Blood Glucose Monitoring in the Intensive Care Unit: Toward Defining Bias and Imprecision Thresholds for (Near) Continuous Sensors

G. Van den Berghe
2014 Clinical Chemistry  
Hyperglycemia is a hallmark of critical illness and is firmly associated with adverse outcomes in both adults and children treated in the intensive care unit (ICU). 2 The association between hyperglycemia and adverse outcomes follows a J-shaped curve, with the lowest risk associated with blood glucoses within the reference intervals for fasting [for adults 80 -110 mg/dL (4.4 -6.1 mmol/L)]. To address causality in such an association, an intervention-based randomized controlled trial (RCT) is
more » ... d trial (RCT) is required, in which one group is randomly allocated to receive insulin titrated to blood glucose reference intervals and the other group to a strategy that tolerates hyperglycemia as determined by the disease process. In 2001, the results of the first RCT performed in surgical ICU patients were published, showing that titrating insulin to maintain blood glucose within the reference intervals for fasting during critical illness reduced morbidity and mortality (1 ). Additionally, a subsequent RCT in adults suffering from nonsurgical critical illnesses (2 ) and another RCT performed in critically ill children, for whom the ageadjusted blood glucose target reference intervals were even lower (3, 4 ), showed reduced morbidity and mortality. Subsequent, so-called repeat studies, performed in the more "real-life" multicenter setting, did not confirm these promising results, and one study showed an even higher mortality with strict blood glucose control (5 ). The reasons for these conflicting results remain debated. Even today, more than 10 years after the first RCT was published, the issue of how to best control blood glucose concentrations with insulin infusion in the ICU continues to be one of the most controversial topics in critical care medicine. A large part of this controversy has been attributed to methodological differences among the studies (6 ). The 3 initial RCTs showing that outcome could be improved by lowering the blood glucose concentrations of adult and pediatric ICU patients used a very accurate and precise instrument to measure blood glucose concentrations in undiluted whole blood samples drawn from an arterial line, and only after extensive training allowed nurses to adjust the insulin infusion based on a customized and validated guideline (1, 7 ) . The subsequent studies did not implement such high degrees of standardization. For example, these later studies allowed the use of all sorts of hand-held blood glucose meters, mostly those that use enzyme-impregnated strips that were not designed for this purpose because the total error was too large to allow titration of the treatment to such narrow target ranges for blood glucose (8 ) . Earlier modeling studies (9 ) previously revealed that, with higher total errors such as those of point-ofcare devices used in the multicenter trials (10 ), the risk of undetected hypoglycemia inevitably increased dramatically, the effective prevention of hyperglycemia was reduced, and the glucose variability increased into the potentially dangerous range (11 ). The impact of such high total errors could thus explain why titrating insulin based on readouts generated by such devices may have caused harm that could, in turn, have offset any potential benefit of strict blood glucose control that was achievable with accurate and precise tools (9 ). These early simulation studies identified a total error of Ͻ10% as mandatory for this purpose, a number that was subsequently incorporated into guidelines for point-of-care blood glucose testing in acute settings (12 ). In the meantime, the critical care community had come to realize that measuring blood glucose concentrations intermittently at a high enough frequency to allow safe blood glucose control was costly and very time-consuming for the ICU nursing staff and, even then, left large windows of time during which no information about the actual blood glucose concentrations was given to the medical team. This insight created a demand for a more continuous display of blood glucose concentrations in the ICU setting. The medical community assumed that devices that could measure blood glucose in a (near)continuous way, with a dis-
doi:10.1373/clinchem.2013.220715 pmid:24491272 fatcat:ibu5bvk7p5e5bcxbwq4rlik6xy