0013-7227/02/$15.00/0
Printed in U.S.A.
The Journal of Clinical Endocrinology & Metabolism 87(7):3162–3168
Copyright © 2002 by The Endocrine Society
Osteopenia in Exercise-Associated Amenorrhea Using
Ballet Dancers as a Model: A Longitudinal Study
MICHELLE P. WARREN, JEANNE BROOKS-GUNN, RICHARD P. FOX, CLAIRE C. HOLDERNESS,
EMILY P. HYLE, AND WILLIAM G. HAMILTON
Departments of Obstetrics and Gynecology, Medicine and Orthopedics, St. Luke’s-Roosevelt Hospital Center and the
Columbia College of Physicians and Surgeons, New York, New York 10032
Few longitudinal studies have investigated the effects of
amenorrhea and amenorrhea plus exercise on bone mineral
density (BMD) of young women. We carried out a 2-yr comparison of dancers and nondancers, both amenorrheic and
normal, that investigated the role of hypothalamic amenorrhea on bone in this context. We studied 111 subjects
(mean age, 22.4 ⴞ 4.6 yr; age of menarche, 14.1 ⴞ 2.2 yr),
including 54 dancers, 22 with hypothalamic amenorrhea,
and 57 nondancers, 22 with hypothalamic amenorrhea. Detailed hormonal and nutritional data were obtained in all
groups to determine possible causal relationship to osteoporosis. The amenorrheic groups, dancers and nondancers,
both showed reduced BMD in the spine, wrist, and foot,
which remained below controls throughout the 2 yr. Only
amenorrheic dancers showed significant changes in spine
BMD (12.1%; P < 0.05) but still remained below controls, and
within this subgroup, only those with delayed menarche
H
YPOTHALAMIC AMENORRHEA IN young women
is associated with reduced bone accretion or premature bone loss during adolescence (1–10), which places
women at high risk for fractures, significant osteopenia,
and severe osteoporosis at menopause. The potential for
reversibility of this problem warrants further investigation, especially due to the current lack of longitudinal
studies.
Research has suggested that exercise, particularly weightbearing exercise, might increase bone accretion in women
with hypothalamic amenorrhea (11) and in young adults and
adolescents (12). However, hypothalamic amenorrhea has
also been increasingly linked to nutritional insults, particularly caloric deprivation, which suggests that the metabolic
axis might be involved with the amenorrhea and osteopenia
in these cases (13–15).
This study investigates the roles of hypothalamic amenorrhea and exercise in the accretion of bone mass during
adolescence and young adulthood, a time frame during
which bone mass is known to increase significantly. We
examined the effects of prolonged amenorrhea on bone mass
in exercising and nonexercising young women. We investigated whether exercise affected the rate of bone accretion and
the functional strength of bones, as measured by the frequency of stress fractures.
Abbreviations: BMD, Bone mineral density; BMI, body mass index;
CV, coefficient(s) of variation; DHEAS, dehydroepiandrosterone sulfate;
EAT, Eating Attitudes Test; PRL, prolactin.
showed a significant increase. The seven amenorrheic subjects (three dancers and four nondancers) who resumed
menses during the study showed an increase in spine and
wrist BMD (17%; P < 0.001) without achieving normalization. Delayed menarche was the only variable that predicted stress fractures (P < 0.005), which we used as a measure of bone functional strength. Analysis of dieting and
nutritional patterns showed higher incidence of dieting behavior in this group, as manifested by higher Eating Attitudes Test scores (16.3 ⴞ 2.00 vs. 11.5 ⴞ 1.45; P < 0.05) and
higher fiber intakes (30.7 ⴞ 3.00 vs. 17.5 ⴞ 2.01 g/24 h; P <
0.001). We concluded that low bone mass occurs in young
women with amenorrhea and delayed menarche, both exercisers and nonexercisers. Crucial bone mass accretion
may be compromised by their reproductive and nutritional
health. (J Clin Endocrinol Metab 87: 3162–3168, 2002)
Materials and Methods
A group of 111 subjects was followed for 2 yr without intervention.
This study compared an exercising group to a group of controls and was
designed to examine the effects of prolonged hypoestrogenism with and
without exercise. Ballet dancers were chosen for the exercising group
because they begin to train before adolescence and are well known to
have both delayed menarche and a high incidence of amenorrhea during
adolescence (7).
Subjects
We followed 111 subjects. The mean age of this group was 22.4 ⫾
4.6 yr, and the mean age of menarche was 14.1 ⫾ 2.2 yr. The exercising
group consisted of 54 ballet dancers, 22 of whom were amenorrheic.
A control group consisted of 57 nonexercising subjects, 22 of whom
had nonexercise-associated hypothalamic amenorrhea. Of these 44
exercising and nonexercising amenorrheics, 26 (59.1%) had delayed
menarche. Volunteers were solicited from national and regional
schools and dance companies, advertisements in college publications,
and physicians’ referrals for hormonal problems and interest in bone
density measurements. These subjects were part of a cross-sectional
study previously reported (5), and 30 subjects were part of a previous
report (1). All subjects were white and from middle to upper class
families. Information was obtained by interview, questionnaire, and
medical examination. Informed consent was obtained by physicians
or nurse practitioners. The procedures followed were approved by
the Institutional Review Board at the St. Luke’s-Roosevelt Hospital
Center. Secondary amenorrhea was defined as amenorrhea of 5 or
more months immediately preceding the study and delayed menarche age as 14 or older. Subjects taking hormones or oral contraceptives for 6 months before the study were eliminated. All subjects
were seen yearly, and all measures were repeated. The educational
levels of the subjects were determined by the seven-point scale developed by Hollingshead and Redlich (16).
