Low-dose exogenous melatonin plus evening dim light and time in bed scheduling advances circadian phase irrespective of measured or estimated dim light melatonin onset time: preliminary findings

Abstract

Study Objectives:

The purpose of the present study was to preliminarily evaluate whether knowing the dim light melatonin onset (DLMO) time is advantageous when treating delayed sleep-wake phase disorder with low-dose melatonin treatment plus behavioral interventions (ie, evening dim light and time in bed scheduling).

Methods:

In this randomized, controlled, double-blind trial, 40 adults with delayed sleep-wake phase disorder were randomly assigned to 4 weeks of 0.5 mg timed to be administered either 3 hours before the DLMO (measured DLMO group, n = 20) or 5 hours before sleep-onset time per actigraphy (estimated DLMO group, n = 20), in conjunction with behavioral interventions. The primary outcome was change in the DLMO (measured in-home). Secondary outcomes included sleep parameters per diary and actigraphy (sleep-onset and -offset times and total sleep time), Morningness-Eveningness Questionnaire, Multidimensional Fatigue Inventory, PROMIS–Sleep Disturbance, PROMIS–Sleep Related Impairment, and Pittsburgh Sleep Quality Index. Mixed-effects models tested for group differences in these outcome.

Results:

After applying the Bonferroni correction for multiple comparisons (significant P value set at < .004), there were significant main effects for visit on all outcomes except for the Pittsburgh Sleep Quality Index and total sleep time per wrist actigraphy and diary. There were no group-by-visit interactions for any of the outcomes (P > .004).

Conclusions:

Scheduled low-dose melatonin plus behavioral interventions may improve many circadian and sleep parameters regardless of whether melatonin administration is scheduled based on estimated or measured DLMO. A larger-scale trial is needed to confirm these preliminary findings.

Clinical Trial Registration:

Registry: ClinicalTrials.gov; Name: The Clinical Utility of Measuring the Circadian Clock in Treatment of Delayed Sleep-Wake Phase Disorder; URL: https://clinicaltrials.gov/study/NCT03715465; Identifier: NCT03715465.

Citation:

Swanson LM, de Sibour T, DuBuc K, et al. Low-dose exogenous melatonin plus evening dim light and time in bed scheduling advances circadian phase irrespective of measured or estimated dim light melatonin onset time: preliminary findings. J Clin Sleep Med. 2024;20(7):1131–1140.

BRIEF SUMMARY

Current Knowledge/Study Rationale: Delayed sleep-wake phase disorder is common and consequential. There is a critical need to understand whether use of field-based circadian markers, such as the time of dim light melatonin onset, enhances treatment outcomes in this disorder.

Study Impact: The study’s results provide preliminary evidence that current clinical practice for treatment of delayed sleep-wake phase disorder, which typically involves timing low-dose melatonin relative to sleep-onset time in conjunction with behavioral interventions, is associated with significant improvements across many clinically relevant circadian and sleep outcomes. Given the challenges inherent in measuring the dim light melatonin onset in clinical practice, it is important for larger-scale trials to more conclusively test whether treatment outcomes are enhanced by its measurement.

INTRODUCTION

The most common circadian rhythm sleep-wake disorder, delayed sleep-wake phase disorder (DSWPD), is estimated to affect up to 7–16% of adolescents and young adults, and more than 10% of patients presenting to sleep clinics with insomnia symptoms.¹ The prototypical patient with DSWPD presents with chronic difficulty falling asleep and waking at required or socially desired times.¹ When free to select their own sleep schedule, patients with DSWPD have significantly later sleep-onset and wake times relative to convention, with preserved sleep quality and normal sleep duration.¹ Delay of the circadian timing system relative to the required or socially desired timing of the sleep episode is posited to be 1 of the major etiological factors.² This is corroborated by numerous studies, which show phase delays ranging from 2 to 6 hours in circadian phase in patients with DSWPD relative to normal sleepers.^3–7

The inability to maintain a sleep-wake schedule consistent with social and occupational requirements often leads to chronic sleep loss and significant daytime impairments. Consequences of DSWPD are wide-ranging and include absenteeism and presenteeism, and impairment in social and family life.⁸ Indeed, role disability is greater, and quality of life is more impaired, in adults with DSWPD relative to other chronic disorders, including sleep apnea, migraine, and depression.⁹ Despite its prevalence, treatment of DSWPD is challenging. Effective treatment requires advancing the circadian phase so that it is optimally aligned with the desired sleep-wake schedule. Treatment options include morning light therapy, hypnotic medications, prescribed sleep-wake scheduling, timed physical activity, and wakefulness-promoting medications. Among the available treatment options, the most recent American Academy of Sleep Medicine Clinical Practice Guideline for treatment of DSWPD in adults recommends only 1 therapy: timed exogenous melatonin.¹⁰ Both strategically timed exogeneous melatonin and light therapy were recommended as a “guideline” treatment option for DSWPD in the first American Academy of Sleep Medicine Practice Parameters published in 2007.¹¹ However, the updated clinical practice guidelines published in 2015 make no recommendation for light therapy due to insufficient evidence and retain strategically timed melatonin as the only recommended treatment for adults.¹⁰ Nevertheless, as a treatment known to produce circadian phase advances, morning bright light therapy may aid in the treatment of DSWPD.¹²

