**The Role of Objectively Measured, Altered Physical Activity Patterns for Body Mass Index Change during Inpatient Treatment in Female Patients with Anorexia Nervosa**

**Celine S. Lehmann 1,2, Tobias Hofmann 1, Ulf Elbelt 1,3, Matthias Rose 1, Christoph U. Correll 2,4,5, Andreas Stengel 1,6,\*,† and Verena Haas 2,\*,†**


Received: 23 August 2018; Accepted: 12 September 2018; Published: 18 September 2018

**Abstract:** Increased physical activity (PA) affects outcomes in patients with anorexia nervosa (AN). To objectively assess PA patterns of hospitalized AN patients in comparison with healthy, outpatient controls (HC), and to analyze the effect of PA on Body Mass Index (BMI) change in patients with AN, we measured PA in 50 female patients with AN (median age = 25 years, range = 18–52 years; mean BMI = 14.4 ± 2.0 kg/m2) at the initiation of inpatient treatment and in 30 female healthy controls (median age = 26 years, range = 19–53 years; mean BMI = 21.3 ± 1.7 kg/m2) using the SenseWear™ armband. Duration of inpatient stay and weight at discharge were abstracted from medical records. Compared with controls, AN patients spent more time in very light-intensity physical activity (VLPA) (median VLPA = 647 vs. 566 min/day, *p* = 0.004) and light-intensity physical activity (LPA) (median LPA = 126 vs. 84 min/day, *p* < 0.001) and less time in moderate-intensity physical activity (MPA) (median MPA = 82 vs. 114 min/day, *p* = 0.022) and vigorous physical activity (VPA) (median VPA = 0 vs. 16 min/day, *p* < 0.001). PA and BMI increase were not associated in a linear model, and BMI increase was mostly explained by lower admission BMI and longer inpatient stay. In a non-linear model, an influence of PA on BMI increase seemed probable (jack knife validation, *r*<sup>2</sup> = 0.203; *p* < 0.001). No direct association was observed between physical inactivity and BMI increase in AN. An altered PA pattern exists in AN patients compared to controls, yet the origin and consequences thereof deserve further investigation.

**Keywords:** accelerometry; eating disorders; motor restlessness; physical inactivity

#### **1. Introduction**

The role of increased physical activity (PA) for the onset and maintenance of anorexia nervosa (AN) is increasingly recognized. Being associated with a longer duration of inpatient treatment [1] and higher rates of a chronic outcome [2] as well as drop-out from treatment [3], increased PA can be regarded as a significant factor in the persistence of the disease [4]. However, high level PA is addressed insufficiently by current research [5]. As a consequence, a deeper understanding of the mechanisms underlying altered PA in AN as well as for the development of suitable therapeutic strategies to manage PA during weight restoration efforts are urgently warranted to improve outcomes for patients with AN.

Elevated levels of physical activity have been observed in 30–80% of patients suffering from AN [6,7], with this high range probably resulting from varying methods of PA measurement [8]. When assessed with subjective measurement tools including exercise questionnaires, patients reported higher total PA in comparison with a control group, yet simultaneous objective PA assessment using actigraphy yielded similar PA levels [9], suggesting that self-report overestimated PA in patients with AN and that objective assessments are needed to obtain accurate results. In addition, PA behavior is complex and has multiple dimensions; therefore, objective quantification of PA targets different components. Previous studies on objectively assessed PA in AN have yielded mixed results, with some reporting no differences in time spent in moderate to vigorous and daytime PA [10], or fidgeting [8], while others reported increased moderate to vigorous PA duration [3] and seated non-exercise PA [11] between AN inpatients and controls.

In a previous study [12], we focused on a potential link between high PA in AN and hypoleptinemia using a multisensor body monitor (Sensewear™ armband) for objective PA detection in hospitalized adults with AN. Results indicated that the use and interpretation of accelerometry, employed to objectively assess PA in AN patients, needs to be developed further and should also include parameters of physical inactivity. Building on the previous findings based on simple step count, the present study focused on a more detailed analysis of an expanded set of objectively measured PA patterns and intensities in adult females with AN, including inactivity parameters and adding a comparison to normal weight controls. We aimed to investigate the relationship between different PA patterns and BMI increase during inpatient treatment. We hypothesized that during inpatient treatment (I), hospitalized adult AN patients show increased low intensity PA in comparison with healthy controls, (II) increased low-level PA and BMI increase are inversely related, and (III) physical inactivity and BMI increase are directly related.

