Criteria abstracted from The Users' Guide to Medical Literature, from the Health Information Research Unit and Clinical Epidemiology and Biostatistics, McMaster University

Highlighted lines and questions below provide links to the pertinent description of criteria in The EBM User's Guide, now available at the Canadian Centres for Health Evidence

Article Reviewed:

Please visit the new Evidence Based Journal Club Reviews

Corticosteroids for severe sepsis and septic shock: a systematic review and meta-analysis.

Annane D, Bellissant E, Bollaert PE, Briegel J, Keh D, Kupfer Y.

Br Med J 2004 329(7464):480. [abstract]

Reviewed by Moreshwar Desai MD, Texas Children's Hospital, Baylor College of Medicine, Houston TX

Review posted March 3, 2005

---

I. Are the results of the study valid?

A. Primary questions:

1. Did the overview address a focused clinical question?

Yes. The objective of the meta-analysis was to assess the effects of steroids on the mortality on patients with severe sepsis and septic shock.

The purpose of this meta-analysis was also to include more recent trials which assessed the role of lower doses (< 300mg of hydrocortisone or equivalent a day) and longer durations of treatments (≥ 5 days) which were not included in the two meta-analyses of the randomized trials published during the period 1966-93.

2. Were the criteria used to select articles for inclusion appropriate?

Yes. The criteria used for including the articles in the meta-analysis were appropriate.

1. Randomized or quasi-randomized trials, with or without blinding, on severe sepsis and septic shock in adults or children were included.
2. Authors included data from trials in sepsis, sepsis syndrome or acute respiratory distress syndrome if separate data were available for septic shock.
3. All studies reporting on i.v treatment with any corticosteroid preparation (for eg. cortisone, hydrocortisone methylprednisolone, betamethasone or dexamethasone) were considered.
4. Length of treatment was clearly defined based on the duration, i.e. long (³5 days) or short (< 5 days) and quantity, i.e. low (< 300 mg of hydrocortisone or its equivalent) or high (≥ 300mg).
5. The control group was clearly defined as those who received standard treatment given either alone or with a placebo.

The criteria were sufficiently general so that author or selector bias would not be inadvertently introduced into the selection of studies.

B. Secondary questions:

3. Is it unlikely that important, relevant studies were missed?

Yes, it is unlikely. The selection process was thorough. Sincere attempts were made to identify all relevant studies regardless of language or publication status (published, unpublished, either in press or in progress). Searches were made in the Cochrane infectious disease group's trials register, and the Cochrane central register of controlled trials in the Cochrane Library (issue 3, 2003).

Searches were done using Medline(1966 - Aug 2003); Embase (1974 - Aug 2003) and LILACS (to Aug 2003).

Appropriate terms and text words like sepsis, septic shock, steroids, corticosteroids, adrenal cortex hormones and glucocorticoids were used to ensure an extensive and a complete search upto August 2003.

The reviewers even checked the reference lists of resulting trials and, when possible, contacted authors to identify any additional published or unpublished data.

4. Was the validity of the included studies appraised?

All the included trials were individually appraised using a methodological scoring system using Cronin et al's "methodologic quality form." The score was generated using eleven discrete criteria, giving a maximum total score of 14.5. The criteria on which scoring was done were 1) Patient selection, 2) Patient characteristics, 3) Randomization, 4) Blinding, 5) Intervention, 6) Contamination, 7) Co-intervention, 8) Explicit description of complications, 9) Withdrawal 10) Intention to treat and adherence to protocol and 11) Explicit definition of sepsis and septic shock. 10 of the 16 selected articles got scores 10 or above, 4 of the 16 got scores between 6 - 9, and 2 of 16 got scores below 6.

Assessment of methodological quality of studies was also done with method recommended by Cochrane Infectious Disease Group. In this method studies were graded as adequate, inadequate or unclear based on generation of allocation sequence and allocation concealment. Methods of blinding were considered as double blind, single blind and open. Loss of followup was also a criterion and graded as adequate, inadequate and unclear.

