Introduction

Light is the most critical environmental factor for circadian rhythmicity and research over several decades indicates that manipulating light and dark exposure may improve clinical outcomes [1–3]. The study of chronotherapies has been galvanised by research demonstrating that the circadian effect of light on humans is primarily mediated by intrinsically photosensitive retinal ganglion cells (ipRGC) that have peak sensitivity to blue light [4–7]. This discovery indicates that it may be feasible to achieve clinical improvements by specifically blocking the blue part of the light spectrum in the evening, so-called virtual darkness, without recourse to prolonged sensory deprivation which is employed in dark therapies [8]. For example, several studies, including small-scale randomised controlled trials (RCTs), have examined the potential benefits of the adjunctive evening use of “blue blocking glasses” (BBG) compared with standard treatment for individuals with insomnia, delayed sleep phase, major depression, postpartum disorders, bipolar disorder, and mania [1,9–15]. At this stage, it is impossible to determine reliably whether the use of BBG is associated with robust, clinically significant effects in certain clinical subgroups. In many instances, the studies recruited small convenience samples (typically 5 to 20 participants selected by the investigators), some intervention groups included cases with a range of different primary diagnoses or problems, some studies lacked a control or comparator group and/or the analyses were reported only for completers. Even allowing for methodological limitations, the interpretation of the benefits of BBG is hampered by the fact that different researcher groups used different rating scales to measure outcomes, e.g., some focused only on changes in sleep patterns, other focused on mood or social functioning. The studies focused on mood disorders showed some putative benefits, but effect sizes varied widely and those studies reporting the largest effects often recruited smaller or less representative clinical samples. Overall, we suggest that cross-study findings are inconsistent in terms of the magnitude of any benefits of BBG or the specific nature of any benefits (BBG could improve sleep or mood but might influence other acute psychiatric symptoms, such as agitation and aggression or suicidality, across a broad range of mental disorders [1,9–15].

Another issue in optimising the effectiveness of BBG and other individual chronotherapy interventions is that many individuals, especially acutely mentally ill inpatients, struggle to adhere with the interventions or lack capacity to participate in research. One way to reduce reliance on individual behaviour and therefore increase access to chronotherapy for severe mental disorders is to consider changing the environment in which the individual is residing. For example, some new-build general hospitals now employ dynamic “circadian” lighting systems that change light spectrum and intensity according to time of day, e.g., evening blue-depleted light [16]. A particular appeal of such innovations is that, once installed, the intervention is “low intensity” in as much as no additional clinical staff is required to manage the system. Our own research has shown that changing the evening light environment in an acute psychiatric unit has positive effects on the circadian, sleep and neurocognitive systems of healthy young adult volunteers [17,18]. However, it is unknown whether any change would translate into better clinical outcomes. Several mental health inpatient projects are underway, but there is limited research on the potential benefits of exposing trans-diagnostic inpatient populations to dynamic, programmable lighting conditions. To our knowledge, only 2 small-scale RCTs have been published [19,20]. Canazei and colleagues (2022) used actigraphy to monitor sleep and rest-activity rhythms in 30 individuals with depressive disorders admitted to inpatient rooms with circadian or standard lighting conditions. The study showed that although exposure to the experimental lighting was associated with some significant improvements in selected sleep and circadian metrics, the duration of admission did not differ between groups (20 versus 21 days). Okkels and colleagues (2020) randomly allocated 54 individuals admitted to an affective disorder unit to a pre-set circadian lighting condition where light intensity and spectrum is changed throughout the 24-h cycle or standard lighting environment. Most inpatients had an International Classifications of Diseases (ICD)-10 diagnosis of depression, but a minority had mania, anxiety, or personality disorders. The intervention group demonstrated nonsignificantly greater improvements in sleep quality and other clinical symptoms, and no differences in length of stay (22 versus 19 days in the control group).

In summary, uncertainties exist regarding the potential extent or magnitude of any improvements in clinical outcomes, including duration of general psychiatric acute admissions, that may be associated with exposure to evening blue-depleted light alongside usual inpatient care and treatment. The only consistent published finding is the lack of clinically significant side effects or adverse events. However, the small sample sizes and selective recruitment strategies mean that many inpatient subpopulations were underrepresented (e.g., psychotic disorders) [21]. Lastly, the trials of acutely ill patients employed eligibility criteria that de facto limited the generalizability of findings to real-world settings, such as the permanent exclusion of inpatients with severe symptoms, complex presentations, or impaired capacity to give immediate consent.

