The artificial pancreas: What’s on the horizon?
And will a fully closed-loop system ever be a reality?
Closed-loop insulin delivery is a rapidly expanding field. At the time of writing in July 2019 no further systems have come onto the market since Medtronic’s MiniMed 670G was approved by the US FDA in September 2016. However, a plethora of artificial pancreas systems and academic algorithms, not to mention open-source algorithms, are in advanced stages of testing and expected to enter the market over the next few years.
What’s in the pipeline?
The company thought to be closest to seeking FDA approval is Tandem, with a system based on its own t:slim X2 insulin pump and the Control-IQ algorithm developed at the University of Virginia. As reported at this year’s ADA Scientific Sessions, the system achieved good results in the iDCL trial, using the Dexcom G6 continuous glucose monitoring (CGM) device.
“All the information we have heard indicates that they are going to be hitting the US market, I think, by the end of the year,” says Dan Finan, Research Director for the JDRF.
The main systems on the horizon for the US market are as follows:
The Beta Bionics Gen3 iLet system, based on the algorithm developed at the University of Boston, features a pump that can operate in an insulin-only, glucagon-only, or bihormonal mode and works with either Dexcom or Sensomics CGM devices. The company is planning a pivotal phase III trial in 2020, testing the bihormonal configuration with dasiglucagon.
Bigfoot Biomedical has a system based on an algorithm developed by one of the company’s founders, and is preparing for a pivotal trial in the USA.
Insulet’s Omnipod Horizon system features the company’s own algorithm, its tubeless insulin pump, and a Dexcom CMG device.
The company DreaMed Diabetes arose from the MD Logic algorithm developed by the team of Moshe Phillip at Schneider Children’s Medical Center in Israel. The company markets its Advisor treatment decision support software, but has entered into collaboration with Medtronic to further develop its artificial pancreas aspirations, with Medtronic expected to produce future systems based on the MD Logic algorithm.
An open-source algorithm enters the regulatory field
As medwireNews previously reported, the JDRF has had a key role in working with the regulatory authorities to adapt the approvals process, which was designed around pharmaceuticals, to accommodate the demands of rapidly iterating medical devices. And this extends to opening avenues for regulatory approval to the DIY community – people who have created open-source control algorithms that work with commercial CGM devices and insulin pumps.
“So the hot-off-the-press news is that, maybe 6 months ago, JDRF and also the Helmsley Charitable Trust announced a significant amount of funding to go towards the development of one of these DIY systems, the one known as Loop, to have another non-profit – Tidepool – develop this professionally,” says Finan.
Ultimately, the algorithm, which runs on the iPhone, will become available through the App Store, and will work with a range of commercially available, in-warranty pumps and CGM devices. At the time of writing, Tidepool has announced three collaborations, so the algorithm will work with the Omnipod and Medtronic insulin pumps, with the Dexcom G6 CGM device, at the point of launch.
“So what we will have soon, in the US market anyway, is this handful of four or five holistic solutions for a hybrid closed-loop control product, but then we will also have a number of combinations for mix and match components,” Finan summarizes.
Other systems and algorithms undergoing testing around the world include a bihormonal system from Inreda, based in the Netherlands, a system from the France-based Diabeloop, and an algorithm from the US-based Dose Safety. The majority of systems use components from multiple manufacturers, with Dexcom CGMs particularly pervasive.
The best algorithm
Finan also highlights “what some people consider to be the best algorithm out there”– the one developed at the University of Cambridge, UK, by the team of Roman Hovorka, whom Finan describes as “a very, very good friend of people with type 1 diabetes.”
The algorithm has been in development for many years, and has been tested in multiple situations and groups of patients; for example, in very young children, pregnant women, noncritical care inpatients with type 2 diabetes, and inpatients receiving nutritional support (these latter two using a fully closed-loop system). And the team has recently published the protocol for a trial assessing a hybrid system over a full 6 months of use in children and adolescents.
Finan notes that Hovorka is “a mathematician by training” and says that “he was probably – looking back on it – a visionary in terms of what you can do with data and how you can train an algorithm to learn, to adapt to changing circumstances in terms of physiology in the patient.”
“It is an understatement,” Finan adds, to say that Hovorka is well aware of the possibilities for getting his algorithm through the regulatory process and onto the market.
