The aim of the present study was to assess linear growth retardation characteristics in adolescent females with AN, with particular attention to the period post weight restoration. Based on our findings in male patients 
we hypothesized that female AN patients would show significant growth retardation upon hospitalization, and that weight restoration would be associated with a significant, although not complete, catch up growth. Our results support this hypothesis, as growth impairment was indeed prevalent in our group of adolescent female AN inpatients, from admission until attainment of final adult height.
The mean height SDS on admission for all 211 patients was −0.285±1.02, significantly lower than expected in a normal population. In the subgroup of patients with complete growth data, the mean pre-morbid height SDS was not significantly different from the expected in a normal population, whereas mean height SDS on admission was significantly lower. These results are in agreement with Swenne et. al. and Nielsen 
(), and suggest that the decreased height SDS observed on admission reflects actual growth retardation, rather than an increased predilection to AN in females with short stature as suggested elsewhere 
. Furthermore, our findings imply that early signs of AN may not be detected in a significant number of patients, and that the diagnosis of the disorder is often delayed.
Patients admitted at age 13 years or younger, and patients admitted less than one year after menarche presented with more severe growth-impairment compared to patients admitted at an older age. A possible explanation of this finding is that some of the older patients may have already accomplished most of their linear growth prior to the onset of AN.
Weight restoration resulted in accelerated linear growth and improvement in height SDS during hospitalization. Still, complete catch-up growth was not achieved, and the mean final height SDS of our patients was significantly lower than expected in a normal population, and not significantly different from height SDS on admission. Considering the pre-morbid records, our data suggest that compromised final adult height was the result of prolonged malnutrition and not related to pre-morbid short stature. This notion is further supported by the finding that BMI-SDS upon admission was negatively correlated with the difference between height-SDS on admission and at final height, implying that patients with a greater degree of malnutrition upon admission had a lesser degree of catch-up growth.
Several hormonal changes may contribute to growth retardation in AN. These include low thyroxine and triiodothyronine, elevated cortisol, and low sex hormone levels. Growth hormone (GH) resistance - characterized by GH hypersecretion, and low serum levels of growth hormone binding protein, insulin-like growth factor I (IGF-I), and insulin-like growth factor binding-protein 3 (IGF-BP3) - has also been related to growth failure in AN patients. These changes are reversible with weight gain, with a gradual increase in thyroid hormone levels, gonadotropins and IGF-I, and a decrease in GH and cortisol levels (19, 32). In addition, it has been suggested that leptin may act as a skeletal growth factor and induce longitudinal growth (33). Since leptin levels are low in underweight AN patients and increase with re-feeding (32), it is possible that hypoleptinemia may contribute to growth retardation in AN, whereas the increase in leptin levels with weight restoration may contribute to catch-up.
Several smaller studies 
() described the occurrence of growth retardation in AN. However, whether growth retardation seen at the time of diagnosis of AN affects final adult height has not yet been established. Two studies suggested that adolescents with AN can realize their height potential. Pfeiffer et al 
assessed final adult height in 13 males and 58 female AN patients, claiming that most patients reached their expected heights. However, final height percentile was compared with height percentile at diagnosis, with no reference to pre-morbid height. Thus, height deceleration prior to admission could have been missed. Moreover, since only limited information was given regarding the patients' age distribution, it is possible that at least some of them reached final height prior to the onset of AN. Prabbhakaran et al 
followed 63 female AN adolescents for one year. They found that height at study entry, as well as the difference between target height and both follow-up height and predicted adult height did not differ between patients and controls, suggesting the preservation of height potential in female AN adolescents. However, no information was given regarding the premorbid height of the AN patients, and final height was not assessed. Furthermore, this study included patients treated as outpatients that might have had a milder disease course compared to the inpatients included in our study. In keeping with this assumption, the investigators reported that girls with longer duration of illness had lower height SDSs and lower predicted height SDSs 
By contrast, the results of Lantzouni et al 
in 16 females with early onset AN, and of Lanes and Soros 
in healthy children after prolonged period of inadequate caloric intake, were similar to ours. Both groups reported decreased height SDS at presentation. Despite accelerated growth following nutritional rehabilitation, the patients did not achieve their predicted final height 
or their target height 
As emphasized earlier, the occurrence of growth impairment in our patients well before hospitalization suggests a considerable delay in the diagnosis of AN. We hypothesize that weight loss in our patients may have been masked by their reduced height: once growth retardation occurs, the severity of malnutrition may be underestimated because weight is related to the “stunted” height rather than to the pre-morbid projected height, resulting in a weight-per-height/BMI SDS that may seem reasonable. In addition, growth impairment may have not been recognized in adolescents whose heights were in the upper percentiles. This is illustrated in - the patient depicted in this figure shifted from the 90th
to the 50th
height percentile well before AN has been noticed, perhaps because she still appeared normal relative to her peers (). Irrespective of the reason, a delay in the diagnosis of AN in adolescent, and perhaps even more so in pre-adolescent females, may result in severe malnutrition and compromised final height. Furthermore, since AN may be associated with irreversible multi-organ damage including brain atrophy 
, osteoporosis 
, and major adverse obstetric outcomes 
, we postulate that stunting of growth may be viewed as a measurable marker of covert tissue injury. Thus, we believe that early diagnosis of AN and early aggressive nutritional rehabilitation are of utmost importance in order to prevent not only growth failure but also irreversible tissue damage with long-term implications.
Growth curve of a representative patient.
It has been suggested that the hypogonadism found in AN may lead to delay in growth-plate closure 
, allowing, in turn, for compensatory growth with the potential for catch-up during later adolescence. Hence, we hypothesized that patients admitted at a younger age, with seemingly significant growth potential, will show the greatest degree of catch-up growth. However, in contrast to our hypothesis, and similar to the findings of Favaro et al 
, these patients actually had the worst outcome regarding final height. Thus, it seems that prolonged undernutrition, arising (albeit not diagnosed) during early adolescence and continuing around the critical time of peak height velocity, may interfere with growth and result in irreversible stunting despite the delay in skeletal maturity.
Our series is one of the largest series describing growth in female AN patients. It provides important pre-morbid, follow-up and final-height data not previously available in other series. Nevertheless, several limitations should be considered. Our study included only inpatient AN female adolescents, so that our findings cannot be generalized to less severe forms of the disorder. Secondly, we did not include matched controls. Thirdly, only a third of the original cohort had final height data, and premorbid height was reported in only a minority, although no difference was shown between patients with and without final height data in any of the demographic and clinical parameters evaluated. Although we made an attempt to contact all former patients, many refused to come for a final assessment, or have not been located. This was likely the result of our department being a tertiary care center that hospitalizes patients from all over Israel, reducing the likelihood of continuation of treatment in our center following discharge. Finally, Tanner stages of sexual development, bone age studies, parental heights, and results of laboratory tests were not recorded in most of the charts. Therefore, we were unable to calculate the pre-morbid predicted adult height and the mid-parental target height, or to investigate the relationships between pubertal stages and adult height.
Despite these limitations, our study provides strong evidence to support our hypothesis. In particular, the strength of our study relies on the data obtained from the subgroup of patients with pre-morbid and final height measurements (), showing normal pre-morbid height followed by growth deceleration and almost no catch-up growth. These patients present an unusual growth pattern that cannot be explained based on genetic height potential or timing of puberty.
It could be argued that growth deceleration and the subsequent catch-up growth observed in our patients may actually represent the effects of delayed puberty per-se. However, inspection of the growth data makes this unlikely. First, the severity of growth failure observed in some of our patients far exceeds the pre-pubertal growth deceleration observed in children with delayed puberty, which is in the range of 4.0–4.5 cm/year 
. For instance, the patient depicted in had virtually no linear growth for the 2 years preceding her hospitalization (). Secondly, whereas patients with delayed puberty may not reach their parental target height 
, they typically regain or even exceed their own childhood growth percentile 
. In contrast, in our patients, catch-up growth was incomplete, and their pre-morbid height percentile was not reached.
In conclusion, our findings emphasize the importance of early detection of AN in female adolescents, as growth retardation occurring during a critical growth period during puberty may be irreversible. Hence, AN should be suspected in every adolescent female showing decelerated linear growth. Secondly, in keeping with previous studies 
, we suggest that weight restoration geared towards restitution of height to the pre-morbid percentile for age should be initiated as early as possible in the treatment of AN adolescents. We hypothesize that in order to achieve maximal catch-up growth, pre-morbid growth data should be obtained, and target weight should be based on the expected, rather than the measured height percentile at the time of diagnosis. Acceptance of weight as appropriate for height as measured at admission, with no consideration of the pre-morbid or potential height, may result in perpetuation of growth retardation, leading eventually to short stature. Larger scale prospective linear studies, including sexual maturation evaluation and endocrine function tests, are required to verify our observations, to further define factors influencing final height, and to establish the optimal therapeutic approach for optimizing catch-up growth.