The Circadian Response to Evening Light Spectra in Early Childhood: Preliminary Insights

Abstract

Although the sensitivity of the circadian system to the characteristics of light (e.g., biological timing, intensity, duration, spectrum) has been well studied in adults, data in early childhood remain limited. Utilizing a crossover, within-subjects design, we examined differences in the circadian response to evening light exposure at two different correlated color temperatures (CCT) in preschool-aged children. Healthy, good sleeping children (n = 10, 3.0 – 5.9 years) completed two 10-day protocols. In each protocol, after maintaining a stable sleep schedule for 7 days, a 3-day in-home dim-light circadian assessment was performed. On the first and third evenings of the in-home protocol, dim-light melatonin onset (DLMO) was assessed. On the second evening, children received a 1-h light exposure of 20 lux from either 2700 K (low CCT) or 5000 K (high CCT) (~9 and ~16 Melanopic equivalent daylight illuminance (mEDI lux), respectively) centered around their habitual bedtime. Children received the remaining light condition during their second protocol, with the order counterbalanced across participants. Salivary melatonin was collected to compute melatonin suppression and circadian phase shift resulting from each experimental light condition. Melatonin suppression across the 1-h light stimulus was significantly greater during exposure to the high CCT light (M = 56.3%, SD = 19.25%) than during the low CCT light (M = 23.90%, SD = 41.06%). Both light conditions resulted in marked delays of circadian timing, but only a small difference (d = −0.25) was observed in the delay between the 5000 K (M = 35.3 min, SD = 34.3 min) and 2700 K (M = 26.7 min, SD = 15.9 min) conditions. Together, these findings add to a growing literature demonstrating high responsivity of the circadian clock to evening light exposure in early childhood and provide preliminary evidence of melatonin suppression sensitivity to differences in light spectrum in preschool-aged children.

INTRODUCTION

Light is the strongest time cue for the mammalian circadian biological clock. Even dim levels of light in the evening can reduce production of the sleep-promoting hormone melatonin and delay the timing of the circadian rhythm (Duffy & Wright, 2005), with large individual differences observed in both adults and children (Hartstein, Diniz Behn, et al., 2023; Phillips et al., 2019). The timing of melatonin production is regulated by the suprachiasmatic nucleus (SCN) in the hypothalamus such that circulating levels typically increase prior to habitual bedtime, are high across the night, and reduce to low daytime levels in the morning (Wright Jr et al., 2005). Light influences sleep and circadian timing primarily through stimulation of the eye’s intrinsically photosensitive retinal ganglion cells (ipRGCs), which are ocular photoreceptors maximally sensitive to short wavelength blue light (~480 nm) that express the photopigment melanopsin(Berson et al., 2002; Brown, 2020; Enezi et al., 2011; Hattar et al., 2002). When stimulated, light exposure information is transmitted from the ipRGCs to the SCN, which in turn modulates the pineal gland’s production of melatonin(Brainard & Hanifin, 2005). The influence of light on the ipRGCs can be quantified via melanopic equivalent daylight illuminance (mEDI), a newly developed SI unit that strongly predicts the circadian response to light exposure in adults (Brown, 2020).

Findings from several studies with children, adolescents, and adults indicate that the sensitivity of the circadian clock to light decreases across development(Eto & Higuchi, 2023). Pre- to mid-pubertal adolescents show significantly greater melatonin suppression than late- to post-pubertal adolescents in response to evening light exposure at three different intensities (50, 150, and 500 lux) (Crowley et al., 2015). Additionally, melatonin is suppressed nearly twice as much for school-aged children compared with their parents following an evening bright light exposure(Higuchi et al., 2014). We previously demonstrated that preschool-aged children’s melatonin suppression response and circadian phase shifting are ultrasensitive to a high correlated color temperature (CCT) light source (5000 K) in the hour before bedtime across a wide range of light intensities. (Hartstein et al., 2022; Hartstein, Diniz Behn, et al., 2023) Specifically, we observed a consistently high level of melatonin suppression (M = 85%) in response to a single light intensity randomly assigned between 5 and 5000 lux, as well as robust delays of circadian phase (M = 56 min), even in response to low illuminance light exposures.

