Home-based dim light melatonin onset assessment among adults with obesity: feasibility and procedural considerations

Abstract

Study Objectives

Dim light melatonin onset (DLMO) is the best-established marker of central circadian phase and may contribute to unraveling the role of the circadian system in obesity. This study evaluated DLMO among individuals with obesity using a home-based assessment and explored its clinical correlates and procedural variations.

Method

Fifty-eight women (mean [SD] age 40.9 [7.8] years) and body mass index (41.4 [6.6] kg/m²) completed a home-based DLMO assessment, measures of sleep quality, diurnal preference, and cardiometabolic parameters. Procedural variations we explored included individualized versus standardized DLMO thresholds, 7 versus 3 days assessment of sleep onset timing (SOT), as well as diary-based, actigraphy-based, or a “combined” method to calculate SOT, and hourly versus half-hourly saliva sample data points. Correlation coefficients and univariate ANOVA models were used for statistical analysis. Bland–Altman plots were used to inform agreement between methods.

Results

DLMO was detected in 98.2% and 89.6% of participants using an individualized or a standardized threshold, respectively. DLMO correlated with SOT but not with body mass index, cardiometabolic parameters, sleep quality, or diurnal preference. A later SOT and a larger phase angle of entrainment (DLMO-SOT) correlated with younger age and with eveningness. Most procedural alternatives showed good agreement with the original methods.

Conclusions

Home-based assessment yielded a high rate of detectable DLMO in women with obesity. Diurnal preference was not correlated with central circadian phase, suggesting that other factors (e.g. behavioral, sociodemographic) may be relevant in chronotype assessment in this population. We offer implications for future research including procedural variations to consider.

Statement of Significance

Evidence from models of circadian misalignment suggest an association between circadian disruption (e.g. delayed circadian phase) and obesity. However, dim light melatonin onset (DLMO), the best-established marker of central circadian phase, has been poorly explored in obesity. Additionally, DLMO assessment remains costly and laborious and individuals with obesity are often excluded from chronobiological studies. While other studies using in-hospital melatonin secretion profiles or home-based DLMO assessments have included individuals with obesity among their participants, a focused analysis on home-based DLMO in this population is lacking. We provide evidence on DLMO assessment using a home-based procedure among participants with obesity and explore its clinical correlates. We also evaluate procedural variations that could facilitate its widespread use in future studies.

Introduction

The circadian system is a complex brain–body interaction network that is involved in the organization of most physiological processes throughout the 24-h cycle [1]. It consists of the suprachiasmatic nucleus (SCN) in the hypothalamus and its interaction with peripheral oscillators located in most tissues throughout the body. Mistiming between the cycle of the components of the circadian system is referred to as “internal circadian misalignment,” while a conflict or mistiming between environmental or behavioral cues and the circadian system cycle is referred to as environmental or behavioral circadian misalignment and can cause adverse health consequences [2–4]. These conflicts represent forms of circadian system disruption that have been linked to adverse cardiometabolic outcomes, including obesity, also in real-life settings [5, 6]. Such observations have drawn considerable attention to the study of the circadian system across biomedical fields. However, methods to obtain objective markers of circadian system timing in humans to study its potential role in obesity remain burdensome, costly, and are not readily available, which hampers progress in the field [6–8].

Dim light melatonin onset (DLMO) is currently the most reliable marker of the central circadian phase [9–11]. It relies on sequential measurements of melatonin levels in saliva or blood samples that are collected throughout the evening/early night under dim light conditions (required to prevent confounding by acute retinal light-induced melatonin suppression) and reflects the biological transition from the day into the night. DLMO occurs when the SCN’s GABA-ergic suppression of the multi-synaptic pathway is removed, leading to the disinhibition of the pineal to release of melatonin into the circulation [12, 13] and can be observed as a sharp increase in melatonin levels above a specified threshold at the beginning of the biological night. DLMO procedures were initially conducted in controlled sleep laboratories or in-hospital settings, which is costly and labor intensive and restricted its widespread use. This is an important limitation in the field because being a reliable marker of central circadian phase, DLMO serves as reference to assess circadian alignment through measures of phase angle of entrainment (e.g. the relationship between DLMO and sleep timing) as well as for conducting studies of chronobiological interventions [14–16]. Due to its relevance and potential applications in research and clinical settings, there have been vigorous and successful efforts to transition into less costly and simpler, yet reliable home-based DLMO methods [9, 17, 18].

Models of circadian misalignment such as night shift work, social jet lag, or in some individuals with an evening chronotype suggest an association between circadian disruption (e.g. delayed circadian phase) and obesity [19, 20]. Indeed, there are reports that melatonin levels may be reduced in individuals with higher body mass index (BMI) [21]. However, most human studies to date have relied on subjective assessments of behavioral preference or behavioral habits and the association between BMI and DLMO is unclear. Moreover, available studies have focused mainly on populations with normal weight and exclude individuals based on BMI criteria, due in part to the concern that the circadian system may be altered in obesity and introduce confounders to DLMO assessments [21, 22]. For example, two studies, one in adolescents/young adults (n = 161) and one in adults (n = 59) without obesity, showed no association between BMI and DLMO [23, 24]. These studies used an interpolation method to determine DLMO at a standardized threshold of 4 pg/mL. By contrast, a small study (n = 35) in adolescents with obesity found an association between a later DLMO and a higher BMI [25]. This study was conducted in-hospital and used a standardized 3 pg/mL threshold to determine DLMO. While other studies using in-hospital melatonin secretion profiles [26] or home-based DLMO assessments have included individuals with overweight or obesity among their participants, a focused analysis on home-based DLMO in this population is lacking [8, 16, 27, 28].

