Can a Smartwatch Detect Sleep Apnea?

A smartwatch or ring that measures blood oxygen rarely misses sleep apnea and often flags people who do not have it. A meta-analysis of 18 studies found 97% sensitivity and 63% specificity overall. In the consumer wearable subgroup of those 18 studies, sensitivity was 93% (95% CI 90 to 96) and specificity 63% (53 to 71). In plain terms, about four in ten people without apnea were wrongly flagged. The authors say the devices suit screening high-risk adults, with every positive confirmed by a sleep test.

How often does an oxygen wearable catch sleep apnea?

Almost always. A 2025 meta-analysis compared oximetry-based devices with polysomnography in adults. It included 18 studies. Overall sensitivity was 97% (95% CI 96 to 97).

Consumer wearables were the smaller subgroup. Their sensitivity was 93% (95% CI 90 to 96). Medical oximetry devices reached 97%.

Sensitivity is the share of people with apnea whom the device flags. A high value means it rarely misses the disorder. The authors say the devices show high sensitivity for ruling out apnea in high-risk adults.

Why are there so many false alarms?

Specificity was only 63%. That is the share of people without apnea whom a device correctly clears. About 37% of them were flagged anyway.

The same pattern held for consumer wearables, at 63% with a wide interval (53 to 71). The authors note this low specificity raises concern about unnecessary healthcare costs.

Their conclusion is that the devices may suit screening in high-risk people if every positive result is confirmed by a gold-standard test. A flagged night on a wrist or ring is a reason for a sleep test. It is not a diagnosis.

Have specific rings and watches been tested?

Two rings have, with different results. The Circul ring (Bodimetrics) was tested in 164 patients suspected of sleep-disordered breathing, mean age 44.8. Against polysomnography, for an apnea-hypopnea index of 5 or more, its best cutoff gave 87.8% sensitivity and 93.8% specificity. For an index of 30 or more it gave 85.7% and 92%. Those cutoffs were picked as the best performers within the same patients, which tends to flatter the result, and the Belun figures below come from children, so the two rings cannot be compared directly.

The Belun ring was tested in 75 children aged 5 to 18 with suspected apnea. Its index correlated moderately with the lab index (r = 0.63). The area under the curve was 0.66. A cutoff of 3 events per hour gave 74.1% sensitivity and 52.4% specificity. The authors say implementation must be approached with caution.

Watches were tested for oxygen accuracy, not apnea. Eighteen healthy young people wore each watch during induced low oxygen. The Apple Watch 8, Samsung Galaxy Watch 5 and Withings ScanWatch each met the 4% root-mean-square limit against a Masimo Radical-7. Apple Watch 8 had the lowest error. Those tests measured saturation, not apnea.

Is the oxygen reading itself reliable for everyone?

Not equally. The apnea studies above do not report results by skin tone. Pulse oximetry research on finger devices does.

In 146 healthy people, the Nellcor N-595 and Masimo Radical 7 both showed errors tied to skin pigment, low perfusion and the degree of low oxygen. With low perfusion, missed diagnosis of hypoxemia was 1.1% in light skin, 8.2% in medium and 21.1% in dark. An earlier study of 21 people found finger oximeters overestimated saturation by 3.56% at 60 to 70% in dark skin against 0.37% in light skin.

Those are clinical finger devices, not wrist or ring wearables. They show why accuracy by skin tone matters. See whether a wearable's blood oxygen reading works the same for everyone. Background on the sensor is in how wearables measure blood oxygen and how accurate the Apple Watch reading is.

What does this not settle?

The consumer subgroup is small and its interval is wide. The two ring studies enrolled patients already suspected of breathing problems at sleep clinics. They do not describe a person wearing a ring at home with no suspicion.

No study here tested a consumer apnea alert against polysomnography in the general public.

Heart rate variability is the other signal in play. Machine-learning models reading HRV from ECG recordings detected sleep apnea at a pooled 79.0 percent sensitivity and 75.0 percent specificity across nine studies and 2,019 people, all on ECG rather than a wrist sensor (Sleep & Breathing, 2025).

Common questions

KnowYourPrime Evidence Registry: Finding KYP-ER-0017: Can Blood Oxygen Monitoring Rule Out Sleep Apnea? · Finding KYP-ER-0001: Smartwatch Blood Oxygen (SpO2) Accuracy Against Arterial Blood (SaO2) · Finding KYP-ER-0019: Blood Oxygen Readings With Poor Circulation · Finding KYP-ER-0023: Apple Watch Blood Oxygen Accuracy Against Medical Devices