Having had the Libre2 fall off during the first few hours of the “Six of the Best” test, I decided to run One Vs Libre2 (the two cost neutral options on the NHS Drug Tariff) head to head against one another.
In this case, I used the Librelink software to scan the Libre2 at allotted times, and Diabox to provide real-time data from the device, to see if there were any major differences.
Introduction and disappointment
Unfortunately, the Libre2 sensor started to perform poorly after approximately 10 days, and had completely failed by day 12. Abbott are again replacing the sensor, however so far, from a purely physical use case, the Libre2 has two failures compared to Dexcom One with zero. Under normal circumstances, I’d expect a user to request a replacement sensor when the scanned values diverge from blood by over 20%, which as mentioned, was day 10.
As a result of the above mentioned sensor failure, while the full set of data presented will show some areas of divergence, I can’t say that it’s fully representative of Libre2..
As already mentioned, there were one or two issues with the Libre2 sensor latterly, and it eventually died early. As a result, while I’m presenting data here, I’d argue that it’s not the highest quality.
Firstly, consensus error grids, with Dexcom One, Libre2 via LibreLink and Libre2 via Diabox overlaid, then comparing all as pairs against each other.
Compared to the data in the first tranche, there is much lower dispersion across all three ways of delivering the data.
Secondly a series of MARDf calculations.
Firstly, for me personally, I’d call the reliability of the Libre2 sensor into question, given that the two I’ve had have either not stayed stuck on, or failed early.
I may just have had a bad couple of sensors though.
Looking at the data, during the period where the system was working as expected, it turns out that the Libre2 generally had a lower variance from the reference than the Dexcom One did.
Comparing the Dexcom One with the Libre2 scans, it’s very clear that this One sensor tended to overstate values, while the Libre2 tended to understate them. In comparison, during the first sensor I tried, the One was much closer to the mid-line of the graph.
What’s also clear from the graphs is that with the real-time data, and importantly the ability to calibrate when necessary, the Libre2 data was again dispersed more around the middle of the graph, as the bias (which incidentally grew in the latter days of the sensor) was negated by the calibration.
Finally, the MARDf data shows a surprise in the values that were obtained with the ONE, compared to the poorly performing Libre2 and the Diabox real-time feed, with the ONE sensor not performing anywhere near as well on this occasion. This really highlights why a single sensor test has some limitations in its validity. It can only ever be indicative.
From the variety of sensors tested, and the results obtained, there are a few takeaways that come out of this.
- Attempting to run n=1 rials comparing sensors with fingerpricks takes its toll on your fingers!
- Single sensor tests are subject to error, just by virtue of sensors not working as they should, whether that’s the difference between the two Dexcom ONE sensors in this and the last comparison, or the Libre2 starting to fail and then stopping with two days of life supposedly remaining.
- Of the sensors I’ve tested, using multiple examples, on me the Libre2 and Dexcom sensors generally work reasonably well. The GlucoRX Aidex, Glucomen Day and Medtrum family of sensors don’t work well for me and are generally inaccurate.
- The quality of the “accuracy studies” quoted by manufacturers is really important if you don’t have the luxury of trying out multiple CGM sensors. Official guidance is needed on how to interpret data, when there is massive variation in the approaches taken.
Overall in this series of tests, the two sets of sensors that have stood above the rest for accuracy on me are once again, the Dexcom ONE/G6 and Libre2.
In terms of application design and usage? They’re all very similar.
And in terms of sharing data? Well if it’s official, then there’s only really one winner amongst those available on NHS prescriptions if real-time CGM is required. That’s the Glucomen Day, which at time of testing was the only one with an officially supported “follow” app. Otherwise, the Libre2 also provides scanned data and alerts, but not rtCGM. Otherwise, ONE works with xDrip and Nightscout, and there are systems like Diabox that also allow the use of realtime following with Libre2.
The hard question is always, “Which system would you choose?”. For me, given I’ve had the luxury of access to all of these systems, I’d say the Dexcom, as it generally provides the best results for me without having to “create” a CGM. Given its results, I could use the Libre2 though.
Of the six of the best, it turns out that it’s much harder to pull out a clear winner than you might expect. It’s just very easy to identify those who don’t come close.