What the Apple Watch ECG can and cannot detect, and the clearance it rests on

The Apple Watch takes a single-lead electrocardiogram, and the phrase carries the whole story of what it can do. A clinical ECG uses twelve leads to view the heart's electrical activity from many angles; the watch records one channel between a finger on the digital crown and the back of the case. That one channel is enough to tell a normal rhythm from atrial fibrillation, and not much else. The FDA cleared it on exactly that basis.

What the clearance actually says

The ECG app reached the US market in 2018 through the FDA's De Novo pathway, which created a new device category — over-the-counter electrocardiograph software — rather than matching it to an existing product [s1]. The clearance is narrow by design. The app "determines the presence of atrial fibrillation (AFib) or sinus rhythm on a classifiable waveform," and the agency's decision document states plainly that it "is not intended to provide a diagnosis" [s1].

The performance goals set for the pivotal study were a sensitivity of 90% and a specificity of 92%, measured against a physician-adjudicated 12-lead ECG [s1]. In the clinical study the false-positive rate was 0.4% and the false-negative rate was 1.7% [s1] — strong numbers, but for a single task: separating atrial fibrillation from normal sinus rhythm in recordings clean enough to read.

The limits are in the label

The same document is unusually direct about what the feature does not do. It "is not intended to detect any other type of arrhythmia," and, in the labeling the agency quotes, "It cannot detect heart attacks" [s1]. A person having a myocardial infarction can take a reassuring watch ECG showing sinus rhythm; the recording says nothing about a blocked coronary artery. The agency's guidance to users is to call emergency services for chest pain regardless of what the watch shows [s1].

Two further caveats matter for anyone reading their own tracing. First, the reassurance is not absolute: the labeling warns that the absence of a notification does not guarantee the absence of an arrhythmia [s1]. Second, a meaningful share of recordings are unusable — the agency noted that approximately one reading in eight came back inconclusive [s1]. An inconclusive result is not a normal result; it is no result.

What independent testing has found

Independent work has broadly supported the watch on its narrow task. A study at two cardiac centres compared 469 paired recordings from the Apple Watch and a standard 12-lead ECG and found no significant differences across the ECG characteristics it examined, with a strong correlation in heart-rate measurement [s2]. The watch's automated interpretation detected atrial fibrillation with a sensitivity of 99.54%, and manual reading of the same tracings reached 100% [s2]. The most common abnormalities the watch flagged were atrial fibrillation, in 37 of the recordings (7.9%), and first-degree atrioventricular block, in 32 (6.8%) [s2].

That last detail is worth dwelling on. The watch's algorithm is built to output only "AFib" or "sinus rhythm," but a trained reader looking at the same single-lead strip can sometimes see more — a prolonged PR interval, ectopic beats — which is why a clinician may act on a watch tracing the app itself scored as sinus. The value of the single lead depends heavily on who is reading it.

How it fits the wider picture

The ECG app is the confirmatory half of the watch's cardiac pairing. The passive irregular-rhythm notification watches the optical pulse sensor in the background and prompts the wearer to take an ECG; the ECG app then records a strip a clinician can review. Neither replaces a diagnosis, and the same regulatory logic now governs the watch's newer notification features: the device may measure or classify a signal, but the clinical meaning is left to a doctor.

For a wearer, the practical reading is modest and useful. A watch ECG can capture a rhythm at the moment symptoms occur — often the hardest thing to catch in a clinic — and hand a real tracing to a physician. It cannot rule out a heart attack, cannot see most arrhythmias, and fails to classify a substantial minority of the recordings people take. Used as a way to document a suspicious rhythm rather than to settle a question, it does the one job it was cleared for.