Breast-feeding and the Risk of Breast Cancer in BRCA1 and BRCA2 Mutation Carriers

Abstract

Background: Several studies have reported that the risk of breast cancer decreases with increasing duration of breast-feeding. Whether breast-feeding is associated with a reduced risk of hereditary breast cancer in women who carry deleterious BRCA1 and BRCA2 mutations is currently unknown. Methods: We conducted a case–control study of women with deleterious mutations in either the BRCA1 or the BRCA2 gene. Study participants, drawn from an international cohort, were matched on the basis of BRCA mutation (BRCA1 [n = 685] or BRCA2 [n = 280]), year of birth (±2 years), and country of residence. The study involved 965 case subjects diagnosed with breast cancer and 965 control subjects who had no history of breast or ovarian cancer. Information on pregnancies and breast-feeding practices was derived from a questionnaire administered to the women during the course of genetic counseling. Conditional logistic regression analyses were used to estimate odds ratios (ORs) for the risk of breast cancer. All statistical tests were two-sided. Results: Among women with BRCA1 mutations, the mean total duration of breast-feeding was statistically significantly shorter for case subjects than for control subjects (6.0 versus 8.7 months, respectively; mean difference = 2.7 months, 95% confidence interval [CI] = 1.4 to 4.0; P <.001). The total duration of breast-feeding was associated with a reduced risk of breast cancer (for each month of breast-feeding, OR = 0.98, 95% CI = 0.97 to 0.99; P_trend <.001). Women with BRCA1 mutations who breast-fed for more than 1 year were less likely to have breast cancer than those who never breast-fed (OR = 0.55, 95% CI = 0.38 to 0.80; P = .001), although no such association was seen for BRCA2 (OR = 0.95, 95% CI = 0.56 to 1.59; P = .83). Conclusions: Women with deleterious BRCA1 mutations who breast-fed for a cumulative total of more than 1 year had a statistically significantly reduced risk of breast cancer.

An early age at first full-term pregnancy and increasing parity are both associated with long-term reductions in the risk of breast cancer in women from the general population ( 1 ). The risk of breast cancer increases transiently after pregnancy and then decreases for a period extending into the postmenopausal years ( 2 ). Women who carry a BRCA1 or BRCA2 mutation have a lifetime risk of breast cancer that may be as high as 80%–90% ( 3 , 4 ), although lower risk estimates (37%–56%) have also been reported ( 5 , 6 ). At least two studies ( 7 , 8 ) have reported that the risk of breast cancer is reduced in women who breast-fed their children and that the association with breast-feeding is greater for premenopausal women than for postmenopausal women ( 7 ). In a large international study ( 8 ), the risk of breast cancer was reduced by 4.3% for women who had a cumulative total of 12 months of breast-feeding and by 27% for women who had a cumulative total of breast-feeding of 55 months or more.

To date, no large studies have examined the relationship between breast-feeding and breast cancer risk among women who carry deleterious mutations in the BRCA1 or BRCA2 gene. To establish whether there is an association between breast-feeding and the risk of breast cancer among women who carry a deleterious mutation in either the BRCA1 or BRCA2 gene, we performed a matched case–control study.

M ethods

Study Population

Eligible subjects were drawn from a database of women who carried a deleterious mutation in either the BRCA1 or the BRCA2 gene. These women have been assessed for genetic risk at 53 centers located throughout North America, Europe, and Israel. All study subjects provided written informed consent for genetic testing. The study was approved by the ethics committees of all participating centers.

For most women, genetic testing was offered initially to those who were affected by either breast or ovarian cancer. Subsequently, testing was offered to their relatives. After a mutation in either the BRCA1 or BRCA2 gene was found in a proband or in her relative, genetic testing was offered to other at-risk women in the family. However, for less than 10% of the study subjects, an affected woman in a family was not available for genetic testing, and an unaffected woman was the first in the family to receive genetic testing.

Mutations in BRCA1 or BRCA2 genes were detected by using several techniques, but all nucleotide sequences were confirmed with direct sequencing of DNA. A woman was eligible for the present study after molecular analysis established that she was a carrier of a known deleterious BRCA1 or BRCA2 mutation.

