Computer-advised insulin infusion in critically ill patients – a randomized controlled trial

Introduction

Tight blood glucose (BG) control has been shown to decrease morbidity and mortality in patients in the surgical ICU [1] but is difficult to achieve using standard insulin infusion protocols. We previously evaluated a software model predictive control (MPC) insulin administration algorithm in postcardiac surgery patients [2]. This study investigated the use of an enhanced MPC algorithm (eMPC) in more severely ill patients over 72 hours.

Methods

Fourteen (seven male) critically ill ventilated medical and surgical patients, mean age 65 years, with an arterial BG > 6.7 mmol/l within 24 hours of ICU admission (RBH) or already receiving insulin infusion, and expected to require mechanical ventilation for more than 72 hours, were treated either with BG control by the standard ICU insulin intravenous infusion protocol [2] or eMPC-advised insulin infusion (n = 6) for 72 hours. The eMPC algorithm, installed on a bedside computer, requires input of current insulin requirements, bodyweight, carbohydrate intake and BG concentration. The algorithm advises the time to next BG sample (up to 4 hours) and the insulin infusion rate, targeted to maintain BG at 4.4–6.1 mmol/l. Patients in the eMPC group had BG measured hourly (for safety) but values were only entered if requested by the algorithm.

Results

The mean (SD) glucose concentration was significantly lower in the eMPC group (6.0 (0.34) vs 7.1 (0.54) mmol/l, P < 0.001). The mean insulin infusion rate was not significantly different (4.1 (2.7) vs 3.1 (1.8) IU/hour, eMPC vs standard care). BG sampling occurred more frequently in the eMPC group, with a mean of every 1.1 vs 1.9 hours (P < 0.05). No patients in either group had any BG measurements <2.2 mmol/l.

Conclusion

The eMPC algorithm was effective in maintaining tight BG control in this more severely ill patient group without any episodes of hypoglycaemia (BG < 2.2 mmol/l), but required more frequent BG measurement.