Calcified Tissue International

, Volume 83, Issue 5, pp 315–323

Factors Influencing Changes in Bone Mineral Density in Patients with Anorexia Nervosa-Related Osteoporosis: The Effect of Hormone Replacement Therapy

Authors

    • Department of RheumatologyCHRU Lille, Hôpital Roger Salengro
  • Jean Vignau
    • Department of AddictologyCHRU Lille, Hôpital La Charité
  • Francis Collier
    • Department of GynecologyCHRU Lille, Hôpital Jeanne de Flandre
  • Bernard Cortet
    • Department of RheumatologyCHRU Lille, Hôpital Roger Salengro
Article

DOI: 10.1007/s00223-008-9173-y

Cite this article as:
Legroux-Gerot, I., Vignau, J., Collier, F. et al. Calcif Tissue Int (2008) 83: 315. doi:10.1007/s00223-008-9173-y

Abstract

The purpose of this longitudinal study was to evaluate factors affecting changes in bone mineral density (BMD) in patients with anorexia nervosa (AN) and osteoporosis and, more particularly, to assess the benefits of hormone replacement therapy (HRT) on BMD in these patients. Our study involved 45 AN patients, 12 of whom had been treated by HRT for 2 years following a diagnosis of osteoporosis by densitometry (WHO criteria). Patients’ mean age was 25.3 ± 6.7 years. Mean duration of illness was 5.7 ± 5.3 years. Serum calcium and phosphate were measured at baseline, as were bone remodeling markers. Osteodensitometry by dual-energy X-ray absorptiometry was performed at inclusion and after 2 years. After 2 years, no significant differences were observed between spine, femoral neck, and total hip BMDs either in the HRT group (P = 0.3, P = 0.59, P = 0.58) or in the nontreatment group (P = 0.17, P = 0.68, P = 0.98). Moreover, there were no significant differences between the two groups when changes in spine, femoral neck, and total hip BMDs at 2 years were compared (P = 0.72, P = 0.95, P = 0.58). In both groups, change in weight at 1 year correlated with change in spine BMD at 2 years (r = 0.35, P = 0.04) and change in total-hip BMD at 2 years (r = 0.35, P = 0.04) but not with change in femoral neck BMD at 2 years. Patients with a body mass index (BMI) ≥ 17 kg/m2 at 2 years showed a significant increase in total-hip BMD when compared with patients with a BMI < 17 kg/m2 (+4.4% ± 6.7 vs. −0.5% ± 6.01, P = 0.03). No significant differences were observed for spine and femoral neck BMD. In patients who had recovered their menstrual cycle, significant increases were observed in spine BMD (+4% ± 6.3 vs. −1.9% ± 5.6, P = 0.008), femoral neck BMD (+3% ± 6.2 vs. −2.4% ± 8, P = 0.05), and total-hip BMD (+3% ± 7.1 vs. −3.7% ± 10, P = 0.04). Prevention of bone loss at 2 years in AN patients treated by HRT was not confirmed in this study. We did confirm that increase in weight at 1 year was the most predictive factor for the improvement of spine and hip BMD at 2 years.

Keywords

OsteoporosisAnorexia nervosaHormone replacement therapy

Over the last few years anorexia nervosa (AN) has become a serious public health issue in industrialized countries. AN occurs with a very high frequency among young women (0.5% prevalence as opposed to 2% for bulimia) and gives rise to serious metabolic complications resulting from malnutrition [1, 2]. One of the complications of the disorder is bone loss. Indeed, osteoporosis may occur in 20–50% of cases depending on the groups observed, with bone fracturing occurring in 44% of patients [17]. In patients who have had the disorder for an average of 5.8 years, the risk of fractures occurring is seven times higher than in healthy women of the same age [8]. Fractures tend to occur at the most common osteoporotic fracture sites, i.e., most frequently at vertebrae and at the radius and the upper extremity of the femur [3, 911].

