摘要：AbstractGlycaemic control has been shown to improve outcome in critically ill patients, but hard to achieve in a safe and effective manner. This paper presents the preliminary results of 8 patients controlled at the University Hospital of Liège under STAR-Liège, an insulin-only version of the model-based STAR glycaemic controller framework. Clinical data is compared with virtual trial simulations of the glycaemic control outcomes for the STAR-Liège protocol, and with the standard of care protocol of this intensive care unit, to assess safety, performance, and compliance of the new protocol.Results show 78% of clinical blood glucose measurements in target band. Only 3% of blood glucose measurements were below 4.4 mmol/L (79 mg/dL), with only 1% mild hypoglycaemia and no severe hypoglycaemia. These results are similar to simulation of the protocol, but slightly higher workload is observed clinically due to nursing choice. Compared to standard protocol virtual trial simulations, STAR-Liège achieved tighter and less variable control with similar safety, and less percentage time in higher blood glucose levels. Clinically, 14% of insulin intervention were increased or decreased from recommendation with median [IQR] change of 1 [1, 2] or -2 [-3, -2] U/hr respectively.Clinical and simulation results show STAR-Liège better controls glycaemia to lower ranges compared to the standard protocol, while ensuring safety. Lower time in higher blood glucose ranges potentially improves patient outcomes. Compliance analysis shows potential nurse fears in protocol changes and different insulin dosing. These results are encouraging for the continuation of the clinical trial realised in this medical intensive care unit and its extension to insulin and nutrition control.
