The present study has adopted a very novel approach to investigating the relationship between nutritional status and psychological symptoms in AN, since it takes into account body composition and biological markers, and not solely weight or BMI, for the nutritional assessment, also adjusting for age and psychotropic treatment. To our knowledge it is the largest study on this topic in AN.
The most important finding of this study of a large sample of severe AN patients is that we did not identify any correlation between the level of depression, anxiety or Obsessive-compulsive disorder (OCD) and any measure of current nutritional status at inclusion, even when taking into account potential confounding factors (age and current psychotropic treatment). This is an unexpected result. Scores for the intensity of depression were however related to the presence of psychotropic treatment and in certain cases to age or duration of illness (univariate analysis).
Three main hypotheses might explain our results that failed to reveal any correlation between the nutritional status and the psychological symptoms that we measured.
First of all, although we studied a large sample, the subjects’ characteristics might have been largely homogeneous in terms of the severity of malnutrition and the reported psychological symptoms, so that we could not identify any relationship between them. Patients were evaluated at admission to inpatient treatment at the most (or one of the most) severe moments in their illness. For example the narrow SD for BMI (mean
1.45) shows the severe malnutrition and the relative homogeneity in the sample. Even the patients that had higher BMI had lost huge amounts of weight (Cf. methods) and thus were all severely malnourished. Consequently there was very little variation in the degree of malnutrition. Despite the fact that levels of depression and anxiety were variable, we were not able to establish any link with levels of malnutrition. There may be a nutrition threshold, whereby psychological state is only affected when a certain degree of nutritional deficiency has been reached.
Second, we evaluated nutritional status in more comprehensive manner and in a larger sample compared to previous studies, and we used relatively a large set of indicators. However, body composition was measured using the BIA which is not a reference method (such as Dual-emission X-ray absorptiometry (DXA) or measurements using 4 compartment models). The severely malnourished status of the patients did not enable transfer to DXA centres for the measures to be performed. Also, the severity of malnutrition was measured by a rough estimation of the difference between the highest lifetime BMI and BMI at inclusion, thus considering weight loss to be linear, and not accounting for duration of illness and weight fluctuations. A more precise measure of these variations should be performed to provide information on this subject.
Third, as hypothesised by certain authors 
depression in AN, rather than having a single aetiology, is likely to be the consequence of various factors; depressive and anxiety symptoms in severely malnourished AN patients could therefore be mainly due to factors other than malnutrition, such as depressive symptoms linked to exhaustion, chronic illness or in some cases premorbid depression.
An interesting yet worrying observation from our study was the frequent use of psychotropic drugs in the treatment of very malnourished patients. More than 36% percent of AN patients admitted were receiving antidepressants. This is unusual, especially in severely malnourished subjects: it is well established that antidepressants are not effective on patients with low BMI 
. These treatments have usually been prescribed before inpatient admission, generally by non-specialized physicians, and they are generally stopped after admission, because they are ineffective. Despite these elements, it is interesting to see that the higher the anxiety or depressive scores, the more likely patients are to be receiving anti-depressants (AD).
How can we understand the link between psychological symptoms and malnutrition in the light of our data and the literature?
In the first stages of the illness, patients report that starvation provides relief from pre-existing anxiety and depressive symptoms. However, in a second stage, these symptoms tend to increase and regardless if malnutrition and starvation continue. If at the beginning of the illness, patients feel better and less anxious although they are starving, this might be due to complex biological and psychological mechanisms: depletion in tryptophan resulting from a strict diet could relieve anxiety, as suggested by Kaye et al. 
. In addition, other effects such as satisfaction at having lost weight, positive reinforcement from peers 
, or battling against hunger as a source of pleasure and control 
, enable them to experience a degree of “well-being”. However, with time, these effects fade and anxiety and depression re-emerge, along with other rituals and obsessions. It is at this stage that patients are usually admitted to hospitalization. This anxio-depressive recrudescence is also explained by several other factors: the regulation and adaptation of the body to all kinds of nutritional deficiencies and hormonal changes, negative comments concerning extreme thinness, exhaustion, chronicity of the illness and the hospitalization itself. Thus the patient can be caught in a vicious circle that drives him/her to ever-lower BMI, sometimes fatal. In fact, the patient adopts again the first strategy, which is starvation, in an attempt to decrease anxiety and depression, as at the beginning of the illness. Unfortunately, this strategy aggravates the anxiety and depression and the vicious circle described by Garner described 
The present study is a pioneer investigation of relationships between various nutritional indicators and psychological symptoms in severely malnourished AN patients. In contrast with theories set out in the literature, we did not identify any correlations between severely malnourished status and psychological symptoms. However these results suggest several lines of research to confirm this finding. The use of even better nutritional indicators is needed, for example DXA instead of BIA for body composition analysis, and other than albumin and prealbumin proteins as serum markers 
. The development of a precise measure of the scale of weight loss could be beneficial. Screening for vitamin and minerals levels could also help to distinguish symptoms resembling depression or anxiety, such as irritability, moodiness, restlessness, etc, associated with malnutrition (vitamin deficiencies, mineral depletion and decreased food intake 
). These could mediate the effect of malnutrition on psychological symptoms more markedly than the variables explored in this study. Clinicians and the treating team of AN, should be aware that there could be confusion in the aetiology of certain malnutrition symptoms that appear as depression and anxiety symptoms.
The cornerstone of treating AN is still nutrition rehabilitation which should be initiated immediately 
. Nutrition rehabilitation should start first in order to decrease immediately physical complications and psychological well-being. In practice managing co-occurent anxiety or depression symptoms in ED patients will include the specific treatment of ED, that could lower a part of anxiety and depressive symptoms by nutrition rehabilitation, withdrawal from binges and purges, specific psychotherapy (individual or family therapy) and work on the social impact of the illness.
Future studies with a longitudinal design and a follow up on the evolution during treatment are needed to explore variations in nutritional status in relation to psychological symptoms using more heterogeneous samples. For instance, future research should consider including a group of healthy controls or a group of recovered subjects, and the use of further assessment scales for psychological symptoms. More studies are needed to confirm our results.