Unveiling the Limitations of Continuous Glucose Monitoring: Insights from Fasting and OGTT with Advanced Clustering Analyses

Authors

Dwaipayan Saha Dr, The University of Texas at El PasoFollow
Preyangsee Dutta Dr., The University of Texas at El PasoFollow
Jehu Narh Apaflo, The University of Texas at El PasoFollow
Joshua Labadah, The University of Texas at El PasoFollow
Zahra FatahimeiabadiFollow
Ulices VillalobosFollow
Irene John TomyFollow
Sudip Bajpeyi Dr., The University of Texas at El PasoFollow

Abstract

Continuous glucose monitoring (CGM) systems offer real-time glucose insights but are limited by temporal lags and variability compared to capillary blood glucose (CBG) and venous blood measurements. These limitations are particularly pronounced during fasting and oral glucose tolerance tests (OGTT), where accurate glucose tracking is critical. However, gaps remain in understanding how CGM performance compares to traditional methods in capturing glucose trends, variability, and clinical outliers. PURPOSE: This study addresses these gaps by evaluating the accuracy, variability, and clustering efficiency of CGM compared to CBG and venous glucose measurements during fasting and OGTT. METHODS: A cohort of 26 participants underwent fasting and OGTT with glucose levels measured via CGM, CBG, and venous blood at seven time points. Time-series analysis, k-means clustering, and DBSCAN identified trends, outliers, and clustering efficiencies across modalities. RESULTS: During fasting, CGM lagged 21 ± 2 minutes behind venous glucose, underestimated variability by 18% (p = 0.01), and failed to capture 12% of rapid glucose transitions. Venous glucose showed the least variability (SD: 13.13), while CGM exhibited the highest (SD: 44.36). In OGTT, CGM demonstrated a mean lag of 14 ± 3 minutes relative to CBG, with peak glucose discrepancies at 60 minutes (CGM: 192.8 ± 58.71 mg/dL; CBG: 179.17 ± 46.3 mg/dL). Clustering revealed that DBSCAN effectively isolated outliers, identifying extreme glucose deviations (CGM > 300 mg/dL or CBG > 275 mg/dL), while k-means smoothed over these anomalies. CONCLUSION: CGM's temporal lag and variability compromise its clinical accuracy, particularly during rapid glucose fluctuations. While CGM excels in detecting trends, CBG and venous glucose provide more reliable point measurements. Enhanced CGM calibration and the integration of advanced clustering methods are essential to improving its diagnostic utility and clinical reliability.

Recommended Citation

Saha, Dwaipayan Dr; Dutta, Preyangsee Dr.; Apaflo, Jehu Narh; Labadah, Joshua; Fatahimeiabadi, Zahra; Villalobos, Ulices; John Tomy, Irene; and Bajpeyi, Sudip Dr. (2025) "Unveiling the Limitations of Continuous Glucose Monitoring: Insights from Fasting and OGTT with Advanced Clustering Analyses," International Journal of Exercise Science: Conference Proceedings: Vol. 2: Iss. 17, Article 110.
Available at: https://digitalcommons.wku.edu/ijesab/vol2/iss17/110