The Theory of Metabolic Jet Lag in Bipolar Disorder: A Pilot Feasibility Study Exploring the Association Between Eating Rhythm Disruption and Illness Burden Parameters

ABSTRACT

Introduction

Changes in energy metabolism are commonly observed in bipolar disorder (BD) and have been linked to a more severe clinical course. ‘Metabolic jet lag’ (MJL) is a state of shift in circadian patterns of energy homeostasis, expressed through behavioral changes such as irregular meal timing. However, the relationship between MJL and the pathophysiology of BD remains unknown. The objective of this study is to determine the feasibility of an investigation assessing the association between MJL, evaluated through eating rhythm disruption, and markers of illness burden in ten individuals with DSM‐5‐defined BD type 1.

Patients and Methods

This is a 14‐day longitudinal, naturalistic study that used the smartphone application ‘RxFood’ to elaborate a personalized, time‐stamped “feedogram” for each participant. Illness burden and severity parameters were collected at baseline, and the Acceptability E‐Scale was administered at the endpoint. Eating rhythm disruption scores were calculated based on a novel analytical approach applied to the “feedogram” data.

Results

The use of ‘RxFood’ demonstrated excellent feasibility and acceptability (94%) in individuals with BD and captured 628 eating occasions over the study period. Eating rhythm disruption scores varied between participants and correlated with the number of manic episodes (r = 0.718, p = 0.019), although the association was not significant after correcting for multiple comparisons.

Discussion

Ecological momentary assessment of eating rhythms is feasible in individuals with BD. Larger investigations evaluating the association between eating rhythm disruption and clinical illness trajectories are needed to delineate the role of MJL in the pathophysiology of BD.

1. Introduction

Bipolar disorder (BD) is a chronic, complex, and prevalent mental illness, characterized by dysfunctions in affective, cognitive, and behavioral domains [1, 2]. For most individuals with BD, currently available treatments are limited in providing remission, and the investigation of new targets for intervention is required [3, 4, 5]. Recently, it has been postulated that circadian rhythm dysfunction is an important marker of unfavorable illness trajectory, including a severe and progressive course of BD [1, 6, 7]. Most individuals with BD exhibit different degrees of circadian desynchronization at different moments of their illness trajectory, [1, 8, 9] and current evidence suggests a bidirectional relationship between circadian disruption and mood symptoms [6, 10].

Previous work by our group described the theory of ‘metabolic jet lag’ (MJL), which refers to a dramatic shift in circadian patterns of energy homeostasis, including hormone release, adipose tissue function, and immune function, and is expressed through behavioral changes such as irregularities in eating behavior [11, 12]. As an indicator of MJL, irregular eating rhythms are associated with a higher risk of metabolic abnormalities such as metabolic syndrome, diabetes mellitus, and some forms of cancer [11, 13]. There is some evidence to suggest that MJL contributes to the underlying pathophysiology of BD and may represent a target for therapeutic intervention [9, 14, 15, 16, 17, 18]. However, the association between eating rhythm dysfunction and the clinical expression of illness burden and trajectory phenotypes in BD remains unknown.

Few studies have investigated eating rhythms in psychiatric populations other than anorexia and bulimia, and methodological limitations often hinder the external validity of the results. For example, preliminary investigations using simple questionnaires suggest that eating rhythm dysfunction is present in BD during acute episodes and euthymia, and is associated with illness severity and poor quality of life [19, 20]. However, the cross‐sectional, retrospective, and manual nature of the assessment of eating rhythms limits its ecological validity. Conversely, there is emerging interest in ecological momentary assessment (EMA), an active data collection approach to measure behavior in a participant's normal environment, typically via smartphones [21, 22]. EMA requires relatively little effort on behalf of the participant and provides naturalistic data with high temporal resolution, overcoming limitations of traditional eating rhythm assessments [23, 24, 25].

To our knowledge, there is no investigation that uses an EMA approach to assess MJL in individuals with BD. Therefore, the objective of this longitudinal cohort study is to determine the feasibility of an EMA approach to evaluate the association between eating rhythm disruption and markers of illness burden in BD. This study presents a novel approach to the analysis of eating rhythms and facilitates further discovery regarding the theory of MJL in BD.

