What's the Difference Between SaO2 and SpO2

SaO2 is oxygen saturation measured directly from an arterial blood sample, an arterial blood gas test, usually run in a lab with a co-oximeter. It is a blood test, which is the real difference between the two numbers. SpO2 is an estimate of that same saturation, produced by a pulse oximeter shining light through tissue and calculating a number from the pattern it detects. Research directly comparing the two under controlled hypoxia has repeatedly found that SpO2 tends to run close to SaO2 at normal saturation but can diverge, usually overestimating, as saturation drops or under certain skin and perfusion conditions.

Why these two numbers get mixed up

Someone looks at a hospital monitor, or a finger clip, or a wearable screen, and sees a percentage labeled oxygen saturation. Somewhere else, a blood draw report lists a different-sounding oxygen saturation number. The instinct is to assume they're the same measurement, just from different machines. They're related, but they're not produced the same way, and that distinction turns out to matter more than it looks.

The confusion isn't really about terminology. It's about trust: which number do you believe when they don't match, and does the gap mean anything clinically, or is it just measurement noise.

SaO2, SpO2 and PaO2 side by side

The three numbers come from two tests. SaO2 and PaO2 both come from one arterial blood sample; SpO2 comes from a pulse oximeter clipped to a finger or built into a watch or ring. In healthy adults SaO2 and SpO2 both normally read 95 to 100 percent and PaO2 75 to 100 mmHg, and a pulse oximeter reading may be 2 to 4 percent higher or lower than the actual blood oxygen level, according to MedlinePlus, the US National Library of Medicine’s consumer health service. It lists 92 percent or lower as the point to contact a provider and 88 percent or lower as the point to seek immediate care.

What SaO2 actually is

SaO2 comes from a blood sample, typically arterial, run through a co-oximeter in a lab. It's a direct measurement of how much hemoglobin in that sample is carrying oxygen. It doesn't rely on light absorption through skin or tissue, and it isn't affected by motion, nail polish, or finger temperature the way a clip-on sensor can be. That's part of why it works as the reference standard that pulse oximeter studies get checked against.

The tradeoff is obvious. Drawing blood means it's not continuous, and it's not something a device on your wrist or finger can do. Every comparison study in this space, including the classic work here, uses SaO2 as the fixed point and asks how well SpO2 tracks it.

That's also the design behind the studies that induce controlled hypoxia in healthy volunteers, lowering inspired oxygen in stages while drawing blood at each plateau and comparing it to what the pulse oximeter shows in real time.

What SpO2 is estimating, and where it slips

SpO2 is the number a pulse oximeter displays, whether that's a hospital finger clip, an emergency department monitor, or a consumer device. It's calculated from how light at two wavelengths is absorbed as blood pulses through tissue, an approach with well-documented limitations around motion artifact, weak perfusion, temperature, and interfering hemoglobin variants. None of that requires a blood draw, which is exactly why it became so widely used.

So SpO2 is always an inference, not a direct count. Research using four different pulse oximeter models under motion and low perfusion conditions found that all four could still detect hypoxia against an SaO2 reference, but with differences in how well each one performed under stress. Separate work on wrist-worn reflective devices, including a validation study of the Withings ScanWatch against arterial blood samples across a range of induced hypoxia levels, and a study testing the Apple Watch Series 7 against arterial saturation during controlled hypoxia induction, similarly used SaO2 plateaus as the comparison point, and found the wearable readings tended to run high relative to that reference.

Anyone curious about how that estimation process physically works, rather than just how it compares to blood values, might find how wearables actually measure blood oxygen useful background.

Where PaO2 fits, and why it keeps coming up

PaO2 is the third number in this family, and it comes out of the same arterial blood draw as SaO2. The difference is what each one counts. SaO2 is the percentage of hemoglobin carrying oxygen. PaO2 is the partial pressure of oxygen dissolved in the arterial plasma, reported in mmHg or torr rather than as a percentage. One is how full the carriers are, the other is how hard the oxygen is pushing to get onto them.

The two are joined by the oxyhemoglobin dissociation curve, which is S-shaped rather than a straight line, and that shape has a consequence that matters for anyone reading a saturation number. Saturation stops tracking oxygenation once oxygen tension is high: above roughly 100 torr, saturation measurements are not sensitive to changes in PaO2 at all (International Anesthesiology Clinics, 1988).

Read practically, that means a normal saturation reading does not establish that oxygenation is unchanged. PaO2 can fall a long way while saturation still sits at 97 or 98 percent, because the curve is nearly flat through that region. The same review gives the clinical version of the problem: a pulse oximeter can take minutes to reveal an esophageal intubation in a patient who was well preoxygenated, precisely because saturation has that much slack in it before it starts to move.

