ARTICLE IN PRESS

Journal of Adolescent Health 000 (2018) 1 6

www.jahonline.org

Original article

Long-Term Outcomes of Adolescent Anorexia Nervosa on Bone

D1X XJessica Mumford, D2X XM.B.B.S.a,b, D3X XMichael Kohn, D4X XM.B.B.S., Ph.D.a,c,
D5X XJulie Briody, D6X XM.Bio.Med.Eng.d, D7X XJane Miskovic-Wheatley, D8X XM.Sc., D.Clin.Psy.a,
D9X XSloane Madden, D10X XPh.D.a, D1X XSimon Clarke, D12X XM.B.B.S.c,e, D13X XAndrew Biggin, D14X XM.B.B.S., Ph.D.f,g,
D15X XAaron Schindeler, D16X XPh.D.h,i,*, and D17X XCraig Munns, D18X XM.B.B.S., Ph.D.f,i
a
Eating Disorder Service, Sydney Children’s Hospital Network, Westmead, New South Wales, Australia
b
School of Rural Medicine, University of New England, Armidale, New South Wales, Australia
c
Adolescent and Young Adult Medicine, Westmead Hospital, Westmead, New South Wales, Australia
d
Nuclear Medicine, The Children’s Hospital at Westmead, Westmead, New South Wales, Australia
e
The Children's Hospital at Westmead, Westmead, New South Wales, Australia
f
Department of Endocrinology, The Children’s Hospital at Westmead, Westmead, New South Wales, Australia
g
Children’s Hospital Westmead Clinical School, University of Sydney, Sydney, New South Wales, Australia
h
Orthopaedic Research Unit, The Children’s Hospital at Westmead, Westmead, New South Wales, Australia
i
Discipline of Paediatrics & Child Health, University of Sydney, Sydney, New South Wales, Australia

Article History: Received February 12, 2018; Accepted July 25, 2018
Keywords: Anorexia nervosa; Bone; Bone density; Bone mineral density; Bone morphology; Dual-energy x-ray absorptiometry; DXA;
Peripheral quantitative computed tomography; pQCT

A B S T R A C T
IMPLICATIONS AND
CONTRIBUTION
Purpose: Anorexia nervosa (AN) is a chronic and life-threatening eating disorder that can have a consid-
erable negative impact on the growing skeleton. We hypothesized that the long-term impact on bone
While prior studies have
health may persist even after normalization of body weight.
shown the acute effects
Methods: 41 females (mean age 21.2 § 2.9 years) with a history of adolescent-onset AN attended a follow-
adolescent AN can have on
up bone health assessment at 5 years (T5, n = 28) or 10 years (T10, n = 13) after their first AN-related hospi-
bone health, this report
tal admission. Assessment included dual-energy x-ray absorptiometry measurements of the total body,
highlights the long-term
lumbar spine, and proximal femur, peripheral quantitative computed tomography at the radius and tibia,
impact that may place indi-
anthropometric measurements, serum biochemistry, fracture history, and a patient questionnaire.
viduals at increased risk of
Results: A recovery in body weight and BMI was seen for both the T5 and T10 cohorts (BMI at intake 16.6,
osteoporosis and fracture
BMI at T5-T10 21.2-21.3). Dual-energy x-ray absorptiometry body composition indicated a recovery of fat
long after recovery from a
mass and lean tissue mass. Total BMD was unaffected, but reductions were seen at the femoral neck and
severely low BMI.
arms. Peripheral quantitative computed tomography showed reduced trabecular and cortical bone in the
radius, and cortical thinning in the tibia. AN patients showed a statistically significant reduction in measures
of radiographic bone health at follow up, although not to a degree that necessitated clinical intervention.
Serum insulin-like growth factor 1 was also positively correlated with total BMD and BMC measures. While
fracture risk was not increased, a subset of participants (8%) showed multiple (>4) fractures.
Conclusion: A longitudinal study of adolescent AN showed persisting negative effects on bone health.
Crown Copyright © 2018 Published by Elsevier Inc. on behalf of Society for Adolescent Health and Medi-
cine. All rights reserved.

Conflicts of interest: Dr Jessica Mumford, Dr Michael Kohn, Ms Julie Briody, Dr Jane Miskovic-Wheatley, Dr Sloane Madden, Dr Simon Clarke and Dr Andrew Biggin declare
no conflict of interest.
* Address correspondence to: Aaron Schindeler, Orthopaedic Research Unit, The Children's Hospital at Westmead, Locked Bag 4001, Westmead, NSW 2145, Australia.
E-mail address: craig.munns@health.nsw.gov.au (C. Munns).

