摘要：AbstractContinuous glucose monitoring (CGM) devices can measure blood glucose levels through interstitial measurements almost continuously (1-5min sampling period). However, they are not as accurate as glucose readings from blood measurements. The relation between tissue and blood glucose is dynamic and the sensor signal can degrade over time. In addition, CGM readings contains high frequency noise and can drift between measurements. However, maintaining continuous glucose monitoring has the potential to improve the level of glycemic control achieved and reduce nurse workload. For this purpose, a simple model was designed and tested to see the effect of inherent CGM error on the insulin therapy protocol, STAR (Stochastic TARgeted).An error model was generated from 9 patients that had one Guardian Real-Time CGM device (Medtronic Minimed, Northridge, CA, USA) inserted into their abdomen as part of an observation trial assesing the accuracy of CGM measurements compared to a blood gas analyser and glucometer readings. A resulting error model was then used to simulate the outcomes if the STAR protocol was guided by CGM values on 183 virtual patients. CGM alarms for hyper- and hypo-glycaemic region were included to improve patient safety acting as 'guardrails'. The STAR CGM protocol gave good performance and reduced workload by ~50%, reducing the number of measurements per day per patient from 13 to 7. The number of hypoglycaemic events increased compared to the current STAR from 0.03% <2.2mmol/L to 0.32%. However, in comparison to other published protocols it is still a very low level of hypoglycaemia and less than clinically acceptable value of 5% <4.0mmol/L. More importantly this study shows great promise for the future of CGM and their use in clinic. With the a newer generation of sensors, specifically designed for the ICU, promising less noise and drift suggesting that a reduced nurse workload without compromising safety or performance is with in reach.
