Daily light exposure habits of youth with migraine: a prospective exploratory study

Abstract

Photophobia is a common symptom in youth with migraine, but it is unknown if this leads to light-avoidant behavior, and its effect on migraine disease burden. We conducted an exploratory study measuring light exposure using wearable light logger pendants in 20 adolescents with migraine (ages 15–21) over one week. On average, participants received recommended light exposure during only 15.1% during the day ($\ge$250 lux melanopic equivalent daytime illuminance from 7:00–17:00) but were generally within recommended light levels in the evening (78.1% from 20:00–23:00 at $\le$10 lux), and at night (99.1% from 0:00–6:00 at $\le$1 lux). We observed differences in light exposure on weekdays versus weekends. Individual variation in circadian delay of light exposure was strongly correlated with the number of headache days per month (Rho = 0.66, p = 0.002). Measuring daily light exposure is feasible in pediatric populations with photophobia and warrants further study.

Introduction

Variation in visual diet (i.e., the intensity and timing of light exposure during daily life) has been associated with health outcomes. Exposure to brighter days and darker nights is known to confer reduced mortality risk¹, perhaps related to the effect that light has upon circadian biology². Insufficient daytime melanopic equivalent daylight illuminance (mEDI; short wavelength “blue light”) and evening mEDI from artificial light both have the potential to disrupt circadian entrainment^3,4. In support of this, multiple studies have associated nighttime screen use with poorer health-related quality of life and sleep disruption^5–7.

Photophobia (i.e., light sensitivity) is a common symptom of multiple neurologic and ophthalmologic conditions, including migraine^8,9, which may influence the visual diet through avoidance of high-intensity visual environments¹⁰. There has, however, been a limited empirical study of how visual diet is altered and potentially influences symptoms in people with photophobia. Recently developed, wearable light loggers now provide the ability to address these questions using quantified measures of visual diet during daily life^11–13. These small, battery-powered devices record the absolute level of light falling upon a detector over a period of many days. These recordings provide the relative amount of light across wavelengths, supporting inferences regarding the types of light encountered (natural vs. artificial) and the biological effect upon different classes of retinal photoreceptors (melanopsin-containing cells vs. cones).

We measured the visual diet for 20 youth with migraine using wearable light loggers that continuously tracked ambient light exposure over 7 days during a typical week. We hypothesized that continuous measurement of the visual diet in youth with migraine would be feasible and indicate associations between visual diet and disease burden in youth with migraine. Specifically, we predicted that photophobia would be associated with lower light exposure and that reduced light exposure during the day and increased light exposure in the evening and at night would be associated with increased headache frequency.

Results

Clinical and Headache Diary data

Twenty youth with migraine participated in this study. Demographics and clinical characteristics are shown (Table 1). Participants were a median age of 17 years old [IQR 16, 19] and were 70% female, and reported a median of 17 [IQR 6, 30] days per month of any headache, and 5 [IQR 2, 15] days per month of bad headache for the previous month at the start of recording. Headache-related disability (measured by 1-month PedMIDAS scores) was moderate on average, with a median PedMIDAS score of 16 [IQR 8, 29]^14,15. Median Visual Light Sensitivity Questionnaire (VLSQ-8) scores were 23 [IQR 19, 27]. This indicates that all had at least mild light sensitivity, with almost half falling into the moderate-to-severe light sensitivity range. Median fear-of-pain (FOPQ-C) score was 39 [IQR 31, 54], placing most youth in the moderate-to-severe range, consistent with chronic pain conditions, including migraine¹⁶. All participants were on at least one, and many were on a combination of pharmacologic agents for headache prevention. Supplements, OnabotulinumtoxinA, antidepressants, and calcitonin gene-related peptide blocking agents were the most common. This is representative of patients seen in the CHOP headache clinic, who have more severe and refractory migraine compared to the general population of adolescents with migraine.

Table 1.

Demographics and Headache Characteristics

Demographics and Headache Characteristics
Age median [IQR]		17 [16, 19]
Sex n (%)		F 14 (70), M 6 (30)
Race Ethnicity	Hispanica n (%)	2 (10)
	Non-Hispanic Black n (%)	2 (10)
	Non-Hispanic White n (%)	15 (75)
	Prefer not to answer n (%)	1 (5)
Validated questionnaires	Headache d/mo. median [IQR]	17 [6, 30]
	Bad Headache d/mo. median [IQR]	5 [2, 15]
	Continuous headache n (%)	12 (60)
	PedMIDAS (1 mo.) median [IQR]	16 [8, 29]
	Moderate/Severe disability n (%)	14 (70)
	FOPQ-C median [IQR]	39 [31, 54]
	Moderate/Severe fear of pain n (%)	16 (80)
	VLSQ-8 median [IQR]	23 [19, 27]
	Moderate/Severe light sensitivity n (%)	7 (35)
	PROMIS - Sleep	Impairmentb	Disturbanceb
	None (%)	7 (35)	10 (50)
	Mild n (%)	3 (15)	1 (5)
	Moderate n (%)	4 (20)	4 (20)
	Severe n (%)	6 (30)	5 (25)

