Lifestyle behaviors, metabolic disturbances, and weight gain in psychiatric inpatients treated with weight gain-associated medication

A naturalistic longitudinal observational study was performed between 2014 and 2017 at the Department of Psychiatry and Psychotherapy, University Hospital, LMU Munich, Germany. Baseline was defined by the start of weight gain-associated medication. The study was approved by the ethics committee of the LMU (protocol number 290.14) and followed the principles of good clinical practice according to the Declaration of Helsinki and its subsequent revisions. All participants provided written informed consent prior to study inclusion.

Adult psychiatric inpatients (18–75 years) were included in the study if they started treatment with one of the following drugs that were associated with weight gain according to Dent et al. [2] and Bak et al. [1]: amisulpride, clozapine, haloperidol, olanzapine, paliperidone, quetiapine, risperidone, citalopram, doxepin, amitriptyline, mirtazapine, paroxetine, lithium, and valproate. All ICD-10 F-diagnoses [41] were included. Patients with severe delusions or cognitive impairments were excluded to ensure valid data from self-rated questionnaires.

At baseline (t0) measures of metSy and BMI were taken by trained clinical personnel. Lifestyle factors were assessed by self-rating questionnaires at baseline and endpoint at week 4 (t2).

All calculations were done using SPSS Statistics 25. Tests were conducted two-sided and the significance threshold was set at α = 0.05. Lifestyle factors were analyzed regarding their baseline levels and change over time which were calculated subtracting t0 from t2. Thus, a positive score represents an increase and a negative score represents a reduction. Descriptive statistics are presented by frequency, mean and standard deviation, or median and interquartile range where applicable. To compare baseline and endpoint values, paired t test was carried out for continuous variables. In case of inhomogeneous variances, Welch test was used. Chi-square test was applied for categorical data. For correlation coefficients, Spearman rank correlation was used. Lifestyle factors at baseline were further contrasted by the dichotomized variables sex, BMI (overweight/obese ≥ 25 kg/m²), and presence of metSy. To test for significance, Student’s t test was used for continuous data and Chi-square test for categorical data. Change of lifestyle factors was tested among these subgroups (women, men, low/normal BMI, overweight/obesity, no metSy, and presence of metSy) using paired t test. Multiple test correction was dispensed due to the exploratory hypothesis-generating character of the study and partly small sample sizes in subgroups. Multiple regression analyses were performed to explore the predictive value of lifestyle factors at baseline and their change from baseline to endpoint for the change of body weight over the treatment period. BMI, sex, presence of metSy, drug dosage, and their two-way interaction with each lifestyle factor were also included in the model.

The total sample of eligible patients at baseline consisted of 163 (60.1% (n = 98) men and 39.9% (n = 65) women) psychiatric inpatients. N = 98 patients participated in the follow-up assessment at week 4, of which n = 52 patients filled out any questionnaires. Demographical and clinical characteristics of the study sample are displayed in Table 1. Noteworthy, the mean percent drug dosage had a high range from 0.05 to 18.6%, where the upper quartile was at 5.3% indicating higher values. Further information on demographical data and clinical characteristics can be found in Barton et al. [17].

Eating habits, food craving, sitting time, and self-esteem did not change from baseline to 4 weeks of treatment. There was a significant increase of scores for attitude towards drugs to a more positive attitude and for affect balance score to a less negative affect, as well as a trend for sweets craving (more craving after treatment). A significant decrease emerged for total sleep quality score towards a less-disturbed sleep quality. Results are displayed in Table 2.

To examine the potential role of medication for the observed changes (see Sect. Lifestyle characteristics at baseline and their change over time), the association with drug dosage was investigated. Change of self-esteem over time was positively correlated with drug dosage at the border of significance (n = 50; Spearman-Rho = 0.27; p = 0.058): the higher the dosage, the higher the improvement of self-esteem. No other lifestyle changes were associated with drug dosage.

