Effect of intravenous corticosteroids on death within 14 days in 10 008 adults with clinically significant head injury: randomised placebo-controlled trial

Roberts et al for the CRASH trial collaborators. Lancet 2004; 364:1321-28. doi:10.1016/S0140-6736(04)17188-2

Clinical Question

• In adults with a head injury, do early corticosteroids compared to placebo reduce death and disability?

Design

• Randomised, controlled trial
• Centralised (21%) and non-centralised (79%) randomised concealed allocation method
• Blinding of clinicians, patients, and data analysts
• Annual interim analyses, with unmasking only if:
  • “proof beyond reasonable doubt of a difference” was found or,
  • new published data removed equipoise and uncertainty regarding corticosteroid use
• Intention-to-treat analysis
• Powered at 90% to detect 2% difference in mortality from baseline of 15%, with accepted type I error of 0.01 (chance of false positive), if 20,000 patients were recruited

Setting

• 239 hospitals from 49 countries
  • Europe 38%, Asia 27%, South America 16%, Africa 14%, North America 4%, Oceania 1%
• April 1999 to May 2004

Population

• Inclusion: adults over 16 years with head injuries, who presented within 8 hours of injury with GCS ≤ 14
• Exclusion: clear indication or contraindication for steroids
• 10,008 patients randomised

Intervention

• Administration of methylprednisolone for 48 hours
  • Loading dose of 2 g over 1 hour in 100 ml 0.9% NaCl
  • Maintenance of 0.4 g per hour for 48 hours in 20 ml per hour 0.9% NaCl

Control

• Identical regime of placebo

Outcome

• Primary outcome: unmasking of randomisation took place after 10,008 patients due to a clear difference favouring placebo for mortality at 2-weeks
  • Corticosteroid group 21% mortality vs placebo group 18% mortality
  • Relative risk 1.18 (95% CI 1.09–1.27; p=0.0001)
  • Absolute risk increase 3.15% (95% CI 1.60%–4.70%; NNH 32)
• Secondary outcome: 6-month follow-up favoured placebo for mortality and severe disability

 

Primary Outcome

Measure	Corticosteroids	Placebo	RR	ARR	NNH
2-week mortality	1052 (21.1%)	893 (17.9%)	1.18 (95% CI 1.09–1.27; p=0.0001)	-3.15% (95% CI 1.60%–4.70%)	31
CI = confidence interval; p = p-value; RR = relative risk; ARR = absolute risk reduction; NNH = number-needed-to-harm.

Secondary Outcomes

Measure	Corticosteroids	Placebo	RR	ARR	NNH
6-month mortality	1248 (25.7%)	1075 (22.3%)	1.15 (95% CI 1.07–1.24; p=0.0001)	-3.40% (95% CI 1.70%–5.10%)	29
6-month mortality or severe disability	1828 (38.1%)	1728 (36.3%)	1.05 (95% CI 0.99–1.10; p=0.079)	-1.80% (95% CI -0.12%–3.72%)	55

Authors’ Conclusions

• “Corticosteroids should not be used routinely to treat head injury, whatever the severity.”

Strengths

• Excellent external validity due to worldwide recruitment and minimal exclusion criteria
• Robust statistical thresholds in planning phase (⍺<0.01, β=90%)
• Primary outcome data available for over 99% of patients
• Rapid publication of important outcome with secondary outcomes published later

Weaknesses

• Halting at interim stage may produce results that are due to an extreme play of chance
• Hyperglycaemia from corticosteroid administration may have unblinded clinicians and / or led to unbalanced management of the two groups (see Lancet comments)
• Cause of death not investigated, so further theories and an explanation for the results are limited

Metadata

Summary author: @DuncanChambler
Summary date: 15 August 2014
Peer-review editor: @davidslessor