Accuracy of blood glucose measurements in the intensive care unit

Introduction

Strict glucose control in critically ill patients may improve survival. According to the Van den Berghe protocol frequent glucose measurements are needed. In the original protocol, however, measurements are made by a blood gas analyzer that is not accessible to our unit. Measurements in the core chemistry laboratory are time and staff consuming and unpractical because of a time delay of about half an hour. According to some other protocols, bedside glucometry is a proposed solution, which is easily accessible and not time consuming. In our preliminary phase of adopting the aforementioned protocol we observed some clinically suspected cases of hypoglycemia, which were not confirmed by the bedside glucometry. However, glucose measurements in the core laboratory confirmed our suspicion. The question arose whether bedside glucose measurements are accurate enough to be performed in so strict a protocol as that of Van der Berghe.

Methods

The blood samples were obtained by arterial line. Capillary blood was taken simultaneously. Analysis was performed by the core chemistry laboratory in arterial plasma as the gold standard (GS). Arterial whole blood and capillary blood samples were analysed by four bedside glocometers: Accu-Chek Active (ACA), Accu-Check Inform (ACI), HemoCue (HC) and Gluco Touch (GT). Results were compared with the GS and presented as the correlation coefficient. Data were expressed as the percent of difference to GS and the precision for each glucometer was expressed as the percent of data in the range of ± 10%. The proportion of overestimated values was calculated for each group and the difference between the groups was evaluated by chi-square test. The average percent of difference to GS for all overestimated values for each group was calculated and expressed as the mean ± SD. The difference between the groups was tested by analysis of variance.

Results

One hundred and forty series of measurements were performed from blood samples of four male and four female critically ill patients. The correlation coefficients and the precision as already defined for each group (a – arterial, c – capillary) were: ACAa – 0.91 and 53%; ACIa – 0.96 and 93%; HCa – 0.91 and 44%; GTa – 0.93 and 13%; ACAc – 0.91 and 2%; ACIc – 0.94 and 91%; HCc – 0.91 and 42%; GTc – 0.85 and 25%. The proportion of overestimated values for ACAa, ACIa and HCa were 27%, 30% and 23%, respectively, and this was significantly less than in the GTa group (98%, P < 0.05). The average percent of difference to the GS of all overestimated values in groups ACAa and GTa were 16 ± 15% and 30 ± 20%, which is significantly greater than in groups HCa and ACIa (4 ± 3% for both; P < 0.05).

Conclusion

In the settings where measurements by blood gas analyzers are not possible and core chemistry laboratory data have an unacceptable time delay, AccuCheck Inform measurements are accurate enough to be performed in strict protocols. HemoCue seems also to be acceptable because of its low probability of overestimating the glucose level. The other two analyzers proved to be inaccurate. The best way to obtain a blood samples is an arterial line; however, one should be aware of physiologic and analytic reasons for discrepancy between results from plasma and whole blood samples. Capillary blood sampling is inaccurate, especially in circumstances such as septic shock.