Aim: Near-patient blood glucose monitoring is an essential component of neonatal intensive care but the analysers currently used are unreliable and inaccurate. The aim of this study was to compare a new glucose electrode-based analyser (EML 105) and a non-wipe reflectance photometry method (Advantage) as opposed to a recognized laboratory reference method (Hexokinase). We also investigated the effect of sample route and haematocrit on the accuracy of the glucose readings obtained by each method of analysis.
Methods: Whole blood glucose concentrations ranging from 0 to 3.5 mmol/l were carefully prepared in a laboratory setting and blood samples from each respective solution were then measured by EML 105 and Advantage analysers. The results obtained were then compared with the corresponding plasma glucose reading obtained by the Hexokinase method, using linear regression analysis. An in vivo study was subsequently performed on 103 neonates, over a 1-y period, using capillary and venous whole blood samples. Whole blood glucose concentration was estimated from each sample using both analysers and compared with the corresponding plasma glucose concentration estimated by the Hexokinase method. Venous blood was centrifuged and haematocrit was estimated using standardized curves. The effect of haematocrit on the agreement between whole blood and plasma glucose was investigated, estimating the degree of correlation on a scatterplot of the results and linear regression analysis.
Results: Both the EML 105 and Hexokinase methods were highly accurate, in vitro, with small proportional biases of 2% and 5%, respectively. However, in vivo, both study analysers overestimated neonatal plasma glucose, ranging from at best 0.45 mmol/l (EML 105 venous) to 0.69 mmol/l (EML capillary). There was no significant difference in the agreement of capillary (GD = 0.12, 95% CI, [-0.32,0.08], p = 0.2) or venous samples (GD = 0.05, 95% CI. [0.09, 0.19], p = 0.49) with plasma glucose when analysed by either study method (GD = glucose difference between study analyser and reference method) However, the venous samples analysed by EML 105 estimated plasma glucose significantly better than capillary samples using the same method of analysis (GD = 0.24, 95% CI. [0.09,0.38], p < 0.01). The relationship between haematocrit and the resultant glucose differences was non-linear with correlation coefficients of r = -0.057 (EML 105 capillary), r = 0.145 (EML 105 venous), r = -0.127 (Advantage capillary) and r = -0.275 (Advantage venous). There was no significant difference in the effect of haematocrit on the performance of EML 105 versus Advantage, regardless of the sample route.
Conclusion: Both EML 105 and Advantage overestimated plasma glucose, with no significant difference in the performance of either analyser, regardless of the route of analysis. Agreement with plasma glucose was better for venous samples but this was only statistically significant when EML 105 capillary and venous results were compared. Haematocrit is not a significant confounding factor towards the performance of either EML 105 or Advantage in neonates, regardless of the route of sampling. The margin of overestimation of blood glucose prohibits the recommendation of both EML 105 and Advantage for routine neonatal glucose screening. The consequences include failure accurately to diagnose hypoglycaemia and delays in the instigation of therapeutic measures, both of which may potentially result in an adverse, long-term, neurodevelopmental outcome.