Skip to main content

Foundational Trials

Eurotherm3235

Image: Shutterstock / ChaNaWiT

Publication

  • Title: Hypothermia for intracranial hypertension after traumatic brain injury
  • Acronym: Eurotherm3235 Trial
  • Year: 2015
  • Journal published in: New England Journal of Medicine
  • Citation: Andrews PJD, Sinclair HL, Rodriguez A, Harris BA, Battison CG, Rhodes JKJ, Murray GD, for the Eurotherm3235 Trial Collaborators. Hypothermia for intracranial hypertension after traumatic brain injury. N Engl J Med. 2015;373:2403-2412.

Context & Rationale

  • Background
    Raised intracranial pressure (ICP) after severe traumatic brain injury (TBI) is associated with secondary brain injury, disability, and death, and is typically managed via a staged escalation pathway (supportive measures, osmotherapy/physiology optimisation, then barbiturates and/or decompressive craniectomy).
  • Research Question/Hypothesis
    Does therapeutic hypothermia targeted to 32–35°C, applied as a preferred “stage 2” ICP-lowering strategy in adults with intracranial hypertension after TBI, improve 6‑month neurological outcome compared with standard care without hypothermia?
  • Why This Matters
    Hypothermia is physiologically plausible (reduces cerebral metabolic demand; can lower ICP), widely available in modern ICUs, and had been supported by heterogeneous pre‑Eurotherm meta-analytic signals; a definitive, pragmatic strategy trial for established intracranial hypertension was needed to inform practice and guideline-level recommendations.

Design & Methods

  • Research Question: In adults with severe TBI and sustained intracranial hypertension despite stage 1 measures, does therapeutic hypothermia (32–35°C) as an early stage 2 treatment improve 6‑month functional outcome compared with standard care without hypothermia?
  • Study Type: Multicentre, pragmatic, parallel-group, randomised trial (1:1 allocation); international (47 centres across 18 countries); ICU/neurocritical care setting; open-label clinical management with blinded primary outcome assessment.
  • Population:
    • Adults with severe TBI managed in ICU with invasive ICP monitoring and an abnormal cranial CT.
    • Required sustained ICP >20 mm Hg for ≥5 minutes despite stage 1 therapy (e.g., sedation/ventilation/physiological optimisation).
    • Randomisation allowed up to 10 days after injury (protocol expanded during the trial; only a minority were enrolled within 12 hours of injury).
    • Key exclusions included: barbiturate infusion before randomisation; already receiving therapeutic hypothermia; normal CT; temperature <34°C on hospital admission; pregnancy; >10 days from injury; or judged unlikely to survive 24 hours.
  • Intervention:
    • Therapeutic hypothermia strategy targeting 32–35°C, titrated to the lowest temperature required (within range) to keep ICP ≤20 mm Hg.
    • Induction: refrigerated 0.9% saline bolus 20–30 mL/kg; maintenance by the site’s usual cooling technique (surface and/or endovascular), with explicit shivering management guidance.
    • Duration: minimum 48 hours; rewarming at 0.25°C/hour once ICP control was maintained (and delayed/paused if ICP rose during rewarming).
    • Escalation: additional (non-temperature) stage 2 measures were introduced if hypothermia alone did not control ICP; stage 3 therapies (barbiturates or decompressive craniectomy) were available if stage 2 measures failed.
  • Comparison:
    • Standard care strategy without therapeutic hypothermia; stage 2 therapies introduced as required to control ICP; stage 3 therapies used if stage 2 failed.
    • Temperature management in control aimed to avoid fever (specific temperature targets and cooling methods were not protocolised as a uniform intervention).
  • Blinding: Clinicians and sites were not blinded (temperature management is intrinsically difficult to mask); 6‑month Extended Glasgow Outcome Scale (GOS‑E) was assessed by an investigator unaware of treatment allocation.
  • Statistics: Target sample size 600 to provide 80% power at a two-sided 5% significance level to detect a 9% absolute reduction in unfavourable outcome (from 60% to 51%); primary analysis used an intention-to-treat ordinal (shift) approach with an adjusted common odds ratio for the GOS‑E.
  • Follow-Up Period: Primary endpoint at 6 months; key in-hospital/early outcomes included ICU/hospital endpoints and complications (e.g., pneumonia within the first 7 days); 28-day (or discharge) Modified Oxford Handicap Scale was also captured.

