Objective: To evaluate the clinical and analytical accuracy of home blood glucose meters.
Research design and methods: Six blood glucose meters--Reflolux S (Boehringer Mannheim, Mannheim, Germany), One Touch II (LifeScan, Milpitas, CA), Glucocard Memory (Menarini, Florence, Italy), Precision QID (Medisense, Cambridge, U.K.), HaemoCue (HaemoCue, Angelholm, Sweden), and Accutrend alpha (Boehringer Mannheim, Mannheim, Germany)--were compared with a reference method (Beckman Glucose Analyzer II) under controlled conditions (glucose clamp technique). Validation of the blood glucose meters was accomplished by clinically oriented approaches (error grid analysis), statistical approaches (variance components analysis), and by the criteria of the American Diabetes Association (ADA), which recommend a target variability of < 5%.
Results: A total of 1,794 blood glucose monitor readings and 299 reference values ranging from 2.2 to 18.2 mmol/l were analyzed (705 readings < 3.89 mmol/l, 839 readings between 3.89 and 9.99 mmol/l, and 250 readings > 9.99 mmol/l). According to error grid analysis, only Reflolux S and Glucocard M had 100% of estimations within the clinically acceptable zones A and B. Assessment of analytical accuracy revealed substantial differences between the glucose meters after separation of the data into defined glycemic ranges. None of the devices met the ADA criteria.
Conclusions: To evaluate accuracy of blood glucose meters, error grid analysis, as well as statistical models, are helpful means and should be performed together. Analytical performance of currently available home blood glucose meters differs substantially within defined glycemic ranges.