Associations between light exposure and sleep timing and sleepiness while awake in a sample of UK adults in everyday life

Abstract

Experimental and interventional studies show that light can regulate sleep timing and sleepiness while awake by setting the phase of circadian rhythms and supporting alertness. The extent to which differences in light exposure explain variations in sleep and sleepiness within and between individuals in everyday life remains less clear. Here, we establish a method to address this deficit, incorporating an open-source wearable wrist-worn light logger (SpectraWear) and smartphone-based online data collection. We use it to simultaneously record longitudinal light exposure (in melanopic equivalent daylight illuminance), sleep timing, and subjective alertness over seven days in a convenience sample of 59 UK adults without externally imposed circadian challenge (e.g., shift work or jetlag). Participants reliably had strong daily rhythms in light exposure but frequently were exposed to less light during the daytime and more light in pre-bedtime and sleep episodes than recommended [T. M. Brown et al., PLoS Biol. 20, e3001571 (2022)]. Prior light exposure over several hours was associated with lower subjective sleepiness with, in particular, brighter light in the late sleep episode and after wake linked to reduced early morning sleepiness (sleep inertia). Higher pre-bedtime light exposure was associated with longer sleep onset latency. Early sleep timing was correlated with more reproducible and robust daily patterns of light exposure and higher daytime/lower night-time light exposure. Our study establishes a method for collecting longitudinal sleep and health/performance data in everyday life and provides evidence of associations between light exposure and important determinants of sleep health and performance.

Keywords: light; melanopic; sleep; sleepiness.

Fig. 1.

Light exposure in everyday life. (A) The difference between melanopic and photopic illuminances (log lx) across different times of day (clock time). The blue line shows the mean with SD. The red dotted line represents the expected relationship for ‘standard’ daylight. (B) Daily profiles of mean melanopic EDI (lx) as a function of clock time, depicted as a single smoothed line for each of 59 individuals. Individuals were represented by blue lowess smoothed fit lines. The x axis shows clock times. (C) Melanopic EDI (log lx) daily profiles (clock times) for workdays (red) and free days (blue). The lines show the mean with SD. (D) Frequency distribution of melanopic EDI as a function of time of day (0.1 h bins). The cumulative frequency of observations is colored according to the recent recommendations (33): melanopic EDI below 1 lx (black), between 1 and 10 lx (dark gray), between 10 and 250 lx (gray), and values above 250 lx (yellow). (E) Violin plots for exposure durations (hours) in three conditions: melanopic EDI below 250 lx in 8:00 to 17:00 daytime (yellow), above 10 lx in 3 h before bedtime (gray), and values above 1 lx during sleeping period (dark gray).

Fig. 2.

(A) Distribution of the sleepiness score collected using the Karolinska Sleepiness Scale (KSS), with 10 being extremely sleepy (1799 observations from 59 participants). (B) KSS distribution across time awake (hours). Time awake was calculated as the duration between KSS reporting time and wake time. (C) KSS in time awake (hours). The box plots show a median with an interquartile range (whiskers showing range). Quadratic fit line was estimated using a mixed model. Model: KSS = time awake + time awake². The model estimates: intercept = 5.0, time awake term coefficient = −0.3 (P < 0.001), and time awake² term coefficient = 0.02 (P < 0.001). (D) KSS across different times of day (clock time). The box plots show median with interquartile range (whiskers showing range). Harmonic fit line was estimated using a mixed model. Model: KSS = sine (2π × time of day/24) + cosine (2π × time of day/24). The model estimates: intercept = 5.0, sine term coefficient = 0.8 (P < 0.001), and cosine term coefficient = 1.1 (P < 0.001). The fit line: mesor = 5.0, amplitude = 1.4, and nadir = 15.5. (E) Association between awakening KSS and sleep duration (hours). The blue line shows quadratic fit line with 95% CI. Data analyzed by the quadratic mixed model. Model: KSS = sleep duration + sleep duration². The model estimates: intercept = 11.2, sleep duration term coefficient = −1.5 (P = 0.001), and sleep duration² term coefficient = 0.09 (P < 0.05). (F) Correlation between KSS and melanopic EDI (log lx) at the minute of KSS recording. Thin lines show the best-fit linear relationship for each participant. The thick blue line shows the population linear fit. Data were analyzed using a linear mixed model with random intercept and slope, melanopic EDI coefficient = −0.5 (P < 0.001). (G) KSS across time awake (hours) for melanopic EDI > 250 lx (red) and melanopic EDI < 250 lx (blue) subgroups. The lines show quadratic fit lines with 95% CI.

