RESUMO
OBJECTIVE: To evaluate and compare the reliability and validity for both normal and high glycaemia values of three models of glycaemia self-analysers: Reflolux, Glucocard and One Touch. DESIGN: Descriptive and crossover study. Validation of a measuring appliance through its comparison with a standard appliance. PARTICIPANTS: The sample was composed of 90 blood specimens chosen by suitability sampling. MEASUREMENTS AND MAIN RESULTS: The glycaemia in the specimens was measured simultaneously by each one of the self-analysers and the central laboratory. There was high reliability (reproducibility) among the self-analysers, with correlation coefficients of 0.98, 0.97 and 0.98, with no statistically significant differences. As to intra-instrument reliability (repeatability), tested by analysing the variances of the differences, no statistically significant differences were found, except in the Glucocard self-analyser which lost precision at high glycaemia levels. Validity or accuracy, measured by the method of taking the average of the differences, showed a systematic bias in the Reflolux model, which had a value of 22.3 mg/dl above the reference model, and in the One Touch of -16.3 mg/dl. The Glucocard was the most accurate with a value of -1.78 mg/dl. CONCLUSIONS: Portable glycaemia dosimeters are very useful in following up diabetic patients because of their high reliability, but are not useful in the diagnosis of diabetes because of their lack of accuracy. Before beginning to use a portable dosimeter, its characteristics in terms of reliability and validity must be known, in order to evaluate correctly its findings.