Bright light therapy timed to the body clock for adults with depression: a plan for a multi-center trial on treatment methods, response predictors, and biological mechanisms

Depressive symptoms place a heavy burden on mood, motivation, and everyday functioning. Yet conventional antidepressants often require weeks to exert their full effects and can burden patients with persistent side effects. Bright Light Therapy (BLT)—which involves repeated exposure to a high-intensity light source—offers a compelling alternative: rapid relief, robust antidepressant impact, and only minimal, transient discomfort. In fact, BLT has now earned a place in clinical guidelines and is widely used in countries across the world—albeit with varied specifications of light intensity, treatment duration, and spectral composition of the light [1 –4]. Below, we review BLT's efficacy across mood disorders, introduce an optimized delivery model, explore predictors of response, and examine the mechanisms underpinning its antidepressant action, all to introduce a multi-center randomized controlled trial (RCT) that investigates the optimization, response prediction, and chronobiological and neurobiological mechanisms of BLT.

The clearest evidence for the efficacy of BLT comes from Seasonal Affective Disorder (SAD). In one of the first trials, Rosenthal and colleagues (1984; N = 29) found that 30–60 min of bright light—2500 to 10,000 lx—each morning lifted mood within just 3 to 5 days [5]. Larger, placebo‐controlled studies soon confirmed these results. In a pivotal randomized controlled trial with 96 SAD patients, 1.5 h of 6000‐lux exposure per day produced remission in 61% of participants, compared with just 32% in a dim‐light control group after 3 weeks [6]. Meta‐analyses of around 20 trials (N≈1200) report large effect sizes (0.60–1.08), with morning administration and intensities of at least 2500 lx yielding the greatest benefit [7, 8].

BLT's efficacy, however, extends well beyond SAD. In non‐seasonal major depressive disorder (MDD), early evidence of BLT was mixed [7], but recent meta-analyses now clearly support its use [9, 10]. A meta‐analysis of 20 RCTs (N = 881) found a moderate pooled effect (SMD ≈ − 0.4), again highlighting greater efficacy at higher intensities (≥ 7500 lx) and with consistent morning dosing [10]. Even in bipolar depression—where early concerns suggested morning light might trigger mania—a meta‐analysis of 11 studies (N = 567) showed significant reductions in depressive symptoms without any increase in manic episodes, provided BLT was used alongside mood stabilizers [11]. Beyond mood disorders, preliminary evidence points to benefits in, e.g., Attention-Deficit/Hyperactivity Disorder (ADHD) [12], Parkinson's disease [13], dementia [14], and cancer-related fatigue [15]. Across this spectrum of indications, patients typically experience relief within 1 to 2 weeks with minimal side effects. Notable is also the synergistic effects that are achieved when combining BLT with pharmacotherapy [16, 17] underscoring its promise as a transdiagnostic adjunctive intervention.

With efficacy well established the challenge shifts from demonstrating that BLT works to refining its delivery. While optimal intensity, spectral composition, and treatment duration remain open questions [9, 10], the greatest impact may come from moving beyond standalone BLT and embedding it within complementary interventions and supportive contexts that together target depressive symptoms. Conventional BLT typically confines patients to their homes or to (semi-)clinical spaces, sitting alone under a bright lamp in an otherwise sterile, low—interaction environment. Although this format reliably delivers the prescribed light dose, it overlooks the potent effects of social engagement [18, 19], behavioral activation [20], and environmental cues [21] on mood and treatment adherence that can amplify and extend BLT's benefits—transforming it from a solitary task into an immersive, habit-forming therapy. To bring this vision to life, this trial introduces the LightCafé [22]—a BLT delivery model set in a café-style space that feels restorative rather than clinical. Here, instead of solitary lamp setups, participants receive their morning light exposure embedded within a rich psychosocial context. Daily guided social conversations with a therapist foster positive affect and counteract isolation, while lifestyle coaching—from diet and exercise to sleep hygiene and routine building—helps translate acute mood changes into durable behavioral change. This multi-faceted approach is thought to facilitate longer lasting and sustainable change in patients struggling with depressive symptoms.

An unpublished observational study by Ottenheijm and Maes reports high response rates and strong patient satisfaction among patients in the LightCafé. However, the current LightCafé implementation follows a non-personalized schedule and does not yet leverage chronobiological insights now central to depression treatment. Evidence shows that aligning light exposure with each individual's circadian phase and minimizing evening light, is critical for consolidating sleep and stabilizing rhythms [23, 24], both of which directly support mood regulation [25]. To address this, the current trial will refine the LightCafé protocol by synchronizing morning light sessions to each participant's sleep–wake rhythm and by providing blue-light-blocking glasses (BBG) to minimize evening light exposure. In marrying the psychosocial power of the café environment with precision chronotherapy, we aim to deliver more effective and personalized BLT.

