We identified 20 trials with 12,303 randomised participants. The effect of corticosteroids on the risk of death was reported in 17 included trials. Due to significant heterogeneity we did not calculate a pooled estimate of the risk of death. The largest trial , with about 80% of all randomised participants, found a significant increase in the risk ratio of death with steroids (95% CI to ) and a relative risk of death or severe disability of (95% CI to ). For infections the pooled risk ratio from five trials was (95% CI to ) and for the ten trials reporting gastrointestinal bleeding (95% CI to ).