Control IQ needs more then a little work

A little after 4 it cuts basel to zero. Then while the dots are going down on a 45 it decides to give basel, then when the Titanic goes underwater and splits in two it goes to zero (WAY LATE) NEEDS A LOT OF WORK

Looking at that graph it seems reasonable to me.

You were dropping, it cut basal.
The rate of dropped was slowing and it predicted that the trend would continue.
Instead, the rate of drop accelerated for who knows why and you went low.

An APS can only predict the future with the data it has in the moment and it knows less than the human it is attached to. No single hormone APS will prevent all hypoglycemia and that is not it’s goal.

I’m also not saying this doesn’t suck. It’s T1D, it has it’s moments. But hardware that gets plugged into a biological being to play with hormones will not be perfect for the predictable future.

1 Like

Until, that is, it knows EXACTLY how many carbs were consumed & when, AND whether fats and proteins were present and how much of those.

The low was destined. It should have never increased the basel from zero. The code is flawed.

I think most people would agree that Control-IQ needs some work, and nobody is going to disagree with you there. Even Tandem knows this, and they’re already working to update it. But it’s still really impressive for a 1st generation system.

I’m having trouble understanding your extreme upset with this particular situation, though, because it looks to me like it all went really smoothly. It reduced basal when you were falling steeply, increased it when you were going up again, and stopped it altogether when you were going low. It only resumed when it thought you were heading up again, and only did so briefly and at a reduced rate.

Remember, it’s not a fully closed loop system yet. The technology just isn’t there, in neither the pump, CGM, nor insulin. The system still relies heavily on us… For the correct inputs, the correct settings, and to take appropriate action when the system is unable to intervene adequately. I assume at some point it gave you the “Control-Iq predicts you will drop below 70” alert, which translates to “we can’t save your butt from this oncoming low, eat some carbs, or hold on tight 'cause this ship is going down”.

We don’t know how low you went, but I think it worked pretty well for you. It certainly looks like it behaved appropriately given what your BG was doing at every turn. It only gave you what looks like half of your programmed basal rate for a total of 15 minutes. I’m guessing you were given approximately 0.15 units of insulin since it first suspended. 0.15 units stretched out over the next 5 hours. (The pump gives 1/10 of the basal rate every 6 minutes, so you got 3 micro-doses of about 0.05 units each.) That amount of insulin wouldn’t have been able to contribute to the immediate low…

With Control-IQ, basal is reduced anytime you’re predicted to be 112-70 mg/dl in the next 30 minutes, and completely stopped anytime you’re predicted to be under 70mg/dl in that same time frame. Basal insulin resumes at a reduced rate as soon as you’re predicted to be over 70 again in the next 30 minutes, which you were, and stays reduced until the prediction reaches target. Delivering that tiny bit of insulin was the right call at that moment given your BG trend. It quickly re-evaluated and changed it’s mind, though. Unfortunately, BG is a fickle beast and doesn’t always behave according to the predictions. If control-iq DIDN’T resume basal when out BG starts turning around, we’d have an entirely different problem… Wicked rebounds!

Now all that said, remember you have a lot of control over how Control-IQ behaves, as it’s largely based on your settings. Control-Iq used your correction factor setting combined with the Dexcom 's prediction to figure out how much insulin it should withhold to bring you back up to the 110 target. Correction factor determines how aggressive it is at withholding basal to prevent lows and raising basal to prevent highs. If you think it should have withheld more insulin, this would be accomplished by lowering your correction factor to a smaller number. Note, I’m not actually suggesting you do that. A brief excursion below 70 doesn’t mean your settings are bad. AID algorithms aren’t targeting 100% TIR yet, just lessening the extremes. But if you really do think Control-iq is behaving badly, then your settings are worth a look.


Tony24, from my experience, Control-IQ worked as it does for me.

Furthermore, Robyn_H’s explanation is spot on from my understanding and use of Control-IQ.

When I started using insulin, I was horrified that glucometer readings were +/- 20%. How could anyone really manage their glucose level with a 20% SWAG.

I just looked up on Dexcom’s web site and see that the G6 is still rated +/- 20% for accuracy. Meaning that if your G6 reads a glucose level of 100, your actual level could be as high as 120 or as low as 80.

This confirms my belief that the functioning of Contro-IQ, in my case/experience is unbelievable.



I’m still making mistakes and learning. But can understand that your settings really matter (what is preset basal rate, does it reflect your needs, is it finely tuned enough with different rates throughout the day and night? What is insulin to carb ratio? What is correction factor?)
Then, having a good carb counting practice, and accounting for fats that may slow down and extend absorption, and the additional carbs that proteins provide.
And the impact that any exercise has, especially if there’s insulin on board.
I’m frankly amazed that Control IQ does as good a job as it does.

1 Like