Unhealthy Diet Quality as a Clinical Phenotype in Bipolar Disorder: A Potential Treatment Target for Improved Mood and Metabolic Outcomes

Abstract

Objectives:

This study aimed to investigate unhealthy diet quality in bipolar disorder (BD) and its association with social, metabolic and clinical features in a cohort of well-characterized adult patients with BD.

Methods:

This study analyzed a subset of 737 adult participants from the Mayo Clinic Bipolar Disorder Biobank (MCBB) who completed the Rapid Eating Assessment for Participants – Shortened Version (REAP-S) diet quality questionnaire, in which scores <32 indicate an unhealthy diet quality (UDQ). Demographic and clinical variables were compared between individuals with unhealthy and healthy diet quality. Regression models were constructed to explore associations between UDQ and demographic variables, bipolar subtype, chronotype, medical comorbidities, psychotropics and differences in social determinants of health-related stressors.

Results:

Individuals with BD and UDQ (78.8 %) were significantly younger, more likely to be male, and to have a diagnosis of BD-I or schizoaffective bipolar subtype. Additionally, UDQ was significantly associated with evening chronotype, binge eating disorder, comorbid anxiety disorders, obesity, hypertension, and unemployment-related stress. A multivariable logistic regression model identified that a higher education level and older age at BD diagnosis were associated with lower odds of UDQ, while an evening chronotype and comorbid anxiety disorders increased the likelihood of UDQ.

Limitations:

The study design limits causality assessment and poses recall bias from self-reported dietary tools.

Conclusions:

These findings highlight the importance of investigating diet quality as a potential therapeutic target in the management of BD to improve mood outcomes and mitigate risk of metabolic disorders.

1. Introduction

Bipolar disorder (BD) is a complex mental illness, often manifesting in young adulthood, characterized by cyclical changes in mood and energy that result in behavioral disturbances, functional impairment and increased risk of death by suicide (1). Individuals with BD are disproportionately affected not only by psychiatric comorbidities but also by co-occurring medical disorders, including obesity, hypertension, insulin resistance, type 2 diabetes mellitus, dyslipidemia, and metabolic syndrome, which further increase the risk of cardiovascular disease and all-cause mortality (2–5). Although the exact pathophysiological pathways leading to the high comorbidity of cardiometabolic disorders in patients with BD are not fully understood, there is evidence supporting the conception of BD as a multi-systemic inflammatory disease. This perspective is reinforced by the observed chronic peripheral and cerebral inflammation that results in an increased cardiometabolic risk (6,7). As additional burdens, lifelong exposure to psychotropic medications with weight-gain potential, frequently used in the treatment of BD, barriers to access health care and unhealthy lifestyle further perpetuate comorbidity illness burden (8–10). The cumulative effect of this combination of factors results in an increased health and economic burden for individuals with BD, including more pronounced psychopathology, suicidality, and medical costs (11,12).

Within the unhealthy lifestyle behaviors observed in patients with BD, diet and eating behaviors are particularly concerning. Unhealthy dietary habits observed in this population include high intake of ultra-processed foods, sugary drinks, and saturated fats, as well as irregular meal patterns, emotional eating, and high rates of eating disorders, are particularly concerning (13,14). Given the significant impact of metabolic comorbidities on BD, there has been an increasing focus on exploring modifiable lifestyle factors that may contribute to their development. In this context, diet quality, a multidimensional concept that reflects an individual’s degree of adherence to a diet characterized by food variety and nutrient-dense choices, has emerged as a critical component, supported by emerging evidence suggesting a strong link between diet, metabolic health and mood and by the significant role of poor diet quality as a mortality risk factor and contributor to cardiovascular disease (15–17). In the absence of standardized definitions and assessment tools, several diet quality questionnaires, often reliant on patient self-report of food and beverage intake, have been developed to evaluate dietary patterns in clinical populations (18). For instance, the American Heart Association (AHA) has previously recommended the use of three dietary screeners to assess patients’ diet quality as a risk reduction strategy for cardiovascular disease (17). Among these, the Rapid Eating Assessment for Participants – Shortened Version (REAP-S) was developed as a self-report scale for clinical and research settings that allows a fast evaluation of diet content and intake in an average week (19,20) and has been used in cohorts with mood disorders (14,21,22).

