Effects of Dynamic Bedroom Lighting On Measures of Sleep and Circadian Rest-Activity Rhythm in Inpatients With Major Depressive Disorder

Bright light therapy is an effective treatment option for seasonal and non-seasonal affective disorders. However up to now, no study has investigated effects of dynamic bedroom lighting in hospitalized patients with major depression. A bedroom lighting system, which automatically delivered articial dawn and dusk and blue-depleted nighttime lighting (DD-N lighting) was installed in a psychiatric ward. Patients with moderate to severe depression were randomly assigned to stay in bedrooms with the new lighting or standard lighting system. Patients wore wrist actimeters during the rst two treatment weeks. Additionally, hospitalization duration and daily psychotropic medication were retrieved from patients’ medical charts. Data from thirty patients were analyzed. Patients under DD-N lighting generally woke up earlier (+20 minutes), slept longer (week 1: +11 min; week 2: +27 min) and showed higher sleep eciency (+2.4%) and shorter periods of nighttime awakenings (-15 minutes). In the second treatment week, patients started sleep and the most active 10-hour period earlier (-33 min and -64 min, respectively). This pilot study gives rst evidence that depressed patients’ sleep and circadian rest/activity system may benet from adjunctive bedroom lighting when starting inpatient treatment.


Introduction
Altered alertness and disturbed circadian rhythms as well as sleep problems are often reported in person with affective disorders 1 and may play a crucial role in the pathogenesis of this disease 2 .
Light acutely alters vigilance in humans 3 . Timed light exposure additionally has the potential to resynchronize the circadian clock 4 and to improve sleep quality 3 . Thus, timed light exposure is a promising non-pharmacological intervention option in the treatment of affective disorders.
Today, light therapy is a rst-line treatment for seasonal affective disorders 5 (SAD), and may be used in treating patients with moderate to severe depression 6,7 . Moreover, combined medication and light therapy showed greater e cacy than pharmacotherapy alone in depressed patients 6 . Despite these promising results, light therapy is still underused in clinical settings today.
Light therapy comprises a daily morning, polychromatic white light exposure of 5000 lux-hours, is fastacting in its antidepressant effect 1,8 , and has a minor side-effect pro le (e.g., headache, eyestrain, nausea, and agitation) occurring in up to 45% of patients only within the rst treatment days 9 .
In 1989, Terman and colleagues 10 conducted a pioneering study and reported bene cial effects of arti cial dawn and dusk in three SAD patients. Later, two landmark studies provided evidence that arti cial dawn is as effective as light therapy in the treatment of SAD 11,12 .

Page 3/24
Dawn simulators are installed at the bed head-side and deliver timed, gradually increasing light intensities in the early morning. They differ in several technical aspects to light therapy devices: light exposure occurs during the sleep-wake transition period over a time span of 30 minutes to 3 hours and light intensities are gradually rising and peaking at the time of planned awakening with 250 to 1,000 lux (applied light dose in dawn simulation is thus substantially lower than in light therapy). Dawn simulators are further perceived as a more natural light intervention, deliver light without side-effects, and thus may be easier and more convenient to use than light therapy devices 13 .
Research in healthy people has broadened our understanding of arti cial dawn effects: it improves sleep inertia [14][15][16] and cognitive performance after awaking 15,17 . In addition, arti cial dawn induces some physiological effects, e.g., a phase advance of the melatonin rhythm 18 , an accelerated decline in skin temperature 16 and a reduced heart rate gradient during sleep-wake transition 19 .
Besides dawn, dusk is a second important natural cue for photic entrainment feeding into the evening oscillator of the biological clock 20 . Although the pioneering study of Terman and colleagues 10 used arti cial dawn and dusk, effects of arti cial dusk were scarcely investigated. To the best of our knowledge only Danilenko and Hommes 21 have compared effects of a rectangular lights on/off pulse with a dusk simulation in the evening in healthy subjects, showing reduced body movements in the rst 40 minutes of sleep but no phase shifting of the circadian clock under arti cial dusk.
Moreover, up to now, only two eld studies have investigated effects of a combined arti cial dawn and dusk intervention 22,23 . In these studies, persons with dementia living in nursing homes were exposed to arti cial dawn and dusk that was tted to individual sleep times. In a rst study 22 , persons with dementia had an earlier sleep onset, shorter sleep latency, longer sleep duration and decreased nighttime activity after three intervention weeks. In a second study 23 , patients with dementia reported better mood upon awakening under arti cial dawn and dusk. However, in this study no effects were observed on actigraphically measured circadian activity rhythm and sleep parameters.
In addition to light therapy and arti cial dawn and dusk, nocturnal lighting may affect humans. Research has shown that blue-depleted nighttime lighting may reduce nighttime arousal 24 and minimize disruptions of circadian and sleep parameters in healthy people [25][26][27] .
It is worth noting, that blue-depleted nighttime lighting can also be generated by wearing blue-light blocking glasses. In a recently published systematic review 28 it was shown that this intervention decreased manic symptoms in bipolar patients but delivered inconclusive results in improving mood in patients with major depression. However, clear evidence was found in this review of blue-light blocking glasses for a reduction in sleep latency in patients with sleep disorders. So far, research has shown bene cial effects of arti cial dawn in SAD patients and of arti cial dawn and dusk in patients with dementia. No study has yet investigated effects of a dawn-dusk simulation and blue-depleted nighttime lighting in hospitalized patients, in particular in patients with affective disorders.