^aFor the two participants who indicated they were Hispanic, both selected “prefer not to answer” for race. ^bSleep impairment measures symptoms of poor sleep, including fatigue and daytime sleepiness, while sleep disturbance measures difficulties falling and staying asleep. PedMIDAS Pediatric Migraine Disability Assessment, moderate/severe was defined as a score of 10 or greater. FOPQ-C Fear of Pain Questionnaire for Children, moderate/severe is defined as a score of 30 or greater, based on FOPQ-C definition; VLSQ-8 visual light sensitivity questionnaire, moderate/severe light sensitivity was defined as a score >24, which is the midpoint score.

All participants had 100% compliance with headache diary prompts. Headache diary responses were consistent with validated questionnaire responses. Youth reported a median of 6 headache days [IQR 3, 7], and 3 migraine days [IQR 1, 6], indicating they had high any headache and bad headache frequency during the recording week. Median daily light sensitivity score was 2 [IQR 1, 3], indicating mild light sensitivity, and median pain score was 4 [0, 6], indicating moderate pain.

Light logging

Overall, compliance with light logger wear was high. Each participant completed 7 days of recording for a total of 140 days. For one participant, 4 days of recording were excluded due to being away from their ActLumus device for more than 2 hours. The participant reported 3 days, but an additional day was excluded because no measurable change in illuminance or change in actigraphy sensors on the ActLumus was captured. This left 136/140 (97.1%) of days with usable light logging data across 20 participants. Seven participants reported using light-blocking lenses (blue light-filtered and/or sunglasses). Of those, only one used light filtering glasses continuously, with the remainder using glasses a maximum of 1-2 hours a day on an as-needed basis.

24-hour light exposure profiles

We evaluated patterns of photopic illuminance and mEDI across the 24-hour circadian cycle (Fig. 1). We compared light exposure on weekdays and weekends, as substantial differences have been noted in adolescent populations given the structure imposed by school¹⁷. Photopic illuminance and mEDI demonstrated high correlation throughout the day (Fig. 1a). However, there was a relative increase in photopic compared to -EDI exposure that was most pronounced from approximately 18:00 to 22:00. As expected, average light exposure levels were delayed by 49 minutes on the weekends compared to the weekdays. This shift was consistent with later and more variable bedtimes, sleep times, and wake-up times, and longer sleep durations reported in the Chronotype Questionnaire (Fig. 1b).

Fig. 1. 24-hr circadian light exposure and sleep patterns.

a Photopic illuminance (orange) and mEDI (light blue) were compared for the weekdays (left) and weekends (right). The difference (photopic – mEDI) is also shown (gray). Each 1-minute timepoint was calculated across a sliding 30-minute window. The central tendency represents the mean, and error bars represent 95% confidence intervals by bootstrap analysis. b Reported bedtime (dark blue vertical line), sleep period (gray bar), and wake times (black line) reported in the Chronotype questionnaire, which allowed participants to report their general bedtimes on weekdays versus weekends. Each row represents one participant during the weekday (left) and weekend (right). Participants are organized in descending order of frequency of any headache days per month. hr hours, mEDI melanopic equivalent daylight illuminance.

To determine if there were any emerging differences in light exposure patterns in youth with migraine based on headache frequency, we compared the temporal profiles of mEDI of youth based on the frequency of any headache and bad headache days per month (Fig. 2). There was a strong correlation between both the frequency of any headache (Rho = 0.66, p = 0.002) and bad headache (Rho = 0.60, p = 0.005) days per month and shifts in temporal light profiles: youth who demonstrated later temporal light profiles reported more any headache and bad headache days per month than those who were exposed to light earlier. This did not appear to be directly related to sleep/wake habits because there was not a clear systematic difference in sleep and wake times between those with higher frequency and lower frequency headache (see Fig. 1b). It was also not related to age, as there was not a significant relationship between age and shifts in temporal light profile (Rho = 0.13, p = 0.581).

Fig. 2. Association between circadian mEDI profile and headache frequency.

Shift in temporal profile of mEDI as a function of any headache days (top) and bad headache days (bottom). Spearman’s Rho with p-value is reported. Negative values indicate participants who had an earlier temporal profile compared to the mean, and positive values indicate participants who had a later temporal profile compared to the mean. Significance was defined as p < 0.05. mEDI melanopic equivalent daylight illuminance, mo month, min minutes.

