Fourteen RCTs were included in the review (n=1,991; range 28 to 499). Quality scores ranged from 6 to 14.5 (where reported). Thirteen RCTs were double-blind, ten had adequate allocation concealment, and five had effect/sample size calculations (where reported).
There were no statistically significant effects of high-dose or low-dose corticosteroid treatment on mortality in patients in severe sepsis and septic shock (13 trials). Corticotrophin response was not significantly different between treatment and control groups (four trials). A statistically significant effect was found for shock resolution in favour of low-dose therapy (OR 2.00, 95% CrI 1.07 to 4.55; r=0.57; seven trials). The odds of shock reversal did not differ according to corticotrophin response (six trials).
Meta-regression did not reveal any statistically significant relationships between mortality and age, or with control arm risk, but trends indicated that treatment efficacy may be dependent on these factors.
Recorded complications of therapy were secondary infections, gastro-intestinal bleeding, and steroid-induced hyperglycaemia, but there were no significant effects arising from either dose category.
The removal of two outlying trials in sensitivity analysis reduced the level of heterogeneity, but failed to alter the main findings. Predicted mortality odds ratios in hypothesised sample sizes of 2,000 and 4,000 patients remained non-statistically significant.
There was no evidence of publication bias.