Overdiagnosis Only a Matter of Time With ECG Watches

John Mandrola, MD


December 08, 2017

It would be easy to rejoice over recent news in mobile health.

First, the FDA cleared a new ECG sensor for the Apple Watch. The Kardia Band allows people to record a single-lead ECG from a watch. That is nifty. And it's an advance over other wrist devices (such as, Fitbit, Garmin, etc) that use photoplethysmography (PPG) and detect only pulse rate. The Kardia band—similar to the AliveCor—lets you see the actual ECG.

Perhaps the bigger news was that Apple announced its entry into cardiac research with the Apple Heart Study. In partnership with Stanford researchers, the study will collect Apple Watch ECG recordings from people who volunteer to share their data. The goal is to learn more about atrial fibrillation (AF) and potentially reduce the burden of stroke. Getting Apple interested in cardiology is big.

The thinking behind mobile heart-rhythm monitoring is simple: Stroke is a major health problem. AF associates with stroke. We have a treatment for stroke reduction (anticoagulants). Thus, it would be great to screen more people to detect AF.

The problem is that simple thinking rarely works in medicine—an observation that was made long before mobile tech. In 1968, two pioneers of public health, Drs Wilson and Jungner,[1] writing for the World Health Organization, noted that in theory, "screening is an admirable method of combating disease. . . . In practice, there are snags."

The two images below from patients with AliveCor devices seen in my clinic depict some of the snags. In both cases, AF was incorrectly classified, and the patients suffered anxiety until I read the patterns as sinus rhythm. Even if there were true AF, it would not have changed therapy. One patient was already taking an anticoagulant, and the other was a vigorous young person with no risk factors for stroke.

Figure 1. ECG reading from AliveCor that triggered a "possible atrial fibrillation" alert [For a larger image, click here]

Figure 2. Another example of an ECG reading from AliveCor that was labeled "possible atrial fibrillation" [For a larger image, click here]

The Upside of the Mobile ECG

Before I tell you about the challenges faced by Apple and Kardia and other mobile innovators, let's first state many of the upsides of this new technology. All these comments came from Twitter conversations:

Businessperson Marc Cuban, who has AF and an AliveCor device, said, "The quick ECG and ability to send it to a doc for quick response lets me have some indication of what's going on."

Dr Mintu Turakhia (Stanford University) a researcher on the Apple Heart Study, wrote that "it's important to patients to be able to see if they have AF (ie, see their own data). That is one of the key drivers of innovation here."

A computer technologist from Romania wrote that his country has poor access to medical care. His father lives 3 hours from a hospital, and sharing ECG data from the device online with doctors allows him to get feedback faster.

And Dr Gopi Dandamudi (Indiana University) highlighted the obvious: "It has to start somewhere." Dandamudi believes that empowering consumers to be actively involved in their healthcare is a good thing in the long run.

These are all excellent points. What's more, I have used the AliveCor device in patients to confirm AF diagnoses, correlate symptoms with a specific rhythm, and monitor the QRS and QT intervals during use of antiarrhythmic drugs.

Then there are the future possibilities: as companies like Apple and Kardia collect millions of data points from watches, it's feasible that these data will feed increasingly smart artificial intelligence, perhaps combined with gene risk scores, and, who knows, in the future, people may receive text warnings minutes or hours before they have a heart attack or stroke.

In the meantime, I am afraid that this technology will lead to mostly hype, angst, overdiagnosis, medicalization of normal life, and harms from overtreatment.

The Challenges of New ECG Technology

The first challenge for this technology is that the causal link between AF episodes and stroke is less certain than you may think.

As Hooman Kamel and colleagues wrote in the journal Stroke, episodes of AF fail to fulfill many of the criteria outlined by the epidemiologist Bradford Hill to determine if one factor (AF) causes another (stroke).[2,3] First, AF clearly fails in the temporality criteria—a substudy of TRENDS observed that the majority of the 40 patients who experienced stroke or systemic embolism had zero AT/AF episodes within the 30 days prior to their event.[4] Second, AF fails in the specificity criteria because many strokes in patients with AF come from either small or large vessel atherosclerosis, not cardiac emboli.[5,6] The third criteria in which AF fails to satisfy causality is the biological gradient (or dose-response curve). There should be greater stroke risk from permanent AF than paroxysmal AF. Most likely there is not. Finally, if AF caused stroke, Hill's accordance with experimental evidence criteria would hold that rhythm-control medicines that reduce AF should also reduce stroke risk. They clearly do not.[7]

These observations cast doubt on the value of AF episodes as a surrogate for stroke risk. And if that is the case, what is the point of searching for it—especially in those without risk factors for stroke?

Another challenge for mobile monitoring of AF is that we don't know whether anticoagulant therapy will benefit patients with asymptomatic shorter-lived AF episodes (so-called subclinical AF). Remember, anticoagulants have been shown effective only in patients with documented "clinical" AF.

We know from observational studies that short-lived episodes of AF in patients with pacemakers associate with a higher risk of stroke.[8,9] A post hoc analysis of ASSERT suggested that this increased risk came almost exclusively from episodes lasting more than 24 hours.[10] We also know more intense monitoring in patients with stroke of undetermined origin leads to more detected AF.[11,12].

But none of these studies tested therapy. It will likely be a lot harder to show benefit for anticoagulants in patients with subclinical AF than in those with longer-lasting symptomatic AF. Consider that in the original TRENDS study, an older cohort (with pacemakers), the annualized stroke rate was a mere 1.2%.[9] Similarly, Quinn and colleagues reported that stroke rates in AF cohorts are often lower than what is presumed.[13]

Baseline stroke rates are crucial because anticoagulants come with potential harms. The challenge of the ongoing ARTESiA trial, which compares apixaban vs aspirin in patients with subclinical AF and risk factors, will be that it's hard to show a net benefit when stroke events occur at low rates.

The scariest part of enhanced AF detection is overdiagnosis and overtreatment. At least in the US, the upcoming watch-driven explosion of AF diagnoses will happen in a fee-for-service environment that pays doctors and hospitals to test and treat.

Downstream testing of patients with AF is no small thing. A couple of examples: an echocardiogram done for AF shows a floppy septum. This leads to a transesophageal echocardiography (TEE), and sometimes more. Palpitations from AF often cause chest pain. This leads to a nuclear stress test, and you know what may come next: a coronary angiogram and possibly a stent.

The lack of overall mortality reduction from cancer screening[14,15] should teach cardiologists a lesson. Finding abnormalities is not always a good thing. Every time a person with a watch-driven encounter with the medical establishment suffers a complication from downstream testing or treatment, this acts as a drag on the net benefits for screening. Turning people into patients should be done with great caution.


The goal of stroke prevention with mobile personal ECGs is reasonable. Giving patients control of their data is, too. And no doubt, future doctors will better understand the normal range of ambient arrhythmias. An AF episode that today begins an overtesting cascade may someday be considered normal.

But once again, I worry that mobile ECGs will distract both patients and doctors. We already have effective ways to prevent stroke. In a study of more than 11,000 high-risk patients, Swedish authors observed a 72% lower risk of stroke in those who ate well, moved, maintained a normal body weight, drank low to moderate amounts of alcohol, and did not smoke.[16] Two recent large observational studies confirmed the cardiovascular benefits of physical activity.[17,18]

I know this may seem like nonspecific, unsexy advice, but lifestyle modification is safe, and everyone agrees it delivers huge health benefits.

The mobile ECG is a nice tool for specific tasks. It's fun. But it's fantasy to think it will improve health outcomes.


Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.