Blood Oxygen & SpO2 · Evidence-Based Health Guide

Blood Oxygen (SpO2): What the Number Means and Where It Fails

A pulse oximeter reading is an estimate, not a measurement — and the size of the error is not the same for everybody. Here is what the number means, how wide the margin really is, and the accuracy gap across skin tones that regulators are still working to close.

86–94%
the true saturation range behind a cleared oximeter reading of 90%1
1.71×
the rate of missed low oxygen in Black patients vs White patients4
10
skin-tone values the FDA now proposes testing against, up from a two-subject minimum5
2.9–6.7%
error range across four consumer smartwatches in one head-to-head test6

1What the number actually is

SpO2 is the percentage of haemoglobin in your blood that is carrying oxygen — estimated from light, not measured from blood.

A pulse oximeter shines red and infrared light through tissue and infers saturation from how much of each colour is absorbed. Oxygenated and deoxygenated haemoglobin absorb the two wavelengths differently, so the ratio gives an estimate. The clinical gold standard, an arterial blood gas, measures saturation directly from a blood sample — written SaO2. SpO2 is the estimate of SaO2. Every limitation below comes from that gap.

2Where readings sit, and where they matter

NHS home-monitoring thresholds. These are general adult guidance and do not replace clinical advice — people with chronic lung disease are often given different personal targets.2
ReadingWhat it meansWhat to do
96% or moreNormal rangeNothing
95%Acceptable — keep an eye on itContinue monitoring
93–94%Below the normal rangeContact your GP
92% or lessLow — needs assessment nowEmergency care

The thresholds are narrow. The error bar is not.

The whole clinically interesting range — normal to emergency — spans about four percentage points. The FDA's own margin on a single cleared-device reading is roughly the same width: a reading of 90% generally means a true saturation somewhere between 86% and 94%.1 That is why no clinician acts on one number in isolation, and neither should you.

3The accuracy gap across skin tones

This is the best-documented limitation of pulse oximetry, and the reason the rules are being rewritten.

In 2020, researchers compared paired pulse-oximeter and arterial blood-gas readings and looked for occult hypoxemia — true arterial saturation below 88% while the oximeter was still showing a reassuring 92–96%. The pattern held across two separate cohorts:3

ICU cohort — Black patients17.0% of readings
ICU cohort — White patients6.2% of readings
Inpatient cohort — Black patients11.7% of readings
Inpatient cohort — White patients3.6% of readings
All four bars are the same unit: the share of paired readings where true saturation was under 88% but the oximeter read 92–96%.

A later systematic review pooled 263,000 paired observations and found the same direction of error. Oximeters overestimated saturation by an average of 1.54 percentage points in Black patients versus 0.62 in White patients, and pooled occult-hypoxemia prevalence was 11.4% versus 6.5% — a prevalence ratio of 1.71, or 71% higher. The reviewers rated the evidence for overestimation in Black patients as moderate strength, and evidence on downstream clinical outcomes as insufficient — the accuracy gap is well established; how much harm it causes is still being quantified.4

The honest nuance: the gap is smallest where wearables operate

The bias grows as saturation falls. Laboratory work found a maximum bias of about 3.6 percentage points in darkly pigmented subjects in the 60–70% saturation range, but under 1 point above 80% saturation.7 In other words, the error is largest exactly where the stakes are highest — critical care — and smallest in the healthy range where a smartwatch spends its time. That does not make it irrelevant to consumers: it means a wrist reading of 94% deserves the same scepticism from everyone, and a borderline reading deserves more follow-up, not less, if you have darker skin.

4What regulators are changing

The testing rules that produced the gap are being rewritten, but the work is not finished.

Under the long-standing approach, a device could be validated on as few as ten people, with darkly pigmented subjects making up 15% of test subjects, or two people, whichever was greater.7 In January 2025 the FDA issued draft guidance — Pulse Oximeters for Medical Purposes — proposing a substantially larger and more representative evidence base: around 150 participants, with skin tone recorded using both a subjective scale (the Monk Skin Tone Scale, 10 values, analysed in three groups: 1–4, 5–7 and 8–10) and an objective optical measure (individual typology angle).5,8

Status: still draft

As of this guide's last update the FDA document remains draft guidance — non-binding recommendations, not a rule in force. Devices cleared before it takes effect were validated under the older standard. If a manufacturer claims comparable performance across skin tones, ask what the labelling actually says.

5Everything else that shifts the number

The FDA's own list of factors that can affect pulse oximeter accuracy.1
  1. Poor circulation. Cold hands, low blood pressure or peripheral vascular disease all weaken the pulse signal the device depends on.
  2. Skin pigmentation. As above — the most studied factor, and the one being regulated.
  3. Skin thickness and temperature. Both change how light travels through the tissue.
  4. Nail polish and artificial nails. A classic and easily fixed source of error on fingertip devices.
  5. Current tobacco use. Carbon monoxide binds haemoglobin and can leave saturation looking normal when it is not.
  6. Motion and fit. Specific to wrist wearables — a loose band and a moving arm are the two most common causes of a nonsense reading.

