Neural-Net Artificial Pancreas: A Randomized Crossover Trial of a First-in-Class Automated Insulin Delivery Algorithm
Primary result · TIRResult — go to the resultDifference 1 percentage pointAuthors' conclusion
In a randomized crossover study, a neural-network encoding of a complex model-predictive control algorithm demonstrated similar performance, at a fraction of the computational demands.
Study context
- Population
- Automated insulin delivery (AID) users with T1D; ages 22–68 years, diabetes duration 7–58 years, 10/5 female/male, baseline HbA1c 5.4%-8.1%17 recruited
- Participants
- 15 users
- Comparison
- Neural-Net Artificial Pancreas vs original UMPC AID algorithm
Methodological detailspilot-feasibility, two consecutive 20-h hotel sessions
- Pilot-feasibility
- Two consecutive 20-h hotel sessions
Primary result
Time-in-range (TIR)
Difference
1 percentage point
By group
As reported in the abstract
Further findings
Coefficient of variation
Time <70 mg/dL
- The two algorithms achieved similar times <70 mg/dL
- 2.0% versus 1.8%
Insulin recommendation under identical inputs
- Average absolute insulin-recommendation difference
- 0.031 U/h
Computational demands
- NAP had sixfold lower computational demands than UMPC
Safety
- No serious adverse events on either controller
Original publication
Neural-Net Artificial Pancreas: A Randomized Crossover Trial of a First-in-Class Automated Insulin Delivery Algorithm. Diabetes Technol Ther. Published 30 May 2024. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →