Analytical performance of glucometers used for routine glucose self-monitoring of diabetic patients.

Background: Glucometry is an essential part of diabetes treatment, but so far, no standard quality control procedure verifying blood glucose meter results is available. In this study, we evaluated the analytical performance of eight glucose meters: GX and Esprit (Bayer Diagn.), MediSense Card Sensor, ExacTech (MediSense) with strips Selfcare (Cambridge Diagn), One Touch Basic, One Touch II, One Touch Profile (Lifescan) and Glucotrend (Boehringer Mannheim/Roche).

Methods: The evaluation included within-run imprecision, linearity, comparison with the laboratory method and calculation of differences between individual glucometers.

Results: Within-run imprecision ranged from 1.5% to 4.5%, linearity assessed as the correlation between measured and calculated glucose concentrations yielded r(2) values from 0.97 to 0.981. Analytical bias of glucose concentration values obtained by the glucometry amounted from 0.14% to 16.9% of values measured by the laboratory method. Bias higher than 5% was found for One Touch Basic, II and Profile meters (however, glucose concentrations in plasma obtained by the laboratory method One Touch meters showed analytical bias from 3.0% to 8.8%). The regression analysis yielded slope values from 0.77 to 1.09 and r(2) values from 0.86 to 0.98. The best correlations with the laboratory method were found for One Touch Basic, II Profile, Glucotrend and Esprit meters. The calculated differences between the individual glucose meters can constitute 0.02-1.49 mmol/l (0.96-26.9%) at glucose concentration 5.55 mmol/l, and 0.16-4.16 mmol/l (0.96-24.96%) at glucose concentration 16.67 mmol/l. Error grid analyses have shown that Glucometers One Touch Basic and One Touch Profile yielded all results in zone A (acceptable). The remaining glucometers yielded 1-7% of results in zones B (insignificant errors), C or D (lack of detection and treatment).

Conclusions: All studied glucometers had both small deviation from laboratory reference values (<10%) and high concurrence with results obtained by the laboratory method.