The mechanisms underlying bone loss in AN are still unclear. Hypoestrogenia plays a role, according to the definition of AN, but it cannot alone account for the entire extent of loss in bone mineral density (BMD) observed in AN patients [5, 7, 12]. Other factors are involved, notably nutritional [1, 7, 1214]. Weight gain generally has a favorable impact on BMD, but it is still debatable whether it is sufficient to reverse the process of osteoporosis. Moreover, as estrogen deficiency is a potential causative factor for low BMD [15], there is a rationale for using hormone replacement therapy (HRT) to prevent bone loss in AN patients. In a previous cross-sectional study, we evaluated the respective roles of estrogen deficiency and nutritional factors in the process of bone loss in AN patients [15]. In that study, we demonstrated that estrogen levels were a factor which could account for the variance in BMD at the lumbar spine. A few studies have been conducted to determine the benefits of HRT in these nonmenopausal patients but with conflicting results. Previous studies in which the effect of HRT on BMD was assessed involved AN patients without low BMD. Now the main interest of HRT for preventing bone loss is for women with low BMD who are at high risk of fracture. Further, the estrogens used in previous studies were conjugated equine estrogens or ethinyl estradiol but not 17β-estradiol, which is currently used in Europe for preventing bone loss. Finally, some of the previous studies on this issue raise methodological concerns because there were not longitudinal.

Thus, the purpose of the present study was to (1) evaluate changes in BMD at 2 years in AN patients, (2) determine whether or not HRT (17β-estradiol) could slow down bone loss in AN patients with osteoporosis, and (3) identify factors influencing changes in BMD.

Patients and Methods

Our study involved 45 patients with AN as defined by DSM-IV criteria [16]. The patients were initially evaluated between December 2001 and November 2004 at the Rheumatology Department in Lille and were followed up for 2 years. Follow-up continued until January 2007. Patients’ mean age was 25.3 ± 6.7 years (range 15–41). Mean duration of illness was 5.7 ± 5.3 years. Mean age at the onset of illness was 19.6 ± 5.5 years. Mean duration of amenorrhea was 3.5 ± 5.1 years, with a mean age at onset of 20.1 ± 5.2 years. None of the patients had primary amenorrhea. Of the 45 patients, 29 were found to have pure restrictive anorexia and 16 exhibited a mixed form of the disorder (bulimia and anorexia). None of the women included in the present study had previously received drugs or medication that could affect bone metabolism.

We recorded patients’ weight and height and determined, by interview, their risk factors for osteoporosis as well as the type of treatment they had been receiving.

Patients with osteoporosis defined by bone densitometric assessment (as defined by the WHO criteria) at inclusion (i.e., T score <2.5 SD) underwent HRT consisting of one to two percutaneous doses of 0.5 mg 17β-estradiol (Estreva®) from day 1 to day 21 (according to the safety profile, mean 0.7 ± 0.1 mg/day), and 10 mg of dydrogesterone 1 cp from day 11 to day 21. These patients were seen in consultation (outpatients) twice a year alternately by both the gynecologist and the rheumatologist participating in the present study (I. L.-G. and F. C.). Thus, each patient was examined by a doctor (rheumatologist or gynecologist) four times a year. At each visit, the patients were thoroughly examined and their compliance was sought through a series of questions. The other patients were also followed up but without treatment. The cumulative dose of estrogens was calculated as mean dose × number day of treatment: 504 ± 72 mg. When vitamin D levels were below the threshold used at the time of the study (i.e., <20 ng/mL), vitamin D supplementation was initiated. The dose varied according to baseline vitamin D level. Similarly, when calcium intake was low, calcium supplementation was provided.

Changes in BMD were compared between patients who had received HRT and those who had not.

The present study was a prospective, longitudinal, controlled (but not randomized) study.

