A pre-post trial to examine biological mechanisms of the effects of time-restricted eating on symptoms and quality of life in bipolar disorder

The circadian timing system consists of a hierarchy of oscillators controlled by a central pacemaker in the suprachiasmatic nucleus (SCN). The SCN is entrained to the environment via a direct pathway from the retina [12]. Coherence among oscillatory cells in central and peripheral systems is maintained by multi-synaptic, autonomic innervation of peripheral systems by the SCN and central clock control of hormones that target the periphery [13, 14]. Disturbances to this coordinated rhythmic harmony are linked to significant mental and physical health risks [13].

In addition to autonomic innervation and hormonal communication, the timing of food intake can also set the phase of peripheral tissues, without impact on the SCN [15], leading to incongruence between central pacemaker and peripheral systems. Under circumstances where food is ingested during the active phase of the day, systemwide synchrony occurs. However, when eating occurs outside of the active phase, the light-entrained SCN and periphery are at odds in their timing. Seminal work in mouse models established that restricting feeding to the active portion of the day (i.e., the dark phase of the day) improves rhythms in the genes regulating circadian function, improves metabolic functioning (e.g., glucose tolerance and insulin resistance) [7–9, 16–19], and increases longevity independent of changes in caloric intake [20]. Thus, by coordinating peripheral oscillators and the SCN during the day, time-restricted feeding (TRF) helps coordinate circadian rhythms between brain cells in the SCN and cells within other organs (e.g., the liver, lungs, and kidneys) [21, 22]. Indeed, converging evidence across studies indicates that at least some metabolic benefits of TRF result from the resetting the phase of peripheral circadian rhythms [21].

Based on these foundational findings in animal models, the benefits of time-restricted eating (TRE) have received significant attention for human health studies. Across at least 14 randomized controlled trials, substantial metabolic benefits have been observed in glucose regulation and, in many studies, decreased body mass index, including several studies with large sample sizes [23]. In a randomized controlled trial of 80 healthy males, TRE upregulation of the ‘clock’ genes BMAL1, CLOCK, and SIRT1 was observed in serum, suggesting that TRE enhances rhythmic amplitude [24]. As observed under TRF in mice, TRE was shown to shift circadian rhythm of clock gene expression in two smaller studies [10, 11].

High satisfaction and adherence are reported by participants assigned to TRE [21, 25] that likely stem from its health benefits and improvements to sleep duration, sleep efficiency, and reductions in sleep timing variability [25, 26]. Participants report improved energy and restedness within a few weeks of beginning TRE that are sustained for at least six months post-intervention [11, 25]. In fact, several studies report that TRE enhances perceived subjective sleep quality and restedness [11, 25, 27]. Across studies, participants assigned to TRE report significantly improved quality of life ratings from baseline to the 6-month follow-up [27–30]. Of note, TRE does not seem to increase hunger [21, 31], reduce muscle mass [32], or negatively impact cognitive function [33].

Considerable evidence supports a strong association between biological rhythm disturbances and the pathogenesis of BD [34]. Disturbed sleep-activity rhythms are principal symptoms observed in manic and depressive episodes [35, 36], and these disturbances continue at lower levels for most people with BD during remission [37–40]. Sleep disturbances predict both BD onset and symptom increases in naturalistic studies of sleep, jet lag, activity, and sleep deprivation [40–45]. These issues predict vulnerability to the disorder, with evidence linking (a) BD to single-nucleotide polymorphisms within multiple clock genes, (b) mood variability to polygenic risk scores related to chronotype and low circadian amplitude, and (c) disruptions in sleep/sleep-activity rhythms [46–48] in family members of those diagnosed with BD. Accordingly, there is a need for interventions that target the circadian timing system to stabilize biological rhythms in patients with BD.

The aim of this study is to examine the mechanisms through which TRE influences symptom improvement in BD. We will assign all participants to TRE and will measure diurnal amplitudes and phases of clock gene expression (mRNA) responsible for circadian rhythm generation at the cellular level, as well as dim light melatonin onset (DLMO), the standard measure of circadian phase. We will examine whether changes in the diurnal amplitude predict improved symptoms and quality of life (QoL)—an ideal test of the circadian rhythm hypothesis of BD. To maximize the ability to show that TRE leads to changes in diurnal rhythms, we will recruit individuals who are (a) diagnosed with BD, (b) have at least some symptoms of circadian rhythm sleep–wake disorders or sleep disorders, and (c) report an eating window of ≥ 12 h.

