I, medical device robotic pancreas

Third-grader Andrew Calabrese carries his backpack everywhere he goes at his San Diego-area school. His backpack isn’t just filled with books, it is carrying his robotic pancreas.

The device, long considered the Holy Grail of Type 1 diabetes technology, wasn’t constructed by a medical-device company. It hasn’t been approved by regulators.

It was put together by his father.

Jason Calabrese, a software engineer, followed instructions that had been shared online to hack an old insulin pump so it could automatically dose the hormone in response to his son’s blood-sugar levels. Mr. Calabrese got the approval of Andrew’s doctor for his son to take the home-built device to school.

The Calabreses aren’t alone. More than 50 people have soldered, tinkered and written software to make such devices for themselves or their children. The systems—known in the industry as artificial pancreases or closed loop systems—have been studied for decades, but improvements to sensor technology for real-time glucose monitoring have made them possible.

The Food and Drug Administration has made approving such devices a priority and several companies are working on them. But the yearslong process of commercial development and regulatory approval is longer than many patients want, and some are technologically savvy enough to do it on their own.

Here is the Kate Linebaugh story, interesting throughout, via Adam Thierer and Eli Dourado.


With the regulatory and liability costs involved, I wonder what the FDA approved devices will cost?

Great post. As a physician, I would comment that this is potentially very dangerous. Small overdoses of insulin can easily kill someone. Receiving lower than the proper dose of insulin in a type 1 diabetic can also lead to death in as short as a few hours. I'm not sure I'd hand over the monitoring of blood sugar and administration of insulin to pcb board and instruction set I downloaded off the internet. If this kid goes into a coma due to a glitch the father will be charge.

Would the monitoring and administration be better if it was done by a 9-year old? What about in the middle of the night?

From the webpage of the OpenAPS project https://openaps.org/

How does OpenAPS work?

OpenAPS is a simplified Artificial Pancreas System (APS) designed to automatically adjust an insulin pump’s basal insulin delivery to keep blood glucose (BG) in a safe range overnight and between meals. It does this by communicating with an insulin pump to obtain details of all recent insulin dosing (basal and boluses), by communicating with a Continuous Glucose Monitor (CGM) to obtain current and recent BG estimates, and by issuing commands to the insulin pump to adjust temporary basal rates as needed. It follows the same basic diabetes math that a person would do to calculate a needed adjustment to their BG – but it’s automated and precise in it’s measurements.

Scott - the point is real people are too dumb to do medicine. That's why we need bureaucrats and politicians to decide for us. \sarc off

Modern micropumps are pretty accurate, and if the kid dies in a bubble at home, is he any better off? Jehovah's Witnesses and children needing blood transfusions also come to mind.

Bonus trivia: the mechanical tube to relieve encephalitis and/or swelling of the brain was invented by a mechanical engineer who had an infant daughter with the condition, and L. Pasteur experimented on himself, as was common back in the day.

He suggested experiments on Brazilian prisoners, but the Emperor, his penpal, denied permission.

You'd probably have him monitored by a physician with a full team of staff 24/7. Can't be too careful. Anything less would be negligence and should subject the parents to serious penalties. How can you put a price on a child's life, right?

What was that about medical errors being the third leading cause of deaths in the United States? Maybe we should get a little more serious about criminal charges for negligent medical workers.

No, but I would never allow my child to have insulin administered via a program I downloaded off some random personal webpage. I'm not saying the FDA is fine just the way it is, but administering insulin through devices and software you put together from Amazon and Sourceforge is a tragedy waiting to happen.

A tragedy, or a triumph?

I respect the fact that you're a clinician, but as a type 1 diabetic adult who has kept up to date with this work, you can't be a moron and implement this. I *wish* I had someone savvy enough to do this with. You don't just cobble this together from Amazon and Sourceforge. Obviously.

Speaking as a type 1 diabetic, can I ask:
If I were normally controlled, taking the right amount of insulin, but then started taking less, what plausible mechanism could there be that would lead to death in a "few hours" after I stopped the right regime? I guess by plausible, I mean, not entirely contrived, and not for abnormal values of "few" or "could"? In 50 years I've forgotten my shots _entirely_ on a handful of occasions, and yes touched 500 mg/dL as a result, which will reduce my life expectency, but is there any data to show I had a decent chance of actually dying there and then? Imagine I stop taking insulin, gorge myself on glucose until I choke, still ... "a few hours". Not saying you are wrong (you are a physician) but is there any research (or even anecdotes!) to support "that" in reality. By "that" I mean a well controlled diabetic dying a few hours because she took less insulin than needed a few hours ago (prior to which all was well.) This is so counter to everything I know, well, I don't want to call you a liar, but is there anything resembling evidence for this?

Can you give specific instances of death from under-dose in a few hours? As a T1D, I have never heard of this possibility, so I am both skeptical and concerned.

As a type one diabetic for over 35 years I would prefer to have the right to make these decisions for myself.

+1 (although only for 24 years)

+1 and T1D for 7 years (adult-onset).

Hack away!

What's stopping you?

Implied in my statement, perhaps inadequately so, was the claim that I should also have the right to pay someone who is better than I at doing so to build one for me.

