The most consequential number in the Samsung Galaxy Ring's sleep apnea capability is 15. Not sensitivity, not specificity, not the price. Fifteen apnea-hypopnea events per hour is the line the cleared algorithm is authorized to draw — and it is the reason the ring will stay silent for roughly half the people who actually have sleep apnea. That is not a defect in the product. It is written into the regulatory language, and the gap between what the clearance says and what buyers will assume it says is the most important thing in this launch.
Our verdict, stated up front: on the narrow population the clearance describes, the ring flagged what the reference test flagged, and it did so about as well as the wrist device running the same algorithm family. Its real advantage over the watch is not measurement quality. It is that people kept it on.
We spent nine weeks running the feature against a scored home sleep apnea test. Here is the protocol, then the data, then the part nobody puts in the press materials.
What the clearance actually authorizes
Samsung's sleep apnea feature reached the US market through a De Novo authorization in February 2024 — a pathway that creates a new device classification rather than clearing a device against an existing predicate. That matters, because it means the regulator evaluated one specific claim on one specific hardware-and-algorithm pairing. It is not a grade awarded to the product as a whole, and it does not transfer to a company's other sensors by association.
The authorized claim is narrow in four ways worth naming individually. It detects signs of moderate-to-severe obstructive sleep apnea — AHI of 15 or higher — not mild disease. It is intended for adults 22 and over who have not already been diagnosed. It requires two qualifying nights of at least four hours' sleep inside a ten-day window before it will produce any result. And its output is a binary: signs detected, or not. It is a triage prompt to see a clinician, not a number you can act on.
Set that threshold against prevalence. Benjafield and colleagues, in a 2019 modelling study in The Lancet Respiratory Medicine, estimated 936 million adults aged 30–69 worldwide with an AHI of 5 or higher, of whom 425 million met the AHI-15 threshold. Slightly more than half of the affected population therefore sits below the line the cleared feature is permitted to detect. A silent result from this device is not evidence of clean breathing. It is evidence that you are probably not in the top half of severity.
How we tested
Eleven participants completed the protocol out of fourteen recruited; three withdrew over ring fit and one night of skin irritation. Each participant wore three devices simultaneously for six consecutive nights, yielding 62 usable nights after we discarded four for reference-signal loss.
The reference was a type III home sleep apnea test — nasal pressure cannula, thoracic effort belt, finger pulse oximetry — scored manually by a registered polysomnographic technologist using AASM criteria with the 3% desaturation hypopnea rule. Participants wore the reference kit on nights one, three and five. The smart ring went on the index finger of the non-dominant hand; the comparison ring on the dominant hand; the watch on the non-dominant wrist. Ring sizing was done with the manufacturer's kit, and we photographed placement each night to check for rotation.
By reference AHI, the panel split four participants under 5, four between 5 and 15, and three at 15 or above.
What we could not test. We had no in-lab polysomnography, which is the actual gold standard; type III testing tends to underestimate AHI because it divides events by recording time rather than sleep time. Our panel had eleven people, which is nowhere near enough to estimate sensitivity or specificity, and we are not going to publish figures that imply otherwise. Our participants were also insufficiently diverse in skin tone to say anything about optical performance across pigmentation — a live concern given that Sjoding and colleagues documented systematically missed hypoxemia in Black patients using hospital-grade pulse oximeters in a 2020 NEJM research letter. If that bias exists in a regulated bedside device, assuming it away in a consumer PPG sensor is not defensible.
What the data showed
On the three participants with reference AHI at or above 15, the ring produced a positive result inside the ten-day window in all three cases. The watch did the same. Neither device flagged any of the four participants in the 5–15 band — correct behaviour under the clearance, and precisely the behaviour that will mislead people.
One participant with a reference AHI of 3.1 received a positive result. With four people in that stratum, we cannot call that a false-positive rate; we can only report that it happened and that the participant subsequently tested negative on a repeat night.
Where we could compare continuous signals, the ring tracked the reference oximeter reasonably: mean bias on 3% oxygen desaturation index was −1.8 events per hour against the reference, with the ring reading low, and nightly SpO2 nadir ran 2.4 percentage points high on average. Directionally consistent, not interchangeable with a scored study.
Failure modes were mundane and predictable. Six nights produced no result because sleep fell under the four-hour minimum. Two nights were lost to ring rotation moving the sensor off the palmar side of the finger. Cold-hand nights — participants who ran their bedrooms below about 17°C — produced visibly noisier perfusion traces, though not enough to void results.
The comparison grid
| Device | US regulatory status | What triggers a result | Window | Nightly data shown |
|---|---|---|---|---|
| Samsung's smart ring | Cleared claim, ring form factor | Signs of AHI ≥ 15 | 2 qualifying nights in 10 | Binary result plus trend metrics |
| Galaxy Watch | De Novo authorised, Feb 2024 | Signs of AHI ≥ 15 | 2 qualifying nights in 10 | Same |
| Apple Watch | Cleared 2024, separate submission | Elevated breathing disturbances | Rolling 30 days | Nightly disturbance metric |
| Oura Ring 4 | No US OSA detection clearance | n/a | n/a | Breathing regularity, average SpO2 |
| Type III home test | Prescription diagnostic | Scored AHI, full spectrum | 1–3 nights | Full respiratory channels |
The row that deserves attention is Apple's. A 30-day window with a different underlying trigger is a different product philosophy: slower to speak, drawing on more nights. Samsung's ten-day window gets to an answer faster. Neither is obviously right, and neither has been compared head-to-head in independent published work that we could find.
Where the ring actually wins
Not on accuracy. On adherence. Across 62 nights, the ring was worn for 61. The watch was worn for 48 — participants took it off for charging, for comfort, or because they disliked a screen on the wrist in bed. A screening algorithm that requires two qualifying nights inside ten days is hostage to whether the device is on the body, and in our panel the finger won that argument decisively.
If you want the cleared feature and you know you will not wear a watch to bed, the ring is the better instrument — for compliance reasons, not sensor reasons.
Who this is for, and who it isn't
It is for undiagnosed adults with a real index of suspicion — witnessed pauses, loud snoring, unrefreshing sleep — who want a low-friction nudge toward a clinician, and who will act on a positive result rather than filing it away.
It is not for anyone already diagnosed or on CPAP; the clearance excludes you and the output is not designed to track therapy. It is not for people with suspected mild apnea, who are exactly the population the AHI-15 line excludes. And it is not a substitute for a diagnostic study — a positive result buys you a referral, not a diagnosis.
Evidence grade for the central claim — that this device detects signs of moderate-to-severe OSA in the population described: Moderate. There is a regulatory review behind it and our small panel was consistent with it. Evidence grade for the claim most buyers will infer — that a silent ring means you do not have sleep apnea: Weak, and we would take Unclear over Weak only because the question has barely been asked properly.
Which brings us back to 15. We opened by calling it the most consequential number here, and after nine weeks of nights we would revise that only slightly: 15 is not a marketing decision Samsung made, it is roughly where the physics of a finger-worn optical sensor and an accelerometer stop being able to separate a shallow breath from a stopped one. The clearance did not choose that boundary so much as ratify it. Read the number that way and the feature is honest work. Read it as a clean bill of health and you have been told nothing about the half of sleep apnea that lives below the line.