Dexcom overreacting to exercise

I’ve been using a Dexcom for about 7 months now, a G5 as of September, and one thing I’ve noticed is that it tends to overreact to exercise. For instance I just got back from a 50 minute strenuous bike ride, and by the last ten minutes of the ride it started giving me <80 alerts with a vertical down arrow. When I got in the house it was showing me at 70, still vertical down, but my fingerstick showed a much more moderate 107. Seeing a significant drop during exercise is expected–and desired: I was at a post-prandial 180 at the start of the ride–but I wish there was a way to have the CGM stay more in line with fingersticks and not overreact or overstate the drop.

Does anyone else have this overreaction problem and know a way to ameliorate it? One thing I’m wondering is whether sensor location affects it. I’m wearing it on my lower abdomen and it may be getting jostled quite a bit by the movement of my legs while I’m riding.

I have had that happen to me many times, @DrBB, especially on bike rides. It seems to me that dehydration and sensor location make a big difference there. Sometimes, on a bike ride, when I see unexpected low numbers, I drink some water and wait a few minutes to see if the numbers correct – frequently, they do. Lately, I’ve been wearing the sensors on the side of my chest and have seem a lot less of that kind of problem. When I wear the sensor lower on my abdomen, I get the kind of results you’re seeing.

I ran into that during my short trial of a couple of Dexcoms. Cardio caused inaccurate low readings for me.

The CGM uses interstitial fluid which has like a 15 minute delay from a finger prick. To check to see there is really an overshoot or error you need to check a finger prick and then 15 minutes later see what the CGM reads.

That’s true @Brian_BSC, though I think the current algorithm seems to try and compensate for that delay (I presume by projecting the likely BG from the combination of calibrations and trend data). That said, it cannot always account for rapid changes, which could happen as a result of intense exercise. Even with that in mind, I have seen results that more closely follow finger-sticks if the sensor is placed in some locations - for me, mostly higher on the abdomen or on the sides of my chest.

1 Like

My understanding is that CGMs have predictive algorithms that are used for hypoglycemia alarms but not for predictive estimates of blood sugar readings. They simply record and display the readings from the sensor.

Well, only Medtronics has “predictive alerts” - presumably a patented technology. Dexcom hasn’t publically spoken of the details of their more recent algorithm, however, it is approved for clinical decisions in some countries (and I believe there is or will be an application for FDA approval for the same), and it clearly more-closely reflect current BG numbers as seen thru observation. Since it still tests only interstitial fluids, results that closely match current BG must be calculated mathematically. Regardless, my experience has been that unless there is fairly rapid change or another mitigating condition (such s dehydration) occurring, the newer Dexcom very closely reflect finger-stick results in the moment, rather than 15 min in the past, especially in the low and normal range.

1 Like

Yes, you are correct @Thas. My point was that no CGM has a predictive algorithm for readings (let alone FDA approval). Only for alerts. My understanding of the Dexcom 505 update was that it was all about increased accuracy and used averaging techniques and was not about trying to get more timely readings consistent with a finger prick.

CGM as a whole are pretty useless for exercise pretty much due to as discussed above and the nature of how they work. The use of glucose in the interstitial tissue instead of in blood. During exercise all kinds of fun can effect what’s going on in this tissue layer, be it dehydration, depletion of reserves, body still suffering and pulling resources from another days exercise, or simply we are using it straight from the blood before it get’s to the tissue.

I’ve consider the 640G and while it is able to turn off basal based on trends from the CGM (it is a very nice device to use for the interface is a huge improvement), it suffers from being Medtronic’s CGM, and also still unable to understand exercise unless you create a favourite for it, which is a touch faster but if you are happy enough a navigating around on an older gen pump it’s just as easy to put on a low basal and/or suspend the pump for a while to help control the lower levels. Will be interesting with the 670G though with it also increasing basal for increased BG trends, but do suspect that will be even more fun with exercise.

In basic though the best use of a CGM during exercise is only for direction of bloods, not for readings and definitely not for doing any corrections from.

I was shoveling out my car in a hotel parking lot (snow plow left a huge pile right behind several vehicles in the lot) and experienced a smiliar situation, @DrBB. I was wet and grumpy and didn’t stop to check with a fingerstick…and of course took on too many carbs while I finished the job. So in a while I had to do a little more exercise to fix the over-correction. :flushed:

Interesting observation @Thas. I doubt that I could get a dex sensor in that location. I have only had success wearing it in the lower abdomen, and I have had @DrBB’s type performance on bike rides.

Regarding the 15 minute lag … well, if you are timing with an hourglass filled with grains of salt, I’ll buy that. However, my experience is once low, the Dex will continue to read much lower than my fingersticks for over an hour. For me, I have to go a bit by experience and figure that if I treat for a low while riding, it will come up, but I will have to perform more frequent fingersticks to make sure I am trending in the right direction.

This is a prime example of why I get worried when I see the 505 SW being touted as part of an AP solution. Without some really good predictive algorithim(s), I have a hard time imagining how the AP will manage.

1 Like

Like everything else D-related, there’s a lot of difference between people’s experiences. Generally, the Dexcom with 505 SW seems to track very close to the results I get from fingersticks, consistent with the 92-96% accuracy listed on the FDA approval for +/- 20% and 84+% accuracy within 15%. I have also seen that on the low end, the new software “Recovers” a lot faster that the earlier algorithm did. There is more lag time on the high end, from my experience. Considering the absorption speeds of current insulins, the “reaction time” of Dexcom with 505 would appear adequate for AP – HOWEVER, I agree that neither the accuracy of the CGM or the speed of the insulin would satisfy the level of control I would like to see from an AP (not that I’m likely to actually SEE an AP :smiley:, but that’s another topic… )

Good suggestion re hydration. I tried it on today’s ride… but then ran afoul of the other X factor that kicks in when exercising outdoors this time of year. Temps in Boston are hovering around freezing today, and so of course I got the adrenal effect–the one where your body says No one exercises when it’s this cold out! We’re being chased by a damn polar bear! and dumps a big load of glucose into the bloodstream so you can hightail it out of danger. So instead of going low, like usual, when I got back my fingerstick was over 200. CGM said 100, so at least it wasn’t shouting warnings at me.

Of course, but if that was the explanation the CGM would be registering higher, not lower, because my level has been descending. It would give a value 15 minutes earlier on the down curve.


I also experience a significant “bounce-back” once the ride is over. Won’t get me all the way to normal from a 45, but I’ve found it’s really easy to overtreat exercise lows if I treat them the way I’d deal with say an overnight hypo. If I can stand to wait 30 minutes or so and test again I often find I can treat 'em lightly or not at all.

In 2010 with the first generation Medtronic cgm when I would start a run I would get a false dip. The cgm would say my BG dropped. And then later in the run it would bounce back up to an accurate level. But that bg fluctuation never actually happened. I tested. Then I just expected a false dip.