As noted in both the Clinical Practice Guideline and its accompanying editorial, there is a significant lack of research on clinical populations with circadian sleep-wake rhythm disorders, and treatment of circadian rhythm sleep-wake disorders such as DSWPD, with a particular need for field-based studies.¹³ There are various treatment paradigms for the administration of melatonin to treat DSWPD, many of which rely on timing melatonin to desired bed time or sleep onset.¹⁴ In clinical practice, melatonin is typically timed to desired bed time, current habitual bed time, or scheduled at a specific time in the evening regardless of the patient’s current sleep-wake schedule. The few available placebo-controlled studies of melatonin for DSWPD in adults, although limited by small sample size and variability regarding the dose and timing of melatonin administration, show improvements in objectively measured sleep parameters, including increased total sleep time, reduced sleep-onset latency, advanced dim light melatonin onset (DLMO) time, and an advanced sleep-onset time for melatonin relative to placebo.^15–18

However, on average, these improvements are relatively modest and the development of treatment protocols that enhance the efficacy of melatonin therapy is needed. Clinical strategies that allow for in-home measurement of the DLMO and the administration of melatonin personalized to each patient’s DLMO hold promise to improve the efficacy of melatonin. Indeed, evidence from laboratory-based studies of healthy adults suggests that melatonin must be administered at the correct time in the phase response curve relative to the individual’s circadian phase to achieve a shift in circadian timing.^19,20 One previous study has demonstrated the importance of the timing of melatonin administration in maximizing phase advances in adults with DSWPD, with a more robust phase advance observed for earlier administration times.¹⁶ Nevertheless, to date, no study has evaluated timing the administration of melatonin to an individual’s circadian time vs typical clinical practice (ie, timing to sleep-onset time) using a randomized design.

In the present study, we preliminarily evaluated the clinical utility of obtaining field-based DLMO in the treatment of DSWPD in a randomized, parallel-controlled, double-blind study comparing 4 weeks of exogenous melatonin administration timed to either 3 hours before measured DLMO (M-DLMO condition) or 3 hours before estimated DLMO (E-DLMO; estimated as 2 hours before habitual sleep onset) plus time in bed scheduling. Specifically, we hypothesized that participants in the M-DLMO condition would show a greater advance of DLMO (primary outcome) relative to the E-DLMO condition. We further hypothesized that participants in the M-DLMO condition would show greater improvements in secondary outcomes relative to E-DLMO, including chronotype, sleep parameters, daytime functioning, and self-reported sleep quality.

METHODS

Design

The study was a randomized, parallel-controlled, double-blind trial of 4 weeks of exogenous melatonin 0.5 mg administered 3 hours before measured DLMO (M-DLMO condition) or 3 hours before estimated DLMO (E-DLMO condition) plus time in bed scheduling. In the E-DLMO condition, the DLMO was estimated to occur 2 hours before sleep-onset time according to wrist actigraphy; thus, participants in this group were scheduled to take exogeneous melatonin 5 hours before their pretreatment average sleep-onset time. The dose of melatonin (0.5 mg) was selected based on published studies,^16,19 which have shown similar phase-shifting effects between smaller (0.3–0.5 mg) and larger (3.0 mg) doses. A smaller dose also minimizes the soporific effects of taking melatonin during waking hours. The timing of melatonin relative to the DLMO in the M-DLMO condition was selected based on previous work showing that earlier administration relative to DLMO produces maximal phase shifts.^16,19 The E-DLMO condition was designed to mimic current clinical practice.

All participants received 4 weeks of 0.5 mg of exogenous melatonin and attended 4 weekly 15-minute treatment sessions with a study clinician. Study clinicians (D.A.C, G.B.R.) were licensed PhD-level clinical psychologists with board certification in behavioral sleep medicine. Participants completed salivary DLMO collection and self-report measures of chronotype, sleep, and daytime functioning before and after treatment. They wore a wrist actigraph and maintained a daily sleep-wake diary throughout the study. Participant enrollment was scheduled so that no active participation in the study occurred across daylight saving time changes.

Study procedures were approved by the University of Michigan Medical School Institutional Review Board. All participants completed written informed consent prior to beginning study procedures. This study was registered on ClinicalTrials.gov (identifier: NCT03715465).