2. Methodology

The study protocol was approved by the Queen's University Institutional Review Board, and all participants provided written informed consent prior to enrollment in the study. A 14‐day longitudinal cohort study was conducted to evaluate the feasibility of an investigation of the association between MJL and illness burden parameters in individuals with BD. Using an EMA approach, a personalized “feedogram” was elaborated using the smartphone application ‘RxFood,’ which collected eating rhythm data based on pictures that participants took of all meals ingested over the study period. Baseline, 1‐week, and end‐point assessments were all carried out virtually, allowing the study to be conducted during the COVID‐19 pandemic.

2.1. Participants

Male and female individuals with DSM‐5‐defined BD type 1 (n = 10) were enrolled in the feasibility study. The inclusion criteria for all participants included the following: (1) age between 18 and 60 years old; (2) being in the inter‐episodic period, defined as not filling criteria for depressive, hypomanic, or manic criteria; (3) low severity of manic symptoms, defined as a Young‐Mania Rating Scale (YMRS) below 8, and low to moderate severity of Montgomery Asberg Depressive Rating Scale (MADRS) below 34; (4) being clinically followed by a psychiatrist or a family doctor; (5) not being acutely suicidal, operationalized by the response to the item of suicidal ideation at MADRS; (6) no changes in medication in the past 4 weeks and no plans of changing medication in the next month; (7) absence of comorbidities with substance use disorders other than tobacco and caffeine; (8) fluency in English.

The exclusion criteria for all participants included the following: (1) past or present diagnosis of any eating disorder; (2) patient currently trying to lose weight; (3) adoption of special diets (e.g., vegan, carnivore, paleo, ketogenic diet); (4) currently practicing intermittent fasting of any type of time‐restricted eating; (5) fasting or planning to fast for any reason in the next 4 weeks; (6) presence of general medical comorbidities with potential impact in appetite (e.g., hypothyroidism; Cushing syndrome; liver or kidney diseases; HIV/AIDS; cancer); (7) documented history of pregnancy or postpartum depression; (8) use of medication with the aim of losing weight (e.g., GLP1‐agonists; bupropion and naltrexone).

2.2. Baseline Clinical Assessments

To confirm the diagnosis of BD type 1 and to assess psychiatric comorbidities, the MINI International Neuropsychiatric Interview (MINI) was used. The Young Mania Rating Scale (YMRS) and Montgomery‐Asberg Depressive Rating Scale (MADRS) were used to evaluate the severity of manic and depressive symptoms at baseline, respectively. Illness burden parameters included: (1) age of onset; (2) duration of illness; (3) number of manic, hypomanic, and depressive episodes; (4) number of hospital admissions; (5) history and number of suicide attempts; (6) functional status. As a potential covariate, quality of diet was assessed using the Rapid Eating Assessment for Patients‐Shortened Version (REAP‐S) [26]. The REAP‐S was previously used to assess the quality of diet in individuals with BD, which was shown to be negatively correlated to circadian sleep disturbance and evening chronotype [27]. In addition, anthropometric variables were collected at baseline, 1‐week, and endpoint meetings, including measurements of weight, height, and waist circumference.

2.3. Ecological Momentary Assessment of Eating Rhythms

As an indirect parameter of MJL, eating rhythm disruption was evaluated both clinically and through EMA. The clinical evaluation of eating rhythm disruptions was conducted using the Biological Rhythm Interview for Assessment in Neuropsychiatry (BRIAN). This is a self‐report instrument that evaluates rhythm disturbance in sleep, activity, social, and eating patterns [28]. It was already used in the Canadian context and in individuals with BD [28, 29]. The ‘RxFood’ app was incorporated to create a history of all the foods and liquids ingested by the participants for a period of 14 days, including the time of meal consumption. Eating occasions were defined as an eating event in time, regardless of the caloric amount that was consumed. This approach complies with the recommendations of Leech et al. [30] and the methodology employed by Buyukkurt et al. [19]. This personalized feedogram assessed eating rhythms (i.e., the timing of meals over the day‐night cycle), which is one parameter to assess the severity of MJL in each participant.

2.4. Procedures

The project was advertised on the NeuroMood Lab social media and the Department of Psychiatry website through a recruitment poster. Advertisements were also posted on Kijiji, Facebook, and Reddit. The volunteers were instructed to contact the research team using a safe email to have a screening phone call scheduled. The phone call had the objective to check eligibility and to provide an explanation about the objective of the study, procedures, potential risks, and benefits for the research. After confirming eligibility, participants completed the letter of information/consent and sociodemographic forms through a protected OneDrive link before the virtual baseline interview was scheduled via Zoom.

