PPG Skin Tone Bias: How Melanin Affects Accuracy in Pulse Oximeters and Wearables
**Pulse oximeters and PPG-based wearables systematically overestimate blood oxygen saturation in individuals with darker skin pigmentation, a bias wit...
- SpO2 & Oximetry
Pulse oximeters and PPG-based wearables systematically overestimate blood oxygen saturation in individuals with darker skin pigmentation, a bias with measurable clinical consequences that has persisted for decades despite being well-documented in the scientific literature. This is not a theoretical concern: Sjoding et al. (2020) demonstrated in a landmark study of over 10,000 paired measurements that Black patients had nearly three times the rate of occult hypoxemia compared to White patients when relying on pulse oximeter readings. Understanding the optical physics behind this bias, its clinical magnitude, and the engineering approaches to mitigate it is essential for anyone designing, deploying, or relying on PPG-based physiological monitoring.
This article examines the biophysics of melanin-light interaction, quantifies the accuracy disparities from published clinical data, reviews the regulatory landscape, and discusses technical solutions for more equitable PPG device design. For background on how PPG technology works, see our introduction to PPG technology.
The Biophysics of Melanin and Light Absorption
Melanin is a broadband chromophore concentrated in the epidermis, the outermost layer of skin. Its primary biological function is photoprotection through ultraviolet light absorption, but it absorbs across the entire visible and near-infrared spectrum relevant to PPG measurements. Two forms are relevant: eumelanin (brown-black, dominant in darker skin) and pheomelanin (yellow-red, more prominent in lighter skin).
The absorption spectrum of melanin decreases steeply and monotonically with increasing wavelength. Tissue-optics reviews model it as a power law in wavelength (Jacques, 2013, doi:10.1088/0031-9155/58/11/R37).
This means melanin absorbs substantially more at shorter wavelengths (green, 530nm) than at longer wavelengths (red, 660nm) or near-infrared (940nm). Darker skin carries much more epidermal melanin, so the extra attenuation is largest for green light and smaller, though still present, in the infrared.
The practical consequence is that melanin acts as an optical filter that attenuates the PPG signal before it reaches the blood-containing dermis and after it returns from tissue to the photodetector. In reflectance-mode PPG (used in wrist wearables), the light passes through the epidermis twice (in and out), doubling the melanin attenuation effect. In transmission-mode PPG (fingertip clip pulse oximeters), the path traverses the epidermis at entry and exit as well.
Critically, melanin attenuation affects the DC (baseline) and AC (pulsatile) components of the PPG signal differently. The DC component, which represents total light absorption by tissue, blood, and melanin, increases substantially with melanin content. The AC component, representing the cardiac-driven pulsatile blood volume change, is attenuated proportionally. This reduces the AC/DC ratio (perfusion index), which is the fundamental quantity from which heart rate and SpO2 are derived.
How Melanin Biases SpO2 Measurement
Pulse oximetry calculates blood oxygen saturation from the ratio of ratios (R):
This ratio is mapped to SpO2 through empirical calibration curves derived from controlled desaturation studies in human volunteers. The critical assumption is that R depends only on the relative concentrations of oxyhemoglobin and deoxyhemoglobin. Melanin violates this assumption.
Because melanin absorbs more at 660nm (red) than at 940nm (infrared), it disproportionately attenuates the red channel's DC component. This alters the R value in a direction that mimics higher oxygen saturation (lower R corresponds to higher SpO2 on the calibration curve). The effect is subtle but clinically significant: a true SaO2 of 88% might produce a pulse oximeter reading of 92-96% in a patient with high melanin content.
Bickler et al. (2005) (doi:10.1097/00000542-200504000-00004) tested three finger pulse oximeters during stable desaturation plateaus (SaO2 60-100%) in 11 subjects with darkly pigmented skin and 10 with light skin. At 60-70% SaO2, the oximeters overestimated saturation by an average of 3.56% in the darkly pigmented subjects, compared with 0.37% in the lightly pigmented subjects. The bias appeared at low saturations, which is precisely when accurate measurement matters most.
Feiner et al. (2007) (doi:10.1213/01.ane.0000285988.35174.d9) extended this work in 36 subjects with a range of skin tones, testing clip-on and adhesive probes for three oximeters (Masimo Radical, Nellcor N-595, Nonin 9700). Dark skin increased bias at low SaO2, adhesive/disposable sensors showed more bias than clip-on types, and estimates differed by up to 10% between instruments in dark-skinned subjects at low SaO2. SaO2 level, sensor type, skin color, and gender all predicted SpO2 error below 80% saturation.
