Family History of Cancer in Relation to Breast Cancer Subtypes in African American Women

Having a mother, sister, or daughter with a breast cancer diagnosis is a well-known risk factor for breast cancer (1). Among African American women, estimates of the relative risk for first-degree family history of breast cancer range from 1.65 to 1.78 (2, 3), similar to findings from studies of European American and Asian women (1, 4, 5). Among studies that reported results separately for estrogen receptor (ER)–positive and ER⁻ breast cancer, most reported similar associations by subtype (3, 6–11), whereas one reported a stronger association with ER⁺ cancer (12) and two reported a stronger relation for ER⁻ breast cancer (13, 14). Only the Black Women's Health Study (BWHS) reported on family history separately for ER⁺ and ER⁻ breast cancer in African American women, with similar increases by subtype, but findings were based on small numbers (3).

A first-degree family history of cancers other than breast cancer may also increase breast cancer risk. Family history of prostate (15, 16), lung (17), ovarian (18), and colon or colorectal cancer (3, 16) have been associated with greater risk of breast cancer in some, but not all studies that examined specific other cancers. In studies that examined combinations of cancers, risk of breast cancer was elevated for family history of breast and prostate cancers (15), breast and ovarian cancers (19, 20), and breast and colorectal cancers (16, 21, 22). Among African American women, family histories of lung cancer (23), colon cancer (3), or both breast and prostate cancer (16) were associated with increased risk of breast cancer.

The objective of this study was to investigate the relation of first-degree family history of breast and other cancers to the risk of ER⁺, ER⁻, and triple-negative breast cancer in African American women.

The African American Breast Cancer Epidemiology and Risk (AMBER) Consortium has been described in detail elsewhere (24). The AMBER Consortium pools data on African American women from two cohort studies, the BWHS and the Multiethnic Cohort Study (MEC), and two case–control studies, the Carolina Breast Cancer Study (CBCS) and the Women's Circle of Health Study (WCHS). Informed consent was provided to each study by its participants. Each study and the consortium were approved by the relevant Institutional Review Boards.

The BWHS is a prospective cohort study that enrolled 59,000 African American women across the United States in 1995 (25). Participants were 21 to 69 years old at baseline when they completed an extensive health questionnaire and are followed with biennial questionnaires for data on incident diagnoses and other factors. Incident breast cancers were identified through self-report on questionnaires or through linkage to state cancer registries. For the AMBER Consortium, a nested case–control study was created; cohort participants without breast cancer were frequency-matched to cases based on age (5-year categories), geographic region, and the most recent completed questionnaire.

The MEC is a prospective cohort study that enrolled men and women in Los Angeles county and Hawaii from 1993 through 1996 (26). Participants were 45 to 75 years at baseline when they completed an extensive questionnaire and have been followed with questionnaires in 1999, 2003, and 2010 to update information. Breast cancer diagnoses are identified through linkage with the Los Angeles County Cancer Surveillance Program and the California Cancer Registry. A nested case–control study of African American women was created to pool MEC data with the AMBER Consortium. Controls, selected from women who had not developed breast cancer, were frequency-matched to cases on age (5-year categories) and the most recent completed questionnaire.

The CBCS is a case–control study that enrolled women in North Carolina from 1993 through 2001 (27). Participants were 20 to 74 years old and were interviewed inperson. Cases were identified through the North Carolina Central Cancer Registry, while controls were identified through Division of Motor Vehicle lists or Health Care Financing Administration lists. Controls were frequency-matched to cases on the basis of age (5-year categories).

The WCHS is a case–control study that enrolled women in New York from 2003 through 2008 and in New Jersey beginning in 2006 (28). Recruitment in New Jersey is ongoing. Participants were 20 to 75 years old and were interviewed in person for data collection. Cases were identified through New York City hospitals and the New Jersey State Cancer Registry, while controls were identified through random digit dialing and community-based recruitment (29). Controls were frequency-matched to cases on the basis of age (5-year categories).

