Menstrual Factors and Breast Cancer Risk

GUNNAR KVALE, MD, AND IVAR HEUCH, DPHIL

Relationships between menstrual factors and breast cancer risk were investigated in a
prospective study of 63,090 Norwegian women. A total of 1565 cases of breast cancer
occurred during follow-up from 1961 through 1980. The risk of breast cancer decreased
with increasing age at menarche (P = 0.06) and increased with increasing age at menopause
(P = 0.005) in analyses adjusted for age, urban/rural place of residence, parity, and age at
first and last birth. The results correspond to an average increase in breast cancer risk of
4.0% for each year of decrease in age at menarche, and an increase in risk of 3.6% for each
year of increase in age at menopause. The protective effect of early menopause was
strongest for breast cancer diagnosed in patients 80 years of age or older. No clear
relationship was seen between menstrual irregularities and breast cancer risk.
Cancer 62:1625-1631,1988.
There exists extensive evidence of a protective effect of low age at menopause on breast cancer
risk. There also are several studies indicating that early menarche is associated with increased
risk, but this evidence is less conclusive.' In the current report, these associations are assessed
using data from a population-based prospective study of 63,090 Norwegian women who were
observed from 1961 through 1980. We previously reported, from the same cohort, a strong
inverse association with parity-parity and more complex relationships with ages at first and last
birth and with lactation.
Materials and Methods

From 1956 through 1959, all women 20 to 69 years of age by January 1, 1956, in four
counties in Norway were invited to attend a screening program for early diagnosis of breast
cancer. Each woman was interviewed according to a standard questionnaire and had a clinical
breast examination camed out by a physician. In addition to demographic data, the questionnaire
elicited information on different reproductive variables (age at menarche and menopause,
number of full-term pregnancies, age at first and last birth, number of abortions, and duration of
lactation). Age at menarche and menopause was defined as age at first and last menstruation,
respectively.
The current study was confined to residents of the three counties of Vestfold, Aust-Agder,
and Nord-Trondelag who were 27 to 69 years of age by January 1, 1956 (total, 92,573 women).
The official personal registration number was retrieved for 85,063 women who were still alive at
the start of follow-up on January 1,196 1. Of these women, 63,090 had attended the screening
program and were interviewed during the years 1956 through 1959, corresponding to an overall
response rate of 74.2% in this group. The response rate was higher in the younger age groups, in
rural areas, among people occupied in farming or forestry, and in Nord-Trondelag where the
screening program was offered three times (Table 1).
The official registration number served as a unique identification of each woman's record and
was used to link follow-up information to our files. Thus, data on cancer cases (obtained from
the Cancer Registry of Norway) and complete information concerning emigration and death
(obtained from the Central Bureau of Statistics) could be added to the records with information