Participants

Participants were recruited from the University of Michigan Sleep Medicine Clinics (including the Behavioral Sleep Medicine Clinics and the Collegiate Sleep Disorders Clinic) and through community advertisements placed throughout greater southeastern Michigan between January 2019 and September 2021. To be eligible for the study, participants had to be 18 years of age or older and meet the International Classification of Sleep Disorders, third edition, diagnostic criteria for DSWPD as follows: (1) evidence of a delayed phase of the sleep-wake pattern on daily sleep diaries and actigraphy maintained for at least 7 days (eg, a ≥ 2 hours delay in the timing of habitual sleep episode between work/school and free days) and (2) report difficulty falling asleep and difficulty awakening at desired/required times for ≥ 3 months. Participants were ineligible if any of the following were present: suspicion of a sleep disorder other than DSWPD per clinical interview, including chronic insomnia (Duke Structured Interview for Sleep Disorders²¹); presence of chronic psychiatric conditions for which melatonin may be contraindicated or which may directly influence sleep (eg, active substance use disorder, bipolar disorder, psychotic disorder, posttraumatic stress disorder) per the Mini International Neuropsychiatric Interview;²² presence of an unstable chronic medical condition directly related to sleep or which may be affected by melatonin (eg, chronic pain, diabetes, hypertension, clotting/bleeding disorders, seizures) per self-report and medical record review; pregnancy (verified via human chorionic gonadotropin urine test) or breastfeeding per self-report; routine nightshift work per self-report; past month travel or planned travel during the study across more than 1 time zone per self-report; use of melatonin in the past month per self-report; and current use of medications that may interfere with the measurement of melatonin (nonsteroidal anti-inflammatory drugs if used daily, and beta-blockers) per self-report and medical record review.²³

Procedures

Eligibility screening

Participants completed a 2-stage eligibility screening. In the first stage, they completed a screening interview, which included the Duke Structured Interview for Sleep Disorders²¹ and the Mini International Neuropsychiatric Interview.²² Female participants of childbearing potential also completed a urine pregnancy test at this time. Participants who were determined to be eligible based on the first screening stage proceeded to the second screening stage, during which time they wore a wrist actigraph (Actiwatch Spectrum; Philips Respironics, Murrysville, PA) and completed a daily sleep diary for at least 7 nights. Final eligibility determination was made by examining the sleep-wake patterns documented on wrist actigraphy and sleep diary (eg, evidence of a ≥ 2 hours delay in the timing of habitual sleep episode between work/school and free days).

Randomization and blinding

Upon study enrollment, randomization was completed by the study coordinator using a sequential treatment assignment software program developed by the Consulting for Statistics, Computing and Analytics Research service at the University of Michigan. Randomization was balanced by baseline average sleep-onset time per wrist actigraphy (earlier than 02:00 hours or 02:00 hours and later) and age (18–29 years, 30–45 years, and > 45 years). Group assignment for 3 participants who were randomized to the M-DLMO condition but whose pretreatment DLMO was not able to be calculated was changed to the E-DLMO condition. Analyses were conducted with and without these 3 participants to ensure that their inclusion did not systematically influence the results. As the results of these analyses with and without these participants were not different, the analyses reported herein include those participants. Participants and study clinicians were blinded to group assignment. Clinicians indicated which group they thought the participant was assigned to as a blinding check.

Intervention

All participants received 4 weeks of 0.5 mg of exogenous melatonin and 4 weekly 15–20-minute treatment sessions with a study clinician. The study was conducted under an “Investigational New Drug” protocol for melatonin (Food and Drug Administration Protocol #141269). The fast-dissolve melatonin tablets used in the study were manufactured by Natrol LLC (Sherman Oaks, CA).

The study coordinator provided the study clinician with the initial clock time at which each participant should take melatonin. The initial clock time of melatonin administration for participants assigned to the M-DLMO condition was set at 3 hours before their measured baseline DLMO. The initial clock time of melatonin administration for participants assigned to the E-DLMO condition was set at 5 hours before their average time of sleep onset at baseline per most recent 7 days of actigraphy.

At the initial treatment session, participants were assigned a set time in bed schedule to maintain for the first week of treatment based on their baseline time in bed schedule and without altering their total time in bed from their baseline average. At each treatment session thereafter, participants were instructed to advance the time of melatonin administration and their time in bed schedule (without altering their total time in bed from their baseline average) until they reached their desired time in bed schedule. Although the general intervention guidelines provided to the study clinicians were that melatonin time and time in bed schedules should be advanced by up to 1 hour per week, in order to approximate typical clinical practice, clinicians were free to adjust melatonin or time in bed schedules using their clinical judgement. Participants were also instructed to dim their ambient lights and all electronic devices to their lowest usable level starting 60 minutes prior to their scheduled bed time. Exogenous melatonin administration time and time in bed schedule were kept stable once the participant achieved their desired sleep-wake schedule. The study clinician provided each participant with a written weekly schedule for exogenous melatonin administration, time in bed, and dim light times. Note that, due to in-person human subjects research restrictions related to the coronavirus pandemic, treatment sessions were shifted in mid-March 2020 from in-person sessions to virtual sessions conducted via Zoom (Zoom Video Communications, San Jose, CA). Twenty-one participants (53%) had completed the study prior to the shift to virtual sessions and 19 participants (48%) completed the study after the shift to virtual sessions. No significant differences were observed on baseline characteristics or outcomes by treatment type; thus, analyses included all participants together regardless of treatment type.

