Depressive disorders, both unipolar (MDD) and bipolar (BD), impact patients and society greatly. In bipolar depression and seasonal affective disorder the episodic nature and periodicity relate to changes in circadian rhythms. Bright light therapy (BLT) is thought to ameliorate symptoms of depression through its influence on circadian rhythms. Effectiveness of BLT has not been thoroughly established in real-world clinical samples.
MethodsFor seventy-four patients with depression Inventory of Depressive Symptoms – Self Rated (IDS-SR) scores were available through Routine Outcome Monitoring (ROM) used in BLT in the outpatient clinic for mood disorders. Patients received one or two weeks of add-on BLT as usual care. Patients suffering from MDD (n = 33, 60.6 % female, mean age 36.1 ± 11.5 years) were compared to patients suffering from BD (n = 41, 70.7 % female, mean age 45.0 ± 14.5 years) and changes in individual symptoms were analyzed for these two groups as well as the whole cohort.
ResultsIDS-SR scores decreased significantly in both groups of patients and did not differ in effect size between the groups. Explorative analyses of the effects on individual items of the IDS-SR showed that items related to core symptoms of depression such a as mood, concentration and energy level showed the largest improvements.
ConclusionSelf-report depressive symptoms in patients suffering from either MDD or BD decreased in this naturalistic cohort after receiving BLT.
Bright light therapy (BLT) is a potential treatment for bipolar depression (BD)1 as well as for major depressive disorder (MDD) with seasonal pattern.2 Both types of depression impact society and individuals greatly and tend to be treated with a combination of psychotherapy and pharmacotherapy.3 BLT may have additional value, as it is thought to ameliorate depressive symptoms through a different mechanism than other treatments.
Patients with MDD and BD often report sleep-related symptoms and both disorders are associated with disturbances in circadian rhythms.3–5 The suprachiasmatic nucleus (SCN) is known to be vital in regulating physiological circadian rhythms and behaviors.6 Light is received by the retina and communicated to the SCN via the retinohypothalamic tract and thought to contribute to the circadian rhythm. The recovery of circadian rhythm in turn is beneficial to several other interrelated systems, such as the autonomic central nervous systems and the sympatho-adrenal system.7 BLT is thought to restore normal function in these systems, contributing to the recovery from depression. The exact mechanism of BLT, however, remains unknown.
Patients with MDD with a seasonal pattern (previously called Seasonal Affective Disorder, or SAD) were first treated with bright light in the 1980′s, with positive treatment effects. Since the first study in 1984,8 several studies have assessed BLT in various patient groups. These studies were heterogenous in terms of methodology (e.g., placebo light groups, blinding), patient groups (e.g., BD, MDD, MDD with a seasonal pattern), and aspects of the intervention itself (e.g., treatment length, duration and intensity of the light, BLT as monotherapy or as an adjunctive to sleep deprivation or pharmacological interventions). This possibly resulted in a great diversity in study findings.
Meta-analyses combining evidence from these studies concluded that BLT using between 7500 and 10.000 lux was a safe treatment option for MDD9,10 and a depressive episode in patients with BD.11,12 BLT appeared to be more effective in alleviating depressive symptoms than dim or red light therapy and was also effective as an add-on therapy13 in depressed BD patients. However, another meta-analysis concluded that effectiveness of BLT in BD could not be robustly proven.14 These meta-analyses all mention heterogeneity of the studies as a caveat. Furthermore, recent changes in pharmacological treatment were often exclusion criteria in clinical trials. It is unclear whether the effectiveness of BLT found in clinical trials is generalizable to real-world settings in which patients often suffer from comorbid disorders and are treated with psychopharmaceuticals.
