Association of Corticosteroid Treatment With Outcomes in Adult Patients With Sepsis: A Systematic Review and Meta-analysis

Fang F, Zhang Y and Tang J et al.

JAMA Internal Medicine. 2019;179(2):213-223. doi:10.1001/jamainternmed.2018.5849


  • In-hospital mortality after sepsis develops ranges from 30-45%
  • Corticosteroids have been used as an adjunct therapy in sepsis since the mid 20th century
  • Evidence remains conflicting as dose, duration, type or overall effect on mortality
  • 2 large RCTs were published in 2018 with conflicting results (ADRENAL and APROCHSS trials) and varying methodology


  • Sepsis is a systemic inflammatory response to infection
  • Standard treatment of sepsis: fluid resus, broad spec abx and inotropic/vasopressor support
  • Glucocorticoids act via 3 mechanisms in sepsis:
    • Inhibition of nitric oxide synthase in vascular smooth muscle cells –> increased contractility and reduction of hypotension
    • Inhibition of phospholipase A2 –> prevention of synthesis of prostaglandin and other inflammatory mediators
    • Direct prevention of release of lysozymes and superoxides from neutrophils
  • Overall effect is to dampen systemic inflammatory response and therefore decrease end organ damage
  • The theory is this should lead to improved patient outcomes; however at present this is not clearly borne out in evidence to date

Primary outcome

Measure the effect of adjunctive glucocorticoid therapy on 28-day mortality vs standard therapy alone or placebo

Secondary outcomes

  • In-hospital mortality; ICU mortality; 90-day mortality
  • Shock reversal day 7; SOFA score
  • ICU/Hospital length of stay
  • Patient reported QoL
  • Vasopressor/ventilation free days to day 28;
  • Rate of adverse events (GI bleed, superinfections, hyperglycaemia, hypernatraemia)

Study Design and Methodology

  • Systematic Review and Meta-Analysis
  • n = 9564 from 37 RCTs between 1968-2018
  • Population: Adults (>18y) diagnosed with sepsis, severe sepsis or septic shock
  • Intervention: any type of corticosteroid (hydrocortisone, prednisolone, methylprednisolone, dexamethasone, betamethasone)
  • Control: either placebo or standard supportive therapy
  • Exclusion:
    • Case reports, case series, observational studies
    • Topical or inhaled corticosteroids
    • All patients received corticosteroids
  • 2 researchers independently selected papers and discussed any discrepancies with third researcher (K statistic generated)
    • 9939 articles/records identified and screened, eventually 37 RCTs selected based on eligibility criteria
  • Data Analysis: Two independent investigators extracted data from the included RCTs into standardized collection forms and created tables for the evidence and outcomes
  • Statistics:
    • Dichotomous variables assessed using Mantel-Haenszel method and expressed as risk ratio (RR)
    • Continuous variables assessed inverse variance random-effects model and were expressed as mean differences (MD)
    • 2 -tailed P value with significance set at 0.05
    • Heterogeneity – X 2and I2 (value of >50% significant)
  • Bias:
    • Publication bias assessed with funnel plot and statistical analysis (Egger/Begg/Harbord tests)
    • All RCTs assessed using Cochrane GRADE tool – categorises studies into high, medium or low risk for bias
  • Power Calculation: trial sequential analysis performed using O’Brien Fleming approach
  • Sensitivity Analysis: conducted for primary outcome only for variables including (but not limited to) trials with small numbers of patients, those published before 2000 and those with non low risk of bias
  • Subgroup Analyses:
    • Dose of corticosteroids
    • Treatment duration
    • Sepsis population subtype
    • Type of corticosteroids
    • Disease severity
    • Year of publication


