The Impact of Skin Pigmentation on Pulse Oximetry SpO2 and Wearable Pulse Rate Accuracy: A Meta-Analysis

Abstract

Abstract Background: Medicine has used photoplethysmography (PPG) with pulse oximetry devices for decades to assess blood oxygenation (SpO2) and pulse rate (PR) and this technology is now being used in consumer devices. Skin pigmentation may influence accuracy, leading to health outcomes disparities. Methods: This meta-analysis identified 23 pulse oximetry studies with 59,684 participants and 197,353 paired observations between SpO2 and arterial blood and 4 wearable PR studies with 176 participants and 140,771 paired observations between PR and electrocardiography. The primary objectives were to evaluate SpO2 and PR accuracy by skin pigmentation group by comparing SpO2 accuracy root-mean-square (Arms) values to regulatory thresholds of 3% and PR 95% Limits of Agreement (LoA) to American National Standards Institute (ANSI), Advancing Safety in Medical Technology (AAMI), and International Electrotechnical Commision (IEC) Standards of ±5bpm. The secondary objectives were to evaluate biases and clinical relevance using mean bias and 95% confidence intervals (CI). Findings: For SpO2, Arms was 3·96%, 4·71%, and 4·15% and the pooled mean bias was 0·70% (95% CI: 0·17 to 1·22), 0·27% (95% CI: -0·64 to 1·19), and 1·27% (95% CI: 0·58 to 1·95) for light, medium, and dark pigmentation, respectively. For PR, the 95% LoA were -16.02 to 13.54, -18.62 to 16·84, and -33.69 to 32.54 and the pooled mean bias was -1·24 bpm (95% CI: -5·31-2·83), -0·89 bpm (95% CI: -3·70-1·93), and -0·57 bpm (95% CI: -9·44-8·29) for light, medium, and dark pigmentation, respectively. Interpretation: The current meta-analysis suggests overall inaccurate SpO2 and PR measurements across all skin pigmentation groups as they exceed FDA guidance and ANSI standard thresholds. Pulse oximeters also exhibit statistically significant overestimation of SpO2 for light and dark skin pigmentation, but no clinically relevant bias. Wearable PR exhibits no statistically significant or clinically relevant bias.