What’s all the fuss about the DIY artificial pancreas?
DIY closed-loop insulin delivery has provoked controversy within the diabetes community, raising questions around patient empowerment, safety and liability, and whether the regulatory process remains fit for purpose in a technological age. medwireNews takes a look at the issues.
The first commercially available artificial pancreas system received FDA approval in September 2016, but this was too little too late for some people with type 1 diabetes: the DIY closed-loop revolution – in which people marry up their continuous glucose monitors and insulin pumps with an open-source control algorithm – was already well advanced.
A tipping point
2018 marked the year the enterprise was afforded a wider audience, with the ADA, for example, allocating it a Saturday afternoon slot in one of the largest lecture theatres at its annual Scientific Sessions. The planning committee member responsible for getting it there was Will Hsu (Joslin Diabetes Center, Boston, Massachusetts, USA).
“I think the DIY community has come to a tipping point,” he says, explaining why he felt the time was ripe for its inclusion. The “critical mass” of people now involved necessitates open discussion, he believes, “much like, for example, the herbal medications out there: patients are using the supplements.”
He says that “by including the program as part of the ADA, it does not necessarily mean that one endorses it, or one has one feeling or another, but we have to face the reality when the clinicians are taking care of the patients.”
Hsu found limited knowledge of the subject among his fellow committee members, implying a clear need for wider exposure, but on the day of the session only a couple of hundred people attended, in a hall designed to hold 2000. “It did look sparse,” he says, “but I think that goes to show you how the greater medical community knows very little about the DIY movement.”
Interestingly, the ATTD, despite its self-declared commitment to new diabetes technologies and treatments, held back from including DIY closed-looping in its 2018 conference program (although it included it in 2019), hinting at the controversy the movement has stirred up within the diabetes healthcare professional community; speaking to individual specialists undercovers everything from cautious advocacy to doom-saying.
Jason Wittmer is a lung specialist from Mercy Medical Center-Des Moines, Iowa, who was drawn into the DIY movement when his son developed type 1 diabetes. He says that, on the one hand, “there are people who are willing to endorse this as full-throatedly as they can within the limits of […] the regulatory system.” But on the other hand, “in my personal opinion, there are some people who would like to see us go away and stop talking.”
The mixed reaction to what is, after all, a patient-led attempt to improve blood glucose management speaks to the myriad issues it has thrown up, covering patient autonomy and safety, medical ethics, and the question of whether the medical regulatory process remains fit for purpose in an era of fast-moving technological advances and informed, empowered patients.
The DIY artificial pancreas movement was born of some very specific issues that people with diabetes were unable to address with current options, with perhaps the best-known example being the ability of one of the movement’s leaders, Dana Lewis, to sleep through the low-blood-glucose alarm on her continuous glucose monitoring device. So with potential device improvements seeming distant – on the other side of a long, slow process of development, testing, and approval – technically savvy patients began to explore what they could do for themselves.
This led to DIY closed-looping, meaning that some people were in possession of a system that automatically monitors blood glucose and adjusts the insulin infusion rate well before the first commercial system was approved.
It’s even tough with these tools. [Patients] still have to think about their diabetes, they’ll still have to do a lot of things to keep themselves healthy.
And the rewards of this technology can be huge.
“Diabetes stinks,” says Wittmer. “It’s tough.”
Even as a physician, he found it hard-going trying to manage his son’s blood glucose.
“[I was] working as hard as I possibly could, along with my wife […] doing all the things that we could to try to keep him healthy, to try to keep his life as normal as we could and to give him the best possible outcomes. And quite frankly I wasn’t cutting it, not to the degree that I wanted to. I was always giving something up. I was giving up [glucose] control, I was giving up quality of life.”
And his medical knowledge, while helpful, also gave him “a very acute sense of how things can end very badly.”
One of the most frightening, albeit very rare, outcomes of type 1 diabetes is “dead in bed,” believed to result from prolonged, severe hypoglycemia during sleep; hence a too-quiet low-blood-glucose alarm was far more than a minor irritation for Dana Lewis.
Wittmer says: “And so after my son was diagnosed – this is honest – I woke up every morning for 8 years, my very first thought was: ‘is he dead this morning?’ ”
Having a continuous glucose monitor, which allows more frequent checks with fewer finger-pricks relative to standard monitoring, helped with daytime glucose control and nighttime reassurance, but only to a degree. He says: “I still remember one day when I couldn’t tell if he was breathing or not and it was just such a horrible sickening feeling until I lifted the covers and I could tell that he was breathing.”