For a wrist or fingertip reading the limit is worth holding onto. A pulse oximeter estimates saturation, and saturation is the flattest, least sensitive part of the signal. Neither a watch nor a clip-on oximeter reports PaO2, and nothing in a consumer device substitutes for a blood gas when the question is what oxygenation is actually doing.

Where the gap between the two shows up most

The overestimation of SaO2 by SpO2 isn't random. The largest pooled comparison cited here combined 23 pulse oximetry studies, 197,353 SpO2 readings each matched to an arterial blood sample, and found a root-mean-square error of 3.96 percent on light skin, 4.71 on medium and 4.15 on dark, every group above the 3 percent threshold used in FDA guidance and industry standards (Journal of Medical Internet Research, 2024). In 903 intensive care patients in England, with 11,018 SpO2 readings checked against arterial blood, five fingertip oximeters read 0.6 to 1.5 points higher on darker skin, and when arterial saturation was 92 percent or lower, 7.6 to 62.2 percent of readings on darker skin still showed above 94 percent, depending on the device, against 1.2 to 26.9 percent on lighter skin (BMJ, 2026). Research has tied part of it to skin pigmentation, with darkly pigmented skin associated with a larger overestimation at lower saturation levels in controlled desaturation studies. A pediatric reflectance-based study similarly found SpO2 overestimated arterial saturation near 90% across participants, with the effect most pronounced at the lowest reflectance (most pigmented) end of the measured range. Ongoing prospective work, including a multisite pediatric cardiac catheterization study measuring skin color via spectrophotometry against directly sampled arterial saturation, is built specifically to test that relationship further in real clinical settings rather than lab-induced hypoxia.

That thread gets explored in more depth in a piece on whether blood oxygen readings work the same for everyone, since the SaO2-SpO2 gap is where that question actually lives in the data.

Measurement site matters too. A comparison of ear canal SpO2 against the standard finger site found a small average difference between the two locations at rest, along with a measurable delay between central and peripheral circulation during breath-holding. Separate issue from the SaO2 comparison, but still relevant to how any SpO2 reading should be read in context.

How closely smartwatch readings match arterial blood, model by model, is rated in KnowYourPrime Evidence Registry finding KYP-ER-0001, and the skin tone question in finding KYP-ER-0002.

Carbon monoxide: when SpO2 reads normal and SaO2 does not

Carbon monoxide is the clearest case where the two numbers are different measurements rather than a measurement and its estimate. A co-oximeter counts oxygen-bound and carbon monoxide-bound hemoglobin separately. A pulse oximeter reads the carbon monoxide-bound kind as if it were carrying oxygen, so in carbon monoxide poisoning it can show a normal number while the blood carries far less oxygen.

In 124 patients evaluated for carbon monoxide poisoning at a US hyperbaric center, measured oxyhemoglobin averaged 88.7 percent and fell as carboxyhemoglobin rose, while finger pulse oximetry averaged 99.2 percent and never read below 92. The gap between the two approximated the carboxyhemoglobin level (Annals of Emergency Medicine, 1997). In 16 exposed adults, pulse oximetry did not fall below 96 percent with carboxyhemoglobin as high as 44 percent (Annals of Emergency Medicine, 1994), and in 30 patients with severe poisoning it consistently overestimated arterial oxygen saturation, leaving laboratory co-oximetry of arterial blood as the only accurate measure (Chest, 1998).

Watches and rings estimate blood oxygen with the same light-based principle. None of the studies cited here tested a consumer wearable during carbon monoxide exposure.

Why this distinction keeps mattering

Clinical decisions get built on the assumption that SpO2 is a close enough stand-in for SaO2, and that's the reason this distinction isn't just a technical footnote. Neonatal intensive care units, for instance, set SpO2 target ranges for preterm infants precisely because continuous arterial sampling isn't practical. A survey of European NICUs found wide variation in what those SpO2 targets actually were, with most units having shifted their limits over the preceding decade. That variation is itself indirect evidence of how much judgment goes into translating an SpO2 estimate into a clinical threshold.

For anyone tracking overnight readings on a personal device, the same SaO2-versus-SpO2 distinction is behind a lot of the confusion about what those numbers mean during sleep, a topic covered separately in how accurate wearable blood oxygen readings are during sleep.

Common questions

KnowYourPrime Evidence Registry: Finding KYP-ER-0001: Smartwatch Blood Oxygen (SpO2) Accuracy Against Arterial Blood (SaO2) · Finding KYP-ER-0023: Apple Watch Blood Oxygen Accuracy Against Medical Devices · Finding KYP-ER-0019: Blood Oxygen Readings With Poor Circulation