Accuracy of the continuous glucose monitoring system in inpatient and outpatient conditions.

OBJECTIVE To evaluate in inpatient and outpatient conditions, using only non-calibration data, the accuracy of the Continuous Glucose Monitoring System (CG-MS). RESEARCH DESIGN AND METHODS Twelve, 21 and 20 type 1 diabetic patients participated in 3- day inhospital, 6-day ambulatory (i.e., 2 sensors worn consecutively) and 3-day ambulatory studies. Capillary glucose tests (7-8/day) served as comparisons. Pairs of data (excluding the 4 daily pairs used automatically by the CGMS for calibrations) were analyzed using correlation coefficient, error -grid analysis, and mean of absolute differences. RESULTS Two third and half of the sensing days were qualified as "optimal", i.e., interpretable by the CGMS program in the inpatient and outpatient studies, respectively. Correlation coefficient was 0.92, 0.81 and 0.73 in the inpatient, 6 day outpatient and 3-day outpatient studies. 98%, 93% and 96% of the data fell in the A or B error-grid "clinically acceptable" zones. Mean absolute differences between sensor and glucometer values were 25 +/- 2, 34 +/- 5 and 32 +/- 5 mg/dl. CONCLUSION CGMS may be used for assessment of glycemic trends, though accuracy is more in the range of glucometers than of laboratory tests. Inhospital utilisation improved accuracy and 6-day rather than 3-day sessions may be more appropriate to evaluate day to day reproducibility of glucose profile.