First in Human Feasibility Study: Automated Insulin Delivery Utilizing a Self-Adapting Algorithm in Adults With Type 1 and Type 2 Diabetes.

Background: This feasibility study assessed a novel self-adapting closed-loop system which does not require carbohydrate announcement, in adults with type 1 and type 2 diabetes.

Methods: Single-arm study, comprising a 14-day run-in using participants' usual insulin therapy with a blinded continuous glucose monitor (CGM), followed by 12 weeks use of the novel closed-loop system. The algorithm adjusted its own parameters after 4, 6, 8, and 10 weeks of use.

Results: Thirty-two participants with type 1 and 10 participants with type 2 diabetes were enrolled. Mean time in range (TIR; % CGM readings = 70-180 mg/dL) was 37.7% at baseline and 55.9% during the intervention period in type 1 diabetes; 17.6% at baseline and 51.5% during the intervention period in type 2 diabetes. Median time <70 mg/dL during the intervention period was 1.1% in type 1 and 0.0% in type 2 diabetes. Median TIR was 65% following the fourth algorithm adaptation. Median daily insulin delivered by manual bolus was 1.0 units in type 1 and 0.0 units in type 2 diabetes, consistent with no meal announcement. There were four serious adverse events: worsening retinopathy, severe hypoglycemia following a period of paused automation, and two hospitalizations unrelated to the device.

Conclusions: A closed-loop algorithm that adjusts its own parameters and requires no meal announcement was feasible in a cohort of adults with type 1 and type 2 diabetes. Clinical benefits were most apparent with the fully adapted algorithm.