Vigorous physical training
1-5 and even moderate exercise
6-9 can interrupt the menstrual cycle, perhaps by suppressing the pulsatile release of gonadotropin-releasing hormone.
10,11
This effect of physical activity may lower a woman's cumulative
exposure to estrogen and progesterone, thereby inhibiting carcinogenesis
in the breast.
12-22
Energy balance might also influence the risk of breast cancer. Caloric
restriction in rodents reduces the proliferative activity of the mammary
glands
23 and inhibits carcinogenesis.
24,25
However, the effect of energy balance, as indicated by energy intake,
body-mass index (the weight in kilograms divided by the square of the
height in meters), and energy expenditure, on the risk of breast cancer
has not been examined thoroughly in humans.
In this study we
evaluated the influence of physical activity, both at work and during
leisure time, on the risk of breast cancer in a cohort of 25,624
premenopausal and postmenopausal women. Data on parity, dietary factors,
and body-mass index allowed adjustment for potentially confounding
factors, and reassessment of physical activity after three to five years
gave an indication of the effect of sustained physical activity on the
risk of breast cancer.
Methods
Study Population
From
1974 to 1978, the National Health Screening Service invited people in
three counties in Norway (Oppland, Sogn og Fjordane, and Finnmark) to
participate in a survey of risk factors for cardiovascular disease. All
women who were 35 to 49 years of age and a random sample of 10 percent
of those who were 20 to 34 years of age were invited. In four
municipalities in Finnmark all women who were 20 to 34 years of age were
invited. A comprehensive description of these populations has been
published previously.
26 A total of 31,556 women were invited to participate, and 28,621 (91 percent) actually did.
All
women in this survey as well as a random sample of women who were 20 to
39 years of age were invited to participate in a second survey three to
five years later (1977 to 1983). Of these 34,378 women, 31,209 (91
percent) participated.
27
This second survey was used as the base line, because no information on
parity and dietary factors was collected during the first survey.
Each
woman received a written invitation to participate, together with a
one-page questionnaire. The participants were asked to answer the
questionnaire and bring it to the clinical examination. At screening,
trained nurses checked the questionnaire for inconsistencies regarding
physical activity and menopausal status, measured weight and height, and
collected blood samples.
During screening in the second survey,
the participants were asked to fill out a food-frequency questionnaire,
to be returned by mail. After one reminder, 25,892 (83 percent) returned
the questionnaire. The energy and fat intakes for each woman were
derived from the sum of all food consumed. The semiquantitative
food-frequency questionnaire that we used has been described in detail
and validated.
28,29
Assessment of Physical Activity
Self-reported
categories of physical activity during leisure hours in the year
preceding each survey were assessed when the women entered the study and
graded from 1 to 4 according to the participant's usual level of
physical activity. A grade of 1 was assigned to those whose leisure time
was spent reading, watching television, or engaging in other sedentary
activities; a grade of 2 to those who spent at least four hours a week
walking, bicycling, or engaging in other types of physical activity; a
grade of 3 to those who spent at least four hours a week exercising to
keep fit and participating in recreational athletics; and a grade of 4
to those who engaged in regular, vigorous training or participating in
competitive sports several times a week. The self-reported level of
physical activity during work hours in the preceding year was also
graded on a four-point scale. A grade of 1 was assigned to those whose
work was mostly sedentary; a grade of 2 to those whose job involved a
lot of walking; a grade of 3 to those whose job required a lot of
lifting and walking; and a grade of 4 to those engaged in heavy manual
labor.
Two identical assessments of leisure-time activity were
made at an interval of three to five years, and the results were
combined for all groups. Women who reported moderate (grade 2) or
regular (grade 3 or 4) exercise during leisure time in the first survey
and regular exercise (grade 3 or 4) in the second survey were
characterized as being consistently physically active. Women who were
sedentary (grade 1) during leisure time in both surveys were
characterized as being consistently sedentary. The women who were
neither consistently sedentary nor consistently active during leisure
time were characterized as being moderately active.
