How do you handle off sensor readings in "auto mode"?

The input for all kind of “loops” is information from the sensor. If this info is no good, how well does the system perform?

And how do people handle this?

Question goes for all type of sensors and all types of automated insulin delivery systems/commercial closed loop systems.

Reason I ask is I am not yet to use a commercial closed loop system, but I will get the chance some day soon.

I´m an experienced CGM-user since 2013 and really, really often my sensor is either:
-way off (like 30-40%)
-drops suddenly for no reason (not compression)
-raises way too high
-stays too low for hours at the time

I hope this will be a thread for all of us who at one point or another will use Medtronics 770G/780G, Tandem t:slim X2, Omnipod 5 or any other insulinpump.

All responses appreciated.


If it’s giving readings that are wrong, I can calibrate but that won’t necessarily fix a sensor problem.

The whole pump algorithm requires accurate data and if your CGM isn’t providing accurate data, your insulin doses will be wrong.

Luckily for me, Dexcom is very good in my use. I rarely restart sensors, don’t need to “pre-soak” and calibration is pointless - they’re very close whenever I check (just not in the middle of rapid changes due to recent eating or bolus).


I’m using the Tandem with Control-IQ. About half of my sensors will read very low for a few hours after I start them. Low as in, below 40 sometimes. I’ll calibrate once or twice, but they can be low for up to five or six hours. I need to turn off Control-IQ, because if I don’t, the Tandem will set my basal rate to zero, and my real BG will go sky-high. Then I have to remember to turn it back on. I set an alarm on my watch to remind me.


With Omnipod 5, manual entries for BG will override your CGM reading and trend for bolus dosing. And the rest of the time, a user with a bad sensor or inaccurate sensor would be advised to run in Manual mode which operates it solely from the basal settings entered in the pump software. Manual entries for BG would still be needed for bolus.

I think it boils down to a question of how much time per day will you have a reliable CGM? Unless your CGM is reasonably accurate most of the time there’s really no point in attempting to use closed loop pumps.


Some people do better if they “pre-soak” the G6 sensor. This means you insert a new sensor 2-3 hours before current one ends.

Being well hydrated can also help.

When I do this, I also do fingersticks and calibrate if necessary.


Medtronic 770G / Guardian 3 here.
Day 7 is the only day where I’m sometimes out of sync between the sensor and finger sticks. When that happens, I pull the sensor and go on manual until it’s time to put the new sensor in on day 8. It’s like a mini-vacation from alarms :slight_smile:


In normal operation all single off crazy readings will simply iron themselves out. If the reading is within 15% high or low the sensor and the meer are considered pretty much in sync depending on the meter. Remember the meter can be just as much off as the sensor. Take the two at each extreme and a reading of 100 might be 85 or 115. At 200 your blood sugar might be 170 to 230 and both extremes are valid.

Likely at 300 it is more like 20% either way. So 240 to 360 and both extremes are valid.

Now sensor companies ask us to calibrate to lower these difference. but they are being changed to read more like your meter. Even if the meter is off.

So here is my advice, worry less about about the degree off unless it is a constant. Hey i go ballistic if my sensor reads low and my meter reads high. I go crazy as my wife.

For me with a Sexi Dexi - day 1 and 10 are funky almost always.


Thank you all for all of your replies. Just keep 'em coming.

All input is food for thought and better decision making for all of us.

1 Like


I am one of those lucky people for whom G6 work pretty well. When I first got my G6, I did occasional finger sticks and found that if my CGM trend arrow was flat that the agreement between CGM and BG was good enough that I was happy to have Control-IQ do its work.

I don’t have to “soak” sensors during start up and, to be honest, now go through most 10-day sensor periods without a single finger stick. Note: I’m not recommending that to anyone … just saying that is what I do.

I certainly don’t understand why some of us get reliable data from the start, others have to “soak” for 12-24 hours, and still others seem to have an ongoing struggle with CGM accuracy.

