Pulse oximeters miss the mark on every skin tone, not just darker ones

Pulse oximeters and wrist wearables both rely on the same trick: shining light through the skin to read what the blood is doing, a technology called photoplethysmography. Because skin pigment absorbs light, there has been growing concern that these devices read less accurately on darker skin, which could feed real disparities in health outcomes if clinicians and users trust flawed numbers. Until now, much of the debate has rested on individual studies pointing in different directions. This review set out to pool the evidence and ask two questions at once: how accurate are blood oxygen readings across skin pigmentation groups, and does the pulse rate from consumer wearables suffer the same problem?

The researchers identified 23 pulse oximetry studies covering 59,684 people, with 197,353 paired comparisons between oximeter readings and arterial blood, the gold standard. They also found 4 wearable pulse rate studies covering 176 people, with 140,771 paired comparisons against electrocardiography. For blood oxygen, accuracy was judged against the 3% regulatory threshold used by the US Food and Drug Administration and industry standards bodies. For pulse rate, the benchmark was agreement within plus or minus 5 beats per minute. Results were analysed separately for light, medium and dark skin pigmentation groups.

Blood oxygen accuracy breached the 3% threshold in every pigmentation group, with error values of 3.96% for light, 4.71% for medium and 4.15% for dark skin. Oximeters significantly overestimated oxygen levels in both light skin, by a pooled 0.70%, and dark skin, by 1.27%, though the authors judge this overestimation may not be clinically relevant; the medium group's bias of 0.27% was not significant. For wearable pulse rate, no group showed a statistically or clinically significant bias, though the spread of agreement was widest in dark pigmentation, ranging from -33.69 to 32.54 beats per minute.

This pooled analysis complicates the common framing that pulse oximeters fail only on darker skin: here, readings breached accuracy standards across all pigmentation groups, and overestimation was significant in light skin too. It also offers some reassurance on consumer wearables, where pulse rate showed no pigmentation bias, a question that had far less evidence behind it. Open questions remain, including why the pulse rate data spread so much more widely in dark pigmentation despite no average bias, and how these findings vary between specific devices. The authors' ties to a wearables company are worth keeping in view when weighing the pulse rate conclusion.

If you use a wearable to track your heart rate, this is a reason to think the reading is not systematically skewed by your skin tone, though the spread in the dark pigmentation data means individual readings can still wander. If you own a fingertip pulse oximeter, this is a reason to treat its number as an estimate rather than a precise measurement, whatever your skin colour, since readings tended to flatter the true figure slightly across groups. The pooled overestimation was small and may not matter clinically, but the pattern of readings breaching accuracy standards across all pigmentation groups is worth knowing. None of this replaces a proper clinical measurement when one is needed.