• Guest - w'd love to know what you think about the forum! Take the 2025 Survey »

'Artificial Pancreas'

Some interesting points raised in this topic.

On the open projects front, it's now possible to run an Open Source AP software set on a mobile phone speaking directly to a pump with no widgets involved. And that will soon be released for the Roche Combo pump, which would make it massively more available in the UK and Europe.

@first14808 - you seem very cynical about this stuff, and yet most of the T1s using both the open systems and medtronics 670G in the US seem to be mostly very impressed with the capabilities and the hands off effects that they provide, which is where they really come into their own. The growth in number of users of the systems over the past two years has been really quite something, and a lack of availability and flexibility of commercial systems continues to drive this take up. In addition, CGMs take readings every 5 minutes at the most and 1 minute at the least. With the current state of the insulins available, you don't need anything more than this, as as @ringi mentioned, half life is your biggest issue, and that's driven by administration technique more than anything else. Until we can administer insulin intravenously, we're likely to be talking between 4 and 8 times the half life of human insulin from the pancreas.

@ringi - the risk of DKA is already there for people using pumps and is not something that is increased by using an AP system. The real risk of ending up in hospital is low, and certainly, adding in the low dose of long acting adds different complications in relation to exercise and IOB that can't be handled completely with an AP's ability to shut off insulin (in current iterations that are single hormone). People talk about this as a risk a lot, but the reality is that it's just one of many things that can go wrong with T1D.

In addition, all the systems being built for use by commercial companies use some level of machine learning to try and interpret data and then set the settings themselves. The 670G does this from two weeks of use pre-looping, and iLet learns as you go (and results in a couple of weeks of high glucose levels as it learns as a result). Even OpenAPS takes your existing pump settings and meal and Glucose data and adjusts its settings to match you better. These are all mechanisms designed to make it less of a bind on the user to get started with these systems.

@donnellysdogs - I beg to differ on the "off the shelf"-ness of systems. Every single item in the Open Source systems is "Off the shelf", as in they need no special modifications to be used. Indeed, Roman Hovorka's systems all use off the shelf systems with added access provided by manufacturers that we as patients have had to reverse engineer.

Access to the items via the NHS is a different question and is driven entirely by cost, and not by whether the items are available, and given the evidence of the Libre, there is now way the NHS would queue up and offer them to people. Something that costs £840 a year is being pushed back on by most CCGs because it "costs too much and doesn't offer enough evidence of benefit". I'd agree though that with your site issues, there's very little an AP could do. It might work quite well with a Diaport though.

@mentat - congratulations. For many, MDI remains tough, and even fixing the underlying issue (for quite a few done by reducing carbs dramatically) doesn't always help. I find that using an APS system makes maintaining a time in range of 90+% significantly less effort, even with lower carbs!
 

Thanks! Actually I lied slightly. I'm using MDI + CGM + Afrezza. I use Afrezza sparingly due to its mind-boggling out-of-pocket price but it helps.

My "underlying issue" was undiagnosed epilepsy, which was causing various cognitive and mood disturbances (but no convulsions). Epilepsy medication has significantly improved that and also made my insulin needs fluctuate much less.
 
Wow. That's a big underlying issue to be diagnosed with! I'm pleased to hear it's helped the diabetes management, and also it goes to show just how much interaction there is between various feedback loops within the body!

Afrezza makes a huge difference on the ease of handling spikes, but agreed it is stupidly expensive, and of course, so does CGM. I guess you're probably doing some variant of sugar surfing?
 
tim2000s said:
Wow. That's a big underlying issue to be diagnosed with! I'm pleased to hear it's helped the diabetes management, and also it goes to show just how much interaction there is between various feedback loops within the body!
Click to expand...

I'll say. I'm pretty confident my T1D is what made the epilepsy develop. Lows in particular would reduce my seizure threshold and then I'd have seizures and as a result my insulin needs would jump around and I'd have more highs and lows and a vicious cycle developed.

tim2000s said:
Afrezza makes a huge difference on the ease of handling spikes, but agreed it is stupidly expensive, and of course, so does CGM. I guess you're probably doing some variant of sugar surfing?
Click to expand...

