Download as a PDF

American Journal of Epidemiology
Copyright O 1996 by The Johns Hopkins University School of Hygiene and Public Health
AD rights reserved
Vol. 143, Mo. 7
Printed in U.SA.
Body Size and Breast Cancer Risk among Women under Age 45 Years
Christine A. Swanson,1 Ralph J. Coates,2 Janet B. Schoenberg,3 Kathleen E. Malone,4 Marilie D. Gammon,5
Janet L. Stanford,4 Irwin J. Shorr,6 Nancy A. Potischman,1 and Louise A. Brinton1
anthropometry; body height; body weight; breast neoplasms
In etiologic studies of breast cancer, anthropometric
measurements may serve as useful biologic markers of
environmental factors, including contemporary and
past diets. In recent years, height has reemerged as a
breast cancer risk factor (1). Frame size, another index
of skeletal dimension, has also been associated with
breast cancer risk (2, 3). As hypothesized by others (4,
5), height may be related to mammary gland mass and,
by inference, to the niimber of ductal stem cells at risk
of transformation. Ecologic studies suggest an association of both height and breast size with breast cancer
incidence (5). Adult stature may reflect early environmental influences such as energy intake during childRecefved for publication January 24, 1995, and in final form
January 24, 1996.
Abbreviation: Cl, confidence Interval.
1
Environmental Epidemiology Branch, National Cancer Institute,
Bethesda, MD.
2
Department of Epidemiology, Rollins School of Public Health,
Emory University, Atlanta, GA.
3
Special Epidemiology Program, New Jersey State Department
of Hearth, Trenton, NJ.
4
Fred Hutchinson Cancer Research Center, Seattle, WA.
6
Division of Epidemiology, Columbia University School of Public
Heath, New York, NY.
6
Olney, MD.
Reprint requests to Dr. Christine A. Swanson, Nutritional Epidemiology Section, Environmental Epidemiology Branch, Division of
Cancer Epidemiology and Genetics, National Cancer Institute, National Institutes of Hearth, EPN Room 443, 6130 Executive Blvd.,
MSC 7374, Bethesda, MD 20892.
hood and adolescence. Sitting height may be particularly sensitive to early diet, and relative sitting height
(i.e., the sitting height-to-standing height ratio) has
been proposed as an indirect index of preadolescent
nutrition. Briefly, long-waisted women tend to experience early menarche, and early maturation may reflect increased nutrition (i.e., excess energy intake)
during childhood (6).
The relation of obesity to breast cancer risk has been
considered in numerous studies. For many years, obesity was widely regarded as a risk factor for breast
cancer. More recently, excess weight has been viewed
as a relatively modest risk factor (7). Furthermore, risk
associated with obesity is, in most studies, limited to
older postmenopausal women. In younger women
(e.g., those younger than age 45 years), obesity appears to be inversely related to risk of the disease.
Initially, the apparent protective effect of obesity in
younger women was attributed in large part to detection bias (8, 9). Later studies indicated that the inverse
association was not adequately explained by difficulties in detection of tumors in heavy women (10, 11).
The protective effect of excess weight may be related
to hormonal changes associated with obesity (e.g.,
reduced exposure to progesterone) (12). Studies comprised mainly of postmenopausal women (13—16)
show a direct relation of upper-body or central adipos-
Downloaded from http://aje.oxfordjournals.org/ at Pennsylvania State University on February 26, 2014
In a multicenter population-based case-control study that included 1,588 cases and 1,394 controls less than
age 45 years, the authors examined the relation of adult body size and breast cancer risk among young
women. Breast cancer patients and healthy controls were identified in Atlanta, Georgia; Seattle/Puget Sound,
Washington; and central New Jersey. Cases were newly diagnosed with in situ or invasive breast cancer during
the period of May 1, 1990, through December 31, 1992. Anthropometric variables thought to reflect early
environmental factors (e.g., height, sitting height, frame size), obesity, and body fat distribution were measured
directly. Height, but not sitting height or frame size, was a breast cancer risk factor. Risk of the disease was
increased 46 percent among women in the fourth quartile of height (>167 cm) compared with women in the
first quartile (<159 cm). Body weight, but not body fat distribution, was related to breast cancer risk. Risk of
the disease was 35 percent lower among women in the highest quartile of Quetelet index (>28.8 kg/m2)
compared with women in the lowest quartile (<22.0 kg/m2). Risk of the disease was increased about 2.1-fold
(95 percent confidence interval 1.2-3.8) among women who were thin and tall compared with women who
were heavy and short. Thus, breast cancer risk was increased substantially among younger women with a
linear body type. Am J Epidemiol 1996; 143:698-706.
Body Size and Breast Cancer Risk in Young Women
ity and breast cancer risk independent of generalized
obesity. To our knowledge, there is only one report of
the relation of body fat distribution and breast cancer
risk among premenopausal women (16).
Few studies of body size and breast cancer risk have
included large numbers of women diagnosed at a
young age. Furthermore, most studies of younger
women have been limited to the assessment of height
and weight. To assess fully the relation between body
size and breast cancer risk among young women, we
undertook a case-control study that included a variety
of anthropometric measurements.
