We evaluated the precision and accuracy of monitoring of the blood glucose level in the laboratory and at the bedside with one of four glucose meters by an experienced operator and by 39 nurses in a teaching hospital. For precision studies aqueous quality control materials were used. A total of 85 blood samples were tested. The precision of the glucose meters (expressed as the coefficient of variation [CV]) in the hands of the experienced operator ranged from 6.7% to 11.1%. The correlation between the values obtained by the experienced operator and the reference values obtained in the laboratory was high (0.95 to 0.98). The precision of the values obtained by the nurses using the meters ranged from 13.7% on medical wards to 45.7% in the intensive care unit (ICU). The correlation between these values and those obtained in the central laboratory ranged from 0.72 to 0.82. Twenty-four percent of the glucose values determined on medical wards and 62% of those determined in the ICU deviated from the reference value by at least 20%. Of the 85 patients 12 (14%) would have received different insulin dosages had the reference value been available at the same time as the glucose meter reading: in 3 of the patients the discrepancy was 6 units of insulin or greater. Continuous quality control of bedside monitoring of the blood glucose level is needed. In addition, personnel who use glucose meters should receive adequate training.