The performance of flash glucose monitoring in critically ill patients with diabetes.

OBJECTIVE Frequent glucose monitoring may improve glycaemic control in critically ill patients with diabetes. We aimed to assess the accuracy of a novel subcutaneous flash glucose monitor (FreeStyle Libre [Abbott Diabetes Care]) in these patients. METHODS We applied the FreeStyle Libre sensor to the upper arm of eight patients with diabetes in the intensive care unit and obtained hourly flash glucose measurements. Duplicate recordings were obtained to assess test-retest reliability. The reference glucose level was measured in arterial or capillary blood. We determined numerical accuracy using Bland- Altman methods, the mean absolute relative difference (MARD) and whether the International Organization for Standardization (ISO) and Clinical and Laboratory Standards Institute Point of Care Testing (CLSI POCT) criteria were met. Clarke error grid (CEG) and surveillance error grid (SEG) analyses were used to determine clinical accuracy. RESULTS We compared 484 duplicate flash glucose measurements and observed a Pearson correlation coefficient of 0.97 and a coefficient of repeatability of 1.6 mmol/L. We studied 185 flash readings paired with arterial glucose levels, and 89 paired with capillary glucose levels. Using the arterial glucose level as the reference, we found a mean bias of 1.4 mmol/L (limits of agreement, -1.7 to 4.5 mmol/L). The MARD was 14% (95% CI, 12%-16%) and the proportion of measurements meeting ISO and CLSI POCT criteria was 64.3% and 56.8%, respectively. The proportions of values within a low-risk zone on CEG and SEG analyses were 97.8% and 99.5%, respectively. Using capillary glucose levels as the reference, we found that numerical and clinical accuracy were lower. CONCLUSIONS The subcutaneous FreeStyle Libre blood glucose measurement system showed high test-retest reliability and acceptable accuracy when compared with arterial blood glucose measurement in critically ill patients with diabetes.