Doc response

[mohe0001]

OK, let me summarize what I have read after exploring this a bit.

Most people prefer Dexcom over Libre.

Sometimes people claim that they believe they got a bad box of sensors (and sometimes a bad transmitter). Dexcom has changed their tech and their manufacturing processes, so this is possible. People say that you should just call Dexcom and have those sensors replaced by them. Dont expect those to show up quickly, so if you are new to the device, you might expect to come up short on sensors (consider having a stash of Libre in case you need extras). Dexcom will replace 3 sensors after you call them on the phone - 3 at one time. So, you may need to call more regularly than that.

I find that I am most likley to run short on supplies right before a Doc visit, which is a disaster.

Dexcom is currently advising you deal with sensor error through the 20/20 rule. Call the nurse line if you want them to walk you through how to deal with sensor error by calibrating. DO NOT calibrate more than once per day.

I think that you want to exercise caution with calibration. It might make things better, but it might also make things worse. I declare a sensor faulty when I see 40 - 60% error. Thats when I, for sure, swap it out for a new one.

Some people think Dex reads a little low, but I hear that about Libre, too. SOme people get very good accuracy. I’m getting good accuracy in my new set of supplies.

If you get variation between sensor and manual stick (or, sensor and lab), you never trust the sensor. The sensor can have error.

20/20 Rule:

If your manual finger stick differs from sensor reading by more than 20 points (when BG is LESS than 80), then calibrate the sensor according to the finger stick.

If your manual finger stick differs from sensor reading by more than 20% (when BG is GREATER than 80), then calibrate the sensor according to the finger stick.

How do they calculate 20%?
Lets do an example so you can learn from my mistakes…

[mohe0001]

Wherever cali = 1, I calibrated. So, you can already see that I calibrate too often in this data set. That is likely to increase sensor error.

EXAMPLE 1.
The first sensor reading from the G6 = 60. Finger stick = 73.
So, they are pretty close to one another.

But, lets check if sensor passes the 20/20 rule at 8:30 AM…
73 is LESS than 80.
73 < 80, so we want the sensor to read within 20 points of the finger stick.

73 - 20 = 53
73 + 20 = 93.

So, if the sensor reads between 53 and 93, it is within Dexcom’s desired accuracy.
My sensor read 60.
PASS.
I should NOT have calibrated here.

[mohe0001]

EXAMPLE 2.
At 10AM, I have sensor = 84 and manual = 135. Does this pass the 20/20 rule?

135 is GREATER than 80. So, we do the calculation a different way.

First, multiple 135 by 20%. To do that, we multiply 135 by 0.2
135*(0.2) = 27 points
27 points is 20% of 135.

So, if our sensor is accurate, we want to see it fall within 27 points of the finger stick.

135 - 27 = 108
135 + 27 = 162

Sensor should fall between 108 and 162 if it is accurate.

Sensor reading = 84.
It FAILS. 84 is NOT in between 108 and 162.

We should calibrate the sensor.
Note in my records that I did NOT calibrate here, so I made a mistake.

If your sensor is failing the 20/20 rule, call the Dexcom nurse help line for advice.

[mohe0001]

I go pretty easy on the sensor. If I see results that are wildly different, like 40%, again and again, then I throw the sensor away and replace it with a new one.

[CJ114]

You may also want to mention to never ever calibrate unless your trend shows a horizontal arrow for at least 15 minutes. You need to do a finger stick for calibration and calibrate only when trend arrow is in horizontal mode for at least 15 minutes.

[mohe0001]

Yea, he/she is a Libre user, so its worth mentioning.

But, I didn’t mention that because Dexcom didn’t mention that, specifically, to me, for G6. G6 is different. This gets confusing. I was, initially, only calibrating during stable sensor behavior. But, that might be the wrong way to do it, according to what they say now.

I was also only calibrating when the values were similar because that’s what I used to do with G4. But, its worth discussion because G6 doesn’t recommend calibration when the numbers are similar. It recommends cali when the numbers are dissimilar. Different, right? You raise a good point. I think it warrants discussion. This has led to confusion, on my part.