Page 4/24
The present prospective pilot study closes this gap. Outcome measures were actigraphically recorded sleep and circadian rest-activity rhythm parameters and clinical health data (i.e., medication and length of hospitalization). Based on current knowledge, we hypothesized an improved sleep quality (i.e., shorter sleep onset latency and higher sleep e ciency) and more stable circadian rest-activity cycles in hospitalized patients residing under this dynamic bedroom lighting. In two double bedrooms in a psychiatric ward, seven tunable-white LED luminaires (WS14-01 FFF LED wallwasher, Projektleuchten GmbH, Dortmund, Germany) were installed at the upper wall opposite to the patients' beds (see Supplementary Materials S1). Each luminaire comprised 24 LEDs (12 Luxeon Z-ES 2200 Kelvin and 12 Luxeon Z-ES 4000 Kelvin). In addition, bed-side reading lights (research prototypes, Projektleuchten GmbH, Dortmund Germany), containing 10 LEDs (Luxeon Z-ES 2200 Kelvin), were installed in the bedrooms as well (see Supplementary Materials S2). To guarantee a consistent lighting design for the whole private area of the patient, the adjacent bathroom lights were further equipped with tunable-white LED luminaires (research prototypes, Projektleuchten GmbH, Dortmund Germany), providing the same light colors at the same daytime period as the bedroom lights. Ambient bedroom and bathroom luminaires were DALI-dimmable and controlled by a central lighting control system (Tridonic GmbH Co KG, Dornbirn, Austria). The reading light was manually adjustable in its intensity by a dimmer installed nearby the bed.
The lighting control strategy was aligned to the daily life of the patients in the psychiatric ward.
Speci cally, patients leave their beds at 07:00h (or were woken by the staff at the latest at 07:00h), so daytime activities (e.g., morning toilet and dressing) typically start at 07:00h, followed by breakfast at 7:30h (served in common rooms outside the bedrooms). Typically, patients reenter their bedrooms during short therapy breaks in the morning and afternoon and after lunch, but stay in public clinic areas during the day otherwise. The bedroom is again the main place of residence until after dinner at around 18:00h.
Commonly, patients get into their beds between 20:00h and 21:00h and read, watch TV or relax. Usually, patients start sleeping after the last caring procedures is nished at 22:00h.
In line with these social routines and to generate a smooth light-induced sleep-wake transition period by ambient bedroom lighting, in the present study arti cial dawn began 35 minutes before daily activities started (at 06:25h) and was implemented with the ambient lighting system. During arti cial dawn, light intensities gradually increased from 0 to 209 lux (horizontally at bed's height) accompanied by risen color temperatures from 2192 Kelvin to 3588 Kelvin. At 07:00h, the light intensity and light color was at its maximum. To stop arti cial dawn, patients had to leave their beds, as the light switch was installed at the entrance of the bedroom.
During the day (from 07:00h to 20:00h), ambient bedroom lighting must be switched manually and provided 209 lux at bed's height with a color temperature of 3588 Kelvin. After 20:00h, ambient room lights were again dimmed automatically (arti cial dusk) in its illuminance and color temperature over a period of 2 hours. At 22:00h, room lights provided 100 lux at bed's height with a color temperature of 2192 Kelvin. In case, patients preferred to use reading lights for zonal illumination of the bed-side area, they were able to stop the arti cial dusk at any time. Reading lights provided 20 lux at bed's height with 2192 Kelvin and could be manually dimmed.
If necessary, room lights could be switched on during the night at any time and provided 100 lux at bedside level with 2192 Kelvin.
The lighting control strategy is graphically summarized in Figure 1. A comprehensive summary of photometrical measures generated by the DD-N lighting system at speci c periods in time, is given in Table 1.