The concept of unhealthy diet phenotype in mood disorders, previously explored in a narrative review by Koning et al. (2022), highlights the presence of dietary patterns composed of low protein intake, high intake of fats and calorically dense foods and a preference for refined carbohydrates in patients with BD (23). In a study comparing people with BD and healthy controls, the former consumed more carbohydrate and ultra-processed sweets (24). Along this line, we have previously reported the presence of higher rates of disordered eating and unhealthy diet quality (UDQ) among patients with BD, in addition to considerable difficulty avoiding unhealthy foods and worse mood outcomes (14,21), which may be the consequence of a combination of intrinsic and environmental factors, such as medication use, related to mood disorders. Overall, UDQ has been associated with an increased risk of psychiatric disorders, including persistence of depressive symptoms and altered treatment response, due in part to the pro-inflammatory potential of these foods (25–28). On the contrary, conceptually, what constitutes a healthy diet quality is the consumption of diverse food sources that provide multiple nutrients that sufficiently satisfy an individual’s nutritional needs (29). For instance, high quality dietary patterns, such as Mediterranean diets, have been previously associated with anti-inflammatory properties and improved mood outcomes in individuals with BD (30–32).

While the exact mechanisms that result in the maladaptive eating behaviors and UDQ frequently observed in mood disorders are not fully known, studies attempting to phenotype these eating behaviors have been conducted (23,33). Disruptions in interoceptive hunger/satiety signals, brain gamma-aminobutyric acid concentrations, and neuroendocrine circadian, and reward systems have been proposed as mechanisms involved in unhealthy dietary patterns in BD. These disturbances interact with other psychiatric symptoms and environmental factors ultimately contributing to suboptimal eating behaviors, food preferences and diet quality (23,34–36). Furthermore, there is an additional bidirectional relationship between dysregulated intrinsic mechanisms involved in the pathophysiology of BD, including inflammation, oxidative stress imbalances, impaired mitochondrial activity and disturbed gut microbiota, and feeding behaviors that are hypercaloric, high-carbohydrate and high-fat (32,37,38). These mechanisms, along with diet and sedentarism, are further impacted by comorbid obesity and cardiometabolic disturbances and the use of weight-liable psychotropic medications, establishing a vicious cycle where UDQ, metabolic risk and mood symptoms reciprocally amplify and reinforce each other (21,39,40). The attempts to develop effective and sustainable lifestyle interventions for BD are complicated by the intricate interaction of socioeconomic, cultural, and biological factors contributing to dietary patterns and eating behaviors.

The extent to which UDQ contributes to and is a consequence of BD is yet to be fully understood. Seeking to generate evidence that allows the development of therapeutic targets for non-pharmacologic interventions for BD, this study aimed to investigate unhealthy diet quality as a discrete clinical phenotype of BD and to explore associations with social and illness-related features in a large cohort of well-characterized patients with BD from the Mayo Clinic Bipolar Disorder Biobank (MCBB).

2. Methods

2.1. Study design and participants

This study represents a sub-analysis of 737 of the 2268 clinically phenotyped individuals living with BD who participated in the MCBB and completed the Rapid Eating Assessment for Participants – Shortened Version (REAP-S) at the time of enrollment (Supplementary Figure 1). The MCBB represents an international multisite collaboration led by Mayo Clinic (Rochester, MN, USA) to create a biorepository to investigate risk biomarkers of bipolar illness and pharmacogenomic markers of treatment response (41). Recruitment sites included Mayo Clinic, Lindner Center of HOPE, University of Minnesota, Universidad Autónoma de Nuevo León (UANL), and the Universidad de los Andes. Criteria for inclusion to the biobank included a diagnosis of types I or II BD or schizoaffective disorder bipolar subtype, confirmed with the Structured Clinical Interview for DSM-IV (SCID-IV), no current suicidal ideation or psychosis and ages 18 through 80. Each enrolled participant provided informed consent, completed questionnaires to obtain a detailed clinical phenotype, including basic anthropometric measurements (i.e., weight, height, body mass index [BMI]), and provided a blood sample for subsequent genomic analysis. Specific assessments conducted in the MCBB have been published elsewhere (41). The study protocol was approved by each participant site’s institutional review board (IRB).