With a prevalence of obesity around 40% among US adults [29], there is an increasing interest in unraveling the role of the circadian system in obesity and its associated comorbidities. However, the feasibility of conducting home-based DLMO assessments among individuals with obesity and methodological alternatives to facilitate its widespread use are not clear. In this study, we report the preliminary results of home-based DLMO assessments among adults with obesity and a broad BMI range. The main objectives of this study were to evaluate whether a home-based DLMO assessment was feasible among individuals with obesity, to explore alternative procedural and analytical approaches to facilitate home-based DLMO assessments, and to evaluate the correlation of DLMO with BMI and other clinical variables. We hypothesized that a home-based DLMO assessment and alternative procedural approaches will be feasible among individuals with obesity. We also hypothesized that a later DLMO would correlate with adverse clinical outcomes, including a higher BMI, blood pressure, heart rate, later diurnal preference, smaller phase angle, as well as lower sleep quality.

Materials and methods

This is a preliminary report of the observational phase of a larger ongoing study (Clinicaltrials.gov # NCT04724668) evaluating the circadian system among adults that had obesity with and without binge eating disorder (BED). Study method and results are reported following the Strengthening the Reporting of Observational Studies in Epidemiology Statement for cross-sectional studies [30]. Study procedures involving participants were performed according to the principles outlined by the Helsinki Declaration and were conducted at the Lindner Center of Hope in Mason, OH. The study protocol was approved by the University of Cincinnati Institutional Review Board. Participants voluntarily signed an informed consent form prior to initiating study procedures.

All participants had obesity (BMI ≥ 30 kg/m²), any gender, ages 18–50 years (inclusive), and participants with and without BED were allowed. Participants were excluded if they had another eating disorder (concurrent night eating disorder was allowed). Participants did not have a severe comorbid psychiatric condition (e.g. severe depressive or manic episode, psychotic symptoms), a current substance use disorder (past month), a significant risk of suicide, or suicidal behavior in the past year. Participants were not currently using light therapy, did not have frequent (e.g. >1/week) use of melatonin in the past month, did not engage in night shift work in the past month, and did not travel across more than one time zone in the 2 weeks prior to screening. Participants were allowed to take psychiatric medications if the dose was stable for 8 weeks prior to screening. Participants were not on a medication known to affect DLMO or the circadian system, including beta-blockers, hypnotic sedatives, anticoagulants, antidiabetic drugs (including GLP-1 receptor agonists), oral corticosteroids, Non-steroidal anti-inflammatory drugs (or were asked to suspend 48 h prior to DLMO procedures), and immunosuppressant medications. Participants did not have a lesion in the oral cavity or a clinically significant unstable medical condition, including seizure or neurodegenerative disorders, thyroid conditions, autoimmune disorders, and cardiovascular disease. Participants were not pregnant or breastfeeding and had not participated in a clinical trial in the past month.

Study procedures

After signing the informed consent form, participants completed a screening evaluation. The Structured Clinical Interview for the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (SCID-5) [31] was used to evaluate current and past psychiatric diagnoses. A medical/psychiatric history and physical exam were performed. After meeting eligibility criteria, participants attended a baseline visit to receive training on the completion of a paper diary to record sleep–wake times and to wear an actigraphy wrist watch (Actiwatch Spectrum Plus, Philips, UK) for 2 weeks [32, 33].

Participants returned for a study visit scheduled at noon to complete a series of assessments and be trained on DLMO procedures that would be conducted at home that same day. Briefly, participants completed self-report instruments, including the morningness-eveningness questionnaire (MEQ) to measure diurnal preference and determine chronotype [34], and the Pittsburgh Sleep Quality Index (PSQI) [35] to assess sleep quality. Paper and/or REDCap-based electronic versions of instruments and scales were used [36]. Vital signs (blood pressure and heart rate) were obtained while sitting after at least 5 min of rest. Height, weight, BMI, and body composition parameters (e.g. fat mass, body water, lean mass, basal metabolic rate) were obtained with participants in light clothing using an FDA cleared Body Composition Analyzer (MC-780 U, TANITA, Japan).

