Results! - The G5-G6 Challenge - An actual comparison

I know there has been a lot of discussion on comparing the G5 vs the G6. And depending on who you listen to, it’s either a lot better, the same, or worse. I even did a poll on it a while back.

But I am not sure if anyone has actually worn them both at the same time and done an actual comparison of numbers! Has anyone worn them simultaneously? If so, I’d love to hear what you thought.

But anyway, I think the only way for me to know which is better for me is to wear them both at the same time and actually compare numbers.

Thanks to @LarissaW this was made possible. Thanks for the sensor donation and the lending of the transmitter and receiver! You’re the best!

I am on day 3 of the comparison right now. I will come back with all the numbers.

I am running the G6 sensors and G5 sensors at the same time, My comparison will span about 3 weeks - 10 days each for the two G6 sensors I have, and 7 days each for the three G5 sensors I will be using.

I know Dexcom has reports for this sort of stuff. But I have no idea how they did their test. Maybe it was only when flat at certain levels, or maybe they did tests when BG was rising and falling, or who knows what they did. Who knows how many BG tests or anything like that. Also, did they compare the G5 and G6 simultaneously during real-life moments?

I mean, lab stuff may not be the same as real life, depending on how they do the tests. Maybe they did not have people sleeping and waking up, working, exercising, eating a bunch of food, drinking coffee, drinking other things, or all of the other possibilities in a lab, all while comparing the G5 or G6. I have no idea how they ran their tests.

But I know how I am doing mine.

The way I decided to do the test was pretty simple, and seemed like an extremely fair way to do it. I test a lot. I mean…a lot. So I am doing it like this.

  • Every BG test that I do - I will write down that BG number and what each CGM is reporting at that moment.

  • I will use the same meter, a Contour Next One. No complaints about using a crappy meter, this thing is top-notch.

  • If I happen to have a BG meter reading that is off, I will test it again. I know when it is off.

  • I will keep both CGM receivers together, so they will both either be picking up their sensor readings or not picking them up. Neither one will have an advantage/disadvantage of getting more or less sensors readings. If one receiver is out of sensor range, they will both be out of sensor range.

  • Nothing like, “Hey no fair, that test does not count. You were rising or falling, too high, too low, or eating or sleeping…”

  • I am not ever going to test my BG just for the sake of doing a CGM comparison. These are all actual BG tests I would normally do. My BG tests are frequent enough and at different time intervals that they are pretty much randomly timed throughout the day. If I test, I will log it. But I am not testing just for the sake of scoring a CGM at that particular moment. That makes it much more fair!

  • But…if I do a quick succession of tests, I am not going to count all of them. Like sometimes I may do 3 or 4 tests within a 15 minute span. I will only count the first one, so that any big “off” difference at that moment is not multiplied for any of the CGM’s.

  • I am not counting my BG tests when running, because I am not carrying two CGM receivers and a notebook with me.

  • Likewise, if I leave the house, I won’t count those tests. Again, not bringing the notebook with me when I leave the house.

  • But I will have a lot of tests and a lot of random times - before eating, after eating, waking up, going to bed, etc, etc.

  • I will put the numbers together in a logical way, like a an overall comparison of differences between the CGM’s and meter.

  • I will also have a comparison for different ranges - like low, normal, and high.

  • I will also group different days, because I heard the first day for the G6 sucks. So I will show all the days together, and then also have all the days broken out individually.

  • I will look at if a CGM tended to be higher or lower more frequently, and look at the average (absolute value) difference as well.

  • I will follow the rules for them both for calibration. I will calibrate the G5 every 12 hours. For the G6, I will use factory calibration and only calibrate it if it is way off. Those are the rules I got from Dexcom! I called them to ask!

Ultimately it comes down to which CGM is closer to my meter for me. Which is more accurate? That is all that matters for me.

Better insertion, no need to calibrate, 10 day wear vs 7 day wear - all seems like marketing BS are not priorities to me. The only thing that really matters to me is accuracy, so that is what I am looking to find out.

Will let you know and share numbers when I am done!

13 Likes

Can you add a Guardian 3 to this experiment? It would be interesting to see how it stacks up to both the G5 and G6.

1 Like

I don’t have a Guardian 3.

There’s got to be a few floating around here from disgruntled 670 users.

It’ll be interesting to see your results. My impression is that G6 is more accurate but it achieves this accuracy, in part, by giving ? or sensor error more frequently. So there are more times when it remains agnostic rather than skewing results.
I’m not sure how Dexcom gets its published accuracy results, but I met someone who was testing the newest generation of Dexcom at the beach one time. She had something like 7 or 8 sensors on her belly, all at the same time. And she was doing her daily life stuff. So I’m guessing that however they test it, at least some of it involves comparing between different sensor values. Im not sure whether they asked her to do additional finger sticks as well, but I’m guessing so?