1054-139X/Crown Copyright © 2018 Published by Elsevier Inc. on behalf of Society for Adolescent Health and Medicine. All rights reserved.
https://doi.org/10.1016/j.jadohealth.2018.07.025
 ARTICLE IN PRESS
2 J. Mumford et al. / Journal of Adolescent Health 00 (2018) 1 6
Interference with bone development, and the attainment of
peak bone mass (PBM) and peak bone strength, can have dramatic
effects on osteoporosis and fracture risk in later life. An increase in
PBM by one standard deviation is estimated to reduce fracture risk
by up to 50% [1]. Conversely, disruptions in attaining PBM, as can
be seen with aberrant pubertal timing, can lead to sustained future
fracture risk [2]. This has led many to view nutrition affecting bone
health as critical for long-term outcomes in bone density mainte-
nance and fracture prevention [3].
Anorexia nervosa (AN) is characterised by the restriction of
energy intake relative to requirements, low body weight, an
abnormal body image, and a fear of weight gain. AN is the
third most common chronic disorder in adolescent females [4]
and is associated with multiple and significant physical compli-
cations. AN can adversely affect bone health and is associated
with low bone density, compromised bone quality, and
increased fracture risk [5]. One of the first studies examining
the relationship between AN and bone health in adolescent
girls showed that areal bone mineral density (aBMD) was sig-
nificantly reduced [6]. Moreover, in this patient cohort aBMD
did not improve after weight recovery in the 6 and 12 months.
In a broader study of a larger cohort where nutritional status,
hematologic, and biochemical parameters were analysed, the
relative bone ages of AN adolescents was significantly reduced
and correlated with the length of illness [7]. Hormones includ-
ing IGF-1 and oestradiol that affect bone health were signifi-
cantly reduced in subjects with AN.
The prior studies used dual-x-ray absorptiometry (DXA) for
measures of aBMD, lean mass and body fat. However, a 2008 study
reported on high-resolution computed tomography (CT) scanning
of the distal radius [8]. AN subjects showed significantly reduced
bone volume density (BV/TV) and trabecular thickness, and higher
trabecular separation. Both cortical and trabecular microarchitec-
ture were subsequently shown to be affected at this site [9].
In a cohort of 418 females (310 with AN, 108 controls), subjects
with AN showed a lifetime prevalence for prior fracture of 31%
compared with 19% in normal-weight controls. Moreover, in this
report, fracture incidence peaked after a diagnosis of AN. The risk
was increased without significant reductions in aBMD [10].
In addition to this well-characterized burden on acute bone
health, we hypothesized that poor bone health and fracture risk
may persist after weight recovery. Future bone health issues may
particularly affect individuals with adolescent AN, as this is a time
of peak bone growth and development as reflected by significant
reductions in BMD Z-scores in the arms, femoral neck, and L1-4
areal BMD. Thus, the main objective of this research project was to
evaluate the natural history of bone health in patients with a previ-
ous diagnosis of adolescent-onset AN at 5 or 10 years post-diagno-
sis. The primary outcome measures were radiographic whole
body DXA and peripheral Quantitative Computed Tomography
(pQCT). Secondary measures included bone turnover markers and
hormones, clinical bone health history, and current eating disorder
and weight status.
Methods
Ethics
This study was approved by the Sydney Children’s Hospitals
Network’s Human Research Ethics Committee (HREC) on the 16th
April, 2014 (reference number 2006/114).
Recruitment
This study sample consisted of patients who had previously
been diagnosed with AN and admitted to and treated by the Eating
Disorder Service at The Children’s Hospital at Westmead, Sydney,
Australia. The diagnosis was made by a multidisciplinary clinical
team and was consistent with criteria of the DSM-IV (Diagnostic
and Statistical Manual of Mental Disorders, 4th Edition) and initial
presentation was before the age of 18 years [11]. Throughout treat-
ment, 198 patients, including 4 males, elected to take part in stud-
ies assessing the outcomes of their eating disorder. These studies
included bone and body composition assessment by DXA and
blood tests [12,13].
Follow up written and phone contact was attempted on 198
patients who had taken part in either of two prior studies. Of these,
84 were unable to be contacted and 57 declined enrolment. Of the
remainder, 16 met one or more of the following exclusion criteria:
male gender (due to small sample size), pregnancy, and absence of
a baseline DXA. Patient information sheets approved by the Syd-
ney Children’s Hospitals Network’s Human Research Ethics Com-
mittee were distributed to all participants and informed consent
was obtained.
Patients were each assessed during a single hospital appoint-
ment. Patients had DXA and pQCT scans performed; anthropomet-
ric measurements (height, weight, and BMI) measured; a blood
test; and also completed questionnaires.
DXA assessment
DXA measurements of the Total Body, AP lumbar spine (LS)
and proximal femur (right) were acquired on a GE-Lunar Prodigy
scanner, using enCore versions 6.1, 8.1, and 8.6 for baseline scans
and version 13.60 for current follow-up. Scan acquisition parame-
ters were the same for all three software versions. Scan acquisition
mode was as suggested by the scanning software. “Thin” and
“Standard” were the only modes used. Scan analysis used the man-
ufacturer’s recommended techniques and enCore13.60 (GE Medi-
cal Systems Lunar, Madison, WI, United States). “Standard” total
body analysis was performed as the “Enhanced” Total Body analy-
sis option is not available on enCore13.60.
Variables analyzed included body height, body weight, aBMD
(g/cm2), projected bone area (BA in cm2), lean tissue mass (LTM),
Fat %, and bone mineral content (BMC) relative to age. Skeletal
sub-regions selected for analysis included the upper limbs (arms),
L2-4 LS, and right femoral neck (RFN). The bone mineral apparent
density (BMAD, g/cm3) of the LS (L1-4) was estimated as the ratio
between BMC and the extrapolated 3-dimensional bone volume
(BV in cm3) using an established method for deriving BV from BA
[14].
pQCT assessment
As a radiographic scanning modality, pQCT allows discrete
analysis of the cortical and trabecular bone compartments. pQCT