Although our data will be preliminary, we also aim to gather information about two other potential biological mechanisms of TRE: changes in pro-inflammatory cytokines and glucose tolerance. Across studies, TRE rapidly attenuates night-time, fasting, and 24-h glucose levels, with effects sustained across 3–5 weeks of TRE as well as improved glucose tolerance, improved insulin sensitivity and resistance, and lower HbA1c within 3–6 weeks [23]. This is of importance given that over 22% of those with BD show insulin resistance and another 33% have type II diabetes; insulin resistance among those with BD predicts robustly worse symptom course, rapid cycling, treatment resistance to lithium [70], and worse neurocognitive performance. Recent pilot work showed that symptoms of BD were reduced substantially among 50% of patients who were treated with metformin, a diabetes medication [71].

TRE has been found to be associated with significant pre-post reduced pro-inflammatory cytokines, and decreased biomarkers indicative of oxidative stress [72, 73]. Therefore, substantial evidence links inflammation to BD. We will assess and control for changes in pro-inflammatory cytokines and glucose tolerance in examining circadian rhythm normalization as a mediator. This will supply critical mechanistic information to explain individual differences in responsivity to TRE.

Our goal is to examine TRE as an adjunct to psychiatric care for BD. Accordingly, participants will be asked to sustain ongoing medical care for BD.

The trial is a non-randomized, single arm (TRE) study. The primary endpoint is immediately after the 8-week-long TRE period (end of week 10). The follow-up assessments will be conducted at 6 months after the baseline, supplemented with post-baseline self-reported assessment at week 16. Hypothesized mechanisms of change will be tested at baseline and at 8 weeks post-baseline. The study setting is remote (online food plan materials, questionnaires, and interviews), supplemented with at-home collection of biological data (DLMO, clock genes, actigraphy, CGM) and assays completed by Quest Diagnostics (CMP, glucose tolerance test).

The inclusion/exclusion criteria are designed to maximize participant safety and the applicability of TRE, and secondarily, to maximize the generalizability of the findings. We will recruit adults aged 18–65 years who meet diagnostic criteria for bipolar I disorder or bipolar II disorder per the Diagnostic Interview for Anxiety, Mood, and OCD and Related Neuropsychiatric Disorders (DIAMOND), but not cyclothymia, BD not otherwise specified, or BD due to another medical condition.

To maximize the ability to determine whether TRE improves sleep- or circadian-related outcomes, the entrance criteria will include at least some current sleep or circadian sleep–wake concerns. We do not expect this to greatly diminish generalizability, as more than 75% of those with BD report sleep or circadian concerns [77]. This inclusion criteria will be met by any one of the following:

Score of ≥ 8 reflective of subsyndromal insomnia on the Insomnia Severity Index (ISI; [78–80])Early chronotype (midpoint of sleep on free days < 4.16 h) or late chronotype (midpoint of sleep on free days > 4.53) on the MCTQDisrupted social activity rhythms, as indicated by a score ≥ 30 on the Brief Social Rhythm Scale (BSRS) [81]Sleep (insomnia, hypersomnolence) or circadian sleep–wake (delayed phase, advanced phase, irregular sleep–wake, non-24-h sleep–wake-type) disorder diagnoses on the Structured Clinical Interview for Sleep Disorders-Revised (SCISD-R; [82]).

Other inclusion criteria include the following: English language adequate for completion of questionnaires or interviews, as indicated by speaking English for at least 10 years, speaking English at home, or proficiency as observed during the screening interview. Receiving medical care for BD (referrals will be provided for those who would like to begin care) Mood-stabilizing medication regimens need to be stable for at least one month. Those with BD II per the DIAMOND will be permitted to engage in the trial if not taking mood-stabilizing medication, as long as their provider has supported this treatment plan and continues to see them at least once every 3 months. Currently eating ≥ 12 h per day at least twice per week Able to operate the camera function and respond to surveys by smartphone (loaner phones will be provided as needed)

