Department of Electrical, Computer and Biomedical Engineering, University of Pavia, Pavia, Italy.
CLOSED LOOP Systems, Prague, Czech Republic, Prague, Czech Republic.
Diabetes Technol Ther. 2020 Feb;22(2):112-120. doi: 10.1089/dia.2019.0375. Epub 2019 Dec 17.
Safety data on Do-It-Yourself Artificial Pancreas Systems are missing. The most widespread in Europe is the AndroidAPS implementation of the OpenAPS algorithm. We used the UVA/Padova Type 1 Diabetes Simulator to in silico test safety and efficacy of this algorithm in different scenarios. We tested five configurations of the AndroidAPS algorithm differing in aggressiveness and patient's interaction with the system. All configurations were tested with insulin sensitivity variation of ±30%. The most promising configurations were tested in real-life scenarios: over- and underestimated bolus by 50%, bolus delivered 15 min before meal, and late bolus delivered 15 min after meal. Continuous Glucose Monitoring (CGM) time in ranges (TIRs) metrics were used to assess the glycemic control. In silico testing showed that open-source closed-loop system AndroidAPS works effectively and safely. The best results were reached if AndroidAPS algorithm worked with microboluses and when half of calculated bolus was issued (mean glycemia 131 mg/dL, SD 27 mg/dL, TIR 91%, time between 54 and 70 mg/dL <1%, and low blood glucose index even <1). The meal bolus over- and underestimation as well as late bolus did not affect the TIR and, importantly, the time between 54 and 70 mg/dL. In silico testing proved that AndroidAPS implementation of the OpenAPS algorithm is safe and effective, and it showed a great potential to be tested in prospective home setting study.
DIY 人工胰腺系统的安全性数据缺失。在欧洲应用最广泛的是 AndroidAPS 实现的 OpenAPS 算法。我们使用 UVA/Padova 1 型糖尿病模拟器对该算法在不同场景下的安全性和有效性进行了计算机模拟测试。我们测试了 AndroidAPS 算法的五种不同激进程度和患者与系统交互方式的配置。所有配置都在胰岛素敏感性变化±30%的情况下进行了测试。最有前途的配置在现实场景中进行了测试：餐前和餐后 50%的超量和低估推注、提前 15 分钟推注和延迟 15 分钟推注。连续血糖监测（CGM）时间范围内（TIR）指标用于评估血糖控制。计算机模拟测试表明，开源闭环系统 AndroidAPS 有效且安全。如果 AndroidAPS 算法与微推注一起工作并且只发布计算推注的一半（平均血糖 131mg/dL，SD 27mg/dL，TIR 91%，54-70mg/dL 之间的时间<1%，低血糖指数甚至<1），则可以达到最佳效果。餐前和餐后推注的高估和低估以及延迟推注均不会影响 TIR，重要的是，54-70mg/dL 之间的时间也不会受到影响。计算机模拟测试证明，OpenAPS 算法的 AndroidAPS 实现是安全有效的，并且具有在前瞻性家庭环境研究中进行测试的巨大潜力。