The sensitivity of the melatonin suppression response to light can also vary with spectral composition, both from monochromatic (Brainard et al., 2001; Brainard et al., 2008; Brown et al., 2021; Gooley et al., 2010; Thapan et al., 2001) and polychromatic light sources. In a study of adult men, a 2-h evening exposure to a fluorescent lamp at 6500 K resulted in greater melatonin suppression and sustained attention than both a fluorescent lamp at 2500 K and an incandescent lamp at 3000 K, all at an illuminance of 40 lux(Chellappa et al., 2011). Furthermore, in a repeated-measures design, young adults had greater melatonin suppression in response to 2 h of evening exposure to an LED computer screen with a short dominant wavelength (460 nm) compared with a long dominant wavelength (620 nm) across both low intensity (80 lux) and high intensity (350 lux) conditions(Green et al., 2017). Short-wavelength light also elicits greater circadian phase shifts than light of longer wavelengths(Lockley et al., 2003; Rahman et al., 2021; Wright & Lack, 2001); however, blue-enriched polychromatic light (17000 K) does not appear to be more effective than standard white polychromatic light (4000 K) for phase shifting responses in adults, at either ~100 lux for both conditions or 4100 K at 5000 lux and 17000 K at 4000 lux (Hanifin et al., 2019; Smith & Eastman, 2009).

Circadian sensitivity to short-wavelength light decreases with age(Gabel et al., 2017). Adolescents, but not adults, show significantly greater melatonin suppression when exposed to 300 lux of light at 5600 K than at 2700 K(Nagare et al., 2019b). Similar findings have been reported in school-aged children. In a between-subjects study, children aged 8.9 +/− 2.2 years who received an evening exposure to 300 lux of light at 6200 K had significantly greater melatonin suppression (81.2%) compared with those exposed to a 3000 K light (58.1%), while no difference in melatonin suppression between conditions was observed for adults(Lee et al., 2018). Additionally, melatonin suppression was greater for children than adults across both light conditions. To our knowledge, no research has examined the sensitivity of the circadian system to differences in the spectrum of evening light in preschool-aged children, a pivotal developmental period for sleep maturation and consolidation and a time when behavioral sleep problems (i.e., bedtime resistance, difficulty falling asleep, nighttime awakenings) often first emerge. Given the high photosensitivity previously demonstrated in this age group(Akacem et al., 2018; Hartstein et al., 2022; Hartstein, Diniz Behn, et al., 2023), data are needed to inform recommendations for parents and other stakeholders on regulating nighttime light exposures to support sleep and circadian health.

Using a within-subjects crossover design, healthy, good sleeping preschool-aged participants underwent a 10-day protocol twice. After maintaining a stable sleep schedule for 7 days, children completed a 3-day in-home dim-light assessment. Dim-light melatonin onset (DLMO) was determined for the first and third evenings of the in-home assessment. On the second evening, children received a 1-h exposure to a light source of either a low CCT (2700 K) or high CCT (5000 K), both set to a photopic illuminance of 20 lux (Table 1). Saliva samples were collected each evening in order to compute acute melatonin suppression and circadian phase shift. We hypothesized that exposure to a higher CCT light stimulus, ~16 mEDI, for 1-h centered at habitual bedtime would result in significantly greater suppression of melatonin and larger delays in circadian timing than exposure to a lower CCT light stimulus, ~9 mEDI, in preschool-aged children.

Table 1.

Characteristics of each experimental light source.

	Low CCT (2700 K)	High CCT (5000 K)
Photopic illuminance (lux)	19.87	19.36
Peak wavelength (nm)	641.2	447.1
Melanopic equivalent daylight illuminance (lux)	9.16	16.41

METHODS

Participants

Interested parents completed a brief online screening questionnaire, followed by a telephone interview to confirm eligibility. Children were eligible if they were between the ages of 3.0–5.9 years with ≥ 10-h parent-reported typical time in bed per night(Hirshkowitz et al., 2015). Children were excluded from participation for parent report of any of the following: clinical sleep disorders; behavioral problems; personal or family history of diagnosed narcolepsy, psychosis, or bipolar disorder; current illness or injury, fever, respiratory infection, asthma, or allergy; pre-term (< 35 weeks) or low birth weight (< 5.5 lb.); current use of caffeine or medications affecting sleep, the circadian system, or light sensitivity; developmental disabilities, epilepsy, or other neurological disorders; head injury involving loss of consciousness; migraine or frequent headaches; travel beyond 2 time zones in the 2 months prior to the first assessment; parent-reported sleep schedules varying > 2-h between weekends and weekdays; regular daytime napping (> 2 times per week); eye disorders or color blindness (verified with Ishihara Color Vision Test). Corrected vision with eyeglasses was permitted (n = 2). Written informed consent was obtained from parents. All study procedures were approved by the University of Colorado Boulder Institutional Review Board and families were compensated for participation.

Thirteen participants enrolled in the study. Of these, one participant who completed the study was excluded from the analyses due to substantial irregularities in their melatonin values (extremely high values with most samples at ≥ 50 pg/mL across both protocols). Additionally, one family (two participants who were siblings) withdrew after completing the first protocol, resulting in 10 participants (5 female, average age = 4.86 years, SD = 0.90 years, 100% Caucasian, Non-Hispanic/Latino) included in the analyses. Several additional exclusions from specific analyses occurred due to experimental error including accidental light exposure; these exclusions are described with the results of the specific analyses so that the number of participants included in each analysis is clear.

