Apple Watch can detect an early sign of heart disease

MacWorld: Developers of the Apple Watch app Cardiogram worked with researchers leading the University of California San Francisco’s Health eHeart study to develop a ResearchKit-based study of their own called mRhythm. On Thursday, Cardiogram and UCSF’s cardiology division are presenting the results of that 14-month study, which collected more than 100 million heart rate data points from more than 6,000 Apple Watch users. Cardiogram developed a machine learning-powered algorithm that can detect atrial fibrillation, which is often asymptomatic.

Cardiogram’s algorithm was tested against an in-hospital test called cardioversion. Patients experiencing atrial fibrillation, which affects one in four people in their lifetime and causes 25 percent of all strokes, wore an Apple Watch while undergoing cardioversion to compare outcomes. Both segments, the cardioversion test and the Apple Watch’s heart rate data, were blinded against whether the patients’ heart rates were normal or abnormal, then sent to Cardiogram’s algorithm. The results: the Apple Watch data detected atrial fibrillation 97 percent of the time.

Apple has been communicating privately with the FDA for years about medical devices and so far the FDA has taken a light touch to Apple but these issues are coming to a head. As with the regulation of DNA tests, the regulation of these devices is going to raise important free speech issues. It’s one thing to ensure that the devices do what they say they do at reasonable accuracy (measure heart rate, identify genes etc.) but regulating what advice may be given on the basis of such readings is problematic. Can the FDA regulate a website that says go see your doctor if your heart rate monitor exhibits these particular readings? Why is an app that tells you the same thing any different?

Hat tip: Samir Varma.


Can the FDA regulate a website that says go see your doctor if your heart rate monitor exhibits these particular readings? Why is an app that tells you the same thing any different?

I bet the FDA can try. Wasn't there a case recently where State authorities went after a woman because she offered weight loss advice on line?

Not that they need to. They just need to make sure Apple has no protection from law suits. Even limited protection. If they claim, even implicitly, that their watch can detect heart attacks, they just need to step back and let the 3% sue.

Repeat 100 times out loud: Overdiagnosis, overdiagnosis, ....

This is a significant concern.

Right. Don't trust an article that confuses "detects 97% of the time" with "97% accuracy."

No mention of the rate of false positives.

I can detect 250% of heart problems by simply looking a person in the eye.

I want my Apple Watch to tell me when I'm drunk.

I want Jeff's Apple Watch to tell my insurance company when Jeff is drunk if he hits my car. I want Jeff's Apple Watch to tell all the medical insurance companies how often he gets drunk so he can pay for his increased health risk.

Not in our lifetimes, but who knows what humans 200 years from now will think reasonable?

Mark, when I hit your car, it'll be on purpose, so it won't matter whether I'm drunk or stone sober!

Drunk-detecting watch already exists:

If we were a rational people with a rational legal regime, health monitors would be great.

In the real world, besides overdiagnoses, we will get lawsuits where someone has a heart attack, and then lawyers go pore over the entire history of the device and say "hey, you could have detected a heart flutter here and warned the patient, pay up!"

I bet that in Singapore health monitors will be a boon for people's health and their health system.

I had the same thought, and in general when medical testing data reports only sensitivity and not specificity, I default to overdiagnosis as well, otherwise why not report both. In this instance, it seems like run of the mill journalistic incompentance:

'Cardiogram trained a deep neural network on the Apple Watch’s heart rate readings and was able to obtain an AUC of 0.97, enabling us to detect atrial fibrillation with 98.04% sensitivity and 90.2% specificity.'

Of course proper caveats, this is a pilot, non-peer reviewed, and unpublished. But that specificity looks decent.

The FDA isn't cracking down on people who do a similar thing for diabetes with DIY hacks (or who share the knowledge online):

So perhaps the cardiogram app folks simply need to make the app's data available and then independently publish the criteria for making a diagnosis. Hackers will close the circle.

How do physicians return the "favor"? By reading medical economics and being an "expert" on the subject. The only thing worse than a narcissistic physician is the know-it-all physician. Apple wants to turn all of us into Woody Allen hypochondriacs.

good for business! ....aaaaand GDP increase ;)

Let's hypothesize a problem and react to the hypothetical.

Doesn't the real world work the other way: Apply to the FDA and see what happens.

You did notice in the post that it said the FDA had been applying a light touch to these matters.

On other news today, President Trump tweeted....

This is just marketing hype from Apple, which Alex chooses to trumpet for some reason..

Nothing magic or new about this AppleWatch App -- it measures your pulse rate. Wow !

There are dozens of retail personal heart rate monitors that do the same thing.