scans of the non dominant radius and tibia were performed using
a Stratec Medizintechnik GmBH XCT 2000 (Stratec Biomedical AG,
Birkenfeld, Germany) and software version 6.00B. To determine
limb dominance, the patient was asked which leg they used to kick
a ball (non dominant leg) and the hand they used to write (domi-
nant arm). Limb length was determined manually using a tape
measure, as required by the scanning software to determine the
distance from a “Reference” position to the measurement sites. For
 ARTICLE IN PRESS
J. Mumford et al. / Journal of Adolescent Health 00 (2018) 1 6
Table 1
vBMD analysis
RADIUS 4% CALCBD Threshold 280
RADIUS 65% CORTBD Threshold 711
TIBIA 4% CALCBD Threshold 169
TIBIA 66% CORTBD Threshold 711
the tibia, distance from the tibial plateau to the highest point of the
medial malleolus was measured, and for the radius, distance from
the distal end of the processus styloideus ulnae to the olecranon
was measured.
A scout scan was used to visualize the correct “Reference” posi-
tion. The majority of individuals showed a lack of continuous scle-
rotic lines on the scout view indicative of completing growth. For
these, the manufacturer recommended “Reference” position was
used (the middle of the articular cartilage at the endplate of the
tibia/fibula and ulna/radial junction). In studies of actively growing
patients, the “Reference” position was selected atop the most
proximal density line surrounding the growth plate so long as this
was proximal to the standard adult position.
Cross-sectional scans of the radius 4% and 65% sites and tibial
4% and 66% sites were performed. Scan speeds of 15 and 20 mm/sec
were used for the radius and tibia, respectively, with a .4 mm voxel
size. The effective radiation dose was 1 microSv per pQCT and the
total effective radiation dose for pQCT during this study was
3 microSv.
Subsequently, scans were analyzed using pre-determined
thresholds and modes (Table 1). The distal epiphysis (4% site)
determines total and trabecular volumetric bone mineral density
(vBMD), BMC, and total bone cross-sectional area (CSA). The diaph-
ysis (65% site for radius and 66% for tibia) determines total bone
CSA, cortical CSA, cortical thickness, cortical vBMD, cortical BMC,
and stress strain index (SSI).
Anthropometric measurements
Height (§.1 cm) was measured with a stadiometer (Holtain Ltd,
UK). Body weight (§.1 kg) was measured in light clothing using
electronic scales (AND FW-150K, Japan). BMI SDS (body mass
index standard deviation score) was obtained from the CDC 2000
growth charts [15].
Serum biochemistry
A single fasting blood test was conducted to measure hormones
associated with bone health as well as bone turnover markers.
Measurements included: calcium, magnesium, phosphorus, free
T3 (fT3), free T4 (fT4), thyroid stimulation hormone, estradiol, leu-
tenizing hormone, follicle stimulating hormone, insulin-like
growth factor-1 (IGF-1), 25-hydroxyvitamin D, serum alkaline
phosphatase, parathyroid hormone, procollagen type 1 N propep-
tide (P1NP), and c-terminal telopeptide of type 1 collagen. The
total of 39 of 41 patients agreed to have blood tests (27/28 in T5
and 12/13 in T10). Baseline biochemistry data was obtained via
electronic medical records, including estradiol, IGF-1, and fT3.
Statistical analysis
Statistical analysis was conducted using SAS Software (SAS
Institute, Cary, NC, United States). Patients were analyzed as a
3
CONTMODE 1 PEELMODE 1 Trabecular area 45% of Total Bone area
CORTMODE 1
CONTMODE 1 PEELMODE 1 Trabecular area 45% of Total Bone area
CORTMODE 1
whole group, and by T5 and T10 subgroups. For subjects under
25 years old, age and ancillary Z-scores were obtained from in-
house pediatric reference data, which is an updated and expanded
dataset (n = 812) of previously published cohorts from Sydney,
Australia [16,17]. GE-Lunar software was also used to calculate
total BMD, LS BMD, and RFN Z scores, which were more applicable
for subjects over 25. Non GE-Lunar Z-score measures exclude the
n = 6 eldest subjects.
pQCT age Z-scores were calculated based on both published
data and self-reported data [18 20]. Due to the lack of normative
tibia data for patients over 19 years of age, Z-scores for patients
over 19 years old were calculated based on the mean and SD refer-
ence values published for 19 year old females.
For DXA and pQCT Z-scores a t-statistic was generated based on
the mean Z-score, sample size, and standard deviation based on a
mean normal Z-score of 0. The t-statistic was used to determine a
2-tailed P-value with a cut-off for statistical significance set to .05.
As the eldest 6 patients were over 25 years old, in-house DXA Z-
scores were only able to be calculated for n = 7 of the T10 group.
For analyzing correlation between bone composition/serum
markers and BMD measures, Spearman correlation coefficients
were calculated between bone parameters and other parameters.
Patient questionnaire
The Eating Disorder Examination Questionnaire (EDE-Q) and a
customized questionnaire (to assess menstrual, fracture, and exer-
cise histories data) were conducted using a secure application and
a mobile device. The EDE-Q is a gold standard measure commonly
used in clinical practice to assess the current eating disorder status
of patients, and calculates scores for common AN-related behav-
iors and concerns: restraint (regarding limiting and avoidance of
eating), eating concern, shape concern and weight concern, as well
as a mean global score [21].
Results
Forty-one patients previously diagnosed and treated for AN
were recruited for reassessment for bone health parameters.
Twenty-eight respondents were within »5 years of initial admis-
sion (T5) and 13 respondents within »10 years of initial admission
(T10). The cohort showed mean EDE-Q (global) scores within one
standard deviation of normal reference data, inconsistent with a
current diagnosis of AN and at low risk of relapse [22] (Table 2).
Body composition analysis shows that AN has long-term effects on
bone health
Analysis of DXA body composition was performed using retro-
spective data (baseline data obtained at the time of the initial diag-
nosis of AN), and from the 5-year or 10-year follow-up data. At
baseline, the mean age of the entire cohort at was 14.6 years (SD
1.5), mean BMI was 16.6 (SD = 1.7) and mean Fat % was 15.9%
 ARTICLE IN PRESS
4 J. Mumford et al. / Journal of Adolescent Health 00 (2018) 1 6

