Within-person modeling of postprandial glucose using multimodal wearable data

Area of Science:

• Metabolic Health
• Wearable Technology
• Personalized Medicine

Background:

• Continuous glucose monitoring (CGM) and wearable sensors generate vast real-world data.
• Existing analytical methods often fail to capture individual variations in daily life.
• A gap exists between sensing capabilities and actionable, individualized insights.

Purpose of the Study:

• To identify within-person, meal-level predictors of postprandial glucose dynamics.
• To analyze the impact of macronutrients, baseline glucose, meal timing, and physical activity on glucose response.
• To bridge the sensing-analysis gap using multimodal data under free-living conditions.

Main Methods:

• Collected multimodal data, including CGM and wrist movement from wearable sensors.
• Analyzed outcomes: peak glucose, time to peak, and area under the glucose curve.
• Employed linear mixed-effects models with within-person centered predictors.

Main Results:

• Net carbohydrate intake strongly correlated with glucose magnitude (p < 0.001).
• Fat intake was associated with delayed glucose response (p = 0.024).
• Baseline glucose impacted all outcomes (p < 0.001); increased wrist movement variability reduced glucose excursions.

Conclusions:

• Individualized postprandial glucose responses can be characterized by integrating multimodal wearable and behavioral data.
• Meal composition and physical activity are key factors in real-world glucose variability.
• This approach offers potential for personalized metabolic health management.