The next step in optimization involves more precise therapy selection upfront, as without knowing who benefits most, we risk offering intensive BLT protocols to unlikely responders. However, to date, we lack hypothesis‐driven investigations into which patient characteristics and behaviors predict BLT response. Secondary analyses in SAD cohorts suggest that individuals presenting with atypical features—such as increased appetite, hypersomnia, and a positive diurnal mood variation—alongside higher baseline BMI tend to experience the most rapid and robust improvement with morning light exposure [26, 27]. Those exhibiting melancholic symptoms—psychomotor retardation, early and late insomnia, appetite loss, pervasive guilt, and worse mood upon awakening—often derive little benefit [26]. Similarly, in treatment-resistant inpatients undergoing combined wake and light therapy, early diurnal symptom shifts, evening chronotype, and higher educational history have been linked to both short- and long-term remission [28]. Yet it remains unknown whether these results generalize across the full spectrum of unipolar and bipolar depressions. Clarifying these predictors is crucial: it would empower clinicians to stratify patients by likelihood of response, and ultimately render BLT a more efficient, cost-effective, and patient-centered antidepressant strategy.

Next, this study aims to examine the mechanisms by which BLT alleviates depressive symptoms, as currently no theory has conclusively explained its effects. The most widely accepted framework—the chronobiological phase-shift hypothesis—proposes that BLT realigns the circadian system with the natural 24-h light–dark cycle [29], correcting rhythm disturbances common in major depressive and bipolar disorders [30 –33], and thereby alleviating depressive symptoms. Although robust phase advances following morning BLT have been reliably documented [34], attempts to link these shifts directly to reductions in depressive symptoms have yielded inconsistent results [35 –40]. Several other studies further challenge a circadian explanation, suggesting that light exerts its mood-enhancing effects through neural pathways independent of the central clock [17, 41 –43], or via directly altering the serotonin release [44, 45]. Consequently, while BLT is often framed as a chronotherapeutic intervention, key questions persist about the true role—and relative importance—of underlying chronobiological mechanisms which we aim to clarify in the current trial.

Light's impact on sleep quality and energy levels may provide an additional—and potentially contrasting—route by which BLT exerts its antidepressant effects. Up to 90% of depressed patients report insomnia, delayed sleep onset, and fragmented rest at night [46, 47], while many also suffer daytime sleepiness [48]. Controlled trials show that daily morning BLT consolidates sleep, shortens sleep-onset latency, and improves alertness during the day [49, 50], but complicated by the multidirectional associations between mood, sleep, and energy levels [51 –53], it remains unclear whether these changes stem from circadian stabilization, light's direct alerting effects [54, 55], a synergistic interplay of both, or an unknown mechanism. By tracking light exposure, sleep, energy, and mood on a day-to-day level in the current trial, we aim to map how these factors interact with each other over time and gain insight into each one's role in BLT's antidepressant mechanism.

Neuroimaging adds a third dimension: though sparse, early studies hint that BLT engages both emotion-processing hubs (amygdala, habenula) and circadian centers (suprachiasmatic nucleus, hypothalamus) [41, 42, 56 –58]. Some RCT studies have evaluated the effects of BLT alone or in combination with other chronotherapeutics (e.g., sleep deprivation) in MDD [59, 60], subthreshold depression in adults [57, 61], in young adults [62, 63], or bipolar depression [64, 65]. They used mostly functional neuroimaging techniques to evaluate brain connectivity at rest and reported that weeks (varying between 3 and 8 weeks) of morning BLT restore connectivity within default-mode, frontoparietal, and salience networks—and that these neural changes predicted and correlated with symptom improvement [57, 61, 62, 64]. Interestingly, these studies bring some indications about the use of these neural parameters as potential biomarkers for disease in such that the neural changes found were linked to specific neurotransmitter systems (amongst others serotonin, dopamine) [62], but also to metabolic changes (glutamate/glutamine) [65]. One needs to mention that these studies have a high heterogeneity in study design and sample type used but also in neuroimaging data analysis approach, i.e., some using specific regions of interest and not the whole brain, i.e., amygdala [60], cingulate cortex [62], and cerebellum [61]. Hence, a clear mechanistic model is hard to determine at this stage. In addition, besides evident functional changes after BLT, there are also structural change indications after BLT administration in, e.g., posttraumatic stress disorder (PTSD) [66] and mild traumatic brain injury (mTBI) cohorts [67, 68]. The existing studies suggest that blue-light exposure can increase gray-matter volume and white-matter integrity [66 –68].

In the current study, we therefore plan to take a multimodal neuroimaging approach by combining both functional and structural change analysis as well as going from whole brain analysis to specific regions of interest and link it to symptom change pre-post BLT. Hence, we believe that this approach will give a more comprehensive mechanistic model of the neural pattern behind BLT in depression. Furthermore, no study to date has combined precise circadian profiling, sleep metrics, and multimodal imaging in a large, multicenter sample and using the light cafe approach which brings the lacking social component into the depression model—a gap our trial is uniquely poised to fill.

The primary objective of this study is to comprehensively assess and compare the effectiveness of three distinct BLT approaches for the treatment of depressive symptoms: (1) standard home-based BLT (Light@Home), (2) BLT at the LightCafé (LightCafé), and (3) BLT at the LightCafé with personalized timing of light and darkness using BBG (LightCafé +).

