Use of arterial blood with bedside glucose reflectance meters in an intensive care unit: are they accurate?

OBJECTIVE: To compare blood glucose values obtained from two different sampling sites (arterial catheter and capillary from finger puncture), which were analyzed by a bedside reflectance meter. A sample was also analyzed by standard methods (oxygen electrode oxidation in the laboratory).
DESIGN: Prospective, cross-sectional clinical study.
SETTING: Cardiovascular intensive care unit (ICU) designed for postoperative open-heart surgery patients in a 1,100-bed medical center. PATIENTS: Sequential sample of 50 patients immediately after open-heart surgery.
MEASUREMENTS AND MAIN RESULTS: The blood glucose concentration of each patient was analyzed on the patient's arrival to the ICU (immediately postoperatively) by three methods: one blood specimen was obtained from an arterial catheter, divided and analyzed either at the bedside by a reflectance meter (glucose method 1) or in the hospital laboratory (glucose method 2); another sample was obtained by lancing the fingertip and the glucose concentration was analyzed at the bedside in the same reflectance meter (glucose method 3). Using paired analyses to compare the mean glucose values of the bedside arterial whole blood sample (method 1) with the arterial serum sample (method 2) demonstrated that the glucose concentration in the arterial whole blood sample (method 1) was significantly (p < .001) higher. For 46 of 50 comparisons, the glucose value in the arterial whole blood sample (method 1) was higher, with a mean difference of 30 mg/dL (1.7 mmol/L). Although the mean difference was reduced to 10 mg/dL (0.6 mmol/L) when the arterial whole blood sample (method 1) was corrected for the hematocrit (i.e., < 35% [< 0.35]), the mean glucose concentration in the arterial whole blood samples (method 1) remained statistically higher (p < .05). The glucose concentration in the arterial serum sample (method 2) was significantly higher than the value determined from the bedside capillary sample (method 3) before (p < .05) and after (p < .001) correction for hematocrit. The difference in mean glucose concentrations between the arterial serum sample (method 2) and bedside capillary sample (method 3) was 9 mg/dL (0.5 mmol/L) when the capillary specimen (method 3) was not corrected for hematocrit. This difference increased to 21 mg/dL (1.2 mmol/L) when low hematocrit values were considered and appropriate adjustments of the glucose values were made. At the bedside, one can accurately correct arterial whole blood glucose values to correspond to laboratory values by the following formula: (0.94 x arterial whole blood glucose) + (4.6 x hematocrit) + (-16.5 x [37 degrees C--patient's temperature])--132 = laboratory glucose value.
CONCLUSIONS: Since arterial whole blood samples give higher glucose results than arterial serum, the use of arterial whole blood in combination with reflectance meters must be recommended with caution. This caution is especially advised if the glucose values obtained with arterial whole blood are used in conjunction with a sliding scale of insulin, which depends on threshold concentrations of glucose. In our hospital, use of arterial whole blood in combination with reflectance meters could have resulted in an incorrect dose of insulin in 31 of 50 patients.