Skip to main content

Foundational Trials

Clifton

Image: Shutterstock / ChaNaWiT

Publication

  • Title: Lack of effect of induction of hypothermia after acute brain injury
  • Acronym: National Acute Brain Injury Study: Hypothermia (NABIS:H)
  • Year: 2001
  • Journal published in: The New England Journal of Medicine
  • Citation: Clifton GL, Miller ER, Choi SC, et al. Lack of effect of induction of hypothermia after acute brain injury. N Engl J Med. 2001;344(8):556-563

Context & Rationale

  • Background
    Moderate systemic hypothermia was biologically plausible for neuroprotection after traumatic brain injury (TBI) and early small trials suggested improved outcomes, but results were inconsistent and potentially confounded by case mix and co-interventions.
  • Research Question/Hypothesis
    Does early induction of moderate hypothermia (target 33°C), initiated within 6 hours of closed head injury and maintained for 48 hours, improve 6-month functional outcome compared with normothermia in comatose adults with severe TBI?
  • Why This Matters
    If effective, hypothermia would represent a widely deployable disease-modifying therapy for severe TBI rather than solely supportive intracranial pressure–directed care.

Design & Methods

  • Research Question: In comatose patients after severe closed head injury, does induced hypothermia (33°C) begun within 6 hours and maintained for 48 hours improve 6-month functional outcome versus normothermia?
  • Study Type: Prospective, multicentre, randomised trial (NABIS:H); stratified by centre and initial Glasgow Coma Scale (GCS); standardised co-interventions including intracranial pressure monitoring and protocolised intracranial hypertension management.
  • Population:
    • Setting: acute hospital care (including ICU) at 11 centres; majority enrolled at 5 centres.
    • Inclusion: age 16–65 years; non-penetrating head injury; GCS 3–8 after resuscitation; ability to initiate cooling within 6 hours after injury.
    • Key exclusions: GCS 3 with unreactive pupils; life-threatening non-brain injury; systolic blood pressure <90 mm Hg after resuscitation; oxygen saturation <94% after resuscitation; active bleeding; pregnancy; major pre-existing comorbidity (e.g., severe heart disease).
  • Intervention:
    • Target bladder temperature 33°C; cooling initiated immediately after randomisation and intended to achieve target within 8 hours after injury; maintained 32.5–34.0°C for 48 hours using surface cooling (including temperature-control pads integrated into a kinetic treatment table) plus adjuncts (ice, iced gastric lavage, room-temperature ventilator air).
    • Rewarming no faster than 0.5°C per 2 hours.
    • Neuromuscular blockade used for 72 hours to prevent shivering in the hypothermia group.
  • Comparison:
    • Normothermia targeted to 37.0°C with otherwise identical standard care (including ICP monitoring, sedation/analgesia, and stepwise intracranial hypertension management).
    • Neuromuscular blockade in the normothermia group as needed for respiratory management.
  • Blinding: Treating clinicians were not blinded (temperature management is not practically concealable); 6-month outcome assessments were performed by examiners unaware of treatment assignment.
  • Statistics: Planned sample size 500; a prespecified power calculation (effect size/alpha/power) was not reported in the available trial materials. Primary analysis was intention-to-treat; interim monitoring included futility stopping rules (trial stopped when probability of detecting a treatment effect at N=500 was <0.01).
  • Follow-Up Period: 6 months for primary functional outcome (Glasgow Outcome Scale; favourable = good recovery or moderate disability).

Key Results

This trial was stopped early. Enrolment stopped after 392 of the planned 500 participants because an interim analysis showed the probability of detecting a treatment effect was <0.01 if the trial continued to its planned sample size.