Adherence to the time of melatonin administration was tracked using a MEMS TrackCap (Sion, Switzerland) monitor, which documented the time at which the melatonin vial was opened. Adherence to scheduled bed-wake times was monitored using wrist actigraphy; noncompliance was addressed by the clinician as needed. Participants were informed that their adherence to the time in bed schedule and timing of melatonin were being tracked as part of the study. Pill counts were completed at each weekly treatment session.

Outcome measures

Dim light melatonin onset:

Home-based salivary DLMO was collected pre- and post-treatment, following validated procedures.²⁴ Post-treatment DLMO was collected the day after participants took their last dose of melatonin. To ensure consistency and to minimize the effects of sleep schedule changes on the weekends (or weekdays off of work), all participants completed the DLMO collection on the third night of their typical workday sleep schedule (eg, the DLMO collection occurred on a Wednesday, Thursday, or Friday night for participants who worked a traditional Monday–Friday schedule). Participants were asked to refrain from using nonsteroidal anti-inflammatory drugs in the 72 hours prior to the DLMO collection and to refrain from using alcohol and caffeine in the 24 hours prior to the DLMO collection. Participants collected their saliva every 30 minutes starting 6 hours before average bed time per wrist actigraphy for the past week, with the last sample collected at average bed time (13 samples in total). The timing of the collection protocol was adjusted at post-treatment so that it occurred 6 hours before their average bed time per wrist actigraphy on the final week of treatment. Participants were provided with blue-blocking glasses (Uvex Skyper, SCT-Orange lens, Honeywell) to wear during the collection period to minimize the impact of any ambient light on melatonin secretion.²⁵ Compliance with dim light (≤ 50 lux) during the saliva collection protocol was monitored using a light-sensing actigraph (Actiwatch Spectrum, Philips Respironics), worn around the neck to measure light exposure close to the face. Participants were provided with a light meter to ensure that their environment was appropriately dim. An MEMS TrackCap monitored when the bottle containing salivettes was opened to determine compliance with the time of sample. Ninety-three percent of all samples were collected under compliant conditions (ie, light levels ≤ 50 lux in 30 minutes before the sample and salivette bottle opened within 5 minutes of scheduled time). Saliva samples were centrifuged, frozen, and shipped on dry ice to SolidPhase, Inc (Portland, ME), which performed direct Novolytix Laboratory radioimmunoassay (formerly Buhlmann, Switzerland; sensitivity of 0.5 pg/mL and intra- and interassay coefficient of variability < 7.5% at 3 pg/mL).^26,27

To determine the DLMO, linear interpolation of the clock time when the melatonin level exceeds the mean of 3 consecutive low daytime values plus twice the standard deviations of these points was calculated (also termed the “3k” method).^23,28 The DLMO was not able to be calculated for 5 participants at pretreatment and 5 participants at post-treatment. Thus, the DLMO sample sizes for pre- and post-treatment, respectively, are as follows: E-DLMO condition, n = 15 and 14; M-DLMO condition, n = 20 and 18.

Sleep assessments:

Participants wore a light-sensing wrist actigraph (Actiwatch Spectrum, Philips Respironics) on their nondominant wrist for the duration of the study. Actigraphy data were collected at a 60-second sampling rate. Actigraphy was scored according to established procedures using Actiware–Sleep software (version 6.0.9; Philips Respironics, Bend, OR) in conjunction with daily sleep/wake diaries.²⁹ The following actigraphy measures were extracted for each night and averaged across 7 days at baseline and post-treatment: sleep-onset time, sleep-offset time, and total sleep time (number of minutes scored as sleep in each rest interval). Participants also maintained a daily sleep-wake diary in which they recorded the following information: bed time, lights out, sleep-onset latency, frequency of nighttime awakenings, wake after sleep onset, final awakening time, and rise time. The following sleep diary measures were summarized across 7 days at pretreatment and post-treatment: sleep-onset time (calculated as lights-out time + sleep-onset latency time), sleep-offset time (time of final awakening), and total sleep time (calculated as [sleep-offset time – sleep-onset time] – wake after sleep onset).