Fig. 3.

Recent history of light exposure and sleepiness and sleep latency. (A–D) The scatterplots show statistical model estimates for melanopic EDI (log lx) means across different exposure durations (hours). Each point is derived from a mixed model with random participant intercepts to estimate sleepiness score, collected using the 10-item Karolinska Sleepiness Scale (KSS). Melanopic EDI with different exposure durations were the main predictors. The models were adjusted for a quadratic association with time awake (h), harmonic association with time of day (radian), and their interactions with light exposure. The x axis shows light exposure duration (hours), while the y axis shows the standardized coefficient of KSS mixed models. (A) All KSS recordings were included in the models, and melanopic EDI histories (for every 30 min) were predictors (the line and CI show polynomial fit of standardized coefficient). The red color represents P < 0.05. In B–D, the predictors were melanopic EDI history (for every 1 min), and outcomes of the models were (B) the first KSS scores within 2 h after awakening, (C) the first KSS scores within 2 h after awakening, but the light exposure history was calculated starting from reported wake time instead KSS recording time, and (D) the first KSS scores within 2 h after awakening, but the light exposure history was calculated between the wake time and KSS recording time.

Fig. 4.

(A) Linear mixed models to compare hourly mean melanopic EDI (log lx) with bedtime on that day. The scatterplot shows statistical models for each hourly mean throughout the day. The x axis shows the time of day (clock time), while the y axis shows the standardized coefficient of the mixed models. The blue dotted line shows zero effect. The red color represents P < 0.05. (B) Linear mixed models to compare daily collected light exposure characteristics with daily sleep states. Mean: The mean of daily melanopic EDI (log lx). Bright duration: Total minutes with melanopic EDI more than 1,000 lx. Mean Mel - Pho: The mean of differences between all daily observations of melanopic EDI (log lx) and photopic illuminance (log lx). The next three columns show melanopic light exposure durations: daytime (8:00 to 17:00) below 250 lx, above 10 lx in the 3 h pre-bedtime, above 1 lx during sleeping period. Last time: The latest time with melanopic EDI more than 1,000 lx. These models were adjusted for daylength, work/rest status, smoking, and alcohol and caffeine consumption on the test day, as well as age, sex and baseline subjective health as covariates. The color scale represents standardized mixed model coefficients, with red being positive effect direction and blue being negative. Bedtime (clock time), wake time (clock time), night sleep duration (hours), daytime sleep duration (hours), subjective sleep quality (score 1 to 5), sleep latency (hours), sleep efficiency (%), and time in bed (hours) were collected using sleep diary data. Bedtime, wake time, daytime sleep duration, and sleep latency coefficients were inversed in the figure. *P < 0.05 and **P < 0.007. (C) Linear regression models to compare weekly light exposure characteristics with weekly sleep states. IS: Interdaily stability of melanopic EDI. IV: Intradaily variability of melanopic EDI. L5: The dimmest 5 h mean. M10: The brightest 10 h mean. Mean: The mean of weekly melanopic EDI (log lx). Bright duration: Average minutes with melanopic EDI more than 1,000 lx per day during the week. These models were adjusted for daylength on the test week, employment status, age, sex, baseline subjective health, baseline smoking, and baseline alcohol and caffeine consumptions. The color scale represents standardized regression coefficients where red is the positive effect direction and blue is negative. Bedtime (clock time), wake time (clock time), night sleep duration (hours), daytime sleep duration (hours), sleep latency (hours), and sleep efficiency (%) were calculated as the weekly average of sleep diary data. Bedtime, wake time, daytime sleep duration, and sleep latency coefficients were inversed in the figure. *P < 0.05 and **P < 0.008. (D) The scatterplot shows statistical mixed model estimates of sleep latency (hours) across melanopic EDI means with different exposure duration (hours). The models were adjusted for a linear association with time awake (h), linear association with bedtime (clock time), and their interactions with light exposure. The x axis shows light exposure duration (hours), while the y axis shows the standardized coefficient of mixed models (for every 1 min). The red color represents P < 0.05.