Weight change in the acute treatment of bipolar I disorder: A naturalistic observational study of psychiatric inpatients

1. Introduction 

Obesity and overweight are associated with increased mortality, cardiovascular disease, type II diabetes, osteoarthritis, and some forms of cancer (Wyatt et al., 2006). Patients with bipolar disorder show increased prevalence rates of overweight and obesity. The prevalence of obesity in patients with bipolar disorders has been reported to range from 20% to 35%, exceeding that in the general population (Elmslie et al., 2001, Elmslie et al., 2000, Fagiolini et al., 2002, Fagiolini et al., 2003, Mokdad et al., 2003). Moreover, individuals with bipolar disorder have a greater prevalence of diabetes mellitus compared with the general population, as much as threefold higher (Cassidy et al., 1999, Lilliker, 1980). Recent study showed that 30% of the bipolar disorder patients met the criteria for metabolic syndrome, and those patients were more likely to report a life time history of suicide attempt (Fagiolini et al., 2005).

The increased use of atypical antipsychotics in both acute and maintenance treatment could exacerbate weight gain and obesity in bipolar patients (Allison and Casey, 2001). Furthermore, combinations of atypical antipsychotics and mood stabilizers have been used to attain remission, to prevent relapse, and to control refractory illness (Frye et al., 2000, Guille et al., 2000), and these combination therapies can increase the incidence of overweight and obesity in bipolar patients.

Weight gain during acute treatment may predict long-term increases in body weight and obesity (Hennen et al., 2004). In patients with bipolar or mixed mania, substantial weight gain after 30 weeks of treatment with olanzapine was predicted by weight increases of 2–3 kg within the first 3 weeks of treatment (Lipkovich et al., 2006). Bipolar patients are at risk for weight gain, particularly during acute treatment (Fagiolini et al., 2002), and this weight gain can itself increase the risk of developing diabetes, hypertension, and coronary heart disease (Colditz et al., 1995, Kawachi, 1999, Rimm et al., 1995, Willett et al., 1995). Moreover, past history of weight gain and loss is a significant risk factor for cardiovascular disease (Lissner et al., 1991). To prevent the morbidity and mortality associated with weight gain in patients with bipolar disorder, it is important to avoid weight gain during short-term treatment for bipolar disorder.

Obesity has been shown to be a multifactorial phenotype influenced by multiple genes and environmental components (Barsh et al., 2000). It is difficult to delineate the factors contributing to weight gain in patients with bipolar disorder; for example, how much of the weight gain in a particular subject was due to medications, the bipolar disorder itself (e.g. genetics, innate energy metabolisms), or environment (e.g. food intake and physical activity). Data on weight change during acute treatment is mostly derived from clinical trials. Although these studies provide valuable information on weight gain and obesity as an adverse event, they may be limited in gauging the extent of weight change and related factors in actual clinical situations, which differ from those of controlled trials. Thus, naturalistic observational studies of large numbers of patients are required to assess weight gain and obesity during acute treatment for bipolar disorder. To identify the extent of acute weight change and its clinical correlates during acute treatment of bipolar disorder, we have conducted a naturalistic observational study during the first 4 weeks of treatment in a closed-ward hospital setting.

2. Methods 

3. Results 

3.1. Weight change and clinical variables 

During the study period, 322 patients with bipolar I disorder were consecutively hospitalized for treatment of acute mania. Of these, 179 patients were included in this naturalistic study; the other 143 patients were not included in the study because of nonconformance to inclusion criteria and/or unavailability of relevant information. The correlations of demographic and clinical characteristics with weight change after 4 weeks of hospitalization are presented in Table 1. The mean age of patients was 33.1±13.2 years, and 91 (50.8%) were male. Overall weight increase after 4 weeks was 2.7±3.0 kg (F=97.901, P<0.001). Of the various demographic and clinical variables, age (r=−0.164, P=0.029), marital status (single 3.5±3.0; divorced and separated 2.6±3.0; married 1.7±2.6, F=6.385, P=0.002), and DOI (r=−0.189, P=0.011) were correlated with weight change. Following stepwise linear regression analysis, only DOI remained significant (change of weight=−0.073⁎DOI+3.161, P=0.011).

