(CANCER RESEARCH 51. 568-572. January 15. 1991]
Endometrial Cancer, Obesity, and Body Fat Distribution1
Harland Austin,2 J. Max Austin, Jr., Edward E. Partridge, Kenneth D. Hatch, and Hugh M. Shingleton
/h'partmenl of Epidemiology, School of Public Health, (.'niversity of Alabama at Birmingham 35294 ¡lì.
A.]; Department of Obstetrics ami (iynecoloxy. L'niversity of
Alabama at Birmingham 35294 ¡E.E. P., K. D. H., H. M. S./; and Southern (iyn Oncology, Birmingham /J. M..(./. Alabama 35205
ABSTRACT
A case-control study »asundertaken to evaluate the roles of obesity
and body fat distribution in the etiology of endometrial cancer. The study
also included an evaluation of the associations of serum estrone, estradiol,
and androstenedionc with obesity, body fat distribution, and endometrial
cancer risk. The study included 168 cases and 334 control subjects
identified at an optometry clinic.
A strong, positive relationship between overall obesity and endometrial
cancer »asfound. Hie relative rate of endometrial cancer for »omenin
the upper 90th percentile of a body mass index compared to those below
the median was estimated as 5.5 with 95% confidence limits of 3.2-9.6.
There was no association between endometrial cancer and the waist to
hip ratio, an index of upper versus lower body fat distribution. A statistical
test of trend across the four quartiles of the waist to hip ratio yielded a
/' value of 0.45 after adjustment for confounding by the body mass index.
On the other hand, there was a statistically significant, independent
positive effect of a high subscapular to tricep skinfold ratio, a measure
of central versus peripheral obesity, on endometrial cancer risk. The
relative rates of endometrial cancer for the second, third, or fourth
quartile compared to the first quartile of this index were 1.5, 1.9, and
2.7, respectively (P = 0.007), after adjustment for the body mass index.
Serum estrone and estradiol, but not androstenedione, were statistically
significantly correlated with the body mass index among control subjects
(r = 0.37 and 0.40 for estrone and estradiol, respectively). On the other
hand, each of the sex hormones was uncorrelated «idithe waist to hip
ratio after adjustment for body mass. The correlations between each of
the three hormones and the subscapular to tricep skinfold ratio among
controls were weak and were not statistically significant (0.10, 0.10, and
0.14 for estrone, estradiol and androstenedione, respectively). Cases had
statistically significantly higher mean serum estrogen and androstenedi
one levels than did controls and these elevations did not simply reflect a
higher prevalence of obesity among them.
The findings are equivocal with respect to fat patterns and endometrial
cancer. \Ve suggest that future epidemiológica! studies of cancer and
body fat distribution more carefully distinguish among the various types
of fat patterns.
for EC (1,2) and are more common among individuals with
UBSO, and since EC is strongly related to degree of obesity,
this study was undertaken to investigate whether women with
UBSO have a higher risk of EC than do women with LBSO.
We have attempted to distinguish between the independent
effects of the amount of body fat and its distribution on EC
risk.
Obese women produce more estrone through aromati/ation
of androstenedione in fat cells and have measurably higher
serum estrogen levels than do the nonobese (9-11). Serum
estrone, estradiol, and androstenedione were measured among
control subjects and their relation with overall obesity and with
its distribution were evaluated.
MATERIALS
AND METHODS
Cases. Endometrial cancer subjects were identified through the serv
ices of the Department of Obstetrics and Gynecology at the university
hospital and through a large, private gynecological-oncological practice
in Birmingham. AL. All cases were diagnosed between June 1984 and
October 1988, and subjects were enrolled into the study between June
1985 and December 1988. The study includes 17 women for whom the
diagnoses of EC were made up to 1 year prior to the commencement
of the study (prevalent cases) and 151 newly diagnosed cases (incident
cases).
Twelve women with EC refused participation in the study. Thus. 168
(93%) of 180 eligible cases participated. All cases were confirmed
histologically. Classifications were as follows: 149 adcnocarcinomas. 6
adenoacanthomas, and 13 adenosquamous carcinomas; 144 stage I. 16
stage II, 5 stage III, 2 stage IV. and 1 stage unknown.
The cases ranged in age from 40 to 82 years with a mean age of
about 64 years; 153 are white and the remaining 15 are black; 17 were
premenopausal at the time of diagnosis and 151 were postmenopausal.
Controls. The controls were selected from among women attending
the university optometry clinic. Almost all women attend this clinic for
routine eye examinations and generally are in good health.
