Journal: my start on Loop

[Kaelan]

By early September, I realized that Loop was simply unable to bring me back from high up when in a hormonal peak.

I think this is why: when I am in a high hormonal peak, I remain with very high insulin needs until I am able to come down much lower. Loop gives me all the insulin it thinks I need, then waits. It does not bring me down enough because not only do I need to come down, but I am also consuming insulin much faster. My ISF is what is telling Loop what to give me to come down, but it is not possible to have an ISF when you are at 200 and another at 120. So, when I override my basal 50%, the only think that Loop is doing is discounting the effect of IOB faster, But it is not increasing the amount of insulin that Loop is administering to me. At least that’s how I understand it.

So I think that is why neither Loop along nor Loop with large overrides can deal with large hormonal peaks like the ones I have.

You might ask: why don’t you just accept to be at 200+ for a larger part of the day, and let Loop do the work over a longer period of time? Even if I am willing to let my A1c get a lot worse than by hand management, the problem is that I feel really bad above 180 or 200: I can’t work well and I can’t focus well. I can’t go to school or do good work when I am too high. And my tests are bad.

So I was stuck. But I was not ready to quit on Loop.

[Kaelan]

My last option with Loop was to compensate for the high peaks with large manual insulin injections. Here, I had two ways to do it. I could use Loop to inject fake carbs. The problem was that it would leave bad data behind (the carbs), when I was doing reports. The other method was inject them by hand but to log them in Nightscout. The problem with that method is that, while you can see a good log, Loop is unable to account for the right IOB.

The best of both worlds would be to inject with Loop so that it can account for IOB, but to get rid of the fake carbs after they are counted out by Loop: you can list them as 2-hour carbs, then, after 2 hours, you can go back and delete them by hand. I though that was the best, and I decided I would do that.

The result: after 4 weeks, it turned out to be pretty good. It ended up being the best (and only) way to take me down from large hormonal peaks. But it is not perfect. First, it compels me to do more management by hand. Second, learning to do this with Loop is tricky, and it ended up at first in really increasing my lows a lot, not as bad a before, but still higher than I would want. Finally, it is really difficult to remember to delete the carbs entries in Loop after 2 hours, and I forget all the time.

[Kaelan]

Unfortunately, when I was starting to try to refine my use of manual injections, the California fires started, and soon created such bad air quality that I was not able to run regularly any more. In turn, this caused my BG to become much harder to manage, and really destroyed my ability to know what was caused by what: my BG became quite chaotic (although my A1C was still reasonable) because of my very irregular exercise patterns.

I was lucky: in mid October, my Swiss friends invited me to come stay with them for a few weeks so that I could run every day. Since school is online, I was able to go. While I was there, with regular exercise and my new Loop + large insulin injections on peaks, my TDD went back down to low 40s (from mid 60s), and my calculated A1c went back down to 6.0-6.1. My low % was about 4-5%, a bit higher than I would like, but much better than before. My STD was a bad as ever, around 42-43.

I am just back in Berkeley. I am up so late because I am jetlagged and I can’t sleep today. My first two days’ BG back are pretty ugly.

[Trying]

Thank you so much, @Kaelan, for updating us. You’ve been on a long journey that many of us have gone on, though perhaps not the hormone tribulations!

I, too, often use FAKE carbs, and label them as such so that Nightscout knows they are fake for report purposes. I will also delete FAKE carbs too quite often as you do. A big hassle.

I’m so glad you are working out and able to tweak Loop to your body. I wonder if you should give the autobolus or freeAps branches a try. It might help with the hormone issues. I use the autobolus branch and believe it helps over the master but I still do FAKE carbs!

Good luck and welcome back to CA! I’m so glad you are able to get your runs in, too, a huge benefit for BG control!! Looking forward to following your journey!

[ClaudnDaye]

FreeAPS is night and day from regular loop…the ability for loop to AUTOMATICALLY give micro boluses when your BGs begin climbing results in (most of the time) smooth nights for us…there are always exceptions, especially when CGM/Sensors are having issues. But if you want better control…get on MicroBolus or AutoBolus…the difference between the two really is that FreeAPS (MicroBolus) gives small boluses than the AutoBolus branch…AutoBolus gives one or two big boluses while FreeAPS gives many small boluses every 5 minutes based on the algorithm in the background.

I HIGHLY recommend it. We have been usingit for over a year and absolutely love it. AND if you don’t like it, there’s a switch to just turn it off and revert back to normal loop.

Definitely what I’ve found is…the biggest area that causes the most issues within loop are having a) incorrect insulin sensitivities set and b) having incorrect basal rates. If those two things are high, roller coasters will ensue…you have to figure out the right insulin sensitivity and the right basal rate for loop to work properly.