Clinical Consequence: Occult Hypoxemia
The seminal study by Sjoding et al. (2020) (doi:10.1056/NEJMc2029240), published in the New England Journal of Medicine, brought the clinical impact of pulse oximetry bias to broad attention. Analyzing 10,789 paired pulse oximetry and arterial blood gas measurements from inpatients receiving supplemental oxygen at the University of Michigan, plus 37,308 pairs from intensive care units at 178 hospitals, they found:
- Occult hypoxemia (SaO2 < 88% with SpO2 92-96%) in the Michigan cohort: 11.7% of measurements in Black patients vs. 3.6% in White patients
- The absolute difference was 8.1 percentage points, meaning roughly 1 in 9 SpO2 readings of 92-96% in Black patients corresponded to true hypoxemia
- The multicenter cohort showed the same pattern (17.0% vs. 6.2%)
These findings have direct treatment implications. Clinical guidelines use SpO2 thresholds of 94% (general wards) and 88-92% (COPD patients) to guide oxygen therapy decisions. If a pulse oximeter reads 94% but the true SaO2 is 87%, the patient may not receive necessary oxygen supplementation or escalation of care.
Subsequent studies have confirmed and extended these findings. Fawzy et al. (2022) (doi:10.1001/jamainternmed.2022.1906) analyzed 32,282 paired measurements from 1,216 patients with COVID-19 in the Johns Hopkins Health System. Occult hypoxemia occurred in 30.2% of Asian, 28.5% of Black, and 29.8% of non-Black Hispanic patients, compared with 17.2% of White patients. Black and Hispanic patients also had significantly delayed or unrecognized eligibility for oxygen-threshold-based COVID-19 therapies, showing that the bias had concrete consequences for treatment.
Impact on Wearable Heart Rate Accuracy
Skin tone bias in PPG extends beyond SpO2 to affect heart rate accuracy, particularly during motion. The evidence for heart rate is mixed, however. Bent et al. (2020) (doi:10.1038/s41746-020-0226-6) systematically tested consumer and research-grade wearables across the full range of skin tones and found no statistically significant difference in heart rate accuracy across skin tones. They did find significant differences between devices and between activities: absolute error during activity was on average 30% higher than at rest. In contrast, Shcherbina et al. (2017) (doi:10.3390/jpm7020003) reported higher device error with darker skin tone, along with higher error in males, at higher BMI, and during walking.
Where a heart rate effect exists, the mechanism differs from SpO2 bias: the issue is signal-to-noise ratio. Melanin attenuation reduces the PPG signal amplitude, lowering the SNR and making the pulsatile signal more susceptible to corruption by motion artifacts. During exercise, when motion artifact power can exceed the pulsatile signal by 10-100 times, the lower baseline SNR in darker skin means the cardiac signal is lost more frequently, leading to higher error rates and more dropout periods.
Colvonen et al. (2020) (doi:10.1093/sleep/zsaa159) argued in Sleep that green-light PPG wearables used in sleep and health research may be less accurate on darker skin. They called for researchers to report participants' skin tone and to validate devices across skin tones, so that wearable-based research does not widen health disparities.
For a detailed comparison of how different PPG wavelengths interact with skin pigmentation, see our guide on green vs red vs infrared PPG.
Regulatory and Standards Landscape
Historical Context
Pulse oximeter accuracy is validated with desaturation studies on healthy volunteers under ISO 80601-2-61. FDA's 2013 510(k) guidance added only a minimal pigmentation requirement: 10 or more healthy subjects, 200 or more paired SpO2-SaO2 data points, and "at least 2 darkly pigmented subjects or 15% of your subject pool, whichever is larger" (FDA, 2013). FDA later noted that this proportion was based on U.S. census race distribution rather than on skin tone.
Validation studies were often conducted predominantly on light-skinned subjects. This created a systematic validation gap where devices could meet regulatory requirements while failing to perform adequately for a substantial portion of the population.
FDA Response
On November 1, 2022, the FDA convened its Anesthesiology and Respiratory Therapy Devices Panel. The panel concluded that clinical evidence showed disparate pulse oximeter performance in patients with dark skin pigmentation, and it recommended standardized skin pigmentation assessment and tighter overall accuracy. The FDA published a discussion paper in November 2023 and reconvened the panel on February 2, 2024.
