OK - I've been using the updated algorithm for 2 sensors now, and it is really depressing how much less accurate this is than before. Is anyone else experiencing this? I keep hoping that at some point the accuracy will just magically improve, but I've only had about 3 readings (out of probably over 100) that were either spot on or within 2-3 points. Every other reading has been about 20 or more points off in either direction, not in a consistent manner, and we're talking meter (not Dex) readings between 80-150. Anyone have any thoughts on this? Previously, I was within less than 10% of my readings the vast majority of the time. Now, I'm lucky if it's within 15%.
Sorry about your experience but for my part I've found the new algorithm to be spot one which I consider no more than 10 points from a finger stick. Hope your next sensors provide better results.
Well, apparently all I had to do was complain about it publicly. Since I wrote this, the last 3 readings have been within 2 or 3 points of my Dex. Maybe I'm on an upswing?
It's likely just a mental worry more than a real drop in accuracy. I'm sure someone with a psychology degree could find a term for this, but I've noticed it with new tech like this.
We expect that an upgrade should have a distinctive benefit, but the Dexcom was already pretty darn accurate, so in reality it probably went from 95% to 96%. We're just more likely to scrutinize the lack of perfectness and blame it on the software, but as we all know, diabetes isn't a precise science. There are so many factors (bad sites, being sick, change in foods, change in habits [esp. around holidays], etc.) that are all affecting the numbers much more than the software upgrade is contributing.
It's just easier to blame the software :)
In other news, shortly after the 505 upgrade, my Dexcom receiver just sucks. It's out of range quite often, but magically comes back in range after checking to make sure it's still on. The Dex rep said this isn't related to the software, and my receiver being more than a year old is probably just dying. Guess it's time to get a new one.
A type of placebo effect??!
You might look at a standard deviation rather than a percentage off. That may be more representative of the errors.
I just got a new Dexcom with the AP algorithm. It's taken a few days, but I can finally say that its bG tracking is more accurate. Did it take you awhile to get consistently accurate readings?
hmm not sure if that was a compliment or insult. lol. we are passive, but not brainless. lol etc.. etc... etc....
Neither. Humor
The MARD dropped to around 8% IIRC, down from something like 12% with the old software. (wanted to put some real meat behind the phrase, "more accurate")
Ok - so I've been using the new software and actually find it less accurate than before. Anyone noticing the same thing?
Yes! Although it seems to be "settling down" now. The first couple of sensors were always at least 20 points off, but now it is usually within 10 points probably 75% of the time. Which is still worse than it was before for me.
Folks having difficulties: Sensor placement and application technique can have a big impact.
Consider that when doing a fingerstick, you're, well, poking your finger and measuring whole blood from a very highly perfused tissue. It's about as close to "real-time" glucose levels as you can get.
The G4 sensor wire, on the other hand, is not exposed directly to whole blood circulation. It instead is bathed in interstitial fluid around fat cells. How fast this glucose is "updated" to reflect levels in the bloodstream depends on how much blood flow that tissue gets, and what is happening mechanically with the tissue.
Turns out that fatty areas that are cold to the touch -- i.e. effective insulating fat -- isn't as well perfused as thinner subcutaneous fat. So, if you've got love handles, for example, or a fair amount of belly fat, these locales will, on average, be less accurate compared to a fingerstick comparing side-by side if BG is changing.
Also, mechanical effects: If the area where the sensor is located is being compressed (you're lying on it, chairback compressing it, etc.) it will take longer for the interstitial fluid to "update" because there can be very attenuated circulation. Try pressing your thumb into the heel of your hand, then take it away, and notice the pale spot that fills back in and gets pink. You get the idea.
So placement is very important. For these reasons, I use the backs of my arms almost exclusively, and have excellent (I think I called the 505 firmware telepathic :-)) results.
Finally, care when inserting the wire is important. I've found that if I'm not careful to make sure the sensor is perfectly flat and flush with the topology of wherever I'm applying it, making sure that the wire goes in at exactly the angle provided by the applicator, I can have a "bad sensor", where the accuracy is wonky until I change it. Not completely nuts, just a lot more large errors on some calibrations. This seems to happen mostly if the wire goes in "shallow"; on the other hand, it's easy to go in at too steep an angle and get a deep insertion that scratches at the underlying muscle, which I try to man-up and use the sensor, but usually within 24-48 hours it's too painful.
Hope this help some of you get better results!
I feel a little nervous using software without understanding the algorithms they are using. I've been needing to read this for a long time, if anybody is interested, here's the link, I think. Biostater algorithm OR, Closed loop
This is my first post. I was having issues similar to what you and others on this group have mentioned. I think I figured out what is going on with the inaccurate numbers with Dexcom's new 505 software version. I discovered that it is not a problem with the Dexcom software in my case.
I went through Dexcom's tech support channels, and ultimately ended up on the phone with one of their supervisors. Since the update, I was having a bunch of issues. False low alarms (my real BG was 154 mg/dl, Dexcom was reading 48 mg/dl), and very inaccurate numbers otherwise. Dexcom did replace that sensor for me, but it turns out that it was not a bad sensor, like I thought. I started having to test more often, because I really didn't trust the numbers I was getting with the Dexcom G4. It would have gotten expensive very quickly, as I was having to test as many as 8 times per day.
The supervisor at Dexcom thought that it may be an issue with my blood tester, and recommended that I get a new blood tester to check the numbers against. I didn't have enough money to buy a replacement blood tester, so I contacted my blood tester manufacturer (Abbott Labs Freestyle, InsuLinx). They sent me out control solution to verify that the tester is working properly, and the control solution showed that the tester was within the specified requirements.
When I got in touch with Freestyle's tech support, they had me read off the lot number that I was using to their support representative. I came to find out that I was using the test strips that DO require coding before using them. In this case, the yellow label bottle that I was using said Code 16 on it. The Freestyle InsuLinx monitor DOES NOT require coding whatsoever, and the label on the bottle is WHITE for the InsuLinx strips. The bad thing is that Abbott labs makes 3 types of test strips, and only the InsuLinx blood test strips are compatible with my tester-however all 3 types of test strips LOOK identical and will function in the InsuLinx tester. The problem was that I was using the Freestyle test strips (wrong strips) on my Freestyle InsuLinx monitor.
I went down to the pharmacy and exchanged the unopened box of strips that I had for the InsuLinx test strips. I then started using the new strips to calibrate the Dexcom G4. Since I have been using the new strips, my blood sugar on the InsuLinx tester has been within 8 mg/dl difference from the Dexcom G4. 8 mg/dl difference is satisfactory performance to me.
The problem was caused because the doctor that I go to had prescribed me the wrong test strips! For all those that are having inaccuracies with the new 505 software, please call your manufacturer of your tester and verify that you are using the correct test strips.
I have contacted Dexcom's supervisor, and Dexcom is now aware of this issue. I really wish that test strip manufacturers would make physically different shaped or colored test strips, if multiple varieties that are NOT compatible will work on the same tester!
I hope this helps people who have had the same issues as I was having with the 505 software.