Androgens impact on psychopathological variables according to CPRS, and EDI 2 scores: In women with bulimia nervosa, and eating disorder not otherwise specified

Bulimia nervosa (BN) is characterized by binge eating, compensatory behavior, over-evaluation of weight and shape, which often co-occur with symptoms of anxiety and depression. Depression is the most common comorbid diagnosis in women with eating disorders. The role of androgens in the pathophysiology of depression has been recognized in recent years. However, the research on psychopathological comorbidity and androgen levels in bulimic disease is sparse. This study aimed to investigate, if there were any correlations between the androgens, testosterone (T), dehydroepiandrosterone sulphate (DHEAS), androstenedione (A4), 5 α -dihydrotestosterone, (5 α - DHT), and test scores of psychopathological variables, in women with bulimia nervosa (BN), eating disorder not otherwise specified of purging subtype (EDNOS-P) assessed by CPRS, and EDI 2. Women with DSM-IV diagnosis of BN (n = 36), EDNOS-P (n = 27), and healthy control subjects (n = 58) evaluated for fifteen psychopathological variables, i.a. depressive symptoms, impulsivity, personal traits, as well as serum androgen levels. All women were euthyroid, and polycystic ovarian syndrome (PCOS) diagnosis was excluded. Although androgen levels were almost equal for all three groups, significant correlations between core psychopathological symptoms (9/ 15) of bulimia nervosa and the most potent endogenous androgen, 5 α -DHT, was found only in the EDNOS-P group. The role of 5 α -DHT in women is not fully elucidated. Both animal and human studies have shown that the brain is able to locally synthesize steroids de novo and is a target of steroid hormones. Maybe these results can be interpreted in the light of differences in androgen receptor variability, metabolism and origin of T and 5 α - DHT.


