Dexcom g6 sensor accuracy drops ~day 10



I’ve read some other threads that have noted a decreasing trend in dexcom g6 accuracy, specifically around day 10. For my past 3 sensors (from 2 separate boxes) I’ve noticed that accuracy is pretty on target (within 20) for the 1st 9 days and then day 10 goes haywire (sensor reading off by 50ish, before and after the 1st calibration id give for the sensor). The first sensor this happened to I just took out. However, for the second and third sensor, I kept them in and extended to see if I could get the sensor back to being accurate via calibrations/waiting it out. Day 11/12ish I found that the sensors bounced back on accuracy (again within 20 of finger sticks). The first sensor lasted 21 days and the second I’m currently on day 17 (And still with great accuracy). This is only for 2 sensors but it is kinda strange that on day 10 I’d thought these sensors were complete goners and yet they still had more than a week of left of good readings for me. So if you hit day 10 scaries, maybe just try to wait it out or try a day of calibrating? If it’s a lost cause anyways might as well try was my mindset

Has anyone else noticed this drop and rebound in accuracy???

@Michel feel free to move this topic or link it if this is covered elsewhere already!


I’m trying to keep the conspiracy theorist side of me quiet right now…

I’m someone who still has a land line, just for perspective.


Land lines are just as easy for the NSA to intercept. No difference.

(In case that helps)



In terms of G6 accuracy, we do factory calibration and very rarely provide any manual calibration.

We find that Day #10 is just as accurate as in the beginning. I expect accuracy far greater than the crappy standards that Dexcom holds itself up to for the G6 (its own topic).

My expectations for the G6 are to be within a single digit difference between the G6 and our Contour Next One meter. This is regardless of the BG level although at level cgm. I do not get concerned over how accurate it is during rapid rises or falls and hence do not get into the lag aspect. Not that those all are not valid - simply not part of my consideration for whether the G6 is accurate for us.

I find the first 12 hours (14 hours if you include the 2 hrs warmup) are outside the bounds of what I consider accurate but enough to show trending information. But certainly not good enough to bolus from. During the initial 12 hours (14 hours including warmup) if a bolus is needed, we will do a fingerstick and bolus from that.

I will also check meter against cgm when the 2 hour warmup has completed and we have 3 or 4 cgm data points so I can see the beginning of a trend. At that point, I expect the G6 to be reasonably close to the meter. I will check again about 12 hours later and at that point, expect the difference to be single digits.

By day #10, we will get drops and gaps in readings but when the cgm is reading, I still expect extremely good accuracy.

So far in our experience, a good sensor will stay good until it fails.

A bad sensor is bad from the get go. As well, a bad sensor can not be resolved via calibrations. (Our experience.)

BTW - New sensor put in last night. Just did the “12 hour” fingerstick check this morning. 6 points difference. Good sensor. The sensor had a bit of bleeding about the size of a dime, more than typical but less than “bad”. So I was not sure if this was going to be a problem on this sensor but it looks like not an issue and all good !!!


So I’m an uninformed conspiracy theorist!! LOL


What percentage gaps/ missed readings do you get, around day 10, and beyond?


I don’t track it. Only eyeball.

I generally ignore small drops and certainly ignore anything where one device (ie - Phone) loses the signal while the other device (ie - Pump) shows a signal. In that case, obviously the problem was the iPhone (IMHO). FWIW - with the G6 and the Basal-IQ update, the pump loses the signal far less then the iPhone loses the signal. It is not uncommon to require the iPhone to restart the Dexcom G6 Mobile App or on occasion do a complete iPhone reboot.

A drop that is the fault of the transmitter/sensor would show up as a drop on both the pump and the iPhone and would typically be a drop in the range of 1~3 hours. If we get two of those on day #9 or day #10 then I am not going to bother extending the sensor for another session. IMHO not worth it. The only condition I will extend the G6 sensor session past its initial 10 days is near perfect conditions right up to the end of the session on day #10.

Bear in mind that we use the data from the cgm to drive the algorithm on the pump. So this is far more then simple trending data for us.

Our last sensor (replaced yesterday evening) was getting minor fluctuations on day #9 and had a drop of maybe 2 hrs on day #10. So when the session ended, we pulled it and replaced it.



As long as one is aware, it’s not a big deal until one is wearing a pump that is making insulin delivery decisions based on those numbers. The Dexcom lag is far less of an issue anyway.

Agreed. Not all initial 12 hours are like this, but it’s not uncommon.

Pretty much agreed. I can’t say it happens on day 10… or that it will happen on every sensor, but I have hit a pretty spotty 12 hour period on a couple of sensors I’ve gotten past day 12. I’ve dome a few calibrations, and it’s tightened everything up. Again, I haven’t been wearing it long enough to speak in anything more than broad generalizations.

Lucky. I had to take one out that had a spot of blood— nothing remarkable. However, I still have a silver dollar-sized bruise, so I guess it was more than the blood that caused the problem.


@Thomas: thanks for the thorough reply. I certainly have less higher expectations from my g6, but after attending the jdrf local summit today, I’m realizing that my finger sticks may not be as accurate as I’d thought. I use freestyle lite and freestyle flash meters (and am going to look into their accuracy because they may be more off than my dexcom g6 sensors).