Light exposure summary metrics

Summary metrics were calculated to capture daily light intensity and light timing based on our findings across the 24-hr light profile, and prior studies^11,13,18,19. Two metrics of photopic illuminance were used to characterize daily light intensity (Fig. 3a). Total photopic luminous exposure was calculated, which represents the integrated exposure to photopic light in a 24-hour period. The mean total luminous exposure across participants was 6.2 klux*hr, with a wide range between 0.2 and 16.9 klux*hr. Time youth spent in bright light was also estimated, which was defined as light exposure of 1000 lux photopic illuminance or greater, as indoor light is typically below 1000 lux though it can vary based on lighting conditions. The mean total time spent in bright light across participants was 42 minutes per day [range 0 to 108 minutes].

Fig. 3. Light exposure summary metrics.

a Light intensity. Photopic illuminance was used to calculate light intensity metrics as these have been used as the standard for designing lighting spaces. Each participant represents the mean photopic luminous exposure (left panel) and the mean time spent in bright light levels (right panel) across the 7-day period for each participant. The mean (black line) is also shown. b mEDI (light blue) was used to measure the timing of light exposure because this is important for circadian entrainment. Percent time spent within recommended mEDI based on time of day: at or above 250 lux mEDI from 6:00–17:00 (left), at or below 10 lux 20:00–23:00 (middle), and at or below 1 lux at 0:00–6:00 (right). Mean across participants (black) and individual participants averaged across 7 days (light blue circle) are shown. mEDI melanopic equivalent daylight illuminance.

Percent time spent within recommended light levels across a 24-hour period was used to capture light timing (Fig. 3b). Different levels of light exposure have been recommended for healthy adults based on the time of day: a minimum light exposure of 250 lux mEDI during daytime hours; a maximum of 10 lux starting three hours before bedtime; and 1 lux or less at night is recommended¹⁹. We therefore calculated the proportion of time within these recommended limits during 7:00–17:00 (daytime), 20:00–23:00 (evening/pre-bed), and 0:00–6:00 (nighttime). Timing was selected based on typical school schedules, the diurnal pattern of the sun at the location of the study. These definitions were supported by the timing of light exposure observed across participants within a 24-hour period as well as reported bedtimes and wake times. Participants spent an average of 15.1% +/− SD 8.0% of hours between 7:00–17:00 exposed to a minimum mEDI of 250 lux (Fig. 3b). Percent time spent within recommended levels improved substantially between 20:00–23:00 and 0:00–6:00, with youth spending an average of 78.1% +/− SD 20.8% and 99.1% +/− SD 2.9% of the time within recommended levels, respectively.

We conducted power analyses to determine sample sizes needed for group comparisons within youth with migraine comparing clinical features across light logger metrics. We found that sample sizes of 75 participants or fewer per group would be sufficient for most comparisons depending on the light metric and clinical variable being considered (see Supplementary Materials).

The impact of environmental factors on light exposure

Across the study, there was variation in multiple environmental factors that could impact light exposure: daylight length ranged from 9.3 to 12.3 hours, temperatures ranged from 14.5 to 70.5°F, and 32% of days had more than trace precipitation ranging from 0.01 to 1.54 inches. To determine the day-to-day influence of these factors on light exposure habits, we assessed daily light metrics as a function of daylight hours, temperature, and precipitation (Table 2).

Table 2.

The influence of environmental factors on light exposure habits

Influence of environmental factors on daily light exposure
Factor	24-hr Photopic Luminous Exposure [klux*hr]	Time spent in bright light [minutes]	% time 7:00 – 17:00 $\ge$ 250 lux mEDI	% time 20:00 – 23:00 $\le$ 10 lux mEDI	% time 0:00 – 6:00 $\le$ 1 lux mEDI
Daylight length [hr]	Rho =−0.05 p = 0.589	Rho =−0.09 p = 0.300	Rho =0.04 p = 0.700	Rho = 0.15 p = 0.08	Rho =−0.21 p = 0.017
Temperature [°F]	Rho = 0.04 p = 0.644	Rho = 0.03 p = 0.702	Rho = 0.02 p = 0.857	Rho =−0.11 p = 0.192	Rho = 0.01 p = 0.914
Precipitation [in.]	Rho =−0.33 p = 1.2e−4	Rho =−0.26 p = 0.002	Rho =−0.29 p = 7.4e−4	Rho = 0.05 p = 0.580	Rho = −0.01 p = 0.953

Spearman’s Rho is reported for each comparison to assess the relationship between summary light metrics and daylight length, temperature, and precipitation on daily light exposure habits. Bolded text indicates significant values, defined as p < 0.05.