Remember…loop is only doing what it’s programmed to do. If it’s doing a crappy job, we have to consider that our settings are just wrong! When dialed in, loop (especially the branch we use, FreeAPS) is wonderful

And on a personal note: Thanks so much for documenting your journey with us. I know that I, as a parent, appreciate this because in the next 10 years or so, I may have Liam reading over your journey as he begins his own journey. So from a parent to a child, thank you for helping my child as you discuss your diabetes journey and struggles.

You’re doing a wonderful job and remember…every bad day is just 1 day…tomorrow has the opportunity to be better! Let it roll off of you like water on wax. Just learn at every opportunity and do the best you can do. I know I speak for the entire community when I say we are all so proud of you and how you’ve grown and how you’re dealing with Diabetes.

[bkh]

Ah, yes. Learning how to work with Loop. I ended up with a strategy not far from Kelan’s. The key fact for me is that Loop determines its dosing based on it’s prediction of where the BG will be 5 hours from now. If it thinks my BG will be in range 5 hours from now it won’t give me extra insulin, and if I tell Loop to give me extra insulin anyway because I know better, it will turn off my basal to try to save me from it’s prediction of a future low. So when I “know” the my BG isn’t coming down but Loop predicts that it will, my main strategy is to adjust the amount and duration of carbs I’ve eaten (or, less often to enter fake carbs) until Loop’s prediction agrees with mine. Then I can tell Loop to dose me with insulin, and it will happily do so. No syringe needed for that. Furthermore, sometimes I hurry the process by entering an even higher carb load to enable Loop to give me an overbolus, then an hour or 2 later if it looks like I actually will end up low I’ll back off on the carbs so that the basal will be cut to help me land in a good range. If I’m over 200, sometimes I’ll do a rage IM bolus and not tell Loop about it at all, with the IM bolus sized to bring my BG down roughly at the rate Loop was wrongly predicting.

Now both of those strategies (adjusting carbs and IM correction bolus) are easier for me than Kaelan because I don’t care if reports of my carb and insulin amounts are accurate. I just care what my BG does. So I don’t generally go back and remove added carbs or enter IM boluses to the records, and that saves me significant work. But yes, to get the time in range that we want, we must manage the dosing because Loop can’t do it all by itself in cases of temporarily increased insulin resistance. I don’t mind, because Loop still saves me a tremendous amount of work. Loop generally is trying to do the right thing, but just not aggressive enough to overcome any temporary insulin resistance, and I can fix that.

[Trying]

Thanks, @bkh, for the “tricks” we can use with Loop. I use these myself. Are you on master branch or autobolus/FreeAPS?

I agree, Loop does what it says it will do, but we need to help it along, and still, it saves me tremendous amount of work, too!

[bkh]

@Trying

Are you on master branch or autobolus/FreeAPS?

Master branch release v2.2.4. It seems likely that some microbolus strategy will eventually be accepted into the release version, but I don’t see a need to rush in given how well I get along with the current version.

[Trying]

@bkh
Yes, I agree, it will eventually be included in master. I am considering trying FreeAPS to see how it compares, but autobolus is doing a pretty good job for me, and I am comfortable using the main Loop branches.

[Michel]

As K’s dad, I need to add one thought: it is such a bloody relief to see Loop suspend at night as he is going down, or when he is too low. For the first time, I feel that college is not going to be the end of the world. Even with the worst difficulties we have had with Loop, it still is a blessing for us. It is a challenge, but a very, very positive one.

[Kaelan]

@Trying

use FAKE carbs, and label them as such so that Nightscout knows they are fake for report purposes.

@Trying, that is a great idea, thank you! As hard as I try, I know I’ll never remember in time for many of the fake carbs I put in.

@Trying

I wonder if you should give the autobolus or freeAps branches a try.

I just started on Autobolus this morning, thanks to your prompting and @ClaudnDaye! I hope to try it for a few weeks, then to try the FreeAPS branch also.

@Trying

I’m so glad you are able to get your runs in

So am I! Thanks very much for all the support: it is really helpful to be encouraged by someone who is ahead of me in the same direction.

@ClaudnDaye

AutoBolus gives one or two big boluses while FreeAPS gives many small boluses

@ClaudnDaye, thanks very much! You had encouraged me earlier too, but I was thinking that I should figure out Loop first. Thanks to you and @Trying, I just started on Autobolus today!

@ClaudnDaye

most issues within loop are having a) incorrect insulin sensitivities set and b) having incorrect basal rates.

I already figured out the first You go up and down like crazy, or you don’t come down. Thanks for the advice.

