J C E M
O N L I N E
Hot Topics in Translational Endocrinology—Endocrine Research
Do Changes in Sex Steroid Hormones Precede or
Follow Increases in Body Weight during the
Menopause Transition? Results from The Study of
Women’s Health Across the Nation
Rachel P. Wildman, Ping G. Tepper, Sybil Crawford, Joel S. Finkelstein,
Kim Sutton-Tyrrell, Rebecca C. Thurston, Nanette Santoro, Barbara Sternfeld,
and Gail A. Greendale
Department of Epidemiology and Population Health (R.P.W.), Albert Einstein College of Medicine, Bronx,
New York 10461; Department of Epidemiology (P.G.T., K.S.-T.), University of Pittsburgh Graduate School of
Public Health, Pittsburgh, Pennsylvania 15261; Division of Preventive and Behavioral Medicine (S.C.),
University of Massachusetts, Worcester, Massachusetts 01655; Endocrine Unit (J.S.F.), Department of
Medicine, Massachusetts General Hospital, Boston, Massachusetts 02114; Department of Psychiatry (RCT),
University of Pittsburgh School of Medicine, Pittsburgh, Pennsylvania 15213; Department of Obstetrics and
Gynecology (N.S.), University of Colorado School of Medicine, Aurora, Colorado 80045; Division of Research,
Kaiser Permanente (B.S.), Oakland, California 94612; and Division of Geriatrics (G.A.G.), David Geffen School
of Medicine, University of California, Los Angeles, Los Angeles, California 90095
Context: Whether menopause-related changes in sex steroids account for midlife weight gain in
women or whether weight drives changes in sex steroids remains unanswered.
Objective: The objective of the study was to characterize the potential reciprocal nature of the associations between sex hormones and their binding protein with waist circumference in midlife women.
Design, Setting, and Participants: The study included 1528 women (mean age 46 yr) with 9 yr of
follow-up across the menopause transition from the observational Study of Women’s Health Across
the Nation.
Main Outcome Measures: Waist circumference, SHBG, testosterone, FSH, and estradiol were
measured.
Results: Current waist circumference predicted future SHBG, testosterone, and FSH but not vice versa.
For each SD higher current waist circumference, at the subsequent visit SHBG was lower by 0.04 – 0.15
2
SD, testosterone was higher by 0.08 – 0.13 SD, and log FSH was lower by 0.15– 0.26 SD. Estradiol results
were distinct from those above, changing direction across the menopause transition. Estradiol and
waist circumference were negatively associated in early menopausal transition stages and positively
associated in later transition stages (for each SD higher current waist circumference, future estradiol was
lower by 0.15 SD in pre- and early perimenopause and higher by 0.38 SD in late peri- and postmenopause;
P for interaction ⬍0.001). In addition, they appeared to be reciprocal, with current waist circumference
associated with future estradiol and current estradiol associated with future waist circumference.
However, associations in the direction of current waist circumference predicting future estradiol levels
were of considerably larger magnitude than the reverse.
Conclusions: These Study of Women’s Health Across the Nation data suggest that the predominant
temporal sequence is that weight gain leads to changes in sex steroids rather than vice versa. (J Clin
Endocrinol Metab 97: E1695–E1704, 2012)
ISSN Print 0021-972X ISSN Online 1945-7197
Printed in U.S.A.
Copyright © 2012 by The Endocrine Society
doi: 10.1210/jc.2012-1614 Received March 8, 2012. Accepted June 4, 2012.
First Published Online June 21, 2012
Abbreviations: BMI, Body mass index; E2, estradiol; LLD, lower limit of detection; SEM,
structural equation modeling; SWAN, Study of Women’s Health Across the Nation; T,
testosterone.
J Clin Endocrinol Metab, September 2012, 97(9):E1695–E1704
jcem.endojournals.org
E1695
E1696
Wildman et al.
Sex Steroids and Adiposity in Midlife Women
n women, the prevalence of obesity increases during
midlife. However, whether this increase is due to aging,
the menopause transition, or both remains unclear (1).
Additionally, it is unclear whether weight gain and/or development of abdominal obesity induces changes in reproductive hormones or vice versa.
Weight gain increases the incidence of diabetes, cancer,
and cardiovascular disease. An improved understanding
of the temporal associations between weight gain and
changing hormone levels may lead to new strategies to
curb weight gain during midlife. Therefore, we characterized the sequential nature of the longitudinal associations
between changes in selected hormones and changes in
waist circumference and body weight in a multiethnic cohort of women undergoing the menopause transition.
I
J Clin Endocrinol Metab, September 2012, 97(9):E1695–E1704
Questionnaire data
Menstrual cycle bleeding history over the prior year was assessed by interview and categorized as follows:
• Premenopause: one or more periods within the previous 3
months and no perceived change in cycle interval.
• Early perimenopause: one or more periods within the past 3
months, with a perceived change in cycle interval.
• Late perimenopause: no bleeding in the previous 3 months but
bleeding in the last 11 months.
• Postmenopause: 12 or more consecutive months of
amenorrhea.
Race-ethnicity was self-assigned. Highest grade level or educational degree attained and smoking history were assessed by
questionnaire. Alcohol consumption was assessed by interview
using a modified version of the Block Food Frequency Questionnaire (3).
Hormone measurement
Materials and Methods
Participants
The Study of Women’s Health Across the Nation (SWAN) is
a multicenter, longitudinal study to characterize the biological
and psychosocial changes with menopause in a communitybased sample (2). From 1996 to 1997, 3302 women aged 42–52
yr were enrolled across seven recruitment sites (Boston, MA;
Chicago, IL; Detroit, MI; Los Angeles, CA; Newark, NJ; Pittsburgh, PA; and Oakland, CA). SWAN eligibility criteria included
an intact uterus, at least one menstrual period in the previous 3
months, at least one ovary, not being pregnant or breast-feeding,
and no oral contraceptive or sex steroid hormone therapy use in
the previous 3 months. The protocol was approved by each site’s
institutional review board, and all women provided written informed consent.
