Accuracy assessment of continuous glucose monitoring by frequent venous blood collection in patients with type 2 diabetes mellitus

Abstract

Objective To assess the accuracy of continuous glucose monitoring system (CGMS) in type 2 diabetic patients by frequent venous blood collection. Methods A total of 11 patients with type 2 diabetes treated from August to October 2010 were enrolled in this study. The Medtronic GOLD™ CGMS was used to monitor the continuous changes of the blood glucose levels for 3 successive days in all subjects, meanwhile the fingertip blood glucose was tested at least 4 times each day for calibration. The venous blood of the subjects was randomly collected every 15 min per hour for 7 successive hours on one day when the subjects wore the sensor, and was tested using the YSI 2300 STAT Plus™ glucose&lactate analyzer. The accuracy was comprehensively evaluated by using statistical methods including the rate of agreement between the sensor readings and the paired YSI values within ±20% and ±30% of deviations, the error grid analysis, absolute relative difference (ARD), Bland-Altman analysis and trend analysis. Correlation analysis of two variables was done with Pearson correlation analysis. Results A total of 319 paired YSI-sensor values were collected from the 11 patients. Total 88.4% of the sensor readings (95%CI: 0.84–0.92) were within ±20% of the YSI values, and 96.9% within ±30% of the YSI values. Clarke error grid analyses showed that the ratios of the YSI-sensor values in region A to region B were 88% and 12%; Consensus error grid analyses showed that the ratios of the YSI-sensor values in region A to region B were 96.2% and 3.8%; Continuous error grid analyses showed that the ratios of the YSI-sensor values in region of accurate reading, benign errors, erroneous reading were 94.4%, 2.8%, 2.8%, respectively. The mean ARD for all subjects was 10.5% and the median ARD was 8.4%. The Bland-Altman analysis detected the mean blood glucose level of 0.47 mmol/L (95%CI: -1.90-3.01 mmol/L). Trend analysis revealed that 82.5% of the difference of the rates of change between the YSI values and sensor readings occurred within the range of 0.06 mmol·L-1·min-1, and only 1.7% of the absolute differences of the paired values were more than 0.17 mmol·L-1·min-1. Conclusions The CGMS has high accuracy both in monitoring the blood glucose at a specific time and reflecting the trend of changes of the blood glucose level. However, actual clinical manifestations should be taken into account for diagnosis of hypoglycemia. Key words: Diabetes mellitus, type 2; Continuous glucose monitoring; Error grid analysis; Accuracy