Background: We verified the analytical performance of strip-based handheld glucose meters (GM) for prescription use, in a comparative split-sample protocol using blood gas samples from a surgical intensive care unit (ICU).
Methods: Freestyle Precision Pro (Abbott), StatStrip Connectivity Meter (Nova), ACCU-CHEK Inform II (Roche) were evaluated for recovery/linearity, imprecision/repeatability. The GMs and the ABL90 (Radiometer) blood gas analyzer (BGA) were tested for relative accuracy vs. the comparator hexokinase glucose-6-phosphate-dehydrogenase (HK/G6PDH) assay on a Cobas c702 analyzer (Roche).
Results: Recovery of spiked glucose was linear up to 19.3 mmol/L (347 mg/dL) with a slope of 0.91-0.94 for all GMs. Repeatability estimated by pooling duplicate measurements on samples below (n=9), in (n=51) or above (n=80) the 4.2-5.9 mM (74-106 mg/dL) range were for Freestyle Precision Pro: 4.2%, 4.0%, 3.6%; StatStrip Connectivity Meter: 4.0%, 4.3%, 4.5%; and ACCU-CHEK Inform II: 1.4%, 2.5%, 3.5%. GMs were in agreement with the comparator method. The BGA outperformed the GMs, with a MARD of 3.9% compared to 6.5%, 5.8% and 4.4% for the FreeStyle, StatStrip and ACCU-CHEK, respectively. Zero % of the BGA results deviated more than the FDA 10% criterion as compared to 9.4%, 3.7% and 2.2% for the FreeStyle, StatStrip and ACCU-CHEK, respectively. For all GMs, icodextrin did not interfere. Variation in the putative influence factors hematocrit and O2 tension could not explain observed differences with the comparator method.
Conclusions: GMs quantified blood glucose in whole blood at about the 10% total error criterion, proposed by the FDA for prescription use.