Can a Smartwatch Detect Sleep Apnea? The FDA-Cleared List and the Real Numbers

The notification lands at 7:14 a.m. Your Health app says it has spotted signs of moderate-to-severe sleep apnea. By 7:20 you’ve read four forum threads, two of which tell you it’s a glitch and two of which tell you to price a CPAP machine.

Here’s the honest answer to the question you’re actually asking: no consumer wearable can diagnose sleep apnea. As of August 2026, exactly three features are cleared by the FDA to screen for it, and a screen returns a probability, not a result.

But the number that matters most isn’t the accuracy figure. It’s the asymmetry buried inside it.

A flag is very likely real. The absence of a flag means almost nothing. Apple says so itself, in the Indications for Use it filed with the FDA: “The absence of a notification is not intended to indicate the absence of sleep apnea.”¹ That sentence is not a critic’s take. It is the manufacturer’s own regulatory language.

This guide works from the primary record — the three clearances, their published sensitivity and specificity with confidence intervals and sample sizes, what each device physically measures, and what to do the morning after a flag. None of it is medical advice, and none of it replaces a clinician.

What the watch handed you is a probability. Turning a probability into a next step is a different job — and for the rest of your health data, that’s what SensAI is for. It won’t tell you whether you have apnea; nothing on your wrist can.

Can a smartwatch detect sleep apnea?

A smartwatch cannot detect sleep apnea in the diagnostic sense. As of August 2026, exactly three consumer wearable features are FDA-cleared to screen for it — Apple’s Sleep Apnea Notification Feature, Samsung’s original Sleep Apnea Feature, and Samsung’s v2.0 update that extends the feature to the Galaxy Ring — and all three are legally barred from providing a diagnosis.

That “exactly three” is checkable. Every one of these clearances lives under FDA product code QZW, “Over-the-counter device to assess risk of sleep apnea,” 21 CFR 868.2378. The product code contains three authorizations from two companies. That’s the entire cleared universe.²

Most of the confusion in this category comes from mixing three different regulatory tiers into one list:

Tier	Devices	Regulatory status	What it can tell you
FDA-cleared OTC screening	Apple Watch Series 9 or later, Ultra 2, SE 3; Samsung Galaxy Watch; Samsung Galaxy Ring (cleared July 2026)	Class II, 21 CFR 868.2378, product code QZW	Risk of moderate-to-severe apnea. Explicitly not a diagnosis2
Prescription home tests	Withings Sleep Rx (K231667), Nox T3s1 (K192469)	Prescription devices in a different regulatory category — Withings is indicated as “an aid for diagnosis”3	Ordered by a clinician, scored against clinical criteria
Unregulated wellness metrics	Oura, WHOOP, Garmin, Fitbit/Pixel — SpO2, breathing variations, respiratory-rate trends	No FDA clearance for sleep apnea. General wellness	Trends. No validated apnea claim, no published sensitivity

The regulation itself draws the line, and it’s worth reading in the FDA’s own words rather than a vendor’s. A QZW device “is not intended to provide a standalone diagnosis, replace traditional methods of diagnosis (e.g., polysomnography), assist clinicians in diagnosing sleep disorders, or be used as an apnea monitor.”²

One correction worth making, because nearly every roundup gets it wrong: the Withings Sleep Rx does not belong in the same tier as Apple and Samsung. It cleared as K231667 in September 2024 under product code MNR — Breathing Frequency Monitor, 21 CFR 868.2375 — it is prescription-only, it sits under your mattress rather than on your wrist, and it is indicated “as an aid for diagnosis of obstructive sleep apnea.”³ Listing it beside a consumer watch isn’t a nitpick; it’s mixing a clinician-ordered recording device with an over-the-counter screener.

What your watch is actually measuring (and it’s not your breathing)

No cleared feature measures airflow. They measure two entirely different downstream shadows of it — which is why their numbers look nothing alike.