@ClaudnDaye

I may have Liam reading over your journey

It shouldn’t be too long now: he is growing up so fast, it’s crazy to think that in a small number of years he’ll be as old as I was when I was diagnosed. Of course, he’ll be so much more experienced by then!

Thanks very much for all your advice now and through all the years. FUD has really helped me a lot. And I would really like to help other kids when they get diagnosed. I know it takes a long time to figure out things. I have been diagnosed for almost 5 years now, and there are a lot of things I haven’t figured out yet.

Of course, my hormone peaks don’t help. But I am pretty sure most boys have the same at my age. I think if we survive them (I mean the lows that come with them when we inject too much the 3rfd or the 4th time), we’ll be able to deal with the stuff that comes later.

[Kaelan]

Here is my latest update. Following @ClaudnDaye and @Trying advice, I switched to the Auto-Bolus branch of Loop just about a month ago. I picked Auto-Bolus vs FreeAPS because my hormone peaks require very large amounts of insulin, and, from the description, it seemed that Auto-Bolus would be ready to inject more at a time, but I don’t know if that is true.

As always, the ramp-up is significant. I am still learning how to adjust to the algorithm. In my first couple of weeks, I ended up low a few too many times: I was up to 10% low time for a while. Then I backed off too much, and I ended up with a bad calculated A1c.

These are a few things I found out:

• first, I figured out very quickly that my regular settings for Loop had been wrong: I had a max 6U per hour for basal, so my regular Loop adjustments before switching were maxing out at 6U per hour. But, in a hormone peak, I could need easily twice as much or more. So that was stupid of me not to think that I would need to adjust that number up a lot. I probably should have been at 15U max per hour, or some such absurdly high number. Of course, the downside of that is safety.

• I was very pleased to see that Auto-Bolus could bring me back down faster than before! It is kinda nice to see tons of injections, every 5 minutes, about 0.5U each often for me, accompanying a peak: all that work that I don’t need to do!

• but I still need to inject myself for large corrections. I am realizing that, in the end, the problem is that, when I get a high hormone peak, basically I am very insulin-insensitive, and my Loop settings are so far off that Auto-Bolus still can’t catch up. In the end, that is my problem: I am oscillating between a state when I am “normal” and one where my sensitivity to insulin is “abnormal”, so far from my normal parameters that Loop can’t get me back into balance.

• the problem, though, is that it is very difficult to figure out what A-B injected for correction, because you have a myriad of injections in, say, the last hour when I am suddenly peaking. So It is easy to make a mistake when you do an extra injection (FAKE carbs etc.), so I can easily end up low if I don’t painfully count everything already injected. It was much faster/easier to figure it out from the main branch, since you could just look at how high your basal had been and for how long.

• so it is more difficult to know what to inject to bring you back: you have to calculate carefully or guesstimate what A-B did, then add to it carefully. The worst is before going to bed, because you don’t want to overdose then. Unfortunately, I have peaks almost every night before bed.

• it appears also easier for A-B to throw you low in a “regular” oscillation. That happened to me once at night. This, of course, is bad: your algo itself is pushing you low.

Right now, after 30 days, I still have not quite figured out how to take advantage of A-B best. I am back to very small amounts of lows, 2% over the last 30 days. But my calculated A1c over that time is very poor, 6.6%: highest I have ever been since diagnosis. I am putting it down to having a hard time learning how to work with A-B.

My main lesson is that, for me, I still have to take into account my own insulin sensitivity changes along the day: A-B won’t do it for me. For instance, yesterday I forgot to inject 2.5U before starting school and up my basal by 40% (I do that as a standard). I ended up fighting a high for all my school time, and injecting lots of extra bolus insulin, without being able to control it. It looks like A-B does what you would do yourself, but it only does it based on you regular, “normal” parameters and so it won’t take care of the moments when you are not “normal”. So I think it causes you to be less active because it looks like it is doing it already, when you actually need to deal with these moments yourself. I know that what I am saying is probably very confusing.

I still like A-B better than the main branch. But, if anything right now I am worse off than before, until I learn to use it properly. I am not discouraged at all tho. I am not frustrated. If there is a way to figure it out I will.

[Trying]

Thanks, @Kaelan, for updating us. I’m glad you decided to try the AB branch but man, I am sorry you are going through so much getting the settings to work for you. The hormone issue is so tough. I can’t imagine having to compensate for this with AB settings or anything else, like MDI! I obviously don’t have this issue. But I do have insulin insensitivity issues from time to time. I noticed Loop will say I still have X amount of IOB still when my BG is already going back up! I decided to try a different insulin duration and am currently using 2 hours (under Settings, Insulin Model, the Walsh Model), and have been using this for nearly a week. It is helping quite a bit but of course, I have to monitor not to get insulin stacking. So far, I have not had any stacking. AB has been doing a good job with suspending insulin for me. I don’t eat frequently so stacking is much less likely to occur for me. I think if you eat frequently (as you may, since you are growing!), it may cause stacking. I know most of the developers frown upon using anything other than 6 hours of insulin duration! Anyway, just another “setting” to think about!! Good luck getting everything configured properly for you!!!