The baseline interview was conducted by a psychiatrist from the research team. It involved the collection of sociodemographic data, illness trajectory data, and the application of MINI. The baseline assessment also included the assessment of the severity of manic and depressive symptoms via YMRS and MADRS, respectively; BRIAN, REAP‐S, and anthropometric variables. In addition, the participants downloaded the ‘RxFood’ app during the baseline interview and were instructed on how to use it as a meal tracking tool. Each participant underwent a simulation with the app in the presence of the research team to ensure that they could successfully carry out meal tracking. If a participant failed to demonstrate competency with meal tracking using ‘RxFood’ after three simulations, they were excluded from the trial. Finally, the virtual 1‐week and end‐point assessments were scheduled at the end of the baseline interview.

During the 1‐week and end‐point assessments, the individuals were instructed to measure their weight and waist circumference in the virtual presence of a member of the research team. During these assessments, participants had the opportunity to provide qualitative feedback on their experience using ‘RxFood.’ At the end‐point assessments, all participants completed a 5‐item Acceptability E‐Scale to determine the acceptability of the ‘RxFood’ app. During the whole duration of the study, the participants received a daily text message to encourage and reinforce adherence to meal self‐tracking. No instructions were given regarding nutrition, rhythms, place, or types of foods to be ingested. The participants were instructed to continue their usual diet and eating habits during the period of study. Figure 1 depicts a summary of the research timeline.

FIGURE 1.

The timeline for the pilot trial investigating the feasibility of a 14‐day EMA study evaluating eating rhythm disruption in individuals with bipolar disorder. Abbreviations: LOIC, Letter of Information and Informed Consent; SF, Sociodemographic Form; MINI, MINI International Neuropsychiatric Interview; YMRS, Young‐ Mania Rating Scale; MADRS, Montgomery Asberg Depression Rating Scale; BRIAN, Biological Rhythm Interview for Assessment in Neuropsychiatry; REAP‐S, Rapid Eating Assessment for Patients‐Shortened Version; EMA, Ecological Momentary Assessment.

2.5. Statistical Analysis

All analyses were conducted in PyCharm [31] (Runtime version: 17.0.9 + 7‐b1087.7 amd64) using libraries Pandas and NumPy. Correlations were conducted using IBM SPSS Statistics version 26.0 [32].

As this is a feasibility study, descriptive statistics were used for clinical and demographic variables. Eating rhythmicity was determined based on the calculation of an eating rhythm disruption (ERD) score, derived from a combination of three subscores that each probe a circadian aspect of eating behavior: (1) meal regulatory score; (2) eating window score; (3) phase angle score. The mean and standard deviation (SD) of the meal time of each meal were calculated per patient across the study days. The SD representing the variability in meal timing for each meal across different days was calculated to determine the meal regularity score. Similarly, to derive the eating window score, the SD of the feeding and fasting window length in hours was calculated for each patient across the study period. The SD of the phase angle score was also computed for each participant from the time difference in hours between the first meal of the day and the sunrise, and the last meal of the day and the sunset. Sunrise and sunset times represent external zeitgebers for eating behavior and were calculated using the Sunrise/sunset calculator accessed from the National Research Council of Canada website [33]. These subscores contributed a multidimensional measure to the ERD score, which was calculated as follows:

$$\mathrm{ERD}=\left({\mathrm{SD}}_{\text{meal time}}+{\mathrm{SD}}_{\text{window}}+{\mathrm{SD}}_{\text{phase angle}}\right)$$

where SD_{meal time} represents the meal regularity score, SD_window represents the eating window score, and SD_{phase angle} represents the phase angle score. A larger ERD score reflects a higher degree of eating rhythm disruption, while a lower ERD score indicates less variation and a higher degree of regularity in meal times, eating windows, and phase angles.

To validate the ERD subscores as measures of eating rhythmicity, a cross‐correlation was conducted using Spearman's ρ. The correlation between ERD score and BRIAN score, illness burden parameters, MADRS, and YMRS scores was also calculated using Spearman's ρ. Point‐biserial correlation was used for the nominal scales with two levels (dichotomous variables). Holm's correction was applied to address multiple comparison problems Holm, [34]. This method compares each p‐value to its adjusted significance level, which is α/(m − i + 1), where α is the significance level (0.05); m is the total number of tests; and i is the rank of the p‐values when sorted in ascending order. Correlation between illness severity parameters and ERD, MADRS, YMRS, and acceptability was conducted by applying Kendall's tau‐b. A paired sample t‐test was conducted to determine if there is a significant change in anthropometric measures, including weight and waist circumference, from baseline to endpoint. The Shapiro–Wilk test was applied to assess the normality of the anthropometric measures [35].

3. Results

3.1. Sociodemographic Characteristics

The characteristics of the study sample are shown in Table 1. The majority of the participants identified as female (50%), were in the age category between 20 and 29 years of age, had full‐time employment, obtained at least a Bachelor's‐level education, and had an annual salary of $50,000 and over. All the participants identified as Caucasian.

TABLE 1.

Sociodemographic characteristics of the study sample.

Characteristic	n
Gender	5
Female	3
Male	2
Non‐binary	10
Total
Age	4
20–29	3
30–39	2
40–49	1
50–59	10
Total
Employment	6
Full‐Time	2
Part‐Time	2
Student	10
Total
Education	2
High School Diploma	1
Community College Diploma	5
Bachelor's Degree	2
Master's Degree
Income	2
$15,000–$24,999	3
$25,000–$49,999	4
$50,000 and over	1
Prefer not to Answer	10
Total
Recruitment Method	3
Reddit	2
Lab Social Media	2
Online Newsletter	2
Poster at Doctor's Office	1
Participated in a Previous Study	10
Total
Ethnicity	10
Caucasian	10
Total

3.2. Baseline Assessments

Illness burden and severity parameters, anthropometric variables, medication use, comorbidities, biological rhythm disturbance, and diet quality data collected at baseline are shown in Table 2. Participants had mean (SD) depressive and manic symptom scores of 16 (9.21) and 4.8 (3.19) according to the MADRS and YMRS, respectively. Illness burden parameters displayed substantial variability between participants, the average duration of illness being 13.4 (12.2) years, and the number of manic, hypomanic, and depressive episodes being 13.2 (18.92), 17.9 (31.44), and 15.3 (13.27), respectively. Participants experienced 5.1 (9.72) hospital admissions on average, with 50% of participants experiencing a history of suicidal ideation and/or attempt. At baseline, 80% of participants indicated mild symptom severity, while 10% experienced moderate and severe symptom severity. Mild functional impairment was observed in 70% of the participants, with the remainder experiencing moderate functional impairment at baseline.

TABLE 2.

Baseline assessment results.

Characteristic	Mean (SD)
MADRS	16 (9.21)
YMRS	4.8 (3.19)
Illness Burden	13.4 (12.2)
Duration of Illness (years)	13.2 (18.92)
Number of Manic Episodes	17.9 (31.44)
Number of Hypomanic Episodes	15.3 (13.27)
Number of Depressive Episodes	5.1 (9.72)
Number of Hospital Admissions	5 (50) a
History of Suicide	4.3 (6.65)
Number of Suicide Attempts
Illness Severity	8 (80) a
Symptom Severity	1 (10) a
Mild	1 (10) a
Moderate	7 (70) a
Severe	3 (30) a
Functional Impairment	0 (0) a
Mild
Moderate
Severe
BRIAN Score	45.4 (11.82)
Sleep	12.6 (4.12)
Activity	12.5 (4.17)
Social	9.6 (2.76)
Eating	10.7 (4.08)
REAP‐S Score	26.1 (5.09)
Anthropometrics	—
BMI	3 (30) a
< 18.5	3 (30) a
18.5–24.9	4 (40) a
25–29.9	92.1 (17.52)
≥ 30
Waist Circumference (cm)
Psychiatric Medication Class	3 (30) a
Anticonvulsant	1 (10) a
Anxiolytic	3 (30) a
Atypical antipsychotic	1 (10) a
Mood stabilizer	1 (10) a
Stimulant
Comorbidities	3 (30) a
GAD	1 (10) a
ADHD	4 (40) a
OCD	1 (10) a
Panic Disorder	1 (10) a
ASD	1 (10) a
Substance Use Disorder
Routine Substance Use	1 (10) a
Alcohol	6 (60) a
Cannabis	—
Nicotine

Abbreviations: ADHD, Attention‐Deficit/Hyperactivity Disorder; ASD, Autism Spectrum Disorder; BMI, Body Mass Index; GAD, Generalized Anxiety Disorder; MADRS, Montgomery‐Asberg Depression Rating Scale; OCD, Obsessive‐Compulsive Disorder; REAP‐S, Rapid Eating Assessment for Patients‐Shortened Version; YMRS, Young‐Mania Rating Scale.

^a Values presented as count (percentage) for categorical variables.

