AI Digital Twins Are Helping People Manage Diabetes and Obesity

Based on logged meals, the app predicts a person’s blood sugar response to those foods. It also makes personalized recommendations throughout the day, such as adjusting portion sizes, choosing a different food combination, or taking a walk after eating. Users can accept or ignore these suggestions: maybe broccoli isn’t their favorite food or they prefer to exercise at a particular time of day. The app uses AI to adapt to their preferences over time. Users can also chat with human coaches if they have specific health questions.

For Buckley, Twin Health has helped him make healthier choices, like replacing frozen and prepackaged breakfast sandwiches with homemade breakfast burritos with low-carb, high-fiber wraps. He no longer drinks soda and walks several miles a day.

“When I started the program, I could barely walk a mile before my back and knee hurt. Now I do six and a half miles every morning,” he says.

He likes to receive instant feedback from the app and also track his biometric data over time. He may notice that his body fat percentage and blood pressure are trending down.

“That’s where I get my motivation to keep walking and doing my job,” he says.

Buckley reached his initial goal weight of 300 pounds and now weighs around 275 pounds. After taking blood pressure medication for decades, her doctor recently suggested a lower dose.

When Twin Health contacted the Cleveland Clinic Health Plan about using its program, endocrinologist Kevin Pantalone was initially skeptical. He decided to conduct a study himself.

“We’ve really struggled to implement lifestyle changes very effectively. Patients often need many treatments to control their diabetes,” he said. “So I was definitely very interested.” Despite the age-old advice to simply exercise more and eat healthily, most Americans struggle to get the recommended amount of weekly physical activity and have difficulty sticking to a healthy diet.

Pantalone and colleagues recruited 150 participants with type 2 diabetes, randomly assigning 100 people to the Twin program and the rest to a control group. On average, participants were 58 years old and suffered from obesity and had a blood sugar level, or A1C, of ​​7.2 percent. A level of 6.5 percent or higher indicates diabetes. The goal of the trial was to see if participants could achieve an HbA1c level below 6.5 percent with fewer medications.

After 12 months, 71 percent of study participants using the Twin app reached that blood sugar level with less medication, compared to just 2 percent of people in the control group. People using Twin also lost more weight: 8.6% of their body weight compared to 4.6% in the control group.

At the start of the study, 41 percent of people using Twin were taking a GLP-1 drug, but by the end of the study, only 6 percent were still taking it. In the control group, 52 percent of participants started with a GLP-1, and by the end of the study, this increased to 63 percent. The results were published last year in the New England Journal of Medicine Catalyst.