Every 3 months a listing of women s40 years of age who had recently
INTRODUCTION
attended the clinic was obtained. Controls were selected from among
Excess body fat is widely recognized as a determinant of EC.1 women on this roster such that their age and race distribution was
similar to that of the cases. Controls included only women with an
Morbidly obese women experience at least a 3-fold increased
intact uterus.
risk of EC compared with women of average or below average
Included in the study are 334 controls. There were 100 women
weight (1-3). Furthermore, it has been reported that persons
identified as eligible but who refused participation (participation rate,
with a deposition of fat predominately in the upper body 77%). The controls ranged in age from 40-79 years with a mean age of
segment, android obesity, experience a higher risk of heart
about 63 years. They include 312 whites, 21 blacks, and 1 oriental who
disease, hypertension, diabetes, and other obesity-related met
was considered as wbite in all analyses. At the time of enrollment. 30
abolic disorders than do persons with a deposition of fat in the of the controls were premenopausal; the remaining 304 were postmen
opausal.
lower body segment, gynecoid obesity (4-8).
AnthropométrieMeasurements. All anthropométrie measurements
Since a history of diabetes and hypertension are risk factors
were done by a certified dietetic technician trained in anthropometry
techniques. The technician obtained the measurements cither at the
Received 6/11/90; accepted 10/25/90.
university or at the subject's home. Height was measured without shoes
The costs of publication of this article were defrayed in part by the payment
of page charges. This article must therefore be hereby marked advertisement in
to the nearest one-quarter inch. Weight was measured with a Seco
accordance with 18 I'.S.C. Section I7.M solely to indicate this fact.
' Supported by a grant from the National Cancer Institute (Rol-C'A39733).
platform scale to the nearest pound. Circumferences were measured
with a cloth tape and recorded to the nearest one-quarter cm. Skinfolds
2 To whom requests for reprints should be addressed, at 202 Tidwell Hall.
School of Public Health. I niversit\ Station. Birmingham. AL 35294.
were measured to the nearest mm with Lange calipers according to the
1The abbreviations used are: EC. endometrial cancer: I'BSO. upper body
method of Jelliffe (12). Three skinfold measurements were obtained
segment obesity: LBSO. lower body segment obesity: BM1. body mass index;
and averaged. All limb circumferences and skinfolds were measured on
\VHR. waist to hip ratio: STR. subscapular to tricep ratio: RR. relative rate: CI.
the right side. A brief description of each measurement and the positions
confidence interval.
568
Downloaded from cancerres.aacrjournals.org on June 16, 2017. © 1991 American Association for Cancer Research.
ENDOMETR1A1. CANCER AND I AT PATTERNS
of the subjects during measurement is available upon request.
The anthropométriemeasurements were obtained prior to hysterec
tomy for 111 cases and were done postoperatively for 57 cases. For
these latter cases, the measurements were done within 6 months of
diagnosis for 31, between 6 and 12 months for 19. and >l year
postdiagnosis for only 7.
A BM1. weight (in kg) divided by height (in m) raised to the power
1.5, is used as an estimate of overall obesity (13). Additionally, the
subscapular plus the triceps skinfolds were used to estimate overall
obesity.
The distribution of LBSO versus L'BSO was evaluated through the
use of the \VHR. the minimum waist divided by the maximum hip
circumference. Other indices of UBSO that we evaluated are the waist
to thigh ratio (the minimum waist divided by the midthigh circumfer
ence) and the midarm circumference divided by the midthigh circum
ference.
In addition to upper versus lower body fat distribution, we evaluated
the relation between EC and indices of central versus peripheral body
fat. The STR and another central obesity index suggested by Hiramatsu
et al. were obtained. The index of Hiramatsu el al. is the sum of the
circumferences of the neck, chest, and abdomen divided by the sum of
the upper arm. the upper thigh, and the calf (14).
Hormonal Assays. Subjects at least 3 years postmenopause not on
replacement estrogens were asked for 20 ml of venous blood. Of such
cases, 52% consented to venipuncture. as did 52% of eligible controls.
Sera were assayed for estrone, estradiol. and androstenedione for 67
cases and 142 controls. The measurements were done by radioimmunoassay by a commercial laboratory (Nichols Institute: Reference Labo
ratories. San Juan Capistrano. CA).
Interviews. Subjects were questioned about their reproductive history,
their medical history, and other aspects of their background and life
styles. Interviews were conducted in-person by the technician.
Statistical Analysis. The exposure odds ratio, an estimate of the RR
of EC for women "exposed" to some factor compared with those
"unexposed" was estimated by unconditional logistic regression (15).