Introduction
BN is an eating disorder characterized by recurrent binge eating episodes and compensatory behaviors, such as purging, misuse of diuretics or laxatives, excessive exercise, or fasting, [1]. Recurrent binge eating episodes are the core feature of BN, and these are characterized by the consumption of an unusually large amount of food associated with the sense of loss of control overeating during the episode [2].
Eating disorder not otherwise specified of purging type (EDNOS-P), fulfils all the criteria for BN except that the binge eating, and inappropriate compensatory behaviour occur at a frequency of less than twice a week or for a duration of less than three months according to the Diagnostic and Statistical Manual of Mental Disorders-IV [2]. The category "eating disorder not otherwise specified'' (EDNOS) in DSM-IV is restricted to eating disorders of clinical severity that do not completely fulfil the criteria for anorexia and bulimia nervosa. Over of the years women can be afflicted with various subtypes of eating disorders. This actual study was performed prospectively by using the DSM-IV, before the introduction of DSM-V. There appears to be sufficient scientific evidence for EDNOS to be considered as an eating-disorder category of substantial clinical severity. [3]. Depressive symptoms, impulsivity, self-defeating, and dissocial personality traits are regarded as the psychopathological core of BN. [4], and these core symptoms seem to be accompanied by an increased risk of, among others, depression, anxiety, low self-esteem. Fear of weight gain, overevaluation of body image, and weight, feeling fat is identified as highly central to BN psychopathology, and these psychopathological variables, create a vulnerability for anxiety and depression. [5][6][7].
Bingeing and purging behavior are often associated with affect and feelings of shame, guilt, and dysphoria, and elevated depressive symptoms increase the risk for onset of bulimic pathology. [6,8]. Varying, estimates of psychiatric comorbidity, 64.1-88.0%, has been demonstrated in women with BN [1,9], particularly mood and anxiety disorders, with depressive disorder being one of the most common comorbid diagnoses [1,10,11].
Recent research suggests that androgens may play a role in the pathophysiology of depression, mood disorders, and associations between testosterone levels, and depressive symptoms has been observed by several groups, [12,13]. Major depressive disorder is more prevalent in women, [14], and an increased well-being, in women as well as in men, has been shown with androgen substitution, [15]. In addition, elevated testosterone levels have been noted in individuals with increased impulsivity, aggression, and in depressed individuals than in the general population. [16]. Carré 2017. For example, hyper-androgenic state, such as polycystic ovarian syndrome (PCOS) seem to be associated to mood disorders, [17,18]. Given the occurrence of high rates of anxiety and/or depression in women with PCOS, [19], women with PCOS diagnosis were excluded in this study.
Besides its reproductive functions, testosterone has many additional effects on the central nervous system [20,21]. Testosterone seems to affect neurotransmitters, including γ -aminobutyric acid (GABA), dopamine, and serotonin (5-HT) which may underlie some of testosterone's protective benefits on psychological well-being, [22]. The major circulating androgens include dehydroepiandrosterone sulphate (DHEAS), dehydroepiandrosterone (DHEA), androstenedione (A4), testosterone (T), and 5α-dihydrotestosterone (5α-DHT) in descending order of serum concentration, though only the latter two bind the androgen receptor (AR). The other three steroids are considered as pro-androgens [23]. The role of 5α-DHT, in females is still unclear, although, T, and 5α-DHT are formed by de novo synthesis in the ovaries, the adrenal cortex, and by peripheral conversion of androgens 4-androstene-3,17-dione (A-4), dehydroepiandrosterone (DHEA) and its sulfate (DHEA-S) in the ovaries, liver, skin, central nervous system (CNS), [24]. Besides, T also the major circulating C19 steroid DHEAS serves as a precursor to 5α-DHT via conversion to T, and A-4, [25]. The 5α-DHT is the most potent endogenous androgen and has a relative binding affinity for the androgen receptor (AR) that is about 4 times that of T, [26].
A large fraction of 5α-DHT produced in the liver and is metabolized to DHT-glucuronide prior to subsequent entry into the circulation, [27]. The anti-depressive effect of testosterone (T) has been, attributed to actions of 5α-androstane-3β,17β-diol, and its aromatized metabolite, estradiol-17ß, [20,28]. Testosterone also seems to enhance dopamine release in the mesolimbic system [29], which may protect against depression and the associated decrease in dopamine activity in reward-related brain pathways. Although, testosterone has been studied in relation to depression, increased impulsivity, aggression, [27], the molecular mechanisms associated with these effects are not clear. Both high, and low levels of testosterone seem to affect eating behavior and mental state. Depressive symptoms are an important risk factor, and consequence of binge eating and purging behavior in bulimia nervosa (BN) [30]. However, whether sex hormones directly or indirectly play a role in the underlying cause of these abnormalities remains unknown. Data on psychopathological variables, in women with BN, as low self-esteem, disturbed interoceptive awareness, body dissatisfaction and co-morbid disorders, and the role of androgens are scarce, and have included either unspecified "bulimic disease" or women with BN only [29].
In a recent study, correlation analysis indicated profound differences in the origin of androgens, T, and 5α-DHT, between women with BN and EDNOS-P [31]. These results might involve different conversion of these steroids in the CNS. Accordingly, the aim of this study was to investigate if there were any correlations between the androgens testosterone (T), dehydroepiandrosterone sulphate (DHEAS), androstenedione (A4), 5α-dihydrotestosterone, (5α-DHT), and psychopathological variables, in women with BN, and EDNOS-P. For the assessments of the psychopathological variables, CPRS and EDI 2 test scores were used.

Subjects and Methods
Women with bulimia and healthy regularly menstruating control subjects, matched for age and body mass index (BMI), were recruited by media advertising and among students or hospital staff. Thirty-six women fulfilled the criteria for bulimia nervosa, including frequent binge eating episodes and the regular use of inappropriate compensatory behaviors such as self-induced vomiting, laxative abuse, or periods of fasting to prevent increase in body weight as defined by DSM-IV. Twenty-seven women who fulfilled all criteria for BN except for that the binge eating, and inappropriate compensatory behavior occurred less than twice a week or for a shorter period than three months and the compensatory mechanism for weight increase was by purging only, allocated into EDNOS-P group. In addition, 58 matched, healthy women were included in the study. Inclusion criteria for all three groups were: Being between 18 and 40 years old, BMI 18-28 kg/m 2 and no medication, including psychotropic drugs, or psychotherapy was allowed. Exclusion criteria for all groups was any on-going psychiatric illnesses; physical conditions known to influence eating pattern and pregnancy and for healthy participants also the presence of eating disorders. All participants had normal thyroid function, and the cortisol levels was within normal range. The biological sex for all participants was as women, and the gender identity was as females.
The study was approved by the local ethics committee at the Karolinska University Hospital, and written informed consent was obtained from all women before participation.