FEV Regarding eating behavior, women showed higher cognitive control than men, and overweight/obese patients and patients with metSy showed higher disinhibition and hunger compared to patients with normal BMI and without metSy (Table 3 ).

Food craving Men had higher fat craving than women. Overweight/obese patients had higher fast food craving, as well as higher sweets and total food craving as a trend (Table 3). Furthermore, patients with metSy showed higher sweets craving as well as a more severe weight cycling profile (Table 3). Women, patients with normal BMI, and patients without metSy showed a significant increase of sweets craving over time (Table 5).

ABS and self-esteem Negative–positive affect balance in men changed significantly to a less negative affective state over time, as well as in patients with normal BMI, and patients without metSy. Self-esteem in men significantly increased over time (Table 5).

Sleep quality Concerning sleep, women had a more severely disturbed sleep profile compared to men and sleep quality score was higher (more disturbed) in patients without metSy (Table 4). In men, sleep quality significantly improved over time, as well as in patients with normal BMI and patients with metSy (Table 5).

Drug attitude Interestingly, a trend for a more positive attitude towards drugs emerged in overweight/obese patients and patients with metSy (Table 4). Drug attitude changed to a more positive evaluation in patients with normal BMI and without metSy (Table 5).

Based on the previous sections, potential relevant covariates were included to predict weight gain by the expression of lifestyle factors.

Baseline lifestyle factors Multiple regression analyses revealed a significant model for experience of hunger at baseline (R² = 0.26; F(9, 62) = 2.37; p = 0.023), but none of the coefficients reached significance. No other significances emerged for baseline lifestyle factors on weight change over time.

Lifestyle factor change The overall model for disinhibition was significant (R² = 0.90; F(9, 7) = 6.60; p = 0.011), while disinhibition change was predictive for weight change independently of covariates, as well as was BMI. Furthermore, disinhibition change interacted with and sex, drug dosage, and the presence of metSy to predict weight change (Table 6, Supplementary table 1, and Supplementary figs. 1 and 2).

Regarding the predictive values of food craving change over time for weight change, some trends emerged for single coefficients (but not for the overall model): sweets craving change interacted with sex, drug dosage, and presence of metSy to predict weight change (Table 6, Supplementary table 1 and Supplementary figs. 3 and 4). Furthermore, a strong trend emerged for the overall model of fast food craving change over time (R² = 0.32; F(9, 40) = 2.06; p = 0.058), while fast food craving change interacted with drug dosage to predict weight change (Table 6 and Supplementary table 1). The total food craving change score interacted with drug dosage to predict weight change (Table 6). The overall model for change of drug attitude was not significant, while change of drug attitude significantly predicted weight change (Table 6). Furthermore, drug attitude change interacted with BMI, drug dosage, and presence of metSy to predict weight change (Table 6).

FEV Eating habits were moderately expressed with regard to norm standards [44], but differed at baseline due to patient characteristics: women showed higher cognitive control and overweight/obese patients and patients with metSy showed especially higher disinhibition (emotional eating) and hunger. Thus, disturbed eating behavior is clearly associated with already present metabolic abnormalities. Prior research confirms that women show higher restriction (cognitive control), but also disinhibition/emotional eating and hunger at the same time [57, 58]. Furthermore, disinhibition and hunger were associated with higher BMI [58, 59]. Sentissi et al. [58] also found a trend for higher disinhibition/hunger and lower restraint (cognitive control) in schizophrenic patients with compared to without metSy. In our study, patients were investigated transdiagnostically and treated with weight gain-associated medication. In line, disinhibition trended for a higher expression under atypical antipsychotics than conventional neuroleptics [58]. As a potential effect of diagnosis or symptoms, individuals with mood disorders showed more cognitive restraint and disinhibition concerning food intake compared to population norms [59], and anhedonia was particularly associated with emotional/uncontrolled eating (disinhibition) [60]. Thus, investigating a drug- and symptom-specific effect for certain eating behaviors in future analyses may be worthwhile to refine population characteristics that can be targeted with interventions on a behavioral basis.

