CRASH trial is based on problematic meta-analysis

BMJ 1999; 319 doi: http://dx.doi.org/10.1136/bmj.319.7209.578 (Published 28 August 1999) Cite this as: BMJ 1999;319:578
  1. B Gregson, principal research associate,
  2. N V Todd, consultant anaesthetist, intensive care,
  3. D Crawford, consultant anaesthetist, intensive care,
  4. C J Gerbe, consultant neurosurgeon,
  5. B Fulton, consultant neurosurgeon,
  6. L Taccon, consultant neurosurgeon,
  7. P J Crawford, consultant neurosurgeon,
  8. R P Sengupta, consultant neurosurgeon
  1. Newcastle General Hospital, Acute Services, Newcastle upon Tyne NE4 6BE

    EDITOR—Kmietowicz reports the start of the CRASH trial, a multicentre trial of steroids for treating head injury.1 The meta-analysis which forms the basis for the trial is a good example of how one small study can potentially expose thousands of patients to unnecessary risk. The dose of steroids used in CRASH (2 g methyl prednisolone over 1 hour followed by 0.39 g/h for 48 hours) may cause fluid retention, raised glucose concentrations, immunosuppression, and gastric erosions with gastrointestinal bleeding. These hazards are particularly important in vulnerable and often ventilated head injured patients.

    The suggestion of possible benefit from steroids comes from the meta-analysis by Alderson and Roberts published in 1997.2 In that meta-analysis the odds ratio for death (0.91, 95% confidence interval 0.74 to 1.12) is heavily influenced by the small trial by Faupel et al in 1976.3 This trial was of poor quality, did not use acceptable randomisation (class B according to Alderson and Roberts), and used discharge outcomes which Alderson and Roberts say may account for “incongruous” results.

    If the Faupel trial is removed from the meta-analysis, steroid treatment does not seem to convey benefits. We recalculated the Peto odds ratio without the Faupel study (figure). If the other trials which do not use adequate concealment are also removed, any possible benefit is eliminated, and indeed the Peto odds ratio would indicate that steroids are potentially harmful (figure).


    Meta-analysis of trials of steroids for head injury recalculated without study by Faupel et al and including only trials with adequate concealment

    Ethics committees considering the safety of high dose steroid treatment in head injury need to consider the significant risks of these doses and weigh them against the vanishingly small benefits that may accrue.


    1. 1.
    2. 2.
    3. 3.