# Fully closed-loop systems: can people with type 1 diabetes just do it? Insights from open-source systems

## TL;DR

This paper reviews open-source automated insulin delivery systems for type 1 diabetes, showing they can manage glucose without meal announcements, but still face practical challenges.

## Contribution

The paper highlights how open-source AID systems can operate without meal announcements, offering insights into advancing fully automated diabetes management.

## Key findings

- Open-source AID systems can manage glucose levels without meal announcements, achieving results comparable to hybrid closed-loop systems.
- Successful implementation requires advanced algorithms, personalization, and clinician involvement.
- Challenges include handling rapid insulin sensitivity changes and expert system configuration.

## Abstract

Automated insulin delivery (AID) systems have significantly advanced diabetes management, progressively reducing user interactions required for optimal glucose management. This review evaluates the current landscape and future potential of AID systems without meal announcement, particularly focusing on real-world insights from open-source AID (OS-AID) technologies. Although commercial AID systems operating in hybrid closed-loop (HCL) mode have improved glycaemic outcomes, they remain dependent on manual meal announcement and user-driven actions, limiting their real-world utility. Current versions of OS-AID systems, developed by the diabetes community, can allow operation without meal announcements, presenting an opportunity to move closer to truly automated diabetes management. Recent clinical trials suggest that OS-AID systems can effectively manage glucose levels without meal announcements, achieving glucose levels comparable with those obtained by AID systems in HCL mode, with the potential of reduced management burden for users. However, practical challenges persist, including the need for expert configuration and handling of rapid changes in insulin sensitivity, such as during exercise or rapid glucose fluctuations following predicted low-glucose. This review synthesises insights from user and healthcare professional experiences, and emerging clinical evidence. It highlights the fact that successful implementation of AID without meal announcement requires advanced algorithmic responsiveness, user personalisation and ongoing clinician engagement. Looking forward, integrating adjunctive therapies, artificial intelligence and enhanced physiological modelling will likely enhance system performance to drive the next generation of diabetes care towards wider adoption and true ‘set-and-forget’ functionality.

The online version contains peer-reviewed but unedited supplementary material including a slideset of the figures for download, available at 10.1007/s00125-025-06644-8.