on reproductive variables. Dates of diagnosis and sites of primary tumor were checked
thoroughly by the Cancer Registry, and these variables and the histologic subtype recorded were
not revised for this study. The Cancer Registry covers the whole population of Norway and the
registration is practically complete for almost all sites, including breast cancer. The follow-up
period extended from January 1, 1961, to December 31,1980. Thus, cancer cases diagnosed at
screening in 1956 through 1959 (60 cases) and those representing an earlier history of breast
cancer ( 140 women) were not included in the study. Data on height and weight were linked to
our files for 63,086 women who also attended a screening program organized by the National
Mass Radiography Service from 1963 through 1975.
The analyses of incidence of breast cancer were adjusted for age at start of follow-up
(with 5-year age groups) and urban/rural place of residence, and in special cases, reproductive
factors other than the variable studied. Thus, parity was considered with six categories (0, 1, 2, 3,
4, and 5), and age at first and last birth with five Categories each (corresponding to 5-year
intervals). The adjustment was made by forming a stratum for each combination of categories of
the covariables. A test for trend in association was carried out by the stratified extension of the
Cochran Armitage statistic. Relative odds were estimated by stratified logistic regression
analyses on scores depending on age at menarche or age at menopause, respectively. The
percentage increase in breast cancer risk per year decrease in age at menarche or per year
increase in age at menopause could be estimated from the slopes of the regression lines. A test
for departure from trend allowed assessment of the appropriateness of the model.
Of the 63,090 participants, 14,216 died and 126 emigrated in the period 196 1 through
1980. In the estimation procedure, a crude correction for censoring was introduced by decreasing
the initial number at risk by half the number of deaths and emigrations occurring among those in
whom breast cancer did not develop. The calculations also included a test for interaction between
the study variable and the separate variables defining strata.The expected numbers of cases in the
various categories under the hypothesis of no association were derived in analyses adjusted for
times until censoring. These numbers also were incorporated in the calculation of two-tailed P
values for linear trend. Strata without cancer cases, or in which all participants were at the same
level of the study variable, provided no information and were deleted. For this reason and
because of missing values, the number of women included may vary according to the degree of
stratification in the statistical analyses.
In analyses involving age at menopause, only 22,151 participants who had passed
menopause at the time of interview and reported their age at menopause could be included. Of
these women, 1491 reported unilateral or bilateral oophorectomy, other operations on the
ovaries, or hysterectomy, and the main analyses were limited to the remaining 20,660 women.
Results
Table 1 shows the distribution of age at menarche over categories defined by
demographic variables. Age at menarche decreased in successive years of birth. The lowest age
at menarche was observed in rural areas, in the county of Nord-Trondelag, and among
professional workers, but differences according to place of residence and occupation were small.
Ever married women were younger at menarche than never married women. Complete
information on age at menopause was obtained only for participants born in 1900 or earlier, and
the distribution of age at menopause showed little variation among different subgroups according
to demographic variables. One thousand five hundred sixty-five cases of breast cancer were

diagnosed during follow-up among the 63,090 women who attended the screening program. One
thousand five hundred twenty-eight cases occurred among the 6 1,36 1 women who reported age
at menarche.
We observed a significant decrease in breast cancer risk with increasing age at menarche
(Table 2). The relative odds estimate was 0.96 per year of increase in age at menarche, in logistic
regression analyses with adjustment for age at start of follow-up and urban / rural place of
residence. Adjustments for panty and age at first and last birth did not influence this estimate,
which corresponds to an average increase in breast cancer risk per year decrease in age at
menarche of 4.0%. Women who were 17 years of age or older at menarche had somewhat higher
breast cancer risk than women who were 16 years of age at menarche. The departure from a
linear trend was not, however, statistically significant.
Five hundred eighty cases of breast cancer occurred among the 20,660 postmenopausal
women who reported age at natural menopause. The association with age at menopause was
highly significant. The relative odds estimate per 5 year increase in age at menopause was 1.19
in logistic regression analysis with adjustment for age, urban / rural place of residence, panty,
and age at first and last birth (Table 3). This estimate corresponds to an average increase in breast
cancer risk of 3.6% per year increase in age at menopause. For the 44 cases diagnosed among
1491 women who had reported unilateral or bilateral oophorectomy, other operations involving
the ovaries, and/or hysterectomy, we observed an odds ratio of 1.25 per 5 year increase in age at
menopause (95% confidence interval, 0.90 to 1.74).
The effects of age at menarche and age at menopause were still obtained in analyses with
additional adjustment for occupation, body mass index (weight/height), or height. Age at
menarche and age at menopause were not correlated in our cohort, and adjustment for either
factor did not alter the effect of the other. The effect of age at menarche was similar for breast
cancers diagnosed at different ages, whereas age at menopause had the strongest effect for
cancers diagnosed in women 80 years of age or older (Table 4). The relative odds estimate for
women reporting irregular menstruation versus other women was 0.93 in analyses with
adjustment for parity and age at menarche (95% confidence interval 0.74 to 1.16).
Of all breast cancer cases included in this study, 98% were verified by histologic
examination. The classification into histologic subtypes was based on information reported to the
Cancer Registry by pathologists in many different laboratories. Thus, it is rather crude. Despite
some variation in risk estimates between subtypes, we observed no significant heterogeneity,
neither with regard to age at menarche nor age at menopause.
It should be noticed that most women in the younger part of the cohort were
premenopausal at the time of interview. In the younger age groups, therefore, women with an
early menopause would be overrepresented among participants who reported their age at
menopause. This will not, in itself, introduce any bias when we study the dependence of cancer
risk on age at menopause, provided that age at the start of follow-up is properly adjusted for and
that the logistic model holds in all age strata. Thus, results based on analyses with more detailed
adjustment for age at the start of follow-up did not differ essentially from the main results with
adjustment for age in 5 year groups, nor did results from analyses restricted to women 60 years
of age or older at the start of follow-up.

