How a maker with Type I diabetes led an open source project to create a free-as-in-code artificial pancreas

Originally published at:


They closed the loop. Result: an artificial pancreas that monitors blood sugar and controls the insulin pump.

Mind Blown - Bill Nye


Usually titles that begin “How a maker…” end with some new kind of machine that turns itself off.


This is an inspirational story that reveals the deep and cruel dysfunction in the US healthcare system.

We’re in the mad max future where people are hacking together their own prosthetics and medical devices because they have no other viable option.

I bought a RaspberryPi and had fun with an emulator project. Imagine entrusting your life to one, tho.


I have a friend who acts as an artificial pancreas to her two type one children. It’s a huge drain on her as is the cost of testing supplies and insulin. That’s our nation in a nutshell. A woman who must spend all her time with her children to keep them alive is struggling to pay for the supplies needed to do so because she must spend all her time keeping them alive.


I really feel for your friend and her children. My sister has been living with type 1 for 30 years. It’s a huge pain and it never goes away.

She has good insurance through her husband’s work, but they are truly wage-slaves. It’s almost a feudal serf relationship - they are tied to a particular employer and can’t leave because of the cost and uncertainty of making a change or entering a potentially less stable position.


It has been a couple of years since I talked to anyone in the industry, but there is a reason pumps require user involvement instead of doing things in code the way every single person would prefer: safety. At least at that time, insulin pumps had continuous glucose monitoring (interstitial, which is less accurate and lags blood levels). That’s not good enough to make dosing calculations, so users had to do blood testing, because the risk of giving too much is too great otherwise. At least if a conscious person chooses the wrong dose, they may be able to call for help.

In principal you could get around this by dosing with both insulin and glucagon, but at least back in 2013 no one had developed a stable, storable form of glucagon to use.


You’re clearly not seeing this from a supply-side economists point of view. Sure, there’s an enormous human cost, and is it worth it? The men with the money seem to think so. As long as we continue to accept their money as payment for our lives, we’re not really disagreeing with them.


As a T1 PWD myself, I have implemented certain parts of the semi-closed-loop system in the past. Currently, since the Continuous Glucose Monitor Dexcom started supporting smart watches, I no longer use any part of the home brew system, but I would if it gave me more utility.

But, what I really want to mention: as annoying as it is, this system is only possible because of the choice by Medtronic not to secure the wireless communications between their remotes and pumps. If the communications had been secured in a robust fashion, we would not be able to control the pump remotely to deliver those tiny boluses of insulin. The thing we chastised the company for yesterday is the only thing making this possible today.
But it won’t be possible long.
All of Medtronic’s newest pumps have foregone most of the wireless capability available in previous models. They only allow an upload of the data from the pump and are not even compatible with the company’s own (admittedly limited) Bluetooth transceiver. Medtronic is now one of only 3 insulin pump choices - and I’d expect that to fall to 2 in the near future.
So, if you want to move to newer pump technology (along with the newer CGM systems Medtronic offers) you won’t be able to control the pump through a system like this any longer. (Not bad for a $7500+ device, huh…)
It’s super frustrating!


The CGM issue is largely a thing of the past. It is now robust enough to make dosing calculations from - and the Dexcom system is FDA approved to do so.
Medtronic’s newest pump - sorry “artificial pancreas” according to the FDA* - uses their own CGM to suspend and increase insulin delivery based on CGM, all automatically. Everything I have found says the system works marvelously, and I am a skeptic here!
It does not introduce glucagon as that would necessitiate an additional infusion site/reservoir and, as you mention, a stable glugagon to infuse…

*thanks to that “artificial pancreas” designation, insurance companies are now refusing coverage citing it as “not medically necessary.” Garrrghh!


Thanks! I’m glad they solved the cgm problem.

The insurers will come around after a bureacratically long delay when they realize “not making people do math when they are not in a good mental or physical state” results in fewer hospitalizations. I hope?

1 Like

As someone in the industry, you’re right.

It all boiled down to having to prove that the system was near-absolute fail safe. Software (which requires validation proving every possibility for the software is mapped along with the hardware reactions), hardware, sensors, warnings, etc must work in a way that manages risks and hazards. The FDA created a set of requirements that was incredible. Most companies have said “screw that!” More work, higher risk, and (due to reimbursement caps) selling the system for the same price.

Medical companies risk getting crucified by the FDA, the press (including BB), and the courts, so are slow and cautious. Yeah, profits are usually good. Often there’s lots of money to be made doing anything difficult, just not this situation. Medical products are difficult because we demand very high quality. For the most part, I agree with FDA regulations that enforce it. There are a few oddball regulations, but for the most part the FDA is trying to help us, not hinder us. Yet compliance is expensive.

PS Everything above is highly simplified. If you want to nitpick details, there are entire websites dedicated to medical device development. For an outsider, it looks bonkers. From the inside, it looks … less bonkers.

PPS I dream of working in another industry some day that isn’t so highly regulated. Less stress.


I don’t use a pump (another Type 1 here), but I thought they all included a “continuous” release of a lente-style insulin in the background.

1 Like

I tend to try to take a Chesterton’s Fence approach to institutional roadblocks. A few years ago when I closed on my house, my lawyer said something like, “It’s a lot of forms, but every one of them is there because someone, at some point, did something that the form was later designed to stop.” As for the FDA: I do believe all the cost-benefit analyses showing that on net we’d save more lives by being less strict, and that it’s in part a problem institutional incentives not quite aligning with optimal outcomes. But I have no doubt that you’re right, and everything is there for a reason.


As if a choice exists.

I have to disagree. My wife, a type-1 diabetic since age 11, uses that system. It is usually right, but sometimes it is dramatically wrong.

Just this morning, it woke us to warn that her blood-sugar level was at 44 and sharply falling (two down arrows displayed). If true, that would mean she was in danger of dying. But a blood test showed that her level was actually above 80.

(Non-diabetics run at about 90. Below 70, your brain gets fuzzy. Below 50, you are unable to make decisions. Below 30, you’re dying.)

Patent troll kills this project in 3…2…

Cool, I learned something new. And it definitely applies to FDA regulations. I have to explain FDA regs to my Chinese colleagues, who for the longest time thought they were nonsense. But I explained that every reg is the result of someone doing something stupid. Rather than wait for someone in China to make the same mistakes, learn from our mistakes.

The true frustration is when a rule is vague or overly broad. You can end up trying to comply with a rule that makes no logical sense in a particular situation. Even though you plainly see it is a good rule made in good faith.

For example, device materials need to be biocompatible. Different levels depending on how touches the patient (i.e. the outside distal end of an instrument should be better than the handle that never touch a patient). The FDA recommends testing the whole device instead of each material independently. Some people take this to mean you must test the whole device. But I maintain that totally throws the dosing during testing, giving false positives or negatives.

If you read further, you see the alternative of testing each material independently is allowed. What gives? I say it’s because testing is really expensive and the FDA is actually trying to help manufactures control costs by suggesting a reasonable cost cutting method. And mot of the time, it works. Not for the stuff I work on. I’d rather be conservative and test each and every material. Plus once you have a database of materials, you don’t have to keep testing devices. In the long run, you save money and are safer.

One rule. Multiply that by thousands and you have what looks like chaos, but is a really a subtly, but not perfectly, tuned system. Chesterton’s Fence in action.

This topic was automatically closed after 5 days. New replies are no longer allowed.