Background
Anorexia nervosa (AN) is a psychiatric illness characterized by body dissatisfaction accompanied by restriction of energy intake leading to a significantly low body weight, and persistent behaviors, including high levels of physical activity, that interfere with weight gain [
1]. Individuals with AN become severely malnourished and develop significant medical complications [
2,
3]. Although weight restoration is a primary intervention for AN, along with psychotherapy and behavioral interventions, there has been little research on optimal refeeding strategies. There is currently no standard, evidence-based clinical refeeding diet for the treatment of anorexia nervosa (AN) and nutritional treatment is focused primarily on calories and less on diet composition [
4]. However, studies in both rats and humans show that diet composition can affect the rate of weight regain following weight loss, impact the ability to maintain at a target weight, and dramatically alter metabolic health [
5‐
7]. Therefore, in order to provide the most efficacious clinical care for AN patients, it is necessary to define the effect of current refeeding diets on markers of metabolic health and use this information to develop optimal AN refeeding diets. Our ultimate goal is to achieve metabolically healthy weight gain in this population.
Given the low overall prevalence of AN in the population and the challenges in treating the illness, it would be difficult to test multiple different refeeding diets in AN patients. However, there is an established rodent model of AN that can be used for this purpose. Activity-based anorexia (ABA) is a well-established paradigm in which restricted time with access to food combined with free access to exercise causes hyperactivity, reduced energy intake, and, ultimately, anorexia in rats [
8] (reviewed in [
9,
10]) and to a lesser extent mice (reviewed in [
11]). Most early studies of the ABA model utilized male rats, provided access to food for 1–2 h per day, and defined anorexia as a body weight at 65–75 % of free feeding or initial body weight. However, because AN in humans is most common in females, there has been a shift towards the use of female rats in the ABA model, primarily to study brain structure and neuronal biology/signaling in AN [
12‐
15]. We endeavored to use recently published data about the diet and physical activity characteristics of female adolescent AN patients [
16] to refine the ABA model in pubertal female rats to more closely mimic the human condition.
The goals of this study were to: 1) define the metabolic effects of a clinical refeeding diet on markers of metabolic health in a modified rodent model of AN and, 2) to develop a rat model of AN that closely mimics the human condition which could be used to screen more effective refeeding diets in the future. We hypothesized that, during the refeeding phase, AN rats would exhibit “catch-up growth” with preferential deposition of ingested energy into adipose depots [
17‐
19], especially visceral adipose, indicative of a higher metabolic risk for cardiovascular disease and associated co-morbidities.
Discussion
The primary goal of this study was to develop a rat model that can be used to evaluate refeeding strategies and potentially improve the process of weight restoration during the treatment of AN. Specifically, it is likely that different dietary regimens could be used to strengthen the biological response to refeeding and potentially improve weight regain and maintenance in this population. Thus, developing a model that mirrors the human condition was of critical importance. Our model, which combined a classic exercise induced anorexia paradigm (free access to a running wheel with timed access to food; [
12]) with the addition of using female adolescent rats genetically predisposed to leanness, led to the development of AN within 4–7 days and was characterized by a loss in both total body mass and body fat. Unlike data from Dixon et al. [
12], rats in the AN group in this study ate significantly less food during the AN induction phase than the control group. In addition, increased physical activity preceded any decrease in energy intake by several days which mirrors the development of human AN in which increased physical activity may be the first apparent symptom prior to onset of dramatic weight loss [
25].
When AN rats were compared to control animals, refeeding on a standard LF diet caused comparable increases in total body weight and subcutaneous fat accumulation. However, leptin concentrations were lower in the AN rats compared to controls and were also lower in AN rats than would be expected based on total body fat. This phenomenon of leptin concentration underestimating body fat stores is similar to that seen in rodent models of obesity whereby formerly obese rats that lose weight do not have restored leptin concentrations after weight regain [
23]. In our AN model, however, the effects are more dramatic than in the models of obesity. Although the underlying mechanism for this effect is unknown, the fact that the same amount of adipose tissue is secreting different amounts of leptin suggests that the adipo-insulin and adipo-leptin axes are perturbed in these animals.
