DIY Artificial Pancreas Systems: Safety concerns and potential risks in perspective?

Published: Posted on

In this post we explore some of the risks and benefits of DIY Artificial Pancreas technologies. In future posts we will take a look at the law, ethics, and policy implications of these.

In a previous post we explained how Do-it-Yourself Artificial Pancreas Systems (DIY APS) are systems made up of four interconnected components:

  • a continuous glucose monitor (CGM) to provide ‘real time’ glucose data;
  • software (AndroidAPS, OpenAPS, or Loop) to interpret CGM readings, perform calculations and send instructions to insulin pumps;
  • a smartphone or small computer to run the software; and
  • an insulin pump to administer insulin therapy based on commands received from the software.

Once connected, these components create a hybrid closed loop system, automatically adjusting and administering insulin doses. The system is a hybrid rather than fully closed loop system as users still need to ‘announce’ meals by manually inputting data. This is then fed into the algorithm along with the data gathered from the CGM.

In this post, we consider some safety concerns raised by DIY APS, focusing on the potential risks posed by each of the four components listed above. We also consider the significant benefits which users report from using these systems. These include an increase in time spent with blood glucose levels in target range, a reduction in the burden of diabetes decision-making, and less anxiety around sleeping.

Continuous glucose monitors (CGM)

Accurate CGM data is critical for the safety of the DIY APS. If the CGM tells the rest of the system that the glucose level is higher than it is, the dose of insulin delivered will be higher than required, potentially causing the user to experience hypoglycaemia (low blood glucose). These episodes can potentially be serious if left untreated. For individuals who have lost the warning signs of a hypo, this could be particularly problematic during the night, where the user is not awake to double-check the reading.

Furthermore, not all CGM devices are designed for using the data as the basis for treatment decisions. Dexcom G6 is one example of a CGM that has received regulatory approval for treatment decisions to be made off the data. By contrast, safety information for Freestyle Libre 2 states that interpretation of the readings should be based on the glucose trends and several sequential readings over time. Moreover, it states that the device ‘must not be used with automated insulin dosing (AID) systems, including closed loop’.


The software part of DIY APS comprises an application (most often a phone app) that can interpret data from devices, perform calculations, and issue instructions to devices. This software is open-source, meaning anyone can inspect, modify, and enhance the source code (as opposed to closed source where only the original authors can modify the source code). One benefit of open source software is that its transparency means any flaws may be quickly detected by other developers and fixed. However, this is not always the case. Like all software, open-source applications can contain undiscovered ‘bugs’ which could cause devices to malfunction. Moreover, as with most technologies, open-source software is vulnerable to attack. Given the potency of insulin, any accidental or malicious changes to the software, which might be unintentionally authorised, could be catastrophic for users.

Mobile phones

Individual users also need to be mindful that their mobile phone should provide a secure ecosystem in which the app can run. This may mean being cautious of running the DIY APS software on their mobile phone alongside other apps, which could be hiding malware that could potentially harvest sensitive data, or render the phone useless. Additionally, users must keep their mobile phones safe. Whilst a low risk, the possibility that a malicious person could steal, or secretly access the phone and change settings in the app, and issue a high dose of insulin is nonetheless present.

These are of course inherent issues with the security of mobile phones, but as people become more integrated with technology the potential for them to experience physical harm is increased.

Insulin pumps

There is no evidence to-date that indicates using DIY APS damages insulin pumps, although all users report an increase in battery usage. Although there is a risk that any time a device uses radio communications, it could be exploited, this is also the case for standard, non-looping pump operations. For this reason, DIY APS instructions advise that, within the pump settings, users limit the maximum amount of insulin that can be automatically delivered to a level which they believe is safest for them. It is believed that these hard limits cannot be circumvented by malicious hackers.

Additional components

Depending upon the type of APS software used, and hardware that a user has available, additional components to those listed above might be required. For example, Loop users need an additional device called a RileyLink, and a small computer such as a Raspberry PI. OpenAPS users need to store their CGM data in the cloud, or in an app that tracks data, such as Nightscout. Although not recommended by the DIY APS community, some users may choose to create a CGM by using an adapter, such as MiaoMiao, with Freestyle Libre 1 (a flash glucose monitoring system available on the NHS).

All of these different components create a complex chain, and DIY APS users may be using systems which vary slightly from those used by others in the community. This means that self-reported safety data for one set-up may not mean that an alternative set-up is equally safe, and users must be mindful of the variables.

Risks in perspective

As we have seen, DIY APS systems are made up of multiple components, and there are potential safety concerns with all of them. CGMs may not read the data appropriately, the software may contain errors or be interfered with maliciously, phone communications are vulnerable to interception and the user needs to be aware of the battery life of their pump. Moreover, to guarantee safety, all components have to work together seamlessly, something they were not originally designed to do.

However, the use of DIY APS also has a number of benefits. People living with diabetes have the daily responsibility of keeping their glucose levels in check. This means testing their glucose levels, either via finger-pricks or CGM, and then considering a huge number of factors which may influence how much insulin to take. They repeat this process multiple times a day, every day. It has been reported that people with type one diabetes make approximately 300 more decisions a day than someone without diabetes. As there is a potential for error with every decision, there are lots of opportunities for people with diabetes to make mistakes when self-administering insulin.

DIY APS reduces some of this decision-making burden and, in reducing the amount of decisions people have to make each day, can help mitigate the risk of human error.  CGMs can monitor glucose levels up to 288 times a day, which is far more than the most diligent person could do (even if they didn’t have to balance every other aspect of their life alongside glucose monitoring!).

As a result of using DIY APS, users say they experience better overall blood glucose management, a reduction in the mental and manual labour of managing type one diabetes, and a decrease in (anxiety around) undetected hypoglycaemia, especially during sleep. The significance of this and the impact on patients’ lives should not be underestimated.

We have begun to engage in dialogue with DIY APS users, clinicians, and regulators to explore and understand further the risks and benefits surrounding these technologies and, importantly, to begin to think about the law, ethics, and policy implications of these. More on this in the coming months.

Written by Victoria Moore and Joseph Roberts


Work on this was generously supported by Wellcome Trust Investigator Award in Humanities and Social Sciences 2019-2024 (Grant No: 212507/Z/18/Z), an ESRC Impact Acceleration Award, and a Quality-related Research grant from Research England.