In addition to the SCID-IV, clinical assessments included a Bipolar Biobank Clinical Questionnaire (BiB-CQ), a self-applied Bipolar Biobank Patient Questionnaire (BiB-PQ) for demographic characteristics, family history and substance use, a single self-reported item from the reduced Morning-Eveningness Questionnaire (rMEQ) to determine diurnal preference (chronotype), the Eating Disorder Diagnostic Scale (EDDS), a modified version of the Stanley Foundation Bipolar Network Life Stress Questionnaire (SFBN), and the REAP-S (20,41–44). To assess the presence of stress related to the five domains of social determinants of health (economic stability, education access and quality, health care access and quality, neighborhood and build environment, social and community context) during the twelve months before the biobank enrollment, we used seven negative stressors from the SFBN as proxies (i.e., lack of family support, unemployment, financial problems, housing problems, educational problems, inadequate health/mental health coverage and problems with access to health care services) (45).

Markers of bipolar illness severity included rapid cycling course within the previous year, history of suicide attempts and/or psychotic symptoms, cumulative anxiety disorders, illicit substance use disorders and nicotine and alcohol dependence 2.2 Definition of unhealthy and healthy diet quality using the REAP-S

Diet quality was measured using the first 13 of 16 items of the REAP-S (individual items can be reviewed in Supplementary Table 1 and Supplementary Figure 2)(19,20). These evaluate the relative intake of specific food groups and respondents answer the frequency in which a determined food group is consumed in an average week (usually/often= 1 point; sometimes= 2 points; rarely/never or does not apply to me= 3 points). The remaining three questions focus on the willingness to make lifestyle changes and are not included in the total score. Total scores for the first thirteen items range from 13 to 39 points and offer an estimate of an individual’s diet quality; higher total scores indicate higher diet quality. A previous study by Johnston et al. (2018) reported a mean score for the average U.S. omnivorous diet of 32 points (20). Based on the latter, and as done in a previous MCCB study, we used a total score of REAP-S ≥32 as a definition of healthy diet quality and scores <32 points to define unhealthy diet quality (21).

2.3. Statistical analysis for clinical variables

In descriptive statistics, we report frequencies with percentages for categorical variables and means with standard deviations for continuous variables. Regression models were used to examine associations between diet quality (healthy vs. unhealthy) and demographic and clinical variables, adjusting for study site. A multivariable logistic regression model predicting UDQ was constructed using Least Absolute Shrinkage and Selection Operator (LASSO) to obtain a parsimonious set of demographic and clinical variables associated with UDQ. LASSO penalizes coefficients and shrinks regression results using a regularization parameter, $\lambda$, selected based on the one-standard-error rule (i.e., the largest value of $\lambda$ such that the mean deviance on left-out data in 10-fold cross-validation is within one standard-error of the minimum). The final independent variables selected for the model included age, education level, bipolar subtype, chronotype (evening vs non-evening), and any anxiety disorder, which were refit to obtain parameter estimates. Finally, separate logistic regression models were conducted for medical comorbidities, psychotropic use, and SDoH stressors as outcome variables, with UDQ as the predictor, adjusting for age, sex, and study site. Odds ratios and 95% confidence intervals are reported. In these models, age was scaled in 10-year increments to improve interpretability. Psychotropic medications were classified into the following groups based on clinical indications: lithium, mood-stabilizing antiepileptics, ADHD medications, antipsychotics, antidepressants, hypnotics, and thyroid medications. All statistical tests were two-tailed with a significance level of 0.05. Analyses were performed using R Statistical Software (v4.4.1; R Core Team 2024).

