Objective: To analyse agreement between two methods for blood glucose measurement in intensive care patients: capillary blood using a reagent strip and glucometer with arterial blood using a blood gas analyser.
Design and setting: Prospective, single-centre, observational study in a 12-bed tertiary referral intensive care unit.
Measurements: Blood glucose levels were measured in consecutive patients using simultaneous measurements of capillary blood samples using glucometry and from a multi-electrode arterial blood gas analyser. An a priori subgroup of patients with tissue hypoperfusion was identified (defined as systolic blood pressure <90 mmHg or vasopressor dependency). A total of 493 paired measurements were obtained; 75 of these were from patients with systemic hypoperfusion.
Results: Overall, the mean difference (bias) was 0.12 mmol/l (2.15 mg/dl) and precision 0.77 mmol/l (13.8 mg/dl); 95% limits of agreement were -0.14 and 1.66 mmol/l (-2.5 and 29.8 mg/dl). In patients with systemic hypoperfusion the bias was 0.24 mmol/l (4.0 mg/dl) and precision 0.9 mmol/l (16.2 mg/dl); 95% limits of agreement -2.05 and 1.58 mmol/l (36.8 and 28.4 mg/dl).
Conclusions: In a general population of intensive care patients, there is statistical agreement between blood glucose measured from capillary blood glucometry and arterial blood gas analysis. However, in patients with systemic hypoperfusion, the accuracy of agreement between these two measurement techniques may be such that that biochemical hypoglycaemia (<2.5 mmol/l, 44.9 mg/dl) may go undetected if used interchangeably.