Eating disorders (ED) are serious mental illnesses which are also characterized by detrimental somatic changes such as reduced physical fitness, altered body weight and body fat percentage, and poor bone health compared to healthy controls [1]. These impairments occur despite higher physical activity levels among patients compared to controls [2,3]. Physical activity is useful for improvement of both physical fitness, bone mineral density (BMD) and body composition [4], but excessive amounts of physical activity can have opposite effects [5,6]. In a public health perspective, physical fitness, especially aerobic fitness, is one of the most important predictors for good health and prevention of lifestyle related diseases [7]. Although focus on physical changes during treatment of ED is often limited to bone health and body composition, the importance of physical fitness for somatic and mental health calls for studies which also address this issue.

Despite higher amounts of physical activity among patients with longstanding ED compared to controls, we found no increase in aerobic capacity and reduced muscular strength among patients across the three DSM-IV ED diagnoses anorexia nervosa (AN), bulimia nervosa (BN) and ED not otherwise specified (EDNOS) [2,8]. This fits well with previous studies on patients with AN [9,10,11]. During weight restoration of adolescents with AN, Waller et al. [12] and Rigaud et al. [13] found only small improvements in aerobic capacity after eight weeks, whereas Fohlin et al. [14] found aerobic capacity to increase into normal values. Treatment approaches including regular, systematic physical activity have been found effective in increasing aerobic capacity among patients with AN [15] and BN [16]. Less is known about changes in muscular strength during treatment. Rigaud et al. [13] found great improvements in muscular strength on a bicycle ergometer exercise test after 45 days of refeeding in patients with AN. These improvements occurred despite unaltered muscle mass and aerobic capacity. Studies using light-to-moderate intensity resistance training as a part of treatment are inconclusive regarding the effects on muscular strength. Chantler et al. [17] found an effect after eight weeks of resistance training, whereas del Valle et al. [18] found no effects after three months of resistance training. Although light-to-moderate intensity, the doses of resistance training were different in these two studies, and it is possible that the lack of improvements in Del Valle et al. [18] can be explained by an inadequate dose of resistance training. To our knowledge, no studies have examined the changes in aerobic capacity and muscular strength among both underweight and normal weight adult patients across the three diagnoses AN, BN and EDNOS.

Several studies have reported low values of bone mineral density (BMD), especially in the lumbar spine and femur neck, among patients with AN compared to controls [19]. In patients with BN, the literature is inconsistent. Sundgot-Borgen et al. [20] found normal BMD values among both patients with BN and healthy controls, whereas others have found low BMD values among both AN and BN patients [21,22,23]. Possible explanations for this inconsistency can be differences in the patients’ previous ED, menstrual irregularities and exercise history. Studies have found a history of AN, especially with adolescent onset, to be a strong predictor of low BMD among patients with BN [21]. In addition, Bratland-Sanda et al. [8] found both current and previous high mechanical loading physical activity, but not general weight-bearing physical activity, to be correlated with BMD in lumbar spine, femur neck and total body. Treatment for low BMD values in patients with ED includes pharmaceutical and nutritional strategies, but today there is lack of knowledge regarding the effectiveness of these strategies [24]. Studies following changes in BMD during treatment have found that weight restoration and menstrual recovery are the most important factors in patients with AN [25,26]. To our knowledge, only one prospective study has examined changes in BMD in patients with BN, and in this study only one patient had osteopenia (i.e., t-score, from −1.0 to −2.5 SD) at baseline [27]. Therefore, it is still unclear how low BMD values changes following treatment in both underweight and normal weight patients with ED.

Body composition measurements separate between fat mass and fat free/lean body mass. The fat mass consists of essential fat mass (i.e., fat mass necessary to maintain normal physiological and cognitive functioning) and storage fat [28]. Behnke developed a theoretical limit for minimal body mass, which states that females have approximately 12% essential body fat [28]. If the body fat percentage drops below this limit, there is a high risk of menstrual irregularities and impaired physiological and cognitive functions [28]. The major weight related public health issue in the western world is obesity, but there is no consensus of when body fat percentage exceeds healthy limits. Several suggest 30–35% to be the upper limit for healthy body fat percentage [28,29,30]. Studies have found reduced brain volume and altered brain function and metabolism among very low weight patients with AN [31,32]. Although these changes can be reversible, abnormalities in brain tissue volume, brain structure and cognitive function have been seen in females with a history of AN [32,33]. Weight restoration is therefore important; however current clinical practice often does not take body composition into account. Existing studies have found weight restoration in patients with AN to consist mainly of increase in fat mass [12,34]. As an immediate effect, Mayer et al. [34] found an altered body fat distribution with more visceral adipose tissue, however the body fat distribution was normalized within one year after weight restoration [35]. Several studies have suggested that restrictions regarding physical activity is necessary to achieve weight restoration [36,37,38], whereas Touyz et al. [39] did not find anaerobic exercise (i.e., strength training) to interfere with weight gain during treatment of AN patients. Only one study has to our knowledge examined changes in body composition during treatment in patients with BN. Sundgot-Borgen et al. [16] found larger enhancement of body composition in patients who exercised compared to patients who received nutritional counseling or cognitive-behavioral therapy.