Self-report questionnaires:

Participants completed self-report questionnaires at pretreatment and post-treatment. The Morningness-Eveningness Questionnaire assessed chronotype; scores on this scale range from 16 to 18.³⁰ Scores ≤ 41 indicate “evening” chronotype, scores between 42 and 58 indicate “intermediate” chronotype, and scores ≥ 59 indicate “morning” chronotype. Sleep quality was assessed by (1) the Pittsburgh Sleep Quality Index (PSQI³¹) global score (scores on this measure range from 0 to 21, with higher scores indicative of worse sleep quality) and (2) the Patient-Reported Outcome Measurement Information System (PROMIS)–Sleep Disturbance scale³² (for this scale, total raw scores are translated into standardized scores such that a mean score of 50 and standard deviation [SD] of 10 represents the average distribution for the US general population). Daytime functioning was measured by the following: (1) PROMIS–Sleep-Related Impairment (SRI) scale,³² which assesses sleep-related daytime impairment (for this scale, total raw scores are translated into standardized scores such that a mean score of 50 and SD of 10 represents the average distribution for the US general population) and (2) the Multidimensional Fatigue Inventory (MFI³³) general fatigue subscale was used to evaluate fatigue (scores on this subscale range from 4 to 20, with higher scores indicative of greater fatigue).

Statistical analyses

A priori power analyses were conducted to determine the sample size needed for the primary outcome of DLMO, based on the number of participants necessary to detect a 1-hour change in DLMO using estimates from a published trial.³⁴ The sample size required to detect this change at 90% power with an alpha of 0.05 was 23 per group. Thus, the targeted sample size was 25 per group.

All analyses were conducted using IBM SPSS Statistics software (version 28; IBM Corporation, Armonk, NY). Analyses were intent-to-treat in design and all participants who provided pretreatment data were retained in the analyses. Season of participation, based on the first day of treatment, was coded as a categorical variable (spring, summer, fall, winter). Compliance with time of melatonin administration was measured for each week of treatment by calculating the difference in minutes between scheduled melatonin administration time for each treatment week and the average time when the melatonin bottle was opened across the treatment week. Compliance with scheduled bed and wake times was measured daily throughout treatment by calculating the difference in minutes between scheduled times for each day and actigraphically assessed bed or wake time. Actigraphically assessed bed or wake times that were > 30 minutes later than scheduled were coded as noncompliant for that day. Fisher’s exact tests (for categorical variables) or Student’s t tests (for age) were used to test whether the conditions differed by demographic characteristics and season of participation. Student’s t tests evaluated between-group differences in the following: initial scheduled melatonin time, time interval between initial scheduled melatonin time and pretreatment DLMO time, time interval between scheduled melatonin time and sleep-onset time per actigraphy for each treatment week, change (in minutes) in scheduled melatonin time for each treatment week, and change (in minutes) in scheduled bed time, wake time, and melatonin time for each treatment week. Linear mixed models tested changes in the outcomes by condition. Models included fixed effects for visit (pretreatment and post-treatment), condition, a visit × condition interaction term, and the covariates age, race, and season, and a random intercept and slope. As multiple tests (n = 12) were conducted, statistical significance was set to P < .004 for the linear mixed models.³⁵

RESULTS

A total of 75 participants completed the initial screening interview and 40 were randomized (see Figure 1 for the participant flowchart). Twenty participants were assigned to the M-DLMO condition and 20 were assigned to the E-DLMO condition. Three participants, 2 from the E-DLMO condition (10%) and 1 from the M-DLMO condition (5%), withdrew from the study after randomization. Enrollment was terminated somewhat short of the target due to the end of the funding period and slower recruitment than anticipated, particularly after the start of the coronavirus pandemic.

Figure 1. Participant flow.

On average, participants were in their mid-20s. More than half reported their race as White (see Table 1 for demographic characteristics at pretreatment). The racial distribution of participants was significantly different between conditions (Fisher’s exact test = 9.48, P = .008). There were no other differences on demographic characteristics or outcome variables at pretreatment between conditions. Study clinicians correctly guessed group assignment 49% of the time (19/39 participants correct); as this value approximates chance (ie, 50%), it suggests successful blinding of study clinicians.

Table 1.

Participant demographic characteristics.

Variable	All Participants (n = 40)	Measured DLMO (n = 20)	Estimated DLMO (n = 20)	P *
Age, mean (SD), y	26.1 (5.9)	25.5 (6.2)	26.7 (5.6)	.508
Sex, n %				.480
Male	11 (27.5)	4 (20)	7 (35)
Female	29 (72.5)	16 (80)	13 (65)
Race, n %				.008
Asian	4 (10)	4 (20)	0 (0)
Black	4 (10)	4 (20)	0 (0)
White	29 (72.5)	11 (55)	18 (90)
Multiracial	3 (7.5)	1 (5)	2 (10)
Ethnicity, n %				1.00
Hispanic/Latino	1 (2.5)	0 (0)	1 (5)
Not Hispanic/Latino	19 (95)	19 (95)	19 (95)
Unknown	1 (2.5)	1 (5)	0 (0)
Employment status,† n %				.878
Full-time	16 (40)	7 (35)	9 (45)
Part-time	10 (25)	4 (20)	6 (30)
Unemployed	12 (30)	4 (35)	5 (25)
Household income,‡ n %				.759
0–$45,000	7 (42.5)	8 (40)	9 (45)
$45,000–$75,000	7 (17.5)	3 (15)	4 (20)
>$75,000	8 (20)	4 (20)	4 (20)
Education, n %				1.00
High school/GED	1 (2.5)	1 (5)	0 (0)
Associates degree	12 (30)	6 (30)	6 (30)
Bachelor’s degree	15 (37.5)	7 (35)	8 (40)
Postgraduate degree	11 (27.5)	5 (25)	6 (30)
Season on first treatment day, n %				1.00
Spring	9 (22.5)	4 (20)	5 (25)
Summer	12 (30)	6 (30)	6 (30)
Fall	7 (17.5)	4 (20)	3 (15)
Winter	12 (30)	6 (30)	6 (30)

*P values from t test for age; Fisher’s exact tests for all categorical variables comparing the 2 conditions. P values < .05 are bolded. †Two participants in the measured DLMO condition declined to provide their employment status. ‡Five participants in the measured DLMO group and 3 participants in the estimated DLMO group declined to provide their income. DLMO = dim light melatonin onset, GED = General Educational Development, SD = standard deviation.

Descriptive characteristics and t-test results for exogenous melatonin and selected sleep variables (including initial scheduled melatonin times, time interval between initial scheduled melatonin time and pretreatment DLMO, time interval between scheduled melatonin time and sleep-onset time per actigraphy), as well as change in scheduled bed, wake, and melatonin time by treatment week are summarized in Table 2 by condition. The initial scheduled melatonin time was significantly later in the E-DLMO condition, and the time interval between the initial scheduled melatonin time and pretreatment DLMO time was significantly longer in the M-DLMO condition. The time interval between scheduled melatonin time and sleep-onset time per actigraphy was significantly different between the groups at each treatment week. Participants in the E-DLMO condition had a significantly greater advance of their scheduled wake and melatonin times from treatment week 1 to week 2. Participants in both conditions reached their desired time in bed schedule, on average, at treatment week 4 (P > .05).

Table 2.

Mean (SD) and t test for selected melatonin and sleep variables.

Variable	Measured DLMO (n = 19)	Estimated DLMO (n = 18)	t Test
Initial scheduled melatonin time (clock time)	20:08 (2:15)	21:33 (1:28)	−2.26*
Pretreatment DLMO time – initial scheduled melatonin time† (hh:mm)	2:56 (0:15)	0:49 (1:18)	6.92**
Sleep-onset time (actigraphy) - scheduled melatonin time (hh:mm)
Week 1	6:37 (1:08)	5:08 (0:44)	4.79**
Week 2	6:23 (1:11)	4:32 (1.17)	4.31**
Week 3	6:28 (1:15)	4:46 (1:12)	3.98**
Week 4	6:37 (1:31)	5:04 (0:54)	3.52**
Change in scheduled bed time (hh:mm)
Week 1 to 2	−0:44 (0:31)	−0:56 (0:28)	1.25
Week 2 to 3	−0:21 (0:37)	−0:28 (0:24)	0.72
Week 3 to 4	−0:21 (0:33)	−0:28 (0:29)	0.77
Change in scheduled wake time (hh:mm)
Week 1 to 2	−0:30 (0:33)	−0:55 (0:30)	2.32*
Week 2 to 3	−0:14 (0:34)	−0:21 (0:25)	0.69
Week 3 to 4	−0:19 (0:30)	−0:13 (0:27)	−0.59
Change in scheduled melatonin time (hh:mm)
Week 1 to 2	−0:19 (0:23)	−0:43 (0:28)	2.93**
Week 2 to 3	−0:21 (0:22)	−0:23 (0:24)	0.37
Week 4 to 5	−0:21 (0:28)	−0:25 (0:26)	0.50

Data are from participants who completed the study. †Estimated DLMO, n = 15. *P < .05, **P < .01. DLMO = dim light melatonin onset, SD = standard deviation.

With respect to melatonin compliance, on average, participants opened the melatonin bottle 6 minutes (SD = 11 minutes) later than scheduled at week 1, 12 minutes (SD = 16 minutes) later than scheduled at week 2, 14 minutes (SD = 16 minutes) later than scheduled at week 3, and 25 minutes (SD = 37 minutes) later than scheduled at week 4. Compliance with bed time was 81.76% (±16.43%) of all nights; compliance with rise times was 84.75% (±15.53%) of all mornings. Compliance was not different by condition for any of the treatment weeks (P > .05). The most common treatment-emergent adverse events reported by participants included sleepiness (reported by 50% of participants), headaches (reported by 20% of participants), and insomnia (reported by 18% of participants).

