Lowering of glucose in critical care: a randomized pilot trial

Similar to cardiac surgery patients, medical-surgical critically ill patients may benefit from intensive insulin therapy. The objectives of this pilot trial were to evaluate the feasibility of a randomized trial of intensive insulin therapy with respect to (a) achieving target glucose values in the 2 ranges of 5 to 7 and 8 to 10 mmol/L and (b) uncovering problems with the protocol in anticipation of a larger trial.

The trial was conducted in a 15-bed medical-surgical university-affiliated intensive care unit (ICU).

We included patients older than 18 years, expected to be in ICU for more than 72 hours, with a glucose value of more than 10 mmol/L within 48 hours of ICU admission. Exclusion criteria were diabetic ketoacidosis, severe hepatic failure or hepatic resection, pancreatitis, glucose of less than 2.2 mmol/L on admission to hospital, insulin infusion on admission to ICU, planned withdrawal of life support, and inability to obtain informed consent. Patients underwent concealed random allocation to a target glucose range of 5 to 7 or 8 to 10 mmol/L using pretested algorithms of insulin infusions. Dedicated glucometer measurement of arterial glucose values was calibrated daily to values measured in the laboratory.

We enrolled 20 patients with a mean (SD) Acute Physiology and Chronic Health Evaluation (APACHE) II score of 32 (10.2); 14 were insulin-dependent pre-ICU, and all were medical admissions. Mean glucose values were different in the 2 groups (7.1 +/- 2.6 vs 9.4 +/- 2.1 mmol/L, P < .001). Although the intensive insulin therapy group had more glucose measurements performed than the control group, a similar proportion of values were within the target range (682 [42.4%] of 1607 values in the 5- to 7-mmol/L range; 250 [38.7%] of 660 values in the 8- to 10-mmol/L range, P = .35). Glucose values of less than 2.5 mmol/L developed 7 times in 5 patients, 4 of whom were in the intensive insulin therapy group; however, no adverse consequences were documented. As expected, there were no differences in clinically important outcomes.

In this pilot trial of ICU patients with high illness severity, glucose values were in the 2 target ranges only 40% of the time, using well-accepted initiation and maintenance insulin infusion algorithms. A large randomized trial of glycemic control is feasible in this population to examine clinically important outcomes, but will require refined insulin algorithms and more comprehensive behavior change strategies to achieve target values.