Publication

• Title: Effect of intravenous corticosteroids on death within 14 days in 10008 adults with clinically significant head injury (MRC CRASH trial): randomised placebo-controlled trial
• Acronym: CRASH (MRC CRASH)
• Year: 2004
• Journal published in: The Lancet
• Citation: CRASH trial collaborators. Effect of intravenous corticosteroids on death within 14 days in 10008 adults with clinically significant head injury (MRC CRASH trial): randomised placebo-controlled trial. Lancet. 2004;364:1321-1328.

Context & Rationale

• Background

  • Traumatic brain injury (TBI) is a leading cause of death and disability in young adults, with secondary injury mechanisms (oedema, raised intracranial pressure, inflammation) contributing to poor outcomes.
  • Glucocorticoids were widely used after head injury to reduce cerebral oedema and intracranial pressure, despite uncertain effects on patient-important endpoints (mortality and disability).
  • Earlier trials and meta-analyses were underpowered and imprecise, leaving substantial uncertainty about benefit or harm and prompting calls for a large definitive trial.
  • Large between-centre and between-country practice variation persisted, making a pragmatic international trial methodologically and clinically salient.
• Research Question/Hypothesis

  • Does early high-dose intravenous methylprednisolone (started within 8 hours of injury) reduce death within 2 weeks and improve longer-term outcome (death and disability at 6 months) compared with placebo in adults with clinically significant head injury?
• Why This Matters

  • Corticosteroids were inexpensive, globally available, and commonly administered; a modest relative effect (benefit or harm) would translate to large absolute population impact.
  • Establishing net clinical benefit (or net harm) required a sufficiently large, methodologically robust trial using hard clinical outcomes rather than surrogate physiology.

Design & Methods

• Research Question: In adults with clinically significant head injury within 8 hours (GCS ≤14), does a 48-hour high-dose methylprednisolone infusion, compared with placebo, reduce death within 2 weeks (and improve death/disability at 6 months, planned)?
• Study Type: Randomised, placebo-controlled, double-blind, pragmatic, international, multicentre trial (239 hospitals in 49 countries; recruitment 1999–2004).
• Population:
  • Setting: Acute hospital care for head injury (emergency/acute admissions), across diverse health systems internationally.
  • Inclusion: Age ≥16 years; clinically significant head injury defined as Glasgow Coma Scale (GCS) score ≤14; randomisation within 8 hours of injury; treating clinician “substantially uncertain” about whether corticosteroids were indicated (uncertainty principle).
  • Exclusion: Clinician judged a clear indication for corticosteroids or a clear contraindication to corticosteroids.
  • Post-randomisation eligibility deviations (included in ITT): 62 patients later found to be <16 years; 21 randomised >8 hours after injury; infusion stopped in 3 at request of a relative.
• Intervention:
  • Methylprednisolone sodium succinate: loading dose 2 g in 100 mL infused over 1 hour, followed by maintenance infusion 0.4 g/hour for 48 hours (20 mL/hour).
  • Treatment packs contained identical-appearing vials prepared to preserve blinding.
• Comparison:
  • Placebo infusions (sterile water) in identical-appearing vials, delivered with the same loading and maintenance infusion schedule as the intervention arm.
• Blinding: Double-blind (participants and clinicians); emergency unmasking via central service/pager when clinically required (allocation unmasked for 24/10008 [0.2%] patients: 15 methylprednisolone, 9 placebo).
• Statistics: Planned sample size 20,000 to detect a 2% absolute difference in mortality (baseline ~15%) with >90% power at two-sided α=0.01; analyses by intention-to-treat; relative risks with 95% CI for overall effect and 99% CI for prespecified subgroup plots; heterogeneity tests via χ² at the 5% level.
• Follow-Up Period: Early outcomes assessed at discharge, death, or 2 weeks (whichever first); planned assessment of death and disability at 6 months using a structured Glasgow Outcome Scale questionnaire (postal/telephone/home visit as needed; translated/validated instruments).