3162
Warren et al. • Osteopenia in Exercise-Associated Amenorrhea
Medical evaluation
A nurse practitioner took a menstrual and hormonal history, including weight fluctuations, dieting behavior, hours spent dancing per week,
and a history of past and present illness. Women with chronic illness
were eliminated from the study. Weight and height measurements and
a brief physical examination were performed by a physician or a nurse
practitioner. Blood for estradiol levels was drawn at random in amenorrheic subjects and in the early follicular phase (d 3–7) in menstruating
subjects. Amenorrheic subjects also had FSH, LH, prolactin (PRL), dehydroepiandrosterone sulfate (DHEAS), and testosterone levels measured. All stress fractures were diagnosed by an orthopedic surgeon on
the basis of symptoms and were confirmed by an x-ray or bone scan.
Nutritional evaluation
Food intake was determined using a 2-d dietary history as well as the
Walter Willet semiquantitative food frequency questionnaire (17). The
food frequency measure is designed to target frequently consumed food
items that contain relatively high values for those nutrient groups
known to either contain or influence calcium intake or its absorption, i.e.
calcium, vitamin D, fiber, and caffeine (18). The 24-h recall food intake
diaries were based on a modified version of that described by Frank et
al. (19). The food intake records and food frequency records were coded
separately, using the established coding protocol and the Nutri-calc
software package (CADME Corp., Rochester, NY).
Activity level
Activity level was determined on the basis of the number of calories
expended per day according to the method of Bouchard et al. (20).
Subjects were asked to fill out a 3-d questionnaire over 2 weekdays and
1 weekend day. Each day of the record was divided into 96 periods of
15 min each. The participant selected a code number from nine categories of listed activities that best represented her level of activity during
each period. Activities were grouped according to similar energy costs
in kilocalories per kilogram per minute, and a median value for each
category was used to compute energy expenditure. Mean kilocalorie
expenditure for 3 d is highly reliable, with a coefficient of variation of
0.96 (P ⬍ 0.01).
Psychometrics and eating disorders
Eating problems were assessed through questionnaires and subject
interviews, including Garner and Garfinkel’s Eating Attitudes Test
(EAT)-26, an abbreviated version of their EAT-40 (5). During a semistructured interview, each subject was also asked to indicate how typical
for her was each of the thoughts or behaviors required by Diagnostic and
Statistical Manual III for a diagnosis of anorexia, bulimia, and atypical
eating disorder. Results were also rescored according to Diagnostic and
Statistical Manual III-R criteria (5).
Measurements of bone density
Bone density was measured at the spine, wrist, and foot. Studies have
shown that spinal BMD is particularly affected in amenorrheic athletes
(21) because high-impact activity appears to have a beneficial effect on
BMD primarily at the hip (22). Bone mass analyses of the spine and
radius were carried out using a DP3 Dual Photon Spine/Femur Scanner
and a SP2 Single Photon Scanner (Lunar Corp., Madison, WI) (5). The
bone density measurement of the foot, specifically the first metatarsal,
was done with the Dual Photon Absorptiometer with a method described previously (5). Dual energy x-ray absorptiometers became available in the middle of the study but were not used for the sake of
longitudinal consistency. The decision to continue with the same equipment was advised by our osteoporosis consultant.
The precision of BMD testing was verified in 16 healthy, normal
women. BMD was measured 10 times with repositioning. Coefficients
of variation (CV) were 1.9%, 1.6%, and 4.0% for spine, wrist, and metatarsal, respectively. To correct for source decay and source change,
calibrations were made daily using a standard that consisted of a block
of tissue-equivalent material with three bone-simulating chambers
(small, medium, large) of known bone mineral content. Mean measurements, sd, and CV were made during the period of study. Percentage
J Clin Endocrinol Metab, July 2002, 87(7):3162–3168 3163
CV for photon standard measurements on DP3 was less than 1%, and
SP2 was less than 2% over the 2 yr for the three chambers. These values
remained constant before and after a source change with a percentage
CV of less than 2% on the DP3 and less than 2% on the SP2. The same
machine was used during the entire study. Subjects were seen at baseline
and yearly for 2 yr, for a total of three visits. All measurements, including
bone density, were done at each visit.
Statistical analysis
We used ANOVA to evaluate differences within and between four
groups to look at the following variables: amenorrhea dancer, normal
cycles dancer, amenorrhea control (nondancer), and normal cycles control (nondancer). We examined these groups across the years of study
while controlling for possible differences between individual subjects.
Data were analyzed using one-way ANOVA at each point in time for the
four groups for all valid data points. We also compared controls (both
dancers and nondancers) to amenorrheic groups (both dancer and nondancer) using independent t tests. For the longitudinal analyses, repeated measures ANOVA were used for the data for subjects with valid
data at all three visits. In addition, percentage increase in BMD at 12 and
24 months was measured by subtracting BMD at baseline from the value
at the later time point and dividing by the baseline value. Multiple
comparisons were made using the Bonferroni method, available in SPSS
software (SPSS, Inc., Chicago, IL), and correlations between variables
and bone mass using Pearson linear correlation. Multiple regressions
were performed to determine the change in BMD from baseline to 24
months controlling for the following variables: changes in weight,
height, caloric intake, hormonal levels, body mass index (BMI), percentage body fat, and kilocalories expended per 24 h. ANOVA was used
to compare differences in the spine, wrist, and foot bone mineral density
(BMD); weight; caloric intake; and other variables in the four groups. We
also examined bone density changes by comparing the slopes of the
regression lines after examining the correlation between BMD and time.
Results
One hundred eleven subjects entered, and 62 completed
the study (22 normal controls, 21 normal dancers, 9 amenorrheic controls, and 10 amenorrheic dancers); 50 subjects
who finished the study had all three bone density measurements at the three time intervals; 7 amenorrheic subjects
resumed menses during the study (3 amenorrheic dancers
and 4 amenorrheic controls). The age of menarche for amenorrheic subjects was significantly older than controls
(14.7 ⫾ 2.3 vs. 13.7 ⫾ 1.9 yr; P ⬍ 0.05) (Table 1). Amenorrheic
dancers were younger than normal controls, and normal
controls were heavier and nearer their normal weight (Table
1). Behavior and nutrition patterns are shown in Table 2.