Apple reads your wrist for tremor, not your chest for breath. Think of it less like a stethoscope and more like a seismometer: the watch isn’t listening to your breathing, it’s picking up the faint mechanical signature your body makes when it fights for a breath. A deep-learning model runs on Apple Watch accelerometer data — not the optical sensor — to calculate breathing disturbance values while you sleep. Apple trained it on over 11,000 nights of watch sensor data recorded concurrently with in-lab polysomnography — the overnight wired-up sleep study that is still the diagnostic gold standard — and home sleep apnea tests.¹

Those values get binned as “Elevated” or “Not Elevated,” analyzed in discrete, consecutive 30-day evaluation windows. “Elevated breathing disturbances” on an Apple Watch means exactly this and nothing more: over that 30-day window, the watch logged more of these wrist-detected interruptions than it considers typical. It is a pattern label, not a measurement of your breathing and not a severity score. You need to wear the watch to bed at least 10 nights out of 30 for the feature to work, and you never start it — it runs opportunistically in the background, and only speaks up if the 30-day pattern crosses its threshold.¹⁴

Samsung reads your blood instead. The v2.0 feature uses the green optical sensor on the underside of the device — the one that reads blood flow through your skin — plus accelerometer data to look for repetitive breathing disruptions that cause significant changes in relative blood oxygenation. It’s on-demand rather than passive, needs two nights of valid data within a 10-day period, and returns a straight binary: if the time-weighted average across those two nights is 15 or more breathing disruptions per hour, you get a positive result and a prompt to contact your doctor.⁵

Both are proxies. And proxies degrade in predictable places.

A separate check on Samsung’s wrist respiratory-rate estimate makes that concrete. In a Samsung-funded validation, researchers led by sleep neurologist Eun Yeon Joo of Samsung Medical Center, Sungkyunkwan University School of Medicine, put a Galaxy Watch on 195 sleep-clinic patients during overnight polysomnography and compared its breathing rate against a nasal thermocouple. Overnight accuracy was good — root-mean-square error of 1.13 breaths per minute. But accuracy rates that exceeded 90% in the normal-to-moderate OSA groups fell to 79.45% in patients with severe OSA.⁶

The wrist gets less reliable in exactly the patients with the most at stake.

How accurate is the Apple Watch sleep apnea notification? The validation table

The Apple Watch sleep apnea notification catches 66.3% of moderate-to-severe cases (95% CI 62.2–70.3) and almost never fires on a healthy sleeper — 98.5% specificity — measured in 1,278 analyzed subjects against a home sleep apnea test.¹ Here is every published accuracy figure in the cleared category, straight from the FDA submissions. Each one is measured against moderate-to-severe sleep apnea, defined as an apnea-hypopnea index of 15 or more events per hour.

Feature	FDA record	Sensitivity (AHI ≥ 15)	Specificity	Sample	Reference standard
Apple Watch SANF	K240929, cleared 2024-09-13	66.3% (95% CI 62.2–70.3)	98.5% (98.0–99.0)	1,499 enrolled; 1,278 analyzed	Nox T3s home sleep apnea test1
Samsung Sleep Apnea Feature (original)	DEN230041, granted 2024-02-06	82.7% (76.7–87.6)	87.7% (83.1–91.4)	620 enrolled; 470 analyzed	Polysomnography7
Samsung Galaxy Watch v2.0	K261011, cleared 2026-07-20	90.8% (86.0–94.4) — 188/207	90.3% (86.1–93.6) — 242/268	475	2-night polysomnography5
Samsung Galaxy Ring v2.0	K261011, cleared 2026-07-20	77.9% (71.5–83.5) — 155/199	95.9% (92.8–97.9) — 259/270	469	2-night polysomnography5
Oura, WHOOP, Garmin, Fitbit	None	Not published	Not published	—	No FDA-cleared apnea claim

Apple’s 66.3% looks low next to Samsung’s 90.8% until you read the design note Apple put in its own summary: “The performance was optimized for high specificity given SANF is designed as an opportunistic detection feature.”¹

That’s a deliberate trade, and a defensible one. Apple’s feature runs unattended in the background for 30 days at a stretch across tens of millions of wrists. A screen that fires too easily at that scale generates a wave of anxious, unnecessary sleep-clinic referrals. Samsung’s runs for two nights because you asked it to, so it can afford to be twitchier.

