Dados do Trabalho

Título

Application of a mathematical actigraphy-based model to predict Dim Light Melatonin Onset, and its insights into chronotype and sleep patterns

Introdução

Circadian phase can be assessed through biological measures such as melatonin secretion, with Dim Light Melatonin Onset (DLMO) being the gold standard. However, the conditions required for DLMO measurement limit its widespread use, highlighting the need for more practical alternatives, such as predicting DLMO using actigraphy data.

Objetivo

Predict DLMO from actigraphy data and explore its relationship with chronotype, diurnal preference, and key actigraphy components.

Métodos

We analyzed actigraphy data from 402 participants (60% female, mean age 49.5±14.9 years) from São Paulo (EPISONO 4th edition). Data were collected over 5 workdays and 2 free days between August 2018 and April 2019, with 26.6% during daylight saving time (BRST). An open-source mathematical model was used to predict DLMO (pDLMO) using activity, sleep-wake states, and light exposure data from actigraphy. A Principal Component Analysis (PCA) identified 5 components: (1) sleep quantity and opportunity, (2) sleep and rhythm timing, (3) sleep quality and continuity, (4) activity during wakefulness, and (5) sleep regularity and rhythm robustness. Correlation analysis was performed between pDLMO and mid-sleep time corrected for sleep debt (MSFsc), self-reported diurnal preference (Morningness-Eveningness Questionnaire), and the PCA components.

Resultados

The average pDLMO was 7:11 p.m. ± 2:22, with an interquartile range from 4:42 p.m. to 9:00 p.m. The data exhibited a bimodal distribution (Dip test value=0.20; p<0.001), with no significant differences related to BRST (p=0.096), gender (p=0.176), or age (p=0.898). pDLMO was significantly associated with MSFsc (R²=0.026; β=0.161; p<0.001), but not with diurnal preference scores (p=0.185). There was a significant association with pDMLO and the "sleep and rhythm timing" component (R²=0.0286; β=0.169; p<0.001).

Conclusões

We predicted DLMO using an actigraphy-based model in a Brazilian sample across a wide age range. pDLMO was significantly correlated with MSFsc, indicating its potential as a chronotype proxy. Its association with "Sleep and rhythm timing" component suggests alignment with actigraphy-derived sleep and rhythm indicators. The lack of correlation with diurnal preference highlights a possible mismatch between behavioral patterns and circadian preferences. Mathematical models using actigraphy data can be practical tools for estimating circadian phase, but require validation for accuracy and reliability.

Palavras -chave

Accelerometry, Circadian Rhythm, Mathematical model, predicted DLMO