Significantly higher EAT scores for dancers and the amenorrheic nondancers indicate that dieting behavior was more
frequent among these groups, although this was significant
only for the latter. These groups also showed a higher incidence of anorexia in the past (significant for control amenorrheics and normal dancers) and higher fiber intake (significant in both exercising and nonexercising amenorrheic
groups). All amenorrheic subjects had normal PRL, testosterone, and DHEAS levels, normal to low LH and FSH levels,
and low estradiol levels so that they fit the hypothalamic
amenorrhea profile (Table 3).
BMD was obtained at yearly intervals (yr 1, 385.9 ⫾ 68.1 d;
yr 2, 386.0 ⫾ 60.5 d). Figure 1 reports data analyzed using
ANOVA at each point in time for all valid data points. BMD
was significantly lower at baseline in both amenorrheic
groups, and BMD remained lower in the spine during the 2
yr of observation (Fig. 1), independent of weight or age and
3164
J Clin Endocrinol Metab, July 2002, 87(7):3162–3168
Warren et al. • Osteopenia in Exercise-Associated Amenorrhea
TABLE 1. Subjects who entered the study
Dancers
No. (total ⫽ 111)
Age (yr)a
Age at menarche (yr)b
Height (cm)
Weight (kg)c
% ideal weightd
% fatd
Controls
Normal
Amenorrheic
Normal
Amenorrheic
32
22.0 ⫾ 4.7
14.3 ⫾ 2.3
164.2 ⫾ 7.2
53.4 ⫾ 6.8
⫺4.4 ⫾ 8.6
23.3 ⫾ 3.4
22
19.2 ⫾ 3.4
15.0 ⫾ 1.5
162.2 ⫾ 6.0
46.6 ⫾ 5.5
⫺14.4 ⫾ 7.6
20.9 ⫾ 3.9
35
23.0 ⫾ 4.8
13.3 ⫾ 1.5
164.7 ⫾ 4.9
57.4 ⫾ 4.9
2.2 ⫾ 8.5
27.0 ⫾ 4.7
22
23.0 ⫾ 4.1
14.6 ⫾ 2.9
162.4 ⫾ 5.6
50.0 ⫾ 5.7
⫺9.0 ⫾ 8.1
21.3 ⫾ 4.8
Data are mean ⫾ SD.
a
Controls vs. amenorrheic dancers; P ⬍ 0.05.
b
Normal controls vs. amenorrheic dancers; P ⬍ 0.05.
c
Normal controls vs. all other groups; P ⬍ 0.05.
d
Normal controls vs. all other groups; P ⬍ 0.01.
TABLE 2. Nutrition of subjects at entry into study
Dancers
EAT-26a
No. of anorexicsb
No. of bulimics
Calories expended (kcal/24 h)c
Calories intake (kcal/24 h)
Fat intake (g/24 h)
Fiber intake (g/24 h)d
Controls
Normal
(n ⫽ 32)
Amenorrheic
(n ⫽ 22)
Normal
(n ⫽ 35)
Amenorrheic
(n ⫽ 22)
13.0 ⫾ 8.7
9
7
2939.0 ⫾ 590.0
1620.2 ⫾ 597.9
55.2 ⫾ 31.5
22.4 ⫾ 18.3
13.1 ⫾ 8.5
6
7
2674.9 ⫾ 614.4
1765.1 ⫾ 642.9
56.8 ⫾ 49.8
27.4 ⫾ 15.0
10.4 ⫾ 9.8
0
8
2577.2 ⫾ 376.1
1737.0 ⫾ 485.8
64.4 ⫾ 29.5
14.4 ⫾ 9.3
21.6 ⫾ 13.9
9
8
2401.9 ⫾ 430.3
1880.0 ⫾ 700.8
54.1 ⫾ 29.2
26.0 ⫾ 12.9
Data are mean ⫾ SD.
a
Amenorrheic controls vs. normal groups; P ⬍ 0.05.
b
Normal controls vs. amenorrheic controls and normal dancers; P ⬍ 0.001.
c
Controls vs. normal dancers; P ⬍ 0.05.
d
Normal controls vs. all amenorrheics; P ⬍ 0.05.
despite the return or onset of menses in seven subjects. The
normal exercising group (dancers) showed significantly
higher bone density in the weight-dependent sites (i.e. spine
and foot) at baseline when compared with amenorrheics of
both groups (Fig. 1); this effect persisted over the 2 yr. These
findings were weight-dependent but were not affected by
controlling for age. The wrist BMD was significantly higher
only at baseline in the normal group, and a significant rise
in wrist BMD was noted in the amenorrheic dancers over the
2 yr. None of the groups studied showed a significant decline
in BMD.
Only the amenorrheic dancers showed a significant rise in
spine BMD (12.1%; Fig. 2), and they were significantly
younger and had an older age of menarche than the other
groups. The increase in spine BMD was related to amenorrhea, delayed menarche, and exercise (kilocalories expended
per 24 h; P ⬍ 0.005). When controlling for age and weight,
a significant increase in BMD occurred only in the amenorrheic dancers (the exercising group) with delayed menarche.
Surprisingly, dancers and nondancers did not differ in
measures of kilocalories expended. However, when broken
down into aerobic and nonaerobic activity, dancers engaged
in significantly more high-intensity aerobic activity, whereas
controls were more active in such low-intensity aerobic activities as shopping, waitressing, etc. so that overall activity
did not differ. Dancers showed more high-intensity aerobic
activity (754.8 vs. 240.6 kcal/24 h; P ⬍ 0.0001), less lowintensity nonaerobic activity (1391.6 vs. 1699.5 kcal/24 h; P ⬍
0.008), and less sleep (286.6 vs. 351.5 kcal/24 h; P ⬍ 0.01).