Protocol

Data collection occurred between May and August of 2022 and 2023. Participants completed two identical 10-day study protocols (Figure 1). For the first 7 days of the protocol, children followed a strict, parent-selected sleep schedule (bedtime and wake time) consisting of ≥ 10 h time in bed. Parents were instructed that children should receive no light exposure (outside of a dim night light) between bedtime and wake time, in order to ensure consistent light/dark timing leading up to the circadian assessment. To support this, researchers covered the windows in children’s bedrooms with black tarps at the start of the protocol to avoid exposure to sunlight during the sleep period. Adherence to the sleep schedule was confirmed through wrist actigraphy with a light sensor, a parent-completed sleep diary, and a daily phone call with the research team. Children wore an actigraph watch (Spectrum Plus, Philips Respironics, Pittsburgh, PA) on their non-dominant wrist throughout the entirety of each study protocol to assess sleep and measure light exposure in 1-min epochs.

Figure 1. Ten-day study protocol.

Children followed a parent-selected sleep schedule for seven days, followed by a three-day dim-light circadian assessment. On Days 8 and 10, an assessment of salivary dim light melatonin onset (DLMO) was performed. On Day 9, children were exposed to light (either 2700K or 5000K at 20 lux) for 1 h starting 30 min before habitual bedtime. Salivary melatonin was measured before, during, and after the light exposure. Participants then repeated the ten-day protocol, receiving the remaining light condition on Day 9. The order of exposure to each light condition was counterbalanced across participants. The times in the figure are provided as an example; actual study sleep schedules were based on habitual sleep schedules and varied across participants.

The final 3 days of each protocol consisted of an in-home dim-light circadian assessment. Researchers converted each participant’s home into a dim-light environment by covering windows with black tarps and ensuring that all light levels in the home were below 10 lux at the child’s angle of gaze. See Hartstein et al. (2023) for specific details on setting up the home(Hartstein, Wong, et al., 2023). The child entered the dim-light environment on Day 8 of the protocol 4.33 h before their habitual bedtime and remained in dim light through the completion of the protocol. Baseline DLMO was assessed that evening by collecting saliva samples in 20- or 30-min intervals from 3.33 h before until 0.83 h past each child’s habitual bedtime. If the first 7 days of actigraphy indicated an average sleep onset latency of more than 30 min, saliva samples were collected for an additional hour each night (n = 2).

On the evening of Day 9, participants received a 1-h light exposure centered on their scheduled bedtime. Children were seated at a custom-built light table consisting of dimmable LED light strips (Solid Apollo LED, Lynwood, WA). Light strips (rated as 2700 K and 5000 K) were arranged on a backplane of a 58 cm × 58 cm × 13 cm deep wooden box with an acrylic diffusing panel on its face. Children were engaged in activities designed to keep them looking down at the light source continuously, such as coloring on clear sheets with dry erase markers or playing with translucent blocks. Light levels at the child’s angle of gaze were recorded every 10 min with a research photometer (ILT 2400; International Light Technologies, Inc., Peabody, MA) and adjusted as needed to maintain a consistent photopic illuminance of 20 lux at the child’s eye. Starting 1.33 h before habitual bedtime, saliva samples were collected on Day 9 before, during, and after the light exposure at the same clock time as the previous evening to assess acute melatonin suppression and recovery of melatonin levels after returning to dim light. Characteristics of each light condition are reported in Table 1 in accordance with the ENLIGHT checklist and guidelines (Supplemental File S1) (Spitschan et al., 2023). Spectral characteristics for each light condition were obtained with a research spectrometer (MSC15, Gigahertz-Optik, Amesbury, MA, USA; calibrated 04/01/2021) and are depicted in Figure 2.

Figure 2. Spectra of each experimental light source.

Each light was set to a photopic illuminance of 20 lux.

Finally, on Day 10 saliva samples were collected from 3.33 h before until 2.33 h after each child’s habitual bedtime, in order to calculate a final DLMO. Saliva samples on Day 10 were collected at the same clock time as those on Day 8 until 0.83 h after bedtime at which point they were collected in 30-min intervals for an additional 1.5 h. Participants then repeated the 10-day protocol (inter-protocol period ranged from 1 to 64 days based on family availability, median = 12 days), receiving the remaining light condition on Day 9 of the second protocol. The order of the light conditions was counterbalanced across participants, ensuring equal distribution of the order across males and females. All saliva samples were collected and stored using our previously published protocols(Hartstein, Wong, et al., 2023). At the end of each data collection period, samples were assayed offsite (SolidPhase, Inc., Portland, Maine, USA).

