In this study, we evaluated the gonadal function of patients suffering from AN by analyzing anthropometric, endocrinological, metabolic and psychological data at the onset of illness, during rehabilitation and after recovery. To our knowledge, this is the first study analyzing clinical and biochemical parameters in the three categories of AN patients with distinct menstrual histories: (i) those with FHA resolving during a stable improvement of the disease conditions, (ii) those with protracted amenorrhea despite BMI and psychological recovery or (iii) persistence of regular menses throughout the course of the disease.

The inclusion and exclusion criteria allowed the selection of patients in whom FHA was apparently related to the critical phase of the AN. It is known that weight and BMI are relevant factors influencing the function of the gonadal axis [13, 25, 26]. However, in our cohort, premorbid BMI, age of disease onset or amount of weight loss was similar among the three groups of patients with variable menstrual history, and in contrast with a previous study which defined premorbid BMI as the main predictor for longer duration of amenorrhea [27]. In addition, despite the patients with persistent amenorrhea experienced the lowest BMI values when compared with the other two groups, they were also the ones to enter FHA at higher BMI values. Indeed, also the logistic regression studying together the factors associated with persistence of FHA found the BMI at which patients experience amenorrhea to be the strongest predictors of the recurrence of menses. This finding, together with the lack of significance of the overall weight loss, points to a higher contribution of the individual predisposition to HPG axis disruption than the effect of the severity of the disease. Accordingly, the last weight at follow-up was statistically higher in patients with persistent amenorrhea even when compared with the weight at which the patients of Group 0 resolved the amenorrhea, thus supporting the thesis that weight is not an independent parameter in determining FHA. Even when we compared Group 2 of the eumenorrheic patients with the other two groups, we found that the weight was not a determining factor for the functionality of the HPG axis. In fact, the lowest BMI at which these patients maintained the menstrual cycle is statistically lower compared to the others.

As far as hormonal parameters are concerned, it is known that leptin correlates with patient's weight and nutritional status and with the amount of fat mass [28,29,30,31,32]. Some authors have questioned such correlations: Miller and colleagues [33] reported hypoleptinemia as independent of fat mass in women with FHA. In our study, we found similar leptin levels at admission, with average values below the lower limit of the healthy population (< 3.6 μg/mL). Leptin concentrations statistically improved at the last follow-up visit only in patients who were menstruating at discharge (Groups 0 and 2). In contrast, the increase of leptin levels was limited and not significant in subjects with persistent amenorrhea (Group 1). Low leptin can account for the decreased pulsatility of LH [30], and the lack of its significant increase justifies the persistence of FHA in Group 1; in this group, low leptin concentrations despite weight recovery are indicative of a different body composition, perhaps due to an unbalanced diet with a low fat content, as previously suggested [33].

Interestingly, the average fT3 values at admission were above the lower reference limit (> 3.7 pmol/L) in eumenorrheic patients (Group 2) and statistically higher (p value = 0.0249) than in the group of patients recovering from amenorrhea (Group 0), despite similar clinical characteristics at admission (Table 1), and tended to be higher than in the group of patients with persistent amenorrhea (Group 1). At the end of rehabilitation, there was a general improvement in fT3 concentration, but they remained below or at the lower limit of the normal range in a consistent number of patients with permanent amenorrhea (Group 1), thus indicating the persistence of a hypothalamic dysfunction despite the increase of body weight which might result from the lack of a significant leptin rise.

Reinehr et al. [34] hypothesized that leptin could be the link between BMI and fT3; this observation appears partially in disagreement with the results of our study because patients maintaining the gonadal function during the critical phase of the disease (Group 2) had low levels of leptin and normal fT3 at admission; moreover, patients with persistent amenorrhea had an improvement in fT3, but not in leptin levels after rehabilitation. Swenne et al. [35] concluded that fT3 can be used as a nutritional indicator and that the severity and the velocity of weight loss are the strongest predictors of fT3 concentrations. Consistently, we found an inverse correlation of fT3 with the amount of weight loss and a positive correlation between fT3 and leptin levels, as previously described [36]. But in addition, here we observed a highly significant positive correlation between fT3 and estradiol, with fT3 levels tending to an earlier normalization before resumption of menses and rise of circulating estradiol in Group 0, thus indicating fT3 rise as an early marker of the resumption of hypothalamic activity.This adds to the evidence already available for IGF1 [37] as opposed to cortisol levels which are reported to be a predictor of hypothalamic inhibition [38]. Still, the usefulness of fT3 levels in this setting should always be put in context of clinical parameters. In eumenorrheic patients of Group 2 at admission, fT3 levels were significantly higher in association with the conserved gonadal function; this combination may indicate the portion of AN patients with a hypothalamic function that is less sensitive to the effects of caloric restriction and weight loss.

The main limitation of our study is represented by the lack of body composition determinations to support and correlate with the leptin measurements, leaving open the possibility for a relevant role of the psychopathological profile of the patients [39]. Finally, only some endocrine/metabolic parameters have been evaluated, but many others (GH/IGF-1, hypothalamus–pituitary–adrenal axis, other adipokines and gastro-intestinal hormones) may contribute to the HPG axis dysfunction. Eating disorder manifestations, quantified by the EDI-2 questionnaire, were similar among the three groups, both at admission and discharge. However, when we evaluated the improvement of the score within the three groups, we found that only the patients recovering gonadal function (Group 0) showed a significant general improvement of the score involving almost all the categories. The patients with persistent amenorrhea (Group 1) had an intermediate improvement. These results are in partial agreement with those of Miller et al. [40].

In conclusion, to our knowledge this study is the first comparing three different categories of AN patients defined on the basis of menstrual activity. We found that the premorbid weight and the weight loss during AN are not the main determinants of the GnRH neuron activity, but a more important role is played by the minimal BMI reached during the critical phase of the disease. Nevertheless, a significant portion of patients maintain menstrual cycles even at BMI significantly lower than the other two groups, indicating the significant role played by individual factors in the definition of the set-point of hypothalamic dysfunction. This view is reinforced by the findings in patients with persistent amenorrhea (Group 1) showing a body weight higher than in patients resuming their menses, both at the amenorrhea onset and during the follow-up, further supporting the hypothesis that persistent FHA may be a consequence of an intrinsic fragility of the HPG axis, thus supporting the multifactorial origin of the FHA in the AN patients [41, 42].

At admission, fT3 levels are statistically higher in eumenorrheic patients and fT3 variations during rehabilitation often precede the restoration of gonadal function in Group 0, thus indicating fT3 rise as an early indicator of FHA recovery.