Having tested a G7 against the Libre3, and found that without calibration, it tended to produce results that I wasn’t satisfied with using just the factory calibration, I thought I’d have another go and see where we got to.
This is one that was ordered from the UK website, and applied according to the user guide on my arm.
The reference data recording wasn’t as methodical as the capture for the previous study, but the results were very far from what I was expecting.
The MARD Vs fingerprick with this device was 22.3% and the MRD was 15.9% (using the same Contour Next blood testing system that was previously used), indicating that in general it read considerably higher than the reference finger pricks (and the G6 it was sitting next to). There are only 28 matched pairs here, so statistically it’s an irrelevance.
As is clear from the grid shown, the dispersion of the very few points collected was quite wide, with only 62% of the readings falling within zone A.
The modified Bland-Altman plot shows the dispersion in relation to the 20/20 boundaries. It’s clear that quite a number were a long way outside. It’s fair to say that this sensor did not perform well.
Additionally, the device fell off after 6 days.
What’s going on?
I’m not sure. This is the second device that I’ve used with readings that were a long way off when using the factory calibration, and also the second that fell off early (one with and one without the overlay patch).
I submitted an issue and have a replacement sensor coming, which I’ll check again. If I still see the same problems, then I assume that I don’t get on with the G7 as well as I did the G6.
But why might that be the case? One of the major differences between the G7 and the G6 is the orientation of the sensor wire, which is now vertical, compared to the angled entry of the G6. It’s possible that this could be interacting with me in a way that causes movement within the “sheath” and affects the results.
Having said all of that, anecdotal observation of the G6 and ONE systems that I have historically used has also recently shown a higher level of variation than I have been used to with the factory calibration. At least once in the life of a sensor, I am now checking and calibrating. Something seems to have changed, but I can only assume it’s me, as I can’t imagine that the devices are being manufactured or set up differently.
The only answer seems to be that further tests are probably required.