Author(s): HUNTER, N., Institution: STANFORD UNIVERSITY, Country: UNITED STATES, Abstract-ID: 2490

INTRODUCTION:
Homeostasis, a biological phenomenon that allows the human body to quickly adapt to external stimuli and maintain high levels of performance, constantly relies on various internal systems. Negative feedback loops and regulatory systems, such as the cardiovascular system, inter-organ communication, and thermodynamics, exhibit rhythmic and reactive modulation. These systems leave “footprints” in biometric data such as heart rate, temperature, heart rate variability (HRV), and sleep. In women, the balance of these systems can also be observed in reproductive health through menstrual cycle–related data. Athletes are excellent models for studying homeostasis due to their frequent and consistent exposure to both eustress and distress. The understanding of these systems is highly valuable to athletes, as their performance, recovery, and athletic ability are directly tied to optimized internal regulation.
METHODS:
hirty-eight female competitive athletes (across a variety of sports), aged 18-34, completed a one-year longitudinal study involving weekly survey responses and daily biometric data collection. Participants provided health, sport, and lifestyle and were equipped with Oura Rings. Weekly surveys provided information regarding stress, illness, travel, exercise and MC information. Daily biometric data measures included temperature, electrocardiogram data, respiratory rate, and SpO2. Novel Gaussian Process (GP) models were fit to digital phenotypes temperature, resting heart rate (RHR), and HRV. Model outputs were annotated with MC data to evaluate patterns and fluctuations in concert with known internal regulatory shifts. Sleep outcomes (efficiency, RHR, duration, and consistency) were analyzed to determine nightly quality and relatedness to cycle timing.
RESULTS:
Of the full cohort, 28 participants reported at least one occurrence of active bleeding. The median cycle length of this subgroup was 29 days with an interquartile range of 6 days (Q1: 26 days, Q3: 32 days). Twenty-three participants had at least 3 cycles, meeting the inclusion threshold for longitudinal cycle pattern analysis. Across this subgroup, 78% of intra-individual cycle regularity (length SD) fell within the clinically healthy range (0-7 days), while 22% of regularity was at least 8 days. GP model outputs revealed high levels of variability in cycle related temperature spike, with amplitude ranges between .2C and 2C. RHR and HRV digital phenotypes resulted in similar outcomes. A direct relationship was observed between average sleep duration and cycle regularity.
CONCLUSION:
These findings highlight the importance of individualized algorithm development when using digital phenotypes to predict menstrual cycle phase. They also suggest that cycle length regularity may serve as a more informative indicator of health than cycle length alone, the current clinical benchmark. The strength of the relationship between sleep and cycle regularity observed in female athletes represents a novel contribution to the field.