Total body weight and body mass index (BMI) are the most common measures of weight restoration. However, BMI and body composition are not necessarily comparable. On one hand, persons with a BMI in the underweight category (i.e., <18.5 kg/m²) can have body fat percentage values within normal range, and persons with BMI in the normal weight or overweight category can actually be low on body fat (i.e., certain athlete populations) [28]. Since the majority of patients with ED are normal weight, possible unhealthy body composition in these patients is not necessarily detected and managed in treatment. In most studies concerning physical fitness, BMD and body composition in patients with ED, the samples consist of adolescents with AN. Therefore it is a need for studies covering adult underweight and normal weight patients across the three diagnoses AN, BN and EDNOS.

Seven patients with BMI < 18.5 (range 14.6–18.4) and 22 patients with BMI ≥ 18.5 (range 19.5–29.3) participated. Four of the patients with BMI ≥ 18.5 were classified as overweight (BMI > 25), but due to the low sample size they are grouped together with the normal weight patients. Table 1 shows no differences between patients with BMI <18.5 and ≥18.5 in EDE and EDI scores, duration of eating disorder, objectively monitored weekly physical activity level at admission and self-reported physical activity level for the four weeks prior to admission. The patients with BMI < 18.5 had longer treatment period at Modum Bad (p < 0.01). During treatment there was a significant reduction in ED psychopathology assessed through EDE global score and EDE total score (p < 0.001) [47].

The main findings of this study were the increase in body fat percentage among patients with normal body fat percentage, and that three of the seven patients with BMI <18.5 at admission were still underweight at discharge. Two of these patients had a BMI below 17.5 at discharge, which is the current cut off value for the AN diagnosis [48]. There was a statistically significant increase in total body weight and total fat mass among patients with BMI above as well as below 18.5, however, the net value was a mean increase of 3.3 kg over a mean treatment period of about 16 weeks for the BMI < 18.5 patients. Although this gave an effect size of 1.9, it is far less than reported in other studies [12,34]. A possible explanation for this is that the patients in the studies of Waller et al. [12] and Mayer et al. [34] had lower mean body weight and BMI compared to the underweight patients of this study. It is important to emphasize that the patients in the current study, although not everyone reached their aimed weight gain, showed significant reduction in ED psychopathology from admission to discharge. As this was the primary treatment goal, it is possible that the weight restoration continues after discharge.

The weight gain among patients with BMI ≥ 18.5 was on average 0.9 kg, which has no clinical relevance. However, the fat mass increased significantly among the patients who already had a normal body fat percentage, and two patients went from a normal body fat percentage (i.e., 20–32%) at admission to an increased risk of obesity related lifestyle diseases at discharge. Although the main focus during treatment was, and still needs to be, management of ED symptoms and psychopathology, there was an inverse relationship between changes in BMI and total body weight, and changes in e.g., body dissatisfaction and bulimia. In a public health perspective, the negative change of body composition among the normal weight patients is an important issue to address.

Changes in android/gynoid (A/G) ratio varied among the individuals from an increase of 88% to a decrease of 24%. The android mass refers to visceral fat mass, and the gynoid fat mass refers to fat mass on the hips. An increase in A/G ratio indicates an altered body fat distribution, this has been shown by Mayer et al. [34]. Previous studies have found hypercortisolemia among adult, but not adolescent, females with AN, and that these high levels of cortisol can contribute to increased A/G ratio [35].

The increase in muscular strength is in accordance with among others Rigaud et al. [13]. Due to the low physical fitness at admission, one of the goals in the physical activity sessions during treatment was to increase both aerobic capacity and muscular strength. The fact that several of the patients had low aerobic capacity despite high levels of vigorous intensity physical activity can possibly be explained by inadequate recovery and low energy availability [8,49]. When the volume of physical activity is reduced and the energy intake is more optimal, the patients should be able to increase in aerobic capacity. However, recovery includes not only energy and nutrient intake, but also sleep and rest. If a person is very psychologically distressed, e.g., during school exam periods or periods with sleep deprivation, the recovery time is prolonged [50]. Thus, in patients with ED, both malnutrition and psychological distress may contribute to prolonged exercise recovery. Physical activity amount (i.e., frequency, duration and intensity) was assessed both objectively and through self-report at admission and discharge. During treatment, the patients kept a physical activity log. In a previous paper, we reported the finding of significantly underreporting of weekly physical activity among inpatients with longstanding eating disorders [2]. In addition, the non-prescribed physical activity both prior to admission and during treatment varied greatly among the patients [2,47]. Although the patients were admitted to an inpatient unit, they were allowed to go outside the hospital for walks etc. The real amount of physical activity could therefore have been higher than reported, and may account for some of the lack of improvements in aerobic capacity.

The increase in lumbar spine BMD among the underweight patients is promising, and in accordance with previous studies having found that weight gain and restoration are important factors for improvement of BMD. The fact that all osteopenic or osteoporotic patients, except two, still had low BMD values at discharge can be due to too short duration of treatment or inadequate treatment approaches. It is possible that other changes regarding the geometry and structure of the bone has occurred, without this being captured by the DXA scanning. Unfortunately, we did not measure hormonal variables such as estrogen and insulin-like growth factor (IGF-1). Physical activity, especially strength training, stimulates the release of growth hormone, which again stimulates the release of IGF-1 [51]. This has been found to be an important predictor of bone microarchitecture independent of body mass in females with AN [52], and future studies should therefore measure changes in hormonal levels when examining changes in BMD among adult underweight females with eating disorders.

The results of this study showed unaltered aerobic capacity, improved muscular strength, and increased body weight and body fat percentage among both underweight and normal weight patients with longstanding ED. Among the underweight patients, there was an increase in lean body mass and lumbar spine BMD. Increased total body mass and BMI were significantly correlated to increased body dissatisfaction, restraint and bulimic symptoms among the patients.