Design
A randomized within-subject, crossover trial design was employed. The protocol for each participant was 5.5 months in length, including 2-week baseline, two 2-month conditions (treatment and control), and 1-month follow-up (see Fig. 1 for study design). The study employed a cross-over design; thus all participants were exposed to both lighting conditions. There was no wash-out period between conditions, as carryover effects of the light were considered negligible, and the effect of either lighting condition was considered to be removed with the removal of the lighting [30]. Sinclair et al. [25] obtained a large effect for the primary outcome. A power analysis (G*Power [31]) undertaken with power (1-β) set at 0.80 (with α = 0.05; [32]) to detect a medium effect size (dz = 0.60) showed a required sample size (within-subjects) of 24.
Participants
Participants were individuals with mild-severe TBI or stroke sustained at least 3 months earlier, living in the community. Inclusion criteria included documented history of mild-severe TBI, or stroke, and self-reported significant fatigue (Fatigue Severity Scale ≥ 4). Exclusion criteria included other medical illness causing fatigue, including other neurological disorders, pre-injury sleep disorders, including obstructive sleep apnea [33] or chronic fatigue syndrome, presence of visual impairments affecting sensitivity and response to light, transmeridian travel within preceding six weeks, current use of prescribed and over-the-counter sleep medications and inability to give informed consent as assessed by the referring clinician. Use of antidepressants was permitted (n = 5) provided a stable dosage was maintained throughout the study.
Procedures
The study was approved by human research ethics committees at Epworth Healthcare (#EH2016-164) and Monash University (#9246). Participants provided written informed consent. No compensation was provided to participants. The study adheres to CONSORT guidelines.
Participants were recruited by hospital or community clinicians, from routine follow-up of people with TBI in a longitudinal outcome study and via advertisement within stroke support organizations. Interested individuals received a study explanation and eligibility screening. Following consent, injury details were obtained from medical records, including injury date, initial Glasgow Coma Scale (GCS) score, post-traumatic amnesia (PTA) duration, other injuries, MRI/CT scan results for TBI patients and for stroke patients date and nature of stroke, CT scan, treatment and medication details. Outcome measures were administered at baseline and monthly intervals (mid- and end of Treatment/Control condition), and one-month follow-up. Participants completed daily sleep and activity diaries, and wore an actigraphy device daily throughout the study.
For randomization, an independent researcher used online randomization software (www.randomization.com), based on random permuted block sizes of two and four, and transcribed allocation sequences onto cards in sealed envelopes, opened after baseline assessment by the study coordinator. Assessments were conducted by a researcher blinded to the lighting condition being received.
Lighting intervention
Participants’ current lighting was assessed before study commencement. Priority for both Treatment and Control lighting installation was given to rooms in which participants spent most time (e.g., lounge, kitchen, bedroom, bathroom). The Colormunki Light Meter (X-Rite, Grand Rapids, MI, USA) was used to measure participants’ home lighting conditions (specific spot measurements at fixed height in vertical (54″) and horizontal (72″) planes) and data analyzed using f.luxometer software (f.lux, Los Angeles, CA, USA). We utilized recently published International Commission on Illumination (CIE) International Standard CIE S026/E:2018 to quantify the lighting [34]. The melanopsin photoreceptor predominantly mediates non-visual responses, and changing these levels was the target of the study. Equivalent Daylight (D65) Illuminance (EDI) was calculated for each photoreceptor, including melanopsin as well as the melanopic Daylight Equivalent Ratio (DER), which expresses melanopic EDI as a function of photopic illuminance (lux); higher melanopic DER values represent relatively greater melanopsin stimulation.
The active lighting intervention consisted of short-wavelength enriched high-intensity white light with correlated color temperature (CCT) of approximately > 5000 K during the day. In the evening, for 3 h prior to sleep, participants were instructed to change which lights they used to reduced intensity, short-wavelength-depleted white light (< 3000 K) provided. Participants were asked to maintain as stable as possible lighting schedule day-to-day, and light exposure was timed relative to individual sleep patterns. Lighting fixtures and lamps were selected to integrate with participants’ existing lighting arrangements. Under some circumstances, fixed spectrum lighting was used, using the concept of ‘day’ and ‘evening’ light. For example, if two circuits existed in a room, one circuit was reserved for day-time high intensity blue-enriched light (e.g. ceiling light) and another for evening and night-time with dimmer blue-depleted light (e.g. table lamp). If this was not possible, table and bedside lamps were provided for evening use. Tunable and programmable lamps were programmed to change lighting automatically at the right time of day (Smart Wi-FI LED Bulb, TP-Link, Shenzhen, China; Genesis DynaSpectrum HealthE LED Lamp, Lighting Science, RI, USA). Participants were given written and verbal instructions on use and timing of lights for each condition. In the sham control condition, lamps were changed as per Treatment condition but did not change in color temperature or intensity from participants’ normal lighting (typically ~ 3000-4000 K). All lighting was commercially available and within safety standards for residential lighting. A qualified electrician changed light fittings or bulbs in participants’ homes. At each monthly visit, participants were asked to reflect upon their compliance with treatment lighting when at home, and transitioning from day to nighttime at the designated hour. Further details of the lighting protocol are documented in a separate paper, with two case studies illustrating our approach (Connolly et al., submitted).
Measures
Baseline and Screening Measures included the following:
Demographics questionnaire: age, gender, educational history, occupational history, ethnicity, living circumstances and whether the individual had a bed partner.