Secondary objectives are: i. The identification of predictors of treatment response, encompassing baseline clinical characteristics, sleep metrics, circadian parameters, and light-related behaviors. ii. Elucidation of underlying mechanisms of BLT's efficacy by investigating the contribution of circadian rhythm changes, enhancing our understanding of the complex interactions between mood, energy levels, and sleep within the context of BLT and exploration of the neural effects of BLT by studying both functional and structural brain changes before and after treatment.

Importantly, we are not testing BLT's efficacy for specific diagnostic categories; rather, we ask whether embedding BLT within a rich psychosocial setting and personalizing its timing can amplify and prolong its antidepressant effect, all while unraveling who benefits most and through which pathways. By moving from "Does BLT work?" to "How can we make it work better, for whom, and why?", this multicenter RCT will lay the groundwork for personalized and embedded BLT protocols and real-world clinical guidelines.

This study constitutes a multi-center, single-blind randomized controlled trial assessing the effects of BLT for individuals with a current depressive episode (either diagnosed with SAD, MDD, or bipolar disorder). The study consists of three phases—the baseline phase, the treatment phase, and the follow-up phase. During the treatment phase, participants will receive BLT through one of three administration strategies—Light@Home, LightCafé, or LightCafé + —with treatment lasting 1 to 3 weeks, depending on their individual response. Baseline assessments occur 1 week prior to treatment, with an additional week for participants undergoing melatonin assessments. Follow-up evaluations occur at 4 weeks, 10 weeks, and 4 months after the start of the treatment.

BLT is administered either at home or in the LightCafé, depending on the treatment condition. LightCafé locations are distributed across two mental healthcare institutions in the following way:

Mental Health Clinic of Eindhoven and the Kempen (GGzE) in EindhovenLightCafé Ketelhuis Boschdijk, EindhovenLightCafé Lucifer Kennedyplein, EindhovenLightCafé Library Veldhoven, Veldhoven

Leids Universitair Behandel- en Expertise Centrum (LUBEC) in Leiden.LightCafé Faculty of Social Sciences, Leiden

This study compares the effectiveness of three distinct BLT administration strategies for treating depression:

I.Home-Based BLT (Light@Home): Patients receive BLT in their own homes.II.BLT in LightCafé (LightCafé): Patients undergo BLT in a café-like environment (LightCafé), which emphasizes lifestyle improvements alongside the therapy.III.Personalized BLT in the LightCafé (LightCafé +): Patients receive BLT in the LightCafé, with treatment timing tailored to their individual sleep–wake pattern and personalized timing of darkness with the use of BBG.

The Light@Home condition serves as an active control condition against which both other conditions will be evaluated. Given that our primary objective is not to measure the effectiveness of BLT in alleviating depressive symptoms per se, but rather to optimize its delivery, the inclusion of a placebo or treatment-as-usual condition in this trial is deemed unwarranted. To assess and possibly control for the difference in expectations of treatments, all participants will complete the Credibility/Expectancy Questionnaire (CEQ; [71]) before treatment onset.

In all study arms, participants are subjected to BLT in adherence to Dutch BLT guidelines [72], utilizing an Innolux LED light Lamp delivering light with an illuminance of 10,000 lx at a distance of 70 cm and a color temperature of 3800 K. BLT sessions, lasting 30 min each, are scheduled from Monday to Friday between 7:30 AM and 10:30 AM.

The number of BLT sessions is contingent upon the individual response of each participant, with a maximum duration of 3 weeks. Initially, participants undergo five consecutive BLT sessions during the workweek. Subsequently, the therapeutic efficacy is evaluated utilizing the QIDS-SR during the Friday session. If the QIDS-SR falls below 6, remission is inferred, leading to the termination of BLT. Alternatively, an additional five sessions are administered. This evaluation and decision process is repeated at the end of the second week. If the QIDS-SR score is still above 6, a third and final week of treatment is provided.

This individualized strategy is intended to reduce the small risk of (hypo)manic switches, which occur in less than 2% of bipolar patients and almost exclusively in those not maintained on mood stabilizers [11, 73]. We will continuously monitor for (hypo)manic symptoms using the Young Mania Rating Scale (YMRS; [74]) and clinical observation: a YMRS score ≥ 13 and clinician‐identified symptoms will lead to immediate treatment cessation, while scores of 10–11 will prompt a joint review with a second clinician to decide on discontinuation [75, 76].

Each BLT session in the LightCafé includes 5–10-min consultations with a professional nurse or psychologist, during which participants discuss various aspects of their lifestyle, such as sleep hygiene, exercise, nutrition, daily routines, social interaction, and overall healthy living. These brief consultations are integrated into the light therapy sessions to provide personalized guidance and support. Additionally, participants are provided with workbooks designed for independent use, enabling them to actively work on and improve their lifestyle habits outside of the sessions. The ultimate goal of the LightCafé approach is to foster a stable rest-activity pattern, promote a more active lifestyle, ensure a fulfilling daily routine, and encourage social interaction, all within a treatment environment that enhances the overall treatment experience.

