233-OR: Occurrence and Impact of Missing Data on CGM Metrics—Analysis of Data from Type 1 Diabetes Exercise Initiative (T1DEXI)

• Published in: Diabetes (New York, N.Y.) Vol. 74; no. Supplement_1; p. 1
• Main Authors: ZHAN, DONGYING, FISHER, SIMON J., ZHANG, XIAOHUA D.
• Format: Journal Article
• Language: English
• Published: New York American Diabetes Association 20.06.2025
• Subjects: Decision making; Diabetes mellitus (insulin dependent); Missing data
• ISSN: 0012-1797; 1939-327X
• DOI: 10.2337/db25-233-OR

Introduction and Objective: Missing CGM data can confound clinical decision making. This study explores the impact of data missingness on CGM metrics from the T1DEXI study. Methods: Study 1: Gaps in CGM data were determined based on time intervals between consecutive readings. Study 2: To evaluate the effect of gap sizes on CGM parameters, a simulation study randomly introduced missing data with gap sizes of 1-10, 11-50, or >50 continuous missing data points while maintaining the same overall missing rate of 20.83% in all groups. CGM glucose metrics including mean, SD, CV, TAR, TIR, and TBR (see Figure B for abbreviations), were compared to a control group with no missing data. Results: In the T1DEXI CGM dataset, gaps of 1-10 account for 15.76% of all missingness but 68.47% of the frequency, gaps of 11-50 account for 41.06% of all missingness and 25.44% of the frequency, and gaps of >50 account for 43.18% of all missingness but only 6.09% of the frequency. The simulation study showed that, compared to smaller missing gaps, larger gaps resulted in higher Mean Absolute Percentage Error (Figure B). Conclusion: Small gaps in CGM data are frequent but cause minimal errors; whereas large data gaps, though less frequent are associated with large errors in CGM metrics particularly in TAR and TBR. These findings highlight the need for accurate prediction of large gaps to improve CGM data interpretation.