Another interesting issue is the discrepancy between energy input and output in all of the groups. As several investigators have reported, groups who exercise (particularly
amenorrheic groups) may develop energy efficiency (23–25).
However, an underestimation of energy intake may also
occur, a problem that has been noted in nutritional surveys.
Despite the increase in dancers’ spine BMD, the BMD of
all amenorrheics (dancer and nondancer) remained below
normal. The seven subjects who resumed menses had a significant increase in BMD (17% in wrist and spine; P ⬍ 0.001)
but did not achieve normalization (Table 4). Multiple regression correlating the change in BMD from baseline to 24
months and controlling for the changes in weight, height,
caloric intake, hormonal levels, BMI, percentage body fat,
and kilocalories expended per 24 showed only an effect of
kilocalories expended at 24 months on spine BMD. Because
the significant changes occurred only in amenorrheic dancers, this was not surprising.
Hormonal values throughout the study are shown in Table
3. Overall, estradiol levels were higher in the normal groups
(dancer and nondancer) vs. amenorrheic groups (dancer and
nondancer) throughout the study, and testosterone and PRL
levels were higher in normal controls at baseline (testosterone significant vs. amenorrheic dancers only). The amenorrheic dancers were the only group to show a significant
increase in estradiol and testosterone levels over time.
A total of 108 stress fractures were diagnosed, with 25
occurring at baseline and 83 during the 2 yr of the study.
Delayed menarche was the only variable that predicted stress
Warren et al. • Osteopenia in Exercise-Associated Amenorrhea
TABLE 3. Hormones and calcium (mean ⫾
FSH (IU/liter)
Baseline
12 months
24 months
LH (IU/liter)a
Baseline
12 months
24 months
Estradiol (pmol/liter)b
Baseline
12 months
24 months
Testosterone (nmol/liter)a
Baseline
12 months
24 months
DHEAS (mmol/liter)
Baseline
12 months
24 months
PRL (ng/mliter)c
Baseline
12 months
24 months
Calcium (mg/24 h)
Baseline
12 months
24 months
SD)
J Clin Endocrinol Metab, July 2002, 87(7):3162–3168 3165
of subjects during study
Normal dancers
Amenorrheic dancers
Normal controls
Amenorrheic controls
6.5 ⫾ 2.1
6.3 ⫾ 3.0
5.0 ⫾ 3.4
8.0 ⫾ 2.1
8.3 ⫾ 4.0
5.8 ⫾ 3.7
6.6 ⫾ 2.4
6.9 ⫾ 3.3
5.2 ⫾ 2.1
1.27 ⫾ 0.03
1.31 ⫾ 0.02
4.08 ⫾ 1.14
5.0 ⫾ 4.6
6.2 ⫾ 12.6
3.3 ⫾ 2.2
3.1 ⫾ 2.2
5.3 ⫾ 3.6
3.7 ⫾ 3.2
6.8 ⫾ 6.1
7.0 ⫾ 12.6
4.8 ⫾ 3.3
3.4 ⫾ 3.3
2.4 ⫾ 2.3
3.9 ⫾ 5.3
238.5 ⫾ 194.7
390.3 ⫾ 433.6
316.3 ⫾ 158.8
90.3 ⫾ 83.3
181.6 ⫾ 146.6
128.6 ⫾ 86.5
243.8 ⫾ 165.9
275.6 ⫾ 264.3
288.8 ⫾ 169.2
112.4 ⫾ 133.4
165.4 ⫾ 143.4
105.7 ⫾ 95.8
1.20 ⫾ 0.74
.78 ⫾ 0.39
1.50 ⫾ 0.69
0.83 ⫾ 0.48
0.85 ⫾ 0.42
1.39 ⫾ 0.65
1.64 ⫾ 0.80
1.16 ⫾ 0.87
1.50 ⫾ 0.64
1.43 ⫾ 0.94
.86 ⫾ 0.50
1.40 ⫾ 0.87
5.7 ⫾ 3.5
5.5 ⫾ 3.2
5.4 ⫾ 2.8
4.7 ⫾ 2.8
5.6 ⫾ 2.7
6.6 ⫾ 5.1
5.7 ⫾ 2.6
6.7 ⫾ 3.3
5.6 ⫾ 3.0
5.7 ⫾ 3.9
6.6 ⫾ 3.7
6.6 ⫾ 4.2
7.1 ⫾ 4.5
5.4 ⫾ 3.4
7.4 ⫾ 4.3
5.4 ⫾ 3.1
3.1 ⫾ 1.5
5.1 ⫾ 2.8
10.7 ⫾ 7.2
7.8 ⫾ 6.3
8.3 ⫾ 5.7
6.1 ⫾ 3.1
4.5 ⫾ 2.5
6.8 ⫾ 5.3
649.5 ⫾ 365.2
735.5 ⫾ 303.2
674.4 ⫾ 342.0
856.1 ⫾ 439.8
742.9 ⫾ 274.0
762.7 ⫾ 357.7
844.5 ⫾ 661.1
642.1 ⫾ 348.2
623.5 ⫾ 295.0
779.4 ⫾ 409.3
865.4 ⫾ 620.9
920.4 ⫾ 595.5
P ⬍ 0.05, normal controls vs. amenorrheic dancers at baseline.
P ⬍ 0.05, normals vs. amenorrheics at baseline, 12 months, and 24 months.
c
P ⬍ 0.05, normal controls vs. all other groups at baseline, and normal controls vs. amenorrheic dancers at 12 months.
a
b
fractures. An increased frequency of stress fractures (P ⬍
0.05) was associated with lower BMDs in the spine. Subjects
with stress fractures showed an older age of menarche
(15.2 ⫾ 2.3 vs. 13.5 ⫾ 1.5 yr; P ⫽ 0.002), more dieting behavior
(EAT scores, 16.3 ⫾ 2.00 vs. 11.5 ⫾ 1.43; P ⬍ 0.05), and higher
amounts of fiber in the diets (30.7 ⫾ 3.3 vs. 17.5 ⫾ 2.01 g; P ⬍
0.001). There was no difference in calcium or fat intake in
these groups.