In the LightCafé + arm, BLT timing is personalized, and participants wear BBG according to their individual sleep–wake rhythms. These adjustments aim to promote sleep regularity and synchronize sleep phase with the endogenous biological night. Before commencing treatment, each participant's current sleep onset time (CSO) is established based on their baseline sleep data obtained from the Consensus Sleep Diary (CSD; [60]. Following this, a target sleep onset time (TSO) is collaboratively determined through a phone call with a member of the study team. During this conversation, the participant's daily routine, social and work obligations, and personal preferences for sleep onset and offset are discussed. The aim is to establish a sleep schedule that accommodates approximately 7–8 h of nightly sleep and fits well within the participant's everyday life.

Based on the relation between CSO and TSO, one of the following timing strategies is employed:Stabilize: When CSO is within 30 min of TSO, the aim is to not shift sleep timing and the circadian clock. Morning BLT is administered approximately 11 h after TSO, and participants wear BBG during the last hour before TSO.Advance: When CSO occurs more than 30 min later than TSO, the aim is to advance the sleep timing and the biological clock by exposing patients to BLT earlier in the morning (around 8–9 h after CSO) and having them wear BBG during the last 2 h before CSO.Delay: In cases where CSO is more than 30 min earlier than TSO, a delay in sleep–wake cycles and the biological clock is the goal. Morning BLT is administered later in the morning (around 11 h after CSO), and BBG are worn in the morning, during the first 2 h after awakening.

The chronobiological strategy is selected and implemented on a weekly basis, based on a reassessment of the CSO, using CSD entries from that week, and the TSO, determined via phone contact. Treatment times will always remain within the 7:30 AM to 10:30 AM window. This time frame was chosen because morning light has been shown to be most effective in reducing depressive symptoms [7]. While some individuals may theoretically benefit from alternative timings, individual adjustments are limited by the practical constraints of the clinical setting. If the prescribed strategy suggests a time outside this window, the closest suitable time within the range is selected.

Participants can withdraw from the study at any time without providing a specific reason. BLT is discontinued upon the patient achieving remission or exhibiting mania symptoms as per the YMRS or clinical assessment. Furthermore, treatment discontinuation can be initiated by clinical staff when they consider it to be in the best interest of the patient.

Treatment adherence is assessed using two complementary methods: EMA and personal light exposure sensors. Each day, participants receive a smartphone questionnaire asking whether they adhered to the light therapy protocol. If not, they are prompted to indicate the reason for non-compliance. In parallel, participants wear personal light exposure sensors throughout the study period to objectively record light exposure and verify whether the prescribed timing and dosage were followed.

To support protocol adherence and increase participant engagement, a feedback session is scheduled after the post-treatment assessment. During this session, participants receive personalized feedback based on their individual data (e.g., daily affect changes, and actigraphy reports) collected during the intervention.

Participants are allowed to continue existing antidepressant treatments alongside BLT, provided these therapies—pharmacological or non-pharmacological—were initiated at least 2 months prior to study entry. This includes antidepressant medications, psychotherapy, EMDR, or other interventions aimed at reducing depressive symptoms. While the initiation of new treatments or dosage adjustments during the study period is discouraged, they are not prohibited. Any such changes will be carefully documented and accounted for in the statistical analysis to control for potential confounding effects.

In the event of a successful treatment, the opportunity for another round of BLT becomes available 3 months after the last treatment session, provided there are no clinical contraindications, and the final follow-up of the study has been completed. If the treatment is considered unsuccessful, official communication will be sent to the referring healthcare provider, and the responsibility for further treatment will rest with them.

Participants in this study are not provided with research subject insurance coverage, as BLT is a non-invasive treatment associated with a low likelihood of mild and short-term side effects.

We assess a range of secondary outcomes, categorized as mood-related and functional outcomes, sleep-related outcomes, circadian parameters, brain imaging parameters, and light-related parameters.

In the Baseline Phase, participants undergo 1 to 2 weeks of measurements to assess their pre-treatment condition. The study begins with a 90-min face-to-face intake appointment (Intake). During this session, participants undergo the M.I.N.I. interview and complete the intake questionnaires. They also receive guidance on the proper use of the sensors and an explanation of how to use the BLT lamp. At the conclusion of the Intake, participants are randomized into one of the three treatment conditions.

One week before the start of therapy—typically on the following Monday—baseline examinations (B1) are conducted online. This includes completing the online questionnaires for B1 and a MADRS assessment via video call. Additionally, participants start with the EMA protocol and begin wearing the actigraphy watch and illuminance sensors. Those undergoing MRI assessments will also complete their baseline MRI scan (MRI-1) in the week before therapy. For participants involved in the melatonin assessments, an additional baseline week is added 2 weeks prior to the start of the treatment. This additional week includes the baseline DLMO assessment (M1) and an evaluation of light sensitivity (M2).

During the treatment phase, participants receive 1, 2, or 3 weeks of BLT within their assigned condition. Every Friday, depressive symptoms are assessed using the QIDS-SR, regardless of whether the participant received treatment that week. Treatment will be continued as long as QIDS-SR-score is higher than 6, with a maximum of 3 weeks. Treatment tolerability and mania symptoms are assessed using the YMRS after each week of BLT. EMA, actigraphy, and light logger usage continue throughout the treatment.