This population-based case-control study (17) was
conducted in three geographic areas covered by cancer
registries: the metropolitan areas of Atlanta, Georgia;
Seattle/Puget Sound, Washington; and five counties in
central New Jersey. The present analysis is based on
women aged 20-44 years who were newly diagnosed
with in situ or invasive breast cancer during the period
May 1, 1990, through December 31, 1992. Hospital
records of eligible patients were abstracted to document details on the clinical and pathologic characteristics of the diagnosed breast cancers. Controls were
frequency matched by geographic area and age to the
expected distribution of cases and were identified
through random digit dialing (18). A 90.5 percent
response rate to the telephone screener was obtained
from 16,254 telephone numbers assessed as residential.
Interviews were obtained from 1,669 of the 1,940
eligible cases (86.0 percent) and 1,505 of the 1,912
eligible controls (78.7 percent). The major reasons for
noninterview were subject refusals (6.6 percent of the
cases and 12.9 percent of the controls) and physician
refusals (5.8 percent of the cases). Among the controls, the overall response rate was 71.2 percent (telephone screener rate times the interview response rate).
Structured in-person interviews, which lasted a median time of 71 minutes, collected detailed information
regarding demographic factors; reproductive and menstrual history; contraceptive behavior; use of exogenous hormones; medical and screening history; adolescent diet; alcohol consumption; physical activity;
smoking; occupation; family history of cancer, and
certain lifestyle factors and opinions about cancer causation. In addition, participants were asked, either at
the time of the personal interview or subsequently, to
complete a 100-item dietary questionnaire.
Interviewers at each of the three centers received
standardized training in anthropometric techniques. At
least twice during the course of the study, the anthropometry instructor evaluated the quality of the data
Am J Epidemiol
Vol. 143, No. 7, 1996
collected at each study center. Accuracy and reliability
of the measurements were determined by having the
anthropometry instructor and interviewers perform repeated measurements on volunteer subjects. Measurements of the interviewers were compared with those of
the instructor, and problems were corrected on site.
Measuring equipment was standardized regularly
throughout the study. As expected, some measurements were more reliable than others. For example,
accuracy and reliability were greatest for height and
weight and least for the skinfold determinations.
The following anthropometric measurements were
made: standing and sitting height were measured to the
nearest 0.1 cm with a custom-made stadiometer using
a height-measuring rod (Seca Corp., Columbia Maryland); elbow and wrist widths to the nearest 0.1 cm
with a caliper (Holtain bicondylar vernier caliper, Holtain Ltd., Seritex Inc., Carlstadt, New Jersey); weight
to the nearest 0.2 kg with a portable digital scale
(Integra 815, Seca Corp., Columbia Maryland); waist
and hip circumferences to the nearest 0.1 cm with a
flexible measuring tape (Lufkin steel measuring tape,
Seritex, Inc., Carlstadt, New Jersey); mid-upper arm
circumference to the nearest 0.1 cm with a nonstretch
plastic tape (insertion circumference tape, Ross Laboratories, Columbus Ohio); and subscapular and triceps skinfolds to the nearest 0.2 mm with skinfold
calipers (Holtain Tanner-Whitehouse skinfold caliper,
Holtain Ltd., Seritex, Inc., Carlstadt, New Jersey).
Mid-upper arm circumference was measured at the
midpoint of the arm between the acromion and olecranon processes. Waist circumference was measured
just superior to the iliac crest of the pelvis; the measurement site was often at the level of the umbilicus.
Hip circumference was measured to include the maximum extension of the buttocks and usually included
underclothing plus a light loose-fitting garment. Triceps skinfolds were measured on the posterior midline
of the arm at the same level as the mid-upper arm
circumference. The subscapular skinfold site was
identified by locating the area just below the inferior
angle of the scapula. Skinfolds, mid-upper arm circumference, and elbow and wrist widths were measured on the right side of the body unless the subject
reported surgery of the right breast or any past or
present injury to the right arm that resulted in pain,
swelling, or deformity. Because of concerns that treatment might affect measurements, cases were asked if
they had received radiation therapy, chemotherapy, or
any other drug therapy since their surgery. Cases were
identified through rapid ascertainment systems, and 84
percent of cases were interviewed and measured
within 6 months of diagnosis.
Downloaded from http://aje.oxfordjournals.org/ at Pennsylvania State University on February 26, 2014
MATERIALS AND METHODS
699
700
Swanson et al.
RESULTS
The median age of both cases and controls was 40
years. The two groups did not differ by race, with 80
percent being white, 15 percent black, and 5 percent
other races. The major risk factors for breast cancer
were a family history of breast cancer, a previous
breast biopsy, limited number of births, a late age at
first birth, and alcohol consumption (table 1). Disease
risk was lower among postmenopausal compared with
premenopausal women. Early age at menarche and
oral contraceptive use were relatively weak risk factors. Cigarette smoking and education were not related
TABLE 1. Distribution of potential risk factors and
associated relative risks of breast cancer among esses and
controls less than age 45 years, United States, 1990-1092
No. of No. Of
cases controls
Risk
factor
RR.
