Abstract
Dear Editor:
Sonmez et al. 1 recently reported observations of inaccuracy for the OneTouch® Select™ blood glucose monitoring system (LifeScan, Milpitas, CA) during a glucose meter evaluation of hypoglycemic blood samples. Although the authors stated that they followed published standards for glucose meter evaluations, our review of their methods identified several important technical discrepancies that may help to explain their observations.
First, the authors used venous blood samples for some (but not all) of the comparisons, even though the labeling for OneTouch Select stipulates that the test strip is intended to measure glucose only in capillary blood. The OneTouch Select system uses an oxygen-sensitive enzyme, glucose oxidase, and is calibrated for the oxygen partial pressure (PO2) that exists in capillary blood. When venous blood is tested with OneTouch Select, a slightly high bias would be expected because of low sample PO2. It is noteworthy that the meters that did relatively poorly in the study (OneTouch Select and EZ Smart) both use glucose oxidase enzymatic methods, whereas the other three meters, which did better, use a glucose dehydrogenase (GDH) enzyme that is not affected by PO2. This choice of the GDH enzyme is a trade-off because the enzyme may be susceptible to other interferences, depending on the GDH co-factor. 2
Second, sodium fluoride was used as a preservative for the venous reference samples, even though this compound has little or no effect on the rate of glycolysis within the first 1–2 h after phlebotomy. 3,4 Because the authors stated that the venous blood samples were “sent to the reference laboratory approximately 30 min after collection” (time of centrifugation and/or time of reference analysis was not specifically stated), one would expect this sample handling to result in relatively lower glucose concentrations in the reference method samples and relatively higher glucose concentrations in the more rapidly tested glucose meter samples. Indeed, an overall positive bias can be seen for most of the meters in Figure 3 in Sonmez et al. 1 This protocol artifact is the reason that the Clinical and Laboratory Standards Institute (CLSI) recommends that blood samples for the reference method be either tested or centrifuged within 5 minutes of the meter test. 5
Finally, although the specific methods are not always clearly stated, the authors may have failed to conform to several other CLSI recommendations, namely, by failing (a) to perform the reference tests in duplicate (before and after each meter test), (b) to verify that these duplicates were within 4% or 4 mg/dL of each other, (c) to test each glucose meter in duplicate, and (d) to compare the mean of the reference method to the individual meter replicates.
In summary, even though the authors claimed that their evaluation protocol complied with published evaluation standards, we identified several discrepancies and possible discrepancies. Although the methods are not always clearly stated, the protocol failed to fully comply with manufacturers' instructions and with certain CLSI recommendations. Because of these discrepancies, we believe the conclusions about glucose meter performance with hypoglycemic blood samples are not supported. Investigators should design studies carefully, following both manufacturers' and CLSI's recommendations, so that they can base conclusions on high-quality results.