This report is based on data from the baseline visit as well as
follow-up visits occurring 3, 6, and 9 yr after the baseline visit.
From the inception cohort of 3302 women, we excluded 253
women who had a hysterectomy or bilateral oophorectomy; 795
who missed at least two of the three follow-up visits; 656 who
reported current hormone use or were in the middle of a 6-month
hormone washout period during any of the three follow-up visits; 19 whose hormone use information was unknown at baseline; two who were pregnant; and women from the Newark site
(n ⫽ 49) due to the fact that Newark did not collect data at visit
9. This left 1528 women for the current analyses.
Physical measures
Height and weight were measured in light clothing without
shoes using calibrated scales. Body mass index was calculated as
weight (kilograms) divided by height (meters) squared. Waist
circumference was measured in nonrestrictive undergarments or
over light clothing, at the narrowest part of the torso as seen from
the anterior aspect. For women in whom a waist narrowing was
difficult to identify, the measure was taken at the smallest horizontal circumference in the area between the ribs and the iliac
crest.
Fasting blood draws were targeted to the follicular phase of
the menstrual cycle (d 2–5) for menstruating women. Two attempts were made to obtain a follicular phase sample. When a
follicular phase sample could not be obtained and in women no
longer menstruating regularly, a random fasting sample was
taken within 90 d of the anniversary date of the baseline visit. For
the purposes of analysis, cycle day of blood draw was dichotomized as within d 2–5 of the menstrual cycle or outside d 2–5/
unknown. Samples were maintained at 4 C until separated and
frozen at ⫺80 C until assays were performed.
Hormone assays were performed at the University of Michigan
Endocrine Laboratory on an ACS-180 automated analyzer (Bayer
Diagnostics Corp., Tarrytown, NY) using a double-antibody
chemiluminescent immunoassay with a solid phase anti-IgG immunoglobulin conjugated to paramagnetic particles, antiligand antibody, and competitive ligand labeled with dimethylacridinium
ester. The FSH assay modified a manual assay kit (Bayer Diagnostics) using two monoclonal antibodies directed to different regions
on the ␤-subunit, with a lower limit of detection (LLD) of 0.4 –1.0
mIU/ml. The estradiol (E2) assay modified the rabbit anti-E2– 6
ACS-180 immunoassay to increase sensitivity, with the visitspecific LLD ranging from 1.0 –7.0 pg/ml. The testosterone (T)
assay modified the rabbit polyclonal anti-T ACS-180 immunoassay, with a LLD of 2.0 –2.2 ng/dl. The SHBG assay was developed on-site using rabbit anti-SHBG antibodies, with a LLD
of 1.9 –3.2 nM. Duplicate E2 assays were conducted with results
reported as the arithmetic mean for each subject, with a coefficient of variation of 3–12%. All other assays were single determinations. Hormone values below the LLD were replaced with
a random value between zero and the LLD. This occurred for at
least one hormone in 3.5% (n ⫽ 53) of women; these women did
not differ from those with values above the LLD in age, body
weight, waist circumference, race-ethnicity, education levels,
menopause status, cycle day of blood draw, or the prevalence of
smoking or alcohol use.
Statistical analysis
Distributions of variables of interest were summarized using
means and SD for normally distributed variables, medians, and
interquartile ranges for skewed variables and frequency counts
for categorical variables.
J Clin Endocrinol Metab, September 2012, 97(9):E1695–E1704
The dynamic sequential relationships between hormones and
waist circumference were examined using structural equation
modeling (SEM). Specifically these equations examined the following three longitudinal relationships: 1) the association of current hormone values and waist circumferences with their respective future values, 2) the associations of current waist
circumference with future hormone values, and 3) the associations of current hormone values with future waist circumference
values. Additionally, concurrent correlations of hormone levels
and waist circumference as well as ethnicity, education, alcohol
use, smoking status, height (all measured at baseline) and age,
cycle day of hormone blood draw, and menopausal status (all
time varying) were adjusted for. Waist circumference associations were additionally adjusted for body weight (time varying).
Because findings for waist circumference and body weight were
similar, only waist circumference results are presented here.
Identical analyses for body weight are available in an appendix
(Supplemental Figs. 1– 4, published on The Endocrine Society’s
Journals Online web site at http://jcem.endojournals.org).
To test whether the sequential relationships between hormones and waist circumference varied by menopausal status, we
selected two subpopulations: 1) pre/early perimenopausal
women (n ⫽ 655): women in this group remained premenopausal or early perimenopausal from baseline through visit 6;
and 2) late perimenopausal/postmenopausal women (n ⫽ 334):
women in this group had transitioned to late perimenopause or
postmenopause by visit 3 and provided data through visit 9.
Analyses were repeated in these subgroups. Effect modification
of the waist circumference/hormone relationships by menopausal status was also tested using generalized estimation equation modeling with SAS Windows 9.2 (SAS Institute, Cary, NC),
rather than SEM, to allow for testing of the overall interaction
effect across the follow-up period (4). Interaction terms between
prior menopausal status and prior hormone values were included
to predict subsequent waist circumference. Conversely, interaction terms between prior menopausal status and prior waist circumference were included to predict subsequent hormone levels.