A home-based DLMO kit was used to assess the circadian phase in a home environment aimed at reducing costs and impacts to patients [18, 37]. This home-based approach correlates highly with laboratory DLMO (r = 0.93, p < .001), and 92% of the assessments are valid among adults [18, 37]. In this study, research staff obtained the diaries and actigraphy watches from the participant for their review. Diary and actigraphy data for sleep onset time (SOT) was obtained to establish DLMO saliva collection timeframe. A visual inspection allowed for corrections and verification of SOT inconsistencies between actigraphy (Actiware) and diary reports. For example, obvious earlier SOT estimation by Actiware software during a low activity period versus diary SOT, as confirmed by visual inspection of locomotor activity data and confirmation by the participant. In this instance, SOT was for example established as the first epoch of at least 5 min of no locomotor activity that is consistent with the diary reported SOT. This “verified” SOT was used to calculate a past 7-day average to establish the timing for DLMO procedures. The research team then trained participants on how to use a DLMO kit at home and were provided with detailed written instructions for the procedures. The kit included a study device with a timer set with alarms that prompted participants during the procedures. Participants were instructed to prepare for procedures 6.5 h prior to the verified SOT average obtained from the prior 7 days and remain under strict dim light conditions with “just enough light to read the instructions” throughout the procedures. Participants were instructed to sit in the 10 min before each sample and abstain from vigorous physical activity, drinks, and food that could alter melatonin levels. Food was not allowed within 10 min prior to saliva collection, when participants were instructed to brush their teeth with an ultra-soft toothbrush (Post-Operation, GUM-Sunstar, Schaumburg, IL) if food was consumed and to rinse with water. They were instructed to dim the screen from electronic devices and avoid their use (watching TV was acceptable at 8 ft from dimmed screen) and to start wearing blue-light blocking glasses (>98% absorption/blocking per manufacturer; UVEX, Honeywell, USA or ABL0311, LVIOE, China). The glasses were expected to mitigate the potential effect of accidental light pulses (>50 lux) on melatonin concentration [18]. Dim light conditions were monitored with a photo sensor at eye-level (primary light exposure monitoring tool) mounted with a strapless actiwatch with light sensing capability (Actiwatch Spectrum Plus, Philips, Cambridge, UK) over the bridge of the blue-light blocking glasses and facing forward. Locomotor activity data on this actiwatch were reviewed to monitor compliance with wearing the blue-light blocking glasses. Participants were also instructed to wear their wrist actiwatch around a watch pillow attached to a necklace with the light sensor facing forward as a second light monitoring tool during DLMO procedures.

Participants received the instruction to initiate saliva sample collection every half hour starting 6 h before verified SOT for a total of 13 samples, with the last sample collected at verified SOT. Previous studies show that DLMO typically occurs before usual SOT [18, 38]. In this study, the track caps Medication Event Monitoring Systems (MEMS, Aardex, Belgium) was utilized as part of the kit that enabled monitoring of the timing at which participants took clean cotton swabs out of the bottle to then proceed to collect saliva samples. Participants kept saliva samples frozen in their freezer and were asked to deliver them to the research team inside a study cooler with icepacks on the next morning. Saliva samples were processed using Novolytix (previously Bühlmann) saliva RIA assays at SolidPhase (Portland, ME), which enables the detection of low (i.e. 0.2 pg/mL) baseline melatonin levels [39]. The DLMO threshold was established as the average of 3 baseline melatonin concentration + two standard deviations (3 BL + 2 SD) based on previous research [40, 41]. A DLMO assessment was considered compliant and suitable for analysis if: (1) saliva collection was conducted according to schedule or deviations from schedule were minimal (<15 min) around the time of a clear increase in melatonin levels above baseline levels per visual inspection; (2) dim light conditions were met according to light exposure at eye level was <50 lux, or if these data were missing (e.g. photo sensor malfunction), a similar threshold could be considered from the actiwatch necklace photo sensor. Results would also be compliant and suitable for analysis if light exposure to >50 lux was brief (<5 min), the blue-light blocking glasses were being worn and it did not occur within 30 min prior to DLMO; and (3) melatonin levels across saliva samples allowed for a DLMO calculation according to the DLMO threshold to be analyzed [18, 37].

Statistical analysis

Pearson correlation coefficients and univariate ANOVA models were used to examine relationships among DLMO, SOT, phase angle, and clinical characteristics. To explore the feasibility of procedural alternatives to decrease the impact and cost of DLMO home-based assessments, we analyzed a shorter SOT assessment time frame (3 vs. 7 days of recording), as well as diary-based, actigraphy-based, or a “combined” verified diary/actigraphy methods to calculate SOT. The feasibility of collecting hourly versus half-hourly saliva sample collection to obtain DLMO results was also explored. Bland–Altman plots were generated to inform agreement between procedural methods [42].

All time variables were included using a 24-h clock format. In the analyses, times after midnight were reformatted to be the actual time plus 24 h (e.g. 1:00 a.m. was reformatted to be 25:00) to correctly examine averages and differences in times. All analyses were conducted with SAS version 9.4 (Cary, NC, USA). Statistical tests were two-sided with a significance level of 0.05. The current sample size analyzed afforded 80% power to detect a clinically meaningful correlation coefficient of at least 0.36 with an alpha = 0.05.

Results

A total of 58 participants successfully completed a home-based assessment prior to the intervention phase of the study and were included in the present analysis. The sociodemographic and clinical characteristics of participants are presented in Table 1 and Table S1. All participants were female, with a mean (SD) age 40.9 (7.8) years. The mean BMI was 41.4 (6.6) kg/m², with a range between 30.8 and 60 kg/m², and 52% of participants had a BMI higher than 40 kg/m².

