Response to Ligtenberg et al.
We thank Ligtenberg et al. (1) for drawing attention to our study, which demonstrated the efficacy and safety of the model predictive control (MPC) algorithm in controlling glycemia in critically ill postsurgery patients (2). We agree that properly designed studies evaluating different treatment approaches are needed. Our contribution was to execute the first prospective multicenter comparison of insulin titration protocols aiming to achieve tight glucose control. We agree that the best feasible approach for a nurse-led algorithm is to include the last two glucose measurements, the so-called dynamic scale protocol (3), and infuse insulin continuously, but this alone does not constitute “state-of-the-art glucose management protocol.” Extensive variations on the theme exist. Additionally, intensity of educational support for nurses and (dis-) continuous nutritional feeding impact the outcome. While the “gold standard” is formed, our study evaluated existing protocols that have been designed for tight glucose control, that have been operational, and that have received institutional support in their respective intensive care units (ICUs). In agreement with the conclusion of a comprehensive review of the literature by Meijering et al. (3), a dynamic scale protocol (Prague) demonstrated comparable glycemic control, and a sliding-scale protocol (Graz) demonstrated inferior glycemic control compared with the MPC algorithm. However, in terms of glucose levels in the target range, the rather complex dynamic scale protocol (as used in London, U.K.) tended to be inferior when compared with the MPC algorithm; this supports our views that tight glucose control in the ICU is not exclusively dependent on the protocol but also on clinical features such as different methods of nutritional provision and intuitive decision making of the ICU staff. A clear advantage of an automated algorithm is the avoidance of the intuitive decision making, integration of nutritional information, and the continuity of the day-and-night operation. An extension of our work for different study populations and the reduction of the sampling frequency are clearly required. An enhanced version of the MPC algorithm using extended blood sampling is currently being tested in a multicenter trial at medical ICUs. We regret that our conclusion that “the MPC algorithm is safe and effective in controlling glycemia in critically ill postsurgery patients” was interpreted as “computer can beat man.” What we meant was that “computer can help man” to implement tight glycemic control and save numerous lives in intensive care medicine.
Footnotes
- DIABETES CARE