What Is the Difference Between SpO2 and SaO2?

SpO2 is an estimate of blood oxygen saturation obtained noninvasively by a pulse oximeter clipped to your finger, while SaO2 is the same measurement taken directly from an arterial blood sample in a lab. They aim to report the same thing: the percentage of hemoglobin in your arterial blood that is carrying oxygen. The difference is in how they get there, and that difference matters more than most people realize, because the two numbers do not always agree.

What Each Number Actually Represents

Both SpO2 and SaO2 describe how saturated your hemoglobin is with oxygen. Hemoglobin is the protein in red blood cells that picks up oxygen in the lungs and delivers it to tissues. If every available hemoglobin binding site is loaded with oxygen, saturation is 100%. In a healthy person breathing room air, saturation typically sits in the mid-to-upper 90s.

SaO2 is measured by drawing blood from an artery, usually the radial artery at the wrist, and running it through a device called a co-oximeter. A co-oximeter shines multiple wavelengths of light through the blood sample and can distinguish between oxygenated hemoglobin, deoxygenated hemoglobin, and abnormal forms like carboxyhemoglobin (hemoglobin bound to carbon monoxide) or methemoglobin. Because it directly analyzes a blood sample, SaO2 is considered the gold standard.

SpO2 is the noninvasive stand-in. A pulse oximeter clips onto your fingertip and shines red and infrared light through the tissue. Oxygenated hemoglobin absorbs more infrared light, while deoxygenated hemoglobin absorbs more red light.¹ The device calculates a ratio from the pulsatile changes in those two signals and maps that ratio onto a calibration curve to produce the SpO2 number on the screen. The “p” in SpO2 stands for “pulse” and exists precisely because this reading is derived from pulse oximetry rather than a direct blood measurement.²

How Pulse Oximeters Are Calibrated

Pulse oximeters are not measuring oxygen saturation from first principles. They are estimating it based on calibration curves built from real human volunteers. During calibration, healthy volunteers breathe lower-than-normal oxygen concentrations while wearing the pulse oximeter sensor. Simultaneously, arterial blood is drawn and analyzed by co-oximetry. The manufacturer records the relationship between the optical ratio the sensor detects and the true SaO2 in the blood sample, and that relationship becomes the lookup table baked into every device of that model.³

There are two important limitations here. First, for ethical reasons, you cannot drop a healthy volunteer’s oxygen saturation dangerously low. Calibration data generally stop around 70–75% saturation. Below that range, the device is extrapolating, and accuracy tends to deteriorate. Second, the volunteer pools used in calibration studies have historically lacked diversity in skin pigmentation, which introduces systematic errors for darker-skinned individuals. Both of these are well-recognized problems in the field.

How Close Are SpO2 and SaO2 in Practice

Manufacturers typically claim pulse oximeters are accurate to within about 2–3% of the true SaO2 value.⁴ In many routine clinical situations, this is good enough. But the agreement is not always tight, and SpO2 tends to overestimate rather than underestimate the true value. A study comparing SpO2 and SaO2 in critical care patients found that SpO2 overestimated SaO2 by about 1% on average, but the 95% limits of agreement ranged from about −5% to +7%.⁵ That means for any individual reading, SpO2 could be as much as 5 points lower or 7 points higher than the true SaO2. When you are making treatment decisions around specific oxygen thresholds, that kind of spread matters.

The two values also tend to diverge more as oxygen levels drop. Under extreme hypoxic conditions simulating high altitude, even a corrected pulse oximeter showed a mean error of about 2% compared with SaO2, though the correlation remained strong.⁶ When hematocrit levels are very low (severe anemia), the error grows further, with some devices showing more than 3% error once oxygen saturation drops below the 70–80% range.⁷

A point worth emphasizing: functional versus fractional saturation can also muddy comparisons. SpO2 reports “functional” saturation, meaning it only looks at the fraction of hemoglobin that can carry oxygen. Co-oximetry can report “fractional” saturation, which includes all hemoglobin, including forms that cannot carry oxygen. In a healthy person, the two are nearly identical. In someone with significant carboxyhemoglobin or methemoglobin, they diverge. Research has found that the consistency between functional and fractional measurements depends on hemoglobin levels and oxygenation status, so treating them as interchangeable is a mistake.⁸

Carbon Monoxide and Methemoglobin

The most dangerous limitation of pulse oximetry involves abnormal hemoglobin variants. Standard pulse oximeters use only two wavelengths of light, which means they can distinguish oxygenated from deoxygenated hemoglobin but cannot detect carboxyhemoglobin or methemoglobin.