Outcome Hypothermia (33°C) Normothermia Effect p value / 95% CI Notes
Primary: poor functional outcome at 6 months (severe disability, vegetative state, or death) 108/190 (57%) 102/178 (57%) RR 1.0 95% CI 0.8 to 1.2; P=0.99 Outcome adjusted for age and admission GCS (analysis set n=368).
Mortality at 6 months 53/190 (28%) 48/178 (27%) RR 1.0 95% CI 0.7 to 1.4; P=0.79 No difference in neuropsychological test battery at 6 months (reported as not different; values not shown).
Protocol delivery: time injury→randomisation 4.3 ± 1.1 h 4.1 ± 1.2 h Not reported Not reported Mean time injury→target 33°C: 8.4 ± 3.0 h (hypothermia group).
Intracranial hypertension burden: ever ICP >30 mm Hg (days 1–4) 41% 59% Not reported P=0.02 Day 2: 14% vs 28% (P=0.002); Day 3: 16% vs 26% (P=0.03).
Hospital days with any recorded complication (per patient) 78% ± 22 70% ± 29 Not reported P=0.005 Composite of 67 predefined complications recorded daily.
Critical hypotension (MAP <70 mm Hg with organ failure for ≥2 consecutive hours) 10% 3% Not reported P=0.01 Safety signal aligned with greater vasopressor exposure in hypothermia group (80% vs 69%; P=0.01).
Bradycardia with hypotension for ≥2 consecutive hours 16% 4% Not reported P=0.04 Clinically relevant hypothermia-associated adverse physiology.
Renal signal: serum creatinine >2.5 mg/dL (221 µmol/L) in first 96 hours 8% 0.3% Not reported P=0.05 Other mean laboratory values reportedly within normal ranges.
Key subgroup: age >45 years, poor outcome 23/26 (88%) 18/26 (69%) RR 1.3 95% CI 1.0 to 1.7; P=0.08 Not statistically significant, but directionally worse; mortality 38% vs 38%.
Exploratory: admission temperature ≤35°C, poor outcome 38/62 (61%) 31/40 (78%) RR 0.8 95% CI 0.6 to 1.0; P=0.09 Hypothermia-on-admission subgroup had greater injury severity; interpretation limited by baseline imbalance.
  • Despite achieving mean hypothermia 33.2 ± 1.0°C for 47.2 ± 3.0 hours and blinded 6-month assessment, induced hypothermia did not improve functional outcome or survival.
  • Hypothermia reduced the proportion of patients with very high ICP (>30 mm Hg) but this surrogate improvement did not translate into better 6-month outcomes.
  • Hypothermia was associated with more haemodynamic complications (critical hypotension; bradycardia with hypotension) and a higher overall complication burden.

Internal Validity

  • Randomisation and allocation: Randomised with stratification by centre and initial GCS; allocation concealment method not reported.
  • Follow-up completeness: 6-month outcomes obtained for 385/392 (98%); adjusted primary analysis reported for 368 due to missing/inaccurate age or admission GCS in 17 plus missing outcomes in 7.
  • Performance/detection bias: Treating teams unblinded (temperature management), but primary outcome assessed by blinded examiners at 6 months.
  • Protocol adherence: Nine patients assigned to hypothermia did not receive hypothermia (protocol violation); in the normothermia group, 35% had temperature ≤35°C at some time during the first 16 hours after injury, reducing physiological separation early after injury.
  • Timing and dose: Injury→target 33°C was 8.4 ± 3.0 hours; if neuroprotection requires much earlier cooling, the tested strategy may have been inherently late for the hypothesised mechanism (biological plausibility issue rather than internal validity per se).
  • Separation of the variable of interest: Mean temperature during first 48 hours: 33.2 ± 1.0°C (hypothermia) vs mean 96-hour temperature 37.2 ± 0.8°C (normothermia); however early spontaneous hypothermia occurred in the control arm (≤35°C in 35% during first 16 hours).
  • Heterogeneity: Severe TBI is clinically heterogeneous (diffuse injury vs mass lesions, variable extracranial injury); randomisation produced broadly similar baseline characteristics, but biologically relevant strata (e.g., admission temperature) complicated interpretation.
  • Outcome assessment and statistical rigour: Objective, prespecified primary endpoint (GOS at 6 months); intention-to-treat analysis used; early stopping for futility reduces risk of false-positive findings but may limit precision for subgroup effects.

Conclusion on Internal Validity: Overall, internal validity appears moderate: randomisation and blinded outcome assessment support causal inference, but incomplete early physiological separation (control hypothermia), protocol violations, and early stopping constrain certainty about small benefits in specific subgroups.

External Validity

  • Population representativeness: Adults 16–65 with severe closed head injury and coma after resuscitation; excludes the physiologically most unstable (persistent hypotension/hypoxaemia) and those with major extracranial life-threatening injury, limiting applicability to polytrauma/shock-heavy cohorts.
  • Applicability: Tested a pragmatic surface-cooling approach achievable in most ICUs, but required initiation within 6 hours and achieved target at ~8 hours on average; settings without rapid trauma systems may struggle to reproduce timing.
  • Practice context: Co-interventions reflected intensive, protocolised TBI care (ICP monitoring and CPP targets), which may not be universal; effect estimates could differ where baseline care is less standardised.

Conclusion on External Validity: Generalisability is moderate for high-resource systems caring for severe closed TBI within organised trauma pathways, and limited for older patients, penetrating trauma, or persistently shocked/hypoxaemic presentations.

