Dexcom launched the Stelo on to the US market on 26th August as the first over the counter 15 day CGM system that doesn’t require a prescription (in the US).
It has very clear warnings that it mustn’t be used for the following:
- People using insulin
- People on dialysis
- People with problematic hypoglycemia
We might ask why not, and what could happen if someone did?
Fortunately, in the US manuals, Dexcom provides a large amount of study data which allows us to compare Stelo to G7 and draw some conclusions.
We’ll look at that data shortly, digging into study design and reported outcomes, plus comparison with the G7, but if that’s TLDR; then here’s the headline.
The Stelo appears to perform as well as the G7 and One+ over a 15 day period instead of a 10 day period. The only reason for not using it according to the bullet points above appears to be regulatory approval and lack of alarms.
This raises a few questions.
1. Would it really matter if a person on insulin used it, if it was all they could get? Dexcom wouldn’t approve, but we all know the saying “needs must”.
2. Is this the start of the opening up of the US market and the removal of all CGM from prescription?
3. When will the G7 and One+ be extended to 15 days?
EASD is only a few weeks away. An announcement there would sure steal the limelight…
Study design
The three major differences between the Stelo and G7 studies were:
- The number of adult participants (we’re ignoring the paediatrics in the G7 for now as they weren’t included in the Stelo, as it’s an over-18s device) – 308 for G7 compared to 130 for Stelo.
- The lack of glucose manipulation in the Stelo study (aimed at type 2 users and the glucose inquisitive, not type 1, so understandable)
- The mix of the study population (not clear what the percentages were in the Stelo study).
Having said that, Dexcom have provided data sets that help to alleviate the differences, by showing:
- Accuracy to comparator reference data
- Concurrence of biosensor readings with the reference analyser
- Trend accuracy
They’ve also included numbers of data points, which helps us to understand the reliability of these metrics.
Comparative Results
Thanks to the data provided in the “Technical Information” appendix of US users guides, we have some data that allows us to look across the two sensors and make comparisons.
Accuracy to Comparator reference data
First up is the accuracy to comparator reference data, grouped by comparator ranges.
As we can see from these two tables, if we weight the G7 values to <70, there is likely to be a higher percentage of values within the specified ranges for the Stelo than the G7.
We have to caveat that with the note that there are far fewer points in the Stelo study, and that without the glucose challenges, the likelihood is that the number of readings below 70 mg/dl contains far more between 54 and 69 as a proportion than below 54, which would lead to a better accuracy percentage.
Concurrence of sensor readings with reference
This analysis shows, for each range of reference blood values, what percentage of sensor readings are in, above and below range.
Once again, the G7 data goes much more deeply into low and high glucose values than the Stelo does, again, down to the target audience of the device.
Given the lack of data at low and high values for Stelo, we can only really look at the ranges 122-160, 161-200 and 201-250 to gain any insight into how the two compare. The table below shows these three sets of data plus the 70-120 for G7 and 81-120 for Stelo as a useful adjunct.
What stands out here is that in the Stelo, the “normal” glycaemia range appears to have a much higher concurrence with the YSI data than the G7. The number of data points is lower, but it opens the question as to whether the population played a part in that number. It would be good to see the data split by T1D and T2D.
Overall though, in the 70-160 range, the Stelo appears to have produced better concurrence with the venous blood than the G7, and that’s over 15 days. For the levels above 160, the differences appear marginal.
Trend Accuracy
Trend accuracy shows how well sensor detected fluctuations match those detected in blood.
Dexcom likes to state things like “When the [Stelo] sensor glucose is rapidly rising (at more than 2mg/dl/min), blood levels are are also trending up (more than 0mg/dl/min) 96.9% of the time”.
This obfuscates the point that when the sensor detects rapidly rising glucose levels, the corresponding blood rise rate matches only 54.2% of the time.
But anyway, on to the results.
Again, looking at the two datasets side by side, and bearing in mind that the number of measurements is around 50% lower, the Stelo appears to outperform the G7, better matching sensor variation rates to blood variation rates at every level.
Sensor life
The final point of note in the user guide is the amount of data over each day of the sensor and the likelihood of a sensor lasting for 15 days.
This is perhaps where it gets most interesting.
Dexcom completed a separate sensor life study, which had 108 participants and 110 sensors. 77.9% of the sensors lasted until day 15.
The stats that come out of this are:
- 1 in 5 chance that a Stelo won’t last 15 days
- Of those that don’t last the full 15 days, 1 in 10 won’t last 12.
It remains to be seen what this means for replacement programs, given that people are paying out of pocket.
Data availability
The final nugget of data from the studies looked at data availability on a day by day basis. Dexcom reported that up until day 10, data availability was always greater than 99%. Over the final five days it drops off to 97.1%.
In terms of the number of readings that means are lost on day 15, we’re talking missing roughly 8 sensor values, or 40 mins worth of data.
Discussion
As we see in the results, given the caveats of the differences between the studies, the numbers from the Stelo are very similar to those of the G7, in some cases appearing to be better.
This shouldn’t be an enormous suprise, given the origins of the Stelo sensor.
What’s perhaps more surprising is that they’re coming from a 15 day version of a sensor that’s previously had a 10-day lifespan.
Members of the diabetes community (#WeAreNotWaiting and those who are) have long extended the wear of their Dexcom sensors and probably aren’t surprised to see that, in a proper study, the numbers come out favourably. Many of us have worn G6 sensors for longer than the recommended lifespan for a long time, with good results.
Ultimately, the data suggests that using the Stelo “off label” wouldn’t present huge challenges for this with T1D. Yes the app doesn’t do alarms, but if you really needed it, I’m sure it’s better than fingerpricks.
The real question is “What is the read across in this data?”.
With such good results from 15 days, and the ongoing commercial pressures that Abbott are applying with their Libre2+ and Libre3+ 15 day sensors, I’d expect a 15 day “G7+” and Dexcom “One++” to be just around the corner. Even at the failure rates stayed, which I’m guessing aren’t too far from those of the Abbott products.
What we’ve not been shown, and probably matters quite a bit for an iCGM, is the performance of the sensor on those extra five days. #WeAreNotWaiting users looping off extended sensors would tell you they should be fine. Dexcom will have to prove it.
And perhaps the G7 will even be available without a prescription…
In fact, there’s a large Diabetes conference coming in just a few weeks. EASD. Perfect for an announcement.
Wouldn’t that steal the limelight?
Leave a Reply