Dexcom G6: how is the algorithm affected by calibrations, and why does it take so long to convince it that it's wrong?

I’ve got a very specific question here.

On those semi-annual occasions where the sensor I insert produces wildly inaccurate readings and I have to calibrate (but where it doesn’t accept my calibration for 12 to 24 hours), what exactly is happening?

Is it that my dozen-plus calibrations performed during this period (repeatedly resulting in “sensor error: enter BG in 15 minutes” and no data) eventually “convinces” the algorithm that I am correct after all? (Why does it take the algorithm so long to realize that I’m not lying, and it needs to listen to me?)

Or, is it that the quality or the quantity of the interstitial fluid drawn into the sensor filament is not adequate (for whatever reason) to produce accurate BG data for that large period of time (12-24 hours) and that the dexcom finally begins displaying data when the results sync up with my finger sticks indicate?

Perhaps there is some other explanation that I cannot figure out?

I’m curious to know what the more tech savvy in the community know.

I don’t worry why. I know that if I enter a “halfway” value that is accepted, then do another calibrations about 10-15 minutes later (with bg retest), it usually corrects itself.

3 Likes

I agree with @MM1’s “halfway” tactic. I also don’t calibrate more than 25-30 mg/dL at one time. Calibrating a large difference (between CGM and fingerstick values) in two or three steps also works for me.

2 Likes

I don’t know how the algorithms work but I do know that it works a heck of a lot better for me since I stopped calibrating it altogether (on my doctors advice)

1 Like

The sensors are calibrated in the factory.
When you calibrate it it is only temporary, then it reverts back to the factory calibration. That’s why if you leave it alone it will correct itself, or if it’s a bad sensor it just won’t work no matter what.

The only way to truly calibrate them is to start it with NO CODE. Then it will work like the g5 where you will need to calibrate every 12 hours.

3 Likes

Ok so I’m my situation, the CGM read 100 but my actual BG was about 230.

When I entered 230, I got a sensor error- recalibrate in 15 minutes. In fifteen minutes my BG is still 230. I do this many, many, many times. Eventually, I have to go to sleep. No readings overnight. In the morning, my BG is 150, I calibrate and finally the readings show up.

Are you suggesting that what I should have done is enter 130, then 160, then 195 and then 230? When you do this, does it step up with you, or leave blank chunks of data? How much time exactly do you allow for between calibrations?

And are you thinking that the dexcom thought all this time that my BG was 100, and it wasn’t until my BG was 150 that it was close enough to be able to accept it? Yeesh.

This is awfully unintuitive, but if it works that’ll be a great help.

I don’t usually see this large of a gap, but if I did, I would calibrate with a 130. If successful, the subsequent CGM readings should move closer to the 130 calibration influence.

I’d then wait about 15 minutes and do another fingerstick. Let’s say it’s 230 again and the CGM now reads 130. Yes, then I’d calibrate with a 160 and the CGM should, within 15 minutes or so, move to near 160.

Repeat this process until the CGM is close enough to the actual BG as measured by a fingerstick.

Having said all that, I think there is a benefit for calibration to take place when BGs are much closer to normal, say 90-110.

Yes, I’ve found that to be true for me. Large calibration jumps seem to bring on premature failures. I think you’ll need to do some personal experimentation to discover what works best for you. Pro tip: keep a journal to record your actions. This will shorten your learning curve. After a while you’ll be able to draw some conclusions that work for you.

Please see correction to this advice below!

2 Likes

Ok - I’m not thinking clearly enough this morning! Please disregard my advice above. It is not right. I am sorry to mislead you.

When your CGM reads 230 and your meter reads 100, I would calibrate first with a 200, wait 15 minutes, fingerstick again, and then calibrate with a value that is 30 less than the next CGM reading until your CGM and meter reading are fairly close.

As the CGM value starts to drop, you can do a few more calibrations, always 30 less than the CGM value.

The key concept here is that the Dexcom algorithm doesn’t like it when the calibration correction is too far from what it thinks the current blood glucose is. So, getting it to fall in line is a multi-step process using smaller steps that it will not object to.

What I wrote about keeping a journal still holds. It’s the best way to learn something quickly.

When your CGM reads 230 and your meter reads 100, I would calibrate first with a 200, wait 15 minutes, fingerstick again, and then calibrate with a value that is 30 less than the next CGM reading until your CGM and meter reading are fairly close.

So in my experience, the moment I enter that first divergent BG value (say “200” on a 230 fingerstick for a 100 CGM reading), there won’t be a next value for me to work off of in fifteen minutes, because the receiver will have stoped displaying any data points until the calibration error is resolved.

Occasionally, one or two stray data points will creep through, but mostly I see nothing during these calibration errors.

