Replies to This Discussion

Permalink Reply by Doug on September 4, 2011 at 8:44pm

The 20% plus 20% argument is BS
The accuracy is +or- 20%
The idea that the difference you see on the dex is due to the difference between ISTF and BG is also BS - the algorithm can be made to accommodate the differences. The navigator surely did.

Ive been wearing the Dex on and off for almost 2 years. The ONLY thing its good for is trends.. Is my BG going up or down ? has the rise peaked and started down yet ? The numbers are OFTEN off by 100% and at least 40% of the time off by 50%. I should call Dex support every time BUT I cant spend an hour or more per crappy sensor downloading data and uploading it and explaining it to the support people. Eventually Ill tire of paying 20% for the sensors. and simply stop ordering them.

The really sad thing is that Dexcom claims they have better accuracy than Medtronic. If Dex is the best then the artificial pancreas project is doomed.

Permalink Reply by jims on September 4, 2011 at 9:22pm

ANother happy customer - oh great. As i said, the industry should be shot over the 20 % BS.

Tracking though is not presently good as reported to fast moving events.

I have made it work for me but that does not mean I am happy either with issues like those raised raised by prior writer.

In addition, Dexcom prone to hide out and not respond which is unfortunate.

Permalink Reply by Doug on September 4, 2011 at 10:20pm

In your opinion after reading one post Im a lunatic ? And full of poop ?
Great argument - clearly a member of the debate team ... Thanks for sharing.

Back on topic - call it whatever you want - algorithm - accuracy - product quality - computer program - how ever you describe it - Dexcom is doing it BADLY by ANY measure, but especially badly when compared to the recently departed former competitor the Navigator...
I used a navigator for 2 or 3 years and it was typically within 10 - 15 points of my finger sticks. it was the exception when it wasn't.

More of my lunacy is the question, if there is NO math ( or algorithm) going on between your calibrations, the ITSF and the readings then why, when the Dex says your BG is 200 and you test and its 95 and you calibrate the 95, does the Dex drop about 15 points to 185 and SLOWLY trend down ? That to me is an algorithm. But you will call it something else...

Doug in Skagit County - All ANY of us can provide are opinions and experiences, which is the point of these kinds of social networking. The FACTS are that twice today the Dex was off by over 100 points.
My opinion is that SUCKS.



.

Permalink Reply by Doug on September 4, 2011 at 10:36pm

Doug in Skagit County-
National Institutes of Health is also apparently full of poop
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2825632/
"Thus calibration algorithms capable of reducing CGM error are important additions to the arsenal of data processing techniques that help transfer raw current into BG readings. Algorithms reducing the time lag are critically important as well."

The Journal of Diabetes Science and Technology also seems to think there is an algorithm to correct between the sensor and the finger sticks ...
http://www.journalofdst.org/March2010/Abstracts/VOL-4-2-ORG16-BEQUE...
"Algorithms for real-time use in continuous glucose monitors are reviewed, including calibration, filtering of noisy signals, glucose predictions for hypoglycemic and hyperglycemic alarms, compensation for capillary blood glucose to sensor time lags, and fault detection for sensor degradation and dropouts. A tutorial on Kalman filtering for real-time estimation, prediction, and lag compensation is presented and demonstrated via simulation examples. A limited number of fault detection methods for signal degradation and dropout have been published, making that an important area for future work."

Permalink Reply by Doug on September 5, 2011 at 7:19am

In order to discuss an algorithm Im required to provide one ? Huh ?

I never said it would be easy for me - Simply that it HAS been done. And that Dexcom should strive to improve theirs.

I thought that this discussion is used by users and POTENTIAL users and they should see opinions from BOTH sides of the discussion.

I guess this discussion is for Dexcom cheerleaders only.

Rah Rah - Dexcom you are doing a great job - keep it up. Dont worry about accuracy. We love you anyway....

Sorry I participated...

Permalink Reply by jbowler on September 4, 2011 at 10:46pm

"The 20% plus 20% argument is BS"

Hum. Is that the statement of someone well versed in statistics? In which case why don't you give the correct answer? Or is it, itself, BS? Ok:

We have two BG meters. Both are presenting us with approximations to our actual, intra-veinous, blood glucose. The FDA requires these meters to be within 20% of the correct (IV) answer some proportion (probably 66%) of the time. So:

reading x = IVBG +/- 20%
reading y = IVBG +/- 20%

Therefore (and, yes, I know this is wrong):

reading (x-y) = (IVBG +/- 20%) - (IVBG +/- 20%) = +/- 40% (of IVBG) 66% of the time.

Ok, that's wrong because (a) the answer is 37% and (b) that assumes that the two meters are acting independently: let's hope they aren't!