Case and Control Subjects

Data were collected on breast-feeding for 1927 women with invasive breast cancer who carried a BRCA1 or BRCA2 mutation (case subjects) and for 2032 women without invasive breast cancer who carried such a mutation (potential control subjects). Case subjects were excluded if they had been diagnosed with ovarian cancer (n = 200), had undergone bilateral oophorectomy prior to diagnosis (n = 188), had incomplete data on pregnancies (n = 48) or on oral contraceptive use (n = 18), leaving 1473 eligible case subjects. Case and control subjects were matched on year of birth (±2 years), country of residence, and gene with the mutation (BRCA1 or BRCA2). Control subjects were excluded if they had been diagnosed with ovarian cancer (n = 403), had undergone prophylactic oophorectomy before the age of the case subject at diagnosis (n = 107), or data were missing on pregnancies (n = 20) or on oral contraceptive use (n = 19). Control subjects were eligible for matching if they were cancer-free and had not had a prophylactic mastectomy before the age of the case subject at diagnosis of breast cancer. In total, 965 matched pairs were generated: 685 case–control pairs who carried BRCA1 mutations and 280 case–control pairs who carried BRCA2 mutations. The only variables considered in assessing exposures were pregnancies, breast-feeding, and oral contraceptive use before the age of diagnosis in the matched case subject. The mean year of diagnosis of the case subjects was 39 years (range = 18–71 years). On average, the time between diagnosis of breast cancer in the case subject and the completion of the study questionnaire was 8.2 years.

In a second analysis, we excluded nulliparous women and matched case and control subjects on parity. Women who had experienced a stillbirth or reported an early infant death were also excluded because they would have had no opportunity to breast-feed the child. After matching for parity, the analysis included 481 case–control pairs who carried BRCA1 mutations and 172 case–control pairs who carried BRCA2 mutations.

Breast-feeding Information

The total months of breast-feeding for each pregnancy was reported. For women who gave an approximate length of breast-feeding (e.g., 3–5 months), the midpoint of duration was used. The year of breast cancer diagnosis and years of all pregnancies were recorded; however, if both occurred during the same calendar year, it was not possible to establish the sequence of breast cancer and pregnancy. Therefore, we included only live births that occurred at least one calendar year before a diagnosis of breast cancer. Twin births were considered to be one birth with respect to breast-feeding. Exposure information on control subjects (e.g., breast-feeding, parity) was restricted to the period before the date of a diagnosis of breast cancer in the matched case subject.

Data and Statistical Analysis

Breast-feeding histories were compared between matched case and control subjects. The paired Student's t test was used for comparing continuous variables among case and control subjects. Conditional logistic regression was used to estimate odds ratios (ORs) in the univariate and multivariate analyses. In the first analysis, parity and oral contraceptive use were used as covariate variables. Because of the strong association between parity and breast-feeding, we performed a second analysis that excluded all nulliparous women, in which we generated a new set of case–control pairs that were matched for parity. All analyses were adjusted for oral contraceptive use. All P values were two-tailed. Statistics were generated using the SAS statistical package, version 8.2 (SAS Institute, Cary, NC).

R esults

Case and control subjects were similar with respect to year of birth, age at first birth, age at last birth, and smoking during breast-feeding years (Table 1 ). Case subjects were, on average, 10 months older than matched control subjects. Case subjects were slightly younger at menarche than control subjects (Table 1 ).

Among women who carried deleterious BRCA1 mutations, the mean total duration of breast-feeding was statistically significantly shorter for case subjects than for control subjects (6.0 versus 8.7 months, respectively; mean difference = 2.7 months, 95% confidence interval = 1.4 to 4.0; P <.001). The mean length of breast-feeding per child was also statistically significantly shorter in case subjects than in control subjects (3.5 versus 4.4 months, respectively; mean difference = 0.9 months, 95% CI = 0.3 to 1.4 months; P = .002). There was a modest but statistically nonsignificant reduced risk of breast cancer associated with ever having breast-fed (OR = 0.81, 95% CI = 0.62 to 1.05; P = .12). However, a statistically significant trend for a reduced risk of breast cancer was associated with total duration of breast-feeding; for each additional month of breast-feeding, the risk decreased by 2% ( P_trend <.001) (Table 2 ). Among women with a deleterious BRCA1 mutation, those who had breast-fed their children for more than 1 year were 45% less likely to have breast cancer than those who had never breast-fed (OR = 0.55, 95% CI = 0.38 to 0.80; P = .001).

