I’m thrilled to sit down with Ivan Kairatov, a renowned biopharma expert with extensive experience in research and development, as well as a deep understanding of technology and innovation in the industry. Today, we’re diving into a groundbreaking study on AI and hidden diabetes risk, exploring how cutting-edge tools like continuous glucose monitoring and machine learning can uncover patterns in blood sugar spikes that traditional tests often miss. We’ll discuss the diverse participants involved, the role of lifestyle and gut health in glucose regulation, and what this means for personalized diabetes care.
Can you start by telling us what this study on AI and diabetes risk is all about and what inspired the focus on glucose spikes?
Thanks for having me. This study is really about looking beyond the standard ways we diagnose diabetes. We wanted to understand the subtle, day-to-day fluctuations in blood sugar—specifically glucose spikes—and how they might signal early risks, even when traditional tests like HbA1c show normal results. The inspiration came from realizing that millions of people might be at risk for prediabetes or type 2 diabetes without knowing it, simply because current methods don’t capture the full picture of glucose regulation. By using AI, we aimed to create personalized risk profiles that could catch these hidden patterns early.
How does this approach differ from the conventional methods used to screen for diabetes?
Conventional methods, like fasting glucose or HbA1c tests, give a snapshot of blood sugar levels over a specific period. They’re useful but miss the dynamic, minute-by-minute changes that happen after meals or during sleep. Our approach uses continuous glucose monitoring, or CGM, combined with AI to analyze real-time data alongside factors like diet, activity, and even gut health. This lets us see individual variability in a way that static tests can’t, offering a much more nuanced view of someone’s risk for diabetes.
Let’s talk about the people who took part in this research. Can you share a bit about who they were and why diversity mattered in this study?
Absolutely. We had over 1,100 participants in the main cohort from a nationwide, remote trial in the U.S., with a final analysis group of 347 individuals ranging from normoglycemic to prediabetic to those with type 2 diabetes. What’s really significant is that nearly half came from historically underrepresented groups in biomedical research. Diversity mattered because diabetes risk and glucose patterns can vary across ethnicities, lifestyles, and socioeconomic backgrounds. Including a broad range of people helps ensure our findings and AI models are relevant to everyone, not just a narrow segment of the population.
I’m intrigued by the technology behind this. Can you walk me through how continuous glucose monitoring played a role in gathering data?
Sure. Continuous glucose monitoring, or CGM, involves a small wearable sensor that tracks blood sugar levels every few minutes throughout the day and night. In this study, participants wore these devices for 10 days, giving us a detailed, real-time look at their glucose fluctuations—like spikes after meals or drops during sleep. Unlike a single blood test, CGM captures the full rhythm of someone’s blood sugar, which was crucial for spotting patterns and differences across groups. It’s a game-changer for understanding individual responses.
Beyond CGM, you also looked at things like diet, sleep, and gut bacteria. How did you collect such a wide range of information?
We took a multimodal approach, meaning we gathered data from multiple sources to get a holistic view. Participants used a food logging app to track their meals, wore fitness trackers for sleep and activity data, and provided samples like stool, blood, and saliva for microbiome and genetic analysis. We even accessed their electronic health records with consent. It was all done remotely, which made it accessible but also required clear instructions and support to ensure accuracy. Combining these layers of data helped us see how lifestyle and biology interact with glucose control.
One of the fascinating findings was about gut bacteria. Can you explain how the gut microbiome connects to blood sugar regulation?
Absolutely, the gut microbiome is a hot topic right now. We found that people with less diverse gut bacteria—measured by something called the Shannon index—tended to have higher average glucose levels and poorer control over spikes. The gut plays a big role in metabolism, including how we process sugars and carbs. A less diverse microbiome might struggle to support healthy digestion and inflammation control, which can worsen blood sugar stability. This connection suggests that nurturing gut health could be a key strategy for managing or preventing diabetes risk.
You also uncovered some interesting insights about lifestyle factors. How do things like diet and exercise influence glucose spikes?
Lifestyle has a huge impact. We saw that higher physical activity was linked to better glucose patterns—fewer spikes and quicker recovery. Diet was more complex. For instance, eating more carbs often led to faster resolution of spikes, meaning blood sugar returned to normal quicker, but it also triggered more frequent and intense spikes in the first place. It’s a double-edged sword. This shows how personalized dietary advice could help—there’s no one-size-fits-all when it comes to managing blood sugar through food and movement.
What surprised you most about the differences in glucose patterns across people with normoglycemia, prediabetes, and type 2 diabetes?
One of the biggest surprises was how prediabetic individuals often showed glucose patterns much closer to normoglycemic folks than to those with type 2 diabetes, especially in terms of spike frequency and intensity. I expected prediabetes to look more like a midpoint between the two, but for many metrics, it leaned toward normal. Another striking finding was nocturnal hypoglycemia—low blood sugar at night—in type 2 diabetes patients. It’s a hidden issue that can be dangerous, and seeing how much longer their spike resolution times were, often over 20 minutes slower, really highlighted the physiological burden of the disease.
Looking ahead, what’s your forecast for the role of AI and multimodal data in transforming diabetes prevention and care?
I’m incredibly optimistic. AI and multimodal data have the potential to revolutionize diabetes care by moving us from a reactive to a proactive model. Imagine a future where wearable tech and personalized algorithms can warn someone of a high-risk glucose pattern before it becomes a problem, or tailor diet and lifestyle plans based on their unique biology, including their gut microbiome and genetics. We’re not there yet—more validation and long-term studies are needed—but I believe within the next decade, this kind of precision medicine will become a cornerstone of preventing and managing diabetes, making care more equitable and effective for everyone.
