Screenshot of the Parts of a CGM
As we can see, most of the limitations of continuous glucose monitoring come from the technology we’re working with, and as such it is instructive to take a look at how the process itself works to be able to clearly define what these limitations are and where exactly they come from.
The clear thread through all of the different mechanisms of directly measuring glucose concentration is that is that all of these different approaches convert what we want to measure, the concentration of glucose, into some sort of electrical signal. That signal contains all of the information we need, but now in a more usable form, one that we can change and interpret in order to eventually give something meaningful to the user.
Thus, from here the goal is to take our electrical signal and process it into something that a computer can more easily understand, and then display that information to a user.
Let’s focus on the first part: going from an electrical signal to a processed signal. This isn’t really a single step, rather it has multiple components that we’ll go through in turn. First, converting from a current signal to a voltage signal. Second, performing an analog to digital conversion. And finally, calibration.
The first step is to convert from our current signal to a voltage signal, which can be done with a simple circuit design: a current source driving a resistor. If you’re familiar with any sort of electric circuits, you’ll know that Ohm’s law dictates that a current driving a resistance produces a voltage, and that the voltage is proportional to the resistance.
The important point is this: we’ve converted our signal from one form into another, in this case from current to voltage. What you’ll notice is that because these signals are directly proportional to each other (the higher the current is, the greater the voltage, and vice versa) we can work with a signal in a more usable form like voltage without losing any information.
However, one place where we do lose information is in the analog to digital conversion. Computers don’t tend to work with sine waves and smooth curves directly. At the most fundamental level, they work with numbers, and there’s not an infinite number of numbers to work with. We have to work with a digital signal, and so we must adapt our continuous curve to one which is quantized, and in the conversion there has to be some loss.
There are ways to limit this effect, but this costs memory or time or some other resource which is not infinite [28]. This is a reliability limitation: there will always be a loss of information and thus a loss of accuracy.
The last step is another limitation, and it’s the consequence of the fact that we’ve been changing the form of what we’re measuring at all of these steps along the way. The current signal is proportional to the concentration of glucose, but this depends on the concentration of the enzymes, the temperature, and many other factors. The conversion from current to voltage depends on the resistance of the circuit. These conversions require us to calibrate our final value to determine what exactly it represent in terms of glucose concentration. This is done through “retrospective calibration” [24] which in essence comes down to making a line or curve of best fit. This defines the relationship between our final result and glucose concentration, but one again, reality creeps in: it won’t be perfect. Thus, our last reliability limitation: the accuracy of calibration.
All in all, the limitations that come directly from the technology are very few, but it’s important to consider that this is in large part because of the scope of the device: it’s not treatment or a cure, it’s just a sensor. It’s biggest limitation, in that regard, is that it’s just monitoring and providing data, and so beyond rubbing up against the limitations of technology, the complications largely come in with how we interpret that data and what we choose to do with it.
References:
[24] Klonoff DC, Ahn D, Drincic A. Continuous glucose monitoring: A review of the technology and clinical use. Diabetes Res Clin Pract. 2017 Nov;133:178-192. doi: 10.1016/j.diabres.2017.08.005. Epub 2017 Sep 1. PMID: 28965029.
[28] Atmel. (n.d.). AVR121: Enhancing ADC Resolution by Oversampling – Microchip Technology. microchip.com. https://ww1.microchip.com/downloads/en/appnotes/doc8003.pdf