Stated plainly: Samsung’s Galaxy Watch v2.0, cleared July 2026, hits 90.8% sensitivity and 90.3% specificity in 475 subjects against two-night polysomnography — and the Galaxy Ring, cleared under the same submission, is the only FDA-cleared smart ring for sleep apnea in existence, at 77.9% sensitivity and 95.9% specificity in 469 subjects.⁵

A smartwatch is also not a substitute for a home sleep apnea test, and Apple’s own trial settles the hierarchy: the watch was measured against a Nox T3s home sleep apnea test as the reference standard, not compared to it as a rival.¹ The home test is prescription-only and clinician-ordered; the watch is an over-the-counter screener that points you toward one.

The outside literature lands in roughly the same place. A 2025 systematic review and meta-analysis of 59 studies — 2,743 participants in the wearable arm — found pooled sensitivity of 0.87 and specificity of 0.86 at AHI > 15, with an AUC of 0.93. The authors’ own conclusion: the role of these devices “is likely going to be to screen, and compliment PSG rather than replacing it.”⁸

Push past detection into grading severity and the picture gets noticeably worse. A 38-study meta-analysis of wearable AI put pooled accuracy for classifying severity level — normal, mild, moderate, severe — at just 0.651, and concluded that current performance “is suboptimal for routine clinical use.”⁹

Rings deserve their own footnote here. A 2026 review of 11 finger-worn devices across 28 articles found that all but one were more accurate at an AHI threshold of 30 than at thresholds of 5 or 15 — meaning a ring is a triage tool for severe apnea, not a detector of mild apnea.¹⁰

One more finding from the literature changes how you should use any of this. Sara H. Browne, MD and Robert L. Owens, MD of the University of California, San Diego — in work funded by a Samsung Electronics research grant — put a Galaxy Watch 4 on 51 patients undergoing in-lab polysomnography. Wrist oximetry tracked reference SpO2 tightly (root-mean-square error 2.54 percentage points) and discriminated severe apnea well on its own — AUC 0.894 for AHI > 30, but only 0.800 for AHI > 15. Add a short sleep questionnaire and both figures jumped, to 0.943 and 0.868.¹¹

In other words, the device performs best with clinical context, not instead of it. Which is the whole argument of this article, arriving from a different direction.

A flag probably means something. No flag means almost nothing.

Sensitivity and specificity sound interchangeable. They produce opposite errors, and the gap between them is what actually determines how to read your result.

Think of it as a metal detector tuned so high it only beeps on solid steel — when it beeps, there is definitely something there; when it stays quiet, you have learned nothing about the coins.

If it flagged you, take it seriously. Specificity of 98.5% means false alarms are rare by construction. In Apple’s 1,499-subject trial, the feature did not falsely notify a single subject with a normal AHI under 5.¹ A notification is not noise. One caveat on reading that in your own life: Apple’s trial population was deliberately enriched for apnea — 559 normal, 362 mild, 216 moderate, 201 severe — so the real-world chance that any given notification is correct also depends on how likely you were to have apnea in the first place.¹

If it didn’t flag you, you have learned very little. Sensitivity of 66.3% means roughly one in three people with moderate-to-severe sleep apnea received no notification at all. And mild apnea sits outside the feature’s intended use entirely — it isn’t missing those cases, it isn’t looking for them.¹

The Galaxy Ring shows the same asymmetry in a milder shape: 95.9% specificity against 77.9% sensitivity, so roughly one in five moderate-to-severe cases goes unflagged.⁵

Two structural blind spots the marketing never surfaces:

• Central sleep apnea is not the target. Samsung’s original clearance is worded explicitly for obstructive sleep apnea, and the meta-analytic literature still lists central apnea detection as an open research problem rather than a solved one.⁷⁹
• The estimates run low. A 2025 real-world validation against AI-scored polysomnography — also Samsung-supported, with two Samsung employees among its authors — found a smartwatch AHI estimate correlated strongly with the reference (ρ = 0.88, ICC = 0.87) and hit 92.3% sensitivity and 92.6% specificity for moderate-to-severe OSA — but plotting each estimate against its true value revealed systematic underestimation of actual AHI, particularly in mild OSA.¹²

So the rule to carry out of this section: a notification is a reason to book an appointment. A silent watch is not a reason to cancel one.

