Comparing Dexcom’s G7 and One+. Is there much difference?

As many people will have seen, Dexcom have been expanding their product line based on the G7 hardware. The replacement for their more budget device, the One, is the One+, and they’re using the same hardware for the Stelo, the first US over the counter CGM.

Whilst the One+ isn’t yet available in the UK, I was lucky enough to be provided one by friends in Europe, as well as a G7 to run it alongside. Thanks Kamil and Inge.

As you can see from the above image, they’re basically the same, and I was able to use two variants of xDrip to capture the data.

Both were applied side-by-side on my arm and the intention was to analyse the difference between the datapoints of the two devices, rather than compared both to fingerpricks.

Feedback from those that have scanned the firmware is that this is also virtually identical

Unfortunately, I only got around 6.5 days from the G7, due to one of the edges coming loose and sticking to the sleeve of a jumper, which resulted in an accidental removal.

This resulted in 1,713 datapoints.

The analysis

With both sensors applied within a minute of one another, the clocks were virtually synchronised and produced datapoints on the five minutes marks of the hour (:05, :10, etc) making comparison relatively easy.

I exported the datapoints for each sensor from xDrip and aligned them using Excel. I then calculated the relative difference and used the statistical functions in Excel to analyse the data.

Relative difference is:

100 * [ (yCGM1 – yCGM2)/mean(yCGM1,yCGM2) ]

Where yCGM1 and yCGM2 are the two datapoints from the CGMs at the same time.

Results

If we plot the data on a graph, we can see that there’s not much difference between the two sensors for the majority of the data set. It’s only really on day one that there’s anything noticeable.

Zooming in on day one, the G7 appears to have had a slightly different experience of insertion compared to that of the one. I wasn’t comparing data with fingerpricks, so there’s not an answer to the question “Is one more accurate?”, but it highlights that day 1 data can be unreliable and vary between sensors.

Looking at the relative difference , the descriptive statistics suggested that we didn’t have a normal distribution, which is backed up by the below histogram.

This time, there appears to be a bimodal distribution, with, once again, the day one variation shown influencing the negative tail.

The mean value of relative difference was -0.169% and the standard deviation 15.965.

Plotting the relative difference of the period of sensor wear demonstrates where most of the variation occurred. Adding bands for ±15% also demonstrates that the vast majority of the difference was well within one standard deviation.

There is also a significant difference on day two, which corresponds to approximately 1.30am and is assumed to be a compression low on the G7.

Discussion

The key takeaways from this analysis are that, as expected, the two sensors are very similar to one another in results, which was anticipated.

In general, values were mostly within a range of ±15% of each other, which is well within the operating range that Dexcom would expect from a single sensor.

Much of the variation (and therefore skew) in the statistical analysis is driven by the significant differences on day one, which highlights the sensor to sensor variation that people see in general after insertion.

Otherwise, the key takeaway here is that, as you’d expect, there is next to no difference between the G7 and the One+. Which is exactly what we want to see.