Accuracy of AccuChek glucose measurement in intensive care patients

Objective: To evaluate the accuracy of the AccuChek Inform point-of-care glucose measurement device as compared with central laboratory glucose measurement. Design. Prospective, observational study. Setting: A ten-bed mixed closed format intensive care unit in a 500-bed general hospital. The unit has a computerized insulin protocol aiming for 81 to 135 mg/dL. Patients: All intensive care unit patients were eligible. Interventions. None. Measurements and Main Results: Paired samples (AccuChek glucose in whole blood calibrated to give whole blood results and central laboratory glucose in serum) were taken simultaneously. In 32 critically ill patients, we obtained the following information: mean +/- standard deviation age 71.6 +/- 11.9 yrs; mean Acute Physiology and Chronic Health Evaluation 11 score at admission 17.8 +/- 6.7; 239 paired samples were taken from arterial catheters. Mean AccuChek whole blood glucose was 126 36 mg/dL (7.0 +/- 2.0 mmol/L); mean central laboratory serum glucose was 137 +/- 38 mg/dL (7.6 +/- 2.1 mmol/L). Mean difference was 11 mg/dL (0.61 mmol/L) (8%) (95% Confidence Interval = 9-13 mg/dL, p < .001). ISO 15197 guideline requires 95% of point-of-care measurements to be within 15 mg/dL margins with reference <75 mg/dL or within 20% if reference is higher. In total, 216 (90.4%) of AccuChek measurements were within ISO 15197 margins. Because AccuChek was calibrated to give whole blood results, we calculated a correction factor of 1.086 from the two mean values to correct whole blood AccuChek into serum-like results. This is almost the same as the correction factor of 1.080 given by Roche Diagnostics. By multiplying AccuChek whole blood results with 1.086, 225 (94.1%) of results were within the ISO 15197 margins. Hematocrit did not influence AccuChek results in the 0.20 to 0.44 range. Beyond this range, there were not enough data to draw conclusions. Conclusions. In critically ill patients, the accuracy of AccuChek glucose measurement calibrated to give serum-like results with blood samples derived from arterial catheters is acceptable but falls short by about 1% of complying with the ISO 15197 guideline. (Crit Care Med 2009; 37:2691-2696)