I am impressed that her practice has several patients on OpenAPS. The location must help!
How well did what she tell you match your personal experience?
I understand your thought about focusing on the math. But I also understand about keeping the algorithm work understandable. What is concerning is that new AIs in chess and Go are not understandable so far in terms of human strategy concepts. Having a closed world that works better but cannot be understood is a bit concerning, in particular when it involves BG management: I would like to understand!
How beta cells are supposed to work certainly isn’t understandable in terms of the concepts we use to manage exogenous insulin either though … concepts like ISF and IC ratios and basal rates are really pretty astoundingly crude tools to try to emulate this cellular process with if you really stop to think about it…
ETA… really if you stop to think about how bass ackwards management concepts really are—- the only insulin on the market ever created that actually works pretty close to the same as natural cellular process (afrezza) doesn’t seem to be catching on because people think it’s dosing is “crude” !!! Argh! It’s the concepts we understand how insulin works that are crude!
In my opinion the reason it isn’t catching on, is the first top market drug in the category was shown to cause lung cancer, which scared many people away and they are lumping the two together.
Additionally, in the physician group the early adopters have adopted, making the leap to the largest chunk of physicians takes years and years. In heart valves for instance (very conservative), most physicians want 20 years of safety data before they switch. It will take many years before the largest group of physicians adopts something that is truly innovative. Many many therapies die on the vine with only the early adopter group having adopted.
I think a big factor is how people just don’t understand it. People are taught that 1u insulin cancels out X grams carbohydrates. That is their gospel. That is what they know to be the truth just like only a few hundred years ago we knew the earth was flat and you could literally sail off the edge of it if you weren’t careful… When something comes along that works completely differently and to which those rules no longer apply it is pretty natural for them to be skeptical.
I agree with you, but as with any revolutionary technology, getting adoption isn’t easy, especially to a risk averse crowd. That is why I am sure their customer acquisition costs are astronomical, and therefore no private equity company is going to step in and fund the adoption.
In that respect I expect that closed loop systems will eventually help people make the logic leap…because my bet is that the better of these systems will wind up completely ditching the 1 unit insulin = X units BG logic.
I think that’s a hard sell to the diabetic population beyond forums like this. My CDE deals with people who would never change their own pump settings. The sense I get from most other forums is a wariness about a Frankenpump that does its own thing in the background and might just make a mistake and kill you. I would hope that evidence of improved control would shift them over, but how quickly? It took a few decades for conventional pump therapy to be considered mainstream. On the other hand, people today are more accustomed to insulin pumps, and to rapid change in general, and to having technology do things for them (“Alexa, give me a bolus”), so I hope the hybrid loops coming onto market will be well received by patients. Personally, I can’t wait for a pump that’s as hands-off as possible.
That’s pretty interesting for two different reasons - first, that an office has more than a single patient using a DIY closed-loop system - surprising, and second, the trend of dialing up basals to have the system operate better. With loop closed and programmed basals dialed up, I assume you must have observed OpenAPS zero temping for significant amounts of time? In a sense, the system was dialed to continuously sort of super-bolus? I can see how that might work “better” although I am not sure if I’d recommend the approach in general. Another question: did you try running auto-tuning on your system? Wonder if auto-tuning would tell you that basals were too high.
As a side note, my programmed basal rates, sensitivity and carb ratio remain all the same closed-loop or open-loop. But, I am an adult, with much less variability overall, and wider error margins compared to kids, so everything is much more predictable and easier.
I assume you must have observed OpenAPS zero temping for significant amounts of time? In a sense, the system was dialed to continuously sort of super-bolus? I can see how that might work “better” although I am not sure if I’d recommend the approach in general.
Yes, this is exactly what we noticed. He was zero-temped for the majority of the day.
I think what I’m realizing is that the basal rate we programmed in was essentially the median TDD we use in our daily life – which is different from the average. Samson’s average insulin use is about 8 units per day, whereas his median is closer to 10, because he has some days where he reaches 12-14 units a day and a fair number where he’s around 6. So my sense is that positioning the basal in the middle of the range does a decent job of keeping up on those high- and low- days. Or it may simply be related to the proportion of the ISF/basal rate and carbF that he uses – maybe there’s no way for openAPS to ramp up enough when he uses such low basal and yet his carbF and ISF aren’t quite as different from a full-grown person’s.
Either way, when we start back up with openAPS (which we probably will if only for the night times), then we’ll start with the newer basals and parameters we’ve calculated, and see if it works. My guess is we’ll run into the same problems we had before, but I’m not sure.
You are probably correct - you may very well find that you’d need to dial up basals again with OpenAPS loop closed. As you already noted, the crude models we are working with are not too bad for adults in a relatively stable situation, but are otherwise lacking.
Another thought about that (and another point adding to the inadequacy of the models we are working with, and against setting up the system to zero temp often). On few occasions when I had zero temping going for longer periods of time, say a few hours, I noticed that Loop’s predictions and bolus recommendations after such event were way off, as if my carb sensitivity temporarily increased by an order of magnitude. Any tiny carb amount would just shoot my bg up in huge increments. The explanation, I think, is that in the absence of any insulin, liver temporarily becomes effectively locked out, and the entire carb intake goes directly to the bloodstream, instead of being partially stored in the liver. Well, as may be obvious from the above statement, I have no expertise in physiology, endocrinology or medicine, I am just making up explanations
I can confirm this phenomenon. Last night because of suddenly decreased basal requirement, LOOP had me on zero basal for 3 hours; despite this the CGM woke me because of 55 mg/dL. I took 3 glucose which normally would raise me to 85, but with no insulin in my bloodstream it sent me to 160.