Each study confirmed incident breast cancer cases with data on ER, progesterone receptor (PR), and HER2 obtained from medical records and/or state cancer registries (24). Cases were classified as ER⁺, ER⁻, and triple-negative (ER⁻/PR⁻/HER2⁻). Of the 5,736 potential cases, ER status was available for 4,520 cases (79%) at the time of this analysis. PR status was available for 4,301 cases (75%); HER2 status was available for fewer cases (2,927; 51%), due to more recent inclusion of HER2 in routine testing. There were no statistically significant differences between women with and without known receptor status by age or family history of breast cancer. In total, there were 3,023 ER⁺ cases, 1,497 ER⁻ cases (including 696 triple-negative cases), and 17,420 controls.

Participants were asked whether any parent, sibling, or child (first-degree relative) had been diagnosed with breast cancer and whether the relative was diagnosed before age 50. Participants were also asked about first-degree family history of ovarian, colorectal, prostate, lung, and cervical cancer and lymphoma or leukemia.

Each study obtained detailed data on most known and suspected risk factors for breast cancer. Variables were centrally harmonized and evaluated as risk factors for breast cancer overall and for ER⁺, ER⁻, and triple-negative breast cancer (24, 30–33).

Multinomial logistic regression models were used to calculate ORs and 95% confidence intervals (CI) for the relation of family history of cancer to the risk of ER⁺, ER⁻, and triple-negative breast cancer. Multivariable models adjusted for the design variables, age (5-year categories), study (BWHS, CBCS, MEC, and WCHS), geographic region (Northeast excluding New Jersey, New Jersey, South, Midwest, and West), and questionnaire time period (1993–1998, 1999–2005, and 2006–2014), and recency of mammogram (never had a mammogram, mammogram within past 2 years, and last mammogram more than 2 years ago). Additional variables were also assessed as potential covariates but were not associated with family history of cancer and did not appreciably change the effect estimates: years of education (<12, 12, 13–15, 16, and >16 years), menopausal status and age at menopause (premenopausal; <45, 45–49, 50–54, and ≥55 years), years of use of postmenopausal estrogen together with progesterone (never used; <5, and ≥5 years), age at menarche (<11, 11–12, 13–14, 15–16, and ≥17 years), body mass index (<18.5, 18.5–24.9, 25–29.9, 30–34.9, 35–39.9, and ≥40 kg/m²), years of oral contraceptive use (never used; <1, 1–9, and ≥10 years), parity (nulliparous; 1, 2, 3, and ≥4 births), age at first birth (<25 and ≥25 years), lactation (parous and never breastfed, parous and ever breastfed), pack years of cigarette smoking (never smoked; <20 and ≥20 pack years), and alcohol consumption (never drinker, former drinker, current drinker of <7 drinks/week, and current drinker of ≥7 drinks/week). The missing indicator method was used to handle missing values for covariates. To test for interaction between family history of cancer at different cancer sites, interactions were examined by introducing cross-product terms into the models (34). Analyses were conducted using SAS 9.3 statistical package (SAS Institute Inc). Random effects meta-analyses of study-specific results were conducted using Stata/SE 11.2 statistical software (StataCorp LP), with results tested for heterogeneity by the Cochran Q statistic (35).

The prevalences of first-degree family history of breast cancer and of cancers other than breast cancer were largely similar across studies (Table 1); the differences were statistically significant due to the large sample size. Among controls, 9.3% had a first-degree relative with breast cancer, and 22.7% had a first-degree relative with a cancer other than breast cancer. Few controls had first-degree family history of both breast cancer and another cancer site (2.9%). Lung cancer was the most common other cancer among first-degree relatives (7.7%), followed by prostate (7.6%) and colorectal cancer (6.2%).