Medical records: date of injury, initial GCS, duration of PTA, other injuries, MRI/CT scan results, date and nature of stroke, treatment and medication details.
Outcome measures
All measures had been previously used in TBI and stroke populations. The primary outcome measure was the Brief Fatigue Inventory (BFI) [35], a 9-item scale used to capture current fatigue levels on a scale of 0 (no fatigue or does not currently interfere) to 10 (bad fatigue that completely interferes with activity/work) in the previous 24 h (state fatigue).
Secondary Outcomes included the following measures.
Fatigue Severity Scale (FSS) [36] is a 9-item self-report measure assessing impact of fatigue on daily activities or trait-like fatigue, on a 7-point scale from 1 (strongly disagree) to 7 (strongly agree). A mean item score ≥ 4 indicates clinically significant fatigue.
Epworth Sleepiness Scale (ESS) includes 8 items assessing a person’s likelihood of falling asleep during everyday activities such as “Watching TV” or “Sitting quietly after a lunch without alcohol” [37]. Score > 10 suggests clinically significant daytime sleepiness [37].
Pittsburgh Sleep Quality Index (PSQI) assesses subjective global sleep quality [38] in past month (e.g. bedtime, sleep duration) and frequency of problems interfering with sleep. Lower scores indicate greater sleep quality. Scores ≥ 8 indicate clinical insomnia [39].
Insomnia Severity Index (ISI) [40] screens for insomnia with 7 questions rated on a 5-point scale, (0 = no problem, 4 = very severe problem). A score of 8–14 indicates subthreshold insomnia, 15–21 clinically moderate, and ≥ 22 severe clinical insomnia.
Psychomotor Vigilance Task (PVT) (10 min) measured reaction time (MATLAB v. R2018b) once between 10am-5 pm at each of the assessment timepoints [41]. Prior research has demonstrated exposure to short wavelength light, decreases reaction time and errors on this task [42].
Hospital Anxiety and Depression Scale (HADS) measured self-reported depression symptoms [43] with the depression (HADS-D) subscale. The 14 items are rated on a 4-point scale, where 0 = “Not at all” and 3 = “Most of the time”.
Participation Objective Participation Subjective (POPS [44]) assessed community, work and social participation, with higher scores indicating greater participation.
Activity diary. A customized activity questionnaire was completed daily at 9 pm and used to calculate number of minutes spent on activity, rest and sleep in four time-blocks (between 9am and 9 pm). Activity encompassed physical and mental activity (e.g. doing chores, reading), rest was “giving the body a break” (e.g. lying down, listening to music) and sleep included napping. Percentage of daily productive activity was calculated as time spent on physical and mental activity relative to rest or sleep.
Side Effects Questionnaire was used to capture side effects experienced, including headache, nausea, cognitive changes, and appetite, at each assessment.
End of Light Therapy Questionnaire was completed at follow-up, to capture participants’ qualitative experiences of the lighting interventions and subjective symptoms.
Actigraphy and sleep diary. Participants recorded sleep and wake times and other sleep phenomena, in a daily sleep diary throughout the study, including time to fall asleep, awakenings after sleep onset, and daytime naps. They also wore wrist actigraphy devices (Actiwatch-2, Actiwatch Spectrum or Actiwatch Spectrum Plus; Philips Respironics, Bend, OR, USA) on the non-dominant wrist, with activity measured in 1-min epoch as sleep or wake. Actigraphic sleep parameters were analysed for sleep episodes identified in the sleep diaries.
Data analysis
Thirty participants were enrolled and 28 randomized, with 24 completing the study and included for analysis (see Fig. 2 CONSORT chart). All variables met assumptions of linearity, homogeneity of variance and had normally distributed residuals. A linear mixed-model analysis was used to model each outcome variable as a linear function of treatment (i.e., end treatment or end control), period (i.e., differences between condition 1 and condition 2 for treatment–control and control-treatment) and sequence (i.e., participants allocated treatment–control vs. participants allocated control-treatment), with participant included as a random variable. The analysis controlled for baseline scores and injury type (TBI and stroke). Random effects were included for participants intercepts. The primary indicator of a treatment effect was interaction of time by treatment group. Results were considered significant if two-tailed p value was < 0.05. Data analysis was performed using RStudio [45] and lme4 [46].
Actigraphy data analysis
Individual actigraphic sleep episodes were inspected and aligned with sleep diaries by inputting of subjective sleep and wake times by an independent researcher blinded to study conditions. When discrepancies ≥ 60 min between actigraphy data and sleep diary entries were identified, the following approach was used [47]: If subjective bedtime was reported as ≥ 60 min before a substantial reduction in activity and light levels, bedtime was adjusted to the time of decreased activity and light; if reported wake time was ≥ 60 min after a substantial increase in activity and light, wake time was shifted to the start of the sustained activity and light increase. Actigraphy data was excluded from analysis in cases of equipment malfunction or insufficient data to determine night time sleep episodes. In total, there were 663 (17%) nights missing (total of 3976) across all study periods; 155 (12%) missing from Treatment, and 171 (15%) from Control.
The following six outcomes were derived, averaged across each study condition for each participant: total sleep time (TST), sleep onset latency (SOL), wake after sleep onset (WASO), sleep efficiency (%), and participants’ average sleep and wake times. Actigraphy-derived sleep parameters obtained during baseline and mid to end condition periods were utilised in linear mixed-model analyses.