| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
Neuroendocrine Unit (M.M., R.P., K.K.M., J.C., A.K.), Massachusetts General Hospital and Harvard Medical School, Pediatric Endocrine Unit (M.M., R.P.) and Adolescent Medicine Unit (M.A.G.), Massachusetts General Hospital for Children and Harvard Medical School, and Harris Center (D.B.H.), Massachusetts General Hospital, Boston, Massachusetts 02114; Wilkins Center for Eating Disorders (D.M.), Greenwich, Connecticut 06831; Bedford Center for Eating Disorders (L.C.), Bedford, New Hampshire 03110; Eating Disorders Center (P.L.), Mercy Hospital, Portland, Maine 04101; and Division of Adolescent Medicine (D.K.K.), Department of Pediatrics, Hospital for Sick Children, Toronto, Canada M5G 1X8
Address all correspondence and requests for reprints to: Madhusmita Misra, M.D., M.P.H., BUL 457, Neuroendocrine Unit, Massachusetts General Hospital, 55 Fruit Street, Boston, Massachusetts 02114. E-mail: mmisra{at}partners.org.
 |
Abstract |
|---|
 Top Abstract IntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
10% increase in body mass index) would predict an increase in these measures of bone density. Methods: In a prospective observational study, lumbar and whole-body (WB) bone density was measured at 0, 6, and 12 months in 34 AN girls aged 12–18 yr and 33 controls. Using Ward’s modification of the Molgaard approach, we determined measures of BMC for BA and BA for height at the lumbar spine and WB and also determined spine bone mineral apparent density and WB BMC adjusted for height.
Results: Girls with AN had lower spine BMC for BA Z-scores (P = 0.0009), and lower WB BA for height Z (P < 0.0001), compared with controls. Menstrual recovery and weight gain in AN (AN-recovered) (median 9 months) resulted in a stabilization of BMD measures, whereas BMD continued to decrease in AN who did not gain weight and recover menses (AN-not recovered). AN-recovered also predicted greater increases in spine BMC for BA and WB BA for height, compared with AN-not recovered (P < 0.05).
Conclusions: Even short-term weight gain with menstrual recovery is associated with a stabilization of BMD measures.
|
Introduction |
|---|
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
We examined changes in spine and WB BMC and bone density measures by DXA after weight gain and menstrual recovery in AN girls and a cohort of normal adolescents over a 1-yr period. We also examined specific changes in bone density measures with menses recovery and weight gain by assessing changes in BMC for BA and BA for height based on the approach by Molgaard et al. (8), modified by Ward et al. (12) for Hologic densitometers. In addition, we examined changes in spine BMAD and WB BMC adjusted for height. Although the use of WB BMC adjusted for lean body mass has been advocated by some (13), recent studies indicate that this is not a good predictor of fracture risk (11) or bone strength as assessed by peripheral QCT (10). Therefore, these data are not reported.
|
Subjects and Methods |
|---|
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
Data from subjects reported in previously published studies by Soyka et al. (3) and Misra et al. (14, 15, 16, 17) were analyzed to determine effects of weight gain with menstrual recovery on changes in bone density measures over a 1-yr period in adolescent girls with AN and healthy adolescent girls 12–18 yr old. Neither study had previously examined changes in specific bone density measures based on recovery status. Follow-up data for bone density were available in 38 of 42 girls with AN and 35 of 40 controls 12–18 yr old. Ten girls with AN and two controls were premenarchal. Among the premenarchal group, only girls with delayed menarche [age > 15.3 yr (mean age at menarche +2 SD for American girls (18)] were included because delayed menarche was likely subsequent to low weight, and onset of menses expected after weight gain. Girls younger than 15.3 yr old were not included because they were still in the normal age range for attaining menarche, and lack of menses could not be attributed to low weight alone. Thus, four girls with AN and two controls were excluded from final analysis, and 34 girls with AN and 33 controls were evaluable for effects of menstrual recovery and weight gain on bone density measures. Characteristics of girls with AN did not differ between the two studies, and neither did those of controls. Mean duration since diagnosis was 11.2 ± 12.4 months.
Girls with AN were recruited through referrals from eating disorder providers and treatment centers in the New England area, whereas healthy adolescents were recruited through mass mailings to local pediatricians and adolescent medicine physicians as well as advertisements in area newspapers. None of the controls had a past or present history of an eating disorder. Body mass index (BMI) SD scores for healthy adolescents were between –0.96 and 1.86, and these subjects had not crossed percentile lines for weight in the year preceding the study per report. The Institutional Review Board of Partners HealthCare approved the study, and informed assent and consent was obtained from all.