Permalink Reply by jbowler on September 5, 2011 at 12:30am

I have to admit I interpreted the comment as a complaint about the accuracy of the Dexcom compared to the Abbott FreeStyle Navigator, probably when compared against Abbott FreeStyle test strips. Abbott claims, here:

http://www.abbottdiabetescare.com/adc_dotcom/url/content/en_US/gene...

That 81.7% of the readings were in zone A ("clinically accurate, would lead to correct treatment decisions) and 16.7% of the readings in zone B ("would lead to benign decisions or no treatment") using the "Clark Error Grid", with a total of 98.4% in either. Dexcom claims (in chapter 10 of the Seven Plus CGM System User's Guide) that 73% of the Dexcom readings were in zone A and 23% in zone B, giving a total of 96% in either.

The Abbott results dropped to 96.4% when they conducted the tests at home, not in the lab. The Dexcom tests were conducted for seven days with one of them in the lab.

We can see the accuracy figures as well as the statistical significance from the Dexcom book, I guess the Abbott data is available somewhere too but it wasn't immediately obvious to me where. Dexcom has 76% of the readings within 20% (of IVBG), 90% within 30% and 95% within 40%.

As Dexcom observes it is more informative to overlay accurate blood glucose readings (not from a finger stick) over the Dexcom trend.

Permalink Reply by Doug on September 5, 2011 at 7:13am

The FDA rules are a different potentially interesting argument.. The point Im making is that its possible to make a more accurate, precise sensor and Dexcom isnt doing it. The fact that they are operating withing the FDA rules just means the FDA has set very low standards. If there was a more accurate option I would be using it.

Permalink Reply by jims on September 5, 2011 at 8:04am

The rules are asinine especially when one is chasing lows. These rules are unacceptable and FDA should know better. On the high side, yes the errors do not have much impact. On low side totally different story.

I am fed up with statistical analysis that attempts to prove some stupid point and not real world.

There are two approaches in this world. Back in early days of Boeing Engineering, they would literally
bend and stress an airplane to see actual wearout - stress factors to prove they had made/met worst case design goals and safety. . There is another pack of clot morons
who do statiatiscal analysis to show why statistical analysis is adequate and why doing actual worst case testing and design are not necessary as statistictal analysis says plane design should be ok.

Which plane do you want to trust your life to?

Permalink Reply by jims on September 5, 2011 at 9:44am

Doug:

Excellent response.

In fact I do get very useable helpful data not withstanding all the worts.

Yes statistical analysis provides a starting point to see if any co-relation and then hard testing needed to validate that the statistical analysis was picking up correct issue.

I do not need spot on readings and I do work with trends and handheld cross checking each other.

Its just the balance of the human interface at Dexcom and the issues like having to keep my unit chilled so it does not jam and stop, watching out for interfereing servers/and wireless gear crap
causing machine to jam up on old reading ( worst in my condo /home) , the startup delay which really can be a whole day and trying to get max use out of sensor.

I am not unhappy but the dancing and waltzing getting it to run 24/7 without major hickups can be fun.

Yes, this is good step forward. But as a 40 year experienced electronics engineer and computer hardware engineer, I see the warts and wrinkles up front while Dexcom offers little explanation/detail; upgrades ( FDA won't let us fix nothing - what you get is it)

If I had a emulator and development system, I would dive in there to see what they actually did both in great works, difficult tradefoffs and how good the manufacturing/testing worst case work was done.

I just wish the interface knowledge sharing, customer interface guidance was better at Dexcom. If the FDA is buggering that up - a pox on both their houses.

The fact one cannot upgrade software issues on a product to make it work correctly to match original approved design spec and not adding new features without massive federal approval checking process is outright shortsighted and unhelpful. Yes regression testing needed always - but surely the FDA can offer process, approval,reporting that promotes fixing the bugs, eases user problems and makes experience better without ham fisted regulation effort locking product to ocean bottom with 10 ton anchors.

As one who has spent thousands on finger stick technology strips - 32 readings a day until I got Dexcom, As great as it helps ; it just seems such a contradiction to be using a $ 900 electronics/computer package with $ 300 odd dollars in sensors per month ( and less if sensor grants one extra use) and using a $ 20 dollar handheld and $ 1 dollar strips to calibrate, keep cgms in line and then get occasional off the wall readings and outside the 20% .

On top of that , ones cost are really the handheld unit 6 to 10 strips a day plus the CGMS and probes. Is that bad - no, it is saving my life and getting my liver and type 2 monster caged after 30 years of being driven wrong.

On top of it, every time there is a crisis blowup and thanks to low speed tracking , I am out with the strips seeing what is tearing loose.

The CGMS really catches a lot of data having to do with digestion times and other idiotic behaviour in my body that finger stick machines cannot do.

Thank you for adding to discussion and excellent thoughts.

Permalink Reply by JamesDG on September 5, 2011 at 11:59am

OK, Lessee...