Note: @CJ114, see how close the values are at 8:30 AM? Even though I have a down arrow, I calibrated because the numbers were so similar. That was a really accurate sensor reading for me. So, I jumped on it. But, then, look how inaccurate the readings become after calibration. It kinda suggests that you want to be cautious about calibration. Others sometimes say that. Dexcom says that. So, there might be something there. Perhaps the arrow played a role. I don’t know what to say. Maybe? In this instance, accuracy eventually increases after flat line calibration. But, I can’t say that is always the case. Exercise caution with calibration is the best I can say. Dexcxom says DO NOT calibrate more than once per day. Some people say NEVER calibrate. Accuracy just seems to jump around a lot during the day…

These are the ones that make me mad. I think these are not uncommon…

They make me mad because sensor data makes system behavior look significantly
better than it actually is. I feel like Dex G6 ‘smooths out’ the peaks and the valleys - kinda falsifying the data. But, I had to take a lot of blood sugars to see that pattern of behavior. I don’t want to do that all the time.

[Marie20]

I calibrate the first day more than once if needed, I just try to at least wait a couple of hours before a second calibration. The only time one froze on me was when it was way off and I calibrated twice within about 15 minutes. (When I first started using them and I didn’t have a clue)

And I don’t care if they think 20% is okay, I don’t. I want mine within 5-10 points. I calibrate it if it is off by more than 5 points. I haven’t had any issues with doing this.
Last night when I checked my sensor which has now been through 2 restarts it was exactly the same number 10 minutes after my meter test. Perfect!

For some reason it really responds differently to different people.

[CJ114]

@Marie20

For some reason it really responds differently to different people

You are a technology goddess. It takes both good technology and a good operator of that technology to make any technological item be productive. Some people just don’t have the interest or the patience to learn how to use technology and when they can’t make it work, blame the technology. Nothing is perfect in this world and those looking for perfection will most likely usually be disappointed. What is available, however, is miles ahead of what we had just a few years ago and certainly does help us improve control of our diabetes if we put in the effort and set our minds to making it work rather than wasting time trying to blame someone or something for less than perfect results.

[Marie20]

Such a great invention! Lol @CJ114, thank you for the complement, but I am no way a technology goddess!!! To be able not to finger stick 4-10 times a day?? Wonderful stuff!

[mohe0001]

There’s no way to calibrate with 40% error. It breaks the device. I don’t really calibrate the G6. I just let it run, and I recognize that it might be sending good data or it might be sending garbage. IDK.

Think of it like this, @CJ114
The problem is that G6 doesn’t lie consistently. So, I just assume its always lying, unless proven otherwise. Its not lying now. I’m gonna see if I have better luck on G5. Having trouble running system automation with this much error. I think its gopfy manufacturing process, but that’s just a guess.

[Marie20]

You could try to do wide of a range in a couple of calibrations? The first do half the difference and the second 2 hours later do the rest? See what happens instead of trying to do it all at once?

It would drive me nuts to be close to accurate and then wildly off, it would be very hard to rely on it.

[YogaO]

FWIW, your math is wonky.

Percentage differences are always based on the sensor reading not percentages of the fingerstick reading.

In this example, the sensor says 84. The acceptable range is 84 +/- 20%.

The math is:

84 x 1.2 = 101
84 x 0.8 = 67

So if the fingerstick is between 67 - 101, Dex would call that sensor reading accurate.

The actual percentage difference is:

135-84 = 51
51/84 = 60.7%

However, this isn’t really important. Again, if you fall outside the =/- 20% range and you have the steady right pointing arrow, you should calibrate to the fingerstick value.

[mohe0001]

I’m glad you checked this again. We are doing our math different.
Let me get out an old fashioned piece of paper.

No, your right. This is important. Let me call them and ask again.
It might not settle anything definitively, because sometimes people call and get different answers to the question, but I’ll see what they say today.

Your calculating 20% off the sensor reading. I’m calculating 20% of the finger stick.

Calling Dex…888-738-3646…

[YogaO]

I’m interested in how you’re coming up with your % differences because I have tried a couple of different ways on a couple of different days and I can’t get your answers.

Let us know what the Dexi techi says …

[mohe0001]

Darn those nurses at the helpline are good. She told me all kinds of stuff I didn’t know.

1. Caculate the 20% off the finger stick, NOT the sensor. Sensor should fall within 20% of the manual finger stick.

2. They DO NOT recommend calibration, in general, when your using the factory calibration from the code. In ‘no code mode,’ cali is recommended.

When your using factory calibration, they recommend calibration according to 20/20, but also say that it won’t have much impact on changing the sensor output. Sensor output is what it is, in accordance with the factory calibration. They expect it to naturally come back into alignment with finger stick eventually if the two values are off.