Hence, there is a need for real-life data to assess effectiveness in the heterogenetic group that is found in routine psychiatric care. In the present study we aimed (1) to assess the effectiveness of BLT in treating unipolar and bipolar depression in a naturalistic outpatient setting, and (2) to explore how individual symptoms of depression react to BLT in an outpatient setting for depression in BD and MDD. We hypothesized that items related to sleep might improve more than others, due to the nature of BLT15 and previous findings that BLT improved sleep quality in other patient groups.16,17
MethodPatients and procedureThis naturalistic study was performed by analyzing anonymized routine outcome monitoring (ROM) data (October 2014 to October 2022) from the database of the outpatient clinic for BD of the regional mental health institution Rivierduinen, Leiden, The Netherlands. In this outpatient clinic, adult patients suffering from MDD and BD are treated. BLT is an add-on treatment option and performed with 10.000 lux of white light (Sunray II, the Sunbox company, Los Angeles, USA18), five mornings a week for 30 min. The total treatment lasts for one to two weeks,1 depending on patients wishes and its effect (in previous years). Measurements took place at the outpatient directly before the start of the first BLT session and directly after the last session of each BLT week, i.e. after the completion of five or, in case of a two-week treatment, ten sessions. The BLT procedure is in concordance with the Dutch protocol for BLT in BD.1 BLT is only initiated after referral by a psychiatrist and in concordance with patient wishes.
InstrumentThe Inventory of depressive symptoms-self rated (IDS-SR)19 is part of the routine outcome monitoring (ROM) before the start of BLT and thereafter every week. The IDS-SR is a validated self-report questionnaire assessing 30 items related to symptoms of depression by scoring on a 4 point Likert scale from zero to three points per item. Its use has been validated in patients suffering from bipolar and unipolar depression.19 Clusters of symptoms can be identified; mood/cognition related symptoms, sleep related symptoms and anxiety related symptoms.20 Remission was achieved if the IDS-SR score was lower than 13 points, indicating the absence of a current depressive episode. Response was defined as >50 % reduction in IDS-SR score. All ROM-data was linked to demographic data, including known lifetime and current diagnosis and in some cases also use of pharmaceuticals.
AnalysisBased on their lifetime clinical diagnosis patients were divided in a BD and a MDD group. Patients suffering primarily from schizoaffective disorder were categorized among those suffering from BD. Treatment effectiveness was analyzed firstly by assessing the change in total IDS-SR-score, using firstly the t-test between the first and last score. IDS-SR scores were normally distributed. Remission and response were calculated from the respective baseline and final scores. The chi-squared test was used to determine whether there was a difference in remission and/or response rates. Using a mixed-effects model all available data points were used, allowing for the inclusion of all available data points, even when some were missing, without the need for imputation. Subgroup analyses were performed with adjustment for age and sex, where appropriate. Significance for changes in IDS-SR scores was determined using a two-tailed p-value threshold of <0.0125, adjusted for multiple testing with the Bonferroni correction. Next, multilevel regression analysis was performed to assess the change in the total score and in scores for separate items of the IDS-SR. In this analysis total and item scores were group-level standardized, and all available time points were used in case a patient had more than two IDS-SR measurements. Significance for item scores was determined based on a two-tailed p-value threshold of <0.001, adjusted for multiple testing using the Bonferroni method. This analysis led to the outcomes depicted in Fig. 2. Statistical analyses were conducted in SPSS version 29 and R version 4.3.1 (R Foundation for Statistical Computing, 2019) using RStudio (R Foundation for Statistical Computing, Vienna, Austria, 2016) with employing the packages ‘lme4’ (version 1.1–35.1) and 'lmerTest' (version 3.1–3).
ResultsIn total 74 patients, were included. Thirty-three (44.6 %) patients suffered from MDD, 18 (24.3 %) from BD type II depression, and 23 (31 %) from BD type I depression – which included three patients with schizoaffective disorder of the bipolar type (see Fig. 1). Fifty patients (67 %) had one or more comorbid psychiatric disorders, 26 patients had two or more comorbid psychiatric disorders. Most frequent were personality disorders (n = 25, 33.7 %) and anxiety disorders (n = 22, 29.7 %). Gender distribution was equal between the BD and MDD group, but age was significantly higher in the MDD group compared to the BD group (mean age 45.0 ± 14.5 years vs 36.1 ± 11.5 years, respectively, t = 2.85 p = 0.006).
Effectiveness of BLT in depression (MDD and BD)The average decrease between the first and last IDS-SR total score in the complete cohort was 11.8 ± 9.2 points, with a large effect size (Cohen's d = 1.0). The mean total score of the first measurement was 34.9 ± 10.5 and the mean of the last score was 23.1 ± 12.9. When all time points were fitted into the standardized multilevel regression model, the average decrease in the whole cohort was 12.1 points (t = −11.4, p < 0.001) which translates into a standardized mean decrease of 0.95 SD (p < 0.001; see Fig. 2). 25.7 % of patients achieved remission and an additional 6.8 % showed response.
