Fully Artificial Pancreas Produces Results Comparable With Hybrid Artificial Pancreas

This new study from Canada compared outcomes in patients living with type 1 diabetes who used 2 closed-loop insulin delivery systems: a fully artificial pancreas that did not require meal input and a hybrid artificial pancreas that required carbohydrate counting.

A new randomized study from Canada has determined that a fully closed-loop insulin delivery system that incorporated the antihyperglycemic pramlintide and did not require meal input for carbohydrate accountability (a fully artificial pancreas) may be capable of producing comparable results vs a closed-loop system that required carbohydrate counting (a hybrid pancreas).

Results of the open-label controlled crossover noninferiority trial among 28 adults living with type 1 diabetes (T1D) conducted at McGill University Health Centre in Montreal were published in The Lancet Digital Health.

“Because the level of literacy and numeracy required for carbohydrate counting can be a barrier for many people with T1D,” the authors wrote, “a fully automated closed-loop system that eliminates the need for carbohydrate counting or meal announcement is highly desirable for T1D.”

Their primary outcome was 24-hour time spent in target range, defined as 3.9 to 10.0 mmol/L for this study, with a 6.0% noninferiority margin. All of the study participants had been living with T1D for at least 1 year, had a 12% or lower glycated hemoglobin (A1C) measure, and had been using an insulin pump for at least 6 months. Each system was used for 2 27-hour periods and there was a 2-day crossover intervention when switching systems.

Among the 24 adults with data from use of the fully artificial pancreas and the hybrid system, times spent in the target range were 74.3% (95% CI, 61.5%-82.8%) and 78.1% (95% CI, 66.3%-87.5%), respectively. This equated to a paired difference of 2.6% (95% CI, –2.4% to 12.2%; noninferiority P = .28).

Although the target range time was slightly lower with the fully artificial pancreas, use of this fully closed-loop system resulted in fewer hypoglycemia-related events compared with the hybrid system: 33% vs 58%. In contrast, participants did not report any instances of nonmild nausea or nonmild bloating when using the hybrid pancreas vs 13% and 4%, respectively, when using the fully artificial pancreas. Gastrointestinal adverse effects were reported by 29% of those using the fully closed-loop system and 8% using the hybrid system.

Participants were enrolled in the study between February 8, 2019, and September 19, 2020. Exclusion criteria included gastroparesis and recent severe diabetic ketoacidosis or hypoglycemia. The mean (SD) patient age was 35 (14) years, mean A1C was 8.1% (1.3%), 50% each were male and female, and the mean diabetes duration was 19 (14) years.

The algorithms each system used differed in that the fully artificial pancreas algorithm was initialized with just the total daily dose and basal rates for insulin while the hybrid artificial pancreas also included insulin-to-carbohydrate ratios. Sensor readings, however, were received every 10 minutes in both systems.

Additional study analyses produced these findings:

  • Less time was spent in the glucose range of 3.9 to 7.8 mmol/L when using the fully closed-loop vs the hybrid closed-loop system: 45.3% vs 53.4%
  • No one in either group spent time below 2.8 mmol/L, below 3.3 mmol/L, or above 16.7 mmol/L
  • The fully artificial pancreas did not lead to any time spent below 3.9 mmol/L, while 1.8% of time with the hybrid pancreas was spent in that range
  • More time was spent above 7.8, 10.0, and 13.9 mmol/L when using the fully closed-loop vs the hybrid closed-loop system: 54.3%, 24.3%, and 2.4% vs 44.3%, 19.8%, and 0.4%, respectively

In addition, although total daily basal insulin units were equivalent between the groups, at 30.5 (95% CI, 19.8-39.9) for the fully closed-loop system and 30.7 (95% CI, 23.4-44.4) for the hybrid system, the latter group did require almost 10 more units of bolus insulin overall, at 18.8 (95% CI, 13.4-23.6) vs 27.9 (95% CI, 18.6-31.1), respectively.

Participants also spent more time overnight in the target range when using the fully artificial pancreas vs the hybrid pancreas, at 90.6% vs 86.5%, but less time during the day in this range, as 66.1% vs 78.6%.

When stressing the importance of their findings, the authors highlighted that their investigation is likely the first to investigate use of a closed-loop system that incorporated pramlintide “to delay meal glucose absorption to match the delays in insulin absorption.”

“This study is an important step toward fully automated closed-loop glucose control for T1D,” the authors concluded. “Considering that the time-in-range with the fully closed-loop system was high, larger studies with the fully closed-loop system in an outpatient setting are needed.”


Tsoukas MA, Majdpour D, Yale J-F, et al. A fully artificial pancreas versus a hybrid artificial pancreas for type 1 diabetes: a single-centre, open-label, randomised controlled, crossover, non-inferiority trial. Lancet Digit Health. 2021;3(11):e723-e732. doi:10.1016/S2589-7500(21)00139-4