Daylight length did not have a significant correlation with light exposure during the day or evening, but was significantly correlated with nighttime light exposure: nights during which individuals spent less time within the maximum recommended 1 lux mEDI tended to happen when the days were shorter. Rainy days were associated with lower light exposure: precipitation was negatively correlated with total luminous exposure (Rho = −0.33, p = 1.2e⁻⁴, moderate effect size), time spent in bright light (Rho = −0.26, p = 0.002, small effect size), and percent time spent at or above 205 lux between 7:00–17:00 (Rho = −0.29, p = 7.4e⁻⁴, small effect size), Precipitation did not show significant correlation with evening and night light exposure. There was no significant correlation between temperature and any light metric.

Participant Feedback

Seventeen (85%) participants agreed or strongly agreed with the statement “I would recommend somebody to participate in this study,” while 3 (15%) strongly disagreed. Specific comments included liking the text reminders for the diary and remembering to wear the devices. They found the text-based diary easy to use. Participants offered ways of improving the study, including making the device smaller and addressing challenges with the headache diary, only being once a day, but experiencing multiple headache spikes a day.

Discussion

We conducted an exploratory study of light exposure during the everyday life of 20 youth with migraine, most with high migraine disease burden. To our knowledge, this is the first study to measure light exposure habits in a population with migraine using wearable light logger technology. We found preliminary evidence that visual diet is associated with headache burden in youth with migraine. Here, we review intriguing trends we observed in youth with migraine in the context of other studies, and how these findings should inform future study design.

Participants spent only about 15% of the daytime (~1.5 hours per day) at or above the recommended minimum daylight levels for healthy adults. By comparison, they spent most of their time within the recommended maximum light levels 3 hours before bed, and during the night (78% and 99% on average, respectively). We suspect the low light exposure levels observed in this study are due to multiple factors. Perhaps one of the largest contributors to low daytime light exposure—not unique to individuals with photophobia—is the tendency in modern societies to live and work in indoor lighting environments that are darker than outdoor environments. Lucas and colleagues conducted a similar study in 59 generally healthy, mostly younger adults, providing an important reference point. They found participants spent 33% of daylight hours at or above recommended light levels, but 66% of the time at or below recommended light levels 3 hours before bed¹⁸. While results across the two studies are generally similar, our participants were exposed to lower light levels at each point in the 24-hour cycle. Additionally, youth with migraine spent less time exposed to bright light: the adults spent an average of 1.7 hours per day under bright light conditions¹⁸ compared to an average of just over 40 minutes per day in our study. These differences are present even though Lucas and colleagues used wrist watches that may underestimate light exposure compared to pendant wear under some circumstances^12,20.

Many factors could be contributing to the differences between our study and the study by Lucas and colleagues. First, given the absence of relevant normative data, it is unknown if adolescents and adults generally have similar light exposure habits. We observed light exposure profiles shifted later on weekends compared to weekdays, indicating schedules imposed by school substantially influence light exposure in this age group. This corresponded with reported sleep/wake times and is consistent with prior studies in adolescents¹⁷. Second, environmental factors could explain differences between the two studies. Multiple studies have reported significantly lower light exposure in the winter compared to the summer, which may be related to photoperiod and weather^11,21. We recorded November through March, biasing our study towards the winter months with shorter daylight hours. However, we did not observe a significant relationship between duration of daylight (which varied by 3 hours across the study) or temperature and daytime light exposure. We did find rainy days corresponded to lower light exposure. Finally, other factors (e.g., participation in outdoor sports) could influence light exposure. We found that bright light exposure was variable across participants, ranging from 0 to 2 hours a day, and could reflect differences in extracurricular activities.

Clearly, simultaneous measurements of study and control populations will be needed to support stronger claims of migraine-related photophobia being associated with a restricted visual diet. We note that recommended light levels are based on expert consensus and appeal to the biology of circadian responses to ocular light¹⁹. While these recommendations were rigorously developed, large observational studies using this metric are needed to determine if there is a broader impact on general health if these recommendations are not met or are only met some of the time.

We observed a strong correlation between later temporal profiles and a greater number of any and bad headache days per month. This means that youth who were exposed to light later in the day reported more frequent headaches. There are myriad possible explanations for this relationship. It is possible, for example, that youth with more frequent and severe headaches are exposed to light later due to waking up with severe symptoms, producing a shifted light schedule. Another explanation is that light exposure is contributing to increased headache frequency, possibly reflecting evening screen use that can disrupt sleep^6,22. This would be consistent with the observation that increased headache frequency in youth is associated with prolonged screen use^23,24. It is also possible that the interaction of fixed school start times and a later chronotype provoke additional headache days. Later chronotype has been associated with more frequent headache in youth²⁴, however, reported sleep/wake times did not appear systematically different across participants with lower versus higher headache frequency. Further study with appropriately powered sample sizes that account for extracurricular activities and consider the direction of effects is needed to confirm and unpack this finding.

