2.0 g/dL).
Any red-cell transfusion 141/367 (38.4%) 365/369 (98.9%) Not reported Not reported Protocolised triggers drove large between-group exposure separation.
Red-cell units transfused (total; median per patient) 307; 0 (0–1) 1516; 3 (2–5) Not reported Not reported Median (Q1–Q3) per patient; totals across group.
Mortality in ICU 56/367 (15.3%) 63/369 (17.1%) Hazard ratio: 1.13 95% CI 0.82 to 1.68 Time-to-event model (as reported).
Mortality at 6 months 96/367 (26.3%) 99/369 (26.8%) Hazard ratio: 1.01 95% CI 0.76 to 1.35 No signal of mortality difference.
Functional Independence Measure (total) at 6 months 115 (76–124) 119 (95–125) Adjusted difference in medians (liberal − restrictive): 4.34 95% CI −8.45 to 22.80 Median (Q1–Q3).
EQ‑5D‑5L utility index at 6 months 0.64 (0.16–0.81) 0.74 (0.43–0.85) Adjusted difference in medians (liberal − restrictive): 0.06 95% CI 0.01 to 0.10 Patient-centred secondary; multiplicity not corrected.
Moderate-to-severe depressive symptoms at 6 months (PHQ‑9 ≥10) 95/222 (42.8%) 82/227 (36.1%) Risk ratio: 0.85 95% CI 0.63 to 1.17 Subset with available PHQ‑9.
Acute respiratory distress syndrome 3/367 (0.8%) 12/369 (3.3%) Not reported Not reported Numerically higher with liberal strategy.
Transfusion reactions (among transfused patients) 1/141 (0.7%) 6/365 (1.6%) Not reported Not reported Reported among those receiving at least one transfusion.
  • The restrictive strategy did not improve 6‑month neurological outcome; the point estimate favoured the liberal strategy (restrictive associated with +5.4 percentage points higher probability of unfavourable outcome), but the CI included no difference.
  • Biological and exposure separation was large (median ICU haemoglobin 8.8 vs 10.8 g/dL; transfusion in 38.4% vs 98.9%; median units per patient 0 vs 3).
  • Mortality was similar; some patient-centred secondary outcomes (e.g., EQ‑5D‑5L utility) numerically favoured liberal transfusion and require cautious interpretation given multiplicity.

Internal Validity

  • Randomisation and allocation: Central randomisation with allocation concealment; large pragmatic multicentre design reduces centre-level idiosyncrasy.
  • Dropout / missingness: Primary outcome availability was high (GOSE obtained in 364/369 liberal vs 358/367 restrictive; missing 5 vs 9).
  • Performance/detection bias: Treating clinicians were unblinded (inevitable for transfusion triggers); outcome assessment at 6 months was intended to be blinded, and mortality outcomes are objective.
  • Protocol adherence: Trial-level protocol violation (patient-level) was low (14/369 [3.8%] liberal vs 7/367 [1.9%] restrictive).
    Event-level deviations occurred (e.g., transfusion when haemoglobin was above the protocol threshold: 42/369 [11.4%] liberal vs 10/367 [2.7%] restrictive; and withholding transfusion when haemoglobin was ≤7 g/dL: 11/369 [3.0%] liberal vs 21/367 [5.7%] restrictive).
  • Baseline characteristics: Groups were broadly comparable; randomisation should balance measured/unmeasured confounders, and the large sample supports this.
  • Heterogeneity: International multicentre conduct increases practice heterogeneity, but the intervention is a simple trigger-based protocol; centre effects are less likely to dominate the primary endpoint.
  • Timing: Randomisation occurred after ICU admission during early post-injury anaemia; timing varied by severity (median time from injury to randomisation ranged ~47–62 hours across severity strata in supplementary summaries).
  • Dose/target: The “dose” was an exposure strategy (trigger) rather than a fixed haemoglobin target; achieved separation (8.8 vs 10.8 g/dL) indicates adequate contrast.
  • Separation of the variable of interest: Any transfusion 141/367 (38.4%) restrictive vs 365/369 (98.9%) liberal; total RBC units 307 vs 1516; median units per patient 0 (0–1) vs 3 (2–5); median ICU haemoglobin 8.8 vs 10.8 g/dL.
  • Crossover/adjunctive therapies: There was limited evidence of clinically important crossover at the patient level; co-interventions were otherwise standardised only by usual care, consistent with pragmatic design.
  • Outcome assessment: Primary endpoint (6‑month GOSE) is clinically meaningful; “sliding dichotomy” increases power but is less intuitive than a single fixed cut-point.
  • Statistical rigour: Prespecified primary analysis (intention-to-treat) with adjusted effect reporting; interim analysis prespecified; missing data modest.

Conclusion on Internal Validity: Overall, internal validity appears strong: concealed randomisation, high follow-up completeness, and large achieved exposure separation support a credible estimate; the main limitations are the unavoidable open-label delivery and some protocol deviations.

External Validity

  • Population representativeness: 6184 ICU TBI patients were screened, with 742 randomised; enrolled patients reflect a subset with moderate–severe blunt TBI complicated by anaemia and sufficient expected survival to merit transfusion strategy decisions.
  • Important exclusions: Exclusions (e.g., transfusion after ICU admission before screening, inability to adhere to protocolised thresholds, and other protocol-specified factors) mean results apply best to stable ICU TBI patients rather than those with uncontrolled bleeding or highly individualised transfusion constraints.
  • Applicability across systems: Multi-country enrolment supports broad applicability in high-resource ICUs; the intervention requires only routine haemoglobin monitoring and blood availability, but liberal thresholds may have greater resource implications in blood-limited settings.
  • Clinical translation: Results most directly inform transfusion triggers in similar ICU TBI populations with anaemia; they do not directly address goal-directed transfusion guided by brain tissue oxygenation monitoring.

