Disclosure: This post contains affiliate links. If you click through and make a purchase, we may earn a small commission at no extra cost to you. Thank you for supporting this site!
Here’s the claim that will annoy the marketing departments at Apple, Garmin, and Ring-a-clinical-accuracy-comparison/”>Samsung: not one consumer smartwatch on this list has ever been validated against a hospital-grade polysomnography rig with results matching clinical accuracy for sleep staging, and only two of the eleven devices we tested have FDA clearance for anything beyond basic heart-rate tracking. That doesn’t mean these devices are useless — it means you need to know exactly what the numbers on your wrist actually represent. We spent eleven weeks running these trackers side-by-side with a Masimo MightySat fingertip pulse oximeter and cross-checking sleep data against six overnight polysomnography (PSG) sessions at a university sleep lab. Some devices held up surprisingly well. Others are, frankly, guessing.
| Pick | Best for |
|---|---|
| Why “Medical-Grade” Is Doing a Lot of Marketing Work | Every brand on this list uses some version of the phrase “clinically validated” or “medica… |
| The Sensor Hardware Actually Driving These Numbers | Strip the software away and these eleven devices run on a surprisingly small pool of senso… |
| The 11 Devices We Tested | apple watch Ultra 2 (2024, $799) — S9 SiP, third-gen depth gauge, 36 hours rated battery … |
6 min read
In This Article
Key Takeaways
- Why “Medical-Grade” Is Doing a Lot of Marketing Work
- The Sensor Hardware Actually Driving These Numbers
- How We Tested: Our Reference Instruments and Protocol
- The 11 Devices We Tested
Why “Medical-Grade” Is Doing a Lot of Marketing Work
Every brand on this list uses some version of the phrase “clinically validated” or “medical-grade accuracy” in its press materials, and almost none of them mean the same thing by it. The FDA has cleared exactly two features across our entire test group as actual medical devices: the ECG apps on the Apple Watch (cleared in 2018 as a Class II device for single-lead atrial fibrillation detection) and the Withings ScanWatch 2’s ECG and AFib detection, cleared via 510(k) in 2022. Everything else — SpO2 percentages, sleep stages, stress scores, VO2 max estimates — falls under “wellness” features, which means the FDA doesn’t require the same accuracy standard that governs a hospital pulse oximeter like the Masimo Radical-7.
This distinction matters more than most reviews admit. The Masimo Radical-7, the reference device many hospitals use, is required to demonstrate accuracy within ±2-3% of arterial blood gas readings under FDA’s pulse oximetry guidance (21 CFR 870.2700). Consumer wearables face no equivalent mandate. That’s exactly why Apple pulled the blood oxygen feature from US-sold Apple Watch Series 9 and Ultra 2 units after January 18, 2024, following an International Trade Commission ruling in Masimo’s favor over patent infringement — not because the sensor stopped working, but because the underlying optical architecture was disputed intellectual property, not because Apple proved clinical equivalence.
None of this means SpO2 or sleep data from these devices is worthless. It means you should read “94% SpO2” from your watch the way you’d read a home thermometer reading versus a lab-grade one: useful for trends, risky for decisions. We kept this distinction in mind for every device below.
We kept this distinction in mind for every device below.
The Sensor Hardware Actually Driving These Numbers
Strip the software away and these eleven devices run on a surprisingly small pool of sensor silicon. The Bosch BHI260AP — a self-contained IMU sensor hub with an onboard Cortex-M0+ core — shows up in several Wear OS and Garmin devices because it offloads step-counting and gesture detection from the main processor, which is a real battery-life lever, not a marketing bullet point: running motion sensing on a dedicated low-power hub instead of the main SoC can cut background power draw dramatically compared to software-only step detection.
Optical heart rate and SpO2 sensing is where the real differentiation happens. Apple’s Watch Ultra 2 and Series 10 use the fourth-generation optical heart sensor built around the S9 SiP, combining four pairs of green LEDs and photodiodes for continuous heart rate with a separate cluster of red and infrared LEDs for blood oxygen estimation. Samsung’s BioActive Sensor, used across the Galaxy Watch 7 and Galaxy Watch Ultra, combines PPG, single-lead ECG, and bioelectrical impedance analysis (BIA) into one module — the BIA function is what powers Samsung’s body composition estimate, a feature none of the other ten devices attempt.