The ORs for a first-degree family history of breast cancer were similar by subtype: ER⁺ cancer (1.76; 95% CI, 1.57–1.97), ER⁻ cancer (1.67; 95% CI, 1.42–1.95), and triple-negative cancer (1.72; 95% CI, 1.38–2.13; Table 2). For each subtype, the ORs were higher if the relative was diagnosed before age 50. For example, for ER⁻ cancer, the OR was 1.96 (95% CI, 1.56–2.46) for having a relative diagnosed with breast cancer before age 50 and 1.46 (95% CI, 1.15–1.86) for a relative diagnosed at 50 or older. The ORs were also somewhat higher if the participant herself was diagnosed with breast cancer before age 45: the ORs for the association of having a first-degree relative diagnosed with breast cancer before age 50 with the risk of breast cancer before age 45 were all greater than 3 for ER⁺, ER⁻, and triple-negative breast cancer.

In assessing the relation of family history of other cancers to breast cancer risk, we looked first at the risk of overall breast cancer. A first-degree family history of breast cancer alone (with no other cancers among first-degree relatives) was associated with a 1.58-fold risk (95% CI, 1.38–1.82; Table 3). The ORs for family history of each of the other cancers alone were close to 1, with the exception of cervical cancer, for which the OR for the association with overall breast cancer risk was 1.53 (0.94–2.47). The risk of breast cancer was tripled in women who had a family history of both breast and prostate cancer (OR = 3.02; 95% CI, 2.19–4.16; P_interaction < 0.01). The OR for a family history of both breast and cervical cancer was 3.56 (95% CI, 0.99–12.85), but this estimate was based on only 7 exposed breast cancer cases. The risk of breast cancer was significantly increased for women with a family history of three or more cancer sites, but only when breast cancer was one of the sites.

Next, we considered the family history of a cancer diagnosis in relation to the risk of specific breast cancer subtypes. The ORs for first-degree family history of breast cancer alone were similar for ER⁺, ER⁻, and triple-negative breast cancer (Table 4). A family history of cervical cancer was associated with ER⁻ (OR = 2.56; 95% CI, 1.44–4.53) and triple-negative (OR = 3.04; 95% CI, 1.57–5.87) breast cancer, but not with ER⁺ cancer. A family history of lung cancer was associated with a 20% increase in the risk of ER⁺ cancer (95% CI, 1.04–1.48), whereas a family history of prostate cancer was associated with a 24% increase in the risk of ER⁺ cancer (95% CI, 1.00–1.44). There were no significant associations with family history of any of the other sites, although the OR for the association of family history of ovarian cancer with the risk of triple-negative breast cancer was elevated (OR = 1.53; 95% CI, 0.89–2.65). The OR for a family history of both breast and prostate cancer was 3.40 (95% CI, 2.42–4.79) for ER⁺ cancer, as compared with 1.62 for breast alone (P_interaction = 0.02), and was 2.09 (95% CI, 1.21–3.63) for ER⁻ cancer, as compared with 1.50 for breast alone (P_interaction = 0.11); for triple-negative breast cancer, the corresponding OR was 1.60, as compared with 1.54 for breast alone (P_interaction = 0.62). The ORs for breast and lung and for breast and colorectal cancer were also higher, although not significantly higher, than for breast cancer alone for some subtypes. Having a family history of breast cancer and two or more other cancer sites was associated with an increased risk of each subtype, with the ORs ranging from 2.42 for ER⁺ to 2.78 for ER⁻ cancer.

The results were similar across the four studies. For example, the ORs for first-degree family history of breast cancer in relation to breast cancer risk were 1.74 (95% CI, 1.51–2.02) in BWHS, 1.59 (95% CI, 1.19–2.14) in CBCS, 1.88 (95% CI, 1.50–2.35) in MEC, and 1.55 (95% CI, 1.22–1.97) in WCHS (P_{heterogeneity} = 0.65). To assess the possibility of recall bias, we examined the associations separately in the case–control and cohort studies. The ORs for first-degree family history of breast cancer in relation to the risk of breast cancer were 1.57 (95% CI, 1.31–1.90) in the case–control studies and 1.80 (95% CI, 1.60–2.03) in the cohort studies (P_{heterogeneity} = 0.23).