Experimental protocol
After a screening visit to determine eligibility, subjects were admitted to the General Clinical Research Center of Massachusetts General Hospital for the baseline visit. Conditions other than AN and medications that may affect bone metabolism excluded subjects from study participation. At the baseline visit, height was measured in triplicate on a single stadiometer at the General Clinical Research Center and averaged. Weight was measured on a single electronic scale in the fasting state. BMI was calculated as the ratio of weight (in kilograms) to height (in square meters). Bone density measures were measured at baseline and repeated at 6 and 12 months. A bone age was performed at baseline and read using methods of Greulich and Pyle (19). Weight gain was defined as a 10% increase in BMI, and menstrual recovery as three menses or more in the previous 6 months. Subjects were characterized as AN recovered (AN-recovered) if they had both weight gain and menstrual recovery (n = 14) and as AN not recovered (AN-not recovered) if they did not have both weight gain and menstrual recovery (n = 20). Median duration of recovery was 9 months.
Bone density and body composition measurement
DXA (4500A fan beam densitometer, software version 11.2; Hologic Waltham, MA) was used to determine BMC and areal BMD at the spine and for the whole body and fat and lean mass. Z-scores for the spine (L1-L4) were generated using the reference database available to Hologic (20). To correct for body size, BMAD for was calculated from lumbar BMC and BA (12). We used methods described by Ward et al. (12) to derive measures of lumbar and WB BMC for BA and BA for height Z-scores. Ward et al. used version 12.1 of the Hologic 4500 scanner in their studies, whereas we used version 11.2. These versions behave very similarly at the spine and also for the whole body except for subjects weighing less than 40 kg [Ref. 21 and Kelly, T., personal communication (Hologic)]. Therefore, two subjects with body weight less than 40 kg were not included in WB DXA analysis. Although we do not expect our data to be affected by the different software versions (Kelly, T. (Hologic), personal communication), we also report a healthy control population of similar age and maturity in this study, with the absolute and Z-score measures of bone density for AN girls being compared against the control population rather than just the reference population. The coefficient of variation (CV) for spine and WB BMD was 1.1 and 0.8% and 2.1 and 1.0% for fat and lean mass, respectively. Longitudinal measurements of bone density were performed on the same scanner for all subjects with the same software version.
Biochemical measurements
We used a RIA to measure estradiol (Diagnostic Systems Laboratories, Inc., Webster, TX; limit of detection 2.2 pg/ml and CV 6.5–8.9%), and an IRMA to measure IGF-I (Nichols Institute Diagnostics, San Juan Capistrano, CA; detection limit 30 ng/ml, CV 3.1–4.6%).
Statistical analysis
All data are presented as mean ± SD. Data were analyzed using the JMP program (version 4; SAS Institute Inc., Cary, NC). When two groups were compared, the Student t test was used to compare means. AN girls were then characterized as AN-recovered or AN-not recovered (see Experimental protocol). For three-group comparisons (AN-recovered vs. AN-not recovered vs. controls), ANOVA was used followed by the Tukey-Kramer test for intergroup comparisons to correct for multiple comparisons. Six subjects (one control, two AN-recovered, and three AN-not recovered) had follow-up data available at 6 months but not 12 months, and a carry-forward analysis was performed for these subjects. P < 0.05 was considered significant, and trends (P values between 0.05 and 0.10) are also reported. We used analyses of covariance (ANCOVAs) to determine significant predictors of changes in bone density measures.
|
Results |
|---|
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
Baseline and some bone density characteristics have been reported previously (3, 14, 15) and are summarized in Tables 1
and 2. At the spine, BMD, BMAD, and corresponding Z-scores were lower in AN girls. Height and height Z-scores did not differ between groups, indicating that lower spine and WB bone density measures in AN were not a consequence of short bones. Spine BA adjusted for height did not differ between the groups; however, lumbar BMC for BA was significantly lower in AN, indicating that lower lumbar bone density in AN indicates light and not thin bones. For the whole body, BMD, BMC to height ratio (BMC/Ht) and its Z-score were lower in AN than controls. WB BA adjusted for height was lower in AN, whereas WB BMC for BA did not differ between groups. Lower WB bone density measures in AN are therefore a consequence of thin rather than short or light bones. Of interest, despite the fact that WB BMC for BA was not significantly lower in AN girls, compared with controls, the ratio of BMC for BA Z-scores to BA for height Z-scores did not differ between the groups, suggesting proportionate decreases in these measures.
