Enhancing Continuous Glucose Monitoring: The Role of AI in Supporting Ambulatory Glucose Profile Interpretation

Artificial intelligence (AI) was a powerful adjunct that increasingly influenced continuous glucose monitoring (CGM) data interpretation. However, the consensus was that it would augment rather than completely replace the ambulatory glucose profile (AGP) shortly. The AGP’s strength was its simplicity and standardization, which made it a trusted common language for providers and patients. Any AI-driven model had to demonstrate clear advantages and reliability to displace that role.

In a 2023 article, ¹ Bergenstal and colleagues argued that AI could accurately assess glycemic control as CGM use grew. They believed that AI analysis of CGM profiles might better gauge an individual’s glycemic status and progression risk, such as transitioning from prediabetes to diabetes. This reflected confidence that AI could integrate various factors (CGM patterns, clinical data) for more detailed assessments than current static metrics. Bergenstal emphasized that exploring CGM data with AI and other health information (genomic, metabolomic) could enhance precision diabetes care. Experts viewed AI as crucial for obtaining deeper insights from CGM data, such as detecting early metabolic decline or enabling personalized therapy, roles that basic AGP could not fulfill independently.

Experts did not suggest that AGP disappear overnight. Its role as a standardized report was well-established and allowed clinicians to understand a patient’s profile quickly. Some commentary highlighted that AGP and metrics like time in range had enhanced CGM data review quality. Replacing it would require extensive validation of AI tools and user training. Many saw a hybrid approach: AI insights layered onto standard reports. A future AGP report might have included AI-generated narratives or alerts and preserved familiarity while leveraging AI benefits. Initial studies with generative pre-trained transformer (GPT-4) showed that clinicians found AI summaries helpful and safe. This suggests that AI could effectively aid interpretation, especially for those struggling with graphs or busy clinicians needing quick insights synopsis.

Diabetes technology experts emphasized clinical validation and oversight. An AI decision-support tool was meant to be seen as “one of many expert opinions” rather than a sole authority. Diabetes management was complex and patient-specific, so AI, while capable of identifying trends, worked best alongside clinician judgment. ² If clinicians disagreed on an insulin adjustment, AI could serve as a tiebreaker or discussion starter. Patients often preferred a human provider to explain AI findings. Experts believed that AI/machine learning (ML) would enhance CGM data interpretation, making it more efficient and personalized, but it would complement rather than replace the AGP framework. The AGP might have evolved to include metrics like “time in tight range” for pregnancy and other targets. ³ AI aided clinicians in prioritizing key data insights for patients.