978

Introduction: Glucose variability (GV) is an independent predictor of ICU and hospital mortality, but the strength of this association is influenced by the mode of assessment. Near-continuous glucose monitoring (CGM) systems can facilitate estimation of GV by providing more frequent blood glucose concentrations. We calculated the correlation between GV indices derived from intermittent readings and those derived from CGM in the same patients over the same period of time Methods: Data from patients recruited for the MANAGE (Manual vs. Automated monitoring Accuracy of Glucose) II trial between July and December 2012 were analyzed. Inclusion criteria were age >18 years, expected ICU length of stay at least 24 hours, and need for a central venous catheter, one lumen of which was connected to an Optiscanner® (Optiscan, Hayward, CA). GV indices (standard deviation [SD], glucose variability index [GVI], glucose lability index [GLI], mean amplitude of glycemic excursions [MAGE], J-index and maximal glucose change [MGC]) were calculated from blood glucose (in mg/dl) measured routinely with a blood gas analyzer and by CGM over the same period. Data are presented as median [interquartile range]. Pearson's correlation coefficient between intermittent and CGM values and the percentage of error ([2.2 x SD of the bias]/mean variability index) were calculated for each metric. Percentage of error defines the degree of interchangeability of two measures Results: Data from 27 patients were analyzed, resulting in a set of 15 [10-16] intermittent and 94 [82-96] CGM-derived measurements per patient. Correlations between values were higher for SD (0.97), J-index (0.94) and GLI (0.93) than for MAGE (0.15), MGC (0.7), and GVI (0.48). The percentage of error between the methods was 50% for SD, 68% for GVI, 113% for MGC, 270% for MAGE, 102% for GLI and 48% for the J-index. Conclusions: Because of the high percentage of error between continuous and intermittent methods, indices need to be adapted when using CGM-derived readings to assess GV. Intermittent readings yielded higher values than CGM for all GV indices, except GVI. SD, J-index seems actually the most reliable indices for evaluating GV in daily practice.