When CGMs lie, and when a deep low is really a nice landing

I have mentioned several times in passing that I think many Dexcom lows, when they come off a steep slope, aren’t really lows. Today I decided to prove it.

My son calibrated his CGM this morning at 8:37 AM and was off by 2: his Dexcom has been very accurate for the past two days (so we can assume that, when stable, his CGM accurately reflects his real BG). Later this morning, he had a major puberty peak that took him to about 250. He injected large quantities of insulin and came down quickly.

When he approached in-range level, he and I decided to start our experiment of the day: he took some carbs in several steps to land as smoothly as he could. At the same time, he fingerpoked every few minutes to be able to follow his real BG. The results are below, and they are shocking :slight_smile:

The blue line is his CGM curve: it comes down steeply, crosses 70 swiftly all the way to a “fairly” deep low of 53, then slowly goes back up. His total time out-of-range appears to be about 30 minutes.

The red line, on the other hand, is his real BG by fingerpoke. It never goes below 69, and hovers above 80 most of the time, while his CGM is showing mild to medium-low BGs!

So the CGM lied. The circumstances when I think the CGM will show a false low are as follows:

  • BG is coming down quickly

  • You are taking in good numbers of carbs to land

In this situation, I think that (a) either the CGM will not see the ingested glucose coming into the system, or (b) the Dexcom algorithm always wants to make sure the CGM number is below the real number as opposed to above it, and may give a worst case approximation – possibly a combination of the two.

Either way, it is clear from this curve that a truly good landing, when using carbs for the landing, may show as a fairly deep low – so let your endocrinologist know when she is scolding you! Or, even better, show her this post and tell her to sign up for the forum:-)


This is what I’ve been suspecting for a while with my son… because he still seems to feel lows but many apparently deep, steep lows do not seem to affect him at all, and when we test him he’s usually not low.

I need to do a systematic test to confirm this…but then the question is, what do you do with this information? Do you recalibrate? Do you tell your Endo to just ignore all the lows on the CGM tracing?

I am not quite sure either :slight_smile: For sure no recalibration is necessary: the CGM is still properly calibrated after all this mishegas.

As for the lows: we typically test every deeper low, and find that some are real. First, the ones that creep down rather than steeply fall are not affected by this phenonemon. Of the ones coming off a steep down slope, some end up being real, maybe 20-25% for us? But, for us, the large majority is not real.

Imho, one thing to do for sure is to prove this fact to the endo with an example such as this one. But I don’t think we can neglect all of these lows altogether – at least for us, because some of ours are real, just not that many :slight_smile:

1 Like

That’s not shocking at all. My Dex is always wrong like that. I will have corrected and be on my way back up and the Dex will be showing me in the 40’s. I am glad other people are noticing this sort of stuff.

While you were gone, a bit of a debate between the value of CGM vs BG:


well we definitely treat – we treat even when he’s on his way down and hasn’t hit the low yet. Our Endo I know for sure is not lining up our Dexcom data with our finger stick data in time… that’s something we should ideally be doing but we actually never get our act together for it in time.

For us the slow drifts are often underestimates of how low he is by about 5 to 10 points – unless it’s at night, in which case he’s typically between 80 and 110 when Dexcom shows him in the 50s and 60s. Those ones are the worst … you can finger test and see he’s normal when Dex shows 60, and then he keeps dropping. Do you just finger test all night? Recalibrate?/ Gah. At some point you have to make a gut decision and figure out how to get some sleep.

1 Like

Couldn’t this just be because the Dex is reading interstitial fluid and the meter is reading actual blood? Isn’t it possible that the carbs went from stomach to blood to cells and are slower to enter the interstistial fluid? I find that sometimes the Dexcom exaggerates lows but also sometimes misses them. I saw a video on YouTube where Dr. Bernstein claims that his BG went to zero, but he didn’t die because the level of sugar in his brain was slightly different (e.g., not zero). I’m not sure I believe this because a) meters can’t read down to zero and b) presumbaly the heart and lungs also need glucose to run (but maybe I’m wrong there), but still, it illustrates that the level of glucose in different tissues can be different at the same time.

Imho it’s totally possible.

But my suspicion goes more to the algorithm that Dexcom uses to predict the BG value. Dexcom obviously tries its best to make sure it is guessing under rather than over. For instance, when you calibrate, if the actual is under the Dexcom BG, the new value will be more or less the one you entered, but it is over the Dexcom BG, the new value will only go halfway there.

Or it may be a mix of the two.

1 Like

I would have to agree with @Michel.

XDrip+ uses a home brewed algorithm different than the Dexcom algorithm, and as I have commented and shown before the xDrip+ algorithm follows my lows very well, both in intensity and duration.

1 Like