Table 1 Distribution of cases and controls and the relative rates with 95' i
confluence internals according to a body mass index
BM1
(kg/m")<28.428.4-32.032.0-36.4>36.4TotalCases30233877168Control83887984334RR"1.00.71
CI0.4.
1.30.7.
2.21.3.
3.9
a From ii logistic regression model including terms for age. race, years of
schooling, and three indicator variables for the quartiles of BMI.
Table 2 Distribution of cases and controls and the relative rates with 95l'Ãconfluence intervals according to the waist to hip ratio
WHRsO.760.76-0.810.81-0.86>0.86TotalCases25354068168Controls81818587334RR°1.01.20.
Cl0.6.0.5,0.6,2
21.82.1
" Adjusted for age. race, years of schooling, and BMI.
Table 3 Distribution afeases and controls and the relative rates with V5'i
confidence intervals according to the suhscapular to triceps skinfold ratio
Cl0.8,
STRf=0.580.58-0.710.71-0.85>0.85TotalCases22334864167Controls85798288334RR°1.01.51.
2.81.0,
3.61.5,4.9
°Adjusted for age. race, years of schooling, and the BMI.
WHR. an index of upper versus lower body fat distribution, is
displayed in Table 2. Since there is a strong, positive correlation
between the BMI and the WHR (r = 0.56), the association
between the WHR and EC is confounded by BMI. There is no
residual association between EC and the WHR after adjustment
The matching factors, age and race, were included in all such analyses.
for BMI (P = 0.45 for the trend test). There also is no associ
Since the cases were less educated than were the controls (see below),
ation between the WHR and EC among subjects above the 50th
years of schooling was included in all logistic models.
For continuous variables, the distribution of the measurement was percentile of BMI. The mean WHR for controls is 0.82. while
the BMI-adjusted mean WHR difference between cases and
examined among the controls and divided into quartiles. Three indi
controls
is 0.005 (P = 0. 45). The WTR and the midarm to
cator variables were used in a model to estimate the RRs for each
quartile. A statistical test of trend was evaluated either by the use of an
midthigh circumference ratio (other measures of UBSO) also
ordinal variable with four equally spaced scores or by use of a contin
were not associated with EC.
uous exposure variable in a logistic regression model. The natural
In contrast to the findings above, there is a positive, inde
logarithm of the steroid measurements was used for statistical testing
pendent association between the STR, an index of central versus
because the untransformed values were asymmetric.
peripheral obesity, and EC (Table 3). A trend test for the
The evaluation of the 3 steroid measurements among the controls
relationship between EC risk and increasing STR yields a P
was done by use of multiple regression techniques and partial correla
value of 0.007. Nearly identical RRs were obtained when ad
tion (16). All /'values are two tailed and 95% CIs are used.
justment for overall obesity was made by use of the sum of the
subscapular and triceps skinfolds rather than by the BMI. The
mean STR among controls is 0.74, while the mean difference
RESULTS
between cases and controls after adjustment for the BMI is 0.05
The mean years of schooling of cases and controls are 11.1 (P = 0.01). The positive association between EC and STR was
and 12.4, respectively. Forty-two % of cases and 25% of controls
evident among subjects in the first 3 BMI quartiles. However,
did not complete high school. Since obesity is more prevalent
among subjects in the fourth BMI quartile the association was
among uneducated women, years of schooling was included in slightly stronger (P = 0.0.3).
all logistic models.
The analysis of the index of Hiramatsu et al. provides addi
The distribution of cases and controls according to the BMI,
tional support for the hypothesis that central obesity is an EC
risk factor. The BMI-adjusted RR for subjects exceeding the
the index of overall obesity, is displayed in Table I. The trend
between EC risk and increasing obesity is highly statistically
90th percentile of this index compared to those below the
median is 1.9 (95% CI, 1.0 to 3.5; P = 0.05). This RR was
significant (P < 0.0001). However, the excess risk is confined
to very obese women. The RR for subjects exceeding the 90th
about the same in each BMI quartile.
percentile compared with those below the median is 5.5 (95%
Steroid Hormones among the Controls. In order to elucidate
CI. 3.2 to 9.6). Similar findings were obtained with the use of the possible mechanisms by which obesity and fat patterns may
the sum of the subscapular and triceps skinfolds as the index
influence EC risk, the distributions of the sex hormones were
of obesity.
evaluated among control subjects. There is a strong, positive
The distribution of subjects according to quartiles of the correlation between the BMI and the logarithm of serum estra569
Downloaded from cancerres.aacrjournals.org on June 16, 2017. © 1991 American Association for Cancer Research.
