The Impact of Body Mass Index on the Clinical Features of Bipolar Disorder: A STEP-BD Study

Abstract

Introduction:

The effects of body mass index (BMI) on the core symptoms of bipolar disorder (BD) and its implications for disease trajectory are largely unexplored.

Objective:

To examine whether BMI impacted hospitalization rate, medical and psychiatric comorbidities, and core symptom domains such as depression and suicidality in BD.

Methods:

Participants (15 years and older) were 2790 BD outpatients enrolled in the longitudinal STEP-BD study; all met DSM-IV criteria for BD-I, BD-II, cyclothymia, BD NOS, or schizoaffective disorder, bipolar subtype. BMI, demographic information, psychiatric and medical comorbidities, and other clinical variables such as bipolarity index, history of electroconvulsive therapy (ECT), and history of suicide attempts were collected at baseline. Longitudinal changes in Montgomery-Åsberg Depression Rating Scale (MADRS) score, Young Mania Rating Scale (YMRS) score, and hospitalizations during the study were also assessed. Depending on the variable of interest, odds-ratios, regression analyses, factor analyses, and graph analyses were applied.

Results:

A robust increase in psychiatric and medical comorbidities was observed, particularly for baseline BMIs>35. A significant relationship was noted between higher BMI and history of suicide attempts, and individuals with BMIs>40 had the highest prevalence of suicide attempts. Obese and overweight individuals had a higher bipolarity index (a questionnaire measuring disease severity) and were more likely to have received ECT. Higher BMIs correlated with worsening trajectory of core depression symptoms and with worsening lassitude and inability to feel.

Conclusions:

In BD participants, elevated BMI was associated with worsening clinical features, including higher rates of suicidality, comorbidities, and core depression symptoms.

Introduction

Bipolar disorder (BD) carries a two- to three-fold higher risk of early mortality compared to the general population¹. Furthermore, recent longitudinal studies found that the mortality gap between individuals with BD and the general population has widened over the past two decades², and that individuals with BD have an average reduced lifespan of nine to 17 years^1,3 due to factors that include increased suicide rate, medical comorbidities, metabolic syndrome, and physical inactivity^1,4-6. Compared to individuals with major depressive disorder or healthy individuals, those with BD are at increased risk for suicidal behavior—a prevalence rate of 29.2% for attempts and 18.8% for completion⁷.

In the past three decades, obesity rates in the United States (US) and worldwide have risen sharply. In the US between 1999-2016, for instance, the average body mass index (BMI) increased in men from 27.7 to 29.1 and in women from 28.2 to 29.6⁸. The National Health and Nutrition Examination Survey (NHANES) for 2015–2016 revealed that the prevalence of obesity was 39.8% among US adults age 20 and older (37.9% for men and 41.1% for women), affecting about 93.3 million individuals⁹.

While obesity is a risk factor for mortality across the disease spectrum¹⁰, individuals with mood disorders appear to be at considerably increased risk^11,12. Furthermore, obesity itself appears to be associated with several psychiatric disorders, including BD¹³. In particular, abdominal obesity associated with overweight was linked to a roughly 12% increase in major depressive disorder and a 47% increase in BD¹³. In individuals with BD, obesity rates have been associated with significantly greater illness burden, poorer outcome measures, and increased premature mortality¹⁴. Interestingly, one of the key contributing factors to the increased obesity rates seen in individuals with BD appears to be depression-related behaviors, including lifestyle modifications, psychological stressors, and the iatrogenic impact of psychotropic treatments¹⁵.

In the US, suicide mortality rates in the general population increased dramatically between 1999 and 2017, echoing similar increases in obesity rates during the same time frame¹⁶. In this context, it is interesting to note that obesity appears to be an important risk factor for depression and affective instability, especially in the US^17,18. Clear evidence of a link between obesity and suicidal behavior, however, remains elusive. For instance, one large, population-based study of individuals of European descent found a positive relationship between suicidal ideation and obesity¹⁹. However, another study found a more nuanced relationship between obesity and suicide, reporting that, in men, high BMI was associated with low risk of attempted or completed suicide but that, in women, high BMI was associated with a higher risk of attempted suicide but a lower risk of completed suicide²⁰. The association between obesity and suicidal ideation has also been shown to be mediated by ethnicity/race, with African-American females demonstrating no significant relationship between weight and suicidal ideation in contrast to European-American females²¹. Another recent study found that certain ethnicities/races were a significant covariate in the relationship between suicide behaviors and obesity in adolescents²². Finally, a large meta-analysis found that higher BMI was associated with decreased suicide risk²³. The mixed scientific evidence suggests that additional, thorough analyses in larger psychiatric samples are needed.

Multiple lines of evidence indicate that obesity, particularly abdominal obesity, is accompanied by changes in pro-inflammatory, metabolic, and oxidative stress signatures that in turn affect CNS cellular circuits and network systems²⁴. Because inflammation has been consistently associated with both depressive states and obesity, it could be a major underlying mechanism to explain the link between BD and BMI. For instance, Bond and colleagues found that higher BMI directly contributed to an inflammatory state in BD that predicted not only mood severity but also course of illness and depressive relapse²⁵. Other studies have suggested that adipokines—cytokines secreted by adipose tissue—regulate metabolism, appetite, and energy expenditure in the body as well as impact frontal cortical development and functional connectivity²⁶. Interestingly, a recent multicenter trial found an association between BMI and subcortical volumes in individuals with BD^27,28, and another study found that higher BMI in adolescents with BD was associated with aberrant frontal neurocircuitry and reduced frontal cortical volumes²⁹. Taken together, these findings suggest that increased low-grade inflammation—triggered by higher BMI—might aggravate disease course in BD^30,31. They also suggest that increases in global obesity rates, including among individuals with BD, may shape the phenotypic presentation of BD in clinical settings³².