|
|
The AN-recovered group had a median duration of recovery of 9 months. The AN-recovered and AN-not recovered groups were comparable for baseline BMI, fat mass, and lean mass (Table 3). The AN-recovered group had significant increases in BMI and fat and lean mass, compared with controls and the AN-not recovered group. Height Z-scores did not change during follow up (–0.00 ± 0.20 in AN-not recovered vs. 0.00 ± 0.21 in AN-recovered and –0.06 ± 0.28 in controls, P value not significant). Height Z-scores increased minimally from 0.22 to 0.39 in premenarchal AN and from 0.81 to 1.21 in premenarchal controls. The difference in height Z-scores over the follow-up period did not differ between girls with AN and controls. No changes in height Z-scores were observed in postmenarchal girls. Changes in bone density over time were similar in premenarchal and postmenarchal AN girls (not reported), and the combined data are presented here.
|
Table 4 indicates changes in lumbar and WB bone measures in the three groups over the follow-up period. The changes in lumbar BMD and BMAD and their Z-scores were significantly lower in the AN-not recovered group, compared with controls, whereas the AN-recovered group did not differ from controls (Fig. 1). When we analyzed the data using ANCOVA with the final bone density measure as the dependent variable, and the initial bone measure and AN status as independent variables, differences between groups were similar to those observed for changes (
) in bone measures over time. In an ANCOVA model including recovery status (AN-not recovered, AN-recovered, and controls), change in height and change in lean mass, recovery status remained a significant predictor of change in lumbar BMC (P = 0.001), even after adjusting for change in height and lean mass. Conversely, recovery status was not a significant predictor of change in lumbar BA. These data suggest that changes in lumbar bone density in AN girls are a consequence of change in BMC rather than change in BA. This is corroborated by the fact that change in lumbar BA for height did not differ between the groups, whereas change in lumbar BMC for BA was higher in the AN-recovered group, compared with the AN-not recovered group. The AN-not recovered group lost lumbar BMC for BA, compared with controls.
|
|
As for the spine, when we analyzed the data using ANCOVA with the final bone density measure as the dependent variable and the initial bone measure and AN status as independent variables, differences between groups were similar to those observed for changes in bone measures over time. In an ANCOVA model including recovery status, change in height, and change in lean mass, recovery status remained a significant predictor of change in WB BA adjusted for height (P < 0.0001), WB BMC (P = 0.003), and WB BMC/Ht and its Z-score (P = 0.003 and < 0.0001), even after adjusting for height and lean mass changes. In contrast, recovery status was not a significant predictor of change in WB BMC adjusted for BA in this model.
In a subset analysis, we also analyzed data based on weight recovery alone without menses recovery (n = 6), compared with no weight gain (n = 14) and weight gain with menses recovery (n = 14). Compared with girls who did not gain weight, girls who gained weight but did not recover menses showed no differences with regard to changes in bone parameters at the lumbar spine. These girls did less well overall than girls who gained weight and resumed menses. Weight gain without menses recovery was associated with significant decreases in lumbar BMC for BA as opposed to girls with AN who gained weight and recovered menses who had an increase in this parameter over time (P < 0.05). For the whole body, weight gain alone was associated with an increase in BA, BMC, and BMC/Ht that was intermediate between those who did not recover weight and those who recovered weight and menses. However, weight gain without menses recovery was associated with significant decreases in BA adjusted for height as opposed to girls with AN who gained weight and resumed menses, who had an increase in this parameter over the follow-up period (P < 0.05).
|
Discussion |
|---|
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
Low BMD with energy restriction in AN girls is consistent with animal models of low BMD with energy restriction (22, 23). Few studies have examined bone density parameters in adolescents with AN prospectively and have looked at both the spine and the whole body (3, 24, 25), particularly using the Molgaard approach [WB only (26)]. In addition, most studies did not simultaneously study controls to compare rates of bone mass accrual in AN with healthy adolescents. In fact, of the studies described, only one compared AN girls with healthy adolescents (3). Most previous prospective studies did not differentiate AN girls by recovery status but reported changes from baseline to follow-up regardless of recovery status, reporting only overall mean weight changes during follow-up (3, 25, 26, 27). In addition, data available are contradictory with regard to changes in bone density with therapy for AN. Several prospective studies in adolescent AN did not observe an increase in bone density over time despite ongoing therapy and an overall increase in weight (3, 25, 26) or reported an increase at the spine but not for the whole body (24). These studies, however, did not differentiate AN girls by menstrual recovery. In contrast, two other studies did report improvement in spine bone density after good treatment outcome (27, 28). In a cross-sectional study, Audi et al. (29) reported a trend toward better bone density in AN girls who recovered weight and menses, compared with nonrecovered girls and girls recovering weight alone, but did not follow the girls prospectively or compare with controls. Nor did they report detailed structural data.
