Continuous Glucose Monitoring-Based Machine Learning Identification of Diurnal Glycemic Patterns and Diabetes...

Area of Science:

• Endocrinology
• Data Science
• Psychosocial Health

Background:

• Type 2 diabetes (T2D) management requires understanding diurnal glycemic patterns.
• Continuous glucose monitoring (CGM) generates extensive data for pattern analysis.
• Diabetes distress is a significant psychosocial factor impacting T2D self-management.

Purpose of the Study:

• To identify distinct diurnal glycemic patterns in adults with T2D using machine learning on CGM data.
• To investigate the association between identified glycemic patterns and diabetes distress.

Main Methods:

• An observational study involving 137 adults with T2D using blinded CGM for 1657 days.
• Unsupervised machine learning (Gaussian mixture modeling) to identify glycemic profiles.
• Diabetes distress assessed using the Diabetes Distress Scale and analyzed via ANCOVA.

Main Results:

• Four glycemic profiles were identified: suboptimal control with nocturnal hypoglycemia (15.8%), suboptimal control with nocturnal hyperglycemia (27.1%), poorly controlled with prolonged hyperglycemia (21.1%), and well-controlled (36.1%).
• Participants with poorly controlled glycemic profiles (Cluster 3) reported significantly higher diabetes distress compared to those with well-controlled profiles (Cluster 4).
• The differences in distress levels corresponded to clinically meaningful categories.

Conclusions:

• Machine learning applied to CGM data can identify distinct glycemic phenotypes in T2D.
• These glycemic phenotypes are associated with varying levels of psychosocial burden, specifically diabetes distress.
• CGM phenotyping offers potential for precision diabetes care, enabling early identification of at-risk individuals and guiding tailored interventions.