Exclusion criteria include the following:Current episode of depression, hypomania or mania, or psychosis (per the DIAMOND).Suicidal ideation coupled with current plan or intent or history of attempt as assessed by the Columbia Suicide Severity Rating Scale (C-SSRS).Eating disorder diagnosis (self-reported history of treatment or diagnosis, Short Eating Disorder Examination Questionnaire (EDE-QS; [83]) scores above clinical concern thresholds for eating disorders (≥ 15) or DIAMOND diagnosis of eating disorder (anorexia, bulimia, binge eating disorder or other specified eating or feeding disorders).Past 3-month alcohol use disorder or substance use disorder (per the DIAMOND)Current shift work or responsibilities, such as providing care, that would chronically disrupt sleep (i.e., > 3 h between 22:00 and 05:00 h for at least 1 day/week) > 5 kg weight change in the past 3 monthsConditions that would interfere with ability to take part in TRE, including type I diabetes, pregnancy, breastfeeding, uncorrected hypo- or hyperthyroidism, or gastrointestinal conditions impairing nutrient absorption.Medical conditions that could confound immune or other study measures, such as HIV, AIDS, multiple sclerosis or lupusMedications contraindicated for fasting; clozapine; glucose-lowering medications; diabetes or obesity-related injections; medications requiring food early in the morning or late in the evening, corticosteroids; GLP-1 agonists (such as Semaglutide) will not be an exclusion criteria if weight is stabilizedCognitive deficits as noted during the initial interview or as indicated by low performance on the Brief version of the Orientation Memory Concentration Test (Weighted, reverse score < 20)Failure to complete screening and baseline questionnaires adequately (see strategies to maximize data quality procedures below).Failure to complete 7 days of dietary logs adequately (e.g., at least 2 entries per day, covering at least a 5-h eating window) during the baseline period.

Recruitment and screening procedures will be parallel with those reported previously [69]. Briefly, participants will be recruited from community and clinic centers, and we will advertise to oversample those in the early stage of their illness (within 5 years of onset). Screening will begin with completion of online informed consent, and then an online screening survey to assess medical and demographic criteria, and to complete self-rated psychiatric screeners for eating disorder symptoms, alcohol, substance and psychosis symptoms. Those who meet criteria within the screening survey will be scheduled for a remote interview consisting of the DIAMOND interview to assess psychiatric diagnoses, and the Structured Clinical Interview for Sleep Disorders-Revised (SCISD-R) [82] to assess sleep disorders (insomnia, hypersomnolence disorder) and circadian sleep–wake disorders (irregular sleep–wake disorder and non-24-h sleep–wake type disorder). Participants will be paid $25 USD per hour for baseline and follow-up assessments but not be paid for other components of the study (screening, coaching and time spent implementing the dietary plan).

As described elsewhere [69], we will use multiple criteria to assess potentially careless responses. These include strategies to detect and respond to inattentive responding, and data cleaning techniques.

As detailed elsewhere [69], risk management procedures are based on online intervention and psychoeducation websites for BD [98, 99]. These will include obtaining a release to contact a primary provider, careful training of all interviewers and coaches in how to assess and respond to high-risk symptoms and suicidality, routine assessments of possible relapse and emergent symptoms, and guidelines for providing sensitive referrals based on the urgency of the situation.

We will be careful to inform participants that our study does not involve dieting and that we expect them to sustain their caloric intake throughout the time following the TRE food plan. TRE will be administered as self-help via online instruction supplemented with optional coaching sessions and participant logging as previously described [69].

To minimize attrition and nonadherence, the food plan period is limited to 8 weeks. Participants are welcome to continue to access food plan materials and follow the food plan after the assigned 8 weeks. Individuals with personal and family experience with BD reviewed intervention content and their suggestions about the usefulness and modification of the material were incorporated [100, 101].

Because previous research has shown that interactive support improves adherence to web interventions [102], participants will be provided with opportunities to schedule 15-min online coaching sessions or to email questions about implementing TRE. We will encourage an initial coaching session. At one, three, five, and seven weeks into the TRE period, they will receive a link offering a chance to contact us or a link to a calendar to schedule a coaching appointment. Participants will be instructed that the scope of this coaching is narrow and will only focus on TRE. Participants will be invited to bring significant others, family members or household members to the coaching sessions if they’d like. Participants can send as many messages as they like using a contact form but will receive only one response per week. Coaching sessions will also be limited to no more than once a week.