Total biological rhythm disturbance scores were 45.4 (11.82) on average, with participants scoring 10.7 (4.08) in the eating rhythm category of the BRIAN. Mean diet quality scores in the cohort were 26.1 (5.09) according to the REAP‐S. Regarding anthropometric data, the majority of participants were in the obese BMI category, with 30% being overweight and 30% being ‘normally’ weighted. All participants were taking psychotropic medication, most commonly anticonvulsants (30%) and atypical antipsychotics (30%). Psychiatric comorbidities included generalized anxiety disorder (30%), obsessive‐compulsive disorder (40%), attention‐deficit hyperactivity disorder (10%), panic disorder (10%), autism spectrum disorder (10%), and addiction (10%). 60% and 10% of participants in the cohort reported regular use of cannabis and alcohol, respectively, characterized by use at least once per week.

3.3. Feasibility and Acceptability

The use of the ‘RxFood’ mobile application demonstrated excellent acceptability according to the results of the Acceptability E‐Scale (Figure 2). There were no dropouts at any point between the baseline and end‐point assessment, and only one participant missed one day of meal tracking. The mean acceptability of ‘RxFood’ was 28.2 on the 30‐point scale (94%), with a standard deviation of 3.26, indicating low variability in acceptability between participants. Five participants (50%) expressed interest in continuing to use the ‘RxFood’ application following completion of the study period; one participant continued tracking voluntarily for over 48 days after the study period had ended.

FIGURE 2.

‘RxFood’ demonstrated a high degree of acceptability according to the 5‐point Acceptability E‐Scale in individuals with bipolar disorder.

Qualitative data support the feasibility of ‘RxFood,’ with participants expressing ease and enjoyment of use during the 1‐week and end‐point assessments. Most participants expressed their experience with ‘RxFood’ as positive, for example: “Meal tracking is easy and actually very satisfying;” “I really enjoy tracking and was surprised at the AI's capability to identify certain foods, vegan products, for example. The nutrition tab didn't always work for me, but other than that, it was smooth;” “Meal tracking was illuminating and showed me a lot about my calorie and nutrient content;” “I enjoyed the report the most, which gave me a breakdown of fruits, vegetables, etc. There was some awkwardness with the app interpreting something wrong, and it wasn't super intuitive about how to go in and edit, but entering food in the first place was super easy, and I only had to go in and fix what the AI picked up 1 or 2 times;” “I really enjoyed using the app and would like to continue past the study period.”

According to the Shapiro–Wilk test statistic, weight and waist circumference did not significantly deviate from normality (test statistic = 0.846, p = 0.053; test statistic = 0.849, p = 0.056, respectively); therefore, the parametric t‐test was used for both. The result of the paired t‐test demonstrated that there was no statistically significant difference in weight (t₍₉₎ = 1.673, p = 0.129) or waist circumference (t₍₉₎ = 1.00, p = 0.343) from baseline to endpoint. This suggests that meal tracking did not have a significant effect on anthropometric parameters in the participants of the trial.

3.4. Eating Rhythm Disruption

A total of 628 eating occasions were captured across the pilot trial. The timing of meals for each participant is shown in Figure 3; individuals demonstrated vastly different meal times and eating windows across the study period.

FIGURE 3.

Eating rhythm data were collected with ‘RxFood;’ each green dot represents an eating occasion across the 14‐day study period.

Table 3 shows the mean and standard deviation of the time for each meal number across the study period for each participant and the variance (σ2) associated with each meal. The number of meals consumed per day ranged from 3 to 10, and the average time that each meal was consumed varied among participants. Among the recorded meals, the highest variance in time occurred for the second (σ2 = 9.48) and third meals (σ2 = 9.14) of the day. Figure 4 depicts the average meal times of each meal number per participant.

TABLE 3.

The mean and standard deviation (in parentheses) of meal times for different meal numbers across the study period, per participant, and the variance of each meal number.