It is certainly true for ALL feedback systems that crappy sensor data results in crummy feedback control. I offer the Boeing 737 Max angle-of-attack sensor …. I also think that as the first-generation closed-loop system, the Tandem Control-IQ intentionally provides a “weaker” feedback signal than many would like. As I recall, it tends to give you only 60% or the ideal correction dose. While I have no inside information, I suspect that part of the reasoning behind that is to avoid over correcting if your sensor happens to be reading HIGHER than your actual BG. Note: if your CGM is reading LOWER than your actual BG it will actually give you less of a correction dose or less basal than C-IQ would give it the sensor was perfectly accurate.

You outlined 4-5 ways that you get funny readings. I would suggest that for each type of “crummy” data you ask two things:

  1. Out of a 10- day sensor life, about how many hours do you see each type of bad data.

  2. For each of those types of aberrant data, which ones will cause C-IQ to overcorrect and which ones will tend to under correct. My instincts … as an engineer, not a CGM professional, tell me that errors that cause C-IQ to under correct are less troubling than those that would cause it to overcoorrect.

Finally, I firmly believe that CGM sensors are going to get better and better and that closed-loop administration of insulin (and at some point glucagon …) are also going to get better and better. Even if you conclude that first-generation C-IQ isn’t for you, I predict that you WILL be happy with a better sensor and a better version of feedback. Will the G7 and its matching version of C-IQ (or equivalents for other CGM sensors and pumps) bevtfat system? To early to tell …

Best of luck,



This is a really good tip!
I´ll do this over the next month or so. I think this will help me decide if any kind of “Loop” will work for me. You got to love an engineer, right? :blush:

1 Like

@MBW, @Michelle43, @John58, @MM1, @Rphil2 and @John_S2:

You all use Dexcom, so this one is for you:

When the trend arrow is missing Dexcom says the sensor values should not be used for medical decisions.

Does Omnipod 5 and Tandem X2 “see” this missing arrow when getting values from the sensor?

Or is it still considererd valid and usable data?

My trend arrow is almost always present. The rare times it has disappeared it’s came back by the next measurement — so, in five minutes. The loss of feedback isn’t really an issue for me, assuming the pump goes to non-guided mode. I know the couple times I’ve attempted to bolus during a trend arrow loss, the pump asks me to input a glucose level, even though it shows up on the pump. So it must not trust itself at that moment!


I could not find the exact calculation but I believe if Tandem finds the arrow is not moving in three five minute cycles it produces a read error. Does anyone know for sure?

1 Like

It would be nice if there was some kind of a box that that you could just turn on and it would completely solve a problem, something like Tesla’s completely autonomous cars or a truly artificial pancreas.

But there isn’t, and they’re never will be. Nothing physical is eternal or can be 100 percent reliable.

Good BG management requires awareness, data and comprehension. Asking questions about failure modes is a good start.

No blood glucose measurement system is trustworthy and reliable, not A1c tests, CGMs or fingerstick glucometers. Of those three, the fingerstick is the more reliable for making management decisions, IF it is used properly.

No automatic pump system is a substitute for a person with diabetes intelligently managing their blood glucose. These are tools. The best experimental pump makes it possible to improve the BG management of a poorly managed PWD but not the BG of an expert manager using more basic tools.

A hybrid closed loop system is designed to TRY TO keep BG in a defined range. It is ineffective in bringing BG into that range from above it or below it. A CGM becomes increasingly inaccurate outside of that range while a finger stick meter continues to work better.

The PWD always has to be ready to assume control using positive glucose or negative insulin correction, exactly the same as with MDI.

With a hybrid closed loop system, an inaccurate sensor reading is handled with a finger stick reading when a meal bolus is given. When confirmed that a high or low CGM is far off, a fingerstick reading is used to calibrate the system and calculate a correction.

With all that being said, I prefer using a CGM and pump to fingersticks and MDI but not because it’s simpler or more accurate. It’s neithe

It is an assistant that can be trusted within limits to alert me when something goes very wrong and to let me know in what direction things are changing. Within those limits it can make simple decisions and report to me what it’s done.

I’m still the boss, I still have to make the hard and strategic decisions. I still have to pay the bills make sure that the stock is an inventory and it’s used properly. I have to know how to run my business and I need to be able to tell when something isn’t working right.