The dropout rate in this study was 45%. Our analyses
showed no difference in age, BMD, weight, change in BMD,
level of exercise, fractures, or menstrual status for the dropouts vs. compliant subjects. A review of the subjects found
that 72% dropped out because of geographic relocation, 8%
due to lack of time, 15% for undetermined reasons, and 5%
due to lack of interest.
To further eliminate the effects of noncompliant subjects
on the longitudinal observations, we examined the subjects
(n ⫽ 50) who had all BMD measurements at 0, 12, and 24
months. These data are presented in Table 5. When examining this compliant group, the amenorrheic group (both
dancer and nondancer) showed slightly lower spine BMD at
baseline when compared with normals (both dancer and
nondancer). Findings were similar to previous analyses with
the following exceptions: an increase in spine BMD was
significant only from 12–24 months for amenorrheic dancers
and not in the first year. This group also showed a significant
decline in calcium intake in the first year (833.3 ⫾ 337.9 vs.
679.5 ⫾ 173.9 mg/24 h; P ⬍ 0.01) and significant increases in
PRL (5.6 ⫾ 1.2 vs. 6.8 ⫾ 1.3 ng/ml; P ⬍ 0.01) as well as
estradiol (111.3 ⫾ 120.5 vs. 184.8 ⫾ 125.6 pmol/liter; P ⬍ 0.05)
during this year. Again, multiple regression demonstrated
an effect of kilocalories expended per 24 h at 24 months,
which would be expected because the dancers were the only
group to show an increase in spine BMD. There was no effect
of caloric or calcium intake, hormonal levels, change in
weight, height, BMI, or percentage body fat.
Discussion
This is the first longitudinal study of this magnitude using
a homogenous group of exercising subjects from one athletic
discipline. Our results show that women with hypothalamic
amenorrhea also have lower bone density that remains below
normal menstruating counterparts over 2 yr. These results
were independent of weight and age and occurred despite
the continuous exercise in one group. The findings are tempered by the large dropout rate but are consistent whether
compliant or noncompliant subjects were included.
Only one group, the amenorrheic dancers, showed a significant change in spinal BMD, and within this group it was
only those with delayed menarche who showed an increase
in spinal BMD. This subgroup with delayed menarche also
showed the most stress fractures and the most distorted
pattern of eating behavior. Despite the increase, their BMD
remained below that of their normal counterparts.
All amenorrheic subjects were deficient in BMD at baseline, which suggests that the bone mass deficiency was
present before the start of this study. The loss of bone or lack
of accretion, therefore, may have occurred at a crucial time
not fully addressed by this study. An increase in spinal BMD
3166
J Clin Endocrinol Metab, July 2002, 87(7):3162–3168
Warren et al. • Osteopenia in Exercise-Associated Amenorrhea
FIG. 2. Percentage increase in spine BMD over a 2-yr period ⫾ SE.
Amenorrheic dancers, P ⬍ 0.05. F, Amenorrheic dancers; ‚, normal
dancers; f, amenorrheic controls; 〫, normal controls.
TABLE 4. Bone density (g/cm2 ⫾
end of the study
Amenorrheic
(n ⫽ 12)
Spine
Baseline
24 months
% change
Wrist
Baseline
24 months
% change
Foot
Baseline
24 months
% change
FIG. 1. BMD over a 2-yr period ⫾ SE. Top, Spine BMD, P ⬍ 0.001,
normals vs. amenorrheics; middle, wrist BMD, P ⬍ 0.05 at baseline
only, amenorrheic dancers vs. other groups; bottom, foot BMD, P ⬍
0.05, normal dancers vs. other groups. Amenorrheic dancers showed
a significant increase in density of wrist and spine, P ⬍ 0.05. (Number
of subjects measured in yr 1, 2, and 3 are in parentheses.) Results were
similar using repeated measures, including only those subjects who
completed all three time points. F, Amenorrheic dancers (n ⫽ 22, 17,
and 10 in yr 1, 2, and 3, respectively); ‚, normal dancers (n ⫽ 32, 22,
and 21 in yr 1, 2, and 3, respectively); f, amenorrheic controls (n ⫽
22, 14, and 9 in yr 1, 2, and 3, respectively); 〫, normal controls (n ⫽
35, 26, and 22 in yr 1, 2, and 3, respectively).
was significant when controlled for age and weight only in
the amenorrheic dancers. Thus, exercise-associated amenorrhea (seen in our dancers) is associated with delayed sexual
development and possibly delayed bone accretion; this increase in BMD may represent a catch-up phase. This is also
suggested by the significant increase in estradiol in the compliant subgroup.
In the study, seven women (three dancers and four controls) resumed menses and showed a significant increase in
their spine and wrist BMD (both 17%; Fig. 3) without normalization. Past studies on athletes (6%; Refs. 1 and 26) and
subjects with anorexia nervosa (0 –19%; Refs. 10, 27, and 28)
have also associated returned menstrual cyclicity with increased bone mass. Significantly, none of the women who
resumed menses achieved normalization of BMD. The lack
a
SE)
studies of subjects at the
Resumption of
menses (n ⫽ 7)
Normal
(n ⫽ 36)
1.16 ⫾ 0.04
1.20 ⫾ 0.04
4.05 ⫾ 1.95
0.99 ⫾ 0.05
1.16 ⫾ 0.06
17.47 ⫾ 4.93a
1.26 ⫾ 0.03
1.31 ⫾ 0.02
4.01 ⫾ 1.19
0.60 ⫾ 0.02
0.62 ⫾ 0.03
3.09 ⫾ 2.37
0.59 ⫾ 0.05
0.67 ⫾ 0.03
17.54 ⫾ 8.75a
0.65 ⫾ 0.01
0.67 ⫾ 0.01
2.85 ⫾ 1.44
0.84 ⫾ 0.03
0.82 ⫾ 0.02
⫺1.85 ⫾ 2.88
0.76 ⫾ 0.03
0.81 ⫾ 0.03
7.05 ⫾ 4.79
0.86 ⫾ 0.02
0.85 ⫾ 0.01
⫺0.74 ⫾ 1.69
P ⬍ 0.001.
of normalization suggests that reduced bone accretion might
result in a permanent failure to achieve peak bone mass,
which underscores the importance of intervention, either
preventative or therapeutic.