#### **2. Subjects and Methods**

#### *2.1. Study Population*

We enrolled 50 female adults with AN who were admitted to the Department of Psychosomatic Medicine at Charité—Universitätsmedizin Berlin for inpatient treatment of AN between 2011 and 2016. Patient inclusion criteria were: A diagnosis of AN according to ICD-10 (International Statistical Classification of Diseases and Related Health Problems, 10th Revision), restrictive, purging or atypical type, as well as a BMI < 17.5 kg/m2. Exclusion criteria were: age <18 years, current pregnancy or a diagnosed psychotic episode. Information about the duration of the illness, comorbidities as well as medication at the beginning and end of the treatment program were retrieved from anamnestic data and medical reports. Between 2015 and 2016, we also recruited 30 sex-matched and similar aged normal weight healthy controls (HC), consisting mostly of clinical staff and relatives thereof. A BMI between 18.5 and 25 kg/m2 served as inclusion criteria. Exclusion criteria were: Any known major medical or psychiatric disease and any condition with significant influence on PA. All participants gave written informed consent, and the study was approved by the institutional ethics committee of the Charité—Universitätsmedizin Berlin.

#### *2.2. Anthropometry*

Weight of all patients was measured to the nearest 0.1 kg via a digital scale (Seca 771, Vogel & Halke, Hamburg, Germany) and height to the nearest 0.5 cm via a stadiometer (Seca 220 Stadiometer, Vogel & Halke, Hamburg, Germany) [13]. Measurements took place in the morning between 7 and 8 a.m. after overnight fasting and in underwear. Weight of the controls was measured after a 2-h fast using a chair scale (MCB300K100M, KERN & Sohn GmbH, Balingen, Germany) and height was measured using a stadiometer (Vogel & Halke). BMI was calculated as kg/m2.

#### *2.3. Bioelectrical Impedance Analysis*

Whole-body bioimpedance was measured by Nutriguard-M (Data Input, Darmstadt, Germany; electrodes: Bianostic-AT, Data Input) as part of the patients' clinical measurements. For bioimpedance analysis (BIA) of the normal-weight controls Biacorpus RX 4004 (MEDICAL HealthCare GmbH, Karlsruhe, Germany; Electrodes: BIA Classictabs, Medical HealthCare GmbH, Karlsruhe, Germany) was used. Patients and controls were weighed after fasting for at least 2 h, voiding and an equilibration period in a supine position. The equilibration period of both AN patients and controls lasted at least 10 min. BIA was carried out in accordance to the manufacturer's instructions, and body composition was calculated with Body Comp software (Version 9.0, Professional Scientific, Medical Health Care GmbH, Karlsruhe, Germany).

#### *2.4. PA Assessment*

PA was measured in AN patients after inpatient admission and inclusion into the study. Using a portable armband device (SenseWear™ PRO3 armband; BodyMedia, Inc., Pittsburgh, PA, USA), PA was continuously detected over a 3-day period (Friday to Sunday). During the time of PA detection, the study population was not restricted regarding their daily physical activity [13]. A day was included into data analysis if the armband had been worn for at least 20.5 h [13]. Measurements of controls took place while they stayed in their usual environment and by using the SenseWear™ PRO3 or the SenseWear™ MF armband. According to a statement of the manufacturer from 15 March 2011, the Sensewear Pro 3 and MF models were shown to be functionally equivalent in terms of sensor technology and data analysis (manufacturers statement on equivalency available on request).

The Sensewear armband is a multi-sensor device worn on the upper dominant arm which enables a continuous physiological PA detection [14] by measuring parameters such as heat flow, galvanic skin response (GSR), body temperature and near-body temperature [15]. An integrated two-axial accelerometer captures the movement of the upper arm as well as the position of the body [14]. The information captured by the five sensors and participant characteristics (age, sex, weight, height, smoker or non-smoker and handedness) [16] are integrated and analyzed by a proprietary software (SenseWear™ Software, Version 8.0, BodyMedia, Inc., Pittsburgh, PA, USA). This program is based upon algorithms of the manufacturer and able to analyze the collected raw data at different metabolic equivalent (MET) values. The latter represents a standardized indicator which is independent of time, body weight and sex [15]. One MET is equivalent to 1 kcal/h/kg body weight and serves as useful parameter to describe the energy expenditure [17] and intensity [15] of a specific activity. The MET value ranges from 1 MET while at rest [17] and 1.1 METs when driving in a car to 2–4 METs when doing housework [15], and can reach maximum values of 20 METs when doing excessive sports [15]. According to previous studies, we used six different MET categories to classify different activity intensities of PA within our AN and control group:

