Anorexia nervosa (AN) is characterized by a difficulty to maintain minimal weight, a fear of gaining weight, a disturbed body image, often associated with denial based on criteria from the Diagnostic and Statistical Manual of Mental Disorders (1). Two subtypes have been descripted, the restricting subtype with primarily loss of weight through significant reductions in caloric intake, and the binge-eating subtype with recurrent binge-eating and purging through self-induced vomiting or laxative misuse. AN is a potentially serious disease; a mortality rate of 5–10% at 10 years has been reported, making it the psychiatric disorder with the highest mortality (2), although more recent studies report less severe mortality rates (3). AN is associated with multiple and severe somatic complications related to malnutrition including bradycardia, hypotension, anemia, and hormonal imbalance (4). AN is also associated with significant psychiatric comorbid conditions, including anxiety, depression, obsessive–compulsive disorders, and excess physical exercise referred as hyperactivity (5, 6).

Low bone mineral density (BMD) is a frequent complication in AN: 38% of patients with AN suffer from osteoporosis and 92% from osteopenia (7). Impaired bone metabolism in AN is multifactorial and includes hormonal changes (8) and reduction of anabolic effect of muscle contraction on the bone, which is directly related to lean body mass (9). Conventional bed rest imposed to patients with AN is probably detrimental on BMD, since immobilization is a well-established risk factor for bone loss (10). Historically, the main treatment of low BMD during AN relies mainly on compensation of calcium-vitamin D depletion (11), but has a limited efficacy in the absence of efficient refeeding (12). Furthermore, the prevalence of vitamin D deficiency among adolescent girls with AN have been reported to be only 2% compared with 24% in healthy controls (13) probably due to an increased use of supplements. Moreover, the benefit of bisphosphonates for the treatment of BMD during AN is not established (14), although Miller et al. have previously reported a significant increase in BMD in adult women with AN with risedronate compared with placebo in a 1-year follow-up study (15). Other studies have reported a decrease in bone turn-over with bisphosphonates, particularly in adolescent patients with AN (16).

Thus, recovery of BMD in AN is described as a slow process with complex interactions between hormonal and nutritional factors (17). There are controversial point of views in the literature regarding the restoration of bone mineralization after recovery in AN. The purpose of this study was to assess changes of BMD at 3 years in patients with AN and to explore the relationships between body composition, physical activity, and BMD at the first visit and at follow-up.

In this study, at the first visit, more than 50% of the population exhibited low BMD either osteopenia or osteoporosis. This result is in accordance with previous studies and confirm the major impact of AN on bone health (18, 19). Moreover, we found a positive correlation between hip BMD and duration of AN as reported in the previous cross-sectional studies (20–22). However, there is emerging evidence that loss of BMD is rapid, occurring relatively early in the disease, even in adolescents with a brief duration of illness. Legroux-Gérot et al. found that spine bone loss correlated negatively with durations of AN in a 1-year follow-up study (23), this finding could be explained by a very rapid process of bone loss at the onset of the disease. In our study, hip BMD was also correlated positively with BMI. This finding is comparable with those reported in the literature (21).

At 3 years follow-up, despite the increase in BMI, bone condition did not significantly change and we did not find any difference between the AN subtypes. Thus, weight recovery did not influence the reestablishment of concrete bone mineralization in our population. Many contradictory results are reported in the literature regarding the reestablishment of bone mineralization after recovery in AN. Some trials have found that weight recovery was related to BMD improvement (18, 24, 25), whereas other recent studies have reported that bone recovery was not influenced by BMI (26, 27). Potential mechanisms involved in the lack of bone recovery after weight restoration are not well understood. Physical hyperactivity has been associated with low bone mass in AN (28). Moreover, increased marrow adiposity, which correlates with low bone density, has been observed in AN (29).

Moreover, we attempted to determine whether or not associations existed between change in BMD at 3 years follow-up and body composition changes. We found a significant positive association between BMD improvement and FM. Indeed, in our population, increase in BMI was mainly related to an increase in FM as previously reported in AN (26). The pathophysiological mechanisms between weight, fat tissue, and bone tissue are increasingly better understood. Leptin is an anorexigenic adipokine mainly produced by fat tissue and has effects on bone. Central leptin is deleterious to the axial skeleton (30, 31) but peripheral leptin has bone anabolic effects with possible osteoclast inhibitory effects (32). AN is often associated with low leptinemia (33), probably an adaptive mechanism to increase appetite, and lower leptin levels are associated with lower FM and bone density measures (34). Furthermore, in a recent study, hypoleptinemia persisted despite the full recovery of FM in an animal model of anorexia (35). Moreover, recent findings are in accordance with our results showing that increasing FM reversed bone loss osteopenia (36).

Lower lean mass is also an important determinant of lower bone density and impaired bone structure in adults and adolescents with AN (37). Previously, Soyka et al. have shown that increases in lean mass following weight gain are strongly predictive of coincident increases in bone density in adolescents with AN (9). In our study, FFM gain was not a significant predictive factor of BMD improvement, even if a trend was observed (2.4 ± 2.9 vs −0.1 ± 2.9 kg, ns). Interestingly, patients exercising had higher FFM gain than inactive patients (+1.3 ± 3.3 vs −0.8 ± 3 kg, p = 0.06). A few studies have suggested that moderate exercise may be protective against osteoporosis in women with AN, whereas pathological hyperactivity may be harmful (38, 39). It is generally accepted that exercise help in maintaining BMD in postmenopausal women and increase BMD of the spine and hip in women with osteopenia and osteoporosis (40). Therefore, some authors have hypothesized that, through increasing lean body mass, physical activity may improve BMD during AN management to compensate in part for the hypogonadism (28). However, in our study, physical activity did not appear as a predictive factor of BMD improvement, maybe because of the small population. Thus, prospective and controlled studies are needed to better understand how adapted physical activity may influence BMD evolution during the treatment of AN and may be integrated in a multimodal preventive or curative treatment of low BMD.

Because of its retrospective nature, the main limit of this study is the lake of correlation with biological parameters such as leptinemia or 25-OH vitamin D levels. Moreover, we did not provide detail on either patient menstrual status or type and intensity of exercise. However, this trial has some strength: the long-term follow-up, the evaluation of bone, body composition, and impact of physical activity which have been poorly studied in AN.

In conclusion, our findings suggest that the reestablishment of normal bone values is not related to the increase in body weight, at least after 3 years. Duration of AN, FM, hormonal factors seem to play an important role in the pathophysiological mechanism of osteoporosis and osteopenia in AN. Further trials are needed to better understand this complex pathways in AN.

This project was approved by the “Local Ethics Committee for Non-Interventional studies” (CERNI, Comité d’Ethique local pour la Recherche Non Interventionnelle). This study complies with the International Declaration of Helsinki and we obtained a written consent from all the patients.

NA, MC, and SG designed the study. NA, PJ, AR, PD, and SG included patients. MC and JC performed BMD and body composition acquisition. NA and SG performed statistical analysis. NA wrote the first draft of the manuscript. All the authors performed critical revision of the manuscript and approved the final version of the manuscript.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.