Latest News

DIY Artificial Pancreas Gaining Widespread Clinical Support

By Deborah Borfitz

December 21, 2021 | People with type 1 diabetes face unique challenges in managing their disease, including, for many decades, the need to review their blood sugar levels and adjust insulin dosing by hand. Creation of an artificial pancreas system connecting glucose sensors and insulin pumps introduced automation, but due to the slow pace of innovation and regulatory hurdles the technology was not routinely available, according to Sufyan Hussain, MRCP, Ph.D., a consultant diabetologist and honorary senior lecturer from King’s College London and Guy’s and St Thomas’ NHS Foundation Trust, who has lived with type 1 diabetes for over 30 years. Online peer communities filled this gap by engineering sophisticated, do-it-yourself (DIY) solutions, which is “citizen science and co-creation at its best.” 

Today, more than 10,000 people worldwide rely on automated insulin delivery systems based on open-source, community-generated algorithms that automatically adjust their insulin dosing in response to glucose and insulin pump data, and, in some cases, information pulled from diet- and exercise-tracking apps, says Hussain. Remarkably, while none of these DIY artificial pancreas systems have been formally tested in long-term clinical trials, there is a wealth of real-world evidence leading to first-of-its-kind guidance on their use that recently published in The Lancet Diabetes & Endocrinology (DOI: 10.1016/S2213-8587(21)00267-9) and has the endorsement of nine professional diabetes organizations. 

The paper was co-led by Hussain along with colleagues from Charité (Berlin, Germany) and Stanford University.

Children with type 1 diabetes, and their parents, account for at least 20% of users of these personalized systems, Hussain says, but they are by no means only for the young and tech-savvy. The user base includes demographic groups seen in traditional clinics, including pregnant women.

A DIY system is an option for pretty much anyone who opts for it, can invest the time to research and perform the work needed to set it up, and “appreciates that they do not have medically regulated status,” continues Hussain.

“Do-it-together” is a more apt description of the philosophy, he adds. “If you don’t have the technical skills, you have a whole community of people out there who are willing to help you.” Once the system is set up, maintaining it isn’t technically difficult but does require adjustments, which is why healthcare professional involvement is also key.

Commercial closed-loop systems have recently become available and may offer an advantage over the DIY versions in terms of their off-the-shelf simplicity, Hussain says. But the regulator-approved artificial pancreas device systems may require additional expense and are accessible only in certain countries. They also do not always allow for adequate customization—a necessity for some people living with type 1 diabetes.

“At present, the DIY approach can give you more options to adapt your treatment,” he says. “You can also mix and match the devices that you want to use and [choose]… how you want to use them.” 

In its simplest form, a DIY artificial pancreas system allows users to control their insulin level from their smartphone. Glucose data from their insulin pump is fed to an app that automatically raises or lowers insulin delivery to a fixed target, somewhat like a programmable thermostat regulates the temperature in their home, explains Hussain.

The system must figure out how much insulin to give, considering multiple confounding variables to avoid blood sugar spikes and crashes. Insulin injected subcutaneously can take 30 minutes to start working but can stay in the body for six hours or even longer, Hussain offers as an example.

On top of that, users tell the system every time they plan to eat and the estimated carbohydrate count, a familiar calculation among people with type 1 diabetes, he adds. This ensures they get a bolus of insulin at just the right time to help prevent a dramatic rise in their blood sugar level. Similarly, users can let the system know an hour or two before they plan to exercise to raise their target insulin level and counteract the fact that they’ll be burning sugar more quickly at the gym.

All of this is part of the newly published guidance on the use of DIY artificial pancreas systems to manage insulin delivery, a collaborative effort among more than 40 healthcare professionals and legal experts. It sets out recommendations allowing healthcare professionals to support such systems as a safe and effective treatment option for type 1 diabetes.

More advanced users of the systems have configured them with other third-party apps, such as online calendars where users have indicated planned visits to the gym, so insulin doses get automatically adjusted at those times and they don’t have to repeatedly enter that information manually, says Hussain. A few savvy users have even created early versions of a fully closed-loop system, with impressive results, which does not require to be told when they are eating.

Knowledgeable Patients

Use of DIY systems for automated insulin delivery has begun to normalize but is not without its critics. Many of the expressed concerns can be due to not fully understanding the system and its potential, says Hussain. “When I first came across the [do-it-yourselfers] many years back, I thought they were sort of a cult. A whole community of people out there that were talking a language I couldn’t speak.”