Biological Assessment

Blood and urinary calcium and phosphate levels were determined (calcemia, phosphatemia, alkaline phosphatases, 24-hour calciuria), as were 25-hydroxyvitamin D (25[OH]D; indeed, the 25[OH]D assay does not distinguish D3 from D2) and parathyroid hormone (PTH) levels. Bone remodeling markers (osteocalcin and bone alkaline phosphatases (BSAPs) for bone formation, serum crosslaps or C-terminal telopeptide crosslinks [CTX] and type I collagen carboxy-terminal telopeptide [ICTP] for bone resorption) were assessed. Sera were stored at −80°C until use. Radioimmunological assays were used to measure osteocalcin (Cis-Bio International, Gif-sur-Yvette, France; normal values 5.2–34.5 ng/mL) and ICTP (Orion Diagnostica, Espoo, Finland; normal values 1.8–5.2 ng/mL) levels. BSAP was determined using a human-specific immunoradiometric method (Hybritech, San Diego, CA; normal values <20.5 ng/mL). Serum CTX was measured using an immunoenzymological method (ELISA; serum CrossLaps One Step; Osteometer Biotech, Herlev, Denmark; normal values 270–3,270 pmol/L). The within-run and run-to-run coefficients of variation were, respectively, ≤5.2% and 2.8% for osteocalcin, ≤9.4% and 6.4% for ICTP, ≤8.1% and 6.7% for BSAP, and ≤8.1% and 5.4% for CTX. We also assessed follicle-stimulating hormone (FSH), luteinizing hormone (LH), estradiol, 8:00 a.m. cortisol, free urinary cortisol, thyroid-stimulating hormone (TSH), triiodothyronine (T3), thyroxine (T4), and prolactin levels, as well as nutritional factors such as somatomedin (insulin-like growth factor I [IGF-I]) and IGF-binding protein 3 (IGF-BP3).

All samples were collected from fasting patients between 8:00 a.m. and 10:00 a.m. and only at inclusion after a fast overnight.

Measurement of BMD

Lumbar spine and hip BMDs were determined by dual-energy X-ray absorptiometry (DXA) using a Hologic (Waltham, MA) QDR 2000 densitometer. Values were compared to those of the control group. Measurements were taken at inclusion and at 2 years. The coefficient of variation was 1% for lumbar spine measurements and 2% for hip measurements. The least significant change with the Hologic device was 34 mg/cm2 at the lumbar spine and 27 mg/cm2 at the hip [17].

Statistical Analysis

Results for both groups were expressed as mean ± standard deviation. Changes in BMD were evaluated using Student’s t-test for paired samples or Wilcoxon’s t-test according to the distribution of BMD at baseline in each group. Changes in BMD between the treatment group (HRT) and the nontreatment group were compared using Student’s t-test for unpaired samples or the Mann–Whitney U-test. When BMD changes were normally distributed, multiple adjustments for confounding factors were made (covariance analysis).

Stepwise multiple regression analysis was used to evaluate the factors affecting the change in BMD.

Results

Characteristics of Patients as a Whole and in Each of the Two Groups

Forty-five patients were included in the study, 12 of whom were treated by HRT on the basis of BMD results measured by DXA (i.e., T score < −2.5) at the lumbar spine and/or hip (Table 1). The nontreatment group is hereafter referred to as the “control” group.
Table 1

Comparison between patients undergoing HRT at inclusion and untreated patients: clinical data

 

Patients without HRT (n = 33)

Patients undergoing HRT (n = 12)

P

Age (years)

25 ± 7

26.2 ± 6

0.6

Age at onset of AN (years)

20.2 ± 6

17.7 ± 7.5

0.18

Duration of AN (years)

4.7 ± 4.3

8.4 ± 6.8

0.03

Age at onset of amenorrhea (years)

20.7 ± 6

18.8 ± 3.1

0.34

Duration of amenorrhea (years)

2 ± 3.3

7.2 ± 7

0.0019

Weight (kg)

43.6 ± 7

39 ± 4.7

0.041

Height (cm)

166.2 ± 5

163.3 ± 6.3

0.12

BMI (kg/m2)

15.7 ± 2.2

14.7 ± 1.7

0.16

At inclusion, 39 patients exhibited amenorrhea and six had irregular cycles. At the 2-year follow-up, 17 patients (37.8%) were still amenorrheic, including five (41.6%) from the treatment group and 12 (36.3%) from the control group. Twenty-eight patients (62.2%) recovered their cycles.