Yeah. I would describe it as "very rough carb counting with a CGM+Afrezza safety net". Due to my epilepsy, my goal is zero hypos per month and I'm pretty close!
 

Thanks for responding Tim. I always feel a bit guilty tagging you into everything relating to AP, but someone's got to blow your trumpet for you.

Thanks again.
 

Oh, I am, and 30yrs spent working in technology often has that effect. There's a lot of hype, and when I saw the BBC article, it looked like a CGM+pump combo, and not especially novel. Then I saw your blog & that you'd been doing the same thing, in the best hacker fashion, and some of your challenges.

But for me, an 'artificial pancreas' should be a neat package users can install and mostly forget about, other than topping up it's insulin tank when prompted. Current systems don't seem to offer that, or interoperability. Vendors like that because it locks users into buying their consumables. That can result in higher costs, and sometimes supply chain issues. I've seen situations where critical components have shortages due to fires, floods, quakes etc disrupting factories.

That's also where the work you and others are doing with the 'Open' APS is great, especially if you can lobby government and big customers like the NHS or US insurers to support it. That would probably mean going the standards route, ie relevant devices having published data exchange formats to support interoperability and development of 3rd party stuff.. Which would get some pushback on safety grounds, or just by vendors who prefer to lock users in. And having worked on standards in the past, there can be an element of time pressure, ie the market & bureaucracy are often resistant to change. That can often lead to inferior standards becoming the official ones.
 
All your points are valid, but also miss the key aspect of living with T1D, which is that the freedom from constantly thinking about it that these systems bring, the fact that many people now feel comfortable sleeping overnight instead of waking up twice to test due to fear of hypos, etc, the improvement in management that they provide without having to spend significant amounts of our time doing it comes in to play.

Very few people can comfortably live with T1 and not end up concerned about either highs or lows, and not have to do anything to deal with it. It is, unfortunately, a condition driven by numbers, and they can be hard to handle.

Whilst the idea that having a nice "easy install" would be the best for end users, we also know how much work is involved in getting these things in place. Simply moving from a pump, to "a pump with enough technology to mean I don't have to ever bolus for meals and still achieve 80% time in range" is truly remarkable. The term "Artificial Pancreas" has been overused, but to be able to have no additional maintenance overhead compared to a pump, but much less input and need to do anything, really is a game changer.

No-one in the open source world is looking to have the platforms formally adopted by anyone. That's a huge leap and as you say, comes with a regulatory burden that is horrific. It's why they all are created with personal liability of n=1. What's perhaps more unexpected is that the work that has gone on by PWD to build our own systems has informed the FDA about the direction they needed to go with the medtech companies and as a result has seen commercial offerings come to market more quickly. It has also driven charities like the JDRF to recognise that there is a strong "maker" community of intelligent people with significant amounts of expertise in Diabetes and managing it that would benefit from open protocols and as a result they launched an initiative around this.

We're not trying to get adopted by the establishment, but we are trying to change the way the establishment thinks about the problems that PWD face and make life easier for ourselves and others.
 
first14808 said:
Oh, I am, and 30yrs spent working in technology often has that effect.
Click to expand...

I'm with you regarding vendor lock-in and the need for open standards. My health relies upon the fact that someone has cracked the authentication algorithm used by the Dexcom G5, so that I can use an app I built which offers more than the vendor one does.

first14808 said:
There's a lot of hype, and when I saw the BBC article, it looked like a CGM+pump combo, and not especially novel.
Click to expand...

I also used to get annoyed when virtually every article about an incremental improvement in AP technology grabbed headlines with "world's first artificial pancreas".

I got over it when I realised that most of these articles are actually quite informative and help the general public understand T1D and increase the chances that the researchers will get more funding etc.


I think this is exactly the way to go with regards to monitoring and reporting - e.g. displaying CGM data on a smartwatch or preparing reports for an Endo. Anyone should be able to write an app either for personal use or for wider distribution; it shouldn't be hard to get one's own health data, preferably in real-time.