After the treatment has been completed, EMA, actigraphy, and light logger usage are extended for an additional week and are discontinued on the Sunday at the end of this additional week. The post-treatment DLMO assessment (M3) and MRI scan (MRI-2) also occur within this first week after BLT.

Following these assessments, a post-treatment evaluation (PT1) is scheduled 4 weeks after the start of treatment, which includes a series of online questionnaires and an online MADRS assessment. During this meeting, participants receive general insights into their EMA data. The online follow-up questionnaires (F1) are sent 10 weeks after the start of treatment, with the second set of follow-up questionnaires (F2) distributed 4 months after the start of treatment, after which study participation is discontinued.

As the EMA is the most burdensome aspect of our study protocol, we carefully evaluated the feasibility of utilizing a subsample of the study population for this component. The primary objective of the EMA is to investigate the cross-lagged effects between sleep parameters and affect during therapy using a Random Intercept Cross-Lagged Panel Model (RI-CLPM). To determine the appropriate sample size for our analysis, we conducted a power analysis using the powRICLPM package in R, following the procedures outlined by Mulder (2023) [91]. The parameters for our analysis included 14 time points, an intraclass correlation coefficient (ICC) of 0.3, a correlation between random intercepts (RI_cor) of 0.3, a within-group correlation of 0.5, and expected effect sizes indicated by a correlation matrix with a moderate effect size of 0.4 for sleep and affect, as well as cross-lagged effects of 0.3. In our power analysis, we examined a sample size range from 100 to 231 participants, incrementing by 10, while employing 1000 Monte Carlo replications to ensure robust power estimates. The results indicated that a sample size of 120 participants would provide a power level of over 90%. As a result, we opted to include only 120 participants in the EMA component of the study. Following the data analysis, we will conduct a post hoc power analysis to evaluate the power of the RI-CLPM. If this analysis reveals insufficient power, we have the option to revert to a mixed model approach, which necessitates fewer participants to achieve similar power levels.

Saliva sampling is optional in the study protocol to reduce participation barriers. The main aim of the saliva sampling is to compare possible shifts in DLMO between conditions. A simulated dataset was created using the simr package in R, with a mean DLMO set at 23:00 and extensive between-person variance (SD = 120 min) and within-person variance (SD = 45 min). Using a mixed-effects model with a random intercept for participants, we simulated a sample of 72 participants measured at two timepoints (pre- and post-treatment). The analysis yielded a power of 96.40% (95% CI 95.05, 97.47) to detect a phase shift of 30 min in DLMO with an alpha level of 0.05. This high power suggests that even with the optional saliva sampling and a sample size of 72 we can confidently assess the effect of BLT on circadian rhythm changes.

For the MRI assessments, our study will focus on a cohort of 60 participants recruited from LUBEC, as GGzE does not have the means to conduct MRI assessments. The main analysis of the brain data will involve a pre- vs. post-BLT comparison, without any comparisons among the three conditions (i.e., no between-subjects effect). A simple sensitivity power analysis, conducted using G*Power software for the intended statistical tests on the neuroimaging data (paired t-test), suggests that with a sample size of N = 60, we can achieve a power of 0.80 at an alpha level of 0.05, anticipating a small to medium effect size (Cohen's d = 0.33), which aligns with findings from recent neuroimaging studies [59, 64, 66, 67].

Similar to the explorative subgroup analysis of the neuroimaging data comparing remitters and non-remitters will be based on past neuroimaging studies with a similar setting and having a sample size of minimal 20 participants per group [59]. This sample size is in adherence with recent publications in the field and accounts for the limitations encountered in a priori power size calculations of neuroimaging data [92, 93].

We intend to include participants year-round for approximately 2 years. The distribution of the recruitment burden is shared between two mental health clinics: GGzE's aim is to include 171 participants, while LUBEC aims for 60. At GGzE, all individuals referred for BLT are approached for potential participation. Previous light therapy studies conducted at this center with the same enrolment strategy have achieved inclusion rates of approximately 100 participants per year. Given that this study is considered more extensive than previous research, we anticipate it will take approximately 2 years to reach the targeted enrolment of 171 participants. At LUBEC, BLT is offered for the first time, and only in conjunction with study participation. Recruitment efforts involve the collaboration with primary care and secondary mental health institutions for the referral of patients.

The rate of participant accrual is reviewed every half year. The number of participants who consented but did not qualify and the number of qualified, enrolled participants who dropped out will also be examined to detect patterns of insufficient recruitment rates. These will be addressed with modified recruitment strategies so that we can meet our recruitment goals.

Eligible participants are randomized using block randomization via the Research Manager data platform [94]. To ensure a balanced distribution among treatment groups and minimize potential biases, participants are stratified based on several key factors, including chronotype (categorized as early, intermediate, or late based on the MEQ [86], clinical diagnosis (differentiating between unipolar and bipolar conditions according to the M.I.N.I. assessment), and participation in the melatonin assessment (indicated as either yes or no).

A randomization table was generated using ResearchManager before the start of the trial. Access to the randomization table is restricted for the whole research team and can only be accessed by an independent researcher of the Eindhoven University of Technology not affiliated with the study. The randomization process in ResearchManager employs permuted block randomization with blocks of six to maintain balanced allocation across study groups.