PID	Meal Number
	1	2	3	4	5	6	7	8	9	10
1	9.92 (1.72)	13.29 (2.72)	13.89 (0.86)	18.14 (1.77)	—	—	—	—	—	—
2	10.11 (2.82)	18.33 (4.14)	20.58 (4.22)	—	—	—	—	—	—	—
3	5.46 (3.66)	8.38 (4.77)	12.44 (4.56)	15.68 (2.56)	17.67 (2.76)	18.85 (1.48)	19.56 (1.29)	20.8 (2.15)	21.27 (1.57)	—
4	8.29 (1.07)	11.65 (1.43)	15.93 (2.62)	17.53 (2.76)	18.59 (1.43)	21.36 (2.71)	—	—	—	—
5	10.45 (3.13)	16.41 (3.71)	18.32 (5.06)	—	—	—	—	—	—	—
6	9.31 (2.17)	13.64 (2.14)	18.91 (2.1)	22.25 (2.23)	21.44 (1.96)	23.01 (0.08)	—	—	—	—
7	10.19 (1.15)	14.3 (2.86)	16.5 (2.96)	19.89 (2.96)	18.53 (2.33)	19.81 (0.81)	21.76 (1.65)	—	—	—
8	9.37 (3.12)	11.7 (2.13)	14.45 (2.81)	16.58 (3.53)	17.33 (2.93)	18.13 (2.46)	18.77 (3.46)	19.55 (5.01)	—	—
9	4.79 (3.22)	7.76 (2.88)	10.06 (3.34)	12.79 (3.84)	15.76 (3.96)	17.42 (2.97)	19.09 (2.4)	20.56 (2.41)	22.46 (1.06)	23.02 (0.47)
10	7.21 (2.4)	13.61 (4.21)	17.31 (3.77)	19.09 (3.06)	19.3 (1.61)	20.92 (1.26)
Variance (σ2)	3.74	9.48	9.14	7.14	2.68	3.32	1.37	0.29	0.35	0

FIGURE 4.

The average time of day that each meal number was consumed across the 14‐day study period by each participant.

The scores for meal regularity, eating window, and phase angle also varied between participants, as shown in Table 4. The degree of eating rhythm disruption varied in the study cohort (σ2 = 4.01). For example, participant 7 demonstrated the least eating rhythm disruption (ERD = 4.73) while participant 5 demonstrated the most disrupted eating rhythm (ERD = 11.03).

TABLE 4.

Meal regularity score, eating window score, phase angle score, and total eating rhythm disruption score per participant, and the variance of each parameter.

Participant	Eating Window Score	Meal Regularity Score	Phase Angle Score	Total ERD Score
1	3.71	1.77	1.80	7.28
2	4.78	3.73	2.43	10.94
3	3.89	2.76	1.98	8.63
4	2.44	2	1.24	5.68
5	4.71	3.97	2.35	11.03
6	2.58	1.78	1.33	5.69
7	1.74	2.1	0.89	4.73
8	3.28	3.18	1.64	8.10
9	3.79	2.65	1.90	8.34
10	3.63	2.72	1.81	8.17
Variance (σ2)	0.85	0.55	0.21	4.01

To validate the ERD subscores as measures of eating rhythmicity, a cross‐correlation was conducted (Figure 5). A significant correlation was identified for eating window scores and meal regularity scores (r = 0.636, p = 0.048), eating window score and phase angle score (r = 0.988, p < 0.001), as well as meal regularity score and phase angle score (r = 0.697, p = 0.025).

FIGURE 5.

Cross‐correlation of the eating rhythm disruption subscores for validation of indices: (A) the relationship between eating window score and meal regularity score; (B) the relationship between eating window score and phase angle score; and (C) the relationship between meal regularity score and phase angle score.

3.5. The Association Between Eating Rhythm Disruption and Clinical Parameters

The correlations between ERD scores, acceptability, and clinical parameters are shown in Table 5. The only significant correlation occurred between the number of manic episodes and the ERD score (r = 0.718, p = 0.019), suggesting that a higher number of manic episodes is associated with increased eating rhythm disruption; however, this correlation was not significant after correcting for multiple comparisons.

TABLE 5.

Correlations between eating rhythm disruption score, acceptability, illness burden parameters, Biological Rhythm Interview of Assessment in Neuropsychiatry score, and illness severity.

	ERD Score		Acceptability
	r	p	r	p
Duration of Illness	−0.249	0.487	0.164	0.650
Number of Manic Episodes	0.718	0.019*	−0.133	0.714
Number of Hypomanic Episodes	0.058	0.873	0.119	0.744
Number of Depressive Episodes	0.239	0.507	0.119	0.743
Number of Admissions	0.232	0.519	0.088	0.809
History of Suicide Attempts	0.063	0.864	0.065	0.859
Number of Suicide Attempts	0.187	0.604	0.189	0.601
BRIAN Total	0.043	0.907	0.585	0.075
BRIAN Eating Subscore	0.360	0.307	0.587	0.074
MADRS	0.309	0.385	0.059	0.871
YMRS	0.210	0.561	−0.107	0.769
Symptom Severity	0.181	0.517	0.129	0.674
Functional Impairment	0.358	0.210	−0.270	0.388

Note: *The level of significance is 0.05, and significant p‐values are marked with an asterisk.