Standard bedroom lighting
Lighting design in the standard double bedrooms followed European lighting standards for hospitals (EN DIN 12464-1) and was implemented with two customary luminaires (‚Pureline Basic', Zumtobel AG, Dornbirn, Austria), which were installed at the wall above the patient`s head. Each luminaire housed two uorescent lamps (Osram T16 39W/840 G5 HO). Reading lights comprised the same uorescent lamp technology and offered 52 lux horizontally (measured at the height of the bed). Following recommendation from Spitschan 29 , photometrical measures are again summarized in Table 1.

Dimensions of the bedrooms and daylight penetration in the bedrooms
All four study bedrooms (two bedrooms with DD-N lighting and standard lighting, respectively) had the same room area (length: 4.60m; width: 2.9m), were located on the second oor and oriented to the east.
Each bedroom had one window with an area of approximately 1.60 square meters. Tall, densely grown trees reduced daylight levels in the bedrooms and direct sunlight penetration in the morning. The daylight factor (i.e., the amount of daylight at 1-meter distance from the window in the bedroom on overcast days) in all four bedrooms was low (1.2% ± 0.8). optic; the alpha-opic irradiances were derived from the spectral irradiance measurements excel-toolbox S026-2018 provided by the CIE; * typical switching times of ambient lights and reading lights were determined from staff's observations; tabulated light spectra are given in Supplementary Materials Table  S1 and Figure S3.

Actigraphy
The rest-activity cycle is a core behavioral measure of the circadian system and often disturbed in depressed patients 30 . An actimeter is a body-worn device, which comprises an accelerometer and allows the measurement of physical activity over periods of days or weeks. By means of these activity data sleep quality parameters 31 and circadian activity rhythm parameters 32 can be calculated. Due to the fact, that sleep parameters derived from actimetry are similar to polysomnography (validation shows high sensitivity but moderate to low speci city 33 ), this method has often been employed in sleep and light impact research 34 .
In the present study, we used a wrist-worn, light-weight (19 grams), water-resistant device (ActiGraph wGT3x-BT; Pensacola, FL, USA). The device was worn for the rst two treatment weeks and recorded three-dimensional activity data with a sample rate of 30 Hz. Additionally, the device included a light sensor, which measured illuminances at wrist level of up to 5,000 lux. To minimize non-wearing times, the actimeter was attached with a non-removable textile bracelet. We used the software ActiLife (Version 6.8.1) to retrieve and process actimetry data (i.e., we aggregated data over time periods of 1 minute) and to run the Cole-Kripke sleep scoring algorithm. For the sleep analysis, usually a daily bed time protocol is needed. In the present study, bed times were estimated from activity data: the individual starting time in bed was de ned as the beginning of the latest 10-min period between 20:00h to 23:00h with lower physical activity compared to the mean daytime activity level of the subject; the individual end time in bed was de ned as the end of the earliest 10-min period between 06:00h to 8:00h with lower activity compared to the mean daytime activity level of the ongoing day. The following sleep parameters were derived for each night: total sleep time (TST; hours), sleep onset latency (SOL; minutes), sleep e ciency (SE; percent), number of wake episodes (WASO; count data) and sleep onset and offset times (clock times). For statistical analysis, we calculated mean values of these sleep parameters separately for the rst and second treatment week. To derive circadian activity rhythm parameters, we applied a nonparametric analysis routine 35 to weekly activity data and calculated the following parameters also separately for each of the two treatment weeks: intra-daily variability (IV), inter-daily stability (IS), relative amplitude (RA), most active 10-hour period (M10), clock time when M10 starts (M10 onset), least active 5-hour period (L5), and clock time when L5 starts (L5 onset).
To compare light exposure levels between the two intervention groups, daytime-speci c light exposure levels were additionally retrieved from the wrist-worn actimeters.