Table 1. Changes in weight after 4 weeks hospitalizations of bipolar disorder according to demographic and clinical characteristics

			N=179 (%)	Change in weight (kg, mean±S.D.)	Statistics	P-value
	Demographic characteristics
	Age (year)	33.1±13.2			r=−0.164	0.029
	Sex	M	91 (50.8)	3.1±3.4
		F	88 (49.2)	2.4±2.6	t=0.131	0.687
	Marital status	Married	67 (37.4)	1.7±2.6
		Single	93 (52.0)	3.5±3.0
		Divorced, separated	19 (10.6)	2.6±3.0	F=6.385	0.002
	Education(year)	13.5±3.1			r=0.117	0.120
	Occupations	Employed	122 (68.2)	2.5±3.0
		Unemployed	57 (31.8)	3.3±3.1	t=1.585	0.115
	Smoking	Smoker	44 (24.6)	2.8±3.6
		Non-smoker	135 (75.4)	2.7±2.8	t=−0.572	0.568
	SES	High	27 (15.1)	3.3±3.6
		Middle	114 (63.7)	2.4±2.7
		Low	38 (21.2)	3.4±3.5	F=2.414	0.092
	Clinical characteristics
	Age of onset (year)	27.5±11.2			r=−0.072	0.341
	Duration of illness (year)	5.6±7.9			r=−0.189	0.011
	Previous admission (number)	1.4±0.9			r=−0.104	0.167
	Previous medications	Yes	114 (63.7)	2.7±3.0
		No	65 (36.3)	2.8±3.1	t=0.222	0.824
	CGI-I (score)	1.65±0.7			r=0.083	0.269
	Baseline weight/BMI
	Weight at baseline (kg)	62.8±11.6			r=0.076	0.314
	BMI at baseline (kg/m2)	22.8±3.3			r=−0.049	0.517
	Degree of obesity at baseline	Normal	102 (57.0)	2.7±3.1
		Overweight	33 (18.4)	2.8±3.6
		Obesity	44 (24.6)	2.9±2.5	F=0.114	0.892
	Overall			2.7±3.0	F=97.901⁎	<0.001

Linear regression with Age, Marital status, and Duration of illness (DOI), stepwise: change of weight=−0.073⁎DOI+3.161, p=0.011. ⁎Repeated measured ANOVA.

At baseline, 44 patients were obese (24.6%); after 4 weeks, 65 subjects (36.3%) were obese, 32 (17.9%) were overweight, 82 (45.8%) were normal weight (Table 2). Forty-five (25.1%) met the criteria of SWG. Among the demographic and clinical characteristics assessed, only age differed significantly between patients with and without SWG (t=14.182, P<0.001); rest of variables except age were not significant (data not shown). Although baseline weight, BMI, and degree of obesity were not significantly correlated with weight change (Table 1), these baseline variables differed significantly among the three subgroups of patients, assorted according to degree of obesity after 4 weeks of acute phase treatment (Table 2). That is, patients who were obese after 4 weeks showed higher baseline weight and BMI than overweight and normal BMI patients. Of the patients with normal weight at baseline, 3.9% became obese after 4 weeks, compared with 54.5% of overweight patients and 97.7% of obese patients at baseline. When we compared subjects who were overweight or obese at baseline with those of normal weight, we found no statistically significant differences among the other demographic and clinical variables, including age, sex, marital status, education, employment, smoking, SES, age of onset, duration of illness, number of previous hospitalizations, history of previous medications, and overall improvement of acute manic symptoms, as defined by the CGI-I (data not shown).

Table 2. Comparing baseline weight, BMI and degree of obesity according to patients with overweight and obesity after 4 weeks hospitalization

			Normal BMI (N=82)	Overweight (N=32)	Obesity (N=65)	Statistics	P-value
	Baseline weight (kg)		55.5±7.9	64.1±7.4	71.4±11.0	F=55.868⁎	<0.001
	Baseline BMI (kg/m2)		20.1±1.9	23.1±1.1	25.9±2.5	F=142.578⁎	<0.001
	Degree of obesity at baseline (N)	Normal (102)	80 (78.4%)	18 (17.7%)	4 (3.9%)
		Overweight (33)	2 (6.1%)	13 (39.4%)	18 (54.5%)
		Obesity (44)	0 (0%)	1 (2.3%)	43 (97.7%)	χ2=147.97	<0.001

⁎ANCOVA, DOI as covariate; log transformations for Baseline weight and BMI were performed.