[Chris]

@Kaelan

But my calculated A1c over that time is very poor, 6.6%: highest I have ever been since diagnosis.

Don’t worry Kaelan, you will get the hang, and don’t worry if your A1c gets just a bit higher. It is hard to maintain everything and your life as you are getting older. Just look at the average teenager that is your age’s A1c. You are rocking it.

[ClaudnDaye]

@Kaelan

I picked Auto-Bolus vs FreeAPS because my hormone peaks require very large amounts of insulin, and, from the description, it seemed that Auto-Bolus would be ready to inject more at a time, but I don’t know if that is true.

FreeAPS can give big boluses, but AB can’t give micro boluses.

Gratz on the success! 6.6% isn’t “very poor” - no matter what you may hear. I think we sometimes become jaded by all our successes and don’t realize that 6’s and 7’s aren’t going to kill us or even cause long-term complications really. The research for hyperglycemia related issues points to (from all that I’ve read) people who stay in the much higher range for much longer periods of time than anyone in this community does. 6.6% is great imo and you shouldn’t beat yourself up over it.

[dm61]

@Kaelan, thanks for the report. Dealing with such huge variability around growing and being active is no easy task and you are doing spectacularly well!

Looks like some users here are trying the “automatic-bolus” (AB) or FreeAPS branches, so I thought it might be useful to go over the algorithmic differences between the standard Loop and the AB (or FreeABS) Loop. The differences are actually more modest than one may think.

First, the total correction dose calculation is based entirely on the BG forecast and is exactly the same in all Loop branches. For example, consider the BG forecast below:

BG is predicted to eventually be 139 mg/dL. This particular user has the correction range set to 85-85 mg/dL, and the insulin sensitivity (ISF) is set to 55 (mg/dL)/U. The total correction dose calculated by Loop is (139-85)/55 = 0.98 U, same in all Loop branches. This value (rounded down to the nearest deliverable value) would show up if the user tapped on the manual bolus icon. What happens next is where the standard and the AB branches differ. In the standard Loop, the dose of 0.98U is delivered as a 30-min long increased temporary basal rate equal to the (programmed rate) + (2 * 0.98U)/h = (programmed rate) + 1.96 U/h. During the first 5 minutes, however, the additional insulin delivered is only 1.96 * (5 min)/(60 min) = 0.16 U, which equals (2 * 5/60) = 16.7% of the total correction dose. In the AB branch, instead of the 16.7% delivered over the next 5 minutes via temp basal, a higher percentage of 40% is delivered in the form of a bolus. In the example above, an automatic bolus is calculated as 0.4*0.98 U = 0.39 U, which is rounded down to the nearest deliverable value of 0.375 U as seen below:

The exact same process is repeated every 5 minutes, resulting in a series of decreasing boluses until the eventual BG is driven to the correction range. Here is a series of 3 such automatic boluses.

By the way, regardless of the branch, Loop displays IOB as “Active Insulin” - there is no reason to count or guess anything, and from that point of view, AB Loop is no more difficult to use than the standard Loop.

As can be seen in the example above, almost the entire correction is delivered by the AB Loop in about 15 minutes, compared to more than 30 minutes in the standard Loop. So, the AB Loop is not “more aggressive” in the sense of delivering more insulin, it is just somewhat faster in delivering automated corrections compared to the standard Loop. The faster up-front correction delivery also means that the time available to counter-act BG dropping down is somewhat shorter, which implies somewhat increased risks of lows. The increased risks should not really be substantially higher - if they are, ISF or something else is probably off.

By the way, where is that 40% coming from? That percentage is a best-guess trade-off that represents a level of confidence a user may have in the Loop predictions. It is a hard-coded value in the AB branch (still easy to adjust in the code if one wishes to do so for some reason), while in FreeAPS the % is an adjustable parameter. Aside from having the % adjustment exposed to the user, I think FreeAPS still delivers a portion of the correction as temp basal, and a portion as a bolus, which objectively does not really make any difference.

[Carol]

@Kaelan

I am not discouraged at all tho. I am not frustrated. If there is a way to figure it out I will.

Just read your post. Love your attitude. I’m sure you will figure this out, and maybe have opportunities to help others learn in the future.