I'm not really an expert on T2DM (I'm T1), but I believe the main issue is going to be that insulin sensitivity is much more variable for T2s than T1s, and will change faster than autosens can keep up with.

The reason for the mixed messaging about treatment decisions is that the Dexcom G4/G5 (they're the same sensor with a different receiver) spent part of its product life not approved for use as a basis for treatment decisions, and was later approved for use as a basis for treatment decisions, but only under certain circumstances. There are some best-practices around CGM sensors that the manufacturers won't tell you about, like pre-soaking sensors, which greatly improve reliability. And, counter-intuitively, a closed-loop system is actually less demanding of sensor precision than a human giving correction boluses would be.

For OpenAPS, getting started involves installing Linux on an IoT board, setting up a server on Heroku, and monitoring log files. Tech savvy is definitely required, but mostly for setup, and occasional troubleshooting; if you have a tech-savvy person willing to put some time into it, the can set it up, show you how to monitor it, and then step away.

I can't answer your questions from a clinical perspective, however I'll have a stab at them. I'm a type 1 diabetic, diagnosed when I was 9.

> Do you know why the applicability to type two diabetes is not as clear as for type one?

I'm only stabbing at this, and it's anecdotal, but I believe it is likely due to the lack of consistency across type 2 diabetics. Whilst there is variation in Type 1 Diabetes, it is largely limited to lifestyle but the basic principle is the same - you don't produce insulin yourself. Type 2 diabetes is often caused by insulin resistance, and I guess varies wildly based on the individual, their lifestyle, and their genes.

> I live with someone who has a CGM (continuous glucose monitor) where the device came in a box with an insert saying "do not use this device to make medical treatment decisions." To see any of the readings you have to install an app whose start screen says it's experimental and, again, can't be used to make treatment decisions. But the user's manual has a chapter titled "Treatment Decisions" that tells you how to use it to decide when to take insulin and how much to take. It's frustrating that the messages have to be so contradictory.

I haven't been given a CGM for longer than a few weeks (NHS rules are a bit restrictive in my area if your control is good). However the rationale for not using the CGM to make decisions is based around how it gauges your blood sugar. Because it's relying on blood glucose levels in the skin tissue I have been told that there is a delay, and basically the CGM operates about 15 minutes behind your current glucose levels. This is ok when you're using it to track glucose levels over a day, but not particularly safe if you're relying on it to make a decision for obvious reasons.

> A final question: how essential is it to be tech-savvy to use this technology?

This is a question I have asked myself a few times. As mentioned above, it's pretty difficult to get a pump and CGM on the NHS in my area without proving that you can't treat yourself adequately with regular injections. I am unwilling to sacrifice my health to have a pop at getting a pump, so I had a look at private options. Whilst initially I think there's a fairly steep learning curve, there is quite a lot of documentation provided, and the community is pretty active and helpful. I suppose it's also helped by the fact that there is only a few pumps on the market (relatively speaking). The issue is that if something goes wrong in your config, which is not uncommon for a homebrew solution, then you have to go back to manually adjusting your insulin.

My SO has a continuous glucose monitoring, and as you say, the reading lags behind by 15-30 minutes compared to measurement from blood. But the blood measurement is a single, instantaneous value while the GCM meter shows the last 8 hour glucose levels plotted and the current upward/downward trend. While it takes a bit of cross-referencing between blood and tissue glucose levels to get used to the readings, it actually provides better information to base decisions on (fast acting) insulin dosage.

The GCM (Freestyle Libre) has been a life-changer and improved long-term management of diabetes. In particular, it has been helpful for maintaining a reasonably flat glucose level throughout the night (because of better information for evening insulin doses), where they previously had high and low peaks (down to dangerous levels a few times).

I suggest you do everything you can to get a continuous monitoring system. Because you can see your glucose level history, you can then make decisions on whether you need an insulin pump or not. If your night time glucose levels are fine, you might not need one. If there are high or low peaks you can't manage, a pump will improve your life.

> it actually provides better information to base decisions on (fast acting) insulin dosage

Would you say that's because you can get a better idea of trends once you have enough data? My understanding has always been that the true value in a CGM is revealing those trends and patterns that are really unique to the person and that are hard to spot with regular glucose testing.

EDIT: I realise that this is basically what you've said. Apologies, brainfart.

> I suggest you do everything you can to get a continuous monitoring system.

I'm currently jumping through the hoops to get myself one now. I can get my hands on one via the NHS fairly easily, it's just the pumps which are hard to get. I've done a lot of work over the past few years to get my control to a place I am happy with (finishing uni and that awful realisation that you are in fact mortal). I also wanna say that I recognise that as someone who is able to control my diabetes using injections it would be wrong for me to demand a pump when there are others who for various reasons can't control their condition. In a perfect world it would be great, but hey, we work with what we've got, right.

Yes, exactly. When you wake up in the morning, you can get a plot of the last night's glucose levels which you normally don't know about with discrete blood measurements. This allowed my SO to adjust evening eating and insulin dose and their timing so that the glucose levels stay within the "good zone" all night (the GCM reader unit has a helpful graph with the good levels highlighted).

