What CGM Data Gaps Are Really Telling Us

AbstractThe aim was to investigate the association between continuous glucose monitoring (CGM) data coverage and glycemic metrics. This study included over 97,000 clinical study participants and real-world data from type 1 or type 2 diabetes treated with multiple daily insulin injections, closed-loop systems, or basal-only insulin regimens. Over 35 million days of CGM data were analyzed with multilevel modeling. Low coverage was observed in 6.4%–10.1% of days and was significantly associated with lower time in range (TIR) across sources (P < 0.001). Each 1% increase in coverage was associated with a within-person increase of 0.07%–0.13% in mean daily TIR (P < 0.001). Our analysis shows that higher daily sensor coverage is significantly associated with higher daily TIR, suggesting that missing CGM data may be missing not-at-random. Although low-coverage days are included in TIR calculations, they contribute fewer measurements and may underrepresent periods of poor glycemic control, potentially leading to a systematic overestimation and bias of overall TIR.CitationCichosz, S. (2026, januar 2). Research Unmuted – Episode 2: What CGM Data Gaps Are Really Telling Us. Research Unmuted. https://doi.org/10.5281/zenodo.18170881ReferencesCichosz SL, Hartvig NV, Kronborg T, Hangaard S, Vestergaard P, Jensen MH. Biases in Glucose Metrics Are Directly Related to Low Coverage of Continuous Glucose Monitoring: Insights from Diverse Populations. Diabetes Technol Ther. 2025 Sep 3. doi: 10.1177/15209156251376007. PMID: 40897432.Tags#Diabetes #Wearables #Digital_Health #Data_Science