*2.2. Subjects*

In the present study, 219 female inpatients (68 with a diagnosis of anorexia nervosa (AN), 79 with obesity (OB), and 72 normal-weight (NW) patients treated for conditions other than eating disorders or obesity, such as adjustment disorders, somatoform disorders, or mild depressive episode) were recruited upon admission to the Department of Psychosomatic Medicine at Charité–Universitätsmedizin Berlin (between February 2012 and July 2018). All patients were at an age of ≥18 years. Current pregnancy or lactation period, malignant disease, treatment with immunomodulatory drugs (e.g., methotrexate, azathioprine, and oral corticosteroids), hypercortisolism, and untreated thyroid dysfunction were exclusion criteria. Moreover, women with psychotic disorders, somatoform or somatic disorders of the gastrointestinal system, and those preceding (e.g., bariatric) surgery of the gastrointestinal system, except for appendectomy and uncomplicated cholecystectomy, were excluded.

#### *2.3. Anthropometric Measurements*

Study enrolment, including clarification of potential exclusion criteria and blood withdrawal, was conducted within four days of admission. Venous blood samples were taken after an overnight fasting period between 7.00 and 8.00 a.m. Patients were permitted to drink a small amount of water, but were advised not to drink coffee, smoke, or exercise before blood withdrawal. On the same morning each patient's actual medication, body height, and weight in light underwear were assessed and BMI (kg/m2) was calculated. Medications and the presence of comorbidities were recorded at admission and discharge. Participants diagnosed with any of the exclusion criteria during their inpatient treatment were excluded.

#### *2.4. Physical Activity and Energy Expenditure Assessment*

To assess PA, we used a SenseWear® Pro3 armband (BodyMedia, Inc., Pittsburgh, PA, USA), which is a two-axis accelerometer that calculates PA by measuring skin temperature, near-body ambient temperature, galvanic skin response, and heat flux [24]. PA was analyzed for three consecutive days starting from Friday, which was the day of the blood withdrawal. Data were accepted if inpatients wore the armband for more than 20.5 h for at least two out of the three days, as described previously [25]. The PA of the patients was not restricted by the medical staff while wearing the accelerometer.

Using a generalized proprietary algorithm developed by the producer, the total amount of steps, metabolic equivalents of tasks per day (MET), level of energy expenditure, and exercise activity thermogenesis (EAT) were directly calculated after reading out the data. As EAT, we defined an activity of more than three metabolic equivalents of task (METs), which refers to moderate- and vigorous-intensity activities according to the 2011 Compendium of Physical Activities [26].

The thermic effect of food (TEF) was estimated as comprising 10% of total energy expenditure (TEE) and calculated as TEE × 0.1 [27]. Since resting energy expenditure (REE), required for the calculation of NEAT, cannot be directly determined by the SenseWear® armband, it was estimated using weight-group-specific REE prediction equations provided by Müller et al. [28]. Non-exercise-related activity (NEAT) was calculated using the formula NEAT = TEE −TEF − REE − EAT.

#### *2.5. Body Composition Measurements*

Bioelectric impedance analysis (BIA) was performed between 10:30 a.m. and 1:00 p.m. on the day of blood withdrawal under standardized conditions in the supine position, after subjects had fasted for at least two hours and had lain for half an hour. Phase angle, fat mass, fat free mass, extracellular mass, and body cell mass were assessed using the equations provided by the manufacturer of the bioelectrical impedance analyzer (Nutrigard-M®, Data Input®, Darmstadt, Germany).

### *2.6. Laboratory Analyses*

Blood was collected in pre-cooled standard EDTA tubes prepared with aprotinin for peptidase inhibition (1.2 Trypsin Inhibitory Unit per 1 mL blood; ICN Pharmaceuticals, Costa Mesa, CA, USA) and immediately submerged in ice. After that, tubes were centrifuged at 4 ◦C for 10 min at 3000× *g* for plasma separation, which was stored at −80 ◦C, until further processing. After enough samples were collected, SPX plasma levels were measured using a commercial enzyme-linked immunosorbent assay (ELISA, catalog # EK-023-81, Phoenix Pharmaceuticals®, Inc., Burlingame, CA, USA). All samples were processed at once. Intra-assay variability was 7.5% and inter-assay variability was <15%. Measurement was performed in January 2019. Every measurement was performed twice, and a mean value was calculated.

#### *2.7. Patient-Reported Outcomes*

All study participants were asked to fill in the following self-reported questionnaires: Perceived Stress Questionnaire (PSQ), Generalized Anxiety Disorder-7 (GAD-7), Patient Health Questionnaire depression scale (PHQ-9), and Eating Disorder Inventory-2 (EDI-2). Results obtained between two days before and five days after the respective blood withdrawals were accepted.

PSQ-20 is a revised 20-item German version [29] of the Perceived Stress Questionnaire (PSQ; 30 items) [30] and is applied to evaluate subjectively perceived stress. It provides four subscales: "worries", "tension", and "joy" as stress responses, and "demands" as the perception of external stressors. It assesses the subjective experience of stress. Cronbach's alpha for the total scale was 0.73 and for the subscales 0.86 ("worries"), 0.86 ("tension"), 0.81 ("joy"), and 0.84 ("demands").

The Generalized Anxiety Disorder Questionnaire (GAD-7) [31] is a part of the Patient Health Questionnaire (PHQ) and an established and widely used 7-item screening instrument for diagnosing general anxiety disorder. It also captures symptoms of social anxiety, posttraumatic stress, and panic disorder. In this study, the German version was used [32]. The Cronbach's alpha for the current sample was 0.87.