Over time I expect the newer “assistants” that I use to become better and to delegate more things to them, as I already do calculations for boluses and corrections.

But I’d be foolish to ever believe that anything is reliable when it depends on software and small parts. Or that it cares about preserving my life.


I’m a G6 user. Like you I get unexplained Highs and lows
This morning is the prime example. Prior to my Dexcom ending My glucose was a 115. Finger stick had it at one 120.
After warm up and coming on line. It was 170. Back to back ginger sticks shows 125 and 130.
Recalibrated Dexcom and it’s been pretty much in line all day. I’m not concerned over any one reading but in the trend data. To me this Dexcom is just another tool in our treatment arsenal.


I use the 770 with a guardian 3 sensor. I marinate the night before it expires. I find that the less bouncing around with the bloodsugars the better chances it will last the full 7 days.

1 Like

To understand correctly, you put in a new sensor the night before, then charge transmitter the next morning, then attach as normal? I’m curious because my sensor readings are always off that first day. Wondering if this will fix that.

1 Like

You are my twin! I’m a former Animas user - and screen of course started to be an issue. I was using the Dexcom G4/meter along side of it - and was doing excellent with staying in control. Had to stop Dexcom due to cost.

Went onto the Libre2 system as it’s affordable out of pocket. And that worked fine - but then where I live in Ontario - was up for a new pump for the next 5+ years (thank you lovely tax payers) … because of the Dexcom $$ - sadly Tandem is out for me - unless I just go commando like before and not bother with ICQ - but I really REALLY wanted to see what this “looping” was all about.

In end, I’ve been trialing the love/hate Medtronic 770g with the AutoMode (with FREE upgrade - yuppers - I love a sale) - and so far - that algorithm is driving me more bonky wonky. Because you can only look at their reports - rather than Glooko/Tidepool which is more user friendly for the non technical person like myself … I’m finding it all highly frustrating. BG’s according to the free CDE help from Medtronic Canada - I’m the cat’s meow, but so not used to being in the higher ranges during the day time, and especially overnight where I’m averaging 9 mmol/l - 162 mg/dl .

As the Medtronic CDE stated, when I tried the manual mode with my usual basal rates that keep me in relatively better zone (or did on my other pumps for some reason) - even if I tweak my basal rates to try to get to that purrfect OCD control I have (my words - not theirs). As they said to me - In Auto Mode basal is delivered based on our SG every five minutes targeting 6.7mmol/L. So no, changing your manual mode basal settings will only affect things in manual mode.

If I can hang on without forking out the $8K for the pump - trying to see if the 780g will be better for my purrfectionism is the whole thing that’s buggering up my anxiety etc. … and this is the statement from the Medtronic CDE - The MM780G will have target option that you can select-either 5.5mmol/L, 6.1mmol.L or 6.7mmol/L and will also add auto correction boluses once your glucose levels go above 6.7mmol/L and auto basal has done all it can.

Sorry for the mmol/L bits up there - usually I will put down mg/dl … but a lazy cat today … but just multiply those numbers by 18 .

So, as someone who usually does what you do and is their own “manual loopable program” (and perhaps there are others the same here) - I’m highly frustrated - and almost ready to just give up pumping altogether and go back to the poor man’s pump that I used for about 3 years prior to going onto technology back in 2008. I find with all this technology, maybe due to my Golden Years :rofl: - that I seem to live/breath diabetes all the time with all these gadgets and never had the mental anguish that I find is happening with technology that goes bing/bong , etc.

Okay - that’s what I did with both the Dexcom G4 and the Libre 2 … I just love the “marinating” bit Dee - you’ve made me have my first laugh of the day - and it’s noon here in my part of the world as I type this out - and try to concentrate on work after being off for 10 days (holiday not ill - though DH is currently battling Covid (still testing positive - ugh - and feeling rough today - so not like him) that he caught while we were away in Florida - it pays to wear a mask / washy washy.

Yep. I have NEVER EVER trusted the first reading when it comes out of “ warm up”. But I also don’t necessarily recalibrate. It depend on the finger stick and the first reading. If with in 25 I’ll let it alone for a half hour to watch the trend. Yep over bolusing will make it drop real fast.