Interestingly, we found that there was a reduction in mEDI relative to photopic illuminance in the evenings. This may reflect differential spectral emissions of artificial lights used in the home as compared to public indoor spaces or sunlit outdoor spaces during the day²⁵. Another possible contributing factor is that participants were using electronic devices in night mode, which purposefully reduces mEDI. In our clinical experience, our patients with migraine often use these “night mode” settings because they feel it improves the tolerability of their devices.

To our knowledge, this is the first study to report objective measurements of light exposure combined with a daily diary to track migraine symptoms in real time. We used a light logger with 10 channels of differing spectral sensitivities, allowing for estimation of photopic and melanopic illuminance. Overall, diary and device compliance were high, providing a complete dataset for analysis, and participants generally had positive feedback on the study design. We speculate that diary compliance was excellent because the questions were delivered via a familiar smartphone messaging system, allowing them to answer directly in a text stream on their phones. Indeed, participants reported liking this feature and finding it easy to use. Participants also had a financial incentive for at least 80% compliance.

Limitations include that this was an exploratory study with a small sample size and the lack of a control group. Furthermore, participants were established patients of a pediatric headache clinic receiving active treatment, limiting the generalizability of the results. There may be bias in that youth willing to participate in the study may be pursuing healthy headache habits, while those still struggling with daily habits may be less likely to participate. While light filtering lens use was captured by survey data, we did not correct our measurements for the continuous use of blue light filtering glasses for one participant. Trends we saw across differences in schedule, photoperiod, and weather did not account for individual variation; thus, further study is needed to confirm these results. We cannot rule out the possibility that participants did not place the light loggers upright on their bedside tables in the evening and thus may be overestimating low light exposure at night; however, we did observe spikes of light overnight in many participants, indicating we were able to capture night light exposure. Finally, we regret that we are unable to provide public access to our raw data due to legal limitations imposed by the consent form.

Studies that include larger sample sizes will allow for comparison within youth with migraine across multiple clinical features, as well as comparison to a control group of migraine-free peers. Prior findings indicate that significant differences between winter and summer light exposure should be found in within-participant sample sizes as small as 3 individuals¹¹. We found a strong correlation with shifts in temporal light profile with our sample size of 20 participants, but other comparisons indicate sample sizes of up to 75 participants per group will be needed, depending on the light metric and clinical variable being considered (see Supplementary Materials). Controlling for time of year may still be critical as there are dramatic differences in light exposure between the summer and winter months at greater latitudes²¹. Although we did not see substantial variation in daytime light exposure, we did not have large enough sample size to control for individual differences. This cycle may contribute to the seasonal variation observed in migraine, where migraine symptoms to be worse in the late fall and winter months^26–29, and seasonal comparison offers a unique opportunity for within-participant comparisons.

Ultimately, non-optimal light exposure may offer a modifiable risk factor for increased disease burden in youth with migraine. If larger observational studies confirm a relationship between migraine disease burden and light exposure, then clinical trials focused on interventions that address visual diet could be pursued. Multiple participants expressed interest in having access to their data, suggesting this may be a viable target for behavioral intervention.

In conclusion, measuring daily light exposure in youth with migraine is feasible and offers a promising avenue for larger observational studies to understand the relationship between visual diet and migraine burden in youth. Our preliminary data demonstrate that youth with migraine spend little time at or above recommended daylight levels but are generally within recommended evening and night darkness levels. Comparison to a matched migraine-free control group is needed to determine if this is related to the high levels of photophobia they reported. We also observed that those with more frequent headache had a delayed light exposure schedule that warrants further exploration.

Methods

Study design

This single-center prospective observational exploratory study was conducted within the pediatric headache program within the Children’s Hospital of Philadelphia (CHOP) neurology department between October 2024 and March 2025. The protocol received approval from the Children’s Hospital of Philadelphia Institutional Review Board (IRB 24-022242) and was in accordance with the Declaration of Helsinki.

Participants

A sample size of 20 was chosen for this exploratory study because prior work has based power analysis calculations on 13 participants and found that sample sizes as low as 3 individuals were sufficient for intra-individual differences between the summer and winter months¹¹. We opted for a larger sample size of 20, given that we were assessing a clinical population and aimed to assess group differences. Potentially eligible participants were identified by screening patients being seen in upcoming headache clinics via chart review. Participants were included if they were between the ages of 10 to 21 years, ICHD-3 defined migraine with or without aura, given by a headache specialist of any headache frequent,y and consented (≥18 years) or participant consent and parental consent (<18 years) to participate in the study. Exclusion criteria were a history of major neurological conditions besides migraine (e.g., history of epilepsy, stroke, multiple sclerosis), a recent history of concussion (<3 months), or starting or weaning off a headache preventive medication (supplement or prescription) without being on a stable dose for at least 1 month prior to enrollment. During our enrollment period, we had 191 potential participants screened from two weekly headache clinics, of which 145 (75.9%) were eligible for the study. Of those eligible, 75 potential participants were contacted. The remainder were not contacted due to the limited availability of devices. Of those, 20/75 (26.7%) were enrolled. Of the 55 not enrolled, the majority (35/55, 63.6%) did not respond to the enrollment telephone call, 2 (3.6%) did not want to wear the devices, 2 (3.6%) the family was going to be away from home, 3 (5.5%) did not give a specific reason, 1 (1.8%) their appointment was switched to a virtual visit and thus were unable to attend the in person visit, and for the remaining 11 (20%) the reason they were not enrolled was not recorded.