6How consumer wearables compare

Four consumer smartwatches, 49 participants, measured against a clinical-grade reference oximeter. Lower error is better.6
Apple Watch Series 72.9 percentage points
Garmin Fenix 6 Pro4.6 percentage points
Withings ScanWatch4.8 percentage points
Garmin Venu 2s6.7 percentage points
All four bars are the same unit: root-mean-square error against the reference device. The same study found 11–31% of attempted readings produced no result at all.

Two things are worth saying plainly about this. First, the spread between the best and worst device was more than twofold — "has an SpO2 sensor" tells you almost nothing about accuracy. Second, this particular study found skin tone had little correlation with error in its sample; that is one study at healthy saturations and does not overturn the clinical literature above, but it is what the data showed and it belongs in the record.

7What a wrist sensor can and cannot do

✓ Reasonable uses

  • Watching your own trend over weeks, where a consistent drop is more informative than any single number
  • Overnight sampling that may prompt you to raise snoring or daytime sleepiness with a doctor
  • Noticing the effect of altitude on a trip
  • Adding context to symptoms you already have

✗ What it must not be used for

  • Deciding whether you need medical care. The margin of error is wider than the decision thresholds
  • Diagnosing sleep apnoea, COPD, asthma or a respiratory infection
  • Titrating home oxygen, or replacing a prescribed oximeter
  • Reassurance when you feel breathless. Symptoms beat the number — breathlessness with a "normal" reading still needs assessment

If you take a reading, take it properly

1. Sit still and rest your arm for a minute first — motion is the biggest single source of junk readings. 2. Warm cold hands. 3. Wear the band snug and one finger-width above the wrist bone. 4. Take two or three readings and look at the pattern, not the first number. 5. Write down what you saw and when, so a clinician has a trend rather than an anecdote.

Browse by category

REXQU Home Health Smartwatches Blood Pressure Watches ECG Smartwatches Smartwatches for Seniors Kids Smartwatches Rugged & Outdoor GPS & Hiking Screenless Fitness Trackers Women's Smartwatches Smartwatches with Flashlight

More health guides

  • Understanding Your Blood Pressure Numbers
  • Can a Smartwatch Detect AFib?
  • Falls in Older Adults: What the Data Says
  • All REXQU health guides

Devices mentioned in this guide

REXQU O2Max, REXQU BioVision, REXQU VitaCore, REXQU AltiTrek — see all at REXQU health smartwatches.

Sources

  1. U.S. Food and Drug Administration — Pulse Oximeter Accuracy and Limitations: FDA Safety Communication (accuracy range, list of factors, prescription vs over-the-counter distinction)
  2. NHS — patient pulse oximeter home-monitoring guidance (reading thresholds)
  3. Sjoding MW, Dickson RP, Iwashyna TJ, Gay SE, Valley TS. Racial Bias in Pulse Oximetry Measurement. New England Journal of Medicine, December 2020 (two cohorts; occult hypoxemia defined as SaO2 <88% with SpO2 92–96%)
  4. Agency for Healthcare Research and Quality — Differential Pulse Oximeter Accuracy, Occult Hypoxemia Prevalence, and Clinical Outcomes by Patient Race/Ethnicity: A Systematic Review (pooled bias, prevalence ratio 1.71, strength-of-evidence ratings)
  5. U.S. Food and Drug Administration — Pulse Oximeters for Medical Purposes: Non-Clinical and Clinical Performance Testing, Labeling, and Premarket Submission Recommendations, draft guidance, January 2025 (Monk Skin Tone Scale, MST groups)
  6. Nelson BW et al. Investigating the accuracy of blood oxygen saturation measurements in common consumer smartwatches. PLOS Digital Health, 2023 (49 participants; RMSE against a clinical reference oximeter)
  7. Anesthesia Patient Safety Foundation — APSF Statement on Pulse Oximetry and Skin Tone (bias by saturation range; historical test-subject requirement; guidance against acting on a single value)
  8. U.S. Food and Drug Administration — FDA Proposes Updated Recommendations to Help Improve Performance of Pulse Oximeters Across Skin Tones, 6 January 2025
Free to share and republish with attribution.
rexqu.com
Not medical advice. This guide is general health information, not a diagnosis or treatment plan. REXQU devices are wellness products, not medical devices, and are not intended to diagnose, treat, cure or prevent any disease. A wrist blood oxygen sensor is not a pulse oximeter cleared for medical use and must not be used to make decisions about medical care or oxygen therapy. If you are breathless or unwell, seek medical help regardless of what any device displays.