Follow-up and Identification of Cases of Breast Cancer
We
followed a total of 25,707 women who had not been given a diagnosis of
cancer before our base-line survey (1977 to 1983). We used the
participants' national 11-digit personal identification numbers to
identify every incident case of breast cancer reported to the Cancer
Registry of Norway and Statistics Norway through the end of follow-up
(December 31, 1994). A total of 98 percent of the cases were verified
histologically. Women in whom cancer developed (n = 72) or who died (n =
11) within the first year of the study were excluded from the analyses
to account for the possibility that undiagnosed cancer or severe illness
might influence the level of physical activity. Through a linkage to
the Central Population Register at Statistics Norway, we obtained
information concerning the reproductive history of each woman, including
the date of birth of each liveborn child through December 31, 1992, and
deaths in the cohort through December 31, 1994.
The ultimate
study cohort consisted of 25,624 women who participated in both surveys
(age range, 20 to 69 years) during 359,930 person-years of follow-up.
Statistical Analysis
Base-line
variables were adjusted for age and compared by analysis of covariance.
Cox proportional-hazards regression analysis was carried out to
investigate the simultaneous effect of physical activity and covariates
on the incidence of breast cancer. To calculate the risk of breast
cancer, women were observed for the development of breast cancer from
entry into the study to the date of diagnosis of any cancer, the time of
death, or the end of follow-up, whichever event came first. In the
analysis, grades 3 and 4 of leisure-time activity were merged because of
the small numbers of women with a grade of 4 in both surveys (48 women
in the first survey and 57 in the second survey). As a reference group
we used women who were sedentary at work or during leisure time.
In
the analyses, we adjusted for age at entry (a continuous variable),
county of residence, number of children, age at birth of first child,
intake of total fat and energy, and body-mass index. Women who reported
that they were premenopausal at base line were treated as premenopausal
until they reached the age of 50 during follow-up, at which time they
were considered postmenopausal. Women who reported that they were
postmenopausal at base line were treated as postmenopausal.
Because
there were few women with breast cancer who were sedentary both at work
and during leisure time, the effect of this combination on the risk of
breast cancer could not be analyzed. All significance tests were
two-tailed, and the level of significance was set at 5 percent. The
analyses were performed with the SAS statistical package version 6.11.
Results
There
were 351 incident cases of breast cancer (100 among premenopausal women
and 251 among postmenopausal women) among 25,624 women. The mean length
of follow-up was 14.0 years (median, 13.7), and the median age at
diagnosis was 54.7 years (range, 36.3 to 68.0).
Table 1Table 1Base-Line Characteristics of the Women According to the Level of Physical Activity in the 1977–1983 Survey.
gives the base-line characteristics of the participants. Two thirds of
the women reported moderate activity during leisure time, whereas 15
percent exercised regularly. Only 14 percent reported being sedentary at
work, whereas 20 percent reported lifting and 5 percent reported doing
heavy manual labor. Women who reported regularly exercising during
leisure time did not differ from women who were inactive during their
leisure time with respect to age at entry or number of children, but
they tended to be taller and to have a lower body-mass index, a
relatively low ratio of total cholesterol to high-density lipoprotein
(HDL) cholesterol in serum, lower serum triglyceride levels, and higher
HDL cholesterol levels. Women whose work involved lifting or heavy
manual labor had a higher body-mass index and more children than those
engaged in sedentary work. Energy intake was positively related to
physical activity, but the association was more pronounced with work
activity than with leisure-time activity.
We analyzed other
possible age-adjusted risk factors for breast cancer at base line and
found a 28 percent increase in risk for each additional 6 cm of height
and a 13 percent reduction in risk for each child. An older maternal age
at the birth of a first child was associated with a borderline increase
in risk, whereas body-mass index (in the group as a whole or in the
subgroups of premenopausal and postmenopausal women), energy intake, and
total fat intake did not influence the overall risk of breast cancer
(data not shown).
Table 2Table 2Adjusted
Relative Risk of Breast Cancer According to the Level of Physical
Activity during Leisure Time and at Work in the 1977–1983 Survey.
shows the relation between the level of leisure-time or work activity
and the overall risk of breast cancer. After adjustment for age and with
the sedentary group as the reference group, the relative risk of breast
cancer was reduced among women whose jobs involved walking, lifting, or
heavy manual labor. Adjustments for other factors (body-mass index,
county of residence, number of children, and height) in addition to age
changed the risk estimates only slightly. Further adjustments for age at
first birth or dietary factors (energy intake, total fat intake, and
fiber intake) did not influence our estimates of relative risk and were
omitted from the final model. A 52 percent reduction in risk was
observed among the women who reported doing heavy manual labor (relative
risk, 0.48; 95 percent confidence interval, 0.25 to 0.92). The overall
adjusted risk of breast cancer decreased in a dose–response manner with
increasing activity level during leisure time (P for trend = 0.04).