3. Results

3.1. Sociodemographic and clinical characteristics

Of the 737 participants included in this analysis; BD participants with UDQ (78.8% of the sample) were significantly younger (38.6±14.2 vs. 43.0±15.3; p<.001) (Table 1). Being married and having an education level greater than high school were significantly associated with healthy diet quality. Participants with UDQ were more likely to have a diagnosis of BD-I or schizoaffective bipolar subtype (65.6% vs. 53.8%; p= .03) and to have received an early BD diagnosis (≤19 years of age) (75.2% vs. 61.9%; p= .003). Evening chronotype, as defined by participants who responded “definitely an evening type” in the MDQ, (53.2% vs. 38.7%; p= .001) was associated with UDQ; however, no statistically significant group differences were observed in terms of current mood episode, history of psychosis, suicide attempts, antidepressant-induced mania or rapid cycling. Psychiatric comorbidities, including anxiety disorders (69.0% vs. 52.3%; p<.001), illicit substance abuse (34.6% vs. 20.3%; p= .002), nicotine dependence (43.3% vs. 30.7%; p=.04) and binge eating disorder (13.9% vs. 8.5%; p= .03), were more common in participants with UDQ. In terms of body composition, participants with UDQ were significantly more likely to have a higher BMI (30.3±7.6 vs. 27.5±6.2; p<.001) and obesity (42.8% vs. 27.9%; p<.001). Figure 1 illustrates the distribution of participants based on BD subtype, diet quality scores and BMI categories, indicating how participants with BD-I have a higher tendency to have a UDQ (65.6% vs. 34.4%) and a BMI > 25 (76.9% vs. 69.3%). Detailed REAP-S item-level comparisons are available in Supplementary Table 1 and Supplementary Figure 2.

Table 1.

Comparison of demographic and clinical characteristics of patients with BD with healthy and unhealthy diet quality (adjusted for site).

Characteristics	Healthy Diet Quality (N= 156)	Unhealthy Diet Quality (N= 581)	Total (N= 737)	p-value
REAP-S score, mean (SD)	33.8 (1.7)	25.8 (3.7)	27.5 (4.7)
Age, mean (SD), N=728	43.0 (15.3)	38.6 (14.2)	39.6 (14.6)	<0.001
Sex, female, N (%)	110 (70.5%)	362 (62.3%)	472 (64.0%)	0.047
Racial identity, White, n (%), N=734	105 (67.7%)	407 (70.3%)	512 (69.8%)	0.17
Ethnicity, Hispanic or Latino, N=728	55 (35.5%)	172 (30.0%)	227 (31.2%)	0.66
Employment, fulltime, N=725	36 (23.5%)	139 (24.3%)	175 (24.1%)	0.89
Education level, > high school, N=709	136 (91.9%)	453 (80.7%)	589 (83.1%)	<0.001
Marital status, married, N=728	78 (50.6%)	209 (36.4%)	287 (39.4%)	<0.001
BDI and schizoaffective bipolar subtype	84 (53.8%)	381 (65.6%)	465 (63.1%)	0.027
Age at diagnosis, ≤19 years, N=700	91 (61.9%)	416 (75.2%)	507 (72.4%)	0.003
Current mood state, N=736				0.24
Manic	7 (4.5%)	38 (6.6%)	45 (6.1%)
Mixed	8 (5.1%)	47 (8.1%)	55 (7.5%)
Hypomanic	15 (9.6%)	53 (9.1%)	68 (9.2%)
Major depressive	58 (37.2%)	236 (40.7%)	294 (39.9%)
Euthymic	68 (43.6%)	206 (35.5%)	274 (37.2%)
History of
Psychosis, N=736	44 (28.2%)	213 (36.7%)	257 (34.9%)	0.071
Suicide attempts, N=725	50 (32.7%)	225 (39.3%)	275 (37.9%)	0.14
Antidepressant-induced mania, N=600	17 (13.9%)	97 (20.3%)	114 (19.0%)	0.12
Rapid cycling, N=602	65 (51.6%)	264 (55.5%)	329 (54.7%)	0.52
Body mass index, mean (SD), N=720	27.5 (6.2)	30.3 (7.6)	29.7 (7.4)	<0.001
BMI Category, N=720				<0.001
<25.0	57 (37.0%)	132 (23.7%)	191 (26.5%)
25.0 – 29.9	54 (35.1%)	190 (33.6%)	244 (33.9%)
30.0 – 39.9	39 (25.3%)	180 (31.8%)	219 (30.4%)
≥40.0	4 (2.6%)	62 (11.0%)	66 (9.2%)
Obesity, N=720	43 (27.9%)	242 (42.8%)	285 (39.6%)	<0.001
Evening chronotype, N=732	60 (38.7%)	307 (53.2%)	367 (50.1%)	<0.001
Psychiatric comorbidities
Anxiety disorders, N=724	80 (52.3%)	394 (69.0%)	474 (65.5%)	<0.001
Illicit substance abuse, N=723	31 (20.3%)	197 (34.6%)	228 (31.5%)	0.002
Nicotine dependence, N=723	47 (30.7%)	247 (43.3%)	294 (40.7%)	0.038
Alcohol dependence, N=722	46 (30.1%)	208 (36.6%)	254 (35.2%)	0.14
ADD/ADHD, N=723	25 (16.3%)	129 (22.6%)	154 (21.3%)	0.31
Binge eating disorder, N=723	13 (8.5%)	79 (13.9%)	92 (12.7%)	0.028

