CREATE: ‘Open-source’ automated insulin delivery gets good results in T1D

Automating insulin delivery with an open-source algorithm that is run on a smartphone proves safe and effective in the management of type 1 diabetes (T1D), as shown in the results of Community Derived Automated Insulin Delivery (CREATE) trial.

Over 6 months, the use of open-source automated insulin delivery (AID) led to a significant 14-percent improvement in the amount of time participants spent in the target blood sugar range, according to principal investigator Dr Martin de Bock, paediatric endocrinologist at the University of Otago, Christchurch, New Zealand, who presented the results at the 82nd American Diabetes Association (ADA) Scientific Sessions.

The mean time in range (TIR) increased from 61.2 percent at baseline to 71.2 percent at week 24. When stratified by age, the mean TIR at the end of the study was 74.5 percent for adults and 67.5 percent for children—an improvement of 9.6 percent and 9.9 percent, respectively, from baseline (p<0.001 for both). The investigators did not detect a treatment effect by age interaction (p=0.56). [ADA 2022, abstract 286-OR]

In comparison, participants who used the sensor augmented pump therapy (SAPT) arm saw no improvement in mean TIR, which decreased slightly from 57.7 percent at baseline to 54.5 percent at week 24. The end-of-study mean TIR for adults and children was 56.5 percent and 52.5 percent for adults and children respectively, with no change from baseline.

“There was a fourfold increase in the number of participants who reached the target TIR of >70 percent using AID [from 15 percent to 60 percent],” according to de Bock, adding that the improvement in TIR in both children and adults using open-source AID was immediate and sustained throughout the study duration.

In terms of safety, de Bock reported that there was no severe hypoglycaemic or diabetic ketoacidosis events that occurred in either arm. However, there were two participants who withdrew from AID due to hardware issues.

CREATE included a total of 97 participants in New Zealand, including 48 children aged 7–15 years and 49 adults aged 16–70 years, all of whom had T1D and had been undergoing insulin pump therapy for at least 6 months prior to randomization. All participants had a mean HbA1c below 10.5 percent.

Closed loop artificial pancreas system

The open-source AID system used in CREATE comprises the OpenAPS algorithm within a modified version of AndroidAPS run on a smartphone, paired with the DANA-i insulin pump and Dexcom G6 continuous glucose monitor (CGM).

OpenAPS is a simplified closed loop artificial pancreas system (APS) that automatically adjusts an insulin pump’s basal insulin delivery to keep blood sugar levels within a safe range overnight and between meals for people with T1D. The algorithm works by collecting data on all recent insulin dosing as well as current and recent blood sugar estimates that will be used to calculate the insulin correction dose. Subsequently, it issues a command to the insulin pump to adjust temporary basal rates, as needed. [https://openaps.org/]

The system was built around the first closed loop system designed to respond to changes in sensor glucose levels and automate insulin delivery developed by Dana Lewis and Scott Leibrand. The duo then worked with Ben West to make the system talk to the pumps and the CGM to adjust the insulin delivery. They shared their work publicly by making it open source, allowing other people to refine the algorithm by testing, updating, and adding features to it. This collective work is now the open-source version of the OpenAPS. [J Diabetes Metab Disord 2020;19:1615-1629]

In a statement, Lewis, who was also involved in CREATE, said: “We were encouraged to see the greatest improvements in TIR were in those who had the lowest time in range at the start of the study. This supports the finding that a wide range of people with diabetes who want to use this technology could benefit from it, regardless of their outcomes with previous therapies.”

Open-source AID is not yet approved by the US Food and Drug Administration, but people worldwide have been using it to manage their diabetes, de Bock added. “Our findings demonstrate that this is a safe and effective technology and adds to the body of evidence supporting use of open-source AID for improving glycaemic outcomes.”