Assessments
The study was prospective, and the assembly of all participant data, and blood samples was completed before DSM-V was applied. Besides DSM-IV, compiled by the American Psychiatric Association [2], the Comprehensive Psychopathological Rating Scale (CPRS), and Eating Disorder Inventory 2 (EDI-2) were used, by a trained research and clinical psychologist. Participants were assessed regarding personality disorders by using the Semi-structured Schedule of Clinical Interview for DSM-IV-TR Axis II Personality Disorders (SCID-II). The SCID-II is a useful tool for measuring personality disorders according to DSM-IV criteria. Thereafter, the psychopathological variables were assessed by Comprehensive Psychopathological Rating Scale (CPRS) which is a semi-structured interview covering a wide range of psychopathological symptoms and signs, to understand their severity. CPRS has good reliability, and validity for these evaluations [32,33].
All participants, inclusive healthy control subjects were evaluated face-to-face, with respect to eating disorders, and all the psychopathological variables in the study. Eating Disorder Inventory 2 (EDI-2) was used which is a self-report questionnaire with 36 items generating four subscale scores (restraint, eating concern, shape concern and weight concern, as well as global score) which is the average of the four subscales. The respondents rate each item on a seven-point rating scale (i.e., 0-6) indicating the number of days out of the previous 28 in which specific behaviour's', attitudes or feelings occurred. The EDI-2 is widely used in eating disorder research and has also been validated for clinical and general populations [34].

Serum analysis
Venous blood samples for serum hormone analysis were taken between 08.00 and 09.00 a.m. in early follicular phase (cycle days 1-5) in normally menstruating subjects and arbitrarily in oligo-/amenorrhoic subjects. Serum concentrations of testosterone (T), dehydroepiandrosterone sulfate (DHEAS), sex hormone-binding globulin (SHBG), were determined by established assay methods and as specified below [35], as were the methods for calculation free testosterone (fT). Serum testosterone (T), and androstenedione (A-4) were determined by radioimmunoassay (RIA) using commercial kits obtained from Diagnostic Products Corp. (Los Angeles, CA, USA) ('Coat-a-Count®': T and A-4). In the assay of A-4 the RIA procedure was preceded by an extra step including extraction with diethyl ether, evaporation and dissolving the residue in a zero-calibrator supplied by the manufacturer of the kit. This step was included to avoid possible cross-reactions with water-soluble androgen metabolite conjugates. Serum levels of sex hormone-binding globulin (SHBG), dehydro-epiandrosterone sulfate (DHEAS), and cortisol, were determined by direct chemiluminescence enzyme immunoassay using commercial kits from Diagnostic Products Corp. (Immu-lite®). Detection limits and within-and between-assay coefficients of variation were respectively 0.1 nmol/l, 6% and 10% for T; 0.14 nmol/l, 5.7% and 8.4% for A-4; 0.2 nmol/l, 6.5% and 8.7% for SHBG; 0.8 μmol/l, 8.2% and 12% for DHEAS; 5,5 nmol/l, 7,8 and 7,8% for cortisol.
Apparent concentrations of free testosterone (fT) were calculated from values for total T, SHBG and a fixed albumin concentration of 40 g/l by successive approximation using a computer program based upon an equation system derived from the law of mass action [36]. Since the ratio between total T and SHBG (T/SHBG ratio, 'free androgen index') also is considered a useful index of biologically active T (ESHRE/ASRM, 2004), this marker was also included in the calculations.