P-values are from regression models with UDQ as the dependent variable, examining its association with each characteristic while adjusting for study site. Abbreviations: SD= standard deviation; BMI= body mass index; ADD/ADHD= Attention deficit disorder/attention deficit and hyperactivity disorder; REAP-S= Rapid Eating Assessment for Participants – Shortened Version.

Figure 1.

Sankey diagram illustrating the relationship between bipolar disorder subtypes, diet quality and body mass index in the studied cohort; the width of each flow bar corresponds with the number of participants in each subgroup.

3.2. Relationship of socio-clinical variables with unhealthy diet quality

A multivariable logistic regression model was constructed to examine the associations between sociodemographic and clinical variables and the likelihood of presenting a UDQ phenotype (Figure 2). Several variables showed significant associations.

Figure 2.

Multivariable logistic regression model predicting unhealthy diet quality.

Older age was associated with lower odds of UDQ (OR: 0.87; 95% CI: 0.76–0.99; p= .031; per 10-year increment), suggesting that younger individuals with BD may be more prone to unhealthy dietary behaviors. Similarly, having a higher education level was associated a substantially reduced likelihood of UDQ (OR: 0.40; 95% CI: 0.19–0.75; p= .007), highlighting a potential protective effect related to education.

On the contrary, participants with a BD-I or schizoaffective diagnosis, when compared with BD-II, were significantly more likely to present UDQ (OR: 1.95; 95% CI: 1.32–2.89; p= .001), possibly reflecting the greater complexity of these subtypes. Evening chronotype was also associated with increased odds of UDQ (OR: 1.86; 95% CI: 1.25–2.78; p= .002), reinforcing the role of circadian rhythm disruptions in dietary behavior. Finally, the presence of comorbid anxiety disorders showed an increased likelihood of UDQ (OR: 1.94; 95% CI: 1.31–2.88; p= .001), supporting previous links between anxiety and suboptimal eating patterns.

3.3. Medical comorbidities, psychotropic use and social determinants of health-related stressors

Out of the 14 medical comorbidities analyzed in this study, only hypertension and obesity had statistically significant associations with UDQ. Participants with UDQ had significantly higher odds of hypertension (OR= 1.95, 95% CI: 1.13–3.53; p= .02) and obesity (OR= 2.07, 95% CI: 1.38, 3.14; p<.001) compared to those with healthy diet quality (Table 2). There were no significant associations between different psychotropics used and UDQ (Table 3). Analysis of social determinants of health-related stressors present in the year prior to study enrollment showed that only unemployment-related stress was significantly associated with UDQ (Table 4). These models were controlled for age, sex and study site.

Table 2.

Medical comorbidities (N=728).

