Blood Oxygen During Sleep: How Accurate Is What Your Watch Shows You

This piece covers what published validation studies report about wrist and ring pulse oximetry accuracy, including overnight and sleep-apnea-related testing, and the known limits of pulse oximetry technology itself. It does not cover medical diagnosis of sleep apnea or other conditions, and it is not a buying guide.

Overnight SpO2 tracking on a consumer wearable has been tested against lab-grade equipment, and the picture that comes back is mixed rather than reassuring or dismissive. A commercial wrist device compared against polysomnography showed measurable error against the reference oximeter, while separate validation work on a smartwatch and a smart ring reported closer agreement under controlled, mostly-still conditions. The gap between those two pictures, controlled stillness versus a real night of sleep, is where the accuracy question actually lives.

The number on the app in the morning versus what happened in your body overnight

People check their sleep app the way they used to check the weather, a quick glance at an overnight low and a general sense of whether last night was 'good.' A dip into the low 90s on the SpO2 graph reads like a fact. But an overnight blood oxygen curve from a wrist or finger sensor is a reconstruction built from light passing through skin and blood, not a direct blood draw. And the research on how well that reconstruction holds up while a person is actually asleep, rather than sitting still in a lab chair, is thinner than the confident little graph suggests.

The tension shows up specifically around sleep tracking, where movement, sensor position, and hours of continuous measurement all stack up in ways a short daytime check never has to deal with.

What the sensor is actually doing, and where that runs into trouble

Pulse oximetry, wrist-worn or finger-clipped, works by shining light through tissue and reading how much of it comes back, then inferring oxygen saturation from that pattern. That method has known, long-documented limits that have nothing to do with a specific brand. A foundational review of the technology lays out performance problems tied to motion, low blood flow to the extremities, and other physiological factors that can throw off the reading regardless of how well the device is built. A separate evaluation of pulse oximeter accuracy in general makes a related point, there's no single perfect reference standard, calibration population matters, and the algorithms converting raw light signal into a displayed number are themselves a source of variability between devices.

Sleep adds its own version of the motion and perfusion problem. A person lying still for a few minutes in a lab chair during a validation test is not the same body as one rolling, curling an arm underneath a pillow, or losing circulation to a wrist for stretches of the night. None of the wearable-specific studies here were designed to isolate that difference directly. But the underlying physics they're all working against, light absorption changing with blood flow and movement, is consistent across the older foundational work and the newer device trials alike.

Skin tone and the accuracy question nobody's watch graph mentions

A separate strand of this research doesn't ask how a specific wearable performs, it asks whether pulse oximetry in general performs the same way across different skin pigmentation. A prospective study measuring pulse oximeter readings against arterial oxygen saturation in subjects grouped by Fitzpatrick skin type found that during stable low-oxygen conditions, devices were more likely to miss hypoxemia in darkly pigmented skin, a pattern separately confirmed in another controlled study that found overestimation at low saturation levels specifically in darker skin. A systematic review pooling tens of thousands of paired measurements across dozens of studies evaluated this same pattern at scale, and asked whether the errors involved were large enough to matter clinically, not just statistically.

Not a wearable-specific quirk, this one. A review of the racial disparity literature traces the documented history of pulse oximeter overestimation in dark skin tones back decades and describes it as a contributor to unrecognized low oxygen going unnoticed in clinical settings. None of the sleep-tracking validation studies in this piece report their results broken down by skin tone, so it's not established here whether the same pattern shows up specifically in overnight wrist or ring SpO2 tracking, only that the underlying sensor technology has a documented history of this exact failure mode elsewhere.

Newer form factors, same open questions

Rings are the newest entrant into this space, and the available validation work on them is earlier-stage. One study describes the design and testing of an adjustable smart ring built to track pulse rate and blood oxygen using the same light-based sensing approach, evaluated on a group of healthy volunteers across a testing sequence that included both staying still and controlled movement. That's a meaningfully different test setup than a full night of sleep, and the study itself notes that validation data during dynamic, real-world conditions has generally been a gap in the ring literature.

Smartwatches aimed at older adults face a related question, since aging changes circulation and skin in ways that could plausibly affect a light-based sensor. A comparison of a smartwatch's blood pressure and oxygen saturation readings against reference devices in adults across a wide age range reported no significant difference when values were averaged, though that comparison was not conducted overnight and doesn't speak to continuous sleep tracking specifically.

Common questions