Analysis

Actigraphy data were scored and analyzed using our previously published methods (LeBourgeois, Carskadon, et al., 2013) in order to calculate the following measures of sleep timing: bedtime (lights out time), sleep onset time, midsleep time, sleep offset time (wake time), sleep onset latency, and sleep duration. Light history on each of the first 7 days of the protocol was calculated as the average across 1-min epochs between actigraphically-determined wake time and bedtime.

Saliva samples were assayed with radioimmunoassay (Bühlmann Laboratories AG, Schöenbuch, Switzerland) with upper and lower detection limits of 50.0 pg/mL and 0.5 pg/mL respectively. Any values outside of that range were recorded as either 0.5 or 50.0 pg/mL. DLMO was computed as the linear interpolated clock time at which melatonin levels crossed (and remained above) 4 pg/mL(Carskadon et al., 1997).

Acute melatonin suppression was determined by comparing the area under the curve (AUC) of the melatonin profile during the light exposure (Day 9) and the melatonin profile during the corresponding 1-h time window on the baseline night (Day 8). AUC was calculated by the trapezoidal method and percent melatonin suppression was subsequently computed as:

$$\mathit{\text{Suppression}}=\frac{{\mathit{\text{AUC}}}_{\mathit{\text{baseline}}}-{\mathit{\text{AUC}}}_{\mathit{\text{light}}}}{{\mathit{\text{AUC}}}_{\mathit{\text{baseline}}}}$$

Percent change in melatonin levels between Days 8 and 9 at the time of each sample collected during the light exposure (10 min, 30 min, and 50 min after light onset), as well as 20 min after light offset, was also computed using the above formula with AUC replaced with melatonin levels at each time point.

The phase shift following the light exposure was calculated as the difference between the baseline DLMO (Day 8) and DLMO on the final evening of the protocol (Day 10). Negative phase shifts indicate a delay in circadian timing and positive phase shifts indicate an advance.

Kolmogorov-Smirnov tests were used to examine the normality of the distributions. Two-way mixed intraclass correlation coefficients (ICC) were calculated to assess consistency in baseline DLMO and bedtime phase angle between participants’ first and second assessments. One sample t-tests were used to determine significant effects on melatonin suppression and phase shift from each light condition. Paired samples t-tests were utilized to examine within-subjects differences between protocols and experimental light conditions on all sleep and circadian variables. Significance testing was conducted using an alpha level of 0.05. Effect sizes are presented for significant results as Cohen’s d.

RESULTS

Actigraphy

Table 2 displays actigraphically-measured sleep and light parameters averaged across Days 1–7 of each protocol. Data from one participant is missing because the family lost the actiwatch on the sixth day of their second protocol (5000 K). Paired samples t-tests revealed no significant differences between the two protocols for any of the sleep variables, confirming that participants maintained consistent sleep timing and duration before each circadian assessment.

Table 2.

Actigraphically-measured sleep and light variables averaged across the first 7 days for each experimental protocol.

	2700 K	5000 K	Paired samples t-test
Bedtime (clock hour)	20:03 (0:22)	20:02 (0:22)	t(8) = 1.27, p = 0.24
Sleep Start Time (clock hour)	20:21 (0:20)	20:19 (0:21)	t(8) = 1.04, p = 0.33
Midsleep Time (clock hour)	1:35 (0:20)	1:35 (0:17)	t(8) = −0.08, p = 0.94
Wake Time (clock hour)	6:49 (0:24)	6:51 (0:19)	t(8) = −0.49, p = 0.64
Sleep Onset Latency (min)	17.4 (10.6)	16.2 (6.9)	t(8) = 0.56, p = 0.59
Sleep Duration (min)	628.2 (18.4)	632.7 (21.2)	t(8) = −0.89, p = 0.40
Light Exposure (lux)	462.0 (272.2)	435.3 (368.5)	t(8) = 0.14, p = 0.89

Note: Table only includes data from 9 participants as the actiwatch was lost on the sixth day of one participant’s second protocol.

Light

Average light exposure from wake time until bedtime across the week preceding the dim-light assessment did not differ between protocols (Table 2). Average photopic illuminance measured at the eye and pointed in the child’s angle of gaze during each experimental light exposure (averaged across 7 measurements taken during the 1-h exposure) did not differ significantly between the 2700 K (M = 19.87 lux, SD = 0.92) and 5000 K (M = 19.36 lux, SD = 1.51) conditions (t(9) = −0.96, p = 0.36). Additionally, average illuminance measured at the eye during each saliva sample collection (excluding those collected during the 1-h light exposure) did not differ throughout the 2700 K (M = 0.57 lux, SD = 0.31) and 5000 K (M = 0.60 lux, SD = 0.17) dim-light protocols (t(9) = 0.33, p = 0.75), indicating that participants were exposed to similar light levels throughout each experimental dim-light assessment.