On January 7, 2025, the FDA issued draft guidance, "Pulse Oximeters for Medical Purposes: Non-Clinical and Clinical Performance Testing, Labeling, and Premarket Submission Recommendations." It recommends larger and more diverse clinical studies that assess skin pigmentation both subjectively (the 10-point Monk Skin Tone scale) and objectively (Individual Typology Angle at the sensor site), as well as analyses showing non-disparate performance across pigmentation groups.
Engineering Solutions and Mitigation Strategies
Multi-Wavelength Approaches
Adding wavelengths beyond the standard red (660nm) and infrared (940nm) pair enables explicit estimation of melanin absorption. A third wavelength at a point where melanin absorbs but hemoglobin species do not differentiate (such as 730nm or an isosbestic wavelength) can provide an independent melanin correction factor.
Masimo's co-oximetry platform uses 7+ wavelengths to separately quantify oxyhemoglobin, deoxyhemoglobin, carboxyhemoglobin, methemoglobin, and total hemoglobin. Additional wavelengths could, in principle, support melanin correction because the system solves a multi-variable optical model rather than relying on a single R-ratio, though such devices are substantially more expensive.
For wearable applications using green-light PPG, adding an infrared channel specifically for melanin estimation has been proposed. The ratio of green to infrared DC levels varies with epidermal melanin content, which could enable adaptive signal processing that adjusts filtering parameters and detection thresholds based on estimated melanin level.
Adaptive Calibration
Skin-tone-adaptive calibration modifies the R-ratio-to-SpO2 lookup table based on an estimate of the subject's melanin content. This can be derived from the DC signal level at one or more wavelengths, from a user-input skin tone setting, or from a brief initial calibration period. The challenge is implementing this adaptively without requiring manual input.
Hardware Design Optimization
Increasing LED optical power can partially compensate for melanin attenuation by driving higher photocurrent at the detector. However, this increases power consumption and thermal dissipation. For wearable applications, LED power is typically constrained to 5-25 mW for green and 1-5 mW for infrared to maintain acceptable battery life.
Optimizing the LED-photodetector geometry can help. Smaller source-detector separations favor shallow tissue sampling, which reduces the relative melanin path length. However, this also reduces pulsatile signal depth and sensitivity to deeper arterial pulsations. The optimal geometry represents a compromise that varies with skin pigmentation.
Increasing photodetector area or using avalanche photodiodes (APDs) with internal gain can improve SNR without increasing LED power. This approach is more practical for clinical devices than for consumer wearables, where component cost and power constraints are tighter. For more on how sensor design affects signal quality, see our algorithms and signal processing resources.
Algorithmic Approaches
Signal processing algorithms can be designed to be more robust to low-SNR conditions. Adaptive filtering with automatically adjusting thresholds, ensemble averaging over more cardiac cycles (reducing random noise at the expense of temporal resolution), and machine learning models trained on diverse populations all contribute to reduced skin tone bias.
Importantly, algorithms trained predominantly on data from light-skinned subjects will inherently perform worse on darker skin because the noise characteristics and signal morphology differ. Training data diversity is as critical as algorithmic sophistication. The MIMIC-III waveform database, widely used for PPG algorithm development, has limited skin tone metadata, making it difficult to ensure balanced training.
Path Forward for Equitable PPG Technology
Addressing PPG skin tone bias requires coordinated efforts across hardware design, algorithm development, clinical validation, and regulatory requirements. The technical solutions exist: multi-wavelength sensing, adaptive calibration, and diverse training data can substantially reduce the bias. The barriers are primarily economic (multi-wavelength sensors cost more), historical (existing calibration curves were built on biased datasets), and regulatory (mandatory skin tone diversity requirements are only now being implemented).
For researchers and engineers working with PPG, the immediate priorities should include reporting accuracy metrics stratified by skin pigmentation in all validation studies, ensuring training and test datasets include adequate representation across skin tones, designing sensor hardware with infrared channels that can serve dual roles in SpO2 measurement and melanin estimation, and advocating for updated standards that mandate equitable device performance.