ENDOMETRIAL
CANCER AND FAT PATTERNS
Table 7 Distribution afeases and controls according to other putative
endometrial cancer risk factors
diol (r = 0.40, P < 0.0001) and of serum estrone (r = 0.37, P
< 0.0001). On the other hand, there is little or no association
(r = 0.08, P = 0.37) between the logarithm of serum androstenedione and the BMI. The mean levels of each hormone among
controls according to two BMI levels are displayed in Table 4.
In contrast to the positive relationship between serum estro
gens and the BMI, there is little association between any sex
hormone and the WHR (Table 5). Serum estradiol levels are
slightly higher among subjects in the fourth WHR quartile,
even after adjustment for the BMI, but the difference is not
statistically significant. The partial correlation coefficients after
adjustment for age, race, years of schooling, and the BMI,
between the WHR and the logarithm of estrone, estradiol, and
androstenedione are 0.02 (P = 0.83), 0.14 (P = 0.11), and
-0.02 (P = 0.85), respectively.
There is a weak, statistically nonsignificant, positive corre
lation between the STR and the sex hormones among the
controls. The partial correlation coefficients for the logarithm
of estrone, estradiol, and androstenedione with the STR are
0.10 (P = 0.23), 0.10 (P = 0.25), and 0.14 (P = 0.09), respec
tively.
Hormonal Levels among Cases and Controls. The BMI-adjusted mean levels of estrone, estradiol, and androstenedione
among cases and controls are displayed in Table 6. For each
hormone, the mean is statistically significantly higher among
cases than it is among controls. The BMI-adjusted RR for
subjects in the fourth quartile of estrone (32 pg/ml) compared
with those below the median (24 pg/ml) is 3.8 (1.7, 8.4; P =
0.001), while that for the fourth quartile of estradiol (10 pg/
ml) is 2.9 (1.3, 6.5; P = 0.01) and that for the fourth quartile
of androstenedone (80 ng/dl) is 3.8 (1.9, 8.0; P = 0.0003).
The relationship between obesity and EC apparently is some-
CI0.6,
Age
(yr)£1112-1314-1516+Pregnancy
al menarche
1.70.4,
1.40.2,
1.21.1,3.40.4.
historyParousNulliparousAge
(yr)<4546-4950-5455+Unopposed
at menopause
1.50.9,
2.71.0.4.00.6,
estrogen ther
apyNone
yr1-45-1011
or <1
+Opposed
therapyNeverEverHistory
estrogen
1.31.6.
diabetesNoYesHistory
of
5.71.6,3.6
hypertensionNoYesCase28884171422629256723121149191605138308682Control5116689273043068
of
" Adjusted for age, race and years of schooling.
Table 4 Mean levels and differences between the logarithm of mean serum
estrone (pg/ml), estradiol (pK/tnt). and androstenedione (ng/dl) according to
obesity status among the controls
BMI>75th
percentile
(/V= 39)
<75th percentile
70.2-0.06(0.10)
24.10.24
7.60.36(0.10)
GV= 103)
Difference" (SE)
(0.084)
P valueEstrone31.0
0.005Estradio]10.7
0.0005Androstenedione62.4
0.53
* Mean difference in logarithm of measurement adjusted for age. race and
years of schooling.
Table 5 Adjusted mean levels and differences between the logarithm of mean
serum estrone (pg/ml). estradiol (pg/ml), and androstenedione (ng/dl) hy the
WHR among the controls
WHR
Estrone
Estradiol
2.60.7,
4.22.1,
10.30.2.
Androstenedione
>75th
percentile<75th
percentile(A'=
101)Difference"
(SE)P
(0.088)0.3410.27.80.18(0.10)0.0969.067.70.06(0.11)0.61
value27.025.60.08
" Difference in mean logarithms of measurements adjusted for age, race, years
of schooling, and BMI.
what secondary to high endogenous estrogen levels. Among the
209 subjects for whom serum estrogen measurements were
available, the RR for subjects in the fourth BMI quartile com
pared with those below the median is 2.7 (1.3, 5.5; P = 0.006).
However, after adjustment for serum estrone, this RR is re
duced to 1.7 (0.8, 3.7). Adjustment for androstenedione, rather
than estrone, does not appreciably change this RR.
Other Risk Factors. The distribution of cases and controls
according to other putative determinants of EC is displayed in
Table 7. Cases had a slightly earlier age at menarche (mean,
12.8 years) than did controls (mean, 13.0 years). EC risk was
nearly doubled among nulliparous women (P = 0.03) and
menopause occurred later in life among cases than it did among
controls (P = 0.01). Long-term use of replacement estrogens
for the menopause was associated with nearly a 5-fold increased
EC risk (P< 0.0001), while the use of a combination estrogen/
progesterone product was associated with a 50% reduction in
EC risk, which, however, was not statistically significant (P =
0.14). A history of diabetes and hypertension also was associ
ated with increased EC risk.
