Pulse Oximetry on Your Wrist: What SpO2 Numbers Actually Mean (and Where They Fail)

Every smartwatch and fitness ring above about $200 now displays a number called SpO2 — your blood oxygen saturation, expressed as a percentage. Most of the time it reads 97, 98, 99, and users learn to ignore it. That is broadly the right response. SpO2 is a red flag, not a training tool. This piece explains what the number is measuring, when it is reliable, when it is not, and the one thing consumer SpO2 is genuinely useful for.

What SpO2 actually measures

Pulse oximetry exploits a small piece of optics. Haemoglobin — the protein in red blood cells that carries oxygen — absorbs light differently depending on whether it is loaded with oxygen or not. Oxygenated haemoglobin absorbs more infrared light and lets through more red light. Deoxygenated haemoglobin does the opposite.

A pulse oximeter shines red light (around 660 nanometres) and infrared light (around 940 nanometres) through pulsatile tissue and measures how much of each colour comes out the other side. The device then computes the ratio of ratios: the pulsatile portion of the red signal over its non-pulsatile portion, divided by the same fraction for infrared. That ratio maps, through a calibration curve, to an estimated saturation percentage.

The result — SpO2 — is an estimate of the percentage of haemoglobin binding sites in your arterial blood that are currently carrying oxygen. Full saturation is 100 percent. A healthy adult at sea level breathing room air will sit somewhere between 95 and 100 percent almost all the time.

Normal, low, and dangerous

Rough thresholds every user of a pulse oximeter should know.

• 95 to 100 percent — normal. A healthy adult at sea level should sit in this range at rest and during light activity.
• 91 to 94 percent — mildly low. Common at moderate altitude, during a cold or acute respiratory illness, and in people with mild chronic lung disease. Not immediately dangerous in most contexts but worth a note.
• Below 90 percent — clinically concerning. This is where oxygen delivery to tissues starts to fall meaningfully. In a hospital, this triggers assessment.
• Below 88 percent — treatment threshold. Sustained readings at this level are the standard cut-off for supplemental oxygen prescription in chronic lung disease. Anyone consistently reading here, or dropping here without obvious explanation, should see a doctor.

Two important nuances. First, SpO2 is not linear with tissue distress. Because of the shape of the oxygen-haemoglobin dissociation curve, saturation stays close to 100 percent even when arterial oxygen tension is falling. By the time SpO2 has dropped ten points, arterial oxygen has dropped much more. Second, saturation and oxygen delivery are not the same thing. You can have normal SpO2 and still be oxygen-starved if your haemoglobin is low, your cardiac output is poor, or your peripheral circulation is compromised.

Why low readings happen

When SpO2 drops meaningfully, there is almost always an explanation. The main ones:

• Altitude. At 8,000 feet, resting SpO2 in an unacclimatised sea-level dweller often sits at 90 to 93 percent. At 14,000 feet, values in the mid-80s are common. This is normal physiology; it is not a disease.
• Chronic lung disease. COPD, pulmonary fibrosis, cystic fibrosis, and severe asthma all impair gas exchange in the alveoli. Persistent daytime readings under 92 percent in a sedentary adult are a reason to see a pulmonologist.
• Obstructive sleep apnea. Repeated overnight desaturations during airway collapse. Daytime SpO2 usually normal. We cover this in a separate piece.
• Acute respiratory infection. Pneumonia, severe influenza, and COVID-19 pneumonitis can all drop saturation. This is the use case that made home pulse oximeters mainstream in 2020.
• Anaemia and carbon monoxide poisoning. Both are dangerous and both are situations in which SpO2 can read misleadingly. Standard pulse oximetry cannot distinguish oxyhaemoglobin from carboxyhaemoglobin, so a person poisoned by carbon monoxide may show a normal SpO2 while critically ill.

Consumer wearable accuracy — the honest picture

Here is where marketing and reality diverge.

Hospital-grade pulse oximeters, clipped to a fingertip with a proper transmissive optical path and calibrated against arterial blood gas measurements, have accuracy specifications of around plus or minus 2 percent in the 70 to 100 percent range. Consumer wrist wearables use reflective pulse oximetry (light bounces back off tissue rather than passing through it), have to fight much more motion artefact, and typically report accuracy in the plus or minus 3 to 5 percent range when everything goes right — and much worse when it does not.

Two use cases behave very differently.

• Trend detection over hours or nights. Reasonable. A wrist SpO2 trace that drops repeatedly overnight is a useful signal, because the pattern matters more than the absolute numbers.
• Spot readings for decision-making. Weak. A single wrist reading of 94 in a healthy person almost certainly reflects a sensor artefact — cold fingers, loose band, tattoo interference, motion. It is not a reason to worry.

Wrist SpO2 is designed for detection of patterns, not for precision on any one measurement. Treat it accordingly.

The 2021 FDA safety communication and skin pigmentation

This deserves its own section because it matters for a lot of users.

In February 2021, the FDA issued a safety communication titled Pulse Oximeter Accuracy and Limitations, warning that pulse oximeters can be less accurate in people with dark skin pigmentation. The communication was triggered by a study published in the New England Journal of Medicine in December 2020 by Sjoding and colleagues, which reviewed data from more than ten thousand hospitalised patients and found that Black patients were nearly three times as likely as white patients to have occult hypoxaemia — arterial oxygen below 88 percent while their pulse oximeter still read 92 to 96 percent.