Given the above, we employed an ethically approved, deferred consent procedure that permitted acute admissions to be randomised to adult inpatient care in a ward with programmable dynamic lighting with a blue-depleted evening light environment (experimental group) or a similar ward with standard light environment (control group). This article reports the key findings. We aimed to examine any differences in mean duration of admission in days per individual according to group allocation (i.e., evening light environment) as the primary outcome, and any between-group differences in clinical symptom and function ratings, risk of or actual incidents of harm to self or others, patient satisfaction, and side effects as secondary outcomes.

Methods

The reporting of this single-centre, two-arm, parallel-group, pragmatic effectiveness randomised controlled trial follows the CONSORT guidelines [22] (S1 CONSORT Checklist). The trial was an investigator-initiated study sponsored by St. Olavs Hospital, Trondheim University Hospital, Trondheim, Norway.

Here, we give an overview of the trial methodology while the online Supporting information provides other relevant information including the CONSORT checklist and further details of consent procedures, data management, and analyses (S1–S3 Files, S1 CONSORT Checklist, S1–S6 Tables, and S1 Fig).

Experimental and control conditions

The construction of the 40-bedded psychiatric inpatient unit was initiated in 2015 and finished in December 2017. The unit is divided between 2 wards built around 2 atriums (Fig 1). Each hospital ward has the same staffing levels, layout, and facilities and 5 rooms in each ward are designated as “psychiatric intensive care” beds. The staff rotated between the wards every 6 weeks.

The lighting fittings and fixtures were identical in both wards (Glamox AS, Oslo, Norway), while the diodes were different. From 07:00 h to 18:00 h, the light spectrum and intensity was similar in the experimental and control conditions. From 18:00 h to 07:00 h, the light intensity (photopic lux) was similar in both wards, but individuals were exposed to a different spectrum of evening light in each ward. While light exposure will vary with the direction of gaze, we performed light assessments using a Mavospec Base light metre (Gossen Foto- und Licht messtechnik GmbH, Nürnberg, Germany) horizontally at eye level (160 cm) at standardised locations in both wards. The melanopic EDI was assessed to be between 7 and 21 in different rooms of the blue-depleted ward, whereas it was assessed to be between 27 and 63 at the same locations in the standard light environment. The standard light environment, see Table 1 in Vethe and colleagues [18] for details about the composition of the evening LE in the 2 wards. We confirmed that LE was similar to Vethe and colleagues [18] before the recruitment was initiated.

1. (a) Experimental condition (Blue-depleted Evening Light Environment): The ward had tuneable light emitting diode (LED) fixtures and the amount of blue light in the ward was tested prior to commencing the RCT. At 18:00 hours (h), the lighting underwent a 30-min transition during which the green and blue LEDs were dimmed to produce blue-depleted amber coloured lighting. At 06:50 h, a 10-min transition programme changed the light colour to normal indoor lighting (3,000 Kelvins of colour temperature) which then continued until 18:00 h. The light intensity was dimmed to 20% (of the maximum) from 23:00 h to 06:50 h.

As well as the LED system, blue-blocking window filters were deployed in the evening. All television sets had permanent blue-blocking filters, and the outdoor area had external lights that block blue light. Use of electronic media was not restricted (unless an individual treatment plan limits access), but patients were provided with blue-blocking screens that could be attached to the front of all electronic devices (lowbluelights.com). If a patient left the blue-depleted unit after 18:30, they were offered blue-blocking glasses to wear (circadianeyewear.com).

1. (b) Control condition (Standard Light Environment): The ward had normal indoor hospital lighting installed. The light intensity was dimmed to 20% (of the maximum) during the night (from 23:00 h to 06:50 h). The light spectrum remained constant throughout the 24-h cycle at 3,000 K.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Unit layout.

(A) Shows the layout of the unit. The experimental ward is shown in yellow and the control ward in grey. The 40 patient rooms are located along the outer walls, and the common areas face the atriums and are divided into 2 wards. There are 8 rooms facing east in the experimental ward, whereas 6 rooms in each ward face north or south. All patient rooms are located on the top floor. The white area is reserved for dining and personnel rooms. (B) Shows a patient-room in the evening blue-depleted light environment.

https://doi.org/10.1371/journal.pmed.1004380.g001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Summary of participant demographics and clinical characteristics for patients in the Intention to Treat population, randomised to admission in a standard light environment or evening blue-depleted light environment.

https://doi.org/10.1371/journal.pmed.1004380.t001

Results

Fig 2 provides an overview of the flow of participants through the clinical trial. The core characteristics of the ITT sample and key clinical outcomes are summarised in Tables 1 and 2 and Figs 3 and 4 (see online Supporting information for additional details for all findings: S1–S6 Tables and S1 Fig).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. CONSORT flow-chart.