Discussion

Our results point to definite associations between age at menarche and age at menopause
and breast cancer risk that may reflect causal relationships with hormonal changes occurring at
these points in life. However, the associations also could arise because of correlations with
different determinants of the disease. factors related to socioeconomic class or other reproductive
variables. Finally, the possibility of bias in study design should be considered.
In a follow-up study, we avoid problems such as selection and recall bias, which may
distort results in casecontrol studies. Age at menarche is influenced by childhood socioeconomic
conditions, and early menarche may indicate a particular childhood environment related to longterm increase in risk of breast cancer, independent of hormonal effects. However, the
associations with age at menarche and age at menopause were still found after adjustment for
occupation, body mass index, or height. Thus, it seems unlikely that the associations observed in
our study can be due to bias or confounding with other risk factors.
Our analyses of the effect of age at menarche on breast cancer risk showed a rather
smooth trend with decreasing risk for increasing age at menarche from 13 to 16 years. However,
the women with menarche before 13 years of age had a somewhat lower risk, and those with
menarche after 16 years of age had higher risk than predicted by the linear trend. For those with
late menarche, similar results were observed for cancer of the uterine corpus. Although the
departure from trend was not statistically significant, these results may reflect hormonal
disturbance among women with very late menarche. They support the contention that absence of
menstruation beyond the 16th year should be considered abnormal. Thus, ovarian disease is found
in a large proportion of women with primary amenorrhea.
Our results regarding age at menarche are consistent with findings from other prospective
studies and a large international case-control study. However, some studies have found no
relationship between age at menarche and breast cancer risk. In a number of case-control studies,
the mean age at menarche was only slightly less among women with breast cancer than among
controls. More recent reports have not been entirely consistent. Some have indicated a general
protective effect of late menarche, I4-l7 whereas others have not 18-19 Certain studies have
shown a definite protective effect among premenopausal women only 20-22 supporting previous
reports of a stronger effect premenopausally and perimenopausally than in postmenopausal
women. However, there also have been studies showing a decreased risk for postmenopausal
breast cancer 0nly 23-24 In our data, the suggested protective effect of late menarche seemed to
be independent of age at breast cancer diagnosis (Table 4).
The failure to identify an effect of age at menarche, especially in case control studies,
may be related to inaccurate recall. However, there are data suggesting that recall of age at
menarche is reasonably accurate 25 Nevertheless, this problem might explain the lack of
association seen in some studies, especially for older breast cancer patients. Furthermore, as the
effect of age at menarche is likely to be weak, one should not expect statistically significant
results in small studies. Thus, the current cumulative evidence definitely suggests that early age
at menarche is a risk factor for breast cancer, both in premenopausal and postmenopausal years.
A history of menstrual irregularities was not strongly associated with breast cancer risk.
Women reporting irregular menstruation had somewhat decreased risk. This finding was not
statistically significant, but is consistent with previous observations from Finland and Italy
suggesting that women with irregular cycles may be at reduced risk. In a Swedish study a deficit
in long cycles was observed among women with breast cancer. To the extent that menstrual

