Comparison of accuracy of three point-of-care glucometers in an adult ICU

Introduction

Obtaining accurate blood glucose levels at the bedside is mandatory to titrate insulin infusions in ICU patients under tight glycemic control. We evaluated concurrently the performance of three point-of-care devices – one blood gas analyzer and two glucometers – in an adult ICU.

Methods

Simultaneously, arterial blood glucose was measured with RapidLab 1265, the Accu-chek Aviva, the Nova StatStrip and in the central laboratory as reference using the hexokinase method. The Bland–Altman approach and a modified Kanji approach [1] were used.

Results

A total of 330 matched analysis were randomly performed in 275 patients. The mean SOFA score was 4.5 (minimum 0; maximum 21). The range of laboratory glucose was 34–526 mg/dl. One patient showed 1,025 mg/dl and was not included in statistical analysis as glucometers all indicated a high out-of-range value. No patient was receiving peritoneal dialysis with icodextrin, and none had a paracetamol overdose. Biases are defined as point-of-care minus laboratory glucose values. These mean biases were -2.9 mg/dl for the RapidLab 1265 blood gas analyzer, -1.2 mg/dl for the Accu-Chek Aviva and -0.3 mg/dl for the Nova StatStrip. The analysis of the 20% discrepancy showed, respectively, zero cases, five cases and one case in the study, while another 22 cases, 40 cases and 19 cases revealed more than 10% discrepancy. See Table 1.

Table 1

Conclusion

The very low biases and the low rate of significant (>20%) discrepancy appear sufficient for safe tight glucose control monitoring in the adult ICU.