95% Cl*
First-degreerelativewith
breast cancer
No
Yes
1,361
227
1,301
83
1.0
2JJ2
1.8-3.0
Previous breast biopsy
No
Yes
1,436
152
1,305
89
1.0
1.49
1.1-2.0
Menstrual statust
Premenopausal
Postmenopausal
1,420
168
1,197
195
1.0
0.67
0.5-0.8
No. of births
£4
3
2
1
0
83
212
581
320
392
116
230
479
280
289
1.0
1.31
1.73
1.65
2.08
0.9-1.8
1.3-2.4
1.2-2.3
1.6-2.9
Age at first birth* (years)
<20
20-24
25-29
£30
214
353
349
279
241
346
317
201
1.0
1.14
1.26
1.55
0.9-1.4
1.0-1.6
1.2-2.0
Age at menarche (years)
£14
13
12
S12
283
426
494
383
289
421
372
310
1.0
1.03
1.35
1.26
0.8-1.3
1.1-1.7
1.0-1.6
372
1,212
391
1,003
1.0
1.28
1.1-1.5
684
292
387
109
112
656
245
339
97
55
1.0
1.15
1.10
1.07
1.96
0.9-1.4
0.9-1.3
0.8-1.4
1.4-2.8
Cigarette smokerll
No
Yes
827
761
705
689
1.0
0.94
0.8-1.1
Education
High school or less
Technical school
Some college
Cotege graduate
Postgraduate work
419
106
419
395
249
382
109
389
327
187
1.0
0.89
0.99
1.12
1.20
0.7-1.2
0.8-1.2
0.9-1.4
0.9-1.5
Oral contraceptive user
No
Yes (£6 months)
Alcohol use§ (drinks/week)
Nondrtnker
1-6.9
7.0-13.9
£14
• Relative rfsks (RR) and 9 5 % confidence Intervals (Cl) adjusted for age
and study site. For some variables, the number of observations does not
equal 2,982 (1,588 cases and 1,394 controls) because of missing values.
t Women who had not had a menstrual period In 6 months were defined
as postmenopausal: natural and surgical menopause were combined.
$ Restricted to parous women.
j Drinkers were defined as women who had drunk more than 12 drinks of
alcoholic beverages In their Bves and also had drunk at least once a month for
8,months or longer.
U Smokers were defined as women who had smoked 100 cigarettes or
mors In their Bves and who had ever smoked on a regular basis for 6 months
or longer.
to risk of the disease. Potential risk factors were included in the multivariate analyses if they were related
to disease risk and were clearly associated with any of
the anthropometry variables. In addition to age and
study site as matching variables, we included race, age
at menarche, oral contraceptive use, parity, and alcoAm J Epidemiol
Vol. 143, No. 7, 1996
Downloaded from http://aje.oxfordjournals.org/ at Pennsylvania State University on February 26, 2014
With the exception of height, sitting height, and
weight, measurements were taken twice, and a third
measurement was made if the difference of the first
two measurements exceeded a predetermined limit.
Obvious outliers were removed before the means of all
replicates were calculated. Sixty cases and 111 controls provided either no anthropometry information or
so few measurements that neither body mass index nor
body fat distribution could be calculated (seven of
these cases and 48 controls were excluded because
they were pregnant or less than 7 months postpartum).
Because controls were identified through telephone
sampling, 21 cases without residential telephones were
eliminated from the analysis. This analysis focused on
1,588 cases and 1,394 controls.
To assess obesity, we evaluated measured weight,
the more commonly reported Quetelet body mass
index (kg/m2), and a body mass index for women
(kg/m 15 ) recommended by the National Center of
Health Statistics (19). The three variables were highly
correlated and similarly related to breast cancer risk;
only Quetelet index and measured weight are reported.
To assess fat distribution patterns, two indices were
derived. The waist-to-hip circumference ratio was
used as an index of upper-body or android obesity
(20). The subscapular-to-triceps skinfold ratio provided an index of central obesity (21), a measure of
subcutaneous fat on the trunk of the body versus the
periphery.
Logistic regression was used to obtain relative risk
estimates (odds ratios) and their 95 percent confidence
intervals (22). Multivariate logistic regression was
used to adjust for potential confounders and to test the
statistical significance of interaction terms. In analyses
involving stage of diagnosis as an outcome, we used
polychotomous logistic regression to compare each
case group simultaneously with the controls (23).
Tests for trend in the logistic analyses were obtained
by categorizing the exposure variable and treating the
scored variables as continuous after eliminating unknown values.
Body Size and Breast Cancer Risk in Young Women
hoi consumption as potential confounders. Addition of
other variables (e.g., family history, smoking, and
education) did not materially alter the risk estimates.
Age was modeled as a continuous variable, but results
were similar when it was included as a categorical
variable.
Height and one measure of frame size were associated with risk (table 2). Height was directly related to
breast cancer risk. Women taller than 167 cm had a 46
percent greater risk compared with women less than
159 cm tall. While a high sitting-to-standing height
ratio was associated with early onset of menarche
(data not shown), breast cancer risk was reduced
somewhat among long-waisted women (p for trend =
0.08). Elbow width was not a breast cancer risk factor.
However large frame size, as assessed by wrist width,
was protective. Risk was 22 percent lower among
women in the fourth quartile of wrist width (>5.3 cm)
compared with women in the lowest category (<4.9
cm).
