Current issues of ACP Journal Club are published in Annals of Internal Medicine


Review: Glucocorticoids reduced mortality in sepsis in recent (post-1997) but not previous (pre-1989) trials, or all trials combined


ACP J Club. 2005 Mar-Apr;142:31. doi:10.7326/ACPJC-2005-142-2-031

Related Content in this Issue
• Companion Abstract and Commentary: Review: Low-dose but not high-dose corticosteroids reduced all-cause mortality in severe sepsis and septic shock

Clinical Impact Ratings

Hospitalists: 7 stars

Critical Care: 6 stars

Source Citation

Minneci PC, Deans KJ, Banks SM, Eichacker PQ, Natanson C. Meta-analysis: the effect of steroids on survival and shock during sepsis depends on the dose. Ann Intern Med. 2004;141:47-56. [PubMed ID: 15238370]



In patients with sepsis, are glucocorticoids more effective than a control intervention for reducing mortality?


Data sources: Studies were identified by searching MEDLINE (1988 to April 2003) {and included trials from a previous meta-analysis}*.

Study selection and assessment: Studies were selected if they were randomized controlled trials (RCTs) that enrolled adult patients with sepsis (documented site or strong suspicion of infection, temperature > 38°C or < 36°C, heart rate > 90 beats/min, respiratory rate > 20 breaths/min, and leukocyte count > 12 × 109 cells/L) or septic shock (hypotension despite fluid resuscitation plus hypoperfusion abnormalities), had a primary endpoint (either the discontinuation of vasopressor therapy or a change in survival), and compared glucocorticoids with a control intervention (antibiotics, vasopressors, or fluids) with or without placebo. Study quality was assessed on method and adequacy of randomization, blinding, completeness of follow-up, adherence to treatment protocols, and cointerventions.

Outcomes: Survival at 14 to 28 days, and reversal of shock.

Main results

14 RCTs (n = 1717) met the selection criteria. RCTs were partitioned according to publication date: previous studies (9 RCTs published before 1989) and recent studies (5 placebo-controlled RCTs published after 1997). 4 of the 5 recent RCTs had complete follow-up and adequate randomization and enrolled patients with vasopressor-dependent septic shock. Previous RCTs used a wider range of inclusion criteria (from “severe infections” to shock). Glucocorticoids used were hydrocortisone, betamethasone, dexamethasone, methylprednisolone, and prednisolone. Recent RCTs administered glucocorticoids later (23 h vs < 2 h, P = 0.02) and for longer courses (6 vs 1 d, P = 0.01) of lower-dose steroid therapy (1209 mg vs 23 975 mg, P = 0.01) than previous RCTs. Using a fixed-effects model, meta-analysis of 13 RCTs (previous and recent) showed that glucocorticoid and control groups did not differ for survival (Table). The groups did not differ for survival in previous RCTs (Table). Of 4 recent RCTs that reported mortality and reversal of shock, more patients who received glucocorticoids survived (Table) and had reversal of shock (3 RCTs [relative benefit range 1.13 to 3.24, 95% CI 0.86 to 7.01]; 1 RCT [hazard ratio 1.54, CI 1.10 to 2.16]) than did those who received the control intervention.


In patients with sepsis, glucocorticoids reduced mortality more than a control intervention in trials published after 1997, but not in trials published before 1989 or in all trials combined.

*Cronin L, Cook DJ, Carlet J, et al. Crit Care Med. 1995;23:1430-9. 7634816

Source of funding: National Institutes of Health.

For correspondence: Dr P.C. Minneci, National Institutes of Health, Bethesda, MD, USA. E-mail

Table. Glucocorticoids (GCs) vs control (antibiotics, vasopressors, or fluids) with or without a placebo for reducing mortality in sepsis and septic shock at 14 to 28 days†

Trials Number of trials (n) Weighted event rates RRR/RRI (95% CI) NNT/NNH (CI)
GCs Control
All trials 13 (1717) 39% 40% RRR: 1% (−10 to 12) NNT: Not significant
Previous trials 9 (1297) 36% 35% RRI: 7% (−7 to 23) NNH: Not significant
Recent trials 4 (420) 47% 57% RRR: 17% (1 to 31) NNT: 11 (6 to 134)

†Previous trials = published before 1989; recent trials = published after 1997. Abbreviations defined in Glossary; weighted event rates, RRR, RRI, NNT, NNH, and CI calculated from data in article using a fixed-effects model.


Both clinicians and researchers have hypothesized 2 distinct mechanisms of CS therapy, which might improve outcomes for patients with severe sepsis. The first premise is that high-dose CSs suppress the deleterious and uncontrolled systemic inflammatory response of severe sepsis. The second theory is that relatively low doses of CSs address the physiologic needs of patients with sepsis-associated adrenal insufficiency.

The 2 reviews by Annane and colleagues and Minneci and colleagues highlight pivotal advances related to CS therapy in sepsis and introduce new clinically important insights. The systematic review by Annane and colleagues confirms earlier findings that high-dose CS therapy in severe sepsis and septic shock does not improve survival and may cause harm as a result of secondary infections. In striking contrast, however, both reviews showed that a 5- to an 11-day course of low-dose CSs (200 to 300 mg of hydrocortisone per d) accelerated shock reversal and improved survival at 28 days, ICU discharge, and hospital discharge in patients with vasopressor-dependent septic shock.

The 17 original studies of these reviews differ extensively. Methodological quality was generally superior in the more recent trials. Among the pool of over 2000 patients, study populations from recent trials were uniformly sicker. Although early studies included an assortment of patients with sepsis, severe sepsis, or septic shock, recent trials focused almost exclusively on patients with vasopressor-dependent septic shock. Control-group mortality rates were correspondingly higher among later trials.

Protocols for CS administration also varied. Early trials used higher daily doses over relatively short periods, while recent trials investigated longer courses of lower doses in the physiologic range. Interestingly, total steroid doses were larger in the earlier trials. Finally, the primary studies measured adverse effects of CS therapy (secondary infection, gastrointestinal bleeding, organ system failure, and hyperglycemia) inconsistently, although each of the systematic reviews provided a comprehensive and quantitative summary.

Despite careful explorations to understand the divergent results between early and recent trials, the only compelling explanation relates to differences in daily dose and duration of therapy. Disparities in quality did not appear to influence study findings but cannot be ruled out. Similarly, the relation between sepsis severity and therapeutic benefit is uncertain. For now, the evidence is clear only for those patients with septic shock, for whom low-dose therapy confers a statistically significant and clinically important survival benefit.

An unresolved issue is the role for adrenal testing to select patients for a course of CS therapy. On this issue, the authors of the 2 reviews disagree. 3 relatively recent trials conducted subgroup analyses in patients with and without biochemical evidence of adrenal insufficiency. Observed differences between the subgroups were relatively small, not statistically significant, and inconsistent across the 3 trials (as shown in the review by Minneci and colleagues). Taking the limitations of subgroup analyses and the prevailing uncertainty about adrenal testing in critically ill patients into consideration, findings of these subgroup analyses do not support withholding CS therapy while waiting for results of adrenal function tests. A widely accepted alternative approach is to test and treat all patients with vasopressor-dependent septic shock (barring contraindications), and to consider weaning from steroids those whose test results (when available) rule out a diagnosis of adrenal insufficiency.

Maureen Meade, MD
Hamilton Health Sciences and McMaster University
Hamilton, Ontario, Canada