Internal inconsistency of Dexcom G6

One thing I’ve noticed with the G6 that I never noticed with the G4 or G5 is that it can provide results that are inconsistent with themselves. For example, I might get 120 with a full down arrow (ie at 120 and dropping fast) and then, when it does the 5 minute refresh, the next result is 128 with a diagonal down arrow (ie at 128 and dropping at a moderate rate). How can my BG rise from 120 to 128 from one reading to the next and yet still be dropping? I see internally inconsistent 5-minute jumps like this I would guess about once a day if not more often.

I would not say that the G6 is necessarily less accurate the the two previous Dexcoms. I find them all pretty good. My sense, though, is that the G4 and G5 kept their story straight whereas the G6 is more volatile.

This shows up in ways that are not, strictly speaking, inconsistent, but that strike me as a little implausible, too. For example, I’ve seen a BG of 220 with double rising arrows, refreshed it 5 minutes later, and seen 218 with a flat arrow. It’s certainly possible that my BG rose rapidly and then leveled off just as rapidly, but it’s not something I ever remember seeing on the previous incarnations of the Dexcom. If my blood sugar was rising rapidly it would take a while to level off.

Has anyone else had this experience? Is there something about the new algorithm that allows it to jump around more?

On the subject of calibrations, I’ve noticed that if it’s off (usually reading consistently lower than my meter) and I calibrate it, it will almost invariably within an hour or so, start being off in the other direction (reading consistently higher than my meter.) It makes me wonder whether calibrating is wise. My experience with calibrating the G4 and G5 was that it was best not to over calibrate, but that occasional calibrations could bring it closer to fingerstick readings without tipping it into going the other way.

These issues are far from dealbreakers for me, but they do leave me puzzled.


It may be that the arrows take into account the last 3-4 readings, not just the last 2. And possible dexcom changed some things between versions.

I have only used G4, so interested if others have noticed what you are seeing on G6.


@MM1 - I don’t pay much attention to the arrow. So am not sure if the “Arrow Algorithm” changed between G5 and G6.

But you are entirely correct in terms of the G5 that the arrow clearly takes more readings into account then the last two.

(And I don’t remember how the G4 operated in this respect.)

Logical to me -

If within the 5 minutes it spikes to 130 and at the 5min interval drops to 128 it is a gradual down

It is looking at what happened in the 5 minutes

My understanding is that the 5 minute interval collects data, and averages/analyzes the data to give the reading. Just like fingersticks can give different readings from the same drop of blood, sensor readings are not precise.

I don’t see the dexcom numbers as being “exact” at any point. I watch the trend, and do BG checks when needed.

Sometimes I see Dexcom as a speedometer in my car. I glance down to see if I’m close to speed limit, not to check if exactly on the nose. Unfortunately I think many are taught to think/assume dexcom being more precise.

I started using the medtronic Sof-sensor in 1996. Horrible accuracy, but it was still useful. The sensors are gradually improving over where we’ve been.

I also come from the perspective of visual BG testing in early 1980s, comparing colored strips to 40-80 point color blocks on the container. For example, would mark in my log that it was 120/180, or 40/80, or 240/400, or list 2 that it was in between. So dexcom is light years ahead !


My wife is the T1 - I remember when she got the first giant glucometer - before that is was urine testing - Using Libre now hoping for medicare approval of Eversense

The SOf-sensors are random number generators (as someone here invented)

The single biggest fraud in consumer history - bar none - IMHO

lol … I drive the same way. Speed limits and speedometers are for “guidance” :wink:

1 Like

They were frustrating, but still helpful for me at the time. Before that I only BG tested before meals, as instructed. So early CGMS was eye opening to what happens after meals.

1 Like

The “arrow” difference may simply be a result of using too few readings from the past; if you jump off a cliff you are definitely heading double arrow down, but when you hit the ground you are instantly flat. Samples taken at 5 minute intervals show the same effect and there is a way to deal with this but likely as not the Dexcom programmers don’t know it (it’s called “sampling theory”, you want the Nyquist-Shannon reference if you look it up on wikipedia.) The programming is non-intuitive; the intuitive answer is just plain wrong [personal communication :frowning: ].

In my experience using xDrip+, not the Dexcom receiver, but using the transmitter internal algorithm the transmitter will re-write history as a result of calibrations. That considerably complicates the analysis - later samples change earlier samples.

So far as calibration is concerned I see the G6 as pretty much accurate within the range 100-180mg/dl when calibrated (and possibly when not) but outside that range it reads higher than one other reference, the Abbott FreeStyle test strips. My current reference is the Ascensia (Bayer) Contour Next One and that seems consistent with the G6 above 100, but I avoid calibrating above 200 all the same. Below 100 the Contour is consistently lower and my own body confirms this.

If the G6 is below 100 it would be dangerous for me to rely on it in critical applications. Fortunately I’m very hypo-aware and that has been protecting me. The G6 trends in this region (below 100) are good, but since I’ve been waving a falling saw around for the last three days I have certainly not relied on the absolute numbers.

John Bowler


MM1, you make an excellent point about finger sticks giving different readings from the same drop of blood and that there’s no sense in expecting the G6 to be “exact.” And it may very well be that particularly early in a session when the G6 doesn’t have as much past data it makes sense that it would jump around a little.

I think the G6 is an excellent device and I wouldn’t want my perplexity about this particular aspect to be taken as a knock against it. I’ve just starting using my G6 with a Tandem pump and I really think it’s great.

The thing that struck me is that the G6 doesn’t seem to act in the same way as the G4 and G5 in this respect. Perhaps it’s because the G6 is showing something a bit closer to the raw data whereas the previous Dexcom iterations smoothed that over more?

As to Tony24’s point, I’m no scientist, but going down up and then back down all within 10 minutes doesn’t sound like a biological process to me. It sounds more like physics. Yes, a projectile could go from 100mph to 0mph in fractions of a second. I don’t think blood glucose could. But, I don’t have expert knowledge and I’m ready to be corrected.