A multiple logistic model containing terms for age, race,
years of schooling, BMI, WHR, STR, and all the variables in
Table 7 was obtained. None of these results, nor those pertain
ing to BMI, WHR, and STR, changed meaningfully.
DISCUSSION
Table 6 Adjusted mean levels and differences between the means of the logarithm
of serum estrone, estradiol. and androstenedione according to case-control status
These results support the hypothesis that overall obesity is a
strong risk factor for EC. However, the excess risk is confined
largely to very obese women; a finding that has been reported
by others (1-3). The fact that obese women have high estrogen
levels provides biological plausibility for the hypothesis that
obesity is a cause of EC because increased stimulation of the
Case (/V = 67)
Control (N = 142)
26.9
68.3
0.28
(0.080)0.0005
Difference in loga
0.25
(0.068)0.0004Estradiol13.18.9
0.42
(0.088)<0.0001Androstenedione90.5
rithms (SE)
P valueEstrone36.4"
* Means arc adjusted for age, race, years of schooling, and the BMI.
570
Downloaded from cancerres.aacrjournals.org on June 16, 2017. © 1991 American Association for Cancer Research.
ENDOMETRIAL
CANCER AND FAT PATTERNS
endometrium by endogenous estrogens, unopposed by proges
terone, is believed to increase EC risk (17).
The nearly exclusive source of estrone among postmenopausal women is through aromatization of plasma androstenedione
in adipose tissue (9). High serum estrone and estradiol levels
among obese women have been reported previously (10, 11). In
the present study, serum estrogen levels among the controls
were elevated only for subjects in the highest BMI quartile; the
levels were similar across the first three quartiles. Thus, the
findings that excess EC risk and elevated serum estrogens were
restricted to the very obese is consistent with the belief that
obesity causes EC by means of excess production of endogenous
estrogens.
Among our controls there was little, or no, association be
tween obesity and serum androstenedione. In fact, subjects in
the upper BMI quartile had slightly lower levels than did those
in the first 3 quartiles. Judd et al. (18) found no association
between obesity and androstenedione (as well as testosterone),
whereas they did report a positive association between obesity
and endogenous estrogens. Thus, the relation between obesity
and increased EC risk probably is the result of increased pro
duction and level of estrogens rather than by increased availa
bility of androstenedione.
Serum estrone, estradiol, and androstenedione were higher
among our cases than among our controls, and these elevated
levels apparently were not due solely to an increased prevalence
of obesity among the cases. On the contrary, our findings
suggest that obesity plays a secondary role to estrone in the
etiology of EC. insofar as adjustment for estrone appreciably
attenuates the strength of the obesity and EC association.
Other investigators have examined serum estrone and estra
diol levels among women with EC. For example, in a study of
35 postmenopausal women with EC and an equal number of
controls matched for age and percentage of ideal weight, Judd
et al. (18) report serum estrone levels of 39 and 35 pg/ml for
cases and controls, respectively. In another study of postmen
opausal women. Judd et al. (10) found a mean serum estrone
level of 34 pg/ml among 16 women with EC and a mean of 28
pg/ml among 10 controls of similar weight. Calanog et al. (19)
reported a mean estrone level of 28 pg/ml among 14 women
with EC and a mean of 24 pg/ml among 5 women without EC.
These investigators considered these differences small and re
ported that they were not statistically significant. In the present
study, serum estrogen levels were higher among cases than
controls and these differences were independent of obesity and
were highly statistically significant. We emphasize that in each
study just described, serum estrone levels were higher among
women with EC. Thus, the overall epidemiological evidence
indicates that postmenopausal women with EC have elevated
endogenous estrogen levels.
Serum androstenedione levels also were higher among cases
than controls in this study. Judd et al. (18) reported a similar
finding in their study of 35 women with EC, but the finding
was not statistically significant. Calanog et al. (19) reported
that the plasma concentration of androstenedione and the in
stantaneous conversion of androstenedione to estrone was in
creased among postmenopausal women with EC. Judd et al.
(20) also had reported that there is an increase in ovarian
secretion of androstenedione and testosterone in postmenopau
sal women with EC. Hence, the data are consistent with the
belief that androstenedione levels are higher among women
with EC. However, since androstenedione is unrelated to obe
sity, its relation with EC probably is independent of obesity.