That posture — treat a wearable signal as an input to a decision, never the decision — is the same one SensAI takes with the HealthKit streams these devices write into. It’s also why the accuracy question deserves more nuance than a spec sheet gives it; our breakdown of how far the accuracy gap actually runs on sleep metrics covers the rest of the picture.

Oura, WHOOP, Garmin and Fitbit: what “breathing disturbances” means when nobody cleared it

None of these four has an FDA clearance for sleep apnea. The QZW product code holds three authorizations from two companies, and neither company is Oura, WHOOP, Garmin, or Google/Fitbit.²

That’s not an accusation — they don’t claim otherwise, and their own documentation is refreshingly blunt about it:

• Oura surfaces overnight blood oxygen saturation and its variation, which it describes as “indicating possible disturbances in your breathing during the night.” The same support page states plainly that “Oura Ring is not a medical device and is not intended to diagnose, treat, cure, monitor, or prevent medical conditions or illnesses.”¹³
• Garmin calls the metric Breathing Variations, derived from the pulse oximeter. The owner’s manual opens with a warning: the device “is not a medical device and is not intended for use in the diagnosis or monitoring of any medical condition.” Garmin frames the insights as awareness of “your sleep environment and general wellness.”¹⁴
• Fitbit and Pixel report Estimated Oxygen Variation, described as tracking overnight events “when your blood oxygen saturation levels varied, which could indicate breathing disturbances.” Google’s help center: “This SpO2 feature is intended for general wellness purposes only and should not be used or relied on for any medical purposes.”¹⁵
• WHOOP measures overnight respiratory rate. Its only FDA clearance on record is K243236, the WHOOP ECG feature cleared in April 2025 — an electrocardiograph clearance, with nothing covering sleep apnea.¹⁶

Here’s the fair reading, and it cuts both ways. Unvalidated does not mean useless. It means unmeasured. There is no published number telling you how many cases these metrics miss, because general wellness is a legal category that requires no validation against polysomnography and no published sensitivity. The absence of an accuracy figure is not the same as a bad one.

What that leaves you with is genuinely valuable trend data and no threshold. If your ring’s overnight oxygen variation has been climbing for weeks and you snore, that’s a reason to talk to a doctor. It is not a test result. For everything else these four platforms do well, our head-to-head across every other metric is the fuller comparison.

Your watch flagged you. Here’s the decision tree.

If your watch flagged you, do three things: export the report from the Health app, book primary care or a sleep physician, and expect a home sleep apnea test or in-lab sleep study — not a CPAP purchase.

None of what follows is medical advice, and a screening notification is not a diagnosis. It’s a prompt to start a process — here’s the process.