3. (Interestingly…) Tylenol is not contraindicated for G6, unless you take more than 1,000 mg every 6 hours - that will break it.

4. IV prep and other cleaning agents people use before inserting the sensor can also break it. They feel that they have been seeing that.

Now, sometimes people get different answers to these questions when they call, so if in doubt, I recommend calling. But, this is their position today.

I think I’m gonna try ‘no code mode.’ I think that I might be getting bad resuslts off the factory calibration, but I’m also reading the IEEE paper just published for G6. Not much practical advice in there, but it exists, so I’m reading it. Its here: Modeling the error of factory-calibrated continuous glucose monitoring sensors: application to Dexcom G6 sensor data | IEEE Conference Publication | IEEE Xplore

[mohe0001]

Ahhh, yes. I do tend to calculate percent difference, because it produces similar results. But, I am not using the formal definition of % diff here because that’s not what Dex recommend. They leave off the % part (they get rid of the denominator), which is fine for our case. Its easier to calculate in your head.

You will see different equations for % diff online. This is the one I am using now. Some months ago, I was using the simpler one. Percent difference - Math Central

When I calculate % Diff, for the numerator, I do Manual - sensor.
That works best because manual is typically the higher number (for me). Its also considered the ‘truer’ number, and that matters. But, if manual is smaller than sensor, I take the absolute value of the numerator. I am also using the average of the 2 numbers in the denominator, instead of the finger stick (which would be conventional). Thats just to give sensor a little bit of a handicap and produce better accuracy measurements in my sensor. My sensor readings need all the help they can get to pass the test. I typically don’t put the absolute value into my computer calculation because I like to see that number come up negative when finger stick is the smaller number…just because that’s unusual for me. It gets my attention to see that negative number. When I see the negative number, I take note, because I’m over-dosing insulin in my system automation.

Thanks, BTW, @YogaO, for checking the arithmetic. I was hoping somebody would. Nobody ever wants to do that, lol. My arithmetic is ALWAYS in doubt. Wish I had an old fashioned calculator. Really need one lately.

The main point (and difference in our calculations) is that I am measuring the error in the sensor from the manual stick. You are measuring the error in the manual stick, from the sensor. Does that make sense?

@YogaO, sorry. I’m too long winded. Takes me forever to say anything.

These are all different ways of getting the same answer. Follow your heart.

https://webassign.net/labsgraceperiod/ncsulcpmech2/appendices/appendixB/appendixB.html

[YogaO]

Thanks for getting clarification.

So to make this a bit easier, here’s my quick way of thinking about this. If fingerstick BG > 80, +/- 20% applies. If fingerstick BG < 80, 20 point rule applies.

Fingerstick = 100, acceptable sensor value range is 80 - 120. Calibrate if outside of this range.

Fingerstick = 75, acceptable sensor value range is 55 - 95. Calibrate if outside of this range.

The CDE I spoke with also said two important things.

If you want a closer range between your fingerstick and the sensor reading, but the sensor is within the 20/20 rule, you can calibrate. However, don’t keep doing this repeatedly as that will “kill” the sensor (due to overcalibration).

[The_Burn]

Thanks for taking the time to go thru this. I doubt I’ll ever have the patience to do this level of detail. I started using the Libre because finger sticks were one of the things keeping me from treating my T2 properly. With the Omnipod (which I start on 1/16), i’m really going to try to make a change. Doc is also putting in paperwork to Medicare to switch to the Dexcom. I am a technology junkie, so i’m hoping that I’ll like playing with these. Thanks again, Russell

[mohe0001]

You probably wont have to do this. I just get a lot of inaccuracy compared to everybody else.

You’ll probably be fine.

But, take a look at it, just in case. I didn’t notice the error for a long time because I always got good numbers from G4. I thought G6 would be the same. But, I could feel when it was off, so I started looking at it closer and playing around with it. I’m 5 months in. I should have noticed within 2 months because my a1c was unusual after starting G6. I didn’t, initially, know what to make of that because sensor data looked exceptional. I was real confused.

[LorraineK]

The thing is, tech is reading glucose through fat, not blood. Although I am just now (after 4+ years since diagnosis) getting set up with CGM, have already been told by my provider that it will be good to sometimes double check readings from that CGM.