Baseline DLMO

We observed a high degree of reliability in baseline DLMO between participants’ first (M = 19:08, SD = 0:37) and second (M = 19:01, SD = 0:44) protocols, which occurred an average of 0.92 h and 1.04 h before scheduled bedtime respectively. The single measure ICC for baseline DLMO was 0.80 with a 95% confidence interval ranging from 0.38 to 0.95 (F(9,9) = 9.01, p = 0.002). For phase angle, the single measure ICC was 0.77, with a 95% confidence interval ranging from 0.31 to 0.94 (F(9.9) = 7.66, p = 0.003).

Average baseline DLMO was similar across participants for the low CCT, 19:07 (SD = 0:47), and for the high CCT conditions, 19:01 (SD = 0:34); (t(9) = 0.66, p = 0.53). As a result, because each participant’s scheduled bedtime was held constant across protocols, phase angle between baseline DLMO and the light exposure (beginning 30 min prior to scheduled bedtime) did not differ between protocols, indicating that the light was presented at consistent circadian times.

Melatonin Suppression

Melatonin suppression for both light conditions was normally distributed. Percent melatonin suppression for one participant during the 5000 K light exposure was > 2 SD from the mean (melatonin levels during the light exposure were 25.4% greater than at the same time on the baseline night). Analyses were conducted with and without this participant and any differences in statistical significance are noted in the text below.

One sample t-tests revealed significant suppression of melatonin during the 5000 K light exposure (t(8) = 8.77, p < 0.001, d = 2.92) but not the 2700 K exposure (t(8) = 1.75, p = 0.12, d = 0.58). Percent melatonin suppression was significantly greater during the 5000 K light (M = 56.3%, SD = 19.25%) compared to the 2700 K light (M = 23.90%, SD = 41.06%), t(8) = −2.42, p = 0.04, d = −0.81 (With outlier included, p = 0.15, d = −0.50).

Melatonin levels between the baseline (Day 8) and light exposure (Day 9) nights were compared at each time point that saliva was collected during the light stimulus (10 min, 30 min, and 50 min after light onset) as well as 20 min after light offset. For the 5000 K light condition, melatonin levels on Day 9 were significantly lower than on the baseline night at each time point analyzed (Figure 3A). For the 2700 K condition, melatonin levels on Day 9 were significantly lower than on the baseline night 30 and 50 min after light onset, with a marginal difference observed after light offset (p = 0.06) (Figure 3B). With the outlier included, melatonin levels were also significantly lower 10 min after light onset (p = 0.05) and 20 min after light offset (p = 0.03) for the 2700 K condition.

Figure 3. Condition averages of melatonin levels on Days 8 and 9 of the (A) high CCT (5000 K) protocol and (B) low CCT (2700 K) protocol (n = 9).

The protocol with the high CCT light (5000 K) is on the left and the low CCT light (2700 K) is on the right. Filled circles represent samples taken during the baseline evening (Day 8) and open circles represent samples taken during the evening of the light exposure (Day 9). The blue and red shaded regions denote the timing of the light exposure. Asterisks denote statistically significant differences in melatonin levels between samples taken at the same clock time on each evening (p < 0.05).

We next compared each data point during and 20 min after the light exposure to the data point collected at the same time during the baseline night in order to examine the percent change in melatonin levels at each time point across conditions. The percent change in melatonin levels was higher during the 5000 K light than the 2700 K light at 10 min (p = 0.051, d = 0.77), 30 min (p = 0.02, d = 1.00), and 50 min (p = 0.02, d = 1.00) after light onset (Figure 4). The percent change in melatonin levels 20 min after light offset did not differ between conditions (p = 0.20, d = 0.47). With the outlier included, the difference at 10 min results in p = 0.24.

Figure 4. Percent change in melatonin levels during and after each experimental light condition compared to dim-light baseline night (n = 9).

The shaded yellow area denotes the timing of the light exposure. Asterisks indicate statistical significance of p < 0.05. The 5000 K light exposure elicited a significantly larger suppression of melatonin 30 and 50 min after light onset compared with the 2700 K light exposure.

Phase Shift

Phase shift data were incomplete for 3 participants whose final DLMOs were not obtained during one of their protocols. One participant was accidentally exposed to a bright light in the afternoon following the 5000 K light exposure and, therefore, no final evening samples were analyzed for that participant. Another participant had an accidental light exposure after 5 samples had been collected on the final evening of the 2700 K protocol. Assays of those 5 samples indicated that DLMO had not yet occurred. Finally, due to experimenter error, the first 4 samples of the final evening of a participant’s 5000 K protocol were not collected; all samples subsequently collected were above 4 pg/mL, so DLMO could not be calculated. Additionally, final DLMO for the 2700 K protocol could not be calculated for one participant with complete sampling because all samples collected that evening were above the 4 pg/mL threshold. This resulted in complete phase shift data across both light conditions for 6 participants (2 female, average age = 5.05 years, SD = 0.93 years, 100% Caucasian, Non-Hispanic/Latino). Given the reduced sample size and small effect size observed, the within-subjects difference in phase shift across the two lighting conditions was not analyzed statistically.