I have not seen this. Maybe Dex stops displaying when you try to calibrate with a 100 when the CGM thinks 230 is correct. I’ve never seen the CGM data display go blank with a calibration that is within 30 mg/dL.

Does your data display blank on every calibration you give it? If so, perhaps someone else with that experience can comment.

In this instance, the device is recognizing that it is in error and it doesn’t know why. It sees that the error is significant and that something is amiss. It reacts by doing nothing. It doesn’t take your word for it that BG = x. Nor does it believe itself. This is not resolved through calibration. This is resolved by throwing the sensor away and starting a new one. It will display values when/if they happen to sync. I would exercise caution in that scenario. I think its better to just toss the thing. Not everything can be fixed through calibration.

You should NEVER be calibrating more than twice. If you see bad error (defined as more than 20%), your allowed to attempt to cali once. They might let you cali again after letting time pass. But, by the letter of the law (and what Dex tech support will tell you) is to define the sensor as “FAILED” if that error is greater than 20%. If you are seeing error much worse than 20%, that’s a definite ‘Throw the sensor away’ situation.

2 Likes

This is a good answer from @Timothy about how to get around this problem if it occurs frequently. I typically ‘no code’ method for this reason.

Yes, I agree with mohe and the Dexcom reps, call and get a new sensor when this happens. I had trouble with my last sensor and the reps said to replace the sensor. I kept calibrating it until it started working again correctly, and then it worked perfectly for the rest of the time. The Dexcom rep sent me a new sensor anyway, because the sensor had been malfunctioning in the beginning.

I had trouble with my last sensor and the reps said to replace the sensor. I kept calibrating it until it started working again correctly, and then it worked perfectly for the rest of the time.

Hi Marilyn

Yes, this is what usually happens with me when I get in this situation.

I have 12-24 hours of no data due to calibration errors. Then after X minutes pass, the sensor starts displaying accurate numbers and is fine for the rest of the 10 day period.

Also, I’ve have had the situation where I call in the problem, they send me a new sensor, and a few hours (or next day) later the situation resolves.

Honestly, I am a bit tired of calling Dexcom on this issue, because this happens with some frequency (2+ times per year). Every time, I have to go through the call center troubleshooting flowchart “script” which is a tedious and time-consuming chore. Customer service training is far too rigid-- I can tell them the entire scenario upfront, including the fact that I’ve called many times before about the same problem-- and yet they’ll still mindlessly go through their troubleshooting flowchart asking me to answer the same questions that I’ve addressed in my initital description of the problem.

You should NEVER be calibrating more than twice. If you see bad error (defined as more than 20%)

In my opinion, this really should be addressed in the app. It’s a user experience issue- why keep asking me to recalibrate every 15 minues for 12-24 hours? At some point, the software needs to give up and realize that what I’m telling it is true. Or, it should admit that the sensor has malfunctioned and shut it down-- not just send me in this 15-minute calibration loop for half a day or more.

Although I must say that if it shut down my sensor automatically, I probably wouldn’t like it, knowing that, in most cases, the sensor begins working accurately

Does your data display blank on every calibration you give it?

Yes, exactly. Sometimes one or two data points will display, but then after ten minutes or so, will go back to “sensor error: recalibrate in 15 minutes”

@schleima Are you restarting your sensors? It’s the only reasonable reason for such a big discrepancy that I can think of.

It’s completely normal for a restarted sensor to be off more than 100 pts. But you can make the calibration process easier on yourself. Remember, appropriate calibration rules still apply. You need to be well-hydrated, your BG trend needs to be flat and stable (don’t restart a sensor after a meal!), and you need to be in range. I try to be as close to 100 mg/dl as possible once the warmup is done, as I find it makes the discrepancy smaller and requires fewer steps to calibrate.

Once you meet those criteria, you have to make sure to make small steps. 30% different than what the Dexcom says is generally acceptable (that’s when their official rules say you should calibrate), and sometimes you can get away with 40%, but that’s pushing it. However, I never change it more than 40 mg/dl at a time. Big jumps confuse the G6, so you have to take baby steps.

The problem comes with the next calibration. You wait 20 minutes and re-test. If your fingerstick has gone UP, you cannot calibrate it with another smaller number. Dexcom, and the T:slim if you’re using it, will just give you an error… But my Xdrip gives a little more information. It will actually say “confused calibration”, because the lower calibration doesn’t make sense when the sensor data KNOWS your BG was going up, not down.

However, so long as you’re staying flat or even trending slightly down, you can get away with another 30-40% calibration.