Our data are comprehensive and prospective for the spine (primarily trabecular bone) and the whole body (primarily cortical bone), and we present comparative data in controls of similar maturity. In addition, we differentiate AN girls based on recovery status rather than collectively providing information for all AN. We have been able to differentiate effects of weight gain alone (associated with improved IGF-I levels) from effects of weight gain with menses recovery (improved IGF-I levels and gonadal recovery) in this prospective study. Because bone mass accrual in adolescence depends greatly on rising levels of IGF-I and estrogen, maximal increases in bone density would be expected when IGF-I levels increase as weight improves and levels of estradiol increase, as indicated by menses recovery. The novelty of our data also lie in the analysis using the Molgaard approach, which has not been previously reported in this group based on recovery status, particularly for the spine, and using adjustments for volumetric bone density while following up subjects longitudinally. Finally, previous studies have not reported baseline prognostic indicators of bone density changes AN, as reported in the accompanying manuscript.
Healthy adolescents continue to accrue bone mass through adolescence, and we observed a 2.8% increase in lumbar BMD over 1 yr in controls. Most studies have demonstrated that short-term weight recovery is not associated with increases in BMD in adolescents with AN (2, 3, 5). Our data are consistent with these findings but, importantly, show that BMD stabilized in these girls in contrast to girls who were not recovering, in whom BMD continued to decrease. At the spine, decreased mineralization (light bones) appears to cause low bone density in AN rather than thin or short bones, given that lumbar BMC for BA was low, whereas height Z-scores and lumbar BA for height did not differ from controls. Recovery was associated with an increase in lumbar BMC for BA and, if sustained, would be expected to eventually result in an increase in lumbar BMD, BMAD, and corresponding Z-scores. A longer period of follow-up is necessary to determine whether BMD increases with continued weight and reproductive function recovery.
For WB bone measures, lower BMD, BMC, and BMC/Ht appear to be a consequence of thin rather than short or light bones, given that WB BA for height was decreased in AN, whereas height Z-scores and BMC for BA were not different from controls. Of note, although WB BMC/Ht Z-scores were overall lower in AN than in controls, this difference was not statistically significant. Interestingly, the ratio of WB BMC/BA Z-scores to WB BA/Ht Z-scores did not differ between the groups, suggesting that reductions in these measures occur in proportion in AN. It is therefore surprising that the reduction in WB BMC/BA Z-scores in AN, compared with controls, did not reach statistical significance. The observation of thin bones contributing to low WB bone density in AN is consistent with data reported by Stone et al. (26); however, these authors did not examine effects of recovery in their subjects. In our study, recovery was associated with significant increases in WB BA for height, suggesting a reversal of bone loss. Of concern, although WB BA for height increased with recovery and there was an increase in absolute BMC, this increase in BMC was not sufficient to keep pace with the increasing BA for height, and BMC for BA decreased with recovery. Studies are necessary to determine whether, with sustained recovery, increases in BMC will catch up with increases in BA. This raises concerns that a transient period of wider but undermineralized bones in girls with AN may occur as they recover weight. The impact of this finding on bone architecture and strength is unclear.
The differential effects at baseline of AN on BA for height and BMC for BA at different skeletal sites are of interest. At the spine, AN girls had a sparing of BA for height, but BMC for BA was lower than in controls, whereas for the whole body, BMC for BA was spared and BA for height was lower than in controls. Therefore, nutritional status and hypogonadism preferentially appear to affect BMC at sites of primarily trabecular bone and BA at sites of primarily cortical bone. The mechanism underlying these effects needs to be further explored.