Table 2 Table 4
EDE-Q results of whole cohort (n = 41) at follow-up pQCT of radius and tibia for whole cohort (n = 41)

Cohort Normative p-value z-score Mean SD 95% CI Mean P

Variable Mean SD Mean SD RADIUS

4% BMC Total -1.0 1.1 -1.3 ¡.6 <.001
Global Score 1.7 1.5 1.6 1.2 .85 4% CSA Total Bone ¡.1 1.2 ¡.5 .3 .56
Restraint 1.5 1.5 1.3 1.3 .36 4% vBMD Trabecular ¡.4 1.1 ¡.67 .0 .04
Eating Concern 1.2 1.5 .6 .9 .01 4% vBMD Total ¡.8 1.2 -1.2 ¡.4 <.001
Shape Concern 2.2 1.7 2.1 1.6 .92 66% BMC Cortical ¡.3 1.1 ¡.7 .2 .06
Weight Concern 1.8 1.6 1.6 1.4 .41 66% Cortical Thickness ¡.4 1.2 ¡.8 .0 .03
66% CSA Cortical ¡.7 1.2 -1.1 ¡.3 <.001
66% CSA Relative Cortical ¡.1 1.1 ¡.5 .2 .45
(SD = 6.6). At the follow-up appointments, the mean age of the 66% CSA Total Bone ¡.5 1.3 ¡.9 .0 .04
66% pSSI ¡.6 1.3 -1.0 ¡.2 .006
entire cohort was 21.2 years (SD 2.9) and the individuals showed
66% vBMD Cortical .0 3.1 -1.0 1.0 .98
improved mean BMI of 21.2 (SD = 2.9) and Fat % of 30.5% (SD = 8.7). TIBIA
At baseline, the entire cohort’s body weight, BMI, Fat %, LTM 4% vBMD Trabecular ¡.1 .6 ¡.3 .1 .26
relative to height, and BA relative to height were lower than con- 66% BMC Cortical .0 .6 ¡.2 .2 .80
trol data (Table 3). There was no significant reduction in total 66% Cortical Thickness ¡.3 .7 ¡.5 ¡.1 .02
66% CSA Cortical ¡.2 .6 ¡.4 .0 .06
BMD. However, BMD in the upper limbs, RFN, and spine (both vol- 66% CSA Total Bone .1 .6 ¡.1 .3 .55
umetric L1 L4 spine and adult GE-Lunar LS Z-scores) were signifi- 66% pSSI .49 .67 .27 .70 <.001
cantly reduced across the cohort. 66% vBMD Cortical .75 .36 .63 .86 <.001
At follow-up, clinical improvement was reflected by body
weight and BMI Z-scores showing no difference from normative PQCT confirms that AN has long-term effects on bone health
standards (Table 3). LTM relative to height increased but remained
low (P = .02), which could explain the relative increase in BMC rel- pQCT data of the tibia and radius for the entire cohort are
ative to LTM. Overall, total BMD was significantly reduced com- shown in Table 4. The radius was particularly affected, with the
pared with normative standards and this was further reflected in epiphyseal (4%) BMC and vBMD being significantly below average
site-specific measures in the spine and upper limbs (Table 3). based on normative values [16,17]. Cortical vBMD was elevated,
Stratified analysis of the T5 and T10 groups separately showed while thickness and cross-sectional area in the diaphysis were sig-
that the mean Z-score for bone measures were still below average nificantly reduced resulting in a reduction in SSI. In the tibia, corti-
(negative) in the T10 cohort (Supplementary Table 1). While less cal vBMD was again increased, while cortical thickness and SSI
measures achieved statistical significance, this was likely due to were reduced.
the reduced statistical power associated with separating the two Subgroup analysis of the T5 and T10 cohorts separately showed
subgroups. Notably, in the T10 group 10 years after the initial similar findings (Supplementary Table 2). Again the radius showed
diagnosis of AN (average age 24.7 years) the mean BMD Z-score more metrics reaching statistical significance compared with the tibia.
was ¡1.3 in the upper limbs and ¡1.9 in the L1 L4 volumetric Both the T5 and T10 cohorts showed a trend towards an average reduc-
spine measures. These data suggest a major long-term impact of tion in tibial cortical thickness, however, this did not reach significance
AN on bone health. presumably due to a lack of statistical power in the subgroup analysis.