The mechanism is optical. Melanin in the epidermis absorbs light at wavelengths overlapping the red and infrared bands the sensor uses. The calibration curves for most existing devices were built primarily on data from lighter-skinned volunteers, so the ratio of ratios the device sees does not map to the same true saturation in darker skin. In practice, pulse oximeters tend to overestimate saturation in patients with more melanin, especially at lower true saturation levels — which is exactly the range where an accurate reading matters most.

Two implications for consumer wearables.

1. If you have darker skin, and your wrist device reports normal SpO2 while you feel short of breath, take the symptom more seriously than the reading. The device may be masking a real drop.
2. Manufacturers have started publishing more diverse validation data since 2021, but not all consumer devices are held to the same standard. FDA-cleared devices with published clinical validation deserve more trust than devices that quietly advertise SpO2 as a wellness feature only.

Two consumer devices with real clearance

The Withings ScanWatch 2 is the wrist device we would trust most for overnight SpO2 trend data. Withings publishes peer-reviewed validation for the pulse oximetry and holds an FDA 510(k) clearance for its breathing disturbance detection algorithm. That regulatory clearance is not the same as a hospital pulse oximeter clearance, and it is not marketed as a diagnostic tool, but it puts the device in the small subset of consumer wearables that has been reviewed by the FDA rather than only marketed at users.

The Fitbit Sense and its refurbished successor use continuous overnight SpO2 rather than spot readings, which is the right architecture for detecting patterns like sleep-disordered breathing. Fitbit publishes validation data on both the pulse oximeter and the sleep-stage algorithm, and its SpO2 estimation range and known limitations are documented — more than most competitors offer.

How pulse oximetry got everywhere

A little history explains the state of the wrist SpO2 market. The modern pulse oximeter emerged from Japanese engineering in the 1970s — Takuo Aoyagi at Nihon Kohden and Susumu Nakajima at Minolta Camera developed the two-wavelength technique that survives today. By the late 1980s, fingertip and earlobe pulse oximeters were standard equipment in anaesthesia and critical care, and they arguably saved more lives that decade than any other single medical monitor.

Consumer wearable SpO2 arrived much later. Apple added it to the Apple Watch Series 6 in 2020. Fitbit and Garmin followed. The COVID-19 pandemic accelerated interest enormously — fingertip pulse oximeters sold out worldwide in early 2020 as home users looked for a way to monitor respiratory illness. Manufacturers built on that demand, and by 2022 nearly every premium wrist wearable advertised SpO2 as a headline feature.

The rush created a mismatch. The underlying algorithm was designed for well-perfused fingertips at rest, not for cold wrists at motion. Consumer accuracy has caught up somewhat as manufacturers have refined optical stacks and adaptive filtering, but the gap between what a hospital device does at 99 percent and what a wrist device does at 90 percent is still real.

Where wrist SpO2 fails hardest: altitude training

If you have ever climbed a big mountain or trained at altitude, someone has probably told you to track your SpO2 as a proxy for hypoxic dose. This is a bad idea with a wrist wearable.

At altitude, resting SpO2 falls into ranges — 88 to 93 percent, then lower — where consumer wrist accuracy is worst. The optical signal is weaker because peripheral vasoconstriction (cold, high altitude) reduces pulsatile flow at the wrist. Motion artefact is high because you are moving. Sensor drift is worse. The number you see on the watch face at 15,000 feet may be four or five percentage points off in either direction.

Serious altitude athletes and mountaineers use fingertip transmissive pulse oximeters, not wrist devices. If you are calibrating a hypoxic tent, tracking acute mountain sickness risk, or timing rest periods on a big peak, use a dedicated fingertip unit with published accuracy data and cross-check against how you feel. The wrist number is background context at best.

How to actually use SpO2 on a wearable

A short list of sane uses.

1. Overnight trend for sleep-disordered breathing. If your wearable shows repeated overnight drops of three to four percentage points, especially paired with snoring or daytime sleepiness, that is a reason to talk to your doctor about a sleep study.
2. Illness monitoring. Persistent daytime SpO2 in the low 90s during a respiratory infection is a reason to escalate care. This is exactly the use case that made home pulse oximeters mainstream in 2020.
3. Longitudinal baseline. Knowing your own normal helps you notice when it is not normal. A month of steady 97 percent daytime readings makes a sustained shift to 93 percent more meaningful.

And a short list of misuses.

• Reacting to a single reading. A 94 in a healthy person on a wrist wearable is almost always artefact.
• Using it to gauge training load. Fitness-related fatigue does not consistently register on SpO2. Use HRV or resting heart rate.
• Using it in place of a fingertip meter during altitude exposure or during an acute illness where every point matters.

The bottom line

SpO2 on your wrist is a screening tool, not a training metric. It is genuinely useful for detecting overnight patterns of sleep-disordered breathing and for keeping an eye on a nasty respiratory infection at home. It is much less useful as a spot reading for a healthy person, and it is worst exactly at the low-saturation ranges where accuracy matters most.

If you have darker skin, treat wrist SpO2 with more suspicion than the number suggests — the FDA and the New England Journal of Medicine agree the calibration is imperfect. If you are training at altitude, use a fingertip meter. And if a healthy person's watch tells you they are at 96 percent while they feel fine, believe how they feel first.