1. Export the report before you do anything else. On your iPhone: Health app → Search → Respiratory → Sleep Apnea Notifications → scroll down → Export PDF.⁴ It carries your breathing-disturbance history, which is the only artifact in this chain a clinician can actually read. Samsung’s app has its own export for the risk assessment and breathing-disruption data.⁵
2. Book a clinician, not a CPAP retailer. Primary care or a sleep physician. In the American Academy of Sleep Medicine’s 2018 position statement, led by Seema Khosla, MD of the North Dakota Center for Sleep, the AASM’s position is that consumer sleep technology “may be utilized to enhance the patient-clinician interaction when presented in the context of an appropriate clinical evaluation” — and that absent validation and FDA clearance, it cannot be used to diagnose or treat sleep disorders.¹⁷
3. Expect one of two tests. Under the AASM diagnostic guideline led by Vishesh K. Kapur, MD, MPH of the University of Washington, polysomnography or a home sleep apnea test with a technically adequate device is recommended for uncomplicated adults whose signs and symptoms indicate increased risk of moderate-to-severe OSA. In-lab polysomnography is recommended instead — not home testing — for patients with any of the following:¹⁸
   • Significant cardiorespiratory disease
   • Potential respiratory muscle weakness from a neuromuscular condition
   • Awake or suspected sleep-related hypoventilation
   • Chronic opioid use
   • History of stroke
   • Severe insomnia
4. A negative home test with continuing symptoms is not the end. The same guideline is explicit: if a single home sleep apnea test is negative, inconclusive, or technically inadequate, polysomnography should be performed.¹⁸
5. Bring specifics, not vibes. Snoring, witnessed breathing pauses, morning headaches, daytime sleepiness, neck circumference, BMI — plus your watch’s PDF. The STOP-Bang questionnaire is the screening instrument most clinicians already reach for; a score of 0–2 marks low risk for moderate-to-severe OSA and 5–8 marks high risk, with the probability of moderate-to-severe OSA climbing from 18% at a score of 0–2 to 60% at a score of 7–8.¹⁹

One limit worth carrying into that appointment: none of the three cleared features is indicated for central sleep apnea. Samsung’s original clearance is the only one that names a subtype, and the subtype it names is obstructive — it is “intended to detect signs of moderate to severe obstructive sleep apnea.”⁷ Apple’s Indications for Use name no subtype at all, claiming only “patterns of breathing disturbances suggestive of moderate-to-severe sleep apnea,” and nowhere claim central apnea.¹ Across both cleared vendors, central apnea is outside what the FDA reviewed.

And the part that overrides everything above: loud snoring, witnessed breathing pauses, or persistent daytime sleepiness warrant evaluation regardless of what your watch says. The watch’s job is to raise the question. It has no power to lower it.

The 425 million people this is actually for

The scale here is the reason a 66.3% screen is still a net win.

A landmark analysis in The Lancet Respiratory Medicine, funded by ResMed and co-authored by three of its employees, estimated that 936 million adults aged 30–69 have mild-to-severe obstructive sleep apnea worldwide, and 425 million have moderate-to-severe.²⁰ It’s the smaller of those two numbers that matters here: moderate-to-severe is the only band these features are cleared to screen for, and everyone whose apnea is mild sits outside their intended use entirely. A large share of that 425 million have never been tested.

Now hold two things side by side. A test with perfect sensitivity that requires a referral, a waitlist, and a night wired up in a lab. And a 66.3%-sensitive screen that runs silently on a device tens of millions of people already wear to bed, costs nothing extra, and asks for no appointment.

The second one surfaces more undiagnosed cases. Not because it’s better — because it actually runs.

Population screening and personal certainty are different jobs. Your watch is very good at the first and was never built for the second. Which means the correct response to a notification is a clinician, and the correct response to a silent watch is to keep paying attention to how you actually feel — including how sleep quality shows up in tomorrow’s session before it shows up anywhere else.

What your recovery data has been telling you all along

Untreated breathing disturbance suppresses HRV night after night with no training explanation.

The autonomic signature is well documented. A meta-analysis of 22 studies covering 2,565 patients with obstructive sleep apnea and 1,089 healthy controls found significantly reduced parasympathetic function in the OSA group: across every standard measure of HRV, severe OSA patients sat below healthy controls — and the more severe the apnea, the wider the gap.²¹ Repeated arousals and desaturations keep sympathetic drive elevated through the night. Every recovery app you own reads that as accumulated fatigue.

So here’s the pattern worth recognizing: HRV chronically below baseline. Resting heart rate creeping up. Sleep duration adequate on paper but recovery scores stubbornly poor. And no training load that explains any of it.

Deload weeks don’t fix it. Neither does more sleep. That’s the tell — and it’s the case where the correct output from a recovery app is not “take an easier week.” It’s “this isn’t a training problem.” This pattern has many causes; apnea is one worth ruling out with a clinician.

