More trials on early treatment with corticosteroids are needed

EDITOR—We agree with the sentiments expressed by Tarnow-Mordi and Mitra in their editorial—namely, that randomised controlled trials with longer term follow up are required to answer the question of how corticosteroids affect long term outcome in preterm babies.1

We have followed up 42 preterm infants enrolled into a randomised controlled trial of early dexamethasone to prevent chronic lung disease.2 Seventeen died before discharge, and a further three died after discharge from hospital. A detailed assessment was performed in 21 of the 22 survivors (one infant lost to follow up) at 30 months of age, corrected for prematurity, with the assessor blinded to the original treatment allocation. Infants underwent a quantitative neurodevelopmental assessment using the revised Bayley scales of infant development.

The table shows our own (unpublished) results alongside data from the only other early (started before 96 hours of age) steroid studies that have reported long term mortality and neurological outcomes (ES Shinwell, personal communication).3 These data add appreciably to those presented by Tarnow-Mordi and Mitra. In our comparatively small study, the effect of early dexamethasone on any of the long term outcomes studied was not significant. But when our results are included in an overview of all the available data, the pooled estimate suggests a statistically and clinically significant adverse effect of dexamethasone on the incidence of cerebral palsy (number needed to harm (NNH)=5) and significant developmental delay (NNH=6).

The potential short term benefits of early steroids (such as decreased mortality or chronic lung disease) must outweigh the potential risks. For example, approximately 14 infants would need to be treated with early dexamethasone to prevent one developing chronic lung disease4; at the same time, treating six babies with dexamethasone would result in one extra case of cerebral palsy or developmental delay, or both.

A meta-analysis of the results of individual studies in this way disguises their heterogeneity in terms of inclusion criteria, dosage, and regimen of steroid used, and the use of other interventions such as antenatal steroids and postnatal surfactant. We recognise that the validity of presenting these results in the form of a pooled estimate may be questioned, and that in each of the studies neurological morbidity was a secondary outcome. These data serve to illustrate the adverse effects and highlight the need for a definitive randomised controlled trial. We would also urge other authors of postnatal steroid trials to publish long term outcomes if they are available.

We believe that it is important to report not only the rate of cerebral palsy but also the incidence of developmental delay and sensory impairment in survivors. Also of importance is careful analysis of all deaths up to the time of neurological assessment and not just until hospital discharge as is often reported. The ongoing debate regarding the value of early treatment with corticosteroids should prompt further large randomised controlled trials, which not only report long term morbidity but are sufficiently powered to detect small, but clinically important, differences.