To develop an adaptive precision framework (ACTIVE-GLU) that characterizes personalized relationships between physical activity (PA) and blood glucose (BG) responses in individuals with type 1 diabetes mellitus (T1DM).
Approach:
Study Design: The study utilizes real-world data from wearable devices and continuous glucose monitoring (CGM) systems to quantify the relationship between varying PA intensities and subsequent BG changes.
Model Development: ACTIVE-GLU employs a rolling-window framework to learn from each participant's historical physiology and behavioral patterns, creating personalized PA-BG profiles.
Data Collection: Data is collected from wearable devices to capture metrics such as step counts and activity intensity, aligned with glucose measurements.
Key Findings:
Non-standard PA often leads to steeper downward BG gradients, indicating increased risk of hypoglycemia.
Existing predictive models primarily focus on structured exercise, limiting their applicability to spontaneous, free-living PA.
ACTIVE-GLU adapts to individual behavioral patterns and physiological responses, improving prediction accuracy.
Interpretation:
The framework aims to enhance understanding of lifestyle factors influencing BG control.
Limitations:
The model's effectiveness is contingent on the quality and granularity of the collected data.
Variability in individual responses to PA and other factors may affect prediction accuracy.
Conclusion:
ACTIVE-GLU represents a significant advancement in personalized diabetes management by integrating real-world PA data with BG monitoring.
