Agreement of consumer wearable pulse oximeters across activity conditions and skin pigmentation

Introduction Wearable pulse oximetry devices are increasingly used in consumer-grade health products and patient monitoring; however, concerns remain regarding their accuracy across diverse populations and non-ideal conditions. Methods This study evaluated the agreement of four consumer wearable pulse oximeters with two clinical comparator monitors in a cohort of 49 adults during seated rest, treadmill walking, and seated recovery conditions. Skin tone was assessed using both the visual Fitzpatrick scale and the Individual Typology Angle (ITA), derived from objective skin color measurements. Device performance was evaluated using signed and absolute measurement error, generalized estimating equations accounting for repeated measurements, and repeated-measures Bland–Altman analysis. Results Agreement varied by device and activity condition. Garmin showed a significant negative signed bias relative to the GE Dash during cooldown (−0.72 percentage points, 95% CI −1.19 to −0.25; Holm-adjusted p = 0.028), while Wellue showed a significant negative signed bias during treadmill activity (−1.03 percentage points, 95% CI −1.52 to −0.53; Holm-adjusted p < 0.001). Wellue also demonstrated significant condition-dependent changes between treadmill and cooldown for both signed and absolute error. No device-specific association between skin pigmentation and signed or absolute error remained significant after correction for multiple comparisons using either continuous ITA or Fitzpatrick grouping. Measurement availability was substantially reduced for Garmin during treadmill activity (82.4%) and was nearly absent for the Apple Watch during treadmill activity (0.4%). Discussion These findings demonstrate that agreement between wearable and clinical pulse oximeters varies by device and activity state, while no statistically significant pigmentation-associated effects were identified after correction for multiple comparisons. Incorporating objective skin tone metrics, diverse participant populations, and activity-aware evaluation protocols will be essential for improving device reliability, supporting equitable digital health technologies, and reducing disparities in wearable-based physiological monitoring.