3.2. Weight change and treatment regimens 

Table 3 shows changes in weight after 4 weeks according to treatment regimens. The most frequently prescribed treatment regimen was a combination of olanzapine and valproate (32 patients, 17.9%); after 4 weeks, these patients experienced a mean weight increase of 3.8±2.9 kg, a greater increase than those experienced by patients treated with olanzapine and lithium (3.3±3.1 kg), risperidone and lithium (3.1±2.7 kg), risperidone and valproate (2.5±3.0 kg), lithium monotherapy (2.2±2.9 kg), haloperidol and lithium (2.0±3.1 kg), olanzapine monotherapy (1.6±3.5 kg) and valproate monotherapy (0.5±2.1 kg). Overall, patients on any kind of atypical antipsychotic showed greater weight gain than those on typical antipsychotics or without antipsychotics (Table 4). Combination treatment with antipsychotics and mood stabilizer resulted in greater weight gains than monotherapy with an antipsychotic or mood stabilizer. There were no significant differences in weight change between patients prescribed or not prescribed a mood stabilizer. Differential changes in weight after 4 weeks in patients receiving combination therapies are shown in Table 5. Olanzapine plus valproate (3.8±2.9 kg) and olanzapine plus lithium (3.3±3.1 kg) showed greater weight gains than olanzapine monotherapy (1.6±3.5 kg), but these differences were not statistically significant (F=1.024, P=0.367). Valproate plus olanzapine (3.8±2.9 kg) and valproate plus risperidone (2.5±3.0 kg) showed significantly greater weight gains than valproate monotherapy (0.5±2.1 kg, F=5.786, P=0.005). Lithium monotherapy and lithium combined with olanzapine, risperidone, or haloperidol did not show statistically significant differences in weight change (F=0.823, P=0.486).

Table 3. Change in weight after 4 weeks hospitalizations according to medication(s) frequently used

		N=179 (%)	Change in weight (kg, mean±S.D.)	Statistics⁎	P-value⁎⁎
	Olanzapine+valproate	32 (17.9)	3.8±2.9	F=5.989	0.015
	Risperidone+lithium	22 (12.3)	3.1±2.7	F=0.195	0.660
	Risperidone+valproate	16 (8.9)	2.5±3.0	F=0.099	0.753
	Olanzapine+lithium	15 (8.4)	3.3±3.1	F=0.306	0.581
	Haloperidol+lithium	14 (7.8)	2.0±3.1	F=1.661	0.199
	Lithium	13 (7.3)	2.2±2.9	F=0.390	0.533
	Valproate	9 (5.0)	0.5±2.1	F=5.688	0.018
	Olanzapine	6 (3.4)	1.6±3.5	F=0.290	0.591
	Others	52 (29.1)	–	–	–

⁎ANCOVA, DOI as covariate. ⁎⁎P value was derived from t-statistic based on the change in weight according to each type medication(s) versus all others combined.

Table 4. Change in weight after 4 weeks hospitalizations according to with or without antipsychotics, mood stabilizers, and any kind of combinations

		N=179 (%)	Change in weight (kg, mean±S.D.)	Statistics⁎	P-value
	Typical antipsychotics	24 (13.4)	2.0±3.2
	Atypical antipsychotics	128 (71.5)	3.2±3.0
	Without antipsychotics	27 (15.1)	1.4±2.5	F=5.486	0.005
	Mood stabilizers	158 (88.3)	2.8±3.0
	Without mood stabilizers	21 (11.7)	2.3±3.2	F=0.288	0.592
	Monotherapy	44 (24.6)	1.8±2.9
	Any combinations	135 (75.4)	3.0±3.0	F=4.644	0.033

⁎ANCOVA, DOI as covariate.

Table 5. Change in weight after 4 weeks hospitalizations according to combination therapies

			N	Change in weight (kg, mean±S.D.)	Statistics⁎	P-value
	Olanzapine	Monotherapy	6	1.6±3.5
		+valproate	32	3.8±2.9
		+lithium	15	3.3±3.1	F=1.024	0.367
	Valproate	Monotherapy	9	0.5±2.1
		+olanzapine	32	3.8±2.9
		+risperidone	16	2.5±3.0	F=5.786	0.005
	Lithium	Monotherapy	13	2.2±2.9
		+olanzapine	15	3.3±3.1
		+risperidone	22	3.1±2.7
		+haloperidol	14	2.0±3.1	F=0.823	0.486

⁎ANCOVA, DOI as covariate.