To examine whether analyses that used a shorter time between visits would result in similar conclusions, models based on
visits 5, 6, and 7 were run (n ⫽ 1424 women with complete data
for all three visits). To evaluate whether results were influenced
by large changes in abdominal girth, we also ran models confined
to the subgroup who had a greater than 10% increase or greater
than 5% decrease in waist circumference between baseline and
visit 9 or between any two adjacent visits.
The Mplus statistical package (1998 –2008 version; Los Angeles, CA) was used for SEM modeling, using the maximum
likelihood procedure with nonnormality robust SE. Model fit was
evaluated by root mean square error of approximation (good fit
if values ⱕ0.05 or 95% confidence interval lower values close to
0 and upper values ⬍0.08); Bentler’s comparative fit index (good
fit if values ⱖ0.95) (5); and standardized root mean square residual (good fit if values ⬍0.05).
Results
Baseline mean age was 46 yr, and mean body mass index
(BMI) was in the overweight range (Table 1). Most women
had greater than a high school education, were never or
former smokers, and reported minimal or no intake of
jcem.endojournals.org
E1697
alcoholic beverages. Table 1 also presents characteristics
of the two menopausal status subsamples. As expected,
the group of women who remained pre- or early perimenopausal through visit 6 were younger than the group
that became late perimenopausal or postmenopausal by
visit 3. Women who transitioned to late peri- or postmenopause by visit 3 also had higher BMI values, were
more likely to be African-American, and were more often
current smokers. Mean weight, waist circumference, and
FSH increased across follow-up, whereas mean estradiol
decreased (Table 2). After small initial declines, serum testosterone levels increased slightly, whereas SHBG levels
remained stable.
SEM results are summarized in the figures. After accounting for the associations between current waist circumference and future waist circumference as well as current hormone levels and future hormone levels (black
arrows), the aims of the SEM models were to discern
whether current waist circumference predicts future hormone level (red arrows) and whether current hormone
level predicts future waist circumference (blue arrows).
Current refers to the value at each study visit and future to
the value at the next study visit. Statistically significant
associations are demarcated with solid lines, with thicker
arrows corresponding to a larger effect size. Dashed lines
represent nonsignificant associations. Hormones (original
values for SHBG and T, and log2 for FSH and estradiol)
and waist size were quantified in units of 1 SD to enable
comparison of the strength of association across hormones. Estimates equating to original values, rather than
standardized units, are presented in Supplemental Table 1.
Figure 1 shows the SEM results for the relationships
between SHBG and waist circumference. Current waist
circumference predicted future SHBG but not vice
versa. Specifically, for each SD increment in current
waist circumference, SHBG was lower by 0.04 – 0.15 SD
at the subsequent visit (Fig. 1, solid red arrows). Corresponding results using original units, rather than standardized units, indicated that for each 1-cm increment in
waist circumference, SHBG was lower by 0.07– 0.25 nM at
the subsequent visit (Supplemental Table 1). At no time
point did current SHBG predict future waist circumference (Fig. 1, dashed blue arrows).
Figure 2 shows the SEM results for the relationships
between T and waist circumference. As with SHBG, current waist circumference predicted future T (Fig. 2, solid
red arrows) but not vice versa (Fig. 2, dashed blue arrows).
Each SD higher current waist circumference was associated
with a 0.08- to 0.13-SD increase in T at the subsequent visit
(actual units: 0.08 – 0.16 mg/dl higher subsequent T for
each centimeter higher current waist circumference; Sup-
E1698
Wildman et al.
Sex Steroids and Adiposity in Midlife Women
J Clin Endocrinol Metab, September 2012, 97(9):E1695–E1704
TABLE 1. Baseline characteristics of the full analytic sample and subgroups used to examine menopausal status
interactions
Mean (SD) or n (%)
Age (yr)
BMI (kg/m2)
Body weight (kg)
Waist circumference (cm)
Race/ethnicity, n (%)
Caucasian
African-American
Chinese
Japanese
Menopause status, n (%)
Premenopausal
Early menopausal
Late perimenopausal
Postmenopausal
Education at V0, n (%)
High school or less
High school
College
Postcollege
Smoking status at V0, n (%)
Never
Former
Current
Alcohol consumption at V0, n (%)
No
One or fewer drinks/d
More than drink/d
Blood drawn in cycle d 2–5, n (%)
Yes
No/unknown
Overall (visit 0)
(n ⴝ 1528)
46.2 (2.6)
28.1 (7.3)
74.7 (21.2)
86.0 (16.4)
Premenoapusal/early
perimenopausal (visit 0)
(n ⴝ 655)
44.7 (1.8)
27.5 (6.7)
73.6 (19.7)
84.6 (15.5)
Late perimenopausal/
postmenopausal (visit 3)
(n ⴝ 334)
51.4 (2.4)
29.4 (8.1)
77.7 (22.6)
89.5 (18.0)
674 (44.1)
511 (33.4)
157 (10.3)
186 (12.2)
308 (47.0)
200 (30.5)
64 (9.8)
83 (12.8)
136 (40.7)
125 (37.4)
33 (9.9)
40 (12.0)
884 (58.8)
620 (41.2)
448 (69.8)
194 (30.2)
162 (48.5)
172 (51.5)
306 (20.2)
499 (32.9)
332 (21.9)
381 (25.1)
114 (17.5)
195 (29.9)
158 (24.2)
186 (28.5)
74 (22.3)
126 (38.0)
59 (17.8)
73 (22.0)
902 (59.4)
383 (25.2)
233 (15.4)
417 (64.4)
147 (22.7)
84 (13.0)
179 (53.9)
96 (28.9)
57 (17.2)
798 (52.4)
637 (41.8)
88 (5.8)
354 (54.4)
258 (39.6)
39 (6.0)
175 (52.6)
133 (39.9)
25 (7.5)
1217 (79.9)
307 (20.1)
559 (85.7)
93 (14.3)
14 (4.7)
285 (95.3)
Visit 3 occurred 3 yr after baseline, and visit 6 occurred 6 yr after baseline. V, Visit.
plemental Table 1). At no time point did current T predict
future waist circumference (Fig. 2, dashed blue arrows).