	Healthy Diet Quality (N= 154)	Unhealthy Diet Quality (N= 574)	Odds Ratio	p-value
Hypertension	20 (13.0%)	104 (18.1%)	1.95 (1.13, 3.53)	0.021
Eczema	7 (4.5%)	59 (10.3%)	2.00 (0.95, 4.84)	0.071
Psoriasis	6 (3.9%)	17 (3.0%)	0.89 (0.35, 2.57)	0.81
Diabetes	9 (5.8%)	47 (8.2%)	1.83 (0.88, 4.19)	0.13
PCOS, N=466	3 (2.8%)	36 (10.1%)	2.77 (0.93, 11.94)	0.11
Thyroid disorders	29 (18.8%)	101 (17.6%)	1.02 (0.64, 1.67)	0.93
Fibromyalgia	10 (6.5%)	33 (5.7%)	0.91 (0.44, 2.05)	0.81
Migraine	32 (20.8%)	179 (31.2%)	1.50 (0.95, 2.42)	0.088
Asthma	28 (18.2%)	113 (19.7%)	0.99 (0.61, 1.62)	0.96
Obstructive sleep apnea	14 (9.1%)	63 (11.0%)	1.39 (0.74, 2.75)	0.31
Irritable bowel syndrome	17 (11.0%)	81 (14.1%)	1.34 (0.76, 2.47)	0.32
Obesity, N=714	42 (27.6%)	239 (42.5%)	2.07 (1.38, 3.14)	<0.001

Odds ratio (OR) of medical comorbidities based on UDQ adjusted for age, sex, and study site. Abbreviations: PCOS: Polycystic ovary syndrome.

Table 3.

Psychotropics at time of study enrollment (N=737).

	Healthy Diet Quality (N= 156)	Unhealthy Diet Quality (N= 581)	Odds Ratio	p-value
Lithium, N=417	38 (43.7%)	151 (45.8%)	0.89 (0.53, 1.49)	0.66
Antiepileptic mood stabilizers, N=606	85 (68.0%)	284 (59.0%)	0.77 (0.50, 1.19)	0.24
Antipsychotics, N=608	74 (63.2%)	305 (62.1%)	0.97 (0.63, 1.48)	0.88
Antidepressants, N=608	65 (53.3%)	233 (47.9%)	0.84 (0.55, 1.27)	0.41
ADHD medications, N=201	10 (24.4%)	58 (36.2%)	2.03 (0.92, 4.80)	0.079
Hypnotics, N=675	18 (12.6%)	87 (16.4%)	1.37 (0.79, 2.49)	0.28

Odds ratio (OR) of psychotropic medications based on UDQ adjusted for age, sex, and study site. Medication groups: Antiepileptic mood stabilizers: valproate, lamotrigine, carbamazepine; antipsychotics: typical and atypical antipsychotics; Antidepressants: SSRI, SNRI, tricyclics, mirtazapine, other antidepressants; ADHD medication: amphetamine salts, non-amphetamine stimulants, bupropion, atomoxetine; Hypnotics: benzodiazepines, non-benzodiazepine sedatives, ramelteon.

Table 4.

Differences in social determinants of health-related stressors (N=719).

	Healthy Diet Quality (N= 156)	Unhealthy Diet Quality (N= 568)	Odds Ratio	p-value
Lack of family support	39 (25.8%)	209 (36.8%)	1.48 (0.96, 2.30)	0.079
Unemployment	25 (16.6%)	160 (28.2%)	1.69 (1.05, 2.80)	0.036
Financial problems	43 (28.5%)	235 (41.4%)	1.52 (0.99, 2.35)	0.059
Health coverage problems	14 (9.3%)	79 (13.9%)	1.17 (0.64, 2.27)	0.62
Health access problems	13 (8.6%)	74 (13.0%)	1.25 (0.68, 2.46)	0.48
Educational problems	13 (8.6%)	60 (10.6%)	0.84 (0.44, 1.72)	0.63
Housing problems	6 (4.0%)	59 (10.4%)	2.05 (0.91, 5.50)	0.11

Odds ratio (OR) of SDoH stressors based on UDQ adjusted for age, sex, and study site.

4. Discussion

To our knowledge, this is one of the first studies to comprehensively investigate unhealthy diet quality as a discrete BD phenotype and its associations with social, clinical and metabolic outcomes in a large, well-characterized cohort. Utilizing the REAP-S in a cohort of patients with BD, we found that those with UDQ, when compared to those with a healthy diet quality, were more likely to be men, younger and to have a lower educational level and more unemployment-related stress. Likewise, participants with UDQ were diagnosed with BD at an earlier age and were more likely to have an evening chronotype, along with a higher prevalence of comorbid conditions, including obesity, hypertension, anxiety disorders, binge eating disorder nicotine dependence, and illicit substance abuse. Despite this, UDQ did not appear to be associated with specific markers of severe BD, such as rapid cycling, a history of suicide attempts, or psychosis.