I have been a computer programmer for medical systems with FDA review and approval. It was a very conservative environment and one with less ego and wild ideas than general programming. The culture was very much about not making mistakes, not killing anyone.

Programmers in other domains, especially fast paced competitive environments are amazed that our code freezes lasted six months. That is six months of no new features, only testing and defect-only fixes.

IMO that is the approach you need to not kill somebody as the cost of progress. If you are willing to kill people, you should not be let near patients, especially kids.

But sadly Dunning–Kruger has to rear its head. I am sure many taking the risk do not understand the risk.

When "not killing somebody" results in even more deaths I'm not so sure "If you are willing to kill people, you should not be let near patients" is so black and white.

Exactly... I can imagine that the FDA is not in an easy position here but still, there's got to be a way to allow patients to make the final call on the risk vs reward problem.

How should I really respond to a dataless response that inaction is the greater risk?

I mean Dunning-Krugers can +1 it till the cows come home, it still won't be an argument.

Confirmation biasers can confirmation bias until the cow comes home, but everyone else will see that their job depends on holding the confirmation bias and will know they can't be trusted

There are two risks in any new treatment. The first is the background risk for the condition and old treatment The second is the risk from the new treatment.

The actual doctor above says he would comment that this new treatment is potentially very dangerous.

We could compare the two, if we had two. Is the normal treatment path also "very dangerous?"

Obviously, yes. The guy says too much or too little insulin is very dangerous. The only question is whether the software or the 9 year old is more accurate

What do you call your cognitive bias?

Experience in the domain.

It's called confirmation bias.

So you would not ask a soil engineer about the risk of a retaining wall?

+1. The FDA is not perfect, but it is hardly the dam of progress it often gets portrayed as. For every apparent fumble (such as the FDA's recent decision on the Duchenne's drug), they prevent plenty of snake oil.

No one is disputing that the FDA prevents snake oil, the dispute is that they are so slow, bureaucratic and cautious that they fail to approve devices and drugs in a timely manner and that they drive the costs up.

I think we can expect that any field deployable medical device is going to be built in multiples and then spend a lot of time in environmental chambers.

We know it works in preliminary testing, now we have simulate walking out of the house and into the snow, trip-fall g-shock, into the warm school. Summer testing as well.

I would want months of that testing before attaching it to my blood stream. YMMV.

"Dunning-Kruger" as a noun is an idiot term used by idiots. Are you an idiot or just trying really hard to look like one? Obviously you have no clue what that study says.

Hmm. Are you trying to win this argument without sharing any knowledge of medical electronics or their risks whatsoever?

Surely syntax or semantics are better than 12 years in development labs.

Who is DK?

Hard to imagine a better advertisement for the criminal immorality of the FDA than this.

1) In the first place, its not an 'artificial pancreas'. The pancreas has dual roles, both as an organ of the digestive system [producing proteases, lipases, amylases for the duodenum] and the endocrine system [insulin, glucagon and a new hormone only recently identified, aprosin https://www.sciencedaily.com/releases/2016/04/160414143904.htm]

2) "Jason Calabrese, a software engineer, followed instructions that had been shared online to hack an old insulin pump" [snip]. So why not call it what it is? A home brew insulin pump.

3) "Hard to imagine a better advertisement for the criminal immorality of the FDA than this" Once again, Tabarrok trolling the mouth breathers.

All he has to do is mention the 'three letter acronym'.

4) If young Andrew dies of insulin shock, will father Jason be arraigned for manslaughter?

Why in the world would the father be guilty of a crime if something goes wrong? We don't charge doctors with manslaughter when something bad happens, they're only subject to civil liability. The only case where criminal charges would make sense is if the device is modified in order to cause harm but that's obviously not the scenario in view.

I guess you didn't Google to see if doctors are charged with manslaughter in cases of gross negligence

Manslaughter for negligence, practicing medicine without a license? If a father attempts to remove his son's lipoma with his home surgery kit and youtube instruction videos and his kid gets septic and dies, should the father not be arrested?

Why so protective of physicians, C D? Isn't it just as bad if a properly licensed doctor negligently screws up and kills someone? Shouldn't that doctor be arrested for negligent manslaughter?

Surgery != Automation of Insulin Injections.

Self monitoring & injection is commonplace and necessary to deal with Type 1 Diabetes - even by minors and their parents.

Self Surgery is not common. It's not even close.

The status-quo state of self-injection has notable flaws. Insulin shock is a serious thing that does kill.
Waiting for the FDA Gods to bless the perfect computer / pump / monitor trio has costs that you are blind to.

All physicians should be strung up to begin with because of the immoral union/monopoly power they use to hold us all in debt to them. But aside from that when they fuck up it really should be criminal since they got all of dem degrees proving how competent they is. So if they do something incompetent it was obviously intentionally malicious.

Your link in number one doesn't work. Anyway regardless of the other functions the pancreas plays, the term 'artificial pancreas' to refer to closed-loop control of blood glucose in diabetics is often used in scholarly journals. Search for the term in pubmed or scholar google.


