Corticosteroids in acute traumatic brain injury
BMJ 1998; 316 doi: https://doi.org/10.1136/bmj.316.7128.396 (Published 31 January 1998) Cite this as: BMJ 1998;316:396
- David W Newell, Associate professor of neurological surgerya,
- Nancy R Temkin, Associate professor of neurological surgery and biostatisticsa,
- Ross Bullock, Professor of neurosurgeryb,
- Sung Choi, Professor of biostatistics and neurosurgeryb
- a Department of Neurological Surgery, University of Washington, Seattle, WA 98195, USA
- b Division of Neurosurgery, Department of Biostatistics, Medical College of Virginia, Richmond, VA 23298, USA
- c UK Cochrane Centre, NHS R&D Programme, Oxford OX2 7LG
- d Institute of Child Health, University of London, London WC1N 1EH
Editor—The findings of the systematic review of randomised controlled trials of corticosteroids in acute traumatic brain injury1 are in keeping with the conclusions reached by the Committee for Guidelines for the Management of Severe Head Injury, based in the United States. They indicate that the available evidence does not show a benefit of corticosteroids in acute severe traumatic brain injury.2 The guidelines committee therefore recommended that corticosteroids should not be given routinely to patients with severe head injury.
We are puzzled why the authors state that considerable uncertainty remains over the effect of corticosteroids and recommend that a trial requiring 20 000 participants is justified. The meta-analysis showed that the odds ratio for death was 0.91, corresponding to a pooled absolute risk reduction for death in head injured patients treated with steroids of 1.8%, which was far from being significant. Moreover, the subgroup analysis, which analysed trials with only the highest quality of concealment of allocation (blinding), showed that the pooled odds ratios were even closer to unity (indicating no effect) and again not significant (summary odds ratio=1.07 for death and 0.97 for death or disability). In essence, therefore, the pooled odds ratio for all of the trials as well as that for the trials with the best randomisation protocols all have relative risks for death and death or disability close to 1. Thus the meta-analysis provides no justification for the authors' conclusion that a larger trial is warranted.
A larger trial conducted recently to evaluate the effect of the 21-amino steroid tirilazad on the outcome in severely head injured patients showed that it conferred no overall benefit.3 4 Subgroup analysis showed that male patients with subarachnoid haemorrhage present on initial computed tomography fared slightly better than the other patients. This trial had obvious advantages over the meta-analysis by being standardised for type of steroid, dose, duration of treatment, interval from injury to treatment, entry criterion, standardised assessment measures, and data management.
On the basis of both the meta-analysis and the trial of tirilazad, we disagree with the authors' recommendation for a trial to detect possible effects of steroids in head injured patients. About 40 000 patients would be needed to detect the observed 1.8% difference in death rates. Current trials of drugs thought to affect mechanisms of brain damage in head injured patients are designed to detect a 10% effect. A trial of 20 000–40 000 patients to detect a small reduction in mortality with steroids would divert resources and patients away from trials of compounds that may have more promise in improving outcome in head injured patients.
References
Authors' reply
- Philip Alderson, Senior registrar in public health medicinec,
- Ian Roberts, Senior lecturer in epidemiologyd
- a Department of Neurological Surgery, University of Washington, Seattle, WA 98195, USA
- b Division of Neurosurgery, Department of Biostatistics, Medical College of Virginia, Richmond, VA 23298, USA
- c UK Cochrane Centre, NHS R&D Programme, Oxford OX2 7LG
- d Institute of Child Health, University of London, London WC1N 1EH
Editor—We agree with Newell et al that the available evidence does not reliably show a benefit from using corticosteroids in severe head injury. As they point out, the risk of death in the group treated with corticosteroids seems to be about 2% lower than that in the control group (95% confidence interval −6% to 2%). While this is compatible with no benefit from steroids, it is also compatible with a small benefit. Worldwide, several million people are treated each year for severe head injury, of whom over one million die and a similar number are disabled.1 If a treatment as widely practicable as corticosteroids could reduce the risk of death by “only” 2% and the risk of disability by a similar amount, treatment of 100 000 patients would avoid 2000 deaths and prevent 2000 disabilities. Such a benefit would be impossible to show reliably without large scale randomised evidence.
Newell et al say that trials are usually designed to detect reductions in the risk of death of ≥10%. Unfortunately, this will inevitably mean that moderate but clinically important effects, both beneficial and harmful, are missed. Indeed, the reason that the Brain Trauma Foundation's guidelines committee, of which Newell et al were members, was unable to recommend treatment standards for 11 of the 14 topics reviewed is that previous trials have been too small to reliably confirm or refute important treatment effects.
Newell et al are concerned that a trial large enough to detect a moderate effect of corticosteroids would “divert resources and patients away from trials of compounds that may have more promise.” Corticosteroids are given in more than half of cases in two thirds of trauma units in the United States.2 It seems more sensible to marshal this clinical heterogeneity in a scientifically defensible manner by conducting a randomised trial than to continue with such uncontrolled experimentation. Bearing in mind the numbers who sustain head injury each year, we are intrigued by the notion that there might be insufficient people to participate in relatively small trials of compounds that may have more promise. If such a high proportion of patients with head injury did participate in clinical trials one might expect that there would be less uncertainty about the effectiveness of interventions.