Phase shift for both light conditions was normally distributed. One sample t-tests revealed significant delays of DLMO time on the final night compared to the baseline night following the 2700 K light exposure (t(5) = −4.12, p = 0.009, d = −1.684) and marginally significant delays following the 5000 K light exposure (t(5) = −2.52, p = 0.053, d = −1.03). Average circadian phase delays of 35.3 min (SD = 34.3 min) and 26.7 min (SD = 15.9 min) were observed following the 5000 K (Figure 5A) and 2700 K (Figure 5B) lights respectively. However, the difference between the two conditions resulted in only a small effect size (d = −0.25).

Figure 5. Condition averages of melatonin levels on the baseline and final night under dim light for the (A) high CCT (5000 K) protocol and (B) low CCT (2700 K) conditions (n = 6).

Filled circles represent samples taken during the baseline evening (Day 8) and open circles represent samples collected during the final evening (Day 10). The dotted line represents the 4 pg/mL threshold used to calculated dim-light melatonin onset (DLMO).

For both melatonin suppression and phase shifting, we observed large individual differences between participants in the magnitude of the responses (Figure 6). The 5000K light condition resulted in acute melatonin suppression ranging from −25.4% to 81.7% and phase shift ranging from a 15.0 min advance to a 69.0 min delay. The 2700K light condition resulted in acute melatonin suppression ranging from −44.8% to 82.5% and phase delay ranging from 3.6 min to 44.4 min.

Figure 6. Individual data points for (A) acute melatonin suppression and (B) circadian phase shift in response to each experimental light condition.

Each point on the x-axis denotes an individual participant. Note the participants in each panel are ordered by the magnitude of the response to the 5000K light condition.

DISCUSSION

In this controlled, within-subjects study we examined the circadian response to evening light of different spectra in early childhood. We observed that both low CCT (2700 K; ~9 mEDI) and high CCT (5000 K; ~16 EDI) light at a photopic illuminance of 20 lux resulted in significantly lower salivary melatonin levels after 30 and 50 min of exposure compared to dim light levels (< 10 lux, M = 0.63 lux at angle of gaze during saliva samples) at the same clock times on the previous baseline evenings. These data add to a growing literature demonstrating that children’s melatonin suppression response is ultrasensitive to evening light. In particular, the present study builds upon our previous work in which a 1-h evening light exposure at 5000 K resulted in robust melatonin suppression and phase delay across a wide range of light intensities (Hartstein et al., 2022; Hartstein, Diniz Behn, et al., 2023). Here we extend those findings to light of a lower CCT and low mEDI. Additionally, we observed more melatonin suppression in response to the high CCT ~16 mEDI light source compared with the low CCT ~9 mEDI light source, indicating sensitivity of the melatonin suppression response to spectral differences and extending previous findings in adolescents and school-aged children to children as young as 3 years. This high photosensitivity in early childhood may be due in part to features of eye anatomy, specifically that children have both larger pupils (under both dim and bright light conditions)(Higuchi et al., 2014) and clearer lenses than adults (Eto et al., 2021), allowing for greater light transmittance to the retina. At 480 nm, the peak sensitivity of the ipRGC, transmission through the lens in school-aged children is 1.18 times higher than for adults(Eto et al., 2021). We did not measure baseline pupil size or the pupillary light response in the present sample and so differences in retinal irradiance across individuals or lighting conditions cannot be determined. However, the differences in circadian photosensitivity observed among adolescents based on pubertal stage(Crowley et al., 2015) suggest there may be additional developmental changes in processes downstream of the eye that warrant further exploration.