Hospital stay and psychotropic medication
Length of hospitalization, daily non-pharmacological treatments (i.e., ergotherapy, physiotherapy, and psychotherapy) and psychiatric medications (i.e., antidepressant, antipsychotic, and sedative medication) were retrieved from the patients' medical chart. To compare treatments with different antidepressants, we calculated equivalent doses referring to Citalopram 36 . A similar procedure was applied to antipsychotic medication 37 and sedative medication 38 with the reference agents Chlorpromazine and Diazepam, respectively. Due to the fact, that pro re nata (PRN) medication was used occasionally (on 28 out of 420 treatment days) in few study participants (DD-N lighting: n=4 in the rst week and n=4 in the second week; standard lighting: n=2 in the rst week and n=4 in the second week), we did not extract PRN medication data from patients' health records.

Study protocol and data collection procedure
Clinical effects of light therapy can be seen in SAD patients after the rst treatment week 13,39 . Even for patients with non-seasonal depression, light therapy induces a strong reduction of symptoms within two weeks 6 . Based on these results and the fact, that depressed patients stayed at the psychiatric ward for an average of 17.5 days ± 7.9, the participants continuously wore a wrist actimeter on their non-dominant hands from the rst day of admission up to fourteen days.
Study participants were patients of a psychiatric hospital with an ICD-10 diagnosis of affective disorder (F32, F33, and F34; ICD-10 40 ). Moreover, patients had to suffer from moderate to severe depression, quanti ed with a Beck Depression Inventory score (BDI-V) 42 of at least 35 at admission. Moreover, inpatient treatment should last at least for 14 days. Patients showing acute suicidal ideation at admission were excluded from study participation.
Written informed consent was given by all subjects prior to study inclusion. The study protocol strictly followed the guidelines outlined in the 1964 Declaration of Helsinki and was approved by the ethical committee of the Medical University of Innsbruck, Austria.
In total, 187 patients were asked for study participation at admission by a psychiatrist over the course of one year. Many of these patients did not meet inclusion criteria or declined to participate (n=129). Consequently, 58 patients were randomly assigned to the two interventions (DD-N lighting: n=29; standard lighting: n=29). Data from 15 patients were lost to follow-up (due to unwillingness to wear the actimeter; usually after a wearing period of one or two days). Finally, thirteen patients had shorter inpatient treatment periods than 14 days and thus were excluded from analysis. Finally, data from 30 patients were subjected to analyses. A detailed study ow diagram of the present study is given in Figure  2.

Statistical Analysis
Descriptive statistics are given as mean and standard deviation (mean ± SD) and gures show averages and 95% con dence intervals of means. To compare demographic variables, length of hospitalization, and non-pharmacological treatments between the two intervention groups either a t-test, chi-squared test or exact Fisher test were run. Two-way mixed analysis of variances (ANOVAs) with the between-subjects factor light intervention and the within-subjects factor treatment week were used for the analysis of actimetric parameters, antidepressant and antipsychotic medication, and light exposure. Due to the fact, that data of equivalent doses of sedative medication differed on the rst treatment day, a mixed ANOVA, controlled for the equivalent dose of sedative medication on the rst treatment day, was run. Effect sizes (partial eta-squared) are given for signi cant effects. All analyses were run with a signi cance level of 0.05 (two-sided).
By means of the software package G*Power 43 (Version 3.1.9.4), a sensitivity analysis for a repeatedmeasures, within-between interaction model was run. For this model and under usual statistical assumptions (power = 80%, alpha = 0.05, correlation among repeated measurements = 0.5, and nonsphericity correction epsilon = 1), a sample size between 16 to 34 subjects allows the detection of moderate intervention effects 44 . This effect size was also found in light therapy studies with people with depressive disorders 6,7 .

