Detailed description of Tandem Control-IQ algorithm?

Where can I find a somewhat detailed description of the Tandem t-slim X2 Control-IQ algorithm regarding what variables (except predicted CGM value in 30 min) the automatic basal rate as well as automatic correction bolus is based on:

  • Personal profiles/manual basal rates
  • Total Daily Dose of insulin (calculated by the pump or given by the user)
  • Insulin Sensitivity Factor
  • Insulin On Board
  • Weight


I believe what you are asking for is proprietary information and won’t be released by Tandem.

1 Like

This explains how the different settings are used.

Interesting that it suggests some “learning”, so the behavior may vary over time.

How does Control-IQ technology work? [video]


Articles in this section


I watched the video. Sheesh. I will never learn all those “B” colors and their meaning. I wonder how many on C-IQ as intended find that it correctly predicts the 30 minute BG level which should be determined by a ‘normalish’ number at around 30" due to the IQ bolus. I think. I feel like I am embarking on a rowboat that starts in the middle of the ocean. !!

You are not meant to know. First even with all of this information, it lacks the fundamental ingredient. How does the algorithm use each variable? Also what variables are not identified.

In my experience (2 years on CIQ), my primary attention is on the alarms I set that alert ME to take action.

For example, suppose I eat, but forget to bolus or underbolus. CIQ will assist to cover my oops, maybe bolus when predicted to hit 180. But I also get bg alert at 150 and can correct my mistake, in addition to what CIQ did, if needed.

From what I’ve found in different manuals and scientific papers I’m pretty sure the algorithm uses ISF, TDD and IOB to calculate the amount of insulin that should be infused to normalize a high glucose, however, I can’t find any information about if the personal profile / manual basal program matters and is used by the algorithm.

Your summary sounds partially true and partially confused with Medtronic and Omnipod 5. A lot of the information you’re asking about is not actually detailed at all, but can be answered just by looking at Tandem’s website, or by reading the quick facts sheets and/or the user manual.

Control-IQ is nearly entirely based off of your own personal basal profile(s) settings, for better or worse. Simply put, the better your settings, the better your medical support team and/or the better you understand how to adjust pump settings, the better your outcome. Assuming your desired outcome aligns with the Control-IQ targets! The most critical settings are correction factor (same as ISF) and the programmed basal rate(s).

Control-IQ works off of actionable ranges, based on where Dexcom predicts you will be in 30 minutes. Your current BG matters little. There are three different programs of actionable ranges to choose from. Within each program, there is a target zone, in which Control-IQ does absolutely nothing and your programmed settings are allowed to run without intervention. If you’re predicted to stray slightly outside the target range, it will either reduce or increase your basal rate, informed by your correction factor. If you’re predicted to go to an extreme, then you get extreme intervention… Either by means of an automatic correction bolus (only 60‰ of the calculated dose as a safety measure, and the calculated does does indeed account for IOB), or a complete insulin suspension at the other end of the spectrum.

It sounds like you’ve already seen this chart, or at least one similar to it, but maybe didn’t fully understand it? This is literally how Control-IQ works. When it talks about increasing or decreasing basal rates, it’s doing so to the basal rate you programmed, not to one it manufactured based on your weight and TDD.

Speaking of weight and TDD… You will have to enter these values when you first set up Control-IQ, but your total daily dose will quickly be substituted with actual data from your pump’s history. This is the ONE AND ONLY bit of “learning” Control-IQ does. Unlike other systems, these are not used to determine insulin DELIVERY, but rather the exact opposite. They’re used to impose safety limitations on the system, which will override any extreme value the system might otherwise deliver. (This information is in the manual.) Control-IQ knows that someone who weighs 80 lbs and only takes 20 units a day is much more sensitive to insulin than someone who weighs 200 lbs and takes 100 units in a day, so Control-IQ is much less aggressive with that more sensitive individual.

Your programmed basal rate is also really critical to get the correct IOB to display. The IOB is calculated as a net value. it only displays whatever insulin on board is in addition to your basal rate. If it withholds any insulin, it is subtracted from the IOB (the excess insulin you had on board is essentially redirected to be used as basal instead). If it gives any additional insulin, it adds it to the IOB, as we would expect. If your programmed basal rate isn’t true, though, it will negatively reflect in your IOB. If your basal rate is set too high, excess insulin on board will be hidden from you and Control-IQ. If it’s set too low, it will look like you have more insulin on board than is actually true. Since IOB is used to inform the choices Control-IQ makes, basal testing is HIGHLY recommended.

If this isn’t enough and you want to go further down the rabbit hole and really get detailed information about how Control-IQ works… You won’t find everything in one place, but you can piece together a lot of information from multiple sources to get a bigger picture.

Tandem partnered with Type Zero to formulate Control-IQ. Type Zero’s automatic insulin dosing algorithm was based off of years of research coming out of the University of Virginia. You can learn a lot about how Control-IQ works now by looking at that original research. One of the greatest contributions of that original research was a brand-new insulin decay model, that is unlike anything else ever used before in a commercial pump system. It more accurately reflects the true 5 hour long action curve of our analog insulins. All other pumps, and even the T:slim X2 when Control-IQ is disabled, use what is known a “curvilinear” model, which can only encorporate the steepest segment of the insulin action curve. This is why Control-IQ has the DIA locked in at 5 hours, instead of the usual 2.5-4 hours users might be accustomed to. That standard curvilinear model is very much like trying to fit a round peg into a square hole. It just doesn’t match real insulin action well, and the users basically has to decide how much of the lingering tail they feel comfortable ignoring. Control-IQ uses the full picture, though, to better account for the potential impact of all IOB

I seem to recall an interesting discussion about this by Stacey Simms on her Juicebox Diabetes Connection (edited) podcast. I believe it was the episode in which she discussed why her family switched to Control-IQ instead of continuing with DIY closed loop for her T1 son. In fact, she’s been a great source of more detailed information since she’s done several interviews with Tandem employees who have insider knowledge.