Table 1.

Participant demographics and clinical characteristics

Demographics and clinical characteristics	Mean (SD) or n (%)	N
Age, years	40.9 (7.8)	58
Sex, female	58 (100%)	58
Ethnicity, non-Hispanic	58 (100%)	58
Race		58
White	47 (81%)
Black	9 (16%)
Mixed	2 (3%)
Stimulant medication	2 (3%)	58
Antidepressant medication	22 (38%)	58
Contraceptive medication	4 (7%)	58
BMI, kg/m2	41.4 (6.6)	58
BMI class, kg/m2		58
30 to <35	10 (17%)
35 to <40	18 (31%)
40 or more	30 (52%)
Fat mass, %	43.7 (4.8)	58
Body water, %	40.2 (3.5)	58
Muscle mass, %	53.3 (4.6)	58
Basal metabolic rate, kcal	1978 (279)	58
Systolic blood pressure, mmHg	118.5 (8.1)	58
Diastolic blood pressure, mmHg	78.1 (5.5)	58
Heart rate, bpm	74.0 (9.3)	58
Sleep onset time, 24-h clock time	23.8 (1.0)	58
DLMO 3 BL + 2 SD, 24-h clock time	20.5 (1.0)	57
DLMO 3 pg/mL, 24-h clock time	20.8 (1.1)	52
DLMO 3 BL + 2 SD (hourly saliva), 24-h clock time	20.4 (0.9)	55
DLMO 3 pg/mL (hourly saliva), 24-h clock time	20.7 (1.2)	51
Phase angle, DLMO-SOT BL, h	3.3 (1.0)	57
Phase angle, DLMO-SOT, pg/mL, h	3.0 (1.1)	52
PSQI global score	5.5 (2.7)	58
MEQ total score	54.2 (9.6)	58
MEQ item #19	13 (22%)	58
Definitely a “morning” type	11 (19%)
Rather more a “morning” than an “evening” type	17 (29%)
Rather more an “evening” than a “morning” type	17 (29%)
Definitely evening type	13 (22%)

Abbreviations: standard deviation (SD); body mass index (BMI); millimeters of mercury (mmHg); beats per minute (bpm); dim light melatonin onset (DLMO); baseline (BL); sleep onset time (SOT); Pittsburgh sleep quality Index (PSQI), morningness-eveningness questionnaire (MEQ)

DLMO assessment and clinical correlates

The mean (SD) verified (combined diary and actigraphy data) SOT (for past 7 days) was 23.8 (1.0) h. The phase angle of entrainment (difference between a DLMO-combined SOT with a 3 BL + 2SD threshold) was on average 3.3 (1.0) h. Light exposure at eye-level during DLMO procedures was 1.5 (3.0) lux. Dim light conditions were met by all participants, although 15/58 (25.8%) experienced at least one brief exposure of >50 lux while wearing the blue-light blocking glasses. Exposed participants experienced 4.3 (5.4) individual exposures lasting 0.8 (0.5) min at 249.6 (967.6) lux. Details on light exposure captured by photo sensors at eye-level and necklace are presented in Table S2. The use of an individualized saliva melatonin level threshold for DLMO (3 BL + 2 SD) yielded detectable results in 98.2% (n = 57/58) of the DLMO assessments. A standardized 3 pg/mL threshold yielded detectable DLMO results in only 89.6% (n = 52/58) of the assessments. One DLMO was not detectable using both threshold methods because of unexplained abnormally high melatonin levels in all samples (>50 pg/mL). The rest of non-detectable results (n = 5) for the 3 pg/mL threshold method were due to persistent saliva melatonin levels above 3 pg/mL. DLMO threshold methods were highly correlated (correlation coefficient 0.81, p < .001). The DLMO calculated with a 3 BL + 2SD threshold occurred earlier compared to the 3 pg/mL threshold (−0.35 [SD = 0.67], p < .001) the Bland–Altman plot shows the bias within methods is small and most assessments are within the limits of agreement. However, even within the limits of agreement, 9 (15%) of the assessments show ≥1 h difference between the methods (Figure 1A).

Figure 1.

Bland–Altman plots illustrating agreement between methods to obtain dim light melatonin onset (DLMO). In (C) individualized DLMO threshold refers to either the average of 3 baseline values + 2 standard deviations (3 BL + 2 SD) threshold for half-hourly melatonin level data points or an average of 2 baseline values + 2 standard deviations threshold for DLMO threshold when using hourly melatonin level data points. Solid horizontal line reflects the mean difference between the two methods. Dashed horizontal lines reflect limits of agreement (mean ± 1.96 SD). Other abbreviations: Picogram/milliliter (pg/mL).

