Peripheral Oxygen Saturation as a Digital Biomarker

Peripheral oxygen saturation is the proportion of haemoglobin in arterial blood that is bound to oxygen, estimated non invasively from how tissue absorbs light at different wavelengths. Written SpO2 to distinguish it from SaO2 measured directly in blood, it is one of the most widely used measurements in medicine and one of the few vital signs a consumer device attempts to reproduce.

The distinction between the clinical measurement and the wearable estimate matters more here than almost anywhere else in this library. Fingertip transmission pulse oximetry passes light through tissue and is a cleared medical measurement. Wrist devices use reflectance, reading light scattered back from tissue with far less signal to work with, and most position their output as a wellness feature rather than a clinical measurement. Both report a percentage, and that shared unit conceals a substantial difference in reliability.

Clinical pulse oximeters shine red and infrared light through a fingertip or earlobe and compare absorption across the pulsatile component of the signal. Wrist worn devices place emitters and detectors on the same surface and measure backscatter, typically sampling during sleep when the wearer is still, because motion degrades the estimate severely.

Accuracy depends on perfusion, motion, device fit and skin pigmentation. The pigmentation effect is documented rather than theoretical: pulse oximetry overestimates saturation more often in patients with darker skin, meaning hypoxaemia can be missed in exactly the group already facing worse outcomes, and analysis has shown this bias has not improved over three decades of device development. Studies collecting SpO2 should record skin tone and treat the measurement as having population dependent error.

In respiratory research SpO2 is used to characterise disease severity, to detect exacerbation and desaturation, and as a safety parameter. In sleep research, overnight desaturation measures sit alongside the apnoea hypopnoea index and describe the physiological consequence of breathing interruption. In trials unrelated to the lungs, it is frequently collected simply as a safety signal.

Nocturnal SpO2 from wearables has the strongest case, because the wearer is still, the recording spans many hours, and the question is usually about pattern rather than absolute value. Daytime spot readings from a wrist device carry much less information. Across uses, the measure is reported alongside a breathlessness instrument such as the mMRC scale or the Borg CR10, since saturation and the sensation of breathlessness diverge frequently.

Wrist based estimation is less accurate than fingertip oximetry, which is itself less accurate than arterial blood gas analysis. Motion, poor perfusion, cold extremities and loose fit all degrade the reading, and skin pigmentation introduces a systematic bias that has persisted across device generations.

Most consumer implementations are positioned as wellness features and are not cleared medical measurements, so clinical decisions should not rest on them. Saturation is also a late signal in deterioration: it can remain normal while a person is working hard to breathe, which is why it should be interpreted alongside breathing rate and symptoms rather than on its own.

• Sjoding MW, et al. Racial bias in pulse oximetry measurement. N Engl J Med. 2020. pubmed.ncbi.nlm.nih.gov
• Tobin MJ, et al. Inaccuracy of pulse oximetry in darker-skinned patients is unchanged across 32 years. Eur Respir J. 2022. pubmed.ncbi.nlm.nih.gov
• Jubran A. Pulse oximetry. Crit Care. 2015. pubmed.ncbi.nlm.nih.gov
• Luks AM, Swenson ER. Pulse oximetry for monitoring patients with COVID-19 at home. Ann Am Thorac Soc. 2020. pubmed.ncbi.nlm.nih.gov