New study reveals hidden diabetes risk beyond A1C
Researchers used continuous glucose monitoring and other health data to create personalized diabetes risk profiles that reveal risk differences even within identical A1C values
The Study
Published: Nature Medicine, January 2025 Where: Scripps Research Translational Institute, La Jolla, CA
The Takeaway
Two people can have identical A1C values but vastly different diabetes risk profiles when you look at the complete picture. This study used AI to analyze continuous glucose monitoring data alongside genetics, gut microbiome, diet, sleep, and activity patterns to create personalized diabetes risk scores. The breakthrough finding: people with the same A1C can have dramatically different risk levels for developing Type 2 diabetes, revealing that our current one-size-fits-all approach to diabetes screening misses critical individual differences.
What It Looked At
Researchers from Scripps conducted a comprehensive digital health study called PROGRESS, initially enrolling 1,137 people across the United States. After applying quality criteria for data completeness, they analyzed 347 participants---174 with normal glucose control, 79 with prediabetes, and 94 with Type 2 diabetes. What made this study unique was its multimodal approach: participants wore continuous glucose monitors for 10 days while also providing data on their diet, sleep, physical activity, genetics, and gut microbiome.
The team analyzed six key glucose spike patterns:
- Average glucose levels throughout the day
- How long it takes glucose to return to baseline after meals
- Time spent with elevated blood sugar
- Overnight low blood sugar episodes
- Daily number of glucose spikes
- Peak glucose levels relative to baseline
Using this wealth of data, they trained an AI model to distinguish between people with normal glucose control and those with Type 2 diabetes, then applied this model to create personalized risk profiles.
What It Found
Different diabetes states show distinct glucose spike patterns. People with Type 2 diabetes had longer glucose spike recovery times and higher overnight low blood sugar compared to those with normal glucose or prediabetes. Interestingly, people with prediabetes fell somewhere in between but were more similar to those with normal glucose than to those with diabetes.
Gut health matters significantly for glucose control. The study found strong associations between gut microbiome diversity and glucose spike metrics. People with more diverse gut bacteria had better glucose control---specifically, those with higher gut diversity had lower average glucose levels and spent less time with elevated blood sugar.
The AI model reveals risk differences within identical A1C values. Here's a notable finding: when researchers applied their multimodal AI model to people with identical A1C values, they found substantial variability in diabetes risk profiles. Two people with the same A1C of 5.7% (the lower threshold for prediabetes) could have risk scores ranging from very low to very high based on their glucose patterns, genetics, lifestyle, and other factors.
Lifestyle factors show measurable impact. Physical activity was negatively correlated with almost all glucose spike metrics, meaning more active people had better glucose control. Conversely, higher carbohydrate intake was associated with more glucose spikes but, surprisingly, faster spike resolution.
Why It Matters
This research offers an interesting perspective on diabetes risk assessment. Currently, doctors rely primarily on A1C tests---a blood test that measures average blood sugar over 2-3 months---to assess diabetes risk and management. But this study shows that two people with identical A1C values can have quite different glucose dynamics and risk profiles.
This suggests that continuous glucose monitoring data combined with lifestyle and biological factors could potentially provide more nuanced risk assessments, though much more research would be needed before such approaches could be implemented clinically.
The study also reinforces the importance of gut health, physical activity, and sleep in glucose control---factors that don't show up in traditional diabetes screening but appear to influence metabolic health.
The research was conducted using a decentralized, digital approach where participants collected all data from home using wearable devices and self-collection kits. This demonstrates the feasibility of comprehensive health monitoring outside traditional clinical settings, though questions remain about the scalability and clinical applicability of such intensive monitoring.
Limitations and Future Research
However, the study has several important limitations. The analysis included only 347 people, which is relatively small for developing predictive models. The researchers used different continuous glucose monitors in their validation cohort, which could affect the reliability of the findings. Additionally, the study was cross-sectional, meaning it captured data at a single point in time rather than tracking people over years to see who actually develops Type 2 diabetes.
Before this type of multimodal risk assessment could be used clinically, researchers would need to conduct much larger, long-term studies to prove that these detailed risk profiles actually predict future diabetes better than current methods. They'd also need to address practical questions about cost, accessibility, and which specific combinations of factors are most important for risk prediction.
Take control of your health with expert guidance
Levels pairs real-time glucose data and comprehensive lab testing with clinician analysis and personalized support—everything you need to turn insights into real health improvements. Click here to get started with Levels.