At inclusion, all of the patients had a body mass index (BMI) < 19 kg/m2. At 2 years, nine patients (20%) had a BMI > 19 kg/m2, while in 36 patients (80%) BMI remained unchanged.

At inclusion, there was no significant difference between the two groups in terms of age (control group 25 ± 7 years, treatment group 26.2 ± 6 years; P = 0.6). On the other hand, anorexia and amenorrhea durations were significantly higher in the treatment group than in the control group (anorexia 8.4 ± 6.8 and 4.7 ± 4.3 years, respectively; P = 0.03; amenorrhea 7.2 ± 7 and 2 ± 3.3 years, respectively; P = 0.0019). Weight was significantly lower in the treatment group (39 ± 4.7 vs. 43.6 ± 7 kg, P = 0.041). No significant difference was observed for BMI.

Biological Assessment

Mean values for serum and urinary calcium and phosphate were normal in both groups, and no significant differences were observed between the groups (Table 2). Mean vitamin D was 28.8 ± 19.5 ng/mL in the control group and 22.5 ± 11.8 ng/mL in the treatment group (P = 0.31). A large proportion (41%) of our population had low vitamin D levels at baseline (i.e., before vitamin D supplementation) according to the criteria used at the time of study (vitamin D level <20 ng/mL). When osteocalcin levels were compared between the two groups, no significant difference was observed (control 16.3 ± 6.7 ng/mL, treatment 13.7 ± 9.2 ng/mL; P = 0.34). Similarly, no significant differences were found for BSAP (P = 0.46), the mean values of which were 10.7 ± 4.6 and 12.3 ± 6.8 μg/L in the control and treatment groups, respectively. When crosslaps values were compared (control 6,234.9 ± 3,251.3 pmol/L, treatment 4,008 ± 2,520.3 pmol/L), the difference was not significant (P = 0.07). The mean ICTP values for the control group (6.6 ± 3.1 μg/L) and the treatment group (5.4 ± 1.8 μg/L) were statistically the same (P = 0.31). Likewise, LH, estradiol, TSH, prolactin, free urinary cortisol, IGF-I, and IGF-BP3 levels remained statistically unchanged across the groups. However, T3, T4, and FSH levels were statistically lower (P = 0.005, P = 0.04, and P = 0.03, respectively) in the treatment group at inclusion.
Table 2

Comparison between patients undergoing HRT at inclusion and untreated patients: biological data

 

Patients without HRT (n = 33)

Patients undergoing HRT (n = 12)

P

Calcemia (mg/L)

93.15 ± 28.2

92.2 ± 20.8

0.62

Phosphatemia (mg/L)

36.9 ± 4.6

38.4 ± 4.9

0.39

24-hour calciuria (mg/24 hours)

105.8 ± 61.6

122.7 ± 87.2

0.57

25(OH)D3 (ng/mL)

28.8 ± 19.5

22.5 ± 11.8

0.31

PTH (pg/mL)

37 ± 20.2

34.1 ± 17.1

0.69

Osteocalcin (ng/mL)

16.3 ± 6.7

13.7 ± 9.2

0.34

Bone alkaline phosphatase (μg/L)

10.7 ± 4.6

12.3 ± 6.8

0.46

ICTP (μg/L)

6.6 ± 3.1

5.4 ± 1.8

0.31

Serum CTX (pmole/L)

6,234.9 ± 3,251.3

4,008 ± 2,520.3

0.07

TSH (μUI/mL)

1.4 ± 0.7

1.3 ± 0.6

0.94

T3 (pmole/L)

3.7 ± 1

2.7 ± 0.4

0.005

T4 (pmole/L)

15.8 ± 3.8

13.1 ± 1.4

0.04

Urinary free cortisol (μg/diuresis)

51 ± 34.4

68.7 ± 35.3

0.24

FSH (UI/L)

4.4 ± 3.9

1.7 ± 2.1

0.03

LH (UI/L)

3.3 ± 5.4

0.7 ± 0.9

0.12

Estradiol (pg/mL)

21 ± 23.6

9.8 ± 10.9

0.14

Prolactin (ng/mL)

13.4 ± 14.5

12.9 ± 9.2

0.9

IGF-I (UI/mL)