Abbreviations: BRIAN, Biological Rhythm Interview and Assessment in Neuropsychiatry; MADRS, Montgomery‐Asberg Depression Rating Scale; YMRS, Young‐Mania Rating Scale.

4. Discussion

This pilot trial was conducted to determine the feasibility of employing an EMA approach in an investigation of the association between eating rhythm disruption and clinical parameters in BD. Together, the pilot trial demonstrated a high degree of acceptability and feasibility in the use of ‘RxFood’ in individuals with BD. This is in line with other EMA investigations, which similarly demonstrate a high degree of feasibility of mobile monitoring in individuals with BD, including those with severe clinical profiles [36, 37, 38, 39]. The current trial employed a novel approach to the assessment of eating rhythm data to determine the degree of MJL in individuals with BD, an approach that can be applied to larger investigations on the topic. Feedogram visualization allowed for a comparison of eating rhythmicity patterns within and between each participant. Further, the cross‐correlation of eating rhythm disruption subscores supports the use of the ERD score as a valid measure of eating rhythm disruption.

The ERD scores varied between participants, suggesting different degrees of MJL in the study cohort. In other work, individuals with BD have been shown to display abnormal appetite and eating rhythms [40, 41], which were more severe when compared to healthy controls, even in pharmacologically naive cases (p < 0.001) [15]. A meta‐analysis of 30 case–control studies found a higher degree of behavioral rhythm disruption (ES: 0.78–1.12), one sub‐score of which was eating rhythmicity, in individuals with BD when compared to healthy controls [42]. The identification of genetic risk factors that are associated with eating rhythmicity in BD further supports the potential role of MJL in the pathophysiology, especially when considering the well‐documented heritability rate of the disorder [43, 44]. In the current investigation, the identification of eating rhythm disruption based on naturalistic evidence builds on the existing body of evidence that is primarily based on retrospective questionnaire approaches.

When considering the clinical relevance of MJL, the only significant association between eating rhythm disruption and clinical correlates occurred with the number of manic episodes. Although the association was not significant after correcting for multiple comparisons, this trend has been observed in other work. For example, in a study with a similar design, Buyukkurt et al. [19], demonstrated that eating regularity correlated with hypomanic symptom severity in the preceding week [19]. The number of past manic or hypomanic episodes has also been associated with impaired glucose metabolism and circadian rhythmicity in individuals with BD, as well as circadian gene mutations in preclinical studies [18, 45, 46, 47, 48].

Disrupted mechanisms involved in circadian and metabolic function may precipitate changes in the neurobiology responsible for mania, supported by the higher prevalence of circadian abnormalities in individuals with BD when compared to major depressive disorder [49, 50, 51]. However, the reverse may also be true. Neuroprogression is related to the cumulative effects of a range of dysfunctional processes contributing to allostatic load, and it is possible that disrupted eating rhythms are a contributor to and a consequence of these phenomena in BD [52, 53]. In this way, the relationship between MJL and illness burden parameters is important to consider in BD care, including the role of eating irregularity in the development of metabolic complications such as obesity.

4.1. Limitations

There are several limitations that temper the results of the current investigation. First, the sample lacked sociodemographic diversity and was not representative of all individuals with BD. This is especially true when considering the impact of cultural factors on the distribution of eating occasions across the day‐night cycle [54, 55]. Further, most participants had a relatively high socioeconomic status and education level; therefore, the feasibility of the ‘RxFood’ application may differ for individuals who do not fall within these groups. The same is true for the clinical status of the participants; the results cannot suggest that the study design will be feasible in patients with a more severe clinical profile. The relatively short length of the mobile monitoring period limits the ability to confer feasibility and potential effects of smartphone EMA approaches in a longer study duration of 3 to 6 months, for example.

The study design and calculation of the ERD scores did not include a consideration for behavioral rhythms such as sleep–wake, rest‐activity, or social rhythms, which inevitably influence meal timing and circadian desynchronization. Similarly, several factors may have influenced eating rhythmicity throughout the study period, including medical comorbidities, routine substance use, and medication, all of which varied across the study cohort. The current study did not measure other parameters of MJL, such as metabolic markers and mitochondrial function, which would have provided additional insight into the degree of MJL. Further, no data was collected on the incidence of significant life events that may have occurred and influenced meal timing during the study period, beyond the brief self‐report assessments at 1‐week and endpoint. To our knowledge, no other study has applied the calculation of a similar ERD score in this clinical population, limiting the ability to make a comparison to the literature.

