Corticosteroid Therapy for Severe Sepsis and Septic Shock

Abstract

To the Editor: The systematic review by Dr Annane and colleagues summarized the results of 22 clinical trials investigating corticosteroid therapy in sepsis. The review suggested that prolonged low-dose corticosteroid therapy ( 5 days, 300 mg/d of hydrocortisone or equivalent) decreases mortality and recommended treating adult patients with vasopressor-dependent septic shock. However, we believe that these recommendations are weakened by possible publication bias and by the relationship between severity of illness and the treatment effect of corticosteroids. The review stated that “the funnel plot did not suggest evidence for publication bias.” However, in a recent systematic review of corticosteroids in sepsis, we demonstrated publication bias within the group of trials investigating prolonged low-dose corticosteroids using an Egger regression test and a funnel plot. Therefore, we recommend performing similar analyses to examine the 12 trials investigating prolonged low-dose corticosteroids included in this review. Findings consistent with publication bias would imply that the overall reported improvement in mortality may be due to inclusion of a disproportionately high number of small trials demonstrating beneficial effects of corticosteroids. Small negative trials may be unavailable for inclusion because they were either not reported or not accepted for publication. Potential publication bias is concerning because the CORTICUS trial, accounting for more than 40% of patients in this subgroup analysis, demonstrated no benefit of low-dose corticosteroids in patients with vasopressor-dependent septic shock. The review also suggested an interaction between mortality in the control group and the effects of corticosteroid therapy (P=.06). We recommend examining the data (excluding the Schumer trial for consistency with the other analyses) for a relationship between severity of illness (measured by control group mortality) and treatment effect, using fixed-effects or unrestricted maximum likelihood models. If significant, this would suggest that low-dose corticosteroids decrease mortality only in more severely ill patient populations and would be consistent with our findings and with reported effects of other anti-inflammatory agents in sepsis. Furthermore, such a relationship might in part explain the lack of benefit in the CORTICUS trial, which had a relatively low control mortality (31%) compared with the other the trials (median [interquartile range], 44% [25%-61%]). Long-course low-dose corticosteroids may improve survival only in severely ill patients with refractory vasopressordependent septic shock. Until further data become available, we believe that decisions to treat severely ill septic patients with corticosteroids should be based on individual patient risk-benefit assessments.