Market and user factors
The sheer number of competing systems is aptly illustrated by the fact that the France/UK-based Cellnovo went into administration and ceased production of its artificial pancreas system components earlier this year, citing “the competitive environment and the challenges of rolling-out a breakthrough system at a sustainable cost.”
This problem is exacerbated by the fairly small size of the market for which these companies are competing. Starting with the relatively low absolute number of people with type 1 diabetes, the number is whittled down further when you account for the proportion who can access the technology, have a healthcare professional who can support them with it, are comfortable using it, have the motivation and ability to manage it, and can afford it.
Finan notes that “there is more and more chatter” about how people with type 2 diabetes could benefit from diabetes technologies, including the artificial pancreas if they are insulin-dependent, but concedes that “the fact of the matter is the penetration is pretty low, even just among type 1s.”
So in a future scenario of multiple systems competing for a relatively small market share, how will healthcare professionals know which one to recommend to their patients? Should these systems go head-to-head?
This would be the ideal for diabetes pharmaceutical agents, but Finan thinks it unlikely in the short term for artificial pancreas systems, if only because the huge variability in components, insulins, and preferred outcome measurements make meaningful comparisons extremely challenging.
“The key is that we want the market to work it out on its own,” he says. “We want to control [people’s blood glucose] to a very, very tight range as much of the time as possible, but at the same time, that is not everything.”
He stresses that the form factor – the physical shape, size, and appearance of the devices – is a major issue. “The little delighters are huge. There are companies out there that make really colorful pumps. That might seem like it is not the right priority, but then again who are we to say?”
Indeed, he sees increasing variety, in both the form factor and the algorithms themselves, as key to expanding the use of artificial pancreas systems.
Although Finan believes that competition between the system designers should continue to drive overall improvements, “at the same time, even the algorithms have some differentiating factors, like, maybe there is one algorithm that is a little bit better for exercise. Maybe there is an algorithm that is – and we are interested in this – a design more for the older population who has a different relationship with technology, or something like that. An algorithm for teens who are notorious – and I don’t blame them – but they are notorious for lousy control and maybe there is something we can do.”
The aim, he says, is to “use advanced mathematical techniques to our advantage,” to ascertain whether it is better to have specific algorithms for specific subpopulations or if an algorithm can be designed to more effectively learn and adapt to varying individual needs.
Will the loop fully close?
The holy grail of artificial pancreas systems is, of course, a fully closed-loop system, which does not require the user to bolus for food or announce exercise. Finan notes that the definition of a fully closed-loop system is somewhat in the eye of the beholder, given that current systems will continue to adjust the insulin infusion rate if users forget to bolus. However, they are not designed to cope with such events and so the outcomes will be suboptimal.
“It is the fact that the current formulation of insulin is just too slow when subcutaneously delivered to catch up” with the insulin demands created by food intake, says Finan. Added to which is the fact that CGM devices measure glucose in the interstitial fluid and so their estimated blood glucose values lag behind the current blood glucose levels and they become less accurate in the hypoglycemic range.
But he says: “There are definitely avenues that we are pursuing to get what we would consider more robust, closer to ideal glucose control without the input from patients from meals and so on.”
Faster-acting insulins are in development, and researchers are exploring other options, such as pramlintide, an amylin analog, which helps to lessen the postprandial blood glucose increase; and glucagon, which Finan says “is not really going to help your meal exclusions per se, but you could be a lot more aggressive with insulin, because you have that robust safety net.”
A currently more speculative approach is to deliver insulin directly to the liver – a much more physiological approach than subcutaneous delivery. Finan notes that this is still in earlier stages of development and there are currently “a lot of problems with it,” but he sees it as part of a package of incremental gains that, along with improvements to the algorithms, will eventually lead to something much closer to a fully closed-loop insulin delivery system.
Finan says that the field was “kind of slow moving for a little while,” but with the FDA now facilitating the approvals process, “it is really ramping up in terms of all the companies being competitive with each other – and that was before the open protocol stuff.”
He adds: “There should be this very interesting dynamic to see how that sort of model takes hold in people that either have already chosen these devices, or maybe now are considering it because they get to mix and match.
“We will see how it plays out, but I think in general the device space is in a pretty good place right now. We will keep driving towards better products, better technology, but there is good momentum.”
medwireNews is an independent medical news service provided by Springer Healthcare. © 2019 Springer Healthcare part of the Springer Nature group