As previously mentioned, our study builds on prior diet quality analyses of this cohort that identified a negative relationship between REAP-S scores and BMI, waist circumference, depressive symptom severity and disordered eating (21). In this same cohort, participants with an evening chronotype were more likely to skip breakfast, have less fruit and vegetable consumption, and have a higher intake of ultra-processed foods (14). Compared to epidemiological data of US adults, where only 60–65% achieved ideal scores in diet quality assessments (17), rates of UDQ in our cohort 78.83% were notoriously higher. Similarities are observed when comparing our findings to the few existing studies focusing on diet quality in BD, including higher scores in proinflammatory dietary patterns in women with BD (46) and euthymic BD patients (47). Rates of overweight/obesity in our cohort (73.%) where also similar as the ones reported in the former study (70%). Finally, it supports the notion that a subset of individuals with mood disorders, in this case individuals with BD, exhibit an increased consumption of unhealthy foods and beverages (23,32).

Our finding of lower educational attainment (≤ high school), a commonly used proxy marker of socioeconomic status, in UDQ BD participants, compared to those with healthy diet quality, is consistent with previous studies in psychiatric and general populations, that have further linked unhealthy diet quality with poorer levels of health and food literacy (48–51). The latter refers to an individual’s dynamic ability to understand nutrition information required to make informed foods choices, including the theoretical and practical knowledge to critically analyze dietary recommendations and to shop and prepare foods (52). In patients with BD and UDQ, the need for critical capacity for making healthy dietary choices may be hindered by low educational attainment combined with the presence of core (hypo)manic and depressive symptoms (i.e., cognitive distortions, impulsivity-related poor decision making, anhedonia and lack of motivation), which can further magnify perceived and factual barriers to maintain a healthy diet (53,54). Additionally, in light of the increased odds of unemployment related stress in UDQ observed in our cohort, it is important to consider the indirect costs of BD, often related to unemployment and lower work productivity. These costs can make up as much as 80% of the financial burden of BD and may contribute to food insecurity and poor food choices (12). Therefore, there is merit in further exploring the impact on mood outcomes of dietitian-led educational sessions focused on food literacy and socioeconomic factors.

Adolescent-onset BD has been associated with a more severe course of illness, including more medical and psychiatric comorbidity compared to healthy youth (6). This apparent vulnerability may be partially explained by chronic systemic inflammation that contributes to the early appearance of cardiovascular diseases (i.e., atherosclerosis, hypertension) (55,56) and further worsened by unhealthy eating behaviors, including pro-inflammatory dietary patterns (35,57). Not surprisingly, individuals with UDQ from our study had an earlier BD diagnosis and higher rates of hypertension and obesity than those with healthy diet quality. Similarly, in our multivariable logistic regression model, a younger age at BD diagnosis remained moderately associated with increased odds of having UDQ. While this could be interpreted as a shorter duration of untreated illness, it also entails earlier onset of mood symptoms, medication side effects, and an overall cumulative effect of illness exposure.

Some of the associations observed in this study could offer insights into the chronobiological mechanisms involved in the role of UDQ in BD. For instance, the association between UDQ and evening chronotype, a marker of circadian system dysfunction previously linked to unhealthy dietary patterns, metabolic imbalances and disordered eating, might be a consequence of a positive feedback loop. In this cycle, evening chronotype influences the timing of meals, interpretation of interoceptive inputs, food cravings and preferences (i.e. seeking foods high in fats and sugar) and eating behaviors, which further disrupt circadian rhythms (14,35,58). Likewise, the associations found with obesity, nicotine and illicit substance use disorders and binge eating disorder suggest a link between impulsivity, a core symptom of (hypo)manic episodes, and UDQ. Previous studies have identified that tobacco users with BD report more craving for high fat foods than non-smokers and the co-occurrence of impulsivity and obesity are associated with a more severe bipolar course of illness (35,59). Investigating chronobiological pathways common to UDQ and BD, such as altered frontal GABAergic function (a key inhibitory neurotransmitter linked to mood instability and impulsivity in both disorders) (36,60) could enhance our understanding of the intricate relationship between diet, mood and metabolic disorders and pave the way for the development of personalized zeitgeber-based dietary interventions.

