Digital physiological biomarkers predict within-person symptom changes in complex chronic illness, 2026, Aitken et al

[Felis Catus]

But that's not what's found here. There are variations, fluctuations, that don't seem common, but we don't see things like sub-20s HRV as we should expect with severe deconditioning. Either HRV isn't actually a useful measure of fitness, or there's something odd about us that should be a clue about how we are different.

I actually know a couple of people with HRV in sub-20s which goes down to single digits. That's based on their wearables' algorithms. Those people don't have ME/CFS, they have other health issues but they work, socialise and have hobbies.

My healthy caregiver has had lower HRV than me since we've had our wearables.

The biggest differences by far are when I am sick, then my HRV can drop in the 30s and my RR shoot up to 100+. Which then that clearly has nothing to do with fitness.

Yes, if I'm in any doubt whether I'm in a crash or have a new infection, that's a telltale sign - unusually high drop in HRV and RHR goes above 100. It happens before the full symptoms develop.

 

[Felis Catus]

What HRV metric is being used here, I can't see it skimming through the paper? Is it some wearable derived score where we don't even know what it's supposed to be measuring or something more standard like SDNN?

Section "How do we calculate your HRV?"

I agree with @Jonathan Edwards . I think we need someone sensible to start these things back from the beginning. I don't see that happening in the "wearable field" and with app providers. You need someone that genuinely wants to understand something not someone who believes they already do and want to sell you that idea. Gather genuine data over many months, don't feed the output to the patients and then let statisticians analyse the data carefully to see if anything pops out.

Yeah. I think you'd also need an activity diary to match the numbers, which would be the hard part of data collection.

 

[EndME]

Section "How do we calculate your HRV?"

Thanks! The study is using a Polar device which I believe provides such data directly so supposedly they are using the RMSSD data provided by Polar (respectively a transformation of it).

Yeah. I think you'd also need an activity diary to match the numbers, which would be the hard part of data collection.

I'm not sure. You'd need a type of symptom diary, but the data should be the activity diary initself (step count etc), of course you wouldn't capture all activities like that (like cognitive activities) but that would sort of be the point in my mind.

 

[OrganicChilli]

My daily HRV rate doesn't tell me anything beyond how well I slept, but I don't need a number for that. However, it dropped by 10 units when I went from moderate to mod-severe and hasn't recovered. I've plotted it over several months and you can clearly see a drop. I don't have symptoms at baseline so I don't feel any worse, but I can do significantly less until symptoms appear.

 

[Felis Catus]

I'm not sure. You'd need a type of symptom diary, but the data should be the activity diary initself (step count etc), of course you wouldn't capture all activities like that (like cognitive activities) but that would sort of be the point in my mind.

Would you be able to tell someone's body position from that? For example, I can't sit for long but there is nothing abnormal in my HR or BP while sitting.

I think missing cognitive activities is a big problem because the more severe you get, the more you're forced into cognitive activities within the capacity you have, simply because anything physical would be even more taxing. The amount and intensity of cognitive activity also varies by person. Perhaps we have different ideas in mind.

 

[MrMagoo]

I think (IIRC) the daily ready score thing in Visible looks at the past few days/week data after your morning measurement.
The reasoning being that because we have PEM it’s useful to know if your system is starting to struggle/go out of sorts.
People seem to frequently misunderstand or want to argue with the readiness score, but my understanding is it’s kind of a “temperature check” of how your HRV stats are trending, so it’s totally normal that you feel ok but your score is “bad” because you might be going towards PEM and conversely you may have a “good” score but feel awful because you’ve been stable (by being stuck in bed for a week).

It’s not supposed to govern your life, just give you a heads up of how your stats are.

Seems like everyone these days wants to be able to “biohack” a digital “mystic Meg” horoscope, as if people are simple computers who only act in defined ways which can be foreseen. Le sigh.

 

[EndME]

Would you be able to tell someone's body position from that? For example, I can't sit for long but there is nothing abnormal in my HR or BP while sitting.

I think you'd track positional differences as well, maybe by having one device on the arm and one on the foot or something of that sort. That shouldn't be too complicated.

I think missing cognitive activities is a big problem because the more severe you get, the more you're forced into cognitive activities within the capacity you have, simply because anything physical would be even more taxing. The amount and intensity of cognitive activity also varies by person. Perhaps we have different ideas in mind.

I was thinking that the goal would be to establish relationships between physical activity, other metrics (temperature, heart rate,...) and ME/CFS symptoms and illness progression to see what relationships, if any, exist. If there's a signal, great, if not then it's maybe a sign to move away from an activity centric model of ME/CFS, which is very much the current model. Thoughts, smells, sounds etc don't anways really fit into an activity model, at least it certainly does not fit the activity model that is being studied in ME/CFS where "activity" relates to "active".

 

[Felis Catus]

I was thinking that the goal would be to establish relationships between physical activity, other metrics (temperature, heart rate,...) and ME/CFS symptoms and illness progression to see what relationships, if any, exist. If there's a signal, great, if not then it's maybe a sign to move away from an activity centric model of ME/CFS, which is very much the current model. Thoughts, smells, sounds etc don't anways really fit into an activity model, at least it certainly does not fit the activity model that is being studied in ME/CFS where "activity" relates to "active".

I crash from too much cognitive exertion which is reflected in my HR and HRV the following days. My HR is normal during those activities and it doesn't change upon commencing or stopping them.

Some people deteriorate while trying to work from bed. Perhaps working had no effect on their illness but I'm not sure how many moderate and severe patients would be able to work from their bed without pushing through. If we were all able to carry on with cognitive activities as before, I wouldn't be mentioning any of this.

So, I'm not sure you could collect data long term (because a few weeks or months wouldn't tell us much) that wouldn't be "contaminated" with cognitive activity/exertion and/or its consequences.

 

[EndME]

My HR is normal during those activities and it doesn't change upon commencing or stopping them.

Yes. I think one should expect so. Smelling, thinking, hearing does not effect HR much.

So, I'm not sure you could collect data long term (because a few weeks or months wouldn't tell us much) that wouldn't be "contaminated" with cognitive activity/exertion and/or its consequences.

But in my mental model that wouldn't necessarily be the goal. But maybe I'm also not thinking quite right. You cannot "uncontaminate" things when certain brain processes, which we have no way of measuring in the first place, are contaminating the data. That's why you'd look for other relationships first. You could also collect additional data, but I think at some point it becomes unfeasible for people to fill that data out on a very regular basis, especially when you want to include severe people.

 

[ME/CFS Science Blog]

The paper writes:

When only prior-day symptom predictors (i.e., lag-1 symptom reports) were included,models achieved AUC values were 0.78 for crash, .73 for fatigue, and 0.83 for brain fog. When morning biometric features were added to the prior-day symptom model, AUCs increased modestly across all outcomes, reaching 0.81 for crash, 0.74 for fatigue, and 0.85 for brain fog.

So the conclusion seems to be that prior-day symptoms are useful predictors but that biometric features add almost no additional information to that.