Why smartwatch-measured blood pressure still isn’t ‘ready for primetime’ (2021)(theverge.com)
theverge.com
Why smartwatch-measured blood pressure still isn’t ‘ready for primetime’ (2021)
https://www.theverge.com/2021/9/16/22677381/smartwatch-blood-pressure-samsung-fitbit-apple
36 comments
Not knowing that you have a heart condition tends to be the most lethal about them. Once doctors know, they can fix a lot but not before.
Usually the first doctors do (beyond asking questions) is start measuring things. I've had a checkout a few years ago. That included the indignity of having to walk around with a thing that measured my blood pressure for a full day. It actually came in a bag that I had to carry around. All fine in the end. At least as fine as 24 hours of monitoring can confirm. But it drove the point home that those numbers were important to figure out if I had any issues.
It also rubbed the notion in that by the numbers I was on a statistical path to a likely early death and that I had the opportunity to change things. I lost some weight by cutting down on junk food and I exercise a bit more. I'm not sure it's enough but I am undeniably in better shape these days.
But it also made two things very obvious to me:
1) that stuff has got to be very obsolete. There has got to be a better way to measure people's vital signs than carrying around a thing that pumps itself up every minute or so just to measure heart pressure. And what about everything else?
2) Why don't we permanently monitor people's telemetry? I do that for servers I maintain. I set alerts when stuff is about to go bad. Why can't I monitor my body in the same way? More importantly, why is my body not monitored until it is too late? Why can't I give my doctor access to all that when I visit once a year for my checkup. Hell, why isn't he calling me when there's an obvious issue. The status quo is waiting for bad stuff to happen until you even get to see a doctor.
Of course the raw numbers are not that useful unless you know how to read them but they are gold to the people that can. It's one thing to go to the doctor with some vague complaints; it's another thing to go to a doctor with a year's worth of data that make it very obvious what is going on.
Usually the first doctors do (beyond asking questions) is start measuring things. I've had a checkout a few years ago. That included the indignity of having to walk around with a thing that measured my blood pressure for a full day. It actually came in a bag that I had to carry around. All fine in the end. At least as fine as 24 hours of monitoring can confirm. But it drove the point home that those numbers were important to figure out if I had any issues.
It also rubbed the notion in that by the numbers I was on a statistical path to a likely early death and that I had the opportunity to change things. I lost some weight by cutting down on junk food and I exercise a bit more. I'm not sure it's enough but I am undeniably in better shape these days.
But it also made two things very obvious to me:
1) that stuff has got to be very obsolete. There has got to be a better way to measure people's vital signs than carrying around a thing that pumps itself up every minute or so just to measure heart pressure. And what about everything else?
2) Why don't we permanently monitor people's telemetry? I do that for servers I maintain. I set alerts when stuff is about to go bad. Why can't I monitor my body in the same way? More importantly, why is my body not monitored until it is too late? Why can't I give my doctor access to all that when I visit once a year for my checkup. Hell, why isn't he calling me when there's an obvious issue. The status quo is waiting for bad stuff to happen until you even get to see a doctor.
Of course the raw numbers are not that useful unless you know how to read them but they are gold to the people that can. It's one thing to go to the doctor with some vague complaints; it's another thing to go to a doctor with a year's worth of data that make it very obvious what is going on.
You what I would practically kill for? Continuous monitoring of blood cholesterol. Then I could do experiments with diet, exercise, etc. Unfortunately I think this has the Theranos "drop-of-blood" problem and you just need a giant blood draw every time.
Years ago I had my cholesterol measured at a work sponsored health event. It was just a finger stick that they put onto a slide and then into a machine even smaller than the Theranos. Probably it's more feasible since it's only 1 test. I don't know what the possibility of getting one of those machines for personal use it though. I actually asked to keep my blood slide out of curiosity, but apparently that wasn't allowed.
I have hypertension, and its very difficult to measure my BP and watch it. Why? Because I absolutely hate the cuffs. I hate the pressure it cause, and it gives me anxiety. To be able to measure my BP with out a cuff? That'd be a game changer for me. And also, having all of this data on the watch has greatly reduced my anxiety, because I check some things and have an idea that I'm actually fine.
So please don't assume your situation to others.
So please don't assume your situation to others.
bowsamic(2)
A bit over a year ago I took the time for little experiment and made practically simultaneous heart rate measurements using a fairly recent FitBit model, another smart watch (forgot which one), the heart rate reading from a medical arm cuff blood pressure monitor and manual counting over a minute. In a rested state, repeated five times.
The arm cuff and manual measurements showed very limited variablity over different measurements and were in very good agreement with each other (like +/- 4 beats per minute). The smart watches were all over the place, showing deviations of up to twenty beats per minute from the former two methods / each other, varying by over 20 beats per minute over consecutive measurements, and not even showing signs of convergence over repeated measurements.
My conclusion (at least for my skin type, which does affect accuracy for watch-based measurements): smartwatch-based heart rate measurements are probably best used as slightly biased random number generators. If tracking blood pressure is much harder than the heart rate, as they say in the article, I don't even want to know.