In carbon monoxide poisoning, hemoglobin bound to carbon monoxide absorbs light similarly to oxygenated hemoglobin. The pulse oximeter reads it as “loaded with oxygen” when it is actually loaded with carbon monoxide. One study found that SpO2 stayed above 96% even when carboxyhemoglobin levels reached 44%.⁹ An elevated carboxyhemoglobin level falsely elevates the SpO2 reading, typically by roughly the amount of carboxyhemoglobin present.¹⁰ A person could be severely poisoned with a reassuring-looking SpO2 on the monitor. Only a co-oximeter analyzing an actual blood sample can detect carboxyhemoglobin and reveal the true SaO2.

Methemoglobin causes a different pattern. As methemoglobin levels rise, pulse oximeter readings drift toward roughly 85% and plateau there, regardless of how much further methemoglobin increases.¹¹ This means SpO2 can either overestimate or underestimate the real saturation, depending on where the patient’s actual oxygen level sits relative to that ~85% plateau.¹² Again, the only way to sort this out is arterial blood gas analysis with co-oximetry.

Skin Pigmentation and the Accuracy Gap

One of the more consequential findings in recent years is that pulse oximeters systematically overestimate oxygen saturation in people with darker skin. A systematic review of the literature found consistent evidence of this overestimation, particularly when actual oxygen levels are low.¹³ The concern is not just academic. If SpO2 reads 94% on the monitor but the true SaO2 is 89%, a clinician might not recognize that the patient is hypoxemic and needs intervention.

The root cause traces back to calibration. Because calibration studies have historically enrolled predominantly light-skinned volunteers, the calibration curves embedded in many devices are optimized for a narrow range of skin tones. Melanin in the skin absorbs light in the same red and near-infrared wavelengths that pulse oximeters use, and different melanin concentrations alter the signal in ways that the calibration curve does not fully account for.³ Regulatory agencies have taken notice, and there is growing pressure on manufacturers to diversify their calibration cohorts and validate accuracy across skin tones.

Poor Perfusion, Cold Hands, and Motion

Pulse oximeters rely on detecting the pulsatile signal of arterial blood flowing through the tissue under the sensor. Anything that weakens that pulse signal can degrade accuracy. Poor peripheral perfusion, whether from shock, vasoconstriction, or heart failure, is one of the most significant limitations of pulse oximetry.¹⁴

Cold extremities are a particularly common cause of poor readings. In children, skin temperature below 30°C was identified as a significant predictor of oximeter inaccuracy.¹⁵ The same principle applies to adults: if your fingers are cold enough that blood flow is significantly reduced, the pulse oximeter has less signal to work with and the reading becomes unreliable.

Motion artifact is another well-known problem. When your finger moves while the oximeter is clipped on, the light signals get contaminated with noise from venous blood sloshing around in the tissue, tissue compression, and optical coupling changes. Traditional pulse oximeters assumed that the only pulsatile light-absorbing component in the optical path was arterial blood. Motion violates that assumption. Newer signal-processing technologies, such as Masimo’s Signal Extraction Technology, use adaptive filtering and digital signal processing to separate the arterial signal from motion noise and have improved accuracy during movement and low perfusion.¹⁶ Waveform morphology analysis offers another approach to automatically reject artifact-contaminated readings.¹⁷

Nail Polish, Tattoos, and Other Everyday Interferences

People often wonder whether nail polish affects pulse oximetry readings. The evidence is reassuring for most situations. A systematic review found that dark-colored polish (black, blue, brown, or purple) caused small reductions in SpO2 readings, but the differences were clinically insignificant in the vast majority of cases.¹⁸ One study did find that black nail polish blocked the signal entirely in a high proportion of attempts, but that appears to be an outlier finding. In an emergency, clinicians can simply rotate the sensor to a different finger or place it sideways on the nail bed.

