Performance of a wearable CORE sensor for nocturnal core temperature monitoring during sleep

Wearable sensors are increasingly used to estimate core body temperature (CBT) during sleep, and their validity is commonly assessed using classical Bland-Altman analysis. However, this approach assumes independence between observations and may be inappropriate for continuous, temporally autocorrelated physiological data. Whether reliance on pointwise Bland-Altman analysis leads to inflated agreement estimates during nocturnal sleep has not been systematically examined. We evaluated the validity and reliability of a wearable CORE sensor against rectal thermistor measurements during a controlled 9-h sleep protocol (22:00-07:00) in 26 healthy young adults (16 males & 10 females). Agreement was assessed using a modified Bland-Altman approach based on time-aggregated nocturnal CBT values to account for temporal autocorrelation. Circadian rhythmicity was characterized using cosinor analysis, and test-retest reliability was examined in a subset of 10 males. Classical pointwise Bland-Altman analysis suggested acceptable agreement (>95% of observations within limits of agreement). In contrast, the modified analysis revealed substantial disagreement, with 49.49% of male and 65.67% of female observations falling outside limits of agreement. Females exhibited significantly earlier circadian phase (p = 0.012) and greater amplitude (p = 0.021) compared with males. Test-retest reliability of the wearable sensor was good (ICC = 0.810). These findings demonstrate that classical Bland-Altman analysis overestimates agreement when applied to continuous nocturnal temperature data. Although the wearable sensor showed acceptable reliability and captured circadian trends, it did not accurately measure absolute nocturnal CBT in either sex.