Bone Mineral Density at Extremely Low Weight in Patients With Anorexia Nervosa

Pauline Bemer; Laura Di Lodovico; Ohanyan Haykanush; Hélène Théodon; Karine Briot; Robert Carlier; Marika Dicembre; Maéva Duquesnoy; Jean-Claude Melchior; Mouna Hanachi


Clin Endocrinol. 2021;95(3):423-429. 

In This Article


In this study on 97 extremely malnourished inpatients with AN, 51% had Z-score ≤-2 SD at lumbar spine and 38% had Z-score ≤-2 SD at femoral neck. Low BMD was significantly associated with low FFM, earlier onset of disease and restricting AN for lumbar spine. FFM, Illness duration and age predicted low BMD at femoral neck. Fractures, whose prevalence was estimated at 10.3%, were predicted by low plasma sodium concentrations, low femoral neck Z-score and illness duration.

This is, to our knowledge, the first study on a large cohort of adult patients at extremely severe stages of malnutrition. On average, participants were older and more severe compared with previous studies on this subject.[8,24–26]

The prevalence of low BMD at lumbar spine was 51% with an average Z-score of −2.2 SD. Pre-existing research reports lower prevalence rates, ranging from 21% to 40% depending on population age and diagnostic criteria (T-scores vs Z-scores).[9–11] Higher prevalence found in the current study may be explained by the adopted criteria for low BMD and by the characteristics of this population, at increased risk of BMD loss by virtue of older age, extreme malnutrition and chronicity.

In line with previous studies, the decrease in BMD was predominant in the axial district. This finding can be explained by the preferential damage of trabecular bone in AN. Predominant BMD loss at trabecular bone sites has been attributed to the profound oestrogen deficiency typically found in AN,[10,14] what also explains the correlation found between amenorrhea duration and BMD decrease in this study.

This study confirms low FFM as a main predictor of BMD loss in AN.[13] Lean mass increases BMD by the tensile forces exerted by muscles on the bone attached[14] and by a two-way communication through molecular signalling between the two tissues.[27] It is thus not surprising to find a parallelism between the loss of BMD and that of muscle strength in extremely malnourished patients with AN, both characterized by a predominant involvement of axial regions.[28]

The correlation between precocity of onset and severity of BMD loss is a well-established finding and can be understood by the attainment of peak bone mass during the adolescent period. In this window, it is crucial to maximize bone accrual, and deficits resulting in a suboptimal peak bone mass can be permanent.[29] Illness duration is an additional independent predictor of low BMD, influencing bone structure through prolonged malnutrition and endocrine impairment.[11]

Restricting type AN is a confirmed factor of BMD decrease. This association may be explained by more important nutritional deficiencies than in patients with binge eating/purging type.[30] Moreover, these patients tend to be more malnourished than patients of binge eating/purging subtype, what may additionally explain this finding.[7]

The association between low sodium levels, loss of trabecular and cortical BMD and fractures is already mentioned in current literature, yet poorly explored. Hyponatremia is a frequent finding in AN, due to excessive water intakes dues to behaviours of potomania,[5] and to a decreased ability of kidney to excrete free water because of low solute intakes.[31] The main effect of hyponatremia on bone is a marked activation of osteoclastic bone resorption.[32] In this study, the association between low sodium, decreased femoral neck BMD and fractures was confirmed, but with relative sparing of lumbar spine. This finding deserves further exploration by molecular studies, in order to confirm or infirm the existence of a cortical pattern of bone damage induced by hyponatremia.

As in previous studies, low vitamin D levels were not associated with low BMD in our population.[7] Vitamin D deficiencies play a role in different forms of bone pathology, such as osteoporosis and osteomalacia, both objectified by BMD loss on bone densitometry. One of the explanations of this negative finding could be vitamin D supplementation,[7] regularly administered in our unit as well as gonadal supplementation (ie oestrogenic transdermal replacement), given their potential benefit on BMD.[34]

Surprisingly, and in contrast with previous studies, BMI did not emerge as a predictor of low Z-scores. This may be due to the small BMI variance of this extreme population. Nevertheless, it is worth noting that a lack of moderation effect of BMI on BMD loss in AN was reported by a recent meta-analysis.[11]

The prevalence of fractures was 10%, creating a discrepancy with current literature that sets prevalence of fractures in AN at 30% in cohorts of adolescents and young adults.[13] This finding is even more surprising taking into account that our population was older, more chronic and more cachectic.[14,15] On the other hand, several caveats deserve to be introduced. First, this study examined fragility fractures that are difficult to diagnose and often asymptomatic.[33–35] The retrospective nature of the study and the absence of a systematic radiologic assessment could thus explain this apparent discrepancy with literature and the potential underestimation of fracture prevalence in this study. It is also worth noting that no guidelines for systematic fracture research exist to date. Finally, despite a consistent body of evidence on the subject, age was not confirmed as an independent factor of fractures in this population, neither was BMI. The negative finding of BMI, on the other hand, may be due to a recruitment bias, since extremely low BMI (<13 kg/m2) represents a major criterion of admission in our unit, specialized in intensive management of extreme malnutrition. Besides confirming a role of low BMD and illness duration on fracture incidence, the results of multivariate analysis also suggest chronic hyponatremia as a factor of fracture risk in extremely severe patients.

The main weakness of this study is its retrospective nature leading to missing data, decreasing the statistical power of results on some anamnestic and biological criteria. In particular, it was not possible better to characterize fractures in terms of age, type and mechanism, nor to use objective assessment tools for diagnosis. However, this did not prevent from finding significant correlates of BMD decrease and fractures in this study that are similar to those found in previous studies. The choice of such a specific population, in the context of a single-centre study, limits the generalizability of results to the ensemble of patients with AN. Moreover, extremely low weights could have increased the risk of soft tissue-related precision errors in BMD calculation,[36] and the dearth of studies on DXA in extremely severe AN prevents any estimation of the magnitude of these imprecisions.

By contrast, focusing on this special, strongly homogeneous subgroup of patients provided an original analysis of the relationship between bone and extremely severe AN.