We thank Drs. Liang and Rice for their insightful comments on our study.1 There is no widely used or accepted reference method for blood glucose; therefore, the reference method used is a potentially confounding variable in studies of glucose meter accuracy. Perhaps the best choice for any study would be the predicate method for the device being studied, which for the Nova StatStrip (Nova Biomedical Corporation, USA) would be the plasma hexokinase method. The choice of reference method, however, needs to be weighed against other logistical aspects of study design. Specifically, cellular glycolysis occurring in the reference sample increases glucose meter bias as a function of time between sample draw and analysis.2 We chose a study design that allowed us to analyze reference samples within 10 min of blood draw (a practice used in studies intended for U.S. Food and Drug Administration review). However, this required using the whole blood glucose oxidase method on a Radiometer ABL90 (Radiometer America Incorporated, USA) in a laboratory located adjacent to the operating room as the reference method. This method was shown to have no systematic bias compared to a plasma hexokinase method, very good precision (CV 2.1%), and be unaffected by common sources of interference such as hematocrit, pH, and PO2.3 In our practice, we have observed that Radiometer ABL90 glucose is interchangeable with Roche (USA) plasma hexokinase glucose.
The narrow range of glucose values observed is a limitation of our study, as we pointed out in the Discussion section.1 Some studies have found poor accuracy of glucose meters at very low and very high values, though these tend to be studies of older glucose meter technologies. Like many institutions, we have moved to more moderate glucose target ranges and adapted protocols to proactively prevent hyperglycemia while minimizing hypoglycemia, and rates of hypoglycemia are very low at our institution. We suspect that this will be an ongoing “problem” in studying glucose meter accuracy with current best practices. Manufacturing low and high glucose samples is one option but not applicable to our study comparing capillary to arterial sampling.
We agree with the third point entirely. Although many studies have reported glucose meter accuracies in the intensive care unit, very few intraoperative studies have been performed. The fact that in our study the bias and outlier rates did not differ significantly between capillary and arterial samples in the operating room suggests that there may be something different about the operating room patient population versus that of the intensive care unit. One possibility is that the vasodilatory effects of general anesthesia result in more accurate capillary sampling in this environment, but other explanations are certainly possible. We hope other groups will continue investigating the accuracy of capillary and arterial glucose meter testing in different critical care environments.
Supported by the Department of Anesthesiology and Perioperative Medicine and Department of Laboratory Medicine and Pathology, Mayo Clinic, Rochester, Minnesota.
Dr. Karon has received travel support from Nova Biomedical Corporation, Waltham, Massachusetts. The other authors declare no competing interests.