All study participants are entered into ResearchManager during intake, and upon confirmation of eligibility and completion of intake questionnaires, they are randomized using the program's algorithms. This process can be conducted by any member of the study team, including principal investigators, researchers, or study interns. During enrollment, researchers simply click a randomization button within ResearchManager, which then automatically assigns the participant to a study arm according to the randomization table.

Participants, researchers, and care providers are informed about the allocated interventions. However, to maintain impartiality in evaluating depressive symptoms, the primary outcome assessors are kept blind to both the treatment arm and the assessment time. These assessors have no other role in the study and, therefore, will remain unaware of the study's design and possible interventions.

N/a—all parties that could benefit from unblinding are already unblinded.

To assess sleep patterns and circadian rhythm parameters, participants wear a medical grade actigraphy watch, the MotionWatch 8 (CamNtech, Cambridge, UK), throughout the treatment period, baseline week, and post-treatment week. This validated medical device contains an internal accelerometer that reliably captures movement data in 30-s epochs. Participants receive clear instructions on how to wear the watch and are encouraged to keep it on during all the waking and sleeping hours to ensure accurate and continuous data collection.

Participants wear a small, box-shaped light sensor developed by the Eindhoven University of Technology [88] near their clavicle from waking to bedtime during the baseline, treatment, and follow-up weeks to assess their light-related behaviors. This device uses a TAOSTCS34725 optical sensor and an onboard EEPROM memory of 64 kb. The sampling window is set to 30 s, meaning the device takes a light measurement every 10 s and averages these for the time window. To ensure optimal data collection, participants are instructed to turn the sensor upside down at bedtime to prevent it from measuring light exposure that does not actually fall on the eyes. Additionally, participants are advised to remove the sensor in wet conditions, as it is not water-resistant.

Momentary affect states, momentary energy levels, and subjective sleep parameters are assessed using an ecological momentary assessment (EMA) approach in a subsample of 120 participants. Throughout the baseline, treatment, and post-treatment weeks, participants receive eight EMA notifications daily via a mobile application called M-Path [97]. Each notification, which takes approximately 1 min to complete, occurs at semi-random intervals across eight 105-min periods throughout the day. Participants have the flexibility to choose their preferred notification schedule, receiving alerts either between 7:00 AM and 9:00 PM, 8:00 AM and 10:00 PM, or 9:00 AM and 11:00 PM. To increase the validity of the momentary states, notifications become inaccessible 30 min after they are sent. In addition to the eight momentary notifications, participants are asked to complete a sleep diary each morning to assess subjective sleep parameters, as well as an evening questionnaire assessing daytime sleep, caffeine and drug intake, and treatment adherence.

In a subset of participants (n = 72), salivary melatonin samples are collected in three additional evening sessions: two during the measurement week before baseline (M1 – DLMO and M2 – Light Sensitivity), and one in the week following treatment (M3—DLMO). Participants collect saliva samples at home by swabbing a cotton swab in their mouth and placing it in a salivette®. Samples will be promptly refrigerated and shipped to the lab within 3 days. In the lab, samples are centrifuged and stored at − 20 °C for analysis. Melatonin concentration in the saliva samples is analyzed in the lab of Chrono@Work B.V. (Groningen, The Netherlands) with a double-antibody radioimmunoassay (RIA) (Direct Saliva Melatonin kit, NovoLytiX GmbH, Switzerland). Specific saliva sampling procedures depend upon the type of assessment (DLMO or Light Sensitivity).

Sixty participants at LUBEC undergo two brain imaging sessions at the 3 Tesla MRI facility of the Leiden Institute for Brain and Cognition (LIBC) of the Leiden University Medical Center (LUMC) in Leiden. The First MRI session takes place the week before BLT, and the second session is conducted the week after BLT completion. Each session has a maximum duration of 60 min (including preparation, positioning of participant, and scanning). The brain scans are performed on a Philips 3 T MRI system (Best, The Netherlands) at the Leiden Institute for Brain and Cognition (LIBC) of the Leiden University Medical Center (LUMC) in Leiden, The Netherlands. The following MRI imaging sequences are used in each session:Structural High-Resolution T1: This sequence captures detailed information about GM, including parameters like THK, VOL, SA, and LGI. It is also essential for subsequent functional scan analysis. For this, we will use a high-resolution T1 weighted image with the following parameters: repetition time (TR) = 7.9 ms, echo time (TE) = 3.5 ms, flip angle = 8°, field of view (FOV) = 250 × 195.83 × 170.5 mm, 155 slices.Diffusion Weighted Imaging (DWI): This sequence enables the assessment of the brain's structural connectivity by examining the primary WM tracts. Data from the DWI is analyzed using a diffusion tensor imaging technique (DTI). For this, we use a DTI high-iso sense sequence with the following parameters: TR = 6500 ms, flip angle = 90°, FOV = 224 × 224 × 125.4 mm, number of slices = 57, shim size = 100 × 100 × 100. SENSE = 2.5. Diffusion-sanitizing gradient encoding is applied in 32 directions with a diffusion weighted factor of b = 1000 s/mm² and two b0 (b = 0) images. Images are acquired parallel to the anterior–posterior commissure.Functional Resting State: This sequence provides information about the functional connectivity of brain networks while at rest. For this, a T2*-weighted using gradient-echo planar imaging (EPI) sequence is used. The parameters entail: TR = 2200 ms, TE = 30 ms, flip angle = 80°, number of slices = 40 transverse, slices thickness = 2.75 mm, FOV = 22 × 220 × 120.725.Structural T2w Image: To better reconstruct the structural images and obtain exact values of THK, VOL, SA, and LGI within the analysis program FreeSurfer (http://surfer.nmr.mgh.harvard.edu/↗), we also perform a structural T2-weighted 3D Flair Image at the end of the scan with the following parameters: TR = 4800 ms; FOV = 250 × 250 × 182,56; 326 slices.