Statistical analysis
Normally distributed data are presented as arithmetic mean and SD. For Table 1 differences in age, body weight, BMI, age of onset of bulimia, differences were tested by factorial ANOVA, duration of bulimia, shown as median and 25th -75th percentile, was tested by Kruskal-Wallis test, and frequency of previous overweight, menstruation data, ever been pregnant, current employment, by chi-2 test.
Since the CPRS, and EDI-2 scores in Tables 2, 4, and Table 5, were not normally distributed, comparisons were performed by Kruskal-Wallis test followed by post hoc test by Mann-Whitney U-test or for correlations, Spearman's rank correlation test, was used. The occurrence of CPRS and EDI-2 zero scores in some women in the study population made logarithmic transformation not feasible. For Table 3, hormone levels showed skew distributions and comparisons were performed by Kruskal-Wallis test. Between two groups comparisons in Table 2 were performed post hoc Mann-Whitney U-test due to actual distribution. The significance level was set at p < 0.05.

Results
Clinical data for women with BN, and for EDNOS, and healthy control subjects are given in Table 1. Actual BMI and frequency of previous overweight were numerically slightly higher in the EDNOS-P group than in bulimic women and the control subjects. This difference was not significant. There were no significant differences between the three groups with respect to menstruation data, educational level, and current employment. Psychopathological variables according to CRPS and EDI-2, are shown in Table 2. Compared to healthy control subjects, both BN and EDNOS-P groups showed significantly higher scores for psychopathological (15/15) variables which is in accordance with the inclusion Table 1 Clinical data, for women with bulimia nervosa (BN), with eating disorder not otherwise specified, purging type (EDNOS-P), and healthy participants. Arithmetic mean + SD, median (P25 -P75) or frequency. See the statistics section. There were no significant differences between the groups.  Psychopathological variables according to CPRS, or EDI-2, in healthy participants, in women with bulimia nervosa (BN) and in women with eating disorder not otherwise specified, purging type (EDNOS-P). Median and range or arithmetic mean + SD. Significances of differences between women with eating disorders and controls are denoted by * = p < 0.05; * * = p < 0.01 and * ** = p < 0.001. Significances of differences between women with BN and EDNOS-P are denoted by a = p < 0.05; b = p < 0.01 and c = p < 0.001. Comparisons performed by Kruskal-Wallis test followed by post hoc test by Mann-Whitney U-test.  Table 3 Previous depression diagnosis, previous treatment for eating disorders (Eds), and disturbed impulse regulation, expressed as per cent of the group in healthy controls, in women with bulimia nervosa (BN) and in women with eating disorder not otherwise specified, purging type (EDNOS-P). Significance of differences between women with eating disorders and controls are denoted by * * = p < 0.01 and * ** = p < 0.001. Significance of differences between BN and EDNOS-P are denoted by b = p < 0.01 and c = p < 0.001. criteria. As expected, women with BN scored significantly higher eating problems, lower self-esteem, interpersonal distrust, disturbed interoceptive awareness, and social insecurity than EDNOS-P. Table 3 shows previous depression diagnosis, previous treatment for eating disorders (Eds), and disturbed impulse regulation. BN group had a significantly higher frequency of previous depression diagnosis compared to those with EDNOS-P, but both eating disordered group had significantly higher previous depression diagnosis vs control subjects. Both groups had depression diagnosis previously, but not the last 2 years prior to inclusion in the study. In addition, ongoing depression diagnosis was an exclusion criterion. Both ED groups had significantly higher ongoing disturbed impulse regulation compared to control subjects. Serum androgens and SHBG, are given in Table 4. There were no significant differences between the three groups. Correlations between 5αdihydrotestosterone, 5α-DHT, and psychopathological variables in the two groups of bulimic women are given in Table 5. Significant positive correlations were found between DHT, and 9/15 of investigated psychopathological variables in the women with EDNOS-P. Interestingly there were no such correlations in the women with BN. Variables with significant correlations to 5α-DHT in EDNOS-P group were: depression diagnosis, low self-esteem, disturbed introceptive consciousness, asceticism, social insecurity, fear of adulthood, interpersonal distrust, eating problem and anxiety. Psychopathological symptoms were almost none existing in control subjects ( Table 2) and none of them fulfilled criteria for manifest pathology diagnosis. Therefore, correlation analysis did not reveal any significances in this group (data not shown). All androgens listed in Table 4, were tested, and the only significant correlation between any other androgens than 5α-DHT, and any psychopathological symptom, was a correlation between A-4 and anxiety in the BN group, (r s =0.41, p < 0.05). No significant correlations between psychological variables and age were found.