I also am not using dexcom data integrated with my pump so I’ve been happy with its accuracy (besides the seemingly strange and off day10s). I only manually calibrate when the blood stick is off by more than 30 and I’ve done multiple blood sticks and used that average.

But it is interesting to hear that if you’ve got a sensor gone bad that there’s not going back for it. I may have just had two anomaly sensors


I’m also putting myself in this category for sure!

And yes I agree with both of you on not checking blood sugar against sensor readings when I’m trending up or down!


I’m going to argue the other side of that. Even if the CGM is wrong by 15 or 20 points, that doesn’t have to cause a big problem with a closed loop system. Because a BG 15 or 20 points off target isn’t unsafe, and anyway the closed loop will do the right thing: If there’s too much insulin the BG will fall below target so the closed loop will cut insulin delivery, and if there is insufficient insulin the BG will rise above target so the closed loop will increase insulin delivery. It’s still doing the right thing, just acting as if your target was 15 or 20 points different from what you’d like.

Of course this depends on a decent closed loop algorithm, and accurate CGM is greatly preferred to inaccurate CGM. I’m just saying that 15 or 20 point errors in CGM readings don’t necessarily preclude successful closed looping.


I don’t think we’re disagreeing. I think we’re talking about different numbers. I’m really thinking about the lag… and the Guardian has got a crucial lag. If it takes my sensor 20 extra minutes just to clear a 50 again, it could mean an additional 40 minutes of no insulin.

15 or 20 points is not going to be that big of a deal… although now I kind of want to debate even that, but the lag can make for a completely different experience.

And just because I can’t not— my Dexcom, for whatever reason, reads about 10 points higher than my BG on a regular basis. I can’t be sure it’s having any kind of significant effect on what Basal-IQ is doing for me, but my gut is that it’s to my benefit. So I’m not entirely sure even that small difference, with a pump that’s targeting a specific blood glucose value, doesn’t have some impact. It, alone, might be inconsequential, but then you throw the lag in on top, and you’ve got… hold on… let me go get some carelink reports…


I agree that lag is a nuisance, especially when it’s variable. It tends to lead to overcorrections in both directions.

If you think that’s a stable typical thing and you don’t like the result, you could change your BG target and alert thresholds to compensate. Or just continue to enjoy the lower A1C the error would give.


I haven’t noticed any drop in accuracy over periods up to about 25 days, however there is a clear increase in sensor failure; that’s why I say 25 days.

Dexcom document this; they have about 5% sensor failure (total) up to 10 days. It isn’t clear what amount of that 5% is wardrobe malfunction and what is a sensor failure reported by the transmitter.

Transmitter sensor failures seem to correspond to the sensor going haywire; those I have experienced include massive single drops in BG readings. They do seem to be recoverable, my last sensor was replaced this morning because I had one recovered sensor failure yesterday mid-afternoon then experienced another early this morning. There comes a point where you have to donate your car to NPR.

It is true that I don’t run sensors to 10 days; I use xDrip and preemtively restart at 9 days. However I’ve lived almost 20 years around conspiracy theorists and know a grassy gnoll when I see one.

What I see is a sensor that maintains a broad level of accuracy but does, at some point, fail. I can’t get any trend in that accuracy, but it’s pretty clear from Dexcom’s published information (user manual) and my own experience that the system - sensor+transmitter - fails at some point.


I do like the result. It’s faster to resume insulin, and that works well for me. I haven’t looked into this though. That’s definitely good to know. Thanks. :slight_smile:


Are you running your G6 on:

  • 100% on Factory Calibration (ie - no manual calibrations at all)
  • Mixed meaning starting on Factory Calibration but then adding one or more calibrations during the session
  • 100% on Manual Calibrations (ie - not putting the sensor code in at all and not using Factory Calibration in the slightest)

For a given sensor session. And specifically about the initial session for a given sensor and not talking about any possible extended sessions.


Mixed. I do put in the code, and I only calibrate if it really is just not a satisfying readout. And, believe it or not, I’m patient. I never mess with it with a moving blood sugar, and it really has to be “just not right” for like a day before I’ll calibrate.

Actually, that’s not entirely true. Between 12-24 hours in, if I don’t like what I’m seeing, I’ll calibrate. What do you think about that??


I’ve been the same way. Then when I get ~day15 or so I’m more liberal with calibrating.

I just put in a new sensor last night and will try the

Also, I’ll record my day 8, 9, 10, 11 BG vs SG to see if my last two sensors with that drop in accuracy around day 10 were just chance and anomalies


If you are restarting, the time of restart would be good information. Some people who restart will restart early. So, day #10 - was that before or after restart?


Do you think restarting may effect accuracy in that time-frame?
So the 1st sensor that went haywire I was planning on restarting midday day 10, but I pulled it
2nd sensor, I restarted day 9
3rd sensor, I restarted day 10 (can’t remember when, but I want to say in the mid-morning)

2nd and 3rd both had inaccurate, wonky day 10s, I believe after restarting in both cases. 1st sensor was inaccurate day 10 without restarting.

This new sensor I’ll record when I restart