Full closed loop open‐source algorithm performance comparison in pigs with diabetes

Abstract

Understanding how automated insulin delivery (AID) algorithm features impact glucose control under full closed loop delivery represents a critical step toward reducing patient burden by eliminating the need for carbohydrate entries at mealtimes. Here, we use a pig model of diabetes to compare AndroidAPS and Loop open‐source AID systems without meal announcements. Overall time‐in‐range (70–180 mg/dl) for AndroidAPS was 58% ± 5%, while time‐in‐range for Loop was 35% ± 5%. The effect of the algorithms on time‐in‐range differed between meals and overnight. During the overnight monitoring period, pigs had an average time‐in‐range of 90% ± 7% when on AndroidAPS compared to 22% ± 8% on Loop. Time‐in‐hypoglycemia also differed significantly during the lunch meal, whereby pigs running AndroidAPS spent an average of 1.4% (+0.4/−0.8)% in hypoglycemia compared to 10% (+3/−6)% for those using Loop. As algorithm design for closed loop systems continues to develop, the strategies employed in the OpenAPS algorithm (known as oref1) as implemented in AndroidAPS for unannounced meals may result in a better overall control for full closed loop systems.

Study of blood glucose and insulin infusion rate in real-time in diabetic rats using an artificial pancreas system

“…Of note, the BGL pattern of a healthy rat showed hardly any fluctuation with continuous insulin infusion albeit mock insulin infusion determined by the OpenAPS algorithm (Fig 5,upper panel). These results support the strategy of frequent supermicrobolus injections of insulin as a part of BGL control algorithm [12].…”

A Systematic Review of Commercial Hybrid Closed-Loop Automated Insulin Delivery Systems

“…Ideally, AID systems would instead be fully closed-loop systems that were completely automated, requiring minimal user input and with no need for prior warning of mealtimes or exercise, as required by hybrid systems [9][10][11]. The sensor lag time and delays in insulin action times have been some of the barriers to achieving this, although progress in this regard has been noted in some open-source AID systems, based on testing in animal models and anecdotal real-world experiences in clinics [12]. An ideal AID system would also be a dualhormone system in order to more closely mimic a biological pancreas and reduce the risk of hypoglycaemia by countering aggressive insulin delivery through the delivery of glucagon in addition to insulin [13].…”

Long-Term Glucose Forecasting for Open-Source Automated Insulin Delivery Systems: A Machine Learning Study with Real-World Variability Analysis

“…However, OpenAPS is uniquely designed to generate predictions based on various scenarios, including whether carbohydrates are fully absorbed, or a meal is consumed but not recorded to the system. These predictions are conditionally blended and heuristically used [ 17 ], such as to produce estimates of the lowest predicted glucose value to be observed over the timeframe relevant for insulin dosing and separately the blended average glucose level over the approximate period when the activity of any additional insulin would be peaking, in order to limit contributions to hypoglycemia while also seeking to minimize hyperglycemia. Therefore, OpenAPS is one such system where an ML-based prediction algorithm could be introduced and blended into the current set of predictions and used alongside the backstop of safety rules used by the system to achieve the highest possible time in the target glucose range (known as “time in range” or TIR) without much hypoglycemia or hyperglycemia.…”