The effects of BMI on the core symptoms of BD and its implications for disease trajectory have, to date, remained largely unexplored. The aims of the study were to assess whether higher BMI affected disease course, severity of canonical symptom domains (depression or affective instability), and disease burden (medical and psychiatric comorbidities) in individuals with BD, using data drawn from the longitudinal, seven-year Systematic Treatment Enhancement Program for Bipolar Disorder (STEP-BD) study³³. Links between higher BMI and proxies of disease severity—such as history of suicide attempts, suicide-related hospitalizations, and prior electroconvulsive therapy (ECT) trials—were also assessed.

Materials and Methods

Participants and Study Interventions

Data for this analysis were drawn from the STEP-BD study, a seven-year, longitudinal study conducted across 22 academic psychiatric centers in the United States (US; NCT00012558). The STEP-BD study followed individuals’ trajectories of symptoms for up to 26 weeks³⁴. Once enrolled, participants could receive care for as long as they were in the program—up to five years—and were monitored systematically, even when they were feeling well. Briefly, BD patients were eligible for inclusion if they were 15 years or older (Best Practice Pathways) or 18 years or older (Randomized Care Pathways) and met DSM-IV criteria for BD-I, BD-II, cyclothymia, BD not otherwise specified, or schizoaffective disorder, bipolar subtype as confirmed by the Mini-International Neuropsychiatric Interview (MINI, version 4.4). Notably, participants in the study could have a number of comorbidities, including but not limited to anxiety disorders, substance use disorders, attention deficit hyperactivity disorder (ADHD), and eating disorders. Patients were treated in a naturalistic treatment setting by study clinicians and were eligible for interventions felt to be clinically indicated by their clinicians. Follow-up by study clinicians was conducted at timepoints and intervals each clinician deemed necessary. The study was approved by the institutional review boards of each separate treatment center. Before inclusion, all participants provided written informed consent to all study procedures. Additional details about the participants and the study design have previously been published³³.

For this particular analysis, the weight and height of 2790 individuals diagnosed with BD who met inclusion criteria for the STEP-BD study were analyzed along with a number of related psychometric measures. Only BD patients who were outpatients at the time of their inclusion and those adhering to study procedures in the STEP-BD study were studied.

Outcome Parameters

Relative body weight

Relative body weight was defined as both a categorical and quantitative variable in order to make these findings comparable with those of previous studies that used both approaches. BMI (a person’s weight in kilograms divided by their squared height in meters) was obtained at baseline and calculated based on participants’ body weight and height measured on the day of enrollment. For categorical analyses, participants were grouped into seven BMI categories based on CDC guidelines³⁵: “underweight” (BMI<18.5); “thin” (BMI≥18.5 to <20); “normal weight” (BMI ≥20-25); “overweight” (BMI ≥25 to <30); “class I obesity” (BMI ≥30 to <35); “class II obesity” (BMI ≥35 to <40); “class III obesity or extreme or severe obesity” (BMI ≥40) (Figure 1).

Figure 1. BMI distribution in the STEP-BD dataset (n=2790).

A. Categorization of body mass index (BMI) in the STEP-BD sample cohort according to CDC guidelines. B. BMI distribution and bipolarity index, a clinician-rated scale that rates cardinal features of the disorder across five domains: signs and symptoms, age of onset, course of illness, response to treatment, and family history. C. BD participants with the highest BMIs had a 1.67 (95% CI 0.86-2.48) increased risk for having had previous ECT treatments. D. A significant relationship was observed between BMI and history of suicide attempts (odds ratio (OR) (shaded region = 95% confidence interval). E. When the group with the lowest BMI was excluded, an increased rate of hospitalizations during the study was observed as a function of BMI.

Demographic variables and neuropsychological scales

The STEP-BD study used a standardized patient assessment tool known as the Clinical Monitoring Form (CMF). The CMF—administered at baseline—comprises nine parts, including the Structured Clinical Interview for DSM-IV (SCID), current mood modules, comorbid medical and psychiatric conditions, medication use, adverse effects, substance use, stressors, and care utilization³⁶. The variables “history of ECT” and “history of suicide attempt” were also collected during the baseline assessment.

The bipolarity index was administered at baseline to assess core attributes of BD. The bipolarity index is a clinician-rated tool that measures cardinal features of BD across five domains: signs and symptoms, age of onset, course of illness, response to treatment, and family history. The clinician assigns a score betweent 0 and 20 to each of these five dimensions, with higher scores representing traits more characteristic of BD³⁷.

For longitudinal measures, the Montgomery-Åsberg Depression Rating Scale (MADRS) and Young Mania Rating Scale (YMRS) were administered by trained raters at baseline and every three months for the first year of study participation, then every six months for the rest of the trial up to 36 months. The severe adverse events (SAE) report was performed serially during visits and was used to investigate hospitalizations and hospitalizations for suicide during the study.

Statistics and Modeling

All statistical analyses were performed in R 4.0.3³⁸. Sample size characteristics are presented with descriptive statistics. Given the high expected inter-relationship between comorbidities, a network-based approach was chosen to assess the influence of BMI on psychiatric and medical comorbidities. A graph model (Watts-Strogatz) was constructed using R studio software 4.2.2, where the size of the nodes represented the percentage of individuals within each BMI category with a particular comorbid diagnosis at baseline, and the thickness of the edges represented the frequency of between-pairs diagnoses. The node sizes and edge weights were normalized to the number of participants and the number of possible edge connections within each BMI category, respectively. From these normalized graphs, the median nodal strength—that is, the sum of the weights of the edges connected to each node—was computed. The higher the median nodal strength, the greater the number of comorbidities exhibited by patients in each BMI category.

