Skip to main content

Randomized controlled study for the evaluation of automated blood glucose control in critically ill patients


Numerous guidelines have been developed and tested to implement tight glycaemic control in intensive care units. However, most of these guidelines still require user interventions and/or intuitive decisions of intensive care unit staff and do not allow the establishment of an automated glycaemic control system. To develop Closed Loop Insulin Infusion for Critically Ill Patients (CLINICIP), the objective of the present study was to perform a randomised controlled trial testing a fully automated algorithm for the establishment of tight glycaemic control in critically ill patients and to compare the results with routine treatment of glycaemia.


In a randomized controlled fashion, 20 patients were included into the trial. Patients were investigated after cardiac surgery and for a period of up to 48 hours. Ten patients were treated according to routine care of glycaemia (change of insulin infusion rate according to actual glucose concentration in 1–4 hourly intervals) or using the fully automated model predictive control (MPC) algorithm as implemented on a laptop computer with hourly adjustment of the insulin infusion rate according to information from blood glucose measurements and under consideration of enteral and parenteral carbohydrate infusions. The glucose target range for both treatments was defined between 80 and 110 mg/dl. In all patients, hourly arterial blood gas analysis was performed and intravenous insulin regimen was adjusted according to general routine care or MPC algorithm, respectively. Analysis according to the last value carried forward was applied as not all patients could be investigated for a period of 48 hours.


Glucose concentrations, after start of glycaemic treatment, were comparable between both groups (routine care: 136 ± 35 mg/dl; MPC: 147 ± 32 mg/dl; mean ± standard deviation [SD], not significant). No hypoglycaemic event (blood glucose < 54 mg/dl) occurred in either group of the patients and none of the experiments in the MPC group had to be terminated because of a malfunction of the algorithm. During routine care treatment an average blood glucose concentration of 145 ± 31 mg/dl (mean ± SD) was measured. Average glucose concentration were clearly lower in the MPC group with 109 ± 25 mg/dl (mean ± SD). Confirming this result, 9.8% of the glucose measurements as obtained using routine care compared with 51.6% of the glucose concentrations as obtained using MPC treatment were within the glycaemic target range.


These findings suggest that the software algorithm MPC allows save implementation of tight glycaemic control in postcardiac surgery critically ill patients.

Author information



Rights and permissions

Reprints and Permissions

About this article

Cite this article

Plank, J., Ellmerer, M., Toller, W. et al. Randomized controlled study for the evaluation of automated blood glucose control in critically ill patients. Crit Care 9, P386 (2005).

Download citation


  • Glucose Concentration
  • Routine Care
  • Model Predictive Control
  • Glucose Measurement
  • Tight Glycaemic Control