If you can't get the NHS to give you one, it might be worth it to get it for at least some period (say 3-6 months) out of pocket. By that time it should have given you insights on your treatment and allow you to better assess if an insulin pump is the right thing for you.

It helps with figuring out long-term trends, but it's also an improvement even for a single snapshot-in-time decision. If your blood glucose is 150mg/dL and you're trying to get it to 100mg/dL, the difference between "150mg/dL and rising 3mg/dL/min" and "150mg/dL and falling 3mg/dL/min" is the difference between wanting insulin and wanting carbohydrates.

> better

What happens when the sensor is not reading correctly? Isn’t that the biggest possible danger? The human “knows” it’s wrong but software?

How do you tell that the glucose reading measured from blood (the old way) is correct? You can't. The new method is not worse than the old in this respect.

And again, you get at least 8 hours of history graphed on the display now. If it doesn't look sensible (w.r.t eating and insulin doses), you need a backup. If the old fashioned blood glucose meter gives a wrong reading, you are much worse off.

Note that the sensors are tested and approved, not the DIY hack the article talks about.

Old way: you repeat the measurement nothing is in your body. and you konw di you feel e.g. hypo. New way: the sensor needle is simply a little off (in stays sticked). You don’t know, the software doesn’t know. The sensor is not magic that works 100% of the time, it ends in a needle sticked in your body, many ways to go wrong, even if it worked a while.

And the danger of “fully automatic” is there.

Just to clarify, we were only discussing monitoring and sensors in this subthread. Not closed loop pump systems.

A sensor failure in closed loop is dangerous. I do not know how the control software detects and deals with this. I have no experience with them.

A closed loop uses the oref0 algorithm, which tweaks your basal levels and therefore cannot react to fast peaks or cause a hypo not easily fixed by having two candies (or giving an alarm if you're not awake). What it does is it detects an upward trend for the last X measurements, sets something such as 300% temporary basal rate, detects the trend reaching the peak, sets a 0% basal rate until you hit the predefined target (for me, 5.5 mmol/l during the day and 6.5 mmol/l at night). Here sudden wrong readings will not cause any harm or danger.

The new oref1 algorithm uses a method called super micro boluses, which can detect a sudden peak or unannounced meal, sees the first jump up, gives you a bolus so it borrows from the basals, setting the basals to 0%. The algorithm works really well for things such as eating a pizza, where you get about 50% of the carbs immediately and the rest + a bit more during the next five hours. So you take a bolus for the food, then when the fat hits in the oref1 will give you tiny boluses.

The latter is more sensitive to sensor failures, so it is on only when you told the system you have some carbs in your body, or if you happen to have a Dexcom G5 or G6 sensor with good noise readings, it will be on all the time and the system knows when the reading is faulty.

Which closed-loop system do you use which can provide both boluses and basals? Can you share? Also do you think Dexcom sensors can be always trusted? Searching, I find something like:

https://forum.fudiabetes.org/t/dexcom-out-of-range-calibrati...

The oref1 smb is in AndroidAPS release candidates and in OpenAPS. You can't trust anything 100% ever, but you start recognizing when they have problems and can cope with the errors.

Thanks. Readers should also note that the main topic is about the closed loop apparent "revolution." In reality it really can be dangerous, exactly because of the explained sensor problems which are real. To clarify to those that don't know:

The sensors are big things with the needle permanently in your body. You don't want three needles permanently to be sure that at least 2 from 3 work. If you have only one, you have a real problem if you trust it blindly.

Yes, it's good that some people try. But there are real risks, due to the current limitations as explained.

I'd say that I myself making all the judgements caused way more harm to my body than what the computer can do. It is that much better. Before I used to have nights when my sugar was in a dangerous hypo for hours during the night, luckily not causing too much brain damage. Now all of these problems are gone. Almost all of the scary hypos are gone. My A1c came down from 7.5% to 5.5%. I'm 88% of the time in range, 10% above the 8.5 mmol/l and 2% hypos (none of them at night) which I can correct fast due to alarms.

Thanks a lot. My question is: is the techology as reliable that somebody who is without any technical background (e.g a child on vacation without the parents) able to depend on it?

I know that most of the population is actually not technical, and that they can easily understand “correct when high, eat when low” but complex technical reasoning can’t be expected from them. In that context, sensors not functioning as in ideal lab conditions is a big issue.

Here in Germany we have a group of diabetics helping each other with the system. Most of us are not so technical, for everybody the system has been a massive improvement compared to their own judgement.

As a type1 you must have a lot of knowledge to be able to survive. The software is a tool you must learn how to use, but takes away lots of burden from the decision making.

> Here in Germany we have a group of diabetics helping each other with the system

Can you give some more details please? I ask for a friend of mine. Thanks.

It's just a loosely connected group of people in Berlin, going to the same doctor's office and being interested in tech.