Given the associations between UDQ and aforementioned social and clinical factors identified in our study and the evidence on the modest beneficial effect of brief diet counseling on diet quality and metabolic health (61), incorporating brief diet quality screenings during mental health visits might be a valuable component of a holistic approach to managing BD. Many of the associations found in this study fall within the realm of modifiable or addressable risk factors to ultimately benefit diet quality. In this line, our study offers several strengths that merit mentioning. First, the studied cohort is a large sample of well-characterized individuals from different demographic backgrounds and a confirmed BD diagnosis, enhancing the findings’ generalizability. Second, we use a brief, previously validated dietary assessment tool that evaluates the total diet rather than single food groups, sensitive to treatment-related changes in food intake and that has been previously recommended by the American Heart Association as an optimal diet screener tool (17,21). Third, the study comprehensively explores the association between diet quality and a range of clinical, demographic, and psychosocial factors, including proxies of social determinants of health. This provides a nuanced understanding of the complex interplay between these variables. Finally, identifying unhealthy diet quality as a distinct clinical phenotype of BD could aid in the development of targeted lifestyle interventions.

While this study provides valuable insights into the associations between unhealthy diet quality and BD, several limitations need to be considered when interpreting our findings. First, our analysis is limited by its cross-sectional design, which impedes the evaluation of causality and the development of specific recommendations to modify clinical practice. Second, the risk of measurement errors and recall bias inherent to memory-based self-reported dietary assessment methods and self-reported clinical diagnoses may affect the accuracy of the measurements. This can limit the reliability of the variables being analyzed, particularly of diet quality, a concept that, additionally, lacks a consensual definition; more accurate measures of dietary intake and quality, such as 24-hour dietary recall diaries, photo-based dietary assessments or AI-powered food recognition apps, could provide a more accurate assessment, ideally in prospective studies. In this line, the absence of standardized diet questionnaires limits the comparability and generalizability of the study’s findings. Third, the relatively young age of our cohort may limit the detection of medical comorbidities, particularly those related to obesity and low diet quality, as their prevalence is typically lower in this age group. Fourth, while diet quality was analyzed using a previously established dichotomous classification, we acknowledge that this approach may reduce the granularity of the data. Finally, the study did not account for other potential confounding factors, such as socioeconomic status, cultural factors, including a direct assessment of social determinants of health, and non-psychiatric medication use, which may influence both diet quality and clinical outcomes.

In conclusion, this study provides compelling evidence to comprehensively identify unhealthy diet quality as a distinct clinical phenotype of bipolar disorder and, by highlighting specific social and clinical factors associated with UDQ, it provides a roadmap to inform the development of targeted future interventions. Our findings reveal a significant association between UDQ and adverse social factors and medical and psychiatric comorbidities in individuals with BD; underscoring the potential of routine clinician-delivered diet assessments during mental health visits to reduce the health impacts of UDQ. Future prospective longitudinal studies should delve deeper into the underlying mechanisms linking diet quality to BD to infer causality. Additionally, in lieu of our findings, large-scale, randomized controlled trials are needed to evaluate the efficacy and sustainability of specific dietary interventions in improving metabolic and mood outcomes in individuals with BD and should leverage recommendations for dietary assessments in clinical settings that are already routinely incorporated in other medical specialties. Based on the associations observed in our study, we suggest that future studies aim towards personalized, multi-component dietary interventions that combine chronobiologic, socioeconomic and cultural factors with clinical data and omics platforms.

Highlights.

• The study characterizes unhealthy diet quality (UDQ) using clinical, sociodemographic, and dietary data, identifying it as a distinct bipolar disorder phenotype.

• Sociodemographic, clinical, and comorbidity factors converge to influence UDQ in bipolar disorder patients.

• The study highlights the potential of exploring tailored dietary interventions as treatment aides in bipolar disorder.