1 ± 0.5

0.7 ± 0.3

0.12

IGF-BP3 (mg/L)

3.7 ± 4.3

1.7 ± 0.2

0.37

BMD Measurements

Osteoporosis was observed in at least one of the three sites (lumbar spine, total hip, and femoral neck) in 12 patients (Table 3). Osteopenia was observed in 25 patients. BMD measurements were normal in eight patients.
Table 3

Comparison between patients undergoing HRT at inclusion and untreated patients: densitometric data

 

Patients without HRT (n = 33)

Patients undergoing HRT (n = 12)

P

Spine BMD

0.96 ± 0.1

0.77 ± 0.07

<0.0001

Z score spine

−0.4 ± 1

−2.3 ± 0.8

<0.0001

T score spine

−0.3 ± 1.4

−2.6 ± 0.7

<0.0001

Neck BMD

0.72 ± 0.1

0.64 ± 0.1

0.0081

Hip BMD

0.8 ± 0.1

0.68 ± 0.12

0.0005

Z score neck

−1.3 ± 1.2

−3.4 ± 0.4

0.04

Z score hip

−1.3 ± 1.2

−3.4 ± 0.4

0.04

T score neck

−1.2 ± 0.9

−2 ± 1.1

0.01

T score hip

−1.1 ± 0.9

−2.2 ± 1

0.018

According to the design of the study, patients who had received HRT had significantly lower BMD measurements at all three sites and at the lumbar spine in particular (control 0.96 ± 0.1 g/cm2, treatment 0.77 ± 0.07 g/cm2; < 0.0001). Mean Z scores in the treatment group were also significantly lower (control −0.4 ± 1, treatment −2.3 ± 0.8; < 0.0001), as were mean T scores (control −0.3 ± 1.4, treatment −2.6 ± 0.7; P < 0.0001). Similar findings were observed for femoral neck BMD measurements (0.72 ± 0.1 vs. 0.64 ± 0.1, P = 0.0081), femoral neck Z scores (−1.3 ± 1.2 vs. −3.4 ± 0.4, P = 0.04), femoral neck T scores (−1.2 ± 0.9 vs. −2 ± 1.1, P = 0.01), total-hip BMD measurements (0.8 ± 0.09 vs. 0.68 ± 0.12, P = 0.0005), hip Z scores (−1.3 ± 1.2 vs. −3.4 ± 0.4, P = 0.04), and total-hip T scores (−1.1 ± 0.9 vs. −2.2 ± 1, P = 0.018).

At the 2-year follow-up, there was no significant difference in change in BMD for lumbar spine, femoral neck, and total hip in the HRT group (P = 0.3, P = 0.59, and P = 0.58, respectively). Similarly, BMD measurements remained statistically unchanged in the nontreatment group (P = 0.17, P = 0.68, and P = 0.98, respectively) (Table 4).
Table 4

Change in BMD at 2 years in the untreated and HRT groups

Variation BMD at 2 years (%)

Patients without HRT (n = 33)

Patients undergoing HRT (n = 12)

P

Δ Spine BMD

0.92 ± 5.2

1.7 ± 6.7

0.72

Δ Femoral neck BMD

0.78 ± 4.7

0.63 ± 7.6

0.95

Δ Total hip BMD

1.7 ± 4.3

0.21 ± 9.3

0.58

Moreover, there were no significant differences between the two groups when changes in spine, femoral neck, and total-hip BMDs at 2 years were compared (P = 0.72, P = 0.95, and P = 0.58, respectively).

After adjustment for parameters for which significant differences were observed between the two groups at inclusion, no significant variation was observed in densitometric measurements for lumbar spine, total hip, and femoral neck. In particular, when data were adjusted for bone surface area, the results remained unchanged.

Relationships Between Change in BMD and Clinicobiological Factors

We tried to determine whether or not correlations existed between change in BMD at 2 years and the various clinical and biological factors measured at baseline. In the entire group of 45 patients, changes in spine BMD and osteocalcin levels were inversely correlated: The lower the osteocalcin levels, the greater the extent of bone loss (r = −0.34, P = 0.04). No correlations were found in the nontreatment group.