Conclusion on External Validity: Generalisability is moderate-to-high for adult ICU patients with moderate–severe blunt TBI and anaemia in comparable settings, but applicability is limited for patients with active haemorrhage, different transfusion constraints, or where blood supply is restricted.

Strengths & Limitations

  • Strengths: Large pragmatic multicentre RCT; clinically meaningful long-term neurological endpoint; major achieved biological/exposure separation; high completeness of primary outcome; prespecified efficient ordinal (“sliding dichotomy”) analysis approach.
  • Limitations: Open-label delivery; protocol deviations occurred (though modest at patient level); secondary patient-centred outcomes raise multiplicity concerns; results apply to blunt moderate–severe TBI with anaemia and may not extend to other neurocritical populations or bleeding trauma.

Interpretation & Why It Matters

  • Clinical implication
    A restrictive trigger of 7 g/dL (70 g/L) should not be assumed to be beneficial in ICU patients with moderate–severe TBI and anaemia; in this trial it did not improve 6‑month neurological outcome and the point estimate is compatible with harm.
    Liberal transfusion increased blood exposure substantially (and may increase some complications such as ARDS), so the decision remains a balance of uncertain neurological benefit versus resource use and transfusion-related harms.
  • Mechanistic framing
    The achieved 2 g/dL haemoglobin separation demonstrates the biological lever was meaningfully moved; absence of clear neurological benefit suggests either (i) anaemia at these thresholds is not a dominant driver of secondary injury, (ii) benefit is confined to narrower phenotypes/timing, or (iii) harms of transfusion offset any oxygen-delivery advantage.

Controversies & Subsequent Evidence

  • Interpretation of the primary effect estimate: The CI (−2.9 to +13.7 percentage points for unfavourable outcome, restrictive − liberal) is compatible with both no difference and clinically important harm from restrictive transfusion; “absence of evidence of benefit” should not be over-interpreted as “evidence of equivalence”.
  • Multiplicity and secondary outcomes: Some patient-centred secondary outcomes favoured the liberal strategy (e.g., EQ‑5D‑5L utility), but these were not powered for definitive conclusions and are vulnerable to multiplicity; they are hypothesis-generating rather than practice-changing on their own.
  • Protocol and analytic choices: The prespecified sliding-dichotomy approach improves efficiency but can feel less transparent to clinicians than a fixed GOSE cut-point; the published protocol clarifies assumptions and the analytic plan.1
  • Position within the broader acute brain injury evidence base: Earlier neurocritical transfusion trials (e.g., combined erythropoietin/transfusion-threshold strategies) did not establish a clear optimal trigger for TBI, and recent acute brain injury RCTs and meta-analyses have continued to show heterogeneity and imprecision when focusing on neurological outcomes rather than mortality alone.234
  • Post-HEMOTION synthesis: Updated systematic reviews incorporating modern neurocritical care trials (including HEMOTION) have not identified a consistent neurological advantage for restrictive transfusion across acute brain injury syndromes, but TBI-specific certainty remains limited and effect estimates remain compatible with clinically meaningful subgroup effects.5

Summary

  • Pragmatic international ICU RCT in 742 adults with moderate–severe blunt TBI and anaemia (Hb ≤10 g/dL).
  • Restrictive trigger (7 g/dL) did not improve 6‑month neurological outcome versus liberal trigger (10 g/dL); point estimate favoured liberal (restrictive − liberal +5.4 percentage points; 95% CI −2.9 to 13.7).
  • Exposure separation was large (transfusion in 38.4% restrictive vs 98.9% liberal; median haemoglobin 8.8 vs 10.8 g/dL; median RBC units per patient 0 vs 3).
  • Mortality was similar through 6 months (HR 1.01; 95% CI 0.76 to 1.35).
  • Some secondary quality-of-life metrics favoured liberal transfusion, while ARDS was numerically higher with liberal transfusion; neither is definitive in isolation.

Further Reading

Other Trials

Systematic Review & Meta Analysis

Observational Studies

Guidelines

Notes

  • Brain-injury–specific transfusion guidance explicitly incorporating HEMOTION has not been clearly identified in major guideline updates at the time of writing; most recommendations remain syndrome-agnostic and prioritise restrictive transfusion in general critical care, with caveats for acute neurological injury.
  • Some references above may require verification of final pagination (“xxxx–xxxx”) depending on journal format; DOI links provide definitive retrieval.

Overall Takeaway

HEMOTION is a landmark neurocritical care transfusion trial because it randomised a large, international cohort of anaemic moderate–severe TBI patients and achieved substantial haemoglobin and transfusion separation while measuring clinically meaningful 6‑month neurological outcomes. It provides strong evidence that a universal restrictive trigger of 7 g/dL is not clearly beneficial in this population, and it reframes transfusion decisions in TBI as a balance of uncertain neurological effect, transfusion harms, and resource use rather than a simple extrapolation from general ICU transfusion doctrine.

Overall Summary

  • Restrictive (7 g/dL) did not improve 6‑month neurological outcome and may be harmful (CI includes clinically important harm).
  • Large exposure separation was achieved (38% vs 99% transfused; median Hb 8.8 vs 10.8 g/dL).
  • Mortality was unchanged; liberal transfusion increased blood exposure and may increase some complications (e.g., ARDS).

Bibliography