Effectiveness of BLT in MDD versus BD and subgroups by age and comorbidity
IDS-SR scores in the BD group showed a mean decrease of 11.6 points (mean first IDS-SR 33.3 ± 10.0 vs mean last IDS-SR 21.7 ± 13.2; p < 0.001) and in the MDD group 12.0 points (mean first IDS-SR 37.0 ± 10.8 vs mean last IDS-SR 25.0 ± 12.5; p < 0.001). There were no significant differences among the two groups in the effects of BLT, neither in effect size nor in terms of response (BD 9.8% vs MDD 3.0%; p = 0.25). A relative higher percentage of patients in the BD group achieved remission compared to the MDD group (34.1% vs 15.2%) however, this did not reach statistical significance (p = 0.06). There were no significant differences observed when subgroups BD I and BD II were compared to the MDD group (Fig. 1). BLT was less effective in younger patients (age group 18–35 years). When stratified by the number of comorbidity diagnoses, there were no significant differences in treatment response, as shown in the forest plot (Fig. 1).”
Effectiveness of BLT on individual symptoms in all patientsThe decrease of the total IDS-SR score and individual symptoms is shown in Fig. 2. Mood items (colored blue in Fig. 2) generally showed the largest decrease. No item score significantly increased during BLT. Separate analysis for BD versus MDD showed no significant difference of effect, both on the level of total score as individual items.
The largest decrease was seen in items related to mood, concentration and self-esteem, which had high average baseline scores (e.g. quality of mood: mean baseline score 1.92, see Fig. 2). Scores on the item decrease in weight and waking up too early did not significantly alter during BLT, but also started lower than other items (mean scores: 0.35 and 0.53 points, respectively). This showed a floor effect in many of the patients, as 59 and 54 patients of the total cohort started off with a score of zero on these respective items, at the start of BLT.
DiscussionIn this naturalistic study we assessed the change in depressive symptoms due to BLT in a real-world cohort of patients suffering from unipolar and bipolar depression. BLT was effective in decreasing the IDS-SR total score significantly in both BD and MDD patient groups suffering from a depressive episode with similar effect sizes as well as similar rates of response and remission.
Meta-analyses into effectiveness of BLT in BD and MDD have concluded that BLT is effective in ameliorating depressive symptoms in BD9,13,21 and MDD9,10,22, but all reported high heterogeneity in clinical studies and rarely included patients with psychiatric comorbidity. One meta-analysis concluded that there was insufficient evidence for BLT.9 In the current study, the focus was on real-world patient groups. Patients with psychiatric comorbidity were included if their clinician referred them to BLT. Besides studies mentioned in the meta-analyses, we are aware of two other real-world studies using a protocol very similar to ours and a placebo-controlled double-blind trial with patients with BD with a longer duration of treatment (6 weeks). Sikkens (2019)23 investigated the effects of BLT on depression in 26 patients (therapy resistant MDD and BD and comorbid disorders) and Van Hout (2020)24 studied the effect on 58 patients with only BD, seasonal pattern. Treatment was between two23 and three weeks long24 with 10.000 lux for 30 min per day for five to seven days per week. The study by Sikkens also included three nights of sleep deprivation during treatment. Van Hout24 found a remission rate of 55 %, based on the Quick Inventory of Depression Scale (QIDS), while Sikkens reported lower response (34.6 %) and remission rates (19.2 %) on the IDS Clinician scale. The remission rate in our study (25.7 %) is lower than the study that included only BD patients24 and is also lower than the study by Sit (2018)25 (68.2 %). This might be due to a longer duration of light therapy in the study of van Hout and Sit, a higher frequency of treatment in the study of Van Hout and to the inclusion of patients without comorbidity in these studies. The remission rate in our BD group (34.1 %) is higher than group average, indicating depressed patients with BD might respond better than MDD patients, which has been previously described.26 The remission rate in our study might also have been influenced by the lack of follow-up after the end of treatment.27 Our methodology and remission rate are closer to the case series of treatment resistant depressed patients with psychiatric comorbidity. The authors of that study concluded that patients with psychiatric comorbidity showed a lower, but still significant, reduction of symptom severity. The difference between the effectiveness in our study and the study of Van Hout et al. could be explained by the difference in comorbidity.24 Interestingly, there was no difference between subgroups when stratified by number of psychiatric comorbid disorders, but BLT seemed to be significantly less effective in patients under 35 years of age.