Results
A total of 16 and 14 patients stayed in bedrooms with standard lighting and DD-N lighting, respectively. Gender distribution, mean age, distribution of ICD-10 diagnoses, and mean depression score (BDI-V) did not differ between the two groups (see Table 2).

Sleep onset latency (SOL)
We could not observe any intervention effect on SOL, all p > .10.  Interaction plots for all sleep parameters are given in the Supplementary Materials S4. In addition, sleep data for the rst three treatment days were compared between the two intervention groups (see Supplementary Materials S5), indicating that actimetric sleep data was comparable between the two groups within the rst inpatient treatment days.

Least active 5-hours (L5)
We could not reveal any intervention effect on L5, all p > .10.

Onset time of L5
Interaction between the factors light intervention and measurement time on onset time of L5 reached signi cance, F(1,28) = 4.250, p = .049; partial η 2 = .132. Post-hoc analyses revealed no differences in the onset times of L5 between the two interventions in treatment week 1, p > .10, but a signi cant difference in treatment week 2, F(1,28) = 4.321; p = .047, partial η 2 = .134, with an earlier onset time under DD-N lighting (23:10h ± 55min) compared to standard lighting (23:49h ± 46min), with a mean difference of 39 min. Under both interventions, onset times of L5 did not differ between the two treatment weeks: both p > .10.
In the Supplementary Materials S6a and S6b, interaction plots for effects on circadian rest-activity rhythm parameters are shown.

Antidepressant medication
An independent-samples t-test was run to determine if there were differences in antidepressant medication on the rst treatment day between the two intervention groups. The equivalent daily dose of Citalopram did not differ at the beginning of inpatient treatment for patients under DD-N lighting (20.82 mg ± 14.13) and standard lighting (20.74 mg ± 14.12), p > .10, see Figure 3.

Non-pharmacological treatments
Following a multimodal treatment approach, all subjects took part in ergo-therapies, physiotherapies, and psychotherapies on work days from 09:00h to 11:00h and 14:00h to 16:00h. The daily mean number of therapies did not differ between the two groups over the 14 treatment days; DD-N lighting: 2.5 therapies ± 0.5 and standard lighting: 2.5 therapies ± 0.4, t(28) = 0.185, p = .767. Additionally, both groups participated equally in the therapies: DD-N lighting: 20% ergo-therapy, 69% physiotherapy, and 11%

