In the February issue of Diabetes Care, Plank et al. (1) reported the results of their computer-assisted model predictive control (MPC) algorithm versus routine glucose management in 60 postoperative thoracosurgical patients in three different hospitals. We agree that better glycemic control is worth aiming for, but we have some doubts concerning the design of the study and, consequently, the conclusion.
When comparing two protocols, both have to be “state-of-the-art.” In the control group, however, the glucose algorithm protocol in the different ICUs was not standardized, the target blood glucose values were not identical, the insulin was given continuously or as bolus injection, and the frequency of glucose measurements was lower than in the MPC algorithm (once every 3 h versus hourly). It is known from the literature that glycemic control can best be achieved with a protocol using continuous insulin infusion combined with frequent blood glucose measurements and that the last two blood glucose values are used to determine the rate of insulin infusion (2).
In our opinion, before one may conclude that “computer can beat man,” this promising MPC algorithm should 1) be compared with the best available nurse-driven protocol, 2) be tested in a more critically ill patient population, i.e., medical ICU patients, and 3) be studied after an adequate power analysis has been performed.