Corticosteroids and Mortality in Children With Bacterial Meningitis
Mongelluzzo J, Mohamad Z, Ten Have TR, Shah SS
JAMA. 2008;299:2048-2055
Summary
The authors note that administering corticosteroids to children with bacterial meningitis due to Haemophilus influenzae type b (Hib) reduced hearing loss if the steroid has been given in the first 24 hours and with or before the first dose of antibiotics.
With the change in bacterial causes of meningitis in children with the advent of vaccines against Hib and pneumococcus, it is less clear whether steroids provide a protective effect in children with bacterial meningitis. A Cochrane review evaluating pediatric studies from 1969 to 2006 did not demonstrate a mortality benefit for steroid use in children with bacterial meningitis.
The current multicenter, observational study used data from 27 freestanding US Children's Hospitals to evaluate whether use of steroids provided a survival benefit in children with bacterial meningitis.
Subjects were younger than 18 years and were discharged from the participating hospitals during 2001-2006. The authors identified subjects by discharge diagnoses, and only subjects with ventriculoperitoneal shunts were excluded. The authors captured administration of any of the following steroids in the first 24 hours after admission: dexamethasone, hydrocortisone, and methylprednisolone.
The primary analyses were completed with proportional hazards models, modeling time to death and time to hospital discharge. Because this was an observational study, patients differed in their likelihood of having received steroids (eg, "sicker" patients may have been more likely to receive steroids).
In order to account for such associations in an observational study, the authors calculated a propensity score -- essentially the "risk" that a patient with a given set of conditions would receive steroids. This propensity score was then used as a control variable in analyses to attempt to diminish the bias that would occur if "sicker" patients or patients who were different in some unmeasured way were more likely to receive steroids.
The authors also conducted analyses by age category: younger than 1 year, 1-5 years, and older than 5 years, and they also adjusted for the clustering within the 27 hospitals. There were 2780 subjects with bacterial meningitis during the study period, and 8.9% (248) received steroids in the first 24 hours of treatment.
The subjects had a median age of 9 months (mean 3.4 years), with 18.1% of cases caused by pneumococcus, 13.7% caused by Staphylococcus species, 11.3% by gram-negative bacteria, and 10.1% by Neisseria meningitides.
On admission, 16% of patients required anticonvulsants, and almost 10% required drugs to support blood pressure. The mortality rate was 4.2%, with 23% of deaths occurring in the first 24 hours. The unadjusted death rate was 6% in the steroid recipients and 4% in those without steroids, but this relative risk difference was not statistically significant. In comparing proportions who survived in Kaplan-Meier analyses, there was no difference in survival between the 2 groups.
Finally, in the proportional hazards model that controlled for clinical factors and the steroid propensity score, there was no difference in mortality between the 2 groups. Similarly, there were no differences in duration of hospitalization between the groups. Also, stratification by age and by causative organism did not reveal a protective or negative association with use of steroids.
The authors concluded that in this multicenter observational study, use of steroids was not associated with mortality or length-of-stay outcomes in patients with bacterial meningitis.
Viewpoint
Although these results are no doubt disappointing to many, they underscore how we must make many treatment decisions without having good data. The authors make a plea for a prospective, randomized trial to provide better data, because even the very elaborate statistical efforts to reduce bias in this study may not have been able to account for differences among patients that would be addressed by randomization. They also point out that this study was limited to survival benefit and suggest that future work (presumably with the same cohort) will involve assessment of neurodevelopmental outcomes. Certainly, being able to improve neurodevelopmental outcomes would still be a tremendous benefit of steroid use in bacterial meningitis. Much remains unknown about this important pediatric treatment question.
Abstract