Data collection

Devices were shipped to the home of participants and, once received, a virtual visit was conducted to review the proper use of the wearable devices and text diary before recording began. At this visit, participants completed baseline questionnaires through REDCap (Research Electronic Data Capture)^30,31 hosted by the institution. The first full day of recording (0:00 – 23:59) was considered “Day 1,” and data collection ran for 7 full days. During these 7 days, diary questions were texted between 19:00 and 23:00 based on participant preference. This broad time range was chosen to encourage compliance with the text-based diary. While this did mean that patients were using their phones at night, their preference was in line with reported daily sleep/wake habits. During data collection, participants continuously wore the light logger while responding to headache diary prompts each evening. The light logger was removed while sleeping to avoid wearing a cord around the neck or having the light logging device being obscured by bedsheets. Participants were instructed to keep the light logger facing upwards on a bedside table to allow for continued recording, consistent with prior light logger studies¹⁸. Participants who were participating in sports during which wearing the ActLumus device was not allowed, or otherwise posed a potential safety hazard, were instructed to keep the device in the area of play with the sensor unobscured. Participants were also instructed not to wear the device while showering or submerged in water. At a scheduled clinic visit after the recording week, participants returned the wearable devices and filled out REDCap-based questionnaires. Instances during which the device was removed for sports practice were reviewed in detail to determine if the device was kept with them.

Some participants wear tinted spectacle lenses (e.g., blue light blocking, sunglasses). We intended to measure the transmittance spectrum of these lenses for each participant, but a technical failure prevented us from recording this data.

Demographic, medical history, clinical characteristics, and treatment were gathered. Standardized and validated survey data included the CHOP Headache Questionnaire³², including the number of any headache days and bad headache days per month. Light sensitivity was measured using the Visual Light Sensitivity Questionnaire (VLSQ-8), which is an 8-question validated measurement of light sensitivity with possible scores ranging from 8 – 40³³. The Fear of Pain Questionnaire for Children (FOPQ-C) is a validated metric that was used to measure fear-avoidant responses to pain. After the week of data collection, headache-related disability over the previous month was assessed with the Pediatric Migraine Disability Assessment (PedMIDAS). PedMIDAS has a maximum score of 80 (1-month version)¹⁵, with scores of 3 or less indicating no disability, 3 – 9 indicating mild disability, 10–16 indicating moderate disability, and >16 indicating severe disability. PROMIS Sleep Disturbance questionnaire was also filled out to capture the perception of sleep quality, and the PROMIS Sleep-Related Impairment questionnaire, which captured symptoms of insufficient sleep (e.g., daytime sleepiness). PROMIS rates no, mild, moderate, and severe symptoms based on T-scores³⁴. Data were also collected on chronotype and weekday versus weekend sleep/wake habits using the standardized chronotype questionnaire³⁵. Participants were compensated $75 for their time, with an additional $75 incentive if they achieved at least 80% compliance with device wear and text-diary completion.

Daily migraine symptoms, acute medication use, and disability related to headache were recorded with a validated text-based Daily Headache Diary³⁶ with additional questions to capture device wear compliance and light-blocking lens use. Daily headache diaries were filled out via text message using the HIPAA-compliant Twilio platform hosted at CHOP. Specific questions included “Have you had a headache today” (yes or no), “Has your headache gotten in the way of your school, home, or social life today” (yes or no), and “Rate your light sensitivity (0–5).” These data were collected to allow for future exploration of the day-to-day relationships between light exposure habits and migraine symptoms. A summary of these data is included in the Supplementary Materials.

Light logger data

Device wear and data collection

ActLumus devices (Condor Instruments, São Paolo, Brazil) are 10-channel wearable light loggers that provide 24-hour continuous collection of photopic illuminance and mEDI with an operating range of 1–100,000 lux (Fig. 4a). ActLumus devices were worn as a pendant around the neck, providing more ecologically valid measurements that are not obscured by shirt sleeves and are closer to the visual plane, which is critical for capturing non-image forming effects of light^12,20. Data were collected at a sampling rate of 1/minute to maximize temporal resolution while still providing for 7 days of continuous recording¹³. Examples of a 24-hour recording weekdays are shown for 4 participants (see Fig. 4b). Though there was some individual variation, we observed overall periods of high light exposure that corresponded with the photoperiod and school schedules. We also observed brief transient periods of increased illuminance during the night across multiple participants, indicating that we were able to capture light habits at night.