Key Results

This trial was stopped early. In May 2004, after the Data Monitoring and Ethics Committee disclosed unmasked interim results showing higher mortality with methylprednisolone, the steering committee stopped recruitment after 10,008 of the planned 20,000 participants.

Outcome	Methylprednisolone	Placebo	Effect	p value / 95% CI	Notes
Death (all-cause) within 2 weeks of randomisation	1052/4985 (21.1%)	893/4979 (17.9%)	RR 1.18	95% CI 1.09 to 1.27; P=0.0001	Mortality data obtained for 9964/10008 (99.6%) randomised patients
Death within 2 weeks (injury severity subgroups; GCS at randomisation)	Severe 3–8: 785/1972 (39.8%) Moderate 9–12: 205/1553 (13.2%) Mild 13–14: 62/1460 (4.2%)	Severe 3–8: 687/1972 (34.8%) Moderate 9–12: 143/1476 (9.7%) Mild 13–14: 63/1531 (4.1%)	Not reported	Heterogeneity χ² 3.05; P=0.22	Prespecified; subgroup RRs displayed graphically (99% CI)
Death within 2 weeks (time from injury to randomisation)	≤1 h: 227/1341 (16.9%) >1 to ≤3 h: 308/1530 (20.1%) >3 to ≤8 h: 517/2114 (24.5%)	≤1 h: 209/1342 (15.6%) >1 to ≤3 h: 296/1560 (19.0%) >3 to ≤8 h: 388/2077 (18.7%)	Not reported	Heterogeneity χ² 6.03; P=0.05	Prespecified; suggests greater hazard when treatment started later
Death within 2 weeks (major extracranial injury)	With: 321/1134 (28%) Without: 731/3851 (19%)	With: 244/1082 (23%) Without: 649/3897 (17%)	Not reported	P=0.27	Test for difference in treatment effect by extracranial injury status
Pneumonia treated with antibiotics (within 2 weeks, in hospital)	653/4890 (13.4%)	604/4871 (12.4%)	Not reported	Not reported	Event denominators varied due to differing data completeness by event
Haematemesis or melaena requiring transfusion (within 2 weeks, in hospital)	77/4885 (1.6%)	61/4877 (1.3%)	Not reported	Not reported	Captured during in-hospital follow-up only
One or more recorded complications (within 2 weeks, in hospital)	2942/4985 (59.0%)	2889/4979 (58.0%)	Not reported	Not reported	Complications were not adjudicated; cause of death was not collected
Death and disability at 6 months (planned)	Not reported	Not reported	Not reported	Not reported	Not reported in the index publication (stated to be reported later)

• • High-dose methylprednisolone increased early all-cause mortality (21.1% vs 17.9%; RR 1.18; 95% CI 1.09 to 1.27; P=0.0001).
  • The mortality hazard did not vary by injury severity (heterogeneity P=0.22) and was greatest in those randomised later after injury (heterogeneity P=0.05).
  • Recorded in-hospital complication rates within 14 days were broadly similar between groups; the mechanism of excess mortality was not established in this report.