By contrast, among women who carried a deleterious BRCA2 mutation, breast-feeding was not associated with a reduced risk of breast cancer (Table 2 ). Although the mean lifetime duration of breast-feeding was slightly shorter among case subjects with BRCA2 mutations than among control subjects (7.6 versus 8.5 months, respectively), the difference was not statistically significant ( P = .50). The mean length of breast-feeding per child was also slightly shorter among case subjects than among control subjects (4.1 versus 4.4 months per child; P = .59). There was no statistically significant association between ever having breast-fed (OR = 1.10, 95% CI = 0.75 to 1.62; P = .63) or breast-feeding for more than 1 year (OR = 0.95, 95% CI = 0.56 to 1.59; P = .83) and risk of breast cancer.

Because there was a strong relationship between parity and breast-feeding, the data were also analyzed after excluding nulliparous women and matching case and control subjects for parity. The result of this matched analysis was essentially the same as that in which we did not match for parity. Women with a deleterious BRCA1 mutation who had breast-fed their children for more than 1 year were statistically significantly less likely to have breast cancer than those who had not breast-fed their children (OR = 0.53, 95% CI = 0.36 to 0.77; P <.001). Women with a deleterious BRCA2 mutation who breast-fed for more than 1 year had a statistically nonsignificantly reduced risk of breast cancer than women who breast-fed for less than 1 year (OR = 0.67, 95% CI = 0.34 to 1.26; P = .25).

Breast-feeding practices differed between subjects from different countries (Table 3 ). In all countries, with the exception of Israel, case subjects carrying deleterious BRCA1 mutations reported a shorter mean duration of breast-feeding than control subjects. Among women from all countries, a consistent reduced risk of breast cancer was associated with breast-feeding for 1 year or more (Table 4 ), but the association was statistically significant only for women from the United States.

D iscussion

We observed that women carrying deleterious BRCA1 mutations had a reduced risk of breast cancer if they breast-fed for 1 year or more. Among women carrying deleterious BRCA2 mutations, breast-feeding was not associated with a reduced risk of breast cancer. However, there were fewer women with BRCA2 mutations than women with BRCA1 mutations. In a previous case–control study from Iceland, breast-feeding was not found to reduce the risk of breast cancer in women with BRCA2 mutations ( 9 ).

In the present study, the reasons for not starting breast-feeding or for stopping breast-feeding were not recorded. It is possible that breast-feeding is not associated with a reduced risk of breast cancer but that women with BRCA1 mutations who have difficulty in breast-feeding are at greater risk for breast cancer than carriers who have no trouble breast-feeding. We previously reported that a high proportion of women who carry a BRCA1 mutation cited poor milk production as the main reason for weaning their infants compared with female relatives who do not carry a BRCA1 mutation ( 10 ). One study of women with nonhereditary breast cancer found that the risk of breast cancer was increased among women who tried to breast-feed but could not ( 11 ).

How breast-feeding is associated with a reduced risk of breast cancer is unclear but may be related to changes in mammary gland differentiation or to effects on breast estrogen levels. The mammary gland develops at puberty and undergoes developmental changes with each pregnancy ( 12 , 13 ). During the first part of each pregnancy, the breast epithelium proliferates rapidly, and then the breast tissue undergoes a final differentiation during the final months of the pregnancy. As a result, the parous breast contains a higher percentage of well-differentiated lobules than the nonparous breast ( 14 ). Lobuloalveolar differentiation is stimulated by estrogen, progesterone, placental lactogen, placental growth hormone, oxytocin, and prolactin. After delivery, lactation begins as a result of increasing prolactin levels. When prolactin levels decline, postlactational regression occurs and the glands involute ( 15 ). Because breast fluid estrogen levels are suppressed for several years after breast-feeding ( 16 ), lactation may diminish breast cancer risk by altering the hormonal milieu directly. Breast-feeding may also reduce the risk of breast cancer indirectly by delaying the re-establishment of ovulation ( 17 ). Indeed, women who breast-fed exclusively during the first 6 months postpartum have a low (1%–5%) rate of ovulation ( 18 ), and the number of lifetime ovulations has been linked to breast cancer risk. In support of this association, we observed a small but statistically significant difference in the age of menarche between our case and control subjects (Table 1 ).