The arm cuff and manual measurements showed very limited variablity over different measurements and were in very good agreement with each other (like +/- 4 beats per minute). The smart watches were all over the place, showing deviations of up to twenty beats per minute from the former two methods / each other, varying by over 20 beats per minute over consecutive measurements, and not even showing signs of convergence over repeated measurements.
My conclusion (at least for my skin type, which does affect accuracy for watch-based measurements): smartwatch-based heart rate measurements are probably best used as slightly biased random number generators. If tracking blood pressure is much harder than the heart rate, as they say in the article, I don't even want to know.
When worn correctly, modern smart watch optical heart rate sensors are actually reasonably accurate for most people. Read the DC Rainmaker reviews where he compares them against highly accurate chest straps. The optical sensors will often lag a bit or get off track for a brief period but overall not too bad.
https://www.dcrainmaker.com/product-reviews/garmin
https://www.dcrainmaker.com/product-reviews/garmin
I'm the CTO at a relatively well established medical device company (EEG/ECG). Before I arrived at my current company I spent a year of my own time tinkering in the BP space. The article, while relatively, light on details gives a pretty good sense of the problem with trying to replicate the traditional BP measurements using the optical methods used by smartwatches.
Smartwatches and similar devices use photoplethysmography (PPG) to measure blood oxygen saturation. The heme units in blood have different reflectance/absorbance spectra depending on their level of oxygen saturation. The only difference between the fingertip pulse oxygen sensors used in clinical settings and those used in smartphones is that the clinical ones use transmission, whereas the watches use reflection. While the methods are different, they arrive at very similar numbers (+/- skin color & racial biases).
Unlike pulse oximeters of yore, smart watches can take many thousands of readings per second. And if you offset your photodiodes by a known distance, you can calculate pulse transit time, and a variety of other measures beyond just your spo2.
The challenge has been in translating these measures back into systolic and diastolic blood pressure. To be accepted by the FDA you need to be able to show that your device is +/- 3mmHg. That's just not doable right now for the reasons mentioned in the article.
That said, the measures that you can get from high-speed PPG are probably not useless. But these measures aren't backed up ~200 years of medical practice. It is a very long road to introduce new measures into the medical system. Not only do you need to get FDA clearance to sell your device, but you need to navigate the reimbursement system to find someone who will pay for your device. Doing this in a 5 year time span would be unheard of. While cardiologists would like a better system than cuff based BP measurement, the upside of a better system is limited by the current medical system. And there's a significant cost and downside risk to anyone pushing for something different.
One company that I spoke to and is still working on the problem comes with an impressive biomedical pedigree, and they're starting in the non-diagnostic "wellness" space.
Another company is using mm-wave radar to measure pulse waves in a different way, but it appears that they've pivoted from being directly challenging the BP status quo into now selling their hardware sensors to more well established medical device companies.
Long story short, I'd be impressed to see blood pressure cuffs retired in my lifetime.
Smartwatches and similar devices use photoplethysmography (PPG) to measure blood oxygen saturation. The heme units in blood have different reflectance/absorbance spectra depending on their level of oxygen saturation. The only difference between the fingertip pulse oxygen sensors used in clinical settings and those used in smartphones is that the clinical ones use transmission, whereas the watches use reflection. While the methods are different, they arrive at very similar numbers (+/- skin color & racial biases).
Unlike pulse oximeters of yore, smart watches can take many thousands of readings per second. And if you offset your photodiodes by a known distance, you can calculate pulse transit time, and a variety of other measures beyond just your spo2.
The challenge has been in translating these measures back into systolic and diastolic blood pressure. To be accepted by the FDA you need to be able to show that your device is +/- 3mmHg. That's just not doable right now for the reasons mentioned in the article.
That said, the measures that you can get from high-speed PPG are probably not useless. But these measures aren't backed up ~200 years of medical practice. It is a very long road to introduce new measures into the medical system. Not only do you need to get FDA clearance to sell your device, but you need to navigate the reimbursement system to find someone who will pay for your device. Doing this in a 5 year time span would be unheard of. While cardiologists would like a better system than cuff based BP measurement, the upside of a better system is limited by the current medical system. And there's a significant cost and downside risk to anyone pushing for something different.
One company that I spoke to and is still working on the problem comes with an impressive biomedical pedigree, and they're starting in the non-diagnostic "wellness" space.
Another company is using mm-wave radar to measure pulse waves in a different way, but it appears that they've pivoted from being directly challenging the BP status quo into now selling their hardware sensors to more well established medical device companies.
Long story short, I'd be impressed to see blood pressure cuffs retired in my lifetime.
I have hypertension which has been treated for years with medication. My measured pressure is now within healthy limits. I have a cuff-based device bought from a UK chemist and I check it most days, at a similar time and after at least ten minutes sitting down (as per jawngee elsewhere on this thread). Even so my BP and pulse vary by a few percent. When I go into a hospital / surgery I get my BP checked properly and because the result is usually in the same ball park as the home device, I tend to trust that if the home device showed a sudden massive variation, then I should consider consulting a doctor. If smartphones are doing a variation check, as the article implies, then they could be useful. But there is no substitute for a proper check using a calibrated device if you are really worried.