No correlation was observed between changes in femoral neck BMD and any of the clinicobiological factors in the whole group. In the nontreatment group, changes in femoral neck BMD and IGF-I levels were inversely correlated: The lower the latter, the greater the extent of bone loss (r = −0.44, P = 0.03).

Changes in hip BMD were found to correlate with both PTH levels (r = −0.32, P = 0.05) and TSH levels (r = −0.34, P = 0.05): The higher the levels of these two hormones, the greater the extent of bone loss. The correlation between hip BMD and TSH was also observed in the nontreatment group (r = −0.38, P = 0.03).

We also sought to determine whether the change in BMD at 2 years correlated with the change in weight at 2 years, both in the total group and in the control group. In the total group, changes in spine and total-hip BMD correlated with the change in weight at 2 years (spine r = 0.38, P = 0.01; total hip r = 0.39, P = 0.0091). Similar results were found in the control group (spine r = 0.45, P = 0.01; total hip r = 0.46, P = 0.009).

We then examined the relationship between change in weight at 1 year and change in BMD at 2 years in both groups. In the total group, correlations were found for spine BMD (r = 0.35, P = 0.04) and total-hip BMD (r = 0.35, P = 0.04) but not for femoral neck BMD. Similar results were observed in the control group, with change in weight at 1 year correlating with change in spine BMD at 2 years (r = 0.39, P = 0.04) and total-hip BMD at 2 years (r = 0.38, P = 0.05).

We were unable to find a correlation between initial BMI and change in BMD at any of the three sites, but this could be accounted for in part by the duration of the course of AN, which was highly variable in our population.

We then looked at mean and median BMI values, which were close to 17 kg/m2. Next, we divided our population into two groups: patients with BMI ≥ 17 kg/m2 at 2 years and those with BMI < 17 kg/m2. We examined changes in BMD at all three target sites in both of these groups (Table 5). Patients with BMI ≥ 17 kg/m2 at 2 years showed a significant increase in total-hip BMD compared to patients with BMI < 17 kg/m2 (P = 0.03). For spine and femoral neck BMDs, there was no significant difference between the groups but a positive trend, with an increase in BMD, was observed. If HRT patients were excluded from the analysis, similar results were observed: no significant difference in spine BMD, a significant increase in total-hip BMD (P = 0.04), and a borderline significant increase in femoral neck BMD (P = 0.06). Moreover, the results remained unchanged even after adjusting changes in BMD to take account of estrogen treatment.
Table 5

Change in BMD at 2 years in patients with BMI ≥7 and <17 kg/m2 at 2 years

%

BMI ≥ 17 kg/m2

BMI < 17 kg/m2

P

Δ Spine BMD

+3.3 ± 5.6

+0.2 ± 6.7

0.15

Δ Femoral neck BMD

+3.9 ± 6.3

−0.02 ± 6.3

0.07

Δ Total hip BMD

+4.4 ± 6.7

−0.5 ± 6.01

0.03

We proceeded in the same way with weight measurements, establishing different weight recovery thresholds (7–9%). No significant difference in spine, total-hip, or femoral neck BMD was observed between patients whose weight recovery was greater or less than the thresholds we established.

We compared changes in BMD in patients who were still amenorrheic at 2 years with those who had recovered their menstrual cycles, excluding patients receiving HRT. In patients who had recovered their menstrual cycle, significant increases were observed in spine BMD (+4% ± 6.3 vs. −1.9% ± 5.6, P = 0.008), femoral neck BMD (+3% ± 6.2 vs. −2.4% ± 8, P = 0.05), and total-hip BMD (+3% ± 7.1 vs. −3.7% ± 10, P = 0.04).