This Apple Watch App can only detect atrial fibrillation (abnormal heart rate) if the heart rate is "consistently" abnormal. That's a huge caveat to its overall usefulness. Most people with a "consistently" abnormal pulse rate will notice it without fancy electronic gadgets.

Nothing magic or new about this iPhone App -- it takes photos. Wow !

There are dozens of retail point and shoot cameras that do the same thing.

'which is often asymptomatic'

Not anymore, seemingly.

'the Apple Watch data detected atrial fibrillation 97 percent of the time'

Which is not that impressive, actually, though certainly useful for those who have health insurance. Unless, of course, such data could be used under the latest Republican plan to demonstrate a pre-existing condition.

'Can the FDA regulate a website that says go see your doctor if your heart rate monitor exhibits these particular readings'

Depends on the claims made on the web site, mainly.

'Why is an app that tells you the same thing any different?'

Depends on the claims, mainly. There is no absolute right to free speech in the U.S., and never has been. The FDA, like many other regulatory agencies, have the duty to enforce the law. Somebody claiming that radium cures any number of ailments, and is selling radium at the same time, is not having their free speech rights restricted if they are told not to make such claims in connection with something they are selling.

‘which is often asymptomatic’


"Consistent atrial fibrillation" (reduced heart function/output) always has symptoms from mild to severe.
AppleWatch only detects the consistent type of atrial fibrillation

One issue is whether the Apple Watch is marketed as being useful for making health decisions. If it is, then Apple will have to submit to FDA review. If it's more of a "fun thing" then Apple is (mostly) off the hook.

Device classification depends on the intended use of the device and also upon indications for use.

"Cardiogram’s algorithm was tested against an in-hospital test called cardioversion. "

But this doesn't even make sense: cardioversion is a treatment, not a test. And because (like most medical procedures) it has risks as well as potential benefits, it wouldn't be performed without a prior diagnostic finding.

Cardioversion is having the paddles attached to your body to shock the heart back into sinus rhythm, or accomplishing the same feat with drugs. I've had the paddles twice. It's disheartening when the journalist so clearly knows nothing about the article's topic.

What if Hillary Clinton had had Le Pen's smoker's voice instead of her own?

As an aside, the Timex Ironman GPS is due out this summer for $99

It doesn't even have heart rate but with long GPS mode battery life, run and bike modes, a win.

Only cucks run.

A company called Karia has a product called AliveCor that is essentially an ECG small enough to fit on the back of your iPhone. It has two small paddles that you touch, and it uses AI to analyze your hearth rhythm for AFib. They got FDA approval.

I would guess that fitbit could do this as well. Fitbit recently updated their HR monitoring to analyze your sleep. Instead of just saying when you were asleep or "restless", it now tells you if you were in light, deep, or REM sleep, based on your heart rate.

All these products should be analyzed against "false positives", but should be allowed to market their products with minimal FDA oversight. The potential is just too great.

I don't doubt that cheap and smart medical electronics will shake up regulation in the future. I just don't think is is as simple as "Apple did it!" or "cancel the FDA."

It will be a messy transition though. I don't think doctors are up going to appreciate calls saying "I got a new app, and it says ..."

Alex is right to bring this up. Obama's FDA smashed 23andme for no apparent reason other than to display its powers to the industry at large. They obliterated a very promising sector, one that could have moved the needle of our understanding, due to ideology. We must be eternally viligant of this kind of overreach.

Hopefully, such actions are less likely during the Trump administration. Sadly, the Trump administration has, like every republican administration since the 1970s, been hounded by the media and is having a very hard time making and inroad into into dismantling the regulatory apparatus that can be used for such evil.

Cut the histrionics. Drop the "Obama" from your paranoia. The FDA has always monitored the medical diagnostics industry. I know, because our diagnostic product underwent a full technical audit, including in-house inspection of source code, in the 1980s. The way it always was.

23andme just came too close to saying that they were doing diagnostics, and not "giving you a cancer risk score 'for fun'."

I'm sorry if an accurate recounting of the abuses of power during the Obama administration hurts your sensitivities. I hope I did not trigger you.

You are a crazy bird.


Your eternal vigilance kinda sucks. 23andMe is selling their services again since April 4


"The Cooper Committee--chaired by Theodore Cooper, M.D., then director of the National Heart and Lung Institute--was organized in 1970 specifically to study medical devices as part of President Richard M. Nixon's endorsement of medical device legislation. The committee recommended that any new legislation be specifically targeted to the device i4ndustry, because devices presented entirely different issues from drugs. It also suggested that different classifications for medical devices be created, which would tailor the regulatory controls to the risks involved."

Sensitivity and specificity numbers please.

Try I am not convinced about the results in terms of false postives.

I am really thankful to the holder of this web page who has shared this
great paragraph at here.

Comments for this post are closed