But it doesn't make as much sense with a closed-loop solution. In a closed loop, the insulin delivery algorithm needs to make assumptions about the characteristics of the sensor and the whole system needs to be tested thoroughly with all the likely failure cases. The more that an AP does instead of the person, the less skilled the person will become at noticing and diagnosing problems. I'm all for making "n=1" solutions easier, though; in fact I would love for there to be a wireless pump programming standard.
 
rob_tech said:
To me this shows that the pressure applied by the people using Openaps etc has made Medtronic develop these new insulin pumps, because they see that this is wanted.
Click to expand...

I don't think it's that simple. Closed loop is simply the next logical step in insulin pump technology. It's a no-brainer that people with type 1 will jump on even a 25% reduction in effort or interrupted nights. So vendors want to provide this tech. But the approval process is a nightmare so they have been taking incremental steps. OpenAPS and Loop have accelerated things by proving to non-technical audiences that the technology is there, and the only actual barrier is the approvals process.
 
rob_tech said:
But how come it is only now that these things have happened.
Click to expand...

This is just my speculation:

1. CGMs are becoming more accurate and lasting longer
2. Wider adoption of CGMs in the USA because of improving insurance coverage
3. Yes, OpenAPS has played a significant role, and should be applauded. I also think Nightscout and their #WeAreNotWaiting message has played a big part too.
 
rob_tech said:
due to a leak from a Medtronic tech who supplied details of open communication channels to these pumps that would allow a certain amount of control of the pump, using small pi computers.
Click to expand...
Not quite true. A guy in the US reverse engineered the protocol using a combination of the carelink stick and a radio man in the middle in order to better capture the data from the pump. It took him five years. That was Ben West. Scott and Dana then worked with him to use it to send instructions to the pump after Dana had issues with overnight hypos. They built that in about 2014.

Nate Raclyeft (who started on OpenAPS) then built Loop with Pete Schwamb in 2016 to provide an iPhone based system that was enabled by the introduction of the Dexcom G5. Again, they reverse engineered the comms between the G5 and the iPhone in order to make it work. Meanwhile, in Germany, Android APS was being built using xDrip as the CGM data source and with direct connectivity to the Sooil Dana R pump - which had an open Bluetooth implementation that could be easily accessed.

Medtronic have had CGM sensors (all be it only recently have they been any good) for a long time, and have apparently worked on APS algorithms for around ten years (they were one of the first applicants for JDRF funding into "Artificial Pancreases"). One of the biggest issues they faced was the quality of the CGM data.

rob_tech said:
However they watched the work done in Openaps and then produced the minimed 640g, which uses a cgm and smart guard to monitor and alter basal levels if blood glucose goes below certain levels.
Click to expand...
No, they didn't. The Medtronic Veos have had suspend on low for a long time (these old x22 pumps could do that) and the PLGS function in the 640G was the logical extension and has been around for longer than OpenAPS has been released. The issue for most companies is that the US regulation is not simply "It does what it says" but also "prove that it does no harm", which is a much higher level of proof.


There's a fairly well established rule of thumb that a CGM sensor needs to return a MARD of less than 10% in order to be safely dosed from, and the reality is that only the Dexcom G4 with 505 algorithm, G5 and Medtronic Guardian 3 have routinely reached that level of performance in day to day life, or in other words, CGM has only been good enough for true artificial pancreas development for about 3-4 years. Adoption of CGM in the US, while wider than Europe, is still remarkably low. The main driver in the US is the alarming cost of A&E visits. Compared to Europe, it's a no brainer to run CGM in place of A&E visits.

The main role that the open source projects have played in terms of the implementation of commercial APS systems is that they have shown that a well implemented algorithm is safe, and that people are not waiting. As a result, the FDA was much less concerned about allowing the 670G to be approved.