In contrast to our obesity findings, there was no association
between EC risk and the WHR after adjustment for BMI.
Furthermore, among the controls, serum estrone, estradiol, and
androstenedione were not independently related to the WHR
after BMI adjustment. Thus, these findings indicate that an
upper versus a lower fat pattern has no effect on EC independent
of overall obesity.
We believe that the lack of association between EC and the
WHR in this study is a valid observation and does not reflect
bias. First, the distribution of BMI among our controls is nearly
identical to that of women of comparable age in the general
population (13). This observation, coupled with our finding of
the anticipated relation between EC and obesity, provides evi
dence that our study is unbiased with respect to an evaluation
of BMI. Therefore, it probably also is unbiased with respect to
the WHR. An absence of association between the WHR and
EC could result from inaccurate anthropométriemeasurements.
However, since these measurements were done by one techni
cian with considerable experience in anthropometry, this, too,
is unlikely. Furthermore, we evaluated the association between
a self-reported history of diabetes and hypertension and the
WHR (adjusting for BMI) and found a strong, positive relation
for both. These findings strongly support the opinion that our
WHR measurements were valid. These findings did not change
when the analysis was restricted to the those cases for whom
the anthropométrie measurements were obtained preoperatively.
There are few other epidemiological studies of EC and body
fat distribution. In a nested case-control study of 63 incident
EC cases identified in a follow-up study, Folsom et al. (21)
reported a statistically significant positive association between
EC and the WHR which disappears after adjustment for BMI.
In a study of 10 EC cases, Lapidus et al. (22) also found no
association with the WHR. On the other hand, in a small casecontrol study (46 EC cases) from Baltimore, Elliott et al.
concluded that there was a BMI-independent effect of the WHR
on EC risk.4
In contrast to these negative EC findings, there apparently is
an independent association between fat pattern and breast can
cer. Ballard-Barbash et al. (23) reported that women in the
highest quartile of a central adiposity index in the Framingham
study experienced about an 80% increase in breast cancer
incidence compared with women in the first 3 quartiles. The
association was independent of overall obesity. Folsom et al.
(24) also found a direct, independent association between breast
cancer incidence and the WHR.
The data concerning sex hormones and fat pattern are com
plicated. In a study of 29 morbidly obese premenopausal
women. Kirschner et al. (25) reported that UBSO subjects had
higher testosterone plasma levels and production rates and had
lower levels of sex hormone-binding globulin. Evans et al. (26)
also reported that UBSO premenopausal women had increased
serum testosterone levels, while Folsom et al. (21) found lower
sex hormone-binding globulin levels in UBSO postmenopausal
women. Thus, the data indicate that UBSO women have excess
androgens in the form of free testosterone. On the other hand.
Kirschner et al. (25), Evans et al. (27), and we did not find
higher plasma levels of androstenedione among UBSO women.
With respect to estrogens. Kirschner et al. found higher plasma
levels of estradiol, but not estrone, among their UBSO subjects.
' E. A. Elliott. G. M. Mátanosla. N. B. Rosenshein, F. C. Grumbine. and E.
L. Diamond. Body fat patterning in women with endometrial cancer. Submitted
for publication. 1990.
571
Downloaded from cancerres.aacrjournals.org on June 16, 2017. © 1991 American Association for Cancer Research.
ENDOMETRIAL
CANCER AND FAT PATTERNS
We, too, found little difference between serum estrone levels
for UBSO versus LBSO but did find about a 30% higher
estradiol level (P = 0.09) among UBSO subjects.
Our finding of a positive, independent association between
EC and STR, in the absence of any such relation with WHR,
was unanticipated. STR is only weakly correlated with the
WHR (r = 0.21) and is virtually uncorrelated with BMI (r =
0.005) in this study. Thus, the STR index apparently is meas
uring an aspect of body fat distribution different from that
measured by the WHR. The WHR is an index of upper versus
lower body fat, whereas the STR is an index of central versus
peripheral body fat. Most epidemiological studies of the meta
bolic complications of body fat distribution have used the WHR
as the fat pattern index. Although some investigators have used
the STR or other indices of central versus peripheral body fat
(28, 29), the WHR and these other indices frequently are used
interchangeably. They all are thought to measure metabolically
active intrabdominal fat (30, 31). A distinction between these
different types of fat patterns perphaps is not important for
diseases such as diabetes and hypertension which are strongly
associated with fat distribution, but the distinction may be
important for cancers, for which the relationship with fat pat
tern is likely to be weaker.
