T:slim X2 has different IOB values if CIQ enabled/disabled?

Here is an example of what I’m confused about.

When CIQ is active my Tandem t:slim x2 shows my IOB as (for example):
Insulin On Board 2.43

If I turn off CIQ without changing anything else then this changes to:
Insulin On Board 3.71 4:31 hrs

The (Insulin) Duration time in my pump’s personal profile is set to 5:00 hours. So it matches the duration value which CIQ ostensibly uses. Because the configuration is the same I expected that the pump would calculate the IOB the same way whether or not CIQ is turned on. :confused:

Does anyone know why on earth turning CIQ on results in a lower IOB value? This makes no sense to me. :confounded:

Also, why does the pump display the “time remaining” for IOB when CIQ is turned off but does not show the remaining time when CIQ is turned on? What’s up with that?

Tandem made a revolutionary change to the insulin decay model when they introduced Control-IQ. It more closely resembles how insulin is actually used by the body.

Previously, every Pump to my knowledge used a linear decay model:


You used up insulin at the same consistent rate. We basically set a number for DIA willy nilly based on when we FELT the insulin was no longer effective. Usually somewhere between 2.5 and 4 hours, after the major insulin peak wore off. But it wasn’t true… There was always a long-acting tail that wasn’t accounted for and our insulin usage is ANYTHING but consistent.

With control-ig, Tandem changed the decay model to be an exponential decay with a delayed start. I don’t have a specific picture from Tandem, but the graph is going to look a lot more like a rollercoaster. Starting out at the baseline, peaking activity 30 minutes after bolus, then slowly tapering off to zero at the five hour mark. It’s actually approaching zero for a very long time, hence your pump will display 0 IOB long before the 5 hour mark, and really aligns well with anyone previously using a DIA between 2.5-3.5 hours. It’s also why they locked the DIA in at 5 hours, because they don’t expect us to understand the change they made here.

EDIT: Came back to say that the very first section of this graph isn’t a true representation, just the closest I could find. The start of the graph actually should be up high, even with the peak, but it doesn’t really start to significantly decay until after 30 minutes or so.

You can see how the new model better aligns with real insulin usage:


There are other complications with Control-IQ, because it is often adjusting your insulin. If it delivers a correction bolus or increases your basal, it will add the extra insulin to your IOB. If it suspends delivery or reduces basal, it will subtract that insulin from the IOB. The ticking clock telling you when your bolus expires is no longer relevant. Your IOB is now a living value telling you and the pump exactly how much potential the insulin you have onboard has to lower your blood sugar.

However, the moment you toggle Control-IQ off, it reverts back to the old linear decay model with the ticking clock you define.

My understanding is that Medtronic pumps use a non-linear decay model. Most likely a polynomial 2’nd or at most 3’rd order approximation. I’m surprised that Tandem would still be using a linear model. :disappointed:

My perspective is that, technically, basal insulin should not be included in the IOB since it is intended to mimic the steady state trickle of a normally functioning pancreas.

However, CIQ probably has to somehow add basal amounts that exceed the basal rate specified in the personal profile as IOB. But I wouldn’t expect it to subtract amounts that weren’t delivered. They simply wouldn’t be added to IOB total. No?

I stand corrected. I had to research this more. Apparently the 670G does indeed use what they’re calling a “curvilinear” decay while in auto-mode, so I’m sure the latest pumps do too… It’s only Omnipod that’s strictly linear now. It’s all very proprietary, though, because this insulin decay model is what’s setting the new automated systems apart. I’ve been struggling to piece together my knowledge from various sources and paint an image others can understand.

Outside of pump automation, though, you literally just set a time all your insulin supposedly disappears. So the point still remains outside the latest models.

For what it’s worth, I found a link to a reprint of an old article by Gary Scheiner apparently in Issue 22, 2010 of Diatribe.

Because the article is over 10 years old it refers to pumps that are no longer with us. But the paragraph below seemed to confirm my recollection that Medtronic pumps use a non-linear algorithm.

Medtronic (and all other pumps that we will discuss) calculates IOB using an algorithmic equation so as to match the normal ebb and flow to insulin’s activity. Only a small amount of insulin is used up during the first 30 minutes, a great deal is used up during the next couple of hours, and only a small amount is used towards the end of the insulin’s action curve.

Yep, I already conceded that I was wrong to describe it as strictly linear… but still the algorithms have evolved and it’s very different with Control-IQ.

Tandem partnered with Type Zero to fine tune the decay algorithm for Control-IQ, and it is different than previous software models. Omnipod is working with Tidepool to figure out their take on it, and Medtronic is doing their own thing… They’re all new, different, and diverging. Notably, unless you have absorbancy issues, 5 hours DIA is too long on older pump software. 3.5-4.5 hours is generally recommended, and tweaked from there. You can’t revert to an old algorithm and expect the same numbers.

FWIW, AIT (Active Insulin Time, same as DIA) in the 670G is one of the few user-adjustable settings in automode and has the greatest effect in making the algorithm more or less aggressive. As such, it bears NO relation to the setting you’d use in manual mode. In manual, mine is set to 4.5 hrs, but when I was trying to get auto to work I had it all the way down to the minimum 2hrs and would have set it lower if I could.

1 Like

I had not yet seen your post when I replied with the reference to Gary Scheiner’s short article.