The relationship between BMI and psychosocial variables was tested with adjusted odds-ratios. To test the influence of BMI on core depressive and manic symptoms, sub-items from the MADRS and YMRS rating scales at the baseline visit were graphically plotted. Linear regressions were then calculated with qualitative BMI as the dependent variable and MADRS/YMRS sub-items as the independent variables covaried for age, sex, and medication weight gain risk. To reduce the influence of medications with known weight gain effects, an individual risk score was calculated, resulting in a sum score of +1 for medications associated with weight gain, zero for those that did not affect weight, and −1 for medications associated with weight loss (hereafter termed ‘med risk score’). To further test the relationship between BMI and core depressive and manic symptoms, an exploratory factor analysis (EFA) of MADRS and YMRS items was performed using the minimum residual approach³⁹ with oblique, oblimin rotations to allow for between-factor correlations. The number of factors was determined using parallel analysis. All items had loadings >= 0.2, and all were included. Hierarchical clustering was used to corroborate factor structures.

Total scores were tracked from the largest factors in the EFA longitudinally over study visits. K-means cluster modeling for longitudinal data was used to identify distinct, homogeneous clusters of factor score trajectories. The proportion of individuals belonging to each cluster was subsequently quantified across BMI categories.

Results

The mean age of the sample was 39.6 years, and 54.4% were female, which matches demographic information for the STEP-BD sample as a whole⁴⁰. The BMI distribution sample categories demonstrated that 1.8% of the 2790 participants were in the “underweight” category (n=51), 3.5% were in the “thin” category (n=98), 29.9% were in the “normal weight” category (n=834), 30.9% were in the “overweight” category (n=861), 18.9% were in the “class I obesity” category (n=526), 8.7% were in the “class II obesity” category (n=242), and 6.4% were in the “class III obesity or extreme or severe obesity” category (n=178) (Figure 1A). The average BMI of this sample was 28.42+/−6.75 (median=27.20). Sex (χ²=130.16, p<0.001) and age (F=86.48, p<0.001) differed significantly across BMI categories; specifically, females were more often underweight, and more males with BD were found in every BMI category above 30 kg/m². Table 1 lists additional demographic variables of our sample across different BMI categories.

TABLE 1.

Demographic variables of the STEP-BD sample across different BMI categories.

	BMI categories							F, χ2	P
	Underweight (<18.5)	Thin (18.5-20)	Normal weight (20-25)	Overweight (25-30)	Class I obesity (30-35)	Class II obesity (35-40)	Class III obesity (>40)
N (%)	51 (1.8)	98 (3.5)	834 (29.9)	861 (30.9)	526 (18.9)	242 (8.7)	178 (6.4)
Demographics
Age, years, mean (SD)	29.9 (10.2)	33.3 (11.6)	36.8 (12.7)	41.2 (12.9)	42.4 (12.4)	40.6 (11.9)	41.5 (11.3)	F = 86.48	<0.001
Female, n (%)	46 (90.2)	86 (87.8)	533 (63.9)	407 (47.3)	271 (51.5)	149 (61.6)	124 (69.7)	χ2 = 130.16	<0.001
Adolescent (age 15-18), n (%)	6 (11.8)	3 (3.1)	28 (3.4)	11 (1.3)	7 (1.3)	4 (1.7)	4 (2.2)	χ2 = 31.93	<0.001
Race, n (%)
White or Caucasian	44 (86.3)	91 (92.8)	763 (91.6)	785 (91.2)	466 (88.4)	207 (85.5)	151 (85.2)	χ2 = 16.6	0.011
Black or African American	4 (7.8)	1 (1.0)	25 (3.0)	40 (4.7)	40 (7.7)	24 (10.0)	22 (11.9)	χ2 = 42.73	<0.001
Native American or American Indian	0 (0)	0 (0)	5 (0.6)	8 (0.9)	5 (1.0)	5 (2.1)	1 (0.6)	χ2 = 6.35	0.39
Asian or Pacific Islander	1 (2.0)	4 (4.1)	27 (3.3)	14 (1.6)	10 (1.9)	0 (0)	1 (0.6)	χ2 = 15.74	0.015
No Primary Race	1 (2.0)	0 (0)	4 (0.5)	4 (0.5)	2 (0.4)	4 (0.2)	2 (1.1)	χ2 = 8.20	0.22
Other	1 (2.0)	2 (2.1)	9 (1.1)	9 (1.1)	3 (0.6)	2 (0.8)	1 (0.6)	χ2 = 3.04	0.8
College degree, n (%)	7 (13.7)	23 (23.7)	229 (27.8)	238 (27.8)	134 (26.1)	42 (17.5)	36 (20.6)	χ2 = 18.85	<0.01
Full time employment, n (%)	25 (71.4)	55 (84.6)	440 (84.0)	493 (91.8)	328 (95.6)	155 (93.9)	130 (93.5)	χ2 = 55.12	<0.001

With regard to core features of the disease, a significant difference in bipolarity index scores was observed (F=4.88, p=0.027), with overweight and obese patients exhibiting higher scores. Two items exhibited an especially strong relationship with BMI: ‘episode characteristics’ (representing likeliness for hypomanic and mixed episodes) (F=23.46, p<0.001) and ‘course of the illness’ (indexing risk of recurrence) (F=6.13, p=0.013) (Table 2, Figure 1B). History of previous ECT treatment also differed significantly across BMI classes (χ²=15.12, p=0.019), with low BMI (<20) groups having the lowest rates of history of ECT treatment (Table 2, Figure 1C). Furthermore, a significant relationship was observed between higher BMI and history of suicide attempts, with individuals in the highest obesity category (BMI>40) exhibiting the highest prevalence of previous suicidal attempts (χ²=17.01, p<0.01) (Table 2; Figure 1D). However, no difference was observed with regard to hospitalization for suicide attempt during the study period between the different BMI groups (Table 2, Figure 1E). Finally, the number of hospitalizations over the course of the study period had a bimodal distribution, with higher overall hospitalization rates at either end of the BMI spectrum, but this finding did not reach statistical significance (χ²=12.43, p=0.053) (Table 2, Figure 1E).

