Reviewed by Lexie CornerJan 8 2025
In a study published in Nature Biomedical Engineering, scientists from Stanford Medicine developed an artificial intelligence-based algorithm capable of identifying three of the four most common subtypes of Type 2 diabetes using data from continuous blood glucose monitors.
Traditionally, diabetes has been classified into two types: Type 1, which typically develops in childhood, and Type 2, which is often associated with obesity and appears later in life. However, researchers have observed that Type 2 diabetes is not uniform, with patients varying in body weight, age at onset, and other characteristics.
It is a tool that people can use to take preventative measures. If the levels trigger a prediabetes warning, for instance, dietary or exercise habits could be adjusted.
Michael Snyder, Professor and Study Co-Lead Author, Stanford Medicine
Snyder is the Stanford W. Ascherman, MD, FACS Professor in Genetics.
According to Snyder, the development of widely accessible technology that identifies diagnostic details could transform diabetes treatment. Approximately 40 million Americans, or 13 % of the population, have been diagnosed with diabetes, and 98 million have prediabetes.
The majority of people with diabetes have Type 2, and they are just called ‘Type 2, but it is more complex than that, and there are different underlying physiologies that lead to the condition.
Tracey McLaughlin, MD, Professor and Study Co-Senior Author, Stanford Medicine
Type 2 diabetes, accounting for 95 % of all diabetes cases, is being further subclassified to better understand the unique physiology underlying each patient’s condition and their risk of related complications, such as cardiovascular, kidney, liver, or eye issues.
McLaughlin said, “This matters, because depending on what type you have, some drugs will work better than others. Our goal was to find a more accessible, on-demand way for people to understand and improve their health.”
Snyder noted that such technology would have been invaluable when he discovered he was prediabetic. “When I found out I was on my way to becoming diabetic, I increased my muscle mass, which is one of the common ways to help decrease sugar in the blood, but it had no effect. That is because I am not traditionally insulin resistant,” he said.
Snyder’s type of Type 2 diabetes stems from beta-cell deficiency, a condition where the cells responsible for producing insulin function improperly.
McLaughlin and Snyder are co-senior authors of the study. Ahmed Metwally, Ph.D., a former postdoctoral scholar at Stanford Medicine and now a Research Scientist at Google, is the lead author.
Delineating Details of Diabetes
Currently, diabetes can be diagnosed through a simple blood draw that measures glucose levels.
McLaughlin said, “But those tests reveal little about the biology underlying high blood sugar. Understanding the physiology behind it requires metabolic tests done in a research setting, but the tests are cumbersome and expensive and not practical for use in the clinic.”
Over-the-counter continuous glucose monitors, however, can detect elevated blood sugar levels and provide more detailed data about underlying metabolic biology.
Insulin, a hormone produced by the pancreas, regulates blood glucose levels by promoting the absorption of glucose by cells for energy. Blood glucose levels increase when the pancreas produces insufficient insulin, a condition called insulin deficiency. Similarly, when cells fail to respond to insulin signals, a condition known as insulin resistance develops, also leading to elevated glucose levels. Insulin resistance is a hallmark of diabetes.
Insulin resistance in the liver or dysfunction in the production of incretin—a hormone secreted by the gut after meals that triggers insulin release from the pancreas—can also result in Type 2 diabetes. Each of these four physiologic subtypes of diabetes may respond to different treatments.
Testing the Algorithm
McLaughlin and Snyder investigated whether data with hidden signals related to these subtypes could be extracted from a widely available device like a continuous glucose monitor.
These monitors, worn on the upper arm, track blood sugar levels in real time. Blood glucose typically rises after consuming glucose drinks, but the intensity and pattern of these spikes vary among individuals.
In a study involving 54 participants, 21 with prediabetes and 33 healthy individuals, researchers used an artificial intelligence algorithm to analyze patterns in blood sugar peaks and dips and correlate them with subtypes of Type 2 diabetes.
Participants underwent an oral glucose tolerance test in a clinical setting and simultaneously used continuous glucose monitors.
People have looked at that for decades and have found certain parameters that indicate insulin resistance or beta cell dysfunction, which are the main drivers of diabetes. But now we have the monitors, and you can get a much more nuanced picture of the glucose pattern, which predicts these subtypes with greater accuracy and can be done at home.
Tracey McLaughlin, MD, Professor and Study Co-Senior Author, Stanford Medicine
The AI-powered algorithm identified metabolic subtypes, such as insulin resistance and beta-cell deficiency, with greater accuracy than traditional metabolic tests when compared to clinical data and biomarkers. The tool correctly predicted subtypes approximately 90 % of the time.
Broadening Accessibility
Other benefits of using the monitor include higher-resolution data for individuals with diabetes or prediabetes.
McLaughlin said, “Even if a person with insulin resistance does not develop diabetes, it is still important to know, because insulin resistance is a risk factor for a variety of other health conditions, like heart disease and fatty liver disease.”
McLaughlin and Snyder hope that the widespread availability of the technology will increase access to care, even in cases where patients are unable to attend a doctor's appointment. They intend to continue testing the algorithm with individuals diagnosed with Type 2 diabetes.
“We also see this technology as a valuable health care tool for people who are economically challenged or geographically isolated and cannot access a health care system,” McLaughlin said.
The work was supported by the Department of Genetics and the Department of Medicine at Stanford.
The National Institutes of Health, the Stanford PHIND Center, the Stanford Diabetes Research Center, the Wellcome Trust, Stanford Lifestyle Medicine, and the American Diabetes Association funded the study.
Journal Reference:
Metwally, A. A., et al. (2024) Prediction of metabolic subphenotypes of type 2 diabetes via continuous glucose monitoring and machine learning. Nature Biomedical Engineering. doi.org/10.1038/s41551-024-01311-6.