Corticosteroid Treatment and Intensive Insulin Therapy for Septic Shock in Adults

Annane. JAMA 2010; 303(4): 341-348. doi:10.1001/jama.2010.2

Clinical Question

  • In patients with septic shock who are treated with corticosteroids, does tight glycaemic control (4.44–6.1 mmol/l) compared with less-tight glycaemic control (<8.3 mmol/l) reduce in-hospital mortality?


  • Septic shock is a characterised by a dysregulated host response to an infection including hypotension that is not corrected by fluid administration alone
  • The administration of corticosteroids (hydrocortisone in particular) is thought to attenuate this response and evidence suggests that it reduces the degree of hypotension although it may not ultimately prevent death
  • A consequence of corticosteroid administration is dysregulation of glucose control, which may be harmful
  • It is possible that intensive insulin therapy, with the aim of tight serum glucose control within normal physiological parameters, may counteract some of the harm from corticosteroids allowing them to be more beneficial in patients with septic shock
  • This trial also investigated the effect of additional fludrocortisone as a secondary outcome as the interaction between glucocorticoids and mineralocorticoids had not been adequately studied in patients with septic shock


  • Randomised, controlled trial
  • Multicentre
  • 2×2 factorial design generating 4 groups of patients
    • Arm 1: intensive insulin therapy vs conventional glucose control
    • Arm 2: fludrocortisone vs no-fludrocortisone
  • Open-label
    • Patients, clinical staff and research staff were aware of allocation
    • No attempt to blind participants or staff
    • No placebo therapy was used
  • Centralised web-based randomisation process with permutation blocks of variable size to conceal allocation prior to recruitment and randomisation
  • Hydrocortisone and other non-study therapies were standardised and controlled in keeping with 2004 Surviving Sepsis Campaign guidance
  • Intention to treat analysis
  • Power calculation based only on Arm 1: intensive insulin therapy vs conventional glucose control
    • Estimated 50% in-hospital mortality in control group
    • Expected absolute reduction of 12.5% (25% relative reduction in mortality)
    • Statistics based upon α = 0.05 and Power (1-β) = 80%


  • 11 Intensive Care Units across France
  • January 2006 to January 2009


  • Inclusion:
    • Adults who met the criteria (ACCP/SCCM 1992) for severe sepsis
    • SOFA score ≥ 8
    • Requirement for vasopressors / inotropes to maintain systolic BP > 90 mmHg or mean BP > 60 mmHg
    • Receiving hydrocortisone 50mg IV every 6 hours
  • Exclusion: pregnancy; moribund patients expected to die imminently
  • 946 patients screened; 509 were included and all were included in primary outcome analysis; 25 were lost to follow-up after hospital discharge (i.e. lost for some secondary outcomes)
  • Groups were well balanced for measured baseline variables
Variable Arm 1: Intensive Arm 1: Conventional Arm 2: With Fludrocortisone Arm 2: Without Fludrocortisone
Age  64 years  64 years  64 years  64 years
SOFA score  10.4 10.8  10.6  10.1
Type of admission 88.5% medical  85.9% medical 89.9% medical  84.9% medical
Blood glucose at recruitment  12.0 mmol/l  11.3 mmol/l  11.8 mmol/l 11.5 mmol/l
Table of baseline characteristics


  • Arm 1: Intensive Insulin Therapy
    • Adapted from the Van den Berghe trial
    • Insulin was initiated when blood glucose > 6.1 mmol/l (110 mg/dl)
    • Commenced at 2 units/hr
    • Blood glucose levels were checked at least hourly for first 3 hours and less frequently thereafter if stable
    • Target range was 4.44–6.1 mmol/l (80–110 mg/dl)
    • Hypoglycaemia (< 4.44 mmol/l [80 mg/dl] ) was treated with 10g glucose
    • Insulin infusions were stopped on discharge from Intensive Care unless the patient presented with insulin dependence
  • Arm 2: Addition of fludrocortisone
    • Hydrocortisone was continued
    • 9-α-fludrocortisone 50 µg was administered via nasogastric tube each morning at 8 am for 7 days


  • Arm 1: Conventional glucose control
    • Physicians were advised to follow the 2004 Surviving Sepsis Campaign guidelines and not to follow the Intensive Insulin Therapy trial protocol
      • “maintenance of blood glucose <150 mg/dL [8.3 mmol/l] after initial stabilization”
    • Treatment dose and route of insulin was at the treating physician’s discretion
  • Arm 2: No addition of fludrocortisone
    • Hydrocortisone was continued
    • Fludrocortisone was not administered
    • No placebo was administered either