We compared changes in BMD at all three target sites in the following four groups: BMI > 17 kg/m2 + menstrual cycle recovery, BMI > 17 kg/m2 + amenorrhea, BMI < 17 kg/m2 + menstrual cycle recovery, BMI < 17 kg/m2 + amenorrhea. Changes in spine BMD were significantly different on the whole (P = 0.02), but a two-by-two comparison could not be performed given the small size of the populations. Changes in femoral neck and total-hip BMD were not significantly different (P = 0.07 and P = 0.06, respectively) (Fig. 1).
https://static-content.springer.com/image/art%3A10.1007%2Fs00223-008-9173-y/MediaObjects/223_2008_9173_Fig1_HTML.gif
Fig. 1

(a) Changes in BMD (expressed as %) at the lumbar spine according to BMI (>17 vs. <17 kg/m2) and menstrual cycle recovery (yes vs. no). (b) Changes in BMD (expressed as %) at the femoral neck according to BMI (>17 vs. <17 kg/m2) and menstrual cycle recovery (yes vs. no). (c) Changes in BMD (expressed as %) at the total hip according to BMI (>17 vs. <17 kg/m2) and menstrual cycle recovery (yes vs. no)

Finally, we were unable to find any correlation between cumulative estrogen exposure and BMD changes at the lumbar spine, total hip, or femoral neck.

Discussion

Very few studies have focused on the prevention of bone loss in AN patients. Here, we report on a longitudinal, controlled study involving 45 AN patients, 12 (26.7%) of whom had exhibited signs of osteoporosis and had undergone HRT. To our knowledge, no previous study has sought to assess the effectiveness of 17β-estradiol treatment in women with AN and osteoporosis. In the nontreatment group, 25 patients (55%) had osteopenia, which is consistent with the data in the literature. After 2 years, 20% of the patients exhibited BMI > 19 kg/m2. We found no significant difference between the treatment group and the nontreatment group when changes in BMD were compared. In the group of patients with osteoporosis, BMDs remained low, despite HRT.

Despite the association between AN and an estrogen deficit and despite the strong correlation between the extent of bone loss and the duration of amenorrhea, previous studies report conflicting findings regarding the effectiveness of estroprogestative (EP) treatment against bone loss [1822] (Table 6). It should be noted that, in most of the studies mentioned, estrogens were administered within the framework of EP contraceptive treatment. Seeman et al. [18] investigated the change in BMD in 16 patients receiving EP treatment in pill form vs. 49 untreated patients. In that retrospective study, an increase in spine BMD was observed in patients in the treatment group but no significant difference was observed between the groups when femoral neck BMDs were compared.
Table 6

Effectiveness of HRT in AN patients as reported in the literature

Reference

Patients

Treatment

Duration of follow-up

Change in BMD

Seeman et al. [18]

16 treated/49

EP pills

31.8 months

↑ Spine BMD

(1.4 ± 0.005 vs. 1.02 ± 0.002; < 0.02), no effect at femoral neck

Klibanski et al. [19]

22 treated/26

16 with CEE, 6 with EP pills

1.5 years

No difference under treatment

Golden et al. [20]

22 treated/28

EP pills

23 months

No difference under treatment

Patel [21]

1 case

Estradiol 2 mg

1 year

↑ Spine BMD 38%, no effect at femoral neck

CEE, conjugated equine estrogen

In a 1995 study involving 48 AN patients with amenorrhea, Klibanski et al. [19] reported on 22 patients who had been treated with HRT (Premarin-Provera®, n = 16) or had undergone EP treatment (35 μg of ethinyl estradiol, n = 6). Mean follow-up was 18 months. Follow-up results failed to show any significant difference in BMD change between the placebo and the estrogen treatment groups, and the gain in BMD was not significant. However, the effectiveness of the treatment depended on the initial weight. When patients whose weight was <70% of their ideal weight were investigated separately, those in the nontreatment group (n = 10) exhibited a significant loss of 20.1 ± 16.2% in spine BMD (< 0.004), as opposed to a 4 ± 8.8% gain in the treatment group (n = 6).

In a study involving 50 female adolescents with a mean age of 16.8 years, Golden et al. [20] investigated changes in BMD in 22 patients after 23 months of EP treatment (20–35 mcg ethinyl estradiol in pill form). At 1 year, the authors failed to find a significant increase in spine or femoral neck BMD, despite a significant gain in weight in both groups.