Key Results

This trial was stopped early. Recruitment was suspended and the trial stopped after 387 participants (vs a planned 600) on the recommendation of the independent monitoring committee because outcomes (including mortality) were worse in the hypothermia group.

Outcome Hypothermia (32–35°C) Standard care Effect p value / 95% CI Notes
Primary: GOS‑E at 6 months (ordinal; collapsed to 6 categories) n=191 (valid)
Dead 67 (35.1%)
Vegetative/lower severe disability 4 (2.1%)
Upper severe disability 55 (28.8%)
Lower moderate disability 16 (8.4%)
Upper moderate disability 22 (11.5%)
Good recovery 27 (14.1%) 		n=189 (valid)
Dead 49 (25.9%)
Vegetative/lower severe disability 7 (3.7%)
Upper severe disability 43 (22.8%)
Lower moderate disability 19 (10.1%)
Upper moderate disability 29 (15.3%)
Good recovery 42 (22.2%) 		Adj common OR 1.53 95% CI 1.02 to 2.30; P=0.04 OR <1 would favour hypothermia; observed shift was towards worse outcomes with hypothermia.
Unfavourable outcome at 6 months (GOS‑E 1–4) 142/191 (74.3%) 120/189 (63.5%) Adj OR 1.69 95% CI 1.06 to 2.70; P=0.03 Dichotomised sensitivity analysis; unfavourable outcome was more common with hypothermia.
Death by 6 months 68/195 (34.9%) 51/192 (26.6%) HR 1.45 95% CI 1.01 to 2.10; P=0.047 Time-to-event analysis; mortality numerically and statistically higher with hypothermia.
Stage 3 ICP therapies required (days 1–7) 84/192 (43.8%) 102/189 (54.0%) Not reported Not reported Despite fewer escalations to stage 3 therapies in the hypothermia group, overall outcomes were worse.
Barbiturate-infusion therapy (days 1–4) 20 patients 41 patients Not reported Not reported Control group used more barbiturates (a stage 3 therapy); decompressive craniectomy use was similar.
Decompressive craniectomy (days 1–4) 27 patients 27 patients Not reported Not reported Numerically identical early decompression use.
Mean core temperature (days 1–7) Not reported Not reported Adj mean difference −2.14°C 95% CI −2.34 to −1.94; P<0.001 Confirms robust temperature separation between groups.
Mean ICP (days 1–7) Not reported Not reported Adj mean difference −0.48 mm Hg 95% CI −2.04 to 1.08; P=0.55 Despite lower temperature, average ICP over 7 days was not meaningfully reduced.
Serious adverse events (SAEs) 33 events 10 events Not reported Not reported Total 43 SAEs in 30 patients; events included haemorrhage, cardiovascular instability, and skin/thermal injury.
  • Despite achieving substantial cooling (adjusted mean core temperature difference −2.14°C), hypothermia did not lower mean ICP over days 1–7 (adjusted mean difference −0.48 mm Hg) and was associated with worse 6‑month functional outcome (adj common OR 1.53) and higher mortality (HR 1.45).
  • Hypothermia reduced escalation to stage 3 therapies (43.8% vs 54.0%), suggesting some strategy-level ICP management impact, but this did not translate into improved patient-centred outcomes.
  • Prespecified subgroup analyses did not identify a consistent population in whom hypothermia improved outcomes (no credible qualitative interaction signal reported).