TABLE 2.

Baseline affective disorder evaluations and their disease trajectories across observed BMI categories.

	BMI categories							F, χ2	P
	Underweight (<18.5)	Thin (18.5-20)	Normal weight (20-25)	Overweight (25-30)	Class I obesity (30-35)	Class II obesity (35-40)	Class III obesity (>40)
Affective disorders evaluation
Number of depressive episodes in the past year, mean (SD)	2.1 (2.9)	2.6 (2.8)	2.3 (3.3)	2.2 (3.5)	2.3 (3.4)	2.5 (4.0)	2.6 (3.5)	F = 0.34	0.56
History of suicide attempt, n (%)	24 (47.1)	31 (33.0)	279 (34.3)	308 (36.4)	201 (40.0)	95 (40.6)	82 (48.0)	χ2 = 17.01	<0.01
Received ECT, n (%)	1 (2.0)	1 (1.0)	47 (5.7)	52 (6.1)	49 (9.4)	17 (7.1)	13 (7.5)	χ2 = 15.12	0.019
Bipolarity index total score, mean (SD)	68.1 (18.9)	72.5 (17.3)	70.2 (16.2)	71.9 (15.8)	71.8 (16.3)	73.2 (15.6)	71.6 (15.4)	F = 4.88	0.027
Episode characteristics, mean (SD)	15.1 (4.9)	15.9 (5.0)	14.9 (5.4)	16.1 (5.1)	16.0 (5.1)	16.8 (4.8)	16.4 (5.2)	F = 23.46	<0.001
Age onset, mean (SD)	16.5 (2.5)	16.8 (3.1)	16.4 (3.4)	16.0 (3.8)	16.2 (3.7)	16.1 (3.5)	15.9 (3.0)	F = 4.90	0.027
Course of illness, mean (SD)	13.5 (6.8)	15.0 (5.0)	14.3 (5.4)	14.9 (5.2)	14.9 (5.2)	15.1 (5.3)	15.0 (5.0)	F = 6.13	0.013
Treatment response, mean (SD)	11.5 (7.6)	13.2 (6.5)	13.2 (6.3)	13.1 (6.3)	13.4 (5.8)	13.6 (5.9)	13.6 (5.9)	F = 3.46	0.063
Family history, mean (SD)	11.1 (7.3)	11.4 (7.4)	11.4 (7.1)	11.6 (6.8)	11.3 (6.9)	11.6 (7.2)	10.7 (7.2)	F = 0.48	0.49
Hospitalization during study, n (%)	10 (19.6)	16 (16.3)	141 (16.9)	147 (17.1)	105 (20.0)	52 (21.5)	47 (26.4)	χ2 = 12.43	0.053
Hospitalization by suicide during the study, n (%)	1 (2.0)	1 (1.0)	26 (3.1)	26 (3.0)	14 (2.7)	5 (2.1)	6 (3.4)	χ2 = 2.42	0.88

Abbreviations: SD: standard deviation; ECT: electroconvulsive therapy

With regard to comorbidities at baseline, no significant difference in history of depressive episodes (χ²=4.58, p=0.6) or number of depressive episodes in the past year was observed across BMI categories (F=0.34, p=0.56). However, number of manic episodes due to a general medical condition or substance-induced (χ²=13.77, p=0.032) and diagnosis of BD-II (χ²=14.09, p=0.029), panic disorder (χ²=17.64, p<0.01), social phobia (χ²=18.68, p<0.01), and post-traumatic stress disorder (PTSD) (χ²=17.9, p<0.01) all differed significantly across BMI groups (Table 3). When a graph-theoretical approach was used to map the relationships between comorbidities, median nodal strength—an aggregate measure of the number of co-occuring diagnoses—revealed a robust increase in psychiatric comorbidities in participants with BMIs>35. With regard to medical comorbidities, a linear increase across all BMI categories was observed, and the highest rates of medical comorbidity were seen in participants with BMIs>35 (Figure 2).

TABLE 3.

Lifetime psychiatric and current medical comorbidities and their trajectories across observed BMI categories.