An interactive feedback session is planned with the participants during the post-treatment assessment where we evaluate their individual study outcomes, focusing mainly on the EMA data. This offers a personalized incentive for continued participation. Online questionnaires and automated reminders are integrated to ensure data retention. Additionally, personal contact with participants is maintained throughout the study to promote adherence to the protocol.

ResearchManager is utilized for data entry and initial storage of all information gathered through questionnaires. Study personnel will complete the study forms for each appointment within the electronic Case Report Form (eCRF), while self-rated questionnaires will be dispatched to participants via email through the platform. In instances where questionnaires are completed on paper, the source documents will be securely stored in a locked filing cabinet with restricted access limited to research personnel. Prior to locking the data into the system, all information will undergo validation checks by either EV, OGRO, or NA. Data collected through ResearchManager will be securely stored in encrypted format on a protected server. Following the conclusion of data collection, it can be exported to Excel sheets, which will then be stored on the local servers of both Eindhoven University of Technology and Leiden University.

EMA data will be gathered using the M-Path application [97] and stored in encrypted format on M-Path's secure server, which complies with Dutch NEN 7510 and international ISO 27001 standards for information security management. Subsequent to the final data collection, data is exported to Excel sheets and stored on local research servers. Saliva collected for melatonin assessments will be dispatched to Chrono@Work B.V. via post by the participants themselves. Melatonin concentrations will be determined in the laboratory, and the saliva sample will then be discarded safely. Data will be shared in Excel sheets via secure email and stored on local research servers. Data collected via actigraphy, or personal light exposure sensors will be transformed into Excel sheets using the Motionware program or the LightlogControl program, respectively. These sheets will also be stored on local servers. MRI data will be acquired at the university hospital and will be stored on the protected local server of Leiden University. All study-related data will be archived for a minimum duration of 15 years.

Information collected about participants during this clinical investigation is treated confidentially. The collection and processing of participants' personal information is limited to what is necessary to ensure the study's scientific practicability, the evaluation of efficacy, adherence, side effects, and the investigational product's safety. All data is pseudo-anonymized using participant identification numbers.

Saliva will be collected from the participants that consent to the melatonin assessments. Melatonin concentration in saliva will be analyzed in the lab of Chrono@Work B.V. (Groningen, The Netherlands) and destroyed afterwards. Saliva cannot be used in future studies.

This analysis plan provides the initial framework for examining primary and secondary outcomes. A significance level of 0.05 will be applied unless otherwise specified, with corrections for multiple comparisons (e.g., Bonferroni) used when necessary. All analyses will follow the intention-to-treat principle, accounting for the nested structure within centers if appropriate. The analyses outlined here may be subject to minor modifications based on advancements in the field. Should any deviations from the original plan occur, these will be clearly documented and justified. Our reporting will adhere to the guidelines set by the CONSORT statement [98].

We will report the descriptive statistics of socio-demographic and clinical variables for the whole sample, per site, and per treatment arm. In specific analyses involving subsamples (e.g., melatonin assessments, EMA, and MRI), we will provide socio-demographic and clinical characteristics of the subsample and report any distinctions from the overall sample.

To compare the three different BLT administration strategies in terms of clinical improvement (measured as the difference in MADRS scores between baseline and post-treatment assessments), we will employ a multilevel mixed-effects modelling approach. MADRS scores will serve as the dependent variable, with a fixed effects for timepoint (pre- and post-treatment), and the interaction between timepoint and treatment arm (Light@Home, LightCafé, and LightCafé +). Random intercepts for participants and study centers will be included to account for individual differences and center-level variability, but only if they improve model fit, as determined by evaluation of the Akaike Information Criterion (AIC) and Bayesian Information Criterion (BIC). Random slopes for time within participants will also be considered and included based on the same criteria. Socio-demographic (age, sex) will be evaluated for potential confounding effects, similar to the credibility and expectancy of the treatment arms as assessed with the CEQ. The primary analysis will focus on the interaction between timepoint and treatment arm to determine whether clinical improvement differs across the three BLT strategies. Furthermore, multilevel logistic regression will be used to test differences in response (yes/no) and remission (yes/no) between treatment arms using the same structure and procedure as above.

N/a—we are not conducting any interim analyses.

Exploratory subgroup analyses will be conducted to examine variations in the effects of different administration strategies on depressive symptoms, sleep, and circadian rhythms across various depression diagnoses. Additionally, the whole brain structure and functional parameters will be compared between the three treatment arms. As a post hoc explorative analysis the functional and structural brain parameters will be compared between remitters vs non-remitters. Multiple comparison corrections for repeated measures will be done.