Which is precisely why SensAI reads HRV, resting heart rate, sleep, and training load as a pattern against your own baseline rather than collapsing them into a single number. A composite score can only tell you the number is low. An LLM reading those streams in context can tell you the number is low for a reason your training doesn’t explain — the distinction that separates “back off this week” from “go see someone.” If you want the mechanics of that reasoning, we’ve written about what sustained sleep debt does to HRV and how to read a wearable signal your training doesn’t explain.

To be explicit about the limit: SensAI does not screen for sleep apnea and makes no diagnostic claim of any kind. What it can tell you is that your recovery data has stopped responding to the levers you control. That is a genuinely useful sentence to walk into a doctor’s office with.

Frequently asked questions

Does the Apple Watch detect sleep apnea?

It screens; it doesn’t detect. The Sleep Apnea Notification Feature cleared in September 2024 with 66.3% sensitivity and 98.5% specificity for moderate-to-severe apnea, and Apple’s own Indications for Use state it “is not intended to diagnose, treat, or aid in the management of sleep apnea.”¹

My Apple Watch said I might have sleep apnea — what now?

Export the PDF report from Health → Search → Respiratory → Sleep Apnea Notifications, then book primary care or a sleep physician.⁴ Expect a home sleep apnea test or in-lab polysomnography depending on your other conditions.¹⁸ Don’t buy equipment before you have a diagnosis.

What do “elevated breathing disturbances” mean on Apple Watch?

Breathing disturbances are accelerometer-derived estimates of transient changes in your breathing pattern, binned as “Elevated” or “Not Elevated” and evaluated over 30-day windows.¹ Apple notes that a few disturbances a night are typical, and that alcohol, upper respiratory illness, and some medications can all push the number up.⁴

Which smartwatches are FDA-cleared for sleep apnea?

Three authorizations, two companies: Apple’s Sleep Apnea Notification Feature (K240929, September 2024), Samsung’s original Sleep Apnea Feature (DEN230041, February 2024), and Samsung’s Sleep Apnea Feature v2.0 covering both the Galaxy Watch and Galaxy Ring (K261011, July 2026).¹⁵⁷

Can the Oura Ring or WHOOP detect sleep apnea?

Neither has an FDA clearance for sleep apnea, and neither publishes a sensitivity figure against polysomnography.² Samsung’s Galaxy Ring, cleared in July 2026 at 77.9% sensitivity and 95.9% specificity in 469 subjects, is the only cleared smart ring in the category.⁵

Is a smartwatch as good as a home sleep apnea test?

No — and Apple’s own trial makes the hierarchy obvious, since it used a Nox T3s home sleep apnea test as the reference standard the watch was measured against.¹ The home test is the yardstick, not the competitor. Home tests are prescription devices ordered by a clinician; watches are over-the-counter screeners.

My watch didn’t flag me but I snore loudly — should I still see a doctor?

Yes. Roughly a third of moderate-to-severe cases got no notification in Apple’s validation study, and mild apnea is outside the feature’s intended use altogether.¹ Symptoms override a silent watch — the same reason SensAI’s coach defers to a clinician when a recovery pattern stops tracking training load.

Can a smartwatch detect central sleep apnea?

No cleared feature is indicated for it. Samsung’s original clearance is written specifically for obstructive sleep apnea, and researchers reviewing the wearable AI literature have called for more work on central apnea detection precisely because the evidence isn’t there.⁷⁹

The bottom line

Your watch is a smoke detector, not a diagnosis — and it’s a smoke detector deliberately tuned not to go off when you make toast. That tuning is why a flag is worth acting on and a silence proves nothing.

Three features are cleared to screen. Everything else on your wrist or finger is a trend line, which is useful and unmeasured at the same time.

A flag is a phone call to a doctor. Symptoms are a phone call to a doctor whether or not anything buzzed.

