Digital twin technology improves diabetes control with artificial pancreas

New technology that allows a University of Virginia-developed artificial pancreas system to adapt to users' changing bodies – and lets users test changes to how the system operates – improved control of their type 1 diabetes, a study has found.

The Adaptive Biobehavioral Control (ABC) technology optimizes the automated insulin delivery system in the artificial pancreas every two weeks while giving users access to a "digital twin" computer simulation to test different approaches to managing their blood sugar. The six-month study found the participants using the ABC technology boosted time spent in a safe blood-sugar range from 72% to 77% and reduced their hemoglobin A1c (average blood-sugar level) from 6.8% to 6.6%.

Artificial pancreas systems require adjustments by those who use them to adapt to a person's changing insulin demands. This is the first study that maps each person to their 'digital twin' in the cloud and enables people with diabetes to experiment with their own data to learn how their artificial pancreas system would react to changes, in a safe simulation environment, before adjusting their system." 

Boris Kovatchev, PhD, Director of the UVA Center for Diabetes Technology

A 'digital twin' for diabetes control

While automated insulin delivery systems like the artificial pancreas have succeeded at helping users better control their type 1 diabetes, the ABC technology is designed to address a pair of unresolved challenges. 

The first is improving blood-sugar control during the day, when more events occur – such as meals and exercise – that cause blood-sugar fluctuations. Second, most users see an initial improvement in the time spent in a safe blood-sugar range but quickly reach a plateau of 70% to 75% of time in range, which the researchers theorize stems from users not adapting well to how the system performs.

The ABC technology is designed to address those challenges in two ways by using "digital twins," a computer simulation that matches users' metabolic system. In addition to optimizing the artificial pancreas based on changing physiology and behavior, it also provides users with an interactive computer simulation where they can test changes to the parameters of how their artificial pancreas system performs, such as adjusting the continuous level of insulin released by their insulin pump overnight.

"Human-machine co-adaptation is critical for conditions like type 1 diabetes where treatment decisions are made both by the artificial pancreas algorithm and the person who wears it. Digital-twin technology is very helpful in facilitating this co-adaptation," Kovatchev said. 

Results published

The researchers have published their findings in the journal npj Digital Medicine. The team consisted of Kovatchev, Patricio Colmegna, Jacopo Pavan, Jenny L. Diaz Castañeda, Maria F. Villa-Tamayo, Chaitanya L. K. Koravi, Giulio Santini, Carlene Alix, Meaghan Stumpf and Sue A. Brown. The study was funded by the National Institute of Diabetes and Digestive and Kidney Diseases, grant RO1 DK085623. A full list of the authors' disclosures is included in the paper.