Percent change in melatonin levels was significantly greater for 5000 K than 2700 K at 30 min and 50 min after light onset. In a previous study with adolescents, average melatonin suppression was greater during exposure to a 5600 K light (200 lux) than a 2700 K light (300 lux), but only after 2 h, 3 h, and 4 h of exposure(Nagare et al., 2019b); no significant difference was observed after only 1-h. Here we observed a significantly stronger suppression response to the 5000K compared to 2700K lights only 30 min into the exposure, again highlighting the high photosensitivity consistently observed in early childhood. Melatonin levels also remained significantly lower than on the baseline night 20 min after light offset for the 5000 K but not the 2700 K light condition, indicating a more sustained response to the higher CCT light source. In humans, the endogenous release of melatonin in the evening initiates processes that promote sleep (Cajochen et al., 2003). In addition to suppressing melatonin, evening exposure to light is associated with increased feelings of alertness and decreased sleepiness (Badia et al., 1991; Chellappa et al., 2011; Lockley et al., 2006; Rahman et al., 2014; Wright Jr et al., 1997), which can negatively impact the transition to sleep (Cajochen et al., 2022; Santhi et al., 2012). In a between-subjects study, school-aged children exposed to a 6200 K light condition had both greater melatonin suppression and lower subjective sleepiness (Karolinska Sleepiness Scale) than those exposed to 3000 K, with both conditions at 300 lux (Lee et al., 2018). The greater and more sustained suppression of melatonin observed in the present study in response to the 5000 K light could be accompanied by decreased sleepiness which may result in greater bedtime resistance or difficulty falling asleep at a parent-selected bedtime. However, suppression of melatonin by light does not necessarily equate with changes to sleep or feelings of sleepiness(Rüger et al., 2005). Healthy adults had greater melatonin suppression when exposed to light with a high melanopic EDI for 1 h in the evening than light with a low melanopic EDI (photopic lux ≈ 60 for both conditions), but no differences were observed between conditions on objectively-measured sleep architecture or sleep quality, subjective sleepiness, or behavioral vigilance (Blume et al., 2022). We previously demonstrated that daytime exposure to light of a higher CCT improves preschoolers’ cognitive performance beyond light of a lower CCT matched at ~250 lux (Hartstein et al., 2018). However, future studies are needed to determine whether evening exposure to light, and the accompanying suppression of melatonin, results in changes to alertness, performance, feelings of sleepiness, and/or sleep quality in early childhood.

Both the 2700 K and 5000 K light conditions resulted in a marked delay of circadian timing. We recently demonstrated that the post-illumination pupil response, the melanopsin-driven component of the pupillary light response, is significantly larger following a blue light compared to a red light of equal photon flux in adolescents, but not school-aged children (Hartstein et al., 2024). This suggests a larger contribution of outer retinal photoreceptors (rods and cones) to the non-visual response to light in childhood, which could have potentially contributed to the pronounced circadian responses following exposure to both light spectra observed in the present study. Although delays were on average 9 min greater after the 5000 K light, we observed only a small effect size for the difference between conditions (d = −0.25). Although our dataset was underpowered to examine statistical significance, several previous studies with adults failed to find significant differences in the circadian phase shift following exposure to white light of different spectra. At higher intensities (4000–5000 lux), large shifts in DLMO timing were observed following both morning (Smith et al., 2009) and evening (Smith & Eastman, 2009) light exposures, with no significant differences at either time between conditions (4100 K compared with 17000 K). In another study comparing light spectra at lower intensities typical of indoor environments (96–123 lux), melatonin suppression was significantly greater during exposure to a 17000 K light source than a 4000 K light, but the resulting delays in circadian timing were not significantly different (Hanifin et al., 2019). Further research is needed to identify the contributions of spectra and intensity to acute vs long-term effects.

An important implication of children’s high evening photosensitivity comes from the fact that bedtimes for young children are typically determined by parents and caregivers. Although healthy adults typically self-select bedtimes an average of 2 h after their DLMO (Burgess et al., 2003), parent-selected bedtimes for preschool-aged children occur an average of only ~40 min after their DLMO(Akacem et al., 2018; Hartstein, Diniz Behn, et al., 2023). For some children, parent-selected bedtime may even occur before their DLMO (LeBourgeois, Carskadon, et al., 2013). With such a small phase angle between DLMO and bedtime in this age group, a light-induced phase delay could push DLMO very close to or even after parent-selected bedtime. As bedtime gets closer to the timing of children’s melatonin onset, children take longer to fall asleep and have increased bedtime resistance (LeBourgeois, Wright, et al., 2013). Therefore, if children’s DLMOs are being pushed later by evening exposure to artificial light (regardless of spectrum and intensity), a mismatch between their bedtime and circadian timing could be contributing to the high prevalence of late sleep timing and behavioral sleep problems commonly observed at this age (Owens, 2007).