There are limitations that hinder the ability to confer a relationship between eating rhythm disruption and the pathophysiology of BD. For example, no healthy control group was included, limiting the contribution to the body of evidence indicating abnormal eating rhythmicity in BD when compared to healthy controls. As a pilot feasibility study, the association between illness burden parameters and eating rhythm disruption is not highly powered, limiting the ability to make a definitive conclusion about the relationship. Further, no data were collected on illness burden or severity parameters during or after the pilot study, limiting the ability to determine the association between eating rhythm disruption and syndromal symptoms and the potential role of eating rhythm disruption as a predictor of illness trajectory or relapse. Finally, the pilot trial did not probe the chronicity of onset between eating rhythm disruption and clinical parameters due to the length of the study period and the lack of prospective follow‐up assessment. In this way, the study design cannot suggest whether the association between eating rhythm disruption and clinical correlates reflects a trait or state relationship.

4.2. Future Directions

Building on this work, the association between eating rhythm disruption and illness burden parameters will be more robustly evaluated in a larger, more generalizable participant population that confirms the feasibility of EMA‐based assessments in a longer study duration. To better delineate the pathophysiological role of MJL in BD, more evidence is needed about the trait or state nature of the association between eating rhythm disruption and illness burden parameters, and the potential implications as a predictor of BD onset, treatment response, and relapse. It remains to be determined if there are one or several shared factors contributing to eating rhythm disruption and the psychopathology of BD. The assessment of other MJL parameters beyond eating rhythms, such as metabolic blood markers or other behavioral rhythms, will contribute to a more robust evaluation of the theory of MJL in BD. EMA approaches have already been applied to the investigation of these measures in BD, including social, activity, and sleep rhythms [56, 57, 58, 59].

There are implications of this work, not only for the theory of MJL in BD, but also for the advancement of eating behavior assessment in psychiatry. Digital phenotyping refers to the identification of smartphone‐derived markers of human behavior, including sleep duration, social activity, geolocation, and screen usage patterns. Meaningful patterns have been identified between digitally‐collected markers and clinical correlates, representing potential to characterize psychiatric subgroups; however, reproducibility is needed before these methods can be applied to clinical practice [60]. These real‐time and data‐driven approaches will enhance our understanding of the multidimensional factors contributing to BD and, in the future, may be applied as diagnostic and clinical monitoring tools for circadian‐related phenomena.

Upon establishing a relationship between eating rhythm disruption and the pathophysiology of BD, EMA may be useful in evaluating the feasibility and efficacy of interventions that target MJL in psychiatry, such as time‐restricted eating and interpersonal social rhythm therapy (Bhatnagar et al. [61]; Guerrero‐Vargas et al., 2021) [61, 62]. Considering the acceptability of ‘RxFood’ in individuals with BD and the high rate of smartphone ownership and use worldwide, the clinical potential of mobile applications also extends to the development of self‐management tools in psychiatry. For example, ecological momentary interventions have demonstrated potential to promote the self‐management of negative emotions, stress, substance use issues, and anxiety; however, more work is needed to reduce participant burden, improve compliance, and ensure that EMA‐based care is high quality [63, 64, 65, 66, 67].

5. Conclusion

In conclusion, the naturalistic assessment of eating rhythms via ‘RxFood’ is feasible in individuals with BD and provides meaningful information about MJL in this clinical population. The association between eating rhythm disruption and illness burden parameters requires further investigation; nonetheless, mobile EMA approaches continue to demonstrate significant promise in clinical and research settings, especially as psychiatry moves towards a precision medicine approach. This study forms the basis for larger investigations to refine the naturalistic assessment of eating rhythm disruption, increase the understanding of MJL in the pathophysiology of BD, and contribute to the development of novel therapeutic approaches that consider lifestyle in psychiatry.

Funding

This study was funded by a Department of Psychiatry Internal Faculty Grant, Queen's University School of Medicine. Elena Koning is supported by a Dean's Doctoral Award, Queen's University. Sareh Panjeh receives funding from the Fundação de Amparo à Pesquisa do Estado de São Paulo [Grant No. #2022/13933–4].

Conflicts of Interest

The authors declare no conflicts of interest.

Data Availability Statement

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