Despite similar increases in total body fat, AN rats exhibited dramatically higher liver fat compared to controls. It is likely that the high accumulation of liver fat is due, at least in part, to increased de novo lipogenesis from excess carbohydrate intake during the refeeding period [
26]. In addition, fat gain appeared to plateau in the AN rats during the second week of refeeding. In models of obesity-associated weight loss and weight regain, exercise appears to decrease the formation of new adipocytes via adipogenesis [
27‐
29]. Thus, we would speculate that the exercise during the AN development phase, combined with the increase in caloric intake during the refeeding period, resulted in decreased formation of new adipocytes early in the regain period. If this is the case: 1) existing adipocytes likely reached their storage capacity during the first week of refeeding, and, after this time, excess nutrients were deposited in non-adipose storage sites, such as the liver, and 2) the lower number of total adipocytes may, in part, explain the lower leptin concentrations observed in AN rats. An alternative explanation is that the cessation of exercise could also contribute to hepatic de novo lipogenesis and the noticeable increase in lipid accumulation in the liver of AN rats, as cessation of daily exercise has been shown to promote hepatic steatosis in hyperphagic/obese rats [
30]. Future studies could include use of tracers to directly address the etiology of lipid accumulation. Regardless of the underlying mechanism, this increase in fat accumulation in the liver is cause for concern, as liver fat is associated with insulin resistance [
31] and the development of type 2 diabetes, independent of total adiposity [
32].
It is not known if patients with AN develop increased liver fat during the refeeding process. Further research is needed to determine if human subjects undergoing weight restoration have similar findings. To counter the negative effects of a LF diet on fat accumulation in the liver, observed on the AN group, it may be optimal to adjust the composition of the diet during weight restoration in AN to a diet with lower carbohydrate and increased fat content. Although previous AN research has shown that the percentage of fat and protein is not lower in acute inpatient AN than in age and sex matched adolescents, the absolute amounts of fat and protein intake are low [
16]. Thus, it is possible that increasing the fat and protein content of an AN refeeding diet may be beneficial. Indeed, during the timed food restriction, induction phase of ABA, a high fat diet can prevent AN in rats [
33,
34]. This effect is due to both the preservation of caloric intake during the AN induction phase and decreased physical activity relative to standard rat chow which is a high carbohydrate food [
34].
We would hypothesize that a diet higher in both fat and protein, and lower in carbohydrate, could ameliorate the fat accumulation observed in the liver in the current model and could also increase the rate of weight gain (by decreasing de novo lipogenesis with a concomitant decrease in TEE). This strategy would agree with the recommendation to limit AN refeeding diets to less than 40 % of calories from carbohydrate to reduce the risk of refeeding syndrome [
35]. However, if both fat and protein are to be used to replace dietary carbohydrate, caution must be exercised as there are selected reports of hyperammonemia with high protein intake after substantial weight loss [
36]. Although these data are from case reports and are rare in AN patients [
37], it is important that refeeding diets with higher protein contents be tested in a pre-clinical model, such as that described herein, before being administered clinically.
Although it is possible that higher fat and protein refeeding diets may be beneficial, it is commonly believed that AN patients have an aversion to fat although the data supporting this assumption are equivocal [
38]. To address this possible concern, we conducted a feasibility study in adolescent females with AN, which showed that a MF diet is both acceptable and palatable to patients with AN and that such a diet can be delivered successfully in an inpatient clinical setting. In the single patient with AN who consumed the MF diet for 10 consecutive days during weight restoration, all study parameters for macronutrient content were achieved in addition to following the program’s standard of care guidelines for calorie changes over time, rate of weight gain, monitoring food avoidance behaviors and enhancing reward systems, and behavioral motivation, etc.
The measured average macronutrient content of the clinical refeeding diet utilized in the treatment program in this study matches that for a similar sample of subjects measured retrospectively at the time of medical hospitalization for malnutrition and bradycardia [
16]. As part of treatment, parents work closely with dieticians to plan the patient’s meals based on a specific number of protein, carbohydrate, and satiety (fat) foods to achieve the daily caloric goal for weight restoration so the macronutrient content of these inpatient diets may be somewhat limited by this process.
The onset of AN and maintenance of the illness is associated with decreased caloric intake over a sustained period of time, ranging from months to years. Further studies are needed to determine optimal macronutrient composition and rate of weight gain for the weight restoration phase of treatment of AN, and if sustained change in dietary composition improves outcomes. It is possible that the cognitive changes associated with AN (body image distortion, fear of fat, harm avoidance and cognitive rigidity) may also respond to changes in dietary composition during weight restoration.
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
EDG participated in the design and execution of the animal studies, and drafted the manuscript; JAH designed, executed, and coordinated both the rodent and human studies, conducted data analysis and interpretation, and drafted the manuscript; JOH designed and coordinated the human studies; PSM contributed to the design, execution, and data analysis/interpretation for the rodent study; ZP performed all statistical analyses. All authors read and approved the final manuscript.