I'm paying super-close attention to my Dexcom CGM (continuous glucose monitor) these days, and not just because I'm a Nightscout Newbie. Nope, ten days ago I also upgraded my Dexcom to the new "advanced algorithm" that supposedly makes the unit more accurate.
News of this advanced algorithm came in a press release Nov. 6, coincidentally the same day as the Dexcom quarterly earnings call, but the FDA had actually approved the new software update on Oct. 21. According to the announcement:
"The new Software 505 features the same advanced algorithm as used in artificial pancreas research around the world. The software will impact the performance of the Dexcom G4 ... already considered the most accurate CGM system on the market today. The software will be made available free of charge to adult patients using the Dexcom G4."
Lots of us Dexcom users in the diabetes community were pretty excited to hear this. But then we started asking ourselves (and each other): So, what does this new algorithm do, exactly?
Dexcom won't disclose details of how the new algorithm calculates differently, uttering that dreaded word "proprietary" when asked how the new software does the job better than before. But Dexcom's Executive VP of Strategy and Corporate Development Steve Pacelli says it wouldn't be far off to assume the algorithm "reduces some of the noise interference" in how the system communicates glucose sensor data. He tells us a majority of the CGM performance difference can be seen on the first day of use, and overall the most noticeable differences in accuracy should be seen when catching hypos and when blood sugars change rapidly.
A Great MARD
What we do know is that this technology is what's used in most of the artificial pancreas trials in which Dexcom's involved (it is indeed the CGM used in a majority of the AP trials going on out there). Accuracy of these glucose measuring devices is gauged using a standard measure called mean absolute relative difference (MARD), and the lower the result, the higher the accuracy. And yes! Dexcom's new algorithm scores much lower than anything we've seen -- the first single-digit MARD of 9.0%, compared to 13% in the previous G4 algorithm.
This is also a huge leap from the past 16% MARD score on Dexcom's Seven Plus, and from the competing Medtronic Enlite sensor, which scores between 13 and 17% depending on which studies you believe. Some PWDs also regularly compare contemporary sensors with the now-defunct Freestyle Navigator CGM, which had a MARD of about 12%.
So yes, this new G4 is the most accurate to date -- especially on the hypo side of the range -- and it's the closest we've come to fingersticks that have a MARD of about 5.6%.
What makes this update even more important is that it's part of Dexcom's bigger game plan -- namely, to get CGM results to the level where they're good enough to be FDA approved for use in dosing insulin. It's the dream of replacing fingersticks for real -- with labeling on the CGM saying it's approved for stand-alone use.
This isn't official yet, though some of us CGM'ers probably do use the continuous readings to make dosing decisions -- hoping for a day when the CGM becomes accurate enough to reduce the number of fingersticks required, and eventually replace them completely.
In these days of high test strip costs and scattered Medicare coverage for CGMs because they're considered "supplemental" devices, this labeling would be a huge boost on both fronts.
Meanwhile, all new G4 receivers being sent out have the upgrade built in, and existing customers are invited to download the Software 505 upgrade online from the Dexcom site. You just plug in your receiver and click to update the firmware inside.
And what have I seen since upgrading my own system? Yep, my beloved CGM does seem to be more accurate and responding more quickly to my blood sugar changes -- especially in the low range. I've also noticed my G4 receiver reflects more dramatically the shifts in blood sugar, more immediately letting me know when these big jumps or drops happen instead of delaying the changes and making them appear more gradual trends.
But unfortunately, on the higher end of the BG spectrum, I'm noticing my results do seem skewed more often and way off from what my meter's showing. This is especially true when it's time to calibrate, and the readings have been 100+ points off. That's a little unnerving for me, as it's not been the case before this update.
Anyhow, it seems I'm not alone. I've seen several people commenting that it's more accurate on that lower end, though some have also reported more fluctuations and quicker arrow changes than before. And others have just not noticed any difference. As with everything in D, your sensor performance may vary.
Here's what we heard from one fellow D-peep Scott Powell in Ohio, who's been type 1 since 1981 and has been using the Dexcom G4 for several years since his Abbott Navigator CGM (no longer available in the U.S.) died.
"Overall I do think the Dexcom 505 software is more accurate. I've only had one or two tests since I updated that were > 10% error. That's much improved over the previous software. Still, I don't think it's as accurate as the Navigator was, but seems to be a definite step in the right direction."
Not Yet for Kids
Note to the wise: this algorithm update is not yet approved for pediatric use! That is, G4 was approved in February for use in kids up to age 17, but the FDA will have to approve any changes to those units separately. Dexcom says that's going to happen by end of the year, which is pretty soon. But in the meantime, anyone with a pediatric version of the G4 (it says so on the back and you get some kid-specific warnings) will have to use the older version of the algorithm. Not all kids are in this boat, as some may have "off-label" adult G4s if their docs were willing to prescribe the CGM before the new FDA approval came down.
Future Pump Issues
On top of that, two new pumps that will integrate the Dexcom CGM that are undergoing FDA review now won't be compatible with this new algorithm: the future Animas Vibe G4 and Tandem t:slim G4 devices, both slated to likely hit the market sometime in the next year or so, won't have the better-performing CGM algorithm. Ugh...
As Dexcom's Pacelli tells us that it's not just a matter of FDA filings based on the older tech; it's an issue built into the integrated insulin pumps. Neither the t:slim G4 nor the Animas Vibe will have the ability to support the new sensor algorithm.
Some science-savvy folks in the online community speculate it's probably because of how the pumps are built -- the t:slim G4's architecture has a built-in USB port that probably limits the newer software, while the the Animas Vibe actually uses an old-school IR port that wouldn't allow for the device to update the algorithm.
As a result, Pacelli tells us t:slimmers and Animas pumpers who get either integrated pump will either have to buy a separate Dexcom G4 receiver to have the advanced algorithm or they'll have to wait for the second generation of these integrated devices that will include the yet-to-be-filed G5 sensor.
Dexcom still plans to submit that next-generation G5 device for FDA review in the coming months, either by end of 2014 or early 2015. That transmitter will likely be the same as the G4 with this new AP algorithm built in, but the unit will talk directly to a smartphone, eliminating the need for a receiver.
Pacelli says this newer algorithm is actually being unveiled more quickly than initially planned. "The FDA is very keen on us running a trial and the Artificial Pancreas needs a (CGM) sensor that is dosing-worthy," he said. "We were going to wait for a future generation to add in this advanced algorithm, but the FDA was ready for that now."
The FDA has shown keen interest in moving Artificial Pancreas technology forward; remember, the regulatory agency in late 2012 approved final guidance for these AP devices and in Fall 2013 approved the Medtronic 530G that's the first device to include an automatic insulin suspend if glucose levels go too low.
And now, Dexcom plans to begin additional AP trials with the new algorithm being used for dosing, Pacelli says.
We're excited to see this technology moving forward in AP trials, and also being released so customers like us can get share the benefits in real-time!