Fig. 4. ActLumus recordings.

a Illustration of the ActLumus light sensor consisting of 10 Channels with different spectral sensitivities that sample at a rate of 1/minute. These measurements are differentially combined to derive mEDI which is important in circadian rhythm signaling and photopic illuminance, that is the basis for image-forming daylight color vision. b A representative weekday of four participants continuously recorded photopic illuminance and mEDI over a 24-hour recording period. While there is variation, light exposure increases to measurable levels ($\ge$1 lux) between 6:00 and 7:00 and then drops below measurable levels between 20:00 and 24:00 that following a typical 24-hour sleep/wake cycle. Participants P1, P12, and P13 have clear brief periods of measured light between 0:00 and 5:00 that demonstrate lights turned on at night are captured. Definitions of daylight hours (7:00–17:00), evening or pre-bed (20:00–23:00), and night hours (0:00 – 6:00) are shown in yellow, blue, and gray, respectively.

Data processing

Data were visually inspected to ensure variation in light exposure and movement based on the actigraphy sensor and excluded if participants reported not wearing and being away from the Actlumus device for more than 2 hours in a day. ActLumus data were pre-processed using ActiLab Software (Condor Instruments, São Paolo, Brazil). Specifically, melanopic and photopic lux values were derived from 10 light-sensing channels that detect different wavelengths of light for every minute of data capture. Clinical data, as well as mEDI and photopic illuminance, were processed using custom software developed in MATLAB (Mathworks). Log transformation was performed on mEDI and photopic illuminance values to accommodate large shifts in illuminance levels for graphical representation and statistical analysis¹¹.

Light metrics

Two features of the visual diet were considered: 1) the intensity of light exposure, 2) the timing of light exposure. Light metrics were chosen that summarize continuous light exposure data to capture these two features of the visual diet^11,37. Light intensity was represented by photopic luminous exposure and time spent exposed to bright light levels. Photopic luminous exposure captures the total 24-hour photopic light exposure (kilolux*hr), which provides a measurement of overall light exposure. Bright light was defined as photopic illuminance >1000 lux because outdoor light ranges from 1000 lux on a cloudy day to 100,000 lux on a bright sunny day, while indoor light is generally below 1000 lux though depends on lighting conditions^37–39. Light timing was defined by the percent time spent within recommended mEDI limits. A minimum 250 lux mEDI is recommended during daytime hours, while a maximum of 10 lux starting three hours before bedtime, and 1 lux or less at night is recommended to support optimal timed melatonin release¹⁹. These recommendations were based on expert-scientific consensus and supported by the sensitivity of human “non-visual” responses to ocular light. In this study, we chose fixed time epochs during daylight hours (7:00–17:00), evening or pre-bed (20:00–23:00), and night hours (0:00–6:00). Fixed epochs were based on the diurnal motion of the sun, the structured schedule imposed by school, and sleep recommendations for consistent and sufficient sleep in this age group based on the American Academy of Sleep Medicine⁴⁰. Gaps in time were left between epochs to allow for some variability across individual schedules. Four representative weekdays from different participants are shown to demonstrate that this timing was reasonable (Fig. 1b).

Shifts in the mean temporal light profile for each participant were estimated by determining the correlation between a shifted version (+/−200 min) of each temporal profile to the mean across all participants and taking the highest correlation value to determine the size of the shift. Maximum correlation using this shifting technique was overall high (mean 0.86), with all but two participants with a fit >0.75. Removing these two participants from the analysis did not significantly change the outcome of the results.

Validation of Light Logger measurements

We validated the tabular spectral sensitivity functions of the ActLumus device. To do so, we measured a standard light source using both the ActLumus and a calibrated spectrophotometer (SpectraScan® PR-670, JADAK, North Syracuse, NY). Light from an incandescent source was delivered via liquid light guide into a light integrating sphere (LabSphere, North Sutton, NH). The spectral radiance of the light source was measured at 2 nm resolution using the PR-670. Combining this spectrum with the ActLumus tabular sensitivity functions (provided by the manufacturer), and converting from radiance to irradiance, provided a model prediction of the ActLumus sensor counts. We then measured the light source using the ActLumus and compared the obtained and predicted sensor counts. We found the ActLumus counts to be in excellent agreement with the prediction (Pearson’s R = 0.998). We were, however, limited to validating 9 of the 10 ActLumus channels, as the PR-670 does not measure the infra-red sampling range of the 10^th channel.

We additionally validated the mEDI value reported by the ActLumus device. The measured spectral irradiance within the sphere was supplied to the CIE S 026 alpha-opic Toolbox⁴¹. The calculated alpha-opic equivalent (D65) illuminance (e.g., mEDI) was 28.12 lux, as compared to a reported mEDI from the ActLumus of 29.08 lux. We considered this agreement in values to be reassuring for this single measurement.