Food craving Men showed higher fat craving, and overweight/obese patients and patients with metSy showed higher fast food craving and sweets craving, respectively. Again, already existing metabolic abnormalities are clearly associated with craving for unhealthy food. In a mixed sample from healthy and psychiatric ill individuals, women craved simple sugars/trans fats more than men, while men craved fast food (saturated fats/high calorie content) more than women [61]. In exploratory analyses of a sample of college students, obese individuals were more likely craving high fats than normal weight participants and high-fat scores were positively associated with BMI [62]. Furthermore, in line with our results, patients with severe mental illness and overweight/obesity showed higher craving for fast food fats [63]. The overlap between fat and fast food craving is difficult to quantify, though possibly considerably large. In this light, our results seem to along with the presented findings from literature.

Weight cycling Presence of metSy was also associated with more severe weight cycling which confirms previous results of higher cardiometabolic risk in weight cyclers [64].

Sleep quality Most patients (88.4%) were poor or chronically disturbed sleepers and sleep quality was more disturbed in women and in patients without metSy. Poor sleep quality has also been found in a psychiatric outpatient sample with a very similar rate of 87.4% and sleep quality was lower in psychiatric in- and out-patients compared to controls [65, 66]. So far, previous studies had not related quality of sleep with metabolic disturbances in psychiatric patients.

Sitting time A study by Alswat et al. [34] demonstrated an association of sedentary behavior with metabolic disturbances: psychiatric inpatients with already existent metSy more likely reported sedentary lifestyle, had higher BMI, hypertension, hyperlipidemia, and diabetes compared to patients without metSy. On the contrary, we did not find such association.

Over time, attitude towards drugs, sleep quality, and negative affect improved, indicating an improvement of psychiatric treatment targets, while craving for sweets marginally deteriorated in the total patient group. Of all lifestyle factors, only improvement of self-esteem strongly trended for a positive association with drug dosage. In a schizophrenic sample, patients associated changes in dietary habits like increase in high carbohydrate and fat intake with medication use [39]. However, this study only assessed the subjective perception of patients of this relationship in a cross-sectional design.

Food craving Kazes et al. [67] found suddenly increased appetite for sweets and fatty food before weight gain after antidepressant onset. In our study, other factors seem to play a role for the changes of lifestyle behaviors during the treatment period: Patients without metSy and normal BMI, as well as women, showed increased sweets craving over time. This leads to the conclusion that patients without metabolic abnormalities experience a deterioration of unhealthy sweets craving over time as opposed to patients with presence of those abnormalities who already show such craving at baseline (see Sect. Clinical patient characteristics are differently associated with lifestyle factors). Consequently, it can be suspected that the increase and intensity of appetite may reach a plateau at some point even if medication is still taken. One study found no changes over time across different food craving types after treatment, but patients with normal weight presented an increase of the complex carbohydrates/proteins craving score [63]. This study also found a change of fast food fats craving during treatment with clozapine which was associated with male sex and baseline BMI [63]. The limitation to the use of only one antipsychotic drug in this study poses a substantial difference to our multi-drug approach.

These findings suggest that patients without metabolic abnormalities are more capable of experiencing improvement of mood and positive drug attitude after treatment. Thus, it is tempting to speculate that patients with such abnormalities might be less responsive to treatment, or experience less positive affect, or are less treatment compliant due to higher metabolic burden.