As Scott Hanselman wrote in June 2016 (at about the time I was gathering the pieces to build my own OpenAPS instance) there were only around 150 people worldwide with DIY systems. That's now significantly more.
 
mentat said:
MDI = multiple daily injections, nothing special. By the way, I was diagnosed at 22 and I can understand what a shock it must be to you.
Click to expand...
Yes I am shocked becuse till it happened to my son....I was not aware of what exactly diabetes is. When my son told me that he thinks he has diabetes , I took it very lightly, I told him I will give him natural foods that will lower his sugar and doctors will give him some medicine like they give for fever etc........and things will be ok.
I skipped a heart beat, when I came to know that he will have to keep injecting needles and not once twice, but whole life.
I cried with pain when I came to know that now he will not eat foods that he liked and that he will have to prick needles whenever he will be hungry.....that he will not sleep peacefully as he will live with fear of low sugar........I thought it was a bad dream and soon I will wake up.........but no........it was bitter truth that I had to accept and he has accepted bravely.He is handsome , intelligent , student of physics.......and I keep asking God........why ...why should this happen to my son and to all who are dears of their parents.......tears are rolling down my eyes as I am sharing my pain here with you. But I have hope...........hope that God will help the scientists and researchers and soon they will come out with some solution to this problem. Hope is my strength .Thanks .
 

I don't miss that aspect at all, in fact quality of life was the main reason why I picked my degree. I'd family & friends in medecine and considered that, but didn't think I could handle delivering bad news.. Gadgets that improved quality of life, however..

So that's why I may seem a tad cynical. Much of this stuff is fundamental control systems engineering. So basically creating an action or set of actions from an input(s). Pacemaker's take signals from sensor leads, and generate responses to the pacing leads. Since introduction, that's gone from allowing life, to allowing pretty normal life with the current generations.

Then some control system analogies. A central heating system takes data from thermostats and triggers heater & pumps to maintain a desired temperature. Or a car would have an O2 sensor linked to an ECU that adjusts fuel/air mix based on load. This is all old-tech, although functionality and complexity's increased over time.

So an artificial pancreas is a sensor, some control logic and a pump to deliver insulin based on sensor information, and some user input for food intake. So functionally, it's pretty much the same as many other control system tasks. I appreciate the advantages, ie being able to sleep.. But to me, it seems obvious and overdue. Pumps and CGMs have been around for quite a long time, but only now it seems the dots are being joined.

No-one in the open source world is looking to have the platforms formally adopted by anyone. That's a huge leap and as you say, comes with a regulatory burden that is horrific. It's why they all are created with personal liability of n=1.
Click to expand...

That, I think is a risk, and also why it's been a struggle. Current solutions are uncoupled, even if components are from the same manufacturer. Currently you have a 'man in the loop' between sensor and action. You take your sensor data, calculate your dosage and instruct your pump to deliver it. So if that goes wrong, it's user error. That's an important principle when there's the potential for litigation, and huge financial awards. Couple the components, and you might sleep, but vendor's general counsel and insurers would be looking at the potential downside if you don't wake up.

That also affects the supply side. So if NHS supplies it, they could get sued. In the US, risk could be split between suppliers and insurers if systems are provided under medical plans.. And those costs can be huge burdens on healthcare, ie increased DOI or malpractice insurance premiums that can feed through to personal premiums. Or just limiting the Open source work, ie at the first hint of litigation, APIs or access gets shut down. And open source doesn't neccessarily mean immunity from litigation, it's just less likely if you've got no assets or insurance for lawyers to go after.

We're not trying to get adopted by the establishment, but we are trying to change the way the establishment thinks about the problems that PWD face and make life easier for ourselves and others.
Click to expand...

And I congratulate you for that!
 

Stay positive! Technology's come a very long way in my lifetime, and is still advancing at a rapid pace. Diabetes has been getting more attention, which often means more research funds. That might not mean a true artificial pancreas this year, but it shouldn't have to be that far off.

Science & Technology's also fun because there's frequently cross-over between different fields. So for example the 'War on Terror' lead to a lot of money put into R&D to counter chemical and biological threats.. which means sensors that can detect those threats. And it can sometimes be a small step to have those looking for glucose or insulin instead of toxins or pathogens. If that's possible, then glucose sensing could become more convenient and cheaper with longer lasting sensors.
 
I agree with you.
 