5. Were assessments of studies reproducible?

Maybe. One reviewer checked all identified titles and abstracts, and three reviewers validated this check. Two pairs of reviewers examined all potential trials, selected eligible trials and graded their methodological quality. Any disagreement was resolved by discussion within the four reviewers.

It is unclear how much variation they had in their rankings because they only state "any disagreement was resolved by discussion." The authors did not report a kappa value for level of agreement.

Personal communication with authors was attempted whenever clarifications were necessary.

6. Were the results similar from study to study?

Results were not similar from study to study. The reviewers calculated weighted treatment effect using fixed effect model across trials. They considered using random effects model in case of heterogeneity. Heterogeneity was determined with chi square test and p ≤ 0.10 was considered significant. Though a sincere attempt was made to identify heterogeneity, the choice of fixed effects model may not have been appropriate and random effects model should have been ideally used across trials.

If effects of steroids on all cause mortality was considered, then the results varied. Out of the 15 trials selected, 8 trials had relative risks of mortality of greater than 1, in favor of control, and 7 trials had relative risk of mortality of less than 1 favoring the treatment group. All trials except Schumer's (1976) trial and the CSG (1963) study had equivocal results with the confidence intervals crossing 1. If the subtotal was considered, then the RR was 0.98 with CI ranging from 0.87 to 1.10 suggesting no difference in all cause mortality in the treatment group compared to the control group. There was however significant heterogeneity across trials (random effects model was used).

In subgroup analysis when effect on mortality was considered with long courses of low dose steroids, all 5 trials seemed to favor treatment, with RR less than 1. The subtotal RR was 0.80 with CI ranging from 0.67 to 0.95. Also there was no heterogeneity in the results. In comparison, when short courses of high dose steroids were considered, the results again suggested no difference between treatment and control groups. There was again significant heterogeneity across the trials (random effects model was used).

The results did not vary when effects of steroids on mortality in ICU in patients with sepsis and septic shock was considered. The relative risks of mortality in all trials was less than 1 favoring treatment. The overall RR was 0.83 with a tight CI ranging from 0.70 to 0.97. There was also no evidence of heterogeneity in the results across all the trials.

Results also did not vary significantly when considering the effects of steroids on shock reversal in patients with severe sepsis and septic shock. When shock reversal at day 7 was considered out of 6 trials all except Bone (1987) trial had RR of higher than 1 favoring treatment. When the subtotal of all 6 trials was considered, the RR was still higher than 1 with CI ranging from 1.06 to 1.40 favoring the treatment group. There was however heterogeneity across the trials (random effects model was used).

The heterogeneity was insignificant when the subgroup of long courses of low dose steroids was considered, and the relative risk and confidence intervals clearly favored the treatment group and the results did not vary between trials.

When the effect of steroids was considered on the outcome of shock reversal at 28 days was considered, there was no heterogeneity and results favored treatment group and the results did not vary across the trials.

II. What are the results?

1. What are the overall results of the review?

Out of 23 trials on steroid use in severe sepsis or septic shock, 16 trials were included by the overview (n=2063).

The overall results can be summarized as follows:

Outcome	Included	RR	95% CI	Heterogeneity?
All cause mortality at 28 days	All trials	0.92	0.75-1.14	Significant
	Long course/low dose	0.80	0.67-0.95	None
	Short course/high dose	0.97	0.72-1.31	Significant
Mortality in ICU	4 trials	0.83	0.70-0.97	None
Mortality in hospital	All trials	0.89	0.71-1.11	Significant
	Long course/low dose	0.83	0.71-0.97	None
	Short course/high dose	0.89	0.57-1.37	Present
Shock reversal at 7 days	All trials	1.43	1.01-2.01	Significant
	Long course/low dose	1.60	1.27-2.03	None
Shock reversal at 28 days	4 trials	1.26	1.04-1.52	Not measured

Note:
RR > 1 favors treatment when outcome of shock reversal in concerned
RR < 1 favors treatment when other outcomes are concerned

Adverse events:

There was no evidence of any increased risk for GI bleeding (10 trials, n=1321; RR=1.16, CI of 0.82 to 1.65 with an insignificant p value of 0.40).