But one morning a couple of months after starting to use DIY closed-looping, “I had gotten up and I had gone downstairs, and I made myself a cup of coffee and I was sitting at the kitchen table and I suddenly had this overwhelming feeling that I hadn’t asked myself that question. And it was just the most amazing feeling in the world to realize I didn’t – that wasn’t my first waking thought.”
For his family, Wittmer says, the impact of the closed-loop technology was “earth-shattering.”
“It’s even tough with these tools,” he stresses. Patients “still have to think about their diabetes, they’ll still have to do a lot of things to keep themselves healthy – you can’t just turn these systems on” and expect they will do everything for you.
But, he says, closed-looping “takes a really bad situation, and it makes it more livable and people can focus on living their lives.”
Now, of course, this technology is becoming commercially available, but community-built glucose-control algorithms started to arrive well ahead of the first regulator-approved system, and the speed at which the DIY algorithms are upgraded, unencumbered by regulatory requirements, mean that they already have numerous features not found in the FDA-approved Medtronic MiniMed 670G hybrid closed-loop system. Wittmer says that with commercial systems “you can’t set your own individualized target, you can’t interact with the system remotely, you can’t notice that your son missed treating for breakfast and enter those carbs from your cellphone as you’re walking down the hall in the hospital, or set a temporary target remotely because you know he has gym class.”
[Closed-looping] takes a really bad situation, and it makes it more livable and people can focus on living their lives.
To which engineer and endocrinologist Rayhan Lal (Stanford University, California, USA) adds: “The other aspect that people may not appreciate is that people don’t always like interacting with medical devices.”
Lal, who himself lives with type 1 diabetes, says: “You could be at dinner, for example, and nowadays everyone is rude and they pull out their cellphones, but doing that is not looked upon as being now a weird or out-of-the-ordinary activity. So with Loop, for example, specifically, all interaction is through the iPhone. So you can just pull out your iPhone at dinner, as opposed to getting the pump out and dosing on the pump, and then maybe someone looks at you and says ‘oh, what’s that?’
“I think people prefer that.”
Does it work?
Over the past couple of years, the DIY community has taken pains to gather data from new users, comparing their glucose control before and after they close the loop. This has revealed significant improvements, particularly in time spent in hyperglycemia during the day, despite necessarily involving a self-selected group of highly motivated people with good baseline glucose control.
But is this too good to be true? Diabetologist John Wilding, from the University of Liverpool, UK, says: “It kind of puzzles me that some very big companies and some very serious scientists have been working steadily to solve this problem for decades, and they are at the point where they can probably get a safe overnight and maybe a safe 1 or 2 days if the patient is reasonably sedentary and with very predictable meals and so on.”
He says: “It just surprises me that people seem to be able to make this work.”
Wilding suggests that, besides being motivated and having the ability to manage the technology, these DIY users may also have relatively stable diabetes, with some residual C-peptide. These differences are removed in a randomized trial, “whereas if you just have people who are self-reporting and saying this works for me, it might work for Joe Bloggs who is able to do this, but it seems unlikely that this would just then apply to everybody.”
An exclusive club
And there lies one sizable problem with DIY closed-looping: however effective it may be for some individuals, there are multiple barriers to being able to use these systems.
First, people have to obtain the hardware, which depending on the system could mean sourcing specific older models of insulin pumps – and demand for these pumps may mean prospective users have to pay over the odds for secondhand, out-of-warranty equipment. In addition to which, they may have to purchase a continuous glucose monitor and intermediate devices such as a Raspberry Pi, and meet ongoing costs such as replacement sensors. These costs may be less for people who can fund some components through their national healthcare system or medical insurance, but are usually considerable.
On top of that there is a knowledge/education barrier, as people have to set up and troubleshoot the system. So effectively, DIY closed-looping is restricted to the well-off, well-educated sector of society, wearyingly similar to the healthcare access issues reported for more conventional areas of medical care.
“That’s a concern for people that are in the community,” says Wittmer. “We recognize the privileged position that we have.”
On the other hand, the amount of access to DIY closed-looping that already exists creates its own problems.
Wittmer cites the technical barrier as a positive thing. “We want people to understand what they’re doing; we want people to understand that the onus is on them to test these devices,” he says. “They’re doing experimentation on themselves or their kids and to say it any other way would be unfair.”
However, Lal’s experience of seeing patients who are DIY closed-looping tells him that plenty of them “actually have no clue how any of it works.” He says: “They just have a friend who happens to be in computers or something like that and sets it up for them.”