Temperature, precipitation, and daylight length

Daily temperature and precipitation values were obtained from the United States National Weather Service for the Philadelphia area from November 2024 to March 2025⁴². Data collected by The United States Naval Observatory was used to determine daylight length in the Philadelphia area from November 2024 to March 2025⁴³.

Data analysis

No a priori sample size calculations were performed, as this was an exploratory study to determine sample size for future studies. Descriptive statistics for continuous variables included median with interquartile range for non-normal continuous distributions and mean with standard deviation for continuous variables with normal distribution. Proportions were reported for categorical variables. Continuous light logger data were graphically represented as the mean with 95% confidence intervals (CI) determined by bootstrap analysis. Statistical comparisons were made with the T-test (for normal data) or Kruskal-Wallis (non-normal) to compare two groups. Spearman’s Rho was used to determine correlations for continuous variables where at least one variable had non-normal data. For climate calculations, the comparison between daylight, temperature, and precipitation was compared across light metrics for each of the 136 participant days recorded, disregarding individual variation. Significance was defined as a p-value < 0.05. Effect sizes were determined based on Spearman’s Rho (non-normal continuous and ordinal variables). Small effect size was defined as ρ $\ge$ 0.1, moderate effect size was defined as ρ $\ge$ 0.3, and large effect size was defined as ρ $\ge$ 0.5, which we have used in past studies⁹.

As a first step to determining how light exposure differs with disease burden in youth with migraine, we conducted analyses to determine sample sizes needed to adequately power such comparisons. We compared light metrics across the following groups: (1) those with at least 15 headache days per month and 8 migraine days per month based on ICHD-3 criteria for chronic migraine compared to those who did not meet this criteria, (2) those with no-to-mild versus moderate-to-severe headache-related disability as defined by PedMIDAS, (3) those with low visual sensitivity (VLSQ-8 $\le$ 24) versus high visual sensitivity (VLSQ-8 > 24), (4) those with no or mild versus moderate-to-severe sleep disturbance defined by PROMIS, (5) those with no or mild versus moderate-to-severe sleep impairment defined by PROMIS. These results are presented in the Supplemental Materials.

Author contributions

C.P.G. conceptualized the project, secured funding, oversaw data collection, developed software to analyze the data, performed the data analysis, interpreted the data, prepared all figures and tables, and wrote and edited the main manuscript. R.S. performed the data collection and wrote the first draft of the methods section. B.M.P. was involved in the conceptualization of the project, organized data collection, oversaw and beta tested the text-based diary, managed clinical and light logger data storage, and provided feedback on data interpretation as well as edited and provided feedback on the manuscript. N.R. provided data collection support and approved the final manuscript. C.L.S. was involved in the conceptualization of the project, provided feedback on the interpretation of the results, and edited and provided feedback on the manuscript. A.D.H. edited and provided feedback on the manuscript and data interpretation, and provided resources for the headache-associated disability metric. G.K.A. was involved in conceptualization of the project, provided support with software development to analyze the data, was involved in interpretation of the light metric data, and edited and provided feedback on the manuscript.

Data availability

Raw data from this study are not publicly available due to legal limitations imposed by the consent form. Mean with standard deviation of different group comparisons are provided in the supplementary section.

Code availability

Matlab code used to analyze the data is available on a publicly available github repository: https://github.com/pattersongentilelab/visualDiet.

Competing interests

C.P.G.: Dr. Patterson Gentile receives salary support for the National Institutes of Health/National Institute of Neurological Disorders and Stroke (K23 NS124986) and the CHOP Institutional Foerderer grant. C.L.S.: Dr. Szperka has received research/grant support from the National Institutes of Health/National Institute of Neurological Disorders and Stroke (K23 NS102521), and PCORI. Dr. Szperka or her institution has received compensation for her consulting work for Eli Lilly, Teva Pharmaceutical Industries Ltd, Upsher-Smith Laboratories, LLC, and AbbVie. A.D.H.: Dr. Hershey or his institution has received compensation for serving as a consultant for AbbVie, Amgen, Biohaven, Eli Lilly, Lundbeck, Supernus, Teva, Theranica, and Upsher-Smith. His institution has also received research support from Amgen, Biohaven, Eli Lilly, Theranica, Upsher-Smith, and the NIH NINDS/NICHDS.G.K.A.: Dr. Aguirre receives funding/grant support from the National Institute of Neurological Disorders and Stroke, the National Eye Institute, and the Binational Science Foundation. R.S., B.M.P., and N.R. received salary support from the CHOP Institutional Foerderer grant.

Supplementary information

The online version contains supplementary material available at 10.1038/s44323-025-00056-y.