---

References

Footnotes

1. U.S. Food and Drug Administration. “510(k) Summary: Sleep Apnea Notification Feature (SANF), Apple Inc.” K240929, decided September 13, 2024. https://www.accessdata.fda.gov/cdrh_docs/pdf24/K240929.pdf ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶ ↩⁷ ↩⁸ ↩⁹ ↩¹⁰ ↩¹¹ ↩¹² ↩¹³ ↩¹⁴ ↩¹⁵ ↩¹⁶ ↩¹⁷

2. U.S. Food and Drug Administration. “Product Classification: Over-The-Counter Device To Assess Risk Of Sleep Apnea (QZW), 21 CFR 868.2378.” CDRH Product Classification Database, accessed August 2026. https://www.accessdata.fda.gov/scripts/cdrh/cfdocs/cfPCD/classification.cfm?ID=QZW — full list of authorizations under the code (three records, two companies): https://api.fda.gov/device/510k.json?search=product_code:QZW&limit=100 ↩ ↩² ↩³ ↩⁴ ↩⁵

3. U.S. Food and Drug Administration. “510(k) Summary: Withings Sleep Rx, Withings.” K231667, decided September 6, 2024. Product code MNR, 21 CFR 868.2375, prescription use. https://www.accessdata.fda.gov/cdrh_docs/pdf23/K231667.pdf ↩ ↩²

4. Apple Inc. “Sleep apnea notifications on your Apple Watch.” Apple Support article 120031, published September 18, 2025. https://support.apple.com/en-us/120031 ↩ ↩² ↩³ ↩⁴

5. U.S. Food and Drug Administration. “510(k) Summary: Sleep Apnea Feature v2.0, Samsung Electronics Co., Ltd.” K261011, decided July 20, 2026. https://www.accessdata.fda.gov/cdrh_docs/pdf26/K261011.pdf ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶ ↩⁷ ↩⁸

6. Jung H, Kim D, Choi J, Joo EY. “Validating a Consumer Smartwatch for Nocturnal Respiratory Rate Measurements in Sleep Monitoring.” Sensors (Basel), 2023; 23(18): 7976. https://pubmed.ncbi.nlm.nih.gov/37766031/ ↩

7. U.S. Food and Drug Administration. “De Novo Classification Request: Sleep Apnea Feature, Samsung Electronics Co., Ltd.” DEN230041, granted February 6, 2024. https://www.accessdata.fda.gov/scripts/cdrh/cfdocs/cfPMN/denovo.cfm?ID=DEN230041 — performance data, indications wording, and sample sizes in the De Novo Decision Summary: https://www.accessdata.fda.gov/cdrh_docs/reviews/DEN230041.pdf ↩ ↩² ↩³ ↩⁴ ↩⁵

8. Hasanain M, Akram A, Kabir A, Javed H, Usman M, Jaffar H, Khan AA, Ali Malik GM, Tabassum M, Faizan M, Anjum MU, Umer M, Surani S. “Diagnostic value of wearable and contactless technologies for detecting obstructive sleep apnea: a systematic review and meta-analysis.” Sleep and Breathing, 2025; 29(6): 368. https://pubmed.ncbi.nlm.nih.gov/41310031/ ↩

9. Abd-Alrazaq A, Aslam H, AlSaad R, Alsahli M, Ahmed A, Damseh R, Aziz S, Sheikh J. “Detection of Sleep Apnea Using Wearable AI: Systematic Review and Meta-Analysis.” Journal of Medical Internet Research, 2024; 26: e58187. https://pubmed.ncbi.nlm.nih.gov/39255014/ ↩ ↩² ↩³

10. Jin Y, Xu J, Yue H, Huang Y, Ma W, Fan X, Wang H, Xu L, Wang J. “Performance evaluation of finger-worn devices for sleep stage classification and sleep apnea detection: a systematic review and meta-analysis.” Journal of Translational Medicine, 2026; 24(1): 873. https://pubmed.ncbi.nlm.nih.gov/42141446/ ↩

11. Browne SH, Vaida F, Umlauf A, Kim J, DeYoung P, Owens RL. “Performance of a commercial smart watch compared to polysomnography reference for overnight continuous oximetry measurement and sleep apnea evaluation.” Journal of Clinical Sleep Medicine, 2024; 20(9): 1479-1488. https://pubmed.ncbi.nlm.nih.gov/38652502/ ↩

