Wearable devices that include pulse oximetry (SpO2) sensing afford the opportunity to capture oxygen saturation measurements from large cohorts in unconstrained living conditions spanning all hours of the day. We report here a cross-sectional analysis of 72 million individual SpO2 values collected from 33,080 individual participants in the Apple Heart and Movement Study, stratified by age, sex, body mass index, home altitude, and other demographic variables. Measurements were aggregated by hour of day for each individual to produce subject-specific 24-hour SpO2 profiles, which were then grouped according to demographic variables for downstream analysis. 24-hour SpO2 profiles exhibited similar circadian patterns for all subgroups, being approximately sinusoidal with nadir near midnight local time, zenith near noon local time, and mean 0.8% lower saturation during overnight hours.
Using SpO2 measurements averaged for each subject into mean nocturnal and daytime SpO2 values, we employed multivariate ordinary least squares regression to quantify population-level trends according to demographic factors including age, BMI, home altitude, and assigned (birth) sex. For the full cohort, regression coefficients obtained from models fit to daytime SpO2 were in close quantitative agreement with the corresponding values from published reference models for awake arterial oxygen saturation measured under controlled laboratory conditions. Compared with daytime SpO2, nocturnal SpO2 regression models yielded higher coefficients of determination and emphasized the effects of age, BMI, and altitude in all subgroups.
Sex-specific subgroup analysis revealed significantly different regression coefficients for age and BMI for females compared with those of males (for both daytime and nocturnal SpO2) although constant terms and altitude regression coefficients did not differ between sexes. Subgroup analysis according to self-reported race/ethnicity was accomplished using two strategies. Incorporating categorical variables encoding self- reported race/ethnicity into the full-cohort linear regression models identified small but statistically signifi- ant differences in daytime SpO2 (largest coefficient corresponding to 0.13% lower SpO2 for Hispanic study participants compared to White participants), but no significant differences between groups for nocturnal SpO2. Additional subgroup analyses comparing separate regression models fit independently to subjects of each race/ethnicity were suggestive of small differences in age- and sex-related trends between groups, but indicated no significant difference in constant terms between any groups for either daytime or nocturnal SpO2 .
The diverse study population, automated background SpO2 measurement collection spanning full 24-hour circadian cycle, and study design utilizing a single smartwatch-enabled sensor type enables the establish- ment of healthy population reference trends outside of clinical settings. Compared with daytime SpO2 measurements, nocturnal SpO2 measurements are systematically lower and emphasize the impact of both demographic factors and altitude. Additionally, the large subject pool in this study supports the detec- tion of small but significant differences in age-dependent decline in SpO2 between sexes under naturalistic conditions.