Discussion
The present study aimed at investigating dynamic ambient bedroom lighting effects in moderately to severely depressed patients at the beginning of inpatient treatment using wrist actigraphy and data from patients' medical charts. As expected, the bedroom lighting system signi cantly increased morning (from 6:25h to 7:00h) and decreased evening and nighttime light exposure (from 20:00h to 6:25h) compared to standard bedroom lighting. In addition, we observed sleep-related and circadian activity rhythm effects generated by the dynamic ambient bedroom lighting system. Those effects were likely not caused by differences in pharmacological and non-pharmacological inpatient treatments in the two intervention Research further showed that depressive disorders are associated with circadian disturbances such as a delayed circadian rhythm 48 . In addition, a delayed peak in daytime activity is typically found in phases of acute depression 49 . One suspected mechanism of action of morning light therapy is its potential to phase advance the circadian system 2 . This effect has already been shown in people with seasonal affective disorder 50 . However, to date there is inconclusive evidence that phase shifting of circadian rhythms is the mechanism of action in the treatment of non-seasonal depression 1 .
In the present study, with some delay (in the second treatment week), the onset times of the daily most active and less active periods were earlier under dynamic bedroom lighting indicating that the intervention not only has in uenced nighttime sleep but also circadian activity rhythm parameters. It should be mentioned, however, that sleep problems and circadian disturbances are mutually dependent 51 and thus reported light effects on the circadian activity rhythm can hardly be separated from light effects on nighttime sleep parameters.
Our study found no effects on inter-daily stability and intra-daily variability of activity rhythms, which is in line with results from a systematic review of Burton and colleagues 52 . It can be assumed, that hospitalization and consequently the social rhythm imposed to patients' daily activities may have masked potential light effects on the stability and variability of daily physical activity cycles.
Altered physical activity is a core feature of depression 46 and depressed patients often show lower daytime activity levels 49,52 and a damped circadian activity pro le 46 . During the course of treatment, these physical activity parameters usually improve 52 . We found similar results in our study. Irrespective of bedroom lighting, patients' daytime physical activity level and their daily activity rhythm amplitude improved over time, indicating a general inpatient treatment response.
In this study, patients under dynamic and standard bedroom lighting were prescribed equivalent daily doses of antidepressant and antipsychotic medication. Moreover, by controlling for group differences at the rst treatment day, both groups also did not differ in sedative medication across the fourteen treatment days.
It is well documented, that only about half of all depressed patients respond to antidepressant medication and about one third experience remission of symptoms 53 . Moreover, clinical response to antidepressant medication can often be observed after weeks 54 , which is problematic for adherence, particularly because signi cant side effects of antidepressant medication (e.g., digestive problems, appetite disturbances, sleep problems, dizziness, or agitation) frequently occur from the beginning of pharmacotherapy. Consequently, supplementary interventions with a fast antidepressant response are needed.
For a more immediate relief of speci c symptoms in depressed patients (e.g., distress, sleeplessness and restlessness), sedatives and antipsychotics are additionally prescribed 55,56 . It was recently shown, that a combined therapy (benzodiazepines + antidepressants) improve depression severity in the early phase of treatment (1-4 weeks) compared to antidepressants alone 57 . In addition, benzodiazepines ameliorate symptoms of insomnia e ciently. However, administration must be balanced judiciously against possible harms 58 .
There are also well recognized non-pharmacological interventions with immediate response in antidepressant treatment, such as electroconvulsive therapy 59 and partial or total sleep deprivation 60 .
However, these treatments are complex and can also have signi cant side effects. Due to its fast-acting response pro le, light therapy has been discussed as further, well-tolerated treatment option for depressed patients 6 . The most recently published meta-analysis con rms the effectiveness of light therapy in non-seasonal depression but also states that more research is needed in severely depressed patients 61 . The present study, for the rst time, provides initial evidence of a fast response in physical activity parameters of depressed inpatients to ambient light treatment.
We did not observe an intervention effect on the length of hospitalization. To date, research has clearly shown that increased sunlight exposure during the day reduced inpatient treatment duration in severely depressed patients 62-65 . It can be hypothesized that in our study the light intervention took place at a different time of day and thus may have not affected the length of hospitalization. Further studies are warranted to substantiate this assumption.
This study has several limitations. First, sample size was small. Although we included 58 patients in our study, only data from 30 subjects could be analyzed. Main cause of the high drop-out rate was the fact, that the wrist-worn actimeter was not well accepted. Moreover, a few patients reported that the actimeter disrupted nighttime sleep. Second, no further objective and subjective sleep measure (e.g., polysomnography, questionnaires) nor circadian phase marker (e.g., melatonin or core body temperature) were recorded to con rm results obtained from wrist actimetry. Third, self-rated depression symptoms were recorded at admission only. Fourth, actimetric data were only available after admission. However, by analyzing actimetry data within the rst three treatment days, we could not observe differences in sleep parameters between the two intervention groups.
To conclude, in this pilot study rst bene cial effects of dynamic bedroom lighting (arti cial dawn and dusk and blue-depleted nighttime lighting) on inpatients with moderate to severe depression were reported. The light intervention showed fast-acting effects on sleep and circadian activity rhythm parameters which cannot be explained by differences in medication, non-pharmacological treatments and daytime light exposure. Reported effects indicate that ambient bedroom light may have signi cant short-term effects in inpatient treatment of persons with major depression. Results further complement research on light therapy in non-seasonal depression. Larger studies are warranted to con rm results from this pilot study and to establish dynamic ambient lighting as an effective additional treatment option.
Declarations Figure 1 Light control strategy for the DD-N bedroom lighting system.

Figure 2
Study ow diagram.

Figure 3
Daily antidepressant medication.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download. supplementarymaterials.docx