Continuous Glucose Monitoring
Frontend.com
33

Thanks for the empathetic recounting. Many of us choose to put the sensor on the back of our arms, where it is less bothersome. The hypoglycemic reaction you mention in people who suffer from type 1 diabetes is also known as “insulin reaction”, which is a side-effect of injecting too much insulin (the therapy for type 1), not diabetes.

Here’s a typical trace: we can see that it’s fairly common for people, who are injecting or pumping insulin like me, to receive too much insulin causing sugars to dip dangerously low. We also observe that without enough insulin, the dots wander dangerously high. Sustained levels above 400 are often treated in hospital. It is a constant challenge to maintain just the right amount of insulin in order to influence glucose levels.

The large red dots represent calibrations. The departure from the smaller dots shows the fidelity of the device against an independent glucose meter, and how it affects performance.

The red larger red circles are calibrations, showing how the receiver’s glucose recommendations may have departed from the reality measured by another glucose meter. Most of the time it’s incredibly accurate, but occasionally it can be 100 points off as shown in the middle panel above. This was rendered using open source software created by and for the patients who use CGM devices. Notice how we’ve chosen to display indicators for how stale/fresh the data is, as well as the last actual change (+7), along with the standard graphs.

Contextual displays. The open cyan dots are the predicted range, and confidence in future glucose values.

What do we do when the data disappears? We find another way to look at the “raw” data. We can visually observe the artificial dips that take place when someone rolls over on the sensor while sleeping. When this happens, the receiver stops providing glucose information, however the human eye can easily spot the false dips that occur once an hour or so. It helps to understand that glucose levels can’t possibly change 50 mg/dL all at once. In this trace we can see fairly strong overlap between the raw values and the glucose values presented by the receiver until 2am. At 2am, the user rolled over, and the glucose data disappeared from the receiver, showing ??? until 1:30pm the next day. The lighter white dots show what’s actually going on in the sensor, suggesting a steady trend, despite a handful of errant dots.

Receiver displays ??? from 2am — 1:30pm. “Raw” data represented as smaller lighter dots show what the sensor is doing regardless of the glucose “recommendation.” This is very useful way to determine how trustworthy the device might be.

I was really glad you discussed your experience of using CGM. A lot of us depend on this technology to survive and stay safe, so we develop a special relationship with it, using it to understand our own bodies in ways that might be surprising. Some CGM users have learned how to use this data to enhance their marathon performance. It will be interesting to see how more of the population might learn similar lessons if we can overcome some of the design challenges together. Perhaps as CGM becomes more popular and accessible, we’ll all learn together.