In addition to reduced BMD, subjects with delayed menarche had an increased frequency of stress fractures when
compared with subjects with normal menarche. Such findings suggest that the delay in sexual development may adversely affect their bone quality and functional strength. Our
data suggest that delayed reproductive development may
attenuate the well known benefits of weight-bearing exercise
on bone mass accretion during adolescence and young adulthood (6). High EAT scores and high-fiber diets in the group
with stress fractures suggest that this group may be striving
to maintain low weight, and nutritional factors may also
influence bone quality. This observation is supported by
research documenting that dancers who developed stress
fractures have significantly abnormal nutritional patterns,
irrespective of reproductive function (29). The findings in
this study are tempered by the large dropout rate of 42%.
Although our analyses did not show differences between the
two groups, it is possible that our results may have been
influenced by variables affecting noncompliance. High drop-
Warren et al. • Osteopenia in Exercise-Associated Amenorrhea
J Clin Endocrinol Metab, July 2002, 87(7):3162–3168 3167
TABLE 5. Subjects followed for 24 months with all BMD (g/cm2
⫾ SE) measurements
Normal
dancers
(n ⫽ 17)
Spinea
Baseline
12 months
24 months
Wrist
Baseline
12 months
24 months
Footd
Baseline
12 months
24 months
Amenorrheic
dancers
(n ⫽ 10)
Normal
controls
(n ⫽ 16)
Amenorrheic
controls
(n ⫽ 7)
1.28 ⫾ 0.05 1.08 ⫾ 0.03 1.28 ⫾ 0.03 1.11 ⫾ 0.09
1.34 ⫾ 0.04 1.16 ⫾ 0.04 1.29 ⫾ 0.02 1.13 ⫾ 0.08
1.31 ⫾ 0.04 1.21 ⫾ 0.03b 1.32 ⫾ 0.03 1.14 ⫾ 0.07
0.65 ⫾ 0.02 0.58 ⫾ 0.03c
0.64 ⫾ 0.02 0.62 ⫾ 0.02
0.66 ⫾ 0.01 0.63 ⫾ 0.02
0.66 ⫾ 0.02 0.62 ⫾ 0.03
0.68 ⫾ 0.01 0.65 ⫾ 0.03
0.67 ⫾ 0.01 0.64 ⫾ 0.04
0.89 ⫾ 0.02 0.79 ⫾ 0.03
0.87 ⫾ 0.02 0.79 ⫾ 0.03
0.89 ⫾ 0.02 0.82 ⫾ 0.02
0.83 ⫾ 0.03 0.86 ⫾ 0.05
0.79 ⫾ 0.02 0.80 ⫾ 0.03
0.81 ⫾ 0.02 0.81 ⫾ 0.03
b
Significant increase in amenorrheic dancers 12–24 months only;
P ⬍ 0.05.
a
Spine BMD P ⬍ 0.01, normal vs. amenorrheic groups.
c
Wrist BMD P ⬍ 0.05, baseline only for amenorrheic dancers.
d
Foot BMD P ⬍ 0.05, normal dancers vs. other groups at all times.
out rates are not uncommon in young populations and are
consistent with other studies (30).
Insufficient caloric intake for the level of activity has been
proposed as a factor in the genesis of exercise-associated
reproductive dysfunction (24, 31–33), and the osteopenia
may also be an adaptive response to chronic low energy
intake. Metabolic factors in response to nutritional insults
might mediate reproductive and bone growth adaptations.
Our observations are consistent with the findings of others
(34) and suggest that poor nutrition or an energy deficit with
an adaptation to large caloric needs is fundamentally linked
with the prolonged amenorrheic state and the osteopenia
(34 –36).
Other recent data also suggest that nutritionally regulated
processes may underlie the osteopenia. Hypercortisolemia
(37–39), IGF-I deficiency (34), or other metabolic adaptations
seen in these athletes may contribute to the osteopenia. Lowered bone formation markers, T3 and IGF-I, are seen in amenorrheic runners but not in controls (34, 35). Because our
original studies on ballet dancers showed normal vitamin D
and parathyroid levels, we suggest that these are not the
causal factors in the osteopenia (7). Two nutritionally dependent hormones, IGF-I and T3, predicted change in trabecular bone mass with estrogen-progestin therapy in past
studies on anorexia nervosa (28). Low leptin levels have been
associated with amenorrhea (31, 40) and with disordered
eating (31), a behavior that our amenorrheic group demonstrated. Recently, leptin receptors have been found on bone
(41), and it has been suggested that leptin may be involved
in the skeletal regulation and may possibly be centrally controlled, most likely in the hypothalamus (42– 44). This provides a possible mechanism for an interaction between the
nutritional and metabolic bone axes and suggests that leptin
may function as a physiological regulator of bone mass.
Therefore, a mechanism other than hypoestrogenism may
also account for the low bone density and the alarming stress
fracture rate seen in women whose amenorrhea is associated
with caloric deficiency, nutritional insults, and exercise (29,
31, 45, 46).