Similar results were again observed for relationships
between waist circumference and FSH (Fig. 3). Current
waist circumference was associated with future FSH (Fig.
3, solid red arrows) but not vice versa (Fig. 3, dashed blue
arrows). For each centimeter increment in current waist
circumference, subsequent FSH values were lower by
0.15– 0.26 SD. Thus, for a woman with average FSH values
at baseline (15.2 mIU/ml), each centimeter increment in
baseline waist circumference was associated with a 0.11
mIU/ml lower FSH at visit 3, whereas for a woman with
TABLE 2. Study characteristics across follow-up
Body weight (kg)
Waist circumference (cm)
E2 (pg/ml)
FSH (mIU/ml)
T (ng/dl)
SHBG (nM)
Menopausal status, n (%)
Premenopausal
Early perimenopausal
Late perimenopausal
Postmenopausal
Visit 0 (baseline)
74.7 (21.2)
86.0 (16.4)
54.8 (33.3– 85.2)
15.2 (11.0 –24.9)
42.0 (30.2–56.7)
40.4 (27.4 –57.6)
884 (58.8)
620 (41.2)
0
0
Visit 3
75.7 (21.1)
87.5 (16.8)
34.6 (21.2–75.0)
26.0 (13.9 –59.3)
34.7 (24.6 – 48.6)
39.1 (25.5–53.6)
Visit 6
76.3 (21.1)
89.0 (16.7)
22.2 (12.8 – 47.1)
58.8 (21.2–94.3)
36.9 (26.5–50.0)
39.5 (26.1–56.3)
Visit 9
76.5 (20.8
90.2 (16.8)
20.3 (15.0 –29.9)
89.5 (58.3–122.6)
37.9 (26.5–51.9)
39.4 (25.6 –58.6)
213 (14.9)
885 (61.8)
162 (11.3)
172 (12.0)
69 (4.6)
586 (39.4)
203 (13.5)
631 (42.4)
14 (1.0)
234 (16.6)
143 (10.1)
1021 (72.3)
Values are mean (SD), median (interquartile range), or n (%). Visit 3 occurred 3 yr after baseline; visit 6 occurred 6 yr after baseline; and visit 9
occurred 9 yr after baseline.
J Clin Endocrinol Metab, September 2012, 97(9):E1695–E1704
jcem.endojournals.org
E1699
FIG. 1. Sequential relationships between waist circumference and SHBG using standardized estimates. Red arrows, Association between current
waist circumference and future SHBG level; Blue arrows, association between current SHBG level and future waist circumference; black arrows,
top association between current SHBG and future SHBG; bottom association between current waist circumference and future waist circumference.
Solid lines indicate statistically significant associations, whereas dashed lines indicate nonsignificant associations. *, P ⬍ 0.05; **, P ⬍ 0.01; ***,
P ⬍ 0.001. The thickness of the arrow corresponds to the magnitude of the association (i.e. thicker arrows mean a larger effect size). Values
provided are standardized regression estimates. Visit 3 occurred 3 yr after baseline; visit 6 occurred 6 yr after baseline; and visit 9 occurred 9 yr
after baseline.
average FSH values at visit 6 (58.8 mIU/ml), each centimeter increment in visit 6 waist circumference was associated with a 0.82 mIU/ml lower FSH at visit 9 (Supplemental Table 1). At no time did current FSH predict future
waist circumference (Fig. 3, dashed blue arrows).
In contrast to SHBG, T, and FSH, the associations between estradiol levels and waist circumference were bidirectional and were different during the early menopause
transition stages (Fig. 4A) than they were in later menopause transition stages (Fig. 4B) (P for interaction from
generalized estimation equation modeling ⬍0.001).
Among premenopausal and early perimenopausal women
(Fig. 4A), greater current waist circumference at baseline
predicted lower future estradiol levels at visit 3; specifically, each centimeter increment in baseline waist circumference was associated with a lower visit 3 log2 estradiol
of 0.15 SD [0.55 pg/ml for a woman of average estradiol
levels (54.8 pg/ml) at baseline]. In contrast, during late
perimenopause and postmenopause, greater current waist
circumference predicted higher future estradiol levels (Fig.
4B): each centimeter increment in waist circumference was
associated with a 0.34- to 0.38-SD higher future estradiol
level [0.69 and 0.22 pg/ml for a woman with average estradiol values at visit 3 (34.6 pg/ml) and visit 6 (22.2 pg/
ml), respectively]. In both the early transition stages (Fig.
4A) and later transition stages (Fig. 4B), there was a small,
FIG. 2. Sequential relationships between waist circumference and T using standardized estimates. Red arrows, Association between current waist
circumference and future SHBG level; blue arrows, association between current SHBG level and future waist circumference; black arrows, top
association between current SHBG and future SHBG; bottom association between current waist circumference and future waist circumference.
Solid lines indicate statistically significant associations, whereas dashed lines indicate nonsignificant associations. *, P ⬍ 0.05; **, P ⬍ 0.01; ***,
P ⬍ 0.001. The thickness of the arrow corresponds to the magnitude of the association (i.e. thicker arrows mean a larger effect size). Values
provided are standardized regression estimates. Visit 3 occurred 3 yr after baseline; visit 6 occurred 6 yr after baseline; and visit 9 occurred 9 yr
after baseline.
E1700
Wildman et al.