Discussion
It has been proposed that BN should be conceived as a complex condition which integrates among many eating dysfunctions, psychopathological symptoms, self-defeating and dissocial personality traits [37,38].
This work sought to demonstrate whether there were any associations between the androgens: A4, DHEAS, DHT, T, and psychopathology in women with BN, EDNOS-P, and healthy control subjects. The major finding was strong correlations between serum concentrations of the most potent endogenous androgen, DHT, and a number of psychopathological variables, in women with EDNOS-P but not in those with BN, (Table 5). This result might indicate differences in androgen metabolism i.a. in the CNS, between subgroups of women with bulimia nervosa.
Both animal and human studies had shown that the brain can locally synthesize steroids de novo and is a target of steroid hormones [39,40]. Significant increase in serotonin transporter binding by testosterone has been shown [41], thus suggesting that testosterone may trigger depressed mood by inducing synaptic serotonergic depletion via increased neuronal re-uptake. It is hypothesized that the interaction between low serotonin and high testosterone levels in the central nervous system has a significant effect on the neural mechanisms involved in the expression of aggressive behavior. Aggressive behaviors have been reported to be more frequent in people with eating disorders, especially BN, than in the general population [42]. Further, high aggression levels have been linked to decreased serotonin (5-HT) neurotransmission [43]. It seems that testosterone modulates serotonergic receptor activity in a way that directly affects aggression, fear, and anxiety [44], and women with BN impulsive behaviors have an inverse relationship to serotonin functioning [45].
Low self-esteem is one of the psychopathological variables, which seems to contribute to body dissatisfaction in women with BN [46,47]. Individuals with higher levels of anxiety, and depressive symptoms seem to use disordered eating as a dysfunctional regulation strategy to cope with emotions [48]. Depression symptom severity, and androgens seem to have a positive linkage [40,49], and dysregulation of hypothalamic-pituitary-adrenal (HPA) system appear to have an important role in the pathophysiology of depression. The adrenal hormones, especially cortisol and DHEA increase during a depressive episode [40], and it is stipulated that DHEA may have neuroprotective effects in the brain, diminishing the liability for cortisol to damage neurons and decreasing the risk for psychopathology, and may have direct effect on depression [50]. High as well as low levels of cortisol in individuals with depression have been observed [51,52]. However, serum cortisol levels were within normal range for the BN and EDNOS-P groups. Further, the body weight is also a factor that may be associated with depression, mediated by both high and low free testosterone levels [53]. Disturbed attitudes towards weight, body shape, and eating play a key role in the origin and the maintenance of eating disorders [54]. There were no significant differences in terms of body weight between the three groups in this study, and with BMI values within normal range. As for all androgens analyzed, even free androgen-index, was similar in the BN, EDNOS-P, and control subjects in the current study.
Testosterone's action is diverse due to its transformation intracellularly into 5α-DHT and estradiol by the enzymes 5a-reductase and aromatase respectively. Several animal experiments suggest that 3βAdiol is anxiety reducing, cognitive enhancing [55], and anti-depressive [56]. Another endogenous ligand of ERβ is 3βAdiol is an estrogenic metabolite of 5α-DHT, [57] and the binding affinity of 3βAdiol is stronger for ERβ than for ERα [58]. Women are almost twice as likely as men to have depression [59][60][61], and this sex differences may be explained by 3βAdiol and Δ5-diol levels. Unlike to T, and free T, the 5α-DHT levels do not change across the menstrual cycle. The blood samples were collected during the follicular phase of the menstrual cycle for all participants.
High A4 levels have been associated with pessimism, anxiety, and suicidality [62]. In current study an association between A4, and anxiety was found, but most of the associations were between 5α-DHT, and several psychopathological variables in women with EDNOS. Women Table 4 Serum androgens and SHBG in women with bulimia nervosa (BN) and with eating disorder not otherwise specified, purging type (EDNOS-P). Median (P25 -P75). There were no significant differences between the groups. Comparisons were performed by Kruskal-Wallis test. See the statistics section.  Table 5 Correlations between serum 5α-dihydrotestosterone, and CPRS, or EDI-2 scores of psychological variables in women with bulimia nervosa (BN) and with eating disorder not otherwise specified, purging type (EDNOS-P). Spearman's rank correlation coefficients * = p < 0.05, * * = p < 0.01 < and * ** = p < 0.001 respectively. See statistics section. with BN seem to exhibit greater general psychopathology than those with EDNOS [63], (Table 2). However, a comprehensive meta-analysis has shown that EDNOS-P was associated with substantial psychological and somatic morbidity, comparable with the specified eating disorders, which also has been indicated earlier by other groups [63,64]. Testosterone can be either reduced by 5α-reductase to the more potent androgen 5α-DHT or aromatized by aromatase to estradiol converting the androgen to estrogen activity. In the brain, 5α-DHT can be further metabolized to 5α-androstane-3α,17β-diol (3α-diol) and to 5αandrostane-3β,17β-diol (3β-diol), neuroactive steroids possessing neuromodulatory activity [65].
The mechanisms for T actions are not fully identified at present but a tentative explanation may be that these strong correlations between psychopathological variables, and 5α-DHT, in women with EDNOS-P, may indicate differences in origin of androgens [31]. Other possible rationales might be that the intracellular control of T and 5α-DHT concentrations is mediated by a host of local metabolic pathways. Organ differences in receptor binding of T and 5α-DHT result, in part, from relative differences in intracellular concentrations of these androgens rather than from differences in receptor affinities alone [66]. In both muscle and fat cells intracellular androgen concentrations (e.g., DHEA, T, and 5α-DHT) are mediated by biochemical mechanisms that tightly control local androgen levels [67]. Furthermore, it has been demonstrated that high concentrations of intracellular T can shift AR binding away from 5α-DHT by mass action [68,69]. As stated above, this prospective study was carried before the introduction of DSM-V. The diagnostic criteria comparative studies between the DSM-IV, and DSM-V have shown that EDNOS-P diagnosis may shift over to BN by using DSM-V, as shown by Fisher et al., (2015), [70]. A subset of women with eating disorders might have physiological differences in T, 5α-DHT, metabolism which might reflect variations in AR receptor distribution, ligand-induced conformational changes to AR that effect stabilization, local hormone synthesis and metabolism, and among many other factors [71,72].