FIG. 3. Onset or return of menses. Spine BMD and percentage increase ⫾ SE in spine over a 2-yr period in subjects with and without
return or onset of menses. Spine BMD, P ⬍ 0.05 at baseline and 24
months; percentage change in spine is significant at 24 months for
those with onset or return of menses, P ⬍ 0.001. *, Resumption or
onset of menses; f, amenorrheic; 〫, normal.
Another surprising finding is the lack of further bone loss
in the amenorrheic subjects. This may be due to factors such
as the low bone turnover reported in this state (35, 36, 47)
which in turn may be induced by a central mechanism protecting against further loss and perhaps involving a metabolic signal such as leptin.
Recent studies suggest that accretion of peak bone mass
during adolescence and young adulthood is essential because at least 40% of bone mass is formed at this time, and
it may be difficult for a woman to accrue additional bone
mass later in life (13–15, 48). Maximizing total bone mass
accretion for young women at risk cannot be implemented
without further understanding of the relationship between
hypothalamic amenorrhea and bone density. More studies
on this unique form of bone loss are needed to improve the
efficacy of prevention and treatment of osteopenia.
Acknowledgments
Received October 6, 2000. Accepted March 21, 2002.
Address all correspondence and requests for reprints to: Michelle P.
Warren, M.D., Department of Obstetrics and Gynecology, PH 16-127,
Columbia University, 622 West 168th Street, New York, New York
10032. E-mail: [email protected].
This research was supported by NIH/National Institute of Child
Health and Human Development Grants R01-HD22171-01 through
R01-HD22171-07.
3168
J Clin Endocrinol Metab, July 2002, 87(7):3162–3168
References
1. Jonnavithula S, Warren MP, Fox RP, Lazaro MI 1993 Bone density is compromised in amenorrheic women despite return of menses: a 2-year study.
Obstet Gynecol 81:669 – 674
2. Drinkwater BL, Nilson K, Chesnut III CH, Bremner WJ, Shainholtz S, Southworth MB 1984 Bone mineral content of amenorrheic and eumenorrheic athletes. N Engl J Med 311:277–281
3. Lindberg JS, Fears WB, Hunt MM, Powell MR, Boll D, Wade CE 1984
Exercise-induced amenorrhea and bone density. Ann Intern Med 101:647– 648
4. Marcus R, Cann C, Madvig P, Minkoff J, Goddard M, Bayer M, Martin M,
Gaudiani L, Haskell W, Genant H 1985 Menstrual function and bone mass in
elite women distance runners. Ann Intern Med 102:158 –163
5. Dhuper S, Warren MP, Brooks-Gunn J, Fox RP 1990 Effects of hormonal status
on bone density in adolescent girls. J Clin Endocrinol Metab 71:1083–1088
6. Warren MP, Brooks-Gunn J, Fox RP, Lancelot C, Newman D, Hamilton WG
1991 Lack of bone accretion and amenorrhea: evidence for a relative osteopenia
in weight bearing bones. J Clin Endocrinol Metab 72:847– 853
7. Warren MP, Brooks-Gunn J, Hamilton LH, Warren LF, Hamilton WG 1986
Scoliosis and fractures in young ballet dancers: relation to delayed menarche
and secondary amenorrhea. N Engl J Med 314:1348 –1353
8. Rigotti NA, Nussbaum SR, Herzog DB, Neer RM 1984 Osteoporosis in
women with anorexia nervosa. N Engl J Med 311:1601–1606
9. Biller BMK, Saxe V, Herzog DB, Rosenthal DI, Holzman S, Klibanski A 1989
Mechanisms of osteoporosis in adult and adolescent women with anorexia
nervosa. J Clin Endocrinol Metab 68:548 –554
10. Rigotti NA, Neer RM, Skates SJ, Herzog DB, Nussbaum SR 1991 The clinical
course of osteoporosis in anorexia nervosa: a longitudinal study of cortical
bone mass. JAMA 265:1133–1138
11. Rubin K, Schirduan V, Gendreau P, Sarfarazi M, Mendola R, Dalsky G 1993
Predictors of axial and peripheral bone mineral density in healthy children and
adolescents, with special attention to the role of puberty. J Pediatr 123:863– 870
12. Barr SI, McKay HA 1998 Nutrition, exercise, and bone status in youth. Int
J Sports Med 8:124 –142
13. Ott SM 1990 Editorial: attainment of peak bone mass. J Clin Endocrinol Metab
71:1082A–1082C
14. Riggs BL, Eastell R 1986 Exercise, hypogonadism and osteopenia. JAMA
256:392–393
15. Kreipe RE, Forbes GB 1990 Osteoporosis: a “new morbidity” for dieting
female adolescents? Pediatrics 86:478 – 480
16. Hollingshead AB, Redlich FC 1958 Social class and mental illness: a community study. New York: John Wiley
17. Willett WC, Sampson L, Stampfer MJ, Rosner B, Bain C, Witschi J, Hennekens CH, Speizer FE 1985 Reproducibility and validity of a semi-quantitative food frequency questionnaire. Am J Epidemiol 122:51– 65
18. Calabrese LH, Kirkendall DT 1983 Nutritional and medical considerations in
dancers. Clin Sports Med 2:539 –548
19. Frank GC, Berenson GS, Schilling PF, Moore MC 1977 Adapting the 24-hr
recall for epidemiologic studies of school children. J Am Diet Assoc 71:26 –31
20. Bouchard C, Tremblay A, Leblanc C, Lortie G, Savard R, Thieriault G 1983
A method to assess energy expenditure in children and adults. Am J Clin Nutr
37:461– 467
21. Rutherford OM 1993 Spine and total body bone mineral density in amenorrheic endurance athletes. J Applied Physiol 74:2904 –2908