I don't know if you've missed the 670G from Medtronic? That is the first commercial system approved by the FDA and it does as described, within conservative bounds. Why conservative bounds? Precisely due to the nature of the above concerns. Unfortunately, when you give a bunch of people a tool that starts down the path of delivering automation, and you hamstring it, people start working around the limitations that were imposed to reduce manufacturer liability. The end result? The 670G has had way more people admitted to hospital due to severe hypoglycaemia than any of the "DIY" systems due to trying to make it achieve what people really wanted.

Whilst I agree that Open Source doesn't necessarily mean litigation free, by providing a code base/kit that anyone can download but that doesn't work until you, as an individual, build it and set it up, entering various pieces of information, it's very difficult to attribute liability to the writers of the code.
 
Part of the issue is that the people who are most likely to be given this type of system by a health provider, are the people with the worse BG control. These are often the people who are least engage and have the least skills to monitor a system. (But who cost the health provider the most in the long term unless their BG is improved.)

But the DIY open source systems are being used by the most engage and skilled people…..

Hence the FDA has to take into account with the approval process the likely skill level of the person using the system, and how likely they are to ignore any unexpected behaviour, rather the investigate it while keeping themselves safe. (Hence I expect that the approval for closed-loop systems will include the level of support, training and remote monitoring that must be provided along with the system.)

(For example, no one on this forum would ignore it if their pump stopped working, but lots of the least engage people would do nothing for a day or two. And with no long half-life insulin in their body..... One study of pumps in the NHS found more DKA in the group who were given the pumps! This study was of people with bad BG control, with half the people getting retrained on MDI, and the other half getting trained on a pump and given a pump.)

The NHS has the same issue with CGMs, in that the people who are likely to benefit most from them, are the least engaged with their health care. Hence to get real savings on complications etc, a complete package of training and support along with the CGM has first to be developer and proven to work with clinical trails. (Good results from the very motivated self-selected people on this forum, can’t be used to predict the results the NHS will get. The DVLA rules stop a basic CGM being justified on the saved cost of test strips.)
 
Last edited:
An artificial pancreas is a lot different than a CGM. Sure both can have problems. But the CGM requires that you also use test strips, usually at least twice/day (Dexcom). So if something does wrong (which is not unusual) you already have something else to use. the AP is a continuous sensor dependent insulin delivery system. So if something does wrong with that you could be in a life threatening situation. You could die in your sleep. You could go into DKA. I don't think one will be approved by our FDA in my lifetime. I really don't.
 
As mentioned, the FDA has already approved the 670G, a Hybrid Closed Loop system and Medtronic has shipped some 20,000 units in the US.

However you look at it, a commercial artificial pancreas is ostensibly a very smart pump, and the smartness (i.e. CGM and Algorithm) can be safety engineered to revert to a dumb pump when the self-tests show issues. Open source systems do similar.

If they suffer a mechanical failure/run out of insulin/etc, then it's kind of irrelevant whether it is smart or dumb. The issues are fixed the same way and the level of life threateningness is the same.


That's right, although the FDA generally doesn't view it as a package, the company providing it absolutely does and as a result each and every 670G was issued with a full training programme as that's how they mitigate the risk of misuse and cover themselves legally.

It's also why Bigfoot have come to an agreement with Abbott to use their next gen Libre product, which is supposed to be a calibration free CGM. Reduce the "moving parts" and simply make the APS a top up the insulin and change the site, with a sensor swap every 14 days. The commercial guys get that they need to make these things useable by the least able or motivated people, but in doing so they also put off the more able and motivated who want more than the commercial systems can offer.
 
tim2000s said:
If they suffer a mechanical failure/run out of insulin/etc, then it's kind of irrelevant whether it is smart or dumb. The issues are fixed the same way and the level of life threateningness is the same.
Click to expand...

The issue is that by being "smart" they are much more like to be used by the people who are least able to cope when they fail..... Like most IT systems I have worked on, the hard problems are all "people problems" not computer problems.

(If I was responsible for the design of pumps etc, I would use an easy to break encryption system (that stills stopped someone without physical access to the pump from changing the systems.) So that I had "legal coverage" if people wrote their own software, while not stopping them doing so. )
 
You must log in or register to reply here.
Menu
Log in
Register
Cookies are required to use this site. You must accept them to continue using the site. Learn More.…