In a study published this week in Diabetes Care, researchers combined an insulin pump, a continuous glucose monitoring receiver and a smart phone with cloud-based algorithms to automatically monitor and adjust a patient’s glucose levels.
The team pointed towards a number of systems that are designed to function as an artificial pancreas, but still require that users input their carbohydrate estimates when they need insulin after meals. The goal for a number of companies and for people within the diabetes community is to create a system that requires minimal or no action from the user.
The study describes the researchers’ attempt to build an adaptive algorithm that improves over time, targeting an accepted glucose level range.
The team enrolled 30 people with Type I diabetes and watched their hemoglobin A1c levels over 12 weeks as they tested their automated system.
Frank Doyle, professor and dean of Harvard’s engineering and applied sciences school, said that they had a “disciplined group” of people with good overall HbA1c levels before the trial and were able to bring the group level even lower after the 12-week study period.
The researchers also observed that less than 10% of insulin delivery recommendations generated by the algorithm were overriden throughout the study and that the greatest changes were made early on in the study, highlighting the algorithm’s ability to improve over time.