22. Khan KM, Warren MP, Stiehl A, McKay HA, Wark JD 1999 Bone mineral
density in active and retired ballet dancers. J Dance Med Science 3:15–23
23. Mulligan K, Butterfield GE 1990 Discrepancies between energy intake and
expenditure in physically active women. Br J Nutr 64:23–36
24. Myerson M, Gutin B, Warren MP, May MT, Contento I, Lee M, Pi-Sunyer
FX, Pierson RN Jr, Brooks-Gunn J 1991 Resting metabolic rate and energy
balance in amenorrheic and eumenorrheic runners. Med Sci Sports Exerc
23:15–22
25. Myerson M, Gutin B, Warren MP, May MT, Contento I, Lee M, Pi-Sunyer
FX, Pierson Jr RN, Brooks-Gunn J 1991 Resting metabolic rate and energy
balance in amenorrheic and eumenorrheic runners. Med Sci Sports Exerc
23:15–22
Warren et al. • Osteopenia in Exercise-Associated Amenorrhea
26. Drinkwater BL, Nilson K, Ott S, Chesnut III CH 1986 Bone mineral density
after resumption of menses in amenorrheic athletes. JAMA 256:380 –382
27. Bachrach LK, Katzman DK, Litt IF, Guido D, Marcus R 1991 Recovery from
osteopenia in adolescent girls with anorexia nervosa. J Clin Endocrinol Metab
72:602– 606
28. Klibanski A, Biller BMK, Schoenfeld DA, Herzog DB, Saxe VC 1995 The
effects of estrogen administration on trabecular bone loss in young women
with anorexia nervosa. J Clin Endocrinol Metab 80:898 –904
29. Frusztajer NT, Dhuper S, Warren MP, Brooks-Gunn J, Fox RP 1990 Nutrition
and the incidence of stress fractures in ballet dancers. Am J Clin Nutr 51:
779 –783
30. Maloney JM, Arent DL, Flack M, McLaughlin-Miley C, Sevilla C, Derman
R 2001 Use of low-dose oral contraceptive containing norethindrone acetate
and ethinyl estradiol in the treatment of moderate acne vulgaris. Clin J Women’s Health 1:123–131
31. Warren MP, Voussoughian F, Geer EB, Hyle EP, Adberg CL, Ramos RH 1999
Functional hypothalamic amenorrhea: hypoleptinemia and disordered eating.
J Clin Endocrinol Metab 84:873– 877
32. Loucks AB, Heath EM 1994 Induction of low-T3 syndrome in exercising
women occurs at a threshold of energy availability. Am J Physiol 266:R817–
R823
33. Loucks AB, Verdun M, Heath EM 1998 Low energy availability, not stress of
exercise, alters LH pulsatility in exercising women. J Appl Physiol 84:37– 46
34. Zanker CL 1999 Bone metabolism in exercise associated amenorrhea: the
importance of nutrition. Br J Sports Med 33:228 –229
35. Zanker CL, Swaine IL 1998 Bone turnover in amenorrheic and eumenorrheic
distance runners. Scand J Med Sci Sports 8:20 –26
36. Zanker CL, Swaine IL 1998 The relationship between bone turnover, oestradiol, and energy balance in women distance runners. Br J Sports Med 32:
167–171
37. Loucks AB, Mortola JF, Girton L, Yen SSC 1989 Alterations in the hypothalamic-pituitary-ovarian and the hypothalamic-pituitary-adrenal axes in athletic women. J Clin Endocrinol Metab 68:402– 411
38. Ding JH, Sheckter CB, Drinkwater BL, Soules MR, Bremner WJ 1988 High
serum cortisol levels in exercise-associated amenorrhea. Ann Intern Med 108:
530 –534
39. Prior JC, Vigna YM, Schechter MT, Burgess AE 1990 Spinal bone loss and
ovulatory disturbances. N Engl J Med 323:1221–1227
40. Miller KK, Parulekar MS, Schoenfeld E, Anderson E, Hubbard J, Klibanski
A, Grinspoon SK 1998 Decreased leptin levels in normal weight women with
hypothalamic amenorrhea: the effects of body composition and nutritional
insults. J Clin Endocrinol Metab 83:2309 –2312
41. Thomas T, Gori F, Burguera B, Spelsberg TC, Jensen MD, Riggs BL, Leptin
acts on human marrow stromal precursor cells to enhance osteoblast differentiation and to inhibit adipocyte differentiation: A potential mechanism for
increased bone mass in obesity. Program of the 80th Annual Meeting of The
Endocrine Society, New Orleans, LA, 1998, p 90 (Abstract OR29-1)
42. Steppan CM, Chidsey-Frink KL, Crawford DT, Ke HZ, Swick AG, Leptin
administration causes bone growth in ob/ob mice. Program of the 80th Annual
Meeting of The Endocrine Society, New Orleans, LA, 1998, p 439 (Abstract
P3-256)
43. Ducy P, Schinke T, Karsenty G 2000 The osteoblast: a sophisticated fibroblast
under central surveillance. Science 289:1501–1504
44. Ducy P, Amling M, Takeda S, Priemel M, Schilling AF, Beil FT, Shen J,
Vinson C, Rueger JM, Karsenty G 2000 Leptin inhibits bone formation
through a hypothalamic relay: a central control of bone mass. Cell 100:197–207
45. Grinspoon S, Miller K, Coyle C, Krempin J, Armstrong C, Pitts S, Herzog
D, Klibanski A 1999 Severity of osteopenia in estrogen-deficient women with
anorexia nervosa and hypothalamic amenorrhea. J Clin Endocrinol Metab
84:2049 –2055
46. Laughlin GA 1999 The role of nutrition in the etiology of functional hypothalamic amenorrhea. Curr Opin Endocrinol Diabet 6:38 – 43
47. Okano H, Mizunuma H, Soda M, Matsui H, Aoki I, Honjo S, Ibuki Y 1995
Effects of exercise and amenorrhea on bone mineral density in teenage runners.
Endocr J 42:271–276
48. Kreipe RE 1992 Bones of today, bones of tomorrow. Am J Dis Child 146:22–25