We explored the correlation between DLMO (3 BL + 2SD threshold), SOT, phase angle of entrainment, and clinical variables (Table 2 and Figure S1, A–C). DLMO showed a statistically significant correlation with the combined SOT (correlation coefficient 0.51, p < .001). DLMO was not correlated with age, BMI (Figure S2, A), body composition parameters, basal metabolic rate, blood pressure, heart rate, or PSQI global score. MEQ total scores did not correlate with DLMO results (Figure S3). However, a later SOT was correlated with a younger age (r = −0.28, p = .03) and with lower MEQ scores (i.e. more eveningness; r = −0.51, p < .001). In addition, a larger phase angle of entrainment (DLMO-SOT) was correlated with a younger age (r = −0.38, p = .003), a later SOT (r = 0.53, p < .001), a lower score on the MEQ (r = −0.31, p = .01), and an earlier DLMO (r = −0.46, p < .001). In this sample, SOT and phase angle of entrainment were not correlated with BMI (Figure S2, B and C), body composition parameters, blood pressure, heart rate, or PSQI scores (Table 2). There was no difference in mean (SD) DLMO (3 BL + 2 SD), between those with and without an antidepressant (20.4 [0.9] vs. 20.5 [1.0], 24-h time, p = .69).

Table 2.

Correlation between DLMO, SOT, phase angle, and clinical characteristics

Clinical characteristics	DLMO (3 BL + 2 SD)			SOT			Phase angle
	Correlation coefficient	P-value	n	Correlation coefficient	P-value	n	Correlation coefficient	P-value	n
Age, years	0.09	.48	57	−0.28	.03	58	−0.38	.003	57
BMI, kg/m2	0.14	.31	57	0.12	.37	58	−0.01	.91	57
Fat mass, %	0.04	.79	57	0.05	.71	58	0.01	.92	57
Body water, %	−0.04	.79	57	−0.04	.77	58	0.00	.99	57
Muscle mass, %	0.02	.90	57	0.00	.99	58	−0.01	.93	57
Basal metabolic rate, kcal	0.09	.49	57	0.14	.29	58	0.05	.71	57
Systolic BL, mmHg	0.13	.34	57	0.01	.93	58	−0.12	.38	57
Diastolic BL, mmHg	0.08	.54	57	−0.06	.64	58	−0.15	.28	57
Heart rate, bpm	−0.10	.46	57	−0.13	.34	58	−0.03	.82	57
SOT	0.51	<.001	57				0.53	<.001	57
PSQI global score	0.05	.72	57	0.16	.24	58	0.13	.35	57
MEQ total	−0.20	.13	57	−0.51	<.001	58	−0.32	.01	57
DLMO, 3 BL + 2 SD				0.51	<.001	57	−0.46	<.001	57

The sleep onset time (SOT) refers to the combined and verified data obtained from actigraphy and diaries over the 7 days prior to dim light melatonin onset (DLMO) procedures. Phase angle reflects the difference between SOT and DLMO calculated using the average of 3 baseline values + 2 standard deviations (3 BL + 2 SD) threshold method. Other abbreviations: years (years); body mass index (BMI); millimeters of mercury (mmHg), beats per minute (bpm); Pittsburgh sleep quality inventory (PSQI), morningness-eveningness questionnaire (MEQ).

SOT assessment methods and recording timeframe

We evaluated three methods for calculating SOT for the past 7 days: actigraphy-based, diary-based, and combined (verified diary/actigraphy). The SOT using the three methods were correlated with each other. However, the SOT using the diary method occurred later compared to the actigraphy method (0.41 h ± 0.91, p = <.001), and the actigraphy SOT occurred earlier compared to the combined method (−0.47 h ± 0.73, p < .001). There was no time difference between the diary and combined SOT methods (Table 3). As shown on the Bland–Altman plots, the agreement was good between the three SOT methods (Figure 2A–C).

Table 3.

DLMO thresholds, sampling rate, and SOT methods correlations

Correlations evaluated	Correlation coefficient or group difference	P-value	N
DLMO, 3 BL + 2 SD vs. 3 pg/mL, correlation	0.81	<.001	52
DLMO difference (3 BL + 2 SD—3 pg/mL), h	−0.35 (0.67)	<.001	52
DLMO hourly, 3 BL + 2 SD vs. 3 pg/mL, correlation	0.70	<.001	50
DLMO hourly difference (3 BL + 2 SD—3 pg/mL), h	−0.35 (0.78)	.003	50
SOT, diary, 7 days vs. watch, 7 days, correlation	0.68	<.001	56
SOT, diary, 7 days vs. combined, correlation	0.90	<.001	55
SOT, watch, 7 days vs. combined, correlation	0.79	<.001	53
SOT difference (diary, 7 days—watch, 7 days), h	0.41 (0.91)	.001	56
SOT difference (diary, 7 days—combined), h	−0.04 (0.46)	.47	55
SOT difference (watch, 7 days—combined), h	−0.47 (0.73)	<.001	53
SOT, diary, 3 days vs. diary 7 days, correlation	0.71	<.001	58
SOT, watch, 3 days vs. watch 7 days, correlation	0.83	<.001	55
SOT, combined, 3 days vs. combined 7 days, correlation	0.88	<.001	56
SOT, diary, 3 days vs. watch, 3 days, correlation	0.63	<.001	55
SOT, diary, 3 days vs. combined, correlation	0.87	<.001	58
SOT, watch, 3 days vs. combined, correlation	0.81	<.001	55
SOT difference (diary, 3 days—watch, 3 days), h	0.24 (1.12)	.12	55
SOT difference (diary, 3 days—combined), h	−0.15 (0.64)	.07	58
SOT difference (watch, 3 days—combined), h	−0.40 (0.80)	<.001	55
Individualized DLMO, hourly vs. 30 min, correlation	0.88	<.001	55
Individualized DLMO time difference (30 min—hourly), h	0.09 (0.45)	.143	55
DLMO 3 pg/mL, hourly vs. 30 min, correlation	0.95	<.001	50
DLMO 3 pg/mL difference (30 min—hourly), h	0.12 (0.38)	.032	50