Weight, adiposity, and body fat distribution
Weight and adiposity, but not body fat distribution,
were related to breast cancer risk (table 3). Current
TABLE 2. Relative risks for breast cancer by quartiles of skeletal dimensions among women less than
age 45 years, United States, 1990-1992
Height (cm)
<159
159-163
164-167
>167
Sitting-to-standing height ratio
<0.518
0.518-0.529
0.530-0.539
>0.539
Elbow width (cm)
<6.0
6.0-6.2
6.3-6.5
>6.5
Wrist width (cm)
<4.9
4.9-5.1
5.2-5.3
>5.3
No. of
cases
No. of
controls
329
429
387
443
341
350
352
351
382
319
409
392
372
341
354
345
339
428
306
335
318
408
486
399
406
370
413
362
290
358
356
390
RR*
95% C I '
pfor trend
1.0-1.6
1.0-1.5
1.2-1.8
0.004
0.99
0.87
0.84
0.8-1.2
0.7-1.1
0.7-1.1
0.079
1.0*
1.09
1.02
1.19
0.8-1.4
0.8-1.3
0.9-1.5
0.269
1.0*
0.82
0.74
0.78
0.7-1.0
0.6-0.9
0.6-1.0
0.034
1.0t
1.28
1.20
1.46
1.0t
* Relative risks (RR) and 95% confidence intervals (CI) adjusted for age, study site, race, menarche, oral
contraceptive use, parity, and alcohol consumption. For some variables, the number of observations does not
equal 2,982 (1,588 cases and 1,394 controls) because of missing values.
t Further adjusted for weight
t Further adjusted for weight and height
Am J Epidemiol
Vol. 143, No. 7, 1996
Downloaded from http://aje.oxfordjournals.org/ at Pennsylvania State University on February 26, 2014
body weight (i.e., weight measured at the interview)
was inversely associated with breast cancer risk. Risk
was reduced 28 percent among women who weighed
more the 77.0 kg compared with women who weighed
less than 58.6 kg. Likewise, women in the highest
quartile of Quetelet index (>28.8 kg/m2) had a 35
percent lower breast cancer risk compared with thin
women (<22.0 kg/m2). The combined effect of height
and weight was greater than that of either variable
alone. Women who were tall (highest quartile) and
thin (lowest quartile) had a relative risk of 2.1 (95
percent confidence interval (CI) 1.2-3.8) compared
with women who were short and heavy. Among extremely heavy women, 24 cases and 40 controls who
weighed more than 110 kg, breast cancer risk was
reduced 62 percent (relative risk = 0.38, 95 percent CI
0.2-0.7). A comparable reduction in risk was not
observed, however, when similar extremes of Quetelet
index were examined.
After adjustment for weight, neither waist nor hip
circumference was related to risk of breast cancer
(data not shown). A predominance of upper body fat,
as measured by the waist-to-hip circumference ratio,
was not associated with risk. Triceps skinfold thickness was not related to breast cancer risk (data not
shown), but subscapular skinfold thickness was in-
Skeletal dimensions
Characteristic
701
702
Swanson et al.
TABLE 3. Relative risks for breast cancer by quartiles of weight, adiposity, and body fat distribution
among women less than age 45 years, United States, 1990-1992
Nad
No. erf
controls
RR»
95%CI»
p tor trend
Weight (kg)
<58.6
58.6-66.0
66.1-77.0
>77.0
442
395
379
372
347
349
347
351
1-Ot
0.81
0.75
0.72
0.7-1.0
0.6-0.9
0.6-0.9
0.003
Quetalet index (kg/m*)
<22.0
22.0-24.6
24.7-28.8
>28.8
481
373
388
346
347
350
348
349
1.0
0.74
0.77
0.65
0.6-0.9
0.6-0.9
0.5-0.8
0.003
Waist-to-hlp circumference ratio
< 0.753
0.753-0.799
0.800-0.858
>0.858
399
429
402
358
342
354
347
351
1.0*
1.06
1.09
0.95
0.9-1.3
0.9-1.4
0.8-1.2
0.729
Subscapular sWntoJd (mm)
<12.2
12.2-17.3
17.4-25.2
>25.2
422
407
378
320
334
343
342
340
1.0*
0.97
0.93
0.74
0.8-1.2
0.7-1.2
0.5-1.0
0.078
Subscapular-to-triceps skinfold
ratio
<0.664
0.664-0.832
0.833-1.03
>1.03
392
384
411
339
339
338
340
340
1.0*
1.02
1.08
0.93
0.8-1.3
0.9-1.4
0.7-1.2
0.702
* Relative risks (RR) and 95% confidence intervals (Cl) adjusted for age, study site, race, menarche, oral
contraceptive use, parity, and alcohol consumption. For some variables, the number of observations does not
equal 2,982 (1,588 cases and 1,394 controls) because of missing values.
t Further adjusted for height
* Further adjusted for weight and height
versely related to risk. Risk was 26 percent lower (95
percent CI 0.5-1.0) among women in the highest quartile of subscapular skinfold thickness (>25.2 mm)
compared with women in the lowest quartile (<12.2
mm). The subscapular-to-triceps skinfold ratio, an indicator of central versus peripheral obesity, showed no
relation to disease risk.
We explored the possibility mat the relation of body
fat distribution to breast cancer risk might be modified
by other anthropometric variables or risk factors. We
did not identify any subgroups, including levels of
weight, Quetelet index, or family history of breast
cancer, in which a predominance of upper-body fat or
central adiposity significantly affected breast cancer
risk.
For each anthropometry variable, we examined the
risk relation by strata of breast cancer risk factors
shown in table 1. The effects of height were remarkably consistent across strata of other risk factors. The
associations of weight and Quetelet index to breast
cancer risk were more variable, but there was no clear
evidence of effect modification by other risk factors.