Sex Steroids and Adiposity in Midlife Women
J Clin Endocrinol Metab, September 2012, 97(9):E1695–E1704
FIG. 3. Sequential relationships between waist circumference and FSH using standardized estimates. Red arrows, Association between current
waist circumference and future SHBG level; blue arrows, association between current SHBG level and future waist circumference; black arrows,
top association between current SHBG and future SHBG; bottom association between current waist circumference and future waist circumference.
Solid lines indicate statistically significant associations, whereas dashed lines indicate nonsignificant associations. *, P ⬍ 0.05;
**, P ⬍ 0.01; ***, P ⬍ 0.001. The thickness of the arrow corresponds to the magnitude of the association (i.e. thicker arrows mean a larger effect
size). Values provided are standardized regression estimates. Visit 3 occurred 3 yr after baseline; visit 6 occurred 6 yr after baseline; and visit 9
occurred 9 yr after baseline.
negative association between the current estradiol measured at visit 3 and future waist circumference measured
at visit 6.
Adjustment of the preceding analyses for body weight
had little effect on the results (data not shown). Additionally, substituting body weight for waist circumference in
the analyses described above did not alter the basic patterns of the associations, although the associations were
less consistent and weaker than those for waist circumference (Supplemental Figs. 1– 4). The notable exception
was the P value for the association between visit 3 T and
visit 6 body weight, which was statistically significant,
even though the standardized estimate was very similar to
the corresponding estimate for waist circumference
(⫺0.02 vs. ⫺0.01, respectively).
Sensitivity analyses (i.e. subset with complete, consecutive data at visits 5, 6, and 7; those with ⬎10% increase
or ⬎5% decrease in waist circumference) yielded results
similar to those presented above (data not shown).
Discussion
We prospectively evaluated the potential reciprocal nature of associations between adiposity and endogenous
sex hormones and SHBG. Our results suggest unidirectional relationships between waist circumference and
body weight with SHBG, T, and FSH, such that the
changes in adiposity precede the changes in SHBG, T,
and FSH, rather than the converse. In contrast, the relationship between waist girth and estradiol appeared
to be bidirectional, suggesting both that adiposity may
lead to changes in estradiol levels and that estradiol
levels may lead to changes in adiposity. Additionally,
the relationship between estradiol and waist circumference (and body weight) varied according to menopause
state such that larger waist circumference and body weight
were associated with lower future estradiol levels early in
the menopausal transition but with higher future estradiol
levels later in the menopausal transition.
Current weight and waist circumference levels predicted lower future SHBG levels throughout the menopause transition. This finding agrees with cross-sectional findings and may be related to adiposity-induced
hyperglycemia, which may suppress hepatic SHBG synthesis (6). In addition, hepatic triglyceride accumulation is associated with lower SHBG levels (7, 8). Because
the prevalence of hyperinsulinemia, hyperglycemia, and
fatty liver are higher with more abdominal fat, it is
plausible that weight gain lowers SHBG by increasing
insulin and glucose and/or by promoting hepatic fat accretion. Such a potential mechanism suggests that increased adiposity precedes reduced SHBG, in accordance
with our current results and with prior interventional
studies in which weight loss preceded increases in SHBG
(10, 11). However, other longitudinal observational studies have not reported associations between SHBG and either BMI or waist circumference (12).
That greater girth and weight predict higher testosterone levels is in accordance with cross-sectional analyses
and with the frequent co-occurrence of hyperandrogenemia and abdominal obesity observed in polycystic ovary
syndrome (13). However, most prior studies have not been
able to discern the sequential nature of the associations
because measurements were made too infrequently and/or
J Clin Endocrinol Metab, September 2012, 97(9):E1695–E1704
jcem.endojournals.org
E1701
ciation is spurious. Taken together, the
current SWAN results suggest that the
predominant sequence is that increases
in adiposity precede increases in T
rather than the reverse. Consistent with
this are data from murine models showing that knockout of the androgen receptor has no influence on fat mass; reported body weight reductions were
due solely to decreases in lean mass
(17). However, data in nonmidlife, female-to-male transsexuals suggests
that testosterone administration may
increase abdominal visceral fat. In these
studies, supraphysiological, exogenous
testosterone administration increased
visceral fat but decreased sc fat, with no
net change in overall BMI or total fat
mass (18, 19). However, because participants received doses of testosterone
that were massively supraphysiological, it is difficult to extrapolate these
findings to normal women.
Our observed negative associations
between weight or waist circumference
with FSH are consistent with other
studies of midlife women showing that
obese women have lower FSH than
nonobese women (20 –22). Given the
FIG. 4. Sequential relationships between waist circumference and estradiol, stratified by
menopausal status using standardized estimates. A, Premenopausal and early perimenopausal
negative feedback between estradiol
women. B, Late perimenopausal and postmenopausal women. Red arrows, Association between
and FSH, higher postmenopausal estracurrent waist circumference and future SHBG level; blue arrows, association between current
diol levels among obese women may exSHBG level and future waist circumference; black arrows, top, association between current SHBG
and future SHBG; bottom association between current waist circumference and future waist
plain their lower FSH levels.
circumference. Solid lines indicate statistically significant associations, whereas dashed lines
In contrast to the overarching patindicate nonsignificant associations. *, P ⬍ 0.05; **, P ⬍ 0.01; ***, P ⬍ 0.001. The thickness of
tern that greater girth and weight drive
the arrow corresponds to the magnitude of the association (i.e. thicker arrows mean a larger
effect size). Values provided are standardized regression estimates. Visit 3 occurred 3 yr after
changes in levels of SHBG, T, and FSH,
baseline; visit 6 occurred 6 yr after baseline; and visit 9 occurred 9 yr after baseline.
for estradiol we found not only that
current waist circumference was assobecause the analyses were limited to concurrent repeated
ciated with future estradiol levels but also the reverse, almeasurements of testosterone and adiposity, rather than
though the effect sizes were much larger in the former
lagged relationships (14, 15). Two prior studies tested for
direction. The negative relation between current estradiol
lagged relationships between testosterone and adiposity
and future waist circumference was observed only in the
measures: one found inconsistent and weak associations,
whereas the other found that higher baseline testosterone interval between SWAN visits 3 and 6, reflecting probable
levels and greater increases in testosterone were associated heterogeneity in the transition from early to late perimenowith moderate increases in the risk of incident obesity, pause because many women transitioned from early to late
perimenopause between these visits. These relative mageven after adjusting for baseline BMI (16).