Pearson correlation coefficients or ANOVA results are provided. Individualized dim light melatonin onset (DLMO) was calculated using an average of 3 baseline values + 2 standard deviation threshold in the half-hourly (30 min) method, and an average of 2 baseline values + 2 standard deviation threshold in the hourly method. Other abbreviations: sleep onset time (SOT); picograms/milliliter (pg/mL).

Figure 2.

Bland–Altman plots showing agreement between sleep onset time (SOT) calculation methods. The SOT refers to the combined and “verified” data obtained from actigraphy (watch) and diaries over the 7 or 3 days prior to dim light melatonin onset (DLMO) procedures. Solid horizontal line reflects the mean difference between the two methods. Dashed horizontal lines reflect limits of agreement (mean ± 1.96 SD).

To evaluate whether shorter SOT recording durations are feasible, the agreement between combined SOT methods for the prior 3- and 7-day data was evaluated. The combined SOT methods using 3- and 7-day data showed good agreement (Figure 2D). Each SOT method obtained with 3 days or 7 days of data recordings showed a statistically significant correlation (Table 2). The three methods for calculating SOT (actigraphy, diary, and combined) using past 3-day data were all correlated to each other. There was no difference between the 3-day diary and actigraphy method or between the diary and combined methods. However, the SOT calculated using the actigraphy method occurred earlier than the combined method (−0.40 h ± 0.80, p < .001) (Table 2).

Hourly versus half-hourly saliva collection

To evaluate whether hourly versus half-hourly saliva sample collection was a feasible alternative to assess DLMO, we evaluated and compared the agreement and correlation between these two methods. For this purpose, the interpolation method to obtain DLMO underwent a blinded re-analysis with hourly saliva melatonin sample collection data points instead of half-hourly data points. In addition, for the hourly DLMO calculation, instead of three baseline values (only 81% detectable results, n = 47/58), we opted for a new modified approach using the average of two baseline values + 2 SD (2 BL + 2 SD) to establish the “individualized” melatonin level threshold for DLMO. This modified hourly sample method yielded 94.8% (n = 55/58) of DLMO assessments that were considered detectable. The three non-detectable results were due to consistent levels above 50 pg/mL (n = 1) and a non-observable DLMO (n = 2). With the use of a standardized 3 pg/mL DLMO threshold method for hourly samples, only 86.2% (n = 50/58) DLMO assessments were detectable. Of eight non-detectable results, one was due to consistent levels above 50 pg/mL (n = 1), while the seven others were due to levels consistently above 3 pg/mL (n = 5) or no observable DLMO (n = 2).

The hourly and half-hourly individualized DLMO methods did not show a time difference. The standardized (3 pg/mL) DLMO half-hourly method occurred slightly later than the hourly method (0.12 ± 0.38 h, p = .032). The hourly and half-hourly DLMO methods showed good agreement (Figure 1B–D) and a statistically significant correlation when utilizing both, the individual and the standardized (3 pg/mL) DLMO thresholds (Table 3). However, even with most assessments within the limits of agreement, 13 (22%) of the hourly assessments show ≥1 h difference between the individualized and standardized methods (Figure 1B).

Discussion

In this study, we investigated the feasibility of home-based DLMO assessment among adults with obesity, evaluated its clinical correlates, and explored the effect of procedural variations. Our findings suggest that the majority of DLMO assessments were detectable and procedural alternatives in terms of SOT assessment methods and saliva sample collection rate showed good agreement with the original method. Our results also indicate that an individualized melatonin level threshold for DLMO may offer advantages over a standardized method. These procedural variations may be adapted according to clinical or research contexts, including costs and participant characteristics. Contrary to our hypothesis, BMI, MEQ scores, and sleep quality were not correlated with DLMO. However, a later SOT and larger phase angle of entrainment were correlated with younger age and eveningness, while a larger phase angle of entrainment was correlated with an earlier DLMO, suggesting the involvement of other factors in the assessment of diurnal preference in this population. Collectively, these results indicate that home-based DLMO assessments yield a high rate of detectable results and procedural variations to the method may be considered for future studies among individuals with obesity.