If it’s comfortably close, I’ll let it go at that point until the next morning. Since my pump keeps me around 110 all night long, I can easily afford to be 30-40 mg/dl off throughout the night. That gives the Dexcom plenty of time to get used to the new calibrations and avoid the “calibrate loop of death”. Then in the morning, I’ll do one more fingerstick, and can calibrate the sensor to exactly right.

So, in your example where Dexcom says 230, but your fingerstick reads 100:. Calibrate to 190. (30% if 230 is 69, which is bigger than 40, so 40 is the biggest I would calibrate.). Wait 20 minutes and re-test. If your BG has risen to more than 100, do not calibrate again. If it’s 100 or lower, calibrate another 30-40% or 40 points lower than whatever current value the CGM reads. It will usually fall slightly lower than the 190 you initially calibrated it to. So if it says 180 now and your fingerstick reads 95, I would take off another 40 points and calibrate to 140. (180 * 30% is still bigger than 40.). Unfortunately, the initial 130 or discrepancy is bigger than I usually have to deal with, except for maybe on day 20 of a sensor, so the difference is still too great for my comfort. I would wait 20 minutes again to do a 3rd calibration for the day. Be sure not to rush the calibration, even when they tell you 15 minutes, the closer together you do them, the more likely they’ll get rejected. Same rules still apply, you have to be stable or falling. Do not enter a lower calibration if your fingerstick has actually gone up. Assuming you’re now at like 90 and the Dexcom is reading 130, you can now calibrate it to 99. (130 * 30% = 39). That’s close enough for me, and I’m going to bed at this point! (I do tend to restart sensors in the evening.). In the morning, I’ll do one more fingerstick and calibrate it to whatever the fingerstick says. It’s been long enough since the last calibration that the fine detail rules no longer apply.

If you’re NOT doing restarts, and getting results more than 100 mg/dl off, there’s a bigger problem going on and you’ll have to figure out what it is. This is not normal of the G6. Not hydrated enough? Sensor in too lean of an area? Drug interaction? Compression low? First day wonkiness and you’d benefit from pre-soaking your sensor before starting it?

If you go through the initial pain in the butt to avoid the calibration loop of death in the first place, then the calibrations will indeed stick and the sensor will behave beautifully until the end of its life.

If you do fall into the calibration loop of death, yes, you won’t have a sensor number to work off of. You’re going to have to proceed with very cautious steps. Even if the Dexcom isn’t showing you data, you should still have a good idea where it’s at. It’s going to be wherever you calibrated to, plus or minus and real time variations. Start writing the numbers down. I wouldn’t try to keep stepping it closer to true at this point. Your focus right then should be getting out of the calibration loop (which should be avoidable in the first place if you followed my earlier instructions).

You tested at 100 and Dex says 230. You tried to calibrate it to 150, and Dex through a calibration error. It is theoretically holdings onto 150, it’s just not displaying it because it’s confused. Write down both 100 and 150. In 20 minutes you test again and you’re at 110. Do NOT try and creep it any closer to 110. You increased 10, so Dex thinks it’s around 160 (150+10). Calibrate it at 160 now. Write down both 110 and 160. Hopefully it takes this calibration because it’s logical, but it might still reject it. Wait another 20 minutes and re-test. Now your fingerstick reads 105. You went down slightly, 5 pts. Dexcom thinks it’s at 155 now (160 - 5), so calibrate to 155. By this point, it’s getting consistent calibrations and should not be rejecting it anymore. If should finally show a data point, so you can resume the actual calibration procedure. I would wait at least an hour to proceed with actual calibration at this point, though, because it’s going to be even more sensitive than usual.

I’ve been restarting the G6 for about 2 years now, so I’ve got plenty of experience with big calibrations and the awful calibration loop of death. Take the time to calibrate slowly and cautiously to avoid the problem all together!

2 Likes

This is incredibly helpful. Thank you for going into such detail.

To answer your question- no, I’m using new sensors every time (my insurance pays for them at 100% so I have no need to stretch them out).

I have long wondered if it’s a site issue, but I have no way of definitively confirming this. I’ve suspected that when the “calibration loop of death” happens, the filament is perhaps unable to draw enough fluid to get accurate readings, which is why after 12-24 hours it has finally soaked up enough juice to begin working again. If this is the case, an early insertion might help (I’ve never tried it). Also, I’ve long stopped paying attention to the stable state (or lack thereof) of my BG when I change sensors, so thanks for the reminder.

1 Like

Definitely try pre-soaking the sensors, then! It would be such an easier solution than fighting the sensor calibration process! Especially if the sensors start reading well the next day for you.

The thing is, this problem doesn’t with any predictability or regularity, (as I said maybe 2x a year or so) so I don’t usually think about pre-soaking. I suppose I could simply change my practice, but then you need to worry about hanging on to the sensor calibration code paper. I guess I could take a snapshot of it before inserting it.