Table 3
Body composition of whole cohort: baseline and follow-up

Baseline (n = 41) Follow-Up (n = 35; n = 41 for Lunar scores)

Mean SD 95% CI mean P Mean SD 95% CI mean P

Age (yrs) 14.6 1.5 14.1 15.0 21.2 2.9 20.3 22.1
Height (cm) 160.7 7.5 158.3 163.1 164.7 5.6 162.9 166.5
Weight (kg) 43.0 6.9 40.8 45.1 57.9 11.0 54.4 61.4
BMI 16.6 1.7 16.0 17.1 21.2 3.2 20.2 22.2
Fat % 15.9 6.6 13.8 17.9 30.5 8.7 27.8 33.3
Height Z .1 1.0 ¡.3 .4 .72 .2 .9 ¡.1 .5 .11
Weight Z ¡1.2 1.1 ¡1.5 ¡.9 <.001 ¡.3 1.4 ¡.7 .2 .19
BMI Z ¡1.6 1.0 ¡1.9 ¡1.2 <.001 ¡.5 1.9 ¡1.1 .1 .11
Fat % Z ¡1.2 .9 ¡1.5 ¡.9 <.001 .2 1.2 ¡.1 .6 .19
LTM / Height Z ¡.8 .9 ¡1.1 ¡.5 <.001 ¡.5 1.3 ¡.9 ¡.1 .02
BA / Height Z ¡.5 .6 ¡.7 ¡.3 <.001 .3 1.0 .0 .6 .03
Total BMC Z ¡.3 1.4 ¡.7 .1 .18 ¡.2 1.2 ¡.6 .3 .47
BMC / LTM Z .2 1.0 ¡.2 .5 .31 1.0 1.7 .5 1.6 <.001
Total BMD Z .0 1.3 ¡.4 .4 .97 ¡.6 .9 ¡.9 ¡.3 .001
LS BMD Z ¡.2 1.2 ¡.6 .8 .29 ¡.4 .8 ¡.6 ¡.1 .01
Arms BMD Z ¡.6 1.1 ¡1.0 ¡.3 <.001 ¡.9 ¡1.2 ¡1.2 ¡.7 <.001
RFN BMD Z ¡.5 1.1 ¡.8 ¡.1 .01 ¡.3 .9 ¡.6 .1 .10
L1-4 BMAD Z ¡.4 1.1 ¡.8 ¡.1 .01 ¡.9 1.0 ¡1.3 ¡.6 <.001
Lunar Total BMD Z ¡.1 .9 ¡.4 .2 .56 .2 .9 ¡.1 .5 .21
Lunar LS BMD Z ¡.4 1.0 ¡.7 ¡.1 .02 ¡.4 .8 ¡.6 ¡.2 .002
Lunar RFN BMD Z ¡.2 .9 ¡.4 .1 .31 .2 .8 .0 .5 .07
 ARTICLE IN PRESS
J. Mumford et al. / Journal of Adolescent Health 00 (2018) 1 6 5