This large study provides convincing evidence that first-degree family history of breast cancer is associated with ER⁺, ER⁻, and triple-negative breast cancer in African American women and that having a relative diagnosed with breast cancer at a young age is a strong predictor of risk. Having a history of breast cancer together with prostate cancer was associated with a further increase in the risk of each subtype. Family history of ovarian cancer was not associated with an increased risk of ER⁺ or ER⁻ cancer, but there was some evidence of a positive association with triple-negative breast cancer. In addition, we observed an unexpected association of family history of cervical cancer with an increased risk of ER⁻ breast cancer.

Previous studies with data on African American women have also shown family history of breast cancer to be a strong risk factor for breast cancer (2, 3, 16, 36–39). Only the BWHS and the Women's CARE study also considered the age of the relative at diagnosis, and both observed a greater risk of breast cancer when the relative was diagnosed at a younger age (2, 3). In the only study to present data in African American women by subtype, the BWHS, a similar increase was observed across subtypes (3).

Findings according to the subtype from other populations have been mixed. The association with family history of breast cancer has been similar across breast cancer subtypes (3, 6–11, 40), stronger for ER⁺ breast cancer (12, 41), and stronger for ER⁻ or triple-negative breast cancer (13, 14, 42). The strongest evidence comes from a pooled analysis of 12 studies in the Breast Cancer Association Consortium, where an association with family history of breast cancer was present across subtypes, but with a stronger association for basal-like breast cancer (40).

Only a few studies of African Americans have examined the relation of family history of cancers other than breast cancer to the risk of breast cancer (3, 16). In the BWHS, a family history of colon cancer was associated with an increased risk of breast cancer, with a relative risk estimate of 1.35, but the study did not consider whether participants also had a family history of breast cancer (3). In the Women's Health Initiative, having a family history of both breast and prostate cancers was associated with a 2.34-fold increase in the risk of breast cancer (16). Neither of these studies presented data by breast cancer subtype. Limited data by subtype are available from other populations. In a predominantly European American population from the Iowa Women's Health Study, a family history of prostate cancer was associated with an increased risk of both ER⁺/PR⁺ and ER⁻/PR⁻ breast cancer (43). A family history of lung cancer was associated with increased risk of hormone receptor–positive breast cancer in a case–control study in China (44). No previous study has reported an association of family history of cervical cancer with an increased risk of breast cancer.

Associations with family history of cancer could be explained in part by environmental or genetic factors shared within families. A family may have similar reproductive habits (45–47), dietary patterns (48), physical activity (49, 50), or body size (51, 52), each of which influences the risk of various cancers (53). Although knowledge of a family history of cancer influences cancer screening, individuals with a family history of cancer do not differ in lifestyle from individuals without knowledge of a family history (54–56). Genetics play a role in breast cancer etiology (57, 58). Heritable mutations in BRCA1 or BRCA2 genes are associated with an increased risk of both breast cancer and ovarian cancer (59, 60). Germline mutations in the BRCA1 and BRCA2 genes have also been associated with an increased risk of prostate and colorectal cancers (59–61), whereas germline mutations in the CHEK2 gene increase risk of breast, prostate, and colon cancers (62).

Our observation of a strong increase in risk among participants with a family history of both breast and prostate cancer may relate to recent genetic findings. A family history of prostate cancer has been associated with mutation in the RNASEL gene in African Americans (63). Mutations in this gene have also been associated with the risk of breast and cervical cancers (64). A potential mechanism linking the RNASEL gene and ER⁻ breast cancer is inflammation. Inflammatory markers have been elevated in studies of hormone-negative cancers (65–67). RNASEL variants have been associated with elevated inflammatory biomarkers (68), and the enzyme encoded by the RNASEL gene has proinflammatory functions (69).