Descriptive characteristics and results of the linear mixed-model analyses (F-tests for visit and visit × condition interaction effects) are shown in Table 3. In the linear mixed models with a significant main effect for visit, parameter estimates (β) for visit are as follows: DLMO (β = –1.18, P = .001; 95% confidence interval [CI] = –1.85, –0.51), MFI (β = –2.44, P = .006; 95% CI = –4.13, –0.76), PROMIS–Sleep Disturbance (β = –7.17, P < .001; 95% CI = –10.90, –3.44), PROMIS-SRI (β = –9.22, P < .001; 95% CI = –13.63, –4.81), sleep-onset time per diary (β = –1.68, P < .001; 95% CI = –2.28, –1.10), sleep-offset time per diary (β = –1.27, P < .001; 95% CI = –1.76, –0.78), sleep-onset time per actigraphy (β = –1.56, P < .001; 95% CI = –2.19, –0.94), and sleep-offset time per actigraphy (β = –1.16, P < .001; 95% CI = –1.65, –0.66). Main effects for visit were not significant for the outcomes of PSQI and total sleep time per actigraphy and sleep diary (P > .004). There were no significant group × time interactions (P > .004).

Table 3.

Mean (SD) and F-value for main outcomes by condition and visit.

Outcome	Measured DLMO (M-DLMO)		Estimated DLMO (E-DLMO)		Condition × Visit Interaction (F-value)	Visit Effect (F-value)
	Pretreatment (n = 20)	Post-treatment (n = 19)	Pretreatment (n = 20)	Post-treatment (n = 18)
Dim light melatonin onset (hh:mm)‡	22:54 (2:13)	21:48 (1:13)	22:33 (1:31)	20:37 (1:20)	1.35	33.06***
Morningness-Eveningness Questionnaire†	35.85 (7.1)	41.78 (6.2)	32.2 (6.4)	43.7 (10.8)	6.66	46.38***
Multidimensional Fatigue Inventory†	12.6 (3.8)	10.6 (4.0)	12.2 (3.0)	9.8 (3.8)	0.20	13.83***
PROMIS–Sleep Disturbance†	50.1 (7.0)	44.5 (7.3)	51.1 (7.7)	43.6 (5.8)	0.57	22.68***
PROMIS–Sleep Related Impairment†	54.9 (7.6)	49.1 (6.3)	56.0 (7.3)	46.6 (8.6)	1.40	23.15***
Pittsburgh Sleep Quality Index†	5.4 (2.4)	6.7 (2.6)	6.9 (3.6)	5.8 (2.3)	4.34	0.13
Diary
Total sleep time (h)	6.9 (0.6)	7.2 (0.7)	6.8 (1.0)	7.3 (0.7)	0.45	9.14
Sleep-onset time (hh:mm)	02:19 (1:55)	00:44 (1:34)	01:45 (2:05)	00:01 (1:41)	0.12	72.64***
Sleep-offset time (hh:mm)	09:45 (1:50)	08:31 (1:19)	09:20 (1:46)	08:02 (1:47)	0.27	61.20***
Actigraphy
Total sleep time (h)	6.5 (0.7)	7.0 (0.8)	6.3 (0.8)	6.9 (0.9)	0.01	17.34***
Sleep-onset time (hh:mm)	02:37 (1:59)	01:08 (1:23)	02:20 (1:40)	00:56 (1:51)	0.01	51.24***
Sleep-offset time (hh:mm)	09:50 (1:41)	08:41 (1:21)	09:32 (1:54)	08:13 (1:36)	0.57	54.69***

†Post-treatment M-DLMO condition, n = 18, due to missing data from 1 participant. ‡Sample sizes are as follows: pretreatment E-DLMO condition, n = 15; post-treatment E-DLMO condition, n = 14; pretreatment M-DLMO condition, n = 20; post-treatment M-DLMO, n = 18. ***P < .001. DLMO = dim light melatonin onset, PROMIS = Patient-Reported Outcome Measurement Information System, SD = standard deviation.

DISCUSSION

This randomized, controlled, double-blind trial assessed the preliminary effectiveness of timing 0.5 mg of exogeneous melatonin to measured DLMO vs sleep-onset time, administered in conjunction with the behavioral interventions of time in bed scheduling and scheduled evening dim light. We found that both treatment strategies produced clinically significant improvements in circadian and sleep parameters; DLMO advanced an average advance of nearly 90 minutes across the sample, with similar advances observed in sleep-onset and sleep-offset times when measured by both sleep diary and wrist actigraphy. Equivalent improvements were also observed in other outcomes, including fatigue, sleep disturbance, and impairment related to sleep.

However, other outcomes, including sleep quality and total sleep time were not significantly different from pre- to post-treatment. Notably, average sleep quality scores on the PSQI were in the range indicative of poor sleep quality at both time points, suggesting that sleep quality remained poor despite the interventions. This may be reflective of the intervention, or of the time frame for which the measure is completed (past 4 weeks), which is different from the other outcome measures (past 1 week).