The Z-score compares the patient’s bone density with the mean for age and gender (based on the Hologic database), and as long as the rate of increase in bone mass over a period of time is as expected for age and gender, the Z-score should not change and the change in Z-score should be zero or close to zero. Among healthy controls, the change in Z-scores for BMD and BMAD at the lumbar spine was negative, but only minimally so, and may reflect subtle differences in our control population in terms of bone mass accrual over a year, compared with the reference population used by Hologic to develop their database. However, the change was close to zero as one may expect. Among AN girls who recovered, an increase in bone density would be expected. However, if the rate of increase in bone density did not approximate that of healthy adolescents, the Z-score would decrease over time instead of staying constant, and as a consequence, the change in Z-score would be negative. Indeed, although BMD and BMAD at the spine increased over the year in recovered AN, the increase was less than that seen in controls over this period, likely because of residual weight deficits. However, the decrease in Z-scores was definitely less than that observed in girls with AN who did not recover and who had a decrease in spine BMD and BMAD over time. We noted similar observations for Z-scores for the whole body (WB BMC/Ht).
Of note, weight gain without menses recovery did not improve bone parameters at the spine, and this underscores the importance of gonadal steroids in optimizing trabecular bone density, whereas WB parameters improved somewhat with weight gain, even without menses recovery, although not to the extent observed with weight gain and menses recovery. This observation underscores the importance of optimizing both weight and gonadal status to improve bone density parameters for the whole body.
Limitations of our study include the fact that duration since diagnosis was variable in our subjects, and duration of illness could not be determined with accuracy based on history. It is concerning, however, that despite a relatively short duration since diagnosis, AN girls had markedly lower bone density than controls. It is also possible that some subjects were on the road to recovery when they were enrolled in the study, although all subjects met criteria for AN at enrollment. We acknowledge that a longer duration of follow-up would be useful to better characterize long-term changes in bone density measures.
In addition, controversy exists regarding the best method to determine bone density in AN. Changes in body composition and fat distribution can cause an over- or underestimation of bone density measures by DXA (30), and significant alterations in body composition are characteristic of AN (31). In addition, Hologic uses an image thresholding algorithm to identify bone pixels and does not use discrete edge detection, and this brings to question the accuracy of bone area measurements as reported by DXA. Other modalities for assessing bone density are being studied in AN, but until these are validated, DXA remains the tool available to assess bone density in this condition. Of note, QCT also indicates lower bone density in AN than controls (32), but follow-up data with QCT are lacking.
Our data indicate that weight gain and menstrual recovery are associated with a halting of deterioration in bone measures, even in the short term, and it is hoped that with sustained recovery, significant increases in bone density parameters will be observed. Our data emphasize the importance of optimizing weight gain in adolescents with AN and indicate the beneficial effects of even short-term increases in weight with resumption of menses.
|
Footnotes |
|---|
Disclosure Statement: All authors have no conflict of interest to report.
First Published Online December 18, 2007
Abbreviations: AN, Anorexia nervosa; ANCOVA, analysis of covariance; BA, bone area; BMAD, bone mineral apparent density; BMC, bone mineral content; BMC/Ht, BMC to height ratio; BMD, bone mineral density; BMI, body mass index; CV, coefficient of variation; DXA, dual-energy x-ray absorptiometry; QCT, quantitative computed tomography; WB, whole-body.
Received June 26, 2007.
Accepted December 7, 2007.
|
References |
|---|
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
This article has been cited by other articles:
 |
|
 |
|
  L. K. Bachrach and L. M. Ward Clinical Review: Bisphosphonate Use in Childhood Osteoporosis J. Clin. Endocrinol. Metab., February 1, 2009; 94(2): 400 - 409. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
R. L. Prince and K. Zhu Whole-Body Dual-Energy X-Ray Absorptiometry Comes of Age: Bone Structural Measures and Their Physiological Determinants in Anorexia Nervosa J. Clin. Endocrinol. Metab., April 1, 2008; 93(4): 1178 - 1180. [Full Text] [PDF]
|
|
|
|
|
|
M. Misra, R. Prabhakaran, K. K. Miller, M. A. Goldstein, D. Mickley, L. Clauss, P. Lockhart, J. Cord, D. B. Herzog, D. K. Katzman, et al. Prognostic Indicators of Changes in Bone Density Measures in Adolescent Girls with Anorexia Nervosa-II J. Clin. Endocrinol. Metab., April 1, 2008; 93(4): 1292 - 1297. [Abstract] [Full Text] [PDF]
|
|
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| Endocrinology | Endocrine Reviews | J. Clin. End. & Metab. |
| Molecular Endocrinology | Recent Prog. Horm. Res. | All Endocrine Journals |
