Innovative AI Model Detects Hidden Diabetes Risks Through Glucose Spike Monitoring

Recent advancements in artificial intelligence have led to the development of a novel approach for predicting diabetes risk by analyzing real-time glucose fluctuations. Traditionally, clinicians have relied on HbA1c tests, which measure average blood sugar levels over several months, to diagnose diabetes or prediabetes. However, these tests do not effectively identify individuals at the highest risk of disease progression.

Researchers at the Scripps Research Institute have created an AI-powered model that leverages continuous glucose monitoring (CGM) data alongside genetic, dietary, physical activity, and gut microbiome information. This comprehensive data integration enables the detection of early metabolic signs that often go unnoticed by standard tests. The model specifically assesses how long it takes for blood sugar levels to return to baseline after spikes, the overnight patterns of glucose, and other dynamic factors.

In a pioneering remote clinical trial, over 1,000 participants across the U.S. wore Dexcom G6 CGMs, tracked their meals and exercise, and submitted biological samples over ten days. This innovative approach allowed researchers to collect detailed personal health insights without requiring in-person visits. The data was used to train an AI system that could distinguish individuals with type 2 diabetes from healthy controls.

Key findings revealed that in diabetic individuals, glucose spikes often take more than 100 minutes to normalize. In contrast, healthier individuals typically regain control within a shorter period. Additionally, a diverse gut microbiome, higher physical activity levels, and lower resting heart rates correlated with better glucose regulation. Interestingly, some pre-diabetic participants exhibited metabolic profiles similar to those with full-blown diabetes, even if their lab values appeared normal, highlighting the potential of this model to enable personalized intervention strategies.

The model’s validation on external data sets further supports its potential for widespread clinical application. Future iterations may empower individuals to use CGMs at home for continuous risk assessment, or aid healthcare professionals in tailoring early treatment plans. Ultimately, this technology offers new hope for early detection, prevention, and personalized management of diabetes, aiming to slow or stop its progression before symptoms appear.

This innovative research underscores the importance of analyzing day-to-day glucose dynamics and the potential of AI to transform diabetes care and prevention.

source: https://medicalxpress.com/news/2025-07-ai-glucose-spikes-reveal-hidden.html