Internal Validity

  • Randomisation and Allocation: Central web-based allocation with minimisation (including a random element) to balance key prognostic factors (centre, age, Glasgow Coma Scale motor score, time from injury, and pupil responsiveness), supporting allocation concealment at enrolment.
  • Drop out / exclusions: 387 randomised; primary outcome (6‑month GOS‑E) was available for 380 (7 missing: 4 hypothermia; 3 control), limiting but not eliminating attrition bias.
  • Performance/Detection Bias: Unblinded temperature management and co-interventions create risk of performance bias; primary outcome assessment was performed by an assessor unaware of allocation, reducing detection bias for the main endpoint.
  • Protocol Adherence: Temperature targets were not perfectly maintained; adherence (days 1–4) was 64.8% in the hypothermia group vs 68.8% in the control group (as reported), indicating meaningful but imperfect protocol delivery.
  • Baseline Characteristics: Groups were broadly comparable at baseline (e.g., mean age 36.7 vs 37.4 years; male 79.5% vs 82.8%; mean ICP at randomisation 24.1 vs 24.8 mm Hg), supporting internal comparability.
  • Heterogeneity: Pragmatic delivery across 47 centres (18 countries) increases care heterogeneity; minimisation by centre reduces imbalance but cannot eliminate treatment-effect dilution or centre-level effect modification.
  • Timing: Mean time from injury to randomisation was ~48 hours in both groups; only ~16% were randomised within 12 hours, meaning this strategy largely tested hypothermia for established intracranial hypertension rather than ultra-early prophylaxis.
  • Dose: Cooling was titrated within 32–35°C to achieve ICP ≤20 mm Hg; whether this range is optimal (or whether deeper, shorter, longer, or different rewarming strategies are safer) remains uncertain and is a core interpretive constraint.
  • Separation of the Variable of Interest: Temperature separation was large (adjusted mean difference −2.14°C; 95% CI −2.34 to −1.94), but physiological separation on ICP was minimal (adjusted mean difference −0.48 mm Hg; 95% CI −2.04 to 1.08), raising mechanistic questions about why outcomes worsened despite cooling.
  • Key Delivery Aspects: Hypothermia was implemented as a strategy choice at “stage 2”; in practice, the control group received more barbiturate infusions (41 vs 20), reflecting different escalation patterns that may confound simplistic “hypothermia vs no hypothermia” interpretations.
  • Crossover: A control patient was explicitly documented as crossing over to hypothermia in the SAE records; overall crossover frequency beyond this was not reported, but crossover appears limited.
  • Outcome Assessment: The GOS‑E is a validated, patient-centred ordinal functional outcome; blinded assessment supports robust outcome ascertainment even in an unblinded ICU intervention trial.
  • Statistical Rigor: The prespecified ordinal primary analysis (adjusted common OR) is efficient for this outcome scale; however, early stopping and failure to reach the planned sample size (387/600) increases risk of effect-size exaggeration and imprecision, especially for secondary endpoints.

Conclusion on Internal Validity: Overall, internal validity is moderate: allocation and blinded endpoint assessment were strong, but unblinded co-intervention patterns, pragmatic heterogeneity, and early stopping constrain causal certainty about the mechanism of harm (even though the harm signal on patient-centred outcomes is clinically compelling).

External Validity

  • Population Representativeness: Participants reflect severe TBI patients requiring ICP monitoring and escalation beyond stage 1 measures; this is a common but resource-intensive neurocritical care population.
  • Applicability: Generalisability is strongest to high-resource ICUs capable of invasive monitoring and protocolised temperature management; applicability is limited in settings without reliable cooling systems, close haemodynamic monitoring, or experience in hypothermia complication prevention.
  • Clinical Strategy Scope: The trial tested hypothermia as a preferred “stage 2” strategy (rather than late rescue, or prophylactic ultra-early cooling), so extrapolation to different indications (e.g., late refractory ICP after multiple failed measures, or prehospital/prophylactic hypothermia) should be made cautiously.

Conclusion on External Validity: External validity is moderate: findings are highly relevant to modern neuro-ICUs managing monitored intracranial hypertension, but less applicable to centres without ICP monitoring infrastructure and to clinical scenarios where hypothermia is used only as late salvage therapy.

Strengths & Limitations

  • Strengths: Pragmatic international trial across many centres; clinically relevant population (established intracranial hypertension); robust patient-centred primary endpoint with blinded assessment; meaningful temperature separation achieved; prespecified subgroup analyses; monitoring committee oversight with predefined safety review.
  • Limitations: Open-label intervention with potential co-intervention differences; hypothermia delivery technique varied by site; incomplete granular reporting of non-temperature stage 2 therapies; modest absolute sample size with early termination; inclusion window allowed enrolment well beyond ultra-early injury phase; mechanistic separation on mean ICP over 7 days was small despite cooling.

Interpretation & Why It Matters

  • Clinical practice
    As a strategy for early stage 2 management of intracranial hypertension after severe TBI, therapeutic hypothermia to 32–35°C was associated with worse long-term neurological outcome and higher mortality; routine use for this indication is difficult to justify outside research settings.
  • Mechanistic insight
    Physiological plausibility (cooling reduces metabolism and can lower ICP) did not translate into benefit; the trial underscores that ICP-lowering surrogates may be insufficient (and potentially misleading) when an intervention has systemic harms (e.g., haemodynamic instability, infection/bleeding risk, electrolyte disturbance, and rewarming-related physiology).
  • Methodology
    Eurotherm3235 is a clear example where a pragmatic “strategy” trial (hypothermia as preferred stage 2 therapy) can yield different conclusions from physiological studies and earlier heterogeneous prophylaxis trials, emphasising the importance of patient-centred outcomes and robust safety monitoring.