We found a statistically significant negative association nitudes are consistent with most studies examining weight
between testosterone at visit 3 and body weight at visit 6, changes in midlife, suggesting that weight gain is primarily
suggesting a possible bidirectional relationship between due to aging rather than to menopause-related estrogen
body weight and T (see Supplemental Fig. 2). However, deficiency, (23–26) and with experimental data showing
given that this association was weak and is not consistent that exogenous estrogen administration in postmenowith prior SWAN data (16), it is possible that this asso- pausal women does not prevent weight gain (27). How-
E1702
Wildman et al.
Sex Steroids and Adiposity in Midlife Women
ever, although the effect sizes were larger for adiposity’s
effects on estradiol than for the reverse, murine models
suggest the estrogen protects against fat accumulation;
abdominal fat in particular is markedly increased in mice
with null mutations of either their estrogen receptor-␣
gene or their aromatase gene (28, 29). This pattern of fat
accumulation is consistent with some human epidemiological studies, suggesting that although menopause may
not be associated with overall weight gain, it may lead to
focused fat accumulation in the abdominal region (30 –33)
and with our finding that estradiol’s associations were
stronger for waist size than for body weight. Although
these murine models are compelling, the effects of life-long
estrogen deficiency, as in these animals, may differ from
estrogen deficiency that develops in midlife as occurs with
the menopause.
In accordance with our results, others have also found
that the relation between waist circumference (or weight)
and estradiol levels differ in early compared with later
menopause transition stages, with a negative association
early and a positive association later in the transition (20,
21, 34 –36). One postulate for this finding is that low
SHBG levels in obese women result in greater estradiol
clearance and therefore lower circulating estradiol levels
(21), which is potentially overwhelmed later in the transition, when estradiol levels are substantially lower, by
peripheral estradiol production via androgen aromatization within adipose tissue. However, our data do not support this hypothesis because when we adjusted our estradiol results for SHBG levels, the negative associations
between weight or waist circumference and estradiol in
premenopausal women persisted. Alternatively, murine
models document anorectic and activity-promoting effects of estradiol (37, 38), raising the possibility that before
menopause, when estradiol levels are higher, the negative
association between waist circumference and estradiol
may result from estradiol-induced decreases in meal size
and increases in physical activity.
The results of this study must be viewed within the
context of its limitations. Waist circumference represents
both abdominal visceral and sc fat. However, existing
studies of computed tomography or magnetic resonance
imaging measures of adiposity in midlife women do not
include the necessary number of repeated measurements
or the large sample of women required to address questions of bidirectionality. Estradiol levels in premenopausal
women vary from hour to hour and day to day; standardization of blood sampling by time of day and day of the
menstrual cycle was implemented to reduce measurement
error. Hormone measurements were quantified by immunoassay rather than by mass spectrometry. Therefore, despite the careful assay development, measurement error at
J Clin Endocrinol Metab, September 2012, 97(9):E1695–E1704
extremely low sex steroid cannot be ruled out. However,
prior cohorts have found similar results and immunoassay
is less prone to measurement error when estradiol levels
are high, as in premenopausal women, which constituted
an appreciable proportion of our sample. Furthermore,
both the estradiol and testosterone assays have been validated against celite chromatography (9). Strengths of the
current study include 9 yr of observation on a well-characterized, race-ethnically diverse cohort, and the use of
SEM methodology, a methodology not used in previous
longitudinal investigations, and particularly well suited to
exploration of bidirectional, or reciprocal, associations.
In conclusion, both current body weight and waist circumference were strongly associated with future endogenous sex hormones and SHBG, whereas current hormones
were either not found to be associated with future adiposity measures or, in the case of estradiol, were considerably
more weakly associated with future adiposity measures.
Therefore, it is likely that the more predominant temporal
sequence among midlife women is that adiposity initiates
changes in sex hormones, rather than vice versa.
Acknowledgments
The content of this manuscript is solely the responsibility of the
authors and does not necessarily represent the official views of
the National Institutes of Health, National Institute of Nursing
Research, Office of Research on Women’s Health, or the National Institutes of Health. Clinical centers included the following: University of Michigan, Ann Arbor, Siobán Harlow, principal investigator (PI) 2011, MaryFran Sowers, PI 1994 –2011;
Massachusetts General Hospital, Boston, MA, Joel Finkelstein,
PI 1999 to present; Robert Neer, PI 1994 to 1999; Rush University, Rush University Medical Center, Chicago, IL, Howard
Kravitz, PI 2009 to present; Lynda Powell, PI 1994 –2009; University of California, Davis/Kaiser Permanente, Ellen Gold, PI;
University of California, Los Angeles, Gail Greendale, PI; Albert
Einstein College of Medicine, Bronx, NY, Carol Derby, PI 2011,
Rachel Wildman, PI 2010 –2011, Nanette Santoro, PI 2004 –
2010; University of Medicine and Dentistry, New Jersey Medical
School, Newark, NJ, Gerson Weiss, PI 1994 –2004; and the University of Pittsburgh, Pittsburgh, PA, Karen Matthews, PI. National Institutes of Health Program Office: National Institute on
Aging, Bethesda, MD, Sherry Sherman, 1994 to present; Marcia
Ory, 1994 –2001; and National Institute of Nursing Research,
Bethesda, MD, program officers. The central laboratory was the
University of Michigan, Ann Arbor, Daniel McConnell (Central
Ligand Assay Satellite Services). The coordinating centers were
the following: University of Pittsburgh, Pittsburgh, PA, Kim Sutton-Tyrrell, PI 2001 to present; New England Research Institutes, Watertown, MA, Sonja McKinlay, PI 1995–2001. The
steering committee: Susan Johnson, current chair; and Chris Gallagher, former chair. We thank the study staff at each site and all
the women who participated in SWAN.