The present findings add to the relatively small body of published controlled trials of exogeneous melatonin for DSWPD. Of note, the DLMO advance observed in this study is comparable to other published trials in which exogeneous melatonin was administered in the late afternoon/early evening for 4 weeks. These trials found an advance of 98 minutes for 5 mg of melatonin administered 5 hours before DLMO³⁴ and 105 minutes for 0.3 or 3 mg melatonin administered between 1.5 to 6.5 hours prior to DLMO,¹⁶ with larger advances observed with earlier administration times relative to DLMO. A third trial reported a smaller phase advance (44 minutes), perhaps due to a later melatonin administration time—1 hour before the desired bed time.¹⁸ Thus, taken together with previous work, our results provide further evidence that an earlier administration time of melatonin may be advantageous for phase advancement.

Our hypothesis that timing exogeneous melatonin to measured DLMO would result in better treatment outcomes vs timing to sleep onset was not supported. There are several possible explanations for our null findings. Other aspects of the protocol, such as closely regulated time in bed, time in bed advancement, and instructions regarding timing of dim light exposure, were shared across conditions and could be responsible for the observed effects. Importantly, participants in the E-DLMO condition had a greater advance of wake time, which may have increased exposure to morning light and contributed to the observed phase advance. Light is the most potent circadian zeitgeiber³⁶ and morning bright light advances the circadian phase, which may aid in DSWPD treatment.¹² Limiting light exposure before bed may have been particularly impactful in this study given evidence that suggests increased sensitivity to evening light in delayed sleep-wake phase disorder.³⁷ Participants in the E-DLMO condition also had a greater advance of melatonin time from the first to second weeks of treatment. Their scheduled melatonin time–sleep-onset interval remained close to 5 hours, whereas the interval was significantly longer (∼6.5 hours, on average) across the treatment for the M-DLMO condition. These differences may explain why participants in this group evidenced such a large phase advance even though their initial scheduled melatonin time was much closer to DLMO vs the M-DLMO group. Finally, it is also possible that the study was underpowered to detect significant between-group differences.

Strengths of the study include objective measurement of circadian timing, confirmation of DSWPD diagnosis using wrist actigraphy, objective monitoring of compliance to melatonin administration time, and multimodal assessment of sleep including actigraphy and sleep diary. We note here that, although confirmation of DSWPD diagnosis using wrist actigraphy is a strength of the study, the precise inclusion/exclusion criteria used in any study of DSWPD have the potential to impact results. Limitations should also be considered. The lack of a control group with respect to melatonin (ie, a group that received the behavioral interventions plus a placebo pill) is a limitation that prevents understanding the additive impact of melatonin in conjunction with behavioral interventions. The DLMO was not able to be calculated for several of the participants, which is a limitation of using this method to characterize circadian timing. Our sample was drawn, in part, from community-based advertisements and not exclusively patients presenting to sleep medicine clinics, which may limit the generalizability of the findings to treatment-seeking patients with DSWPD. Other limitations include a smaller sample size; the potential for seasonal confounding; possible effects of the menstrual cycle on the outcomes for female participants; age effects (participants were not matched on age and our sample includes both younger and older participants); less tightly controlled aspects of the protocol, meant to approximate typical clinical practice (ie, change in scheduled melatonin time and time in bed schedule were driven by the study clinician’s judgment); and important unquantified factors (eg, morning light exposure) that may account for the findings.

Overall, the present work suggests that timing melatonin to DLMO and to sleep-onset time, when administered in conjunction with behavioral interventions, may produce improvements in the circadian and sleep parameters that are most relevant to quality of life in DSWPD. Given the challenges inherent in measuring DLMO in clinical practice, our results provide preliminary evidence that larger-scale trials are warranted to more conclusively determine the role of DLMO measurement in DSWPD treatment. It is also important to note that several elements of the research protocol may have improved treatment outcomes beyond melatonin alone, including the behavioral aspects of the intervention, other factors known to impact circadian timing (eg, morning light exposure), weekly follow-up with a study clinician, and feedback to the study clinician regarding adherence to melatonin timing and time in bed schedule. Potential next steps in this line of work include a larger-scale trial, future dismantling studies to inform which aspects of the treatment protocol are “key ingredients,” as well as translation into a digitally based therapy, such as a smartphone application, to enhance treatment uptake.

DISCLOSURE STATEMENT

All authors have seen and approved the manuscript. Work for this study was performed at the University of Michigan. This study was funded by the American Academy of Sleep Medicine Foundation Strategic Research Award (Principal Investigator: Swanson). Additional support for this work was provided by the Michigan Institute for Clinical and Health Research at the University of Michigan (NIH, UM1TR004404). This study was conducted under a “Investigational New Drug” protocol for melatonin (Food and Drug Administration Protocol #141269; sponsor: Swanson). Dr. Burgess serves on the scientific advisory board for Natrol, LLC, and was a consultant for F. Hoffmann-La Roche Ltd. The other authors report no conflicts of interest.

ABBREVIATIONS

CI

confidence interval

DLMO

dim light melatonin onset

DSWPD

delayed sleep-wake phase disorder

E-DLMO

estimated dim light melatonin onset

M-DLMO

measured dim light melatonin onset

PROMIS

Patient-Reported Outcome Measurement Information System

SD

standard deviation