irregularities indicate anovulatory cycles, these results are consistent with the hypothesis that
regular ovulatory cycles may increase breast cancer risk.
Early data support previous reports of a protective effect of early menopause
1,14,16,20,21,30,31 Some studies 32-33 have indicated that this relationship may be strongest in
older women, and our results confirm that the protective effect lasts until old age (Table 4).
However, although the highest risk estimate was observed in the oldest age group, we found no
consistent trend towards a stronger protective effect with increasing age at diagnosis. Because
information on age at menopause was obtained 2 to 5 years before the start of follow-up, it was
not possible to investigate the effect of menopause for cancer occurring in perimenopausal
women.
Only 169 of the women with known age at menopause reported surgical removal of both
ovaries. Thus, we could not adequately study the effect of age at artificial menopause on the risk
of breast cancer. The estimate of the association with age at menopause in the group of 1491
women who reported unilateral or bilateral oophorectomy, other operations involving the
ovaries, and/or hysterectomy, was similar to the effect among the women with natural
menopause shown in (Table 3).
The association with age at menopause was highly significant and seemed stronger than
the effect of age at menarche, but this may be due to the larger range of variation in years for
menopause. Our relative odds estimates imply an increase in breast cancer risk per year decrease
in age at menarche equal to 4.0%. The corresponding increase in risk per year increase in age at
menopause is 3.6%. These figures are, however, subject to large sampling errors. The 95%
confidence intervals for the estimates associated with age at menarche and age at menopause
were 0.0% to 8.2% and 1.1% to 6.0%, respectively. Thus, although the effects of one additional
year of menstrual activity on the risk of postmenopausal breast cancer seems to be of the same
order of magnitude at the start and end of the reproductive period, our results also are consistent
with substantial differences in effect.
Our estimate of the effect of age at menopause agrees with recent results of Hiatt and
Fireman 30 who observed that for every year earlier the menstruation ceased, a woman's risk of
breast cancer was reduced by 3.0%. Average estimates per year have not been reported in other
studies, but from the relative risks given, most studies seem to have found somewhat stronger
effect of the order of 5% to 6% per year of difference in age at natural menopause.
On the basis of a mathematical model, which gave a good prediction of age-specific
incidence curves in different populations, moolgavker et al. found that an age at menarche of 10
years carried a twofold risk over an age at menarche of 16 years. Similarly, menopause at 52
years of age was predicted to carry a twofold risk over menopause at 42 years of age. These
estimates correspond to an increase in risk of 12% per year decrease in age at menarche, and 7%
increase in risk per year increase in age at menopause. Most studies, including our own, have
found weaker effects. If the assumptions in the model are correct, this discrepancy might indicate
inaccurate recall, which would tend to weaken observed relationships, especially for menarche.
A relationship between the total number of years of menstrual activity and risk of breast
cancer has been reported previously. It has been proposed that breast cancer risk may be directly
related to the total number of sterile menstrual cycles, and that the effect of various reproductive
factors could be explained by their contribution towards a reduction in the number of such
cycles. Our results regarding menarche and menopause are consistent with this theory. However,
the protective effect of a pregnancy seems to be stronger than expected under this hypothesis.
Thus, from the relative risk estimates of the effect of paritywe found that each new full term

pregnancy resulted in a reduction in breast cancer risk of 16%. Further, the age when a
pregnancy occurs and the length of intervals between pregnancies may be important. There is
still no general agreement on the effect of lactation on breast cancer risk. However, a simple
relationship explained by the duration of postpartum amenorrhoea, which correlates well with
the duration of lactation, seem unlikely. Thus, we strongly recommend that each reproductive
variable be studied separately. Analyses of such indices as total number of sterile menstrual
cycles may conceal, rather than disclose, important associations.
Associations with early menarche and late menopause have been previously related to a
higher frequency of anovulatory cycle. However, Wallace observed that women with early
menarche established a pattern of regular cycle intervals more promptly than women with late
menarche did, and early onset of ovulatory cycles has been observed in women with early
menarche, case-control study, Henderson found that breast cancer patients established regular
cycles significantly earlier than controls. Data from a study by MacMahon suggest that the high
breast cancer risk in women with early menarche is most likely explained by both a longer
duration and a higher level of exposure to estrogens experienced by these women. Women with a
late menopause are exposed to normal endogenous estrogen and progesterone for a longer time
period than women with early menopause, and also may experience a larger number of
anovulatory cycles4' resulting in a lack of cyclic progesterone. However, whereas the evidence
linking estrogens to risk of breast cancer is abundant, the role of progesterone in the cause of
breast cancer is less clear
In conclusion, our data support previous reports of an inverse association of breast cancer
risk with age at menarche and a positive association with age at menopause, The results are
consistent with the hypothesis that the effects of these variables are due to their contribution to
the overall duration of menstrual activity, which determines the duration of exposure to high
levels of ovarian hormones. These effects may, in particular, depend on the total lifetime number
of ovulatory cycles.