Glucontrol: Implementation and testing of open source platform for APS clinical trials

DELFINA ARAMBARRI; FABRICIO GARELLI; LEANDRO MENDOZA; NICOLÁS ROSALES

Madrid

Conferencia; 13th International Conference on Advanced Technologies & Treatments for Diabetes Conference; 2020

Background and aimsWithin the framework of clinical trials of artificial pancreas systems (APS), the development of reliable platforms is crucial. On the basis of the experience conducted in 2017 in the Hospital Italiano of Buenos Aires, where a new controller designed in collaboration between the University of La Plata and the Technologic Institute of Buenos Aires (ITBA) was tested [1], the need for a platform of its own arises. Thanks to the support by the University of Virginia, the controller was implemented on a first instance on the DiAs platform. The obtained results were satisfactory of a first approach to glucose control without carbohydrate counting. To further develop the ARG algorithm, the design of an own platform was carried out in view of future trials. MethodsTaking into account the use of a smartphone as a supporting device and the development of DIY APS, the development of the GluControl platform (Fig. 1) was based on the Android APS (AAPS) [2]. This system was designed from the Open APS features showing wide acceptance by both the diabetic subjects and medical community, and also the academic community. Initially, the connection with the same commercial devices as those used in clinical trials were considered: Roche Accu-Chek Spirit Combo insulin pump and DexCom G4 Platinum CGM. To run the system, a Motorola Moto G5 cell phone was used, in which the controller and a user interface were developed. Remote monitoring of the main variables of the system was implemented using NightScout [3]. A a patient- and physician-friendly interface was considered important but also an access to the controller´s internal variables to monitor its performance.ResultsThe base system (GluControl, Dexcom G4 and Roche Spirit Combo) was tested during 240 hs, showing a 7% of disconnections. Recently, a closed loop version with the ARG algorithm including disconnection mitigation techniques has been tested, showing on average a relative error of -5,37% of total insulin delivered compared to Matlab version implementation of the controller. Considering clinical trials results with DiAs, which showed an error of -11,51% versus Matlab, the difference between implementations led to an error of 6,94%.ConclusionsIn this work, an Artificial Pancreas system called GluControl is introduced as an open alternative to be employed in clinical trials. Based on AndroidAPS, it adds the ARG algorithm and its input/output layer used in DiAS, in which the disconnections are treated and the online database is updated. It also includes a monitoring system destined to be used with several patients. Taking into account further applications, an advanced version of GluControl was tested using another CGM: Freestyle Libre and a Bluetooth adapter, MiaoMiao reader.