FUDiabetes

Dexcom G5 Calibrations

@RCA221 Look in settings>xDrip+ Display Settings>Graph Settings and it’s the first item

I went thru the settings several times and missed it. There it was in plain site. Probably missed it because it said graph settings and I didn’t think it was a graph and would be there. You sure know Xdrip backward and every which way,

Getting much better at actually using Xdrip and soon will be trying to restart transmitter and I think I have got this. But a couple of more technical questions.

  1. Raw Data. A couple of days ago Dexcom was down for about an hour so I got no readings on my Dexcom receiver yet I got Xdrip -Native readings on my phone. Any idea why Xrip phone was up and Dexcom receiver was down since they both receive Dexcom raw data? And do you know what the principle differences are between Dexcom and Xdrip adjusting raw data?

  2. Calibration. Why not calibrate G5 more than 2 times a day, approximately every 12 hours apart? When not calibrate them when they are even 10% off from blood? I’ve had a situation when Dexcom receiver and Xdrip -Native were 10 apart and they stayed 10 apart for hours until the next calibration. Feels like I should have calibrated hours earlier and this wouldn’t have happened. Would simply like to understand what is internally involved in calibrations.

At the moment, for me, Dexcom is almost always lower that Xdrip. Based on several criteria I developed, Xdrip is more accurate in 5 of 6 categories. As you suggested, I am trying to limit the number of calibrations and doing them for Dexcom and Xdrip at the same time and with the same finger stick reading.

I don’t know how much of this applies for xdrip.

But for Dexcom functionality, the algorithm only keeps a limited number of calibrations. I don’t recall maybe the last six calibrations? Eventually the newer calibrations cause the older ones to drop off. So if you calibrate very frequently then the range of time covered by the calibrations becomes compressed. This leads to the algorithm that uses this data to be less accurate.

I’ve heard that too, “as you add a calibration then the older one drops off”. Why should I be concerned about older calibrations dropping off? I would rather the most recent calibrations be given more weight. And I have no idea why Dexcom advises you to adjust only twice per day.

That was what I tried to explain above. I may have worded it poorly.

Let me elaborate on what Thomas was saying.

It’s a math thing. The algorithm needs to estimate how quickly the sensor signal is weakening as the body’s foreign-body response slowly encapsulates the sensor wire: over time even if your BG is a constant 100, the raw electrical signal from the sensor will decrease as the body gradually “rejects” the sensor wire.

Think of a graph where the horizontal axis is time, and the vertical axis is the glucose reading. Now put a pair of glucose measurements, at 8am and 10am, and draw a line through them. Glucose measurements aren’t exact, they have some error. Let’s say there’s a 5% error (in real life it may be much more). So for the 8am measurement, put dots 5% higher and 5% lower. Same with the right side measurement. Draw a line from the high dot on the left to the low dot on the right. Draw a second line from the low dot on the left to the high dot on the right. You now have a sideways hourglass or bowtie shape, and the “true” line should be somewhere in there.

Repeat this exercise, but with the BG measurements at 8am and 8pm. See how much narrower the bowtie is? There’s less of an error range as the algorithm extrapolates into the future. That’s why they recommend not calibrating frequently. Keep 'em separated in time to get a more accurate prediction of the sensor signal decay.

1 Like

Thanks Thomas and bkh. I was away but appreciate your help and explanations. Generally I like to know what 'm doing before I do it. That’s why I wanted to understand why I should calibrate less frequently. I never heard of it as a result of accounting for sensor error decay. That is very plausible in terms of the frequency issue. So if I routinely calibrate more frequently, you think the readings will be off because I am ruining one item of their algorithm which is the decay item?

Yesterday my Dexcom and Xdrip -Native, which I’ll just refer to as Xdrip from now on, were off for 3-5 hours by 9-10 points. By not calibrating I could see the 9-10 points continue for hours.

I personally think that the sensor wire reading decay is probably pretty insignificant to all the other items in the algorithm although that may be the item that supports their position that you should not calibrate more frequently.

Do either of you have a personal preference for using the Xdrip app with Dexcom native readings or with Xdrip readings? If so is it because you tested, as I am trying to do, or you have information about their algorithms and feel one is superior to the other?

I don’t know whether I’m typical, but for me I don’t worry about a 10-point difference. It is small compared with the BG swings I’m experiencing just from being alive, and it isn’t enough to significantly change what I do. Whether I’m at 60 or 70, if my BG is dropping I’ll take some glucose, and if my BG is rising I’ll watch a bit closely for the next half hour. If I’m at 150 or 160 and falling, I’ll look at IOB and decide whether it’s enough or I should stack a little bit more insulin. If I’m at 150 or 160 and rising I’ll do the same, but I’m more likely to decide to take some extra insulin. So I don’t worry about a difference of 10 points.