Food craving The interaction of fast food craving change with drug dosage showed a strong predictive effect, but the non-parametric correlation was weak. It is arguable whether especially low and high values can be considered as expectable values or as outliers, which complicates the decision for parametric or non-parametric test approach. Here, we decided to include such values in the parametric and non-parametric analyses as they seem clinically reasonable and reach high values in the interaction term due to its multiplicative nature. Like with disinhibition, the craving change itself over time per se was not present, but a subliminal effect in interaction with drug dosage was sufficient to indicated weight gain. The interaction of sweets craving change with sex, drug dosage, and presence of metSy showed a considerable trend. Sweets craving deteriorated over time and its change interacted with relevant variables. Like with disinhibition, the presence of metSy moderated the relationship of eating behavior and weight change, i.e., absent metabolic disturbances resulted in weight gain under deterioration of eating habits and vice versa in presence of metabolic disturbances. This may indicate the higher capacities for metabolic and eating deteriorations among initially unaffected patients, while present disturbances may not lead to further measurable deterioration over such short time. Previously, Garriga et al. [63] found increases in fast food craving to be predictive for weight gain after clozapine treatment in patients with severe mental illness. In another study, the food craving total score as well as the preoccupation with food and loss of control subscales at baseline were significantly associated with weight gain after antipsychotic intake only in females [57]. To our knowledge, these and our results are the first to provide insights in potential targets for behavioral monitoring and interventions, but more research is needed.

Many widely used psychotropic drugs are associated with weight gain, but effective treatment alternatives are lacking. Therefore, it is desirable to investigate potential risk factors for such weight gain to improve the management of unwanted side effects. The here-presented results serve as hypothesis-generating insights that should be verified in future studies and build the grounds for the development of interventions to combat weight gain.

Several limitations need to be considered. First, the study included a follow-up time of only 4 weeks. This approach was based on previous literature pointing to a special relevance of early weight gain during the first 4 weeks of treatment for continuing weight gain after [18]. Thus, determining risk for early weight gain is of special interest to deduce early intervention measures. Nevertheless, not being able to reproduce the long-term effects found in the literature poses a shortcoming of this study. Interestingly, even though the first weeks of treatment are often affected by dosage finding, the dose seems less relevant for metabolic effects in most cases [68]. Prior medication at any time before the study was not taken into account. Even though the patients entered the study drug-free regarding weight gain-associated medication, they likely were not drug-naive. This may have influenced the baseline body weight already and metabolic alterations may be long-lasting processes which complicates an estimation of the possible impact of previous medication intake throughout life. Thus, the results may be biased towards the relationship of lifestyle variables with weight gain attributable to the medication intake. However, following a naturalistic approach, this patient collective represents the clinical situation where patients have likely been treated before weight gain-associated medication may be given regardless of the BMI. Thus, baseline BMI was considered in the analyses to estimate weight gain and possible further weight gain. The results present some consistency. However, the overall picture is somewhat scattered. Due to the unexpectedly large number of drop-outs during the study, the application of an imputation method was dispensed. The reasons for drop-out were not registered systematically. Main reasons were lost to follow up due to discharge from the clinic and lacking patient motivation. Furthermore, the FEV as was added as a follow-up measure during the course of the study explaining the small amount of available values. It is debatable whether non-significant findings are a result of type II error, i.e., lack of statistical power, or of truly non-existent associations. Therefore, such findings should not be generalized to the target population, but may serve as indicators for aspects to be replicated in future studies. In this light, the almost significant trends can rather be interpreted as a positive finding. Generally, the present results should be viewed as preliminary due to potential uncontrolled bias. In future, verification in a bigger sample is needed to see whether a more elaborate pattern emerges or whether these single parameters that showed significant associations should be selected, specifically. Regarding the operationalization of eating behavior, the FEV has to be criticized for its high correlation of the disinhibition and hunger scale, i.e., their limited discriminability [44]. Furthermore, no actual measurement of food intake was carried out, while craving and eating behavior are only indirect surrogates of the actual amount of food eaten. Thus, it is difficult to attribute weight gain to caloric intake. On the other hand, using such scales provides clearer indications for behavioral intervention targets. Finally, other factors beyond the presented clinical and lifestyle characteristics should be considered in future studies. Since a relationship of factors like childhood trauma and socioeconomic variables with weight gain was established, those parameters should be taken into account [69, 70]. In the present study, socioeconomic factors were measured, but their evaluation was disregarded due to questionable validity of the instrument. A currently running project investigates the effects of both aspects for weight gain.