Over the past eight months, Nemaura have announced a number of outcomes of their clinical trials, all of which have suggested that after a tricky start, they may be moving in the right direction. But if we look at what these results show, I believe they raise as many questions as the answers they provide.
Taking a step back and looking first at the data from January, where the following was stated:
Results indicate an overall MARD (Mean Absolute Relative Difference) of 13.76% over a broad dynamic glucose concentration range. However, up to 70% of the data from the study paired between sugarBEAT® and the venous blood glucose concentration achieved an average MARD of 10.28%, denoting even greater accuracy. A MARD of 10% is deemed to be sufficient for making therapeutic decisions. No serious or major device related adverse events were noted.
70% of the data showed an average MARD of 10.28%. Is that 70% of the participants, 70% of the time, 70% of something else? And what about that headline figure? 13.76% isn’t great in a clinical environment, and sits well outside other CGM systems.
If it’s 70% of time, that’s 70% of “up to” 14 hours per day wearing the system. That equates to 40% of the overall amount of time in which the trial was undertaken that the MARD was 10.28%. And even if it’s 70% of wear time, that’s less than ten hours per day. That’s not a reassuring number.
Given that it’s advertised as an “up to 24 hour patch”, and in a 14 hour period, only 70% of the “something” was it at 10.28%, what happens if you wear it for the other 10 hours per day? The 14 hours per day limitation of the trial leads me to question what happens to the accuracy outside of 14 hours of wear. Maybe I’m just suspicious.
More recently, Nemaura announced interim results from home trials. It’s worth reading the full details:
The interim results evaluated data from 25 patients, split approximately equally between Type I and Type II diabetics, each wearing sugarBEAT whilst going about their daily home/work routine for 1 or 2 non-consecutive days, for a total of 36 patient days. Each patient recorded up to 5 finger prick readings per day, at approximately equal intervals, over periods up to 14 hours. Readings were measured using the Abbott Freestyle Optimum neo Blood glucose meter (BGM). The results analysed 121 matched pair points between the BGM and sugarBEAT®, and indicated 84.3% of the data points had an overall MARD (Mean absolute relative deviation) of 10.63%, and an overall nominal MARD of 16.3% (compared with 14.8%, 16.3% and 18% for Eversense, Dexcom G5 and Abbott Libre Pro respectively*).
Once again, this is a small sample. 25 patients, 36 patient days, a maximum wear time of 14 days and “up to” 5 finger pricks per day. In anyone’s book, this can’t be seen to be statistically significant.
In addition, the Clinical Trial Presentation references the recent in-home trial of Dexcom, Senseonics and Libre Pro, whilst ignoring the major differences between that trial and this, namely that in those cases, the reported MARD numbers were for 24 hours of wear, not “up to 14 hours” and that there were 6.5 times as many compared data points. It’s quite a significant slight of hand being undertaken by the marketing department.
For comparison, matching 126 data points across 25 participants is only 12 more than I had in my month with the Medtrum versus the Dexcom G5 with one participant.
When analysing a selection of the data points, you get a great number, but when analysing the full set, it broadens widely, and once again, this is “up to 14 hours”. It’s worth noting that a period of “up to 14 hours” can be as few as one, accounting for the warm up period, and with no indication of when, could also account for a more “helpful” period such as 9pm to 11am, when one would expect to see little rapid deviation in glucose levels.
Overall then, what’s been presented leaves me with a lot questions and presents a picture that the data has been deliberately presented in a favourable fashion. That’s fine for a marketing piece, but not so good for a clinical trial.
The same questions arise about the 14 hour duration. Why?
- Is there an issue with the 14-24 hour period in terms of accuracy?
- Is there an issue beyond 14 hours with the adhesive?
- Does the transmitter not hold charge well?
- Does the mechanism used to monitor glucose levels cause some kind of reaction that is alleviated by reducing the wear time?
If you are claiming up to 24 hours wear then why wouldn’t you use 24 hours in your clinical trial from which you intend to publish the results?
So what am I taking away from this? I’m hoping that the full suite of testing being undertaken for the FDA approval provides far greater insight into the effectiveness of this solution. I still like the idea of it, but what’s been published so far doesn’t make me all that confident about the solution, because of the number of questions that it raises.
Let’s hope I’m wrong!