	BMI categories							F, χ2	P
	Underweight (<18.5)	Thin (18.5-20)	Normal weight (20-25)	Overweight (25-30)	Class I obesity (30-35)	Class II obesity (35-40)	Class III obesity (>40)
Psychiatric comorbidities, lifetime, n (%)
Major depressive episode	46 (95.8)	91 (94.8)	773 (94.6)	784 (93.4)	486 (96.0)	222 (94.5)	163 (95.3)	χ2 = 4.58	0.6
Manic episode	34 (70.8)	58 (60.4)	489 (59.9)	561 (66.9)	330 (65.2)	162 (68.9)	114 (66.7)	χ2 = 13.77	0.032
Hypomanic episode	18 (37.5)	46 (47.9)	396 (48.5)	355 (42.3)	236 (46.6)	107 (45.5)	69 (40.4)	χ2 = 9.84	0.13
Bipolar Type II	8 (16.7)	23 (24.0)	213 (26.1)	167 (19.9)	105 (20.8)	44 (18.7)	32 (18.7)	χ2 = 14.09	0.029
Anxiety	15 (31.3)	25 (26.0)	186 (22.8)	197 (23.5)	117 (23.1)	53 (22.6)	52 (30.4)	χ2 = 6.72	0.35
Panic disorder	11 (22.9)	21 (21.9)	169 (20.7)	180 (21.5)	109 (21.5)	63 (26.8)	58 (33.9)	χ2 = 17.63	<0.01
Social phobia	12 (25)	22 (22.9)	171 (20.9)	202 (24.1)	100 (19.8)	51 (21.7)	59 (34.5)	χ2 = 18.68	<0.01
OCD	7 (14.6)	11 (11.5)	108 (13.2)	101 (12.0)	49 (9.7)	25 (10.6)	27 (15.8)	χ2 = 6.72	0.35
PTSD	10 (20.8)	23 (24.0)	156 (19.1)	156 (18.6)	94 (18.6)	56 (23.8)	53 (31.0)	χ2 = 17.92	<0.01
Alcohol dependence or abuse	19 (39.6)	49 (51.0)	344 (42.1)	386 (46.0)	221 (43.7)	99 (42.1)	75 (43.9)	χ2 = 5.23	0.51
Drug dependence or abuse	14 (29.2)	36 (37.5)	238 (29.1)	248 (29.6)	139 (27.5)	73 (31.1)	51 (29.8)	χ2 = 4.30	0.64
Smoker, n (%)	16 (31.4)	43 (43.9)	281 (33.7)	268 (31.1)	153 (29.1)	65 (26.9)	43 (24.2)	χ2 = 17.1	< 0.01
Psychotic disorder	4 (8.3)	10 (10.4)	42 (5.1)	54 (6.4)	28 (5.5)	20 (8.5)	14 (8.2)	χ2 = 8.40	0.21
Bipolar Type I with psychotic features	10 (20.8)	11 (11.5)	159 (19.5)	177 (21.1)	117 (23.1)	39 (16.6)	38 (22.2)	χ2 = 10.20	0.12
Bulimia nervosa	3 (6.3)	13 (13.5)	68 (8.3)	41 (4.9)	26 (5.1)	20 (8.5)	16 (9.4)	χ2 = 19.51	<0.01
Anorexia nervosa	8 (16.7)	11 (11.5)	30 (3.7)	26 (3.1)	6 (1.2)	5 (2.1)	2 (1.2)	χ2 = 58.82	<0.001
ADHD	10 (20.8)	13 (13.5)	147 (18.0)	148 (17.6)	88 (17.4)	42 (17.9)	33 (19.3)	χ2 = 1.85	0.93
Medication use (current), n (%)
Mood stabilizer	0 (0)	2 (2.1)	24 (2.9)	19 (2.2)	13 (2.5)	5 (2.1)	6 (3.4)	χ2 = 2.97	0.81
Stimulant	4 (7.8)	7 (7.2)	23 (2.8)	37 (4.3)	22 (4.3)	12 (5.0)	7 (4.0)	χ2 = 8.67	0.19
Antipsychotic (typical or atypical)	2 (3.9)	0 (0)	19 (2.3)	24 (2.8)	15 (2.9)	7 (2.9)	7 (4.0)	χ2 = 4.81	0.57
Anticonvulsant	0 (0)	0 (0)	6 (0.7)	4 (0.5)	10 (1.9)	1 (0.4)	4 (2.3)	χ2 = 13.97	0.03
Antidepressant (SSRI, SNRI, TCA, MAOI, atypical ADT)	1 (2.0)	6 (6.2)	28 (3.4)	32 (3.7)	12 (2.3)	9 (3.7)	6 (3.4)	χ2 = 4.82	0.57
Anxiolytic (benzodiazepine, beta-blockers, antihistamine)	1 (2.0)	6 (6.2)	65 (7.8)	86 (10.0)	65 (12.6)	40 (16.6)	20 (11.4)	χ2 = 26.32	<0.001
Complex Pharmacotherapy (≥ 2 psychotropic meds, %)	1 (2.0)	3 (3.1)	29 (3.5)	24 (2.8)	21 (4.1)	9 (3.7)	6 (3.4)	χ2 = 2.1	0.91

Abbreviations: OCD: obsessive compulsive disorder; PTSD: post-traumatic stress disorder; ADHD: attention deficit hyperactivity disorder; SSRI: selective serotonin reuptake inhibitor; SNRI: serotonin norepinephrine reuptake inhibitor; TCA: tricyclic antidepressant; MAOI: monoamine oxidase inhibitor; ADT: antidepressant therapy.

Note that full diagnoses with specifiers were made according to DSM criteria: depressive episode (major depressive episode, mood disorder due to a general medical condition, substance induced, or not otherwise specified); manic episode (due to a general medical condition, or substance induced); hypomania (due to a general medical condition, or substance induced); anxiety (generalized anxiety disorder, with panic attacks, or substance induced); panic disorder (with or without agoraphobia); obsessive compulsive disorder (OCD) (due to a general medical condition, or substance induced); psychotic disorder (due to a general medical condition, substance induced, delusional, schizophrenia, schizoaffective, schizophreniform, brief psychotic disorder); bipolar Type I with psychotic features; bulimia nervosa (purge or non-purge type); anorexia nervosa (binge eating, restricting type); attention-deficit hyperactivity disorder (ADHD) (child, adolescent, or adult).

Figure 2. Correlation metrics of the impact of body mass index (BMI) on psychiatric and medical comorbidities in the STEP-BD sample (n=2790).

Median nodal strength is a proxy for the whole network density of the comorbidities. The higher the density, the higher the network connections. Mapping comorbidities revealed a sharp increase in psychiatric comorbidities for those with BMIs above 30 kg/m². Patients in the underweight and overweight categories exhibited more comorbidities than patients within the 20–30 kg/m² weight ranges. For medical comorbidities, a continuous increase in asthma, migraine, and hypertension was observed.