In this study, DLMO assessment was detectable in 98.1% of study participants, which is consistent with a previous report where the home-based procedures were initially validated against a laboratory-based DLMO (92%) [18, 37]. These results support the use of a home-based DLMO assessment in adults with obesity, with a wide BMI range (30.8–60 kg/m²), where over 50% of participants were classified as having morbid obesity. Remarkably, BMI was not correlated with DLMO, SOT, or phase angle of entrainment. This observation is consistent with two previous studies among normal weight adults [23, 24],, and contrast with an inverse correlation between BMI and a DLMO obtained in controlled laboratory conditions found among adults (BMI range 16–38 kg/m²) [16]. Our findings also differ from a study in adolescents with obesity that reported a correlation between BMI and DLMO assessment under controlled laboratory conditions [25]. Additionally, diurnal preference was not correlated with DLMO, which contrasts with the inverse correlation observed between MEQ scores and DLMO among adults in the general population [43]. Compared to data from the general population across comparable age groups [43], the DLMO (mean DLMO 20:30 h, and the latest obtained at 22:30 h) in our study was not delayed, which may partially explain the lack of correlation between DLMO, MEQ scores, and other cardiometabolic parameters. Notably, a previous study among US adults (n = 9004) reported SOT to occur around 23:02 h [44], which is 46 min earlier than our sample. Indeed, a later SOT and a larger phase angle (DLMO to SOT) was correlated with a lower MEQ score (indicating eveningness). In a previous small study (n = 12) among adults (BMI not reported) with similar circadian phases but different diurnal preferences, eveningness was also associated with a larger phase angle [45]. These findings further suggest that rather than central circadian phase, other internal and external factors (e.g. behavioral, social, and work-related demands) may influence diurnal preference, SOT, and phase angle in this population.

A series of procedural alternatives were explored to facilitate DLMO assessment among individuals with obesity. Consistent with previous reports, our results indicate that the use of an individualized threshold (3 BL + 2 SD) captured more DLMOs and thus may be preferred compared to the use of a standardized threshold (3 pg/mL) to calculate DLMO [41]. This is also supported by the observations that the DLMO time difference between the hourly and half-hourly comparisons between the individualized and standardized thresholds was small but statistically significant, and 15% (half-hourly) and 22% (hourly) of the assessments differed by ≥1 h, which could be relevant in clinical and research scenarios.

In this study, three methods (actigraphy, diary, and combined) to determine SOT for 3 or 7 days of data recording were evaluated. Agreement between methods was good and all methods were correlated. However, actigraphy versus the combined (verified actigraphy + diary) method occurred earlier and showed a small but statistically significant difference in time for both 3- and 7-day data recording time. Previous reports indicate that, depending on the individual, actigraphy software tends to overestimate or underestimate sleep duration [46]. It may, for example, inaccurately identify periods of wakefulness with low activity as sleep, which may impact the automated calculation of SOT. The verified approach offers the advantage of an objective and manually verifiable record of the actigraphy data and the diary data [47]. Of note, the influence of wearing an actigraphy watch on the quality and accuracy of diary reporting was not evaluated. When there is a need for decreased cost or impact to patients, it would be useful for researchers and clinicians to consider a less expensive (e.g. diary) or shorter alternative [48]. However, in optimal conditions, it would be recommended to utilize the verified assessment that combines actigraphy and diary of SOT for at least 7 days.

When considering saliva sample collection frequency, our results indicate that the modified hourly DLMO threshold method shows good agreement and correlates well with half-hourly DLMO [41]. Of note, when restricting the analysis to only using hourly data, we modified the method to include only 2 baseline values instead of 3 because the rate of detectable DLMO assessments was low using the latter. This indicates that hourly DLMO may be a reasonable alternative when there is a need to decrease cost and/or participant cost [9]. However, a small decrease in the percentage of detectable DLMO results with the hourly method should be considered when opting to implement this alternative.

There are several limitations to our study that should be considered. The sample size is limited and included only female participants. Despite targeted recruitment efforts, we were not able to recruit male participants at this stage of the study. Potential male participants did not meet eligibility criteria, disclosed no interest, or a lack of time to conduct procedures. This should be considered for the generalizability of results and when evaluating the feasibility of future studies. Although it was not the objective of this study, the lack of a laboratory-based DLMO limits the evaluation of the validity of the home-based assessment compared to the best-established method. In addition, participants met strict eligibility criteria that may have an impact on overall DLMO assessments. For example, the exclusion of individuals working night shifts, using certain medications thought to influence DLMO assessments, or having unstable medical conditions frequently associated with obesity (e.g. diabetes, hypertension) may influence results. Light exposure during DLMO procedures measured exclusively in photopic lux and not in melanopic equivalent daytime illuminance is a limitation that could be addressed in future studies. In addition, the influence of conditions like BED or night eating disorder on DLMO was not explored as it was out of the scope of the present study and will be analyzed and reported when the main study is completed. In addition, weekdays and weekends and their influence on SOT used to establish DLMO saliva sample collection timeframe was not evaluated.

Collectively, our findings indicate that detectable home-based DLMO assessments can be obtained among adults with obesity [22]. We have also identified procedural alternatives that may be considered in future studies according to needs and potential limitations. Characterizing the role of the circadian system in obesity and how obesity may impact DLMO is also relevant to study conditions with which it is strongly associated (e.g. diabetes, cardiovascular illness, psychiatric disorders) [49]. Moreover, home-based DLMO as an objective marker of central circadian phase and has served to develop phase response curve studies for chronobiological interventions [14]. The continued efforts to develop DLMO assessments that are simpler and broadly accessible may eventually facilitate its use to inform the implementation of chronobiological interventions for a myriad of circadian-related health problems, including obesity [50–52].