Women who exercised at least four hours a week during leisure time had a
37 percent reduction in the risk of breast cancer (relative risk, 0.63;
95 percent confidence interval, 0.42 to 0.95).
When the group was divided according to menopausal status (
Table 3Table 3Adjusted
Relative Risk of Breast Cancer According to Menopausal Status and the
Level of Physical Activity in the 1977–1983 Survey.), a
consistently inverse association was observed between the level of
leisure-time activity and the premenopausal risk of breast cancer; the
adjusted relative risk declined to 0.77 (95 percent confidence interval,
0.46 to 1.27) and further to 0.53 (95 percent confidence interval, 0.25
to 1.14) as the level of activity increased (P for trend = 0.10). A
weaker association was observed between the level of leisure-time
activity and the postmenopausal risk of breast cancer. The inverse
association between the level of activity at work and the risk of breast
cancer was also pronounced among premenopausal women; among
premenopausal women whose jobs involved lifting or heavy manual labor,
the relative risk was 0.48 (95 percent confidence interval, 0.24 to
0.95).
We also divided the cohort into women who were younger
than 45 years of age at entry and those who were 45 or older. Among
those younger than 45 years at entry for whom data were complete (of
whom breast cancer developed in 138; mean age at diagnosis, 48.3 years),
the adjusted relative risk declined to 0.80 (95 percent confidence
interval, 0.52 to 1.22) and further to 0.38 (95 percent confidence
interval, 0.19 to 0.79) as the level of activity during leisure time
increased (P for trend = 0.01). The respective adjusted relative risks
were 1.03 (95 percent confidence interval, 0.72 to 1.48) and 0.84 (95
percent confidence interval, 0.51 to 1.39) (P for trend = 0.54) among
those for whom data were complete who were 45 years of age or older at
entry (of whom breast cancer developed in 208; mean age at diagnosis,
58.2 years). These values indicate that physical activity had a
protective effect, particularly with respect to the risk of breast
cancer before and soon after menopause.
We examined models stratified according to body-mass index (
Table 4Table 4Adjusted
Relative Risk of Breast Cancer According to Body-Mass Index and the
Level of Physical Activity during Leisure Time in the 1977–1983 Survey.).
Among lean (body-mass index, <22.8), regularly exercising women, the
risk of breast cancer was reduced by 72 percent (relative risk, 0.28;
95 percent confidence interval, 0.11 to 0.70). No such association was
observed in the middle or upper thirds of body-mass index among
regularly exercising women. In models stratified according to both
body-mass index and menopausal status, this association was seen among
both premenopausal and postmenopausal lean women (data not shown).
In
the second survey 61.2 percent of the participants reported the same
level of leisure-time activity as in the first survey, 23.5 percent
reported an increased level, and 15.3 percent reported a reduced level.
By combining these two assessments of leisure-time activity, we observed
that the relative risk declined to 0.23 (95 percent confidence
interval, 0.09 to 0.60) as the level of sustained activity increased in
lean (body-mass index, <22.8) women (P for trend = 0.002) (
Table 5Table 5Adjusted
Relative Risk of Breast Cancer According to Body-Mass Index and Overall
Level of Physical Activity during Leisure Time in the 1974–1978 and
1977–1983 Surveys.). This protective effect across
increasing levels of sustained leisure-time activity was observed in
both lean premenopausal women (relative risk, 0.23; 95 percent
confidence interval, 0.06 to 0.88; P for linear trend = 0.02) and lean
postmenopausal women (relative risk, 0.24; 95 percent confidence
interval, 0.06 to 0.96; P for linear trend = 0.03).
Discussion
Our
results support the idea that physical activity protects against breast
cancer, particularly among premenopausal and younger postmenopausal
women. Activity during both leisure time and work reduced the overall
risk. There was a significant inverse dose–response relation between
leisure-time activity and the risk of breast cancer. The protective
effect was evident among lean premenopausal and postmenopausal women,
and repeated assessment emphasized the preventive effect of physical
activity.