BRCA1 is critical to both appropriate proliferation and differentiation within the mammary gland. In the murine mammary gland, Brca1 expression is increased during puberty and pregnancy, possibly to limit proliferation and to promote differentiation ( 19 ). There is also evidence from human cell culture systems that BRCA1 suppresses estrogen-mediated breast cell proliferation ( 20 ). BRCA1 is also important in mammary development; mammary tissue from mice with a mammary-specific Brca1 deletion does not develop normally and does not differentiate during pregnancy and lactation ( 21 ). Results from other animal models have demonstrated that the susceptibility of the mammary gland to cancer is related to the rate of proliferation of breast epithelial cells and is inversely related to the degree of differentiation ( 13 ).

Individuals with low levels of BRCA1 may have increased breast epithelial cell proliferation in response to the increased estrogen exposure of pregnancy. Russo et al. ( 14 , 22 ) reported that breast tissue from parous women with a family history of breast cancer or a BRCA1 mutation does not differentiate normally and resembles breast tissue from nulliparous women.

We used a case–control approach to study the relationship between breast-feeding and breast cancer. Because of our large sample size, we were able to assess the effect separately for women who carry deleterious BRCA1 or BRCA2 mutations. We controlled for parity, which is strongly correlated with breast-feeding and is a risk factor for hereditary breast cancer ( 23 ), by making two different adjustments. The results were similar regardless of which adjustment for parity was used.

Several possible biases could affect our results. Information on breast-feeding was obtained after childbearing was completed; for case subjects, information was obtained after the diagnosis of breast cancer. If there is differential reporting of breast-feeding by case and control subjects, then recall bias is possible. However, there is no obvious reason why case subjects should be less likely to recall breast-feeding their children than control subjects. Furthermore, study subjects did not know their BRCA mutation status when their children were born, and if recall bias were present, it is not clear why the association should be restricted to BRCA1 carriers.

Our data support the hypothesis that hormonal and reproductive factors modify the risk of breast cancer among women with BRCA1 mutations ( 24 ). We found that 1 or more years of breast-feeding in women with deleterious BRCA1 mutations was associated with a reduction in breast cancer risk of 45%—an effect that is much greater than that seen in the general population ( 8 ). In a large multicenter case–control study of patients with breast cancer from 30 countries, breast-feeding for 12 months or more was associated with a reduced risk of breast cancer of 4.3% ( 8 ). In our study, we found no association between breast cancer risk and breast-feeding for women with BRCA2 mutations; the difference between women with BRCA1 and BRCA2 mutations may reflect underlying differences in the pathogenesis of cancers associated with the two genes. However, because our sample of women with BRCA2 mutations was small, it is premature to conclude that a modest reduced risk is not present in this subgroup as well.

Table 1.

Characteristics of control and case subjects

Variable	Control subjects	Case subjects	P†
		(n = 965) mean (SD * )		(n = 965) mean (SD)
Mutation
BRCA1, %	71.0	71.0
BRCA2, %	29.0	29.0
Country of residence
Canada	337	337
Israel	72	72
Poland	117	117
United Kingdom	9	9
Sweden	12	12
United States	418	418
Year of birth	1952.9 (10.1)	1952.1 (10.7)	.11
Age at diagnosis, y	N/A §	39.1 (7.7)
Age at menarche, y	12.9 (1.5)	12.8 (1.5)	.03
Age at first birth, y	23.9 (4.5)	23.9 (4.6)	.86
Age at last birth, y	30.0 (5.1)	29.9 (5.1)	.44
Mean parity	1.95 (1.6)	1.87 (1.4)	.24
Nulliparous, %	208 (21.6)	204 (21.2)	.82
Oral contraceptive use, ever (%)	666 (69.0)	673 (69.7)	.73
No. of women who smoked‡	327 (34.8)	346 (36.7)	.40
Ethnicity
Jewish	288 (29.8)	247 (25.6)
French Canadian	95 (9.9)	108 (11.2)
Other white	564 (58.5)	581 (60.3)
Black	9 (0.9)	15 (1.6)
Other	5 (0.5)	7 (0.7)
Missing	4 (0.4)	7 (0.7)	.23

*

SD = standard deviation.

†

The t test was used for mean age and mean parity, and chi-square test was used for the remaining analyses.

‡

Data on smoking during childbearing years were missing for 46 women.

§

N/A = not applicable.

Table 2.