Other factors that can influence SpO2 readings include ambient light (bright overhead surgical lights or direct sunlight can interfere with the sensor), electromagnetic interference from nearby equipment, and even sensor placement. A comprehensive review of accuracy-affecting factors noted that high heart rates above 150 beats per minute, high skin temperatures, and the specific location of the sensor on the body can all distort the photoplethysmographic signal.¹⁹ Dark tattoo ink on the finger can also absorb light and reduce accuracy, though this comes up less frequently than other sources of error.

When Clinicians Rely on SaO2 Instead

Given SpO2’s limitations, the question becomes: when is the noninvasive number good enough, and when do you need a blood draw? For routine monitoring of a stable patient, SpO2 is usually sufficient. It gives continuous, real-time data without any needles, and trends in SpO2 over time are often more useful than any single number.

But SaO2 from an arterial blood gas is preferred when the stakes are higher or the clinical picture does not add up. An ICU study found that because SpO2 does not reliably predict SaO2, relying on it alone could leave clinicians uncertain whether a patient truly needs more oxygen or is actually in a hyperoxic state.⁴ The researchers recommended arterial blood gas analysis whenever precise oxygen management is needed. Similarly, when using SpO2 to calculate oxygenation ratios used in severity scoring for pneumonia patients, the noninvasive estimate misclassified roughly 20–30% of patients who fell near clinical decision thresholds.²⁰

An arterial blood gas also provides far more information than just SaO2. It tells you the partial pressure of oxygen and carbon dioxide, blood pH, bicarbonate levels, and the concentrations of abnormal hemoglobin species. None of that is available from a pulse oximeter. In suspected carbon monoxide poisoning, acid-base disorders, or unexplained respiratory failure, there is no substitute for the full picture an arterial blood draw provides.

The Risks of Arterial Blood Draws

One reason pulse oximetry is so widely used is that arterial puncture, while generally safe, is not risk-free. A large historical cohort study analyzing over 473,000 arterial punctures found a major complication rate of about 0.14%, with embolisms or thrombosis accounting for roughly half of those complications.²¹ Patients on blood-thinning medications had a higher complication rate. Minor complications like hematomas and pain are much more common than major ones. Reducing the total number of arterial blood draws per patient has been shown to allow cost-efficient care with lower complication risk.²²

This trade-off is the practical reality in hospitals: SpO2 monitoring is cheap, painless, and continuous, but approximate. SaO2 is precise and information-rich, but invasive, intermittent, and not without risk. Good clinical care involves knowing which tool fits the moment.

SpO2 in Sickle Cell Disease

Sickle cell disease offers an instructive example of how the SpO2-SaO2 relationship can shift. Sickle hemoglobin has a right-shifted oxygen dissociation curve, meaning it releases oxygen to tissues more readily at a given oxygen tension. Because pulse oximeters were calibrated on people with normal hemoglobin, there was longstanding concern that SpO2 readings might be unreliable in sickle cell patients. Research has shown that SpO2 slightly overestimates the fractional oxyhemoglobin percentage (by about 3.4 percentage points on average) but almost always accurately estimated SaO2, with underestimation averaging only about 1.1 percentage points. The error was never large enough to misclassify a hypoxemic patient as normal or a normal patient as hypoxemic.²³ That is reassuring, though clinicians caring for sickle cell patients during acute chest syndrome still often confirm with co-oximetry, since the clinical stakes during a crisis are high.²⁴

Home Pulse Oximeters

Consumer-grade pulse oximeters became widely purchased during the COVID-19 pandemic. These devices use the same basic two-wavelength principle as hospital-grade models, but their calibration, sensor quality, and signal processing are not always held to the same standards. Hospital devices tend to have better motion-artifact rejection and more robust calibration data. Consumer devices are fine for general trend monitoring (noticing whether your SpO2 is generally in a normal range versus dropping), but should not be trusted to the exact percentage point.

A few practical tips if you are using one at home: warm your hands first, sit still, and wait for the reading to stabilize for at least 10–15 seconds before noting the number. Dark nail polish probably will not cause a meaningful error, but removing it from the measurement finger eliminates one variable. If you consistently get readings below 94% while feeling well, it is worth checking with a clinician rather than assuming the device is wrong. And if you have any condition that involves abnormal hemoglobin, remember that the device is blind to those variants and the number on the screen may not mean what you think it means.