FUDiabetes

Tslim Control IQ - How Do They Calculate Insulin On Board

Thanks. Much more clear for me and I agree.

But I like it that way, other than preferring a more realistic way to calculate decay.

Do you disagree with anything in my post just before your post? Because I thought I had found the best way to utilize and understand IOB.

No, I completely agree with how you are using it, it just isn’t that helpful for us, and we use Control IQ the same way you do. We definitely don’t wait for it to do its thing on highs, but it works really well for us during the night. I really don’t think that the IOB information helps make any treatment decisions for us however, which it could if it took into account more information, such as Eric has hypothesized. Like you are 250, it normally takes 3 units to get you back to baseline, and you have 4 units active so you should consider x grams of carbs if you haven’t recently eaten. That would be pretty cool.

1 Like

Here is a very simple improvement. Currently Omnipod (and I assume the rest of the pumps), only ask you duration during setup for the use of calculating IOB.

In addition to asking users for their insulin duration, pumps could also ask users their insulin onset time, and insulin peak time .

It could use 3 linear formulas instead of just 1, for a much better IOB calculation.

Kind of like this:

Very simple for setup, just 2 more things to consider, but it would give a better picture of IOB.

Then also it could factor in high BG and/or carbs.

1 Like

Ok, glad we agree although we disagree on its usefulness.

The way I look at it, Control-IQ mostly uses basal insulin to get you back to target (although it will start using bolus insulin if it predicts you will go higher than 180). There are so many things that can change your BG that it would not want to deliver bolus insulin because once delivered you cannot take it back. On the other hand with basal insulin, if all of a sudden things change they will just change the basal insulin adjustment 5 minutes later. What I like about it is that it tells me the correct bolus to take so I can take it if I want. And, in telling you the correct bolus insulin to take if you want it also considers your profile basals to be correct but also tells you that you would go high or low based on the many basal adjustments it has made trying to keep you in control.

1 Like

Totally agree. As Chris said they will probably get there too. Control IQ was a great step. I can’t believe how it has kept me in a narrow target range while sleeping.

2 Likes

@Eric, @Chris, et al:

Here is a very detailed (that is, mathematical) article from the OpenAPS folks about insulin activity, IOB, etc that includes many of the things that @Eric suggested.

While I’m still wading through this myself, it is clear that they are able to generate reasonably sophisticated models and curves with only a couple of parameters.

https://openaps.readthedocs.io/en/latest/docs/While%20You%20Wait%20For%20Gear/understanding-insulin-on-board-calculations.html

In fairness to Tandem, Medtronics, and Omnipod, I certainly do not know what their internal IOB vs time model is. It’s entirely possible that it is more sophisticated than a simple linear decay. I certainly not given myself a bolus and then carefully recorded IOB every 5 minutes over the next 5 hours.

Stay safe!

John

1 Like

I checked it out when I first got the Omnipod. It really is a straight-line depletion. Pretty much the
y = mx + b
formula with a negative slope.

1 Like

@Eric:

Thanks for that information. You’ve inspired me to spot check my t:Slim X2 to see if they are also using a linear decay.

Even if they are using a simple linear decay, I still find IOB to be a useful number to look at and factor in.

I wish, however, that their “bolus wizard” factored in the trajectory of the current BG reading when calculating a bolus. For example, if I’m at 200 mg/dL and rising fast I probably want a major correction … but if I’m at the same value and falling, I may not add a correction at all.

Stay safe!

John

1 Like

Great Discussion. I’m new to the forum. My Type 1 spouse is using CIQ the last 5 months with success. I’m greatly interested in predicting the performance of the TANDEM X2 pump function.
I have search for various TANDEM Type Zero algorithm documentation with minimal success but thier is enough info capture some of the algorithm and logic condition that is supposedly used by TANDEM X2. Its seems that the Tandem pump incorporates and creates significant amount data to time predictive BG levels magnitude and threshold time cross over conditions. In a perfect world a person’s BG would follow and never dip below or above the pump predictive BG curve estimates. A full predictive model would account for carbohydrate inputs too. I would like to know what those predictive plots look like and visualize measured BG (CGM) time values meet or exceed threshold conditions. As I see it the limitation with the Tandem IOB number only value is it can be used to estimate how much addition BG correction will occur in a near term time line (less then 5hrs) and for perfect steady state time condition.

2 Likes

Welcome to the Forum @JAB JAB, it sounds like you will be a wonderful addition to our menagerie of members. I look forward to additional insightful discussions. Please look around, we have many technical members and always welcome new ones!

1 Like

In a perfect world we would have insulin that is almost as fast as a functioning pancreas !!

@JAB:

While this may sound odd, to me this is an exciting time to be a T1D. We have some great tools right now that are only going to be getting better in the coming years.

I know that many wish that the initial Control-IQ was more aggressive in bringing down high glucose levels. In fact, I suspect that you’ve read a number of the “tricks” folks have found to make that happen.

I personally believe that Tandem tried hard to NOT make their first closed-loop offering overly aggressive.

After all, if a bunch of us began showing up in the ER flopping like fish or worse, FDA approval would be pulled and then where would we be?

We all know that a bolus of insulin can cause glucose levels to plummet … and the only “knob” that can be turned to slow (but not really reverse) than trend is to slow or stop basal insulin. Well, we can eat something appropriate, but the pump can’t control that …

I DO expect that we will see rapid and significant improvements in control algorithms and I also expect that we will see dual channel insulin/glucagon pumps that WILL allow tight, but safe, regulation of glucose levels.

Welcome aboard and stay safe!

John

2 Likes

Thank you for the welcome. As time permits I’m working on a control simulation project to force me to better understand how a insulin pump with control functions responds to BG CGM measurements. This will be a low fidelity simulation at first. If this works then I like to incorporate some of the TANDEM X2 logic and settings to better understand what is going on with BG correction and effects of BG function overshooting, undershooting time lines due to pump / CGM settings.

For me this is a big learning effort on many fronts and beyond my normal ability. It is forcing me to learn
new items and processes.

1 Like

@JAB:

It sounds as if you have a solid engineering or hard sciences background. As a (retired) engineer myself, I think that you will find those skills useful.

The variability in the human/biology side of things is, to me, hard to adjust to:
You (or your wife …) can eat the same thing at the same time with the same insulin dose … and you are likely to see 10 different results. Being able to accept and deal with that can be a challenge, but it is part of the game that we all play. Keeping your sense of humor and not getting to frustrated help a lot.

Best of luck,

John

2 Likes

I must smile reading your thoughts. These uncertainties/variables is what makes this disease so hard to deal with. Any way …I do have an unpracticed control background. Unfortunately I have not worked numerical manipulation or simulation programming in many years. While their is some usefully approaches to statistically handle uncertainties I’m not ready for that at this time. I still trying in code to getting a handle on the basics. I’m using GNU Octave for programming the algorithms and displays to set input/output/results. Fortunately there is a lot of useful code on line that is aiding my effort.

1 Like