When an additional analysis of MADRS sub-items from baseline visits controlled for sex, age, and med risk score was performed, results indicated significant group differences across BMI categories for ‘reduced sleep’ (F=3.04, p=0.0024), ‘concentration difficulties’ (F=2.14, p=0.033), ‘lassitude’ (F=5.6, p<0.001), and ‘inability to feel’ (F=2.68, p=0.0074) (Table 4). An increase in the ‘lassitude’ and ‘inability to feel’ items correlated with higher BMI, while the ‘concentration difficulties’ item had a bimodal distribution; specifically, it was higher in participants with BMIs<18.5 but increased linearly within the other BMI groups. The item ‘reported sadness’ also increased continously with BMI but only reached trendwise significance (Figure 3). When YMRS sub-items from baseline visits were analyzed, significant group differences across all BMI categories were observed for ‘irritability’ (F=2.2, p=0.028), ‘disruptive-aggressive behavior’ (F=2.96, p=0.0031), and ‘appearance’ (F=6.43, p<0.001) (Table 4). ‘Appearance’, ‘irritability’, and ‘disruptive-aggressive behavior’ had a bimodal distribution (Figure 4). Notably, our analysis also indicated that age, sex, and the use of medications associated with weight gain or weight loss correlated with differences in MADRS and YMRS items (Table 4).

TABLE 4.

Analysis of MADRS and YMRS sub-items from baseline visits controlled for sex, age, and med risk score.

MADRS	BMI		Age		Sex		Med risk score*
	T	P	T	P	T	P	T	P
Apparent sadness	0.91	0.36	0.31	0.76	1.79	0.073	−2.73	0.0064
Reported sadness	1.84	0.067	−1.32	0.19	0.36	0.72	−3.58	0.0003
Inner tension	0.47	0.64	−3.59	0.0003	1.61	0.11	−3.54	0.0004
Reduced sleep	3.04	0.0024	−1.51	0.13	3.38	0.0007	−1.43	0.15
Reduced appetite	−1.79	0.073	−3.22	0.0013	0.56	0.57	−0.28	0.78
Concentration difficulties	2.14	0.033	−2.35	0.019	3.83	0.0001	−3.74	0.0002
Lassitude	5.60	<0.0001	−1.82	0.068	3.81	0.0001	−1.44	0.15
Inability to feel	2.68	0.0074	−2.76	0.0058	0.30	0.76	−2.09	0.037
Pessimistic thoughts	1.43	0.15	−1.82	0.069	0.034	0.97	−3.72	0.0002
Suicide thoughts	1.10	0.27	−0.24	0.81	−0.17	0.86	−2.73	0.0064
YMRS
Elevated mood	−0.30	0.77	−3.23	0.0013	0.49	0.62	−3.09	0.002
Increased motor activity-energy	−0.45	0.66	−4.64	<0.0001	0.42	0.67	−3.86	0.0001
Sexual interest	−0.54	0.59	−2.40	0.016	−2.28	0.023	−1.47	0.14
Sleep	1.83	0.068	−2.06	0.039	2.48	0.013	−1.44	0.15
Irritability	2.20	0.028	−5.41	<0.0001	2.41	0.016	−4.14	<0.0001
Speech	0.49	0.62	−0.88	0.38	1.63	0.10	−5.08	<0.0001
Language-thought disorder	−0.45	0.65	−4.23	<0.0001	2.27	0.024	−4.05	0.0001
Content	−0.50	0.61	−1.19	0.24	−0.068	0.85	−2.86	0.0043
Disruptive-aggressive behavior	2.96	0.0031	−4.87	<0.0001	1.53	0.13	−3.55	0.0004
Appearance	6.43	<0.0001	0.038	0.97	0.50	0.62	0.19	0.85
Insight	0.65	0.52	1.15	0.25	−2.09	0.037	0.017	0.99

^*Med risk score: Calculated sum score of +1 for medications associated with weight gain, zero for those that did not affect weight, and −1 for medications associated with weight loss.

MADRS: Montgomery-Åsberg Depression Rating Scale; YMRS: Young Mania Rating Scale.

Figure 3. Individual baseline Montgomery–Åsberg Depression Rating Scale (MADRS) items with body mass index (BMI) in the STEP-BD sample (n=2790).

Center: Radar plot of individual MADRS items across all participants at baseline according to standardized BMI groups. Bar charts: Significant differences across BMI groups were found for ‘reduced sleep’, ‘concentration difficulties’, ‘lassitude’, and ‘inability to feel’ (see Table 4). The red, yellow, and blue arcs denote grouping of sub-items resulting from the exploratory factor analysis. The red factor, which comprised ‘reported sadness’, ‘pessimistic thoughts’, ‘apparent sadness’, ‘suicidal thoughts’, ‘inability to feel’, and ‘lassitude’, was labeled as “Core depressive symptoms” and used for the subsequent longitudinal analysis, as described in Figure 5.

Figure 4. Individual baseline Young Mania Rating Scale (YMRS) items with body mass index (BMI) in the STEP-BD sample (n=2790).

Center: Radar plot of individual YMRS items across all participants at baseline according to standardized BMI groups. Bar charts: Significant differences across BMI groups were found for ‘irritability’, ‘disruptive aggressive behavior’, and ‘appearance’ (see Table 4). The red, yellow, and blue arcs denote grouping of subitems resulting from the exploratory factor analysis. The red factor, which comprised ‘increased motor activity and energy’, ‘elevated mood, speech, sleep and sexual interest’, was labeled as “Affective instability” and used for the subsequent longitudinal analysis, as described in Figure 5.