Analyses of data from The Cancer Genome Atlas and other genomic data suggest that there are etiologic links between ovarian cancer and basal-like breast cancer (70–73), which is primarily composed of triple-negative tumors. Consistent with those data, we observed a nonsignificant 53% increase in the risk of triple-negative breast cancer associated with a first-degree family history of ovarian cancer. Genomic analyses also suggest that there are biologic similarities between basal-like breast cancer and lung cancer (72, 73). In our data, the risk of triple-negative breast cancer was significantly increased for a first-degree family history of lung cancer only in the presence of a first-degree family history of breast cancer.

African American women experience a higher prevalence of early-onset breast cancer and of ER⁻ breast cancer compared with European American women (74–76). Because of the large sample size, we were able to informatively assess breast cancer risk by age and for breast cancer subtypes. We were also able to assess family history of cancers other than breast cancer. We controlled for multiple potential confounding factors. The self-report of family history of participants may have been incomplete and could have been subject to recall bias in the case–control studies. However, previous validation studies have shown that self-reported family cancer histories for first-degree relatives are accurate for breast cancer (77), and the results were similar across our studies, which included cohort studies in which family history data were provided before the participant was diagnosed with breast cancer. In addition, the prevalence of family history of breast cancer among controls in the AMBER Consortium (9.3%) was similar to the prevalence in other studies (78, 79). We did not have data on all cancer sites that may be of interest, such as the endometrium and pancreas.

In summary, the present findings suggest that family history of cancers other than the breast cancer may indicate a higher inherited genetic susceptibility to breast cancer. Women who had both breast and prostate cancer–affected family members had a particularly high risk of both ER⁺ and ER⁻ breast cancer. Greater surveillance and counseling for additional screening may be warranted.

No potential conflicts of interest were disclosed.

The results do not necessarily represent the views of or an official position held by the sponsors.

Conception and design: T.N. Bethea, L. Rosenberg, J.R. Palmer

Development of methodology: T.N. Bethea, L. Rosenberg, J.R. Palmer

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): L. Rosenberg, E.V. Bandera, C.B. Ambrosone, J.R. Palmer

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): T.N. Bethea, L. Rosenberg, N. Castro-Webb, K.L. Lunetta, L.E. Sucheston-Campbell, E.A. Ruiz-Narváez, M.A. Troester, C.B. Ambrosone, J.R. Palmer

Writing, review, and/or revision of the manuscript: T.N. Bethea, L. Rosenberg, K.L. Lunetta, L.E. Sucheston-Campbell, E.A. Ruiz-Narváez, M. Charlot, S.-Y. Park, E.V. Bandera, M.A. Troester, C.B. Ambrosone, J.R. Palmer

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): T.N. Bethea, N. Castro-Webb, M. Charlot

Study supervision: K.L. Lunetta, J.R. Palmer

The authors thank the participants and staff of the contributing studies. Data on breast cancer pathology were obtained from several state cancer registries (AZ, CA, CO, CT, DE, DC, FL, GA, IL, IN, KY, LA, MD, MA, MI, NJ, NY, NC, OK, PA, SC, TN, TX, and VA).

The AMBER Consortium was supported by grant P01CA151135 from the NCI (to all authors). Financial support to the collaborating studies was provided by the NCI through grants R01CA058420 (to L. Rosenberg and J.R. Palmer), UM1CA164974 (to L. Rosenberg, J.R. Palmer, E.A. Ruiz-Narváez, T.N. Bethea, and M. Charlot), P50CA58223 (to M.A. Troester), R37CA054281 (to S.-Y. Park), U01CA164973 (to S.-Y. Park), and R01CA100598 (to C. Ambrosone, E.V. Bandera, and L.E. Sucheston-Campbell) and by the University Cancer Research Fund of North Carolina (to M.A. Troester).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.