The overall reduction in the risk of breast cancer among active women is consistent with findings in other cohort
15,17 and case–control
19-22 studies, but at variance with the findings of a few others.
30,31 In one of these discrepant studies,
31
most of the women were older than in the present study and breast
cancer was diagnosed mainly among postmenopausal women. In the other,
30
physical activity at college was assessed 35 to 70 years before the
diagnosis of breast cancer, and no adjustments were made for potential
confounding factors. Our finding of a protective effect of work-related
activity on the risk of breast cancer is also in agreement with other
studies.
18,32,33
Precise
assessment of physical activity is difficult in a population-based
cohort. The accuracy of the levels of leisure-time activity reported on
the questionnaire that we used has been validated previously.
34-36 Since the level of leisure-time activity correlates with the degree of physical fitness,
34,36
our observation that recreationally active women tended to be leaner
than inactive women and had serum lipid profiles associated with regular
exercise strengthens the validity of our assessments. Energy intake was
also positively related to both leisure-time and work activities,
particularly work activities.
Repeated assessment of leisure-time
activity is important in any analysis of the effect of sustained
activity on the risk of breast cancer. The protective effect was notable
among lean women who were consistently active during their leisure
time. In combining the two assessments for each woman, we may also have
increased the precision of our assessment of physical-activity levels,
but we cannot differentiate the effect of sustained activity from any
misclassification.
The population-based approach and the high
participation rate in our study reduced selection bias. The almost
complete reporting of incident cases of breast cancer also strengthens
our results. Age at menarche was not available and could have confounded
our results, but this is not likely, since an increased risk of only 4
percent was observed for each year of earlier age at menarche in a
similar study population in Norway.
37
Information
about the use of hormonal contraceptives was not available, although
recent meta-analyses suggest that there is only a small increase in the
risk of breast cancer among the youngest women who commonly use hormonal
contraceptives.
38 It is probable that this information would not have confounded our results to any large extent.
How does physical activity influence the development of breast cancer? The propensity to be physically active may be inherited,
39
so the genotype may influence both physical activity and the
predisposition to breast cancer. Social and cultural influences on
exercise and energy balance seem to be more important than genetic
factors,
39,40
which points to leisure-time activity as an independent and modifiable
variable with regard to its effect on the risk of breast cancer.
A
reduction in the cumulative exposure to cyclic estrogens and
progesterone may in part explain the preventive effect of both
leisure-time and work activity. Over the long term, vigorous training
and moderate leisure-time activity may decrease estradiol and
progesterone secretion,
3,6,41 reduce the length of the luteal phase,
10,42 induce anovulation,
7,8,41,43 delay menarche,
4,5 and cause secondary amenorrhea.
2,12
Physical
activity influences energy balance, and experimental studies have shown
that calorie restrictions inhibit mammary carcinogenesis.
24,25,44
Anthropometric measures such as height, body-mass index, and weight
gain have been used as biomarkers of calorie intake, and increased
values have been reported to be risk factors for breast cancer in
humans.
45-48 A diet involving a high energy intake has also been associated with early age at menarche,
5,49
and this finding supports the hypothesis that increased net energy may
increase the cumulative hormonal levels that are of importance for
carcinogenesis of the breast. Women who were active during leisure time
reported only a slightly higher total energy intake than sedentary
women, and they tended to be leaner, indicating that their net available
energy was lower. The greater protective effect of leisure-time
activity against breast cancer in lean women indicates that there may be
an optimal energy balance that inhibits mammary carcinogenesis.
Triglycerides
are known to displace estradiol from its tight binding to the sex
hormone–binding globulin, which is found in low levels in obese women,
50
and thus triglycerides increase levels of free estradiol. Serum levels
of triglycerides were higher in sedentary women than in women who were
more active during their leisure time; thus, exposure to estrogen may be
greater in inactive women. This underscores the importance of avoiding
obesity if physical activity is to have an optimal inhibitory effect on
the risk of breast cancer.
Original article at: http://www.nejm.org/doi/full/10.1056/NEJM199705013361801#t=articleTop