Figure 5 further assesses the trajectory clustering analysis of core depressive and affective instability symptoms derived from the MADRS (Panel A) and the YMRS (Panel B) over 36 months of the STEP-BD study. Participants were divided into four clusters based on how symptoms changed over time. In the most severe groups, depressive or affective instability symptoms were high at baseline, and little or no improvement was seen over time. In the mildest groups, symptoms of depression or affective instability improved quickly or were typically low throughout the study (see Figure 5 for additional details regarding each cluster). The analysis indicated that the percentage of individuals with a milder trajectory of depressive symptoms decreased as BMI increased. As regards affective instability symptoms, the group with BMIs>40 had the smallest percentage of participants with a milder symptom trajectory.

Figure 5.

Trajectory clustering analysis of (PANEL A) core depressive symptoms and (PANEL B) affective instability symptoms over 36 months. (PANEL A) Four clusters of trajectories were observed for the factors ‘core depressive symptoms’ derived from the factor analysis. Panels on the right summarize the proportion of patients in each cluster according to BMI. The dark purple cluster (A) started with high symptom scores at the initial visit and maintained high symptom scores over time. Patients in the second cluster (B) in lighter colors exhibited high symptom scores throughout the study, whereas patients in lightest purple colors improved quickly (C) or had low symptom scores (D) throughout the study. Note that the percentage of patients with low symptom scores decreased as BMI increased. (PANEL B) Four clusters of trajectories were observed for ‘affective instability’ items in the factor analysis. Patients in the darkest green cluster (A) started with high symptom scores at the initial visit and improved little over time. Patients in the next lighter green cluster (B) started with low symptom scores at the initial visit but worsened over time. Patients in the lightest green clusters started with high core affective instability symptoms (C) at the initial visit and improved rapidly or had low affective instability (D). Again, the percentage of patients with low ‘affective instability’ decreased as BMI increased.

Discussion

This study used data drawn from the longitudinal STEP-BD study to assess whether higher BMI affects symptomatology, severity, disease course, and disease burden in individuals with BD. BMI was found to affect several key variables, including severity of BD symptoms, increased risk of suicide attempts, number of medical and psychiatric comorbidities, and history of ECT trials.

In particular, the present study found that overweight and obese patients with BD had worse scores on the bipolarity index, a questionnaire measuring disease severity that was administered as part of the baseline assessment. Higher BMI was associated with history of greater likelihood of episodes with hypomanic and mixed characteristics as well as history of higher recurrence rates. Furthermore, BMI was correlated with changes in prevalence for several psychiatric comorbidities; specifically groups with greater BMIs had a higher prevalence of panic disorder, social phobia, OCD, and PTSD. Manic episodes due to medical problems or induced by substance use were more prevalent in underweight populations. However, correlational analyses indicated an increase in general psychiatric comorbidities in individuals with BMIs>35. Overall, a general increase in medical and psychiatric comorbidities was observed that reflects impaired health status risk for individuals in high BMI groups, though it should be noted that comorbidity analyses were not controlled for medications associated with weight gain. As noted in Table 4, sex, age, and pharmacological interventions also impacted many of the symptom domains measured, reinforcing the need for individualized treatment approaches.

The study also found that individuals with higher BMIs were more likely to have a history of treatment with ECT. Because BD treatment guidelines have historically supported ECT use in patients refractory to pharmacological treatment⁴¹, this finding suggests that higher BMI is associated with a more severe course of BD. With regard to hospitalization rates over the course of the study, the findings suggest a non-statistically significant trend of increasing hospitalizations for individuals with BD and extremely low or high BMI; however, when underweight patients were excluded, an increased rate of hospitalizations during the study was observed as a function of BMI. Interestingly, a previous study similarly found that obese patients with BD were more likely to be hospitalized for depression than non-obese patients, but it did not assess hospitalization rates in patients with low BMI¹².

The MADRS items ‘lassitude’ and ‘inability to feel’ linearily increased with higher BMI. The ‘concentration difficulties’ item had a bimodal distribution, which echoes the literature indicating neurocognitive problems in both extremes of the BMI spectrum^42,43. Interestigly, there was no difference regarding prevalence or frequency of depressive episodes when comparing different BMI groups. Taken together, these results suggest that higher BMI may correlate with worsening severity, rather than frequency, of depressive symptoms in individuals with BD. On the YMRS, BMI was associated with the ‘irritability’, ‘disruptive/aggressive behavior’, and ‘appearance’ items, which displayed bimodal distributions and increases at both ends of the BMI spectrum. While intriguing, these exploratory results need to be investigated with more sophisticated tools than sub-items of standardized mood questionnaires. Nevertheless, the findings could be used to guide future, targeted investigations of symptoms that worsen alongside weight gain, and they identify important symptoms that should be monitored in individuals with BD experiencing weight gain.

This study also investigated the trajectory of affective instability symptoms (derived from a scale to assess mania) and core depressive symptoms within different BMI groups over 36 months of the STEP-BD study. The results indicate that, as BMI increased, the proportion of individuals with milder core depressive symptoms decreased. The analysis also revealed a smaller percentage of BD patients with mild affective instability symptom trajectories in the group with BMIs>40. These findings support previous studies indicating a more severe disease trajectory in obese patients with BD¹². Interestingly, a significant association was observed between the use of medications associated with weight gain and some symptom domains assessed by the MADRS and YMRS that were not associated with BMI; these included ‘apparent sadness’, ‘reported sadness’, ‘inner tension’, ‘pessimistic thoughts’, ‘suicidal thoughts’, ‘elevated mood and energy’, ‘speech’, and ‘language-thought disorder’. This suggests that iatrogenic weight gain is not the only factor associated with worsening symptoms and outcomes.

In addition, the present study found a correlation between higher BMI and increased prevalence of past suicide attempts, with the highest prevalence found in the group with BMIs>40. However, the interaction between baseline BMI and suicide-related hospitalization over the 36-month course of the study period was not significant. These results corroborate findings from a previous outpatient study that found that obese patients with BD were more likely to have a previous suicide attempt than those with normal BMI⁴⁴.