Table 5
Spearman correlation coefficients for relationships between DXA bone parameters and other parameters at time of follow-up

Variable Total BMC Z Total BMD Z LS BMD Z Arms BMD Z RFN BMD Z L1-4 BMAD Z
b b a a a
BMI Z .78 .57 .38 .50 .40 .02
Weight Z .84a .56b .37a .49 .38a ¡.06
LTM Z .66b .41a .21 .43b .11 ¡.35a
Estradiol .06 .08 .24 .05 .21 .39a
FT3 .10 .04 .43b .19 ¡.01 .22
IGF-1 .35a .41a .46a .35a .40a .31
C-terminal telopeptide of type 1 collagen ¡.04 .01 .08 .17 .04 .13
P1NP ¡.10 .03 .25 .00 ¡.09 .39a
25-OH Calciferol ¡.06 ¡.14 ¡.17 ¡.06 ¡.35c ¡.35a
Parathyroid hormone .23 .17 .12 .15 .06 .11
Hours of Exercise/Week .30 .24 .23 .37a .06 ¡.02
Intensity of Exercise ¡.01 ¡.03 .02 ¡.06 .15 .23
EDE-Q Global Score .14 .00 .02 .08 ¡.09 ¡.14
a
P < .05
b
P < .01
c
P = .052