Conclusions
A strong correlation between the most potent endogenous androgen, 5α-DHT and depressive symptoms, low self-esteem as well as seven other psychopathological variables in women with EDNOS-P, was found.
The current results, and the possibility of different origin of androgens in women with EDNOS-P vs BN and control subjects, may be linked to T and/or 5α-DHT's metabolism, receptor binding, bioavailability, activity in the CNS as well as in other target organs. Androgen effects on the pathophysiology of bulimia nervosa and subgroups remain to be further elucidated.

The strengths of the study
Research within the area of well-defined group of women with bulimia nervosa, and co-existing psychopathological variables. Women with BN, EDNOS-P, and healthy participants were evaluated face-toface, with respect to eating disorders, by a trained research and clinical psychologist. Each group was homogenous from a clinical point of view. The hormone samples were taken during the follicle phase of menstruation, and the analysis performed in the same laboratory, and under the same conditions, by laboratory expert with extensive knowledge in steroid biochemistry. All participants had body weight within the same range, were euthyroid, and women with PCOS were excluded.

Limitations
The study has some methodological limitations. First, the groups were not too large. Secondly, the results were based on single point hormone concentrations. Hormone analysis repeatedly over a period would have strengthened the results.

Funding
This work was supported financially by the Swedish Research Council (20324), the Stockholm County Council and Karolinska Institutet, Stockholm, Sweden.

Declaration of interest
The authors report no declarations of interest.

Data Availability
Data will be made available on request.