Yet these people are using out-of-warranty versions of devices that are themselves “relatively new and imperfect,” according to Wilding, and potentially without the support of a doctor who fully understands the system.
Even if using an approved system under expert guidance, patients still have to be aware of multiple issues.
“It is the same caveats as driverless cars,” says Wilding. “There are warnings all over them saying keep your hands on the wheel and keep your eyes on the road because it is not infallible. It requires a hand on the wheel and an eye on the road the whole time, otherwise inevitably it will crash, one way or the other.”
Who’s responsible anyway?
In the event of a crash, is it the individual patient, their doctor, or the device manufacturer while still under warranty who is ultimately responsible? Anecdotally, this is one issue that gives some doctors major reservations about supporting patients with a DIY artificial pancreas, particularly in the case of parents using the technology to treat children.
Wilding would be “very worried” about this scenario. “I think there are all sorts of complex medical, and other legal issues around that so I think it would be difficult to offer support in that situation,” he says.
As a diabetes specialist with no specific expertise in technology, Wilding’s approach would be to refer the patient to a diabetes technology expert. Such a person is Hood Thabit (University of Manchester, UK), a long-time member of the closed-loop research group led by Roman Hovorka, responsible for developing the Cambridge closed-loop algorithm.
He says that, at the moment, the onus is still on the individual clinical teams to decide what support they are prepared to give, and what liability they are prepared to risk.
But he says: “I know there is work being done behind the scenes to try […] to produce a guideline that could help clinicians decide what to do in those scenarios.
“They’re moving this forward because they know that a lot of healthcare professionals are asking the same question. They’re in a bit of a dilemma at the moment: they can’t say no to a person with type 1 diabetes in front of them who’s saying, ‘I’m using this system can you still look after me?’ ”
“That is our duty to look after them – I won’t say no – but at the same time what is our responsibility?”
What should a doctor do?
As someone with expertise in artificial pancreas systems, Thabit is willing to support patients using a DIY version. But, he says, “I make sure that first of all the individuals who are doing this, to put it simply, have a good head on their shoulders, so they are reasonable, they know exactly what they’re going into.”
He stresses that it is also important to make patients aware of the limitations of closed-loop delivery, such as the risk of insulin stacking, leading to hypoglycemia, if delivery is too rapid. Like Wilding, Thabit uses the analogy of the driverless car. “These are devices which are not infallible, so any pumps, any [continuous glucose monitoring device], any, even approved, closed-loop systems can fail, so the individual needs to know when they can switch off the system, go back to pens and then manage it themselves.”
He also advises patients to manage their insulin manually when they are poorly or stressed: “Those kind of physiological responses can be a bit too much for even the best control algorithm because of the limitations we still have with the insulin kinetics.”
“It’s almost like a guidance of when not to do things rather than how to do it,” he says.
Although broadly supportive of DIY closed-looping, Thabit would not actively encourage interested patients to adopt it. Rather he would guide them towards approved diabetes technology, “mainly because I know the evidence behind it – we’ve seen the data.”
But he stresses the importance of managing patients’ expectations. “I don't want them to think that this is going to be taking away their diabetes. It still involves a lot of work, and I always have a very honest conversation before I start any form of technology with a person with type 1 diabetes.”
He adds: “Diabetes technology in the past has had problems with that, where the expectation, the hype, was basically too much, that when people saw the reality they were extremely disappointed.”
Playing regulatory catch-up
Many of the above-mentioned problems – safety issues, accessibility, liability – would be solved or at least greatly reduced if DIY closed-looping could be brought into the regulatory framework.
But the regulatory system is built around the approval of medications, so focuses on exhaustive efficacy and safety testing of a molecule that will then remain unchanged. Technology, by contrast, does not carry the risk for surprise serious adverse effects or drug interactions, and the software is constantly tweaked to improve performance and add features.
“How do you tie that into a regulatory process which is used to: this is the regulated version and if you want another one, you have to come back?” asks Rachel Connor, the JDRF’s Director of Research Partnerships.
The charity has become actively involved with this problem partly because of the importance of the artificial pancreas to people with type 1 diabetes, but also, says Connor, because “there is a lot of uncertainty for the clinical care community, and how to interact with the DIY community.”
She says: “If they don’t have the kind of information and the kind of framework that they are used to working in, of course there is more uncertainty there. So to move it forward we need to reduce those uncertainties.”
In response, the JDRF announced the launch of its Open-Protocol Automated Insulin Delivery System Initiative, in 2017. One of the initiative’s aims is to persuade device manufacturers to use open-protocol communication systems as standard.