One reason that I check my BP is that I have underlying issues with my heart's electrical systems: right bundle branch block [0]. This is visible on a professional ECG. I've had quiet a few of those following a ventricular fibrillation episode which involved an air ambulance ride to a cardiac ICU where I had a defibrillator implanted. These underlying conditions are not typically detected on the few (n == 1) smart devices I've tried that offer an ECG-lite check.
[0] https://en.wikipedia.org/wiki/Right_bundle_branch_block
One reason that I check my BP is that I have underlying issues with my heart's electrical systems: right bundle branch block [0]. This is visible on a professional ECG. I've had quiet a few of those following a ventricular fibrillation episode which involved an air ambulance ride to a cardiac ICU where I had a defibrillator implanted. These underlying conditions are not typically detected on the few (n == 1) smart devices I've tried that offer an ECG-lite check.
[0] https://en.wikipedia.org/wiki/Right_bundle_branch_block
Another promising tech for this is using ultrasonic waves to measure the diameter of the artery:
https://www.nature.com/articles/d41586-020-01807-6
https://www.nature.com/articles/d41586-020-01807-6
A lot of these smartwatch sensors have questionable accuracy. No doubt Samsung will rush out something that hardly works.
I don't get how the traditional cuff method actually works. It's not clear to me how having the thing on your arm tells you what the pressure is inside your blood vessels. Isn't your arm made of things that can stress, therefore the pressure is not the same on either side of it? Same as how how do you measure the pressure inside a tyre or other pressure vessel? The point of the vessel is to make the pressure not the same as outside.
To put it simply, it measures how much pressure is required to stop your pulse being detected. The higher your blood pressure, the more pressure required to stop the pulse in the arm
It's apprently not very accurate but I suppose good enough.
https://www.acc.org/latest-in-cardiology/articles/2017/07/24...
https://www.acc.org/latest-in-cardiology/articles/2017/07/24...
So I can actually feel my pulse in my upper arm presumably when the device is in it's measurement range so I've always presumed that it's sensing that with a pressure transducer and a simpleish algorithm, but I never thought to ask about how it actually works, and now I wonder if other people can feel them working.
Wont the stiffness and amount of tissue around the vessels change that ? And thus won't it be different from person to person not because the blood has a different pressure but because of the way we are measuring it.
It doesn't just stop the pulse from being detected, it stops the blood from passing into the limb where the cuff is applied.
That's going to mostly depend on how much pressure is behind the blood.
That's going to mostly depend on how much pressure is behind the blood.
I never knew that, thanks. So when the machine ramps up the pressure, then ramps up some more, then some more, you know you're in trouble?
I've always asked myself the question how this works, but actually never looked it up. Thank you for this explanation!
Brilliant and concise, thanks!
You measure the pressure you need to apply, to stop the blood going through the cuff.
It’s probably not possible to measure blood pressure perfectly. +- 5mmHg seems to be a completely normal error, probably even more. But it seems to be good enough.
It’s probably not possible to measure blood pressure perfectly. +- 5mmHg seems to be a completely normal error, probably even more. But it seems to be good enough.
The gold standard for BP measurement is direct measurement via an arterial catheter - an invasive and relatively risky procedure.
For a device to be approved as a Class 1 medical device for blood pressure measurement, the FDA wants +/- 3mmHg to an established BP measuring method.
For a device to be approved as a Class 1 medical device for blood pressure measurement, the FDA wants +/- 3mmHg to an established BP measuring method.
Wow, I didn’t know that.
I still think though, that in a common setup the +-3mm is impossible to reach. People don’t sit perfectly still, may flex their muscles during the measurement, etc
I still think though, that in a common setup the +-3mm is impossible to reach. People don’t sit perfectly still, may flex their muscles during the measurement, etc
It is a standard way to measure it. So, you might just say that it measures "blood pressure on the upper arm" and that it is taken as a measure. Like temperature with a thermometer.
Like, is the air pressure/temperature/contamination everywhere the same around my city? Well, obviously not, but we make do with what we have.
Like, is the air pressure/temperature/contamination everywhere the same around my city? Well, obviously not, but we make do with what we have.
Medical sensors are fascinating, the one the really blew my mind in simplicity is a light based blood oxygen meter.
There's a much better location to do continuous 24/7 blood pressure monitoring with a much smaller sensor that could eventually be made unobtrusive with development: the earlobe
Must be some kind of stigma for American market or unwilling investment into "not inventing anything new" attitude.
Must be some kind of stigma for American market or unwilling investment into "not inventing anything new" attitude.
"It needs calibration"
Having a heart condition is a very easy gateway to health anxiety. I had to stop wearing my apple watch, my wife keeps the pulse oximeter under lock and key, etc. because I became way too focused on the numbers while ignoring the context. It became a very bad feedback loop.
I don't want to monitor my BP 24/7.