Internal Validity

• • Randomisation and Allocation:
    • Allocation was concealed using either central telephone randomisation (2141/10008 [21%]) or local numbered treatment packs (7867/10008 [79%]) prepared from a centrally generated random sequence.
    • The approach was designed to minimise selection bias in emergency care; baseline prognostic variables were balanced across groups.
  • Dropout or Exclusions:
    • Two-week mortality outcome was available for 9964/10008 (99.6%) randomised patients (4985 methylprednisolone; 4979 placebo).
    • Eligibility deviations detected after randomisation (age <16 years, randomisation >8 hours) were included in intention-to-treat analyses, preserving randomisation.
  • Performance/Detection Bias:
    • Double-blinding was maintained; emergency unmasking occurred in 24/10008 (0.2%) patients, primarily to permit steroid treatment for newly recognised conditions.
    • Primary endpoint (mortality) is objective and relatively robust to ascertainment bias.
  • Protocol Adherence:
    • Adherence was known for 9848/10008 (98%) patients; 9748/9848 (99%) received the full loading dose.
    • Because of early death or discharge, 8286/9848 (83%) received at least 24 hours of maintenance infusion.
  • Baseline Characteristics:
    • Severity spectrum was broad (severe GCS 3–8: 40% vs 39%; moderate 9–12: 31% vs 30%; mild 13–14: 29% vs 31%).
    • Time to randomisation was early (median 3 hours; IQR 1–5), within the hypothesised therapeutic window for limiting secondary injury.
  • Heterogeneity:
    • No evidence that mortality effect differed by prespecified injury severity strata (heterogeneity P=0.22).
    • Evidence of differential effect by time from injury to randomisation (heterogeneity P=0.05), raising a biologically and clinically relevant interaction hypothesis.
  • Timing:
    • Intervention delivery was time-critical and generally achieved (randomisation within 8 hours by design; 21 patients randomised >8 hours).
  • Dose:
    • The tested regimen was a fixed, very high-dose methylprednisolone strategy (2 g loading, then 0.4 g/hour for 48 hours), reflecting contemporary practice extrapolated from other neurotrauma contexts.
    • Internal validity is strongest for this specific dosing strategy; it does not directly test lower-dose, shorter-course, or physiologically guided steroid regimens.
  • Separation of the Variable of Interest:
    • Intervention: methylprednisolone 2 g over 1 hour then 0.4 g/hour for 48 hours.
    • Control: identical placebo loading and maintenance infusions.
    • Unmasking was rare (0.2%), supporting clear separation between arms.
  • Key Delivery Aspects:
    • Eligibility was based on clinician uncertainty, enhancing pragmatic enrolment but potentially excluding patients with strong pre-existing indications/contraindications (a trade-off between generalisability and “pure” biological inclusion criteria).
    • Concomitant therapies and physiological mediators (e.g., glucose, intracranial pressure) were not comprehensively captured, limiting mechanistic inference.
  • Outcome Assessment:
    • Mortality was ascertained to 2 weeks with minimal missingness.
    • Non-fatal events after discharge within 14 days were not recorded, which may underestimate event rates but is unlikely to explain the mortality divergence.
  • Statistical Rigor:
    • Analyses were intention-to-treat with prespecified subgroup structure and appropriate confidence intervals; however, early stopping for harm can influence effect-size estimation and is best interpreted with caution for magnitude (not direction) of effect.

Conclusion on Internal Validity: Overall, internal validity for the observed increase in 2-week mortality appears strong (concealed allocation, double blinding, near-complete outcome ascertainment, objective primary endpoint), with the main limitations being early stopping and limited measurement of co-interventions/mechanistic mediators.

External Validity

• • Population Representativeness:
    • Large international recruitment (239 hospitals, 49 countries) supports broad applicability across diverse resource settings.
    • Included a wide severity spectrum (mild to severe) and substantial extracranial injury burden (23% vs 22% with major extracranial injury), reflecting real-world trauma case-mix.
  • Applicability:
    • Directly applicable to routine early administration of high-dose methylprednisolone in adult clinically significant head injury within 8 hours.
    • Extrapolation to paediatric populations, to patients randomised beyond 8 hours, or to alternative steroid molecules/doses/physiology-guided approaches is uncertain (not tested).

Conclusion on External Validity: External validity is high for adult hospital-managed head injury within 8 hours across heterogeneous global settings; applicability is limited to the tested high-dose methylprednisolone regimen and does not establish effects of other corticosteroid strategies.

