Research on Dynamic Assessment of Glucose‐Lipid Metabolism and Personalized Drug Response Prediction Based on Wearable Multimodal Sensing

This study establishes a dynamic assessment framework for glucose-lipid metabolism using wearable multimodal sensing (optical glucose, EDA, HRV, sleep, energy expenditure) to predict metabolic responses following oral anti-obesity medication (AOMs) treatment. Data from 380 overweight/obese individuals were collected over 24 consecutive weeks to construct behavior-metabolism coupling features. The TFT-Mixture model was applied to predict short-term weight changes and glucose improvement trends (R² = 0.83). SHAP analysis further examined the impact of lifestyle features (evening step count, sleep duration) on drug efficacy variability. This study reveals the contribution of multimodal behavioral phenotypes to AOMs treatment response, providing a technical pathway for intelligent personalized medication management.