  • Primary outcome: 28-day mortality significantly lower with steroids (26.3% intervention vs 29.2% control)
    • RR (0.90; 95% CI, 0.82-0.98; I2 = 27%)
  • Secondary outcomes:
    • No difference in 90 day mortality (3 trials: RR, 0.94; 95% CI, 0.85-1.03, I2 = 27%)
    • However in-hospital (RR, 0.88; 95% CI, 0.79-0.99)and ICU mortality (RR, 0.85; 95% CI, 0.77-0.94, I2 = 0%) significantly better with steroids
    • Increased reversal of shock and lower SOFA score at day 7
    • Decreased length of ICU stay but no effect on hospital stay
    • Increase number of vasopressor free days but not ventilator free days
    • No trial reported on QOL
  • Adverse events:
    • Any adverse event – no difference
    • GI bleed – no difference
    • Superadded infection – no difference
    • Hyperglycaemia – significantly higher in intervention group (RR, 1.19; 95% CI, 1.08-1.30)
    • Hypernatraemia – significantly higher in intervention group (RR, 1.57; 95% CI, 1.24-1.99)
  • Subgroup analyses:
    • No single subgroup had any significant effect on degree of heterogeneity (including mortality of control group)
    • Long course (>4 days) significantly better 28-day mortality than short course (<4 days)
    • Low dose (<400mg/d) significantly better 28-day mortality than high dose
  • Heterogeneity:
    • I2 value for the primary outcome = 27% indicating a moderate degree of heterogeneity
    • Values for secondary outcomes range from 0 – 68%, indicating a significant impact of heterogeneity on the conclusions drawn
  • Publication bias: funnel plot showed a significant degree of publication bias again impacting on results and conclusions


The authors note that while they found a statistically significant association between the use of adjunctive corticosteroids and reduction in 28 day mortality, this was not reflected in 90-day mortality. Whilst the number of patients used to reach these conclusions differ (9564 vs 5238), this would seem to suggest a short term benefit only for the use of steroids in sepsis. Nevertheless the results do show a trend towards favouring their useage; the question remains how best to maximise their value for patient outcomes. The data presented here support the use of a low-dose (<100mg/d) long course (>4d) regime; however by their own admission the authors note that a credible association with mortality is not found in this study.

It could be argued that 28-day mortality has questionable clinical value as a measure of patient outcomes. Perhaps a more valuable outcome would be longer term mortality, as at present there is a sense that steroids merely delay the inevitable, especially given the lack of reduction in 90 day mortality. However this study strongly supports the idea that steroids in an ICU setting speed recovery from organ dysfunction as evidenced by the statistically significant results for shock reversal, SOFA score reduction, time to ICU discharge, ICU mortality and vasopressor free days.

This review and meta-analysis represents the most comprehensive assessment of the available evidence to date. Through the incorporation of the 2018 APPROCHSS and ADRENAL trials (which together provided more patients than all previous RCTs combined) this is the first meta-analsysis on the subject sufficiently powered to show a statistically significant reduction in mortality. Unfortunately, the degree of publication bias and heterogeneity identified is substantial and must be considered when drawing any conclusions from this work. In particular the authors themselves note a wide variety in the reporting of adverse events and thus advise caution when interpreting their data.


  • Large sample size – >9000 patients from 37 RCTs; previous meta-analyses reached a maximum of approx. 4400 patients
  • Powered to reach statistically significant conclusions unlike previous meta-analyses on the subject
  • Rigorous study design and methodology
  • Sources of potential bias eliminated where possible, or identified where not possible
  • Includes data on a wide number of secondary outcomes


  • Subject to publication bias
  • Substantial degree of heterogeneity
  • Value of primary outcome in a clinical setting?
  • Subgroup analyses/secondary outcomes underpowered to reach meaningful conclusions
  • No patient reported outcomes
  • Does not examine delay between onset of sepsis and commencement of steroids – could this be a key factor in improving outcomes?
  • Includes RCTs from 1968 – 2018 – over this time management of sepsis has evolved to such a degree that comparison of studies 50 years apart is confounded by differences in standard supportive treatment

Future Work

  • VICTAS trial ongoing (Vitamin C, Thiamine and Steroids in Sepsis)
  • Suggestion of benefit of ‘personalized medicine’ approach – e.g. metabolomics, gene studies to identify those individuals who would most benefit from steroids


  • Most comprehensive meta-analysis on the subject shows significant evidence of short term benefit but inconclusive long term value – steroids may buy some time
  • Low-dose (<100mg/d) long-course (>4d) appears to be best regime and merits further study
  • Acceptable complications profile but further analysis required due to large variation in reporting
  • Steroids in sepsis is not a one size fits all approach – in future may benefit from ‘personalized’ approach
  • I feel there is now sufficient evidence to support routine administration of steroids in an ITU setting for patients with sepsis, although at present this may only be a temporary measure

Summary by Dr T Woodhead. Journal Club Meeting 06 February 2020.

Sepsis image from:



Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

Blog at

Up ↑

%d bloggers like this: