Algorithm

The algorithm acts as the “brain” of the CLINICIP system: it uses glucose measurements, previously administered insulin rates, and carbohydrate content of parenteral and enteral nutrition to calculate subject-specific time-varying insulin needs.

The key components developed within the project include the model-predictive control (MPC) algorithm representing the core of the control software and the graphical user interface (GUI) to allow interaction between clinical staff and the MPC.

An innovative step was the development of a simulation environment providing a physiological model of the glucose insulin metabolism. This metabolic simulator was used for tuning the control software and predicting outcomes of clinical testing. A synthetic population of 56 critically ill patients (29 patients treated at surgical and 27 patients treated at medical ICUs) included in the simulation environment was developed from clinical data obtained during clinical testing of the MPC at the four ICUs involved in the project.

Real clinical trial outcomes were compared with the simulated studies to demonstrate the validity of the simulation environment and its synthetic population. Very similar results were obtained (see Fig.5), supporting the use of computer simulation, or “in silico” testing, in medical device development to reduce the demand for time and resource intensive clinical trials.

Fig.5: Blood glucose profiles (mean±SD) of the critically ill synthetic population (N=56) controlled by MPC during real clinical trials (blue circle) and during simulated experiments (red circle). Black dashed lines represent target blood glucose range (4.4-6.1 mmol/L)