'artificial pancreas': redolent of 21stC Bull Shit Marketing. It does, like, one [1] thing out of doubtless thousands the [real] pancreas executes every day.

Like calling a Briggs & Stratton 2 cycle lawnmower a Ferrari Daytona

Of interest on the topic: http://blogs.plos.org/dnascience/2016/05/12/gene-therapy-for-type-1-diabetes-preclinical-promise/

"This device has performed an illegal operation, & will now shut down"

"Malfunction 54" is quite famous in the medical programming community.

I recommend The Incidental Economist for those who wish to read about the intersection of economics and medicine.

How economics shows how corrupt and evil practitioners of medicine are?

It’s very a nice and useful piece of posts. Iam glad that you
shared this helpful information with us. Please keep us up to date like this.
Thanks for sharing. jual bibit bunga

As the co-creator of OpenAPS and a long-time MR reader, I thought I'd weigh in with some comments on the discussion to date. Tyler/Alex, feel free to use this in a "from the comments" post, or LMK if you're interested in exploring this topic more deeply.

Type 1 Diabetes is a unique disease, in that it is perhaps the only disease where patients (often young children, but T1D can be diagnosed at any age) or their parents are sent home with lethal quantities of a life-saving medication and asked to properly calculate the "right" dose dozens of times a day. As a person living with T1D, you know that calculating that dose incorrectly can put you into a coma or even kill you if you do it wrong and don't take corrective action in time. But you also know that it is impossible to stay awake 24/7, and that despite doing everything humanly possible, "dead in bed" syndrome is still a significant risk. Many people living with T1D have experienced at least one severe hypoglycemic episode, which means their blood sugar was so low they required assistance, and/or experienced seizures or unconsciousness.

Current FDA-approved insulin pumps routinely overdose insulin, because they are preprogrammed to continue delivering basal insulin regardless of blood sugar. It is far safer to stop overdosing a lethal drug when blood sugar is low and dropping, but unfortunately there are no FDA-approved pumps that do that well. In light of that, we and over 50 other individuals have now decided (after literally years of development and testing) that the safest thing to do for our own safety or that of our loved ones is to build our own systems to instruct our FDA-approved insulin pumps to discontinue insulin when blood sugar is too low or falling too fast, and to administer a slightly higher rate of insulin when blood sugar is too high or rising too fast.

To do so safely, those using OpenAPS (the DIY system described in the referenced article) follow a number of design principles that ensure that even if we make the worst possible series of mistakes, the OpenAPS system is not capable of administering a life-threatening quantity of insulin. For example, all dosing commands sent to the pump are sent as a temporary infusion rates over a period of time rather than a single quantity all at once, and the maximum allowable rate is hard-coded, on the insulin pump itself as well as in our code, at a rate that is too low to do serious harm, even if administered for an extended period of time. Also, we have designed OpenAPS to fail safely: any failure of the system means that the insulin pump reverts to its normally scheduled insulin delivery, and the patient is no worse off than they would've been without an OpenAPS system.

As a result of the unique perspective that people living with Type 1 Diabetes have on reducing the *overall* risk to the patient (not just the risk from action, but the risk from inaction as well), we have been able to design and build "artificial pancreas" systems (APSs) that are far safer than any alternative available today, and which appear to actually be more effective than the first systems that will come to market in 2017. This also helps ensure that people with diabetes will hopefully always have a choice as to selecting the right combination of devices (or lack thereof) to help them best manage their lives with diabetes.

As any long-time reader of MR will know, there have been countless examples of the FDA's slowness to approve new drugs resulting in more harm than good. The fact that obviously safer insulin pump systems with limited APS technology have been approved in Europe but not the US is further evidence of this thesis, but there are other events to make note of that are perhaps more encouraging. Since we started working on OpenAPS and its predecessors, the FDA's center for devices has proven extremely willing to engage constructively with both medical device makers and patient groups like ourselves (and the Nightscout project community) to speed up approval of devices that improve patient safety. They still require detailed clinical trial data from companies wishing to market and distribute APS devices to the general public, but they also appear to recognize that projects like OpenAPS serve a valuable role in advancing real-world research and development of the technology. As a result of that, and the fact that the FDA does not have authority to regulate patients' private off-label use of FDA-approved medical devices, the FDA has been remarkably restrained in their reaction to projects like OpenAPS, while engaging constructively to ensure that we are all working toward the common goal of making T1D treatments safer for patients in every way we can. By showing such restraint, they have indeed shown a way that patients can make the final call on how to deal with the risks of action and inaction that they have to live with every day.

We are definitely taking many small steps toward a much better world.

Very good info. Lucky me I ran across your blog by
chance (stumbleupon). I've saved as a favorite for later!

It's an amazing paragraph in support of all the internet viewers; they will take benefit
from it I am sure.

If you would like to increase your know-how just keep visiting this website and be updated with the hottest news
posted here.

I do not know if it's just me or if everyone else experiencing issues with your blog.

It appears as if some of the text in your posts are running off the screen. Can someone else please provide feedback and let me know if this is happening to them too?
This could be a issue with my browser because I've had this happen previously.

Comments for this post are closed