Background: Asymptomatic hypoglycemia in neonates may contribute to neurologic deficits during development. Whole-blood glucose sensors are often imprecise and inaccurate at the low glucose concentrations found in neonates.
Subjects and methods: In this study, a glucose sensor using a mutated glucose dehydrogenase that does not cross-react significantly with maltose was evaluated at three pediatric centers. Blood samples (n=575) from infants less than 30 days of age (hematocrit 23-70%) were analyzed using six reagent lots on three ACCU-CHEK(®) meters (Roche Diagnostics, Indianapolis, IN): the Inform II, Performa, and Aviva. Reference glucose level was determined in duplicate in perchloric acid extracts using a coupled hexokinase procedure.
Results: Imprecision of glucose measurement using stable control materials ranged from 2.0% to 3.1% (coefficient of variation) using the glucose meters and from 0.8% to 5.3% (coefficient of variation) in perchloric acid-treated controls. The difference between meter glucose values and reference values showed a slight dependence on hematocrit from 23% to 70% (r=-0.391, P<0.001) but not in the typical range of neonatal hematocrit from 45% to 70% (r=-0.036, P=0.239). Linear regression of the aggregated results yielded the following relationship: Meter glucose=0.99×Reference Glucose+0.04; r(2)=0.976; Syx=0.249. Receiver-operator characteristic analysis of the data using 2.2 mmol/L as the reference threshold for hypoglycemia yielded an area under the curve value of 0.993. All infants with a glucose level of <2.2 mmol/L were detected (100% sensitivity) when the meter glucose value was below 2.8 mmol/L.
Conclusions: These data indicate that the modified ACCU-CHEK chemistry may be used effectively in neonatal settings to detect clinically significant hypoglycemia.