Although largely speculative at this stage, a number of putative mechanisms suggest a link between BMI, suicide attempt, and BD, including, but not limited to, factors such as obesity, overactivation of the hypothalamic pituitary adrenal (HPA) axis, and heightened inflammatory and metabolic state ^11,44. Other factors such as lower serum levels of leptin—an adipocyte hormone that inhibits food intake and increases energy expenditure—have also been implicated in both suicide attempt and obesity⁴⁵. Results from the present study add important context to previously mixed evidence in BD and suggest that overweight and obesity should be an important consideration for clinicians treating BD.

As noted above, previous studies found that BMI contributed to significantly increased levels of inflammation that, in turn, worsened disease course in BD^25,30. Accumulation of adipose tissue—considered an endocrine organ—leads to constant low-grade inflammation throughout the peripheral and central nervous systems. In BD, altered cytokine levels throughout symptomatic and asymptomatic periods have been linked to obesity and other comorbid disorders such as coronary artery disease and insulin resistance⁴⁶. These shifts could account for the increased likelihood that those with higher BMIs experience depressive relapse³⁰. Further research is needed to determine the longitudinal impact of changing cytokine levels on worsening symptom patterns.

Another finding of interest is that sex differences were present in many of the MADRS (‘reduced sleep’, ‘concentration difficulties’, ‘lassitude’) and YMRS (‘irritability’) items. Previous studies have similarly noted sex differences in BD, particularly regarding the frequency of manic episodes⁴⁷ and rates of obesity⁴⁸. Sex differences are also prevalent in states that may contribute to higher BMI and certain BD symptoms, such as inflammation and microbiome composition⁴⁹. Further research is needed to determine the exact role of sex and its interaction with BMI for these subsets of symptoms. Significant age differences were also noted for all measures associated with BMI except for ‘appearance’. Unfortunately, little research exists regarding the impact of aging or sex differences on BMI and BD, but changes in psychosocial factors such as increased likelihood of exposure to stressful events or biological events, as well as shifts in hormone levels, could help explain such differences^50,51.

This study also found that smoking remains a significant risk factor and impacts BMI (see Table 3). Interestingly, a previous STEP-BD study⁴⁰ found that cigarette smoking was greatly elevated in the STEP-BD cohort and appeared to be associated with increased risk of suicide attempts, suicide measures, and impulsiveness. In addition, prospective findings from a STEP-BD ancillary study found that observed suicide attempts were more frequent in a group of smokers compared to a group that did not smoke⁵². Thus, while the significant adverse medical consequences of smoking are already well known, STEP-BD data suggest that its association with poor outcome in BD also needs to be appreciated.

Strengths of the current study include the large sample size, real-world data applicability, long-term continuity of care, the use of rigorous statistical methods, and the thoroughness of data collection. However, several key limitations should also be noted. First, because some of the data were obtained in a cross-sectional matter, causality cannot be inferred from this study design. Future research should investigate longitudinal or prospective outcomes in order to verify our cross-sectional findings. Second, this was an exploratory analysis, and some of the sub-items derived from mood questionares require more sophisticated tools in order to be fully substantiated. Third, because body weight and BMI were collected at the beginning of the study, the analysis did not account for fluctuations in an individual’s weight over the course of the STEP-BD study. Fourth, because the study was conducted in 22 research centers across the United States, the generalizability of these findings to other countries needs to be confirmed by similar longitudinal international studies. Fifth, BMI cannot account for the location of adiposity (subcutaneous vs. visceral) or distribution of fat, both of which differentially impact inflammatory markers such as C-reactive protein (CRP), interleukin-6 (IL-6), and tumor necrosis factor alpha (TNF-α)^53-55. For instance, another study found that visceral adipose tissue was a key mediator between inflammation, depression, and obesity⁵⁶. Although waist circumference or waist-to-hip ratio may be better predictors of inflammation, these data were unfortunately not available for the STEP-BD dataset. Future research should collect multiple indicators of weight distribution in order to further assess the relationship between weight, inflammation, and BD. Finally, comorbidity analyses did not control for medications associated with weight gain. This is particularly important because previous studies have shown that rates of comorbid conditions—including metabolic syndrome, obesity, and diabetes as well as lifetime drug and alcohol use disorders—are significantly higher in individuals with BD. This and other factors may encompass iatrogenic risks that may, in themselves, be independent risk factors for obesity⁵⁷. However, apart from anxiolytics and anticonvulsants, no differences were noted in prescribing patterns across BMI categories, suggesting that these did not vary much between BMI categories at study inclusion.

Taken together, these findings indicate that higher BMI affected disease course and severity of BD. Specifically, higher BMI at baseline assessment was linked to more serious presentation of mood symptom domains, increased prevalence of previous suicide attempts, increased history of ECT treatments, and other detrimental impacts on mental and physical comorbidities. Some of these changes—such as disease burden associated with multiple comorbidities—were bidirectional, with more drastic presentations in both underweight and overweight BD participants, suggesting that either BMI extreme poses increased risk. The differential impact of underweight and overweight BMI categories on BD symptom presentation suggests that divergent biological mechanisms and social factors, as well as an increase in medical comorbidities in both of these categories, likely cause bidirectional changes that contribute to worsening outcomes in BD. These findings reinforce the importance of accounting for body weight when treating individuals with BD, underscore the importance of addressing overweight and obesity as a factor in mental health, and support personalized approaches in neuropsychiatric treatment.

Data Availability Statement

Data used in the preparation of this article were obtained from the limited access datasets distributed from the NIH-supported STEP-BD study. The ClinicalTrials.gov identifier is NCT00012558 (https://clinicaltrials.gov/ct2/show/NCT00012558).