4. Discussion 

Obesity has been correlated with poor outcome in patients with bipolar I disorder (Fagiolini et al., 2003). Since acute weight gain during acute treatment may predict overweight and obesity during maintenance treatment, as well as medical mortality and morbidity, it is important to determine the factors associated with weight gain and subsequent obesity. We have therefore assessed weight gain during acute phase treatment in 179 bipolar patients in a closed-ward hospital setting.

Several studies have reported correlates of obesity in bipolar disorder. For example, in a study of 644 outpatients with bipolar disorder, gender, geographical location, comorbid binge-eating disorder, age, income level, and comorbid medical disorders were found to be clinical correlates of overweight and obesity (McElroy et al., 2002). In bipolar patients entering maintenance therapy, the clinical and demographic characteristics related to obesity included years of education, number of previous depressive and manic episodes, baseline scores on the depression scale, and duration of acute episode (Fagiolini et al., 2003). Increase in BMI during acute and maintenance treatment, however, was not related to age, ethnicity, age at first mood episode, duration of illness, education, marital status, employment, or number of previous manic episodes (Fagiolini et al., 2002). It is difficult to directly compare our findings with those of these earlier studies, since we analyzed data from patients consecutively admitted to a tertiary hospital for the treatment of acute mania and evaluated short-term weight changes. We found that correlates of weight change during acute treatment were age, marital status, and DOI, with the latter remaining significant after linear regression analysis, although its correlation was not robust (correlation coefficient, −0.19). We did not, however, observe significant correlations between weight change and gender, socioeconomic status, number of previous admissions, and history of previous psychotropic medications. When we compared patients with and without SWG, we found that only age differed significantly. Taken together, these results indicate that weight changes in bipolar patients during acute treatment may not be largely affected by clinical and demographic factors.

Initially, 24.6% of patients were obese, while, after 4 weeks of treatment, 36.3% were obese; these findings are comparable to previously reported statistics (Fagiolini et al., 2002, Fagiolini et al., 2003). In Korean general population, the prevalence of obesity was reported to be 31.8% (Korean Ministry of Health and Welfare 2002). Taking into account this data, obesity rate after 4 weeks of hospitalization was higher than that of the general population while the rate at baseline was lower than that of the general population of Korea. When we compared rates of obesity at baseline and after treatment, we found they differed by 11.7%. Most obese patients at baseline remained obese, while 3.9% of patients with normal weight at baseline and 54.5% of initially overweight patients became obese after 4 weeks of treatment. Interestingly, we found that weight gain was independent of degree of obesity at baseline, in that patients with normal weight at baseline (57%) gained 2.7 kg; those who were overweight (18.4%) gained 2.8 kg; and those who were obese (24.6%) gained 2.9 kg. This finding differs from previous results, showing that obese individuals do not experience significant weight gain during acute treatment (Fagiolini et al., 2002).

Drug-induced changes in food preference can lead to excessive energy intake, largely as a result of a high intake of sucrose (Elmslie et al., 2001). Our study results, derived from a closed-ward hospital setting where diet was controlled by a central unit and in which all patients received the same diet, suggest that these factors might not be decisive during acute treatment in hospital settings. Bipolar patients may have an intrinsic vulnerability to gain weight regardless of environmental factors, as suggested by findings showing an intrinsic relationship between abnormal glucose metabolism and bipolar I disorder, based on the increased prevalence of type 2 diabetes mellitus, independent of the effects of age, race, gender, medication, and body mass (Regenold et al., 2002).

We did not observe a significant correlation between weight change and clinical improvement, measured by CGI-I score, over time. Thus, the suggestion that excessive weight gain is a necessary consequence of symptom improvement seems irrelevant in clinical practice. Psychotropic medications have been thought to play a substantial role in weight gain during acute and maintenance treatment of bipolar I disorder. Patients treated with antipsychotic drugs were more obese than patients not receiving these drugs and their body fat was more centrally distributed (Elmslie et al., 2000). Data from controlled clinical trials of mood stabilizers and antipsychotics have shown that weight gain is one of the most frequent reasons for non-adherence, particularly for patients taking atypical antipsychotics. In addition to the latter, other drugs, including lithium, valproate, and carbamazepine, have all been associated with weight gain (Keck and McElroy, 2003). Although randomized controlled trial data can provide important information from which to determine risks and extent of weight gain attributable to specific medications, they cannot determine weight gain in actual clinical practice, due to the diversity of treatment regimens, including the frequent use of combinations of antipsychotics and mood-stabilizers. Weight gain associated with atypical antipsychotics may be even more important in bipolar disorder than in schizophrenia, due in part to the use of mood stabilizers that contribute to weight gain (Guille et al., 2000). In our naturalistic treatment setting, we found that the combination of olanzapine and valproate was most frequently used in our study subjects and showed the highest weight gain, higher than the sum of each individually, suggesting that combination therapy may synergistically enhance the potential for weight gain of each individual medication.