J Clin Endocrinol Metab, September 2012, 97(9):E1695–E1704
Address all correspondence and requests for reprints to:
Rachel P. Wildman, Ph.D., M.P.H., F.A.H.A., Associate Professor, Department of Epidemiology, Population Health, Albert Einstein College of Medicine, 1300 Morris Park
Avenue, Belfer Room 1309, Bronx, New York 10461. E-mail:
[email protected].
The Study of Women’s Health Across the Nation has grant support from the National Institutes of Health, Department of Health
and Human Services through the National Institute on Aging, the
National Institute of Nursing Research, and the National Institutes
of Health Office of Research on Women’s Health (Grants
NR004061; AG012505, AG012535, AG012531, AG012539,
AG012546, AG012553, AG012554, and AG012495).
Disclosure Summary: The authors have nothing to disclose.
References
1. Wildman RP, Sowers MR 2011 Adiposity and the menopausal transition. Obstet Gynceol Clin North Am 38:441– 454
2. Sowers MR 2000 SWAN: a multicenter, multiethnic, communitybased cohort study of women and the menopause transition. In:
Lobo RA, Kelsey J, and Marcus J, eds. Menopause: biology and
pathobiology. San Diego: Academic Press; 175–188
3. Block G, Thompson FE, Hartman AM, Larkin FA, Guire KE 1992
Comparison of two dietary questionnaires validated against multiple dietary records collected during a 1-year period. J Am Diet Assoc
92:686 – 693
4. Zeger SL, Liang KY, Albert PS 1988 Models for longitudinal data:
a generalized estimating equation approach. Biometrics 44:1049 –
1060
5. Bentler PM 1990 Comparative fit indexes in structural models. Psycholog Bull 107:238 –246
6. Selva DM, Hogeveen KN, Innis SM, Hammond GL 2007 Monosaccharide-induced lipogenesis regulates the human hepatic sex hormone-binding globulin gene. J Clin Invest 117:3979 –3987
7. Stefan N, Schick F, Häring HU 2009 Sex hormone-binding globulin
and risk of type 2 diabetes. N Engl J Med 361:2675–2676; author
reply 2677–2678
8. Shin JY, Kim SK, Lee MY, Kim HS, Ye BI, Shin YG, Baik SK, Chung
CH 2011 Serum sex hormone-binding globulin levels are independently associated with nonalcoholic fatty liver disease in people with
type 2 diabetes. Diabetes Res Clin Pract 94:156 –162
9. Anderson DC, Lasley BL, Risher RA, Shepherd JH, Newman L,
Hendrickx AG 1976 Transplacental gradients of sex-hormonebinding globulin in human and simian pregnancy. Clin Endocrinol
(Oxf) 5:657– 669
10. Tolino A, Gambardella V, Caccavale C, D’Ettore A, Giannotti F,
D’Antò V, De Falco CL 2005 Evaluation of ovarian functionality
after a dietary treatment in obese women with polycystic ovary syndrome. Eur J Obstet Gynecol Reprod Biol 119:87–93
11. Kopp HP, Krzyzanowska K, Schernthaner GH, Kriwanek S, Schernthaner G 2006 Relationship of androgens to insulin resistance and
chronic inflammation in morbidly obese premenopausal women:
studies before and after vertical banded gastroplasty. Obes Surg
16:1214 –1220
12. Sternfeld B, Liu K, Quesenberry Jr CP, Wang H, Jiang SF, Daviglus
M, Fornage M, Lewis CE, Mahan J, Schreiner PJ, Schwartz SM,
Sidney S, Williams OD, Siscovick DS 2008 Changes over 14 years in
androgenicity and body mass index in a biracial cohort of reproductive-age women. J Clin Endocrinol Metab 93:2158 –2165
13. Pasquali R 2006 Obesity and androgens: facts and perspectives.
Fertil Steril 85:1319 –1340
14. Guthrie JR, Dennerstein L, Taffe JR, Ebeling PR, Randolph JF,
Burger HG, Wark JD 2003 Central abdominal fat and endogenous
jcem.endojournals.org
15.
16.
17.
18.
19.
20.
21.
22.
23.
24.
25.
26.
27.
28.
29.
30.
31.
32.
