Precision-to-tolerance capability: an important consideration in tight glucose control

To assess clinical implications of bias and variance of point-of-care glucometric measurements in cardiac surgery patients with wide variations in postoperative hematocrit.

Point-of-care glucose measurements were compared with values from laboratory analysis of the same sample of whole blood obtained from cardiac patients early on postoperative days 1 and 2. Twenty nurses collected 89 arterial blood samples from 58 patients during a 4-month period. Bias was measured by using difference scores between paired measurements. Patients were grouped within 5% increments according to hematocrit, and analysis of variance was used to test for differences. Variation was analyzed by precision-to-tolerance analysis within 3 euglycemic tolerance ranges.

Laboratory glucose values were 62 to 224 mg/dL; point-of-care measures were 83 to 253 mg/dL. Bias was 10.85 mg/dL across all hematocrit groups. Pairs of laboratory and point-of-care glucose values differed significantly (t(174) = 10.03; P < .001). Bias increased from -2.83 mg/dL for patients with hematocrits exceeding 39% to +16.71 mg/dL for patients with hematocrits between 20% and 24%. The standard deviation of difference scores was 11.59 mg/dL overall. The difference between 5% hematocrit groups was significant (F(4) = 4.11; P = .004). Precision-to-tolerance capability ratios for specification limits of 70 to 300, 90 to 140, and 80 to 110 mg/dL were 0.30, 1.39, and 2.32, respectively.

The direction of bias change between hematocrit groupings was the direction predicted in the manufacturer's information. Precision-to-tolerance measures indicated that the point-of-care equipment was not suitable for testing glucose within the planned "tighter" glycemic standards.