4 Likes

Eric, will be watching and waiting for your results as I use the G5 and have been hesitant to change to G6. But I would if G6 was definitively more accurate. I just spent a couple of months comparing accuracy of Dexcom G5 with Native readings (Dexcom) and Dexcom G5 with non-Native readings (Xdrip). I also used a Contour Next which I think is the gold star.

If you collect all the readings in a simple table (date, time, Contour readings, G5 readings, G6 readings) then you can do lots of different analyses on them even at a later date. I used Excel as my spreadsheet. I kept my Excel table totally separate from my analyses and instead used a clean table to populate the analysis. You could then do whatever analysis you want on the data, whenever you want, using Pivot Tables.

Some of the analysis I did which you might consider when you compare G5 and G6,

  • List item

Analyze the data in several ranges. I looked at which was more accurate at different ranges (<90, 91-120,121-140,>140).

  1. List item

Are readings higher or lower compare to each other most of the time? Overall I found that Dexcom readings were less than Xdrip readings 85% of the time.

  • List item

If individual readings for both G5 and G6 were less than 5% different from Contour Next, you could flag this, as we could live with either being that close to the Contour Next.

Anyway it was very informative and actually fun to do. I learned some very interesting things as I imagine you will as well.

1 Like

I’m interested in these results, too. From my non-scientific comparison, I find G6 to be more accurate than G5, but as @TiaG mentioned, there are tradeoffs. For me, G6 seems to fail more frequently. It isn’t rare for a G6 to fail even before warmup completes! Still, I’ll take the tradeoffs.

3 Likes

Don’t lose the other tests though, because they will give evidence about whether an erroneous CGM converges rapidly to an accurate value, or does it stay wrong.

1 Like

I had to switch from G5 to G6 end of last year. To see what is coming, all of last November I ran G5 and G6 in parallel, and I used 2 different bg meters. So I had (nearly) always 2 CGM values, and occasionally, when doing blood tests, 4 values. As the G5 must be calibrated with one of the bg systems, the comparison may seem unfair. But we must compare the systems as they are really used. Well, to my surprise, the G6 did very well with “just the factory calibration” when looking on data for days 3-9, and sometimes for days 2-10 of each sensor. Never on day 1, though: I was reading with a newer “8G…” type transmitter. It tries to adjust for the biasses that occur in the first hours of embedding the sensor. At least for me, that did not work well at all, and resulted in the first hours in deviations to bg in the range of minus 40-70 mg/dl. So I needed another 5 hours on average after the 2 hr warm-up in order to use the values. (I recently tried an older 81 series tranmitter which worked a bit better for me ). With G5 these days-1 were much less dramatic, plus there is a possibility to avoid day-1 with G5 via overlapping use of old and new sensor (“pre-soaking”, suggested e.g. by OpenAPS looping pioneer Dana Lewis in her e-book). My conclusion as of today: With G5 you can produce uninterupted reliable results. With G6 you mostly have very good and easy days, but depending on the transmitter you use (and usually you do not have a choice there) you can have very troublesome days occasionally where values are for several hours not reliable at all. (Note: I do not know who all is on this website. I am type 1, and looping with relatively tight control. If your situation vastly differs it might not matter as much if a reading is for instance 50 too low. Also the magnitude of deviations in initial hours of embedding a new sensor probably varies between individuals, and maybe their exercise level and other factors also matter in the relevant hours. ).

7 Likes

Ok, marketing BS to you because you don’t care about those things. But those factors listed above certainly impact a lot of patients’ (and their families’) quality of life, ya? I don’t know if I would call that marketing BS or knock it to think those are a priority for others. :woman_shrugging:

8 Likes

Totally agree! There have to be sooo many factors that impact accuracy and which CGM works best for each individual.

Thanks so much for sharing your takeaways from your comparison! I’m excited to see @Eric’s results and also kind of wish I could directly compare the two systems!! It definitely sounds like a bunch of work up front to compare different cgms but I’m sure lets you feel more secure with the one you’ve got… wouldn’t it be nice if there was a way that each diabetic could trial out and compare cgm systems before having to commit to one!!

2 Likes

Point taken. I did not state it well. I have revised my post and am saying it like this now:

6 Likes

G6 Data Revisions: I am new to G6 and my previous G5 transmitter has a few days left on it so I am running both devices plus a FreeStyle Libre. (I am something of a nervous nelly but I don’t have a lot of confidence in G5; even though it is hands-down the better CGM it can sometimes leave me in the lurch especially during exercise)
An interesting thing I wondered about is how the G6 makes such beautiful and smooth curves. I am also Looping using the Dexcom data to control my pump; this makes me especially nervous trying out a new G6 when the G5 seemed to be doing the job. My Loop software keeps a running graph of the Dexcom data and I notice a lot of noise on the Loop graph that I don’t see on the Dexcom graph. I use an iPhone 11 to read the Dex and to run Loop.

