TOOLBOX: Thanks to advances in wearable biosensors, a modern cyclist can seem like a Formula 1 car in the vast array of data available during and after a ride. Here’s a closer look the current state of wearable sensor tech, and some considerations in deciding what might help you in health, training, and racing.
Today, I’ll go through some major concepts when it comes to evaluating a wearable biosensor. This article is NOT a review of any individual sensor.
I also highly recommend the two articles I’ve placed in the References section if you want a deeper dive into the emerging wearables technology and their potential in sport science (Ye et al. 2020; Shei et al. 2022).
Accurate and/or Reliable?
One of the biggest concepts to understand about a sensor, whether it is a bathroom scale or a wearable sensor, is accuracy/validity and reliability/precision. These terms are often incorrectly used interchangeably or wrongly defined, so let’s clear this up with an example of you wanting an altimeter to measure the true elevation at the top of Mount Everest (8,848 m):
Accuracy/validity: Does the sensor give you the true value that you’re trying to measure? In the Everest example, does it actually give you the value of 8,848 m at the summit or does it give you 8,200 m?
Reliability/precision: If you take multiple readings with the same sensor, does it give you the same value? For the Everest example, a sensor that gives you 8,200 m each time is NOT accurate/valid, but it IS reliable/precise.
What do you actually want in a sensor? Well, of course the ideal is that it is both accurate AND reliable. But in some contexts, the goal or optimum may vary. For example, if it’s a heart rate monitor, I’d probably place higher value on accuracy because heart rate by itself can vary greatly day to day (stress, sleep, caffeine, etc.) and I actually want to know what my heart is doing.
In contrast, if I was looking at a bathroom scale or a power meter, I’d probably place higher value on reliability. In those cases, I’m only comparing myself to myself and I’m more interested in tracking variations over time. So I’m more caring about how much weight I’ve gained over a month (e.g., 3 kg) than I care that I weight 65 kg. Similarly, if I am training with power, I may care more that my threshold power increased by 20 W over a training cycle even if it is reading 220 W rather than the actual accurate value of 250 W.
I have previously highlighted the issue of accuracy and reliability in a video article about the CORE body temperature monitor.
Direct vs Derived Value?
Another important concept to understand about biometrics is whether a sensor is directly measuring a value, or if it is taking a directly measured value and then putting it through a complex algorithm to give you a value.
The most common example is probably heart rate. We know that heart rate measured via a chest strap is very accurate and reliable, and that is an example of a direct measure. That is, when you see your heart rate on your bike computer, it is exactly that.
However, one challenge with many new biosensors is that they take that original measure and then use software to try to mathematically predict many things. Using heart rate as a starting measure, this can include heart rate variability, oxygen saturation, recovery, VO2max, etc.
Why is this a problem? The further you get from the original measure, the more you are relying on algorithms and assumptions.
Assumptions are exactly that, they assume that you fall into the general middle of a population in response, whereas we know that individual responses can vary greatly. For example, you’ve all likely seen the 220 – age equation for your max heart rate. Besides the fact that there is no scientific basis for this equation, this leads to the charts at gyms suggesting exercise zones. However, my directly measured maximal heart rate has always been much lower than predicted by this equation by 20-30 bpm, so any predictions for me based on this assumption would be wildly incorrect.
Another problem with algorithms is that they’re by and large black boxes. Each sensor and company has their proprietary algorithms, such that we don’t even know what measure or measures they may be using to derive a value, let alone the equations themselves. That makes it challenging to trust any derived biometric value in my opinion like a recovery score or VO2max.
Who’s the Target Population?
A final related issue to assumptions and algorithms is WHO you’re being compared to in that large database. Keep in mind that most of wearable biosensor development is driven from a health care and clinical imperative, not from an athletic perspective. This can lead to a case where a sensor is first developed for a relatively passive and sedentary population, and then laterally “shoehorned” into an athletic application.
The challenge with this can be that the sensor may be very accurate and reliable in relatively sedentary use or with mild exercise, but that may not always be directly extrapolated to it being as accurate or reliable with heavy exercise, movement, and sweating.
Another potential challenge is that the large population database that is being used for comparison is from a largely sedentary population with potential clinical issues, versus a healthier and highly active population. So it may be a case of comparing apples and oranges.
Summary
Again, as a sport scientist, there is a huge potential for wearable tech, and this article is not meant to disparage the field as a whole or any particular unit. What I’ve hopefully provided instead are three key considerations for you to think about when deciding whether to jump in and buy/adopt one of the many new wearable biosensors coming out onto the market, each with bold claims for how they will advance your training.
Have fun and ride fast!
References
Shei R-J, Holder IG, Oumsang AS, et al (2022) Wearable activity trackers–advanced technology or advanced marketing? Eur J Appl Physiol 122:1975–1990. https://doi.org/10.1007/s00421-022-04951-1
Ye S, Feng S, Huang L, Bian S (2020) Recent Progress in Wearable Biosensors: From Healthcare Monitoring to Sports Analytics. Biosensors 10:205. https://doi.org/10.3390/bios10120205
Like PEZ? Why not subscribe to our weekly newsletter to receive updates and reminders on what’s cool in road cycling?