Strengths & Limitations

• Strengths:
  • Very large sample (10,008 randomised) with high completeness of the primary outcome (99.6%).
  • International, multicentre pragmatic design across 49 countries, enhancing generalisability.
  • Double-blind, placebo-controlled delivery with rare unmasking (0.2%), supporting unbiased mortality ascertainment.
  • Prespecified subgroup framework and robust estimation using relative risks with confidence intervals.
• Limitations:
  • Stopped early for harm, which can affect effect-size estimation and limits planned precision for smaller subgroup interactions.
  • The index report provides early (2-week) outcomes; planned 6-month disability outcomes were not reported in this publication.
  • Cause of death, detailed clinical course, and granular co-interventions (including intracranial pressure management and metabolic mediators) were not systematically collected, constraining causal explanation for excess mortality.
  • Non-fatal outcomes after discharge within 14 days were not recorded.

Interpretation & Why It Matters

• Clinical and methodological implications

  • Routine early high-dose methylprednisolone in adult clinically significant head injury increases early mortality and should not be used for this indication.
  • CRASH demonstrates that biologically plausible therapies (reducing oedema/inflammation) can be clinically harmful when tested against hard endpoints in adequately powered trials.
  • The trial exemplifies the value of large pragmatic emergency-care RCTs: broad eligibility, concealed allocation, and patient-important outcomes can decisively overturn entrenched practice.

Controversies & Other Evidence

• • Why did mortality increase?
    • The index publication reported no large rise in recorded infectious complications or gastrointestinal bleeding, and the mechanism of excess mortality remained unresolved.
    • A contemporaneous Lancet commentary highlighted that limited information on concomitant extracranial injuries, clinical course, and concurrent interventions made it difficult to explain how steroids caused harm despite the indisputable mortality signal.¹
  • Early stopping and interpretation:
    • Correspondence in The Lancet addressed interpretability issues arising from early stopping and emphasised the importance of understanding timing and biological plausibility when reconciling CRASH with prior smaller trials.²³
  • Synthesis of evidence:
    • After CRASH, the updated Cochrane review concluded that corticosteroids should not be used routinely in acute traumatic brain injury because of increased mortality.⁴
  • Guidelines and practice change:
    • Subsequent evidence-based guidelines incorporated CRASH and recommend against corticosteroids for improving outcome or reducing intracranial pressure in severe traumatic brain injury.⁵

Summary

• • Large, pragmatic, double-blind international RCT randomised 10,008 adults (≤8 hours; GCS ≤14) to high-dose methylprednisolone vs placebo.
  • Trial stopped early after interim analyses identified excess mortality with methylprednisolone.
  • Two-week mortality was higher with methylprednisolone (21.1% vs 17.9%; RR 1.18; 95% CI 1.09 to 1.27; P=0.0001).
  • Excess mortality did not differ by injury severity and appeared greatest in patients randomised later after injury (heterogeneity P=0.05).
  • Recorded in-hospital complications within 14 days were broadly similar; the mechanism of increased mortality was not established in the index report.

Overall Takeaway

CRASH decisively demonstrated that early high-dose methylprednisolone after clinically significant head injury increases short-term mortality, overturning a long-standing pathophysiological rationale and practice pattern. Its scale, blinding, and near-complete follow-up make the mortality signal highly credible, and it has directly shaped modern guidelines by establishing that routine corticosteroids are harmful in acute TBI.

Bibliography

• Sauerland S, Maegele M. A CRASH landing in severe head injury. Lancet. 2004;364:1291-1292.
• Douglas IS, Sava J. Possible explanations for the results of CRASH. Lancet. 2005;365(9455):212-214.
• Peto R. Possible explanations for the results of CRASH. Lancet. 2005;365(9455):212-214.
• Alderson P, Roberts I. Corticosteroids for acute traumatic brain injury. Cochrane Database Syst Rev. 2005;(1):CD000196.
• Carney N, Totten AM, O’Reilly C, et al. Guidelines for the Management of Severe Traumatic Brain Injury, Fourth Edition. Neurosurgery. 2017;80(1):6-15.