E1703
hormones during the menopausal transition. Fertil Steril 79:1335–
1340
Sowers MF, Beebe JL, McConnell D, Randolph J, Jannausch M
2001 Testosterone concentrations in women aged 25–50 years: associations with lifestyle, body composition, and ovarian status. Am J
Epidemiol 153:256 –264
Sutton-Tyrrell K, Zhao X, Santoro N, Lasley B, Sowers M, Johnston
J, Mackey R, Matthews K 2010 Reproductive hormones and obesity: 9 years of observation from the Study of Women’s Health
Across the Nation. Am J Epidemiol 171:1203–1213
Callewaert F, Venken K, Ophoff J, De Gendt K, Torcasio A, van
Lenthe GH, Van Oosterwyck H, Boonen S, Bouillon R, Verhoeven
G, Vanderschueren D 2009 Differential regulation of bone and body
composition in male mice with combined inactivation of androgen
and estrogen receptor-␣. FASEB J 23:232–240
Elbers JM, Giltay EJ, Teerlink T, Scheffer PG, Asscheman H, Seidell
JC, Gooren LJ 2003 Effects of sex steroids on components of the
insulin resistance syndrome in transsexual subjects. Clin Endocrinol
(Oxf) 58:562–571
Giltay EJ, Elbers JM, Gooren LJ, Emeis JJ, Kooistra T, Asscheman
H, Stehouwer CD 1998 Visceral fat accumulation is an important
determinant of PAI-1 levels in young, nonobese men and women:
modulation by cross-sex hormone administration. Arterioscler
Thromb Vasc Biol 18:1716 –1722
Freeman EW, Sammel MD, Gracia CR, Kapoor S, Lin H, Liu L,
Nelson DB 2005 Follicular phase hormone levels and menstrual
bleeding status in the approach to menopause. Fertil Steril 83:383–
392
Freeman EW, Sammel MD, Lin H, Gracia CR 2010 Obesity and
reproductive hormone levels in the transition to menopause. Menopause 17:718 –726
Burger HG, Dudley EC, Hopper JL, Groome N, Guthrie JR, Green
A, Dennerstein L 1999 Prospectively measured levels of serum follicle-stimulating hormone, estradiol, and the dimeric inhibins during
the menopausal transition in a population-based cohort of women.
J Clin Endocrinol Metab 84:4025– 4030
Crawford SL, Casey VA, Avis NE, McKinlay SM 2000 A longitudinal study of weight and the menopause transition: results from the
Massachusetts Women’s Health Study. Menopause 7:96 –104
Davies KM, Heaney RP, Recker RR, Barger-Lux MJ, Lappe JM
2001 Hormones, weight change and menopause. Int J Obes Relat
Metab Disord 25:874 – 879
Guthrie JR, Dennerstein L, Dudley EC 1999 Weight gain and the
menopause: a 5-year prospective study. Climacteric 2:205–211
Wing RR, Matthews KA, Kuller LH, Meilahn EN, Plantinga PL
1991 Weight gain at the time of menopause. Arch Intern Med 151:
97–102
Kongnyuy EJN, RJ, Flight HKI, Rees MC 2007 Oestrogen and progestogen hormone replacement therapy for peri-menopausal and
post-menopausal women: weight and body fat distribution. Cochrane Database Syst Rev 7
Heine PA, Taylor JA, Iwamoto GA, Lubahn DB, Cooke PS 2000
Increased adipose tissue in male and female estrogen receptor-alpha
knockout mice. Proc Natl Acad Sci USA 97:12729 –12734
Jones ME, Thorburn AW, Britt KL, Hewitt KN, Wreford NG, Proietto J, Oz OK, Leury BJ, Robertson KM, Yao S, Simpson ER 2000
Aromatase-deficient (ArKO) mice have a phenotype of increased
adiposity. Proc Natl Acad Sci USA 97:12735–12740
Toth MJ, Tchernof A, Sites CK, Poehlman ET 2000 Menopauserelated changes in body fat distribution. Ann NY Acad Sci 904:502–
506
Donato GB, Fuchs SC, Oppermann K, Bastos C, Spritzer PM 2006
Association between menopause status and central adiposity measured at different cutoffs of waist circumference and waist-to-hip
ratio. Menopause 13:280 –285
Janssen I, Powell LH, Crawford S, Lasley B, Sutton-Tyrrell K 2008
Menopause and the metabolic syndrome: the Study of Women’s
Health Across the Nation. Arch Intern Med 168:1568 –1575
E1704
Wildman et al.
Sex Steroids and Adiposity in Midlife Women
33. Pasquali R, Casimirri F, Pascal G, Tortelli O, Morselli Labate A,
Bertazzo D, Vicennati V, Gaddi A 1997 Influence of menopause on
blood cholesterol levels in women: the role of body composition, fat
distribution and hormonal milieu. Virgilio Menopause Health
Group. J Intern Med 241:195–203
34. Randolph Jr JF, Sowers M, Bondarenko IV, Harlow SD, Luborsky
JL, Little RJ 2004 Change in estradiol and follicle-stimulating hormone across the early menopausal transition: effects of ethnicity and
age. J Clin Endocrinol Metab 89:1555–1561
35. Randolph Jr JF, Sowers M, Gold EB, Mohr BA, Luborsky J, Santoro
N, McConnell DS, Finkelstein JS, Korenman SG, Matthews KA,
J Clin Endocrinol Metab, September 2012, 97(9):E1695–E1704
Sternfeld B, Lasley BL 2003 Reproductive hormones in the early
menopausal transition: relationship to ethnicity, body size, and
menopausal status. J Clin Endocrinol Metab 88:1516 –1522
36. Randolph Jr JF, Zheng H, Sowers MR, Crandall C, Crawford S,
Gold EB, Vuga M 2011 Change in follicle-stimulating hormone and
estradiol across the menopausal transition: effect of age at the final
menstrual period. J Clin Endocrinol Metab 96:746 –754
37. Eckel LA 2004 Estradiol: a rhythmic, inhibitory, indirect control of
meal size. Physiol Behav 82:35– 41
38. Gorzek JF, Hendrickson KC, Forstner JP, Rixen JL, Moran AL,
Lowe DA 2007 Estradiol and tamoxifen reverse ovariectomy-induced physical inactivity in mice. Med Sci Sports Exerc 39:248 –256