Furthermore, we will explore the relationship between treatment adherence (as measured by self-report and light sensor data) and treatment effectiveness. Adherence will be operationalized as the percentage of completed light therapy sessions (based on sensor data, the timing of light exposure relative to the prescribed schedule, and daily self-reported compliance). These adherence indicators will be added as covariates in regression models predicting changes in depressive symptoms, sleep, and circadian phase markers. Primary outcome models will also include medication type and dose as time-varying covariates to test for interaction with BLT. Exploratory subgroup comparisons (e.g., on versus off antidepressant) may further probe whether concurrent treatments alter BLT's therapeutic impact.

The data analysis is an intent-to-treat analysis with participants analyzed as randomized, regardless of adherence to the treatment protocol. Additionally, completer analyses will be conducted, including only those participants who have both pre-treatment and post-treatment data. Furthermore, per protocol analyses will focus on participants who completed at least 80% of the treatment (i.e., attending 4 out of 5 days of BLT per week). Any substantial differences between these analytical approaches will be evaluated and reported.

We will describe the extent and patterns of missing MADRS scores and covariates overall and by treatment arm. To evaluate whether missingness is completely at random (MCAR), we will apply Little's MCAR test and examine associations between missingness and key variables (e.g., age, sex, diagnosis, depression severity, treatment expectancy). Under a plausible missing-at-random (MAR) assumption, we will perform multiple imputation by chained Eqs. (10 imputations) as a sensitivity analysis, refitting the primary multilevel mixed-effects models in each imputed dataset and pooling estimates via Rubin's rules. We will then compare results from the complete-case analysis and the MAR sensitivity analysis to our main intent-to-treat (observed-data) findings. Any notable discrepancies in effect estimates, model fit, or inferential conclusions across these approaches will be reported.

Participant-level data (from consenting participants) and the corresponding statistical code will be made available upon request. Parts of the data, i.e., data published in peer-reviewed journals will be stored in the public repository. The aim of this step is that other researchers can have access to include these datasets in meta-analyses, reviews, or similar or enable replication of our study. This data is pseudonymized and will not contain any personal information. The complete data set will be stored on the local research institutional research servers in a pseudonymized fashion. We will adhere to the requirements of the DataverseNL platform for data archiving.

Principal Investigators

Members: Niki Antypa and Yvonne de Kort

Past Members: Luc Schlangen

Roles and Responsibilities: Overall study governance Annual update report to the National Ethical Review Board All (S)AE reporting to the Ethical Review Board Finance oversight, budget administration, contractual negotiations Provide scientific advice to Lead Investigators

Meetings: Weekly with lead investigators and monthly with the steering committee.

Lead Investigators

Members: Emma Visser and Oana Georgiana Rus-Oswald

Roles and Responsibilities: Site-level recruitment, informed consent and enrolment CRF completion, data query resolution and source-data verification Patient follow-up per protocol schedule Drafting and revising protocol sections, writing study progress reports Coordinating and hosting Steering Committee meetings

Meetings: Weekly with lead investigators and monthly with the steering committee.

Steering Committee

Current Members: A.J.W. van der Does, M.C.M. Gordijn, M.C. Marcelis, T. Seetsen, C.J.P. Simons

Past Members: P.P. Oomen.

Roles & Responsibilities:

Final sign-off of protocol and any substantive amendments Ongoing review of study progress: recruitment, retention, data completeness Approve operational changes to facilitate trial conduct Resolve major operational or scientific disputes

Meetings: Monthly meetings with the principle investigators lead investigators and steering committee

This study does not include a Data Monitoring Committee as BLT is a low-risk intervention and study procedures are straightforward and do not require ongoing oversight.

Adverse events (AE) are defined as any undesirable experience occurring to a subject during the study, whether or not considered related to BLT. All AE reported spontaneously by the subject or observed by the investigator, or their staff will be recorded. Additionally, a weekly screening of AE will be conducted. All AEs will be followed until they have abated, or until a stable situation has been reached. Depending on the event, follow-up may require additional tests or medical procedures as indicated, and/or referral to the general physician or a medical specialist.

N/a—no auditing.

A "substantial amendment" is defined as an amendment to the terms of the Medical Research Ethics Committee (MREC) application, or to the protocol or any other supporting documentation, that is likely to affect to a significant degree: The safety or physical or mental integrity of the subjects of the trial. The scientific value of the trial. The conduct or management of the trial; or. The quality or safety of any intervention used in the trial.

All substantial amendments will be submitted for approval to the MREC and to the competent authority. For non-substantial amendments, only a notification will be sent to the accredited MREC, which will be recorded and filed by the sponsor.

The results of the study will be submitted for publication in international peer-reviewed journals adhering to applicable privacy laws and regulations. The main study findings will be communicated to study participants after data collection is complete and data have been analyzed.

The trial is currently enrolling participants, starting from February 9, 2024, with recruitment expected to continue for about 2 years. We are following protocol version 3.1.0, dated October 13, 2023.