Various claims have been made that prolonged exposure to blue light leads to poor sleep or eye disorders, such as eye strain or age-related macular degeneration, with varying degrees of evidentiary support. Blue light-blocking glasses, with lenses that filter out short-wavelength light, can reduce the negative impacts of evening light exposure on the circadian system in both adults(Sasseville et al., 2006) and adolescents(van der Lely et al., 2015), although the lens tint and subsequent efficacy in blocking light of higher circadian stimulation can vary greatly across available products(Mason et al., 2022). When viewing light-emitting screens at night, adolescents had higher melatonin concentrations when wearing orange-tinted glasses than clear glasses or no glasses(Figueiro & Overington, 2016; van der Lely et al., 2015). Light adjusting software (e.g., f.lux, Apple Night Shift) also claims to limit the circadian stimulation of screen-based media devices through reductions in the amount of short-wavelength light emitted. However, studies report minimal effects of such software on sleep or melatonin production across adolescents and adults(Heath et al., 2014; Heo et al., 2017; Smidt et al., 2022). In young adults, melatonin suppression did not differ following a 2-h exposure to a tablet computer set to either a “more warm” (low CCT) or “less warm” (high CCT) color temperature at ~70 lux (Nagare et al., 2019a). Although blue light-blocking glasses are already being marketed for children of all ages, no research to date has examined the effectiveness of either blue light-blocking glasses or screen-adjusting software to prevent circadian disruption in young children, with their heightened sensitivity to light before bedtime. Our present findings, in which both light spectra resulted in significant reductions in melatonin and later circadian timing, suggest that either method to limit blue light exposure from screen use may reduce, but would likely not eliminate, the circadian impact for preschoolers. Future research should explore the efficacy of these and other strategies (e.g., dimming light intensity regardless of spectrum) to mitigate disruptions to sleep and the circadian system in this age group.

Although utilized in the present study, correlated color temperature is not necessarily an accurate metric to describe the strength of circadian stimulation from a light source. Two light sources with the same CCT can have vastly dissimilar spectral power distributions, with differences in peak wavelength and melanopic EDI. However, most consumer lightbulbs are currently labeled with CCT to indicate perceived color and warmth. The use of CCT to distinguish between the spectra of the experimental light sources for the current study was chosen to maximize ecological validity. Both 2700 K (warm/soft white) and 5000 K (cool/daylight) are common CCTs at which lightbulbs for homes are currently sold, thus simulating lighting conditions to which children are frequently exposed in their natural home environments. Exposure to a higher average light CCT at home is associated with a later DLMO for both school-aged children and adults(Higuchi et al., 2016), highlighting the importance of understanding the circadian impacts of lighting conditions to which young children are exposed.

Recent recommendations for daytime, evening, and nighttime indoor light exposure to support sleep and circadian rhythms for healthy adults state that, starting at least 3 h before bedtime, individuals should be exposed to melanopic EDIs of no greater than 10 lux as measured at the eye(Brown et al., 2022). Melanopic EDI for the high and low CCT light conditions in the present study was measured as ~16 lux and ~9 lux respectively. The findings of substantial reductions in melatonin and phase delay across both light conditions, in conjunction with our previous findings of a robust circadian response after exposure to evening light of even lower melanopic EDI (Hartstein et al., 2022; Hartstein, Diniz Behn, et al., 2023), suggest that these recommendations likely need to be adjusted for children in order to account for greater evening photosensitivity in this age group.

The repeated measures design of this study allowed us to control for differences in the endogenous circadian period length and any heightened sensitivity to light related to the dim-light protocol when comparing across the two light conditions within subjects. However, we observed large differences between participants in both the melatonin suppression and phase shifting responses. This is consistent with previous work in adults in which the intensity of light needed to elicit the half maximal melatonin suppression response was found to differ by as much as an order of magnitude between individuals(Phillips et al., 2019). We also previously reported large inter-individual variability in the phase shifting response to light in preschool-aged children. Across low, medium, and high intensities, light of similar intensities presented within a narrow circadian window elicited phase delays differing by 30–40 min across children (Hartstein, Diniz Behn, et al., 2023). A number of factors have been proposed to underlie individual differences in photosensitivity, including eye color, light history, pupil size, chronotype, and genetics(Chellappa, 2020; Swope et al., 2023). Further research is needed to understand the characteristics underlying these differences in young children and how they change across development.

Some limitations of this study should be noted. First, although a within-subjects design was employed, the study sample included only 10 participants overall, and 6 in the phase shift analysis, and therefore may be underpowered to detect small differences between conditions. Additionally, the study sample was highly homogenous regarding socioeconomic status, parent-reported race and ethnicity, and child health and developmental status. As such, our findings may not be generalizable to children from other populations. Recruitment presented a challenge as some families were hesitant regarding the study length and degree of commitment involved and others found it difficult to schedule two 10-day protocols around summer activities and vacations. Despite this, the rate of attrition was remarkably low. Of the 13 families that began the study, 12 (92%) completed both 10-day protocols, demonstrating the feasibility of conducting rigorous within-subjects circadian research with young children.

In summary, we present evidence that the circadian clock in early childhood is sensitive to differences in evening light spectrum, with increased suppression of melatonin observed in response to light of a higher CCT (greater mEDI). However, both light conditions at a photopic illuminance of only 20 lux resulted in delays of melatonin onset, adding to a growing literature demonstrating the potential of evening light exposure to contribute to behavioral sleep problems in early childhood through impacts on circadian timing. As parents seek to navigate children’s growing use of light-emitting media devices and evening exposure to artificial light, our data point to light spectrum as a modifiable component of children’s environments that may contribute to sleep and circadian health.