Correlation of body composition and serum biomarkers with bone
density
Bone turnover markers were investigated as an attempt to under-
stand bone metabolism at the point of follow up, and to help explain
the bone phenotype seen on DXA and pQCT. These markers are used
in adults to guide therapy, although there are caveats to their use.
BMI, body weight, and LTM Z-scores all showed high correla-
tions with BMC (.66 .84) and bone mineral density (.41 .57)
(Table 5). These data suggest lower body weight and BMI may be
limited predictive tools for reductions in bone mass.
Most of the serum markers showed poor correlation or no sta-
tistical significance with respect to total body or regional BMD
measures. IGF-1 showed the greatest correlation with BMD (both
total and site specific for lumber spine, upper limb, and RFN areal
BMD), ranging from .35 to .46 at all sites (P < .01). For volumetric
BMAD measures, a significant correlation was seen with P1NP and
estradiol. 25-OH calcifierol showed a negative correlation with
BMD at select sites. Notably, clinical assessment of all of the indi-
vidual values showed none of the values to be in a range that
would lead to clinical concern.
A significant positive correlation was seen between upper limb
BMD and hours of exercise per week (.37, P < .05). No correlation
was seen between BMD and with the EDE-Q global score or other
self-reported exercise measures tested by the EDE-Q.
Patients report a normal rate of fractures
Self-reporting of fractures showed that 36% of women diag-
nosed with adolescent AN sustained at least one fracture, and that
8% had multiple fractures (>4) (Supplementary Table 3). While the
total fracture rate was within an expected range [23,24], it is
acknowledged that three patients experienced an unexpectedly
high number of fractures (4 6).
Discussion
This research has been the first to assess the long-term out-
comes of adolescent-onset AN on bone health in a longitudinally
assessed cohort. Analysis of the baseline data for the combined
cohort of adolescent girls diagnosed with AN shows considerable
reductions in BMI, body weight, Fat %, and LTM consistent with a
severe eating disorder. While BMI and body weight returned to
healthy values in both the 5-year and 10-year cohorts, bone health
appears to be impeded compared to normative standards. These
results support earlier studies showing adolescent AN impacts
long-term bone health and has long-term effects [1,2].
The fracture prevalence in this study cohort was 36% and was
similar in all groups, and was within the observed range of between
30% and 50% for the general female population [23,24]. Other
reports, in contrast, have demonstrated an increased prevalence of
fracture after adolescent-onset AN by making comparisons with
healthy control populations [10,25]. The prevalence of multiple frac-
tures in this study cohort, however, can be considered as concern-
ing, particularly with a subset of patients having had an abnormally
high number of fractures. While the prevalence of multiple fractures
was more than four times greater in T10 than in T5 A larger sample
size and study of more than 10 years duration are required to defini-
tively demonstrate if the impact on the bone health of adolescent
AN confers an increased fracture risk [26]. One feasible possibility is
to follow up on the T5 group in five years, which would also enable
a longitudinal comparison with this study.
The pQCT data provides insight into the mineral density of bone.
An increase in the mineral density of bone with alteration of the
mechanostat set-point of bone, would be expected to result in reduc-
tion in trabecular vBMD, increased cortical vBMD and reduction in
cortical thickness and cross sectional area. Such findings have been
reported in osteogenesis imperfecta [27]. These are precisely the find-
ings in the current study and suggests that the increase in bone min-
eral density may be a major etiological factor in the reduction in
bone strength seen in AN. In boys with Duchenne muscular dystro-
phy high mineral density is felt to result from low bone turnover
[28]. The DXA data may also support this. These data showed a reduc-
tion in bone area and reduction in LS (trabecular) values. The reduc-
tion in total body aBMD is likely a reflection of a mismatch between
the rates of increases in bone area and BMC, with BMC increases lag-
ging during recovery. Also, the reduction in BMC/LTM at follow-up
may reflect a bone that is unable to respond appropriately to the
stress of muscle pull, due to its high mineral density.
One unexpected finding was that serum IGF-1 showed a strong
correlation with bone density measures. IGF-1 is a circulating hor-
mone with a critical role in childhood growth that is purported to
have an anabolic effect on adults. It is a primary mediator of the
effects of growth hormone. Limited literature exists relating to
IGF-1 and bone, however, in obese adults IGF-1 was reported to
correlate with higher bone mineral density and the bone marker
osteocalcin [29]. In another recent study, low serum IGF-1 was
found to be a marker for frailty, lower BMD, and reduced handgrip
 ARTICLE IN PRESS
6 J. Mumford et al. / Journal of Adolescent Health 00 (2018) 1 6
strength in both genders in the middle aged and elderly [30], sug-
gesting its generic utility as a marker of musculoskeletal health.
One may postulate that low IGF1 could result in reduction in bone
turnover (modeling and remodeling) and could be responsible for
the increased bone density discussed above.
While this study has resulted in novel findings regarding long-
term bone health in a longitudinal cohort, the call-back response
rate was low, limiting the scope of the study. Long-term patient
call-back is difficult to achieve, with individuals being lost to fol-
low up, moving out of the area, or being disinclined to participate
in additional research studies many years later. While there is a
potential for selection bias in the subset of responders, these indi-
viduals otherwise appear healthy and as a group show no evidence
of current AN. Another issue with AN-focused studies is clinical
heterogeneity between patients that cannot realistically be con-
trolled for. Factors that can influence outcomes include length of
illness, severity of malnutrition, age of onset, and time to restore
menses. Other factors such as medical comorbidities, smoking, and
contraceptive use were not scrutinized, with the latter a confound-
ing variable for estradiol measurement. In order to understand the
effects of AN on long-term osteoporosis and fracture risk, a longer
study term with detailed fracture outcomes would be required,
however, the feasibility of such a study remains debatable.
In conclusion, this study confirms persistent effects of AN on bone
