Evaluating methods for calibrating continuous glucose monitor data using fasting capillary and venous glucose in individuals without diabetes mellitus
OA Version
Citation
Abstract
BACKGROUND: Accurate calibration of continuous glucose monitoring (CGM) systems is critical for reliable glucose monitoring. This study evaluated the differences between interstitial fluid glucose levels measured by CGM systems and venous and capillary blood glucose (BG) levels during fasting in individuals without diabetes. Additionally, the study assesses the accuracy of various CGM calibration methods, comparing single venous BG calibration, single capillary BG calibration, and multi-capillary BG calibration.
METHODS: This study was conducted at the Institut für Diabetes Technologie, Forschungs- und Entwicklungsgesellschaft mbH an der Universität Ulm, Germany, and included 34 participants without diabetes mellitus. Each participant wore two FreeStyle Libre CGM systems for 14 days, with glucose data collected every 15 minutes. Capillary glucose measurements were taken using the CONTOUR NEXT ONE system. Calibration methods were evaluated using mean absolute relative difference (MARD), Bland-Altman plots, mean differences, and limits of agreement (LOA). Correlations between mean glucose levels derived from each calibration method and HbA1c were also analyzed using multi-capillary BG calibration as the gold standard to determine the reliability of calibration methods.
RESULTS: In the fasting condition, the MARD for uncalibrated CGM glucose vs. venous or capillary BG was 10.37% and 10.48% respectively, and improved to 10.08% and 10.31% after averaging two CGM devices worn by the same individual simultaneously. Accuracy for using the individual uncalibrated CGM glucose (MARD = 7.88%) was not improved by single venous or capillary BG calibration when compared to multi-capillary BG calibration. However, MARD improved to from 7.88% to 6.39% after averaging two uncalibrated CGM devices worn by the same individual simultaneously. Mean glucose derived from uncalibrated (average of two) CGM did not show a positive correlation with HbA1c (r = -0.14); whereas mean glucose derived from multi-capillary BG calibration was weakly correlated with HbA1c (r = 0.30). Although MARD was not improved by calibrating CGM with a single capillary or venous BG measurement, these simple calibrations did result in CGM mean glucose that was weakly correlated with HbA1c (r = 0.13-0.26, depending on methodology). CONCLUSION: This study highlighted the differences in glucose measurements between interstitial, venous, and capillary compartments. Our results did not suggest that calibrating CGM tracings to a single venous or single capillary blood glucose would improve accuracy of individual CGM measurements, but did lead to improved CGM mean glucose correlation with HbA1c. We also observed improvement in accuracy when participants wore two CGM devices. This observation has implications for newer CGM technology in development using microneedles that takes advantage of multiple glucose sensors on the same device.
Description
2025