With two sensors now used and a month of data captured, I felt it was time for further feedback on the newest CGM system on the block. Sadly it’s not all sweetness and light.
To summarise, whilst the app is easy to use, the system is much more comfortable to insert than the Dexcom G4 and G5 sensors and the team at MEdtrum are a very pleasant bunch, the variation from blood isn’t congruent with what we might consider to be a modern CGM and the life of the sensors seems to be rather variable.
The headline is that the Mean Absolute Relative Variation from Blood (MARVfB) over the two sensors was 14.9%, compared to 10.7% on the Dexcom G5. My first sensor lasted 11 days, my second stopped providing data when it expired at 14 and the sensor filament pulled out when I removed the transmitter to attempt a restart, and the third was no longer available to the app at 9 days, in spite of attempting to change the transmitter. The third one also saw a significant degradation in the adhesive, well ahead of the Dexcom G5 it was sitting next to.
The basic conclusion that I draw from this is that there seems to be variability in the quality of the sensors that I used, although all were supposedly from the same batch.
So lets dig into the data and some observations…
At a basic level, the A6 did a job. It acted as a CGM, but over the time period, it variation undid it. The high level numbers show:
What we can take from this is that the A6 was never anywhere near the 9% value stated in their literature, indeed, overall, this performed more like a Dexcom Seven or the early Medtronic (pre-Enlite) sensors. The feedback that I’ve heard is that no-one has yet managed to replicate the claims that have been made, and I can believe that. Obviously, my second Dexcom sensor wasn’t as good as the first either, but the variance possibly says more about when I was testing blood than about the sensors themselves, so even with that in mind, it suggests that the A6 has less tolerance for unhygienic calibration (i.e. having a flat trace and in range when calibrating).
If we look at the error grids for the two sensors, we see a similar story:
The A6 clearly has greater scatter and more outliers than the G5, although the two outliers with the greatest risk were from the G5.
Looking at the risk data presented above the graphs, we can see that in this data collection, 81.6% of the A6 readings presented no risk as a result of the variation, while 14.9% presented a very slight risk and 3.5% a slight risk. The Dexcom turned in figures of 86.8%, 9.7% and 2.6% by comparison.
What we can also see is that the A6 read higher than the reference blood value on average, while the G5 read lower, on average.
Comparing the sensor data directly:
You can see that across the lifespan of the two sensors, at any point in time there’s quite a noticeable difference between the amounts that the sensors were off from the blood tests. As already mentioned, on the second sensor, both were a lot worse than the first pair. This really highlights the differences between the two and also demonstrates why you may not want to closed loop with an A6 sensor. Discussions within the T1D community have already raised that the A6 didn’t pick up lows all that well, and given that it’s supposed to be used with a low glucose suspend system in their sensor-enhanced pump, I’m not sure I’d be confident that it would suspend for predicted lows successfully.
So that’s the data. I think it’s pretty self explanatory. But what of the other observations mentioned? Well, if you’ll just hang on a sec…..
I mentioned that I was unable to extend my second sensor after recharging the transmitter due to the transmitter pulling the sensor filament out when I detached it. What did I mean? The pictures of the sensor below show you.
The sensor is composed of a foot plate and the sensor filament. When you use the applicator, the sensor filament is injected through the plate, a small “cogged” plate is twisted to hold the filament to the base, and the needle retracts. You can see each of these components in the pictures below.
You can clearly see the metal lock ring in place at the base of the sensor.
The filament section which is passed though the base and locked in place by the applicator.
A view from above showing the lock ring and the plastic lock points into which it locks to hold the filament in place.
While I applaud this engineering, as it is pretty impressive, if the applicator doesn’t quite lock that ring in place when you apply the sensor, then when you take the transmitter off to charge, the filament comes with it. That’s what happened on sensor two. Sensor one and three locked okay as I was able to remove the transmitter without losing that central section.
Which brings me back to questions about the consistency in quality of the devices. Now I mark that it’s a very small subset of the market, in that I am one person and I’ve only used three sensors, but only one has lasted for the published fourteen day claim, and one wasn’t properly connected when it was inserted. I’ll leave you to draw your own conclusions there, but I know what mine are.
While I’m pleased to see Medtrum enter the CGM market and provide some competition, I’m a little disappointed by the level of the technology. It’s not up there with what you get from Dexcom, Libre or Medtronic today, and that’s a large gap to recover. The rechargeable transmitter is a welcome inclusion, and replicates what Medtronic provide, although the charge duration seems better in the A6.
The downside for me is the accuracy, as already mentioned, and the consistency across the sensors in production. While all reached their warranted period, the implication that the sensor will last 14 days (by virtue of the setting in the app and the discussions that various members of Medtrum’s staff have had with members of the public) doesn’t match my experience with two out of three of the sensors I’ve tried. The implication is, of course, that if a sensor lasts less than 14 days then it becomes more expensive to the end user, so the upfront price tag becomes less important.
Whilst the pricing is good, coming in at the same level (more or less) as the Libre pricing on the sensors, with the advent of the G6 from Dexcom and a pricing model that has made Dexcom cheaper, if not easily affordable, you have to ask yourself what’s most important to you. Upfront price, accuracy or lifespan. For me accuracy and lifespan matter (as a self funder), so I wouldn’t choose this system over either Libre or Dexcom as a result.
As a footnote, I’ve heard recently that the Bristol’s Paediatric Diabetes team have started to recommend this system (presumably as funded), and if that”s true, given my experience of this, I’d really like to understand the evidence that they’ve used to make that decision.
If you asked me whether I’d use this long term, I’d have to say no. There’s clear work to be done to bring it up to speed with the current best in class, and in its current form, I simply couldn’t trust it enough to give me reasonable values. As always, your mileage (and Diabetes) may vary…