health. Despite normalization of body weight, the persisting reduced
bone density seen in these individuals may lead to negative long-
term consequences on bone health, fracture risk and incidence, and
the future onset of osteoporosis. pQCT and DXA suggest that high cor-
tical bone density may be at the heart of the bone phenotype in AN,
although this may be a secondary adaptive change related to cortical
thinning. Further work is required to explore the impact on bone
health and to determine the mechanisms involved. Larger sample
sizes, longer follow-up times, and mixed community-dwelling cohorts
are needed to more thoroughly delineate the long-term effects, to
confirm whether bone morphology can normalize over time, and to
ascertain whether fracture trends are abnormal for this population.
Acknowledgements
This study was performed with the assistance of clinical staff
from the Departments of Adolescent Medicine and Endocrinology
at The Children’s Hospital at Westmead. We would also like to
acknowledge the study participants who donated their time.
Funding Sources
Dr Aaron Schindeler has received research support from
AMGEN, Novartis AG, and N8 Medical for work on bone health
unrelated to this project. Dr Craig Munns has received research
support from Alexion, Novartis AG and is a consultant for Alexion
and Novartis AG for bone research unrelated to this project.
Supplementary data
Supplementary data related to this article can be found at
doi:10.1016/j.jadohealth.2018.07.025.
References
[1] Bonjour JP, Chevalley T, Ferrari S, Rizzoli R. The importance and relevance of
peak bone mass in the prevalence of osteoporosis. Salud Publica Mex 2009;51
(Suppl 1):S5–17.
[2] Bonjour JP, Chevalley T. Pubertal timing, peak bone mass and fragility fracture
risk. BoneKey-Osteovision 2007;4:30–48.
[3] Weaver CM. The role of nutrition on optimizing peak bone mass. Asia Pac J
Clin Nutr 2008;17(Suppl 1):135–7.
[4] Yeo M, Hughes E. Eating disorders - early identification in general practice.
Aust Fam Physician 2011;40(3):108–11.
[5] Misra M, Golden NH, Katzman DK. State of the art systematic review of bone
disease in anorexia nervosa. Int J Eat Disord 2016;49(3):276–92.
[6] Soyka LA, Misra M, Frenchman A, et al. Abnormal bone mineral accrual in ado-
lescent girls with anorexia nervosa. J Clin Endocrinol Metab 2002;87
(9):4177–85.
[7] Misra M, Aggarwal A, Miller KK, et al. Effects of anorexia nervosa on clinical,
hematologic, biochemical, and bone density parameters in community-dwell-
ing adolescent girls. Pediatrics 2004;114(6):1574–83.
[8] Bredella MA, Misra M, Miller KK, et al. Distal radius in adolescent girls with
anorexia nervosa: trabecular structure analysis with high-resolution flat-
panel volume CT. Radiology 2008;249(3):938–46.
[9] Faje AT, Karim L, Taylor A, et al. Adolescent girls with anorexia nervosa have
impaired cortical and trabecular microarchitecture and lower estimated bone
strength at the distal radius. J Clin Endocrinol Metab 2013;98(5):1923–9.
[10] Faje AT, Fazeli PK, Miller KK, et al. Fracture risk and areal bone mineral density in
adolescent females with anorexia nervosa. Int J Eat Disord 2014;47(5):458–66.
[11] American Psychiatric Association. Diagnostic and statistical manual of mental
disorders. 5th ed Arlington, VA: American Psychiatric Publishing.
[12] Haas V, Allen J, Kohn M, et al. Total body protein in healthy adolescent girls:
Validation of estimates derived from simpler measures with neutron activa-
tion analysis. AJCN 2007;85:66–72.
[13] Madden S, Miskovic-Wheatley J, Wallis A, et al. A randomized controlled trial
of in-patient treatment for anorexia nervosa in medically unstable adoles-
cents. Psychol Med 2015;45(2):415–27.
[14] Carter DR, Bouxsein ML, Marcus R. New approaches for interpreting projected
bone densitometry data. J Bone Miner Res 1992;7(2):137–45.
[15] Kuczmarski RJ, Ogden CL, Grummer-Strawn LM, et al. CDC growth charts:
United States. Adv Data 2000;314:1–27.
[16] Lu PW, Briody JN, Ogle GD, et al. Bone mineral density of total body, spine, and
femoral neck in children and young adults: a cross-sectional and longitudinal
study. J Bone Miner Res 1994;9(9):1451–8.
[17] Ho€gler W, Briody J, Woodhead HJ, et al. Importance of lean mass in the inter-
pretation of total body densitometry in children and adolescents. J Pediatr
2003;143(1):81–8.
[18] Moyer-Mileur LJ, Quick JL, Murray MA. Peripheral quantitative computed
tomography of the tibia: pediatric reference values. J Clin Densitom 2008;11
(2):283–94.
[19] Rauch F, Scho € nau E. Peripheral quantitative computed tomography of the dis-
tal radius in young subjects new reference data and interpretation of
results. J Musculosk Neuron Interact 2005;5(2):119–26.
[20] Rauch F, Scho € nau E. Peripheral quantitative computed tomography of the
proximal radius in young subjects new reference data and interpretation of
results. J Musculosk Neuron Interact 2008;8(3):217–26.
[21] Fairburn CG, Beglin SJ. Assessment of eating disorders: Interview or self-
report questionnaire. Int J Eat Disord 1994;16:363–70.
[22] Roux H, Ali A, Lambert S, et al. Predictive factors of dropout from inpatient
treatment for anorexia nervosa. BMC Psychiatry 2016;16(1):339.
[23] Cooper C, Dennison EM, Leufkens HG, et al. Epidemiology of childhood frac-
tures in Britain: a study using the general practice research database. J Bone
Miner Res 2004;19(12):1976–81.
[24] Miller KK, Grinspoon SK, Ciampa J, et al. Medical findings in outpatients with
anorexia nervosa. Arch Intern Med 2005;165(5):561–6.
[25] Divasta AD, Feldman HA, Gordon CM. Vertebral fracture assessment in adoles-
cents and young women with anorexia nervosa: a case series. J Clin Densitom
2014;17(1):207–11.
[26] Lucas AR, Melton LJ 3rd, Crowson CS, O'Fallon WM. Long-term fracture risk
among women with anorexia nervosa: a population-based cohort study.
Mayo Clin Proc 1999;74(10):972–7.
[27] Caouette C, Ikin N, Villemure I, et al. Geometry reconstruction method for
patient-specific finite element models for the assessment of tibia fracture risk
in osteogenesis imperfecta. Med Biol Eng Comput 2017;55(4):549–60.
[28] Misof BM, Roschger P, McMillan HJ, et al. Histomorphometry and bone matrix
mineralization before and after bisphosphonate treatment in boys with duch-
enne muscular dystrophy: a paired transiliac biopsy study. J Bone Miner Res
2016;31(5):1060–9.
[29] Fornari R, Marocco C, Francomano D, et al. Insulin growth factor-1 correlates
with higher bone mineral density and lower inflammation status in obese
adult subjects. Eat Weight Disord 2017 Mar 7. [Epub ahead of print].
[30] Chen LY, Wu YH, Liu LK, et al. Association among serum insulin-like growth
factor-1, frailty, muscle mass, bone mineral density, and physical performance
among community-dwelling middle-aged and older adults in Taiwan. Rejuve-
nation Res 2017 May 9. [Epub ahead of print].