Risk of breast cancer associated with duration of breast-feeding in women carrying deleterious BRCA1 and BRCA2 mutations

Breast-feeding duration	Univariate analysis					Multivariable analysis †
		OR *	(95% CI) *	P	OR	(95% CI)	P
BRCA1
Breast-feeding
None	1.00	(referent)		1.00	(referent)
≤1 y	0.89	(0.70 to 1.15)	.40	0.89	(0.68 to 1.17)	.41
>1 y	0.56	(0.41 to 0.76)	<.001	0.55	(0.38 to 0.80)	.001
Trend per month	0.98	(0.97 to 0.99)	<.001	0.98	(0.97 to 0.99)	<.001
BRCA2
Breast-feeding
None	1.00	(referent)		1.00	(referent)
≤1 y	1.18	(0.74 to 1.75)	.42	1.12	(0.73 to 1.71)	.61
>1 y	1.03	(0.66 to 1.60)	.91	0.95	(0.56 to 1.59)	.83
Trend per month	1.00	(0.99 to 1.01)	.51	0.99	(0.98 to 1.01)	.29

*

OR = odds ratio; CI = confidence interval.

†

Adjusted for oral contraceptive use and parity.

Table 3.

Duration of breast-feeding among parous control and case subjects from different countries according to deleterious BRCA mutation

Country	No. of pairs	Total months of breast-feeding					Months of breast-feeding per child
				Control subjects	Case subjects	P	Control subjects	Case subjects	P
BRCA1 mutation
Canada	206	7.4	6.4	.32	3.9	3.6	.55
Israel	50	8.1	9.3	.49	3.6	3.9	.70
Poland	116	10.3	7.8	.13	4.3	3.8	.44
United Kingdom	6	3.2	2.6	.85	2.4	0.8	.45
Sweden	10	26.4	9.1	.23	8.1	5.4	.35
United States	297	8.7	4.5	<.001	4.8	3.1	.001
All	685	8.7	6.0	<.001	4.4	3.5	.002
BRCA2 mutation
Canada	131	7.5	5.0	.10	4.3	2.8	.04
Israel	22	8.3	10.5	.62	2.8	4.9	.44
Poland	1	14.0	3.0	–	7.0	1.5	–
United Kingdom	3	0	13.8	.33	0	4.5	.40
Sweden	2	9.0	11.8	.67	6.0	7.2	.75
United States	121	9.7	9.7	.99	5.0	5.3	.72
All	280	8.5	7.6	.50	4.4	4.1	.59

Table 4.

Odds ratios (95% confidence intervals) for risk of breast cancer associated with breast-feeding for 12 months or longer among women who carry a deleterious BRCA1 mutation, by country

Country *	No. of pairs	OR (95% CI) †	P
Canada	206	0.55 (0.29 to 1.07)	.08
Israel	50	0.85 (0.18 to 4.17)	.85
Poland	116	0.70 (0.27 to 1.82)	.46
United States	297	0.46 (0.26 to 0.81)	.007

*

There were too few pairs from Sweden (n = 20) and from the United Kingdom (n = 12) to generate estimates for these countries.

†

OR = odds ratio; CI = confidence interval.

Supported by grants from the National Institutes of Health (RO1 CA63682, CA63678, CA61231, CA74415, and CA55914), the National Cancer Institute of Canada Breast Cancer Initiative, the Department of Defence (DAMD17-94-J-4299 and DAMD17-94-J-4260, DAMD17-94-J-4340), the Canadian Genetic Diseases Network, the Canadian Breast Cancer Foundation (Ontario Chapter, British Columbia and Yukon Division), the Fonds de la Recherche en Santé du Québec, the Swedish Cancer Society, the Swedish Research Council (VR), the Breast Cancer Research Foundation (BW), and the State of Nebraska Cancer and Smoking-Related Diseases Research Program (LB595). Also supported by revenue from Nebraska cigarette taxes awarded to Creighton University by the Nebraska Department of Health and Human Services. We also thank Drs. Å. Borg, F. Couch, M. J. Dabney, D. Easton, A. Eisen, C. Eng, G. Evans, D. Fishman, D. Gilchrist, K. Hughes, O. T. Johannsson, B. Karlan, E. Lemire, N. Loman, W. McKinnon, J. McLennan, S. Merajver, G. Mills, K. Offit, O. I. Olopade, H. Olsson, M. Osborne, D. Provencher, G. Rennert, T. Rebbeck, H. Saal, J. Scalia, S. Tischler, N. Tung, T. Wagner, E. Warner, and G. Weitzel for their contribution of patients to this study.