The inverse associations of wrist width and subscapular skinfold thickness with risk were clearly limited
to nulliparous women. Relative risks of breast cancer
across increasing quartiles of wrist width were 0.7,
0.6, and 0.4 (p for trend = 0.0016). The risk estimates
for increasing quartiles of subscapular skinfold thickness were 0.7, 0.5, and 0.3 (p for trend = 0.0003).
Among parous women, the estimates for increasing
quartiles of wrist width were 0.9, 0.8, and 0.9. The
corresponding values for subscapular skinfold thickness were 1.1, 1.1, and 1.0.
Height, weight, and breast size
In other analyses, we focused on effects of breast
size, height, and weight. Among controls, we crossAm J Epidemiol
Vol. 143, No. 7, 1996
Downloaded from http://aje.oxfordjournals.org/ at Pennsylvania State University on February 26, 2014
cases
Characteristic
Body Size and Breast Cancer Risk in Young Women
Adiposity and the menstrual cycle
We explored the possibility that the protective effect
of excess weight was related to its association with
menstrual factors (e.g., onset and regularity of menses). The reduced breast cancer risk associated with
greater weight could not be explained by age at menarche because the heavier women had an earlier age at
menarche, which was associated with increased risk.
The women were asked when after menarche their
periods became "regular or predictable." Women who
reported that they could never predict when their menstrual period would start were classified as having
irregular cycles. Irregularity was used as a crude indicator of anovulatory cycles. Risk of breast cancer was
34 percent lower (95 percent CI 0.5-0.8) among
women whose menstrual cycles never became regular
compared with women who reported regular menstrual
cycles. While controls with irregular menstrual cycles
were heavier than controls who were regular, both
adiposity and regularity of menstrual cycles predicted
breast cancer risk (table 5).
Detection bias
We examined the issue of detection bias. Thirty-six
percent of cases were diagnosed with regional or distant tumors. The percentage of advanced tumors was
higher for women in the highest quartile of Quetelet
index compared with women in the lowest quartile (41
versus 31 percent). If tumors are more difficult to
detect in obese women, then one would expect the
protective effect of excess weight to be restricted to
small tumors. The inverse association of excess weight
and risk was, in fact, limited to in situ and local tumors
and was most pronounced among women with in situ
cancers (table 6). Next, we examined the risk relation
by the three main detection methods_(mammography,
routine self-examination, and accidental discovery by
the woman or her partner). This approach suggested
that the protective effect of excess weight could not be
attributed entirely to detection bias. One would expect
that obesity could not mask the disease if a relatively
sensitive method such as mammography were used to
detect tumors. Among women with in situ and local
cancers, the inverse association persisted when the
disease was detected by the most sensitive of the three
methods, mammography.
To evaluate the possibility of selection bias, we
compared the height and body mass index (kg/m2) of
a subgroup of eligible cases and controls who were
classified as nonrespondents. After physician refusals
were excluded, subjects who could not be interviewed
personally were contacted and asked to participate in a
5-minute telephone or mailed interview that included
questions about current height and weight 1 year previously. A total of 34 cases and 122 controls agreed to
participate. Among the "nonrespondents," cases were
taller than controls (164.8 vs. 163.2 cm, p = 0.20) and
had a slightly higher body mass index (24.3 vs. 24.0
kg/m2, p = 0.68).
TABLE 4. Relative risks for breast cancer by bra cup size according to level of Quetelet Index (kg/m2)
among women less than age 45 years, United States, 1990-1992
Quetelet Index 224.7
Quetelet Index <24.7
Bra
cup
size
No. of
cases
No. of
controls
RR*
95% CI*
A
B
C
D
201
436
166
43
191
343
132
22
1.0
1.26
1.23
1.95
1.0-1.6
0.9-1.7
1.1-3.4
No. of
cases
No. of
controls
RR*
95% CI*
60
326
217
130
49
309
214
117
1.0
0.81
0.78
0.84
0.5-1.2
0.5-1.2
0.5-1.3
* Relative risks (RR) and 95% confidence intervals (CI) adjusted for age, study site, race, menarche, oral
contraceptive use, parity, alcohol consumption, height, and weight (continuous). For some variables, the number
of observations does not equal 2,982 (1,588 cases and 1,394 controls) because of missing values.
Am J Epidemiol
Vol. 143, No. 7, 1996
Downloaded from http://aje.oxfordjournals.org/ at Pennsylvania State University on February 26, 2014
classified height and weight quartiles with bra cup size
(A, B, C, D). Height was somewhat related to breast
size. For example, cup size A was reported by 17.2
percent of controls and became less common across
increasing quartiles of height (19.8, 18.3, 16.0, and
15.8 percent). Weight was clearly related to bra size
(Spearman r = 0.37, p = 0.0001). The distribution of
cup size A across increasing quartiles of weight was
34.4, 20.5, 11.3, and 3.8 percent. After adjustment for
height, weight, and other potential confounders, the
risk estimates for increasing levels of bra cup size
were 1.1, 1.0, and 1.2 (p for trend = 0.37). When the
analysis was stratified by two categories of body mass
index (lower and upper quartiles of Quetelet index), a
pronounced association of breast size and risk was
observed in thin women (table 4), with bra cup size D
being associated with a twofold increased risk compared with thin women with cup size A. Bra size was
not a risk factor among heavier women.
703
704
Swanson et al.
