Efficacy and Safety of Three Insulin Protocols in Medical Intensive Care Unit Patients

• Efficacy: mean of patients' median blood glucose during ICU stay [ Time Frame: ICU Stay ]
  
• Safety: incidence of hypoglycemia (≤40 mg/dL)during ICU stay [ Time Frame: ICU stay ]
  

• Death from any cause [ Time Frame: ICU and in-hospital ]
  
• ICU and hospital length of stay [ Time Frame: ICU and in-hospital ]
  
• Five-day variation in SOFA score [ Time Frame: Five days ]
  
• 90-day health status [ Time Frame: 90 days ]
  
• Days on mechanical ventilation
  
• Initiation of dialysis
  
• Days on antibiotics
  
• Days on inotropic of vasopressor support
  
• Necessity of red blood cell transfusions
  
• Bilirubin peak
  
• Platelets through
  

Strict glycemic control has been recommended for critically ill patients. However, its implementation may face difficulties with increased nursing workload, inadequate glucose control and higher risk of hypoglycemia.

We designed a computer guided protocol to adjust endovenous insulin infusion aiming glucose levels between 100mg/dl and 130 mg/dl. This trial evaluates the efficacy and safety of this protocol (algorithm A), compared to a standard endovenous insulin infusion protocol (algorithm B) and a conventional subcutaneous insulin protocol (algorithm C).

Methods : MICU patients with at least one blood glucose ≥ 150 mg/dL and who are on mechanical ventilation, or had SIRS, or are admitted because of trauma or burn will be randomized to one of the following treatments: algorithm A - continuous insulin infusion with adjustments guided by hand held device or desktop software targeting glucose levels between 100mg/dL-130mg/dL; algorithm B - continuous insulin aiming glucose levels between 80mg/dl-110mg/dl using Van den Berghe's insulin protocol; algorithm C - conventional treatment - intermittent subcutaneous administration of insulin if blood glucose levels exceeds 150mg/dL; insulin will be administered as IV boluses in hypotensive patients.

The randomization list was generated in blocks of six by computer software. Patients will be randomly assigned in a 1:1:1 ratio to have their glucose controlled by one of the three insulin algorithms with the use of a central, computerized system accessed by Internet, permitting concealment of allocation list. Randomization will be stratified according to study site.

The study is planned to enroll 165 patients in order to have 80% power to detect a 20mg/dl difference in blood glucose means between groups, assuming standard deviation equal to 33 mg/dl and two-tailed alpha equal to 0.05. Efficacy will be measured by the mean of patients' median blood glucose and safety measured by the incidence of hypoglycemia (≤40 mg/dL). Analysis will follow the intention-to-treat principle.