But I’d also point out that you referenced the 670G and seemed to imply that non-linear was new & unique to the newer pumps and algorithms. That is not how I understand it. Medtronic was apparently using non-linear IOB in it’s Paradigm pumps 10 years ago. This is why I’m really surprised that Tandem would still be using non-linear outside of CIQ.

I’m considering trying to set up my 630g to mirror my Tandem config and see how much the IOB values differ. I am curious how it might compare.

Previously, before using t.slimx2 with CIQ, IOB was very important to me for calculating corrections to avoid stacking. Since I use Sleep Mode 24/7 and CIQ does not give corrections, I was trying to base my corrections on IOB but found that there was no consistency to the number and, of course, no time indicated. It puzzles me that IOB is still there but offers no useful information. At least, no information I find useful for corrections. So, why is IOB there?

Not one pump manufacturer has spelled out exactly how their algorithms are defined. It’s left to outside people to extrapolate those details. I’ve been going of of things I’ve discovered by listening into investor info, following the hype pre-FDA approval, reading other people’s extrapolations, and certain conversations I’ve had with the design team at Tandem.

I’m half-guilty of over-simplifying things, Your knowledge base is more extensive than your question indicated to me. And also, I learned some new info myself this afternoon. The article I was reading when I conceded on the “linear” point was comparing the Omnipod, 670g, and X2, hence my specific mention of the 670g…

I honestly did not know pumps were previously using “curvilinear” instead of just linear. In all my readings and conversation I’ve heard them referenced as simply “linear”. It’s honestly splitting hairs, though, and I can see why it’s been shortened.

Curvilinear is still a simple equation. It’s still a defined rate that expires at a specific time. The Control-IQ algorithm is different. A 10 unit bolus will last longer than a 0.5 unit bolus, because of the exponential decay. The closer to the five hour mark you look, the more similar the two boluses will look. But if you look at other time segments, the difference will become apparent. The 0.5 unit bolus will quickly devolve into an IOB too small to register, which is displayed as “0”, whereas you’ll still see IOB at the 3 hour mark after the 10 unit bolus… Plus/minus whatever changes CIQ made to your delivery.

Yep. It is highly variable. Like I tried to explain earlier, it’s a living value now. Constantly changing. The clock doesn’t matter.

It is indeed still useful information, though. If anything, it should be easier for us.

Let me ask, why does the time the insulin was delivered matter to you? I unit is 1 unit, and carries the potential of your correction factor, so why does the clock matter? Do you place different value on the IOB if it was delivered at different times? Have you from previous years of experience learned that the IOB display is more accurate at some times than others? I’m assuming the answer is something along the lines of “the previous decay models were inaccurate, but I’ve learned to accommodate for them.”

Control-IQ has made advances to correct that notion. It is all about insulin potential now. If your correction factor is 1:50, and the IOB says 1.5… you currently have enough insulin POTENTIAL on board to bring you from 175 to 100 mg/dl. You don’t have to stop and fret about exactly when that insulin was delivered, so you can determine it’s value. Most of that IOB was probably from your bolus, but some of that insulin was delivered repeatedly every 6 minutes as basal adjustments. There’s no one single clock counting down. A new clock starts every six minutes, and they all run simultaneously.

The IOB indicated the value of insulin in your system that is in excess of your basal, not just simply your last bolus. IOB is reduced if basal is withheld, because that suspension reduced the BG-lowering potential of all the insulin in play at that given moment. Likewise, IOB increases if extra basal is given, because you now how increased potential to lower BG. We’re blurring the lines between defined basal and bolus roles.

1 Like

@Robyn_H, it is not the time of IOB that matters to me but the amount. And the amount seems to vary constantly. Yes, the amount may be enough to bring me into range at one point but several minutes later it may not be right and may in one instance bring me low, in another, leave me higher than I would want to be. I suppose the thing to do is not to correct myself. However, the range I strive for is not the range that CIQ is set for.

I am in the process of following your suggestion to adjust the basals as soon as my new CGM transmitter arrives. Had a bad one and no backup and have had a dumb pump throughout the holidays.

I love your explanations. They have been most understandable and helpful. Thanks.

I was editing my above post as you were writing this last response. Maybe go back to read the last paragraph or two i changed about insulin potential?

I also suspect you’ll gain more confidence in trusting the IOB once your settings are situated correctly. I think your correction factor probably still needs tweaking if you’re getting erratic results.

You might also just be overthinking it. I think we sometimes run into a problem with being TOO informed and trying to out-think the technology. Remember, this stuff is written to bring good control to the average user, not as carrot dangling for the privileged. They’re putting in more effort on the design end to require less thinking on the user end. Could you try just giving the IOB a little blind faith for a bit? Maybe only give corrections with the bolus wizard, and let it do the thinking?

Also, I’m under no disillusion that Control-IQ is perfect. There are significant changes coming in the near future, and maybe you’ll be more satisfied then. I don’t think you’ll ever see the return of the countdown timer, though. I don’t think it’s feasible when the system is always changing your IOB.

And for the record, you can indeed see exactly when the last bolus was delivered and how much if you click on the top, right hand corner of the pump to open the status display. It’s only a partial picture, though. That won’t include any changes CIQ made.

Yes, @Robyn_H, I am overthinking it as well as unable to give up control. I gave CIQ full control for the initial 3 months last year and ended up with a higher A1c. I would be happy with the same A1c as I had before IF there were less work achieving it. Night times are beautiful; day times, not so much. I will resume tweaking basals. Appreciate your help.

1 Like