The severity of eating disorder symptoms was evaluated using the Eating Disorder Inventory-2 (EDI-2) [33], which is a widely established tool to assess eating disorder pathology in patients suffering from anorexia and bulimia nervosa. It consists of 64 items and encompasses eight subscales, measuring "drive for thinness", "bulimia", "body dissatisfaction", "ineffectiveness", "perfectionism", "interpersonal distrust", "interoceptive awareness", and "maturity fears". In our study, sum scores ranging from zero to 100 were created. Moreover, we employed the German translation of the second version [34] and interpreted the first eight, above-mentioned subscales of the EDI-2. The Cronbach's alpha for the total scale was 0.96, and for the subscales: 0.91 ("drive for thinness", "bulimia", and "body dissatisfaction"), 0.90 ("ineffectiveness"), 0.80 ("perfectionism"), 0.82 ("interpersonal distrust"), 0.83 ("interoceptive awareness"), and 0.73 ("maturity fears").

To assess the severity of depressive symptoms, we used the German version [35] of the PHQ depression scale (PHQ-9) [36]. It consists of nine items that represent the DSM-IV diagnostic criteria for depressive disorders, and its scores range from zero to 27, with scores of ≥10 indicating major depression with a specificity of 0.92 and sensitivity of 0.80 regarding a meta-analysis [37] of 17 validation studies in different languages. The Cronbach's alpha for the current sample was 0.86.

#### *2.8. Statistical Analyses*

All statistical analyses were conducted using IBM SPSS Statistics® Version 27.0.0.0 (IBM® Corp, Armonk, NY, USA).

Three groups were created, according to the medical diagnosis and BMI: an anorexia nervosa group (AN) with women diagnosed with anorexia nervosa (n = 68), an obesity group (OB) consisting of patients with a BMI of ≥30.0 kg/m2, and a normal weight group (NW) with a BMI between 18.5 kg/m<sup>2</sup> and 25.0 kg/m2 and without a diagnosed eating disorder.

Regarding the explorative design of this study, we established a cut-off of three standard deviations from the mean SPX level to identify outliers. During data analysis, three outliers (two in the anorexia nervosa group and one in the obesity group) were detected and excluded from further statistical analyses, which resulted in a study population of 219 women.

To investigate differences between the three groups, between-group comparisons were made using the Kruskal–Wallis test for non-parametric and one-way ANOVA for parametric data. To assess the frequency distributions between the groups, an overall chisquared test was performed. In case of significant differences, pairwise comparisons using a chi-squared test were added. Correlations were assessed using Pearson's for normally, and Spearman's analysis for non-normally, distributed data. Due to the exploratory approach, we decided not to perform multiple linear regressions. The correlations and differences between groups were considered significant when *p* < 0.05. Due to the explorative design of the study, no corrections for multiple testing were applied.

#### **3. Results**

#### *3.1. Demographic, Socioeconomic, and Medical Characteristics of the Study Population*

Demographic and socioeconomic characteristics, comorbidities, and medication of study participants are outlined in Table 1. The AN group was significantly younger than both the OB (*p* < 0.001) and NW groups (*p* < 0.001; Table 1). By definition, patients with AN displayed a lower BMI than patients with OB (*p* < 0.001) and NW subjects (*p* < 0.001), and the NW group had a lower BMI than the OB group (*p* < 0.001; Table 1). Regarding socioeconomic status, the highest proportion of subjects living in a partnership was observed in the NW group. Furthermore, the OB group showed the lowest level of education, as indicated by a lower rate of university entrance diplomas than AN (*p* < 0.01) and NW (*p* < 0.05) and of any other school-leaving qualification than NW (*p* < 0.01; Table 1). NW women were also less often currently unemployed than OB and AN (*p* < 0.05; Table 1).

**Table 1.** Demographic and socioeconomic characteristics, comorbidities, and medication of study patients.



**Table 1.** *Cont.*

Data are expressed as absolute numbers with percentages in parentheses. Differences between groups were assessed using Kruskal–Wallis (age and BMI) and χ<sup>2</sup> tests. Significant differences (without correction for multiplicity) between the AN and OB groups are displayed as \* (*p* < 0.05), \*\* (*p* < 0.01), or \*\*\* (*p* < 0.001); between the AN and NW groups as # (*p* < 0.05), ## (*p* < 0.01), or ### (*p* < 0.001), and between the NW and OB groups as + (*p* < 0.05), ++ (*p* < 0.01), or +++ (*p* < 0.001). Abbreviations: AN, anorexia nervosa; DPP-4, dipeptidyl peptidase-4 inhibitor; GLP-1, glucagon-like peptide-1; NW, normal weight; OB, obesity; SSRI, selective serotonin reuptake inhibitors; SNRI, serotonin-norepinephrine reuptake inhibitors.

As expected, type 2 diabetes mellitus, impaired glucose tolerance, insulin resistance, arterial hypertension, hyperuricemia, and fatty liver disease (*p* < 0.001), as well as hypertriglyceridemia *(p* < 0.01), were more common in patients with OB than in AN and NW (Table 1). No significant differences were found between groups in terms of medication taken, except for antidiabetics other than insulin and DPP-4-antagonists/GLP-1 analogs (mostly metformin), which were more common in OB than NW and AN (*p* < 0.01) and for opioids (*p* < 0.01) and other psychopharmacological medication (*p* < 0.05), which were more common in NW and OB than AN. In the NW group, tricyclic antidepressants were more often prescribed than in the AN group (*p* < 0.05, Table 1).