If you navigate to the Tandem diabetes webpage, then choose the “providers” tab, you’ll find much more involved information than on the patient side of things. They even share all the clinical research, which contain commentary on the algorithm workings.

And even farther down the rabbit hole yet… They give their investors, or potential investors, way more information than they share elsewhere with the patients. They share the exciting science advancements add a way to entice new investors. I’ve been stalking every investor update for years for exactly this reason. There’s a financials website called Motley Fool, which posts text transcripts of all the back meetings. You can search as far back as 2017 for Control-IQ information.


Great extended comment, Robyn. Just need to correct one relatively unimportant item. Stacey Simms’ podcast is called Diabetes Connections. The Juicebox podcast is put out by her rival, Scott Benner.

Your knowledge of the Tandem Control IQ is impressive! I may end up on a Tandem successor system if Loop ever becomes non-viable and I maintain my marbles as I age. :wink:


Oh no, my wires are getting crossed. Lol. When I keep cramming stuff into my limited brain, some stuff must get pushed out to make room. Thanks for the clarification!


Terrific clarification @Robyn_H !

One thing that it appears to me that Tandem failed to do, is have a sliding scale of CF so that the amount of insulin required to correct a high of, for example 18 mmol/L or 325 mg/dL, is far more than to correct from a high of 9 (165).
Not that I see those high numbers very often, but it does happen on occasion and the higher I start with, the less sensitive I seem to be to the effects of insulin; for me to get from 9 to 6 is about one unit. To get from 18 to six is more like six units of insulin!
Different current glucose levels = different correction factors. The picture of an unstable system. I don’t know exactly how this could be calculated. (Of course for the user it leads inevitably to over-bolusing out of frustration . . )

Totally agree. It’s a fairly common criticism of the system, too.

There are two updates coming in the near future for Control-IQ. The first one is small and just system improvements. “Housekeeping”, they’ve called it. I’m guessing it’s probably the version they just got in Australia. This one also might have the timers for the activity modes, but I don’t know that for sure. Maybe @tedos can tell us if she’s got that.

For the second update, they promise more customization options, but have yet to specify what they are. I’m hoping for this variable Correction Factor and lower BG targets. Medtronic got approved for 100, but Omnipod didn’t even try for less than 110, so I don’t know if Tandem will push for it.

No no timers that I have noticed. But we have manual bolus indicators on the main screen.

Variable correction factor would be useful, it’s obvious the correction factor isn’t a one size fits all, you need a stronger dose to correct a higher BG.


Yes a variable correction would be good maybe they will figure that out for future iterations.

I’ve had my sugar go high and after trying to correct it several times, it alarms and tells me I tried and tried to bring this stupid sugar down but it’s not budging so check your site.

( not in so many words, but is something like that )

1 Like

This is amazing information from all who posted what to me is ‘inside’ information. Even if it reads like Greek to me! All I know, and my endo would agree, is that somehow my educated guesses, and that is what they are, bring my highs, regardless how high, down to about 100. I am lousy with carb counting and the endo group knows that and tells me they can work with it. However, it looks like I may have to get into at least some basics beyond my ‘guesses’! I do have my ‘numbers’ plugged in for sensitivity, etc. So they must be close enough. You guys sound like PhDs in Tandem C-IQ. I had to look up the acronym! That took time and effort on your part. I am in awe.

1 Like

Re high BG and need for higher insulin. My endo, now retired, told me long ago to adjust upward on dosing for higher BG and to realize it will take longer to come down. If I overdose, the line will go down like steep ski slope and I know I best head off the crash. I have learned to let it come down more gradually. With a good dose not too much and not too little. But that is me doing educated guess again.
T1D is not a simple disease and the pump companies need to understand that there is not a straight line textbook approach that fits everyone. That is what has taken them way too long to accept. They know it but they want the T1D to fit the pump criteria vs the opposite.

Truer words never spoken! And this disease changes day to day and week to week, requiring fine adjustments (and sometimes heavy adjustments) to manage.

One thing I have learned:

I now have THREE different profiles which I can choose between, depending on the type of day I’m having or expecting. I call them “Normal”, “Active” which lowers my basal by half, increases my correction factor (more correction for a given unit of insulin) and increases the amount of carbs to insulin. And then there’s “Lazy” when I know I’ll not be able to move much, and it does basically the opposite of “Active”.
I can even use these profiles for a few hours at a time to adjust to my changing insulin needs.

What I would LOVE would be if I could make these changes happen automatically, simply by selecting Exercise or Sleep activity mode.

1 Like

I call mine “IQ” (normal), “low IQ”, and “high IQ”. So I get stupid or brilliant, depending on the day. :rofl:

I love these categories. I will ask my trainer to help me do this with C-IQ. And I like both so am not sure which category group I will use. Ha Ha!! “Lazy” may win, if I’m honest, however. Or “laptop addiction time”. Even if it’s volunteer work on the laptop!

A couple of points to add to Robyn’s excellent summary:

  • CIQ acts on predicted BG values rather than current. In other words, just as it doesn’t wait for you to go low before it reduces or suspends basal, it also doesn’t wait until you actually hit the high BG thresholds before it increases basal or delivers a correction.
  • however, unlike DIY systems which also account for ‘carbs on board’ and carb absorption rates in their BG prediction curve, CIQ only looks at the BG trend. Thus there are scenarios where the BG trend over the last 30 min is to go low, and CIQ suspends even though you have entered carbs which you know will keep your BG up.