Management common to both groups

  • Physicians were advised to follow the 2004 Surviving Sepsis Campaign guidelines for all non-trial therapies
  • Hydrocortisone was administered as 50 mg intravenously every 6 hours to all patients for 7 days from recruitment


  • Good separation between groups was achieved, with statistically significant lower blood glucose observed in the intervention group compared to the control group
  • Primary outcome: There was no statistically significant difference in the in-hospital mortality between the Intensive Insulin Therapy and the Conventional Glucose Control
    • Intensive Insulin Therapy: in-hospital mortality 45.9%
    • Conventional Glucose Control: in hospital mortality 42.9%
    • Absolute Risk Increase: 2.97% (95% CI -5.66% to 11.60%; P = 0.53)
    • Relative Risk: 1.07 (95% CI 0.88 to 1.30)
    • Hazard Ratio (using time to death analysis and not just ‘dead/alive’): 1.04 (95% CI 0.8 to 1.34; P = 0.78)
  • Secondary outcome:
    • There was no evidence of interaction with fludrocortisone (P = 0.31)
      • In-hospital mortality (with vs without): 42.9% vs 45.8%
      • Absolute Risk Reduction: 2.98% (95% CI -5.66% to 11.61%)
    • No pre-defined subgroup analysis identified any benefit from intensive insulin therapy or fludrocortisone administration
    • Median length of stay in ICU and hospital was not different
    • Duration of vasopressor therapy and mechanical ventilation was not different
    • The incidence of hypoglycaemia was significantly higher in the Intensive Insulin Therapy group
      • Intensive Insulin Therapy: 16.4%
      • Conventional Glucose Control: 7.8%
      • Absolute Risk Increase: 8.6% (95% CI 2.97% to 14.23%)
      • Number-needed-to-harm: 12

Authors’ Conclusions

  • This trial provides no evidence to support intensive insulin therapy (blood glucose target 80 to 110 mg/dl [4.44 to 6.1 mmol/l] ) for patients with septic shock being treated with corticosteroids


  • The hypothesis is reasonable given the described theory (but note comment in Weaknesses below), and the randomised, controlled trial design is arguably the best method to investigate this hypothesis
  • The multi-centre design increases external generalisation
  • The centralised, web-based randomisation process is the gold standard for concealing group allocation and preventing allocation bias
  • The power calculation was based on previous data, although in retrospect it was over optimistic to expect the intervention to reduce mortality by 25%
  • The statistical methods and intention-to-treat analysis plan were appropriate for the data types
  • Attempts were made to standardise non-study care according to published guidelines, which may reduce any ascertainment bias due to the lack of blinding, although the compliance with this was not reported
  • Good separation was demonstrated between the groups with regards to blood glucose levels, increasing the internal validity of the results
  • The primary outcome is objective and binary, reducing observer bias and simplifying assessment and analysis
  • The small confidence intervals around the null-effect point suggest that a true benefit or harm is unlikely, which strengthens the conclusion of this trial


  • The hypothesis tested tight glucose control with a target of 4.4 to 6.1 mmol/l against less-tight glucose control with a target of < 8.3 mmol/l, which are arguably not that dissimilar and perhaps the theory has not been adequately tested with this trial
    • Compare this with NICE-SUGAR‘s relaxed glucose control with a target of 8.0 to 10.0 mmol/l
  • The open-label design may introduce significant biases as clinical staff may unconsciously treat the patients differently depending upon their group allocation, which usually exaggerates the effect in favour of the intervention group
  • The time-frame for the primary outcome – in-hospital events – introduces the possibility of distortion of the results due to ‘clipping’
    • If patients get home but still die more often then this significant outcome will be missed
    • 30-day mortality is frequently used for a fairer assessment of outcome but this requires more trial resource to follow-up discharged patients

The Bottom Line

  • This trial provides evidence against Intensive Insulin Therapy in patients administered hydrocortisone for septic shock, and furthermore there is no benefit from additional fludrocortisone
  • The narrow confidence intervals suggest that any true benefit is small and this must be balanced against the significant incidence of hypoglycaemia (NNH 12)
  • This adds to the theory regarding steroids in sepsis – if they are of benefit, counteracting the possibly harmful hyperglycaemia with tight glycaemic control does not make any difference to short-term patient outcomes compared to less-tight glycaemic control

External Links


Summary author: Duncan Chambler
Summary date: 3 January 2018
Peer-review editor: David Slessor

Leave a Reply