Also keep in mind that, scientifically speaking, all BG meters have error in their measurement (5% or 10% is typical). They do not tell me my actual BG even though I have a strong unconscious bias to accept the BG meter number as if it were the exact truth. So maybe the CGM is right and the BG meter is wrong, rather than the other way around. More likely they’re both wrong, maybe in the same direction or maybe in opposite directions. However, note that over multiple calibrations the CGM algorithm tends to average out the errors in glucometer measurements, whereas each BG meter measurement has 100% of the embedded error from that individual test strip.

I’m not an Xdrip user, so I can’t offer experience on your last question.

2 Likes

Same

Same

Same (lol)

I didn’t quite make the point I wanted about Dexcom and Xdrip-Native.

Meant to emphasize that if not re-calibrated and you are significantly off (I used 9-10 point but I could have used 20-25 points) then bg can stay 20-25 points off for 3-5 hours. Key question is what is really the harm in calibrating more than 2 times a day?

It still seems like you are expecting a greater degree of accuracy than is likely reasonable to expect.

Not the part about the Dexcom being 20 or 25 points off from your meter on one check. But then expecting the Dexcom to maintain that difference down the road and consistently be off exactly in that same 20~25 range difference is IMHO expecting more than the technology can deliver.

At the end of the day it is your body and your Dexcom and it is fairly obvious there is a wide variation on how this all works from person to person.

Try it out. Log it. Keep track of what it does. Then decide what is the best approach for you.

[quote=“Thomas, post:32, topic:8788, full:true”]

I have been on the G5 for almost two years. I think we all wish that our CGMs were more accurate. My G5 is pretty accurate compared to my Contour Next. I tracked every reading for six months and for all six months it averaged an accuracy of +/- 6-7%. I think that is better than even Dexcom claims. Yet it can still be 30 or 40 points off at times. So I still test before eating and before bed as dosing or sleeping when G5 can be that inaccurate could be dangerous. I am amazed that people dose or sleep based on their CGM and even more when I hear a statement that the CGM can be more accurate than a meter. I probably am not going to change my dosing or sleeping based on a CGM, either Dexcom or Xdrip.

In any event, the purpose of this exercise was for me to compare Dexcom to Xdrip(-Native). So far Xdrip(-Native) appears to be more accurate. These are number for this week’

image Comparison Dexcon to Xdrip-Native Readings

If one gives me more accurate numbers, then I want to use the most accurate but it is unlikely to impacting my dosing.

Your insight has been very welcome. Recently started on this sight and I really appreciate your posts and comments.

I cannot link Xdrip to my Contour Next meter. When I go to Data Sync, Glucose Meters, Use Bluetooth Meter…I cannot change the off button to on. Know what step I am apparently missing?
Is it because I have an older Medtronic Paradigm pump and not 630G, 640G or 670G?

Also, it does not find my Contour Next meter when I scan.

BTW, Xdrip -Native seems to be more accurate than Dexcom, at least so far.

image Comparison Dexcon to Xdrip-Native Readings

Readings Xdrip

39 All readings 56%
26 If diff CGMs >5% 62%
12 If diff CGMs >10% 75%

A bit of a surprise.

Are you using the Contour Next One?

You also need to have at least one finger stick reading on the meter, and it needs to be in discover mode.

Yes that’s the meter I amusing. I haven’t seen discover mode in the settings.

Seup/ then what? pump options,reminder, date,time,sound,autolog,target,trends rage,high and low lomits,language, customer service

autolog/autolog is on. Don’t know where else it may be. Not in pump options.N

Thanks.

Discover mode (or pairing mode) is on the Contour meter.

Press the on button on the meter and keep it pressed until the strip light blinks blue and a Bluetooth symbol is on the display

Sorry, I have Contour Next Link. Looked the same as Contour Next One. Maybe I need CNOne.

Yes, it appears that Link connects only to pump and One connects via BlueTooth. I assume I can use test strips that I bought for Link. I don’t think Contour Next differentiates.

I promised I’d provide a summary of my comparison of Dexcom G5 readings and Xdrip readings. Per Doc’s suggestion, I used same the same blood sticks and calibrated about twice a day at the same time for both. Total of 46 readings.
Compared to each other, Xdrip readings were higher (or Dexcom readings were lower) 85% of the time.

Compared to Contour Next,
Xdrip was closer (more accurate?) 60% of the time and Dexcom was closer (more accurate?) 40% of the time.
Average of differences was Xdrip,7.75% and Dexcom,9.25%.
Thus overall Xdrip was more accurate in terms of average numbers and also variability.

Based on bg ranges:
image

Based on this limited population, Xdrip readings were more accurate when compared to Contour Next.

I plan on doing one more test after inserting a new senor. If anyone has any interest in the results, I’d be happy to share.