Does skin tone affect wearable heart rate and blood oxygen accuracy?

This guide pulls together what the research on skin pigmentation and optical sensors actually shows, separate from two related questions covered elsewhere: why wrist heart rate loses accuracy during exercise (a motion and lag story that applies regardless of skin tone) and how trustworthy a smartwatch's blood oxygen reading is overall (a use-case story that applies regardless of who's wearing it). This one is about the skin-tone variable specifically, in isolation.

How an optical sensor "sees" through skin in the first place

Nearly every consumer heart rate sensor — Apple Watch, WHOOP, Garmin, Fitbit, Oura, Samsung — uses photoplethysmography (PPG). LEDs built into the back of the device shine light into the skin, and a photodiode measures how much of that light reflects back. Blood absorbs more light when a pulse of it arrives at the vessel, so the reflected signal rises and falls with every heartbeat, and the algorithm converts that oscillation into a heart rate. Blood oxygen sensors work on the same basic principle, but compare two wavelengths (red and infrared) because oxygenated and deoxygenated hemoglobin absorb each one differently.

Melanin — the pigment that makes skin darker — absorbs light at many of the same wavelengths these sensors use, most notably the green light favored for heart rate because it's cheap to produce and detects pulsatile blood flow well in average conditions. More melanin in the light's path means less of it reaches the blood vessels under the skin and less of a clean signal bounces back, which gives the device's algorithm a weaker, noisier starting point before any calculation happens. That's the physical mechanism behind every skin-tone accuracy question in this space — it isn't a software bug, it's an optical physics problem that software has to compensate for.

The blood-oxygen case is the clearest and most established

Pulse oximetry's skin-tone bias has the longest research history of any metric covered here, almost entirely from clinical fingertip devices rather than consumer wearables. A widely cited 2020 New England Journal of Medicine study found that hospital-grade pulse oximeters missed clinically significant hypoxemia (dangerously low oxygen that a blood draw would have caught) roughly three times more often in Black patients than in white patients reading the same device the same way. The FDA has since named skin pigmentation directly, alongside motion and poor circulation, as a known source of pulse oximeter error in a public safety communication.

Consumer wrist and ring devices use the same optical principle, but the FDA has also noted that most accuracy validation for these specific products hasn't included enough participants with darker skin tones to know precisely how large the gap is on any given watch or ring. The direction of the clinical finding — readings running falsely reassuring, not falsely alarming, in darker skin — is the reasonable working assumption until that testing catches up. The full accuracy picture for consumer SpO2, including where the feature is genuinely useful despite this bias, is covered in more depth elsewhere.

The heart rate evidence is a genuinely mixed picture — here's what actually happened in each study

Unlike blood oxygen, heart rate research on skin tone doesn't point in one consistent direction. Several studies have looked for an effect and found essentially none; others, mostly ones specifically designed around exercise intensity, found a real one.

Two things stand out across that table. First, the effect, where it exists, concentrates in exercise, not rest — the same pattern seen with motion artifact generally, which may be compounding rather than coincidental: a weaker baseline signal from reduced light penetration has less margin to spare when motion and intensity are also degrading the reading. Second, the size of the effect varies a lot by device — the 2026 four-brand comparison found Apple and Garmin noticeably more resilient than Fitbit and Samsung in the same conditions, which argues against treating "wearables" as one undifferentiated category on this question.

Why the studies don't agree with each other

A few real methodological differences explain most of the disagreement, rather than one set of researchers being wrong:

• Rest versus exercise. Studies confined mostly to resting or light-activity conditions (Duke, the 2025 Garmin/Polar comparison) tend to find little or no skin-tone effect; studies built specifically around higher-intensity exercise tend to find one. A sensor with just enough signal margin at rest can run out of margin once motion and intensity are added on top of reduced light penetration.
• Device and algorithm differences. Manufacturers use different wavelengths, different numbers of LEDs and photodiodes, and different proprietary signal-processing algorithms to compensate for noisy input. The 2026 four-brand study is the clearest evidence that this isn't a single industry-wide number — some algorithms compensate for weak signal better than others.
• Sample size and population. Several of these studies run 50-60 participants, which is enough to detect a large effect but can miss a smaller, real one — and a null result in a smaller study isn't the same as proof of no effect.
• How skin tone itself is measured. Most of this research uses self-reported Fitzpatrick scale (a six-category scale originally designed for sunburn/UV sensitivity, not for optical sensor calibration), which is a coarser and less objective measure than something like a spectrophotometer reading of actual melanin content — a limitation several of the papers themselves acknowledge.

The equity dimension this research keeps surfacing

None of this research frames itself as an argument against wearables generally — the Duke authors' main message was that activity type matters more than skin tone for the device they tested, and newer studies finding a real exercise-intensity effect still found some devices handling it much better than others. The recurring theme across the field instead is a fairness-in-measurement argument: if a sensor's error is correlated with a demographic characteristic rather than randomly distributed, the people least served by a device are disproportionately the people relying on health data that's quietly less reliable for them — the same equity concern the FDA raised about clinical pulse oximeters applies, in a less-studied but structurally similar form, to consumer wearables. That's a reason for the industry to keep validating across skin tones as cheap LED sensors keep improving, not a reason to distrust every reading today.

A practical way to check your own readings

1. Spot-check against a method that doesn't depend on optical skin sensing. A chest-strap ECG monitor or a manual 30-second pulse count (times two) gives you an independent reference that isn't subject to the same bias, at rest and during a hard effort both.
2. Run the check during intensity, not just at rest. Every study above that found a real gap found it specifically during exercise — a quiet morning spot-check at rest is the condition least likely to reveal a problem even if one exists for you during a workout.
3. Watch for your device's own confidence or signal-quality flag, if it has one. Many platforms silently downweight or discard low-confidence readings rather than displaying them — a string of missing data points during hard efforts is itself informative.
4. Judge trends, not single readings. A heart rate number that looks implausibly low or high for one interval, out of an otherwise consistent session, is more likely sensor noise than a sudden cardiac change — the same logic that applies to cadence lock and other exercise-specific artifacts.
5. Treat any blood-oxygen reading as a trend indicator, not a precise number, especially overnight averages rather than single daytime checks — SpO2's accuracy story is its own, more settled topic with its own caveats.

When it stops being a measurement question

A genuinely irregular heartbeat, chest pain, fainting, or breathlessness at rest are reasons to see a doctor and to trust a resting heart rate reading that's corroborated by how you feel, regardless of what any single wearable displays — a device disagreeing with another device is a sensor-accuracy story; your own body disagreeing with how it normally behaves is a different one.

The bottom line

Blood oxygen bias by skin tone is well established; heart rate bias is real under specific conditions (mainly higher-intensity exercise) and device-dependent rather than universal, with some brands showing meaningfully less error than others in the one direct four-brand comparison run so far. Vita pulls heart rate, HRV and SpO2 from whichever source you connect — WHOOP, Apple Health, or both — into one recovery picture built to weight your own trend over time rather than any single noisy reading, and if a number looks off, Vita's AI coach can help you reason through whether it's worth a spot-check or just normal day-to-day noise.