In summary, in this study we found a positive association
between EC and overall obesity which appears to be due to high
serum estrogen levels among the obese. There was no associa
tion between EC and the WHR and little, or no, association
between the WHR and the sex hormones. Although there was
no association with UBSO, a statistically significant excess EC
risk was found for central obesity. This excess risk was not
explained by high estrogen levels. In the future, studies of cancer
in relation to body fat distribution should consider additional
measures of fat pattern other than the WHR.
ACKNOWLEDGMENTS
The authors thank Cathy Day for her considerable help with this
study and Dr. Robert Kleinstein for allowing us access to patients at
the optometry clinic.
REFERENCES
1. Wynder, E. L.. Escher. G. C., and Manici. N. An epidemiológica! investiga
tion of cancer of the endometrium. Cancer (Phila.), 19: 489-520, 1966.
2. Elwood. J. M., Cole. P.. Rothman. K. J.. and Kaplan. S. D. Epidemiology
of endometrial cancer. J. Nati. Cancer Inst.. 59: 1055-1060. 1977.
3. Kelsey, J. L.. LiVolsi. V. A., Holford. T. R.. Fischer. D. B., Mostow. E. D.,
Schwartz, P. E.. O'Conner, T., and White, C. A case-control study of cancer
of the endometrium. Am. J. Epidemic].. 116: 333-342, 1982.
4. Lapidus. L.. Bengtsson. C., Larsson, B.. Pennert, K., and Rybo. E. Distri
bution of adipose tissue and risk for cardiovascular disease and death: a 12
year follow-up of participants in the population study of women in Gothen
burg. Sweden. Br. Med. J., 289: 1257-1261. 1984.
5. Hartz, A. J., Rupley, D. C., and Rimm, A. A. The association of girth
measurements with disease in 32.856 women. Am. J. Epidemiol.. 119: 7180. 1984.
6. Vague. J. The degree of masculine differentiation of obesities: a factor
determining predisposition to diabetes, atherosclerosis, gout, and uric calculous disease. Am. J. Clin. Nutr.. 4: 20-34. 1956.
7. Van Gaal. L.. Rillaerts, E.. Creten. \V.. and De Leeuw, I. Relationship of
body fat distribution pattern to atherogenic risk factors in NIDDM. Prelim
inary results. Diabetes Care. //: 103-106, 1988.
8. Kissebah, A. H.. and Peiris, A.,N. Biology of regional body fat distribution
to non-insulin-dependent diabetes mellitus. Diabetes Metab. Rev.. 5:83-109,
1989.
9. MacDonald, P. C., and Siiteri, P. K. The relationship between the extraglandular production of estrone and the occurrence of endometrial neoplasia.
Gynecol. Oncol., 2: 259-263. 1974.
10. Judd. H. L.. Lucas. W. E.. and Yen. S. S. Serum 17 beta-estradiol and
estrone levels in postmenopausal women with and without endometrial
cancer. J. Clin. Endocrinol. Metab., 43: 272-278. 1976.
11. Davidson. B. J.. Gambone. J. C.. Lagasse. L. D., Castaido. J. V\'., Hammond.
G. L., Siiteri, P. K., and Judd, H. L. Free estradiol in postmenopausal women
with and without endometrial cancer. J. Clin. Endocrinol. Melab.. 52: 404408, 1981.
12. Jelliffe. D. B. The assessment of the nutritional status of the community.
WHO Monogr. Ser.. 53: 50-78. 1966.
13. National Center for Health Statistics. Obese and Overweight Adults in the
United States. National Health Survey. Series 11. No. 230. DHHS publica
tion (PHS) 83-1680. Washington. DC: Government Printing Office. 1983.
14. Hiramatsu. R., Voshidak. K., and Sato, T. A body measurement to evaluate
the pattern of fat distribution in central obesity. A screening and monitoring
technique for Cushing's syndrome. J. Am. Med. Assoc., 250: 3174-3178.
1983.
15. Hosmcr, D. W.. and Lcmcshow. S. Applied Logistic Regression. New York:
John Wiley and Sons, 1989.
16. Snedecor. G. W.. and Cochran, W. Statistical Methods. Ed. 6. Ames, IA:
Iowa State University Press, 1967.
17. Key, T. J..and Pike, M. C. The dose-effect relationship between "unopposed"
estrogens and cndometrial mitotic rate: its central role in explaining and
predicting cndometrial cancer risk. Br. J. Cancer. 57: 205-212. 1988.
18. Judd, H. L.. Davidson. B. J.. Frumar, A. M., Shamonki. I. M., Lagasse, L.
D., and Ballon, S. C. Serum androgens and estrogens in postmenopausal
women with and without endometrial cancer. Am. J. Obstet. Gynecol.. 136:
859-871. 1980.