Right now, we are playing catch-up.
Connor says the initiative “came about because there is a recognition that not just the DIYers but lots of other people are innovating with how you can take that sensor data and use it to drive changes in insulin dosing. But the way the protocols are currently locked down to devices means that innovation space is pretty much blocked.”
As well as allowing more patients to benefit from this innovation, standardizing open-protocol communication systems will mean that patients can choose which devices they wish to connect into an open- or closed-loop system, rather than having to rely on specific hard-to-find or out-of-warranty devices.
Sanjoy Dutta, Assistant Vice President of Research & International Partnerships for the JDRF, also emphasizes that of the current commercial algorithms – whether approved or in development – “even if they are not one size fits all, it is probably three sizes fits all or something like that.”
So “one big advantage of doing this is there is a level of personalization, personalized medicine that we can deliver to open protocol through writing these cross-talking algorithms, the DIY algorithms.”
The JDRF initiative also aims to work toward an open-protocol regulatory pathway, although Dutta concedes that, in a sense, “that ship has sailed,” given the number of people already using DIY closed-loop. “So right now, we are playing catch-up,” he says, stressing that this is primarily about patient safety.
This initiative is the second layer of an envisaged three-tiered framework for improving diabetes technology approvals.
The first tier and third tiers both directly involve the regulators, whom Dutta describes as “so embracing and helpful right now.” The first involves a simplified pathway for diabetes technology approvals, a tangible result of which was the FDA approval of the MiniMed 670G system, which was based on an open-label study, with no control group, conducted in 124 patients over 3 months.
The third tier, explains Dutta, will cover elements such as “miniaturized pump, ultra-rapid acting insulin, fully closed-loop system – maybe a second hormone because in type 1 people, the insulin is not the only thing that is messed up.”
Can DIY be brought into the regulatory framework?
So if the ship has already sailed, can the DIY community be persuaded into the regulatory framework? Is this even a realistic goal for something born partly of people’s frustrations with the slow-moving approvals process, and which is frequently improved and updated?
“I suppose one thing is actually these software technologies are maturing now,” observes Connor.
She suggests that although the DIY algorithms are still undergoing constant development, they have now reached a stage when “people who aren’t the uber-users” would probably be content using a version that remained unchanged for 6 months or more, meaning that “perhaps a broader population of people with type 1 could start accessing some of this innovation.”
Connor adds: “I think this is something that the regulators need to get on top of, but also the people developing these systems, probably to get comfortable with maybe a compromise of: How long can we sit with this, which isn’t necessarily as perfect as we could make it today, but can we live with it until we have got a significantly more perfect one and take it through another framework?”
Dutta points out that the recently agreed approval process for artificial pancreas systems is much faster than that for medications, because there is no requirement for animal studies. “There is an in silico modelling and simulation algorithm with thousands of patients’ data in there that is FDA approved,” he says. “You have to run it through this to prove what the outcomes are going to be and it can jump into a clinical study.”
But even with this slimmed-down process, the costs of formal clinical trials are likely beyond the reach of the DIY community. There, again, charities such as the JDRF can have a role; Connor stresses that the majority of their money comes from the diabetes community, motivating them to make funding decisions based partly on the wishes of that community, as well as on the most promising science.
There is always going to be someone doing something in their garden shed, which is not quite what is happening in a framework.
And she says: “JDRF has a track record of working with literally anyone who can advance things for people with type 1 diabetes, so we are completely agnostic to the source of innovation.”
She concedes that the funding application process for DIY innovators is less well established than for academic and industry researchers. “I think it is fair to say a lot of funders are slightly working out how to do this.
“But at the same time our goal is to get things going for people with type 1, so that shouldn’t be a barrier to our community interacting with JDRF and seeing how we can move things forward.”
However, many people, both within and outside of the DIY community, doubt that DIY closed-looping will ever fully fall under the regulatory umbrella, instead envisaging two streams: one for larger, less frequent algorithm updates that arrive via accepted approval pathways and one comprising frequent updates and innovations for the “uber-users” who will be – still – not waiting.
Connor is philosophical about this. “That is the case in every field, whether you are talking about hearing loss or limb replacement technology,” she says. “There is always going to be someone doing something in their garden shed, which is not quite what is happening in a framework.
“I think it is about making the broadest church possible, which can allow more people to benefit from the innovations, feel safe in doing so, and allow everyone to move forward together.”
medwireNews is an independent medical news service provided by Springer Healthcare. © 2019 Springer Healthcare part of the Springer Nature group