https://doi.org/10.1371/journal.pmed.1004380.g002

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Primary and secondary outcomes for patients (N = 476) in the Intention to Treat population, allocated to admission in a standard light environment (n = 244) or evening blue-depleted light environment (n = 232).

https://doi.org/10.1371/journal.pmed.1004380.t002

Clinical outcomes

As summarised in Table 2, there were no statistically significant differences in the duration of admission in individuals allocated to the evening Blue-depleted light environment (Mean: 7.1, CI [6.1, 8.1]; Median: 3.71) as compared with the Standard light environment (Mean: 6.72; CI [5.8, 7.5]; Median: 3.68) estimated mean difference: 0.4 days CI [−0.9, 1.9]; p = 0.523, nor in the effect of lighting environment in the duration of admission according to diagnostic subtypes (p = 0.22) or number of diagnoses per individual (p = 0.21) (Figs 3 and 4). See S3 Table for observed values and S4 Table for subgroup analyses for duration of admission. There were no relevant differences in duration of admission between the ITT and Per Protocol samples (see S5 Table) or after extending the follow-up period for participants who remained admitted after data collected concluded (S6 Table).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Proportion of patients by duration of admission for preselected diagnostic categories.

Red line and shaded area = Evening blue-depleted light environment with 95% CI. Blue line and shaded area = Standard light environment with 95% CI. (A) Data for all participants (N = 476). (B) Data for patients diagnosed with a psychotic episode (n = 87). (C) Data for patients diagnosed with a manic episode (n = 31). (D) Data for patients diagnosed with a severe depressive episode (n = 28). (E) Data for patients with all other diagnoses (n = 329). There were no statistically significant differences in the effect of the evening blue-depleted light environment across the subgroups (p-value for interaction = 0.239).

https://doi.org/10.1371/journal.pmed.1004380.g003

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Proportion of patients by duration of admission for number of psychiatric diagnoses.

Red line and shaded area = Evening blue-depleted light environment with 95% CI. Blue line and shaded area = Standard light environment with 95% CI. (A) Data for no psychiatric diagnoses (n = 33). (B) Data for patients with one psychiatric diagnosis (n = 268). (C) Data for patients with 2 psychiatric diagnoses (n = 129). (D) Data for patients diagnosed with 3 or more psychiatric diagnoses (n = 45). There were no statistically significant differences in the effect of the evening blue-depleted light environment across the subgroups (p-value for interaction = 0.209).

https://doi.org/10.1371/journal.pmed.1004380.g004

Compared with the Standard light environment, individuals allocated to the Blue-depleted light environment showed significantly higher improvement during admission (Clinical Global Impressions scale-Improvement, CGI-I, estimated mean difference: 0.28 (95% CI [0.02, 0.54]; p = 0.035), NNT: 12 (95% CI [6, 61]) and lower illness severity at discharge (Clinical Global Impressions Scale-Severity of illness, CGI-S, estimated mean difference: −0.18 (95% CI [−0.34, −0.02]; p = 0.029), NNT: 7 (95% [CI 4, 22)]).

The BVC ratings of aggressive behaviours were significantly lower in the Blue-depleted light environment compared with the Standard light environment. The estimated difference in predicted serious events per 100 days: −2.99 (95% CI [−5.00, −0.98]; p = 0.003), NNT: 9 (95% CI: [7, 15]).

There were no significant between-group differences in actual violence episodes (estimated mean difference: 0.07 (95% CI [−0.09, 0.23]; p = 0.414)), risk of suicide (estimated mean difference: 1.07 (95% CI [0.69, 1.67]; p = 0.752)), or the probability of changing from involuntary to voluntary status (estimated mean difference: 0.43 (95% CI [0.13, 1.42]; p = 0.167)). Likewise, side effects (estimated mean difference: 0.04 (95% CI [−0.09, 0.18]; p = 0.532)) (see also S1 Fig for details) and satisfaction ratings did not differ between groups (estimated mean difference: 0.10 (95% CI [−0.10, 0.31]; p = 0.322)) (Table 2).

Discussion

In this RCT, we tested if there are any benefits of modifying the evening light environment in an acute psychiatric hospital according to chronobiological principles. We found no differences between the Blue-depleted light environment and Standard light environment in our primary outcome of duration of admission. However, there were benefits observed by clinicians on ratings of clinical improvement (CGI) and aggressive behaviour (BVC). In terms of clinical significance, the NNTs for greater clinical improvements and lower levels of aggressive behaviour are similar to other low intensity interventions employed in psychiatry and in general medicine that are included in NICE guidelines. Further, these clinical benefits were achieved without any discernible differences in patient-rated side effects or treatment satisfaction levels. Moreover, we did not find any association with lighting conditions and transition from involuntary to voluntary status, nor any differences in actual violent incidents or suicidality, though the latter were low-frequency events, and it is likely that the trial was insufficiently powered to detect such events.