There was no reported increase in risk for superinfection (12 trials, n=1705; RR=0.93 with CI of 0.73 to 1.18 and a insignificant p value of 0.54).

There was also no evidence of an increase in the risk for hyperglycemia (6 trials, n=608; RR of 0.93, with CI of 0.84 to 1.78 and an insignificant p value of 0.30.

2. How precise were the results?

The confidence intervals of each of the results are mentioned above.

When all trials and the subgroup of short courses of high dose steroids were taken there was no significant difference in the outcomes of mortality at 28 days or mortality in hospital. Though all the CI was precise, they all included 1.

Interestingly when trials with long course of low dose steroids were considered and the same outcomes were measured, the 95% CI did not include 1 and the RR was less than 1 favoring treatment. However if one looks more critically at the confidence intervals, they were 0.67-0.95, 0.70-0.97 and 0.71-0.97 for all cause mortality at 28 days, mortality in ICU and in hospital mortality respectively. This means that we can be certain with 95% confidence that the true RR for mortality reduction in the ICU with low dose steroids was somewhere between 0.67 to 0.97. Though precise, if true RR is 0.97, then that would not be very clinically significant risk reduction.

Again when shock reversal is concerned all RR are more than 1 favoring treatment, and none of the CI include 1, the 95% CI are between 1.01-2.01 and 1.04-1.52 for shock reversal at 7 and 28 days respectively. This though precise one could argue that if true RR was 1.01 then that would not be a clinically significant RR.

Clearly more trials have to be conducted and other factors need to be considered (like checking for adrenal insufficiency).

III. Will the results help me in caring for my patients?

1. Can the results be applied to my patient care?

Maybe not.

Only two trials included children. The CSG trial in 1963 and the Slusher et al in 1996.

The CSG trial had a RR of mortality at 28 days and inhospital mortality of 1.53, favoring control. With respect to adverse effects, the Confidence Intervals for GI bleed, super infections and hyperglycemia included 1. However the CSG trial had a very low methodological score of 4.5.

The study done by Slusher in 1996 had a score of 10. The RR of mortality was 1.5 again favoring control, with short courses of high dose steroids. The results were the same of in hospital mortality in patients with severe sepsis and septic shock. The confidence interval however included 1.

It is therefore impossible to say whether steroids would be beneficial or harmful without an RCT that focuses on the question of steroid use in sepsis and septic shock in the pediatric age group. Such a trial is reportedly in progress.

2. Were all clinically important outcomes considered?

Mostly yes. The primary outcome measure was all cause mortality at 28 days. Secondary outcome measures included mortality in the ICU and in hospital.

Number of patients with reversal of shock was also considered at 7 and 28 days, which was a secondary outcome measure.

More importantly the number of patients with adverse events such as superinfection, hyperglycemia and GI bleeding were also considered as secondary outcome measure.

3. Are the benefits worth the harms and costs?

Unfortunately not in pediatric sepsis or septic shock. The meta-analysis did not provide proof of any benefits for the use of steroids (based on the 2 RCTs) in the pediatric population.

Though only one study (CSG 1963) looked at adverse effects for instance GI bleed and superinfection, and though results could be considered statistically equivocal, there was a clear trend towards GI bleeding, with the use of steroids. Risks of giving steroids outweighed the benefits from all trials conducted so far. None of the trials have made a convincing case to use steroids yet.

The authors did allude to the issue of adrenal insufficiency, though the paper did not evaluate it formally. The paper does conclude that there were beneficial results in adults receiving long courses of low dose steroids but did not analyze the subgroup of those who did not respond to corticotrophin stim test.

Comments

Back to the EB Journal Club Index