Therefore, HRT does not seem to attenuate bone loss in AN. This finding has already been mentioned by several authors and is confirmed in our study. One possible explanation for these results could be that the doses used were too low to have a favorable impact on BMD. Indeed, HRT is known for its poor adherence. One worry was that AN women are concerned about weight gain. The second concern was related to the occurrence of mastalgia, which was also one factor that limited the use of an optimal dose. It is certain that these two findings explained why the mean dose of estradiol used was low (0.7 ± 0.1 mg/day). Moreover, menstrual cycle recovery while undergoing treatment may also contribute to poor judgment on the part of adolescents since it could create a false sense of security even though their weight is too low and delay recovery of the normal menstrual cycle.

Several authors have investigated change in BMD in this category of patient. Their studies often involved few patients [6, 19, 2226]; and while some authors have reported a recovery in BMD in line with the recovery in weight, others have mentioned a persistently low bone mass even after the patient is cured (recovery of weight and menstrual cycles). In our study, patients with BMI ≥ 17 kg/m2 at 2 years exhibited a significant increase in total-hip BMD compared to patients with BMI < 17 kg/m2 (P = 0.03). Similar findings have also been reported by Hotta et al. [22]. In investigating the change in BMD in 29 AN patients over a period of 11–46 months, the authors found an increase in hip BMD in patients who had BMI ≥ 16.4 ± 0.3 kg/m2, even when they were amenorrheic. As was the case in our study, this result was not found in the case of spine BMDs. The authors also reported the existence of a correlation between bone loss and duration of anorexia and amenorrhea. Also, in that study, no correlation was found between BMD and biological markers.

Various predictive factors of bone loss have already been identified. In a previous study involving 113 AN patients [15], duration of amenorrhea seemed to play a key role in accounting for the decrease in BMD at both the spine and hip; but IGF-I also seemed to contribute to bone loss at the hip, while estradiol seemed to have a greater impact on bone loss at the spine. Other authors have also observed that lean body mass was a predictor of change in BMD [27]. Our findings suggest that menstrual cycle recovery and weight recovery are important predictors of gain in bone mass at the spine and at the hip, respectively. These results have also been reported in a recent study. Miller et al. [27] investigated change in BMD in 75 AN patients, followed up for 13.5 ± 1 months. Weight recovery was accompanied by a significant increase in hip BMD (P = 0.05) only in patients who had not been receiving any form of oral contraception. Menstrual cycle recovery was found to be a relevant predictor of an increase in spine BMD (P = 0.02). It is interesting to note that the effect persisted despite the adjustment for weight, which demonstrates a clear hormonal effect on the recovery of bone mass, independent of nutritional factors. The authors also observed a significant increase in spine BMD (+3.6%) and hip BMD (+2.1%) in patients who had recovered their weight and who were no longer amenorrheic, as opposed to those whose weight remained low and who remained amenorrheic (−2.7% and −2.6%, P = 0.02 for the two sites). However, the P level was no longer significant (P = 0.064) when the results were annualized (in the study, the duration between visits was variable).

There are obvious limitations to the present study: The patients were not randomized, and those who had undergone HRT were somewhat older and had longer durations of disease and amenorrhea. However, the results were the same after adjustments were made for these confounding factors. Further, the size of the study population was small, and a longer follow-up would be necessary to better detect changes in BMD in these patients. All of the patients did not receive the same dose of the HRT, and one can also question the observance of treatment in this population, which is certainly less strict than among menopausal women.

In conclusion, we were unable to demonstrate prevention of bone loss at 2 years in a small cohort of AN patients treated by HRT. We did confirm that increase in weight at 1 year was the most predictive factor for the recovery of spine and hip BMD at 2 years. Weight recovery was found to be a predictor of an increase in bone mass at the hip, and patients who had BMI ≥ 17 kg/m2 at 2 years also exhibited a significant increase in total-hip BMD when compared to patients who had BMI < 17 kg/m2. Also, at the spine, menstrual cycle recovery seemed to be an important factor in the recovery of bone mass.

Copyright information

© Springer Science+Business Media, LLC 2008