We acknowledge that our primary outcome measure and the design of this RCT was ambitious. Nevertheless, we believe these choices were justified as inpatient admissions are the largest contributor to the cost of psychiatric care. A reduction in occupied bed-days would not only be welcomed by patients but also offset the cost of new lighting systems. Further, we wanted to test if there were any benefits of the lighting conditions in a real-world clinical setting. A defining characteristic of a public health system that takes all admissions from a defined catchment area is that the inpatient population is heterogeneous, and the throughput of admissions is rapid, with short lengths of stay being the norm. This was true of our study setting with 1,118 admissions over 8 months and a median duration of admission of only about 4 days. Employing a pragmatic effectiveness design with deferred consent meant that this RCT recruited nearly half of all the patients admitted from a population of 300,000 people, including involuntary admissions and many other individuals with severe or complex problems who would normally be excluded from research studies. As such, this RCT has greater external validity, and findings are more generalizable, than efficacy studies undertaken in specialist clinics and other selective research settings that tend to recruit small homogenous samples.

While we set a very high threshold for finding between-group differences in the duration of acute admissions, it is notable that this is the third RCT exploring lighting in psychiatric inpatient settings that has failed to find a significant reduction in the length of admission [19,20]. Length of acute admission and decisions about hospital discharge are not only based on the clinician and patient observations regarding progress, but also on structural and practical issues, such as vacancies for suitable placements at other hospital facilities units and the need to accommodate new acute admissions. While the duration of admission was an obvious choice as the primary outcome, it is also true that these other factors may have reduced the precision of the findings. The absence of significant between-group differences in outcomes related to involuntary status, suicidality or violent incidents may be explained as these were low-frequency events, which reduced the statistical power to find differences. As such, we can only conclude that exposure to the blue-depleted light environment did not exacerbate any of these problems.

The key differences we identified between the light environments were that exposure to the blue-depleted light environment was associated with better clinical evaluations of outcome, greater improvement, and lower levels of aggression during admission. The NNTs were between 7 and 12 for these outcomes, which matches that of other low-intensity interventions in clinical psychology or medicine [39,40]. The CGI-S and CGI-I are widely employed in RCTs in inpatient units and have clearly defined anchor points [27,28,30,31], and the BVC is a measure of observed behaviour that has been validated and implemented in more than 20 countries [41]. Our findings, therefore, suggest some cause for optimism that modifying the evening light environment in a psychiatric inpatient unit could benefit a wide range of acute admitted inpatients. However, we propose two potential research directions that may enhance the outcomes in the future: (i) determining the optimal intervention dosage and enhancing adherence; and (ii) balancing global versus specific outcome measures and beginning to disentangle moderators and mediators of response.

We have previously shown that the same light environment as employed in the current RCT produced changes in sleep, circadian rhythms, and arousal in healthy young adults [17,18]. Further, there were no negative effects in that population, nor in nurses’ ability to work in the blue-depleted light environment at the present unit [42], or elsewhere [43]. However, it is possible that aspects of the “dosage” we employed are suboptimal. The evening light in the current trial was between 7 and 21 melanopic lux [18], whereas new guidelines suggest that 10 melanopic lux is the maximum intensity to avoid negative impact on the circadian system in healthy individuals [44]. The melanopic lux we employed is lower than in previous trials [19,20], but we were mindful that patients in an acute psychiatric ward may have an altered light sensitivity compared with the normal population [45–48]. Moreover, we note that Henriksen and colleagues found a large effect on manic symptom severity after 7 days of evening use of BBG as an adjunct to standard treatments, which is a longer exposure than most patients in our trial. It has also been argued that combining chronotherapeutic interventions may be synergistic [21,49], although our ambition was to test the isolated benefit of only modifying the evening light environment. It will be important for future research to test if changing the dose by increasing the number of days the patients are exposed and/or by adding other chronotherapeutic interventions, such as morning bright light exposure in non-manic patients, will influence key clinical outcomes.