Strengths & Limitations

  • Strengths: Multicentre randomised design; prespecified protocol for temperature targets, maintenance and rewarming; ICP monitoring and protocolised intracranial hypertension management; high 6-month follow-up; blinded primary outcome assessment; explicit safety monitoring with independent stopping rules.
  • Limitations: Treating clinicians unblinded; early control-group hypothermia (≤35°C in 35% during first 16 hours) may have diluted treatment separation; time-to-target 33°C averaged 8.4 hours (potentially late for neuroprotection); nine protocol violations in the hypothermia arm; early stopping limits precision for subgroup effects.

Interpretation & Why It Matters

  • Clinical implication
    Routine induction of systemic hypothermia to 33°C (achieved by ~8 hours post-injury and maintained 48 hours) should not be used as a neuroprotective strategy in unselected severe TBI, given neutral functional outcome and increased complication burden; its effect on ICP suggests a narrower role (if any) as a rescue adjunct for refractory intracranial hypertension rather than prophylactic neuroprotection.
  • Conceptual contribution
    The dissociation between improved ICP surrogates and unchanged 6-month clinical outcome reinforces that physiological endpoints are insufficient substitutes for patient-centred outcomes in TBI trials.

Controversies & Subsequent Evidence

  • Surrogate endpoint fallacy: The trial demonstrated fewer episodes of very high ICP with hypothermia, but no improvement in 6-month GOS; this was highlighted as evidence that ICP is an unreliable surrogate for efficacy in neurocritical care interventions 1.
  • Timing as a mechanistic critique: Editorial commentary noted that effective neuroprotection in animal models typically requires cooling within ~90 minutes, whereas this strategy achieved target temperature at ~8 hours on average, raising the possibility that biologically “late” cooling was destined to be ineffective in most patients 1.
  • Monitoring temperature at the brain level: Published correspondence criticised the absence of direct brain temperature monitoring and questioned whether patient heterogeneity (injury types) could have obscured a benefit in specific lesion patterns 2.
  • Later large RCTs reinforced lack of benefit (and potential harm) in key use-cases:
    • ICP-targeted hypothermia (Eurotherm3235) showed worse 6-month functional outcome with hypothermia (adjusted common OR 1.53; 95% CI 1.02 to 2.30; P=0.04) despite reduced need for stage 3 ICP therapies 3.
    • Early prophylactic hypothermia (POLAR-RCT) did not improve favourable 6-month outcomes (48.8% vs 49.1%; RR 0.99; 95% CI 0.82 to 1.19; P=0.94) 4.
  • Guidelines incorporated the negative adult evidence: Brain Trauma Foundation 4th edition recommendations do not support early short-term prophylactic hypothermia to improve outcomes 5.

Summary

  • In 392 adults (16–65 years) with severe closed TBI (GCS 3–8), induced hypothermia to 33°C maintained for 48 hours did not improve 6-month functional outcome (poor outcome 57% in both groups).
  • Mortality at 6 months was similar (28% vs 27%), but hypothermia increased haemodynamic complications and overall complication-days.
  • Hypothermia reduced the proportion of patients experiencing very high ICP (>30 mm Hg), yet this did not translate into better clinical outcomes.
  • Early stopping for futility strengthens the inference that this tested strategy does not provide clinically meaningful benefit in unselected severe TBI.
  • Subgroup signals (worse outcomes in older patients; potential interaction with spontaneous admission hypothermia) were hypothesis-generating rather than definitive.

Further Reading

Other Trials

Systematic Review & Meta Analysis

Observational Studies

Guidelines

Notes

  • Where a guideline is primarily distributed as an organisational PDF or web resource rather than a journal article, the link points to the authoritative hosting page.

Overall Takeaway

This landmark trial demonstrated that a pragmatic strategy of induced systemic hypothermia to 33°C—achieved by approximately 8 hours after injury and maintained for 48 hours—does not improve 6-month outcomes after severe traumatic brain injury and increases clinically important complications. It shifted the field away from routine prophylactic cooling for neuroprotection and reinforced the necessity of patient-centred outcomes (rather than ICP surrogates) to judge efficacy in neurocritical care.

Overall Summary

  • Induced hypothermia to 33°C (target achieved at 8.4 ± 3.0 hours) did not improve 6-month functional outcome (poor outcome 57% vs 57%) and increased complications
  • Physiological improvement in very high ICP episodes did not translate into clinical benefit
  • Later large RCTs and guidelines largely corroborated neutral or harmful effects for prophylactic or ICP-targeted systemic hypothermia

Bibliography