Controversies & Subsequent Evidence

  • Strategy vs “add-on” interpretation: The protocol framed hypothermia as a preferred stage 2 ICP-lowering strategy (with other stage 2 measures added if hypothermia failed), meaning the trial compared clinical pathways rather than a simple adjunct to an otherwise identical care bundle.1
  • Early stopping and sample size uncertainty: Planned enrolment was not reached, and the study stopped for harm, increasing the risk of overestimating effect sizes and limiting confidence in secondary outcomes and mechanistic explanations; the NIHR programme report provides additional detail on conduct and interpretive constraints beyond the main manuscript.2
  • Position within the hypothermia-in-TBI evidence base: Earlier RCTs of prophylactic hypothermia in adults and children were neutral or harmful overall, and Eurotherm3235’s harm signal for intracranial hypertension management aligns with the broader trajectory of evidence rather than standing as an isolated outlier.345
  • Subsequent trial evidence: The POLAR trial (early prophylactic hypothermia) also did not improve neurological outcomes, reinforcing the conclusion that routine cooling below normothermia is not an effective neuroprotective strategy in unselected severe TBI populations.6
  • Ongoing interest in selected indications: Trials continue to examine longer-duration mild hypothermia for refractory intracranial hypertension (often later and more selected than Eurotherm’s strategy), reflecting persistent uncertainty about whether any niche population benefits; interpretation remains cautious given competing risks and mixed trial signals.7
  • Meta-analytic synthesis after Eurotherm: Contemporary meta-analyses report substantial heterogeneity (population selection, cooling depth/duration, rewarming, and co-interventions), and generally do not demonstrate a consistent improvement in patient-centred outcomes that outweighs complication risk; conclusions depend strongly on trial quality and intervention specifics.8910
  • Guideline impact: Guidelines and expert consensus documents have increasingly emphasised strict fever avoidance and targeted normothermia rather than routine hypothermia, and have not endorsed prophylactic hypothermia as an outcome-improving strategy in severe TBI; any use below normothermia is generally framed as highly selected and evidence-limited.1112131415

Summary

  • In adults with severe TBI and sustained intracranial hypertension, therapeutic hypothermia (32–35°C) as a preferred stage 2 ICP management strategy led to worse 6‑month functional outcomes (adj common OR 1.53; 95% CI 1.02 to 2.30).
  • Mortality was higher with hypothermia (34.9% vs 26.6%; HR 1.45; 95% CI 1.01 to 2.10).
  • Cooling achieved strong temperature separation (−2.14°C), but did not meaningfully reduce mean ICP over 7 days (−0.48 mm Hg).
  • Hypothermia reduced escalation to stage 3 therapies (43.8% vs 54.0%), indicating some strategic effect on escalation pathways despite worse outcomes.
  • Serious adverse events were more frequent with hypothermia (33 vs 10 events), reinforcing that systemic harms can outweigh physiological benefits.

Further Reading

Other Trials

Systematic Review & Meta Analysis

Observational Studies

Guidelines

Notes

  • In Eurotherm3235, effect estimates for functional outcomes were reported such that an odds ratio <1 would favour hypothermia; interpretation should explicitly account for directionality.
  • Several key secondary clinical outcomes (e.g., stage 3 therapy use, serious adverse events) were reported as counts without formal effect estimates; avoid over-interpreting these without prespecified adjusted analyses.

Overall Takeaway

Eurotherm3235 is a landmark because it directly tested a widely used, physiologically appealing ICU strategy—therapeutic hypothermia as an early stage 2 treatment for intracranial hypertension—against standard care using a patient-centred neurological endpoint. Despite achieving substantial cooling, hypothermia worsened long-term outcomes and increased mortality, powerfully shifting modern neurocritical care towards fever prevention and targeted normothermia rather than routine hypothermia for this indication.

Overall Summary

  • For adults with severe TBI and sustained intracranial hypertension, therapeutic hypothermia (32–35°C) as a stage 2 strategy worsened 6‑month functional outcome and increased mortality, despite robust temperature separation and fewer escalations to stage 3 therapies.

Bibliography