12. Kim D, Han JY, Jung H, Song DY, Lee C, Ryu G, Hong SD, Kim HY, Jung YG. “AI-Enhanced Smartwatch AHI Estimation and AI-Scored Polysomnography for Obstructive Sleep Apnea: Real-World Validation.” Nature and Science of Sleep, 2025; 17: 2297-2307. https://pubmed.ncbi.nlm.nih.gov/41018852/ ↩

13. Oura Health Oy. “Blood Oxygen Sensing (SpO2).” Oura Member Care, last updated July 27, 2026. https://support.ouraring.com/hc/en-us/articles/7328398760851-Blood-Oxygen-Sensing-SpO2 ↩

14. Garmin Ltd. “Breathing Variations.” fenix 7 Standard/Solar/Pro Series Owner’s Manual, August 2025. https://www8.garmin.com/manuals/webhelp/GUID-C001C335-A8EC-4A41-AB0E-BAC434259F92/EN-US/GUID-636632DF-E3E8-4C15-936A-CFB7EA2B5C39.html ↩

15. Google LLC. “How do I track blood oxygen saturation (SpO2) with my Fitbit device?” Google Health Help Center, accessed August 2026. https://support.google.com/fitbit/answer/14237927 ↩

16. U.S. Food and Drug Administration. “510(k) Premarket Notification: WHOOP ECG (electrocardiogram) Feature (1.0), Whoop, Inc.” K243236, decided April 4, 2025. https://www.accessdata.fda.gov/scripts/cdrh/cfdocs/cfPMN/pmn.cfm?ID=K243236 ↩

17. Khosla S, Deak MC, Gault D, Goldstein CA, Hwang D, Kwon Y, O’Hearn D, Schutte-Rodin S, Yurcheshen M, Rosen IM, Kirsch DB, Chervin RD, Carden KA, Ramar K, Aurora RN, Kristo DA, Malhotra RK, Martin JL, Olson EJ, Rosen CL, Rowley JA. “Consumer Sleep Technology: An American Academy of Sleep Medicine Position Statement.” Journal of Clinical Sleep Medicine, 2018; 14(5): 877-880. https://pubmed.ncbi.nlm.nih.gov/29734997/ ↩

18. Kapur VK, Auckley DH, Chowdhuri S, Kuhlmann DC, Mehra R, Ramar K, Harrod CG. “Clinical Practice Guideline for Diagnostic Testing for Adult Obstructive Sleep Apnea: An American Academy of Sleep Medicine Clinical Practice Guideline.” Journal of Clinical Sleep Medicine, 2017; 13(3): 479-504. https://pubmed.ncbi.nlm.nih.gov/28162150/ ↩ ↩² ↩³

19. Chung F, Abdullah HR, Liao P. “STOP-Bang Questionnaire: A Practical Approach to Screen for Obstructive Sleep Apnea.” Chest, 2016; 149(3): 631-638. https://pubmed.ncbi.nlm.nih.gov/26378880/ ↩

20. Benjafield AV, Ayas NT, Eastwood PR, Heinzer R, Ip MSM, Morrell MJ, Nunez CM, Patel SR, Penzel T, Pepin JL, Peppard PE, Sinha S, Tufik S, Valentine K, Malhotra A. “Estimation of the global prevalence and burden of obstructive sleep apnoea: a literature-based analysis.” The Lancet Respiratory Medicine, 2019; 7(8): 687-698. https://pubmed.ncbi.nlm.nih.gov/31300334/ ↩

21. Wang Z, Jiang F, Xiao J, Chen L, Zhang Y, Li J, Yi Y, Min W, Su L, Liu X, Zou Z. “Heart rate variability changes in patients with obstructive sleep apnea: A systematic review and meta-analysis.” Journal of Sleep Research, 2023; 32(1): e13708. https://pubmed.ncbi.nlm.nih.gov/36070876/ ↩