Zhang, B., Fedor, S., Stickgold, R., Freedman, M., Nathan, V., Viswanath, V. K., Rahman, S., Duffy, J. F., Czeisler, C. A., & Picard, R. W. (2025, June). Estimating central circadian phase with wrist-worn spectrophotometer in ambulatory environment. Presented at SLEEP 2025: 38th Annual Meeting of the Associated Professional Sleep Societies, Seattle, WA, United States.

Abstract

Introduction: The central circadian pacemaker regulates essential physiological processes from cardiometabolism to mood and neurobehavioral performance. Accurate estimation of central circadian phase has broad implications for improving public health. Traditional laboratory-based methods are time- and resource-intensive. Advances in wearable technology, including wrist-worn spectrophotometers, enable continuous, non-invasive monitoring of light exposure with spectral information. This study evaluates the feasibility of using melanopic irradiance measured on the wrist to estimate central circadian phase in ambulatory settings.

Results: Melanopic irradiance-based estimation outperformed photopic illuminance- and habitual sleep onset-based estimations, achieving mean absolute errors of 40-43 minutes versus 50-52 minutes and 70 minutes, respectively (p<.001, Friedman test). Lin’s concordance correlation coefficients for melanopic irradiance ranged from 0.94-0.95, significantly higher than 0.89-0.90 for photopic illuminance and 0.80 for habitual sleep onset (p<.001, Friedman test). Furthermore, 84% of participants had phase estimation within one hour of their saliva DLMO using melanopic irradiance, a 16% improvement over using photopic illuminance, and a 60% improvement over using habitual sleep onset.

Methods: We monitored healthy non-shift-worker adults (N=25, age: 18-54) wearing a wrist spectrophotometer continuously for two weeks while adhering to their habitual schedules. The devices recorded melanopic irradiance, photopic illuminance, rest-activity data, and off-wrist intervals concurrently. At the end of the two weeks, saliva samples were collected under dim-light conditions to assess dim-light melatonin onset (DLMO), serving as the gold-standard reference for central circadian phase. We evaluated the DLMO estimation performance of three van der Pol oscillator models using melanopic irradiance and photopic illuminance as inputs. Habitual sleep onset minus two hours was used as a baseline comparator.

Conclusions: This study demonstrates the superiority of melanopic irradiance-based methods for estimating central circadian phase in ambulatory environments. The estimation performance improvement of using melanopic irradiance is statistically significant but marginal, and its clinical impact requires further investigation. Future work should also test these methods among shift workers and patients with sleep and circadian rhythm disorders. Overall, the current results support incorporating spectral information, in addition to photopic illuminance and sleep timing, to improve the estimation of central circadian phase from wearables.