The detectability of failure is an excellent point. Anybody who uses the hardware can confirm it's not 100%. I think your point helps me re-frame the glucagon as more of an insurance backstop for when we accidentally hit the gas a little hard, rather than a permission slip to constantly be going too fast and constantly be slamming on the brakes.

Even in this framing, it still feels like an extraordinarily valuable addition, and relatively low risk. It's also, of course, more to add to the patient's maintenance, but might help them or their caregivers sleep at night.

> It's also, of course, more to add to the patient's maintenance

I agree with all that you've said, and this point in particular is extremely important. It's also the reason I moved from a DIY system like the one mentioned here, to a commercial system, once the latter was available. There is simply less hardware and software to juggle with the commercial system. There are fewer knobs, bells, and whistles, meaning I might not be able to tweak things to be in as tight control as might be possible with a DIY system (though with risks!), but overall it's been "good enough" for me, and greatly reduces the cognitive burden of having T1D. My experience clearly doesn't match everyone's, but considering I'm typically someone who loves to tinker, and has plenty of T1D experience (engineer, 34 years with T1D), I'm sure I'm far from the only one that feels this way. My glycemic control isn't significantly better than it was when I did it via constant monitoring and mental math, but the cognitive and emotional burden is much lower.

Yea you would need really good failure detection if you were going to "hit the gas" with a bunch of insulin. Part of the solution is going to be controlling risk via the dosing algorithm itself, so you never get in those situations where you are at risk of severe hypo in the event of a (glucagon) site failure.

> However, if your brakes (glucagon) can sometimes fail completely, that could cause you to die almost immediately

Failure detection is via alarms to trigger patient action based on the continuous glucose monitor (which has a different set of reliability issues) as well as patient symptoms.

Hypoglycemia becomes symptomatic long before blood sugar is low enough to result in death or serious debilitation and T1D patients know their symptoms well. The risks are not nearly as dramatic as you're suggesting as one isn't/shouldn't be relying on glucagon to prevent severe hypoglycemia, I don't think any system is designed or being conceived to operate in such conditions.

Hypoglycemia isn't really much of a problem anymore with current CGMs and pumps.

> Hypoglycemia becomes symptomatic long before blood sugar is low enough to result in death or serious debilitation and T1D patients know their symptoms well.

There is a what seems to be a significant number of people who don't "feel their lows."

> Hypoglycemia isn't really much of a problem anymore with current CGMs and pumps.

Current CGM's can still require hours of "warm up", and many current pumps still must be removed for things like swimming so they don't get penetrated with water.

> There is a what seems to be a significant number of people who don't "feel their lows."

Severe hypoglycemia to the point of what was described (death) is not reported in any of the recent device studies.

Level 2 or moderate hypoglycemia, very different from death, is reported at < 0.5% in recent closed loop system studies.

> Current CGM's can still require hours of "warm up", and many current pumps still must be removed for things like swimming so they don't get penetrated with water.

Current CGMs are water resistant but conveniently one is also not administering insulin while swimming either. The bionic pancreas is also dependent on CGMs and has the same limitations.

I'm really not sure what point you're getting at. Hypoglycemia is not what's being improved upon with current advancements, it's time in target.

> Severe hypoglycemia to the point of what was described (death) is not reported in any of the recent device studies.

Are there large-scale studies that show this for a dual hormone control algorithm (the context of this thread)?

You seem to be misunderstanding how these devices work.

Bihormonal pumps do not mean continuous infusions of both insulin and glucagon. The pumps pulse insulin when you're high and glucagon when you're low. They're not both administered at the same time or continuously infused in a "balanced state".

The context in this thread:

> However, if your brakes (glucagon) can sometimes fail completely

A bihormonal system would not result in more insulin being administered than an insulin-only system for a given blood sugar, if the glucagon pump fails we would have an insulin-only system where we have plenty of safety data. There is no mechanism by which a bihormonal system has higher risk of hypoglycemia than existing closed loop insulin system.

To clarify, the context of this thread / what I was originally responding to was:

> In this case, you could be more aggressive in either direction of pushing BG, because you have a safety net.

OK thanks, I get your points. What I was getting at is a disagreement with "Hypoglycemia isn't really much of a problem anymore with current CGMs and pumps." Because lots of people on current CGMs and pumps still deal with hypoglycemia, despite these pumps and CGMs making the situation so much better than otherwise.

> Hypoglycemia becomes symptomatic long before blood sugar is low enough to result in death or serious debilitation and T1D patients know their symptoms well.

Often times in an acute setting, yes. However patients who have had diabetes (T1 or T2) for a long time often lose a lot of their hypoglycemia sensitivity and symptoms. It's not nearly that simple.

Hypoglycaemia during sleep is the big one. My hypo sensitivity isn't great, but if I'm awake I'll always notice before it gets really dangerous. Sleeping is a different story though. The days when even a slight low would jolt me wide awake in a sweat are gone though, and now I'm much more reliant on sensor alarms to not just sink silently deeper.

I'm a type 1 insulin dependent. The three low tech "hacks" I've been happy with are the following. First I take a sublingual Melatonin most nights, Melatonin upregulates the insulin receptors and lowers my insulin requirements about 40% by my guess. The second one is dietary I add olive oil to my lunch and dinner, I feel this provides my body a reserve of non glucose energy. Finally I use a very small dose of cannabis most days, I like to get a puff or two and night, THC protects nerve cells from Hypoxia so I feel this keeps my brain cells going when my blood sugar gets low. I feel these three things, give me a leg up on my long term blood sugar control. I am thankful there are hackers and diy opensource initiatives. The CGM readers here in America Dexcom, and Libre both have crazy bad user interfaces. Libre will only let you pair to a single device, I had their device fail leaving a functional sensor in my arm, a quick idea I searched for opensource libre reader app and found two. One worked and started reading the sensor. Dexcom has the issue of being a 10 day use cycle so you run out on varying days of the week. Both take the FDA mandate to have low blood sugar alarms as a blank check to overide any controls about sound or do not disturb to bother about countdown to a new sensor. I liked that external libre2 reader as it was the only device or official app that can be silenced, but their rigid only pair with one device still angers me, what if my loved one want's to be able to scan my sensor? Low blood sugars are challenging as they affect my brain and I can answer an amazing amount of questions about my blood sugar wrong if I get too low, my brain trying to preserve energy can be dangerous at times. This was an issue more before CGM. Dexcom decides that it is OK to have a completely automatic warning at any hour of the night "your sensor will expire in six hours!!!", that warning has little to do with my care as it is too late to influence refill compliance and seems to have been ordered by the executives to some how improve their pRofItiBiLiTY. I am so much happier on a third party app with the silent reader as an extra. Sorry if this was long winded, being insulin dependent has bee a challenge over 40 years. The first 10 years they hadn't figured out that insulin reactions are much more subtle on human insulin than the older pig and cow derived ones were.