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


Descriptive Results

Table 1 describes the characteristics of the population included in the study. In total, fifty-one per cent of patients had a lumbar spine Z-score ≤-2.0 SD and 38% had a femoral neck Z-score ≤-2.0 SD. Lumbar spine Z-score was significantly lower than femoral neck Z-score (t(95)=0.24; P<.01) (Table 1).

Factors Associated With Low BMD

Table 2 illustrates correlations between Z-scores and clinical-biological variables. Some factors showed a significant correlation with lumbar spine Z-score, such as age at onset (P<.05). Others, such as age, had a correlation with femoral neck Z-score (P<.05). BMI and duration of amenorrhea were associated with both femoral neck Z-score and lumbar spine Z-score (P<.01).

In the first linear regression model, age at onset, restricting AN and FFM emerged as significant predictors of lumbar spine Z-score, explaining up to 64% of the variance in the studied population (Table 3).

The second linear regression model retained age, duration of illness and FFM as independent predictors of Z-scores at femoral neck. This model explained 63% of variance in the studied population (Table 4).

Antidepressants and PPI had no significant impact on BMD, as well as combined oestrogen/progesterone contraceptive pills (P>.05).

Factors Associated With Fractures

Fragility fractures were reported by 10 patients (10.3%) of whom 7 reported a vertebral topography. Femoral neck Z-scores were significantly lower than patients without fractures, but no significant difference in lumbar spine Z-score was found between the two subgroups (P>.05). Patients with fractures were older, had longer duration of illness, lower sodium plasma concentrations and higher alkaline phosphatase concentrations compared with those without fractures (P<.05), while no difference in BMI was found between the two groups (P>.05). Comparisons between patients with vs without fractures are reported in Table 5.

The logistic regression model encompassing fracture risk as the dependent variable showed an ability correctly to classify 44.4% of patients with fractures and 98.6% of patients without fractures. Hyponatremia and Z-score at femoral neck were the most significant predictors of fractures in our population (Table 6).