The clinical stakes are high. PPG technology is expanding from hospital pulse oximeters to billions of consumer wearable devices used for health monitoring, sleep tracking, and chronic disease management. If these devices systematically underperform for people with darker skin, they risk widening rather than narrowing health disparities. Getting the engineering right is both a technical challenge and an ethical imperative.
For additional context on how different PPG conditions affect measurement accuracy, explore our conditions reference.
References
- Bickler, P. E., Feiner, J. R., & Severinghaus, J. W. (2005). Effects of skin pigmentation on pulse oximeter accuracy at low saturation. Anesthesiology, 102(4), 715-719. https://doi.org/10.1097/00000542-200504000-00004
- Feiner, J. R., Severinghaus, J. W., & Bickler, P. E. (2007). Dark skin decreases the accuracy of pulse oximeters at low oxygen saturation: the effects of oximeter probe type and gender. Anesthesia & Analgesia, 105(6 Suppl), S18-S23. https://doi.org/10.1213/01.ane.0000285988.35174.d9
- Sjoding, M. W., Dickson, R. P., Iwashyna, T. J., Gay, S. E., & Valley, T. S. (2020). Racial bias in pulse oximetry measurement. New England Journal of Medicine, 383(25), 2477-2478. https://doi.org/10.1056/NEJMc2029240
- Fawzy, A., et al. (2022). Racial and ethnic discrepancy in pulse oximetry and delayed identification of treatment eligibility among patients with COVID-19. JAMA Internal Medicine, 182(7), 730-738. https://doi.org/10.1001/jamainternmed.2022.1906
- Bent, B., Goldstein, B. A., Kibbe, W. A., & Dunn, J. P. (2020). Investigating sources of inaccuracy in wearable optical heart rate sensors. npj Digital Medicine, 3, 18. https://doi.org/10.1038/s41746-020-0226-6
- Shcherbina, A., et al. (2017). Accuracy in wrist-worn, sensor-based measurements of heart rate and energy expenditure in a diverse cohort. Journal of Personalized Medicine, 7(2), 3. https://doi.org/10.3390/jpm7020003
- Colvonen, P. J., DeYoung, P. N., Bosompra, N. A., & Owens, R. L. (2020). Limiting racial disparities and bias for wearable devices in health science research. Sleep, 43(10), zsaa159. https://doi.org/10.1093/sleep/zsaa159
- Jacques, S. L. (2013). Optical properties of biological tissues: a review. Physics in Medicine and Biology, 58(11), R37-R61. https://doi.org/10.1088/0031-9155/58/11/R37
- U.S. Food and Drug Administration. (2013). Pulse oximeters – Premarket notification submissions [510(k)s]: Guidance for industry and Food and Drug Administration staff. https://www.fda.gov/regulatory-information/search-fda-guidance-documents/pulse-oximeters-premarket-notification-submissions-510ks-guidance-industry-and-food-and-drug
- U.S. Food and Drug Administration. (2025). Pulse oximeters for medical purposes: Non-clinical and clinical performance testing, labeling, and premarket submission recommendations (draft guidance). https://www.fda.gov/regulatory-information/search-fda-guidance-documents/pulse-oximeters-medical-purposes-non-clinical-and-clinical-performance-testing-labeling-and
Related articles
Pulse Oximeter Accuracy in Dark Skin: FDA Guidance and What Clinicians Should Know in 2025 and 2026
Pulse oximeter bias across skin tones remains a live clinical and regulatory issue. This guide explains what the FDA said in 2025, what remained true as of April 21, 2026, and what clinicians and patients should do in practice.
Wearable Heart Rate Accuracy and Skin Tone
Does skin tone affect wearable heart rate accuracy? What studies of PPG sensors and melanin show across Fitzpatrick skin types, and where gaps remain.
Temperature Effects on PPG Signals: How Cold, Heat, and Skin Temperature Alter Accuracy
**Temperature is one of the most impactful environmental confounders in photoplethysmography, capable of reducing PPG signal amplitude by over 90% in ...
Stress Tracker Accuracy Guide: What Wearables Actually Measure
Independent review of wearable stress tracker accuracy covering HRV methods, sensor limitations, and how to interpret stress scores from Garmin, Fitbit, and Apple Watch.
PPG Inter-Beat Interval Accuracy: IBI Precision vs ECG Gold Standard
**Inter-beat interval (IBI) measurement from photoplethysmography achieves mean absolute errors of 2-10 ms compared to ECG under resting conditions, b...