Author contributions

Francisco Romo-Nava (Conceptualization [lead], Formal analysis [equal], Funding acquisition [lead], Investigation [lead], Methodology [lead], Project administration [lead], Resources [lead], Supervision [lead], Validation [lead], Writing—original draft [lead], Writing—review & editing [lead]), Helen J. Burgess (Conceptualization [equal], Methodology [equal], Supervision [equal], Validation [equal], Writing—original draft [equal], Writing—review & editing [equal]), Thomas J. Blom (Data curation [equal], Formal analysis [lead], Methodology [equal], Visualization [equal], Writing—original draft [equal], Writing—review & editing [equal]), Georgi Georgiev (Data curation [equal], Investigation [equal], Supervision [equal], Validation [equal], Writing—original draft [equal], Writing—review & editing [equal]), Jakyb Stoddard (Data curation [equal], Investigation [equal], Validation [equal], Writing—review & editing [equal]), Elly McMillan (Data curation [equal], Investigation [equal], Validation [equal], Writing—original draft [equal], Writing—review & editing [equal]), Nicole N. Mori (Investigation [equal], Supervision [equal], Writing—review & editing [equal]), Christina Charnas (Data curation [equal], Investigation [equal], Validation [equal], Writing—original draft [equal], Writing—review & editing [equal]), Anna I. Guerdjikova (Investigation [equal], Writing—original draft [equal], Writing—review & editing [equal]), Robert K. McNamara (Conceptualization [equal], Data curation [equal], Investigation [equal], Methodology [equal], Supervision [equal], Writing—original draft [equal], Writing—review & editing [equal]), Jeffrey A. Welge (Data curation [equal], Formal analysis [equal], Investigation [equal], Methodology [equal], Supervision [equal], Writing—original draft [equal], Writing—review & editing [equal]), Carlos M. Grilo (Conceptualization [equal], Investigation [equal], Methodology [equal], Supervision [equal], Writing—original draft [equal], Writing—review & editing [equal]), Frank A.J.L. Scheer (Conceptualization [equal], Data curation [equal], Investigation [equal], Methodology [equal], Supervision [equal], Validation [equal], Visualization [equal], Writing—original draft [equal], Writing—review & editing [equal]), Susan L. McElroy (Conceptualization [equal], Investigation [equal], Methodology [equal], Supervision [equal], Writing—original draft [equal], Writing—review & editing [equal]).

Funding

F.R.N. was supported in part by NIH grant K23MH120503 and 1R61MH133770-01A1. F.A.J.L.S. was supported in part by NIH grants R01HL140574 and R01HL153969. C.M.G. was supported, in part, by NIH grants R01 DK49587, R01 DK114075, R01 DK 121551, and R01 DK112771. REDCap database software was supported by the National Center for Advancing Translational Sciences of the National Institutes of Health, under Award Number UL1TR001425.

Disclosure statement

Financial disclosure: H.J.B. serves on the scientific advisory board for Natrol. C.M.G. reports no competing interests but reports several broader interests which did not influence this paper including honoraria for lectures, CME activities, and presentations at scientific conferences and Royalties from Guilford Press and Taylor & Francis Publishers for academic books. F.A.J.L.S. has received consulting fees from the University of Alabama at Birmingham and Morehouse School of Medicine. F.A.J.L.S. interests were reviewed and managed by Brigham and Women’s Hospital and Partners HealthCare in accordance with their conflict-of-interest policies. F.A.J.L.S. consultancies are not related to the current work. S.L.M. is or has been a consultant to or member of the scientific advisory boards of F. Hoffmann-La Roche Ltd. Idorsia, Myriad, Novo Nordisk, Otsuka, Sipnose, Sunovion, and Takeda. She is or has been a principal or co-investigator on studies sponsored by Brainsway, Idorsia, Janssen, Marriott Foundation, Myriad, National Institute of Mental Health, Novo Nordisk, Otsuka, and Sunovion. She is also an inventor on United States Patent No. 6323236 B2, Use of Sulfamate Derivatives for Treating Impulse Control Disorders, and along with the patent’s assignee, University of Cincinnati, Cincinnati, OH, has received payments from Johnson & Johnson, which has exclusive rights under the patent.

Non-financial disclosure: F.R.N. is supported in part by NIMH grants K23MH120503 and 1R61MH133770-01A1; is the inventor on a U.S. Patent and Trademark Office patent # 10857356. F.A.J.L.S. was supported in part by NIH grants R01HL140574 and R01HL153969. F.A.J.L.S. served on the Board of Directors for the Sleep Research Society. F.A.J.L.S. interests were reviewed and managed by Brigham and Women’s Hospital and Partners HealthCare in accordance with their conflict-of-interest policies. C.M.G. was supported, in part, by NIH grants R01 DK49587, R01 DK114075, R01 DK 121551, and R01 DK112771.

Data availability

The data that supports the findings of this study are available from the corresponding author upon reasonable request.