I started paying very close attention to the G6 and I have convinced myself that this is happening: G6 will put out the current data point; Loop will echo that number. Come back to the screens a few minutes later and Dex graph looks very pretty. Loop graph is noisy. For an hour or two I tracked both devices very carefully. Dex reported the following series of reading which Loop echoed and plotted:
142, 139, 135, 135, 131, 132,142

Going back to the Dexcom graph a bit later shows the following numbers for exactly the same time stamps:
142, 139, 136, 135, 134, 132, 138

and twenty minutes after that the final 3 numbers were now, magically, shown as
134, 135 (had been 132) and 138.

No wonder the charts track so nice when Dex is all done with it!

The other thing I have noted, but this is only day 8 of my first sensor, the G6 seems to track ALL the data at a much reduced curve slope. Low numbers around 50 all match up pretty well. High numbers around 150 (as indicated on the G6) tend to look more like 200 on the G5, and Contour Next fingersticks confirm this offset. With sensor number one, I have a simple formula that helps me guess what G5 and Contour Next would report. The data is important to me, since I use G6 to control my loop. The shallow slope changes ‘apparent’ insulin correction factor, carb ratios, and where I set Low and High alarms on the Dex. Obviously, Basal numbers are not affected.

My Simple Equation gives me correlation with most static values within reason
Take the G6 number as N. Subtract 50 from N and multiply the result by 0.4.
Add N to that result and you have a good approximation of what G5 would have said.

In other words: The G5 result is about N + (N-50) * 0.4

4 Likes

Not sure I follow all this I looked at BG DATA TABLE and noticed that they had changed from when I previously looked at them. Then I found out the the normal setting is for the BG data to be “smoothed” after a calibration. So there is a setting where you always keep the calibrations the same and do not smooth them. This might help. I am also very happy with the G5 and reluctant to switch to G6. Xdrip is the cats meow.

1 Like

I like xDrip also, but it only works on Android near as I can tell. To close the loop I need constant data updates and I have not figured out a way to make xDrip data appear seamlessly on the iPhone; Loop with Insulet Pods only works on iPhones.

The thing I like best about xDrip is that it ignores the ‘failure’ of battery on the transmitter. My G5 transmitters tend to run on for weeks and weeks after the Dexcom receiver reports end-of-battery. The Low Battery Alert seems to track 90 day lifetime very closely… more closely than battery voltage report from xDrip!

1 Like

I haven’t run on Loop but I know that at the moment Xdrip is only Android. I hear Spike however is very similar to Xdrip.

I like so many things about Xdrip it is impossible to even rank them. I also use the G5. I used to want to extend the transmitter life which you can do on Xdrip but if I do that I will be out of warranty on the next or the one following that. So I don’t bother to extend them… I do love seeing the Voltage and Resistance numbers on the status page so I can see if it is going to fail early. I love that you don’t have to do a thing and the sensor works until you decide to change it.

1 Like

I also love xdrip+. I too use Loop on iphone though so now rely on the Dexcom app on the same iPhone. I still use xdrip+ though on my android (I carry 2 phones!) as the ui is better than anything else on mobile. ANDROIDAPS is also great which shows iob, etc, on android.

I wasn’t aware that the dexcom app changed past BGs! I never compared the actual data but you are right, @BenD, this must be why I also noticed the smooth graph on dexcom. I actually rarely open the dexcom app though (except when there is a sensor failure!) so don’t rely on its graph.

Yes. For a graphical picture of what is happening see this post:

I superimposed recorded values from SugarMate over the 3-hour delayed values reported to Apple Health from the Dexcom app. The adjustment is consistent with your two sets of numbers [I editted the stuff below to align the numbers and remove the spurious double space in @BillKast’s original post.
I also added emphasis in italics.
]:

What I don’t know yet is whether the adjustment happens in the transmitter or the receiver. Logically it would be the receiver; Dexcom delays an update to Apple Health for three hours, presumably so that it can let that adjustment pan out. The transmitter retains 3 hours of readings, so if you are about to rewrite history probably better not do it for three hours.

This is interesting.

I have been doing something kinda similar with the Dexcom G6 and the Abbott Libre.

I use the Dexcom G6 with the Tandem t2:slim (now with control IQ). I like it a lot.

However, I swim every day about 5000 to 6000 yards and I need to know what my sugar is when I swim. The Dexcom does not work in the pool so I use an Abbott Libre which I can scan at poolside to get an instant reading.

I also use a Medtronic 670G pump when swimming since it is waterproof and the Tandem is not. I have found that I need insulin during my workouts (i.e. insulin to get glucose into those hungry muscle cells).

Anyway, the Dexcom accuracy is much better than the Libre, that I am sure. Part of it appears to be a more significant lag for the Libre compared to the Dexcom. It may be that the interstitial
BG in the arm (where the Libre is located) lags the interstitial BG in the abdomen.

2 Likes

The G6 transmitter has no problems with water (I’m a diver and one has yet to fail on me). You can use a mobile with the Dexcom app to read it, though even my ruggedised mobile failed after an extended salt-water session (photos, partying, etc); the charger port got salted up so I could only use wireless charging :partying_face:

1 Like