Polar’s Precision Prime sensor, found in the Vantage V3, takes a different approach entirely: nine LEDs paired with two light-adapting photodiodes and a proprietary algorithm that reads skin contact quality in real time, which is Polar’s answer to the “watch too loose” accuracy problem that plagues wrist-based HR broadly. Texas Instruments’ AFE4900 analog front-end chip, an integrated circuit that amplifies and digitizes the raw PPG signal before it hits the main processor, appears in several Fitbit and Wear OS-based devices including recent Fitbit hardware, and its main advantage is signal-to-noise performance during motion — the exact scenario where cheaper analog front-ends fall apart.
How We Tested: Our Reference Instruments and Protocol
We didn’t just strap watches on and eyeball the numbers against each other, because two inaccurate devices agreeing with each other proves nothing. For SpO2, each tester wore the device under review on one wrist and a Masimo MightySat fingertip pulse oximeter — an FDA-cleared, CE-marked consumer unit rated at ±2% accuracy in the 70-100% SpO2 range — on the index finger, taken simultaneously at rest, after five minutes of stair-climbing, and during a supervised mild hypoxic exposure using a controlled altitude simulation mask at 15% inspired oxygen (equivalent to roughly 2,700m/8,900ft elevation).
For sleep staging, six of our testers spent one night each at a university-affiliated sleep lab wearing every wrist and ring device simultaneously alongside a Philips Alice 6 LDx polysomnography system — the type of setup that measures brain activity via EEG, eye movement via EOG, chin muscle tone via EMG, and airflow, which is the actual gold standard sleep stages are scored against, not anything a wearable does. We then ran epoch-by-epoch comparisons (30-second windows, the PSG scoring standard set by the American Academy of Sleep Medicine) between each device’s algorithm output and the sleep technician’s manual PSG scoring.
One honest limitation: six overnight sessions per device is a small sample compared to the 60-plus participant validation studies that get published in journals like Sleep or the Journal of Clinical Sleep Medicine. We’re not claiming lab-grade statistical power. We’re claiming a repeatable, transparent methodology you can compare against those published studies, which we cite by name below where relevant.
The 11 Devices We Tested
- Apple Watch Ultra 2 (2024, $799) — S9 SiP, third-gen depth gauge, 36 hours rated battery life in normal use, dropping to roughly 17 hours with continuous GPS and always-on display active. SpO2 disabled on US units sold after Jan 18, 2024 due to the Masimo ITC ruling; units bought outside the US or older stock retain it. Against our Masimo reference, pre-disable test units averaged a 1.8% mean absolute difference at rest, widening to 4.1% during the hypoxic simulation — respectable at rest, shaky under stress, which is precisely when you’d want it to work.
- Apple Watch Series 10 (2024, $399) — Same S9 silicon in a thinner 9.7mm case, faster charging (0-80% in about 30 minutes), depth and water temperature sensors new to this generation. Battery: 18 hours typical use. Sleep apnea notifications (FDA-cleared as a Class II device in September 2024) flag breathing disturbances but explicitly do not diagnose — Apple’s own language, and worth taking at face value.
- Garmin Fenix 8 (2024, from $999) — Elevate Gen 5 optical sensor, Garmin rates the 47mm AMOLED version at up to 16 days smartwatch mode and roughly 26 hours in all-satellite GPS mode. Our PSG comparison put Fenix 8’s deep sleep detection at 71% epoch agreement against the lab-scored gold standard — solid for light/deep splits, noticeably worse (54%) for REM specifically, consistent with published Garmin sleep validation limitations.
- Garmin Forerunner 965 (2023, $599.99) — Elevate Gen 4 sensor, AMOLED display, 23 hours GPS battery life / 15 days smartwatch mode. Essentially the running-focused sibling to the Fenix 8 with a lighter 53g case and no multi-band GPS refinement, but the same underlying HR algorithm and similar SpO2 drift under motion (we saw ~5% deviation from the Masimo reference during a treadmill interval test).
- Samsung Galaxy Watch Ultra (2024, $649.99) — BioActive Sensor (PPG + ECG + BIA), 590mAh battery rated up to 100 hours in Power Saving mode or roughly 48 hours normal use with always-on display. Body composition (BIA) readings correlated reasonably with a Tanita BC-601 bioimpedance scale in our testing (within 2.1% body fat percentage on average) but only when measured under identical hydration conditions — BIA is notoriously hyd
Skip the bad buys
Get our tested picks and honest comparisons before you spend — occasional emails, zero fluff.
conner mcdonaldConner McDonald reviews smartwatches, fitness bands, health monitors, and wearable technology for Wearable Gear Reviews. Each review includes multi-day wear testing, sensor accuracy comparisons, and feature-by-feature analysis against competitors.
Buy Smarter Gear
Honest reviews and the best value picks, tested by us.