I've been on the DexCom (7 Plus) since 2009, May. Long enough that pretty much all of the varying-experience scenarios reported here are inside my experience.

Doug, the only really important question I'd ask is, "Are you happier and better off managing the diabetes with the DexCom than you were without it?" If yes (...which is, by the way, my answer...), then even with its clear and (experience-obvious as opposed to not-quite-to-promises-)obvious shortcomings, it is well worth its expense and its incumbent frustrations. If not, then you're under no obligation to continue using it--ultimately (and I'm certain you know this, but it bears repeating...), you get to choose how you manage the disease.

For me, the DexCom has been a major godsend. I've seen the wide swings from finger-stick readings, but having the continuous data stream, and especially being able to see the trends is what I'm talking about. My A1c's are improved, and the incidence of BG crashes, while not zero (...although, it was zero for the first 18 months I used the CGM...) is way down from the three or four call-the-parameds incidents a year I was having before the CGM

Now where algorithms and such are concerned, somebody in the thread mentioned Kalman filtering. I have no direct view inside DexCom's engineering, but I know damn well they have to use some form of predictive number-crunching (which is precisely what Kalman filtering is an example of), since it's well known and understood that interstitial glucose levels lag (by 5 minutes...10 minutes...?) BG levels. There are also variables like where the sensor wire is inserted and whether or not that particular tissue is vascularized per DexCom design specs. (I have lipomas on my abdomen, and when I inadvertently stick a sensor into one of them, I usually have to toss it and start again within 24 hours--the infamous "???"s, doncha know...!)

Now in this context, I think the comments about the FDA are spot on. What they are really saying, without saying it (or, for that matter, permitting DexCom or anybody else to say it) is, "We don't trust you adolescent, pus-brained idiot, bone-headed members-of the-unwashed-masses, who have never been attorned into the brotherhood of Medical Validity Evaluators, to make any kind of judgment regarding your use of all these wide and varied things that you can only have access to if we say you can (...we, of course, being the only people on the face of the planet qualified to make such judgments on your, or anyone else's behalves...). So just shut up and trust us--that's what we're here for!" Frankly, I’d love to be able to write an app for my smart phone that acted as the DexCom receiver--one less gizmo to hang on my belt! But having no access to the design specs, or any of the engineering, or even what proprietary variation (if it's as I suspect) of BlueTooth protocol they're using, and since DexCom can't offer an Android app without getting massive FDA approval, any discussion of improving things on my own, or in the company of like-minded enthusiasts, is moot. Open source will never to happen with anything the FDA has hegemony over.

A sinecured bureaucrat, regardless his bona fides, is not, never has been, and never will be, a substitute for the synergy of an educated, informed, and motivated client ("patient ") working with a competent, informed, and interested professional consultant ("doctor" (...which, by the way, comes from the latin for "teacher"...)). All of the long, sordid history of bureaucratic interference under the aegis of "protecting the public" has not, in any demonstrable way I can see, ever managed to add to the "public welfare" (all the trumpet-voluntary protestations to the contrary notwithstanding), and in many demonstrable instances, has subtracted form it egregiously.

So maybe it's time to start a new thread about the efficacy and validity of government-managed health care--this thread being about DexCom accuracy, and all. Speaking once again from my own personal experience, I do find that the DexCom's accuracy, especially in the low ranges, is more than adequate to make my worries about BG crashes way less than they used to be, and to provide, overall, a significant improvement in my ability to manage the disease. I'm sticking with what works for me!

Permalink Reply by jims on September 5, 2011 at 9:56pm

After thinking on this response I finally twigged on what for me is an important issue:

Originally I was on 75/25 humalog when I first got my dexcom. After a month or two; my body decided to throw off all the excess insulin and Doctor put me on small doses of humalog Lispro standard. AT that point my BG moved very slowly for first time in 30 years and smoothly like waves on placid lake. In addition, Dexcom tracking and use became far better and easier.

When on the stinking 75/25 along with some exercise would see BG really scram fast. Far faster than dexcom could follow. To that end when on that crap; I would describe the Dexcom performance as terrible and could not catch error conditions fast enough.

SHould a person have a twichy fast moving system, special care and intuition is needed to use. Just becasue above writer has fast enough response does not mean others do. This is test equipment to help resolve medical conditions.

Yes this does job for me and what other choices does one really have.

The data and non existant graphs that should accompany said device listing its honest capabilitities, speed of response and expected speed on interstitial tissue do not exist nor stated.

When I buy a 10 grand oscilloscope I get all that data and can understand performance of said device and how to apply to my test problem and limitations. With Dexcom and I am sure for others, it is all crapshoot and guesswork and patience to work thru and figger out how best to settle down and collect data and what it means.

• ‹ Previous
• 1
• …
• 3
• 4
• 5
• 6
• Next ›
• Page