Coaches will receive guidelines and will meet with licensed clinical psychologists regularly throughout the intervention for support and to ensure that their coaching maintains fidelity to TRE and the dietary changes. Coaches will aim to provide a supportive, nonjudgmental environment where participants can discuss barriers and difficulties in implementing dietary plans and can receive behavioral tips. The focus of coaching will be on collaborative problem solving and empowerment in addressing barriers. Coaches will aim to help participants lessen self-criticism, to avoid overly ambitious or restrictive dietary changes, and to be sensitive to mood fluctuations that may make following this type of program difficult. Coaches will aim to help participants identify material in blogs, video clips, or other barriers materials that might help address the concerns raised.

Participants will have access to a website of information designed to address barriers to following their food plan (e.g., coordinating timing of food with family meals or holidays), improving sleep, or addressing light exposure. Barriers were identified by reviewing previous studies using TRE, by conducting a survey of individuals with BD who had tried TRE, and by reviewing concerns raised by those with lived experience of BD and/or following dietary programs. For each major barrier, brief tips are available, which vary in format (written, audio, or video); all tips are designed to take 2 min or less. To help participants access relevant materials, they will be sent a quick survey at the end of weeks 3, 5, 7, and 9 in the intervention period or post baseline asking if they had experienced any barriers and providing the option to view tips relevant to the barriers endorsed.

To support intent-to-treat analyses, we will attempt to gather final outcome assessments for all participants who discontinue the study (except those who explicitly discontinued or were withdrawn on ethical grounds. This final outcome assessment will include an intervention acceptability survey, self-rated mania, depression, and QoL; and, if participants are willing, the YMRS, MADRS, and the LIFE. The trial coordinator will attempt to contact participants who withdraw by phone to ask the main reasons for discontinuation. Should adverse events be discovered, those will be reported.

Before conducting analyses of hypotheses, we will examine whether values appear to be missing at random. If this missing value imputation assumption is supported, we will impute data using 20 resamples, using demographic (sex/gender, age), baseline symptom severity, and symptom history variables (number of episodes) to predict missing scores. For comparison, we will perform analyses without imputed values. Data cleaning will be completed as described above in data quality. Statistical analyses will be conducted in accordance with the International Conference on Harmonization E9 statistical principles.

Before conducting analyses of hypotheses, we will carefully examine potential confounds. We will avoid adding any covariates that were entangled with our core mechanisms (e.g., sleep and circadian processes). With this in mind, we will only consider covariates for inclusion in models if they could be confounding our main effects– that is, those that are significantly correlated with both our IV (treatment adherence) and DVs (symptoms and QoL). Power analyses were based on inclusion of two covariates, and we would hope not to exceed this number of covariates. If more than two potential covariates are identified that appear significantly and uniquely correlated with both the IV and DVs, we would redo our power analyses, but at this point, we would be able to consider the covariance matrices, allowing for more careful estimation of true statistical power. We would consider these issues without generating the models of how our treatment variables actually influence symptoms and QoL, so as to avoid p-hacking or “fishing”. Only if those power analyses suggested that we could include more than 2 variables as confounds would we proceed to add more covariates; otherwise, we would plan to choose the two covariates that appear to hold the highest unique correlations with IV and DVs.

Analyses will be conducted using two-tailed tests with alpha set to .05.

Our core hypotheses will be tested using two parallel structural equation models (SEM) to consider mania and depression separately. In each, we will examine paths from TRE adherence levels to a latent variable for clock gene amplitude (based on 3 genes) to symptoms to QoL, controlling expected covariates and evaluating the potential confound of glucose tolerance mediation.

We have planned analyses based on availability of data from 125 participants, allowing for attrition of 25 participants. We assume moderate effects for diurnal amplitude on symptoms and QoL (f2 = .25) because smaller effects are inconsistent with the circadian rhythm theory. Assuming f2 = .40 (symptoms to QoL), power to detect the indirect effect of change in amplitude of gene expression on QoL via symptom change is .81, and for full model (RMSEA) > .90 [130, 131]. Intervention-related effects will be estimated using an intention-to-treat approach, in which we will examine final scores (where available) for all participants.