In a pooled analysis of clinical trials for acute mania, mean changes in weight among patients treated with risperidone alone or in combination with mood stabilizers ranged from +0.1 to 1.7 kg, whereas patients treated with olanzapine alone or in combination gained from +1.65 to +4.0 kg (Chue and Kovacs, 2003). The combination of olanzapine and divalproex or lithium was associated with a weight gain of 2 kg over 18 months, compared with a loss of 1.8 kg in patients receiving divalproex or lithium monotherapy (Tohen et al., 2004). Moreover, 26% of patients treated with olanzapine combination therapy gained at least 7% of baseline weight, compared with 6% of those treated with olanzapine monotherapy (Hennen et al., 2004). In a retrospective study of 50 consecutive treatment trials of bipolar patients receiving risperidone, olanzapine, or clozapine as an add-on to standard mood stabilizers, weight gain with olanzapine (4.0 kg) was significantly greater than with risperidone (2.4 kg) for 4 weeks, and the combination of divalproex and atypical antipsychotics caused more weight gain than did the combination with lithium (Guille et al., 2000). We found that the combination of olanzapine with valproate or lithium led to greater weight gains than olanzapine monotherapy. Although it is difficult to compare our results with those of previous studies, our findings suggest that weight gain during acute treatment is due, at least in part, to the frequent use of combinations to rapidly control acute symptoms.

Our study had several limitations. First, its duration (4 weeks) is relatively brief, especially if weight gain is time-dependent. Second, several important variables previously reported to be associated with overweight and obesity in bipolar disorder were not fully assessed in our study, such as the number of previous depressive episodes and dosage of medications. Thirdly, we assumed that the closed-ward setting would minimize inter-individual differences in energy intake and expenditure. However, we did not objectively measure individual food intake and level of physical activity, thus it is difficult to state clearly how much food intake and physical activity were controlled. Even so, our study was performed in the closed-ward setting, where inter-individual dietary differences and level of physical activity were probably not so significant in comparison to previous studies of bipolar outpatients (Elmslie et al., 2001, Elmslie et al., 2000, McElroy et al., 2002). Fourthly, we could not exclude the possibility that the choice of a specific treatment for a specific patient had influenced the outcome because the treatment medication assignment was not randomized. For instance, patients with a higher risk of gaining weight might have been taking medications with a lower risk of weight gain. Lastly, our study results may not be generally applicable, in that our study subjects were patients in acutely manic episodes with at least moderate severity of symptoms and they were recruited from one specific tertiary hospital. In addition, the propensity to overweight or obesity may vary by ethnicity.

In summary, we have addressed weight gain during acute treatment in patients with bipolar disorder and its clinical correlates. Our findings were consistent with results suggesting that bipolar disorder is often accompanied by significant weight increase during the acute phase treatment, due in part to the use of atypical antipsychotics and combination treatments. In addition, significant weight gains during acute treatment may not be robustly affected by clinical and demographic variables, including baseline weight and BMI. That is, every bipolar patient has the propensity to gain weight in clinical practice, particularly during acute treatment. The association between weight gain and certain medications has remained an important concern in patients with bipolar disorder, especially those receiving combination treatments. Among treatment regimens, the combination of olanzapine and valproate was associated with the greatest weight gain, and the mean change of weight exceeded the sum of each individual medication. The specific mechanisms of interaction by which mood stabilizers and atypical antipsychotics induce weight gain are unclear, especially given the limited synaptic activity of mood stabilizers (Stanton, 1995). In addition, longer-term studies are required to determine the extent of weight gain and its correlating attributes, especially in the context of diverse treatment regimens in real-world clinical practice.