TABLE 5. Relative risks for breast cancer by quartiles of Quetelet Index (kg/m2) according to regularity
of menstrual periods among women less than age 45 years, United States, 1990-1992
Quetelet
index
<22.0
22.0-24.6
24.7-28.8
>28.8
Ever regular
No. of
cases
NO. Of
controls
pp.,
446
347
364
320
321
321
312
310
1.0
0.75
0.81
0.68
Never regular
95% Cl*
0.6-0.9
0.7-1.0
0.5-0.8
No. of
cases
No. of
controls
RR*
95% Cl*
23
20
18
21
22
25
30
34
0.74
0.58
0.44
0.44
0.4-1.4
0.3-1.1
0.2-0.8
0.2-0.8
• Relative risks (RR) and 95% confidence Intervals (Cl) adjusted for age, study site, race, menarche, oral
contraceptive use, parity, alcohol consumption, and height For some variables, the number of observations does
not equal 2,982 (1,588 cases and 1,394 controls) because of missing values.
Method of detection
Quetelet
Index
(kg/m»)
Total
RR*
Routine self-examination
Mammography
Exposed
casest
Exposed
RR
Accidental discovery
Exposed
Exposed
RR
1.0
0.55
0.24*
0.37*
15
9
1.0
0.88
0.58
0.28
1.0
0.77
0.68*
0.71
83
63
56
59
1.0
0.77
0.64*
0.66*
83
66
54
55
1.0
0.69
0.96
0.63
70
48
67
46
1.0
0.90
1.02
1.27
51
46
52
65
RR
In situ
<22.0
22.0-24.6
24.7-28.8
>28.8
1.0
0.59*
0.54*
0.47*
80
51
46
41
1.0
0.44*
0.50*
0.51*
46
22
25
26
4
6
7
6
4
2
Local
<22.0
22.0-24.6
24.7-28.8
>28.8
1.0
0.75*
0.70*
0.63*
241
188
172
158
43
34
34
25
1.0
0.68
0.69
0.49*
Regional/distant
<22.0
22.0-24.6
24.7-28.8
>28.8
1.0
0.81
1.05
0.92
149
123
158
142
1.0
1.29
1.63
0.69
9
13
16
7
* Relative risks are adjusted for age and study site.
t Excludes 39 cases with unknown disease stage at diagnosis.
* The 95% confidence interval excludes unity.
DISCUSSION
It is unclear why height should emerge repeatedly as
a predictor of breast cancer, particularly in relatively
affluent population groups (9, 16, 24-28) unlikely to
have been exposed to energy-restricted diets during
critical growth periods. In this study, the association
was not explained by potential confounders such as
alcohol consumption, oral contraceptive use, parity, or
education. Ballard-Barbash (29) proposed that genetic
and environmental factors, including diet, may influence the hormones that regulate epiphysial closure and
thus attained height. Presumably, these hormones or
perhaps certain endocrine profiles also would be related to subsequent risk of breast cancer. Other investigators have hypothesized that adult height may be
related to mammary density or mammary gland size,
which, in turn, may be related to breast cancer risk (4).
In our investigation, the increased risk associated with
height did not reflect effects of mammary gland size as
estimated by bra size. It is possible, however, that bra
size is an inadequate proxy for mammary gland mass
except at the very low range of its distribution.
We observed a direct relation of breast size and
breast cancer risk in thin women. Breast size, as estimated by bra cup size, is not an accepted risk factor for
breast cancer. In fact, the majority of analytic epidemiologic studies show no association (5). The breast is
comprised mainly of adipose tissue and using bra cup
size as a proxy for mammary gland size undoubtedly
results in considerable misclassification. Perhaps misAm J Epidemiol
Vol. 143, No. 7, 1996
Downloaded from http://aje.oxfordjournals.org/ at Pennsylvania State University on February 26, 2014
TABLE 6. Relative risks for breast cancer by quartlles of Quetelet Index (kg/m*) according to stage of
disease and method of detection among women less than age 45 years, United States, 1990-1992
Body Size and Breast Cancer Risk in Young Women
Am J Epidemiol
Vol. 143, No. 7, 1996
seen in prospective studies with extended follow-up
(26, 27). It is not likely that the lower weight of cases
compared with controls was due to treatment effects
since weight gain rather than loss is a reported side
effect of adjuvant chemotherapy (37). If this were the
case, we would have underestimated the protective
effect of excess weight. We considered the possibility
that treatment effects might explain why we did not
observe an inverse association of weight and risk
among women with regional/distant disease. Ninety
percent of these women received chemotherapy prior
to the interview compared with 54 percent of women
with local disease and less than 1 percent of women
with in situ cancer. Cases were asked to recall their
weight 1 year prior to the interview. The difference
between recalled weight and interview weight indicated that women with regional/distant disease did not
gain more weight during the interval compared with
women with less advanced breast cancer (in situ, 2.4
kg; local, 2.3 kg; regional distant, 1.9 kg).
Similar to our results, waist-to-hip ratio was not associated with risk of breast cancer in premenopausal
women in a study by Bruning et al. (16), even though a
predominance of upper body fat has been associated with a biochemical profile (e.g., decreased sex
hormone-binding globulin, increased bioavailable estradiol, hypertriglyceridemia) thought to increase disease
risk (16, 38). Sellers et al. (39) observed that upper-body
obesity was associated with risk among postmenopausal
women with a family history of breast cancer. In our
study, family history did not act as an effect modifier.