It took a little time to understand and appreciate the engineering feat citizen scientists had pulled off—aided by the educational manuals, tools, and resources simultaneously developed to improve the user interface, he says. “It is just a reminder to us as physicians that we have to stay open-minded.”

As he often tells medical students, about a 5% of patients in any specialty are likely to know more about their medical condition than their doctor. “It’s important to work with [patients] and use their knowledge and understanding in a positive way… to co-produce solutions together,” says Hussain.

Self-creating an artificial pancreas system takes time to build, even with the step-by-step instruction manual, so it’s not for everyone, he adds. But the how-tos are so well explained that “it’s very hard for things to go wrong. The system is based on safety and your fall back if technical aspects do not work is reverting to what you were doing manually anyway.” 

Algorithm Development

The genesis of the DIY artificial pancreas systems began about a decade ago when a group of worried parents created Nightscout to remotely view their child’s glucose data while they slept. With that system, glucose sensor readings get broadcast to the cloud, so the next logical step was to try using all that data to build an alternative insulin delivery mechanism.

In 2011, a security flaw in Medtronic’s implanted insulin pumps made it possible for individuals to hack into the device to remotely control insulin delivery, he says. While the security breach never caused widespread problems, it revealed the connection point for adjusting insulin doses—at least those pumps using wireless technology such as Bluetooth. 

Increasing numbers of pump manufacturers are now quietly supportive of the DIY movement, as evidenced by their embrace of open protocols that permit people to decipher how the remote signaling on their devices work. The community-generated algorithm initially followed established rules about how to correct for dips and spikes in blood sugar levels, says Hussain, but they grew increasingly more sophisticated with time and use—including options to deliver micro-boluses and automatically start sending insulin when a meal is eaten and stop when blood sugars are going down—based on the same principles as commercial automated delivery systems.

Since every individual is slightly different, the parameters can be flexibly altered, he adds, which is one of the bonuses of open-source systems. Commercial systems are engineered around the average user, as determined by clinical studies, but real-world users can vary widely in their requirements. 

Two main families of open-source automated insulin delivery algorithms are publicly available, and they all make predictions about future glucose concentrations detected by continuous glucose sensors, says Hussain. As with any system, they have mitigation strategies should the needs of individuals differ from the established settings and, in some instances, users can also optimize settings based on past model error. 

The algorithms first became available in 2014, a couple of years before the first commercial system received U.S. Food and Drug Administration approval. Newer iterations of the algorithms get released periodically, he adds, and the community-based approach allows users to upgrade as needed and desired.

Beyond Diabetes

As Hussain sees it, the “next revolution” in personalized medicine will be co-creation of solutions with people living with a condition and able to be ambassadors of their own healthcare. This will naturally include greater use of patient-reported outcomes and donated data in research studies, he says, citing the OPEN project that is working to establish an evidence base surrounding the impact of DIY artificial pancreas systems in people with diabetes as well as healthcare systems.

The project will be gaining insights from people with type 1 diabetes regarding how insulin delivery solutions need to accommodate the realities of their everyday life, such as fasting, exercise habits, and menstrual cycles, say Hussain. As a diabetologist, he has seen first-hand that the knowledge gained directly from people with type 1 diabetes can be just as impactful as what’s being generated in clinical trials.

As another example, Hussain notes he was part of a study newly published in the Journal of Diabetes Investigation (DOI: 10.1111/jdi.13720) looking at how people with type 1 diabetes, across different countries, participate in the 30 days of sunrise-to-sunset fasts associated with Ramadan using such systems. It offers practical, community-generated insights on successfully managing this high-risk situation where proper insulin adjustments can prevent the development of potentially life-threatening hypoglycemia or hyperglycemia. Such research helps guide clinicians on how to adjust insulin and guide people with diabetes more generally, including situations where therapies are via injections. 

Commercial device makers are no doubt paying attention to the growing body of citizen science work, he says. But the pace of innovation within large corporation organizations tends to be slower, both because of bureaucratic hurdles and the necessity of conducting trials to market their products.

Fast forward another decade, and the power of community action could well spread to other areas of medicine while the necessity of DIY systems for automated insulin delivery potentially diminishes as commercial systems improve and become cheaper to use, says Hussain. One of his colleagues is now interested in exploring the redesign of hearing aids for young wearers, for instance, which could involve the development of self-management apps. 

Load more comments