We found the largest treatment effect for the core symptoms of depression such as sad mood and loss of energy, which were also items that scored highest at baseline. We anticipated that especially symptoms related to sleep would diminish, based on light therapy's influence on circadian rhythm15,16 and previous reports in other patient groups,17 but found no support in our data for this hypothesis. One previous placebo controlled study found that sleep quality improved in depressed patients with BD during adjunctive BLT, but also in the placebo group,25 indicating that the effects on sleep of BLT might be less than expected. Others, for example, have found expectations of BLT to be predictive for effect as well.28 Other explanations are the large variability of sleeping problems, and possible floor effects on the individual items. We are not aware of studies investigating the effect of BLT on separate symptoms using the scores from a depressive scale. One study protocol mentions using the Pittsburgh sleeping questionnaire to assess sleep quality before and after BLT,29 but results have yet to be published.
This naturalistic study has multiple strengths, first and foremost the relevance for comparable real-world outpatient settings. Second, the use of the IDS-SR enabled us to assess how individual symptoms changed in severity. However, there are also some limitations to consider. The protocol is dynamic, allowing patients and their physicians to evaluate progress after one week of treatment and decide whether to extend the treatment for an additional week. While this approach poses challenges for standardization and comparability in a clinical trial context, it reflects how BLT is applied in routine clinical practice. Attrition and response bias might have influenced our results, due to patients with absolutely no effect leaving the treatment before the second measurement. Due to design and the anonymous nature of the data, it is impossible to assess this, or other possible reasons for missing measurements. Furthermore, the IDS-SR which we considered to be baseline, might not have been the true baseline and the end of therapy measurement might not have been exactly timed with the end of therapy. This study lacks a longer-term follow-up assessment after treatment. An additional limitation was that we could not determine whether BLT was as an adjunct to pharmacological or psychological treatments. It is not possible to distinguish between the effects of the treatment, placebo effects, or symptom reductions due to regression to the mean, since our study lacks a placebo control group. Lastly, the single-item analysis was probably influenced by a floor effect, showing little effect due to many patients already scoring zero at baseline.
Future studies could benefit from data pooling between outpatient clinics to prospectively gather data, achieving larger sample sizes and minimizing missing data. Including other (observer-rated) scales and more frequent follow-up measurements could have provided further insights into symptom improvement, quality of life, and the longer-term persistence of BLT effects. More elaborate subgroup analyses based on depression subtypes and psychiatric comorbidities could further enhance understanding of BLT's effectiveness in different patient groups. A longer follow-up would also enable the creation of trajectories to observe change over time more comprehensively.
ConclusionDepressive symptoms in patients suffering from BD or MDD decrease, on average, equally after BLT for one or two weeks in a naturalistic setting. Individual symptom scores decrease overall with the most profound change in items scoring core symptoms of depression.
Ethical considerationsThe scientific committee of de department of psychiatry approved the analysis of anonymized data on 16th of May 2022 after receiving the protocol titled “Efficacy of light therapy in bipolar and unipolar depression according to Dutch protocol”. The use of anonymized Routine Outcome Monitoring (ROM) data from GGZ Rivierduinen and the LUMC was approved by the ethics review board of the LUMC.
FundingThis research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.
Consent for publicationNot applicable
Availability of data and materialThe data are not publicly available, but can be requested after approval of a complete data analysis plan by the scientific committee of GGZ Rivierduinen.
Conflict of interestAll authors declare no conflict of interest.
CRediT authorship contribution statementM.A. Riedinger: Methodology, Data curation, Formal analysis, Writing – original draft, Writing – review & editing. G.E. van Son: Data curation, Writing – review & editing. N.J.A. van der Wee: Supervision, Writing – review & editing. E.J. Giltay: Methodology, Data curation, Formal analysis, Writing – review & editing. M. de Leeuw: Supervision, Project administration, Writing – review & editing.
We thank A.J.C.G.D. Kluck for assisting in data cleaning and preliminary analyses for his master thesis.