In summary, we observed a positive association
between adult height and breast cancer risk among
women with early-onset disease. In agreement with
other reports, it appears that adult height can be a risk
factor in a population unlikely to have experienced
caloric or other nutrient restriction during critical
growth periods. We could only crudely assess the
relation of height, mammary gland size, and risk of
breast cancer, but the height effect could not be explained by an association with bra size. The risk estimates associated with the combined effects of tallness
and thinness suggest that surface area and metabolic
rate may be important. While weight data obtained
after diagnosis must be interpreted cautiously, our
study provides further evidence that breast cancer risk
is lower among heavy compared with thin women with
early-onset disease. The protective effect of excess
weight was not adequately explained by detection bias
and did not appear to be related either to disease or to
treatment effects. Young obese women may have a
reduced risk of breast cancer because they have more
anovulatory cycles and thus lower exposure to estrogen and progesterone. Our results did not support this
Downloaded from http://aje.oxfordjournals.org/ at Pennsylvania State University on February 26, 2014
classification is reduced when the analysis is restricted
to thin women. Interestingly, one of the few casecontrol studies to show a positive association of breast
size and risk was of Japanese women (30).
Increased sitting height has been proposed as an index
of early maturity and may reflect nutrition during childhood (6). Long-waisted women in our study did experience early menarche, but increased relative sitting height
was not related to increased disease risk. In fact, the
relative sitting height of cases was slightly less than that
of controls. A similar observation was made in two
previous studies (2, 3). If relative sitting height is associated with decreased risk of breast cancer, then the other
component of height (i.e., total height minus sitting
height) must be associated with slightly increased risk.
Perhaps, in future studies, attention should be focused on
leg length rather than sitting height.
In contrast to two previous studies comprised
mainly of postmenopausal women (2, 3), frame size
was not directly related to risk of breast cancer in our
study population of younger women. We cannot explain the observation of an inverse association between wrist width and disease risk or why the association was limited to nulliparous women.
In agreement with numerous studies conducted in
high-risk countries (8, 10, 31-35), we observed that
the risk of breast cancer among young women was
lower among heavy compared with lean women. We
also noted a dramatic reduction in risk among a small
number of extremely heavy women (>110 kg). The
effects of marked obesity should be explored in future
studies. To the extent that our assessment of menstrual
regularity (i.e., ever vs. never) reflected ovulation
(36), the protective effect of excess weight could not
be explained by increased frequency of anovulatory
cycles. Previous reports (10, 11) and this study indicate that the protective effect of excess weight is not
fully explained by detection bias. One argument
against detection bias is provided in a report by Pathak
and Whittemore (35), who examined the incidence of
breast cancer in high-, medium-, and low-risk countries. Breast cancer incidence was inversely associated
with body mass index among women in high-risk
countries and directly associated with risk in mediumand low-risk populations. As noted by the authors, it
seems unlikely that obesity would mask detection of
disease in high-risk but not in medium-risk countries
because controls in both populations had similar distributions of body mass index.
It is unlikely that the inverse association between
weight and breast cancer risk was due to disease
effects. The relation was least pronounced in women
with more advanced disease. Furthermore, the inverse
association of weight and breast cancer risk has been
705
706
Swanson et al.
hypothesis, but our method of assessing regularity of
ovulation was imprecise. Body fat distribution was not
associated with risk of breast cancer in young women.
This observation was unexpected given that upperbody obesity is associated with a hormonal profile
thought to increase breast cancer risk. As is the case
with generalized obesity, body fat distribution also
appears to have a different effect in younger versus
older women. To date, there are no established biologic explanations for these age-related differences.
ACKNOWLEDGMENTS
REFERENCES
1. Adami HO, Adams G, Boyle P, et al. Breast cancer etiology.
Int. J Cancer 1990;5:22-39.
2. Brinkley D, Carpenter RG, Haybittle JL. An anthropometric
study of women with cancer. Br J Prev Soc Med 1971^5:65-75.
3. Swanson CA, Jones DY, Schatzkin A, et al. Breast cancer risk
assessed by anthropometry in the NHANES I Epidemiological
Follow-up Study. Cancer Res 1988;48:5363-7.
4. Albanes E, Winick M. Arc cell number and cell proliferation
risk factors for cancer? J Natl Cancer Inst 1988;80:772-5.
5. Trichopoulos D, Lipman RD. Mammary gland mass and
breast cancer risk. Epidemiology 1992;3:523-6.
6. Micozzi MS. Nutrition, body size, and breast cancer. Yearbook Phys Anthropol 1985;28:175-206.
7. Harris JR, Lippman ME, Veronesi U, et al. Breast cancer. Part
1. N Engl J Med 1992;327:319-28.
8. Willett WC, Browne ML, Bain C, et al. Relative weight and
risk of breast cancer among premenopausal women. Am J
Epidemiol 1985;122:731-4O.
9. Swanson CA, Brinton LA, Taylor PR, et al. Body size and
breast cancer risk assessed in women participating in the
Breast Cancer Detection Demonstration Project. Am J Epidemiol 1989;13O:l 133-41.
10. London SJ, Colditz GA, Stampfer MJ, et al. Prospective study
of relative weight, height, and risk of breast cancer. JAMA
1989;262:2853-8.