19. Calanog. A.. Sail. S. S.. Gordon, G. G.. and Southern. A. L. Androstencdione
metabolism in patients with cndometrial cancer. Am. J. Obstct. Gynecol..
129: 553-556. 1977.
20. Judd, H. L., Lucas, W. E.. and Yen, S. S. Effect of oophorectomy on
circulating testosterone and androstenedionc levels in patients with endo
metrial cancer. Am. J. Obstet. Gynecol.. 118: 793-798, 1974.
21. Folsom, A. R., Kaye, S. A.. Potter, J. D., and Prineas, R. J. Association of
incident carcinoma of the endometrium with body weight and fat distribution
in older women: early findings of the Iowa women's health study. C'anccr
Res.. 49:6828-6831. 1989.
22. Lapidus, L.. Helgesson. O., Merck, C.. and Bjorntorp, P. Adipose tissue
distribution and female carcinomas. A 12-year follow-up of participants in
the population study of women in Gothenburg, Sweden. Int. J. Obesity. 12:
361-368. 1988.
23. Ballard-Barbash. R.. Schatzkin, A., Carter, C. L.. Kannel, W. B., Kreger, B.
E.. D'Agostino. R. B.. Splansky, G. L., Anderson, K. M.. and Helsel. W. E.
Body fat distribution and breast cancer in the Framingham study. J. Nati.
Cancer Inst.. 82: 286-290. 1990.
24. Folsom. A. R.. Kaye. S. A.. Prineas. J. D.. Potter. J. D., Gapstur. S. M.. and
Wallace, R. B. Increased incidence of carcinoma of the breast associated with
abdominal adiposity in post menopausa! women. Am. J. Epidemiol.. 131:
794-803. 1990.
25. Kirschner. M. A.. Samojlik, E.. Drejka. M.. Schneider. G., and Erte!, N.
Androgen-estrogen metabolism in women with upper body versus lower body
obesity. J. Clin. Endocrinol. Metab.. 70:473-479. 1990.
26. Evans, D. J., Barth, J. H.. Burke, C. W. Body fat topography in women with
androgenD. excess.
Int. J. Obesity,
12: 157-162,
27. Evans.
J., Hoffman,
R. G.." Kalkhoff.
R. K..1988.
and Kissebah. A. H. Rela
28.
29.
30.
31.
tionship of androgenic activitiy to body fat topography, fat cell morphology,
and metabolic observations in premenopausal women. J. Clin. Endocrinol.
Metab., 57:304-310, 1983.
Foster, C. J., Weinsier, R. D.. Birch. R., Norris. D. J., Bernstein, R. S..
Wang. J., Pierson, R. N.. and Vanitallie, T. B. Obesity and serum lipids: an
evaluation of the relative contribution of body fat and fat distribution to lipid
levels. Int. J. Obesity.//: 151-161. 1987.
Baumgartner. R. N.. Roche, A. F.. Chumlea. W. C.. Siervogel, R. M.. and
Glueck. C'. J. Fatness and fat patterns: associations with plasma lipids and
blood pressure in adults. 18 to 57 years of age. Am. J. Epidemiol.. 126: 614628. 1987.
Kissebah. A. H.. Vydelingum. N.. Murray. R., Evans, D.. Hart/. A.. Kalkhoff.
R. K.. and Adams. P. W. Relation of body fat distribution to metabolic
complications of obesity. J. Clin. Endocrinol. Metab.. 54: 254-259, 1982.
Kratkiewski. M.. Bjorntorp. P..Sjostrom. L.. and Smith. I'. Impact of obesity
on metabolism in men and women: importance of regional adipose tissue
distribution. J. Clin. Invest.. 72: 1150-1162. 1983.
572
Downloaded from cancerres.aacrjournals.org on June 16, 2017. © 1991 American Association for Cancer Research.
Endometrial Cancer, Obesity, and Body Fat Distribution
Harland Austin, J. Max Austin, Jr., Edward E. Partridge, et al.
Cancer Res 1991;51:568-572.
Updated version
E-mail alerts
Reprints and
Subscriptions
Permissions
Access the most recent version of this article at:
http://cancerres.aacrjournals.org/content/51/2/568
Sign up to receive free email-alerts related to this article or journal.
To order reprints of this article or to subscribe to the journal, contact the AACR Publications
Department at [email protected].
To request permission to re-use all or part of this article, contact the AACR Publications
Department at [email protected].
Downloaded from cancerres.aacrjournals.org on June 16, 2017. © 1991 American Association for Cancer Research.