On a related theme to dosage, is the issue of adherence with the intervention. Previous trials of psychiatric hospital lighting have been limited to single hospital rooms [20] or have had limited control over other light sources [19] that could impact adherence. We had a specific focus on securing adherence, such as installing the light environment in patient rooms, bathrooms, and common areas in unit, deploying filters in front of windows in patient rooms after 18:10 h and permanent filters on TV screens, and offering blue-blocking filters on mobile devices. However, due to the unit’s location close to the Arctic Circle, daylight varies substantially throughout the year. Therefore, we included patients over a fixed season from October to April as continued recruitment in the summer half-year would introduce a bias where we could not control the bright light exposure in the morning or evening. Although the RCT design would cause this to be evenly distributed in the two light environments, it could introduce bias specifically to the experimental condition in this study. First, patients would be exposed to daylight if they entered the atriums in the evening, which could reduce the effects of the evening blue-depleted light environment. To mitigate this issue, lights in the outside recreation/smoking area were blue-depleted and BBGs were available at the doors when leaving the ward. However, patients may have chosen not to use the BBG, and other adherence issues could have attenuated the observed effects. It is currently unknown how high exposure to ordinary evening light is sufficient to disrupt any clinical effect of evening blue depletion. However, a 15-s pulse of bright light exposure (9,500 lux) has been found to induce circadian phase delays in healthy individuals [50]. Secondly, there were 8 rooms in the experimental ward that faced east that could receive morning light exposure during the summer half-year, which could selectively impact the clinical outcomes for patients allocated to these rooms [51]. Finally, it would not be feasible to continue recruitment to an RCT in an acute ward during a time with vacation substitutes for the original staff. Thus, the trial results may not generalise to times of the year when sunset occurs late in the evening or at night.

Many RCTs of neuropsychiatric patients employ CGI ratings to measure outcomes, even in disorder-specific studies [27,29,52,53]. The advantage is that it is a clinically valid measure that clinicians can incorporate into day-to-day practice with minimal disruption [30]. These global outcome measures are especially useful in large-scale pragmatic RCTs with heterogenous samples. The 476 participants in the current trial reported 125 different ICD-10 primary and secondary diagnoses with many presenting with comorbid mental and physical disorders. It was unfeasible to employ disorder-specific rating scales for such a broad and complex range of clinical presentations. However, using global outcome measures to assess the benefits of relatively low-intensity interventions risks missing subtle but important effects the intervention may have on both specific symptoms and/or transdiagnostic processes such as sleep-wake regulation [13]. Further studies should strive to determine how specific patient characteristics, including mental disorder type, nature of sleep-wake disturbances, illness episode duration, and the interaction with certain medications [45–48,54,55] may influence the efficacy of the intervention. Additionally, age-related factors could also modulate the intervention’s effectiveness [56]. This complex array of potential moderators underscores the challenges clinicians currently face in assessing patient suitability for this intervention and may ultimately influence whether the intervention is employed in general psychiatric settings or restricted to specialist units.

In addition to the above issues, other limitations should be acknowledged. First, the nature of this study meant it was impossible to blind patients or clinical staff to the lighting condition, but although we cannot rule out a potential bias, we have no evidence that conscious or unconscious bias would explain these findings, either in the decision to discharge a patient or in any of the clinical evaluations. Moreover, we cannot know specifically the degree to which the intervention may have affected those recording the outcomes, as opposed to the subject participants. Related, we found significant differences in the observations of behaviours and evaluations of clinical staff, but we did not find significant differences in objective outcomes of violent incidents. However, violent incidents were low-frequency events to which the trial was insufficiently powered, and we note that although it did not reach significance, the number of registered incidents was 80% higher in the standard light environment. The primary outcome was heavily skewed. As described in the pre-publication Statistical Analytical Plan, this was expected, and bootstrapping was used to address this issue. The use of deferred consent meant that some patients left the unit before they could be approached regarding study participation and others did not sign the consent form. Therefore, about 50% of all admitted patients were included in the ITT population. There was a variable degree of data incompleteness at baseline, which may be attributable to the patients’ compromised capacity to provide data.

Changing the evening light environment moderately enhances the clinical outcomes of standard acute psychiatric hospital inpatient care and treatment with estimated NNTs between 7 and 12 for clinical state, improvement, and aggressive behaviours. These benefits, coupled with the absence of side effects and the low intensity of the intervention, indicate that there is a reasonable case for broader adoption of this strategy, particularly in new units where the lighting system is being installed for the first time and is not replacing previous systems. Of course, further research is required on the nature and magnitude of benefits of evening blue-depleted light environment and future research should consider whether the dose of light exposure and adherence to the condition improves patient outcomes, reduces length of stay, and/or identifies specific patient groups or symptoms that benefit more than others from admission in a blue-depleted light environment.

Supporting information

Acknowledgments

The authors would like to thank the patient user group, Regionalt brukerutvalg Helse Midt-Norge, for its contributions when designing the study, and the staff at the acute psychiatric unit at St. Olavs Hospital for their contributions to the study.