In critical illness hypo-and hyperglycemia have a negative influence on patient outcome. Continuous glucose monitoring (CGM) could help in early detection of hypo-and hyperglycemia. A requirement for these new methods is an acceptable accuracy and precision in clinical practice. In this pilot study we prospectively evaluated the accuracy and precision of two CGM sensors (subcutaneous sensor: Sentrino®, Medtronic and intravasal sensor: Glucoclear®, Edwards) in 20 patients on a cardio-surgical ICU in a head to head comparison. CGM data were recorded for up to 48 h and values were compared with blood-gas-analysis (BGA) values, analysed with Bland-Altman-plots and color-coded surveillance error-grids. Shown are means ± standard deviations. In total 270/255 intravasal/subcutaneous pairs with BGA-values were analysed. The average runtime of the sensors was 28.4 ± 6.4 h. Correlation with BGA values yielded a correlation coefficient of 0.76 (subcutaneous sensor) and 0.92 (intravasal sensor). The Bland Altman Plots revealed an accuracy of 2.5 mg/dl, and a precision of + 43.0 mg/dl to - 38.0 mg/dl (subcutaneous sensor) and an accuracy of - 6.0 mg/dl, and a precision of + 12.4 mg/dl to - 24.4 mg/dl (intravasal sensor). No severe hypoglycemic event, defined as BG level below 40 mg/dl, occurred during treatment. Both sensors showed good accuracy in comparison to the BGA values, however they differ regarding precision, which in case of the subcutaneous sensor is considerable high.