11. Brinton LA, Swanson CA. Height and weight at various ages
and risk of breast cancer. Ann Epidemiol 1992;2:597-6O9.
12. Key TJA, Pike MC. The role of oestrogens and progestagens
in the epidemiology and prevention of breast cancer. Eur J
Cancer Clin Oncol 1988;24:29-43.
13. Ballard-Barbash R, Schatzkin A, Carter CL, et al. Body fat
distribution and breast cancer in the Framingham Study. J Natl
Cancer Inst 1990;82:286-90.
14. Schapira DV, Kumar NB, Lyman GH, et al. Abdominal obesity and breast cancer risk. Ann Intern Med 1990;l 12:182-6.
15. Folsom AR, Kaye SA, Prineas RJ, et al. Increased incidence of
carcinoma of the breast associated with abdominal adiposity in
postmenopausal women. Am J Epidemiol 1990;131:794-803.
16. Bruning PF, Bonfrer JMG Hart AAM, et al. Body measurements, estrogen availability and the risk of human breast
cancer: a case-control study. Int J Cancer 1992^1:14-19.
17. Brinton LA, Daling JR, Ljff JM, et al. Oral contraceptives and
breast cancer risk among younger women. J Natl Cancer Inst
Am J Epidemiol Vol. 143, No. 7, 1996
Downloaded from http://aje.oxfordjournals.org/ at Pennsylvania State University on February 26, 2014
The authors gratefully acknowledge Mary McAdams for
assistance with data management and analysis.
1995;87:827-35.
18. Waksberg J. Sampling methods for random digit dialing. J Am
Stat Assoc 1978;73:40-6.
19. National Center for Health Statistics. Obese and overweight
adults in the United States. Hyattsvillc, MD: National Center
for Health Statistics, 1983. (Vital and health statistics. Series
11, no. 230) (DHHS publication no. (PHS) 83-1680).
20. Bjomtorp P. The association between obesity, adipose tissue
distribution and disease. Acta Med Scand 1988;723 (Suppl.):
121-34.
21. Austin H, Austin JM Jr, Partridge EE, et al. Endometrial
cancer, obesity, and body fat distribution. Cancer Res 1991;
51:568-72.
22. Breslow NE, Day NE, eds. Statistical methods in cancer
research. Vol 1. The analysis of case-control studies. Lyon,
France: International Agency for Research on Cancer, 1980.
(IARC scientific publication no. 32).
23. Dubin N, Pasternack BS. Risk assessment for case-control
subgroups by polychotomous logistic regression. Am J Epidemiol 1986;123:1011-17.
24. De Waard F, Baanders-Van Halewijn EA. A prospective study
in general practice on breast-cancer risk in postmenopausal
women. Int J Cancer 1974; 14:153-60.
25. Kalish LA. Relationships of body size with breast cancer. J
Clin Oncol 1984;2:287-93.
26. Tornbcrg SA, Holm LE, Carstensen JM. Breast cancer risk in
relation to serum cholesterol, serum beta-lipoprotein, height,
weight, and blood pressure. Acta Oncol 1988;27:31-7.
27. Tretli S. Height and weight in relation to breast cancer morbidity and mortality. A prospective study of 570,000 women
in Norway. Int J Cancer 1989;44:23-30.
28. De Stavola BL, Wang DY, Allen DS, et al. The association of
height, weight, menstrual and reproductive events with breast
cancer results from two prospective studies on the island of
Guernsey (United Kingdom). Cancer Causes Control 1993;4:
331-40.
29. Ballard-Barbash R. Anthropometry and breast cancer: body
size—a moving target. Cancer 1994;74:1090-1100.
30. Wynder EL. Identification of women at high risk for breast
cancer. Cancer 1969;24:1235-40.
31. Helmrich SP, Shapiro S, Rosenberg L, et al. Risk factors for
breast cancer. Am J Epidemiol 1983;117:35-45.
32. KampertJB, Whittemore AS, Paffenbarger RS. Combined effect
of cruldbearing, menstrual events, and body size on age-specific
breast cancer risk. Am J Epidemiol 1988;128:962-79.
33. Bouchardy C, Le MG, Hill C. Risk factors for breast cancer
according to age at diagnosis in a French case-control study. J
Clin Epidemiol 1990;43:267-75.
34. Vatten LJ, Kvinnsland S. Prospective study of height, body
mass index and risk of breast cancer. Acta Oncol 1992;31:
195-200.
35. Pathak DR, Whittemore AS. Combined effects of body size,
parity, and menstrual events on breast cancer incidence in
seven countries. Am J Epidemiol 1992;135:153-68.
36. Pasquali R, Antenucci D, Casimirri F, et al. Clinical and
hormonal characteristics of obese amenorrheic hyperandrogenic women before and after weight loss. J Clin Endocrinol
Metabol 1989;68:173-9.
37. Demark-Wahnefried W, Winer EP, Rimer BK. Why women
gain weight with adjuvant chemotherapy for breast cancer. J
Clin Oncol 1993;11:1418-29.
38. Evans DJ, Hoffman RG, Kalhoff RK, et al. Relationship of
body fat topography to insulin sensitivity and metabolic profiles in premenopausal women. Metabolism 1984;33:68-75.
39. Sellers TA, Kushi LH, Potter JD, et al. Effect of family history,
body-fat distribution, and reproductive factors on the risk of
postmenopausal breast cancer. N Engl J Med 1992;326:1323-9.